Journal of Neurolinguistics 31 (2014) 42e54
Contents lists available at ScienceDirect
Journal of Neurolinguistics
journal homepage: www.elsevier.com/locate/
jneuroling
Effects of semantic constraint and cloze
probability on Chinese classifier-noun
agreement
Chia-Ju Chou a, Hsu-Wen Huang b, Chia-Lin Lee c,
Chia-Ying Lee a, d, e, *
a
Institute of Cognitive Neuroscience, National Yang-Ming University, Taiwan
Department of Applied Chinese Language and Culture, National Taiwan Normal University, Taiwan
c
Institute of Linguistics, National Taiwan University, Taiwan
d
Institute of Linguistics, Academia Sinica, Taiwan
e
Institute of Cognitive Neuroscience, National Central University, Taiwan
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 11 February 2014
Received in revised form 24 June 2014
Accepted 25 June 2014
Available online
This study aims to examine when and how readers make use of
top-down information to predict or integrate upcoming words by
utilizing the characteristics of Chinese classifier-noun agreement,
as measured by event-related potentials (ERPs). Constraint
strength of classifiers (strong and weak) and cloze probability of
the pairing noun (high, low, implausible) was manipulated.
Weakly constrained classifiers elicited a less positive P200 and an
enhanced frontal negativity than strongly constrained classifiers,
suggesting that readers used the preceding classifier to predict the
upcoming noun, even before the pairing noun appeared. For ERPs
elicited by the pairing nouns, there was a significant interaction
between semantic constraint and cloze probability for the N400.
For nouns following the weakly constrained classifiers, there was a
graded cloze probability effect on the N400 (High < Low < Imp).
For nouns following the strongly constrained classifiers, both low
cloze and implausible nouns elicited larger N400s than high cloze
nouns; however, there was no difference between low cloze and
implausible nouns. The critical comparison for the constraint effect
of low cloze nouns was found for the N400 but not for frontal
Keywords:
Constraint
Cloze probability
Classifier-noun agreement
N400
Frontal negativity
* Corresponding author. Institute of Linguistics, Academia Sinica, 128, Section 2, Academia Road 115, Taipei, Taiwan.
Tel.: þ886 2 2652 5031; fax: þ886 2 2785 6622.
E-mail address: [email protected] (C.-Y. Lee).
http://dx.doi.org/10.1016/j.jneuroling.2014.06.003
0911-6044/© 2014 Elsevier Ltd. All rights reserved.
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
43
positivity, suggesting that the N400 reflects a joint effect of both
benefit and cost of prediction.
© 2014 Elsevier Ltd. All rights reserved.
1. Introduction
Studies have demonstrated that processing a word can be influenced by its preceding context.
Readers are usually faster and more accurate in processing words that are congruent with their preceding context (Duffy, Henderson, & Morris, 1989; Stanovich & West, 1981). By recording eye movements during natural reading, fixation and gaze durations are usually shorter for highly expected
words than for unexpected words that are embedded in the sentences (Dambacher, Goellner,
Nuthmann, Jacobs, & Kliegl, 2008; Kliegl, Grabner, Rolfs, & Engbert, 2004; Rayner, Ashby, Pollatsek,
& Reichle, 2004). These findings suggest that contextual information plays an important role in language comprehension. However, it remains unclear when and how readers make use of contextual
information to predict or integrate the meaning of upcoming words. The present study aims to address
this issue by using the unique characteristics of Chinese classifier-noun agreement with event-related
potentials (ERPs), which provides great temporal resolution. Additionally, several ERP components
(such as P200, N400, and frontal positivity) can be used to index various stages of cognitive processing.
In the ERP literature, contextual effect is usually evaluated by manipulating the degree of fit or
semantic congruency between the context and upcoming words (Kutas & Hillyard, 1980a, 1980b, 1984),
word predictability (Dambacher & Kliegl, 2007; Dambacher, Kliegl, Hofmann, & Jacobs, 2006; Van
Petten & Kutas, 1990), or sentential constraint (Federmeier, Wlotko, De Ochoa-Dewald, & Kutas,
2007; Hoeks, Stowe, & Doedens, 2004; Wlotko & Federmeier, 2007). Despite the various ways to
evaluate contextual influences, empirically they are determined by the cloze probability, which is
measured by calculating the percentage of people who complete a sentence frame with a particular
word (Taylor, 1953). A well-replicated finding indicates that N400 amplitudes are inversely proportional to the cloze probability. The reduction of N400 amplitude is found with words that can be easily
integrated into the preceding word, sentence, or discourse context (van Berkum, Hagoort, & Brown,
1999; Kutas & Hillyard, 1980a, 1980b; Van Petten & Kutas, 1990, 1991). When a word in a context
has a higher cloze probability, there is more of a reduction in N400 amplitude when compared to an
unexpected word (Kutas & Hillyard, 1984).
However, the reduced N400 for high cloze probability words in sentences may either reflect the use
of contextual information to predict and pre-activate the upcoming word (predictive view), or the ease
of integrating a word into its preceding context (integrative view). The major difference between these
two views is that the predictive view assumes contextual information can be used in an anticipatory or
predictive manner to exert its effect starting from the early processing stages of word recognition, such
as early perceptual features analysis, to the later stages of lexical activation and selection (Federmeier,
2007; Lee, Liu, & Chou, 2013; Lee, Liu, & Tsai, 2012), whereas the integrative view assumes that language comprehension is mainly based on the post-lexical semantic integration of each embedded word
in the sentence (Fodor, 1983; Schwanenflugel & Shoben, 1985; Van Petten & Luka, 2012).
To further examine how contextual constraint affects the processing of unexpected words,
Federmeier and Kutas (1999) manipulated the sentential constraint and degree of semantic overlapping between an unexpected ending word and its best competition. For example, the sentence,
“They wanted to make the hotel look more like a tropical resort, so along the driveway, they planted rows of
…” can be completed by one of the three ending types: (1) the expected final word, established by cloze
probability, e.g., palms; (2) within-category violation, an unexpected item from the expected semantic
category, e.g., pines; or (3) between-category violation, an unexpected item from a different semantic
category, e.g., tulips. Based on the integrative view, there would be no difference between within- and
between-category violations, as both contain features that are not coherent within the context.
However, the data from this study demonstrated not only that the expected item elicited a smaller
N400 than either violation type, but the within-category violation also elicited a smaller N400 than the
between-category violation. Moreover, this pattern was mainly found when the final words were
44
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
presented in strongly constrained context. Thus, strongly constrained sentences led to a stronger
prediction for the expected item than did more weakly constrained sentences, thereby supporting the
predictive view.
