Article review

Stuck with a difficult assignment? No time to get your paper done? Feeling confused? If you’re looking for reliable and timely help for assignments, you’ve come to the right place. We promise 100% original, plagiarism-free papers custom-written for you. Yes, we write every assignment from scratch and it’s solely custom-made for you.

Order a Similar Paper Order a Different Paper

Please read each article and fill out the form for each article

Article Extract Form



Journal, date, pages:

Institution of primary author:

Most important previous studies mentioned in literature review:

Why the aforementioned articles are most important:

Research question:


Study design:

Sample size and selection procedures:

Dependent variable(s):

Primary independent variable(s):

Analytic or statistical technique used:

Primary findings:

Weakness of study:

Strength of study:

Other important issues:

Conceptual knowledge modulates memory recognition of common
items: The selective role of item-typicality

Cristiane Souza1 & Margarida V. Garrido1 & Oleksandr V. Horchak1 & Joana C. Carmo2

Accepted: 16 July 2021
# The Psychonomic Society, Inc. 2021

This work examines the influence of stored conceptual knowledge (i.e., schema and item-typicality) on conscious memory
processes. Specifically, we tested whether item-typicality selectively modulates recollection and familiarity-based memories as
a function of the availability of a categorical schema during encoding. Experiment 1 manipulated both encoding type (categorical
vs. perceptual) and item-typicality (typical vs. atypical) in a single Remember-Know paradigm. Experiment 2 replicated and
extended the previous study with a complementary source-memory task. In both experiments, we observed that typical items led
to more Guess responses, while atypical items led to more Remember responses. These findings support the idea that the
activation of a congruent categorical schema selectively enhances familiarity-based memories, likely due to the bypassing of
the activated mechanisms for novel information. In contrast, atypical items improved recollective-based memories only, sug-
gesting that their lesser fit with the stored prototype might have triggered those novelty processing mechanisms. Moreover,
atypical items enhanced memory in the categorical condition for both item recognition and recollection memories only, sug-
gesting an episodic gain due to inconsistency/novelty. The source memory results gave further credence to the argument that
“Remember” judgments were based on truly recollective experiences and presented the same interaction between encoding type
and item-typicality observed in recollective-based memories. Overall, the results suggest that the supposedly opposite conceptual
knowledge effects actually coexist and interact, albeit selectively, in the modulation of recollection and familiarity processes.

Keywords Recollection . Familiarity . Schemas . Item-typicality . Declarative memories


Declarative memory rests on explicit long-term storage sys-
tems of meaningful representations that can be consciously
retrieved. Episodic memory refers to our capability to main-
tain vivid representations of contextually relevant details of
the events (e.g., remembering the precise details about our first
visit to our best friend’s home) and is associated with
autonoetic (self-based) conscious awareness while re-
experiencing memories (Bastin et al., 2019; Liu et al., 2020;
Tulving, 2000, 2002; Yonelinas et al., 2010). Semantic mem-
ory constitutes a general knowledge that is abstracted from our
experiences (e.g., the basic social rules when having dinner at

someone’s home) and is related to noetic (factual-based) con-
sciousness (Tulving, 1985, 2002).

Episodic and semantic memories rest on different process-
es and neural substrates. Likewise, recollection and
familiarity-based processes associated with memory recogni-
tion entail distinct operations supported by different brain re-
gions (Gardiner, 1988; Tulving, 1972, 2000; Yonelinas, 2002;
Yonelinas et al., 2010; but see also Migo et al., 2012, and
Wixted & Mickes, 2010, for a single-process model perspec-
tive on how both recollection and familiarity support recogni-
tion). Recollection processes are characterized by a controlled
and effortful vivid recovery. These processes are embedded
with self-related conscious awareness while re-experiencing
memories and are supported by hippocampus structures
(Tulving, 1985, 2000; Yonelinas, 2002). Familiarity refers to
an economical and less demanding process involving factual-
based conscious awareness. This process is driven by holistic
operations (i.e., unicity) that support the retrieval of known
information (see Ozubko et al., 2017; Wang et al., 2018;
Yonelinas et al., 2010), and is supposedly hippocampal-inde-
pendent. Therefore, the reported dissociation between

* Cristiane Souza
[email protected]

1 Iscte- Instituto Universitário de Lisboa, Cis-Iscte, Av. das Forças
Armadas, 1649-026 Lisboa, Portugal

2 Faculdade de Psicologia, Universidade de Lisboa, Lisbon, Portugal

Memory & Cognition

episodic and semantic memories resembles, both functionally
and structurally, the contrast between recollection and
familiarity-based processes (Czernochowski et al., 2004;
Tulving, 2002; Vargha-Khadem et al., 2003; Wang et al.,
2018). The present study examines how these two processes
involved in recognition memory are distinctly influenced by
different types of conceptual knowledge (i.e., schema and

Recent studies have shown the advantage of stored sche-
matic knowledge availability (i.e., schema) on the formation
and retrieval of memories (Liu et al., 2016; Tse et al., 2007;
Tse et al., 2011; van Kesteren et al., 2014; van Kesteren, Beul,
et al., 2013a; van Kesteren, Rijpkema, et al., 2013b; Yamada
& Itsukushima, 2013). For instance, information congruent
with previously learned schemata has been shown to engage
cortical regions and was better retrieved than incongruent in-
formation (e.g., Dudai et al., 2015; van Kesteren et al., 2010,
2014; van Kesteren, Beul, et al., 2013a; van Kesteren,
Rijpkema, et al., 2013b), suggesting the rapid integration of
this type of information into the semantic system. In contrast,
information that is incongruent with a prior schema engages
brain regions and their connectivities, which are classically
associated with the episodic system (van Kesteren et al.,
2010, 2014). Critically, information that is incongruent with
a schema was also shown to improve subsequent memories
despite being more susceptible to forgetting with time
(Bonasia et al., 2018).

Moreover, the debate on the role of prior schema becomes
even more intricate depending on whether prior schema facil-
itation for congruent information is considered a generalized
process in declarative memories or whether it is regarded as
selective for specific memory processes. The facilitation effect
of a prior schema for congruent items has been reported in
situations where previous abstract schematic knowledge en-
hances familiarity-based memories compared to recollective
ones (see Carr et al., 2013; Mäntylä, 1997; Rajaram, 1998). Of
particular interest, Mäntylä (1997) explored the effect of dis-
tinct encoding types on different memory processes by con-
trasting a relational encoding task (based on similarities with
the prior conceptual knowledge) with a distinctive encoding
task based on item-specific information (i.e., how distinctive a
face is in contrast with others). Specifically, this was tested
during a face recognition memory task with the Remember-
Know paradigm. In this paradigm, the phenomenological
judgment regarding memory experience (Remember vs.
Know responses) is obtained together with item-recognition
scores. Remember responses usually reflect recollection while
Know responses capture a factual-based sense of familiarity
(Gardiner, 1988; Gardiner et al., 1998; Tulving, 2000;
Yonelinas et al., 2010). The results of Mäntylä’s study showed
an increase in Know responses in relational encoding and an
increase in Remember responses in distinctive encoding con-
ditions (Mäntylä, 1997). Thus, it seems that the availability of

a schema during learning leads to a selective increase in
familiarity-based memories only. Moreover, the advantage
of distinctive encoding over schema availability in
recollective memories suggests that the schema advantage is
not observed in such memory process.

The schema effect is considered controversial from a cog-
nitive perspective, namely given the mixed-effects reported in
category learning literature (De Brigard et al., 2017; Harris &
Rehder, 2006; Sakamoto & Love, 2004; Yin et al., 2019).
According to this literature, a category can be viewed as a
schema, an abstract, experienced-based, flexible, and contin-
uously updated associative knowledge structure (see Gosh &
Gilboa, 2014). Following this analogy, Sakamoto and Love
(2004) investigated how consistency with a new categorical
schema affects memory. The authors concluded that the rec-
ognition of items that are inconsistent with the category is
improved because they violate knowledge structures (rules)
inherent to the schema regularities. On the other hand, recent
studies on category learning demonstrated that consistency
with a newly learned category improved recognition and en-
hanced false alarms (De Brigard et al., 2017; Yin et al., 2019).
Therefore, the role of categorical stored representations in
memory retrieval needs to be further scrutinized.

Categorical prototypes are understood to be schematic
knowledge constituting an abstraction and an average repre-
sentation of the attributes of the category (Murphy, 2002;
Murphy & Medin, 1985). According to classical models of
concepts and semantic organization, typicality – a property
underlying semantic organization – influences the categoriza-
tion process and declarative memories (Keller & Kellas, 1978;
Rips et al., 1973; Rosch et al., 1976). Typicality refers to how
good an exemplar is in representing its own category, which is
determined by the match of each of its features with the pro-
totypical stored representation (Lin & Murphy, 1997; Medin
et al., 2007; Rosch & Mervis, 1975). Typical items are good
exemplars, that is, those closer to the abstract representation in
memory (e.g., prototypes). In contrast, atypical items have less
fit with the categorical prototype and share more attributes
with other categories (Mervis et al., 1976; Murphy &
Medin, 1985; Rosch & Mervis, 1975).

Like the schema effects, the activation of stored knowledge
regarding the prototype (item-typicality) also shapes declara-
tive memories, although in a different way. In fact, the con-
ceptual distinctiveness of atypical items seems to improve
recognition and recollection processes (Alves & Raposo,
2015; Graesser et al., 1980; Vakil et al., 2003; but also see
Schmidt, 1996, Experiment 5, for different results). For in-
stance, using a Remember-Know paradigm, Alves and
Raposo (2015) manipulated item-typicality (i.e., typical vs.
atypical) and the congruence between the item name and the
category (e.g., robin/bird). The results showed that atypical
items (e.g., “penguin” as a “bird”) enhanced overall recogni-
tion and remember (recollection-based) responses.

Mem Cogn


Notably, this item-typicality effect on memory seems to be
similar to the facilitation effect of incongruent items observed
in the categorical learning literature (see Sakamoto & Love,
2004). Following this reasoning, some authors have argued
that items that do not fit the schema seem to recruit the sys-
tems involved in processing new information, which would
not be engaged when the information fits the schema (see
Bonasia et al., 2018; Dudai et al., 2015; Nadel et al., 2012;
Yonelinas et al., 2010). Consequently, these items would be
better retrieved due to the involvement of the episodic system.
In a recent study, Höltje et al. (2019) simultaneously exam-
ined the effects of categorical schema consistency and
prototypicality on recognition memory. Participants were re-
quired to evaluate the consistency between the items and the
category (e.g., consistent pair: doll-toy; inconsistent pair:
mango-toy). The items also varied in their prototypicality
(e.g., high typicality: doll; low typicality: marble). After a
24-h delay, participants recognized better the items that were
consistent with the available schemata and no item-typicality
effects were observed. These results suggest that the effect of
categorical schema congruency seems to be affecting memory
recognition, independently of item typicality.

In sum, the abovementioned findings suggest the influence
of different types of stored conceptual knowledge (i.e., acti-
vation of prior schemata and item-typicality) on memory in
apparently conflicting ways. Schema-consistent information
seems to enhance episodic memory retrieval (Höltje et al.,
2019; van Kesteren, Rijpkema, et al., 2013b; van Kesteren
et al., 2014; but see Mäntylä, 1997, and Sakamoto & Love,
2004, for opposing results). Likewise, information that is not
(or is less) consistent with the schema (e.g., atypical items that
have little fit with their categorical prototype) also seems to
enhance episodic memory retrieval (Alves & Raposo, 2015;
Bonasia et al., 2018; Dudai et al., 2015; but see Höltje et al.,
2019, for different results). In the current paper, we argue that
these differences may result from the nature of the memory
processes involved during recognition.

The current studies

The current studies were designed to examine how two sup-
posedly opposite prior conceptual knowledge effects – cate-
gorical schema consistency and item-typicality – act and in-
teract on both recollective and familiarity-based memories.
Using a single paradigm, we explore how item-typicality
modulates these memory processes in an encoding condition
that activates the categorical schema as compared to a percep-
tual encoding condition. Item-typicality is expected to impact
conscious retrieval because of its relevance for the semantic
organization of categorical processing (Medin et al., 2007;
Rosch and Mervis, 1975). Specifically, atypical items are ex-
pected to enhance Remember responses because they trigger a
specific mechanism involved during novelty encoding

(Bonasia et al., 2018; Dudai et al., 2015). In contrast, the
activation of a categorical congruent schema is expected to
enhance memories based on familiarity for typical items due
to the bypassing of crucial mechanisms activated for novel
information (see Dudai et al., 2015). Therefore, the interaction
between both types of prior conceptual knowledge will be
further inspected.

Experiment 1 explored the described prior conceptual
knowledge effects on both recollection and familiarity pro-
cesses using a Remember-Know paradigm. Experiment 2 rep-
licated Experiment 1 with an additional source memory task,
further looking into the recollective experiences. To our
knowledge, the simultaneous examination of both categorical
encoding-schema activation and item-typicality, as well as
their interaction, on both recollection and familiarity-based
processes constitutes an innovative effort. We expect that this
research might help advance our understanding of how these
two opposing prior conceptual knowledge effects impact the
two different memory processes and whether they interact and
influence each other.

Experiment 1: Exploring the conceptual
knowledge modulation of conscious memory

Experiment 1 examined the role of item-typicality on con-
scious memory processes (i.e., recollection and familiarity)
as a function of the activation of the stored categorical schema
using the Remember-Know paradigm (Tulving, 1985). This
paradigm allows the direct comparison between recollection
and familiarity-based memories within a single task (see
Gardiner, 1988; Rajaram, 1993; Tulving, 1985; but see
Wixted & Squire, 2010). The encoding type modulation
contrasted a categorical condition (i.e., activating prior con-
ceptual abstract knowledge) with a perceptual condition (i.e.,
eliciting perceptual detailed information). The item-typicality
manipulation contrasted typical items (i.e., with a good fit
with their prototype) with atypical ones (i.e., less fitting with
the prototype).



Sample size (N = 38) was determined a priori (G*Power soft-
ware) using as reference the effect size ηp

2 = .14 and a power
of 1-β = 0.95 from a study by Carr et al. (2013), which inves-
tigated the effect of encoding type on conscious recollection.
Forty-six adults, with normal or corrected vision (38 females;
Mage = 19.57 years, SDage = 4.94; Mschooling = 12.36 years,
SDschooling = 1.24) volunteered for this study in exchange for
course credit. Four participants were excluded due to their

Mem Cogn

very low accuracy (less than 30%), one participant did not
finish the task, and three additional participants were
discarded due to a technical problem. The final sample includ-
ed 38 participants.


The stimulus materials for the encoding manipulation
consisted of 96 images of common items, selected from a
normalized database (Souza et al., 2021). The original items
belonged to eight well-studied superordinate categories (from
Santi et al., 2015) from living (fruits, vegetables, mammals,
birds) and non-living (vehicles, clothes, kitchen utensils, and
musical instruments) domains rated on commonly reported
dimensions in normative studies using such type of stimuli
(Souza et al., 2020). Stimuli selection was based on their rat-
ings on item-typicality on a 7-point scale (low: M = 4.65, SD =
0.93; high: M = 6.58, SD = 0.93, t(94)= -13.90, p < .001, dz =
1.42, 90% CI [1.18,1.66]) and controlled for arousal, t(94)= –
1.546, p = .125; valence, t(94) = -1.783, p = .08; and visual
complexity, t(94) = .807, p = .422. A different sample of 48
images (from the same semantic categories) from the same
database was selected for the recognition task and presented
as New items. Old and new items were matched on the same
variables used in the item selection for encoding (all ps >


We used a within-participants design with two encoding
(Categorical vs. Perceptual) and two item-typicality (Typical
vs. Atypical) as independent variables and conscious recollec-
tion judgments (Remember vs. Know vs. Guess) as the de-
pendent variable.

The study followed an ethical protocol approved by the
Ethics Board of the host institution. Participants were in-
formed about the goals and tasks of the study and provided
signed informed consent. The experiment was conducted in
sessions with one to five participants who completed the tasks
in separate cubicles.

During the encoding phase, participants were asked to clas-
sify the 96 images presented in two counterbalanced tasks
(i.e., 48 images without repetitions for each): a perceptual,
episodic-like encoding task (e.g., “how complex is the ob-
ject?”) using a 6-point scale (from 1 – not complex at all to
6 – very complex) and a semantic-like categorical encoding
task with six forced-choice response options (e.g., “is this a:
vegetable/ mammal/ vehicle/ clothes/ musical instruments/
fruit”?). The order of the category options was randomized
across trials. Item-typicality was manipulated in both
encoding tasks, with half of the items being typical and half
atypical (e.g., “dog” for typical and “dolphin” for atypical
exemplars of mammals). All images were presented in a

randomized order within each encoding task. The images
were also counterbalanced between encoding tasks across

After a 20-min interval (plus 5 min of instructions), partic-
ipants were again presented with the 96 images (Old items)
together with 48 new images (New items). Participants were
asked to recognize each image (i.e., Yes-No forced-choice)
and, if the “Yes” response was given, to provide Remember-
Know phenomenological judgments (e.g., “Do I Remember/
Know/Guess1 seeing the image?”) about the recognized im-
ages (see Gardiner, 1988). Detailed instructions are provided
in Appendix A. (Fig. 1)

E-Prime 2.0 software was used to present the stimuli and to
record participants’ responses. To ensure that participants un-
derstood the instructions, the experiment started with a train-
ing phase (five practice trials in each condition), where their
doubts and questions were addressed.

Data analysis

All statistical analyses were conducted with R Version 4.0.2
(R Core Team, 2019).2 The effects of prior conceptual knowl-
edge on Remember-Know-Guess (RKG) judgments were an-
alyzed with Bayesian mixed-effects multinomial regression
models with encoding type, item-typicality, and their interac-
tion as the predictors of interest. For the Bayesian analysis, all
effects with a 95% credible interval that did not include zero
and a probability of direction (pd) value of 97.5% or higher
were considered significant. When appropriate, follow-up
analyses were conducted to obtain simple effects. Additional
analyses of response times (RTs) during encoding and
overall accuracy during the recognition phase were also
conducted. Statistical details for all the analyses can be
found in Appendix B.

Results and discussion

To confirm the influence of item-typicality on recollection and
familiarity-based memories and its interaction with encoding
type, we fitted a model that estimated fixed effects of encoding
condition, item-typicality, and their interaction; by-
participants varying intercepts and by-participant varying
slopes for encoding condition, typicality condition, as well
as the interaction term, including the correlation of these
terms. In addition, we included varying intercepts for items
in the model to preclude the possibility that something unique

1 Guess responses involve a low confidence inferential judgment and an un-
certainty conscious state (Gardiner et al., 1998). This response option was used
to disentangle the Remember versus Know dichotomic judgments.
2 The package tidyverse (Wickham et al., 2019) was used for data processing;
the packages lme4 (Bates et al., 2015), lmerTest (Kuznetsova et al., 2017),
brms (Bürkner, 2017, 2018), and bayestestR (Makowski et al., 2019) were
used for statistical analyses.

Mem Cogn

about a particular item may influence responses to that item
and, therefore, undermine the analysis’s generalizability. This
way, we constructed a model with a maximal random effects
structure justified by the design (see Barr et al., 2013, for
discussion). If the “maximal” model failed to converge or
was found to be overfitted (e.g., a singular fit warning in R),
we first checked whether the model successfully converged
with a random-effects structure for which no slope-intercept
correlation term is specified (to minimize risks of model re-
duction). Only when this did not help did we reduce the model
by removing a random slope that was causing convergence
problems. Throughout the paper, the fixed effects predictors
were deviation coded (–1 = categorical encoding or typical
item; 1 = perceptual encoding and atypical item) to facilitate

the interpretation of main effects in the presence of interac-
tions. If the presence of a significant interaction was
established, follow-up analyses were performed (1) by
looking at the effect of encoding condition for atypical and
typical items separately, and (2) by looking at the effect of
item-typicality for categorical and perceptual encoding types
separately. Specifically, dummy coding of the encoding con-
dition and item-typicality factors were used to obtain simple

Response times during encoding

The time participants took to classify images during the
encoding phase was analyzed using a linear mixed-effects

Fig. 1. Remember-Know paradigm (adapted from Mäntylä, 1997) ma-
nipulated by Encoding Type and Item-typicality (Experiment 1). Note.
The encoding phase comprises two blocks (categorical vs. perceptual),
counterbalanced between participants. In Experiment 1, the response op-
tions of the categorical condition were presented in a randomized order
across trials. The recognition phase includes a conscious recollection

phase in which participants were asked to provide phenomenological
judgments about their memories (Remember/Know/Guess responses).
When the participants respond “yes,” the subsequent slide presents the
R/K/G judgments question. Otherwise, the trial ends with a final blank

Mem Cogn

regression model (similar to Horchak & Garrido, 2020a,
2020b) This analysis was conducted to understand better
how encoding type (categorical vs. perceptual) and item type
(typical vs. atypical) tap into attentional resources required to
perform the classification tasks. The results of the best con-
verging linear mixed-effects regression model showed that
RTs were faster in the perceptual condition (M = 1,388, SD
= 668) than in the categorical condition (M = 1,416, SD =
676). Further statistical details on this analysis can be found
in Appendix B.

Overall recognition

Participants’ overall recognition accuracy was 73%. The
mixed-effects logistic regression model showed that perceptu-
al condition led to higher recognition accuracy. Moreover,
there was a significant increase in recognition accuracy for
atypical items particularly in the categorical encoding condi-
tion. This finding might reflect an advantage in cases when
there is a violation of the prototype during learning (Bonasia
et al., 2018; Sakamoto & Love, 2004), which might have
engaged the systems involved in processing novelty (see
Dudai et al., 2015), namely the episodic one. Of note, percep-
tual condition alone seems to have engaged the episodic sys-
tem, and hence no differences or little gain was observed for
atypical items in this condition. Further statistical details on
this analysis can be found in Appendix B.

Phenomenological judgments of conscious memories

The package brms (Bürkner, 2017, 2018) was used, and spe-
cifically, the categorial function, to analyze the ternary re-
sponse variable “Know” versus “Remember” and “Guess”
with a Bayesian mixed-effects multinomial regression model.3

The brm’s default non-informative priors for fixed (i.e.,
encoding type and item type) and random (i.e., participants
and items) effects were used. A summary of the results is
provided in Fig. 2.

Know versus Remember The results revealed a significant
effect for the encoding factor (estimate = 0.20, 95%
Bayesian credible interval = [0.02; 0.38], pd = 98.37%), indi-
cating that the log-odds of providing a “Remember” response
in the perceptual encoding condition increased relative to the
categorical condition. Results for the item-typicality factor
with a 95% credible interval included zero, but a probability
of direction above a threshold of 97.5% (estimate = 0.16, 95%

Bayesian credible interval = [0.00; 0.32], pd = 97.53%). These
results suggest the advantage of “Remember” responses in the
atypical item condition relative to the typical item condition.

Importantly, there was also evidence for a two-way inter-
action between encoding type and item-typicality (estimate =
− 0.16, 95% Bayesian credible interval = [− 0.32; − 0.05], pd
= 99.60%). A separate Bayesian mixed-effects logistic regres-
sion model showed that encoding type was not a significant
predictor for atypical items (estimate = − 0.03, 95% Bayesian
credible interval = [− 0.21; 0.16], pd = 62.80%). However,
encoding type was a significant predictor for typical items
(estimate = 0.39, 95% Bayesian credible interval = [0.18;
0.62], pd = 100.00%), with a log-odds increase of the
“Remember” responses during the perceptual encoding, as
compared to categorical encoding. When broken up by
encoding factor, the results demonstrated that the effect of
item-typicality for perceptual encoding was not significant
(estimate = − 0.05, 95% Bayesian credible interval = [−
0.23; 0.13], pd = 68.57%). However, there was a reliable
effect of item-typicality for categorical encoding (estimate =
0.36, 95% Bayesian credible interval = [0.14; 0.59], pd =
99.90%), with a log-odds increase of “Remember” responses
when items were atypical rather than typical.

The effects observed for Remember responses mirror the
ones found for the overall recognition accuracy and show that
it was the perceptual encoding condition (but not categorical)
that improved recollection. This finding is consistent with the
selective role of prior schematic knowledge in memories
(Mäntylä, 1997). Although apparently contradicting the pre-
viously documented advantage of schema activation in epi-
sodic retrieval (Liu et al., 2016; Tse et al., 2007; Tse et al.,
2011; van Kesteren et al., 2014; van Kesteren, Beul, et al.,
2013a; van Kesteren, Rijpkema, et al., 2013b; Yamada &
Itsukushima, 2013), such findings should be interpreted with
caution since our encoding conditions did not mirror the usual
schema-consistency manipulations and because the observed
differences on encoding demands render the conditions not
entirely comparable.

Still, the present results of item-typicality main effect rep-
licate the advantage of the atypical items’ distinctiveness in
recollection (Alves & Raposo, 2015). Finally, the advantage
of atypical items in increasing the amount of remember judg-
ments in the categorical encoding reflects the potential activa-
tion of the episodic system given the novelty of atypical items
(see Bonasia et al., 2018; Dudai et al., 2015). This effect is
specific for recollective-based memories.

Know versus Guess The results indicated a significant effect
for the encoding factor (estimate = − 0.52, 95% Bayesian
credible interval = [−0.79; − 0.27], pd = 100%), in that the
log-odds of providing a “Guess” response in the perceptual
encoding condition decreased relative to the categorical con-
dition. The role of the typicality factor for “Guess” responses

3 We opted for Bayesian analysis as the lme4 package (Bates et al., 2015)
currently does not support the analysis that requires the estimation of mixed
multinomial logistic regression models in which the outcome categorical var-
iable has more than two levels.

Mem Cogn

(estimate = − 0.20, 95% Bayesian credible interval = [−0.41;
0.01], pd = 96.57%) was not significant (see Fig. 2). Finally,
the analysis estimated the interaction effect (encoding type by
item-typicality) for “Guess” responses to be non-significant
(estimate = 0.01, 95% Bayesian credible interval = [−0.17;
0.19], pd = 55.10%).

The activation of the stored schema, in the case of the
categorical encoding, led to an increase of “Guess” responses,
which is consistent with the selective role of the schema for
familiarity-based memories (Mäntylä, 1997), likely due to the
bypassing of mechanisms engaged in the processing of nov-
elty (see Dudai et al., 2015). Such a finding is also in line with
previous research showing increased levels of false alarms for
category-consistent memories (De Brigard et al., 2017; Yin
et al., 2019), with typical items increasing guessing.

However, the influence of prior conceptual knowledge on
conscious awareness of declarative memories may have de-
rived from the different demands of the two encoding tasks. It
is well established that Remember and Know responses might
be differently affected by several variables (e.g., level of
processing, Gardiner, 1988; Java & Gregg, 1997; type of
stimuli, Dalla Barba, 1997; Gardiner & Java, 1990;
instructions, McCabe & Geraci, 2009; and aging, Koen &
Yonelinas, 2014; see McCabe et al., 2009, for a review). Of
especial interest is the case of varying attentional demands
(Curran, 2004; Gardiner & Parkin, 1990). For instance,
divided attention during encoding is likely to decrease
remembering accuracy (Dewhurst et al., 2005). In our
categorical encoding task, participants had to monitor
six counterbalanced response options while visually
inspecting the items, thus disproportionally increasing
the attentional resources required for successful task
performance (compared to the perceptual encoding task).

Finally, it is important to replicate Experiment 1,
balancing the level of difficulty and attention demands
involved in both encoding tasks. Moreover, it is crucial
to further validate the Remember judgments as a truly
recollective experience. Therefore, complementary
source memory information could help to discriminate
between general and vivid representations (see Java &
Gregg, 1997; Tulving, 1985).

Experiment 2: Contrasting the encoding type
and item-typicality on conscious recollection
and the quality of recollective experience

Experiment 2 replicates and extends Experiment 1 with a few
modifications. First, the interaction effect of the encoding type
versus item-typicality was examined with a larger sample.
Second, we tried to control the potential impact of executive
processes and attentional resources on memory (Curran, 2004;
Gardiner & Parkin, 1990) by balancing the demands of the
categorical and perceptual encoding tasks. Additionally, we
expanded the number of images presented during the
encoding phase to increase the amount of collected RKG
judgments. Finally, we examined whether Remember judg-
ments actually reflect recollective experience (see Guo et al.,
2006), disentangled from overconfidence effects (Guo et al.,
2006; Hicks et al., 2002). To this end, we included a source
forced-choice identification task (McCabe & Geraci, 2009)
and a source description task for all Remember responses
(Gardiner et al., 1998; Java & Gregg, 1997). As a direct
recollective-based measure (Guo et al., 2006), we expected
that the source memory task’s results would mirror the pattern

Fig. 2. Proportions of “Remember”, “Know,” and “Guess” responses as a function of Item-typicality and Encoding type in Experiment 1. Note. Overall,
there were 1372 responses (52%) for “Remember”, 943 responses (35%) for “Know” and 347 responses (13%) for “Guess”

Mem Cogn

of influence of prior conceptual knowledge observed for
Remember responses.



A sample of 78 participants was determined based on a power
analysis (G*Power) using a medium effect size (d = 0.5;
Cohen, 1988; Miles & Shevlin, 2001) and a power 1-β =
0.80.4 Eighty-seven participants (Mage = 25.09 years, SD =
6.35; Mschooling = 14.77 years, SD = 2.61; 67 female)
volunteered for this study in exchange for course credit. This
experiment followed the same previously approved Ethical
protocol described in Experiment 1. None of the participants
was excluded from the sample.


The stimuli (N = 160) and their selection followed the same
procedure as in Experiment 1. For each encoding task 80
images were used (without repetitions), with 20 images per
category. Their selection was based on mean contrasts of the
ratings provided in a 7-point scale on item-typicality (low: M
= 4.75, SD = 0.01; high: M = 6.39, SD = 0.03, t(158) = -16.14,
p < .001, dz = -1.280, 90% CI [1.10, 1.45] while controlling
for arousal, t(158)= -1.074, p = .284; valence, t(158) = -1.472,
p = .143; aesthetical appeal, t(158)=-1.475, p = .142; and
visual complexity, t(158) = 1.12, p = .264. A different sample
of 106 new images was selected for both phases of the recog-
nition task, with Old and New items matched on the same
criteria as Experiment 1 (all ps > .498).


We used the same paradigm as in Experiment 1 with a few
variations. First, we presented a higher number of items dur-
ing the encoding phase (N = 160). Second, we narrowed the
response options for both encoding tasks. Specifically, for the
categorical encoding, we used a four forced response, this
time with fixed categories (e.g., “is this a: vegetable/ mammal/
vehicle/ clothes”?). Accordingly, the scale for perceptual
encoding ranged from 1 – not complex to 4 – very complex.
The item categories were counterbalanced between encoding
tasks and between participants.

The recognition task consisted of two phases. Recognition
phase 1 (Rec1), with 96 old and 64 new items, and

Recognition phase 2 (Rec2), with 64 old and 42 new items,
different from those used in Rec1. During this phase, and
following a Remember response, a source memory task re-
quired participants to (1) identify in which task the item was
presented (first or second task; i.e., categorical or perceptual;
counterbalanced; McCabe & Geraci, 2009); and (2) provide a
detailed memory description associated with the previous ex-
perience with the item during the encoding phase (adapted
from Gardiner et al., 1998; Java & Gregg, 1997) by writing
which details they remembered (i.e., particular associations
they made, the way they evaluated the images, item order,
etc.) about their first contact with each image (see detailed
instructions in Appendix A). Everything else was kept similar
to Experiment 1.

Results and discussion

Response times during encoding

The analysis followed the same procedures as Experiment 1
(see Appendix B for detailed RTs and accuracy analyses). The
best converging linear mixed-effects regression model dem-
onstrated that, in contrast to Experiment 1, RTs became faster
in the categorical condition (M = 819, SD = 501) than in the
perceptual condition (M = 908, SD = 574).

Overall accuracy of Recognition phase 1

Participants’ overall recognition accuracy was 84%. The
mixed-effects logistic regression model showed similar results
to Experiment 1 (see Appendix B for further details). These
results give further credence to the idea that the perceptual
condition is a better predictor for recognition accuracy
(Mäntylä, 1997). Furthermore, the item-typicality effect was
robust, with atypical items enhancing recognition (as in Alves
& Raposo, 2015). These results are consistent with findings
showing the influence of low-fit prototypical information on
the categorical condition only (see Sakamoto & Love, 2004).

Phenomenological judgments of conscious memories
of Recognition phase 1

The same multilevel model was fit as in Experiment 1. The
summary of results is presented in Fig. 3.

Know versus Remember The mixed-effects multinomial re-
gression analysis revealed a significant effect for the encoding
type factor (estimate = 0.19, 95% Bayesian credible interval =
[0.06; 0.33], pd = 99.70%), indicating that the log-odds of
providing a “Remember” response in the perceptual encoding
condition increased relative to the categorical condition. This
time, the results were also significant for the item-typicality
factor (estimate = 0.17, 95% Bayesian credible interval =

4 None of the previous studies on visual memory using the Remember-Know
paradigm reported an interaction between these conceptual knowledge vari-
ables (i.e., Encoding and Item-typicality) in conscious recollection. Therefore,
in order to provide a reliable sample criterium for such an interaction we used
the standard medium effect size reported in statistical literature (Cohen, 1988;
Miles & Shevlin, 2001).

Mem Cogn

[0.05; 0.30], pd = 99.78%), in that there was an advantage in
proportion of “Remember” responses for atypical items, as
compared to typical. There was also a significant two-way
interaction between encoding type and item-typicality (esti-
mate = − 0.11, 95% Bayesian credible interval = [− 0.19; −
0.03], pd = 99.73%). Follow-up analyses showed that, similar
to Experiment 1, the type of encoding was not a significant
predictor for atypical items (estimate = 0.08, 95% Bayesian
credible interval = [− 0.08; 0.25], pd = 84.47%). However,
encoding type was again a significant predictor for typical
items (estimate = 0.30, 95% Bayesian credible interval =
[0.14; 0.47], pd = 100.00%), with a log-odds increase of the
“Remember” responses during the perceptual encoding, as
compared to categorical encoding. When broken up by
encoding factor, the results were again in line with those ob-
tained in Experiment 1. Specifically, the effect of item-
typicality was not significant for perceptual encoding (esti-
mate = 0.06, 95% Bayesian credible interval = [− 0.07;
0.21], pd = 81.70%). However, it was significant for categor-
ical encoding (estimate = 0.27, 95% Bayesian credible interval
= [0.13; 0.43], pd = 100.00%), with a log-odds increase of
“Remember” responses for atypical items rather than typical
items. Such results clearly corroborate the findings observed
in Experiment 1, this time with a robust item-typicality effect.

Know versus Guess The results showed that encoding type
was a significant predictor of participants’ responses (estimate
= − 0.31, 95% Bayesian credible interval = [−0.45; − 0.17], pd
= 100%), in that the log-odds of providing a “Guess” response
in the perceptual encoding condition decreased relative to cat-
egorical condition. This time, there was also a significant main
effect of item-typicality for “Guess” responses (estimate = −
0.21, 95% Bayesian credible interval = [−0.34; − 0.07], pd =

99.83%), reflecting the fact that atypical items led to less
“Guess” responses than typical items. Finally, and in line with
the results of Experiment 1, there was no evidence for the
interaction between encoding type and item-typicality for
“Guess” responses (estimate = 0.01, 95% Bayesian credible
interval = [−0.09; 0.12], pd = 59.87%).

