Content uploaded by Dirk Wulff
Author content
All content in this area was uploaded by Dirk Wulff on Feb 15, 2021
Content may be subject to copyright.
Understanding the aging lexicon by linking individuals’ experience,
semantic networks, and cognitive performance
Dirk U. Wulff1,2, Simon De Deyne3, Samuel Aeschbach1, and Rui Mata1,2
1University of Basel
2Max Planck Institute for Human Development
3University of Melbourne
People undergo many idiosyncratic experiences throughout their lives that may contribute to
individual differences in the size and structure of their knowledge representations. Ultimately,
these can have important implications for individuals’ cognitive performance. We review evi-
dence that suggests a relationship between individual experiences, the size and structure of se-
mantic representations, as well as individual and age differences in cognitive performance. We
conclude that the extent to which experience-dependent changes in semantic representations
contribute to individual differences in cognitive aging remains unclear. To help fill this gap,
we outline an empirical agenda involving the concurrent assessment of large-scale semantic
networks and cognitive performance in younger and older adults, and present preliminary data
to establish the feasibility and limitations of such empirical approaches.
Keywords: semantic networks, cognitive aging, individual differences
From childhood and adolescence onwards, the average hu-
man reads a couple of books each year, watches hundreds of
hours of TV, and spends many hours on social media, lead-
ing to the accumulation of a large and, potentially, largely
unique, set of experiences during a lifetime (Twenge et al.,
2019). To what extent does the accumulation of such id-
iosyncratic experiences contribute to individual differences
in thought and judgment across the life span?
Aging research has long realized the importance not only
of describing the modal changes in cognition across the life
span but also that "cognitive development in adulthood and
old age differs substantially from person to person and is mal-
leable within individuals" (Lindenberger, 2014). Despite the
field’s direct acknowledgment of inter-individual differences,
we still know little about the sources of such differences and
the extent to which idiosyncratic life experiences contribute
to cognitive performance. Crucially, some voices have raised
the possibility that cumulative experience is a major factor
Dirk U. Wulffhttps://orcid.org/0000-0002-4008-8022 Simon
De Deyne https://orcid.org/0000-0002-7899-6210 Rui Mata
https://orcid.org/0000-0002-1679-906X
We are grateful to Laura Wiles for editing the manuscript. This
work was supported by a grant from the Swiss Science Foundation
(100015_197315) to Dirk U. Wulff.
Correspondence concerning this article should be addressed
to Dirk U. Wulff, Department of Psychology, University of
Basel, Missionsstrasse 60-62, 4055 Basel, Switzerland. E-mail:
dirk.wulff@gmail.com
underlying typical age-related patterns, such as decreased
memory performance with increased age (Buchler & Reder,
2007; Ramscar et al., 2014).
This article contributes to this effort by reviewing evi-
dence on the effects of cumulative experience on cognition.
We conclude that the existing literature has not fully estab-
lished the links between experience, cognitive representa-
tion, and cognitive performance. One central limit of past
work has been a lack of direct assessment of individuals’
mental representations. As a result, we propose an empiri-
cal agenda to fill this gap and present a study that illustrates
the feasibility and limitations of our approach, which aims to
elicit large-scale semantic networks from single individuals.
Finally, we discuss the limitations and implications of the
outlined research agenda to understand individual and age
differences.
From Experience to Cognitive Performance: An
Overview of the Current Literature
In this section, we assess the current state of the psycho-
logical literature that has investigated the links between en-
vironmental exposure, the size and structure of semantic rep-
resentations, and cognitive performance across the life span
(see Figure 1). Our goal is both to provide a brief overview
of the current literature and to highlight the main gaps that
must be filled to answer the question above concerning the
role of individual experience in cognitive representations and
performance.
There are several lines of research supporting a close link
between cumulative experience and cognitive performance
2WULFF, DE DEYNE, AESCHBACH, AND MATA
Environment
Representation
Performance
Crystallized abilities
Fluid abilities
Life course
(a)
(b) (c)
Figure 1
Research pathways that have linked environment, represen-
tation, and cognitive performance across the lifespan. Path-
way (a) represents research studying the association be-
tween the environment and cognitive performance without
explicit consideration of the underlying mental representa-
tions. Pathway (b) represents research on the association
between environmental exposure and mental representations.
Pathway (c) represents research on the association between
mental representations and cognitive performance.
(see Figure 1, path a: Environment–Performance), of which
three stand out. First, research that summarizes the asso-
ciation between print exposure and reading performance in
children and young adults underlines the tremendous impact
of experience on linguistic proficiency. Specifically, meta-
analytic results show moderate to strong correlations be-
tween print exposure and reading comprehension or spelling
(Mol & Bus, 2011). This work supports the idea of an "up-
ward spiral", such that individuals who are more exposed to
written text become more proficient in reading and compre-
hension, which, in turn, leads to increased print exposure,
further increasing linguistic proficiency. Second, one of the
best-documented findings in aging research concerns the ob-
served increases in crystallized abilities across the life span,
as reflected in vocabulary size (Verhaeghen, 2003, for a meta-
analysis). This work suggests that even as adults, individ-
uals continue to expand their vocabulary store throughout
their lives (Brysbaert et al., 2016), which is compatible with
the idea that such vocabulary growth is related to individ-
uals’ lifetime formal or informal education. Notably, such
increases in knowledge seem to have general implications
for cognitive performance. For example, older adults are
more likely to rely on prediction in reading because of their
additional reading experience (Huettig & Pickering, 2019).
Third, research on expertise suggests that increased cumu-
lative experience leads to domain-specific memory perfor-
mance (Sala & Gobet, 2017, for a meta-analysis), such as, for
instance, experienced chess players being able to memorize
both realistic and random chess position better than novices
(Gobet & Simon, 1996). All in all, such findings coalesce to
make a strong case for the importance of environmental ef-
fects on cognitive performance and suggest that these effects
are cumulative and can be domain-specific.
