Implicit word learning benefits from semantic richness: electrophysiological and behavioral evidence.

Abstract

Words differ considerably in the amount of associated semantic information. Despite the crucial role of meaning in language, it is still unclear whether and how this variability modulates language learning. Here, we provide initial evidence demonstrating that implicit learning in repetition priming is influenced by the amount of semantic features associated with a given word. Electroencephalographic recordings were obtained while participants performed a visual lexical decision task; the complete stimulus set was repeated once. Repetition priming effects on performance accuracy and the N400 component of the event-related brain potential were enhanced for words with many semantic features. These findings suggest a novel and important impact of the richness of semantic representations on learning and plasticity within the lexical-conceptual system; they are discussed in their relevance for assumptions concerning basic mechanisms underlying word learning.

Full text

Page 1

Journal of Experimental Psychology:
Learning, Memory, and Cognition
Implicit Word Learning Benefits From Semantic Richness:
Electrophysiological and Behavioral Evidence
Milena Rabovsky, Werner Sommer, and Rasha Abdel Rahman
Online First Publication, September 19, 2011. doi: 10.1037/a0025646
CITATION
Rabovsky, M., Sommer, W., & Abdel Rahman, R. (2011, September 19). Implicit Word
Learning Benefits From Semantic Richness: Electrophysiological and Behavioral Evidence.
Journal of Experimental Psychology: Learning, Memory, and Cognition. Advance online
publication. doi: 10.1037/a0025646

Page 2

RESEARCH REPORT
Implicit Word Learning Benefits From Semantic Richness:
Electrophysiological and Behavioral Evidence
Milena Rabovsky, Werner Sommer, and Rasha Abdel Rahman
Humboldt-Universita ¨t zu Berlin
Words differ considerably in the amount of associated semantic information. Despite the crucial role of
meaning in language, it is still unclear whether and how this variability modulates language learning.
Here, we provide initial evidence demonstrating that implicit learning in repetition priming is influenced
by the amount of semantic features associated with a given word. Electroencephalographic recordings
were obtained while participants performed a visual lexical decision task; the complete stimulus set was
repeated once. Repetition priming effects on performance accuracy and the N400 component of the
event-related brain potential were enhanced for words with many semantic features. These findings
suggest a novel and important impact of the richness of semantic representations on learning and
plasticity within the lexical-conceptual system; they are discussed in their relevance for assumptions
concerning basic mechanisms underlying word learning.
Keywords: semantic features, implicit learning, repetition priming, visual word processing, ERPs
Language ultimately aims to convey meaning. Yet individual
words differ widely in the amount of associated semantic infor-
mation. Given the key role of meaning in language, the richness of
semantic representations might be expected to drive language
processing and language learning. Indeed, recent evidence sug-
gests that semantic richness, as quantified, for example, by the
number of associates generated in free-association tasks (Nelson,
McEvoy, & Schreiber, 2004), the diversity of contexts in which a
word appears (Adelman, Brown, & Quesada, 2006), or the number
of associated semantic features as produced in feature-listing tasks
(McRae, Cree, Seidenberg, & McNorgan, 2005), facilitates visual
word processing in both semantic and nonsemantic tasks (Dun ˜a-
beitia, A´viles, & Carreiras, 2008; Pexman, Hargreaves, Edwards,
Henry, & Goodyear, 2007; Pexman, Holyk, & Monfils, 2003;
Pexman, Lupker, & Hino, 2002). However, as yet it remains
unclear whether and how the richness of semantic representations
may influence language learning. In the present study, we explored
this issue by investigating influences of the amount of associated
semantic features on implicit word learning as reflected in repeti-
tion priming.
Implicit learning occurs incidentally during information pro-
cessing and is often assumed to be based on prediction errors.
Specifically, it has been suggested that the brain incessantly and
automatically anticipates upcoming events based on an
experience-derived internal model of the environment. Deviations
between anticipated and factual events are assumed to drive ad-
aptations of internal representations to reduce future prediction
errors and optimize processing (den Ouden, Friston, Daw, McIn-
tosh, & Stephan, 2009; Friston, 2009; McClelland, 1994;
McLaren, 1989; Schultz & Dickinson, 2000). Connectionist mod-
els of cognitive processes have implemented this assumption by
learning rules that are based on error back-propagation (McClel-
land, 1994). Thus, the connection weight adjustments considered
to underlie learning are proportional to the discrepancy between
model-generated and correct output, representing predicted and
factual information, respectively. Similarly, on a neural level,
plasticity may be induced via prediction error signals altering
synaptic efficacy (e.g., den Ouden et al., 2009; Friston, 2009).
A well-established measure of implicit learning is repetition
priming, that is, the processing facilitation caused by the repeated
encounter with a given stimulus (Schacter & Graf, 1986) that often
goes along with reductions in brain activity (Henson, 2003). Prim-
ing effects are often considered to reflect an increase in the
accessibility of the representations involved in processing the
repeated stimulus (Graf & Mandler, 1984; Henson, 2003; Ratcliff,
Hockley, & McKoon, 1985; Stark & McClelland, 2000; Wiggs &
Martin, 1998). From the perspective of connectionism (Becker,
Moscovitch, Behrmann, & Joordens, 1997; Stark & McClelland,
2000) or predictive coding (Friston, 2009), such increases in
accessibility can be viewed as consequences of the continuous
adaptation of the system aiming to reduce future prediction errors.
Each encounter induces a refinement of the connections involved
Milena Rabovsky, Werner Sommer, and Rasha Abdel Rahman, Depart-
ment of Psychology, Humboldt-Universita ¨t zu Berlin, Berlin, Germany.
This work was supported by a scholarship from the Berlin School of
Mind and Brain to Milena Rabovsky and by German Research Foundation
Grants AB 277/3 and 5 to Rasha Abdel Rahman. We would like to thank
Melih Bakirtas for assisting in data acquisition and Daniel Schad for
helping with the generalized linear mixed model analyses.
Correspondence concerning this article should be addressed to Milena
Rabovsky, Department of Psychology, Humboldt-Universita ¨t zu Berlin,
Rudower Chaussee 18, 12489 Berlin, Germany. E-mail: milena
.rabovsky@hu-berlin.de
Journal of Experimental Psychology:
Learning, Memory, and Cognition
2011, Vol. ●●, No. ●, 000–000
© 2011 American Psychological Association
0278-7393/11/$12.00DOI: 10.1037/a0025646
1