In sum, categorical encoding improved familiarity-based
memories only, likely due to the economical processing relat-
ed to the activation of a schema, suggesting the recruitment of
the semantic system only. This result is compatible with the
schema effect (e.g., van Kesteren et al., 2010, van Kesteren,
Beul, et al., 2013a, van Kesteren et al., 2014), which seems to
be selective depending on the nature of the memory processes
involved. Perceptive encoding, in contrast, enhanced recollec-
tion (e.g., Mäntylä, 1997). Furthermore, the observed item-
typicality effects were also selective regarding the memory
types, in that they seem to only affect recollection (Alves &
Raposo, 2015; but see Höltje et al., 2019). Finally, item-
typicality improved recollection only for categorically
encoded items. This is arguably the case because atypical
items have a small fit with their categorical prototype, which
might lead to an inconsistency effect that enhances episodic
memories (Alves & Raposo, 2015; Bonasia et al., 2018; Dudai
et al., 2015; Sakamoto & Love, 2004).

Overall accuracy of Recognition phase 2

Participants’ overall recognition accuracy was 77%. The best
converging logistic mixed-effects regression model followed
the same steps as in Recognition Phase 1. The results are
essentially the same as those observed in both previous recog-
nition results, presenting the expected main effects and

Fig. 3. Proportions of “Remember”, “Know,” and “Guess” responses as a function Item-typicality and Encoding type in Experiment 2 (Rec1). Note.
Overall, there were 4,603 responses (65%) for “Remember,” 1,742 responses (25%) for “Know,” and 711 responses (10%) for “Guess”

Mem Cogn

confirming the interaction effect observed before (see
Appendix B for further details on this analysis).

Phenomenological judgments of conscious memories
of Recognition phase 2

The modeling followed the same steps indicated in
Experiment 1. The summary of results is presented in Fig. 4.

The results from Rec2 replicate the item-typicality effect
for Remember, with more Remember responses for atypical
items (see summary of results in Appendix B). For Guess
responses, the expected encoding type effect was observed,
with more guessing for categorical encoding, compared to
perceptual encoding. At the same time, we observed a signif-
icant decrease in the amount of Remember responses (47%) as
compared to 52% and 65% in Experiment 1 and Rec1, respec-
tively, which might have prevented us from observing the
exact same pattern of results found in Experiment 1 and in
Rec1. It is possible that participants became less committed or
motivated for the task in this last phase and tried to avoid the
burden of giving descriptive source responses. Likewise, this
second memory test might have reactivated traces from previ-
ous learning (see Antony et al., 2017; Potts & Shanks, 2012).

Source memory

The source information tasks in Rec2 inspected the source-
type responses as indicators of the detailed and vivid memo-
ries regarding the item and self-related experience with the
item during encoding (adapted from Gardiner et al., 1998).
Below, we present the results for source accuracy in the
task-order identification and the source description question.

Source accuracy

Overall, 2,064 source-type responses associated with
Remember responses were analyzed. False recognition (i.e.,
New items evaluated as Old) was approximately 3% (54 re-
sponses). The responses associated with correct recognition
(97%; 2,010 responses) were the focus of the following anal-
ysis. Participants were highly accurate in identifying in which
task the items were presented (M = .92, SD = .26). More than
half (.54) of the correctly identified items in the task-order
question were presented in the perceptual condition and the
remaining (.46) in the categorical condition. Likewise, more
than half of these items (.56) were atypical, and the remaining
(.44) were typical.

The analysis of the prior conceptual knowledge effects was
conducted using a repeated-measures Anova (2 Encoding and
2 Item-typicality) based on the absolute frequencies of each
correct response for each condition per participant.
Bonferroni’s pairwise adjustment was used to contrast condi-
tions. Post hoc analysis was run using t-tests to inspect the
direction of interaction effects. Responses from 77 partici-
pants were included in this analysis, given that a technical
problem led to the loss of ten participants. The results showed
a main effect of encoding, F(1, 76) = 6.416, p = .013, ηp

2 =
.08, 90% CI [.01, .18] with greater accuracy for perceptual (M
= 6.01, SE = .46) than categorical encoding (M = 5.10, SE =
.41), and a main effect of item-typicality, F(1, 76) = 28.861, p
< .001, ηp

2 = .275, 90% CI [.14, .40] with higher accuracy for
atypical items (M = 6.22, SE = .43) than for typical ones (M =
4.89, SE = .40). The interaction effect was also significant,
F(1, 76) = 10.353, p = .002, ηp

2 = .120, 90% CI [.03, .24],
with increased accuracy of source task for atypical items
encoded in categorical conditions (Atypical: M = 6.19, SE =

Fig. 4. Proportions of “Remember”, “Know,” and “Guess” responses as a function Item-typicality and Encoding type in Experiment 2 (Rec2). Note.
Overall, there were 2,010 responses (47%) for “Remember,” 1,686 responses (39%) for “Know,” and 605 responses (14%) for “Guess”

Mem Cogn

.47, Typical: M = 4.01, SE = .41; t(76) = -6.642, p < .001, dz =
1.07, 90% CI [0.766, 1.368]). No difference was observed for
perceptual encoding, t(76) = -1.222, p = .226.

Source descriptions

The 2,010 source descriptions related to correct
Remember responses were analyzed by two trained
judges based on previously established categories (see
Gardiner, 1988; Gardiner et al., 1998). The a priori
established categories and results of source description
are presented in Table 1. The high occurrence of “Item
evaluation” and “Personal Associations” categories of
source information reaffirms that detailed remembering
was strongly related to the experience of recollection,
being a marker of episodic-like processing.

Regarding prior conceptual knowledge modulation on
source description, distinct rmAnova including 2 encoding
type and 2 item-typicality as within-participant variables
were calculated considering the proportions of source de-
scriptions in item evaluation and personal association (the
categories that were more frequent). An item-typicality
main effect was observed for item evaluation, F(1, 84) =
11.59, p < .001, ηp

2 = .121, 90% CI [.03, .23] and for
personal association, F(1,84) = 10.07, p = .002, ηp

2 =
.107, 90% CI [.02, .21], whereby atypical items prompted
higher item evaluation (MAtypical =.14, SE= .01; MTypical =
.01, SE = .01) and personal associations (MAtypical = 0.12,
SE = .01; MTypical = .078, SE = .01) than typical ones.
Moreover, there was no encoding type effect or interaction
with item-typicality. In other words, distinctive exemplars
of categories seem to be directly related to the enhance-
ment of particular details related to the recollective expe-
rience during source descriptions.

General discussion

The present studies aimed to systematically investigate con-
tradictory findings regarding the influence of prior conceptual
knowledge (see van Kesteren et al., 2010, 2014; but for
opposing results, see Mäntylä, 1997; Sakamoto & Love,
2004) on memory, using the classic Remember-Know para-
digm (Tulving, 1985). To this end, two experiments explored
the idea that item-typicality effects may differentially affect
recollective and familiarity-based memories, particularly as a
function of the availability of a stored schema. Our main pre-
diction was that atypical items would selectively enhance rec-
ollection due to the activation of specific mechanisms
supporting novelty processing (Bonasia et al., 2018; Dudai
et al., 2015). Moreover, we explored how item-typicality
could impact conscious memory processes as a function of
encoding types by comparing recollection and familiarity-
based memories for typical or less typical items depending
on whether they were encoded categorically (schema activa-
tion) or perceptually (non-schematic). Experiment 2 replicated
and extended Experiment 1 by including a second recognition
phase with a source memory task. It was predicted that the
pattern of source accuracy responses would be similar to the
one observed for remember responses regarding the prior con-
ceptual knowledge interaction effect, since both reflect the
engagement in recollection processes.

Overall, the results showed enhanced recognition accuracy
for atypical items in both experiments, in line with previous
evidence on the facilitation effect of atypical items for episod-
ic retrieval (Alves & Raposo, 2015; Graesser et al., 1980;
although not gathering consensus in memory studies, see
Schmidt, 1996).

Regarding the phenomenological judgments, we observed
the selective advantage of perceptual encoding on recollection
as reported by Mäntylä (1997). Notably, as expected, item-

Table 1. Descriptive information (category names, definition, and examples) and percentages for each descriptive response category from the Source
Description task

Code Category Description (%)

IE Item evaluation When the response refers to the assessment of the item in the task, for example, “evaluated as complex”; “the item
was in the animals’ category”


ASS Private/personal

When the response refers to some specific experience related to the item representation, for example, “associated
with the bus that I take to go to the university”; “I found it funny”


AP Item appearance When the response refers to the appearance of the item, for example, “I found the color different”; “Size and position
were unusual”


M Mistake When the response was restricted only to number 5 (key used to end response); when the text was not readable (e.g.,


TP Task position When the response refers to the position of the item in the task, for example, “I remember coming after a monkey”;
“Appeared in training”


TE Task event When the response refers to an event related to the presentation of the item during encoding, for example, “I called
the experimenter at the time”; “I dropped a pen when I saw the image”


K Know When the answer did not indicate details of the recall, for example, “nothing in particular”; “do not know” 1

Note. The column (%) corresponds to the percentage of response types considering the amount of remembering

Mem Cogn

typicality differentially modulated recollection by the advan-
tage of atypical information in selectively increasing
recollection-based memories, as compared to low-
confidence familiarity-based memories. These results corrob-
orate previous findings regarding the advantage of distinctive-
ness in promoting recollection-based memories (Alves &
Raposo, 2015; Rajaram, 1998; Watier & Collin, 2012). The
present findings also indicate that the improvement of
recollection-based memories due to the low typicality of the
materials may reflect the recruitment of the episodic system
when processing information that is novel or violates the
stored prototypical representation (see Bonasia et al., 2018;
Dudai et al., 2015; Yonelinas et al., 2010), and is probably
related to hippocampal involvement (Nadel & Moscovitch,
1997; Sekeres et al., 2018; Yonelinas et al., 2010, 2019).
The event-related potential (ERP) data reported by Höltje
et al. (2019) also showed increased N400 amplitude according
to the lower fit of the items with the categorical schema
encoded (i.e., inconsistent > atypical > typical). This finding
supports the idea that less typical information is less consistent
(i.e., violating expectations) with the activated categorical
schema (prototype) than highly typical information (see
Bonasia et al., 2018; Dudai et al., 2015).

Furthermore, typical items increased familiarity-based
judgments associated with low confidence and vagueness.
The activation of typical items for familiarity-based responses
is only partially in line with the schema-consistency advantage
hypothesis (van Kesteren et al., 2010; van Kesteren, Beul,
et al., 2013a), an advantage that was not observed for
recollective memories. This finding suggests that the semantic
system alone might be engaged in bypassing the episodic
system (Dudai et al., 2015). Moreover, it supports the idea
that if the semanticized information is sufficient in a given
situation (or in the absence of distinctive and vivid infor-
mation), then the cortically instantiated abstract version of
memory will be recruited (Sekeres et al., 2017, 2018; van
Kesteren & Meeter, 2020). The simultaneous observation
of both schema and typicality effects helps to clarify prior
conflicting findings reported in the literature (Alves &
Raposo, 2015; Höltje et al., 2019; van Kesteren,
Rijpkema, et al., 2013b), and suggests that these appar-
ently contradictory effects coexist but act selectively upon
either type of memory processes.

Few studies have simultaneously explored these memory
conceptual knowledge effects in the context of previously
stored categories, and report contradictory results (Alves &
Raposo, 2015; Höltje et al., 2019). For example, our findings
differ from those observed by Höltje et al. (2019), which re-
port the schema advantage and the absence of typicality ef-
fects in memory recognition. However, these differences
might result from relevant procedural differences, namely dis-
tinct tasks and different retention intervals. For instance, rec-
ognition tasks (as those used in Höltje et al., 2019) are known

to involve both recollective and familiarity-based processes at
the same time, which is not the case of the different conscious
judgments required in the Remember-Know task (Gardiner,
1988; Yonelinas et al., 2010). Moreover, larger retention times
(as those in Höltje et al., 2019), including sleeping, are known
to improve consolidation processes (semanticization) due to
reactivation of hippocampal structures and cortical regions
(Dudai et al., 2015; Sekeres et al., 2017) and may enhance
prior conceptual knowledge effects (as in van Kesteren et al.,

Interestingly, when both types of prior conceptual knowl-
edge interacted, atypical items boosted the probability of pro-
viding Remember responses only for the categorical condi-
tion. This finding suggests that atypical information activates
episodic content, which was likely already recruited in the
perceptual condition. Thus, no further gain associated with
the recruitment of the episodic system was observed for per-
ceptually encoded items. This interaction effect is noteworthy
as it points to the importance of the specific stimuli used rather
than the learning and encoding settings alone (see Dudai et al.,

Together, these results suggest that distinct memory types
might be co-activated and implicated in learning, with their
available representations interacting according to materials,
consolidation times, environmental demands, or behavioral
requirements (see Nadel, 2020; Nadel et al., 2012; Renoult
et al., 2019). Additionally, the results provided by the
source-type task and source descriptions showed that
recollection-based memories are influenced by distinctive-
ness, indicating that the overlap between the source judgments
and the actual remember judgments is neither by chance nor
motivated by overconfidence feelings (see Guo et al., 2006;
Hicks et al., 2002).

However, there are some issues to be addressed in future
work. First, the differences between categorical versus percep-
tual conditions might reflect different task demands involved
in each encoding. Moreover, our effort to balance both
encoding conditions in Experiment 2 was not entirely success-
ful. Secondly, the inspection of response times during
encoding in Experiment 1 showed that participants were over-
all faster in the perceptual condition, while in Experiment 2,
the reverse was observed. However, this had no significant
influence on the results during the recognition phase, which
were consistent across experiments. Therefore, the observed
differences in RTs during the encoding phase are unlikely to
explain the recognition phase results since the overall recog-
nition accuracy was always higher for perceptual encoding
than for categorical encoding. Finally, previous studies on
schema-congruency usually use word/sentence stimuli (e.g.,
Höltje et al., 2019; van Kesteren et al., 2014), while our stud-
ies examined abstract knowledge using visual materials. Since
words are more abstract stimuli than images, they may present
a stronger influence of semantic activation in facilitating

Mem Cogn

retrieval. Therefore, our results should be replicated with dif-
ferent stimuli.


The overall role of semantic knowledge in cognitive processes
has been repeatedly reported in clinical and healthy samples
(Nadel et al., 2012; Souza et al., 2016; Toichi & Kamio, 2003;
van Kesteren, Rijpkema, et al., 2013b). However, prior con-
ceptual knowledge, such as schemata and prototypical infor-
mation, both semantic in nature, seem to influence learning
differently (e.g., Alves & Raposo, 2015; Höltje et al., 2019;
Mäntylä, 1997; Sakamoto & Love, 2004; van Kesteren, Beul,
et al., 2013a). Our results provide important insights into the
selective influence of prior conceptual knowledge in both
recollective- and familiarity-based memories when a schema
is available during learning and/or when it is violated.
Notably, recollection was influenced by low item-typicality
and by whether the categorical schema was activated or not.
These findings circumscribe the general advantage of congru-
ent schemas because this advantage was observed for
familiarity-base memories only. Finally, the role of atypical
information was also reiterated for vivid recollection-based
memories, particularly when the categorical schema was acti-
vated during encoding.


Detailed instruction of RKG judgments

In this phase, you will be presented with one image at a time,
and your task is to say if you HAVE SEEN these images
BEFORE, during the first part of this session.

Press “S” (yes) if you have seen the image before.
Press “N” (no) if you have not seen the image.
When you claim to have seen the image before, you will

then be asked to ASSESS YOUR recall experience, as:
REMEMBER: This answer implies the ability to become

aware of some aspects of what happened or what was experi-
enced when the image was presented. In other words, press

REMEMBER when details related to remembering seeing
the image comes to mind as a particular association (i.e.,
something more personal when you saw the item), the appear-
ance of the image itself, its position in the task (i.e., what came
before and after the image), or something that happened when
you saw that image.

KNOW: This answer implies knowing that the image was
presented previously in this task, but you cannot consciously
remember anything about its specific occurrence. In other words,
press KNOW when you are sure that the image was presented,
but you cannot evoke any particular details about its occurrence.

GUESS: This answer implies that when you answered
“yes” previously, you tried to guess that you saw the image
before. In other words, just press GUESS when your answer
“yes” was really guessing, with very little confidence.

For a better understanding of the task, here are some

REMEMBER: If you were asked about the last film you
saw, your answer would be based on a memory like “I remem-
ber”; which requires becoming aware of specific details of
past experience.

KNOW: When you recognize someone on the street, but
you do not remember who the person is or where you know
the person from, you can only experience a feeling of famil-
iarity without becoming aware of a particular event or experi-
ence with the person in question.

GUESS: When you say that you remember someone, but
you are just trying to guess that you know him/her without
much confidence.

If you have any QUESTIONS about how to classify the types
of memory you have, please ask the EXPERIMENTER to
EXPLAIN. A training phase will help you to understand the task


Experiment 1

Response times (RTs) during Encoding

For this analysis, trials with RTs faster than 300 ms or slower
than 3,000 ms were excluded. Furthermore, trials with RTs
2.5 SDs or higher from the relevant condition means were
discarded. Finally, RTs were standardized by subtracting the
mean and dividing by the SD for analysis. The model was
estimated using ML and BOBYQA optimizer; with encoding
condition and typicality condition and their interaction con-
sidered as fixed effects, by-participant and by-item random
intercepts, and a by-participant slope for encoding condition
and typicality condition. The results of the best converging
linear mixed-effects regression model showed that there was
a main effect of encoding condition (estimate = − 0.05, SE =
0.03, t = − 2.04, p =.048, 95% CI [−0.10, 0.00]) in that re-
sponse times were faster in the perceptual condition (M =
1,388, SD = 668) compared to categorical condition (M =
1,416, SD = 676). There was also a main effect of typicality
condition (estimate = 0.08, SE = 0.02, t = 3.36, p =.001, 95%
CI [0.03, 0.12]) in that response times were slower in the
atypical condition (M = 1,445, SD = 676) than in the typical

Mem Cogn

condition (M = 1,361, SD = 666). Finally, there was no evi-
dence for an interaction between the two factors (estimate = −
0.01, SE = 0.01, t = − 0.68, p =.495, 95% CI [− 0.04, 0.02]).

Overall accuracy of Recognition

The binary response variable “Incorrect Response” versus
“Correct Response” was analyzed with a mixed-effects logis-
tic regression model, using the lme4 package (Bates et al.,
2015), and specifically the binomial (link = “logit”) function.
The best converging model, estimated using ML and
BOBYQA optimizer, included encoding condition (categori-
cal vs. perceptual) and typicality condition (typical item vs.
atypical item) and their interaction as fixed effects; by-
participant and by-item random intercepts, and by-
participant slopes for encoding condition and typicality con-
dition as random effects. The results of the mixed-effects lo-
gistic regression model showed a significant main effect of
encoding condition (estimate = 0.54, SE = 0.13, z = 4.25, p
< .001, 95% CI [0.29, 0.78]), with more correct responses in
the perceptual condition (M = 0.80, SD = 0.40), compared to
categorical condition (M = 0.66, SD = 0.47). There was no
main effect of typicality condition (estimate = 0.12, SE = 0.11,
z = − 1.04, p = .298, 95% CI [− 0.10, 0.33]). Furthermore,
there was a significant interaction between the two factors
(estimate = − 0.17, SE = 0.05, z = − 3.37, p = .001, 95% CI
[− 0.27, − 0.07]). When broken up by the encoding type fac-
tor, follow-up comparisons showed that atypical items (M =
0.71, SD = 0.46) were recognized more accurately than typical
items (M = 0.62, SD = 0.49) during the categorical encoding
(estimate = 0.29, = 0.12, z = 2.42, p = .015, 95% CI [0.05,
0.52]). However, there was almost no difference in recogni-
tion rates for atypical (M = 0.79, SD = 0.40) and typical (M =
0.80, SD = 0.40) items during the perceptual encoding (esti-
mate = − 0.05, SE = 0.12, z = − 0.43, p =.666, 95% CI [− 0.30,
0.19]). Finally, the segregation of the data by item-typicality
revealed that participants were more accurate to recognize
typical items during the perceptual (M = 0.80, SD = 0.40)
encoding than during the categorical (M = 0.62, SD = 0.49)
encoding (estimate = 0.71, SE = 0.14, z = 5.20, p < .001, 95%
CI [0.44, 0.97]). Similarly, participants were also more accu-
rate to recognize atypical items during the perceptual (M =
0.79, SD = 0.40) encoding than during the categorical (M =
0.71, SD = 0.46) encoding (estimate = 0.37, SE = 0.14, z =
2.69, p = .007, 95% CI [0.10, 0.63]).

Experiment 2

Response times (RTs) during encoding

Similar to Experiment 1, we analyzed the time participants
took to classify typical and atypical images during the
encoding phase using a linear mixed-effects regression model.

Trimming procedures related to outlier treatment and RT stan-
dardization were the same as in Experiment 1.

This model was estimated using ML and BOBYQA opti-
mizer; with encoding condition and typicality condition and
their interaction considered as fixed effects, by-participant and
by-item random intercepts, and a by-participant slope for
encoding condition and typicality condition). The best con-
verging linear mixed-effects regression model demonstrated a
main effect of encoding type (estimate = 0.09, SE = 0.02, t =
4.48, p < .001, 95% CI [0.05, 0.13]) in that response times
were overall slower in the perceptual condition (M = 908, SD
= 574) compared to categorical condition (M = 819, SD =
501). There was also a main effect of item-typicality (estimate
= 0.05, SE = 0.01, t = 4.48, p < .001, 95% CI [0.03, 0.17]) in
that response times were slower in the atypical condition (M =
886, SD = 552) compared to the typical condition (M = 841,
SD = 526). However, there was a strong evidence for an in-
teraction between the two factors (estimate = − 0.06, SE =
0.01, t = − 6.51, p < .001, 95% CI [− 0.08, − 0.04]). Follow-
up analyses with a dummy-coded item-typicality factor
showed that participants took significantly more time to judge
typical items during the perceptual (M = 914, SD = 578)
encoding than during the categorical (M = 770, SD = 460)
encoding (estimate = 0.15, SE = 0.02, t = 6.69, p < .001,
95% CI [0.11, 0.19]). Interestingly, however, the same pattern
did not hold true for atypical items, in that participants did not
significantly differ in their response times during the percep-
tual (M = 903, SD = 569) encoding, compared to categorical
(M = 870, SD = 535) encoding (estimate = 0.04, SE = 0.01, t =
1.56, p =.122, 95% CI [− 0.01, 0.08]). When broken up by the
encoding type factor, follow-up comparisons showed that
atypical items (M = 870, SD = 535) were responded to more
slowly than typical items (M = 770, SD = 460) during the
categorical encoding (estimate = 0.11, SE = 0.12, z = 7.55, p
< .001, 95% CI [0.08, 0.14]). However, the difference in re-
sponse times for atypical (M = 903, SD = 569) and typical (M
= 914, SD = 578) items during the perceptual encoding was
negligible (estimate = − 0.01, SE = 0.01, t = − 0.44, p =.658,
95% CI [− 0.03, 0.02]).

Overall accuracy of Recognition phase 1

These analyses followed similar procedures from Experiment
1. In the present analysis, the lme4 package (Bates et al., 2015)
was applied, and specifically, the binomial (link = “logit”)
function was used to analyze the binary response variable
“Incorrect Response” versus “Correct Response” with a
mixed-effects logistic regression model. The best converging
model (estimated using ML and BOBYQA optimizer) includ-
ed encoding condition (categorical vs. perceptual) and item-
typicality condition (typical item vs. atypical item) and their
interaction as fixed effects; by-participant and by-item random

Mem Cogn

intercepts, and by-participant slopes for encoding condition
and item-typicality condition as random effects.

The results of the mixed-effects logistic regression model
showed a significant main effect of encoding type (estimate =
0.43, SE = 0.08, z = 5.61, p < .001, 95% CI [0.28, 0.57]) with
more correct responses in the perceptual condition (M = 0.88,
SD = 0.32) compared to categorical condition (M = 0.80, SD =
0.40). This time, there was a reliable main effect of item-
typicality (estimate = 0.23, SE = 0.06, z = 3.66, p < .001,
95% CI [0.11, 0.35]), reflecting the fact that participants’ ac-
curacy was higher when they processed atypical items (M =
0.87, SD = 0.34) rather than typical items (M = 0.82, SD =
0.39). Finally, there was also a significant interaction between
the two factors (estimate = − 0.10, SE = 0.04, z = − 2.84, p =
.004, 95% CI [− 0.17, − 0.03]). When broken up by the
encoding type factor, follow-up comparisons showed that
atypical items (M = 0.85, SD = 0.36) were recognized more
accurately than typical items (M = 0.76, SD = 0.43) during the
categorical encoding (estimate = 0.33, SE = 0.07, z = 4.89, p <
.001, 95% CI [0.20, 0.46]). However, and similar to
Experiment 1, the differences in recognition rates were not
statistically different for atypical (M = 0.89, SD = 0.31) and
typical (M = 0.87, SD = 0.33) items during the perceptual
encoding (estimate = 0.13, SE = 0.8, z = 1.65, p =.098, 95%
CI [− 0.02, 0.27]). Finally, the segregation of the data by item-
typicality revealed that participants were more accurate to rec-
ognize typical items during the perceptual (M = 0.87, SD =
0.33) encoding than during the categorical (M = 0.76, SD =
0.43) encoding (estimate = 0.53, SE = 0.08, z = 6.44, p < .001,
95% CI [0.37, 0.69]). In a similar way, participants were also
more accurate to recognize atypical items during the percep-
tual (M = 0.89, SD = 0.31) encoding than during the categor-
ical (M = 0.85, SD = 0.36) encoding (estimate = 0.32, SE =
0.09, z = 3.76, p <.001, 95% CI [0.15, 0.49]).

Overall accuracy of Recognition phase 2

The same statistical procedures as in Experiment 2 were used.
The best converging logistic mixed-effects regression model
to analyze error rates was the same as in Recognition Phase 1.
The results showed a significant main effect of encoding type
(estimate = 0.36, SE = 0.07, z = 4.89, p < .001, 95% CI [0.22,
0.50]) with more correct responses in the perceptual condition
(M = 0.82, SD = 0.38) compared to categorical condition (M =
0.72, SD = 0.45). Similarly, there was a significant main effect
of typicality condition (estimate = 0.23, SE = 0.06, z = 3.45, p
< .001, 95% CI [0.10, 0.36]), with more correct responses for
atypical items (M = 0.80, SD = 0.40) than typical items (M =
0.74, SD = 0.44). Finally, there was also evidence for a sig-
nificant interaction between the two factors (estimate = − 0.15,
SE = 0.04, z = − 3.98, p < .001, 95% CI [− 0.23, − 0.08]).
When broken up by the encoding type factor, follow-up com-
parisons showed that atypical items (M = 0.78, SD = 0.42)

were recognized more accurately than typical items (M = 0.67,
SD = 0.47) during the categorical encoding (estimate = 0.38,
SE = 0.07, z = 5.20, p < .001, 95% CI [0.24, 0.53]). Again, the
differences in recognition rates were negligible for atypical (M
= 0.83, SD = 0.38) and typical (M = 0.82, SD = 0.38) items
during the perceptual encoding (estimate = 0.07, SE = 0.8, z =
0.93, p =.352, 95% CI [− 0.08, 0.23]). Finally, and in line with
previous results, the segregation of the data by item-typicality
revealed that participants were more accurate to recognize
typical items during the perceptual (M = 0.82, SD = 0.38)
encoding than during the categorical (M = 0.67, SD = 0.47)
encoding (estimate = 0.51, SE = 0.08, z = 6.32, p < .001, 95%
CI [0.36, 0.67]). Similarly, participants were also more accu-
rate to recognize atypical items during the perceptual (M =
0.83, SD = 0.38) encoding than during the categorical (M =
0.78, SD = 0.42) encoding (estimate = 0.20, SE = 0.09, z =
2.40, p = .016, 95% CI [0.04, 0.37]).

Phenomenological judgments of conscious memories
of Recognition phase 2

Know versus Remember The mixed-effects multinomial re-
gression analysis demonstrated that, unlike before, there was
no significant effect of encoding type factor (estimate = 0.11,
95% Bayesian credible interval = [−0.00; 0.23], pd = 97.20%).
However, there was a significant main effect of item-typicality
factor (estimate = 0.24, 95% Bayesian credible interval =
[0.11; 0.38], pd = 99.97%), in that there again was an advan-
tage in proportion of “Remember” responses for atypical
items relative to typical ones. Unlike before, there was no
evidence for an interaction between the two factors (estimate
= − 0.06, 95% Bayesian credible interval = [− 0.15; 0.03], pd
= 91.15%).

Know versus Guess The mixed-effects multinomial regression
analysis showed that encoding type was a significant predictor
of participants’ responses (estimate = − 0.21, 95% Bayesian
credible interval = [−0.34; − 0.08], pd = 99.90%), in that the
log-odds of providing a “Guess” response in the perceptual
encoding condition decreased relative to categorical condi-
tion. The evidence for the effect of item-typicality factor for
“Guess” responses was minimal in that the probability of di-
rection was above 97.5% but a 95% credible interval included
zero (estimate = − 0.15, 95% Bayesian credible interval =
[−0.30; − 0.00], pd = 97.87%). Most interestingly, however,
the analysis showed that this time there was a reliable evi-
dence for the interaction between encoding type and item-
typicality for “Guess” responses (estimate = 0.19, 95%
Bayesian credible interval = [0.08; 0.31], pd = 99.95%). A
separate Bayesian mixed-effects logistic regression model
with a dummy-coded item-typicality factor demonstrated that
the type of encoding was not a significant predictor for atyp-
ical items (estimate = − 0.01, 95% Bayesian credible interval

Mem Cogn

= [− 0.18; 0.17], pd = 84.47%). However, encoding type was a
significant predictor for typical items (estimate = − 0.40, 95%
Bayesian credible interval = [− 0.57; − 0.24], pd = 100.00%),
with a log-odds decrease of the “Guess” responses during the
perceptual encoding, as compared to categorical encoding.
When broken up by encoding factor, the results showed that
the effect of item-typicality was not significant for perceptual
encoding (estimate = 0.04, 95% Bayesian credible interval =
[− 0.15; 0.24], pd = 65.80%). However, it was significant for
categorical encoding (estimate = − 0.35, 95% Bayesian cred-
ible interval = [− 0.53; − 0.18], pd = 100.00%), with a log-
odds decrease of “Guess” responses for atypical items rather
than typical items.

Acknowledgements The authors would like to thank some of their stu-
dents for their assistance in recruiting participants, namely, Maria Ana
Gonçalves, Ana Marta Carvalho, Sara Pimenta, and Catarina Santos, for
their valuable assistance in this research. We would also like to thank the
experts who provided comments and suggestions that helped improve the
current paper.

Open practice statement None of the experiments were pre-registered
in an open-source database. The main data are available online at the
Open Science Framework, link <

Funding This research was supported by the Fundação para a Ciência e
Tecnologia, Portugal, with grants awarded to CAS [PD/BD/128249/
2016], OVH [SFRH/BPD/115533/2016], and JCC [FCT, I.P.: Norma
Transitória DL57/2016/CP1439/CT02]. The funders had no role in study
design, data collection and analysis, or preparation of the manuscript.


Alves, M., & Raposo, A. (2015). Is it a bird? Differential effects of
concept typicality on semantic memory and episodic recollection.
Revista Portuguesa de Psicologia, 44, 65–79.

Antony, J. W., Ferreira, C. S., Norman, K. A., & Wimber, M. (2017).
Retrieval as a Fast Route to Memory Consolidation. Trends in
Cognitive Sciences, 21(8), 573–576.

Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random effects
structure for confirmatory hypothesis testing: Keep it maximal.
Journal of Memory and Language, 68(3), 255–278. https://doi.

Bastin, C., Besson, G., Simon, J., Delhaye, E., Geurten, M., Willems, S.,
et al. (2019). An integrative memory model of recollection and
familiarity to understand memory deficits. Behav. Brain Sci. 42:

Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fitting Linear
Mixed-Effects Models Using lme4. Journal of Statistical Software,
67(1), 1-48.

Bonasia, K., Sekeres, M. J., Gilboa, A., Grady, C. L., Winocur, G., &
Moscovitch, M. (2018). Prior knowledge modulates the neural sub-
strates of encoding and retrieving naturalistic events at short and
long delays. Neurobiology of Learning and Memory, 153, 26–39.

Bürkner, P.-C. (2017). brms: An R package for Bayesian multilevel
models using Stan. Journal of Statistical Software, 80(1). https://

Bürkner, P.-C. (2018). Advanced Bayesian multilevel modeling with the
R package brms. The R Journal, 10, 395–411.

Carr, V. A., Engel, S. A., & Knowlton, B. J. (2013). Top-down modula-
tion of hippocampal encoding activity as measured by high-
resolution functional MRI. Neuropsychologia, 51(10), 1829–1837.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences
(2nd ed.). Erlbaum.

Curran, T. (2004). Effects of attention and confidence on the hypothe-
sized ERP correlates of recollection and familiarity.
Neuropsychologia, 42(8), 1088–1106.

Czernochowski, D., Brinkmann, M., Mecklinger, A. & Johansson, M.
(2004). When binding matters: an ERP analysis of the development
of recollection and familiarity. In: A. Mecklinger, H. Zimmer, U.
Linderberger (Eds.). Bound in Memory. Insights from Behavioral
and Neuropsychological Studies, Aachen: Shaker Verlag, S. pp
93-128. Available at
2004_.pdf. Accessed Jul 2021

Dalla Barba, G. (1997). Recognition memory and recollective experience
in Alzheimer’s Disease. Memory, 5(6), 657–672.

De Brigard, F., Brady, T. F., Ruzic, L., & Schacter, D. L. (2017).
Tracking the emergence of memories: A category-learning para-
digm to explore schema-driven recognition. Memory & Cognition,
45(1), 105–120.

Dewhurst, S., Barry, C., & Holmes, S. (2005). Exploring the false recog-
nition of category exemplars: Effects of divided attention and ex-
plicit generation. European Journal of Cognitive Psychology, 17(6),

Dudai, Y., Karni, A., & Born, J. (2015). The Consolidation and
Transformation of Memory. Neuron, 88(1), 20–32.

Gardiner, J. M. (1988). Functional aspects of recollective experience.
Memory & Cognition, 16(4), 309–313.

Gardiner, J. M., & Java, R. I. (1990). Recollective experience in word and
nonword recognition. Memory & Cognition, 18(1), 23–30. https://

Gardiner, J. M., & Parkin, A. J. (1990). Attention and recollective expe-
rience in recognition memory. Memory & Cognition, 18(6), 579–

Gardiner, J. M., Ramponi, C., & Richardson-Klavehn, A. (1998).
Experiences of remembering, knowing, and guessing.
Consciousness and Cognition, 7(1), 1–26.

Ghosh, V. E., & Gilboa, A. (2014). What is a memory schema? A his-
torical perspective on current neuroscience literature.
Neuropsychologia, 53, 104–114.

Graesser, A. C., Woll, S. B., Kowalski, D. J., & Smith, D. A. (1980).
Memory for typical and atypical actions in scripted activities.
Journal of Experimental Psychology: Human Learning &
Memory, 6(5), 503–515.

Guo, C., Duan, L., Li, W., & Paller, K. A. (2006). Distinguishing source
memory and item memory: Brain potentials at encoding and retriev-
al. Brain Research, 1118(1), 142–154.

Harris, H. D., & Rehder, B. (2006). Modeling category learning with
exemplars and prior knowledge. In R. Sun & N. Miyake (Eds.),

Mem Cogn

Proceedings of the 28th Annual Conference of the Cognitive Science
Society (pp. 1440-1445). Erlbaum.

Hicks, J. L., Marsh, R. L., & Ritschel, L. (2002). The role of recollection
and partial information in source monitoring. Journal of
Experimental Psychology: Learning, Memory, and Cognition,
28(3), 503–508.

