DISTRIBUTION OF DIFFERENT TYPES OF HESITATION PHENOMENA 117
Distribution of Different Types of Hesitation
Phenomena and Their Patterns of Correlations (1)
Tomohito Ishikawa
Abstract
The present study addressed two research questions. The first question was: Is the distribution of different types statistically significantly different from uniform?, which was originally posed by Shriberg (1994) in her study of first language (L1) hesitation phenome- na. The second research question was: What are the relationships between different types of hesitation phenomena as well as the relationships between different types of hesitation phenomena and perceptions of task difficulty? In order to answer these research questions, a sub-corpus was analyzed, which was originally used in Ishikawa (2008). The results showed that similar to Shriberg's (1994) results, distributions of different types of hesitation phenomena were far from uniform. Several significant correlations were also found between hesitation phenomena and perceptions of task difficulty and
among different types of hesitation categories. Of particular signifi- cance were statistically significant correlations that are related to highly frequent categories of hesitation phenomena.
118
Second language (L2) task performance has recently been a major concern of L2 teachers and researchers. One of the possible reasons for this is ascribed to a recent focus on the effect of manipulating task complexity on L2 performance. Many researchers in fact made effort to describe various aspects of L2 speech performance such as accuracy, complexity, and fluency so as to examine the effect of manipulating task complexity and to find out possible underlying factors affecting those speech characteristics. Among the aforementioned dimensions of L2 speech, the present study focuses on L2 disfluency phenomena and examines their distributions of different types of disfluency phenomena and correlations with other possible factors.
According to Ellis and Barkhuizen (2005), fluency measures can be divided into two major categories. One category of utterance fluency (Segalowitz, 2010) is related to speed of speaking (i.e., temporal variables) and the other category is related to repair fluency. In contrast to the relatively straightforward definitions and operationalizations of the temporal aspect of L2 fluency such as speech rate, the repair aspect of L2 fluency needs consistent descriptive and classification schemes. In order to solve this problem, the present article adopts the disfluency classification system developed by Shriberg (1994; also see Bear, Dowding, Shriberg, & Price, 1993). Shriberg's descriptive scheme is relatively easy to apply to L2 data though minor modifications were made to her scheme in the present study, which will be mentioned later in this article.
The present study focused on distributions of seven categories of Shriberg's (1994) classification scheme of hesitation. They were: (1) deletion (DEL); (2) repetition (REP); (3) substitution (SUB); (4) insertion (INS); (5) filled pauses (FP); (6) hybrid (HYB); and (7) articulation (ART). Samples of these hesitation categories are provided in the methodology section.
In the present study, two research questions were put forth. The first
DISTRIBUTION OF DIFFERENT TYPES OF HESITATION PHENOMENA 119 question was: Is the distribution of different types statistically significantly different from uniform? The null hypothesis is that each type will occur equally; the alternative hypothesis is that some types are more likely than others. This research question and the hypotheses were addressed in Shriberg (1994), which used L1 corpus, and in her case the null hypothesis was rejected. In the present study, this research question was addressed by using a simple frequency measure and a ratio measure. The second research question was: What are the relationships between different types of hesitation phenomena as well as the relationships between different types of hesitation phenomena and perceptions of task difficulty? The second research question is more exploratory and descriptive. Thus, the goals of this article are three- fold: to describe and examine (1) distributions of different types of hesitation phenomena, and (2) correlations among different types of hesitation and their correlations with perceptions of task difficulty. The purpose of achieving these goals is to compare patterns of L1 and L2 phenomena and also provide descriptions of L2 hesitation phenomena, which, as far as the author knows, is not available to L2 researchers and teachers sufficiently. In order to do so, data on a task were used, which was originally used as a control task when examining the effect of manipulating task complexity on L2 speech performance (Ishikawa, 2008).
Methodology
Participants
The participants for the present study were college students learning English as an L2, whose Ll was Japanese (N = 24; male = 2, female = 22).
They were either English majors or English for Academic Purposes (EAP) Economics students at two universities in Tokyo. The participants'
120
proficiency levels ranged approximately between low to high intermediate.
The proficiency test scores of the English majors at a prestigious university were not available, but they participated in subject matter classes taught in English. The minimum score of the EAP Economics students at the other university was 450 on the paper-based TOEFL. Their ages ranged from 18 to 35 years old, with an average age of 22.2 years old. Their mean length of stay in English-speaking countries was approximately 8 months and the mean length of English instruction prior to this study was approximately 11 years.