Page 3

in processing the stimulus. These adaptations bring about the
observed repetition-dependent performance benefit and reductions
of cortical activity. It has also been suggested that repetition
priming “reflects the very same connection adjustment process that
gives rise to fluent reading ability” (Stark & McClelland, 2000, p.
965), further indicating repetition priming to provide a well-suited
measure of the processes underlying implicit word learning.
Directly relevant to the question whether such learning may be
modulated by semantic richness is the connectionist attractor
model of lexical-conceptual processing proposed by Cree, McRae,
and McNorgan (1999). This model relies on an error-based learn-
ing rule that is sensitive to the amount of semantic features
associated with a given word and yields substantial positive cor-
relations between the number of features and the computed error
driving connection adaptations. Even though the model has pri-
marily been used to account for performance in speeded lexical
tasks at the endpoint of learning, it predicts enhanced adaptation,
and thus enhanced implicit learning, for words with many semantic
features.
In the present study, we investigated influences of the amount of
associated semantic features on implicit word learning with a
repetition priming design. Participants performed lexical decisions
on visual words differing in the amount of semantic features
according to the feature production norms of McRae et al. (2005;
cf. Table 1 and the Appendix). These norms were obtained by
asking participants to list semantic features (e.g., “is small,” “can
swim,” etc.) for concrete words. The number of features listed for
each word provides a measure of semantic richness, with partici-
pants listing more features for some words (e.g., 16 for car) than
for others (e.g., seven for cork). The complete stimulus set was
presented twice, and implicit learning was assessed by repetition
priming effects on performance as well as event-related brain
potentials (ERPs). Repetition priming has been consistently shown
to reduce the amplitude of the N400 component, a negative-going
ERP deflection at centro-parietal electrode sites with a maximum
of about 400 ms (e.g., Doyle, Rugg, & Wales, 1996). The N400
has been related to semantic processing (Kutas & Federmeier,
2011) and—relevant for present purposes—also to the implicit
prediction error assumed in connectionist models (McClelland,
1994). From this perspective, repetition-induced N400 amplitude
reductions may reflect reduced error values for repeated stimuli
due to connection adjustments triggered by the previous presenta-
tion. On the basis of the above-mentioned model of Cree et al.
(1999), we predicted semantic richness to enhance adaptation.
Specifically, repetition priming effects on performance and ERPs
should be augmented for words associated with many as compared
to few semantic features.
Method
Participants
Twenty-four native English speakers (12 women) with mean
age of 25 years (range ? 19–32) were paid €7 (?$10 U.S.) per
hour for taking part in the study, after giving written informed
consent. Participants had normal or corrected-to-normal visual
acuity; 20 of them were right-handed.
Materials
Stimuli were 160 concrete English nouns (80 per semantic
richness condition; cf. the Appendix) and 160 pseudowords.
Within the word stimuli, the number of semantic features (McRae
et al., 2005) varied across two levels (M ? 16.1 vs. 9.3), while the
number of associates, familiarity, concreteness, length, word fre-
quency, and number of orthographic neighbors, phonological
neighbors, phonemes, and syllables were controlled (all Fs ? 1;
please see Table 1). Most stimulus characteristics were taken from
McRae et al. (2005). Number of associates and concreteness
values were taken from Nelson et al. (2004). The word frequency
values were retrieved from the English Lexicon Project (Balota et
al., 2007) and represent log-transformed frequencies based on the
HAL corpus (Lund & Burgess, 1996). CELEX-based frequency
values were not intentionally matched between conditions, but post
hoc comparison indicated that the difference was not statistically
significant. Pseudowords were constructed by recombining the
letters of the word stimuli (e.g., osnop from spoon). Pseudowords
were pronounceable but orthographically less typical, as assessed
by bigram frequencies and orthographic neighborhood size, Fs (1,
318) ? 9.2, ps ? .01 (Balota et al., 2007).
Procedure
Participants sat in a dimly lit, sound-attenuated, and electrically
shielded chamber. Stimuli were presented in black on a light blue
screen. Each trial began with a fixation cross, shown for 1.5 s,
followed by a letter string, terminated with a response or when 3 s
had elapsed. Immediately thereafter, the next trial started. Partic-
ipants were instructed to indicate as fast and accurately as possible
whether the letter string was a word or not by pressing one of two
buttons with their left or right index finger. Response hand-to-
stimulus assignments were counterbalanced. The complete stimu-
lus set was presented twice, in two successive blocks. Each block
was subdivided into two parts each consisting of 80 words (40 per
semantic richness condition) and 80 pseudowords that were pre-
sented in a different random order for each participant and block.
The order of the two parts was identical for the two successive
blocks, counterbalanced across participants. Thus, the lag between
subsequent presentations of the same word varied randomly be-
tween 160 and 480 intermediate words. Overall, the experiment
comprised 640 trials, subdivided into 16 blocks of 40 trials each,
separated by short breaks.
Table 1
Stimulus Characteristics
FeaturesMany Few
Number of features
Number of associates
Familiarity
Concreteness
Length (number of letters)
Word frequency
Number of orthographic neighbors
Number of phonological neighbors
Number of phonemes
Number of syllables
16.1
13.6
6.3
6.1
5.4
8.3
6.6
14.4
4.4
1.6
9.3
13.6
6.1
6.0
5.3
8.3
6.4
14.4
4.3
1.6
2
RABOVSKY, SOMMER, AND ABDEL RAHMAN

