BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 – 1 60
a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m
w w w. e l s e v i e r. c o m / l o c a t e / b r a i n r e s
Research Report
An event-related potential study of the concreteness effect
between Chinese nouns and verbs
Pei-Shu Tsai a,d , Brenda H.-Y. Yu b,d , Chia-Ying Lee c,d , Ovid J.-L. Tzeng c,d ,
Daisy L. Hung b,d , Denise H. Wu b,d,⁎
a
Institute of Neuroscience, National Yang-Ming University, Beitou, Taipei 112, Taiwan
Institute of Cognitive Neuroscience, National Central University, Jhongli, Taoyuan County 320, Taiwan
c
Institute of Linguistics, Academia Sinica, Nankang, Taipei 115, Taiwan
d
Laboratories for Cognitive Neuroscience, National Yang-Ming University, Beitou, Taipei 112, Taiwan
b
A R T I C LE I N FO
AB S T R A C T
Article history:
The effect of concreteness has been heavily studied on nouns. However, there are scant
Accepted 29 October 2008
reports on the effect for verbs. The present research independently manipulated
Available online 14 November 2008
concreteness and word class of Chinese disyllabic words in tasks that required different
depths of semantic processing: a lexical decision task and a semantic relatedness judgment
Keywords:
task. The results replicated the concreteness effect for nouns, indicating that concrete
Grammatical class
nouns elicited larger N400 responses than abstract nouns with a broad distribution over the
Imageability
scalp, irrespective of the task demands. Similar to the findings from English unambiguous
N400
verbs, the concreteness effect for Chinese verbs was also robustly observed from fontal to
ERP
posterior electrodes in both tasks. These results suggest that when Chinese nouns and verbs
Topographic distribution
are typical and unambiguous in both meanings and word classes, the similar topographic
distributions of the N400 components reflect the same underlying cause(s) of the
concreteness effect for these two word classes.
© 2008 Elsevier B.V. All rights reserved.
1.
Introduction
The differences between nouns and verbs have been heavily
examined in the previous research, with abundant evidence
from neuropsychological studies, behavioral experiments,
neurophysiological findings, and developmental observations
between the two word classes. In neuropsychological studies,
a double dissociation between deficits of nouns and verbs has
been reported on aphasic patients at the lexical level (e.g.,
Caramazza and Hillis, 1991; Chen and Bates, 1998; McCarthy
and Warrington, 1985; Miceli et al., 1984; Zingeser and Berndt,
1990) and the sublexical level (Bates et al., 1991). In behavioral
experiments using picture naming, verbs are named slower
than nouns even when possible affecting factors are strictly
controlled (Szekely et al., 2005). In neurophysiological
research with event-related potentials (ERPs), nouns are
associated with more intensified N400 than verbs over
centro-parietal sites, and verbs elicit enhanced positivity at
the left frontal sites (Federmeier et al., 2000; Lee and
Federmeier, 2006). In developmental research, it is generally
accepted that children learn many more nouns, and also
earlier, than verbs in vocabulary acquisition (Gentner, 1978,
1982; Nelson, 1973; but see Choi and Gopnik, 1995; Tardiff,
1996).
⁎ Corresponding author. Institute of Cognitive Neuroscience, National Central University, No.300, Jhongda Rd., Jhongli City, Taoyuan
County 32001, Taiwan. Fax: +886 3 4263502.
E-mail address: [email protected] (D.H. Wu).
0006-8993/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.brainres.2008.10.080
150
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 –16 0
To account for such distinction between nouns and verbs,
different characteristics of word classes have been emphasized.
Phonologically, the stress pattern of verbs in English is less
predictable, with shorter duration in sentences and fewer
syllables (in terms of word length) than nouns, which demands
more resources in language acquisition (Black and Chiat, 2003;
Kelly, 1992). Morphologically, nouns and verbs carry different
types of information. For example, temporal information such
as tense is often carried on verbs, while plurality is often
marked on nouns. Syntactically, the order of nouns and verbs in
a sentence is fixed, and the different positions occupied by
nouns and verbs are not interchangeable. In addition, the
construction of a sentence (the argument structure) is determined by the verb (see Goldberg, 2003; Kemmerer, 2006). Once
the verb is selected, the nouns then complete the sentence
and serve different roles (e.g., the agent and the patient)
according to the semantic constraints designated by the verb.
Semantically, nouns encode entities, while verbs refer to
relations between entities (Gentner, 1981; Langacker, 1987). It
has been proposed that the slower developmental trend of
verbs than nouns is caused by the greater conceptual complexity of verbs as compared to nouns (Gentner, 1982). As a result,
only children at a certain age with enough cognitive ability can
master the concepts represented by verbs. Also from a semantic
viewpoint but different from the assumption of conceptual
complexity, Pulvermüller et al. (1999) attributed the neurophysiological differences between nouns and verbs to different
semantic networks (e.g., motor and visual associations) supporting the distinctions between these two word classes.
Aligned with the discussion on the relationship between word
class and semantics, Lee and Federmeier (2006) pointed out
based on ERP data that the effect of word class could be
intertwined by multiple dimensions such as semantic ambiguity. Black and Chiat (2003) reviewed the evidence from
different perspectives and concluded that an interaction
between all the aspects mentioned above had possibly contributed to the observed dissociation between nouns and verbs.
Among the studies examining the contribution of semantic
factors to the word class effect, most research has concentrated
on concreteness/imageability between nouns and verbs (e.g.,
Bird et al., 2001, 2003; Lee and Federmeier, 2008; Zhang et al.,
2006). Concrete words are processed faster than abstract words
when behavioral measures are used (Kounios and Holcomb,
1994; West and Holcomb, 2000). Physiologically, a larger N400
component is elicited by concrete than by abstract words
(Kounios and Holcomb, 1994; Lee and Federmeier, 2008; West
and Holcomb, 2000). Such effect of concreteness can be
explained by a dual-coding theory (Paivio, 1986), which suggests
that the processing of concrete words involves two systems: a
verbal semantic system and an imagistic system, whereas the
processing of abstract words lacks the assistance from the
imagistic system (but see Schwanenflugel et al., 1988 for a
different view). Further support for the dual-coding theory from
ERP (Kounios and Holcomb 1994; Swaab et al., 2002) and fMRI
studies (Giesbrecht et al., 2004) illustrate that the neural signals
observed in the frontal sites is associated with imageability
effects, whereas the effects observed in the central–posterior
sites are associated with processing in the verbal system.