Höltje, G., Lubahn, B., & Mecklinger, A. (2019). The congruent, the
incongruent, and the unexpected: Event-related potentials unveil
the processes involved in schematic encoding. Neuropsychologia,
131, 285–293.

Horchak, O. V., & Garrido, M. V. (2020a). Dropping bowling balls on
tomatoes: Representations of object state-changes during sentence
processing. Journal of Experimental Psychology: Learning,
Memory, and Cognition.

Horchak, O. V., & Garrido, M. V. (2020b). Explicit (Not Implicit)
Attitudes Mediate the Focus of Attention During Sentence
Processing. Frontiers in Psychology, 11.

Java, R. I., & Gregg, V. H. (1997). What do people actually remember
(and know) in “Remember/Know” experiments? European Journal
of Cognitive Psychology, 9(2), 187–197.

Keller, D., & Kellas, G. (1978). Typicality as a dimension of encoding.
Journal of Experimental Psychology: Human Learning and
Memory, 4(1), 78–85.

Koen, J. D., & Yonelinas, A. P. (2014). The effects of healthy aging,
amnestic mild cognitive impairment, and Alzheimer’s Disease on
recollection and familiarity: A meta-analytic review.
Neuropsychology Review, 24(3), 332–354.

Kuznetsova, A., Brockhoff, P.B., Christensen, R.H.B. (2017). “lmerTest
Package: Tests in Linear Mixed Effects Models.” Journal of
Statistical Software, 82(13), 1-26.

Lin, E. L., & Murphy, G. L. (1997). Effects of background knowledge on
object categorization and part detection. Journal of Experimental
Psychology: Human Perception and Performance, 23(4), 1153–

Liu, Z.-X., Grady, C., & Moscovitch, M. (2016). Effects of prior-
knowledge on brain activation and connectivity during associative
memory encoding. Cerebral Cortex, bhw047.

Liu T., Xing, M. & Bai, X. (2020). Part-List Cues Hinder Familiarity but
Not Recollection in Item Recognition: Behavioral and Event-
Related Potential Evidence. Front Psychology 11:561899. https://

Makowski, D., Ben-Shachar, M. S., Chen, S. H. A., & Lüdecke, D.
(2019). Indices of Effect Existence and Significance in the
Bayesian Framework. Frontiers in Psychology, 10:2767. https://

Mäntylä, T. (1997). Recollections of faces: Remembering differences and
knowing similarities. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 23(5), 1203–1216. https://doi.

McCabe, D. P., & Geraci, L. D. (2009). The influence of instructions and
terminology on the accuracy of remember–know judgments.
Consciousness and Cognition, 18(2), 401–413.

McCabe, D. P., Roediger III, H. L., McDaniel, M. A., & Balota, D. A.
(2009). Aging reduces veridical remembering but increases false
remembering: Neuropsychological test correlates of remember–
know judgments. Neuropsychologia, 47(11), 2164–2173. https://

Medin, D. L., Unsworth, S. J., & Hirschfeld, L. (2007). Culture, catego-
rization and reasoning. In S. Kitayama, & D. Cohen (Eds.),
Handbook of cultural psychology (pp. 615-644). Guilford Press.

Mervis, C. B., Catlin, J., & Rosch, E. (1976). Relationships among good-
ness-of-example, category norms, and word frequency. Bulletin of
the Psychonomic Society, 7(3), 283–284.

Migo, E. M., Mayes, A. R., & Montaldi, D. (2012). Measuring recollec-
tion and familiarity: Improving the remember/know procedure.
Consciousness and Cognition, 21(3), 1435–1455.

Miles, J., & Shevlin, M. (2001). Applying regression and correlation: A
guide for students and researchers. SAGE.

Murphy, G. L. (2002). The big book of concepts. MIT Press.
Murphy, G. L., & Medin, D. L. (1985). The role of theories in conceptual

coherence. Psychological Review, 92(3), 289–316.

Nadel, L. (2020). What Is a Memory That It Can Be Changed?
Neuroscience of Enduring Change, 11–24.

Nadel, L., & Moscovitch, M. (1997). Memory consolidation, retrograde
amnesia and the hippocampal complex. Current Opinion in
Neurobiology, 7(2), 217–227.

Nadel, L., Hupbach, A., Gomez, R., & Newman-Smith, K. (2012).
Memory formation, consolidation and transformation.
Neuroscience & Biobehavioral Reviews, 36(7), 1640–1645.

Ozubko, J. D., Moscovitch, M., & Winocur, G. (2017). The influence of
recollection and familiarity in the formation and updating of asso-
ciative representations. Learning & Memory, 24(7), 298–309.

Potts, R., & Shanks, D. R. (2012). Can testing immunize memories
against interference? Journal of Experimental Psychology:
Learning, Memory, and Cognition, 38(6), 1780–1785. https://doi.

R Core Team (2019). R: A language and environment for statistical com-
puting. R Foundation for Statistical Computing. Retrieved from

Rajaram, S. (1993). Remembering and knowing: Two means of access to
the personal past. Memory & Cognition, 21(1), 89–102. https://doi.

Rajaram, S. (1998). The effects of conceptual salience and perceptual
distinctiveness on conscious recollection. Psychonomic Bulletin &
Review, 5(1), 71–78.

Renoult, L., Irish, M., Moscovitch, M., & Rugg, M. D. (2019). From
knowing to remembering: The semantic–episodic distinction.
Trends in Cognitive Sciences, 23(12), 1041–1057.

Rips, L. J., Shoben, E. J., & Smith, E. E. (1973). Semantic distance and
the verification of semantic relations. Journal of Verbal Learning
and Verbal Behavior, 12(1), 1–20.

Rosch, E., & Mervis, C. B. (1975). Family resemblances: Studies in the
internal structure of categories. Cognitive Psychology, 7(4), 573–

Rosch, E., Simpson, C., & Miller, R. S. (1976). Structural bases of typ-
icality effects. Journal of Experimental Psychology: Human
Perception and Performance, 2(4), 491–502.

Sakamoto, Y. & Love, B. C. (2004). Schematic Influences on Category
Learning and Recognition Memory. Journal of Experimental
Psychology: General.

Santi, A., Raposo, A., & Marques, J. F. (2015). Superordinate and do-
main category structure: Evidence from typicality ratings. Revista

Mem Cogn

Portuguesa de Psicologia, 44, 8–108.

Schmidt, S. R. (1996). Category typicality effects in episodic memory:
Testing models of distinctiveness. Memory & Cognition, 24(5),

Sekeres, M. J., Moscovitch, M., & Winocur, G. (2017). Mechanisms of
memory consolidation and transformation. In Axmacher N., Rasch
B. (eds) Cognitive neuroscience of memory consolidation. Studies in
Neuroscience, Psychology and Behavioral Economics. Springer,

Sekeres, M. J., Winocur, G., Moscovitch, M., Anderson, J. A. E.,
Pishdadian, S., Martin Wojtowicz, J., … Grady, C. L. (2018).
Changes in patterns of neural activity underlie a time-dependent
transformation of memory in rats and humans. Hippocampus,
28(10), 745–764.

Souza, C., Coco, M. I., Pinho, S., Filipe, C. N., & Carmo, J. C. (2016).
Contextual effects on visual short-term memory in high-functioning
autism spectrum disorders. Research in Autism Spectrum Disorders,
32, 64–69.

Souza, C., Garrido, M. V., & Carmo, J. C. (2020). A Systematic Review
of Normative Studies Using Images of Common Objects. Frontiers
in Psychology, 11.

Souza, C., Garrido, M. V., Saraiva, M., & Carmo, J. C. (2021). RealPic:
Picture norms of real-world common items. Behavior Research

Toichi, M., & Kamio, Y. (2003). Long-Term Memory in High-
Functioning Autism: Controversy on Episodic Memory in Autism
Reconsidered. Journal of Autism and Developmental Disorders,
33(2), 151–161.

Tse, D., Langston, R. F., Kakeyama, M., Bethus, I., Spooner, P. A.,
Wood, E. R., … Morris, R. G. M. (2007). Schemas and memory
consolidation. Science, 316(5821), 76–82.

Tse, D., Takeuchi, T., Kakeyama, M., Kajii, Y., Okuno, H., Tohyama, C.,
… Morris, R. G. M. (2011). Schema-Dependent Gene Activation
and Memory Encoding in Neocortex. Science, 333(6044), 891–895.

Tulving, E. (1972). Episodic and semantic memory. In: E. Tulving & W.
Donaldson (eds.), Organization of memory. Academic Press.

Tulving, E. (1985). How many memory systems are there? American
Psychologist, 40(4), 385–398.

Tulving, E. (2000). Memory: Overview. In: A. Kazdin (ed.),
Encyclopedia of Psychology, 5, 161-162. American Psychological
Association and Oxford University Press.

Tulving, E. (2002). Episodic Memory: From Mind to Brain. Annual
Review of Psychology, 53(1), 1–25.

Vakil, E., Sharot, T., Markowitz, M., Aberbuch, S., & Groswasser, Z.
(2003). Script memory for typical and atypical actions: controls
versus patients with severe closed-head injury. Brain Injury,
17(10), 825–833.

van Kesteren, M. T. R., & Meeter, M. (2020). How to optimize knowl-
edge construction in the brain. Npj Science of Learning, 5(1). https://

van Kesteren, M. T. R., Rijpkema, M., Ruiter, D. J., & Fernandez, G.
(2010). Retrieval of associative information congruent with prior
knowledge is related to increased medial prefrontal activity and
connectivity. Journal of Neuroscience, 30(47), 15888–15894.

van Kesteren, M. T. R., Beul, S. F., Takashima, A., Henson, R. N., Ruiter,
D. J., & Fernández, G. (2013a). Differential roles for medial pre-
frontal and medial temporal cortices in schema-dependent encoding:

From congruent to incongruent. Neuropsychologia, 51(12), 2352–

van Kesteren, M. T. R., Rijpkema, M., Ruiter, D. J., & Fernández, G.
(2013b). Consolidation differentially modulates schema effects on
memory for items and associations. PLoS One, 8(2): e56155. https://

van Kesteren, M. T. R., Rijpkema, M., Ruiter, D. J., Morris, R. G. M., &
Fernández, G. (2014). Building on prior knowledge: Schema-
dependent encoding processes relate to academic performance.
Journal of Cognitive Neuroscience, 26(10), 2250–2261. https://

Vargha-Khadem, F., Salmond, C. H., Watkins, K. E., Friston, K. J.,
Gadian, D. G., & Mishkin, M. (2003). Developmental amnesia:
Effect of age at injury. Proceedings of the National Academy of
Sciences, 100(17), 10055–10060.

Watier, N., & Collin, C. (2012). The effects of distinctiveness on memory
and metamemory for face–name associations. Memory, 20(1), 73–

Wickham, H., Averick, M., Bryan, J., Chang, W., McGowan, L. D. A.,
François, R., … & Kuhn, M. (2019). Welcome to the tidyverse.
Journal of Open Source Software, 4(43), 1686,

Wixted, J. T., & Mickes, L. (2010). A continuous dual-process model of
remember/know judgments. Psychological Review, 117(4), 1025–

Wixted, J. T., & Squire, L. R. (2010). The role of the human hippocampus
in familiarity-based and recollection-based recognition memory.
Behavioural Brain Research, 215(2), 197–208.

Yamada, R., & Itsukushima, Y. (2013). The effects of schema on recog-
nition memories and subjective experiences for actions and objects.
Japanese Psychological Research, n/a–n/a.

Yin, S., O’Neill, K., Brady, T.F., & De Brigard, F. (2019). The effect of
category learning on recognition memory: A signal detection theory
analysis. Proceedings of the 41stAnnual Meeting of the Cognitive
Science Society (pp. 3165-3171). Available in: https://people.duke.
accessed on November 27th, 2020.

Yonelinas, A. P. (2002). Components of episodic memory: The contribu-
tion of recollection and familiarity. In A. Baddeley, J. P. Aggleton,
& M. A. Conway (Eds.), Episodic memory: New directions in
research (p. 31–52). Oxford University Press.
1093/acprof:oso/9780198508809.003.0003, ISBN-13:

Yonelinas, A. P., Aly, M., Wang, W.-C., & Koen, J. D. (2010).
Recollection and familiarity: Examining controversial assumptions
and new directions. Hippocampus, 20(11), 1178–1194. https://doi.

Yonelinas, A. P., Ranganath, C., Ekstrom, A. D., & Wiltgen, B. J. (2019).
A contextual binding theory of episodic memory: systems consoli-
dation reconsidered. Nature Reviews Neuroscience, 20(6), 364–375.

Wang, W. C., Brashier, N. M., Wing, E. A., Marsh, E. J., & Cabeza, R.
(2018). Knowledge supports memory retrieval through familiarity,
not recollection. Neuropsychologia, 113, 14–21.

Publisher’s note Springer Nature remains neutral with regard to jurisdic-
tional claims in published maps and institutional affiliations.

Mem Cogn

  • Conceptual knowledge modulates memory recognition of common items: The selective role of item-typicality
    • Abstract
    • Introduction
      • The current studies
    • Experiment 1: Exploring the conceptual knowledge modulation of conscious memory processes
      • Method
        • Participants
        • Stimuli
        • Procedure
        • Data analysis
      • Results and discussion
        • Response times during encoding
        • Overall recognition
        • Phenomenological judgments of conscious memories
    • Experiment 2: Contrasting the encoding type and item-typicality on conscious recollection and the quality of recollective experience
      • Methods
        • Participants
        • Stimuli
        • Procedure
      • Results and discussion
        • Response times during encoding
        • Overall accuracy of Recognition phase 1
        • Phenomenological judgments of conscious memories of Recognition phase 1
      • Overall accuracy of Recognition phase 2
      • Phenomenological judgments of conscious memories of Recognition phase 2
      • Source memory
        • Source accuracy
        • Source descriptions
      • General discussion
      • Conclusion
      • Detailed instruction of RKG judgments
      • Experiment 1
        • Response times (RTs) during Encoding
        • Overall accuracy of Recognition
      • Experiment 2
        • Response times (RTs) during encoding
        • Overall accuracy of Recognition phase 1
        • Overall accuracy of Recognition phase 2
        • Phenomenological judgments of conscious memories of Recognition phase 2
    • References

published: 14 September 2021

doi: 10.3389/fnagi.2021.683908

Edited by:

Dennis Qing Wang,
Southern Medical University, China

Reviewed by:
Hudson Sousa Buck,

University of São Paulo, Brazil
Siri-Maria Kamp,

University of Trier, Germany

Ricardo J. Alejandro

[email protected]

Nico Bunzeck
[email protected]

Received: 22 March 2021
Accepted: 16 June 2021

Published: 14 September 2021

Alejandro RJ, Packard PA,

Steiger TK, Fuentemilla L and
Bunzeck N (2021) Semantic

Congruence Drives Long-Term
Memory and Similarly Affects Neural

Retrieval Dynamics in Young and
Older Adults.

Front. Aging Neurosci. 13:683908.
doi: 10.3389/fnagi.2021.683908

Semantic Congruence Drives
Long-Term Memory and Similarly
Affects Neural Retrieval Dynamics in
Young and Older Adults
Ricardo J. Alejandro1,2*, Pau A. Packard1,3, Tineke K. Steiger1, Lluis Fuentemilla4,5,6 and
Nico Bunzeck1,7*

1Department of Psychology, University of Lübeck, Lübeck, Germany, 2Department of Experimental Psychology, Ghent
University, Ghent, Belgium, 3Center for Brain and Cognition, Department of Information and Communication Technologies,
Universitat Pompeu Fabra Roc Boronat, Barcelona, Spain, 4Cognition and Brain Plasticity Group, Bellvitge Biomedical
Research Institute (IDIBELL), Hospitalet de Llobregat, Barcelona, Spain, 5Department of Cognition, Development and
Educational Psychology, University of Barcelona, Barcelona, Spain, 6Institute of Neurosciences, University of Barcelona,
Barcelona, Spain, 7Center of Brain, Behavior and Metabolism (CBBM), University of Lübeck Ratzeburger Allee, Lübeck,

Learning novel information can be promoted if it is congruent with already stored
knowledge. This so-called semantic congruence effect has been broadly studied in
healthy young adults with a focus on neural encoding mechanisms. However, the
impacts on retrieval, and possible impairments during healthy aging, which is typically
associated with changes in declarative long-term memory, remain unclear. To investigate
these issues, we used a previously established paradigm in healthy young and older
humans with a focus on the neural activity at a final retrieval stage as measured
with electroencephalography (EEG). In both age groups, semantic congruence at
encoding enhanced subsequent long-term recognition memory of words. Compatible
with this observation, semantic congruence led to differences in event-related potentials
(ERPs) at retrieval, and this effect was not modulated by age. Specifically, congruence
modulated old/new ERPs at a fronto-central (Fz) and left parietal (P3) electrode in a
late (400–600 ms) time window, which has previously been associated with recognition
memory processes. Importantly, ERPs to old items also correlated with the positive
effect of semantic congruence on long-term memory independent of age. Together,
our findings suggest that semantic congruence drives subsequent recognition memory
across the lifespan through changes in neural retrieval processes.

Keywords: aging, ERP, EEG, congruence effect, long-term memory


Learning novel information can be promoted if it is congruent with already stored long-term
knowledge (Craik and Tulving, 1975; Hall and Geis, 1980; Atienza et al., 2011; Tse et al., 2011;
Packard et al., 2017). In cognitive psychology, this so-called ‘‘congruence effect’’ has been
explained through the integration of information into knowledge structures or schemas (see also

Frontiers in Aging Neuroscience | 1 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

Piaget, 1952). On the basis of functional imaging studies,
including electroencephalography (EEG) and functional
magnetic resonance imaging (fMRI), recent work demonstrated
that encoding-specific processes play an essential role (see
below). However, it remains unclear how semantic congruence
during encoding changes retrieval dynamics and whether these
processes change during healthy aging, which is known to be
characterized by impairments of declarative long-term memory.

In typical experiments on the long-term effects of semantic
congruence, a semantic cue, for instance, a word such as
‘‘instrument,’’ predicts the presentation of a target that can be
semantically congruent, for instance ‘‘guitar,’’ or incongruent,
for instance, ‘‘tree’’ (Packard et al., 2017, 2020). fMRI studies
suggest that the long-term memory advantage for congruent
items directly relates to a modulation in connectivity between
the prefrontal cortex (PFC) and medial temporal lobe (MTL,
including the hippocampus; van Kesteren et al., 2010, 2013;
Sommer, 2017). According to the ‘‘schema-linked interactions
between medial prefrontal and medial temporal regions’’
(SLIMM) model (van Kesteren et al., 2012), the medial prefrontal
cortex (mPFC) ‘‘resonates’’ with congruent information and
therefore inhibits MTL activity in order to drive semantic
integration (see also van Kesteren et al., 2012, 2013, 2014). EEG
studies could provide further evidence for encoding specific
effects by showing that semantic congruence accelerates the onset
of the event-related potentials (ERPs) for successful memory
encoding (Packard et al., 2017). Moreover, semantic congruence
at encoding leads to differences in ERPs starting at around 400 ms
after stimulus onset, as well as theta (4–8 Hz), alpha (8–13 Hz),
and beta band (14–20 Hz) oscillations (Packard et al., 2020).
Importantly, these congruence-related ERPs predicted increases
in memory performance for congruent items, further suggesting
that ERPs and neural oscillations underlie the congruence effect
(Höltje et al., 2019; Packard et al., 2020).

While little is known about the neural dynamics of
congruence-dependent memory retrieval, electrophysiological
studies indicate specific correlates of recognition memory. For
instance, post-stimulus ERPs during retrieval typically show a
more positive deflection for correctly identified ‘‘old’’ items as
compared to correctly identified ‘‘new’’ items [i.e., the ‘‘ERP
Old-New Effect’’ (Rugg, 1995; Danker et al., 2008)]. Moreover,
dual-process models suggest that recognition can be associated
with specific details or associations of the encoding episode
(i.e., recollection), or the absence of such recollective experience
(i.e., familiarity; Krantz et al., 1974; Jacoby and Dallas, 1981;
Yonelinas, 2001), and both aspects appear to be linked to
different ERP components (Düzel et al., 1997; Curran, 2000).
Familiarity-based recognition is typically associated with a
midfrontal ERP component peaking between 300 and 500 ms,
often labeled the FN400 (Rugg and Curran, 2007; Bridger
et al., 2012). Recollection based recognition memory, on the
other hand, is associated with later ERP components, typically
observed from around 400–800 ms at left parietal electrodes
(Sanquist et al., 1980; Düzel et al., 1997; Curran, 2000; Rugg
and Curran, 2007; Danker et al., 2008). Additionally, both
components were linked to confidence level (sure, unsure) at
retrieval: item memory strength is associated with the FN400 and

source memory strength is associated with the late positive
complex (LPC; Woroch and Gonsalves, 2010; Wynn et al., 2020).

Similar to ERPs, neural oscillations in specific frequency
bands, namely theta, alpha, and beta, are thought to be crucial
for memory retrieval (Klimesch, 1999; Fell and Axmacher, 2011).
Specifically, theta power increases (i.e., theta synchronization)
in combination with alpha power decreases (i.e., alpha
desynchronization) are associated with enhanced memory
performance (Klimesch, 1999; Sauseng et al., 2002; Klimesch
et al., 2004). Moreover, the theta frequency band has been linked
to successful encoding and retrieval of semantic information
(Klimesch et al., 1997; Bastiaansen et al., 2008), with higher
amplitudes for recollection than for familiarity (Klimesch et al.,
2001). Likewise, the alpha frequency band also seems to play
a role for semantic information at encoding and retrieval, as
well as for sensory input, expectancy, and attentional processes
(Klimesch et al., 1997; Klimesch, 1997, 1999). Other studies,
using combined EEG-fMRI, suggest that theta-alpha oscillations
bind the hippocampus, PFC, and striatum during recollection
(Herweg et al., 2016). Finally, beta oscillations have been linked
to thalamocortical coupling during long-term memory retrieval
(Staudigl et al., 2012). Together, frontal ERPs as well as theta,
alpha, and beta oscillations play a role in memory retrieval but
their possible modulation through semantic congruence remains

Finally, while most previous studies on semantic congruence
have focused on younger participants (i.e., 18–35 years), potential
age-related changes and associated neural mechanisms need
further investigation. While age-related impairments could be
expected on the basis of well-described memory deficits in older
adults, it is also clear that semantic memory (i.e., long-term
memory for facts independent of time and date) is often
preserved until old age (Hedden and Gabrieli, 2004). Indeed,
we could show a preserved semantic congruence effect in older
adults (Packard et al., 2020), which is compatible with others
showing a relatively small effect of aging on semantic relatedness
and associated memory deficits (Crespo-Garcia et al., 2012).
However, congruence-related ERPs and neural oscillations in the
theta, alpha, and beta range (at encoding) were less pronounced
in older subjects indicating age-related neural changes in the
absence of behavioral deficits (Packard et al., 2020).

In this study, we used EEG to investigate the effects of
semantic congruence on subsequent long-term memory,
retrieval mechanisms, and possible age-related changes.
Knowing that semantic congruence promotes long-term
memory in both age groups and that at encoding there was
an effect of age on the electrophysiological measures (Packard
et al., 2020), we hypothesized: (a) a modulation of retrieval
specific ERPs as well as theta, alpha, and beta oscillations; and
(b) an age-dependent effect on the underlying neural processes
(i.e., group differences: young vs. older subjects). Note that
the behavioral data have already been published together with
encoding specific EEG activity (Packard et al., 2020). In this
study, we re-analyzed the behavioral effects and focused on EEG
activity at retrieval. We first employed cluster-based permutation
analyses that included all EEG electrodes (see below). Following
this rather conservative approach, we focused on the fronto-

Frontiers in Aging Neuroscience | 2 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

central electrode Fz and left parietal electrode P3 for both ERP
and time frequency (TF) analysis (see “Materials and Methods”
section) since the retrieval of information is particularly related
to prefrontal and parietal activity (Rugg and Curran, 2007;
Preston and Eichenbaum, 2013). Moreover, the PFC is one of
the brain regions exhibiting pronounced age-related changes in
terms of structure and function (Cabeza et al., 2002; Rajah and
D’Esposito, 2005; Craik and Grady, 2009).


Twenty-four young (15 females, mean age = 22.54,
SD = 2.83 years) and 26 older human subjects (16 females,
mean age = 64.42, SD = 6.56 years) participated in this study.
As described previously (Packard et al., 2020), all participants
were right-handed, had a normal or corrected-to-normal vision
(including color vision), and reported no history of neurological
or psychiatric disorders, or current medical problems (excluding
blood pressure). The cognitive abilities of older participants
were assessed using the Montreal Cognitive Assessment (MoCA)
version 7 (Nasreddine et al., 2005), where all participants had
a score of 22 or higher, which is considered a cut-off value for
Mild Cognitive Impairment (MCI; Freitas et al., 2013). For the
ERP analysis, two subjects had to be excluded, and for the TF
analysis, five subjects had to be excluded for different reasons
(see below).

Participants were recruited through local newspaper
announcements or the database of the University of Lübeck
(Greiner, 2015). All participants signed a written informed
consent and received monetary compensation. The study was
approved by the local ethical committee of the University of
Lübeck, Germany, and in accordance with the Declaration of

Behavioral Procedures
The experimental paradigm was as described previously (Packard
et al., 2020). Briefly, stimuli consisted of 66 categorical six word
lists (Packard et al., 2017), selected from category norms (Battig
and Montague, 1969; Yoon et al., 2004) translated into German.
Each list consisted of the six most typical instances (e.g., Banana,
Pear, Grape, Strawberry, Apple, Orange) of a semantic category

(e.g., Fruit). The total number of words was 396, all of them
were presented in individual encoding trials (see below). The test
phase (recognition) included a total of 396 Old-word (all items
presented at encoding) and 396 New-word trials (Figure 1).

During the encoding phase, each trial started with a fixation
cross in the middle of the screen for a random duration of
2,000–3,000 ms. Subsequently, the name of a semantic category
was displayed (white background, blue font) for 1,500 ms, which
was followed by a fixation cross for 2,000 ms. Finally, the target
word was displayed (white background, green font) for 1,000 ms.
During the presentation of the target word, the participants
pressed a button to indicate whether the word was congruent
(left-hand click) or incongruent (right-hand click) with the
semantic category. The condition was congruent if the target
word fitted the semantic category (Craik and Tulving, 1975),
for example the category ‘‘insect’’ followed by the target word
‘‘spider.’’ The condition was incongruent when the target word
did not belong to the semantic category, for example, the category
‘‘musical instrument’’ and the target word ‘‘rose.’’

The encoding phase lasted about 50 min and included 396
one-word trials, presented in random order. For each category,
three words (out of six) belonged to the semantic category
(semantically congruent), the remaining three were randomly
selected from other categories (semantically incongruent), giving
a total of 198 congruent and 198 incongruent stimuli presented
during encoding.

Following the encoding phase, participants performed a
short (5 min) distraction task, where they had to solve
simple arithmetic operations (additions and subtractions). The
distraction task prevented the participants from rehearsing the
words seen during encoding and prevented recency effects that
can contribute to memory.

Finally, the test phase had a duration of approximately 60 min.
Here, participants were shown a fixation cross at the beginning
of each trial for 1,500 ms, and subsequently, a word (either
an Old-word or a New-word item) was displayed (green font,
neutral background) for a maximum of 4,000 ms. Participants
had to indicate via button press whether they judged the word
as ‘‘sure old,’’ ‘‘guess old,’’ ‘‘guess new,’’ or ‘‘sure new.’’ The
button press determined the end of the trial and it started the
presentation of the next trial (i.e., a fixation cross). Participants
were allowed to take a break every 50 trials.

FIGURE 1 | Experimental paradigm. Encoding phase: word pairs are shown (semantic category + target word). Test phase: all target words from the encoding
phase were presented randomly intermixed with novel distractors (new words). Figure adapted from Packard et al. (2020).

Frontiers in Aging Neuroscience | 3 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

Statistical Analyses of Memory Results
We performed a 2 × 2 repeated measures Analysis of Variance
(ANOVA) using Jamovi Version (The jamovi project,
2020), with encoding condition (Congruent vs. Incongruent)
as within-subjects factor, and age group (Young vs. Older) as
between-subjects factor. Dependent variables were response rates
and reaction times (RTs). We included only high-confidence
responses in the tests (see below), with α (type I error rate) set
to 0.05, η2p to estimate effect sizes, and Bayes Factors (BF10)
to evaluate evidence comparing the alternative hypothesis
(1) model to the null (0) model. For effects due to interactions of
variables, evidence was evaluated by comparing the BF10 of the
model with the interaction against the BF10 of the model with
only the main effects (i.e., BF10 interaction model/BF10 Main
effects model). Note that Bayes factors indicate evidence in favor
of the alternative vs. the null hypothesis given the empirical data.
Bayes factors between 1 and 3 indicate anecdotal evidence, 3
and 10 moderate evidence, 10 and 30 strong evidence, 30 and
100 very strong evidence, and >100 extreme evidence in favor of
the alternative hypothesis. Conversely, 1/3 indicates anecdotal
evidence, 1/10–1/3 moderate evidence, 1/30–1/10 strong
evidence, 1/100–1/30 very strong evidence, and <1/100 extreme
evidence in favor of the null hypothesis. One indicates no
evidence (Schönbrodt and Wagenmakers, 2018; Lakens et al.,

As mentioned in our previous work (Packard et al., 2020),
response accuracy during the encoding phase was very high (see
Table 1). Although there is a significant difference in congruence
judgments driven by age (i.e., younger participants had higher
accuracy identifying incongruent items than congruent items,
while older participants had similar accuracy for both types
of trials, see Table 1, and ‘‘Results’’ section), our analysis
of the test phase only included trials correctly identified as
congruent or incongruent during encoding (see below). Since
older participants are prone to more memory errors while
retrieving recently learned information with high confidence
(Dodson and Krueger, 2006; Dodson et al., 2007a,b; Chua et al.,
2009; Shing et al., 2009), we included only high-confidence
responses in the analyses. Specifically, corrected Hit Rates (CHR)
were calculated by subtracting the proportion of False Alarm
(FA; erroneous ‘‘old’’ response to a ‘‘new’’ item) responses
from the proportion of hits (correct ‘‘old’’ responses to old
words). Only high-confidence responses were included in the
calculation of the CHR. Note that FAs could not be classified as
congruent and incongruent, because they corresponded to words
not presented during encoding.

Post hoc t-tests were used to evaluate significant interactions
detected in the ANOVAs. The Bonferroni correction was used
to account for multiple comparisons, lowering the significance

TABLE 1 | Proportion of correct responses and standard deviations (in brackets)
during the encoding phase.

Age group Trial type Accuracy (%)

Younger Congruent 91.6 (0.07)
Incongruent 94.8 (0.06)

Older Congruent 93.9 (0.04)
Incongruent 95.2 (0.03)

level according to the amount of post hoc tests performed for
each ANOVA. Tests which did not reach the Bonferroni-adjusted
significance levels are stated as non-significant.

EEG Recordings and Analyses
As described in our previous study (Packard et al., 2020),
EEG activity during retrieval was acquired using BrainAmp
amplifiers, an EasyCap system (BrainProducts GmbH, Munich,
Germany), and BrainVision Recorder (version 1.03.0003). We
used 32 standard active scalp electrodes, and four electrodes for
monitoring vertical and horizontal eye movement (VEOG and
HEOG). Electrode impedances were maintained under 20 k�.
Electrode FCz was used as reference and AFz served as a
ground electrode. Data were re-referenced offline to electrode
Oz, since re-referencing to the average, although convenient,
can alter or suppress the representations of effects with a
broad scalp distribution. Furthermore, the average reference
delivers waveforms and scalp distributions dependent on the
study-specific electrode locations, making it difficult to compare
results across studies; it is therefore preferably recommended for
high-density montages (Dien, 1998; Luck, 2005). On the other
hand, electrode Oz is sensitive to brain activity, but it is located
far from the zone of interest (frontal), being thus a suitable
reference to measure the full amplitude of the effect in frontal
areas and avoid channel distortion (Luck, 2005).

The sampling rate for data acquisition was 500 Hz. The
recordings were high-pass (0.1 Hz) and low-pass (240 Hz)
filtered online. The open-source EEGLAB (Delorme and Makeig,
2004) toolbox (version 2019), under a customized MATLAB
(version R2019b; The MathWorks) environment, was used for
preprocessing the EEG data offline.

All trials were limited to a length of 800 ms post-stimulus for
epoching. This restriction was necessary since the response to
a word ended the trial with the disappearance of the word and
the presentation of a fixation cross (i.e., starting the next trial,
see ‘‘Materials and Methods’’ section). A longer trial duration
could cause the trials from earlier responders (<800 ms) to
capture the brain responses to the presentation of the stimulus
of the subsequent trial. All trials were epoched accordingly
and downsampled to 125 Hz, the latter to reduce computation
time, file sizes, and file reading/writing time, without significant
loss of information (Seth, 2010; Cohen, 2014). Trials with
amplitudes exceeding 100 µV were rejected offline as they were
considered artifacts. Eye blinking, saccades, heart beating, and
muscle movement artifacts were identified with Independent
Component Analysis (ICA; Makeig et al., 1996; Jung et al., 2000;
Delorme and Makeig, 2004), implemented in EEGLAB (Makeig
et al., 1996; Jung et al., 2000). The artifactual components were
selected by visual inspection of scalp maps (head topographies),
power spectrum, and ERP plots, and were later removed from
the data.

After cluster-based permutation analyses that included all
electrodes (see below), post hoc, we focused on the fronto-central
electrodes Fz and P3 (both ERPs and TF, see ‘‘Introduction’’
section), based on previous studies investigating retrieval
dynamics (Wilding and Rugg, 1996; Curran, 2000; Rugg and
Curran, 2007; Diana et al., 2011).

Frontiers in Aging Neuroscience | 4 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

ERP Analysis
The ERP analysis was in accordance with our previous work
(Packard et al., 2020). Here, data were low-pass filtered offline
using the recommended windowed-sinc FIR filter (Widmann
et al., 2015), with a Hamming window, the cut-off frequency
at 40 Hz, filter order at 166, implemented in EEGLAB, with
no additional high-pass filtering, since the online high-pass
filtering was considered sufficient. We analyzed the ERPs by
extracting event-locked EEG epochs of 900 ms, starting 100 ms
before (baseline signal) and ending 800 ms after stimulus onset.
Major artifacts, trials with amplifier saturation, and bad channels
were visually identified and removed (maximum four channels,
mean = 0.76). ICA (Makeig et al., 1996; Delorme and Makeig,
2004) was performed and finally, bad channels were interpolated.
Otherwise, the preprocessing was performed as described in
section ‘‘EEG Recordings and Analyses’’.

For our EEG-data analysis, we focused on those trials that
were: (a) correctly classified during encoding as congruent or
incongruent; and (b) correctly classified in the test phase as
old-word with high-confidence or correctly rejected as new
items with high confidence. All other trials (FA, etc.) were
not further analyzed. Trial numbers per condition for both the
ERP and TF analysis are shown in Table 2. The difference in
trial numbers between ERP and TF analysis is due to slightly
different preprocessing routines (see ‘‘Materials and Methods’’

One young and one older participant had to be excluded
from the analysis due to excessively noisy data, or since they
were regarded as behavioral or electric potential outliers as
compared to their age group [identified with Jamovi (The
jamovi project, 2020) using a step of 1.5× Interquartile Range].
Therefore, the number of included subjects was not identical to
our previous work (Packard et al., 2020), in which we analyzed
EEG data from the encoding phase. Here, for the ERP analysis, we
included 25 old and 23 younger subjects. Fieldtrip (Oostenveld
et al., 2011) and customized MATLAB scripts were used for
statistical data analysis via a two-tailed non-parametric cluster-
based permutation test (Maris and Oostenveld, 2007) to identify
differences between the conditions (congruent vs. incongruent
trials). The test included all time points between 0 and 800 ms at
27 (out of 28) scalp electrodes, the reference electrode Oz was
not considered for the analysis since its activity was canceled
out during the re-referencing pre-processing. For every sample
(every channel∗time-pair), the conditions were compared using
a t-test. All the samples scoring higher than a specified threshold
(0.05) were selected and grouped into clusters, based on temporal

and spatial adjacency. The threshold of 0.05 is not the type I
error rate for the statistical test, it is a cut-off value for choosing
a sample as a member of a cluster. We chose this threshold
in accordance with recommendations (Maris and Oostenveld,
2007) and previous literature (Steiger et al., 2019; Packard et al.,
2020). Then the sum of t-values for each cluster was calculated
to obtain the cluster statistics, and the maximum cluster-level
statistics was taken as the test statistic which we used to assess
the difference between the conditions.