Page 4

EEG Recording and Analysis
The electroencephalogram (EEG) was recorded with Ag/AgCl
electrodes from 62 sites according to the extended 10-20 system
and referenced to the left mastoid. Electrode impedance was kept
below 5 k?. Bandpass of amplifiers (Brainamps) was 0.032–70
Hz; sampling rate was 500 Hz. Offline, the EEG was transformed
to average reference, recommended as being less biased than other
common references (Picton et al., 2000). Eyeblink artifacts were
removed with a spatiotemporal dipole modeling procedure using
BESA software. After applying a 30-Hz low-pass filter, the con-
tinuous EEG was segmented into epochs of 1 s, including a 200-ms
prestimulus baseline. Trials with remaining artifacts or incorrect or
missing responses were discarded.
ERP analyses focused on the N400 component, a negative ERP
wave at centro-parietal sites peaking at about 400 ms. In accor-
dance with the literature on ERP repetition effects (e.g., Doyle et
al., 1996), we selected a cluster of centro-parietal sites (Cz, CP1,
CPz, CP2, Pz) and analyzed mean amplitudes at these sites be-
tween 350 and 450 ms as well as between 450 and 550 ms.
Amplitude values were submitted to repeated measures analyses of
variance (ANOVAs) including the factors Features (many vs.
few), Repetition (first vs. second presentation), and Electrode Site
(Cz, CP1, CPz, CP2, Pz). In addition to subject-based (F1) anal-
yses, we also conducted item-based (F2) ANOVAs, for which
ERP amplitudes were averaged over subjects instead of over items
(Hutzler et al., 2004).
Results
Performance
Because error rates (ERs) were very low, violating the ANOVA
precondition of normal distribution, they were analyzed using
logistic generalized linear mixed models, which do not entail such
constraints and provide the additional advantage of allowing for
simultaneous inclusion of both subjects and items as crossed
random factors (Baayen, Davidson, & Bates, 2008; Schad, Nuth-
mann, & Engbert, 2010); response times (RTs) were analyzed
accordingly.
Error rates.
Analyses revealed no main effects of Features
(? ? .33, SE ? .24, p ? .18) or Repetition (? ? ?.03, SE ? .12,
p ? .80). Importantly, we found the predicted interaction between
Repetition and Features (? ? .55, SE ? .23, p ? .019; please see
Figure 1, bottom). ERs decreased with repetition for words with
many features (? ? .30, SE ? .16, p ? .057), while there was no
repetition-induced facilitation for words with few features (? ?
.24, SE ? .17, p ? .16). Accordingly, while there was no effect of
Features on accuracy during the first presentation (? ? .05, SE ?
.27, p ? .84), ERs were significantly lower for words with many
as compared to words with few semantic features during the
second presentation (? ? .60, SE ? .27, p ? .026).
Response times.
RTs did not differ between words with
many versus few semantic features (? ? .006, SE ? .015, t ?
0.434). Words were responded to faster during the second as
compared to the first presentation (? ? .113, SE ? .008, t ?
Figure 1.
Event-related potential waveforms at a centro-parietal electrode site (CPz) and topographical distribution of
repetition effects (first minus second presentation) at peak (480 ms). The topography in the center depicts t
values corresponding to the interaction between Repetition and Features (Repetition effects [first minus second
presentation] for words with many vs. few semantic features); for df ? 23, p ? .05 if t ? 2.069. Bottom: Error
rates (ERs; bars) and response times (RTs; dots) for lexical decisions. Error bars depict standard errors of the
mean.
Repetition priming effects for words with many (left) versus few (right) semantic features. Top:
3
SEMANTIC RICHNESS ENHANCES IMPLICIT WORD LEARNING