Based on the finding that nouns tended to be rated as more
imageable than verbs when they were mixed in a questionnaire,
Bird et al. (2001) proposed that it was the semantic difference on
the concreteness/imageability dimension that gave rise to the
word class effects. When imageability was controlled, the
dissociation between nouns and verbs was eliminated on
aphasic patients who were previously diagnosed with verb
deficits (Bird et al., 2003). However, recent findings have
demonstrated the difference between nouns and verbs even
when the concreteness/imageability is taken into account. For
example, Berndt et al. (2002) reported a double dissociation
between word class and imageability effects on patients. Similarly, Bedny and Thompson-Schill (2006) investigated the
effects of word class and imageability independently in a
semantic judgment task with fMRI. They found that the left
superior temporal gyrus showed greater activation for verbs
than for nouns after the imageability was matched across the
two word classes. These results support the idea that the noun–
verb dissociation and the semantic aspect (concreteness/imageability) of such distinction can be dissociated from each other.
Zhang et al. (2006) provided indirect evidence for the notion
that nouns and verbs were processed differently when the
concreteness between the two word classes was considered.
Specifically, they employed Chinese stimuli in a lexical
decision task in which they compared the concreteness effects
for nouns and verbs separately. Their results indicated that
concrete nouns were associated with a large N400 component
that was broadly distributed over the scalp, including the
frontal, central, and posterior sites. On the other hand, the
concreteness effects for verbs were restricted in the left
centro-parietal region. One possible explanation for these
findings, as Zhang et al. have noted, is that the contrast of
concreteness might be weaker in verbs than in nouns.
Nevertheless, as the authors have stated, the question
remains open in terms of whether the processing of concreteness is different between the two word classes.
Although Zhang et al. (2006) showed different patterns of
the concreteness effect between nouns and verbs, Lee and
Federmeier (2008), using English stimuli with a relatedness
judgment task, offered a different explanation for the semantic distinction between word classes. Lee and Federmeier
compared the concreteness effect, which was reflected in the
N400 component of ERP, for English nouns and verbs. They
found concreteness effects in the frontal and centro-parietal
electrodes for unambiguous verbs, whereas for verbs that
were syntactically and semantically ambiguous, the frontal
effects were not observed. In other words, for ambiguous
verbs, the concreteness effects were observed only at central–
posterior sites. Lee and Federmeier interpreted the inconsistent findings between Chinese and English verbs as a result
of different linguistic characteristics between languages. They
postulated that because Chinese verbs were potentially more
class-ambiguous than English verbs, no frontal effect was
observed for Chinese verbs as it was not observed for English
class/semantic–ambiguous verbs.
Because Zhang et al. (2006) did not directly compare the
concreteness effect between Chinese nouns and verbs, the
question of whether the concreteness of the two word classes
is processed differently remains unanswered. If the activation
of the two word classes involves distinct semantic networks,
the topographic distribution of the N400 patterns should be
different between nouns and verbs. If the concreteness effect in
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 – 1 60
the two grammatical categories results from a general source,
what Zhang et al. have observed could be a reflection of
syntactic and/or semantic ambiguity of the stimuli they have
selected (as the pattern observed for ambiguous verbs in Lee
and Federmeier, 2008). Moreover, even if the meaning and word
class of the stimuli employed by Zhang et al. are unambiguous
at the lexical level, the sublexical structure of the experimental
materials is also an important factor that needs to be
considered. As Hsu et al. (2004) pointed out, the syntactic
class of a Chinese disyllabic compound can be different from
the syntactic categories of the constituent characters. Specifically, approximately 70% of the words in Chinese vocabulary
are disyllabic compounds (Zhou and Marslen-Wilson, 1995; for
a review on Chinese compound processing, see Myers, 2006).
These disyllabic compounds are formed by two characters.
Each character represents one syllable, and for most of the
time, each character also represents one morpheme with a
word class. The independent combination of the word class of
the first and the second characters in a disyllabic word results
in four types of nominal compounds: [NN] (e.g., qingyi ‘friendship’), [NV] (e.g., qiupai ‘racket’), [VN] (e.g., shuidai ‘sleeping
bag’), and [VV] (e.g., zhuchi ‘priest of a temple’). Similarly, there
are also four variants of verbal compounds, including [NN] (e.g.,
wuse ‘to seek’), [NV] (e.g., fenshua ‘to whitewash’), [VN] (e.g.,
wajiao ‘to recruit talents from other corporations’), and [VV]
(e.g., kuqi ‘to cry’). Hsu et al. reported a syntactic category
typicality effect in a categorization task, which indicated that
the nominal NN compounds and the verbal VV compounds
were processed faster than all the other subtypes of compounds. Because the report of Zhang et al. (2006) did not
provide us clues as to what types of disyllabic compounds were
used in the experiment, it was possible that they did not use the
most typical nouns and verbs as stimuli. As a result, the
different patterns observed in their study could have reflected
the effect of word-class ambiguity/typicality rather than a
noun–verb distinction.
In addition to the characteristics of the stimuli that are
discussed above, task demand is also a potential factor
influencing the topographic distribution of the concreteness
effect reported in the previous studies. It has been demonstrated that the effect of concreteness is pronounced in the
frontal sites in studies that explicitly require imagery processing (West and Holcomb, 2000) and semantic judgment (Lee
and Federmeier, 2008; West and Holcomb, 2000). On the other
hand, studies that do not find significant concreteness effects
in the frontal electrodes for verbs adopt a lexical decision task
(Zhang et al., 2006). The present study employed a lexical
decision and a semantic relatedness judgment task in
Experiments 1 and 2, respectively, in order to determine
whether the effects of word class and concreteness are
modulated by different task demands.