To calculate the significance probability, we used the Monte
Carlo method. Random partitions (random samples are extracted
from both conditions and put together in a subset, the remaining
samples are placed into another subset) were created and the
test statistics described above were calculated on those random
partitions. This procedure was repeated 1,000 times to generate
a histogram of the test statistics. The p-value was then obtained
with the proportion of cluster statistics in the random partitions
exceeding the one calculated from the observed data. Clusters
were formed from samples with p-values lower than α (0.05); we
considered only effects with at least three significant neighboring
channels, based on triangulation. Note that significant results
from a cluster-based permutation test provide information for
rejecting the null hypothesis (absence of an effect), rather than
an explanation of the extent of a cluster, which depends on
several factors and requires further interpretation (Maris and
Oostenveld, 2007; Maris, 2012; Sassenhagen and Draschkow,

Time-Frequency Analysis
For the TF analysis, data were low-pass filtered offline using
the recommended windowed-sinc FIR filter (Widmann et al.,
2015), with a Hamming window, the cut-off frequency at 35 Hz,
filter order at 166, implemented in EEGLAB, with no additional
high-pass filtering. Major atypical artifacts, trials with amplifier
saturation, and bad channels were visually identified and
removed (maximum four channels per participant, mean = 2.34).
ICA (Makeig et al., 1996; Delorme and Makeig, 2004) was
performed and finally, bad channels were interpolated. Five
young and one older participant had to be excluded from the
TF analysis due to excessively noisy data, or since they were
regarded as behavioral or spectral power outliers as compared
to their age group [identified with Jamovi (The jamovi project,
2020) using a step of 1.5× Interquartile Range]. Therefore, the
number of included subjects was not identical to our previous
work (Packard et al., 2020), in which we analyzed EEG data
from the encoding phase. Here, for the TF analysis, we included

TABLE 2 | Average trial numbers and standard deviations (in brackets) for EEG Analyses.

Age group Trial type Rate ERP analysis number of trials TF analysis number of trials

Younger Sure Congruent Hit 0.63 (0.16) 95.87 (30.84) 100.14 (32.08)
Sure Incongruent Hit 0.38 (0.17) 58.83 (26.29) 60.95 (26.77)
Sure False Alarm 0.04 (0.03) 12.09 (10.49) 10.52 (7.61)
Sure Correct Rejection 0.59 (0.21) 104.09 (67.74) 105.36 (73.24)

Older Sure Congruent Hit 0.63 (0.19) 102.16 (29.82) 98.14 (27.13)
Sure Incongruent Hit 0.41 (0.19) 65.00 (29.04) 57.68 (22.25)
Sure False Alarm 0.06 (0.08) 24.80 (34.80) 16.64 (29.45)
Sure Correct Rejection 0.79 (0.17) 42.91 (43.13) 51.90 (47.63)

Frontiers in Aging Neuroscience | 5 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

25 old and 20 younger subjects. Otherwise, the preprocessing
was performed as described in section ‘‘EEG Recordings and

The TF decomposition was conducted spanning the
frequencies from 2 to 30 Hz, in steps of 0.25 Hz, from 500 ms
before stimulus onset to 800 ms after stimulus onset, in steps
of 8 ms, convolving each single-trial time series with complex
Morlet wavelets (4 cycles). The epoch length was extended
(using data reflection) to 2,000 ms pre-stimulus, and to 2,000 ms
post-stimulus presentation, to ensure that the time window
of interest (−500–800 ms) was not interfered with by edge
artifacts (Debener et al., 2005; Herrmann et al., 2005; Cohen,
2014, 2017a,b). The average power was obtained across trials.
Baseline correction was applied from 500 ms before stimulus
onset to 200 ms before stimulus onset, to facilitate interpretation
and statistical analyses, and to avoid post-stimulus activity from
being averaged into the baseline estimate as much as possible
(Cohen, 2017a). The values thus obtained indicated the change
in power as compared to the power during the baseline period,
that is, with a scale in dB, a value of 0 would indicate no change
with respect to baseline.

In order to identify differences between conditions
(congruent vs. incongruent), we ran a two-tailed non-parametric
cluster-based permutation test (Maris and Oostenveld, 2007) on
the frequency range from 2 Hz to 30 Hz. The test is performed
similarly as described for the ERPs (section ‘‘ERP Analysis’’), the
difference is that for TF analysis the spectral dimension is added,
the samples are thus channel∗frequency∗time-triplets.


Behavioral Findings
Accuracy of congruence judgement (correctly identifying
congruent items as congruent, and incongruent items as

incongruent) during the encoding phase was analyzed in a
2 × 2 repeated measures ANOVA. The analysis showed a main
effect of congruence (F(1,46) = 24.21, p < 0.001, η2p = 0.345,
BF10 = 636.02), no significant main effect of age (F(1,46) = 0.94,
p = 0.34, η2p = 0.02, BF10 = 0.58), and a significant congruence by
age interaction (F(1,46) = 4.71, p = 0.04, η2p = 0.09, BF10 = 1.64).
Subsequent post hoc t-tests showed a significant difference in
congruence for younger participants (t = −4.91, p < 0.001),
which was absent in older participants (t = −1.99, p = 0.32). See
Table 1 for proportion of correct responses during encoding.

We analyzed the proportions of high-confidence (Sure) correct
responses during the test phase in a 2 × 2 repeated measures
ANOVA. The results are shown in Figure 2A, contrasting the
congruent and incongruent responses for younger (Y. Cong
and Y. Inc.) and for older (O. Cong and O. Inc.) participants.
The analysis showed a significant main effect for congruence
(F(1,46) = 264.09, p < 0.001, η2p = 0.85, BF10 = 5.1 × 10

with a higher CHR for congruent words (mean = 0.57, Standard
Error of the Mean (SEM) = 0.02), than for incongruent words
(mean = 0.33, SEM = 0.02). The data did not reveal a main effect
of age (F(1,46) = 0.123, p = 0.727, η2p = 0.003, BF10 = 0.26), and
no congruence by age interaction (F(1,46) = 0.468, p = 0.497,
η2p = 0.010, BF10 = 0.33). This behavioral analysis confirms our
previous results (Packard et al., 2020).

Reaction Times (RTs)
RTs during retrieval of all correct old responses (high and low
confidence) were analyzed in a 2 × 2 repeated measures ANOVA
and revealed a main effect of congruence (F(1,46) = 21.12,
p < 0.001, η2p = 0.32, BF10 = 613.44), a main effect of age
(F(1,46) = 6.06, p = 0.02, η2p = 0.12, BF10 = 2.31), but no significant
congruence by age interaction (F(1,46) = 0.75, p = 0.39, η2p = 0.02,
BF10 = 0.25). The main effects were driven by overall slower RTs

FIGURE 2 | (A) Behavioral results. Memory performance was significantly higher for congruent items in both age groups. (B) Reaction Times (RT) during recognition
of congruent and incongruent items. Correctly responding to incongruent items took longer in both groups, and older participants had overall slower RTs. Error bars
show one Standard Error of the Mean (SEM). Abbreviations: Y., young, O., older, Cong., congruent, Inc., incongruent, n.s., not significant. ***Indicates statistical
significance at p < 0.001 and *p < 0.05.

Frontiers in Aging Neuroscience | 6 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

in older subjects and overall faster RTs to congruent items, as
shown in Figure 2B. When analyzing only the high confidence
old responses, a slightly different pattern emerged: there was
a main effect of age (F(1,46) = 175, p < 0.001, η2p = 0.79,
BF10 = 2.10 × 1010), no main effect of congruence (F(1,46) = 2.52,
p = 0.12, η2p = 0.05, BF10 = 0.65), and no congruence by age
interaction (F(1,46) = 0.15, p = 0.70, η2p = 0.003, BF10 = 0.31).

EEG Findings
ERP Cluster Analysis
A Monte Carlo cluster-based permutation test was performed
on high-confidence responses for correctly recognized old
congruent (sure hits) vs. old incongruent words (sure hits), of
young and older participants grouped together, from 0 ms to
800 ms after word onset (i.e., main effect of congruence on
old responses). The analysis revealed a significant difference
between conditions (p = 0.004), that appears to be emphasized
approximately in the time window from 450 to 550 ms, with a
mainly frontal topography but also including central and parietal
electrodes (see Figure 3A). The contrast for the main effect
age, i.e., young vs. older subjects collapsed across congruent and
incongruent sure hits, did not reveal any significant effects. The
interaction of congruence and age, as quantified by old congruent
minus old incongruent in young participants vs. old congruent
minus old incongruent in older participants, also did not reveal
any significant results.

ERP Analyses
Effects of Congruence on Old/New Differences at Fz
In order to more thoroughly characterize the ERPs with respect
to typical old/new retrieval dynamics, we studied the waveforms
recorded by the electrode Fz and subsequently P3 (see below;
Figures 3B,C). After visual inspection of the ERPs, we focused
our analysis on an early (100–200 ms) and a late (400–600 ms)
time window. Since retrieval effects are often characterized
by ERP differences to correctly identified old (hits) vs. new
items (correct rejections), we computed difference waves for old
congruent vs. new items, and old incongruent vs. new items at
Fz for both age groups (see Figure 4A). The averaged old/new

differences in both time windows were analyzed separately with
2 × 2 ANOVAs with the factors age and congruence.

The early time window did not show a significant main effect
of congruence (F(1,46) = 2.85, p = 0.10, η2p = 0.06, BF10 = 0.64),
there was no significant main effect of age (F(1,46) = 0.18,
p = 0.68, η2p = 0.004, BF10 = 0.53), and no congruence by age
interaction (F(1,46) = 1.87, p = 0.18, η2p = 0.04, BF10 = 0.62).
The late time window revealed a significant main effect of
congruence (F(1,46) = 7.19, p = 0.01, η2p = 0.14, BF10 = 3.33),
but no significant main effect of age (F(1,46) = 0.01, p = 0.91,
η2p = 0.000, BF10 = 0.47), and no significant interaction of
congruence by age (F(1,46) = 2.67, p = 0.11, η2p = , BF10 = 0.81).
The main effect of congruence was driven by more negative
deflections in the old/new difference wave to congruent items
(Figure 4A).

Effects of Congruence on Old/New Differences at P3
Similar to Fz, averaged old/new differences from both time
windows at P3 (see Figure 4B) were analyzed separately with
2 × 2 ANOVAs with the factors age and congruence. The
early time window did not show a significant main effect of
congruence (F(1,46) = 0.68, p = 0.41, η2p = 0.02, BF10 = 0.27), no
significant main effect of age (F(1,46) = 0.83, p = 0.37, η2p = 0.02,
BF10 = 0.62), and no congruence by age interaction (F(1,46) = 1.31,
p = 0.26, η2p = 0.03, BF10 = 0.48). The late time window
showed a significant main effect of congruence (F(1,46) = 5.36,
p = 0.03, η2p = 0.10, BF10 = 1.54), no significant main effect
of age (F(1,46) = 2.78, p = 0.10, η2p = 0.06, BF10 = 1.003), and
no significant interaction of congruence by age (F(1,46) = 0.84,
p = 0.37, η2p = 0.02, BF10 = 0.35). The main effect of congruence
was driven by more negative deflections in the old/new difference
wave to congruent items (Figure 4B).

TF Cluster Analysis
As with the ERP analysis, a Monte Carlo cluster-based
permutation test was performed on the high-confidence
responses for correctly recognized old (sure hits) congruent vs.
incongruent words (sure hits), comparing changes in spectral
power elicited by word cues of young and older participants
grouped together, from 0 ms to 800 ms after word onset. The

FIGURE 3 | (A) Scalp topographies during the time-span of the most prominent clusters due to congruence. Electrodes belonging to the cluster are highlighted
with asterisks. (B) ERP responses (electrode Fz) for three conditions (Congruent vs. Incongruent vs. Correct Rejections) and age groups (Younger vs. Older adults).
(C) ERP responses (electrode P3) for three conditions (Congruent vs. Incongruent vs. Correct Rejections) and age groups (Younger vs. Older adults). The analyses
for both electrodes focused on an early (100–200 ms) and a late (400–600 ms) time window.

Frontiers in Aging Neuroscience | 7 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

FIGURE 4 | Difference waves for both conditions (Congruent Hits minus Correct Rejection, and Incongruent Hits minus Correct Rejection) for both age groups
(Younger and Older participants) for electrodes (A) Fz, and (B) P3.

analysis revealed no statistically significant effects (p > 0.05).
Similarly, the contrast for the main effect of age (i.e., young
vs. older subjects collapsed across congruent and incongruent
sure hits), and the interaction (i.e., congruent minus incongruent
in young participants vs. congruent minus incongruent in
older participants), neither revealed any significant effects
(p > 0.05).

TF Analysis for Fz
We followed the rather conservative cluster-based permutation
by a specific TF analysis of the electrode Fz. After visual
inspection (Figure 5A), we compared the relative change in
power for the alpha and theta frequency bands. Specifically, we
focused on a time window from 100 to 250 ms, from 4 to 8 Hz
for theta (black rectangles in Figure 5A), and on a time window
from 200 to 600 ms, from 10 to 14 Hz for alpha (white rectangles
in Figure 5A).

The test on the early theta effect revealed a main effect of
age (F(1,43) = 5.91, p = 0.02, η2p = 0.12, BF10 = 2.35), but no
main effect of congruence (F(1,43) = 0.03, p = 0.859, η2p = 0.001,
BF10 = 0.22), and no congruence by age interaction (F(1,43) = 2.85,
p = 0.098, η2p = 0.06, BF10 = 1.10). The main effect of age was
driven by enhanced theta synchronization for older participants
(Figure 5A). The analysis on the later alpha effect showed a
significant main effect of age (F(1,43) = 3.94, p = 0.05, η2p = 0.08,
BF10 = 1.20), but no main effect of congruence (F(1,43) = 3.46,
p = 0.07, η2p = 0.08, BF10 = 1.04), and no congruence by age
interaction (F(1,43) = 0.14, p = 0.71, η2p = 0.003, BF10 = 0.33).
The main effect of age was driven by lower alpha power for the
younger group (see Figure 5A).

TF Analysis for P3
Compatible with our ERP analysis, we explored TF effects at
P3. Visual inspection (see Figure 5B) also revealed an early
theta effect (100–250 ms, black rectangles in Figure 5B) and a
later alpha effect, which was most pronounced from 400–700 ms
(white rectangles in Figure 5B).

The test on the early theta effect revealed a significant main
effect of age (F(1,43) = 6.65, p = 0.01, η2p = 0.13, BF10 = 3.31), no
main effect of congruence (F(1, 43) = 0.39, p = 0.54, η2p = 0.009,

BF10 = 0.32), and no congruence by age interaction (F(1,43) = 0.04,
p = 0.85, η2p = 0.001, BF10 = 0.31). Similar to Fz, the main effect
of age was driven by enhanced theta synchronization for older
participants. The analysis on the later alpha effect showed no
significant main effect of age (F(1,43) = 2.07, p = 0.16, η2p = 0.05,
BF10 = 0.98), no main effect of congruence (F(1,43) = 0.98, p = 0.33,
η2p = 0.22, BF10 = 0.31), and no congruence by age interaction
(F(1,43) = 0.82, p = 0.37, η2p = 0.02, BF10 = 0.26).

Correlations of ERPs and Behavior
To investigate possible correlates between neural activity
specifically to old items and the (congruence-driven) memory
benefit, we ran a partial correlation for the late time window
at both electrodes (Fz and P3). We used the difference of the
mean ERP amplitudes (congruent ERPs to old items minus
incongruent ERPs to old items) as the independent variable and
the memory benefit by congruence (congruent high-confidence
CHR minus incongruent high-confidence CHR) as the
dependent variable. For Fz, the correlation was significant in the
late time window (see Figure 6A, r = −0.29, p = 0.047). Here,
pronounced memory advantages by congruence were associated
with large ERP amplitude differences between congruent and
incongruent old items. For P3, no significant correlations could
be revealed in the late time window (r = −0.21, p = 0.15,
see Figure 6B).

Note that we focused our correlation analyses only on the
late time window since ERPs in the early time window did not
show a significant (all p > 0.05) modulation by congruence. The
same rationale applies to the TF domain, where we also did not
observe a modulation of alpha and theta by congruence in any
time window.


This study investigated the neural processes underlying the
congruence effect with a focus on the retrieval dynamics
and possible age-related changes. The behavioral results show
that semantic congruence during encoding promotes memory
retrieval in both younger and older adults. Compatible with this
observation, congruence led to differences in the ERP retrieval

Frontiers in Aging Neuroscience | 8 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

FIGURE 5 | TF Analysis for electrodes (A) Fz and (B) P3. For Fz, the theta frequency band (black rectangles) was analyzed from 4 to 8 Hz, from 100 to 250 ms, and the
alpha frequency band (white rectangles) was analyzed from 10 to 14 Hz, from 200 to 600 ms, for young and older participants. For electrode P3 theta (black rectangles)
was analyzed from 4 to 8 Hz, from 100 to 250 ms, and alpha (white rectangles) was analyzed from 10 to 14 Hz, from 400 to 700 ms, for young and older participants.

FIGURE 6 | Partial correlations (controlling for age) between the mean ERP amplitudes (Congruent ERP to old items minus Incongruent ERP to old items) as the
independent variable, and the memory benefit by congruence (congruent high-confidence CHR minus incongruent high-confidence CHR) as the dependent variable,
for the late time window for the electrode (A) Fz, and (B) P3.

old/new effect in a time window from 400 to 600 ms at a
frontal and parietal electrode in both age groups. Importantly,
the behavioral benefit of semantic congruence correlated with
neural activity (ERPs) in this time window, pointing towards
a direct relationship. Our findings suggest that semantic
congruence drives long-term recognition memory through

modulations of retrieval dynamics that are preserved across
the lifespan.

At the behavioral level, we confirmed our previous work
(Packard et al., 2020), showing that congruent items were
better remembered than incongruent items in both age groups.
From a general point of view, this is in line with a wealth

Frontiers in Aging Neuroscience | 9 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

of studies (Schulman, 1974; Craik and Tulving, 1975; Packard
et al., 2017) demonstrating that semantic congruence drives
long-term memory. With regard to healthy aging, it is clear
that several aspects of long-term memory, including episodic
memory, decline as age progresses and this might directly
relate to neural degeneration within the prefrontal cortex and
medial temporal lobe (Hedden and Gabrieli, 2004; Nyberg
et al., 2012; Ofen and Shing, 2013). Therefore, a reduction of
the congruence effect could have been expected. However, it
has also been suggested that episodic and semantic aspects of
long-term memory may show differential age effects with the
latter being less pronounced (Ofen and Shing, 2013). In other
words, semantic long-term memory is rather stable throughout
the lifespan, which may help to explain the absence of age-related
differences in the semantic congruence effect in our study.
Indeed, and as we have argued before (Packard et al., 2020), this
observation is compatible with a previous EEG study (Crespo-
Garcia et al., 2012) suggesting a relatively small effect of aging on
semantic relatedness and associated memory deficits. However,
in some studies, congruence effects were reported to be impaired
during healthy aging (Amer et al., 2018, 2019), and this was
associated with additional brain activation in older as compared
to younger adults (Amer et al., 2019). Since there is only a
limited number of published studies on age-related changes in
the effect of congruence on long-term memory, further research
is needed.

Congruent items were recognized faster than incongruent
items (i.e., longer RTs for incongruent words, see Figure 2B)
in both age groups, further suggesting that congruent
information is retrieved faster and more efficiently than
incongruent information. This is in line with the ‘‘depth
of processing’’ account (Craik and Tulving, 1975), stating
that the integration of congruent information into previous
knowledge facilitates subsequent recall since only a portion of
the initial information (semantic cue) is needed to extract and
complete the representation from memory. Even so, it does
not rule out the alternative transfer-appropriate processing
account (Morris et al., 1977; Roediger, 2008). Although the
RT advantage for congruent items was independent of age,
older participants had overall longer RTs at retrieval for both
congruent and incongruent items. Such age-dependent and
therefore characteristic delays have often been described in
the memory literature (Salthouse, 1996; Park et al., 2002; Luo
and Craik, 2008). Accordingly, processing speed is notably
slower at an older age, and this might be attributable to
age-related loss of neural connections (Raz, 2000), changes in
neurotransmitter systems (Tromp et al., 2015), impaired neural
processing (Salthouse, 2013), and reduced attentional capabilities
(Rodrigues and Pandeirada, 2014).

At the neural level, we observed congruence-dependent
effects on retrieval-related old/new processes. Specifically, within
a late time window (400–600 ms) incongruent items were
associated with more negative deflections in the old/new ERPs
at a fronto-central (Fz) and parietal electrode (P3). Although
EEG has a poor spatial resolution, the activity of Fz presumably
reflects activity in underlying frontal brain regions, which
fits to a role of the frontal cortex in semantic congruence.

The SLIMM model (van Kesteren et al., 2012) suggests that
semantically congruent information leads to resonance in the
mPFC, which, as a consequence, inhibits MTL activity in order
to drive semantic integration. While initial evidence appears to
be compatible with SLIMM (van Kesteren et al., 2012, 2013,
2014), others suggest that both the mPFC and MTL together
drive semantic integration (McKenzie et al., 2013, 2014; Preston
and Eichenbaum, 2013; Gilboa and Marlatte, 2017; Liu et al.,
2017; van Kesteren et al., 2020). Although we cannot resolve
the precise role of the mPFC and hippocampus on the basis
of our EEG data, our findings demonstrate that congruence
during encoding modulates subsequent retrieval dynamics at
frontal electrodes.

The left parietal old/new ERP effect, on the other hand,
may reflect components typically associated with recollection-
based recognition memory (Tulving, 1985; Düzel et al., 1997).
Such an interpretation is further underlined by the fact that
we only included high-confidence responses in our analysis,
which are most likely based on recollective experiences rather
than familiarity judgments, and the absence of other frontal
components indicative of familiarity (Curran, 2000; Rugg and
Curran, 2007). Both old/new effects (at Fz and P3) were
associated with more positive deflections for incongruent items,
and the correlation analysis for the ERP responses to old
items showed that memory benefits by congruence directly
relates to the ERP differences in the late time window (see
Figure 6A). In other words, the more pronounced the semantic
congruence effect, the larger the ERP differences for congruent
vs. incongruent items. Although this analysis does not allow any
causal inferences, it further points towards a direct relationship
between subsequent recognition by congruence and neural
processes especially in the late time window at retrieval. Since the
correlation was observed across all participants when partializing
out age, this suggests a common underlying neural mechanism in
both age groups.

Although ERPs indexing successful retrieval are often found
after 200 ms, earlier retrieval ERPs have also been detected
(Bunzeck et al., 2009; Apitz and Bunzeck, 2013). Furthermore,
it is interesting to note that although deep-processing has
been previously found to increase both late positive and
high-confidence retrieval and recollection (Voss and Paller,
2017), here, the items identified as semantically congruent led
to greater high-confidence retrieval, but a less pronounced later
component. While recollection specific late ERP components
may vary in their scalp distribution, they typically have a centro-
parietal (and not frontal) topography (Rugg and Curran, 2007;
Friedman, 2013). Moreover, the effect of congruence on old/new
differences found in parietal (P3), and fronto-central (Fz)
locations further supports our interpretation that congruence
modulates retrieval through recollection processes. Together,
there are apparent differences between previous ERP studies on
recognition memory and our work. Therefore, further research
is needed to determine more precisely which components of
retrieval are increased or facilitated when recognizing items
encoded within a congruent semantic context.

For both the early and late time windows, older participants’
ERP responses exhibited greater amplitudes as compared to the

Frontiers in Aging Neuroscience | 10 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

younger group. This may be indicative of compensatory
mechanisms in order to achieve the same behavioral
performance. Indeed, previous work (Cabeza et al., 2002)
suggests that during episodic memory retrieval, high performing
older subjects (i.e., those that did not differ at the behavioral
level from younger subjects) recruit bilateral instead of unilateral
prefrontal brain regions. In accordance with this observation,
the ‘‘Compensation-Related Utilization of Neural Circuits
Hypothesis’’ (CRUNCH; Reuter-Lorenz and Cappell, 2008)
suggests that an increase in neural activity in older adults can
compensate for age-related cognitive decline if the task is not too
demanding. Therefore, recruiting additional or alternative neural
resources (as indicated by enhanced ERP amplitudes) might
explain the absence of age-related behavioral differences in our
study. Along the same lines, a discrepancy between behavioral
and physiological responses is not uncommon (Cabeza et al.,
1997; Mark and Rugg, 1998; Trott et al., 1999), including studies
on semantic memory (Fjell et al., 2005; Duarte et al., 2006). One
possibility is that neuroimaging techniques are sensitive enough
to detect physiological effects, such as age-related functional or
structural changes, which may not necessarily be apparent in
behavioral tests. For example, abnormal EEG measures during
the early stages of Alzheimer’s disease can predict a severe
decline in cognitive functions even when behavioral changes are
not yet evident (Helkala et al., 1991).

With regard to neural oscillations, we observed clear alpha
and theta power effects during retrieval of congruent and
incongruent items that significantly differed between age groups
(Figure 5). While alpha power decreases were more pronounced
in the younger group, theta power increases were more
pronounced in older participants. In general terms, a reciprocal
variation in theta and alpha power during retrieval might relate
to memory processes (Klimesch, 1999), since their interaction
is believed to facilitate information transfer between working
memory and long-term memory (Sauseng et al., 2002). Other
studies have suggested that theta-alpha oscillations bind the
hippocampus, prefrontal cortex, and striatum during recollection
(Herweg et al., 2016). Specifically, the suppression of alpha power
(desynchronization) is associated with attentional and semantic
memory processes during retrieval (Klimesch et al., 1997;
Klimesch, 1999). Theta, on the other hand, has been associated
with several aspects of encoding and retrieval including the
support of associative memory (Herweg et al., 2020). In our task,
the processing of sensory information (congruent or incongruent
stimuli) requires a semantic evaluation, to extract a meaning and
possible associations to prior knowledge.

The observed age affect, as expressed in less decrease
of alpha power but enhanced theta power in the older
subjects (Figure 5), might indicate age-dependent reductions
in attentional processing that might be compensated by higher
retrieval efforts in order to achieve the same behavioral
performance (Cabeza et al., 2018). In line with our observation,
the alpha frequency band reduces with age (Nussbaum, 1997;
Rizzo et al., 2017; Knyazeva et al., 2018), at a rate of ∼0.08 Hz
per year after the age of 60 (Pedley and Miller, 1983), and an
increase in alpha desynchronization has been associated with the
recruitment of additional attentional resources in participants

on early stages of cognitive decline (Deiber et al., 2015). Theta
power, on the other hand, appears with a increased power in
older subjects, compared to younger controls (Silverman et al.,
1955; Nussbaum, 1997; Rizzo et al., 2017). Moreover, older adults
with MCI and Alzheimer’s disease patients have both shown a
remarkable decrease in alpha power, and an excessive increase in
theta power (Jelic et al., 1996; Rossini et al., 2006).

Finally, neural theta, alpha, and beta oscillations did not show
significant differences while retrieving semantically congruent vs.
incongruent information. This was unexpected since all three
frequency bands have previously been associated with learning
and memory processes (Fell and Axmacher, 2011; Hanslmayr
and Staudigl, 2014; Herweg et al., 2020). Since it is difficult to
precisely pinpoint such a null-finding, we refrain from further
speculating about the possible reasons.

To conclude, semantic congruence drives subsequent
long-term recognition memory across the lifespan, and this
effect could be related to neural activity at frontal and left
parietal electrodes in a time window that has previously
been associated with recollection-based recognition memory.
Together with a correlation of ERP responses and behavior, this
indicates that neural retrieval processes play a significant role in
the memory advantage by semantic congruence. As such, our
work gives novel insights into the underlying neurophysiological
mechanisms of the semantic congruence effect across the life


The raw data supporting the conclusions of this article will be
made available by the authors upon reasonable request.


The studies involving human participants were reviewed and
approved by the Ethics Committee at the University of Lübeck.
The patients/participants provided their written informed
consent to participate in this study.


PP, LF, and NB designed the study. PP and TS collected the
data. RA and PP ran the analyses. All authors participated
in discussion to interpret the results. RA and NB wrote the
article, and all the authors participated in revising it. All authors
contributed to the article and approved the submitted version.


This work was supported by the German Research Foundation
(Deutsche Forschungsgemeinschaft, Grant BU 2670/7-1 and
2670/7-2 to NB).


We are grateful to Maxi-Sophie Kuhlmey and Ramona Reineke
for their support in collecting data.

Frontiers in Aging Neuroscience | 11 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging


Amer, T., Giovanello, K. S., Grady, C. L., and Hasher, L. (2018). Age differences in
memory for meaningful and arbitrary associations: a memory retrieval account.
Psychol. Aging 33, 74–81. doi: 10.1037/pag0000220

Amer, T., Giovanello, K. S., Nichol, D. R., Hasher, L., and Grady, C. L. (2019).
Neural correlates of enhanced memory for meaningful associations with age.
Cereb. Cortex 29, 4568–4579. doi: 10.1093/cercor/bhy334

Apitz, T., and Bunzeck, N. (2013). Dopamine controls the neural dynamics
of memory signals and retrieval accuracy. Neuropsychopharmacology 38,
2409–2417. doi: 10.1038/npp.2013.141

Atienza, M., Crespo-Garcia, M., and Cantero, J. L. (2011). Semantic congruence
enhances memory of episodic associations: role of theta oscillations. J. Cogn.
Neurosci. 23, 75–90. doi: 10.1162/jocn.2009.21358

Bastiaansen, M. C. M., Oostenveld, R., Jensen, O., and Hagoort, P. (2008). I see
what you mean: theta power increases are involved in the retrieval of lexical
semantic information. Brain Lang. 106, 15–28. doi: 10.1016/j.bandl.2007.

Battig, W. F., and Montague, W. E. (1969). Category norms of verbal items in
56 categories a replication and extension of the Connecticut category norms.
J. Exp. Psychol. 80, 1–46. doi: 10.1037/h0027577

Bridger, E. K., Bader, R., Kriukova, O., Unger, K., and Mecklinger, A. (2012).
The FN400 is functionally distinct from the N400. NeuroImage 63, 1334–1342.
doi: 10.1016/j.neuroimage.2012.07.047

Bunzeck, N., Doeller, C. F., Fuentemilla, L., Dolan, R. J., and Duzel, E. (2009).
Reward motivation accelerates the onset of neural novelty signals in humans to
85 milliseconds. Curr. Biol. 19, 1294–1300. doi: 10.1016/j.cub.2009.06.021

Cabeza, R., Albert, M., Belleville, S., Craik, F. I. M., Duarte, A., Grady, C. L., et al.
(2018). Maintenance, reserve and compensation: the cognitive neuroscience
of healthy ageing. Nat. Rev. Neurosci. 19, 701–710. doi: 10.1038/s41583-018-

Cabeza, R., Anderson, N. D., Locantore, J. K., and McIntosh, A. R. (2002).
Aging gracefully: compensatory brain activity in high-performing older adults.
NeuroImage 17, 1394–1402. doi: 10.1006/nimg.2002.1280

Cabeza, R., Grady, C. L., Nyberg, L., McIntosh, A. R., Tulving, E., Kapur, S.,
et al. (1997). Age-related differences in neural activity during memory encoding
and retrieval: a positron emission tomography study. J. Neurosci. 17, 391–400.
doi: 10.1523/JNEUROSCI.17-01-00391.1997

Chua, E. F., Schacter, D. L., and Sperling, R. A. (2009). Neural basis for
recognition confidence in younger and older adults. Psychol. Aging 24, 139–153.
doi: 10.1037/a0014029

Cohen, M. X. (2017a). Rigor and replication in time-frequency analyses
of cognitive electrophysiology data. Int. J. Psychophysiol. 111, 80–87.
doi: 10.1016/j.ijpsycho.2016.02.001

Cohen, M. X. (2017b). MATLAB for Brain and Cognitive Scientists. Cambridge,
MA: MIT Press.

Cohen, M. X. (2014). ‘‘Analyzing neural time series data: theory and practice,’’
in Issues in Clinical and Cognitive Neuropsychology CN – QP363.3 .C633 2014,
(Cambridge, MA: The MIT Press).

Craik, F. I. M., and Grady, C. L. (2009). ‘‘Aging, memory and frontal lobe
functioning,’’ in Principles of Frontal Lobe Function, eds D. Stuss, and
R. T. Knight (Oxford, United Kingdom: Oxford University Press), 528–541.

Craik, F. I. M., and Tulving, E. (1975). Depth of processing and the
retention of words in episodic memory. J. Exp. Psychol. Gen. 104, 268–294.
doi: 10.1037/0096-3445.104.3.268

Crespo-Garcia, M., Cantero, J. L., and Atienza, M. (2012). Effects of semantic
relatedness on age-related associative memory deficits: the role of theta
oscillations. NeuroImage 61, 1235–1248. doi: 10.1016/j.neuroimage.2012.

Curran, T. (2000). Brain potentials of recollection and familiarity. Mem. Cogn. 28,
923–938. doi: 10.3758/bf03209340

Danker, J. F., Hwang, G. M., Gauthier, L., Geller, A., Kahana, M. J., Sekuler, R.,
et al. (2008). Characterizing the ERP Old-New effect in a short-term
memory task. Psychophysiology 45, 784–793. doi: 10.1111/j.1469-8986.2008.

Debener, S., Ullsperger, M., Siegel, M., Fiehler, K., Von Cramon, D. Y.,
Engel, A. K., et al. (2005). Trial-by-trial coupling of concurrent
electroencephalogram and functional magnetic resonance imaging identifies

the dynamics of performance monitoring. J. Neurosci. 25, 11730–11737.
doi: 10.1523/JNEUROSCI.3286-05.2005

Deiber, M. P., Meziane, H. B., Hasler, R., Rodriguez, C., Toma, S., Ackermann, M.,
et al. (2015). Attention and working memory-related EEG markers of subtle
cognitive deterioration in healthy elderly individuals. J. Alzheimers Dis. 47,
335–349. doi: 10.3233/JAD-150111

Delorme, A., and Makeig, S. (2004). EEGLAB: an open source toolbox for analysis
of single-trial EEG dynamics including independent component analysis.
J. Neurosci. Methods 134, 9–21. doi: 10.1016/j.jneumeth.2003.10.009

Diana, R. A., Van den Boom, W., Yonelinas, A. P., and Ranganath, C. (2011). ERP
correlates of source memory: unitized source information increases familiarity-
based retrieval. Brain Res. 1367, 278–286. doi: 10.1016/j.brainres.2010.