Materials
Originally three pedagogic tasks were employed in Ishikawa (2008). In the current paper, only the control task (see Appendix A), which the results of the present study are based on, will be described. The task was a one-way monologic task. In performing it, the learner played the role of a company's general manager, who was in charge of a new section that was established "a week ago." The learner was asked to report to the company president about hypothetical human relationships of his/her section members. At the time of reporting, the president was not available, so after a three-minute planning time, the learner had to report to the president by leaving a message on the president's answering machine. In the task, the learner simply described the current human relationships between the section members represented on the task sheet. The nature of human relationship was represented by different types of lines on the task sheet; namely, the solid line (originally colored in blue) between section members represented a good human relationship, and the broken line (originally colored in red) a bad one. Essentially, necessary information to transfer to the company president (i.e., human relationships between the section members) was available in the learner's task materials.
Therefore, the learner's main task was to transfer given information to the
DISTRIBUTION OF DIFFERENT TYPES OF HESITATION PHENOMENA 121 president. In order to elicit sufficient L2 speech data for analysis, a setting was chosen, where the manager was in charge of four members rather than two or three members.
Procedures
The experiment was conducted on an individual basis. Before the experiment began, the researcher told the participant to perform three tasks, one of which was the task relevant to the present study. The researcher then passed the task-instructions sheet to the participant. Both instructions sheets were originally written in their first language (i.e., Japanese). The participant was also informed in advance that the answering machine would stop after three minutes but they did not have to keep talking for three minutes. After each planning session, the researcher read out the message on the answering machine ("Hi, you have reached xxx-xxx-xxxx. I cannot come to the phone right now. Please leave a message after the beep. Thanks."). A beep sound followed and the recording started. After completing the speaking task, the learner responded to a brief written questionnaire, which was created to assess the learner's subjective perceptions of task difficulty (see below for details of the questionnaire items).
Hesitation measures
The present study employed seven hesitation measures based on Sbhriberg's classification system (Shriberg, 1994). Her system is originally developed for same-turn disfluencies and is designed to handle only within- sentence disfluencies. Other speech phenomena not considered in her system include: silent pauses, pathological stuttering, discourse markers, uncorrected errors (Shriberg, 1994, p. 2, 6-7, chapter 4, and some specific cases are mentioned in this thesis). The coding methods described in the present article
were almost identical to Shriberg's though some minor modifications were made.
Shriberg distinguishes basic disfluency types from complex ones, which contain more than one disfluency episode (an example is given below).
Although Shriberg treats both simple and complex disfluencies separately, I decomposed complex disfluencies into their disfluency components and counted each disfluency episode separately. The decomposition was necessary in order to add consistency in counting the number of disfluencies, which previous studies did not clarify criteria about, which is criticized by Ellis and Barkhuizen (2005). A sample of complex disfluencies given by Shriberg (1994, p. 58) is:
he she she liked it
This is regarded as a complex disfluency because: (1) "he" is substituted by the fist "she," which is the output of the first repair; (2) the output of the first repair operation (i.e., "she") is also an input to the second repair operation. That is, the output of the first repair operation is repeated (i.e., repaired) by the second "she." Formally, the sample above is represented as follows:
he she she liked it [R[ s . s]. r ]
The two sets of square brackets in the above example indicate that two disfluency components are involved in this case (i.e., "he she" and "she she").
Such cases are called complex disfluencies because the first "she" is involved in two disfluency components. In the present study, however, I did not
DISTRIBUTION OF DIFFERENT TYPES OF HESITATION PHENOMENA123 distinguish simple and complex disfluencies as Shriberg did, and decomposed complex ones into simple ones. For instance, rather than treating the above example as a single complex disfluency, I treated it as two disfluencies or two disfluency episodes. In other words, the number of disfluency episodes is equal to the number of interruption points.
Following Shriberg's classification scheme of hesitation, seven disfluency measures were employed in the present study. Table 1 presents the seven disfluency categories and their actual samples from my corpus: 1) deletion (DEL); 2) repetition (REP); 3) substitution (SUB); 4) insertion (INS); 5) filled pauses (FP); 6) hybrid (HYB); and 7) articulation (ART). Shriberg (1994) also includes a category conjunction (CON); in the present study, this category was not analyzed partly because this is a category based on a grammatical category of conjunction. This means that even when conjunctions are repeated, the episodes were not coded as REP but CON. By eliminating the CON category, the present study respected patterns of hesitations; therefore, the above example was coded as REP not CON.