Other supporting evidence comes from Federmeier et al. (2007), who tried to separate two types of
contextual effects by manipulating sentential constraint and the cloze probability of ending words in
sentences. Sentences were subdivided into strongly constrained and weakly constrained types, and
each ended with a high cloze probability (highly expected) word and a low cloze probability (unexpected but plausible) word (Note that the strength of sentential constraint was defined by the cloze
probability of the best completion for a sentence frame). Strongly constrained sentences were defined
as those with at least 70% agreement for the best completion, whereas weakly constrained sentences
were defined as sentences with no more than 40% agreement for the best completion. For example, in
the strongly constrained sentence “He brought her a pearl necklace for her birthday,” 91% of participants
would complete this sentence with the word “birthday”. However, some sentence frames may lack
agreement with respect to their completion. For example, the weakly constrained sentence “He looked
worried because he might have broken his arm” can be completed with a rather divergent end word,
because the best completion word “arm” only has a cloze probability of 36%. The data from this study
revealed that N400 amplitudes were graded by the cloze probability but were unaffected by constraint,
suggesting that the N400 primarily indexes the benefit of supportive contextual information for word
processing. Meanwhile, unexpected words completing strongly constrained sentence frames elicited
enhanced frontal positivity when compared to completing weakly constrained sentence frames. This
effect was interpreted as the need to suppress wrong expectations, and reflected the possible cost
associated with processing unexpected words in strongly constrained contexts (Federmeier et al.,
2007).
Wlotko and Federmeier (2007) further applied the same set of stimuli while using a split visual field
paradigm to examine whether the two hemispheres preferred different modes of language comprehension. Their results revealed P200 sensitivity to constraint; words in strongly constrained contexts
elicited a larger P200 than those in less predictive contexts, but only when presented in the right visual
field. Conversely, the N400 responses for both visual fields departed from the typical pattern where
ERP amplitude was graded by cloze probability. Expected endings in strongly and weakly constrained
contexts were facilitated to a similar degree with right visual field presentation, while expected
endings in weakly constrained contexts were not facilitated compared to unexpected endings in the
same context with left visual field presentation. However, the larger frontal positivity for unexpected
endings in strongly constrained contexts observed in Federmeier et al. (2007) with central presentation
was not seen in either visual field (Wlotko & Federmeier, 2007).
These studies demonstrated that it was possible to dissociate the two types of contextual effects, as
well as sentence frame constraint and cloze probability of completion, in different ERP components. A
major caveat of these studies is that the strength of constraint is also defined by the cloze probability of
the most-favored completion. Thus, both sentential constraint and cloze probability are related but
partially independent measures. In particular, since only strongly constrained sentences could have a
high cloze probability of completion, the constraint effect on the best completion would naturally be
confounded with the cloze probability effect. For example, Federmeier et al. (2007) and Wlotko and
Federmeier (2007) reported that cloze probabilities of best completions for strongly and weakly
constrained frames were 91% and 36.2%, respectively. Since constraint and cloze probability are
inherently confounded for expected completions, the only way to evaluate the effects of constraint is to
compare the processing of unpredictable items for both complete strongly and weakly constrained
sentences (average cloze probability was 3% for unexpected items in both of strongly and weakly
constraining contexts). Therefore, it was not possible to evaluate the effects of sentential constraint and
cloze probability in an orthogonal manner in these studies.
The present study aims to overcome this problem by using Chinese classifier-noun agreement. In
Mandarin Chinese, whenever a noun is preceded by a number or a demonstrative, a classifier must
come in between. For example, “one person” or “this person” in English shall be 一個人 (yi1 ge ren2, oneclassifier person) or 這個人 (zhe4 ge ren2, this-classifier person) in Chinese. The basic structure of a
classifier phrase consists of a number, a classifier, and a noun. In the traditional view, Chinese classifiers
are often not distinguished from measure words in the discussion of Chinese grammar. Chao (1968,
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
45
584e620) refers to classifiers as individual measures and subsumes them under the rubric of measure
words. Duffy et al. (1989) stated that “any measure word can be a classifier.” More recently, Tai (1992)
pointed out that there is an important distinction between the notion that classifiers can only classify a limited and specific group of nouns, while measure words can be used as a measure for a wide
variety of nouns. Tai (1992) also suggested that a “classifier categorizes a class of nouns by picking out
some salient perceptual properties, whether physically or functionally based, which are permanently
associated with the entities named by the class of nouns; a measure word does not categorize but denotes
the quantity of the entity named by a noun.” Therefore, Chinese classifiers are said to carry meaning (to a
greater or lesser extent) about the semantic features of the entities being classified. Ahrens (1994)
argued that the use of Chinese classifiers in modern Mandarin is semantically motivated, although
not fully predictable. This is supported by recent research using semantic judgment and word-picture
matching to demonstrate the connection between classifier usage and the structure of conceptual
categories (Gao & Malt, 2009; Kuo & Sera, 2009; Tien, Tzeng, & Huang, 2002). Studies also indicate that
classifier categories may gain the benefit of predicting succeeding nouns that shared a classifier
(Saalbach & Imai, 2007; Srinivasan, 2010; Zhang & Schmitt, 1998). However, there is some specific
correspondence between classifiers and nouns. For example, books generally take the classifier 本 ben3,
flat objects such as 紙 (paper) take 張 zhang1, and animals usually take 隻 zhi1. Within these categories
are further subdivisions; while most animals take 隻 zhi1, domestic animals (such as 牛, cow) take 頭
tou2, long and flexible animals (such as 蛇, snake) take 條 tiao2, and horses take 匹 pi3. In other words,
classifiers vary in how specific they are. Some (such as 頂 ding3 for hats) could only be used with a few
nouns, and thus provide high semantic constraint to their paring nouns. Others (such as 條 tiao2 for
long and flexible things, one-dimensional things, or abstract items like a news report) are much less
restricted and thus provide relatively low constraint to their pairing nouns. No matter for high or low
constraining classifiers, all plausible nouns can be measured by their cloze probability. For example, a
weakly constrained classifier, 瓶 ping2 (bottle), constrained nouns, including low cloze words, 膠水jiao1
shuei3 (glume; cloze ¼ 3.45%), moderately cloze words, 水shui3 (water; cloze ¼ 51.70%), and the most
preferred noun 飲料yin3 liao4 (drink; cloze ¼ 72.41%). Although the factors that govern which classifiers are paired with what nouns have been the subject of debate among linguists, the specificity of
Chinese classifier-noun agreement poses an interesting testing ground for dissociating the effects of
classifier semantic constraint and the cloze probability of pairing nouns.
In this study, we presented Chinese classifier-noun pairs that were manipulated to have two levels
of classifier constraint strength (strong and weak) and three levels of cloze probability for the pairing
noun (high, low, implausible). Most importantly, the three levels of cloze probability were well
matched between strongly and weakly constrained conditions in order to dissociate the constraint and
cloze probability effects in a set of ERP components.