Most, if not all, psychologists will find it trivial to state
that the effects of cumulative experience are somehow me-
diated by its effects on mental representations and processes
(Figure 1, path b, Environment–Representation). Despite the
truism and the increasing consensus that the lexical-semantic
space continues to be shaped by personal linguistic experi-
ence throughout the life span (Rodd, 2020), only recently
have researchers started to probe more deeply into the ef-
fects of cumulative experience on the structure of knowl-
edge representations. Some researchers have adopted graph-
based approaches to capture potential structural changes to
the mental lexicon that occur across the life span as a func-
tion of experience (Wulffet al., 2019). The current litera-
ture suggests that cumulative experience has implications for
not only the size but also the structure of mental representa-
tions (Dubossarsky et al., 2017). For example, Dubossarsky
and colleagues conducted a network analysis of free asso-
ciation data from thousands of individuals (10 to 84 years
of age), and found older adults’ semantic networks were
less connected (i.e., the words in the network have lower
average degrees), less organized (i.e., the words in the net-
work have a lower average local clustering coefficient), and
less efficient (i.e., the shortest path length between any two
words in the network is greater) relative to those of younger
adults. Crucially, new methods are becoming available that
promise ever greater insight and precision in mapping indi-
vidual and group differences in the structure of mental repre-
sentations (Morais et al., 2013; Zemla & Austerweil, 2018).
Table 1 presents an overview of a few approaches to measur-
ing individual-level mental representations and we discuss
in more detail pros and cons of each approach in the next
section (cf. "An Empirical Agenda").
Finally, research is accumulating that establishes direct
links between the structure of semantic representations and
cognitive performance (Figure 1, path c, Representation–
Performance). For example, various studies on memory re-
call show impaired performance when words represented as
nodes within a semantic network have lower clustering (Nel-
son et al., 1993). More recent work has expanded this line
of research to understand how lexical and semantic structure
is crucial to individual cognitive performance in various do-
mains, such as intelligence and creativity (Kenett & Faust,
2019; Li et al., 2020). For a review, see Wulffet al. (2019).
These more recent studies are particularly relevant because
they have started relying on individual or small group esti-
mates of mental representations and how macro-properties
SEMANTIC NETWORKS AND COGNITIVE AGING 3
Table 1
Approaches to measuring individual-level semantic networks
Paradigm Description Scope Comparability References
Verbal fluency Individuals generate as many items
from a given category as they can in
a fixed period of time
Limited High (Zemla & Austerweil,
2018)
Relatedness judgments Individuals rate the relation, i.e., sim-
ilarity, between pairs of items
Limited High (Benedek et al., 2017;
Roads & Love, 2020)
Free association (snow-
ball)
Individuals generate one or more as-
sociations to word cues, which are
participant generated
Broad Low (Morais et al., 2013)
Free association (fixed
list)
Individuals generate one or more as-
sociations to experimenter-generated
word cues
Broad High see below
Note.Scope =Ability of the paradigm to provide coverage of a large set of semantic categories; Comparability =Ability of the paradigm to
provide comparable coverage of semantic representation from different individuals.
of representational networks impact cognitive performance.
However, one should note that this work has not established
a direct link between an individual’s cumulative experience
and the size or structure of the mental representations.
All in all, these different lines of research support the idea
that cumulative personal experience plays a crucial role in
determining individual differences in cognitive performance
across the life span. However, much of this work is corre-
lational (Nation, 2017, for a similar critique) and not carried
out at the level of the individual. As a result, the mechanisms
that tie cumulative exposure to performance are still under
investigation. In what follows, we propose that to truly un-
derstand the consequences of individual experience on men-
tal representation and cognitive performance, the concurrent
assessment of individuals’ unique environments, as well as
their mental representations and cognitive performance, is
needed.
An Empirical Agenda
As outlined above, the current literature has made large
strides towards understanding the components linking cumu-
lative experience to cognitive performance: We know that
younger and older adults differ in the amounts and kinds of
experiences made, the contents and structure of mental rep-
resentations, and that there are systematic age and individ-
ual differences in both fluid and crystallized performance.
However, presently, we cannot confidently estimate the por-
tion of individual differences in performance across the life
span due to the accumulation of specific types of experi-
ence. What is missing, in our mind, is a concerted empiri-
cal agenda that investigates the Environment–Representation
and the Representation–Performance pathways at the level of
the individual, thus allowing us to quantify how a certain type
or quantity of experience translates into the size or structure
of cognitive representations, and, ultimately, into differences
in cognitive performance. This goal presents challenges on
several levels, and we discuss two major ones below: first,
the challenge of measuring individuals’ environments, and,
second, the challenge of capturing the content and structure
of single individuals’ mental representations.
Concerning the challenge of measuring individuals’ en-
vironments, one main limiting factor is currently the lack
of available data. Put simply, the field lacks context-aware
longitudinal projects that provide a characterization of the
environments experienced by individuals over time. Al-
though unprecedented, large amounts of contextualized text
and speech data are now available to scientists (e.g., Love
et al., 2017; Schröter & Schroeder, 2017), few of these data
sets present data on the level of the individual or distin-
guish between age groups. Crucially, any individual’s envi-
ronment consists of not only linguistic information, but also
rich multimodal sensorial information (e.g., De Deyne et al.,
2021). Thus, a major challenge for future research is to cre-
ate individual-annotated, multimodal corpora accessible for
research. There is optimism in the field that the rise of the
quantified-self movement and the availability of new "digital
tracing" methods can provide multiple data streams to feed
computational modeling efforts that use these data to create
models of individual’s mental representations (Wulffet al.,
2019). These models would be crucial to generate expecta-
tions about the actual estimates of individuals’ mental repre-
sentations.