Dien, J. (1998). Issues in the application of the average reference: Review, critiques
and recommendations. Behav. Res. Methods Instrument Comput. 30, 34–43.
doi: 10.3758/BF03209414

Dodson, C. S., and Krueger, L. E. (2006). I misremember it well: why
older adults are unreliable eyewitnesses. Psychon. Bull. Rev. 13, 770–775.
doi: 10.3758/bf03193995

Dodson, C. S., Bawa, S., and Krueger, L. E. (2007a). Aging, metamemory and
high-confidence errors: a misrecollection account. Psychol. Aging 22, 122–133.
doi: 10.1037/0882-7974.22.1.122

Dodson, C. S., Bawa, S., and Slotnick, S. D. (2007b). Aging, source memory
and misrecollections. J. Exp. Psychol. Learn. Mem. Cogn. 33, 169–181.
doi: 10.1037/0278-7393.33.1.169

Duarte, A., Ranganath, C., Trujillo, C., and Knight, R. T. (2006). Intact recollection
memory in high-performing older adults: ERP and behavioral evidence.
J. Cogn. Neurosci. 18, 33–47. doi: 10.1162/089892906775249988

Düzel, E., Yonelinas, A. P., Mangun, G. R., Heinze, H. J., and Tulving, E. (1997).
Event-related brain potential correlates of two states of conscious awareness in
memory. Proc. Natl. Acad. Sci. U S A 94, 5973–5978. doi: 10.1073/pnas.94.11.

Fell, J., and Axmacher, N. (2011). The role of phase synchronization in memory
processes. Nat. Rev. Neurosci. 12, 105–118. doi: 10.1038/nrn2979

Fjell, A. M., Walhovd, K. B., and Reinvang, I. (2005). Age-differences in verbal
recognition memory revealed by ERP. Clin. EEG Neurosci. 36, 176–187.
doi: 10.1177/155005940503600308

Freitas, S., Simões, M. R., Alves, L., and Santana, I. (2013). Montreal
cognitive assessment: validation study for mild cognitive impairment and
Alzheimer disease. Alzheimer Dis. Assoc. Dis. 27, 37–43. doi: 10.1097/WAD.

Friedman, D. (2013). The cognitive aging of episodic memory: a view based on the
event-related brain potential. Front. Behav. Neurosci. 7:111. doi: 10.3389/fnbeh.

Gilboa, A., and Marlatte, H. (2017). Neurobiology of schemas and schema-
mediated memory. Trends Cogn. Sci. 21, 618–631. doi: 10.1016/j.tics.2017.

Greiner, B. (2015). Subject pool recruitment procedures: organizing experiments
with ORSEE. J. Econom. Sci. Assoc. 1, 114–125. doi: 10.1007/s40881-015-0004-4

Hall, D. M., and Geis, M. F. (1980). Congruity and elaboration in free and cued
recall. J. Exp. Psychol. Hum. Learn. Mem. 6, 778–784. doi: 10.1037/0278-7393.

Hanslmayr, S., and Staudigl, T. (2014). How brain oscillations form memories—a
processing based perspective on oscillatory subsequent memory effects.
NeuroImage 85, 648–655. doi: 10.1016/j.neuroimage.2013.05.121

Hedden, T., and Gabrieli, J. D. E. (2004). Insights into the ageing mind: a view from
cognitive neuroscience. Nat. Rev. Neurosci. 5, 87–96. doi: 10.1038/nrn1323

Helkala, E. L., Laulumaa, V., Soininen, H., Partanen, J., and Riekkinen, P. J. (1991).
Different patterns of cognitive decline related to normal or deteriorating EEG
in a 3-year follow-up study of patients with Alzheimer’s disease. Neurology 41,
528–532. doi: 10.1212/wnl.41.4.528

Herrmann, C. S., Grigutsch, M., and Busch, N. A. (2005). ‘‘EEG oscillations
and wavelet analysis,’’ in Event-Related Potentials: A Methods Handbook,
(Cambridge, MA, USA: MIT Press), 404.

Herweg, N. A., Apitz, T., Leicht, G., Mulert, C., Fuentemilla, L., Bunzeck, N.,
et al. (2016). Theta-alpha oscillations bind the hippocampus, prefrontal cortex
and striatum during recollection: evidence from simultaneous EEG-fMRI.
J. Neurosci. 36, 3579–3587. doi: 10.1523/JNEUROSCI.3629-15.2016

Frontiers in Aging Neuroscience | 12 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

Herweg, N. A., Solomon, E. A., and Kahana, M. J. (2020). Theta oscillations in
human memory. Trends Cogn. Sci. 24, 208–227. doi: 10.1016/j.tics.2019.12.006

Höltje, G., Lubahn, B., and Mecklinger, A. (2019). The congruent, the incongruent
and the unexpected: event-related potentials unveil the processes involved
in schematic encoding. Neuropsychologia 131, 285–293. doi: 10.1016/j.

Jacoby, L. L., and Dallas, M. (1981). On the relationship between autobiographical
memory and perceptual learning. J. Exp. Psychol. Gen. 110, 306–340.
doi: 10.1037//0096-3445.110.3.306

Jelic, V., Shigeta, M., Julin, P., Almkvist, O., Winblad, B., Wahlund, L. O.,
et al. (1996). Quantitative electroencephalography power and coherence in
Alzheimer’s disease and mild cognitive impairment. Dement. Geriatr. Cogn.
Disord. 7, 314–323. doi: 10.1159/000106897

Jung, T. P., Makeig, S., Humphries, C., Lee, T. W., Mckeown, M. J., Iragui, V., et al.
(2000). Removing electroencephalographic artifacts by blind source separation.
Psychophysiology 37, 163–178. doi: 10.1111/1469-8986.3720163

Klimesch, W. (1997). EEG-alpha rhythms and memory processes. Int.
J. Psychophysiol. 26, 319–340. doi: 10.1016/s0167-8760(97)00773-3

Klimesch, W. (1999). EEG alpha and theta oscillations reflect cognitive and
memory performance: a review and analysis. Brain Res. Rev. 29, 169–195.
doi: 10.1016/s0165-0173(98)00056-3

Klimesch, W., Doppelmayr, M., Schimke, H., and Ripper, B. (1997).
Theta synchronization and alpha desynchronization in a memory task.
Psychophysiology 34, 169–176. doi: 10.1111/j.1469-8986.1997.tb02128.x

Klimesch, W., Doppelmayr, M., Yonelinas, A., Kroll, N. E. A., Lazzara, M.,
Röhm, D., et al. (2001). Theta synchronization during episodic retrieval:
neural correlates of conscious awareness. Cogn. Brain Res. 12, 33–38.
doi: 10.1016/s0926-6410(01)00024-6

Klimesch, W., Schack, B., Schabus, M., Doppelmayr, M., Gruber, W., Sauseng, P.,
et al. (2004). Phase-locked alpha and theta oscillations generate the
P1–N1 complex and are related to memory performance. Cogn. Brain Res. 19,
302–316. doi: 10.1016/j.cogbrainres.2003.11.016

Knyazeva, M. G., Barzegaran, E., Vildavski, V. Y., and Demonet, J. F. (2018).
Aging of human alpha rhythm. Neurobiol. Aging 69, 261–273. doi: 10.1016/j.

Krantz, D. H., Atkinson, R. C., Luce, R. D., and Suppes, P. Eds. (1974).
Contemporary Developments in Mathematical Psychology: I. Learning, Memory
and Thinking. San Francisco, CA: W. H. Freeman.

Lakens, D., McLatchie, N., Isager, P. M., Scheel, A. M., and Dienes, Z. (2020).
Improving inferences about null effects with bayes factors and equivalence
tests. J. Gerontol. B. Psychol. Sci. Soc. Sci. 75, 45–57. doi: 10.1093/geronb/

Liu, Z.-X., Grady, C., and Moscovitch, M. (2017). Effects of prior-knowledge on
brain activation and connectivity during associative memory encoding. Cereb.
Cortex 27, 1991–2009. doi: 10.1093/cercor/bhw047

Luck, S. J. (2005). ‘‘An introduction to the event-related potential technique,’’ in
Cognitive Neuroscience CN—QP376.5 .L83 2005, (Cambridge, MA, USA: MIT

Luo, L., and Craik, F. I. M. (2008). Aging and memory: a cognitive approach. Can.
J. Psychiatry 53, 346–353. doi: 10.1177/070674370805300603

Makeig, S., Bell, A. J., Jung, T.-P., and Sejnowski, T. J. (1996). Independent
component analysis of electroencephalographic data. Adv. Neural Inform. Proc.
Sys. 8, 145–151.

Maris, E. (2012). Statistical testing in electrophysiological studies: statistical testing
in electrophysiological studies. Psychophysiology 49, 549–565. doi: 10.1111/j.

Maris, E., and Oostenveld, R. (2007). Nonparametric statistical testing of EEG-
and MEG-data. J. Neurosci. Methods 64, 177–190. doi: 10.1016/j.jneumeth.

Mark, R. E., and Rugg, M. D. (1998). Age effects on brain activity associated with
episodic memory retrieval. An electrophysiological study. Brain 121, 861–873.
doi: 10.1093/brain/121.5.861

McKenzie, S., Frank, A. J., Kinsky, N. R., Porter, B., Rivière, P. D., Eichenbaum, H.,
et al. (2014). Hippocampal representation of related and opposing memories
develop within distinct, hierarchically organized neural schemas. Neuron 83,
202–215. doi: 10.1016/j.neuron.2014.05.019

McKenzie, S., Robinson, N. T. M., Herrera, L., Churchill, J. C., and Eichenbaum, H.
(2013). Learning causes reorganization of neuronal firing patterns to

represent related experiences within a hippocampal schema. J. Neurosci. 33,
10243–10256. doi: 10.1523/JNEUROSCI.0879-13.2013

Morris, C. D., Bransford, J. D., and Franks, J. J. (1977). Levels of processing versus
transfer appropriate processing. J. Verb. Learn. Verb. Behav. 16, 519–533.

Nasreddine, Z. S., Phillips, N. A., Bédirian, V., Charbonneau, S., Whitehead, V.,
Collin, I., et al. (2005). The montreal cognitive assessment, MoCA: a brief
screening tool for mild cognitive impairment. J. Am. Geriatr. Soc. 53, 695–699.
doi: 10.1111/j.1532-5415.2005.53221.x

Nussbaum, P. D. (1997). ‘‘Electroencephalography in the elderly,’’ in Handbook
of NeuroPsychol. Aging (Berlin, Germany: Springer Science+Business Media),

Nyberg, L., Lövdén, M., Riklund, K., Lindenberger, U., and Bäckman, L. (2012).
Memory aging and brain maintenance. Trends Cogn. Sci. 16, 292–305.
doi: 10.1016/j.tics.2012.04.005

Ofen, N., and Shing, Y. L. (2013). From perception to memory: Changes in
memory systems across the lifespan. Neurosci. Biobehav. Rev. 37, 2258–2267.
doi: 10.1016/j.neubiorev.2013.04.006

Oostenveld, R., Fries, P., Maris, E., and Schoffelen, J.-M. (2011). FieldTrip:
open source software for advanced analysis of MEG, EEG and
invasive electrophysiological data. Comput. Intell. Neurosci. 2011, 1–9.
doi: 10.1155/2011/156869

Packard, P. A., Rodríguez-Fornells, A., Bunzeck, N., Nicolás, B., de Diego-
Balaguer, R., Fuentemilla, L., et al. (2017). Semantic congruence accelerates
the onset of the neural signals of successful memory encoding. J. Neurosci. 37,
291–301. doi: 10.1523/JNEUROSCI.1622-16.2016

Packard, P. A., Steiger, T. K., Fuentemilla, L., and Bunzeck, N. (2020). Neural
oscillations and event-related potentials reveal how semantic congruence drives
long-term memory in both young and older humans. Sci. Rep. 10:9116.
doi: 10.1038/s41598-020-65872-7

Park, D. C., Lautenschlager, G., Hedden, T., Davidson, N. S., Smith, A. D.,
Smith, P. K., et al. (2002). Models of visuospatial and verbal memory across the
adult life span. Psychol. Aging 17, 299–320. doi: 10.1037/0882-7974.17.2.299

Pedley, T. A., and Miller, J. A. (1983). Clinical neurophysiology of aging and
dementia. Adv. Neurol. 38, 31–39.

Piaget, J. (1952). The Origins of Intelligence in Children. (New York, NY:
International Universities Press, Inc).

Preston, A. R., and Eichenbaum, H. (2013). Interplay of hippocampus and
prefrontal cortex in memory. Curr. Biol. 23, R764–R773. doi: 10.1016/j.cub.

Rajah, M. N., and D’Esposito, M. (2005). Region-specific changes in prefrontal
function with age: a review of PET and fMRI studies on working and episodic
memory. Brain 128, 1964–1983. doi: 10.1093/brain/awh608

Raz, N. (2000). ‘‘Aging of the brain and its impact on cognitive performance:
integration of structural and functional findings,’’ in The Handbook of Aging
and Cognition, 2nd edition, eds F. I. M. Craik and T. A. Salthouse (Mahwah,
NJ: Lawrence Erlbaum Associate Publishers), 1–90.

Reuter-Lorenz, P. A., and Cappell, K. A. (2008). Neurocognitive aging and the
compensation hypothesis. Curr. Direct. Psychol. Sci. 17, 177–182. doi: 10.1111/j.

Rizzo, M., Anderson, S., and Fritzsch, B. (2017). ‘‘The Wiley handbook on the
aging mind and brain,’’ in The Wiley Handbook on the Aging Mind and Brain,
eds M. Rizzo, S. Anderson, and B. Fritzsch (Hoboken, NJ: John Wiley and

Rodrigues, P. F. S., and Pandeirada, J. N. S. (2014). Attention and working memory
in elderly: the influence of a distracting environment. Cogn. Process. 16, 97–109.
doi: 10.1007/s10339-014-0628-y

Roediger, H. L., III (2008). Relativity of remembering: why the laws of memory
vanished. Ann. Rev. Psychol. 59, 225–254. doi: 10.1146/annurev.psych.57.

Rossini, P. M., Del Percio, C., Pasqualetti, P., Cassetta, E., Binetti, G., Dal Forno, G.,
et al. (2006). Conversion from mild cognitive impairment to Alzheimer’s
disease is predicted by sources and coherence of brain electroencephalography
rhythms. Neuroscience 143, 793–803. doi: 10.1016/j.neuroscience.2006.08.049

Rugg, M. D. (1995). ‘‘ERP studies of memory,’’ in Electrophysiology of Mind: Event-
Related Brain Potentials and Cognition (Oxford, United Kingdom: Oxford
University Press), 133–170.

Rugg, M. D., and Curran, T. (2007). Event-related potentials and recognition
memory. Trends Cogn. Sci. 11, 251–257. doi: 10.1016/j.tics.2007.04.004

Frontiers in Aging Neuroscience | 13 September 2021 | Volume 13 | Article 683908

Alejandro et al. Congruence Effect During Healthy Aging

Salthouse, T. A. (1996). The processing-speed theory of adult age differences in
cognition. Psychol. Rev. 103, 403–428. doi: 10.1037/0033-295x.103.3.403

Salthouse, T. A. (2013). ‘‘Mechanisms of age-cognition relations in adulthood,‘‘ in
Mechanisms of Age-Cognition Relations in Adulthood, (New York: Psychology

Sanquist, T. F., Rohrbaugh, J. W., Syndulko, K., and Lindsley, D. B.
(1980). Electrocortical signs of levels of processing: perceptual analysis and
recognition memory. Psychophysiology 17, 568–576. doi: 10.1111/j.1469-8986.

Sassenhagen, J., and Draschkow, D. (2019). Cluster based permutation tests of
MEG/EEG data do not establish significance of effect latency or location.
Psychophysiology 56:e13335. doi: 10.1111/psyp.13335

Sauseng, P., Klimesch, W., Gruber, W., Doppelmayr, M., Stadler, W., Schabus, M.,
et al. (2002). The interplay between theta and alpha oscillations in the human
electroencephalogram reflects the transfer of information between memory
systems. Neurosci. Lett. 324, 121–124. doi: 10.1016/s0304-3940(02)00225-2

Schönbrodt, F. D., and Wagenmakers, E.-J. (2018). Bayes factor design
analysis: planning for compelling evidence. Psychon. Bull. Rev. 25, 128–142.
doi: 10.3758/s13423-017-1230-y

Schulman, A. I. (1974). Memory for words recently classified. Mem. Cognit. 2,
47–52. doi: 10.3758/BF03197491

Seth, A. K. (2010). A MATLAB toolbox for Granger causal connectivity analysis.
J. Neurosci. Methods 186, 262–273. doi: 10.1016/j.jneumeth.2009.11.020

Shing, Y. L., Werkle-Bergner, M., Li, S.-C., and Lindenberger, U. (2009).
Committing memory errors with high confidence: older adults do but children
don’t. Memory 17, 169–179. doi: 10.1080/09658210802190596

Silverman, A. J., Busse, E. W., and Barnes, R. H. (1955). Studies in
the processes of aging: electroencephalographic findings in 400 elderly
subjects. Electroencephalogr. Clin. Neurophysiol. 7, 67–74. doi: 10.1016/0013-

Sommer, T. (2017). The emergence of knowledge and how it supports
the memory for novel related information. Cereb. Cortex 27, 1906–1921.
doi: 10.1093/cercor/bhw031

Staudigl, T., Zaehle, T., Voges, J., Hanslmayr, S., Esslinger, C., Hinrichs, H.,
et al. (2012). Memory signals from the thalamus: early thalamocortical
phase synchronization entrains gamma oscillations during long-term memory
retrieval. Neuropsychologia 50, 3519–3527. doi: 10.1016/j.neuropsychologia.

Steiger, T. K., Herweg, N. A., Menz, M. M., and Bunzeck, N. (2019). Working
memory performance in the elderly relates to theta-alpha oscillations and
is predicted by parahippocampal and striatal integrity. Sci. Rep. 9, 1–11.
doi: 10.1038/s41598-018-36793-3

The jamovi project (2020). Jamovi 1.0.8 (Computer software). Available online at:

Tromp, D., Dufour, A., Lithfous, S., Pebayle, T., and Després, O. (2015). Episodic
memory in normal aging and Alzheimer disease: insights from imaging and
behavioral studies. Ageing Res. Rev. 24, 232–262. doi: 10.1016/j.arr.2015.08.006

Trott, C. T., Friedman, D., Ritter, W., Fabiani, M., and Snodgrass, J. G. (1999).
Episodic priming and memory for temporal source: event-related potentials
reveal age-related differences in prefrontal functioning. Psychol. Aging 14,
390–413. doi: 10.1037//0882-7974.14.3.390

Tse, D., Takeuchi, T., Kakeyama, M., Kajii, Y., Okuno, H., Tohyama, C., et al.
(2011). Schema-dependent gene activation and memory encoding in neocortex.
Science 333, 891–895. doi: 10.1126/science.1205274

Tulving, E. (1985). Memory and consciousness. Can. Psychol./Psychol. Can. 26,
1–12. doi: 10.1037/h0080017

van Kesteren, M. T. R., Fernandez, G., Norris, D. G., and Hermans, E. J. (2010).
Persistent schema-dependent hippocampal-neocortical connectivity during

memory encoding and postencoding rest in humans. Proc. Natl. Acad. Sci. U
S A 107, 7550–7555. doi: 10.1073/pnas.0914892107

van Kesteren, M. T. R., Beul, S. F., Takashima, A., Henson, R. N.,
Ruiter, D. J., and Fernández, G., et al. (2013). Differential roles for medial
prefrontal and medial temporal cortices in schema-dependent encoding: from
congruent to incongruent. Neuropsychologia 51, 2352–2359. doi: 10.1016/j.

van Kesteren, M. T. R., Rignanese, P., Gianferrara, P. G., Krabbendam, L.,
and Meeter, M. (2020). Congruency and reactivation aid memory
integration through reinstatement of prior knowledge. Sci. Rep. 10:4776.
doi: 10.1038/s41598-020-61737-1

van Kesteren, M. T. R., Rijpkema, M., Ruiter, D. J., Morris, R. G. M., and
Fernández, G. (2014). Building on prior knowledge: schema-dependent
encoding processes relate to academic performance. J. Cogn. Neurosci. 26,
2250–2261. doi: 10.1162/jocn_a_00630

van Kesteren, M. T. R., Ruiter, D. J., Fernández, G., and Henson, R. N. (2012).
How schema and novelty augment memory formation. Trends Neurosci. 35,
211–219. doi: 10.1016/j.tins.2012.02.001

Voss, J., and Paller, K. (2017). Neural Substrates of Remembering: Event-Related
Potential Studies. Amsterdam, Netherlands: Elsevier.

Widmann, A., Schröger, E., and Maess, B. (2015). Digital filter design for
electrophysiological data—a practical approach. J. Neurosci. Methods 250,
34–46. doi: 10.1016/j.jneumeth.2014.08.002

Wilding, E. L., and Rugg, M. D. (1996). An event-related potential study of
recognition memory with and without retrieval of source. Brain 119, 889–905.
doi: 10.1093/brain/119.3.889

Woroch, B., and Gonsalves, B. D. (2010). Event-related potential correlates of item
and source memory strength. Brain Res. 1317, 180–191. doi: 10.1016/j.brainres.

Wynn, S. C., Kessels, R. P. C., and Schutter, D. J. L. G. (2020).
Electrocortical indices of subjectively perceived confidence in episodic
memory. Int. J. Psychophysiol. 151, 18–24. doi: 10.1016/j.ijpsycho.2020.

Yonelinas, A. P. (2001). Components of episodic memory: the contribution of
recollection and familiarity. Philos. Trans. R. Soc. Lond. B Biol. Sci. 356,
1363–1374. doi: 10.1098/rstb.2001.0939

Yoon, C., Feinberg, F., Hu, P., Gutchess, A. H., Hedden, T., Chen, H.-Y. M., et al.
(2004). Category norms as a function of culture and age: comparisons of item
responses to 105 categories by american and chinese adults. Psychol. Aging 19,
379–393. doi: 10.1037/0882-7974.19.3.379

Conflict of Interest: The authors declare that the research was conducted in the
absence of any commercial or financial relationships that could be construed as a
potential conflict of interest.

Publisher’s Note: All claims expressed in this article are solely those of the authors
and do not necessarily represent those of their affiliated organizations, or those of
the publisher, the editors and the reviewers. Any product that may be evaluated in
this article, or claim that may be made by its manufacturer, is not guaranteed or
endorsed by the publisher.

Copyright © 2021 Alejandro, Packard, Steiger, Fuentemilla and Bunzeck. This is an
open-access article distributed under the terms of the Creative Commons Attribution
License (CC BY). The use, distribution or reproduction in other forums is permitted,
provided the original author(s) and the copyright owner(s) are credited and that the
original publication in this journal is cited, in accordance with accepted academic
practice. No use, distribution or reproduction is permitted which does not comply
with these terms.

Frontiers in Aging Neuroscience | 14 September 2021 | Volume 13 | Article 683908

  • Semantic Congruence Drives Long-Term Memory and Similarly Affects Neural Retrieval Dynamics in Young and Older Adults
      • Participants
      • Behavioral Procedures
      • Statistical Analyses of Memory Results
      • EEG Recordings and Analyses
      • ERP Analysis
      • Time-Frequency Analysis
      • Behavioral Findings
        • Encoding
        • Retrieval
      • EEG Findings
        • ERP Cluster Analysis
        • ERP Analyses
        • TF Cluster Analysis
        • TF Analysis for Fz
        • TF Analysis for P3
        • Correlations of ERPs and Behavior



Scientific Reports | (2020) 10:16968 |

Schematic memories develop
quickly, but are not expressed
unless necessary
Alexa Tompary1*, WenXi Zhou2 & Lila Davachi3,4

Episodic memory retrieval is increasingly influenced by schematic information as memories mature,
but it is unclear whether this is due to the slow formation of schemas over time, or the slow forgetting
of the episodes. To address this, we separately probed memory for newly learned schemas as well
as their influence on episodic memory decisions. In this experiment, participants encoded images
from two categories, with the location of images in each category drawn from a different spatial
distribution. They could thus learn schemas of category locations by encoding specific episodes. We
found that images that were more consistent with these distributions were more precisely retrieved,
and this schematic influence increased over time. However, memory for the schema distribution,
measured using generalization to novel images, also became less precise over time. This incongruity
suggests that schemas form rapidly, but their influence on episodic retrieval is dictated by the need to
bolster fading memory representations.

Decades of work in cognitive psychology has demonstrated that schemas can enhance1,2 memory formation. But
while the impact of schemas on new memories is well documented, it is less clear both when schema knowledge
is solidified (accessibility) and when it begins to exert an influence on episodic memories (expression). In other
words, when is a schema formed, and when is it used?.

It is thought that a schema encompasses commonalities across multiple unique experiences, but that the
specific details of each experience are lost over time. Indeed, formal definitions of a schema require that it (1) is
constructed from multiple episodes, and (2) lacks episode-specific details3. How do such structured memories
develop? Neuroscientific theories of systems-level consolidation posit that successful retrieval of episodic memo-
ries is initially supported by the hippocampus, but, over time, memories are supported by distributed cortical
regions through incremental, coordinated reactivation of memories across the hippocampus and cortex4,5. This
process is thought to underpin the slow extraction and cortical representation of statistical regularities common
across overlapping episodes6. Richards and colleagues sought evidence for these theories by simultaneously prob-
ing episodic memory and schematic memory in a water maze experiment. After 30 days, rodents’ swim patterns
more closely matched the learned schema of platform locations, despite weaker memory for specific platforms,
consistent with the idea that schematic memory emerges as memory for specific episodes are forgotten7. We have
shown in prior work that neural patterns in the human hippocampus and medial prefrontal cortex come to reflect
overlap across associative memories after a week, which may support the gradual organization of discrete episodes
into schematic knowledge8. A slowly emerging schematic memory can also explain a wide and diverse family of
behavioral observations that both time and sleep benefit generalization across overlapping experiences, including
(but not limited to) overlapping episodes, statistical regularities, and category members with shared features9–11.
Furthermore, after a delay or a night of sleep, schema-consistent episodes are more likely to be remembered12,13
and remembered with greater precision14 relative to schema-inconsistent episodes.

However, the slow development of schemas is difficult to reconcile with observations that schemas can form
rapidly and are used simultaneously with, or soon after, the encoding of its constituent episodes. Indeed, some
of the earliest observations of schematic memory come from perception research, operationalizing a schema
as a common set of spatial properties across visual displays that participants learn and use within minutes15,16.
Furthermore, newly developed schemas have been shown to influence novel, schema-consistent memories dur-
ing and immediately after encoding the memories that make up the schema12,14,17. These findings suggest that


1Department of Psychology, University of Pennsylvania, Philadelphia, PA 19104, USA. 2Center for Neural
Science, New York University, New York, NY 10003, USA. 3Department of Psychology, Columbia University, New
York, NY 10027, USA. 4Nathan Kline Institute for Psychiatric Research, Orangeburg, NY 10962, USA. *email:
[email protected]



Scientific Reports | (2020) 10:16968 |

schemas can develop quickly and in the absence of time-dependent processing, in contrast to predictions of
systems-level consolidation models.

How might these two different accounts—the slow strengthening of schemas over time, versus their immedi-
ate formation and use—be reconciled? One approach is to disentangle the accessibility of a schematic memory
from its expression during episodic retrieval. One hypothesis would predict that episodic memories are accessible
immediately, while a schema based on those episodes takes time to develop, thus explaining the increasing influ-
ence of schematic representations during episodic retrieval over time (Fig. 1a). A second hypothesis is that both
schematic representations and episodic memories develop concurrently during encoding, but episodic memories
are forgotten over time while schematic memory remains stable (Fig. 1b). According to this view, both types of
memory have formed, but which one is expressed depends on when memory is tested. Soon after encoding, the
episodic memories dominate retrieval because they are more strongly remembered. At this point, the schema
representation is not as necessary to support precise retrieval, so its influence on retrieval is weaker. However, as
the strength of episodic memories decreases, the schematic memory becomes more involved in episodic retrieval.

Both hypotheses predict that schematic memory is more likely to be expressed during episodic memory
retrieval after a delay, when episodic memories are relatively weaker (Fig. 1c). However, these hypotheses offer
different predictions about the strength of memory for the schemas themselves over time. The first hypothesis
predicts weak or no schematic memory early on, while the second hypothesis predicts intact and strong sche-
matic memory. Ideally, a pure assessment of schematic memory would minimize interference from its constitu-
ent episodes. One way to do this is by probing how a schema is recruited to make decisions about novel, yet
related, information (i.e. generalization). While past work has shown that schemas generalize to new stimuli
immediately after encoding, and this generalization is sometimes enhanced after a period of consolidation9,11,18,
the relationship between generalization of a schema and memory for its constituent episodes is less studied.
Quantifying changes in this relationship over time would provide the critical comparison needed to understand
the development of a schema separately from its expression during episodic retrieval.

Figure 1. Two hypotheses of time-dependent changes in the retrieval of newly encoded episodic and schematic
memories, assuming successful encoding of the episodes. (a) After learning its constituent episodes, a schematic
memory emerges over time, while the strength of its constituent episodic memories decreases. This emergence
is thought to be driven by the slow extraction of overlapping information across different episodes. (b) During
learning, a schematic memory forms concurrently with the episodes that it comprises. Over time, it remains
stable as its constituent memories are forgotten. (c) Both hypotheses predict that retrieval of a schema’s
constituent episodes would be more affected by the schema over time, indicated by the yellow-blue gradients



Scientific Reports | (2020) 10:16968 |

In this behavioral experiment, we sought to tease apart the strength of schemas from their influence on con-
stituent episodic memories. Specifically, we developed an experiment to (1) investigate the influence of schematic
memory on episodic retrieval over time, as those component memories become less precise, and (2) probe the
use of schema knowledge when episodic memory retrieval is not required, using a generalization task with new
memoranda. Participants encoded and retrieved associations between images from two categories (animals or
objects) and their locations along a ring (Fig. 2a). The locations of images for each category were drawn from a
probability function centered at opposite ends of the ring (Fig. 2e). Thus participants could learn both episodic
memories (specific image locations) and schematic memories (the likely location of animals and objects). At
immediate and delayed retrieval tests, participants were probed for the location of each image (Fig. 2c, d).
Delayed retrieval was tested either 24 h or 1 week after encoding in separate groups of participants.

To track episodic and schematic memories separately, we developed novel behavioral measures in addition to
adapting ones used in past work. First, we defined an image’s consistency with its category’s ‘location schema’ as
a continuous measure, which differs from typical manipulations of schema-consistency. Specifically, an image’s
schema-consistency varied continuously as images could be closer to or farther from the center of its category’s
spatial distribution (Fig. 3a, top). Continuous-report protocols, commonly used to measure the fidelity of long-
term memory19,20 and recently employed to demonstrate that the precision of schema-consistent memories is

Figure 2. Experiment design. (a) In Session 1, participants completed three cycles of image-location encoding,
followed by an immediate location memory test. Participants completed a delayed test 1 week after encoding
or 24 h after encoding. (b) During encoding, participants dragged images onto their associated location on a
ring, indicated by a red mark. (c) During the immediate test, participants dragged each image onto its retrieved
location on the ring. The image then was presented at its correct location. (d) The delayed test was identical
to the immediate test, but instead of a feedback phase, participants rated their confidence for their location
memory. (e) Radial plot of distributions of animal and object locations for two participants. The locations of
images in each category were determined by sampling from two cosine distributions centered around opposite
points on the ring.



Scientific Reports | (2020) 10:16968 |

preserved over time14, enabled us to track the episodic precision of each association over time (Fig. 3a, bottom
left). To assess the use of a schema during episodic retrieval, we integrated error and schema-consistency into a
novel, subject-unique measure of ‘schema reliance’ (Fig. 3a, bottom right). To compare these measures with other
schema acquisition findings, we adapted a measure used by Richards and colleagues7 to infer rodents’ schematic
memory for the distribution of platform locations (Fig. 3b). Critically, during the delayed test, participants were
also asked to guess where novel images would be located (Fig. 3c). Because these images were never encoded,
guesses about their locations can be used to probe the precision of schematic memories while reducing the use
of specific episodes.

Episodic memory over time. Our primary measure of episodic precision for all image-location associa-
tions was defined as the angular distance, or error, between an image’s encoded and retrieved location (Fig. 3a,
bottom left). Performance was above chance (average error was reliably lower than 90°) at both time points in
both the 24-h and 1-week groups (all t < −5.11, all p < 0.001, all d < −0.95). We directly compared error in both
groups as a function of time in a trial-level mixed effects model with group (24-h, 1-week), time (immediate,
delayed retrieval), and their interaction as discrete predictors of error (Supplementary Table  4). We found an
effect of group (t(56.93) = −2.23, p = 0.03) and of time (t(56.51) = −5.03, p < 0.001). These effects were qualified by an
interaction between group and time (t(56.51) = 4.18, p < 0.001). This interaction was driven by increased error at
the delayed test relative to the immediate test for the 1-week group (t(58.6) = − 6.47, p < 0.001), but not the 24-h
group (t(59.3) = − 0.59, p = 0.56; Fig. 4a). A direct comparison between groups also revealed increased error in the
1-weeek group relative to the 24-h group after the delayed test (t(59.1) = − 3.61, p < 0.001) but not immediately after

Figure 3. Analysis approach. (a) Top: Example retrieval measures for a relatively more schema-inconsistent
animal trial (orange) and a relatively more schema-consistent object trial (teal). Schema consistency was
operationalized as an image’s angular distance from the center of its category distribution; images ranged
continuously in their schema-consistency. Bottom left: Error was defined as the angular distance between an
image’s encoded and retrieved locations. Bottom right: the correlation of error and schema-consistency across
all trials was used as a subject-level measure of schema reliance. For each subject, this was computed across all
tested images, separately for the immediate and delayed tests. (b) Subject-level measure of memory for a schema
used by Richards and colleagues7, operationalized as the Kullback–Leibler divergence (DKL) between encoded
and retrieved locations, here adapted for use with angle values and separately conducted for animal and object
locations. (c) Generalization of a schema was defined as a novel image’s angular distance from the center of its
category distribution, where lower values signify images placed closer to its category center.



Scientific Reports | (2020) 10:16968 |

encoding (t(59.1) = − 0.95, p = 0.35). This confirms the expectation that episodic memory was less precise after one
week relative to after 24 h. Recognition of the images was also less accurate after one week (see Supplementary

Relationship between schema consistency and episodic memory. After confirming that the pre-
cision of episodic memories decreased over time, we next examined how schema-consistency interacted with
this decreased precision, hypothesizing that schematic memory would have a greater influence at a longer delay.
Schema-consistency was operationalized as a continuous measure of the angular distance between an item’s
encoded location and the center of its category distribution, where smaller values indicate images that are closer
to other category members and thus are more consistent with the category’s location schema (Fig.  3a, top).
With the model from the prior section that included group (24-h, 1-week), time (immediate, delayed) and their
interaction, we added schema-consistency (continuous predictor), and its interaction with group, time, and
group by time as fixed effects (Fig.  4b, Supplementary Table  5A). Episodic error, again defined as the angle
between the encoded and retrieved locations of each image, was the dependent variable. We found a significant
effect of schema-consistency (t(55.46) = 7.63, p < 0.001), and no interaction between group and schema-consist-
ency (t(55.46) = 0.07, p = 0.95). This suggests that, when considering both immediate and delayed tests together,
schema-consistent images were more precisely retrieved than schema-inconsistent images, and this effect was
similar across the two groups. Critically, there was a reliable interaction between time and schema-consistency
(t(13099.1) = −2.15, p = 0.03), suggesting that the manner in which schema-consistency influenced error differed
over time. There was no 3-way interaction between time, schema-consistency and group (t(13099.1) = 1.14, p = 0.25).
Separate plots of error by time and schema-consistency for each participant are shown in Supplementary Fig. 3.