Task-difficulty questionnaire
The task-difficulty questionnaire was adopted from Robinson (2001b), who included items assessing interest, ability, motivation, stress, and perceived task difficulty. In addition to these, the present study included three items assessing concentration, response to time pressure, and anxiety. The questionnaire was used to check whether task complexity conceptualized by the task designer is matched by perceptions of difficulty on the part of the L2 learner.
1) Concentration (this task required concentration/this task did not require concentration)
Table 1 Shriberg's disfluency type and samples from my corpus taken from Ishikawa (2010) and modified by the author)
(Table 1 is
Disfluency type Samples from my corpus
DEL
so th- he can figure out what is going on [d-.]
and Kate and John s- ah ok I have to report the relationships between Kate John ah Nick and Cathy [cdddd-.f]
'but he but because of his lost of important document [cd .]
REP
Joe came late came late for a meeting [rr.rr]
so because of ah because of the Beth's showing up later for the meeting [rr.frr]
SUB
INS FP
~till Joe come came to the meeting [s.s]
'Mike was confusing confused about the next presentation [s.s]
Joe was working with secretary Ann ah af- before trouble [s-.s]
and she is actually she is a strict person about the time [rr.irr]
'Beth showed up later for the meeting the important meeting [rr.rir]
um we also have a problem with Cathy and John [.f]
HYB (example of s + i)
i^Ann did not have relationships good relationship with Bob [s.is]
ART 'this is nam Minami calling [r-- .r]
'secretary pa- Cathy [r-- .r]
Note. DEL = Deletions; REP = Repetitions; SUB FP = Filled pauses; ART = Articulations; HYB =
= Substitutions; INS = Hybrids.
Insertions;
2) Time pressure (I did not feel time pressure during task performance/I felt time pressure during task performance)
3) Anxiety (this task made me anxious/this task did not make me anxious) 4) Stress (I felt frustrated during task performance/I did not feel frustrated
during task performance)
5) Difficulty (this task was easy/this task was difficult)
6) Interest (this task was interesting/this task was not interesting) 7) Ability (I did not do this task well/I did this task very well)
8) Motivation (I want to do tasks like this/I don't want to do tasks like this)
DISTRIBUTION OF DIFFERENT TYPES OF HESITATION PHENOMENA 125 The learner was asked to respond to each item along a ten-point Likert scale (ranging from 0 to 9). The questionnaire items were written in the learners' L1. In coding the data, a score of 0 was assigned to the semantically most negative learner response to each questionnaire item (e.g., no time pressure, no anxiety) and a score of 9 to the semantically most positive
learner response (e.g., highest time pressure, highest anxiety) to each questionnaire item.
Results
Figure 1 shows distributions of different types of hesitation phenomena and their raw frequencies. In total, 388 disfluency phenomena are identified.
It is clear from Figure 1 that hesitation phenomena are not produced equally.
Not surprisingly, Filled Pauses (FP) is the most frequent hesitation phenomena. The second most frequent type of hesitation phenomena is Repetitions (REP), which is followed by Deletions (DEL) and Substitutions
(SUB). The occurrences of the categories of Insertions (INS), Articulations (ART) and Hybrids (HYB) are much less frequent.
82001173
a
g 150
7 U U
° 100
0 150
7 z0
FP REP DEL SUB INS ART HYB
Types of hesitation
Figure 1. Distributions of different types of hesitation phenomena DEL = Deletions; REP = Repetitions; SUB = Substitutions; INS = Insertions;
FP = Filled pauses; ART = Articulations; HYB = Hybrids.
126
As stated above, the first question was: Is the distribution of types statistically significantly different from uniform? The null hypothesis is that each type will occur equally; the alternative hypothesis is that some types are more likely than others. In order to answer the first research question, the Chi square analysis was conducted. The results of the analysis was statistically significant (x2(6) = 443.099, p<.01) . Subsequently, in order to locate statistically significant differences between the hesitation categories, Ryan's method was applied to the dataset, using Tanaka and Nakano's (2012) js-STAR 2012. Table 2 presents the results of the analyses. As is clear from Table 2, except for the three comparisons between INS, ART and HYB, all the comparisons were statistically significant.