Page 5

13.977), and this effect was not modulated by the amount of
semantic features (? ? ?.012, SE ? .016, t ? ?0.731; see
Figure 1, bottom).
Electrophysiology
ERP waveforms and topographies are depicted in Figure 1, top.
The typical N400 repetition effect, with less negative amplitude for
repeated stimuli, was well pronounced for words with many fea-
tures but faint for words with few features. An ANOVA of mean
ERP amplitudes between 350 and 450 ms revealed no main effect
of Features (F1 ? 1, F2 ? 1). Repetition was not significant by
subjects, F1 (1, 23) ? 2.48, p ? .129, but was significant by items,
F2 (1, 158) ? 14.27, p ? 001. Importantly, we again obtained the
predicted interaction between Repetition and Features, F1 (1,
23) ? 10.27, p ? .004; F2 (1, 158) ? 12.13, p ? .001 (please see
Figure 1 for a t map depicting the topographical distribution of the
interaction), with a significant Repetition effect for words with
many features, F1(1, 23) ? 13.45, p ? .001; F2 (1, 79) ? 27.74,
p ? .001, but no such effect for words with few features (F1 ? 1,
F2 ? 1). Accordingly, while amplitudes were significantly more
negative for words with many features during the first presenta-
tion, F1 (1, 23) ? 5.12, p ? .033; F2 (1, 158) ? 4.99, p ? .027,
there was a Features effect into the opposite direction for repeated
stimuli, F1 (1, 23) ? 6.77, p ? .016; F2 (1, 158) ? 6.12, p ? .014.
In the segment between 450 and 550 ms, there was no main
effect of Features, F1 (1, 23) ? 1.31, p ? .264; F2 ? 1, but a
significant main effect of Repetition, F1 (1, 23) ? 6.31, p ? .019;
F2 (1, 158) ? 36.98, p ? .001. In this time window, the interaction
between Repetition and Features was not significant by subjects,
F1 (1, 23) ? 2.29, p ? .144, and only marginally significant by
items, F2 (1, 158) ? 3.04, p ? .083. Following up this trend for
an interaction in the item-based analysis revealed that even though
Repetition modulated ERPs more strongly for words with many
features than for words with few features (0.88 vs. 0.49 ?V),
Repetition effects were significant for both words with many, F2
(1, 79) ? 30.35, p ? .001, and words with few semantic features,
F2 (1, 79) ? 9.49, p ? .003.
Discussion
The present study provides the first evidence that the richness of
semantic representations modulates word learning. Specifically,
we examined influences of the amount of semantic features asso-
ciated with a given word on implicit visual word learning as
assessed by repetition priming. We found enhanced repetition
priming for words with many as compared to few semantic fea-
tures in both lexical decision accuracy and ERPs. Thus, the rich-
ness of semantic representations advances repetition-induced
changes in word processing considered to reflect implicit learning.
Even though the present study has focused on semantic influ-
ences on repetition priming, before discussing our findings in more
detail we would like to note that we did not find effects of the
number of features on performance during the first presentation of
the stimuli as might have been expected based on previous evi-
dence showing facilitated processing of words with richer seman-
tic representations (e.g., Pexman et al., 2002). The absence of
feature effects on performance during the first presentation may
have been caused by task characteristics: Because words and
pseudowords considerably differed in terms of orthographical typ-
icality (e.g., bigram frequency), participants might have performed
the lexical decision task based on low-level statistical regularities,
processing the stimuli rather shallowly. Thus, the activation of
semantic features may not have been necessary to solve the task,
and there may not have been sufficient time to influence perfor-
mance. Such an account is in line with previous evidence showing
stronger influences of semantic features for more difficult tasks
requiring longer processing or access to semantic representations
(Pexman et al., 2002, 2003, 2007). Interestingly, in contrast to
lexical decision performance when processing the stimuli for the
first time, ERP modulations during the first presentation (see
Amsel, 2011; Kounios et al., 2009; Mu ¨ller, Dun ˜abeitia, & Car-
reiras, 2010; Rabovsky, Sommer, & Abdel Rahman, 2011, in
press, for related evidence) and, important for present purposes,
the enhancement of repetition priming were observed despite the
presumed shallow processing.
Why was an influence of semantic richness on repetition prim-
ing seen in ERs and ERPs but not in RTs? We suggest that factors
other than the repetition of specific words also contributed to the
RT decrease from the first to the second presentation. Such factors
may be related to the progression of the task in general, such as
increasing familiarity with the task and the environment, decision-
related stimulus–response mappings, motor execution, and so on.
This is indicated by analyses including the additional factor Pre-
sentation Order (first vs. second half of the stimulus set, within
each presentation of the complete set): A highly significant effect
of Presentation Order was found for RTs, F(1, 23) ? 8.83, p ?
.007, but not for ERs (F ? 1) or N400 amplitudes, F(1, 23) ? 1.70,
p ? .205. As the stimulus set was subdivided into two parts for
both the first and second blocks (with presentation order counter-
balanced across participants; see the Method section, above), such
effects of Presentation Order cannot be due to specific item char-
acteristics. Thus, as the experiment progressed, processing speed
was apparently enhanced not only by word repetition but also by
a number of factors unrelated to the repetition of specific words
and hence also unrelated to the number of features of these words.
Repetition effects on ERs and N400 amplitudes, on the other hand,
were seemingly less influenced by these unspecific factors and
more strongly influenced by the repetition of specific words. This
might have made these measures more sensitive to influences of
the richness of semantic representations on repetition priming.
There are alternative views of the mechanisms underlying rep-
etition priming (Henson, 2003). It is often assumed that repetition
facilitates processing by increasing the accessibility of the repre-
sentations involved in processing the repeated stimuli (Graf &
Mandler, 1984; Ratcliff et al., 1985; Stark & McClelland, 2000;
Wiggs & Martin, 1998), and we would like to suggest that seman-
tic richness enhances this repetition-induced increase in accessi-
bility. Importantly, however, repetition effects may also be due to
the formation of a more direct link between stimuli and responses
(Dobbins, Schnyder, Verfaellie, & Schacter, 2004). Such a link can
allow for bypassing or curtailing processing stages involved in
initial processing. Critically, in the frame of this direct link ac-
count, it might be assumed that the obtained influences of semantic
richness on N400 repetition effects reflect a differential repetition-
induced reduction of semantic involvement. Specifically, upon
initial presentation, enhanced N400 amplitudes for words with
many features might indicate semantic involvement, whereas se-
4
RABOVSKY, SOMMER, AND ABDEL RAHMAN