We next computed planned separate models for each group (24-h, 1-week) to understand the interaction
between time and schema-consistency (Supplementary Table 5B-C). We focused first on the 1-week group, find-
ing a reliable interaction between time and schema-consistency (t(6713.07) = −2.25, p = 0.03) in addition to main
effects of schema-consistency (t(28.34) = 5.24, p < 0.001) and time (t(29.12) = −5.21, p < 0.001). While images that were
more schema-consistent were more precisely remembered both at the immediate test (t(27.55) = 3.83, p = 0.001)
and delayed test (t(28.76) = 4.53, p < 0.001), the interaction with time suggests that the influence of schema consist-
ency on error was even stronger at the delayed test. In other words, schemas benefited consistent memories over
inconsistent ones at both time points, but after a delay, schema-consistent memories were even more precise
than inconsistent ones—despite an overall reduction in precision at the delayed test relative to the immediate
one. This suggests that schemas help to preserve consistent memories over time14,21,22.

By contrast, in the 24-h group, there was a strong effect of schema-consistency (t(26.92) = 5.53, p < 0.001)
but no main effect of time (t(26.51) = − 1.24, p = 0.23) and no interaction between time and schema-consistency
(t(6392.95) = − 0.79, p = 0.43). There was a positive relationship between schema-consistency and reduced error
at both tests (immediate: t(27) = 4.93, p < 0.001; delayed: t(28.61) = 5.01, p < 0.001). Thus, although these memories
underwent a 24-h period of consolidation, there was no additional benefit of schema-consistency on episodic
memory precision at that time period. When considered together with the observations from the 1-week group,
it seems that participants in the 1-week group predominantly contributed to the interaction between time and
schema-consistency found across both groups. However, because this effect was not reliably different across
groups (indicated by the lack of a three-way interaction between time, schema-consistency, and group), it cannot
be conclusively claimed that the 1-week group solely contributed to this effect.

Observations from the 1-week group suggest that schemas influence episodic memories more strongly over
time, but not in the 24-h group. Interestingly, the overall precision of memories, as measured by the distance
between their encoded and retrieved locations, did not decrease after a 24-h delay, suggesting that the influence
of a schema may not grow after a delay if the constituent memories giving rise to the schema remain strong. In
other words, schematic influence was strongest one week after encoding, when retrieval was less precise overall
relative to the immediate test or relative to the 24-h group. This raises an interesting question—are schematic
influences on episodic memory time-dependent or strength-dependent? To answer this question, we leveraged
the confidence ratings collected during the delayed test. In both groups, high-confident (HC) hits were more
precisely retrieved than low-confident (LC) hits and misses (Supplementary Fig. 1A), suggesting that confidence
ratings reflected memory strength. We then examined whether confidence modulated the influence of schemas on
episodic retrieval. Across both groups, there was a stronger relationship between schema-consistency and error
for LC hits and misses relative to HC hits (Supplementary Fig. 1B, Supplementary Fig. 4), mirroring the stronger
relationship between schema-consistency and error observed at the delayed test relative to the immediate test.

We also found an intriguing non-linear relationship between average error and schema reliance across indi-
viduals. Specifically, participants with moderate average error were more influenced by the location schemas,
while participants with the most and least precise memory were less likely to be influenced by the schemas (see
Supplementary Results, Supplementary Fig. 2; Supplementary Table 9). This relationship was observed in both
groups, at both tests, further highlighting the possibility that the degree of schematic influence may be governed
less by the passage of time and more by the strength of the episodes retrieved.

Memory for encoded distribution. The above results suggest that participants increasingly use knowl-
edge for the learned schemas over time, resulting in more precise memory for images that are most consistent
with their category’s location schema. We next wanted to probe memory for the schemas themselves to ask
whether it increased over time, adopting an approach used by a recent paper7. This approach infers memory for
schemas by considering memory for all encoded images as a distribution of locations. We hypothesized that after
a delay, participants’ distribution of retrieved locations would more closely match their distribution of encoded
locations, despite lower episodic precision, consistent with Richards and colleagues’ observations.



Scientific Reports | (2020) 10:16968 |

Figure 4. Precision by time and schema-consistency. (a) Average error over time. Points represent participants.
Error bars signify standard error of the mean (SEM) across participants. Statistics reflect results of trial-level
mixed-effects model comparisons. (b) The relationship between schema-consistency and episodic precision.
Points represent trials. Error ribbons represent 95% confidence interval (CI). (c) Divergence between encoded
and retrieved locations for each participant. Points represent participants. Error bars signify SEM. **p < 0.01.



Scientific Reports | (2020) 10:16968 |

To test this hypothesis, we calculated the divergence (as measured by Kullback–Leibler divergence, Fig. 3b)
between each participant’s encoded and retrieved locations, separately for the immediate and delayed tests
(Fig. 4c). A 2 (group: 24-h, 1-week) × 2 (time: immediate, delayed) ANOVA revealed no main effect of group
(F(1, 55) = 0.02, p = 0.90, ηp2 = 0.002), no main effect of time (F(1, 55) = 0.78, p = 0.38, ηp2 = 0.01), and a reliable group
by time interaction (F(1, 55) = 7.67, p = 0.008, ηp2 = 0.12). This interaction was driven by a difference in how the
distributions diverged over time across the two groups: participants’ distributions of retrieved locations were
less similar to the distribution of encoded locations after a delay in the 1-week group (t(28) = −2.80, p = 0.009,
d = −0.57) but not the 24-h group (t(27) = 1.26, p = 0.22, d = 0.16). One participant in the 1-week group had a
relatively high divergence score at the delayed test (although not considered a statistical outlier); after removing
this subject, the difference in divergence over time in this group remained statistically significant (t(27) = −2.65,
p = 0.01, d = −0.49).

There was no reliable difference in divergence across groups at either test (immediate: t(55) = 1.06, p = 0.29,
d = 0.27; delayed: t(28) = −1.23, p = 0.22, d = −0.33). However, when excluding data from three outlier participants
in the 24-h group (> 3 SD from the group mean), participants in the 24-h group diverged from encoding signifi-
cantly less than participants in the 1-week group during the delayed test (t(52) = −2.69, p = 0.01, d = −0.74). This
suggests that the longer the delay, the less participants’ distribution of retrieved locations matched the presented
distribution. In summary, participants’ location memory did not grow more similar to the statistical pattern
comprising the encoded locations over time; rather, it diverged with this pattern, in contrast to what Richards
and colleagues observed (see Discussion).

Schema generalization: a ‘pure’ test of schema memory. So far, we have demonstrated that par-
ticipants’ memory is more likely to be influenced by a learned schema in cases when episodic memory strength
(measured here as episodic precision) is reduced. This implies that the use of a schema during episodic retrieval
depends on whether or not it is needed to bolster memory decisions. However, these observations and, in our
read, the related literature in this area adopt a measure of schema memory that may be bolstered by access to
individual constituent episodic memories during retrieval. Thus, in order to more directly measure the forma-
tion of the schemas, independent from relying on specific episodic memories, we probed if and how participants
generalized the schemas to novel images, thus decreasing the need to retrieve specific, precise memory episodes.

To this end, novel images were interleaved with old images during the delayed test and participants were
instructed to place them on the ring using any information they learned during encoding. Generalization was
operationalized as the angular distance between each novel image and its category’s center location, where lower
values indicates guesses were closer to the center and thus indicate better generalization (Fig. 3c). Average gener-
alization in both groups was reliably above chance (24-h: t(27) = −8.30, p < 0.001, d = − 1.57; 1-week: t(28) = −5.01,
p < 0.001, d = −0.93). Given the observed differences in memory precision between groups at the delayed test,
we next asked whether the placement of new images also differed by group, using a mixed-effects model with
trial-level generalization as the dependent variable and group (24-h, 1-week) as the independent variable (Sup-
plementary Table 6). Interestingly, we found that participants in the 24-h group placed new images closer to
their category centers than participants in the 1-week group (t(54.72) = −2.60, p = 0.01; Fig. 5a). This suggests that
participants’ memory of a schema became less precise over time and shows forgetting of the presented distribu-
tion. This observation is especially interesting when considered against the influence of schemas on episodic
memory: on one hand, memory for the schemas degraded between 24 h and 1 week after encoding (Fig. 5a),
but over the same time frame, they increasingly influenced memory for old images (Fig. 4b, delayed test). This
suggests that over time, schemas are increasingly used to support the retrieval of episodic memories despite less
precise memory for the schemas themselves.

Relationship between schema generalization and episodic memory. In the 1-week group, par-
ticipants’ generalization of a schema was lower than that of the 24-h group, even though their episodic precision
was more influenced by schematic memory. This discrepancy led us to wonder whether the manner in which
participants retrieved specific episodes was related to their schema generalization. One possibility is that par-
ticipants with more precise episodic memories would be better equipped to generalize schematic information
to new items. Another possibility is that participants who relied most on schemas during episodic retrieval
would be better at generalizing the schemas to new images. To differentiate between these two possibilities, we
separately correlated participant’s average generalization to new images with two behavioral measures that con-
sidered memory for the old images: their average episodic precision and their schema reliance (as measured by
the correlation between schema-consistency and episodic precision for each participant, Fig. 3a bottom right).

Focusing first on the relationship between precision and generalization, we computed a linear regression
with group and participants’ average error at the delayed test as independent variables and participants’ average
generalization (distance between novel images’ locations and their category center) as the dependent meas-
ure. This regression found that precision positively correlated with generalization (β = 0.69, SE = 0.09, t = 7.99,
p < 0.001) with an effect of group (β = 12.75, SE = 6.29, t = −2.03, p = 0.048) and a reliable interaction (β = −0.18,
SE = 0.08, t = −2.11, p = 0.04). In both groups, participants with greater error placed new images further from
their category centers (1-week: r(27) = 0.78, p < 0.001, 24-h: r(27) = 0.71, p < 0.001; difference in correlations: z = 0.55,
p = 0.58; Fig. 5b), although the significant interaction indicates that schema-consistency explained more vari-
ance in precision in the 1-week group. This suggests that participants with more precise episodic memories also
exhibited better memory for the corresponding schemas. Alternatively, since the majority of encoded images
were schema-consistent (i.e. close to their category center), better memory for these constituents of the schemas
may also reflect better learning of schemas themselves. These interpretations are not mutually exclusive, but



Scientific Reports | (2020) 10:16968 |

interestingly, both conflict with the notion that forgetting of details of specific episodes allows for the generaliza-
tion of information across them23.

In a separate regression, we entered group and schema-reliance (each participant’s correlation between
schema-consistency and episodic precision) as predictors of new image generalization (Fig. 5c). There was
a strong negative correlation between generalization and schema-reliance (β = −62.47, SE = 12.06, t = −2.77,
p = 0.008), an effect of group (β = −8.10, SE = 2.69, t = − 3.01, p = 0.004), and no reliable interaction (β = 10.33,
SE = 12.06, t = −0.86, p = 0.40). In both groups, participants with greater schema reliance placed new images
closer to their respective category centers (1-week: r(27) = −0.75, p < 0.001; 24-h: r(27) = −0.42, p = 0.02; difference
in correlations: z = 1.87, p = 0.06). This suggests that participants with better generalization, or better memory
of the schemas, were also more influenced by schemas when retrieving specific episodes.

We next asked whether episodic precision or schema reliance was uniquely associated with the placement of
new images. This was important to consider given the non-linear relationship between episodic precision and
schema-reliance observed across all participants (Supplementary Fig. 2). To do this, we computed a linear regres-
sion with participants’ average generalization as a dependent variable, and three predictors: episodic precision,
schema reliance, and group. This revealed an effect of precision (β = 0.53, SE = 0.06, t = 8.28, p < 0.001) and of
schema-reliance (β = −49.92, SE = 8.02, t = −6.22, p < 0.001), but no reliable effect of group (β = − 1.02, SE = 1.48,
t = −0.687, p = 0.50). Taken together, these observations demonstrate that both episodic and schema-influenced
memory were associated with participants’ generalization of schemas, regardless of whether this generalization
task occurred 24 h or 1 week after encoding.

The goal of this study was to better understand both the emergence and expression of schemas over time. To
do this, we developed behavioral measures to separately quantify memory for a schema and its influence on the
retrieval of specific episodes, and we used these measures to track changes in episodic and schematic memory
over time. We found that memory for specific episodes was more precise when the episodes were consistent
with a learned schema, and this modulation by schema-consistency was stronger when memories were weaker
overall. In other words, schema-consistency more strongly mapped onto variation across memories that were
tested after a longer delay and memories that were remembered with low confidence. However, despite this
increasing influence of schemas over time, participants’ memory for the schemas themselves declined over
time as well, as reflected by the distribution of their retrieved responses and their ability to generalize to novel

Figure 5. New items. (a) Generalization during the delayed test for the 24-h and 1-week groups. (b)
Correlation between the precision of episodic memories and schema generalization across participants. (c)
Correlation between schema-reliance during retrieval (correlation between schema-consistency and episodic
precision) and schema generalization across participants. Points represent participants. Error bars signify SEM.
*p < 0.05.



Scientific Reports | (2020) 10:16968 |

images. The observations from these measures can be distilled into two conclusions: although memory for new
schemas themselves decline over time, they more strongly influence episodic retrieval when those episodes are
weakly remembered.

Schematic influences on episodic memory. We first focused on when schemas are expressed by meas-
uring how they influenced the precision of episodic memories. We investigated this relationship across two
factors that reflected differences in episodic memory strength—the delay between encoding and retrieval, and
participants’ reported confidence. We found a strong relationship between schema-consistency and precision,
where images that were closer to their category center were retrieved more precisely on the immediate test.
This relationship became even stronger after 1-week of consolidation, but not after 24-h. This is consistent with
observations that schemas facilitate memory retrieval for more consolidated memories13,21,22. Moreover, we rep-
licate observations from a similar study, in which participants learned that most images belonging to the same
category were located in a certain quadrant of a ring. In that study, the precision of location memories was pre-
served over 48 h for images whose locations were consistent with their category’s location schema, but decreased
for images located in a different quadrant14. These findings, however, could be interpreted in one of two ways:
schemas were strengthened over time, becoming more influential over episodic retrieval (Fig. 1a), or, as episodes
weakened over time, participants increasingly relied on schemas that had formed during encoding (Fig. 1b). Our
results provide evidence for the latter interpretation, suggesting that in circumstances where episodic memories
were less precise overall (at the delayed test for the 1-week group only), retrieval of each location relied more on
schematic memory. Another finding that supports this interpretation is that participants’ reported confidence
was also associated with their schema reliance. Specifically, the relationship between schema consistency and
error was stronger for low-confident hits relative to high-confident hits both at 24 h and at 1 week. This suggests
that reliance on schema information in this experiment was primarily driven by the strength of each episode:
weaker memories, either tested 1 week after encoding or remembered with low confidence, were more prone to
influence from schemas.

Schematic memories are forgotten over time. If the expression of a schema is altered based on the
strength of the episodes being retrieved, it becomes difficult to disentangle whether schemas become more influ-
ential over episodic memory retrieval because they are formed and strengthened over time, or because episodic
precision decays over time. To disentangle these possibilities, we developed a generalization test in order to
probe memory for the schemas themselves while minimizing the retrieval of specific episodes. We found that
this generalization was better in the 24-h group relative to the 1-week group, suggesting that knowledge of the
schema memories actually decreased over this time interval. Furthermore, a separate measure of schematic
memory used in prior work7 paralleled the results of this generalization test. By characterizing memory for a
schema as the divergence between participants’ distribution of retrieved locations relative to the locations they
had encoded, we found that memories for schemas decayed over time in the 1-week group. However, less precise
memory for the schemas was accompanied by an increase in their expression—specifically, relative to the 24-h
group, the 1-week group had less precise memory of the schemas, but their episodic precision grew more influ-
enced by the schemas over time. This demonstrates that, over time, increases in the expression of schemas—in
other words, the extent to which they influence episodic retrieval—is not necessarily driven by their slow devel-
opment, but rather by their growing need. The need for schematic memory seems to depend on the strength
of episodic memories, and because these memories decay rapidly over time, retrieval is increasingly reliant on
schematic rather than episodic memory.

The finding that schema memory decreased over time was surprising, as we had predicted that there would be
a stabilization or even improvement in schema knowledge over time. Indeed, there is ample past work in humans
showing that time and sleep benefits the extraction and generalization of statistical regularities10,11,24, the neural
and behavioral integration of overlapping associations8,9,25, and new word or grammar learning18,26–28. Of particu-
lar relevance are observations of enhanced category learning after a delay19,20—improvements in the classification
of new stimuli over time, using newly learned categories, precisely mirror the learning of a category-location
mapping that we have operationalized here as schematic memory. One reason why our findings may conflict with
this past work is that the experimental time windows adopted are different. Many studies, in particular those that
study the effect of sleep, compare generalization measured directly after encoding and after only one night of
sleep, while in our experiment, new images were introduced in both groups one day or one week after encoding.
In our study, participants’ ability to generalize may have improved between encoding and the first delayed test,
but we did not capture changes in behavior associated with the first night of sleep. In other work that tracked
changes in generalization beyond the first night of sleep, participants’ ability to generalize a rule to new stimuli
did not improve but rather remained stable31. It may be that many time- and sleep-dependent improvements in
generalization, in particular for newly created schemas, may not be permanent, but rather emerge after a night
of sleep and then decay at a slower rate than memory for episodes. A slower rate of decay for schematic memory
would account for the continued and even increasing use of schematic memory to aid the retrieval of specific
episodes, and may explain observations that schematic memory persists in the absence of any specific episodes
when retrieval is tested at even longer retention intervals32.

Another potential reason why schema knowledge decreased over time in our experiment, rather than increas-
ing or remaining stable, is that the experiences that formed the schemas were learned all at once in a compact
encoding session. In an experiment in which image-location associations were frequently re-encoded across
multiple sessions, recall of the schema-related information remained stable for up to 302 days33. Furthermore, our
results differ from a similar protocol that directly inspired the development of our own experiment, conducted in
rodents studying schema knowledge, or pattern extraction of the locations of platforms in a water maze7. In this



Scientific Reports | (2020) 10:16968 |

study, rodents’ swim patterns more closely matched the distribution of platform locations, while crossing fewer
specific platform locations, after 30 days relative to 1 day after training. Although we also find that schematic
memories increasingly influence episodic retrieval over time, our finding that, when tested separately, schematic
memory decays over time, is difficult to reconcile with this result. One reason for the discrepancy with these
two past experiments is that participants in our experiment may have been given sufficient encoding trials to
learn the location of each image and also extract the pattern of locations by category by the end of encoding.
Indeed, if this procedure is thought to be akin to a categorization task, human participants are easily able to learn
to classify complex stimuli along multiple dimensions over the course of encoding34. Moreover, recent spatial
memory experiments in humans show that in a compact encoding session, location schemas can be learned
immediately35,36 and decay over time35, in line with our findings. There is also new evidence in rodents that neural
signatures of schematic memories in the prefrontal cortex become activated during encoding, yet only mature
with time37, suggesting that the groundwork for the extraction of patterns and formation of a schema is present
immediately and does not require time to emerge. It could be that, in the absence of a strenuous encoding task,
and in cases where encoding is spaced over time, consolidation processes allow for the gradual development of
schematic knowledge over time6. Indeed, spaced and repeated encoding may capture the emergence of knowl-
edge in a manner that is more ecologically valid than the schemas learned in our experiment. However, such
manipulations would be unable to tease apart the acquisition of a schema from its influence on the retrieval of
specific episodes, a benefit that only our protocol provides.

Time‑dependent changes in episodic and schematic memory. Taken together, the results reported
here suggest that while the precision of schematic memories decreases over time, they nevertheless become
increasingly influential in episodic memory retrieval. This pattern of results partially supports the hypothesis
that as episodes weaken over time, their retrieval relies more on retained schematic information (Fig.  1b). In
contrast to this account, however, we also found that memory for the schemas grew weaker over time, rather
than remaining stable. This discrepancy calls for an adjustment, where both episodic and schematic memories
are formed during learning, but in the absence of any further stabilization or reinforcement, schemas decay at
a slower rate than that of their constituent episodes (Fig.  6a). As mentioned above, a slower rate of decay for
schemas over episodes would still result in their increasing influence on episodic retrieval (Fig. 6b). Note that
a decrease or decay in memory strength in this framework could be driven by multiple phenomena: forgetting
may arise from either decay of the original memory trace or decreasing accessibility to the trace. Our experimen-
tal design cannot adjudicate between these two possibilities, but future work may be helpful for understanding
the differential trajectories of forgetting in schematic versus episodic memories; for instance, episodic memories
may be more susceptible to interference, leading to faster forgetting.

Another way to characterize memory for a schema is to probe how it relates to memory for the episodes
that gave rise to it. For example, participants with precise knowledge of the central distribution of a category
may also have precise memories of each image’s location. Or, participants whose episodic precision was most
influenced by schemas may have better memory for the schemas themselves. To test these two possibilities, we
developed separate measures of episodic precision and schema-reliance for each participant and assessed their
contribution to the generalization of new items across participants. We found that, at both 24 h and 1 week,
good generalization was associated with both high episodic precision and with a stronger reliance on schematic

Figure 6. Alternative theory of episodic and schematic memory over time. (a) During learning, a schematic
memory forms concurrently with the episodes that it comprises. Over time, it decays, but at a slower rate relative
to the decay of its constituent memories. This is different from both hypotheses presented in the introduction,
which posited that schematic memory would either grow or remain stable over time while episodes decayed. (b)
Like the two hypotheses originally presented, retrieval of a schema’s constituent episodes would be more affected
by the schema over time, indicated by the yellow-blue gradient.



Scientific Reports | (2020) 10:16968 |

memory. This suggests that the relationship between episodic and schematic memory is bidirectional: while
schematic information can aid episodic retrieval, the memory for a schema is also related to memory of its con-
stituent episodes. It further suggests that these two forms of memory develop together, co-exist and can be used
interchangeably. Note that even though generalization trials consisted of novel images, generalization of a schema
is only an indirect measurement of its strength. It is possible that participants may have relied on memories for
encoded images to make their guesses for new images—akin to an exemplar model of categorization38—and
worse generalization over time may be driven by forgetting of those individual images. Interestingly, this finding
contradicts the notion that schematic or general memories are formed through the forgetting of idiosyncratic
details of overlapping episodes23. We instead find that better memory for the episodes that give rise to a schema
are associated with a more precise use of the schema when making decisions about novel information. One
interpretation of this finding is that participants who were more engaged in the tasks exhibited more precise
memory and better generalization. However, the fact that reliance on schemas during retrieval also was associ-
ated with generalization makes this interpretation unlikely. Rather, it suggests that a specific pattern of retrieval
errors, guided by schematic information, was related to knowledge of the schema itself. While there are several
experimental protocols that measure both the influence of a schema and the memory for its constituent epi-
sodic memories31,39, there have been few attempts to understand the relationship between the two. Future work
is therefore needed to understand the different ways by which schematic information can be extracted from
multiple distinct experiences.

Relationship to other literatures. In our study, we defined a schema as a distribution of locations along
a ring at which images of a particular category were likely to be encoded. The majority of encoded images were
located near the central tendency of this distribution, with some images encoded farther away. This is different
from much of the schema literature in two ways: (1) our measures of schema-consistency and memory precision
are continuous, measured by each image’s proximity to the central location of all images of the same category,
and the distance between its encoded and retrieved location, respectively, and (2) participants needed to map
prior knowledge of category membership onto new, trial-specific spatial information to learn a new schema
about the locations of images. Past uses of continuous retrieval reports have enabled researchers to distinguish
the precision of episodic memories from their overall retrieval success40,41, reveal subtle memory deficits in
healthy aging and in patients with altered MTL function42, and map separate neural contributions to differ-
ent aspects of memory retrieval, such as precision, confidence, and vividness43,44. In contrast, most research in
schematic memory tends to discretize memoranda as schema-consistent or not, using prior knowledge that par-
ticipants already know, like famous faces, word pairs that are semantically related, or dot patterns that resemble
letters45–47, although there are numerous exceptions11,16,17,31. Despite these differences, we find that images that
are closer to their category’s central location are more precisely remembered relative to ones that are farther
away, consistent with many past observations that schemas facilitate memory for consistent information1,47–50.

The use of a newly formed schema in our experiment bears interesting resemblance to what is known about
associative learning. By combining prior knowledge of categories with new memories of image-location associa-
tions, participants were able to learn that a general location on the ring was associated with a particular category
and use that knowledge to generalize to new images and protect memory for related individual locations. Simi-
larly, there is an abundance of work demonstrating that associative learning can generalize to similar stimuli51–53.
Interestingly, associative learning has been examined at different levels of resolution, suggesting that it can be
disentangled into specific, exemplar-based representations that do not generalize and more adaptive associations
that may generalize to similar stimuli54, a dichotomy that strikingly parallels the dissociation between episodic
and schematic memory. Despite these similarities, there seem to be fundamental differences between schema
formation and associative learning. First, schemas are thought to be formed from multiple units3, which in our
case is the building of an association out of a distribution of individual image-location pairs. This basis in multiple
episodes is what separates schematic memory from associative learning of one-to-one mappings between pairs
or chains of stimuli, like stimulus-stimulus or stimulus-outcome pairings. Furthermore, unlike in traditional
associative learning protocols, the category location is not directly trained or reinforced—it is only induced
through the combination of participants’ prior knowledge of categories and memory for specific image locations.
To date, it is unclear whether the manner in which a more general memory is formed, be it a schema or a broad
pattern of associative learning, has consequences for how it is employed. Future empirical comparisons between
schema formation and associative learning may better elucidate how differences in their acquisition give rise to
similarities in how they guide future behavior.

Our investigations of the relationship between schemas and episodes can be re-framed as a question of how
memory is reconstructed out of multiple sources of information55,56. In line with a reconstruction account of
memory, integration across schematic and episodic information can either enhance or impair memories, depend-
ing on whether that information conflicts. In line with this framework, past work has shown that prior knowledge
can both enhance and distort episodic retrieval47,57,58. In the present study, episodic precision could be affected
both by specific episodic memory for an image’s location and by schematic knowledge of the distribution of
locations of each category. Integrating across these two sources of information would result in better memory
for images that were closer to its category center, since the episodic and schematic memories are providing
closely-aligning and reinforcing information. At the same time, memory for an image far from its category’s
center would be less precise, because the schematic information is in conflict with the episodic information. Our
observation that the precision of memory positively correlates with its proximity to its category center, regardless
of the time tested or confidence, provides direct support for this reconstruction account. Another hypothesis
that a reconstruction account offers is that memory for images located far from their category center would be
biased towards it14,59,60. Such a systematic distortion in memory would reflect a strong influence of the schema



Scientific Reports | (2020) 10:16968 |

that is at odds with the true experience. Future experiments designed to capture such biases would provide a
complementary approach towards understanding how schemas influence episodic retrieval.

Limitations and avenues for future work. There are a few limitations to this study to be considered.
First, all encoded images belonged to a schema, so we do not have a measure of ‘pure episodic retrieval’ in the
absence of schematic influences. A third condition, with images from a third category that were located uni-
formly around the ring, would be needed to establish a pure baseline measure of episodic precision over time.
While not the focus of the current study, without such a baseline, we cannot know whether differences in mem-
ory precision by schema-consistency are facilitated or impaired relative to precision in the absence of a schema.
In a similar item-location memory experiment61 that includes items clustered by a location schema along with
items in a no-schema condition, continuous retrieval reports were decomposed into precision and accessibility
(the likelihood of retrieving a particular memory rather than guessing). In this study, Berens and colleagues
report that memories associated with a schema were more accessible, but less precise, relative to memories in
the no-schema condition. At first glance, this finding may seem at odds with our results that schema-consistent
memories are more precise than schema-inconsistent ones. However, this apparent discrepancy is well explained
by two factors: (1) the use of different precision measures, where ours encompassed all items, while Berens and
colleagues calculated precision only from items considered to be true memories, excluding likely guesses, and
(2) the notion that retrieval of an item relies both on memory of its category’s location schema and memory of
its specific location. When considering both of these factors, items with no related schema (like those presented
in Berens et al.) are indeed less likely to be remembered, or less accessible, because they have no schema memory
supporting them. However, the memories that are accessible are likely those that are the strongest and therefore
the most precise. By contrast, items with an associated schema are more likely to be remembered (more acces-
sible), but with less precision, because the knowledge of the schema additionally supports these memories and
thus memory for the specific location need not be as strong. In our data, this reliance on the location schema
underlies our finding that schema-consistent images are more precisely remembered than schema-inconsistent
ones, highlighting that memory is affected by the distribution of the schema. One test of this interpretation
would be to analyze precision for all items from Berens et al.’s schema condition as a function of their distance
from the center of their location schema, as we have done in the current analyses. When precision is computed
over all trials, including likely guesses, we predict that memory for schema-consistent items would be more pre-
cise than that of no-schema items, as these memories would have the additional support of the related schema
such that both strong memories and guesses would result in retrieval close to their encoded location. In con-
trast, memory for schema-inconsistent items would be less precise than no-schema items, because guesses for
schema-inconsistent items would be in conflict with memory for the related schema, leading to distortion60. A
second limitation is that, when possible, we tested for within-subject relationships between schema-consistency
and episodic precision, and we also were able to leverage differences between the immediate and delayed tests
within each group as within-participant controls. However, many of our critical comparisons of performance, in
particular comparisons of schema generalization, were computed using between-subjects analyses. Because of
this, any behavioral differences between these conditions may be driven in part by variance across participants,
rather than differences in memory over time. Note, however, that the performance of these two groups at the
immediate test was matched in various ways: their episodic precision did not reliably differ, the divergence of
their remembered locations from the distribution that generated their encoded locations did not differ, and
neither group exhibited a reliable relationship between schema-reliance and average error across participants.
While these between-subjects analyses are useful, future work will be better equipped to quantify differences in
forgetting of episodic and schematic memory using a fully within-subject design.

It is also important to note that our measure of generalization to new images was developed to minimize
the use of specific episodes when assessing the precision of knowledge for the location schemas. However, it is
likely that participants’ memory for specific image locations still informed their performance on this task—for
example, participants may have placed a new image of a tiger at the location that a lion was encoded. Although
we did not systematically vary the semantic similarity of category members, which would have allowed us to
test this possibility, we did observe a correlation between episodic precision and generalization, suggesting that
less precise memory for episodes would explain less precise memory for the schemas over time. However, the
observation that generalization decreased over time is still interesting to consider in light of the observation
that participants’ memory for encoded images was increasingly influenced by their schema knowledge after a
delay. If it is the case that both episodic precision and schematic knowledge decreased over time, the notion that
participants increasingly relied on schemas over the same time interval suggests a more complex relationship
between the two sources of memory. Since schema knowledge by definition is built by accumulating information
across separate episodes, it is likely that no behavioral measure is capable of assessing the acquisition of schema
knowledge in the complete absence of memory for its episodes—thus, we aimed to minimize the influence of
specific episodes, rather than completely eradicate it. Another disadvantage of the generalization measure was
that it assumed participants’ schema knowledge was reflected by their estimate of the center of each category’s
location schema, such that images placed closer from the center was interpreted as better schema knowledge.
An alternative explanation is that participants who happened to be more variable in their responses, perhaps
due to fatigue, placed images anywhere in the category’s section but not necessarily close to its center. If so, our
measure fails to capture the extent of their schema knowledge. Future work, perhaps with more sophisticated
approaches for measuring schema knowledge, may be able to isolate the bidirectional interactions between
schema knowledge and episodic memory.

In summary, we provide evidence that the formation of a schema can be disentangled from its expres-
sion, demonstrating that schemas are increasingly used as episodic memory decays over time, even as precise



Scientific Reports | (2020) 10:16968 |

knowledge of the schemas decays. We provide a comparison of various behavioral measures used to track
memory for specific episodes and assess knowledge of the schemas that emerge through their integration, with
an eye towards improving the analytical tools to disentangle the acquisition and transformation of these forms
of memory in future work. Finally, we propose that considering the relative strengths of schematic and episodic
memory provides testable predictions about how this information is integrated to support retrieval.

Participants. Participants were recruited from the New York University (NYU) subject database and were
compensated with class credit or money. 29 subjects (9 males) participated in the 1-week group and 28 (7 males)
in the 24-h group. Eligible subjects were: (a) between 18 and 30 years old, (b) right-handed, (c) native English
speakers or English speaking for 10 years, (d) normal or corrected to normal vision and (e) normal color vision.
Subjects provided written Informed Consent at the beginning of the study. The research was approved by the
University Committee of Activities Involving Human Subjects (UCAIHB) at New York University. All methods
were carried out in accordance with relevant guidelines and regulations.

Stimuli. Images. 168 color images of everyday objects were used in this study. The images were selected
either from Hemera Photo-Objects 10000 Premium Image Collection (http://www.bmsoft ware
aphot oobje cts10 000.htm) or from Google Image searches. Images were cropped to 400 × 600 pixels such that
each object was centered and occupied a consistent proportion of the image. For each subject, the images were
randomly divided into two sets: 120 for encoding, and 48 for novel foils in the delayed test.

Locations. For each subject, the center of one category’s distribution was randomly chosen within 0 to 2π radi-
ans. Angles within 0.0873 radians (5°) of the vertical and horizontal axes were excluded. The second category’s
center was located π radians away (on the opposite side of the ring). Then, for each category, 60 values were
randomly drawn from a cosine distribution centered around the chosen angle:

These randomly drawn values became the angular locations along the ring, where θ = 0 pointed to the right and
θ = π/2 pointed to the top of the screen. For each category, the resulting 60 locations were randomly paired with
the 60 images chosen for encoding.

The use of cosine functions ensured that when collapsing across category, the distribution of all 120 locations
was roughly uniform and participants could not learn to bias their mouse movements towards particular sections
of the ring. Separate chi squared tests for each participant confirmed that the overall distribution of images was
not reliably different from a uniform distribution in both experiments (all χ2 < 15.50, all p > 0.07).

Software. Stimulus presentation code was written in Matlab 2014 using the Psychophysics Toolbox extension62.
All analyses and statistical tests were conducted using R63.

Procedure. The experiment took place in a soundproof testing room. The experiment consisted of two ses-
sions, with Session 2 taking place one week after Session 1 in the 1-week group and 24 h after Session 1 in the
24-h group (Fig. 2a). All other aspects of the design are identical across the two groups. In Session 1, subjects
completed encoding and immediate retrieval. In Session 2, subjects completed delayed retrieval. Instructions for
each task were given right before the task both verbally by the experimenter and in written form on the screen.

Encoding. Subjects were instructed to learn the association between each image and its location along the ring.
They were told that their memory for the images’ locations would be tested later in the experiment. Encoding
was divided into six 60-trial blocks, with blocks 1–2, 3–4, and 5–6 forming three cycles of stimulus presentation.
Within each learning cycle, all image-location pairs were presented once in random order. All pairs were thus
viewed 3 times. Each block was separated by a one-minute break.

On each trial (Fig. 2b), an image was presented at the center of the screen for 2 s. Then, a ring with a radius of
400 pixels and width of 40 pixels appeared with a smaller version of the image (4% of its original size) at its center.
The location paired with the image was marked on the ring by a red line. Subjects were asked to move the image
to the red line and click the left mouse button when the image was at the target location. The mouse cursor was
linked to the center of the image such that any mouse movement moved the image. To ensure that participants
were engaged during the encoding task, clicks for images that were 40 pixels or closer to the target location were
interpreted as valid responses. All others would trigger a warning sound that prompted the subjects to move
the image closer to its location. After a valid response, or if no valid response was made within 3 s, the image
would appear at its associated location. The image remained at its location for 2 or more seconds depending on
the speed of the response, so that the duration of all trials was matched at 7 s. Trials were separated by a 0.5 s
fixation period. The mouse position was reset to the center of the ring at the beginning of the subsequent trial.