Concerning the challenge of mapping individuals’ mental
representations, and as discussed above, there are now differ-
ent approaches to investigating the size and structure of men-
tal representations (see Table 1). However, not all techniques
are equally suited to uncover the full breadth of a single indi-
vidual’s mental representation (scope) while ensuring com-
4WULFF, DE DEYNE, AESCHBACH, AND MATA
parability between individuals in the lexical-semantic space
covered (comparability). For example, verbal fluency tasks
are limited to specific categories (e.g., animals) and the struc-
ture of a single semantic category may not be representative
of the entire semantic network (De Deyne & Storms, 2008).
In addition, learning about within-category structure is not
very informative about the links between categories. In turn,
other methods, such as relatedness judgments, can be power-
ful in providing comparability between individuals because
participants can be presented with the same pairs of concepts.
However, asking individuals to provide ratings to all possi-
ble pairs becomes prohibitively demanding with even small
sets of stimuli because this implies thousands of paired-
comparisons (Wulffet al., 2018, October 29)1. Compared
to other methods, free associations are relatively economi-
cal and can provide high scope. However, free-association
methods that ask individuals to generate associations in a
snowball method (Morais et al., 2013) may reduce compa-
rability across participants due to path-dependency in any
individual’s search of the representational store. In com-
parison, free associations with fixed, experimenter-generated
lists may provide a better choice in both scope and compara-
bility. One should note that free association has also been the
method of choice for most previous large-scale assessments
of aggregate semantic networks (De Deyne & Storms, 2008;
Steyvers et al., 2004).
All in all, our review of the literature, as well as assess-
ment of the challenges above, highlights several future direc-
tions: first, providing a better description of the idiosyncratic
experiences of individuals that can inform computational and
learning models of linguistic and semantic cognition, and
second, mapping the mental representations of single indi-
viduals that can later be matched to expectations about the
role of experience in cognitive performance and the potential
consequences. In what follows, we provide our own attempt
at tackling the second challenge of mapping individual se-
mantic networks and linking them to cognitive performance.
An example study for the Representation–Performance
pathway: MySWOW
We designed and piloted the My Small World of Words
(MySWOW) study to assess the feasibility of mapping indi-
vidual semantic networks from word associations and ana-
lyzing how they are linked to cognitive performance.
The procedure of our MySWOW study was inspired by
ongoing efforts to obtain word association norms for several
languages in a large online citizen-science project, the Small
World of Words (SWOW) study (https://smallworldofwords.
org). The project currently includes data from over half a
million participants in 15 languages. So far, this project has
resulted in a set of useful linguistic resources for both Dutch
(De Deyne et al., 2012) and English (De Deyne et al., 2018)
and will release additional resources for other languages
including German (https://smallworldofwords.org/de) in the
near future. The adoption of the same procedure for the large,
population-based SWOW and the smaller study of individu-
als, MySWOW, promises future assessments of the compara-
bility between the results from aggregate networks and those
of single individuals or groups (e.g., younger, older).
We largely adopted the same procedure used in SWOW,
which asks participants to provide three associates to a given
cue (e.g., "cat"). However, SWOW typically presents vol-
unteers with 18 cues that provide responses in the course
of minutes. In contrast, we asked each individual partici-
pant to provide answers to thousands of cues over the course
of weeks. Specifically, each participants in our study was
asked to provide three associations to a total of 3,000 unique
cues and 600 repeated cues, resulting in a total of 10,800 re-
sponses per participant. The repeated cues were included in
order to assess the consistency of participants’ associations.
We obtained data for four younger (aged 24 to 28) and
four older (aged 68 to 70) native German speakers. In an ini-
tial session, participants were briefed in person and received
instructions concerning the dedicated online tool that they
could use to complete the word associations. Participants
then completed the word associations from home over the
course of 2-4 weeks. After finishing the word-association
task, participants returned to the laboratory for cognitive test-
ing on several memory and linguistic tasks. In particular, we
focused on verbal fluency, paired associated, and episodic
memory tasks that have often been used to estimate and un-
derstand individual and age differences in cognitive perfor-
mance (Ramscar et al., 2017; Zemla & Austerweil, 2018)2.
Differences in the semantic networks of younger and
older individuals
In what follows, we document our efforts to obtain in-
dividual semantic networks from word association data in
MySWOW and report a qualitative comparison of the results
with those from past work, which suggests structural differ-
ences between the semantic representations of younger and
older adults (Dubossarsky et al., 2017; Wulffet al., 2019).
For each individual, we created uni-partite networks by
placing weighted edges between responses and correspond-
ing cues. This resulted in eight individual networks con-
taining 4,836 to 6,461 nodes, which is comparable in size
to previously obtained large-scale aggregate associative net-
works (Morais et al., 2013; Steyvers et al., 2004). Figure
2 illustrates the network of one individual, highlighting the
1A recent alternative approach (Roads & Love, 2020) that relies
on ranking and active learning algorithms can provide a solution to
the problem of scale.
2We provide a full account of our design, participants, and
methods in a companion data paper https://www.overleaf.com/read/
bdqyfgzsrgpr)
SEMANTIC NETWORKS AND COGNITIVE AGING 5
most central words according to PageRank and the underly-
ing structure by identifying clusters extracted using the Lou-
vain method (Blondel et al., 2008).
Networks strongly overlapped in content with each other
and with existing SWOW data sets. Table 2 presents the ten
most central words according to node degree for each of the
individual networks. Words in bold are also among the most
central words in either the English (De Deyne et al., 2019) or
the German3SWOW data. Considering only the 2,111 words
shared between all individual networks, the degree distribu-
tions of individual networks showed average correlations of
r=.57. Consistent with earlier work suggesting a progres-
sion of network differentiation with age, we found correla-
tions to be smaller between older adults (.48 <r< .58) than
between younger adults (.58 <r< .71).