2. Materials and methods
2.1. Participants
Twenty-three right-handed native Chinese speakers were paid to participate in this study. After
artifact rejection, four participants were excluded from the analysis due to the extremely small number
of valid trials in their data (less than 12 trials in at least one of the conditions). The final analysis was
conducted with 19 participants (9 males, mean age 22 years, age range 18e26 years). All participants
reported normal vision without any history of neurological or psychiatric disorders. Written consent
was obtained from all participants.
2.2. Experimental design and materials
The semantic constraint (weak versus strong) of classifiers and the cloze probability of the pairing
noun (high cloze, low cloze, or implausible [zero cloze probability]) were manipulated in a 2-by-3
factorial design. One hundred and twenty classifiers were chosen. Half of these classifiers were strongly
constrained classifiers and the other half were weakly constrained classifiers, as determined by a
norming study of Chinese classifiers regarding the number of nouns that can be followed by a specific
46
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
classifier. Every classifier would appear twice throughout the whole experiment; once it would be
paired with a plausible noun and the other time it would be paired with an implausible noun. The
plausible classifier-noun pairs were further divided into two conditions; half were paired with a highly
expected noun (high cloze probability) and the other half were paired with an unexpected but plausible
noun (low cloze probability) (see Table 1). To prevent the participant from being aware that second
appearance of a classifier was either followed by a plausible or an implausible noun, one hundred and
twenty filler pairs were added. For those 120 fillers, half of the classifiers was always followed by
plausible nouns and the other half was always followed by implausible nouns, to break the rule.
To determine the strength of constraint for Chinese classifiers and the cloze probability for their
pairing nouns, a norming procedure for 184 Chinese classifiers was conducted with 116 native Chinese
speakers (age range 18e25 years) from National Yang-Ming University and National Cheng-Chi University in Taipei, Taiwan. The 184 classifiers were selected from the Handbook of Common Quantifiers
(Version 3, 1997) and the Academia Sinica Balanced Corpus of Modern Chinese (Huang & Chen, 1998),
and were divided into four lists of 46 classifiers each. Each list was completed by 29 participants. For
rating subjective familiarity, participants were instructed to imagine how often the classifier was used
based on their experience, and were also asked to perform the subjective rating based on the 7-point
scale, ranging from 1 (unknown) to 7 (extremely familiar). For rating strength of constraint, participants were asked to estimate how many nouns could be followed by each classifier based on a 5-point
scale, ranging from 1 (could not think of any nouns) to 5 (more than four nouns).
For the cloze probability norming of the following noun, participants were first asked to complete
the fragment of “numeral þ classifier þ ___,” with the first noun coming to their mind. In addition, we
followed a procedure similar to that used in Federmeier et al. (2007), which instructed participants to
give two or three additional plausible completions if possible. Thus, the cloze probability could be
calculated not only for the best completion, but also for the “next best” completion. For example,
among 29 participants, 25 of them completed the classifier 頂 ding3 with 帽子 (mao4 zi, hat) as the best
completion and 3 of them completed the classifier 頂 ding3 with帽子 (mao4 zi, hat) as the next best
completion. The cloze probabilities of 帽子mao4 zi for being the first and the next best completions are
86.21% (25/29) and 10.34% (3/29), respectively. In addition to the cloze probability of best completion,
we also summed up the best and the next best cloze probabilities to index the overall cloze probabilities for a specific noun. For example, for 帽子, its overall cloze probabilities for being a highly expected item is 96.55% (86.21% þ 10.34%) (see Table 1 and Table 2 for sample materials). This procedure
Table 1
Example of the stimuli for each condition and their characteristics
Conditions
Classifier-noun
agreement pairs
Classifier
Constraint
rating
Noun
Familiarity
(1e5 scale) (1e7 scale)
一頂帽子
one ding3 hat
Low cloze nouns
一架飛機
one jia4 airplane
Implausible nouns 一頂船員/一架國王
one ding3 sailor/one
jia4 king
WC High cloze nouns
一瓶飲料
one ping2 drink
Low cloze nouns
一條馬路
one tiao2 road
Implausible nouns 一瓶學分/一條首都
one ping2 credit/one
tiao2 capital
SC
High cloze nouns
“Next best” Overall
Frequency
First
cloze (%)
completion cloze (%)
cloze (%)
3.23
(0.40)
3.02
(0.41)
3.13
(0.41)
4.13
(0.81)
3.67
(0.68)
3.90
(0.77)
63.79
(13.55)
2.07
(2.65)
0.00
10.34
(9.01)
2.18
(1.69)
0.00
73.79
(12.20)
4.25
(1.48)
0.00
123.23
(102.62)
111.77
(93.39)
124.30
(43.39)
4.31
(0.37)
4.51
(0.36)
4.41
(0.38)
5.78
(0.61)
5.88
(0.56)
5.83
(0.59)
57.47
(18.92)
1.03
(1.84)
0.00
18.16
(12.14)
2.64
(1.42)
0.00
75.63
(14.03)
3.68
(0.87)
0.00
128.07
(101.96)
114.07
(93.99)
123.82
(45.11)
Note. SC: strongly constrained; WC: weakly constrained. All scales reported as mean with standard deviation in parentheses.
First completion cloze reflects the percentage of participants who completed a given classifier with the first noun coming to their
mind. “Nest best” cloze reflects the cloze probability for additional plausible completions, and overall cloze reflects the sum of
first completion cloze and “next best” cloze.
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
47
allowed us to avoid the possibility of unexpected items within the set of the next best completion that
participants could use to complete the fragment. The total amount of possible completions produced
across all the participants for a specific classifier also served as an index for the strength of constraint.
From the resulting database, strongly or weakly constrained classifiers were defined by whether
they had a constraining strength less or more than 3.8. In addition, the high cloze probability nouns
were chosen from those nouns that had a cloze value of 50%, which was based on the overall cloze
probability. The low cloze probability nouns were chosen from those nouns that had a cloze value of
6.9% or less. The implausible nouns were chosen outside the set of possible completions and therefore
had a cloze value of zero. The classifiers across four conditions were matched for subjective familiarity,
and the completing nouns were matched for word frequency. The examples and characteristics of the
six types of classifier-noun pairing conditions are listed in Table 1.