Page 6

mantic involvement—and hence N400 amplitudes—might be
close to floor level for words with few features. During repetitions,
when responses are faster, semantic involvement might be reduced
for words with many features; in contrast, N400 amplitudes to
words with few features might be already near floor during initial
presentations and therefore not be much further diminished by
repetition.
However, it does not seem that there was a floor issue for words
with few semantic features during the first presentation because,
during the second presentation, words with many features elicited
even smaller N400 amplitudes than words with few features. It does
not seem plausible that during repetition, words with many semantic
features would entail less semantic involvement than words with few
semantic features. Furthermore, the plausibility of attributing the
differential repetition effect to a floor issue with the few feature
stimuli is called into question also by the finding that ERs for both
kinds of words did not differ during the first presentation, the inter-
action between repetition and features being obtained nonetheless.
Thus, influences of semantic richness on repetition priming seem to
reflect semantic influences on repetition-induced increases in acces-
sibility (Graf & Mandler, 1984; Ratcliff et al., 1985; Stark & McClel-
land, 2000; Wiggs & Martin, 1998) rather than differential curtailing
of semantic involvement due to rapid response learning (Dobbins et
al, 2004).
It is also important to note that while repetition priming was
enhanced for words with many features, it was present for words
with few features as well. First, even though ERs did not decrease
with repetition for words with few semantic features, repetition
enhanced response speed for all words in equal measure with a
benefit of 42 ms for words with few and 39 ms for words with
many features. As this considerable RT decrease was accompanied
by constant ERs (speaking against the possibility of a speed–
accuracy tradeoff), it indicates a genuine repetition benefit in
performance for words with few features as well (even though, of
course, diminished as compared to the combination of a decrease
in both RTs and ERs for words with many features). Similarly, in
the ERP data, we observed repetition effects for words with few
features as well, even though attenuated as compared to the well-
pronounced ERP repetition effects for words with many features.
Thus, both performance and ERP data are in line with the notion
that repetition priming is enhanced for words with many features
but occurs for words with few features as well.
As noted in the introduction, such influences of semantic rich-
ness on repetition priming are directly predicted by the connec-
tionist attractor model of lexical-conceptual processing proposed
by Cree et al. (1999). This model relies on an error back-
propagation learning rule that is sensitive to semantic richness; the
number of semantic features associated with a given word is
positively correlated with the overall error used as a basis for
connection adaptations. Although the model has not yet been
applied to semantic influences on learning and plasticity, it pre-
dicts enhanced connection adjustments and hence enhanced im-
plicit learning for words with many semantic features.
The observed influences of semantic richness seem to be par-
ticularly telling in relation to a study by Stark and McClelland
(2000), which raised issues concerning the basic mechanisms
underlying word learning. The authors found larger repetition
priming effects for words than for nonwords and took their find-
ings to challenge most connectionist reading models, namely,
those relying on error-driven learning. In these models, less trained
stimuli, such as nonwords, should produce larger error values,
which should result in enhanced adaptation and thus enhanced
implicit learning. However, the present results suggest that the
predicted enhancement of nonword learning due to the sparseness
of prior training may have been offset by an opposing mechanism,
namely, the enhancement of learning of meaningful words relative
to nonwords, due to the obviously much richer semantic represen-
tations of the former. Crucially, as influences of semantic richness
on implicit learning are in line with an error-based model of
lexical-conceptual processing (Cree et al., 1999), they may resolve
the apparent contradiction between the priming results found by
Stark and McClelland and error-based models of reading.
It remains to be explored how far the reported driving influence
of semantic richness on implicit visual word learning generalizes
to other perceptual modalities (e.g., visual vs. auditory) and do-
mains (e.g., language vs. object recognition). Given the crucial role
of meaning in guiding human behavior, driving influences of
semantic richness on learning may go beyond visual word learn-
ing. However, further research is required to examine this sugges-
tion. The present results seem to mark a promising first step and
should be taken into account when aiming to understand and
describe reading development (Seidenberg & McClelland, 1989).
It is important to note, however, that we do not want to make
statements concerning the learning of new words, that is, adding
new representations to memory. Our focus is on investigating a
learning process that presumably involves the update of internal
models of probabilities of occurrence of the encountered (previ-
ously known) words. Within the frame of the complementary
systems approach to learning and memory (Davis & Gaskell, 2009;
McClelland, McNaughton, & O’Reilly, 1995), the investigated
learning process presumably does not correspond to the fast ac-
quisition of new associations and specific episodes as supported by
the hippocampus; instead, we suggest that it relates to the slow
cortical learning induced by the slight adaptation of the cortical
synapses that are directly involved in processing the corresponding
stimulus. Such subtle adaptations of cortical connections presum-
ably take place for all stimuli, which are processed by the system,
including previously unknown words (Stark & McClelland, 2000);
for unknown words, however, they are supposedly not sufficient
for building stable representations.
In this context, we would also like to note that there is debate
concerning the subdivision of memory into distinct systems and
specifically which variable should be regarded as critical in dis-
tinguishing between the presumed systems (e.g., consciousness,
intention, binding, new associations; see, e.g., Reder, Park, &
Kieffaber, 2009, for review). The present study does not intend to
contribute to this debate. Rather, we discuss our findings as an
instance of implicit learning (and slow cortical instead of fast
hippocampus-driven learning), based on using repetition priming
during lexical decisions as a typical implicit memory task, with no
information about subsequent repetitions given during initial ex-
posure, and the speeded lexical decision task offering little incen-
tive and time for intentional recollection upon repetition. Regard-
ing our findings as an instance of implicit learning is also
consistent with a study by Schott, Richardson-Klavehn, Heinze,
and Du ¨zel (2002) aiming to disentangle ERP correlates of implicit
versus explicit memory encoding, which found that enhanced
N400 amplitudes during initial exposure predicted priming but not
5
SEMANTIC RICHNESS ENHANCES IMPLICIT WORD LEARNING