In line with previous work (e.g., Dubossarsky et al., 2017;
Wulffet al., 2018, October 29; Wulffet al., 2016), we com-
pared the macroscopic structure of younger and older adults’
networks in terms of their size N, average degree hki, aver-
age local clustering coefficient C, and average shortest path
length L(see Table 3). On average, older adults’ networks
were larger (NOA =5921 vs. NY A =5133), had lower de-
grees (hkiOA =1.39 vs. hkiY A =1.66), lower clustering
coefficients (COA =0.053 vs. CYA =0.115), and larger
shortest path lengths (LOA =8.87 vs. CY A =7.45) than
those of younger adults. Except for one older adult, whose
network had structural characteristics more similar to those
of the younger adults (Row 5 in Table 3), the same patterns
were found for all possible pairwise comparisons of younger
and older adults’ networks. Moreover, the same patterns also
emerged when we focused on the shared networks consisting
of the 2,111 nodes present in all networks (Table 3).
Overall, the observed differences between younger and
older adults’ individual networks are well-aligned with the
results of previous age-comparisons on the basis of aggre-
gate association networks (Dubossarsky et al., 2017), aggre-
gate verbal fluency networks (Wulffet al., 2018, October 29),
and individual similarity rating networks (Wulffet al., 2018,
October 29).
Linking individual networks to cognitive performance
We wanted to show the feasibility of using the networks
described above to understand individual cognitive perfor-
mance. Individual networks can be linked to cognitive per-
formance at different levels. At the network level, macro-
scopic properties of the network, such as average degree or
clustering, could be used to predict overall performance in a
given task (Kenett & Faust, 2019). At the node level, micro-
scopic properties, such as node degree or the shortest path
length between two nodes, can be used to predict trial-level
performance, such as the order of retrievals in a verbal flu-
ency task. There is value in understanding both kinds of
links: Network-level links may help understand individual
differences in cognitive performance that emerge globally
from structural differences in semantic networks, whereas
node-level links may help us understand when and how in-
dividual differences materialize for specific stimuli or con-
texts. The small sample size (N=8) of our pilot study limits
comparisons at the network-level, and therefore what follows
will focus on associations of node-level links and cognitive
performance. Specifically, we report how specific node char-
acteristics estimated from our individual networks—such as
node centrality, measured using PageRank, and node simi-
larity, measured using Katz’ walk similarity (De Deyne et
al., 2016)—correspond to response patterns in two verbal flu-
ency tasks, a paired associative learning task, and an episodic
memory task.
In the animal and letter verbal fluency tasks, individuals
were asked to retrieve as many animal words or words start-
ing with the letter S, respectively, as they could, within 5
minutes. The retrieved words had substantially higher cen-
trality than words that were not retrieved. Specifically, for
seven out of eight individuals, retrieved animals were more
central in their respective networks as compared to other an-
imals contained in the respective networks that were not re-
trieved by the individual (Figure 3A). Similarly, for all eight
individuals, retrievals of words starting with the letter S had
higher centrality in the respective networks than other words
starting with any same letter contained in their networks
(Figure 3B). We also found words occurring directly adja-
cent within retrieval sequences to be more similar to each
other than words more distant from each other. Comparing
directly adjacent words (Lag 1) to words that were three re-
sponses apart, this was the case for every individual in the
animal fluency task (Figure 3C) and for six of the eight indi-
viduals in the letter fluency task (Figure 3D). In the episodic
memory task, participants freely recalled words from previ-
ously studied word lists. Both retrieved words and intrusions
tended to have, on average, higher centrality than missing
words (Figure 3E). This pattern held for five out of eight in-
dividuals. Furthermore, both retrieved words and intrusions
were, on average, more similar to other retrieved words than
to missing words (Figure 3F) for seven of eight individuals.
Finally, in the associate recall task, which required individu-
als to retrieve previously learned word pairs, retrieved word
pairs had, on average and for all eight individuals, higher
similarity than words pairs that were not retrieved (Figure
3G).
Each of these links demonstrates that individual networks
can be used to predict an individual’s pattern of behavior
in several cognitive measures. The critical question arising
from this framework, however, is whether individual net-
works reveal idiosyncratic differences that can be used to un-
derstand an individual’s behavior better than would be pos-
sible using an aggregate representation derived from the re-
3German SWOW data was downloaded on January 25, 2021.
6WULFF, DE DEYNE, AESCHBACH, AND MATA
Figure 2
Associative network of one participant. Node size represents PageRank centrality, colors communities detected using the
Louvain algorithm (Blondel, Guillaume, Lambiotte, &Lefebvre, 2008). Words were translated from the original German
and show the most important words in the respective community, with size representing the word’s importance in the overall
network as determined by PageRank.
Table 2
Most Central (Degree) Words in Individuals’ Networks
Younger adults Older adults
24 years 27 years 27 years 28 years 68 years 68 years 69 years 70 years
man music money money human Germany music state
music food write music country instrument1development country
work work beautiful water animal military money goods
change school food Germany water car exact first name
water money school work occupation water wood computer
family clothing read food work country make instrument
learn write important furniture food food school family
wood instrument economy instrument car name computer university
church name large love instrument computer children church
love summer family clothes male child work month
Note. Words in bold face are also among the ten most central words in the German (money,music,work,school,food,water,car,love,
green,important) or English (money,food,water,car,music,green,red,love,work,old) SWOW data sets. 1musical instrument
sponses of many individuals. To assess this, we compared
the magnitude of the effects presented in Figure 3, e.g., the
PageRank of retrieved versus non-retrieved words in verbal
fluency tasks, which were derived using individuals’ per-
sonal networks, against the magnitude of effects resulting
from using an aggregate derived from combining the re-
sponses from all eight individuals. This analysis revealed
that the aggregate network produced, on average, equally
large, if not larger effects than individual networks (see black
lines in Figure 3).