2.3. Procedure
Participants were seated in front of a computer monitor at a distance of approximately 75 cm in an
acoustically shielded room. Each trial began with a fixation cross appearing at the center of the screen
for 500 ms, followed by a variable inter-stimulus interval (ISI) between 200 and 350 ms. The classifier
was presented in the center of the screen for 400 ms with a 600 ms ISI. The noun was then presented for
400 ms with a 600 ms ISI. Participants were encouraged to minimize eye movements, blinks, and muscle
movements during this period. After the 600 ms ISI, the cue “?” was presented on the screen to instruct
participants to perform an acceptability judgment for the classifier-noun pair based on a 5-point scale,
with 1 equaling “totally unacceptable,” and 5 equaling “highly expected and totally acceptable.” If a
response was made, the next trial began after an ISI of 1500 ms; if no response was made, the next trial
began 4 s after the onset of the cue. The recording session began with a short set of practice trials in
order to acclimate the participants to the task. The main experimental session was divided into five
blocks that lasted 5e8 min each (2 practice trials with 72 trials/session), with participants taking a short
rest between each block. Different sequences were randomly assigned to each participant. Participants
would view each classifier twice and there were at least 50 different trials in between each classifier in
order to reduce repetition effects. The whole experiment lasted approximately 40 min.
Table 2
Example classifiers and cloze probabilities of the best completion and other completions
Classifier
頂
ding3
帖
tie3
Subjective rating
of constraint
Number of
completion
Best completion
(cloze %)
Completions (meaning, cloze %)
假髮(wig, 24.14), 斗笠(bamboo hat, 6.90), 王
冠(crown, 10.34), 浴帽(shower cap, 3.45)
喜帖(wedding invitation, 24.14), 秘方(secret recipe,
10.34), 字(word, 3.45), 書信(letter, 3.45), 請
柬(invitation, 3.45), 戰帖(challenge, 3.45), 邀請
函(invitation, 3.45), 書法(calligraphy, 3.45)
雜誌(magazine, 24.14), 課本(textbook, 17.24), 小
說(fiction, 13.79), 漫畫(comics, 13.79), 冊
子(pamphlet,10.34), 畫冊(picture album, 10.34), 相
簿(album, 6.90), 簿子(notebook, 6.90), 日記(diary,
3.45), 目錄(catalogue, 3.45), 字帖(copybook for
calligraphy, 3.45), 作業簿(homework, 3.45), 參考
書(reference book, 3.45), 報刊(newspaper, 3.45), 遊
記(travel notes, 3.45), 講義(handout, 3.45)
湯(soup, 37.93), 麵(noodle, 34.48), 粥(congee,
10.34), 湯麵(noodle soup, 10.34), 豆花(tofu pudding,
6.90), 茶水(water, 6.90), 米粉(rice noodles, 6.90), 麵
線(thin noodles, 6.90), 菜(dish, 6.90), 牛肉麵(beef
noodles soup, 3.45), 紅豆(red bean, 3.45), 濃湯(thick
soup, 3.45), 湯藥(Chinese medicine, 3.45), 冰(ice,
6.90), 湯圓(rice ball, 3.45), 稀飯(rice porridge, 3.45),
開水(boiled water, 3.45), 碗粿(salty rice pudding,
6.90), 陽春麵(plain noodles soup, 3.45), 乾
麵(noodles, 3.45)
3.41
5
帽子(hat, 96.55)
3.00
9
藥(medicine, 55.17)
本
ben3
4.52
17
書(book, 93.10),
碗
wan3
4.59
22
飯(rice, 75.86)
48
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
2.4. EEG recording and processing
Electroencephalograms (EEG) were recorded from 64 sintered Ag/AgCl electrodes (QuickCap,
Neuromedical Supplies, Sterling, Texas, USA) with a common vertex reference located between Cz and
CPz. The data were re-referenced off-line to the average of the right and left mastoids for further
analysis. EEG was continuously recorded and digitized at a rate of 500 Hz. The signal was amplified by
SYNAMPS2 (Neuroscan Inc., El Paso, Texas, USA) with the band-pass set at 0.05e100 Hz. Vertical eye
movements were recorded by a pair of electrodes placed on the supraorbital and infraorbital ridges of
the left eye. Horizontal eye movements were recorded by electrodes placed lateral to the outer canthus
of the right and left eyes. A ground electrode was placed on the forehead, anterior to the FZ electrode.
Electrode impedance remained below 5 kU.
For the off-line analysis, the EEG was epoched from 100 ms before the onset of the classifier and the
target noun to 900 ms after. Mean amplitude in the selected latency window (see Results) were
measured with respect to the 100 ms pre-stimulus baseline. Trials contaminated by eye movement or
with voltage variations larger than 60 mV were rejected prior to averaging the trials into ERPs for each
condition. Band-pass filter of 0.1 and 30 Hz (zero phase shift mode, 12 dB) were employed. ERPs were
calculated for each subject and each condition.
3. Results
3.1. Behavioral results
Plausibility judgment data were subjected to a repeated-measures analysis of variance (ANOVA)
with two levels of Constraint (strongly and weakly constrained) and three levels of Cloze probability
(high, low, and implausible). There were main effects of Constraint (F(1, 18) ¼ 47.71, p < .01) and Cloze
probability (F(2, 36) ¼ 1741.90, p < .01), as well as a Constraint by Cloze probability interaction (F(2,
36) ¼ 101.61, p < .01). Further analysis revealed a significant main effect of Cloze probability in both
strongly (F(2, 36) ¼ 4865.78, p < .01) and weakly constrained (F(2, 36) ¼ 5983.46, p < .01) conditions.
For the strongly constrained condition, high cloze probability nouns were rated as being more plausible
on average (mean ¼ 4.50, SD ¼ 0.43) than low cloze probability nouns (mean ¼ 3.58, SD ¼ 0.43, F(1,
18) ¼ 727.79, p < .01) and implausible nouns (mean ¼ 1.23, SD ¼ 0.15, F(1, 18) ¼ 9151.70, p < .01). For the
weakly constrained condition, the mean score for the high cloze probability nouns (mean ¼ 4.67,
SD ¼ 0.26) was slightly higher than that for low cloze probability nouns (mean ¼ 4.25, SD ¼ 0.38, F(1,
18) ¼ 151.09, p < .01) and also higher than that for implausible nouns (mean ¼ 1.23, SD ¼ 0.12, F(1,
18) ¼ 10056.80, p < .01).
3.2. Event-related potentials
To evaluate the semantic constraining effect of classifiers and how the semantic constraint
modulated the cloze probability of the pairing noun, we analyzed the ERPs elicited by classifiers and by
the pairing nouns, respectively. The mean amplitude of ERPs components, including P200, N400, and
frontal negativity, from the selected electrodes served as the dependent measure in repeated-measures
ANOVAs. For each ANOVA, the Greenhouse-Geisser adjustment for degrees of freedom was applied to
correct for the violations of sphericity associated with repeated measures. Accordingly, for all the F
tests with more than one degree of freedom in the numerator, the corrected p-value was reported. The
post-hoc tests were conducted using Tukey's procedure.
