Chapter 3. Experiment 2: The influence of verb bias on ORC processing
3.1.2. Experimental design
This study aimed to manipulate three types of verb bias, Direct Object (DO),
Sentential Complement (SC), and Equilibrium Balanced (EQ) bias verb, which are
followed by a SRC, leading to the following syntactic structure of the target sentences:
Subject + DO/SC/EQ-bias main verb + SRC structure (RC verb + embedded RC noun +
RC marker DE + head noun). In order to make the sentences with SC-bias verb more
complete, in SC-bias condition, the SRC was followed by a verb, leading to the
structure of sentential complement. The current experiment shared the same verbs with
the first experiment. The classification of verb bias has already been defined in the first
experiment.
The construction of target sentences and fillers
The present experiment consisted of 41 target sentences with the manipulation of 14 DO-bias, 13 SC-bias, and 14 EQ-bias verbs as well as 42 fillers (Table2). Each
critical region of the SRC, were matched for word length, word frequency and the associations between critical words with no differences between conditions. The
embedded RC noun, RC verb, and head noun were all two-character words. The word frequency of 1st noun and head noun in each condition was low-medium word
frequency which has a frequency of 65 per million with no differences between groups
(embedded RC noun: F = 1.0, p > .05; head noun: F= 0.99, p > .05). To further control
the word predictability, the association between words were computed using word2vec
(Mikolov et al., 2013). The values of the association between words, , as shown in Table
3, including the association between 1st noun and RC verb (p > .05), and the association
between head noun and main verb ( F= 1.461, p >.05) , RC verb ( F= 1.11, p >.05), or
embedded RC noun( F= 0.328, p >.05 has to be lower than the value 0.3 (more than 0.3:
high associations between words, 0.2-0.1: medium-low associations, lower than 0.1:
low associations) with no differences between groups.
Additionally, 42 filler sentences were also created for preventing the participants
from developing strategies, including three syntactic structures—a. simple SVO
structure: subject + main verb + (adjectives) object b. subject + main verb + sentential
complement with embedded ORC/SRC, and c. subject + main verb + sentential
complement (Table 2).
Table 3.2. Examples of target and filler sentences
Target sentences
Main Clause Subjective Relative Clause (SRC) Sentential Complement
他 EQ-bias:
看見
招募 志工 的 導演
Types of filler sentences Examples subject + main verb + (adjectives) object 他蒐集證據。
subject + main verb + sentential
complement with embedded ORC/SRC
他懷疑老師讚美的學生作弊。
subject + main verb + sentential complement
他聽說產品出現瑕疵。
Table 3.3. The characteristics of the stimuli Frequency (per million words)
One-third of the experimental sentences were followed by one true/false comprehension
question to ensure that participants pay attention on reading comprehension during the
experiment.
3.1.3 Predictions
Based on the result of the Experiment 1 in which verb bias did influence the online
ORC processing, it was expected that verb bias would also affect the SRC processing.
Inasmuch as the syntactic structure of target sentences adopted in the current
experimental design violates the typical SVO word order in Mandarin, the role of word
order should also be taken into consideration for the evaluation of verb bias effect in the
present study. Two general hypotheses were made as follows. Firstly, if only verb bias
plays the role in SRC processing, it was expected that SRC following DO-bias verb
would be more difficult to process than that following SC-bias verb. Secondly, if both
“word order” and “verb bias” influence the SRC processing, processing difficulty might
arise in either condition since all the main verbs were followed by the other verb (RC
verb).
In addition, this processing difficulty may be shown on RC verb and head noun.
Predictions on the RC verb and head noun for each condition were listed as follows:
I. RC verb
Considering the verb bias effect demonstrated in the ORC processing, verb
bias effect in the SRC processing may be first reflected by RC verb in the contrast
between DO-bias and SC-bias conditions. RC verb in the sentence with DO-bias
verbs were expected more likely to be followed by a direct object and thus may
elicit a greater positivity of P600 to reflect the grammatical violation. Nevertheless, if “word order” also plays a role, RC verb in the sentence with SC-bias verbs may
also cause processing difficulty since it can be assigned by various grammatical
roles and which may result in a greater negativity of N400.
II. Head noun
While encountering a DO-bias verb, parsers would try to search for the direct
object for the main verb. Larger N400 on head noun under DO-bias condition than
that under SC-bias condition might be expected to reflect such a processing cost.