The similar or improved performance of the aggregate
compared to individuals’ own networks could be due to dif-
ferences in both the reliability and precision of measurement
of the different types of networks. First, on average, only .53
of three possible associations given to repeated cues were
also given during the first encounter of the same cue, indicat-
ing, at best, moderate levels of internal consistency. Further-
more, focusing on the 2,111 shared nodes, we found individ-
SEMANTIC NETWORKS AND COGNITIVE AGING 7
Table 3
Macroscropic differences in semantic network structure in terms of degree (k), clustering coefficient (C) and average path
length (L) for full networks and common networks based on nodes shared among participants.
Group Age Full network Common network
|V| hkiC L |V| hkiC L
Young 24 5780 1.52 .091 8.15 2111 1.51 .153 7.66
Young 27 4836 1.73 .115 7.13 2111 1.77 .174 6.70
Young 27 4920 1.65 .119 7.36 2111 1.66 .202 6.81
Young 28 4995 1.72 .136 7.14 2111 1.81 .220 6.54
Old 68 5275 1.68 .059 6.54 2111 1.66 .110 5.94
Old 68 6461 1.31 .045 9.33 2111 1.23 .072 8.86
Old 69 6157 1.39 .053 8.37 2111 1.38 .093 7.76
Old 70 5792 1.20 .055 11.3 2111 1.10 .114 10.8
uals’ networks to share considerably more variance with the
aggregate, as indicated by average correlations of ¯r=.61 and
¯r=.55 for PageRank and cosine distributions, respectively,
than they shared, on average, with the networks of other indi-
viduals (PageRank: ¯r=.34; Cosine: ¯r=.33). These results
suggests that the individual networks are a noisy measure-
ment of a shared component that is more reliably captured
by an aggregate network. Second, the aggregate network
is given access to more detailed information on the relative
strength of associations between nodes in the networks. For
individuals, only 2.2% to 5.5% of word associations occurred
more than once, and these repetitions occurred a maximum
of two or three times per individual. For the aggregate rep-
resentation, however, 20.5% of associations occurred more
than once, with counts reaching as high as 15, leading to
a more graded pattern of edge strengths between network
nodes in the aggregate relative to the individual networks.
One should note that this implies that the comparison of in-
dividual and aggregate networks is largely confounded with
a comparison of weighted and unweighted networks.
What have we learned from MySWOW?
To summarize, our approach was successful in obtaining
large-scale individual networks for a small sample of eight
participants, which included both younger and older adults,
suggesting that the approach can be used to capture the men-
tal representations of individuals from several age groups.
Individuals’ networks were highly similar to each other in
terms of content, and these contents were qualitatively sim-
ilar to those of past efforts to produce aggregate networks,
such as SWOW. Nevertheless, the results from the 8 partic-
ipants suggested differences in their structural composition,
showing age-related differences similar to these found in past
work (Dubossarsky et al., 2017; Wulffet al., 2018, Octo-
ber 29). Despite the small sample size, our approach seems
promising in offering a potential window into adult age dif-
ferences in the content and structure of the aging lexicon and
the individual differences that can arise across adulthood as
a function of cumulative experience. Naturally, future work
needs to strive to collect data from more individuals to assess
whether network-level properties are systematically related
to differences in cognitive performance between individuals.
Our results also showed that estimates of node central-
ity and relatedness derived from single individuals’ networks
exhibited strong links to individuals’ responses in four cog-
nitive tasks, including two verbal fluency and two memory
tasks. These results demonstrate that it is feasible, in princi-
ple, to map information from individual semantic networks
onto additional linguistic and memory performance at the
single participant level.
Our results demonstrate that our approach is, nevertheless,
limited. Most crucially, our results suggest that individual
networks were less powerful in accounting for individuals’
response patterns in the additional measures of linguistic and
memory performance than an aggregate network composed
of all individuals’ responses. The advantage of aggregate
over individual networks presents a major challenge to the
claim that it is important to assess individual networks to
understand individual differences in cognitive performance.
Before one fully dismisses our claim, one should distinguish
two explanations for these findings that have different im-
plications. First, one possibility is that our approach was
not able to capture relevant individual differences in the con-
tent of the semantic representations, for example, because of
the use of a pool of cues containing mostly high-frequency
words for which individual differences in semantic represen-
tations may be small. In this case, the breadth of cues used
would need to be expanded so that the networks generated
are better able to capture the idiosyncratic nature of individ-
uals’ representations. Second, another possibility is that our
approach may not have been able to measure the strength
of association between various nodes (i.e., edges) with suffi-
cient precision and reliability. In this case, additional work
8WULFF, DE DEYNE, AESCHBACH, AND MATA
Retrieved
Not
retrieved
0.9
1.0
1.1
1.2
1.3
1.4
Centrality
A
Retrieved
Not
retrieved
0.9
1.0
1.1
1.2
1.3
1.4
Centrality
B
Lag 1
Lag 2
Lag 3
.0
.1
.2
.3
.4
.5
Similarity
C
Lag 1
Lag 2
Lag 3
.0
.1
.2
.3
.4
.5
Similarity
D
Retrieved
Missing
Intrusion
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
Centrality
E
Retrieved
Missing
Intrusion
.00
.05
.10
.15
Similarity
F
Retrieved
Not
retrieved
.00
.05
.10
.15
.20
Similarity
G
Figure 3
Links between mental representations and cognitive performance for the four younger (blue lines) and the four older adults
(yellow lines). The panels show the difference in PageRank centrality between retrieved and non-retrieved words in animal
(A) and letter fluency (B), the average cosine similarity of words one, two, and three words apart in animal (C) and letter
fluency (D), the PageRank centrality of retrieved, missing, and intrusion words in the episodic memory task (E), the average
cosine similarity of retrieved, missing, and intrusion words to (other) retrieved words in the episodic memory task (F), and the
average cosine similarity of retrieved versus non-retrieved word pairs in the associative recall task (G). The black line shows
the effects under the aggregate representaions. Centrality is measured as PageRank times the number of nodes in the nodes in
the network, to account for differences in network sizes.