2.
Results
2.1.
Experiment 1 — lexical decision
2.1.1.
Behavioral results
All participants performed the task with high accuracy, with a
mean of 97% ranging from 92% to 100%. A paired t test of the
151
error rate showed that there was no accuracy difference
between the word and nonword conditions (words: 97%,
nonwords: 97%, t(23) = .250, p = .805). The reaction times (RTs)
for the word condition (mean = 657 ms) were faster than those
for the nonword condition (mean = 728 ms), t(23) = 7.877,
p b .001. Separate repeated two-way analyses of variance
(ANOVAs) were applied to compare the accuracy and RTs of
the four word conditions testing the variables of concreteness
(abstract/concrete) and word class (noun/verb). The analysis of
accuracy showed that the main effects of word class and
concreteness did not reach significance (both ps N .116). The
interaction between concreteness and word class was not
significant, either (F b 1). These results were likely due to the
overall high accuracy in the lexical decision task. The ANOVA of
RTs showed a significant main effect of word class (F(1,23)=20.878,
p b .001), with faster RTs for nouns than those for verbs. This
pattern was consistent with those reported in the literature
comparing noun/verb processing in lexical decision tasks (e.g.,
Sereno and Jongman, 1997). The main effect of concreteness
and the interaction between word class and concreteness
were not significant (both ps N .549).
2.1.2.
N400
The grand mean ERPs elicited by abstract and concrete words
are shown in Fig. 1, which also illustrates the approximate
locations of the electrode sites. The time window of N400 was
set at 300–500 ms post stimulus onset based on moving
window calculations. A repeated four-way ANOVA was performed with the variables of concreteness (concrete/abstract),
word class (noun/verb), hemisphere (left/midline/right), and
caudality (frontal/central/posterior). The result indicated a
significant main effect of concreteness (F(1,23) = 14.858,
p b .005), showing that the amplitude of N400 for concrete
words was more negative than that for abstract words. The
main effect of word class did not reach significance, nor did
the interaction between word class and concreteness (Fs b 1).
The interaction between word class and hemisphere was
significant (F(2, 46) = 5.263, p b .05), indicating that the N400
was marginally weaker for verbs than for nouns in the
right hemisphere (RH) (F(1, 23) = 3.730, p = .066) (Electrode
locations are illustrated in Fig. 1). The interaction between concreteness and hemisphere was also significant
(F(2, 46) = 4.086, p b .05). Post hoc comparisons showed that
the differences between concrete and abstract words were
consistently significant across the two hemispheres and the
midline, with the strongest effects in the midline and the
left hemisphere (LH) (left, midline, right, t(24) = 3.6, 4.1, 3.0,
respectively, ps b .01). The effect of caudality, however, did
not interact with concreteness and word class (F(2, 46) = .173,
p = .74).
When we controlled the sublexical composition of the
materials in this experiment, the main effect of word class was
diminished compared with previous findings in Zhang et al.
(2006). However, in order to directly compare our results with
previous Chinese findings on N400, we followed the comparisons made in Zhang et al. by conducting two separate
repeated measures ANOVAs at the midline and the lateral
sites for each of the two word classes. The effects for nouns
and verbs are presented in Figs. 2 and 3, respectively. The
analyses at the midline considered a factor of concreteness
152
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 –16 0
Fig. 1 – Approximate locations of electrode sites.
and a factor of electrode site with six levels. The analyses at
the lateral sites included three factors: concreteness (concrete/abstract), hemisphere (left/right), and electrode site with
17 levels. For all analyses, robust effects of concreteness were
observed for both nouns and verbs at both the midline and
lateral sites (nouns: midline: F(1,23) = 13,413 p b .005; lateral:
F(1,23) = 11.441, p b .005; verbs: midline: F(1,23) = 5.012, p b .05;
lateral: F(1,23) = 5.298, p b .05). The interactions between concreteness and electrode site for the midline and lateral site
analyses were not significant (ps N .05).
We also followed the analyses in Lee and Federmeier (2008)
by dividing the electrode sites into frontal and centro-parietal
areas and examining the concreteness and word class effects
in these areas separately. For comparisons in the frontal area,
there were three main factors: concreteness (concrete/
abstract), word class (noun/verb), and electrode site with 14
levels (represented by electrodes labeled F1–F6, FC1–FC6, and
Fz, FCz). A repeated measures ANOVA indicated a main effect
of concreteness (F(1, 23) = 11.842, p b .005), and this effect did
not interact with word class (F b 1). The main effect of word
class was not significant (F b 1). The interaction among the
three variables (concreteness, word class, and electrode site)
was not significant, either (F b 1).
As for the effects in the central–posterior region, the
analysis was performed by considering factors of concreteness, word class, and electrode site with 26 levels (including
electrodes of C1–C6, CP1–CP6, P1–P6, PO3–PO6, and Cz-POz).
The results revealed similar patterns observed from the
frontal region. In other words, there was a significant main
effect of concreteness (F(1, 23) = 15.073, p b .005), but there was
no significant effect of word class (F b 1), as well as no
interaction between concreteness and word class (F b 1).
There was a significant interaction between electrode site
and concreteness (F(25, 575) = 3.152, p b .05), indicating that
concrete words were associated with larger amplitude than
abstract words in the LH and the midline (the comparisons
were significant for every electrode site on the LH and the
midline, ts N 2, ps b .05). Paired t tests showed that the concreteness effect was present in a relatively smaller distribution in
the RH (significant at electrodes C4, C6, CP2, CP4, CP6, and P4, t
(23) = 2.902, 2.116, 3.291, 2.443, 2.882, and 3.308, respectively,
ps b .05). There was also an interaction between electrode site
and word class (F(25, 575) = 2.853, p b .05), showing that nouns
elicited larger negativity than verbs did in the right posterior
area (at electrodes CP4, CP6, P6, PO4, and PO6, t(23) = 2.270,
3.743, 2.683, 3.969, and 2.268, respectively, ps b .05).