Page 7

explicit memory during subsequent presentation, nicely fitting
with our finding of larger N400 amplitudes for those words for
which implicit learning was enhanced.
In conclusion, the enhancement of repetition priming for words
with many semantic features seems to reveal a natural account of
the evolvement of semantic richness benefits in visual word pro-
cessing over time (Dun ˜abeitia et al., 2008; Pexman et al., 2002,
2003, 2007). If the amount of semantic features associated with a
given word enhances learning during every single encounter, re-
peated presentations should naturally entail the observed benefit.
In a nutshell, semantically rich words benefit more from each
encounter, as expressed in the biblical parable of the talents: “For
to everyone who has, more shall be given, and he will have an
abundance” (Matthew 25:29, New American Standard Bible).
References
Adelman, J. S., Brown, G. D., & Quesada, J. F. (2006). Contextual
diversity, not word frequency, determines word-naming and lexical
decision times. Psychological Science, 17, 814–823. doi:10.1111/
j.1467-9280.2006.01787.x
Amsel, B. D. (2011). Tracking real-time neural activation of conceptual
knowledge using single-trial event-related potentials. Neuropsychologia,
49, 970–983. doi:10.1016/j.neuropsychologia.2011.01.003
Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects
modeling with crossed random effects for subjects and items. Journal of
Memory and Language, 59, 390–412. doi:10.1016/j.jml.2007.12.005
Balota, D. A., Yap, M. J., Cortese, M. J., Hutchison, K. A., Kessler, B.,
Loftis, B., . . . Treiman, R. (2007). The English Lexicon Project. Behav-
ior Research Methods, 39, 445–459. doi:10.3758/BF03193014
Becker, S., Moscovitch, M., Behrmann, M., & Joordens, S. (1997). Long-
term semantic priming: A computational account and empirical evi-
dence. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 23, 1059–1082. doi:10.1037/0278-7393.23.5.1059
Cree, G. S., McRae, K., & McNorgan, C. (1999). An attractor model of
lexical conceptual processing: Simulating semantic priming. Cognitive
Science, 23, 371–414. doi:10.1207/s15516709cog2303_4
Davis, M. H., & Gaskell, M. G. (2009). A complementary systems account
of word learning: Neural and behavioral evidence. Philosophical Trans-
actions of the Royal Society of London, Series B, 364, 3773–3800.
doi:10.1098/rstb.2009.0111
den Ouden, H. E. M., Friston, K. J., Daw, N. D., McIntosh, A. R., &
Stephan, K. E. (2009). A dual role for prediction error in associative
learning. Cerebral Cortex, 19, 1175–1185. doi:10.1093/cercor/bhn161
Dobbins, I. G., Schnyder, D. M., Verfaellie, M., & Schacter, D. L. (2004,
March 18). Cortical activity reductions during repetition priming can
result from rapid response learning. Nature, 428, 316–319. doi:10.1038/
nature02400
Doyle, M. C., Rugg, M. D., & Wells, T. (1996). A comparison of the
electrophysiological effects of formal and repetition priming. Psycho-
physiology, 33, 132–147. doi:10.1111/j.1469-8986.1996.tb02117.x
Dun ˜abeitia, J. A., A´viles, A., & Carreiras, M. (2008). NoA’s ark: Influence
of the number of associates in visual word recognition. Psychonomic
Bulletin & Review, 15, 1072–1077. doi:10.3758/PBR.15.6.1072
Friston, K. J. (2009). The free-energy principle: A rough guide to the
brain? Trends in Cognitive Sciences, 13, 293–301. doi:10.1016/
j.tics.2009.04.005
Graf, P., & Mandler, G. (1984). Activation makes words more accessible,
but not necessarily more retrievable. Journal of Verbal Learning &
Verbal Behavior, 23, 553–568. doi:10.1016/S0022-5371(84)90346-3
Henson, R. N. (2003). Neuroimaging studies of priming. Progress in
Neurobiology, 70, 53–81. doi:10.1016/S0301-0082(03)00086-8
Hutzler, F., Bergmann, J., Conrad, M., Kronbichler, M., Stenneken, P., &
Jacobs, A. M. (2004). Inhibitory effects of first syllable-frequency in