may be needed that considers both obtaining larger sets of re-
sponses per cue, having participants provide responses to the
same cue on repeated occasions, or asking individuals to pro-
vide additional ratings (Roads & Love, 2020). In our view,
these possibilities speak to the need for future work that guar-
antees appropriate scope and reliability rather than a clear
rejection of the idea that individual networks can be helpful
to understanding the link between experience and cognitive
performance.
General Discussion
Older adults tend to perform worse on a broad set of cog-
nitive tasks, and such findings are commonly attributed to a
decline in fluid cognitive abilities (Healey & Kahana, 2016;
Salthouse, 2010). However, some have argued that changes
in the underlying size and structure of representations can
contribute to age differences in cognitive performance, for
example, due to activation-spreading across many targets in
memory (i.e., fan effect; Buchler and Reder, 2007) or dif-
ficulties in discrimination learning between many similar
items (Ramscar et al., 2017), which can be a direct conse-
quence of individuals’ cumulative interaction with the phys-
ical, linguistic, and social world. In this paper, we proposed
an empirical agenda to provide an estimate of the extent
to which idiosyncratic experiences matter for the size and
structure of mental representations and, ultimately, individ-
ual and age differences in cognitive performance. Despite
the promise we see in this approach, a number of other, major
challenges remain.
First, our suggested approach of mapping individual se-
mantic networks is open to the criticism that no elicitation of
individuals’ knowledge store is independent of the process
by which the representation is accessed. As a result, any in-
dividual or age differences detected cannot be unequivocally
assigned to the nature of association in the knowledge store
(representation), but can in principle result from the process
by which this representation is searched and accessed (Jones,
Hills, et al., 2015; Kenett et al., 2020; Kraemer et al., 2021,
February 10; Siew et al., 2019). So far, there seems to be no
clear consensus concerning the extent to which representa-
tion and process are entangled in the kinds of tasks presented
here (Abbott et al., 2015; Jones, Hills, et al., 2015). Recent
studies have addressed this question empirically and found
SEMANTIC NETWORKS AND COGNITIVE AGING 9
that the retrieval processes might underestimate semantic ef-
fects due to frequency biases, which need be mitigated us-
ing appropriate transformations to associative strength mea-
sures (De Deyne et al., 2019). More generally, it is un-
likely that a single approach can provide a definitive reso-
lution to this conundrum and there are two possible ways
forward that we would like to emphasize. One potential way
to distinguish the role of structure versus process could be
to use multiple methods of assessment (e.g., word associa-
tion, verbal fluency, relatedness judgments) to elicit mental
representations. To the extent that these different assessment
methods are associated with different search processes and
strategies, convergent evidence could provide some support
for the role of representation. A related approach could in-
volve the use of the same elicitation method under different
explicitly instructed search strategies that could equate these
across participants (Wulffet al., 2020, January 16; Wulffet
al., 2013). A second direction could involve using neuro-
science techniques to distinguish between search (i.e., con-
trol) processes and representational components (Hoffman &
Morcom, 2018), in particular to the extent that individuals’
representational space has a signature in the functional orga-
nization of the brain (Huth et al., 2012).
Second, our approach proposed a dedicated experimental
procedure (i.e., word-association task) to capture the mental
representations of single individuals. However, there may
be more efficient and powerful alternatives that should be
explored, for example massive recording and analysis of
naturalistic stimuli and behavior, including both linguistic
sources such as speech (Mehl, 2017) and text production
(Banda et al., 2020), and non-linguistic ones such as senso-
rial input (Sadeghi et al., 2015). These approaches could be-
come particularly powerful as such data become increasingly
connected with other information about individuals, for ex-
ample, if these are collected with informed consent in larger
surveys or longitudinal household panels.
Third, the field has seen considerable progress in the com-
putational modeling of the acquisition of mental representa-
tions (Jones, Willits, et al., 2015) as well as search (Abbott
et al., 2015). However, neither we nor others have used such
models to systematically investigate individual differences.
Ideally, the deployment of large-scale empirical efforts as the
one we discussed above will be preceded by simulation and
modeling work that can provide a guide concerning which
structures and magnitudes of differences are most likely to
be expected and captured with which cognitive paradigm.
Fourth, and finally, most of the research discussed above
(including our own approach) is correlational in nature, in
that it aims to establish a correlation between individual ex-
perience and the contents and structure of mental represen-
tation or cognitive performance. In order to help establish
causality, however, one would need a comparison of individ-
uals assigned to different environmental exposures through
either natural or, ideally, controlled experiments. There are
a few candidate training strategies, including training of spe-
cific physical or virtual environments (Miller et al., 2013),
artificial or natural languages (Pothos, 2007), and even com-
plex narratives (Heusser et al., 2021)). Regardless of the ex-
act type of information and mode of exposure, it remains a
challenge to obtain meaningful and reliable individual esti-
mates of the emergent mental representations and their po-
tential effects across the life span. We hope our review of
past work and our own proposal contribute to this larger ef-
fort.