2.1.3.
Summary and discussion
Consistent with previous neurophysiological findings that
only select nouns as stimuli (e.g., Holcomb et al., 1999; Kounios
and Holcomb, 1994; West and Holcomb, 2000), the concrete
nouns were associated with larger N400 amplitude than
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 – 1 60
153
Fig. 2 – Grand average ERPs at nine representative electrode sites to concrete nouns (dashed line) and abstract nouns (solid line)
in a lexical decision task. The approximate locations of these sites can be seen in Fig. 1. The concreteness effects are
distributed at all electrode sites.
abstract nouns in Experiment 1. Zhang et al. (2006) reported a
significant concreteness effect for nouns with a widespread
distribution across the scalp. In contrast, their concreteness
effect for verbs was only marginally significant (p = .074) during
300–400 ms over the centro-parietal electrode sites. Our results
on nouns were consistent with the findings of Zhang et al.
However, different from their findings of a centro-parietal
distribution of the concreteness effect for verbs, our results
showed that the concreteness effects for verbs were significant over the whole head like the effects for nouns were. Our
data indicate that the N400 reflects degree of concreteness
(with larger amplitude elicited by higher concreteness)
distributed over the whole scalp for both nouns and verbs.
The neurophysiological patterns between concrete and
abstract words in the frontal and posterior electrodes could
be accounted for by the previous hypotheses of separate sets of
semantic features associated with concreteness processing
(Kounios and Holcomb, 1994; Swaab et al., 2002). As the
literature has suggested, the effect in the posterior sites corresponds to the typical verbal semantic processing, whereas the
effect in the frontal electrodes reflects the processing in the
imagistic system. The assumption of such hypothesis is that
the topographical differences of the effects observed reflect the
distribution of the different semantic networks involved.
However, compared with previous findings that report different topographical distribution of the concreteness effect
between nouns and verbs (Zhang et al., 2006), our findings in
Experiment 1 do not show support for this distinction between
the two word classes. If semantics is one of the important
factors that bring about the topographical distinction between
nouns and verbs, the interaction between word classes and
topographic distribution should be more easily detected in a
semantic task. In order to further verify the differences found
by Zhang et al. between nouns and verbs, we performed a
semantic relatedness judgment task in Experiment 2 that
directly instructed participants to process the semantic
information of the stimuli. If the distinction between nouns
and verbs revealed in Zhang et al. is driven by semantic
features associated with the two word classes, we expect this
effect to be amplified by the semantic task in Experiment 2.
2.2.
Experiment 2 — semantic relatedness judgment
2.2.1.
Behavioral results
Since the participants were instructed to make responses to
the second word, rather than to the first critical stimulus, in a
word pair, no behavioral data were available to assess the
effects of concreteness and word class. The overall accuracy of
154
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 –16 0
Fig. 3 – Grand average ERPs at nine representative electrode sites to concrete verbs (dashed line) and abstract verbs (solid line)
in a lexical decision task. The concreteness effects are reflected in N400 (300–500 ms), with concrete verbs associated with
more negative potentials than abstract verbs.
the participants' responses to the second word was almost
ceiling, with a mean 95% ranging from 88% to 99%. The high
accuracy assured that the participants paid attention to the
stimuli during the experiment.
2.2.2.
N400
The time window for N400 was set from 300 to 500 ms post
stimulus onset. The results of a four-way ANOVA with repeated measures among variables of concreteness (concrete/
abstract), word class (noun/verb), hemisphere (left/midline/
right), and caudality (frontal/central/posterior) revealed that
the main effect of concreteness was significant (F(1,23)
= 13.153, p b .005). Specifically, the N400 for concrete words
was more negative-going than that for abstract words (concrete: −.17 μV, abstract: .82 μV). Although the main effect of
word class did not reach significance (F(1,23) = 1.690, p = .206),
the interaction between this factor and hemisphere reached
significance (F(2,46) = 6.345, p b .05), indicating a trend for
nouns to elicit more negativity and thus larger amplitude
than verbs in the LH (nouns: .10 μV, verbs: .78 μV, t(23) = −2.199,
p b .05). The effect of caudality was not modulated by the
two experimental factors, namely, concreteness and word
class (F(2, 46) = 1.559, p = .23).
Similar to the analyses in Experiment 1, we separated the
comparisons for nouns and verbs respectively as shown in
Figs. 4 and 5. Two separate repeated-measures ANOVAs were
conducted at the midline and the lateral sites in order to
compare the effect of laterality. Also similar to the findings in
Experiment 1, we found concreteness effects in both the
midline and the lateral sites for both nouns and verbs (nouns:
midline: F(1,23) = 7.528, p b .05; lateral: F(1,23) = 10.541, p b .005;
verbs: midline: F(1,23) = 3.553, p = .072; lateral: F(1,23) = 5.902,
p b .05). The interactions between concreteness and electrode
locations were not significant (ps N .05).
Following Lee and Federmeier (2008), we also made planned
comparisons for the concreteness effects in the fontal and
central–posterior areas in order to test the caudality effect.
Consistent with our previous findings in Experiment 1 using
LDT, we found a main effect of concreteness in the frontal area
(F(1, 23) = 11.935, p b .005). Neither the main effect of word class
nor the interaction between this factor and concreteness was
significant (word class main effect, F(1, 23) = 2.444, p = .132;
word class by concreteness interaction, F(1, 23) = 1.468, p = .238).
There was an interaction between word class and electrode
site (F(13, 299) = 3.238, p b .05), indicating that nouns were more
negative than verbs in the left frontal channels (at F1, F3, F5,
and FC1, t(23) = −2.447, − 2.383, −2.533, and −2.391, respectively,
ps b .05).