lexical decision: An event-related potential study. Neuroscience Letters,
372, 179–184. doi:10.1016/j.neulet.2004.07.050
Kounios, J., Green, D. L., Payne, L., Fleck, J. I., Grondin, R., & McRae,
K. (2009). Semantic richness and the activation of concepts in semantic
memory: Evidence from event-related potentials. Brain Research, 1282,
95–102. doi:10.1016/j.brainres.2009.05.092
Kutas, M., & Federmeier, K. D. (2011). Thirty years and counting: Finding
meaning in the N400 component of the event-related brain potential
(ERP). Annual Review of Psychology, 62, 621–647. doi:10.1146/
annurev.psych.093008.131123
Lund, K., & Burgess, C. (1996). Producing high-dimensional semantic
spaces from lexical co-occurrence. Behavior Research Methods, Instru-
ments, & Computers, 28, 203–208. doi:10.3758/BF03204766
McClelland, J. L. (1994). The interaction of nature and nurture in devel-
opment: A parallel distributed processing perspective. In P. Bertelson, P.
Eelen, & G. d’Yewalle (Eds.), International perspectives on psycholog-
ical science: Vol. I. Leading themes (pp. 57–88). Hillsdale, NJ: Erlbaum.
McClelland, J. L., McNaughton, B. L., & O’Reilly, R. C. (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. Psychological Review, 102, 419–457.
doi:10.1037/0033-295X.102.3.419
McLaren, I. (1989). The computational unit as an assembly of neurons: An
implementation of an error correcting learning algorithm. In R. Durbin,
C. Miall, & G. Mitchison (Eds.), The computing neuron (pp. 160–178).
Amsterdam, the Netherlands: Addison-Wesley.
McRae, K., Cree, G. S., Seidenberg, M. S., & McNorgan, C. (2005). Semantic
featureproductionnormsforalargesetoflivingandnonlivingthings.Behavior
Research Methods, 37, 547–559. doi:10.3758/BF03192726
Mu ¨ller, O., Dun ˜abeitia, J. A., & Carreiras, M. (2010). Orthographic and
associative neighborhood density effects: What is shared, what is dif-
ferent? Psychophysiology, 47, 455–466.
Nelson, D. L., McEvoy, C. L., & Schreiber, T. A. (2004). The University
of South Florida free association, rhyme, and word fragment norms.
Behavior Research Methods, Instruments, & Computers, 36, 402–407.
doi:10.3758/BF03195588
Pexman, P. M., Hargreaves, I. S., Edwards, J. D., Henry, L. C., &
Goodyear, B. G. (2007). The neural consequences of semantic richness:
When more comes to mind, less activation is observed. Psychological
Science, 18, 401–406. doi:10.1111/j.1467-9280.2007.01913.x
Pexman, P. M., Holyk, G. G., & Monfils, M.-H. (2003). Number-of-
features effects and semantic processing. Memory & Cognition, 31,
842–855. doi:10.3758/BF03196439
Pexman, P. M., Lupker, S. J., & Hino, Y. (2002). The impact of feedback
semantics on visual word recognition: Number-of-features effects in
lexical decision and naming tasks. Psychonomic Bulletin & Review, 9,
542–549. doi:10.3758/BF03196311
Picton, T. W., Bentin, S., Berg, P., Donchin, E., Hillyard, S. A., Johnson,
R., Jr., . . . Taylor, M. J. (2000). Guidelines for using human event-
related potentials to study cognition: Recording standards and publica-
tion criteria. Psychophysiology, 37, 127–152. doi:10.1111/1469-
8986.3720127
Rabovsky, M., Sommer, W., & Abdel Rahman, R. (2011). The time course
of semantic richness effects in visual word recognition. Manuscript
submitted for publication.
Rabovsky, M., Sommer, W., & Abdel Rahman, R. (in press). Depth of
conceptual knowledge modulates visual processes during word reading.
Journal of Cognitive Neuroscience.
Ratcliff, R., Hockley, W., & McKoon, W. (1985). Components of activa-
tion: Repetition and priming effects in lexical decision and recognition.
Journal of Experimental Psychology: General, 114, 435–450. doi:
10.1037//0096-3445.114.4.435
Reder, L. M., Park, H., & Kieffaber, P. D. (2009). Memory systems do not
6
RABOVSKY, SOMMER, AND ABDEL RAHMAN