Conclusion
We have argued that quantifying individual and age dif-
ferences in the size and structure of human knowledge is
important because this represents a missing link in estimat-
ing the role of cumulative experience in cognitive perfor-
mance. We specifically proposed an empirical agenda that
combines tracking individuals’ idiosyncratic experiences and
broad mapping of their mental representations and cognitive
performance. We hope such steps can move us closer to un-
derstanding the role of the environment in the mental lexicon
and its implications for cognitive aging.
References
Abbott, J. T., Austerweil, J. L., & Griffiths, T. L. (2015). Ran-
dom walks on semantic networks can resemble op-
timal foraging. Psychological Review.
Banda, J. M., Tekumalla, R., Wang, G., Yu, J., Liu, T.,
Ding, Y., Artemova, K., Tutubalina, E., & Chow-
ell, G. (2020). A large-scale covid-19 twitter chat-
ter dataset for open scientific research – an interna-
tional collaboration.
Benedek, M., Kenett, Y. N., Umdasch, K., Anaki, D., Faust,
M., & Neubauer, A. C. (2017). How semantic mem-
ory structure and intelligence contribute to creative
thought: A network science approach. Thinking &
Reasoning,23(2), 158–183.
Blondel, V. D., Guillaume, J.-L., Lambiotte, R., & Lefebvre,
E. (2008). Fast unfolding of communities in large
networks. Journal of statistical mechanics: theory
and experiment,2008(10), P10008.
Brysbaert, M., Stevens, M., Mandera, P., & Keuleers, E.
(2016). How Many Words Do We Know? Prac-
tical Estimates of Vocabulary Size Dependent on
Word Definition, the Degree of Language Input
and the Participant’s Age. Frontiers in psychology,
7(021006), 55–11.
Buchler, N. E. G., & Reder, L. M. (2007). Modeling age-
related memory deficits: A two-parameter solution.
Psychology and aging,22(1), 104–121.
10 WULFF, DE DEYNE, AESCHBACH, AND MATA
De Deyne, S., Navarro, D. J., Perfors, A., & Storms, G.
(2016). Structure at every scale: A semantic net-
work account of the similarities between unre-
lated concepts. Journal of Experimental Psychol-
ogy: General,145(9), 1228.
De Deyne, S., Navarro, D. J., & Storms, G. (2012). Bet-
ter explanations of lexical and semantic cognition
using networks derived from continued rather than
single-word associations. Behavior Research Meth-
ods,45(2), 480–498.
De Deyne, S., Navarro, D. J., Collell, G., & Perfors, A.
(2021). Visual and affective multimodal models of
word meaning in language and mind. Cognitive Sci-
ence,45(1), e12922.
De Deyne, S., Navarro, D. J., Perfors, A., Brysbaert, M., &
Storms, G. (2018). The “Small World of Words”
English word association norms for over 12,000 cue
words, 1–20.
De Deyne, S., Navarro, D. J., Perfors, A., Brysbaert, M.,
& Storms, G. (2019). The “small world of words”
english word association norms for over 12,000
cue words. Behavior research methods,51(3), 987–
1006.
De Deyne, S., & Storms, G. (2008). Word associations:
Norms for 1,424 dutch words in a continuous task.
Behavior research methods,40(1), 198–205.
Dubossarsky, H., De Deyne, S., & Hills, T. T. (2017). Quan-
tifying the structure of free association networks
across the life span. Developmental psychology,
53(8), 1560.
Gobet, F., & Simon, H. A. (1996). Recall of random and dis-
torted chess positions: Implications for the theory
of expertise. Memory &cognition,24(4), 493–503.
Healey, M. K., & Kahana, M. J. (2016). A four-component
model of age-related memory change. Psychologi-
cal Review,123(1), 23.
Heusser, A. C., Fitzpatrick, P. C., & Manning, J. R. (2021).
Geometric models reveal behavioural and neural
signatures of transforming experiences into mem-
ories. Nature Human Behaviour, 1–26.
Hoffman, P., & Morcom, A. M. (2018). Age-related changes
in the neural networks supporting semantic cogni-
tion: A meta-analysis of 47 functional neuroimag-
ing studies. Neuroscience &Biobehavioral Re-
views,84, 134–150.
Huettig, F., & Pickering, M. J. (2019). Literacy Advantages
Beyond Reading: Prediction of Spoken Language.
Trends in Cognitive Sciences,23(6), 464–475.
Huth, A. G., Nishimoto, S., Vu, A. T., & Gallant, J. L. (2012).
A Continuous Semantic Space Describes the Repre-
sentation of Thousands of Object and Action Cat-
egories across the Human Brain. Neuron,76(6),
1210–1224.
Jones, M. N., Willits, J., & Dennis, S. (2015). Models of se-
mantic memory. The Oxford Handbook of . . .
Jones, M. N., Hills, T. T., & Todd, P. M. (2015). Hidden pro-
cesses in structural representations: A reply to ab-
bott, austerweil, and griffiths (2015).
Kenett, Y. N., Beckage, N. M., Siew, C. S., & Wulff, D. U.
(2020). Cognitive network science: A new frontier.
Complexity.
Kenett, Y. N., & Faust, M. (2019). A Semantic Network Car-
tography of the Creative Mind. Trends in Cognitive
Sciences,23(4), 271–274.
Kraemer, P. M., Wulff, D. U., & Gluth, S. (2021, February
10). A sequential sampling account of semantic re-
latedness decisions. https://doi.org/10.31234/osf.io/
ksa2g
Li, H., N, K. Y., Kaixiang, Z., Cheng, L., Rongcan, Z., Tin-
grui, Y., Tengbin, H., & Jiang, Q. (2020). The re-
lation between semantic memory structure, asso-
ciative abilities, and verbal and figural creativity.
Thinking &Reasoning,0(0), 1–26.
Lindenberger, U. (2014). Human cognitive aging: corriger la
fortune? Science,346(6209), 572–578.