3.2.1. Classifiers
Fig. 1 shows the ERP grand average to both strongly and weakly constrained classifiers. All conditions elicited early sensory components, such as the N1 and P200. These components were followed by
a sustained frontal negativity, which appeared to be more negative for low than for constrained
classifiers. Based on previous studies (Federmeier et al., 2007), we measured the mean amplitudes for
the P200 between 170 and 250 ms (peaking at 210 ms) and for the frontal negativity between 300 and
700 ms on nine frontal electrodes (FZ, FCZ, CZ, F3/4, FC3/4, and C3/4). For the ERPs elicited by classifiers,
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
49
Fig. 1. ERP grand average for Classifiers at frontal and central sites. Negative values are plotted on the positive y-axis here and in all
subsequent figures.
a repeated-measure ANOVA was performed on the mean amplitudes of the P200 and frontal negativity
with Constraint (strongly versus weakly) and electrode site (nine frontal electrode sites) as withinsubject factors.
3.2.1.1. P200 (170e250 ms). The data revealed a significant main effect of Constraint (F(1, 18) ¼ 7.93,
p < .05) and a Constraint-by-electrode interaction (F(8, 144) ¼ 3.68, p < .05). Strongly constrained
classifiers elicited more positive P200s than weakly constrained classifiers. Follow-up comparisons
revealed that the constraining effect was significant at almost every electrode site (p < .001), except for
CZ and C4.
3.2.1.2. Frontal negativity (300e700 ms). The data revealed a significant main effect of Constraint (F(1,
18) ¼ 10.1, p < .01). Weakly constrained classifiers elicited enhanced negativity.
3.2.2. Nouns
Fig. 2 displays the ERP grand average to high cloze probability, low cloze probability, and implausible
nouns that completed the strongly and weakly constrained classifiers. Based on visual inspection, all
conditions elicited the early N1 and P200, which were then followed by a broadly distributed N400 that
was largest at the centro-posterior sites. The N400 amplitude appeared to be more negative for
implausible nouns no matter the constraint condition. There was also a broadly distributed late positive
component between 600 and 900 ms. Three components were identified for further analysis. The first
one is the P200 (170e250 ms) over nine frontal electrodes (FZ, FCZ, CZ, F3/4, FC3/4, and C3/4). The second
and third components are the N400 (300e500 ms) and frontal positivity (600e900 ms), which were
measured on 15 electrodes (FZ, FCZ, CZ, CPZ, PZ, F3/4, FC3/4, C3/4, CP3/4, and P3/4) over the entire head
and these electrodes were further divided into three levels of Laterality (left lateral, medial, and right
lateral electrode sites) and five levels of Anteriority (frontal, frontal-central, central, central-parietal, and
parietal electrode sites). For the ERPs elicited by the pairing nouns, a repeated-measure ANOVA was
performed on the mean amplitudes of the P200, N400, and frontal positivity with Constraint (strongly
versus weakly), Cloze probability (high, low, and implausible), and electrode site (P200:9 electrodes,
N400 and frontal positivity, 3 levels of Laterality by 5 levels of Anteriority) as within-subject factors.
3.2.2.1. P200 (170e250 ms). The analysis revealed neither a significant main effect (Constraint: F(1,
18) ¼ 0.02, p ¼ .88; Cloze probability: F(2, 36) ¼ 2.32, p ¼ .12) nor any significant interactions (all F < 1).
3.2.2.2. N400 (300e500 ms). There was a significant main effect of Cloze probability (F(2, 36) ¼ 14.18,
p < .01) and a Constraint by Cloze probability interaction (F(2, 36) ¼ 5.15, p < .05), but there was no
50
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
Fig. 2. ERP grand average for nouns from midline scalp sites, arranged from frontal at the top to parietal at the bottom. High cloze
probability words: solid black lines; low cloze probability words: dashed lines; implausible words: solid gray line.
reliable main effect of Constraint (F(1, 18) < 0.01, p ¼ .96). Post-hoc tests revealed that cloze probability
effects were significant for both strongly constrained (F(2, 36) ¼ 23.78, p < .01) and weakly constrained
conditions (F(2, 36) ¼ 10.11, p < .01). For the strongly constrained (SC) condition, both unexpected
nouns (low cloze probability and implausible nouns) elicited a greater N400 than expected nouns did
(high cloze probability; (SC-High versus SC-Low, F(1,18) ¼ 40.99, p < .01; SC-High versus SC-Imp, F(1,
18) ¼ 29.4, p < .01). The difference between low cloze probability and implausible nouns was not
reliable (SC-Low versus SC-Imp: F(1, 18) ¼ 0.96, p ¼ .33). For the weakly constrained (WC) condition,
implausible nouns elicited a more negative N400 than both high and low cloze probability nouns (High
< Low < Imp; WC-High versus WC-Low: F(1, 18) ¼ 4.36, p < .05; WC-High versus WC-Imp: F(1,
18) ¼ 20.19, p < .01; WC-Low versus WC-Imp: F(1, 18) ¼ 5.78, p < .05)). In Federmeier et al. (2007), the
most critical comparison was the constraint effect for low cloze words. Here, we performed this same
comparison and demonstrated a significant Constraint effect for low cloze probability nouns (F(1,
18) ¼ 6.42, p < .05). However, this effect was not found for high cloze probability (F(1, 18) ¼ 3.17, p ¼ .08)
or implausible nouns (F(1, 18) ¼ 0.72, p ¼ .40). There was also a significant Cloze probability-byLaterality interaction (F(4, 72) ¼ 7.07, p < .01), and a Cloze probability-by-Anteriority interaction
(F(8, 144) ¼ 7.89, p < .01). Follow-up comparisons revealed that the cloze probability effect was significant at all electrode sites (all p < .01), and was most prominent at central-parietal sites located off to
the right from the center of the scalp.