Page 8

divide on consciousness: Reinterpreting memory in terms of activation
and binding. Psychological Bulletin, 135, 23–49. doi:10.1037/a0013974
Schacter, D. L., & Graf, P. (1986). Preserved learning in amnesic patients:
Perspectives from research on direct priming. Journal of Clinical and
ExperimentalNeuropsychology,
01688638608405192
Schad, D. J., Nuthmann, A., & Engbert, R. (2010). Eye movements during
reading of randomly shuffled text. Vision Research, 50, 2600–2616.
doi:10.1016/j.visres.2010.08.005
Schott, B., Richardson-Klavehn, A., Heinze, H. J., & Du ¨zel, E. (2002).
Perceptual priming versus explicit memory: Dissociable neural corre-
lates at encoding. Journal of Cognitive Neuroscience, 14, 578–592.
doi:10.1162/08989290260045828
8,
727–743.doi:10.1080/
Schultz, W., & Dickinson, A. (2000). Neuronal coding of prediction
errors. Annual Review of Neuroscience, 23, 473–500. doi:10.1146/
annurev.neuro.23.1.473
Seidenberg, M. S., & McClelland, J. L. (1989). A distributed, develop-
mental model of word recognition and naming. Psychological Review,
96, 523–568. doi:10.1037/0033-295X.96.4.523
Stark, C. E. L., & McClelland, J. L. (2000). Repetition priming of words,
pseudowords, and nonwords. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 26, 945–972. doi:10.1037/0278-
7393.26.4.945
Wiggs, C. L., & Martin, A. (1998). Properties and mechanisms of behav-
ioral priming. Current Opinion in Neurobiology, 8, 227–233. doi:
10.1016/S0959-4388(98)80144-X
Appendix
Stimuli: Words With Few Features, Words With Many Features, and Pseudowords
Few features Pseudowords Many features Pseudowords
ball
barrel
basket
belt
biscuit
boat
bucket
buckle
bull
cabinet
cape
cathedral
cellar
chapel
cherry
chisel
clam
clamp
cloak
coin
cork
crow
dish
dove
drill
emerald
envelope
fence
gown
guitar
harp
hawk
helmet
hook
hornet
hut
inn
mirror
nightgown
lalb
reralb
kasteb
telb
cistibu
atob
cebtuk
celbuk
lulb
nebaict
acep
daltacher
lacrel
hecalp
hyrcer
helcis
almc
plamc
ackol
ocni
ockr
wroc
hisd
odev
rilld
medreal
olpeneve
enfec
nowg
igartu
phra
whak
ethlem
okho
nethor
thu
nin
orrirm
thnongwig
balloon
barn
basement
boots
bra
bread
cabbage
cake
cannon
canoe
carpet
carrot
cat
cheese
coat
couch
cougar
cow
crown
cucumber
cup
deer
desk
dog
eagle
elephant
duck
fawn
flea
football
frog
garlic
gate
goat
grape
hammer
hare
house
lobster
olanolb
ranb
estenamb
ostbo
rab
ardeb
becagab
ecka
nocnan
ocena
trepac
tracor
tac
esehec
acot
hucco
uragoc
cwo
wrocn
mucrebuc
ucp
edre
ksed
ogd
lagee
hetapnel
uckd
wanf
elfa
olfolbta
gorf
ilgrac
egta
otga
repga
mehram
erha
sueho
trelbos
(Appendix continues)
7
SEMANTIC RICHNESS ENHANCES IMPLICIT WORD LEARNING

Page 9

Appendix (continued)
Few featuresPseudowordsMany features Pseudowords
oak
owl
pajamas
pepper
pie
pier
pin
pine
porcupine
pumpkin
raccoon
raspberry
rattle
razor
rice
rock
rocket
sack
salmon
scarf
shell
shield
slippers
snail
spade
spatula
stick
stone
table
taxi
tent
toilet
tomato
toy
truck
umbrella
vest
vine
walnut
whale
worm
koa
wlo
japsaam
prepep
ipe
erip
inp
ipen
nirpocupe
minkupp
norcaco
persybrar
telart
orraz
cire
crok
certok
kacs
lomnsa
crafs
lelsh
ledsih
spespirl
alins
peads
lastaup
scikt
tseon
belat
aixt
tetn
ilotte
mootat
yto
kurct
marblule
tves
eniv
luntaw
helwa
rowm
marble
missile
napkin
necklace
olive
peach
pen
pencil
penguin
pickle
pig
pistol
pony
potato
radio
rake
rat
robe
robin
rooster
ruler
sandals
screwdriver
seagull
sheep
shirt
shoes
sink
sofa
spider
spoon
squirrel
sword
tangerine
tiger
toad
trousers
turtle
typewriter
wasp
whistle
rembla
imselis
knapni
canklece
eivlo
cepha
enp
nelpic
inupneg
cepilk
ipg
spiotl
nyop
ottopa
aidor
eark
tra
breo
brino
orserot
erurl
nassald
werdirsverc
ugalels
hespe
thris
hesos
nisk
foas
drepis
osnop
resluriq
wrods
garneetin
griet
odat
restusor
rettlu
priwetret
psaw
shelwit
Received February 11, 2011
Revision received August 8, 2011
Accepted August 12, 2011 ?
8
RABOVSKY, SOMMER, AND ABDEL RAHMAN