With regard to the analyses for the central–posterior sites,
there was also a significant main effect of concreteness (F(1,
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 – 1 60
155
Fig. 4 – Grand average ERPs at nine representative electrode sites to concrete nouns (dashed line) and abstract nouns (solid line)
in a semantic relatedness judgment task. The concreteness effects are distributed over all electrode sites.
23) = 15.404, p b .005). Consistent with the patterns found over
the frontal sites, there was no effect of word class (F(1, 23) =
1.480, p = .236) and no effect of word class by concreteness
interaction (F b 1). In addition, there was a word class by
electrode site interaction (F(25, 575) = 3.334, p b .05). Paired t test
showed that nouns elicited larger negativity than verbs did
primarily in the left central area of the brain (at C1, C3, C5, CP5,
and P1 electrode, t(23) = − 2.645, −2.758, −2.388, −2.623, and
−2.127, respectively, ps b .05).
2.2.3.
Summary and discussion
Compared with the findings in Experiment 1 which employed
the lexical decision task, the effect of concreteness was also
pronounced on the N400 component in Experiment 2 where
the task was more semantic-demanding. Moreover, we
replicated the effect of concreteness for nouns as it was
frequently reported in the previous literature (Kounios and
Holcomb, 1994; Kutas and Federmeier, 2000; West and
Holcomb, 2000; Zhang et al., 2006). Extending from the findings,
our study on verbs also indicated that concrete verbs elicited
larger amplitude than abstract verbs in the frontal, central, and
posterior electrodes. This pattern and distribution of effects
were similar to those reported for nouns.
3.
General discussion
In the previous studies, Zhang et al. (2006) employed
concrete and abstract Chinese materials in a lexical decision
task. Their data showed distinct topographic activation
patterns between nouns and verbs. They observed that for
nouns, the topographic distribution of the concreteness
effect spread over the whole head, whereas for verbs, the
activation was limited to the left centro-parietal region. Lee
and Federmeier (2008), using English materials in a semantic
relatedness judgment task, found that class-unambiguous
verbs elicited concreteness effects distributed all over the
scalp, but verbs that were semantically and syntactically
ambiguous did not produce a frontal distribution of the
concreteness effect. The present study selected typical and
unambiguous Chinese nouns and verbs that were controlled
for sublexical composition of word class. The effects of
concreteness between nouns and verbs were compared in a
lexical decision and a semantic relatedness judgment task.
The results showed a robust effect of concreteness for both
word classes in the frontal and centro-parietal sites. These
patterns were present in the two experiments, indicating
156
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 –16 0
Fig. 5 – Grand average ERPs at nine representative electrode sites to concrete verbs (dashed line) and abstract verbs (solid line)
in semantic relatedness judgment. The concreteness effects are reflected in N400 (300–500 ms), with concrete verbs
associated with more negative potentials than abstract verbs.
that the processing of concreteness was independent of task
demands.
Across the two experiments in the current study, our
results are compatible with the neuropsychological findings
from patient data in Bird et al. (2003). In their research, the
authors reported cases who exhibited verb deficits in naming.
However, when Bird et al. divided their stimuli into subsets of
high- and low-imagery groups, the patients' performances
appeared to be a disadvantage in producing less imageable
words, and the performances for verbs were not specifically
worse than those for nouns. Similarly, in our study, when the
concreteness of the stimuli was manipulated, the effect of
word class was absent, and the interaction between these two
factors was not observed. It should be noted that, however, the
current results are not necessarily incompatible with the word
class effects independent of the concreteness effects that were
reported in previous studies (e.g., Bedny and ThompsonSchill, 2006; Berndt et al., 2002). It is still possible that the
noun/verb distinction could be observed on another ERP component and/or in another task. Further research is clearly
needed to shed light on this issue.
In conclusion, the current study re-examined the concreteness effects and extended previous research (e.g., Holcomb
et al., 1999; Kounios and Holcomb, 1994; West and Holcomb,
2000; Zhang et al., 2006) with direct and independent manipulation of concreteness and word class of Chinese stimuli. The
concrete words elicited larger amplitude than the abstract
words in the frontal and centro-parietal sites in both lexical
decision and semantic relatedness judgment that demanded
different depths of semantic processing. Moreover, such concreteness effect was robust regardless of word class, and no
interaction was found between concreteness and word class.
These results are consistent with the findings from English
materials (Lee and Federmeier, 2008) and highlight the influence of the characteristics of stimuli in making crosslinguistic comparisons. The current findings also support the
idea that the concreteness effects of nouns and verbs are
generated from the same sources in the brain.
4.
Experimental procedures
4.1.
Pre-experiment rating for concreteness
In order to manipulate the concreteness of Chinese nouns and
verbs in the present study, we obtained the subjective rating of
157
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 – 1 60
concreteness from a group of participants in a pretest. We
prepared 588 disyllabic Chinese compound words, with 293
nouns and 295 verbs. Twenty participants were instructed to
rate the concreteness of each word on a seven-point scale (1:
most abstract; 7: most concrete). We adopted the definition of
concreteness suggested by Zhang et al. (2003): a concept that
can be directly perceived, or with image that can be touched or
felt. This definition was also adopted by previous studies
(Bedny and Thompson-Schill, 2006; Zhang et al., 2006). The
mean concreteness rating for each item was calculated. Words
with mean concreteness scores b4 were classified as abstract;
words with mean scores N5 were classified as concrete. From
the rating, we selected four types of words, each with 35 words
in the category: concrete nouns, abstract nouns, concrete
verbs, and abstract verbs.
4.2.
Experiment 1
4.2.1. Participants
Twenty-eight undergraduate and graduate students (16 males
and 12 females, mean age 22, range 19–26) of National Central
University were paid for participation. All of these participants
were native speakers of Mandarin Chinese with no brain
injury or neurological disease. They were all right-handed
according to self-report and the Edinburgh Handedness
Inventory (Oldfield, 1971), and had normal or correctedto-normal vision.