3.2.2.3. Frontal positivity (600e900 ms). The results indicated a main effect of Cloze probability (F(2,
36) ¼ 3.45, p < .05) and a marginally significant interaction between Constraint and Cloze probability
(F(2, 36) ¼ 3.15, p ¼ .06). There was no main effect of Constraint (F(1, 18) ¼ 1.08, p ¼ 1.31). Follow-up
comparisons revealed that high cloze probability nouns elicited a larger positivity than implausible
nouns (F(1, 18) ¼ 7.09, p < .01) in the strongly constrained condition, and this effect was most prominent at central-parietal electrode sites. There was no difference between high cloze probability and
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
51
low cloze probability nouns (F(1, 18) ¼ 3.81, p ¼ .06), and no difference between low cloze probability
and implausible nouns (F(1, 18) ¼ 0.51, p ¼ .48). For the weakly constrained condition, low cloze
probability nouns elicited a larger positivity than implausible nouns (F(1, 18) ¼ 7.03, p < .01). However,
there was no difference between high cloze probability and low cloze probability nouns (F(1, 18) ¼ 2.53,
p ¼ .12) and no difference between high cloze noun and implausible nouns (F(1, 18) ¼ 1.13, p ¼ .30). To
examine the critical contrast reported by Federmeier et al. (2007), which showed that the less expected
continuations within strongly constrained sentence frames elicited the largest positivity, our comparisons revealed that the effect of constraint was observed for high cloze probability nouns. High cloze
probability nouns that completed the strongly constrained classifiers elicited a larger positivity than
those with weakly constrained classifiers (SC-High versus WC-High: F(1, 18) ¼ 4.87, p < .05). The
constraint effect for low cloze probability nouns was not significant (SC-Low versus WC-Low; F(1,
18) ¼ 1.78, p ¼ .19).
4. Discussion
This study aimed to examine when and how the semantic constraint of Chinese classifiers may
affect the processing of their following nouns by using ERP components to index the multiple stages of
cognitive processing. In each experimental trial, participants would first read a classifier with either a
strong or weak constraint for its following noun. Later, a high cloze probability noun, a low cloze
probability noun, or an implausible noun would then appear to complete the classifier-noun phrase.
The current design allowed us to examine the brain responses elicited by classifiers and their pairing
nouns separately. In particular, comparing the brain responses elicited by strongly and weakly constrained classifiers provided a great opportunity to examine whether the brain makes use of the
preceding classifier to predict the upcoming noun, even before the pairing noun is shown.
The ERPs evoked by classifiers revealed a significant effect of constraint on the P200 and frontal
negativity. The weakly constrained classifiers that could be completed with a larger set of nouns
produced a less positive P200 and a more negative frontal negativity than the strongly constrained
classifiers that could only be completed with a limited set of nouns. Studies have suggested the sustained frontal negativity may reflect working memory demands and the need to maintain and select
among candidate items during recollection (King & Kutas, 1995; Lee & Federmeier, 2006, 2009; Rugg,
Allan, & Birch, 2000). For example, King and Kutas (1995) manipulated whether the main subject was
the subject (subjectesubject relative [SS] sentences), or the object (subject-object relative [SO] sentences) in a relative clause. They found an enhanced sustained negativity for the main verb in SO
sentences relative to SS sentences. Furthermore, such a frontal effect was greater in participants with
good comprehension than in participants with poor comprehension. In contrast to the SS sentences
wherein the head noun was almost immediately assigned its appropriate thematic role, the head noun
in SO sentences must be stored in working memory. Therefore, the frontal negativity may in part be
related to differences in memory storage requirements. Lee and Federmeier's series of studies
embedded noun/verb (NV) homographs in syntactic-constrained but semantically neutral contexts
and found that NV homographs elicited a sustained frontal negativity when compared to matched
unambiguous words (Lee & Federmeier, 2006, 2009). The sustained frontal negativity has also been
observed in other cases where there are multiple referents for a class unambiguous word (Nieuwland &
Van Berkum, 2006). In our study, the strongly constrained classifier may lead to a strong prediction for
its following nouns. However, for the weakly constrained classifiers, participants might need to
maintain all possible candidates for further selection before the noun appears, eliciting a larger frontal
negativity compared to strongly constrained classifiers.
We further measured the ERPs elicited by pairing nouns to examine whether the semantic
constraint of the classifier would affect the processing of pairing nouns at multiple stages, as previous
studies have suggested that contextual information may facilitate the perceptual processing of the
upcoming word, as measured by the P200 (Lee et al., 2012; Wlotko & Federmeier, 2007), and reveal
benefit and cost of semantic predictions in the N400 and frontal positivity responses (DeLong, Urbach,
Groppe, & Kutas, 2011; Federmeier, 2007). In the early time window of the P200, our data revealed that
the ERPs elicited by the nouns revealed neither semantic constraint nor cloze probability effects.
Studies using the split visual field paradigm have demonstrated that the P200 is larger (i.e., more
52
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
positive) for strongly constrained sentence endings for right but not left visual field presentations
(Federmeier, Mai, & Kutas, 2005; Wlotko & Federmeier, 2007). However, using the same set of stimuli
as Wlotko and Federmeier (2007), Federmeier et al. (2007) found no constraint effect on the P200 with
central presentation. It was suggested that only the left hemisphere could use contextual information
to predict the perceptual features of the upcoming words and, with a feature match, could engender an
enhanced P200. Therefore, it might be difficult to demonstrate the contextual effect on the P200 with
the central presentation. Nonetheless, previous studies have observed the P200 effect with central
presentation (Barber, Vergara, & Carreiras, 2004; Hsu, Tsai, Lee, & Tzeng, 2009; Lee et al., 2012). For
example, by manipulating the cloze probability of the ending word in a sentence, Lee et al. (2012)
demonstrated a predictability effect on the P200, in which low-predictability words elicited a less
positive P200 than high-predictability words. In the current study, ERPs to preceding classifiers also
showed a semantic constraint effect on the P200, even though the pairing nouns had not appeared yet.
Therefore, the constraint effect of classifiers on the P200 should have nothing to do with feature
matching. This is congruent with Wlotko and Federmeier (2007), who found that the P200 was
identical for both expected and unexpected items in a strongly constrained context, suggesting that the
P200 was not affected by whether or not the predicted word was actually presented. Thus, the P200
may be sensitive to the state change, rather than the feature matching, between expectation and the
item actually presented (Wlotko & Federmeier, 2007).
When using context to predict upcoming words, there are potentially two types of processing
consequences: the benefits of prediction and the costs of incorrect predictions. Although it has been well
recognized that the N400 is sensitive to cloze probability and semantic congruency (DeLong, Urbach, &
Kutas, 2005; Kutas & Hillyard, 1984; Wlotko & Federmeier, 2007), it is difficult to determine whether the
reduced N400 for high cloze and semantically congruent words reflects the benefit of prediction or
whether the enhanced N400 for low cloze words or semantically incongruent words reflects the cost of
incorrect predictions. Recent studies suggest that two ERPs components, the N400 and frontal positivity,
might reflect these two processes. In the work of Federmeier et al. (2007), although semantic constraint
and cloze probability were confounded for high cloze completions, low cloze completions under
strongly and weakly constrained frames were carefully matched for cloze probability. This critical
contrast revealed no N400 difference between these two conditions, and suggested no additional effect
of sentential constraint on the N400. However, a selective enhancement of the frontal positivity for low
cloze completions that were embedded in strongly constrained sentences was found. Thus, the authors
suggested that the frontal positivity appeared to reflect the disconfirmation of semantically based
predictions. The difference in the response to less expected words might be caused by the presence of a
preferred competitor in the strong constraint condition. In this circumstance, additional resources might
be needed to override or suppress a strong prediction for a different word or concept. Similar conclusions were reached by demonstrating the contextual constraining effect on sentence completion, suggesting that in strongly constrained sentence contexts, the less expected the word, the larger the frontal
positivity (DeLong et al., 2011). DeLong et al. (2011) also reported a similar observation and proposed
that the N400 and frontal positivity might reflect the benefits of confirmed predictions and the costs of
disconfirmed predictions about upcoming words in sentences or discourse.