Table 3 shows descriptive statistics for the occurrences of different types of hesitation phenomena per T-unit. Similarly to the previous results of distributions of different types of hesitation phenomena, the results of the occurrences of different types of hesitation phenomena per T-unit showed an identical order of occurrences of different types of hesitation phenomena, which is shown in Figure 2. Table 3 also shows that individual differences in the production of different types of hesitation phenomena can be observed in FP, REP, and DEL, as shown by relatively larger standard deviations.
In order to examine statistically significant differences, normality of the distributions were tested using the Shapiro-Wilk tests. The results showed that except for the measure of REP per T-unit, all the measures were not normally distributed. Therefore, Friedman's ANOVA was applied to the dataset. The result was statistically significant (x2(6)= 101.816, p<.01, effect size = 20.783) . Subsequently, individual Friedman's tests were applied to the dataset. Subsequently, two-tailed Wilcoxon tests were applied to the dataset with the significance level being adjusted to .00238 (= .05/21) as a result of the Bonferroni method. Table 4 summarizes the results of the two-
Table 2
DISTRIBUTION OF DIFFERENT TYPES OF HESITATION PHENOMENA 127 Results of Ryan's Method: Critical Ratio, P-Value, and Nominal Alpha Level
Comparison Critical ratio p-value Nominal alpha level
FP>REP FP>DEL FP>SUB FP>INS FP>ART FP>HYB REP>DEL REP>SUB REP>INS REP>ART REP>HYB DEL>SUB DEL>INS DEL>ART DEL>HYB SUB>INS SUB>ART SUB>HYB INS=ART INS=HYB ART=HYB
3.95 7.10 10.06 12.30 12.52 12.52 3.23 6.54 9.22 9.49 9.49 3.44 6.57 6.91 6.91 3.50 3.94 3.94 .29 .29 -.32
p<.0002 p<.0002 p<.0002 p<.0002 p<.0002 p<.0002 p<.0002 p<.0002 p<.0002 p<.0002 p<.0002 p=.0006 p<.0002 p<.0002 p<.0002 p=.0004 p<.0002 p<.0002 ns
ns ns
.01429 .00714 .00476 .00357 .00238 .00238 .01429 .00714 .00476 .00286 .00286 .01429 .00714 .00357 .00357 .01429 .00476 .00476 .00714 .00714 .01429
Note. DEL/T = Deletions per T-unit; REP/T = Repetitions per T-unit; SUB = Substitutions per T-unit; INS = Insertions per T-unit; FP/T = Filled pauses per T unit; ART/T = Articulations per T-unit; HYB/T = Hybrids per T-unit.
tailed Wilcoxon tests. The results showed that there were six non-significant pairs, which were between adjacent categories (e.g., FP/T vs. REP; REP vs.
DEL and so on) plus a single non-adjacent pair between INS/T versus HYB/
T.
Table 5 shows Spearman correlation coefficients of different types of hesitation phenomena per T-unit and perceptions of task difficulty. DEL/T
Table 3 Descriptive Statistics for the Occurrences of Different Types of Hesitation Phenomena per T-unit
M SD
DEL/T REP/T SUB/T INS/T FP/T ART/T HYB/T
.19 .33 .09 .02 .52 .02 .02
.14 .24 .07 .04 .39 .03 .05
Note. DEL/T = Deletions per T-unit; REP/T = Repetitions per T-unit; SUB = Substitutions per T unit; INS = Insertions per T-unit; FP/T = Filled pauses per T-unit; ART/T = Articulations per T-unit; HYB/T = Hybrids per T-unit.
Figure 2.