4.2.2. Materials
Based on the ratings from the pretest, 140 disyllabic Chinese
words were selected. The stimuli consisted of 35 of each of the
four word types: (1) concrete nouns (e.g., plum blossom), (2)
abstract nouns (e.g., friendship), (3) concrete verbs (e.g., cry),
and (4) abstract verbs (e.g., forget). Furthermore, since all the
stimuli were composed of two Chinese characters, we also
controlled the sublexical status of the characters by choosing
noun–noun combination for nouns and verb–verb combination for verbs to result in the most typical nouns and verbs in
Chinese (Hsu et al., 2004). Visual complexity was controlled by
equating the number of the strokes of the first and the second
characters across the four types of words. Stimulus examples
and the descriptive statistics for the four conditions are provided in Table 1.
The frequency of the words employed in the present
experiment was obtained from the Academia Sinica Balanced Corpus of Modern Chinese (http://www.sinica.edu.tw/
SinicaCorpus), which comprised over five million Chinese
words collected from newspapers and other written texts, as
well as some transcriptions of discourse. A two-way analysis of variance (ANOVA) applied to log frequency with two
levels of concreteness (concrete/abstract) and two levels of
word class (noun/verb) indicated that there were significant
differences between nouns and verbs (noun 2.14, verb 1.86;
F(1,136) = 5.337, p b .05) and between concrete and abstract
words (concrete 1.75, abstract 2.23; F(1,136) = 15.375, p b .001).
The interaction was not significant (F b 1). Gernsbacher (1984)
suggested that familiarity is a better predictor of word
performance than printed word frequency. Following this
suggestion, the familiarity of each stimulus was obtained by
instructing the participants to consider their daily language
usage in all aspects (reading, listening, speaking, and writing)
and to indicate their overall familiarity to each stimulus on a
seven-point scale (1: least familiar; 7: most familiar). Twentyeight participants were given no time limit in the rating task.
There was no difference in familiarity between nouns and
verbs (noun: 5.72, verb: 5.62; F(1,136) = 1.619, p = .205), or
between concrete and abstract words (concrete: 5.69, abstract:
5.64; F b 1). The interaction was not significant (F b 1).
One hundred and forty nonwords were created by repairing the characters in the word condition, irrespective of
maintaining the same position of each character, to form
meaningless combinations. These 140 nonwords were
included in the lexical decision task to equate the numbers
of “yes” and “no” trials.
4.2.3. Procedure
A lexical decision task was employed. A practice block of 20
trials preceded the experimental session, and the experimental session was divided into seven blocks. Each block contained 42 trials. The first two trials in each block were fillers, in
case the participants were not fully prepared at the beginning
of each block. The procedure for each trial was as the
following: First, a fixation cross was presented centrally on
the screen for 400 ms. After the fixation cross disappeared for
100 ms, a target word or a nonword appeared horizontally at
the center of the screen for 1500 ms. The participants were
Table 1 – Stimulus materials
Trials
Sub-class
Example
Concreteness
Log frequency
Familiarity
Char1 stroke
Char2 stroke
Concrete noun
Abstract noun
Concrete verb
Abstract verb
35
NN
Meihua (plum blossom)
6.12 (.75)
1.75 (.62)
5.76 (.50)
10 (5)
12 (4)
35
NN
Qingyi (friendship)
3.17 (.42)
2.30 (.64)
5.68 (.54)
11 (5)
11 (4)
35
VV
Kuqi (cry)
5.43 (.40)
1.47 (.54)
5.63 (.35)
12 (4)
12 (4)
35
VV
Wangji (forget)
3.49 (.37)
1.87 (.77)
5.60 (.54)
12 (5)
12 (4)
Char = character.
An example is provided for each type of target words. Standard deviation is provided in parentheses. The stimuli are manipulated for concreteness
and word class. The sub-class, familiarity and visual complexity (character stroke) are controlled.
158
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 –16 0
instructed to judge whether the target was a word or not as
quickly and accurately as possible within 1500 ms. If the target
was a word, they should press the key “l”; if not, they should
press the key “s”. Since eye movements would interfere with
the EEG signals which we recorded, the participants were
required not to make any horizontal or vertical eye movements during the time they silently viewed the target. After
the participant made a manual response or the time limit was
reached, a letter “B” appeared centrally on the screen for
1500 ms, a period during which the participant was allowed to
blink. Each block was approximately 3–4 min, and the entire
experiment lasted for 20–30 min.
4.2.4. EEG recording parameters
We used the SynAmp2 system (NeuroScan) for the ERP recording. The changes in the evoked potentials were recorded at 66
different electrode sites. The impedance of each electrode was
kept below 5 kΩ. The sampling (A/D) rate was 250 Hz. The gain
on the amplifier, the resolution of the analog-to-digital
conversion, was 1000 Hz. A low pass filter was set at 70 Hz
and a high pass filter at .05 Hz. The reference electrodes were
placed at the right and left mastoids (M1 and M2). The ground
electrode was at the site between FPZ and FZ. HEOGs placed at
the outer canthi of the eyes and VEOGs placed above and
below the left eye were bipolar channels that were used to
record ocular movements.
4.2.5. Data analysis
All the trials with incorrect responses and extraneous RTs (i.e.,
outside the range of mean RT ± 2.5 standard deviation) were
excluded from further statistical analyses. This resulted in the
deletion of 2.6% of the data.
The remaining trials were further analyzed with the software
Scan 4.3.2. ERPs were computed time-locked to the stimulus
onset. A low-pass filter was set at 20 Hz. All scalp electrodes
except VEOGs and HEOGs were then re-referenced to the average
of the right (M1) and the left (M2) mastoids. The waveforms of all
epochs were corrected relative to the baseline which was the
mean amplitude of the 100 ms period before the target onset.
Trials with artifacts of ocular movements (i.e., with the
amplitudes of VEOG and HEOG exceeding ± 100 μV) were
excluded. Trials that contained other noise, such as muscular
or environmental artifacts, were rejected at ±60 μV in the other
channels. Data which preserved more than 16 trials in each
condition were preserved for further statistical analyses. Four
participants failed to reach this criterion and were excluded
from further analyses. Trial loss averaged 7.0%.