Our study, on the other hand, demonstrated an interaction between semantic constraint and cloze
probability for the N400, indicating that the N400 might reflect a joint effect for both the benefit and
cost of the prediction. ERPs elicited by the pairing nouns revealed different patterns of cloze probability
on the N400 when completing the strongly and weakly constrained classifiers. For the weakly constrained condition, there was a graded effect of cloze probability (High < Low < Imp) on the N400. For
the strongly constrained condition, highly expected nouns also elicited a smaller N400 than did both
low cloze and implausible nouns. However, there was no significant difference between low cloze and
implausible words (High < Low ¼ Implausible). The reduced N400 for high cloze words, irrespective of
highly constrained or weakly constrained conditions, can be interpreted as a benefit of contextual
information that facilitates the semantic processing of expected words. However, the difference between low cloze and implausible words was only significant in weakly constraint condition. This might
be due to the strongly constrained classifier forming a strong prediction for the best completion.
Therefore, the low cloze noun showed a greater cost in the strongly constrained than in the weakly
constrained condition, and elicited an enhanced N400 that made it act like an implausible noun. This is
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
53
further supported by the post-hoc comparison, which revealed a constraint effect for low cloze nouns,
but not for high cloze probability or implausible nouns. The plausible but unexpected nouns following
the strongly constrained classifiers elicited larger N400 amplitudes than those following the weakly
constrained classifiers did. Therefore, the constraint effect for low cloze words suggests such a “cost”
only emerges in the processing of low cloze nouns in strongly constrained condition.
In addition, if frontal positivity reflects the cost associated with the violation of expectance in
strongly constrained contexts or recovery from the violation, implausible or low cloze nouns should
elicit a greater positivity than high cloze nouns, especially in the strongly constrained condition. This is
not the case in our findings. Our data revealed that the high cloze nouns elicited a greater positivity
than low cloze nouns, especially in the strongly constrained condition. In fact, the nature of the frontal
positivity is not yet precisely understood. Some studies have failed to find a constraint effect on frontal
positivity. For example, Wlotko and Federmeier (2007) used the same stimuli as in their prior study
(Federmeier et al., 2007), but failed to observe a constraint effect on the frontal positivity when stimuli
were lateralized to either the right or left visual fields. Thornhill and Van Petten (2012) asked participants to read a set of strongly and weakly constrained sentence frames that were completed by either
the best completion, the low cloze ending that was semantically related to the best completion, or the
low cloze ending that was semantically unrelated to the best completion. They also found no difference
for the low cloze endings of both strongly and weakly constrained sentence frames.
In summary, by utilizing the characteristics of Chinese classifier-noun agreement, the current
experimental design allowed us to evaluate how readers make use of the semantic constraint of
classifiers to predict the following noun as well as how such a constraint modulates the processing of
the following nouns. The data indicates that, when reading a strongly constraining classifier, readers
tend to form a strong prediction for its following noun. Therefore, the brain response elicited by the
classifier showed a reduced frontal negativity. When encountering the plausible but unexpected noun
(low cloze noun), there was an extra cost for such a wrong prediction, demonstrated by a similar increase in N400 amplitude. As for reading weakly constrained classifiers, readers tended to activate a set
of possible candidates, reflected in the enhanced frontal negativity elicited by the classifier and the
graded cloze probability effect on the N400 elicited by the pairing nouns. However, this study failed to
observe the constraint effect on frontal positivity. Further studies are needed to clarify the nature of
frontal positivity.
Acknowledgments
This study was funded by National Science Council (NSC 101-2628-H-001-006-MY3). We thank
Kara Federmeier for her helpful suggestions, and Chun-Hsien Hsu, Ying-Ying Zheng, Chih-Ying Huang,
and Yo-Ning Liu for technical support during the collection of the experiment data. We thank Ya-Ning
Chang, Chih-Ting Chang, Pei-Chun Chao, and two anonymous reviewers for their valuable comments
and suggestions on the earlier version of this paper.
References
Ahrens, K. (1994). Classifier production in normals and aphasics. Journal of Chinese Linguistics, 22, 202e247.
Barber, H., Vergara, M., & Carreiras, M. (2004). Syllable-frequency effects in visual word recognition: evidence from ERPs.
Neuroreport, 15, 545e548.
van Berkum, J. J., Hagoort, P., & Brown, C. M. (1999). Semantic integration in sentences and discourse: evidence from the N400.
Journal of Cognitive Neuroscience, 11, 657e671.
Chao, Y. R. (1968). A grammar of spoken Chinese. University of California Press.
Committe, N. L. (1997). Handbook of common quantifiers (Vol. 17). Taipei, Taiwan: National Languages Committe.
Dambacher, M., Goellner, K., Nuthmann, A., Jacobs, A., & Kliegl, R. (2008). Frequency and predictability effects on event-related
potentials and eye-movements. International Journal of Psychology, 43, 46e47.
Dambacher, M., & Kliegl, R. (2007). Synchronizing timelines: relations between fixation durations and N400 amplitudes during
sentence reading. Brain Research, 1155, 147e162.
Dambacher, M., Kliegl, R., Hofmann, M., & Jacobs, A. M. (2006). Frequency and predictability effects on event-related potentials
during reading. Brain Research, 1084, 89e103.
DeLong, K. A., Urbach, T. P., Groppe, D. M., & Kutas, M. (2011). Overlapping dual ERP responses to low cloze probability sentence
continuations. Psychophysiology, 48, 1203e1207.
DeLong, K. A., Urbach, T. P., & Kutas, M. (2005). Probabilistic word pre-activation during language comprehension inferred from
electrical brain activity. Nature Neuroscience, 8, 1117e1121.
54
C.-J. Chou et al. / Journal of Neurolinguistics 31 (2014) 42e54
Duffy, S. A., Henderson, J. M., & Morris, R. K. (1989). Semantic facilitation of lexical access during sentence processing. Journal of
Experimental Psychology. Learning, Memory, and Cognition, 15, 791e801.