0.6
'E0 .5 H
a)
a0.4
E3
N E 0.3
O 0 0.2
t•. N p t 2 0.1
0
FP/T REP/T DEL/T SUB/T INS/T ART/T HYB/T
Distributions of occurrences of different types of hesitation phenomena per T-unit DEL/T = Deletions per T-unit; REP/T =
Repetitions per T-unit; SUB = Substitutions per T-unit; INS = Insertions per T-unit; FP/T = Filled pauses per T-unit; ART/T = Articulations per T-unit; HYB/T = Hybrids per T unit.
correlated statistically significantly with Anxiety (rs= .48, p = .02) and Difficulty (rs= .42, p =.04). This means that learners who felt anxious and had difficulties produced more deletions per T-unit. Motivation also correlated
DISTRIBUTION OF DIFFERENT TYPES OF HESITATION PHENOMENA 129 Table 4 Results of the Two-Tailed Wilcoxon Test
Comparison z p-value Effect size
FP/T=REP/T FP/T>DEL/T FP/T>SUB/T FP/T>INS/T FP/T>ART/T FP/T>HYB/T REP/T=DEL/T REP/T>SUB/T REP/T>INS/T REP/T>ART/T REP/T>HYB/T DEL/T=SUB/T DEL/T>INS/T DEL/T>ART/T DEL/T>HYB/T SUB/T=INS/T SUB/T>ART/T SUB/T>HYB/T INS/T=ART/T INS/T=HYB/T ART/T=HYB/T
-2 .625 -3.621 -4 .136 -4 .198 -4 .198 -4 .229 -2.767 -3 .638 -4 .168 -4 .108 -4 .109 -2.505 -3 .946 -3 .922 -3 .921 -2 .937 -3.239 -3 .096
- .271 - .350 - .421
ns p=.00029 p=.00004 p=.00003 p=.00003 p=.00002 ns p=.00028 p=.00003 p=.00004 p=.00004 ns p=.00008 p=.00009 p=.00009 ns p=.00120 p=.00196 ns
ns ns
- .739 - .844 - .857 -.857 -.863
- .743 - .851 - .839 -.839
- .805 - .801 - .800
- .661 - .632
Note. DEL/T = Deletions per T-unit; REP/T = Repetitions per T-unit; SUB = Substitutions per T-unit; INS = Insertions per T-unit; FP/T = Filled pauses per T-unit; ART/T = Articulations per T-unit; HYB/T = Hybrids per T-unit.
statistically significantly with REP/T (rs= -.46, p = .03), FP/T (rs= -.46, p = .02), and ART/T (rs= -.48, p = .02), which means that learners who produced fewer deletions, filled pauses, and articulations per T-unit felt more motivated to perform similar tasks in the future. SUB/T also correlated statistically
significantly with Time Pressure (rs= .40, p = .05), which means that learners who produced more substitutions per T unit felt more time pressure. Finally, HYB/T correlated statistically significantly with Concentration (rs= -.46, p =
130
Table 5 Spearman Correlation Coefficients of Different Types of Hesitation Phenomena per T-unit and Perceptions of Task Difficulty
CON TMP ANX FRS DIF INT ABL MTV
DEL/T REP/T SUB/T INS/T FP/T ART/T HYB/T
.13 - .09 - .27 .17 .05 .00
.24 .08 .40*
.37 .18 .02 .19
.21 - .02 .09 .18 .26 - .20
.23 .22 - .06 .24 -.11 .25 .01
.21 .11 .07 -.03
.08 - .09
.25 - .22 - .05 .07 -.30 - .11 .17
- .23 .16 - .05 .03 .18 .00 .00
- .10 - .46 - .19 -.29
.17 Note. *p<.05; N = 24; DEL/T = Deletions per T-unit; REP/T = Repetitions per T-unit; SUB = Substitutions per T-unit; INS = Insertions per T-unit; FP/T = Filled pauses per T-unit; ART/T = Articulations per T-unit; HYB/T = Hybrids per T-unit;
CON = Concentration; TMP = Time Pressure; ANX = Anxiety; FRU = Frustration;
DIF = Difficulty; INT = Interest; ABL = Ability; MTV = Motivation.
.02). This means that learners who concentrated on performing the task produced fewer hybrids per T-unit.
Table 6 shows Spearman correlation coefficients between different types of hesitation phenomena per T-unit. FP/T correlated statistically significantly with REP/T (rs= .53, p = .01) and SUB/T (rs= .41, p = .05), which means that learners who produced more filled pauses also produced repetitions and substitutions. ART/T also correlated statistically significantly with INS/T (rs=
.41, p = .02). This means that learners who produced more articulations also produced more insertions.