However, since SC-bias verb is more likely to followed by sentential
complements, the head noun might be mistaken as the subject of the sentential
complement or the object of the main verb. Such a processing difficulty may result
in thematic-role ambiguity and higher working memory demand and indexed by
the greater negativity of frontal negativity.
Aside from DO-bias and SC-bias verbs, EQ-bias verbs do not have a clear tendency of taking either more direct objects or sentential complements. The
sentences with EQ-bias verbs may serve as a baseline, as compared to the other
two conditions to see if they may exhibit a similar processing pattern to either
DO-bias or SC-DO-bias condition.
In order to delve into the role that each verb bias plays in the ORC processing,
differences between conditions were compared – SC-DO, SC- EQ, and DO-EQ bias
contrast.
3.1.4 Procedure
The procedure of the current study is as same as that of the first experiment
+
3.1.5 EEG recording and data analysis
The current study follows the same EEG recording and data analysis as the first
experiment.
3.2 Result
Three participants were excluded from further analysis due to the insufficient valid
trails; therefore, this study analyzed the total data of twenty-eight participants.
3.2.1 Accuracy of comprehension test
The overall accuracy of comprehension test was 95% (SD=0.04, range: 92%-100%),
showing that participants did not have difficulty understanding the sentences and had
pay attention in the experiment.
3.2.2 Statistical analysis for ERPs data:
cluster-based random permutation analysis
To evaluate the temporal and topographical differences between conditions (SC vs.
DO, SC vs. EQ, and DO vs. EQ), the cluster-based random permutation analysis was
conducted on each of the following critical regions – RC verb, embedded RC noun,
DE and head noun – for the mean amplitudes of two epochs, N400 from 250 to 500ms,
frontal negativity from 500 to 1000ms. The procedure of cluster-based random
permutation analysis was explained in details in Chapter2.
3.2.3 ERPs result (n = 28)
The cluster-based permutation analysis
Contrasts on RC verb
Figure 3.1 showed the grand-averaged ERMs waveforms elicited by RC verb in
three types of verb bias conditions. Visual inspection shows that RC verb in the
sentence with SC-bias verb elicited larger N400 than that in the sentence with DO-bias
verb, but it elicited a larger positive-going waveform than that in DO-bias condition in
the later time window. For the analysis of peak latency in DO-EQ and SC-EQ contrast,
a significant delay in N400 latency was shown in both DO-EQ (p < 0.01) and SC-EQ (p
< 0.01) contrast. It was found an 50ms delay in N400 peak latency on the EQ-bias
condition as compared to the DO-bias and SC-bias conditions. The cluster-based
random permutation analysis (Figure 3.2) revealed that SC-DO contrast yielded
significant negative cluster in the time window of 250-500ms (318-500ms, p<0.01).
DO-SC contrast yielded significant positive cluster in left fronto-central region
(773-956ms, p<0.01) in later time window. Both DO-bias and SC-bias condition yielded
larger positive clusters than the EQ-bias condition.
Figure 3.1. Grand averaged ERMs of the RC verb for DO-bias, SC-bias, and EQ- bias conditions.
Figure 3.2 Topographic maps of RC verb for SC-DO contrast in N400(250-500ms),
and DO-SC contrast in late time window (500-1000ms).
Note. Asterisks represent the significant differences for the contrasts.
Contrast on embedded RC noun
Figure 3.3 presented the grand-averaged ERMs waveforms elicited by embedded
RC noun in three types of verb bias conditions. The result showed that no differences
were found in DO-SC contrast; however, embedded RC noun in both the sentence with
SC-bias verb and DO-bias verb elicited larger negative-going waveform than that in
sentence with EQ-bias verb. The cluster-based random permutation analysis (Figure
3.4) revealed that in the time window of 250-500ms, there was no difference in DO-SC
contrast, but significant negative cluster in DO-EQ (373-500ms, p<0.001) and SC-EQ
were found in both DO-EQ contrast 747ms p<0.001) and SC-EQ contrast
(500-854ms, p<0.001).
Figure 3.3. Grand averaged ERMs of embedded RC noun for DO-bias, SC-bias, and EQ-bias conditions.
Figure 3.4. Topographic maps of embedded RC noun for DO-SC, DO-EQ, and SC-EQ contrasts in N400 (250-300ms) and late time window (500-1000ms).
Note. Asterisks represent the significant differences for the contrasts.