Immediate retrieval. Directly after completing the encoding phase, participants completed the immediate
retrieval test (Fig.  2c). In this test, subjects were asked to retrieve the locations associated with the presented
images and instructed to guess if they could not remember the correct location. They were also instructed to
use the provided feedback as a final opportunity to learn the image-location pairs in preparation for a future

f (x) =
cos (x − center) + 1




Scientific Reports | (2020) 10:16968 |

memory test. All 120 image-location pairs were tested in random order, divided into two 60-trial blocks sepa-
rated by a one-minute break.

The presentation, timing, and mouse movements in this task were identical to the parameters from encoding,
with the exception that the red line marking the location of each image was not shown. Any click on the ring
was categorized as a valid response. Mouse clicks when the image was not on the ring would trigger a warning
sound to prompt the subjects to move the image onto the ring. If a valid response was made, or if no response
was made within 3 s, the image would appear at the location paired with it. This feedback was included in order
to minimize potential retrieval-induced error in the delayed test. Importantly, immediate retrieval differed from
delayed retrieval in that no novel images were shown (see below); this was to minimize participants’ attention to
the location schemas and reduce the influence of this knowledge on the consolidation of memories of specific
item-location associations.

Delayed retrieval. During the delayed retrieval test (Fig. 2d), participants viewed all 120 pairs interleaved with
the 48 novel foils. The order of presentation was pseudo-randomized so that no more than two consecutive
trials were of foils. Subjects were asked to retrieve the locations associated with the encoded images and were
instructed to guess if they could not remember. For foils, they were asked to make their best guess about what
its location could be, using any information they had learned during the encoding task. The 168 trials were
divided into three 56-trial blocks separated by 1-min breaks. The presentation, timing, mouse cursor, and warn-
ing sounds in this task were identical to the parameters from immediate retrieval, with the exception that images
were not presented at their correct locations after a response was made.

Instead, after each response was recorded, participants rated their confidence for the location they had chosen
with a 4-point scale. Options ranged from 4 (“very sure”) to 1 (“not sure”). If participants considered the image
to be a foil, they were instructed not to use the confidence scale and instead press “new”. The five options were
mapped to “z”, “x”, “c”, “v” and “b”, and the mapping was counterbalanced between subjects. Once a rating was
chosen, or if no rating was made within 3 s, the scale would disappear and the trial ended. Trials were separated
by a 0.5 s fixation period.

Analysis. Error. The precision of memory for the locations associated with each image was evaluated by
quantifying the magnitude of error for each trial. Error was defined as the absolute value of the angle between
an image’s encoded location and its retrieved location, with reference to the center of the ring. The error for a
given image ranged from 0 to 180 degrees, where smaller values indicate more precise memory retrieval. Trial-
specific error values were either entered into mixed-effects models, or averaged across trials by retrieval test or
confidence for across-participant correlations and visualization of mixed-effects models. Comparisons to chance
performance were conducted with two-tailed one-sample t-tests of participants’ average error against 90 degrees,
which is the average error that reflects random guessing.

Schema‑consistency. Schema-consistency was operationalized as the consistency between an image’s loca-
tion and the ‘location schema’ that could be learned for each category. This was computed as the distance
between  each image’s encoded location and the center of the distribution of locations corresponding to the
image’s category. In mixed-effects models predicting error, schema-consistency was entered as a continuous,
fixed-effects independent variable.

Generalization. Generalization of a schema was operationalized using the locations of new images. For each
trial, generalization was defined as the angular distance between each image and its corresponding category’s
center location. The generalization for a given image could range from 0 to 180 degrees, where lower values
indicate guesses that are closer to the center of an image’s category and thus better generalization. Comparisons
to chance performance were conducted with a two-tailed, one-sample t-test of participants’ average generaliza-
tion against 90 degrees, which is the average distance from a category center that reflects random guessing. The
comparison of generalization across groups was conducted with a mixed-effects model using group (1-week,
24-h) as a discrete fixed effect and generalization as a continuous dependent variable.

Divergence analyses. To calculate the Kullback–Leibler divergence (DKL) between the distributions of locations
in different tasks, the image locations of each task (encoding, immediate retrieval, and delayed retrieval) were
converted into probability density functions to be comparable with the distributions used to generate encoded
locations for each category. For each task, the locations were first divided into 36 10-degree bins of the ring. The
resulting distribution of locations per bin were normalized to sum to 1, resulting in a probability density func-
tion of the locations around the ring. Each function was smoothed by convolution with a Gaussian distribution
(σ = 22°) in order to avoid bins with no locations, which would lead to extreme DKL values. While the choice of σ
was not motivated by prior work, the specific σ was not critical as using different values produced similar results.
For any two distributions, the DKL was calculated with the following equation:

DKL was computed to compare the distribution of encoded images with the retrieval distributions separately for
the immediate and delayed retrieval tests. Greater DKL indicates a larger mismatch in the distribution of images
between two tasks.

DKL(PEncoded||PRetrieved) =



PEncoded(x) ln







Scientific Reports | (2020) 10:16968 |

Three participants in the 24-h group were statistical outliers (> 3 SD from the group mean), and one par-
ticipant in the 1-week group exhibited a relatively high divergence score at the delayed test (based on visual
inspection). To account for these outlier participants, we report the differences in divergence over time and
across groups with and without these participants included. Furthermore, these participants were not statistical
outliers in any other analysis reported in this manuscript, and excluding them does not meaningfully change
the observed results.

Mixed‑effects models. Because participants were given a limited amount of time to encode and retrieve each
trial, the number of completed trials per participant varied (See Supplementary Tables 1, 2 and 3 for response
rates by task, group, and confidence). Because of this, all analyses were conducted at the trial level with mixed-
effects linear models, except for across-participant correlational analyses and comparisons against chance per-
formance. This included analyses of the effects of schema-consistency, time and confidence on error (Fig.  4a,
b, S1A, B) and group differences in generalization (Fig. 5a). When needed, model convergence warnings were
avoided by scaling error and schema-consistency across all trials and participants to be centered around 0 with
SD = 1. Participant intercepts and slope terms for each included predictor variable were modeled as random
effects. The mixed-effects models were computed with lme464 and significance of a given contrast was obtained
using Satterthwaite’s method, resulting in t statistics and corresponding p values. Analysis of variance was con-
ducted for models containing discrete predictors with more than two levels (i.e. Supplementary Fig. 1A, Sup-
plementary Fig. 2) using the lmerTest package65. Estimated marginal means (EMMs) were computed using the
emmeans package66 to test simple effects. To facilitate visualization and interpretation of the results, all data and
model fits were un-scaled, and analyses with discrete independent variables predicting error (Figs. 4a, 5a, S1A)
were plotted as error averaged within each participant.

Across‑participant correlations with generalization. Generalization to the new images was used to index varia-
tions in participants’ schema knowledge. This was operationalized as the mean of the distances between all new
images’ guessed locations and their category center. Lower values indicate closer placement of images to their
‘category schema’. Linear regressions were used to compute across-participant relationships between generaliza-
tion, average error and schema-reliance. Pearson correlations were used to quantify the relationship between
generalization and average error or schema-consistency for specific groups and time points.

Effect sizes. Effect sizes were reported for all effects, including Cohen’s d for all t-tests and partial η2 for main
effects or interactions of ANOVAs, computed with sjstats67. Tables of the estimates and standard error of the
fixed and random effects of each reported model can be found in the supplementary materials.

Data availability
All data and analysis code that support the findings of this study are available at https :// .

Received: 20 January 2020; Accepted: 21 September 2020

1. Alba, J. W. & Hasher, L. Is memory schematic?. Psychol. Bull. 93, 203–231 (1983).
2. Stangor, C. & McMillan, D. Memory for expectancy-congruent and expectancy-incongruent information: A review of the social

and social developmental literatures. Psychol. Bull. 111, 42–61 (1992).
3. Ghosh, V. E. & Gilboa, A. What is a memory schema? A historical perspective on current neuroscience literature. Neuropsychologia

53, 104–114 (2014).
4. Nadel, L., Samsonovich, A., Ryan, L. & Moscovitch, M. Multiple trace theory of human memory: Computational, neuroimaging,

and neuropsychological results. Hippocampus 10, 352–368 (2000).
5. Winocur, G., Moscovitch, M. & Bontempi, B. Memory formation and long-term retention in humans and animals: Convergence

towards a transformation account of hippocampal–neocortical interactions. Neuropsychologia 48, 2339–2356 (2010).
6. McClelland, J. L., McNaughton, B. L. & O’Reilly, R. C. Why there are complementary learning systems in the hippocampus and

neocortex: Insights from the successes and failures of connectionist models of learning and memory. Psychol. Rev. 102, 419–457

7. Richards, B. A. et al. Patterns across multiple memories are identified over time. Nat. Neurosci. 17, 981–986 (2014).
8. Tompary, A. & Davachi, L. Consolidation promotes the emergence of representational overlap in the hippocampus and medial

prefrontal cortex. Neuron 96, 228-241.e5 (2017).
9. Ellenbogen, J. M., Hu, P. T., Payne, J. D., Titone, D. & Walker, M. P. Human relational memory requires time and sleep. Proc. Natl.

Acad. Sci. 104, 7723–7728 (2007).
10. Djonlagic, I. et al. Sleep enhances category learning. Learn. Mem. 16, 751–755 (2009).
11. Durrant, S. J., Taylor, C., Cairney, S. & Lewis, P. A. Sleep-dependent consolidation of statistical learning. Neuropsychologia 49,

1322–1331 (2011).
12. van Kesteren, M. T. R., Rijpkema, M., Ruiter, D. J. & Fernández, G. Consolidation differentially modulates schema effects on

memory for items and associations. PLoS ONE 8, e56155 (2013).
13. Durrant, S. J., Cairney, S. A., McDermott, C. & Lewis, P. A. Schema-conformant memories are preferentially consolidated during

REM sleep. Neurobiol. Learn. Mem. 122, 41–50 (2015).
14. Richter, F. R., Bays, P. M., Jeyarathnarajah, P. & Simons, J. S. Flexible updating of dynamic knowledge structures. Sci. Rep. 9, 2272

15. Posner, M. I. & Keele, S. W. On the genesis of abstract ideas. J. Exp. Psychol. 77, 353 (1968).
16. Brown, B. R. & Evans, S. H. Perceptual learning in pattern discrimination tasks with two and three schema categories. Psychon.

Sci. 15, 101–103 (1969).
17. Kumaran, D., Summerfield, J. J., Hassabis, D. & Maguire, E. A. Tracking the emergence of conceptual knowledge during human

decision making. Neuron 63, 889–901 (2009).



Scientific Reports | (2020) 10:16968 |

18. Tamminen, J., Davis, M. H., Merkx, M. & Rastle, K. The role of memory consolidation in generalisation of new linguistic informa-
tion. Cognition 125, 107–112 (2012).

19. Magnussen, S. & Dyrnes, S. High-fidelity perceptual long-term memory. Psychol. Sci. 5, 99–102 (1994).
20. Brady, T. F., Konkle, T., Gill, J., Oliva, A. & Alvarez, G. A. Visual long-term memory has the same limit on fidelity as visual working

memory. Psychol. Sci. 24, 981–990 (2013).
21. van der Linden, M., Berkers, R. M. W. J., Morris, R. G. M. & Fernández, G. Angular gyrus involvement at encoding and retrieval

is associated with durable but less specific memories. J. Neurosci. 37, 9474–9485 (2017).
22. Bonasia, K. et al. Prior knowledge modulates the neural substrates of encoding and retrieving naturalistic events at short and long

delays. Neurobiol. Learn. Mem. 153, 26–39 (2018).
23. Richards, B. A. & Frankland, P. W. The persistence and transience of memory. Neuron 94, 1071–1084 (2017).
24. Wagner, U., Gais, S., Haider, H., Verleger, R. & Born, J. Sleep inspires insight. Nature 427, 352–355 (2004).
25. Lau, H., Tucker, M. A. & Fishbein, W. Daytime napping: Effects on human direct associative and relational memory. Neurobiol.

Learn. Mem. 93, 554–560 (2010).
26. Dumay, N. & Gaskell, M. G. Sleep-associated changes in the mental representation of spoken words. Psychol. Sci. 18, 35–39 (2007).
27. Fenn, K. M., Nusbaum, H. C. & Margoliash, D. Consolidation during sleep of perceptual learning of spoken language. Nature 425,

614–616 (2003).
28. Batterink, L. J., Oudiette, D., Reber, P. J. & Paller, K. A. Sleep facilitates learning a new linguistic rule. Neuropsychologia 65, 169–179

29. Maddox, W. T. et al. The effects of sleep deprivation on information-integration categorization performance. Sleep J. Sleep Sleep

Disord. Res. 32, 1439–1448 (2009).
30. Hennies, N., Lewis, P. A., Durrant, S. J., Cousins, J. N. & Lambon Ralph, M. A. Time-but not sleep-dependent consolidation

promotes the emergence of cross-modal conceptual representations. Neuropsychologia 63, 116–123 (2014).
31. Sweegers, C. C. G. & Talamini, L. M. Generalization from episodic memories across time: A route for semantic knowledge acquisi-

tion. Cortex 59, 49–61 (2014).
32. Lutz, N. D., Diekelmann, S., Hinse-Stern, P., Born, J. & Rauss, K. Sleep supports the slow abstraction of gist from visual perceptual

memories. Sci. Rep. 7, 42950 (2017).
33. Sommer, T. The emergence of knowledge and how it supports the memory for novel related information. Cereb. Cortex. 4, 1906–

1921 (2017).
34. Ashby, F. G. & Maddox, W. T. human category learning. Annu. Rev. Psychol. 56, 149–178 (2005).
35. Antony, J., Stiver, C., Graves, K., Osborne, J. & Turk-Browne, N. Spatial Gist Extraction During Human Memory Consolidation.

PsyArXiv (2020).
36. Graves, K. N., Antony, J. & Turk-Browne, N. Finding the pattern: Online extraction of spatial structure during virtual navigation.

Psychol. Sci. 31, 1183–1190 (2020).
37. Kitamura, T. et al. Engrams and circuits crucial for systems consolidation of a memory. Science 356, 73–78 (2017).
38. Nosofsky, R. M. Attention, similarity, and the identification–categorization relationship. J. Exp. Psychol. Gen. 115, 39–57 (1986).
39. Sakamoto, Y. & Love, B. C. Schematic influences on category learning and recognition memory. J. Exp. Psychol. Gen. 133, 534–553

40. Harlow, I. M. & Donaldson, D. I. Source accuracy data reveal the thresholded nature of human episodic memory. Psychon. Bull.

Rev. 20, 318–325 (2013).
41. Harlow, I. M. & Yonelinas, A. P. Distinguishing between the success and precision of recollection. Memory 24, 114–127 (2016).
42. Nilakantan, A. S., Bridge, D. J., VanHaerents, S. & Voss, J. L. Distinguishing the precision of spatial recollection from its success:

Evidence from healthy aging and unilateral mesial temporal lobe resection. Neuropsychologia 119, 101–106 (2018).
43. Nilakantan, A. S., Bridge, D. J., Gagnon, E. P., VanHaerents, S. A. & Voss, J. L. Stimulation of the posterior cortical-hippocampal

network enhances precision of memory recollection. Curr. Biol. 27, 465–470 (2017).
44. Richter, F. R., Cooper, R. A., Bays, P. M. & Simons, J. S. Distinct neural mechanisms underlie the success, precision, and vividness

of episodic memory. eLife 5, e18260 (2016).
45. Posner, M. I., Goldsmith, R. & Welton, K. E. Jr. Perceived distance and the classification of distorted patterns. J. Exp. Psychol. 73,

28 (1967).
46. Goodman, G. S. Picture memory: How the action schema affects retention. Cognit. Psychol. 12, 473–495 (1980).
47. Bein, O. et al. Delineating the effect of semantic congruency on episodic memory: The role of integration and relatedness. PLoS

ONE 10, e0115624 (2015).
48. Bransford, J. D. & Johnson, M. K. Contextual prerequisites for understanding: Some investigations of comprehension and recall.

J. Verbal Learn. Verbal Behav. 11, 717–726 (1972).
49. Kole, J. A. & Healy, A. F. Using prior knowledge to minimize interference when learning large amounts of information. Mem.

Cognit. 35, 124–137 (2007).
50. DeWitt, M. R., Knight, J. B., Hicks, J. L. & Ball, B. H. The effects of prior knowledge on the encoding of episodic contextual details.

Psychon. Bull. Rev. 19, 251–257 (2012).
51. Guttman, N. & Kalish, H. I. Discriminability and stimulus generalization. J. Exp. Psychol. 51, 79–88 (1956).
52. Shepard, R. Toward a universal law of generalization for psychological science. Science 237, 1317–1323 (1987).
53. Dunsmoor, J. E., Martin, A. & LaBar, K. S. Role of conceptual knowledge in learning and retention of conditioned fear. Biol. Psychol.

89, 300–305 (2012).
54. van Osselaer, S. M. J., Janiszewski, C. & Cunha, M. Stimulus generalization in two associative learning processes. J. Exp. Psychol.

Learn. Mem. Cogn. 30, 626–638 (2004).
55. Schacter, D. L., Guerin, S. A. & St. Jacques, P. L. Memory distortion: An adaptive perspective. Trends Cogn. Sci. 15, 467–474 (2011).
56. Hemmer, P. & Persaud, K. Interaction between categorical knowledge and episodic memory across domains. Front. Psychol. 5, 584

57. Hemmer, P. & Steyvers, M. Integrating episodic memories and prior knowledge at multiple levels of abstraction. Psychon. Bull.

Rev. 16, 80–87 (2009).
58. Persaud, K. & Hemmer, P. The dynamics of fidelity over the time course of long-term memory. Cognit. Psychol. 88, 1–21 (2016).
59. Brady, T. F., Schacter, D. L. & Alvarez, G. A. The Adaptive Nature of False Memories is Revealed by Gist‑Based Distortion of True

Memories. PsyArXiv (2018).
60. Tompary, A. & Thompson-Schill, S. L. Semantic Influences on Episodic Memory Distortions. PsyArXiv (2020).
61. Berens, S. C., Richards, B. A. & Horner, A. J. Dissociating memory accessibility and precision in forgetting. Nat. Hum. Behav. 4,

866–877 (2020).
62. Brainard, D. H. The psychophysics toolbox. Spat. Vis. 10, 433–436 (1997).
63. R Core Team. R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2015).
64. Bates, D., Mächler, M., Bolker, B. & Walker, S. Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67, 1–48 (2015).
65. Kuznetsova, A., Brockhoff, P. B. & Christensen, R. H. B. lmerTest package: Tests in linear mixed effects models. J. Stat. Softw. 82,

1–26 (2017).
66. Lenth, R. emmeans: Estimated Marginal Means, aka Least‑Squares Means (2018).
67. Lüdecke, D. sjstats: Statistical Functions for Regression Models (Version 0.18.0). (2020).



Scientific Reports | (2020) 10:16968 |

We thank C. Gasser for help with data collection and V. Murty, E. Cowan, C. Smith, and O. Bein for helpful
discussions. The study was supported by 5R01MH074692-12 to L.D. The funders had no role in the study design,
data collection and analysis, decision to publish or preparation of the manuscript.

Author contributions
A.T., W.Z. and L.D. conceived and designed the study. W.Z. performed the research. W.Z. and A.T. analyzed the
data. All authors helped to interpret the data. A.T., W.Z. and L.D. wrote the manuscript. All authors contributed to
and approved the final manuscript. Pilot results from this study were presented in a master’s thesis written by W.Z.

Competing interests
The authors declare no competing interests.

Additional information
Supplementary information is available for this paper at https :// 8-020-73952 -x.

Correspondence and requests for materials should be addressed to A.T.

Reprints and permissions information is available at

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or

format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.

© The Author(s) 2020

  • Schematic memories develop quickly, but are not expressed unless necessary
    • Results
      • Episodic memory over time.
      • Relationship between schema consistency and episodic memory.
      • Memory for encoded distribution.
      • Schema generalization: a ‘pure’ test of schema memory.
      • Relationship between schema generalization and episodic memory.
    • Discussion
      • Schematic influences on episodic memory.
      • Schematic memories are forgotten over time.
      • Time-dependent changes in episodic and schematic memory.
      • Relationship to other literatures.
      • Limitations and avenues for future work.
    • Methods
      • Participants.
      • Stimuli.
        • Images.
        • Locations.
        • Software.
      • Procedure.
        • Encoding.
        • Immediate retrieval.
        • Delayed retrieval.
      • Analysis.
        • Error.
        • Schema-consistency.
        • Generalization.
        • Divergence analyses.
        • Mixed-effects models.
        • Across-participant correlations with generalization.
        • Effect sizes.
    • References
    • Acknowledgements

Author’s personal copy

How schema and novelty augment
memory formation
Marlieke T.R. van Kesteren1,2, Dirk J. Ruiter2, Guillén Fernández1,3* and
Richard N. Henson4*

Donders Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen, P.O. Box 9101, 6500 HB, Nijmegen,

The Netherlands

Department of Anatomy, Radboud University Nijmegen Medical Centre, P.O. Box 9101, 6500 HB, Nijmegen, The Netherlands

Department of Cognitive Neuroscience, Radboud University Nijmegen Medical Centre, P.O. Box 9101, 6500 HB, Nijmegen,

The Netherlands

MRC Cognition & Brain Sciences Unit, Cambridge, CB2 7EF, UK

Information that is congruent with existing knowledge
(a schema) is usually better remembered than less con-
gruent information. Only recently, however, has the role
of schemas in memory been studied from a systems
neuroscience perspective. Moreover, incongruent (nov-
el) information is also sometimes better remembered.
Here, we review lesion and neuroimaging findings in
animals and humans that relate to this apparent para-
doxical relationship between schema and novelty. In
addition, we sketch a framework relating key brain
regions in medial temporal lobe (MTL) and medial pre-
frontal cortex (mPFC) during encoding, consolidation
and retrieval of information as a function of its congru-
ency with existing information represented in neocor-
tex. An important aspect of this framework is the
efficiency of learning enabled by congruency-dependent
MTL–mPFC interactions.

The existence of prior knowledge to which new information
can be related generally improves memory for that infor-
mation. Although the role of such schemas in learning has
long been studied in psychology (Box 1), this role has only
recently been studied in neuroscience [1,2]. In particular,
whereas structures within the MTL, such as the hippo-
campus, have long been implicated in the learning of
declarative information [3], recent neuroscientific data
have implicated an additional, time-dependent involve-
ment of the mPFC [4,5], particularly when new informa-
tion is congruent with a schema [6–8].

A second line of research has studied how the novelty of
information can also improve its retention (Box 2). This
raises the question of when information conforming to a
schema (congruent information) is remembered better or
worse than information that does not (unrelated, or incon-
gruent, information) [1,9], a question that has important
implications for optimising learning in educational
settings [10]. Below, we review recent neuroscientific
research addressing this question before presenting a

new framework that tries to explain the complex relation-
ship between schema, novelty and memory.

Review of schema in systems neuroscience of memory
Several theories exist about how new information becomes
consolidated into memory [1,11–13]. The so-called stan-
dard systems-level theory of consolidation [14] proposes



Declarative memory: memories that can be declared, that is, have a

propositional truth value (events or facts), normally associated with conscious

recall as distinct from procedural (non-declarative) memories such as skill-

learning, which cannot be verbalised and are often expressed unconsciously.

Episodic or instance memory: declarative memory for a specific event in space

and time, which normally includes other contextual information present at that

time (e.g. internal thoughts and states). We use instance to refer to a specific

pattern of neocortical activity that is bound to an index in the MTL according to

our SLIMM framework; we use episodic more generally to refer to memories

with contextual information, which is often incidental (i.e. non-recurring, not

part of an existing schema).

mPFC (medial prefrontal cortex): medial aspect in the prefrontal cortex,

encompassing Brodmann areas (BA) 10, 11 and 32 in humans, and prelimbic,

infralimbic and anterior cingulate cortex in rodents.

MTL (medial temporal lobe): part of the brain comprising hippocampus,

perirhinal and enthorhinal cortices and parahippocampal gyrus.

Neocortex: association cortex that stores elements of a memory trace (visual,

spatial, auditory, somatosensory, emotional, etc.). Note that the mPFC is part

of the neocortex anatomically, but not considered to represent memory

elements in the present framework.

Novelty: response to information that is not expected or predicted in a given

context on the basis of prior experience. Note that we distinguish here between

two types of novelty (Box 2): unrelated information that does not strongly

match any schema, and incongruent information that is inconsistent with a

dominant schema. Within the present SLIMM framework, only the latter

improves memory, and note that this type of novelty cannot exist without a

schema (i.e. the two concepts are intimately related).

Reactivation: reinstatement of a memory trace, either by online re-encountering

of similar information or by replaying the memory trace during offline periods.

Resonance: neural state of co-activity of multiple mental representations

(possibly across multiple brain regions), most probably bound via coherent

(synchronous) activity.

Schema: network of neocortical representations that are strongly intercon-

nected and that can affect online and offline information processing.

Semantic or schematic memory: general, factual declarative memory that

captures regularities extracted from multiple encounters (instances) over time,

and divorced from accompanying episodic details. We use schematic to refer to a

(resonating) pattern of activity produced by strong connections within neocortex

(i.e. an activated schema) within our SLIMM framework; we use semantic more

generally to refer to acontextual knowledge that people possess.

Systems consolidation: time-dependent and offline process by which connec-

tions between elements of a memory trace in the neocortex are strengthened

so they are retained over the long term, independently of MTL structures such

as the hippocampus.

Corresponding author: van Kesteren, M.T.R. ([email protected])
Keywords: schema; novelty; memory formation; medial temporal lobe; medial
prefrontal cortex; prediction error.

* These authors made equal contributions.

0166-2236/$ – see front matter � 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.tins.2012.02.001 Trends in Neurosciences, April 2012, Vol. 35, No. 4 211

Author’s personal copy

that new (declarative) information is initially dependent
on MTL structures such as the hippocampus, but over time
(possibly through reactivation [15], e.g. during sleep
[16,17]) this information becomes relatively more depen-
dent on the neocortex. This proposal is based on evidence
that MTL lesions not only impair the ability to form new
memories (anterograde amnesia) but also impair the abili-
ty to retrieve memories formed within a period before the
lesion (retrograde amnesia) [18]. After consolidation, long-
term memories are believed to be represented by networks
of interconnected neocortical brain regions representing
the constituents of those memories, retrieval of which has
become independent of the MTL (although see below and
[19]). More recently, an additional role has been suggested
for the mPFC in such consolidation [20–22], consistent
with evidence of offline replay of learning-related brain
activity in mPFC (as well as MTL) [23–25] and by its
prominent anatomical location within memory-related
brain networks [26–28].

It has been suggested that the presence of a schema, in
terms of a pre-existing network of interconnected neocor-
tical representations (see Glossary), accelerates consolida-
tion [2]. For example, a lesion study in rodents showed that
memories congruent with a pre-learned spatial schema
(Figure 1a) became hippocampally independent after only
48 h (Figure 1b) [29], whereas memories that lacked a prior
schema were still hippocampally dependent. In addition,
functional imaging in humans during a period of rest
shortly after encoding revealed decreased hippocampal–
mPFC functional coupling for more versus less congruent
information (Figure 1d) [6], whereas successful retrieval of
congruent information was associated with increased func-
tional coupling between mPFC and a neocortical region
coding that information (Figure 1e) [7]. A schema thus
appears to act as a catalyst for consolidation, affecting
interactions between mPFC, MTL and other neocortical
regions, and possibly increasing the likelihood or effective-
ness of replay of congruent information [1,30].

A schema can also influence processes occurring during
initial acquisition. For example, functional imaging
showed increased activity in mPFC for more versus less
congruent information immediately after encoding in
rodents (Figure 1c) and increased MTL–mPFC coupling
in humans for less congruent information during encoding,
related to the strength of the schema (Figure 1d) [6]. These
results are consistent with a large body of evidence that
MTL–mPFC interactions, along with activity in other
brain regions [31–33], are important for successful encod-
ing and retrieval [34–39]. They are also consistent with
more general claims that the mPFC is important for mak-
ing online predictions (e.g. during perception [40,41]) en-
abled by schemas, whereas the MTL is important for
detecting the type of novelty [42–44] associated with an
incongruent schema (Box 2).

Although there is much debate about whether patients
with MTL damage can form new memories [45], and in
many situations they appear to be unable to do so (antero-
grade amnesia), they can still show a congruency benefit
[46] and there are certain situations in which they appear
to be able to learn new information [47–52]; these situa-
tions are possibly related to the existence of schemas. In
particular, recent data have suggested that such patients
can learn some information as well as controls can [53–55]
– so-called fast (cortical) mapping [54] – which may relate
to schemas (see below). Damage to the mPFC, by contrast,
has been associated with a reduced ability to filter and
integrate incoming information, resulting in confabulation
[56], a lack of a congruency benefit [46] and more errors
during retrieval [57], which may reflect an inability to
utilise schema (Box 1; although see also [58]). These
observations, along with lesion data in rodents [59], sug-
gest that memories mediated by MTL and mPFC might be
different in nature, ranging from more detailed, episodic
memories (instances) supported by MTL, to more general,
semantic (schematic) memories integrated by mPFC, as
expanded below.

Box 1. History of schema research

The term schema was introduced from the philosophical work of Kant

to developmental and cognitive psychology during the early 20th

century by Piaget and Bartlett, respectively [83,84], and refers loosely

to an abstract, structured mental representation. This concept led to a

cascade of both empirical behavioural research [85,86] and theoretical

developments in artificial intelligence and connectionist modelling

[68,77,87]; it also influenced educational theory [88].

A primary focus of behavioural research was how schemas aid the

retrieval of complex information by providing a scaffold for organis-

ing retrieval of that information. This reconstructive aspect of

memory offered a natural explanation of biases and false memories

that occur from an over-reliance on schema [83]. Importantly

however, schemas may also affect the encoding and consolidation

of memories ([68]; see the main text). For example, the superior recall

of schema-congruent information cannot always be explained by

facilitated retrieval [46] (e.g. by generation of schema-related

information at test, followed by episodic recognition of information

present at study [89]).

A primary focus of connectionist modelling was the extraction of

regularities from exposure to new information (instances) during

learning [90]. A core problem here is the stability–plasticity dilemma,

which is the degree to which a new instance should alter existing

knowledge about a class of instances (schema) without destabilizing

such knowledge. One solution to this problem (adopted in adaptive

resonance theory [69]) was a global parameter (vigilance) that

determined whether or not a new instance needs to be represented

separately as a function of its similarity to existing schemas. Another

solution was to draw on different learning rules in complementary

learning systems, in particular a fast-learning system (in MTL) that

stores unique instances, which can then be replayed in an interleaved

fashion to a slower-learning system (in neocortex) that extracts their

commonalities [30,77].

Enthusiasm for schema research waned since the 1980s, partly

because of the overextended definition of schema that arose from the

explosion of interest and partly because of some apparently contra-

dictory behavioural results, where novel information (that does not

conform to a schema) can sometimes be remembered well (Box 2).

Nonetheless, there has been a recent revival of interest in schema

within the neuroscience community [1,6–8,29]. Here, the concept of a

schema is simpler than in previous psychological research, operatio-

nalised, for example, as a familiar spatial layout such as the relationship

between a number of locations within an arena in which a rat expects to

find food [29], or as whether a word that must be associated with a

novel visual stimulus is congruent with a simultaneously presented

tactile stimulus [7]. Our present (neuroscientific) concept of a schema

therefore refers simply to a network of neocortical representations that

are strongly interconnected, activation of which affects processing of

new information, as expanded in the main text.

Opinion Trends in Neurosciences April 2012, Vol. 35, No. 4


Author’s personal copy

Memory for incongruent (novel) information can also be
enhanced (Box 2). This novelty advantage has been asso-
ciated with greater MTL activity during encoding [60–64].
Moreover, the precise type of novelty is likely to be impor-
tant [42], for example whether information is novel be-
cause it is incongruent with an existing schema or because
it is unrelated to any existing schemas. Here we focus on
enhanced memory owing to the former type of novelty (or
prediction error; Box 2), although the latter type of novelty
(such as a completely new environment for a rodent [65])
might also improve memory through other means, such as
arousal, reward and dopamine release [66,67]. Although
the role of novelty has been acknowledged by some schema
theorists (e.g. in terms of schemas being used to direct
attention to novel aspects of an experience [68]), there is no
clear consensus, at least within neuroscientific theories,
about the precise conditions under which memory is super-
ior for congruent or incongruent information. Below, we

outline a framework termed SLIMM (schema-linked inter-
actions between medial prefrontal and medial temporal
regions) that draws on two complementary modes for
learning new information, determined by MTL–mPFC
interactions, to reconcile the facilitatory effects of schema
and novelty on memory.

A new framework relating schema and novelty to
SLIMM extends standard consolidation theory, in terms of a
time-dependent shift from MTL to neocortical representa-
tions, by adding a third component – the mPFC – that acts to
accelerate direct neocortical learning independent of the
MTL. Within SLIMM, the main function of the mPFC is
to detect the congruency of new information with existing
information in neocortex, which we term resonance (akin to
adaptive resonance theory, ART [69]) in the sense that
congruent information resonates with existing information.

Box 2. Novelty and prediction error

Novelty has long been suspected as an important factor in learning

[91]. For example, people are often better able to remember an item

that deviated from its prevailing context [92]. Conversely, there would

seem little (e.g. metabolic) sense in the brain encoding information

that is already fully predicted. For example, there is no need to encode

the presence of your BathToy each time you enter your Bathroom

(Figure 2a), assuming you always find it there. This is consistent with

so-called predictive coding models of memory [81,93], in which the

key factor that drives learning is the amount of prediction error (PE).

Clearly, schemas still play an essential role, in that predictions are

based on such knowledge. This perspective seems to entail greater

learning for incongruent than for congruent information, however, in

opposition to schema theories (Box 1). However, the precise

predictions depend on the nature of the learned information and

how it is subsequently retrieved, as expanded below.

From a Bayesian perspective, PE can be viewed as the divergence

between prior and likelihood probability distributions. Thus, a familiar

location would establish prior probabilities over the objects one

expects to encounter there, whereas the (noisy) sensory input would

provide the likelihood that certain objects are in fact present (Figure I).

If one encounters a novel object in a novel location, such that both the

likelihood and prior distributions are imprecise (flat), PE will be low

(Figure Ia), at least relative to a familiar object in a novel location

(Figure Ib). Thus, maximal overall novelty does not necessarily entail

maximal learning; indeed, novel stimuli are often less well associated

with unpredictive contexts than are familiar stimuli [82].

Alternatively, when a familiar object (e.g. Cake) occurs unexpect-

edly in a familiar context (e.g. Bathroom) PE will be high (Figure Ic).

This situation corresponds to a maximal match–mismatch [42],

where an initial match (recognition of BathRoom) does not match

other information (Cake). High PE results in substantial learning, that

is, updating of the prior distribution to more closely match the

posterior distribution. This can improve subsequent episodic

recognition of the object by virtue of reactivating a distinctive

context (Bathroom) when that object is repeated [81,82]. However,

memory will not always be improved: if cued with the location

instead, the overlap between the new predictions (updated prior)

and the object representation still may not be sufficient for Cake to be

recalled. This contrasts with finding a PlasticDuck in your BathRoom

(Figure Id), for which PE will be low (assuming PlasticDucks and

BathToys have similar representations), but the updated prior for

your Bathroom will overlap with the PlasticDuck representation,

allowing it to be recalled. Thus, although incongruent information

may produce the greatest PE, the accuracy of subsequent retrieval of

that information will depend on how it is cued. This may be one

reason why an additional system (e.g. in the MTL) is needed to store

incongruent instances, in case they recur and become important for

extraction of new schemas (see the main text).