In the statistical analyses of the ERP data, in addition to
the factors of concreteness and word class, the locations of
electrodes in different hemispheres and caudality were also
included. The three levels of hemisphere included the
electrodes on the left side (the electrodes labeled with an
odd number), the midline (the electrodes labeled with Z),
and the electrodes on the right side (those labeled with an
even number). For the three levels of caudality, the electrodes on the F and FC lines were chosen as the frontal sites,
the electrodes on the C and CP lines as the central sites, and
the electrodes on the P and PO lines as the posterior sites.
That is, 40 channels were separated into 9 areas for
analyses. For all analyses using repeated measures, a
Greenhouse–Geisser correction was applied (Greenhouse
and Geisser, 1959), and the Greenhouse–Geisser corrected p
values were reported.
4.3.
Experiment 2
4.3.1. Participants
Twenty-six students (11 males and 15 females, mean age 22,
range 19–29) of National Central University were paid to
participate in this experiment. These participants met the
same criteria used in Experiment 1. None of these participants
were in the pretests or Experiment 1.
4.3.2. Materials
There were 280 word pairs in this experiment. Half of them
were semantically related pairs; the other half was semantically unrelated pairs. Among the semantically related pairs,
the same 140 words used in Experiment 1 were treated as the
first word in each word pair. Then, 140 disyllabic words were
added to form the second word in the related word pair.
Another 280 words were fillers that formed the 140 semantically unrelated pairs.
The semantic relatedness of a word pair was determined by a subjective rating collected from 10 participants who were not in any of the other tests in the
present study. The participants were required to rate the
semantic relatedness between the first and second words
on a seven-point scale (1: least related; 7: most related).
Experimental stimuli were selected based on the rating
scores to form related pairs (mean relatedness 5.85, range
3.8–6.8) and unrelated pairs (mean relatedness 1.35, range
1–2.9).
4.3.3. Procedure
Twenty practice trials were given to the participants prior to
the experiment. The whole experiment was divided into
seven blocks. Each block contained 40 trials. A fixation cross
was presented centrally on the screen for 300 ms. A blank of
200 ms followed the fixation cross. The first word colored in
white then appeared horizontally at the center of the screen
(visual angle = 1.2°) for 1500 ms against a black background.
The actual size of each character was 1.2 cm high and 1.2 cm
wide. During the 1500 ms of the stimulus presentation, the
participants were instructed to think about the meaning of
the target word. Because eye movements would interfere
with the EEG signals, the participants were required not to
move their eyes during this period. Five hundred ms after
the target word disappeared, the second word appeared. The
participants were asked to judge whether the second word
was semantically related to the first word by pressing the
key “l” or “s” to indicate a “yes” or “no” response, respectively. The second word lasted on the screen for at most
2000 ms or disappeared when the participant made a
manual response. The screen remained blank for 500 ms,
and then the letter “B” appeared for 1500 ms, during which
the participants were allowed to make eye movements.
Each block lasted approximately 4–5 min. The participants
were allowed to rest between blocks and then self initiated
the next block. The whole experiment generally lasted for
30–35 min.
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 – 1 60
4.3.4. EEG recording parameters
The apparatus and the EEG procedures were identical to those
in Experiment 1.
4.3.5. Data analysis
All the trials with incorrect responses and extraneous RTs (i.e.,
outside the range of mean RT ± 2.5 standard deviation) were
excluded from further statistical analyses. This resulted in the
deletion of 4.8% of the data.
The pre-processing of the ERP data was similar to that in
Experiment 1. After performing artifact rejections, only the
data which preserved more than 16 trials in each condition
were carried on to further statistical analyses. Two participants failed to reach this criterion and were excluded. Trial
loss was 18%.
Acknowledgments
The first and second authors have an equal contribution to the
present research. The authors would like to thank the
anonymous reviewers for comments, Dr. Shih-kuen Cheng
and Chun-Hsien Hsu for suggestions on ERP analyses, and
Acer Chang for advice on methods for improving quality of the
figures. The research was supported by Academia Sinica (AS94-TP-C06).
REFERENCES
Bates, E., Chen, S., Tzeng, O.J.-L., Li, P., Opei, M., 1991. The
noun–verb problem in Chinese aphasia. Brain Lang. 41 (2),
203–233.
Bedny, M., Thompson-Schill, S.L., 2006. Neuroanatomically
separable effects of imageability and grammatical class during
single-word comprehension. Brain Lang. 98, 127–139.
Berndt, R.S., Haendiges, A.N., Burton, M.W., Mitchum, C.C., 2002.
Grammatical class and imageability in aphasic word production:
their effects are independent. J. Neurolinguist. 15, 353–371.
Bird, H., Franklin, S., Howard, D., 2001. Age of acquisition and
imageability ratings for a large set of words, including verbs
and function words. Behav. Res. Methods Instrum. Comput. 33
(1), 73–79.
Bird, H., Howard, D., Franklin, S., 2003. Verbs and nouns: the
importance of being imageable. J. Neurolinguist. 16, 113–149.
Black, M., Chiat, S., 2003. Noun–verb dissociations: a multi-faceted
phenomenon. J. Neurolinguist. 16, 231–250.
Caramazza, A., Hillis, A.E., 1991. Lexical organization of nouns and
verbs in the brain. Nature 349 (6312), 788–790.
Chen, S., Bates, E., 1998. The dissociation between nouns and verbs
in Broca's and Wernicke's aphasia: findings from Chinese.
Aphasiology 12, 5–36.
Choi, S., Gopnik, A., 1995. Early acquisition of verbs in Korean: a
cross-linguistic study. J. Child Lang. 22, 497–530.
Federmeier, K.D., Segal, J.B., Lombrozo, T., Kutas, M., 2000. Brain
responses to nouns, verbs and class ambiguous words in
context. Brain 123 (12), 2552–2566.
Gentner, D., 1978. On relational meaning: the acquisition of verb
meaning. Child Dev. 49, 988–998.