Contrast on DE
Figure 3.5 showed the grand-averaged ERMs waveforms elicited by DE. DE in the
sentence with DO-bias verb elicited larger N400 than that in the sentence with SC-bias
verb and EQ-bias verb. The cluster-based random permutation analysis (Figure 3.6)
revealed DE in DO-bias condition yielded significant negative clusters than that in
SC-bias condition (310-500ms, p< 0.01) and EQ-SC-bias condition (372-500ms, p<0.01) in the
time window of 250-500ms, but no significant negative clusters in DO-SC contrast and
DO-EQ contrast were found in the later time window.
Figure 3. 5. Grand averaged ERMs of DE for DO-bias, SC-bias, and EQ-bias conditions.
Figure 6. Topographic maps of DE for DO-SC, DO-EQ, and SC-EQ contrasts in N400 (250-500ms) and late time window (500-1000ms).
Note. Asterisks represent the significant differences for the contrasts.
Contrast on head noun
Figure 3.7 presented the grand-averaged ERMs waveforms elicited by head noun.
Head noun in the sentence with DO-bias verb elicited larger negative-going waveform
than that in the sentence with SC-bias verb that in the time window of 250-500ms;
however, head noun in SC-bias condition elicited larger positive-going waveform than
that in DO-bias condition in the later time window of 500-1000ms. The cluster-based
random permutation analysis (Figure 3.8) revealed that in the time window of
250-500ms, head noun in the DO-bias condition yielded significant larger negative cluster
than that in the SC-bias condition (324-493ms, p<0.001) and EQ-bias condition
(334-500ms, p< 0.01). Head noun in SC-bias condition elicited larger positive cluster in
posterior region than that in EQ-bias condition (329-444ms). Yet, in the later time
window, head noun in the DO-bias condition yielded significant larger positive cluster
than that in the SC-bias condition (540-879ms, p<0.001). No significant clusters were
found in DO-EQ contrast.
Figure 3.7. Grand averaged ERMs of head noun for DO-bias, SC-bias, and EQ-bias conditions.
Figure 3.8. Topographic maps of head noun for DO-SC, DO-EQ, and SC-EQ contrasts in N400 (250-500ms) and late time window (500-1000ms).
Note. Asterisks represent the significant differences for the contrasts.
3.3 Discussion- The incremental influence of verb bias on SRC processing
The present study aimed to examine the verb bias effect on the incremental SRC
processing on P600, frontal positivity, frontal negativity, and N400 components. The
cluster-based permutation analysis was performed to characterize the spatial and
temporal dynamics of the verb bias effect on SRC processing by contrasting DO-SC,
DO-EQ and SC-EQ bias conditions. Consistent with the result of the Experiment1, the
result demonstrated the verb bias effect on SRC processing. It was first reflected on the
RC verb in DO-SC contrast -- larger frontal positivity on RC verb following DO-bias
verb and larger N400 on RC verb following SC-bias verb. Such an effect on the initial
structure of RC lasted on the subsequent RC marker DE and head noun in the DO-SC
contrast. Moreover, the result also called attention to the significant role of word order
in Mandarin.
Table 3.4. Summary of the ERPs result on DO-SC contrast
The result of DO-SC contrast was summarized in Table 3.4 The marked ones were
specifically pointed out for the following discussion. (For the complete result, please
refer to Appendix III). Followings are the detailed discussions over the ERPs result on
RC verb, DE, and head noun with the focus on the DO-SC contrast.
The processing difficulty between SC-bias verb and DO-bias verb condition were
first shown on the RC verb, as indexed by late frontal positivity in DO-SC contrast and
larger N400 in SC-DO contrast. RC verb in DO-bias condition elicited larger frontal ERP
component
RC verb 1st noun DE Head noun
N4000 339-500ms 300-463ms 310-500ms 324-493ms late
components
773-956ms (positivity)
No difference No difference 540-879ms (negativity)
positivity in the late time window (773-956ms) than that in SC-bias condition. Since the
markers for the tense and aspect are absent in Mandarin, two interpretations of the syntactic function can be made on the processing of RC verb following DO-bias verb.
One is the “verb”, the other is the “normalization of the verb” functioning as the subject
or noun modifier. The dominant word order of Mandarin syntactic structure is SVO. If
word order precedes over the verb bias effect, RC verb interpreted as the “verb” violates
the Mandarin word order. It might be reflected by typical P600. However, if both verb
bias and word order play important roles, RC verb which can be seen as a “verb” and
“participle” can be interpreted as a non-preferred structure. The processing difficulty of
non-preferred but grammatical structure might be reflected by frontal positivity. The
result corresponded to the second hypothesis. RC verb in the DO-bias condition is a less
preferable sentence pattern. Less preferable sentence structure reflected by frontal
positivity was consistent with the result of RC verb in experiment 1 and the prior
studies (Kann & Swaab, 2003; Leone-Fernandez et al, 2011).