Federmeier, K. D. (2007). Thinking ahead: the role and roots of prediction in language comprehension. Psychophysiology, 44,
491e505.
Federmeier, K. D., & Kutas, M. (1999). A rose by any other name: long-term memory structure and sentence processing. Journal
of Memory and Language, 41, 469e495.
Federmeier, K. D., Mai, H., & Kutas, M. (2005). Both sides get the point: hemispheric sensitivities to sentential constraint.
Memory and Cognition, 33, 871e886.
Federmeier, K. D., Wlotko, E. W., De Ochoa-Dewald, E., & Kutas, M. (2007). Multiple effects of sentential constraint on word
processing. Brain Research, 1146, 75e84.
Fodor, J. A. (1983). The modularity of mind: An essay on faculty psycholgy. The MIT Press.
Gao, M. Y., & Malt, B. C. (2009). Mental representation and cognitive consequences of Chinese individual classifiers. Language
and Cognitive Processes, 24, 1124e1179.
Hoeks, J. C., Stowe, L. A., & Doedens, G. (2004). Seeing words in context: the interaction of lexical and sentence level information
during reading. Cognitive Brain Research, 19, 59e73.
Hsu, C. H., Tsai, J. L., Lee, C. Y., & Tzeng, O. J. (2009). Orthographic combinability and phonological consistency effects in reading
Chinese phonograms: an event-related potential study. Brain and Language, 108, 56e66.
Huang, C. R., & Chen, K. J. (1998). Academia Sinica balanced corpus (version 3). Taipei,Taiwan: Academia Sinica.
King, J. W., & Kutas, M. (1995). Who did what and when? using word- and clause-level ERPs to monitor working memory usage
in reading. Journal of Cognitive Neuroscience, 7, 376e395.
Kliegl, R., Grabner, E., Rolfs, M., & Engbert, R. (2004). Length, frequency, and predictability effects of words on eye movements in
reading. European Journal of Cognitive Psychology, 16, 262e284.
Kuo, J. Y., & Sera, M. D. (2009). Classifier effects on human categorization: the role of shape classifiers in Mandarin Chinese.
Journal of East Asian Linguistics, 18, 1e19.
Kutas, M., & Hillyard, S. A. (1980a). Event-related brain potentials to semantically inappropriate and surprisingly large words.
Biological Psychology, 11, 99e116.
Kutas, M., & Hillyard, S. A. (1980b). Reading senseless sentences: brain potentials reflect semantic incongruity. Science, 207,
203e205.
Kutas, M., & Hillyard, S. A. (1984). Brain potentials during reading reflect word expectancy and semantic association. Nature, 307,
161e163.
Lee, C. L., & Federmeier, K. D. (2006). To mind the mind: an event-related potential study of word class and semantic ambiguity.
Brain Research, 1081, 191e202.
Lee, C. L., & Federmeier, K. D. (2009). Wave-ering: an ERP study of syntactic and semantic context effects on ambiguity resolution for noun/verb homographs. Journal of Memory and Language, 61, 538e555.
Lee, C.-Y., Liu, Y.-N., & Chou, C. J. (2013). Contextual predictability facilitates early orthographic processing and semantic integration in visual word recognition: an Event-related potential study. In F. Shi, & G. Peng (Eds.), East Flows the Great River:
Festschrift in Honor of Prof. William S-Y. Wang’s 80th birthday (pp. 169e186). Hong Kong: City University of Hong Kong Press.
Lee, C. Y., Liu, Y. N., & Tsai, J. L. (2012). The time course of contextual effects on visual word recognition. Frontiers in Psychology, 3,
285.
Nieuwland, M. S., & Van Berkum, J. J. (2006). Individual differences and contextual bias in pronoun resolution: evidence from
ERPs. Brain Research, 1118, 155e167.
Rayner, K., Ashby, J., Pollatsek, A., & Reichle, E. D. (2004). The effects of frequency and predictability on eye fixations in reading:
implications for the E-Z reader model. Journal of Experimental Psychology: Human Perception and Performance, 30, 720e732.
Rugg, M. D., Allan, K., & Birch, C. S. (2000). Electrophysiological evidence for the modulation of retrieval orientation by depth of
study processing. Journal of Cognitive Neuroscience, 12, 664e678.
Saalbach, H., & Imai, M. (2007). Scope of linguistic influence: does a classifier system alter object concepts? Journal of Experimental Psychology: General, 136, 485.
Schwanenflugel, P. J., & Shoben, E. J. (1985). The influence of sentence constraint on the Scope of facilitation for upcoming
words. Journal of Memory and Language, 24, 232e252.
Srinivasan, M. (2010). Do classifiers predict differences in cognitive processing? A study of nominal classification in Mandarin
Chinese. Language and Cognition, 2, 177e190.
Stanovich, K. E., & West, R. F. (1981). The effect of sentence context on ongoing word recognition: tests of a two-process theory.
Journal of Experimental Psychology: Human Perception and Performance, 7, 658e672.
Tai, J. H. (1992). Variation in classifier systems across Chinese dialects: towards a cognition-based semantic approach. Chinese
Language and Linguistics, 1, 587e608.
Taylor, W. L. (1953). “Cloze Procedure”: a new tool for measuring readability. Journalism Quarterly, 30, 415e433.
Thornhill, D. E., & Van Petten, C. (2012). Lexical versus conceptual anticipation during sentence processing: frontal positivity and
N400 ERP components. International Journal of Psychophysiology, 83, 382e392.
Tien, Y. M., Tzeng, O. J. L., & Hung, D. L. (2002). Semantic and cognitive basis of Chinese classifiers: a functional approach.
Language and Linguistics, 3, 101e132.
Van Petten, C., & Kutas, M. (1990). Interactions between sentence context and word frequency in event-related brain potentials.
Memory and Cognition, 18, 380e393.
Van Petten, C., & Kutas, M. (1991). Influences of semantic and syntactic context on open- and closed-class words. Memory and
Cognition, 19, 95e112.
Van Petten, C., & Luka, B. J. (2012). Prediction during language comprehension: benefits, costs, and ERP components. International Journal of Psychophysiology, 83, 176e190.
Wlotko, E. W., & Federmeier, K. D. (2007). Finding the right word: hemispheric asymmetries in the use of sentence context
information. Neuropsychologia, 45, 3001e3014.
Zhang, S., & Schmitt, B. (1998). Language-dependent classification: the mental representation of classifiers in cognition,
memory, and ad evaluations. Journal of Experimental Psychology: Applied, 4, 375.

Proyecto docente

ENGLISH TEST - TUESDAY, 10 FEBRUARY GRADE 4 1

Additional potential constraint in a reverse Monte

BAHASA INGGERIS