Analyzing Temporal Variables

To answer research question two, which was in regard to the difference of the temporal variables of students’ oral fluency, several indicators were chosen for detailed analysis.

The present study incorporated Lennon’s (1990) research framework and Riggenbach’s (1991) research model of the temporal variables concerning speaking fluency. These two researchers have been prominent in this field in that they help gain better insights into which temporal variable could better predict the overall

performance of fluency as perceived by raters. Their definition and powerful indicators of fluency have be widely employed in later studies which investigated speaking fluency from a theory-based method and through empirical evidence.

Based on their works, Lennon (1990) and Riggenbach (1991) have proposed several indicators which could better represent participants’ fluency perceived by others. These indicators included (1) rate of speech, (2) amount of speech, and (3) filled and unfilled pauses. The reason these indicators could be generalized to the presents study is that they are suitable for analyzing participants’ oral fluency in the form of speech test. Previous studies which adopted these indicators explored the oral fluency of EFL or ESL learners, which made them also applicable for the present study. Furthermore, they could be utilized here also because they evaluate oral fluency on the basis of objective labeling and calculation, which could generate empirical evidence for quantitative research. The details of these indicators would presented in the following:

(1) Rate of speech: In the light of rate of speech, Lennon’s (1990) model and definition were chosen in this study. In the present study, the rate of speech was defined as “the number of words per minute,” which is consistent

31

with that stated in Lennon’s (1990) study. In his study, the words he counted only involved “pruned words,” exclusive of self-corrected words and those addressing to the examiner. Among the twelve variables Lennon (1990) employed to investigate participants’ oral fluency in his study, he specifically concluded that “pruned words per minute” was an effective temporal

component which could be used to measure perceived fluency. This study would thus incorporate his model for measurement.

To more accurately calculate students’ rate of speech , a specific condition in Riggenbach’s (1991) study was also implemented in this study, in which she formulized that any silence (unfilled pause) which exceeds three seconds will only be counted as three second. For instance, an unfilled pause which extends to the length of 10 seconds would only be counted as three seconds in this study. In this way, a more representative counting of the rate of speech could be attained (Riggenbach, 1991).

(2) Amount of speech: The amount of speech is defined as the total number of words or semantic units (Riggenbach, 1991), and this variable reflects students’ total articulation of correct words. When analyzing the rate of speech, the increase in the amount of speech actually contributed greatly to the rise in speech rate.

The concept of amount of speech is very similar to that of rate of speech, but only the amount of speech could reveal participants’ improvement in articulating more ideas and details in their speech. For example, a student who utters 60 words in 60 seconds and a student who utters 10 words in 10 seconds and remains silent afterwards may have similar rate of speech, while the latter obviously convey more ideas in his speech. In this light, by looking into

32

amount of speech as an independent indicator, participants who contain more meaningful expressions in their speech could be specified. The increase in speech content could in turn be deemed progress of their speaking fluency (Riggenbach, 1991). This indicator was thus adopted in this research.

(3) Frequency of pauses: In response to the analysis of frequency of pauses, both filled pauses and unfilled pauses were taken into consideration in this study. According to Lennon (1999), he specifically stated that filled pause was a salient indicator to determine the dysfluency of participants’ speech.

Riggenbach (1991), on the other hand, claimed that unfilled pauses could be a significant fluency indicator. Such results were similarly verified in Freed’s (1995) research, in which native raters were found to judge participants’ with faster speech rate and fewer pauses as more fluent speakers. The present study would target both kinds of pauses as a means to reveal participants’ progress after receiving the training of ETA throughout a semester.

In the present study, filled pause suggests nonlexical words, which do not comprise lexical information. Based on Lennon’s (1999) study, he termed these fillers as hesitation markers. They were mostly realized as “uh” and “er”

in the recordings. Furthermore, in order to precisely distinguish filled pauses, this kind of fillers which exceeded the length of 0.2 seconds were taken into account. Unfilled pauses, on the other hand, were unambiguously composed of pure silence. In accord with filled pauses, the silence which is over 0.2 seconds was considered as unfilled pauses (Lennon, 1999).

In order to analyze the temporal variables mentioned above, the recordings were quantitatively analyzed through an audio editing program which is called PRAAT.

This specific program has been widely employed to analyze and compare the

33

recordings of participant’s oral speech in a number of previous studies which target language fluency (De Jong, Steinel, Florijn, Schoonen, & Hulstijn, 2013), so it is looked upon as an effective and accurate tool for the analysis of speech. After all the recording files were imported into this audio editing program, and they would be transformed into visually accessible sound wave diagrams. These files would then be annotated and edited, revealing the number of words that are included in a single test section of a participant.

The interface of the program of PRAAT was shown below in figure 1:

Figure 1 The Interface of PRAAT

34

3.6 Qualitative analysis