Discussion
The goals of the present study were to describe and examine (1) distributions of different types of hesitation phenomena, and (2) correlations
among different types of hesitation and their correlations with perceptions of
DISTRIBUTION OF DIFFERENT TYPES OF HESITATION PHENOMENA 131 Table 6Spearman Correlation Coefficients Between Different Types of
Hesitation Phenomena per T-unit
REP/T SUB/T INS/T FP/T ART/T HYB/T
DEL/T REP/T SUB/T INS/T FP/T ART/T
.28 .03
.24
.12 .28 - .21
.19
.05
.17 .38 - .04
.48 -.02
.08 .14 .25 - .19 .11 -.19
Note. *p<.05; **p<.01; N = 24; DEL/T = Deletions per T-unit; REP/T = Repetitions per T-unit; SUB = Substitutions per T-unit; INS = Insertions per T-unit;
FP/T = Filled pauses per T-unit; ART/T = Articulations per T-unit; HYB/T = Hybrids per T-unit.
task difficulty. Regarding the first goal, the results of the distribution of different types of L2 hesitation phenomena revealed that the relative order of the occurrences of different types of L2 hesitation phenomena was far from uniform and quite similar to the one of L1 hesitation phenomena. Although direct comparisons between the L1 and the L2 data are impossible, the difference between L1 and L2 may be of quantitative kind rather than qualitative one. If this is the case, when bilingual speakers speak in two languages, there should be significant high correlations between hesitation phenomena in the two languages plus more frequent hesitation phenomena in
the L2 than in the L1 (see Segalowitz, 2010).
Another remarkable finding was that filled pauses were the most frequent kind of hesitation phenomena. There seem to be possible reasons for the high frequency of the use of filled pauses. In the L1 literature, filled pauses are claimed to have an important communicative function to announce that the speaker is initiating delays in speaking and some researchers even consider them as "conventional English words" (Clark & Fox Tree, 2002). In the case of the present study, the speaker produced not only English filled pauses but
132
also Japanese filed pauses. Intuitively, the functions of both English and Japanese filled pauses are quite similar, such as using filled pauses to signal processing delays to implicate that the speaker is formulating speech plan, search appropriate lexical items and so on. If this is the case, it means that the speaker seems to learn how to use filled pauses in the L2 or even transfer L1 filled pauses into L2 communication in order to serve the communicative function of announcing the speaking delay although whether they are using filled pauses appropriately is another issue. Additionally, relatively larger standard deviation of the use of filled pauses may imply that there are individual differences issues in the use of filled pauses.
The second most frequent hesitation phenomena were self-repetitions.
Dornyei and Kormos (1998) classify repetitions as one of the two major categories of stalling mechanisms (i.e., lexicalized/non-lexicalized pauses and repetitions). From this perspective, the use of repetitions and the
aforementioned filled pauses are quite similar (i.e., stalling mechanism).
From a psycholinguistic perspective, the use of both filled pauses and repetitions serves as time-saving devices, which also increase available attentional demands, as a result of which, the speaker can for instance search lexical items, review what they have said, plan ahead. Given the similar functions, what is not clear is why or when the speaker chooses to use filled pauses or repetitions.
The third most frequent type of hesitation was deletions. Deletions can be deletions of part of what the speaker said or false starts. Given that L2 speakers generally lack L2 linguistic knowledge and automaticity, they tend to select problematic items more often and tend to delete them. These three major types of hesitation phenomena accounted for approximately 90% of the whole hesitation phenomena in the present study.
The results of correlation analyses also suggested some links between
DISTRIBUTION OF DIFFERENT TYPES OF HESITATION PHENOMENA 133 hesitation phenomena and post-task-elicited perceptions of task difficulty.
Remarkable and consistent was the link between hesitation phenomena and the learner's motivation to perform similar tasks in the future, especially FP and REP given their high frequencies of occurrences. Thus, the more the speaker hesitated, the less motivated the learner felt after task performance. If this is generalizable, increasing task complexity from simple to complex would have positive effects on the learner's motivation to perform similar tasks in the future because it is assumed that the number of hesitation would increase as task demands is also increased. This is an empirical issue, which should be pursued in future studies.
Finally, correlations between the hesitation categories also suggest that FP and REP, both of which occur with higher frequency than the other hesitation types, were positively correlated. This means that repeaters tend to use filled pauses as well, which might suggest problem-solving mechanism for time pressure or for online planning (e.g., Dornyei & Kormos, 1998; Ellis, 2005).