Love, R., Dembry, C., Hardie, A., Brezina, V., & McEnery, T.
(2017). The spoken bnc2014: Designing and build-
ing a spoken corpus of everyday conversations. In-
ternational Journal of Corpus Linguistics,22(3),
319–344.
Mehl, M. R. (2017). The electronically activated recorder
(ear): A method for the naturalistic observation of
daily social behavior [PMID: 28529411]. Current
Directions in Psychological Science,26(2), 184–
190. https://doi.org/10.1177/0963721416680611
Miller, J. F., Neufang, M., Solway, A., Brandt, A., Trippel,
M., Mader, I., Hefft, S., Merkow, M., Polyn, S. M.,
Jacobs, J., Kahana, M. J., & Schulze-Bonhage, A.
(2013). Neural Activity in Human Hippocampal
Formation Reveals the Spatial Context of Retrieved
Memories. Science,342(6162), 1111–1114.
Mol, S. E., & Bus, A. G. (2011). To Read or Not to Read:
A Meta-Analysis of Print Exposure From Infancy
to Early Adulthood. Psychological Bulletin,137(2),
267–296.
Morais, A. S., Olsson, H., & Schooler, L. J. (2013). Map-
ping the structure of semantic memory. Cognitive
science,37(1), 125–145.
Nation, K. (2017). Nurturing a lexical legacy: reading experi-
ence is critical for the development of word reading
skill. NPJ science of learning, 1–.
Nelson, D. L., Bennett, D. J., Gee, N. R., Schreiber, T. A., &
McKinney, V. M. (1993). Implicit memory: Effects
of network size and interconnectivity on cued re-
call. Journal of experimental psychology. Learning,
memory, and cognition,19(4), 747–64.
SEMANTIC NETWORKS AND COGNITIVE AGING 11
Pothos, E. M. (2007). Theories of artificial grammar learn-
ing. Psychological Bulletin,133(2), 227–244.
Ramscar, M., Hendrix, P., Shaoul, C., Milin, P., & Baayen, H.
(2014). The myth of cognitive decline: Non-linear
dynamics of lifelong learning. Topics in cognitive
science,6(1), 5–42.
Ramscar, M., Sun, C. C., Hendrix, P., & Baayen, H. (2017).
The mismeasurement of mind: Life-span changes in
paired-associate-learning scores reflect the “cost” of
learning, not cognitive decline. Psychological sci-
ence,28(8), 1171–1179.
Roads, B. D., & Love, B. C. (2020). Enriching imagenet with
human similarity judgments and psychological em-
beddings. arXiv preprint arXiv:2011.11015.
Rodd, J. M. (2020). Settling Into Semantic Space: An
Ambiguity-Focused Account of Word-Meaning
Access. Perspectives on Psychological Science,
174569161988586–17.
Sadeghi, Z., McClelland, J. L., & Hoffman, P. (2015). You
shall know an object by the company it keeps: An
investigation of semantic representations derived
from object co-occurrence in visual scenes. Neu-
ropsychologia,76(3), 52–61.
Sala, G., & Gobet, F. (2017). Experts’ memory superior-
ity for domain-specific random material generalizes
across fields of expertise: A meta-analysis, 1–11.
Salthouse, T. A. (2010). Selective review of cognitive aging.
Journal of the International neuropsychological So-
ciety,16(5), 754–760.
Schröter, P., & Schroeder, S. (2017). The developmental lexi-
con project: A behavioral database to investigate vi-
sual word recognition across the lifespan. Behavior
Research Methods,49(6), 2183–2203.
Siew, C. S., Wulff, D. U., Beckage, N. M., & Kenett, Y. N.
(2019). Cognitive network science: A review of re-
search on cognition through the lens of network rep-
resentations, processes, and dynamics. Complexity,
2019.
Steyvers, M., Shiffrin, R. M., & Nelson, D. L. (2004). Word
association spaces for predicting semantic similar-
ity effects in episodic memory. Experimental cogni-
tive psychology and its applications: Festschrift in
honor of Lyle Bourne, Walter Kintsch, and Thomas
Landauer, 237–249.
Twenge, J. M., Martin, G. N., & Spitzberg, B. H. (2019).
Trends in U.S. Adolescents’ media use, 1976–2016:
The rise of digital media, the decline of TV, and the
(near) demise of print. Psychology of Popular Me-
dia Culture,8(4), 329–345.
Verhaeghen, P. (2003). Aging and vocabulary scores: A
meta-analysis. Psychology and aging,18(2), 332–9.
Wulff, D. U., De Deyne, S., Jones, M. N., Mata, R., & Aging
Lexicon Consortium. (2019). New Perspectives on
the Aging Lexicon. Trends in Cognitive Sciences,
23(8), 686–698.
Wulff, D. U., Hills, T., & Hertwig, R. (2020, January 16).
Memory is one representation not many: Evidence
against wormholes in memory. https://doi.org/10.
31234/osf.io/b5ynj
Wulff, D. U., Hills, T., & Mata, R. (2018, October 29).
Structural differences in the semantic networks of
younger and older adults. https://doi.org/10.31234/
osf.io/s73dp
Wulff, D. U., Hills, T. T., & Hertwig, R. (2013). Worm holes
in memory: Is memory one representation or many?
Proceedings of the Annual Meeting of the Cognitive
Science Society,35(35).
Wulff, D. U., Hills, T. T., Lachman, M., & Mata, R. (2016).
The aging lexicon: Differences in the semantic net-
works of younger and older adults. Proceedings of
the Annual Meeting of the Cognitive Science Soci-
ety,38(38).
Zemla, J. C., & Austerweil, J. L. (2018). Estimating seman-
tic networks of groups and individuals from fluency
data. Computational Brain &Behavior, 1–23.