Gentner, D., 1981. Some interesting differences between verbs and
nouns. Cogn. Brain Theory 4 (2), 161–178.
Gentner, D., 1982. Why nouns are learned before verbs:
linguistic relativity versus natural partitioning. Lang. Dev. 2,
301–333.
159
Gernsbacher, M.A., 1984. Resolving 20 years of inconsistent
interactions between lexical familiarity and orthography,
imageability, and polysemy. J. Exp. Psychol. Gen. 113,
256–281.
Giesbrecht, B., Camblim, C.C., Swaab, T.Y., 2004. Separable effects
of semantic priming and imageability on word processing in
human cortex. Cereb. Cortex 14, 521–529.
Greenhouse, S.W., Geisser, S., 1959. On methods in the analysis of
profile data. Psychometrika 24, 95–112.
Goldberg, A.E., 2003. Constructions: a new theoretical approach to
language. Trends Cogn. Sci. 7 (5), 219–224.
Holcomb, P.J., Kounios, J., Anderson, J.E., West, W.C., 1999.
Dual-coding, context-availability, and concreteness effects in
sentence comprehension: an electrophysiological
investigation. J. Exp. Psychol. Learn. Mem. Cogn. 25, 721–742.
Hsu, J.C.-F., Tzeng, O.J.-L., Hung, D.L., 2004. Syntactic effects on
constituent components in word recognition: evidence from
compound nouns and compound verbs in Chinese. Univ.
Maryland Working Pap. Linguistics 13, 173–193.
Kelly, M.H., 1992. Using sound to solve syntactic problems: the role
of phonology in grammatical category assignments. Psychol.
Rev. 99 (2), 349–364.
Kemmerer, D., 2006. Action verbs, argument structure
constructions, and the mirror neuron system. In: Arbib, M.A.
(Ed.), In Action to Language via the Mirror Neuron System.
Cambridge University Press, Cambridge, UK, pp. 347–373.
Kounios, J., Holcomb, P.J., 1994. Concreteness effects in semantic
processing: ERP evidence supporting dual-coding theory. J. Exp.
Psychol. 20 (4), 804–823.
Kutas, M., Federmeier, K.D., 2000. Electrophysiology reveals
semantic memory use in language comprehension. Trends
Cogn. Sci. 4 (12), 463–470.
Langacker, R.W., 1987. Foundations of Cognitive Grammar, I,
Theoretical Prerequisites. Stanford University Press, Stanford.
Lee, C., Federmeier, K.D., 2006. To mind the mind: an event-related
potential study of word class and word class ambiguity. Brain
Res. 1081, 191–202.
Lee, C., Federmeier, K.D., 2008. To watch, to see, and to differ: an
event-related potential study of concreteness effects as a
function of word class and lexical ambiguity. Brain Lang. 96,
59–68.
McCarthy, R., Warrington, E.K., 1985. Category specificity in
an agrammatic patient: the relative impairment of verb
retrieval and comprehension. Neuropsychologia 23, 709–727.
Miceli, G., Silveri, M.C., Caramazza, A., 1984. On the basis for the
agrammatic's difficulty in producing main verbs. Cortex 20,
207–240.
Myers, J., 2006. Processing Chinese compounds: a survey of the
literature. In: Libben, G., Jarema, G. (Eds.), In the Representation
and Processing of Compound Words. Oxford University Press,
Oxford, pp. 169–196.
Nelson, K., 1973. Structure and strategy in learning to talk. Monogr.
Soc. Res. Child Dev. 38 (1–2), 1–135.
Oldfield, R.C., 1971. The assessment and analysis of handedness:
the Edinburgh inventory. Neuropsychologia 9, 97–113.
Paivio, A., 1986. Mental Representations: A Dual Coding Approach.
Oxford University Press, New York.
Pulvermüller, F., Lutzenberger, W., Preissl, H., 1999. Nouns and
verbs in the intact brain: evidence from event-related
potentials and high-frequency cortical responses. Cereb.
Cortex 9 (5), 497–506.
Schwanenflugel, P.J., Harnishfeger, K.K., Stowe, R.W., 1988.
Context availability and lexical decisions for abstract and
concrete words. J. Mem. Lang. 27, 499–520.
Sereno, J.A., Jongman, A., 1997. Processing of English inflectional
morphology. Mem. Cogn. 25 (4), 425–437.
Swaab, T.Y., Baynes, K., Knight, R.T., 2002. Separable effects of
priming and imageability on word processing: an ERP study.
Cogn. Brain Res. 15, 99–103.
160
BR A I N R ES E A RC H 1 2 5 3 ( 2 00 9 ) 1 4 9 –16 0
Szekely, A., D'Amico, S., Devescovi, A., Federmeier, K., Herron, D.,
Iyer, G., et al., 2005. Timed action and object naming. Cortex 41,
7–25.
Tardiff, T., 1996. Nouns are not always learned before verbs:
evidence from Mandarin speaker's early vocabularies.
Dev. Psychol. 32 (3), 492–504.
West, W.C., Holcomb, P.J., 2000. Imaginal, semantic, and
surface-level processing of concrete and abstract words: an
electrophysiological investigation. J. Cogn. Neurosci. 12 (6),
1024–1037.
Zhang, Q., Ding, J., Guo, C., Wang, Z., 2003. ERP difference between
processing of nouns and verbs. [in Chinese]. Acta Psychologia
Sinica 35 (6), 753–760.
Zhang, Q., Guo, C.Y., Ding, J.H., Wang, Z.Y., 2006. Concreteness
effects in the processing of Chinese words. Brain Lang. 96 (1),
59–68.
Zhou, X., Marslen-Wilson, W., 1995. Morphological structure in
the Chinese mental lexicon. Lang. Cogn. Processes 10 (6), 545–600.
Zingeser, L.B., Berndt, R.S., 1990. Retrieval of nouns and verbs in
agrammatism and anomia. Brain Lang. 39, 14–32.