Conclusion
Analysis of L2 hesitation phenomena is still its infancy despite the recent increasing attention to assessment of different aspects of L2 speaking phenomena. The purposes of the present study were to compare distributional patterns of different types of L1 and L2 hesitation phenomena and also present patterns of correlations between among hesitation types as well as correlations between hesitation types and perceptions of task difficulty.
Generally speaking, I believe these purposes were accomplished. However, any study has limitations and the present study is not an exception. First of all, the number of participants was relatively small. By increasing the number of participants, the conclusion to be drawn would be more reliable. Secondly,
the present study used database, which are based on a single task. Employing different tasks would allow cross-task comparisons and new aspects of hesitation phenomena might be revealed. Thirdly, the present study used different ratio measure than Shriberg (1994), which used hesitation episodes per word as opposed to the ratio measure used in the present study, hesitation episodes per T-unit. The use of the same measure would have allowed more direct comparisons. It is hoped that future studies will overcome these limitations and provide new insights into L2 hesitation phenomena.
References
Bear, J., Dowding, J., Shriberg, E., & Price, P.J. (1993). A system for labeling self- repairs in speech. SRI Technical Note 522.
Clark, H., & Fox Tree, J. (2002). Using uh and urn in spontaneous speaking. Cogni- tion, 84, 73-111.
Dornyei, Z., & Kormos, J. (1998). Problem-solving mechanisms in L2 communica-
tion: A psycholinguistic perspective. Studies in Second Language Acquisition, 20, 349-385.
Ellis, R. (Ed.) (2005). Planning and task performance in a second language. Amster- dam/Philadelphia: John Benjamins.
Ellis, R., & Barkhuizen, G. (2005). Analyzing learner language. Oxford: Oxford University Press.
Ishikawa, T. (2008). The effect of task demands of intentional reasoning on L2 speech performance. The Journal of Asia TEFL, 5, 29-64.
Ishikawa, T. (2010). Coding disfluency phenomena for a fluency measure in TBLT research. Journal of Soka Women's College, 40, 101-130.
Segalowitz, N. (2010). The cognitive bases of second language fluency. NY:
Routledge.
Shriberg, E. (1994). Preliminaries to a theory of speech disfluencies. Unpublished
doctoral dissertation, University of California, Berkeley.
DISTRIBUTION OF DIFFERENT TYPES OF HESITATION PHENOMENA 135 Tanaka, S., & Nakano, H. (2012). js-STAR. [software]. Available from http://www.
kisnet.or.jp/nappa/software/star/freq/chisq_lxj.htm
Author Note
I thank Professor Peter Robinson at Aoyama Gakuin University for his consistent encouragement throughout my academic life. Thank are also due to Dr. Elizabeth Shriberg at Speech Technology & Research Laboratory, CA , for her useful advice on disfluency analysis and coding, and Dr. Francesc Tarres at the Universitat Politecnica de Catalunya, Spain, for his permission to use his face database in the present study, which was created by F. Tarres and A. Rama, or "GTAV Face Database" available at http://gps-tsc.upc.es/
GTAV/ResearchAreas/UPCFaceDatabase/GTAVFaceDatabase.htm).
Appendix A. No-reasoning task
Section Staff
Section Chief
' ,,
,
„, .:,, t--„,,,.
Nick
Section Staff
Appendix B. Instructions for the No Reasoning Task
In this task, you will play the role of a company's general manager. Today you are supposed to report to the president about the human relations between people in your section, which was established a week ago. In order to report to the president, you decided to make a phone call but the president was not at home. Now, you have decided to leave a message on his answer machine.
Performance goal: As general manager, report to t about the human relations of the new section by leaving a message on the president's answer machine.
Below is an explanation of your task.
You will be given a task sheet. The task sheet c information:
contains the following
DISTRIBUTION OF DIH---±RENT TYPES OF HESITATION PHENOMENA 137
• Persons' names , pictures, and positions in the section.
• Links between the section members' pictures: Good human relations are represented by blue solid lines and non-good human relations by red broken lines.
human relations, names, pictures, and positions
Information on the task sheet
When you report to the president, please keep in mind the following points
• Describe the current human relationships of the office members to the president. Assume that the president does not have any information about the section.
If you have questions, you can ask now. If you have no questions, please read the "Performance goal" one more time. You have three-minute planning time. Also, note that the maximum time length for recording of the answering machine is three minutes.