However, processing difficulty on RC verb in the sentence with SC-bias verb was
reflected by larger N400 on RC verb in SC-bias condition than that in DO-bias
condition. Sentential complement following the SC-bias verb can be represented by
three syntactic structures. Firstly, in Mandarin, the subject can be omitted, so the main
verb can be followed by a sentential complement without subject( e.g.他擔心(他)提出
證據會被報復。) The RC verb in this case could be interpreted as the main verb of the sentential complement. Secondly, the main verb can be followed by the sentence which
begins with the normalized verb (e.g.他擔心提出證據會對公司造成傷害。) RC verb
“提出” could be seen as the normalization of the verb. Thirdly, the RC structure as the embedded subject of the sentential complement or as the direct object. Thus, multiple
syntactic roles could be assigned to this RC verb in SC-bias condition. The variability
of syntactic structure following the SC-bias verb provides an advantage for processing
ongoing syntactic structure; therefore, the parsers would not have difficulty.
Nevertheless, although multiple grammatical roles can be assigned to the “verb” (RC verb) following SC-bias verb, the appearance of this “second verb” (RC verb) following
the main verb still exhibited processing difficulty, as reflected by the larger N400
elicited by RC verb in SC-bias condition than that in DO-bias condition. Hence, the
word order SVO in Mandarin might also play a significant role in SRCs processing to
account for the difficulty of integrating the “second verb” (RC verb) to the preceding
information (main verb: bias verb). The word order and the characteristics of
SC-bias verb competed in terms of the role in sentence processing, bringing about the
integration difficulty on the RC verb following SC-bias verb.
This verb bias effects were also evident on the subsequent RC marker DE and head
noun, as those demonstrated in the Experiment1. In Experiment1, the verb bias effect
which was first shown on the RC verb lasts on its subsequent words, RC marker DE and
head noun, as respectively reflected by larger N400 in DO-SC contrast and larger
frontal negativity in DO-SC contrast. The long-lasting effects also appeared in current
study. Firstly, both RC marker DE and head noun in DO-bias condition elicited larger
N400 than that in SC-bias condition. In parsing the sentence with DO-bias verb, parsers
tried to look for the actual direct object for the main verb. Before reaching the head
noun, the parsers experienced one re-analysis; that is, the previous embedded RC noun
was not the direct object of the main verb, but the head noun. The costs arise from the
process in which the accruing semantic or syntactic contextual representation must be
overridden or revised. When parsing the head noun in DO bias condition, the parsers
finally found the correspondent direct object for the main verb. This cost was reflected
by larger N400 on DE and head noun in DO-SC contrast. Secondly, head noun in
SC-bias condition elicited a larger sustained negativity from 400ms than that in DO-SC-bias
condition. It implied that parsers experienced different processing difficulty in parsing
SRC sentence with SC-bias verb than that with DO-bias verb. The parsers constantly
looked for the corresponding direct object for the main verb when parsing DO-bias
condition. In contrast, when reading sentence with SC-bias verb, parsers expected that
the sentential complement follows the SC-bias verb, so they had to find the subject of
the sentential complement and maintain the information that the RC intended to modify
(e,g. subject). However, there were two possible thematic role assignments on the head
noun. One is SRC as the direct object of the main verb, leading to the interpretation of
the head noun as the direct object. The other is SRC as the subject of the sentential
complement, leading to the interpretation of the head noun as the subject. Thus, the
sustained negativity elicited by head noun in SC-bias condition than that in DO-bias
condition might be related to the thematic-role ambiguity resulting in the cognitive
demands for recruitment of additional memory resources.
Sustained negativities have been conceived as indexing cognitive demands (King
& Kutas, 1995; Yang et al., 2010). Prior studies have found that the structure which is
more difficult for thematic role assignments and requires high memory demand
produces frontal negativity in both English (King & Kutas, 1995) and Chinese (Yang et
al., 2010). For instance, the study of King & Kutas (1995) indicated that a larger
sustained frontal negativity on the early region of ORC (e.g. The reporter who “the
senator” harshly attacked admitted the error.) as compared to SRC (e.g. The reporter
who “harshly attacked” the senator admitted the error.) reflected the larger memory
who “harshly attacked” the senator admitted the error.) reflected the larger memory