Novel context, novel object



Key: PE=1.1










Novel context, familiar object











Unexpected, familiar object











Expected, familiar object(a) (b) (c) (d)

TRENDS in Neurosciences

Figure I. Bayesian perspective on prediction-error-driven learningBayesian perspective on prediction-error-driven learningBayesian perspective on prediction-error-

driven learning[81]. The curves represent probability distributions, for example over a dimension of objects (ordered by similarity). The red line represents the likelihood

of an object being present, given (bottom-up, noisy) sensory evidence; the solid blue line represents the prior distribution, given (top-down) predictions from the current

context (e.g. location in the environment); the dotted blue line represents the posterior probability of objects being present (and resembles the updated priors that

would result from the learning experience). PE refers to the prediction error – the divergence between prior and likelihood distributions – whereas RP refers to recall

prospect – proportional to the posterior probability (from updated priors) of retrieving the object when cueing with the previous context (both PE and RP have arbitrary

units). (a) A novel object in a novel context, with flat (imprecise) prior and likelihood distributions (akin to the new sequence condition of [42], source memory for

unfamiliar proverbs of [82], and the unrelated case in the main text). Although maximally novel overall, PE is relatively low and little can be learned. (b) A familiar object

in a novel context (akin to the familiar proverbs of [82]), where PE is increased relative to (a). (c) A familiar object that is not expected in a familiar context, giving highest

PE (akin to the changed condition of [42] and incongruent case in the main text). Because of the residual divergence between posterior and likelihood distributions,

however, RP is lower than in (d), which corresponds to a familiar object that is expected in a familiar context (akin to the old sequence condition of [42], and the

congruent condition in the main text). This has low PE, but high RP, given high overlap between posterior and likelihood distributions.

Opinion Trends in Neurosciences April 2012, Vol. 35, No. 4


Author’s personal copy

Greater resonance leads to greater mPFC activity, which in
turn is assumed to potentiate direct connections between
neocortical representations (e.g. through phase synchroni-
sation [70]). Note that these are the same connections
assumed to be more gradually strengthened in the absence
of such mPFC input as in standard consolidation theory; the

mPFC thus accelerates neocortical learning [54]. Important-
ly, mPFC is assumed to have a reciprocal relationship
with MTL, such that mPFC activity inhibits MTL activity
[71,72]. This relates to the assumption that MTL automati-
cally captures new experiences [73] except, according to
SLIMM, when inhibition from mPFC means that the new


HPC Key:











70 IEG counts in PrL


























††† †††


F1-L1 F2-L2


F4-L4F5-L5 F6-L6



l c















schema group




(b) (c)

(d) (e)
schema group

Post-encoding rest


TRENDS in Neurosciences

Figure 1. Overview of (a–c) rodent and (d–e) human data on schema and memory. (a) Rodent studies have used an event arena in which rodents initially learn a number of

flavour–location associations [8,29]. A photo of the area and a schematic of the six locations (L1–6) of the wells and their association with six different flavours (F1–6) are

shown. The different landmarks used to navigate in the arena are also shown. (b) After learning such a schema, rats showed rapid hippocampal independence (after 48 h,

but not after 3 h) of new flavour–location associations within the same arena [29]. The graph shows data for hippocampally lesioned (HPC) versus control animals,

represented as percentage dig time in the correct well. A separate group of rodents who had not learned the initial schema did not show such rapid consolidation of the new

associations (data not shown). (c) In a later study [8], the expression of two immediate early genes (IEGs), zinc finger protein 225 (Zif268, left panel) and activity-regulated

cytoskeletal protein (Arc, right panel), was higher immediately after encoding of the new associations (NPA, new paired associates), relative to retrieval of the original

associations (OPA, old paired associates), learning of associations in a completely new area (NM, new map) and the performance of caged control (CC) animals, in prelimbic

(PrL) structures (equivalent to human mPFC) and in the hippocampus (not shown). (d) In humans, mPFC–hippocampal connectivity was greater, both while participants

watched a movie (i.e. during the encoding period) and during a resting period shortly afterwards (i.e. post-encoding rest period), the less congruent that movie was with the

first part of the movie watched the previous day (i.e. the inconsistent schema group) [6]. (e) In a later study [7], mPFC activity and the connectivity between mPFC and a

neocortical (somatosensory) region representing the schema-related information were higher during retrieval of information congruent with a schema than of information

incongruent with a schema. For an explanation of these different effects of congruency on regional activity and inter-regional connectivity, see Figure 2. Reproduced, with

permission, from [29] (a,b), [8] (c), [6] (d) and [7] (e).

Opinion Trends in Neurosciences April 2012, Vol. 35, No. 4


Author’s personal copy

information can be related via a schema. Only when there is
low resonance (or high prediction error; Box 2), as occurs for
incongruent information, will the MTL bind those elements
into an instance (e.g. via a unique index in hippocampus
[74], given its pattern separation capability).


As an example, imagine that you encounter a model duck
(PlasticDuck) in your bathroom (Figure 2a) that resembles
your favourite bath rubber duck (BathToy) but that has not
been encountered in your bathroom before. According to
SLIMM, your memory for this new (congruent) pairing of
PlasticDuck and BathRoom is likely to be good because you
already possess an association between BathToy and Plas-
ticDuck and between BathToy and BathRoom (the schema)
to which the new PlasticDuck can be related. The simulta-
neous perception of PlasticDuck and BathRoom activates
their corresponding neocortical representations, and this
activity spreads to other strongly connected neocortical
representations, such as BathToy, owing to previously
learned associations. These strong connections mean that
the BathToy, BathRoom and PlasticDuck representations
resonate (e.g. via synchronous oscillations [75]). This reso-
nance is detected by the mPFC [8,76], which then potenti-
ates the strengthening of neocortical connections between
the resonating representations, leading specifically to fast
learning of a new, direct connection between BathRoom
and PlasticDuck (i.e. good learning). The high activity in
mPFC also inhibits activity in MTL [6] such that no indi-
rect association is made between PlasticDuck and Bath-
Room via a new MTL instance.

Conversely, the same PlasticDuck encountered in a
Bakery (Figure 2c) will produce a strong novelty effect
(prediction error; Box 2) because such objects are not

normally expected there. In this (incongruent) case,
SLIMM predicts that you are also likely to remember
the pairing of PlasticDuck and Bakery, but for a different
reason. The lack of any strong pre-existing connections,
direct or indirect, between PlasticDuck and Bakery repre-
sentations leads to little resonance in the neocortical net-
work. Thus, mPFC is not activated, MTL is not inhibited
and the MTL serves to bind the active representations of
PlasticDuck and Bakery via a new instance. This leads to
good (episodic) encoding [8] that is sensitive to MTL dis-
ruption [46,50,54].

Finally, if you encounter PlasticDuck in a ToyShop
(Figure 2b), a location assumed to be only loosely related
to BathToy (less congruent), neither a specific schema nor a
prediction error is likely to be evoked, and memory for that
encounter is predicted to be poor. This is because there is
weak resonance, requiring increased MTL–mPFC interac-
tions for resolution, as both try to encode the memory [6].
Consequently, neither is strongly activated and there is
neither good schematic (mPFC-mediated) nor good in-
stance (MTL-mediated) encoding, leading to poor memory.

Retrieval before consolidation

Imagine walking back into the BathRoom shortly after
encoding the congruent case. Activation of the BathRoom
representation will lead to processes similar to those at
encoding: reactivation of BathToy (schema) and hence
PlasticDuck representations, resonance, mPFC activation
and further strengthening of the direct neocortical connec-
tion between PlasticDuck and BathRoom. Note that con-
current activation of other elements of the schema (e.g.
BathToy) can explain the bias towards remembering sche-
matic aspects of PlasticDuck. Similar processes are as-
sumed to happen during replay, when the BathRoom




(a) (b) (c)




Less congruent Incongruent





Stimulus /

Not activePlasticDuck BathRoom ToyShop Bakery




TRENDS in Neurosciences

Figure 2. Schematic depiction of the SLIMM model during encoding. Interactions between the mPFC, MTL and neocortex (indicated by the grey-green plane) during

encoding of associations between a familiar object (PlasticDuck) and a familiar environment, which is either (a) a BathRoom, providing a congruent schema by virtue of a

similar BathToy kept there, (b) a ToyShop, for which the schema is less congruent, in that a BathToy is only loosely related, or (c) a Bakery, in which a BathToy is not part of

the schema. In the congruent case, the neocortical representations of PlasticDuck and BathRoom are activated by their perception, and BathToy is activated by its existing

associations to both. The mPFC is activated by the resonance (synchronous co-activity) of these representations, and therefore potentiates neocortical connections between

all of them, resulting in a new direct connection between PlasticDuck and BathRoom representations (whereas other connections in the schema may have already reached

maximum strength). The mPFC additionally inhibits the MTL (indicated by the inhibitory connection). In the incongruent case, the lack of resonance between activated

neocortical representations means that mPFC is not activated, MTL is not inhibited and the new association between PlasticDuck and Bakery is stored instead via a separate

instance in MTL. In the less congruent case, in which there is only a weak connection between BathToy and ToyShop representations (indicated by a weaker arrow), there is

only partial activation of the BathToy representation, and hence partial resonance, and greater inhibition from mPFC to MTL is required to resolve this intermediate state.

Hence, memory encoding is less effective.

Opinion Trends in Neurosciences April 2012, Vol. 35, No. 4


Author’s personal copy

representation is reactivated by internal processes rather
than by sensory input [30].

In the incongruent case, walking into the Bakery leads
to retrieval of an instance from the MTL, which entails
reactivation of not only the PlasticDuck representation but
also other incidental (episodic) representations that were
present at encoding (see below and Figure 3d). Walking
into the ToyShop, by contrast, only leads to weak reactiva-
tion of PlasticDuck, given only weak neocortical connec-
tions and the low likelihood of the MTL having encoded an
instance. Note, however, that if PlasticDuck and ToyShop
are repeatedly experienced together, gradual strengthen-
ing of neocortical–neocortical connections can eventually
lead to effective storage in long-term memory (see the next
section), as in standard consolidation theory.

Retrieval after consolidation

After a longer delay, the outcome depends on whether
consolidation has occurred, that is, whether there has been
repeated reactivation of the crucial representations, either
by re-exposure to both or by offline replay. Such reactivation
is particularly likely for the congruent case, resulting in the

connection between the PlasticDuck and BathRoom repre-
sentations reaching an asymptote (Figure 3a). In this case,
cueing by Bathroom still activates the mPFC through reso-
nance [4,7,8], but because no further neocortical strength-
ening is needed, the mPFC is not necessary for retrieval.

For the incongruent case, there are two possibilities. If
PlasticDuck and Bakery have been repeatedly reactivated
(Figure 3c), multiple instances will be encoded by the MTL
(in addition to gradual cortical learning). The greater
number of such MTL instances also increases the likeli-
hood of offline reactivation of these representations, allow-
ing commonalities across instances to effectively be
extracted by gradual learning [30,77]. Thus, eventually
the PlasticDuck is no longer incongruent but has become
part of the Bakery schema. However, even if PlasticDuck
and Bakery have not been reactivated, such that no direct
neocortical–neocortical connection exists (Figure 3b), re-
trieval of PlasticDuck can still occur after a long delay via
retrieval of the MTL instance. Although not predicted by
standard consolidation theory, this possibility is consistent
with other theories and evidence that some remote episodic
memories are MTL-dependent [19,28].



l (





without reactivation


with reactivation







(d) (e)

(a) (b) (c)



Resonance Inhibition New





Stimulus /

Not active

Unrelated object
with schema

Unrelated object
without schema



TRENDS in Neurosciences

PlasticDuck BathRoom BakeryBathToy

Figure 3. Schematic depiction of the SLIMM model during (a–c) memory retrieval and (d,e) selective encoding. (a–c) mPFC, MTL and neocortical interactions during

retrieval of associated objects after consolidation when cued by perception of the familiar BathRoom or Bakery representations from Figure 2. (a) In the congruent case,

PlasticDuck is likely to be recalled owing to high activity of its representation following activation spread from the BathRoom (and indirectly from the BathToy)

representation. (b) In one incongruent case, recall of PlasticDuck can occur through retrieval of the MTL instance (episodic recall), although this may be rare (see the text). (c)

Alternatively, if there have been repeated reactivations of the PlasticDuck and Bakery representations during the delay, for example by their repeated co-occurrence in the

environment, recall of PlasticDuck can occur owing to a direct connection from the Bakery representation (i.e. PlasticDuck has now become part of the Bakery schema). (d)

Associations with incidental, unrelated events (e.g. mobile phone ringing) are suppressed when not consistent with the dominant (e.g. BathRoom) schema in the congruent

case, and are hence not well encoded. (e) By contrast, in the incongruent case, when all activated representations are bound into the same instance, such associations are

encoded by the MTL.

Opinion Trends in Neurosciences April 2012, Vol. 35, No. 4


Author’s personal copy

Selective encoding and interference

If memories can be retrieved via indices within the MTL
system, as in Figure 3c, why is an additional mPFC system
needed? Our proposal is that an additional learning system
is necessary to overcome the high levels of interference
resulting from multiple MTL instances sharing common
elements. The function of the mPFC is then to select the
most relevant elements of an experience (those congruent
with existing schema) during both encoding and retrieval.
Thus, the mPFC not only detects resonance but also
amplifies activity in congruent representations by addi-
tionally suppressing activity in representations inconsis-
tent with the dominant schema (possibly through an
attractor-type mechanism [78]). Imagine, for example, that
on encountering PlasticDuck in your BathRoom, your mo-
bile phone rings (Figure 3d). Because telephone calls are
not particularly related to the BathRoom schema, any
connections between the MobilePhone representation
and the other active representations are de-potentiated
(Figure 3d). In this way, only information that is related to
the dominant (active) schema is effectively selected for
direct neocortical learning. This automatic highlighting
of schema-relevant information is likely to maximise the
efficiency of learning of new information [79]. By contrast,
when experiencing the PlasticDuck and MobilePhone call
in the Bakery, where there is no dominant schema activat-
ed (Figure 3e), all of these elements are bound into a single
instance by the MTL (i.e. incidental, episodic details, such
as the phone call, are better remembered in the incongru-
ent case). This mPFC amplification is also important for
reducing interference during retrieval by focusing on
representations congruent with existing knowledge. This
might explain why patients with mPFC lesions often con-
fabulate, retrieving semantic or episodic information not
directly relevant to the retrieval cue [56].

Predictions of SLIMM
SLIMM provides several predictions for future experi-
ments in both healthy subjects and subjects with MTL
or mPFC damage. Foremost, it predicts that memory
performance in healthy subjects can be a non-linear func-
tion of congruency, with better (schematic) memory for
congruent items mediated by mPFC and better (instance)
memory for incongruent items mediated by MTL. Howev-
er, because the nature of the memories underlying perfor-
mance at either end of this congruency dimension differs,
the precise shape of this function will depend on the nature
of the retrieval test. Free recall or cued recall, for example,
may show only an advantage for congruent items (particu-
larly if a generate-and-recognise strategy is used; Box 1),
whereas tests of incidental episodic detail (unrelated to a
schema), such as recognition or source memory tests, may
show an advantage for incongruent items.

For future neuroimaging experiments, the framework
predicts that MTL and mPFC will show differential activi-
ty patterns and functional coupling (both between each
other and with neocortical regions representing compo-
nents of the memory) as a function of congruency during
encoding, offline replay and retrieval. During encoding and
replay, mPFC activity is predicted to increase with con-
gruency, MTL activity is predicted to decrease with

congruency, and mPFC–MTL coupling is predicted to be
maximal for partially congruent conditions when mPFC
and MTL are both partially activated (Figure 2b). After
consolidation, initially incongruent information will en-
gage mPFC (because it has effectively become incorporated
in the schema), whereas successful retrieval of unconsoli-
dated incongruent information will still engage the MTL
(Figure 3).

Damage to either the mPFC or the MTL is expected to
disrupt the balance between the two types of learning
described above. Selective MTL damage is predicted to
disrupt episodic encoding and produce complete retrograde
amnesia for instances [19], along with temporally graded
retrograde amnesia sparing those memories that have al-
ready been consolidated [18]. However, the still-functioning
mPFC will continue to encode information congruent with
prior knowledge (producing congruency effects [46]) via
strengthening of neocortical connections between novel in-
formation and existing schemas. Conversely, mPFC damage
will disrupt schematic encoding of information and hence
lead to absence of a congruency effect, because all memories
will be stored as instances by the MTL. This will result in
difficulties in integrating new information into a schema
and increased interference during retrieval of information
(confabulation). For information acquired shortly before the
mPFC lesion (recent memories), there will still be a congru-
ency effect, because congruent information has been consol-
idated into neocortical networks in an accelerated manner
relative to incongruent information. However, there may be
a brief period of retrograde amnesia for highly congruent
information acquired very shortly before the mPFC lesion,
when no instances were likely to be encoded and consolida-
tion has not yet occurred. For more remote memories al-
ready consolidated in neocortex, the mPFC lesion should
have no effect (unlike, e.g., damage to anterior, lateral
temporal lobes [80]), nor should mPFC lesions affect long-
term instances still indexed by the intact MTL. In sum, MTL
and mPFC lesions will produce specific problems encoding
new-instance and schematic memories respectively, and
differential retrograde amnesia gradients for recent and
remote memories as a function of congruency.

Conclusions and future directions
Our aim has been to integrate research and theories on
schema, novelty and the contributions of the MTL and
mPFC to memory formation within a single framework.
SLIMM is broadly consistent with a number of other
consolidation theories [11–13,19], but makes the role of
schema, mPFC and mPFC–MTL interactions more explic-
it. We accept that the framework is simplistic (e.g. when
assuming mechanisms that are not yet fully empirically
tested, such as resonance detection by mPFC), and faces
problems with some existing data (Box 3). Nonetheless, at
a minimum, SLIMM should help the understanding and
interrelation of previous, sometimes paradoxical, findings
in the neuroscientific and psychological literature. We hope
it will also prompt future behavioural, neuroimaging and
lesion studies that test the predictions outlined above. We
believe that these developments will be of fundamental
importance for optimising life-long learning and education,
and for treating learning and memory disorders.

Opinion Trends in Neurosciences April 2012, Vol. 35, No. 4


Author’s personal copy

We wish to thank Bernhard Staresina, Pierre Gagnepain, Maria Wimber,
Andrea Greve, Atsuko Takashima, Marijn Kroes, and three anonymous
reviewers for helpful comments on the manuscript. We also wish to thank
Vincent Schoots and Simon Strangeways for graphical assistance.
M.T.R.v.K. is funded by Radboud University Medical Centre, and this
work was made possible by the UK Experimental Psychology Society.
R.N.H. is funded by the UK Medical Research Council

1 Wang, S.H. and Morris, R.G. (2010) Hippocampal-neocortical

interactions in memory formation, consolidation, and
reconsolidation. Annu. Rev. Psychol. 61, 49–79

2 Morris, R.G. (2006) Elements of a neurobiological theory of
hippocampal function: the role of synaptic plasticity, synaptic
tagging and schemas. Eur. J. Neurosci. 23, 2829–2846

3 Squire, L.R. (1992) Memory and the hippocampus: a synthesis from
findings with rats, monkeys, and humans. Psychol. Rev. 99, 195–231

4 Takashima, A. et al. (2006) Declarative memory consolidation in
humans: a prospective functional magnetic resonance imaging
study. Proc. Natl. Acad. Sci. U.S.A. 103, 756–761

5 Frankland, P.W. and Bontempi, B. (2006) Fast track to the medial
prefrontal cortex. Proc. Natl. Acad. Sci. U.S.A. 103, 509–510

6 van Kesteren, M.T. et al. (2010) Persistent schema-dependent
hippocampal–neocortical connectivity during memory encoding and
postencoding rest in humans. Proc. Natl. Acad. Sci. U.S.A. 107,

7 van Kesteren, M.T. et al. (2010) Retrieval of associative information
congruent with prior knowledge is related to increased medial
prefrontal activity and connectivity. J. Neurosci. 30, 15888–15894

8 Tse, D. et al. (2011) Schema-dependent gene activation and memory
encoding in neocortex. Science 333, 891–895

9 van Kesteren, M.T. et al. (2012) Schema-dependent neocortical
connectivity during information processing. In Encyclopedia of the
Sciences of Learning (Seel, N., ed.), pp. 2953–2956, Springer

10 Ruiter, D.J. et al. (2010) How to achieve synergy between medical
education and cognitive neuroscience? An exercise on prior knowledge
in understanding. Adv. Health Sci. Educ. Theory Pract. DOI: 10.1007/

11 Squire, L.R. and Bayley, P.J. (2007) The neuroscience of remote
memory. Curr. Opin. Neurobiol. 17, 185–196

12 McKenzie, S. and Eichenbaum, H. (2011) Consolidation and
reconsolidation: two lives of memories? Neuron 71, 224–233

13 Winocur, G. et al. (2010) Memory formation and long-term retention
in humans and animals: convergence towards a transformation
account of hippocampal-neocortical interactions. Neuropsychologia
48, 2339–2356

14 Marr, D. (1970) A theory for cerebral neocortex. Proc. R. Soc. Lond. B:
Biol. Sci. 176, 161–234

15 Rasch, B. and Born, J. (2007) Maintaining memories by reactivation.
Curr. Opin. Neurobiol. 17, 698–703

16 Skaggs, W.E. and McNaughton, B.L. (1996) Replay of neuronal firing
sequences in rat hippocampus during sleep following spatial
experience. Science 271, 1870–1873

17 Wilson, M.A. and McNaughton, B.L. (1994) Reactivation of
hippocampal ensemble memories during sleep. Science 265, 676–679

18 Scoville, W.B. and Milner, B. (1957) Loss of recent memory after
bilateral hippocampal lesions. J. Neurol. Neurosurg. Psychiatry 20,

19 Moscovitch, M. et al. (2006) The cognitive neuroscience of remote
episodic, semantic and spatial memory. Curr. Opin. Neurobiol. 16,

20 Frankland, P.W. and Bontempi, B. (2005) The organization of recent
and remote memories. Nat. Rev. Neurosci. 6, 119–130

21 Maviel, T. et al. (2004) Sites of neocortical reorganization critical for
remote spatial memory. Science 305, 96–99

22 Bontempi, B. et al. (1999) Time-dependent reorganization of brain
circuitry underlying long-term memory storage. Nature 400, 671–675

23 Wierzynski, C.M. et al. (2009) State-dependent spike–timing
relationships between hippocampal and prefrontal circuits during
sleep. Neuron 61, 587–596

24 Takehara-Nishiuchi, K. and McNaughton, B.L. (2008) Spontaneous
changes of neocortical code for associative memory during
consolidation. Science 322, 960–963

25 Euston, D.R. et al. (2007) Fast-forward playback of recent memory
sequences in prefrontal cortex during sleep. Science 318, 1147–1150

26 Ongur, D. and Price, J.L. (2000) The organization of networks within
the orbital and medial prefrontal cortex of rats, monkeys and humans.
Cereb. Cortex 10, 206–219

27 Pandya, D.N. et al. (1981) Efferent connections of the cingulate gyrus in
the rhesus monkey. Exp. Brain Res. 42, 319–330

28 Maguire, E.A. (2001) Neuroimaging studies of autobiographical event
memory. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 356, 1441–1451

29 Tse, D. et al. (2007) Schemas and memory consolidation. Science 316,

30 Lewis, P.A. and Durrant, S.J. (2011) Overlapping memory replay
during sleep builds cognitive schemata. Trends Cogn. Sci. 15, 343–351

31 Staresina, B.P. et al. (2009) Event congruency enhances episodic
memory encoding through semantic elaboration and relational
binding. Cereb. Cortex 19, 1198–1207

32 Atienza, M. et al. (2011) Semantic congruence enhances memory of
episodic associations: role of theta oscillations. J. Cogn. Neurosci. 23,

33 Crespo-Garcia, M. et al. (2010) Functional neural networks underlying
semantic encoding of associative memories. Neuroimage 50, 1258–1270

34 Siapas, A.G. et al. (2005) Prefrontal phase locking to hippocampal theta
oscillations. Neuron 46, 141–151

35 Benchenane, K. et al. (2010) Coherent theta oscillations and
reorganization of spike timing in the hippocampal–prefrontal
network upon learning. Neuron 66, 921–936

36 Peyrache, A. et al. (2009) Replay of rule-learning related neural
patterns in the prefrontal cortex during sleep. Nat. Neurosci. 12,

37 Paz, R. et al. (2008) Theta synchronizes the activity of medial prefrontal
neurons during learning. Learn. Mem. 15, 524–531

Box 3. Outstanding questions

� How precisely does the mPFC detect resonance, amplify con-
gruent representations and suppress less congruent representa-

tions? How does it then potentiate synaptic changes between

resonating neocortical representations (e.g. in terms of synaptic

tagging or neurotransmitters)?

� How precisely do the mPFC and the hippocampus interact: does
the mPFC only inhibit the MTL, or is there mutual competition? Do

they interact differently during online experience and during

offline replay (when instances may be retrieved from MTL)?

� Why do temporary lesions of the hippocampus impair encoding
of schema-congruent information [94]? Does this happen only

when the schemas are spatial, given evidence that hippocampus

also represents spatial information in the rodent?

� How does the role of MTL in SLIMM relate to other theories of
MTL function, such as scene construction [95] and future

simulation [96]? Is reconsolidation related to integration of new

information into schemas, and will blocking of reconsolidation

thus also affect schema-related memories [12]?

� Do different subparts of the mPFC have different functions? And
how are the mPFC subparts in the rodent related to those in

human mPFC?

� What are the precise memory deficits following mPFC lesions, for
example in terms of interference, transient retrograde amnesia

(for congruent information) and possibly even encoding of greater

episodic detail than in controls?

� How do the effects of schemas vary across development? For
example, given the relatively slow maturation of PFC relative to

other brain regions [97], does the ability to use schema change

from childhood to early adulthood? And does healthy ageing

reduce the mPFC efficiency in learning new schema-congruent


� How are more complex or structured schemas represented in the
brain, and how can their formation be manipulated (during

encoding, consolidation and retrieval) to optimise learning and


Opinion Trends in Neurosciences April 2012, Vol. 35, No. 4


Author’s personal copy

38 Ranganath, C. et al. (2005) Functional connectivity with the
hippocampus during successful memory formation. Hippocampus 15,

39 Kumaran, D. et al. (2009) Tracking the emergence of conceptual
knowledge during human decision making. Neuron 63, 889–901

40 Bar, M. (2009) The proactive brain: memory for predictions. Philos.
Trans. R. Soc. Lond. B: Biol. Sci. 364, 1235–1243

41 Summerfield, C. and Koechlin, E. (2008) A neural representation of
prior information during perceptual inference. Neuron 59, 336–347

42 Kumaran, D. and Maguire, E.A. (2009) Novelty signals: a window into
hippocampal information processing. Trends Cogn. Sci. 63, 889–901

43 Knight, R. (1996) Contribution of human hippocampal region to
novelty detection. Nature 383, 256–259

44 Strange, B.A. et al. (2000) Brain mechanisms for detecting perceptual,
semantic, and emotional deviance. Neuroimage 12, 425–433

45 Gabrieli, J.D. (1998) Cognitive neuroscience of human memory. Annu.
Rev. Psychol. 49, 87–115

46 Kan, I.P. et al. (2009) Contribution of prior semantic knowledge to new
episodic learning in amnesia. J. Cogn. Neurosci. 21, 938–944

47 Rosenbaum, R.S. et al. (2005) The case of K.C.: contributions of a
memory-impaired person to memory theory. Neuropsychologia 43,

48 Vargha-Khadem, F. et al. (1997) Differential effects of early
hippocampal pathology on episodic and semantic memory. Science
277, 376–380

49 O’Kane, G. et al. (2004) Evidence for semantic learning in profound
amnesia: an investigation with patient H.M. Hippocampus 14,

50 Tramoni, E. et al. (2011) Long-term consolidation of declarative
memory: insight from temporal lobe epilepsy. Brain 134, 816–831

51 Stark, C. et al. (2005) New semantic learning and generalization in a
patient with amnesia. Neuropsychology 19, 139–151

52 Bayley, P.J. et al. (2008) New semantic learning in patients with large
medial temporal lobe lesions. Hippocampus 18, 575–583

53 Duff, M.C. et al. (2006) Development of shared information
in communication despite hippocampal amnesia. Nat. Neurosci. 9,

54 Sharon, T. et al. (2011) Rapid neocortical acquisition of long-term
arbitrary associations independent of the hippocampus. Proc. Natl.
Acad. Sci. U.S.A. 108, 1146–1151

55 Goshen-Gottstein, Y. et al. (2000) Intact implicit memory for newly
formed verbal associations in amnesic patients following single study
trials. Neuropsychology 14, 570–578

56 Schnider, A. (2003) Spontaneous confabulation and the adaptation of
thought to ongoing reality. Nat. Rev. Neurosci. 4, 662–671

57 Gilboa, A. et al. (2009) Ventromedial prefrontal cortex lesions produce
early functional alterations during remote memory retrieval. J.
Neurosci. 29, 4871–4881

58 Bird, C.M. et al. (2004) The impact of extensive medial frontal lobe
damage on ‘theory of mind’ and cognition. Brain 127, 914–928

59 Wang, S.H. et al. (2009) The precision of remote context memories does
not require the hippocampus. Nat. Neurosci. 12, 253–255

60 Kirchhoff, B.A. et al. (2000) Prefrontal–temporal circuitry for episodic
encoding and subsequent memory. J. Neurosci. 20, 6173–6180

61 Nyberg, L. (2005) Any novelty in hippocampal formation and memory?
Curr. Opin. Neurol. 18, 424–428

62 Detterman, D.K. (1975) The von Restorff effect and induced amnesia:
production by manipulation of sound intensity. J. Exp. Psychol. Hum.
Learn. 1, 614–628

63 Kumaran, D. and Maguire, E.A. (2007) Match mismatch processes
underlie human hippocampal responses to associative novelty. J.
Neurosci. 27, 8517–8524

64 Fenker, D.B. et al. (2008) Novel scenes improve recollection and recall
of words. J. Cogn. Neurosci. 20, 1250–1265

65 Wang, S.H. et al. (2010) Relevance of synaptic tagging and capture to
the persistence of long-term potentiation and everyday spatial
memory. Proc. Natl. Acad. Sci. U.S.A. 107, 19537–19542

66 Lisman, J.E. and Grace, A.A. (2005) The hippocampal–VTA loop:
controlling the entry of information into long-term memory. Neuron
46, 703–713

67 Lisman, J. et al. (2011) A neoHebbian framework for episodic memory;
role of dopamine-dependent late LTP. Trends Neurosci. 34, 536–547

68 Alba, J.W. and Hasher, L. (1983) Is memory schematic? Psychol. Bull.
93, 203–231

69 Grossberg, S. (1987) Competitive learning – from interactive activation
to adaptive resonance. Cogn. Sci. 11, 23–63

70 Fell, J. and Axmacher, N. (2011) The role of phase synchronization in
memory processes. Nat. Rev. Neurosci. 12, 105–118

71 Laroche, S. et al. (2000) Plasticity at hippocampal to prefrontal cortex
synapses: dual roles in working memory and consolidation.
Hippocampus 10, 438–446

72 Colgin, L.L. (2011) Oscillations and hippocampal-prefrontal
synchrony. Curr. Opin. Neurobiol. 21, 467–474

73 Moscovitch, M. (2008) The hippocampus as a ‘stupid’, domain-specific
module: implications for theories of recent and remote memory, and of
imagination. Can. J. Exp. Psychol. 62, 62–79

74 Murre, J.M. (1996) TraceLink: a model of amnesia and consolidation of
memory. Hippocampus 6, 675–684

75 Engel, A.K. et al. (2001) Dynamic predictions: oscillations and
synchrony in top-down processing. Nat. Rev. Neurosci. 2, 704–716

76 Maguire, E.A. et al. (1999) The functional neuroanatomy of
comprehension and memory: the importance of prior knowledge.
Brain 122, 1839–1850

77 McClelland, J.L. et al. (1995) Why there are complementary learning
systems in the hippocampus and neocortex: insights from the successes
and failures of connectionist models of learning and memory. Psychol.
Rev. 102, 419–457

78 McKenna, T.M. et al. (1994) The brain as a dynamic physical system.
Neuroscience 60, 587–605

79 Sweller, J. (1988) Cognitive load during problem-solving – effects on
learning. Cogn. Sci. 12, 257–285

80 Patterson, K. et al. (2007) Where do you know what you know? The
representation of semantic knowledge in the human brain. Nat. Rev.
Neurosci. 8, 976–987

81 Henson, R.N. and Gagnepain, P. (2010) Predictive, interactive multiple
memory systems. Hippocampus 20, 1315–1326

82 Poppenk, J. et al. (2011) Revisiting the novelty effect: when familiarity,
not novelty, enhances memory. J. Exp. Psychol. Learn. Mem. Cogn. 36,

83 Bartlett, F.C. (1932) Remembering: A Study in Experimental and
Social Psychology, Cambridge University Press

84 Piaget, J. (1926) The Child’s Conception of the World, Littlefield,

85 Bransford, J.D. et al. (2000) How People Learn: Brain, Mind,
Experience and School, National Academy Press

86 Johnson-Laird, P.N. (1983) Mental Models: Towards a Cognitive
Science of Language, Inference, and Consciousness, Harvard
University Press

87 Neisser, U. (1976) Cognitive Psychology, Appleton-Century-Crofts
88 Anderson, R.C. (1984) Role of the reader’s schema in comprehension,

learning, and memory. In Learning to Read in American Schools: Basic
Readers and Content Texts (Anderson, R.C. et al., eds), pp. 243–257,

89 Watkins, M.J. and Gardiner, J.M. (1979) Appreciation of generate–
recognize theory of recall. J. Verb. Learn. Verb. Behav. 18, 687–704

90 McClelland, J.L. and Rumelhart, D.E. (1986) Pyschological and
Biological Models, In Parallel Distributed Processing: Explorations
in the Microstructure of Cognition (Vol. 2), MIT Press

91 Tulving, E. and Kroll, N. (1995) Novelty assessment in the brain and
long-term-memory Encoding. Psychon. Bull. Rev. 2, 387–390

92 von Restorff, H. (1933) Über die Wirkung von Bereichsbildungen im
Spurenfeld. Psychol. Forsch. 18, 299–342

93 Friston, K. (2005) A theory of cortical responses. Philos. Trans. R. Soc.
Lond. B: Biol. Sci. 360, 815–836

94 Bethus, I. et al. (2010) Dopamine and memory: modulation of the
persistence of memory for novel hippocampal NMDA receptor-
dependent paired associates. J. Neurosci. 30, 1610–1618

95 Hassabis, D. and Maguire, E.A. (2007) Deconstructing episodic
memory with construction. Trends Cogn. Sci. 11, 299–306

96 Schacter, D.L. and Addis, D.R. (2009) On the nature of medial temporal
lobe contributions to the constructive simulation of future events.
Philos. Trans. R. Soc. Lond. B: Biol. Sci. 364, 1245–1253

97 Casey, B.J. et al. (2000) Structural and functional brain development
and its relation to cognitive development. Biol. Psychol. 54, 241–257

Opinion Trends in Neurosciences April 2012, Vol. 35, No. 4


We’ve proficient writers who can handle both short and long papers, be they academic or non-academic papers, on topics ranging from soup to nuts (both literally and as the saying goes, if you know what we mean). We know how much you care about your grades and academic success. That's why we ensure the highest quality for your assignment. We're ready to help you even in the most critical situation. We're the perfect solution for all your writing needs.

Get a 15% discount on your order using the following coupon code SAVE15

Order a Similar Paper Order a Different Paper