Data Analysis

This section primarily addresses two types of data analysis methods. One is the methods of quantitative analysis, including the computational measures of semantic similarity and the statistical analysis for investigating congruency effects on L2 collocation processing. The other is the methods of quantitative analysis, utilized to collect more in-depth details of learners’ responses and processing patterns. Both types of methods were conducted complementarily so as to collect multi-faceted data and to verify different layers of the research results.

In the present study, three sub-studies involved quantitative analysis. The first sub-study examined the applicability of semantic similarity measures in distinguishing semantic relations. The similarity values of word pairs in categories of different semantic relations were compared based on statistical average and standard deviation. Effectiveness evaluation on semantic similarity measures was made in light of statistical differences in similarity values. The second sub-study applied semantic similarity measures for congruency classification. The third sub-study investigated congruency effects on L2 collocation learning. Both experimental results were examined with the one-way analysis of variance (one-way ANOVA) via SPSS

software. In the second sub-study, the semantic similarity values of L2 collocates and transferred words from L1 counterparts were compared between congruency categories so as to evaluate and refine the properties of congruency. In the third sub-study, the independent variables involved collocational congruency and learners’

proficiency level, whereas the dependent variable being learners’ collocation performance. Test results of three participant groups were statistically processed to explicate the discrepancy between collocational types and participant groups. In both experimental analyses, one-way ANOVA with between-group and within-group variances were applied to evaluate significance of difference, where the threshold of significance level was set at p < α= 0.05.

The fourth sub-study surveyed the participants’ responses to the collocation use questionnaire and involved both quantitative and qualitative analyses. Both reliability and validity were considered in questionnaire data analysis, as well as responsive differences in individual items and in the foci of the questionnaire. After eliminating a few of invalid items of the responses, Cronbach’s Alpha was applied to verify the reliability of the questionnaire. In general, higher reliability was indicated as the

Alpha values reached above 0.6. Taking the same screening steps, KMO & Bartlett’s Test was applied to verify the validity of the questionnaire results. If KMO values fell

As statistic data led to numeric evidences for research topics verification, the explicative features of the qualitative research method further provided in-depth description of learners’ viewpoints to shed lights on various research perspectives.

The present study used think-aloud method as a research instrument to systematically document the orchestration of successive thinking process, knowledge application and meaning construction by L2 learners in collocational processing. The think-aloud

protocol helped reveal learners’ verbalized thinking process and actional activities, which led to more understanding of learners’ dynamic foci, successive knowledge construction, as well as the interaction between lexical decision and mental network.

To document the paralinguistic aspects of verbal discourse, it was important to have a system of transcribing think-aloud data. The transcripts had been produced according to standards of written discourse. Transcription of think-aloud data also necessitated some conventions such as abbreviations and symbols because the informants who thought aloud in collocational processing not only uttered their thoughts, but also performed several non-verbal actions such as facial expression. The actional activities offered additional information to infer mental activities. Overall, through think-aloud, the researcher prompted learners to reveal information which provided clues for the types of processes. Actional information also revealed if learners tackled the collocation problems with frustration or confidence.

The transcription system subsumed the static device of written symbols to represent dynamic movements of think-aloud. To keep the transcripts as concise as possible, the abbreviations and symbols listed in Table 3.5.1 were adopted. After transcription of the verbalized thoughts, the researcher searched the cognitive patterns and the factors of congruency processing. All of the patterns and factors were evaluated and categorized into priming patterns, processing strategies and contributing factors.

Table 3.5.1 Think-Aloud Transcription Conventions (adapted from Silverman’s (2001))

Symbol Explanation

[] Left brackets indicate the point at which a current speaker's talk is overlapped by another's talk.

@ The at sign represents laughing of the speaker.

● A big dot under a lexical item means that the speaker is required to think aloud.

. A period indicates a stopping fall in tone.<>

, A comma indicates a continuing intonation.

? A question mark indicates rising intonation.

＝ Equal signs, one at the end of a line and one at the beginning, indicates no gap between the end of one turn and the beginning of the next by the same speaker.

(0.4) Numbers in parentheses indicate elapsed time in silence in tenths of a second.

(.) A dot in parentheses indicates a tiny gap, probably no more than one tenth of a second.

_ Underscoring indicates some form of stress, via pitch and/or amplitude.

:: Colons indicate prolongation of the immediately prior sound. The length of the row of colons indicates the length of the prolongation.

WORD Capitals (except at the beginning of lines) indicate especially loud sounds relative to the surrounding talk.

.hhhh

A row of h's prefixed by a dot indicates an in-breath; without a dot, an out-breath. The length of the row of h's indicates the length of the in-breath or out-breath.

( ) Empty parentheses indicate the transcriber's inability to hear what was said.

Vocal noises.

(word) Parenthesized words are possible hearings.

(( )) Double parentheses enclose comments interpolated by the researcher.

>< Angled brackets show talk that is noticeably faster than the surrounding talk.

CHAPTER FOUR RESULT ANALYSIS AND DISCUSSION

This chapter addresses the quantitative sub-studies of applying semantic measures to congruency effects on L2 collocation performance, as well as the qualitative sub-studies in soliciting factors of learners’ cognitive processing and extracting patterns of cross-linguistic collocational priming. In the first section, applicability of semantic similarity measures was verified by examining their evaluative results in distinguishing semantic relations between word pairs. Based on the successful evaluation, the second section then applied semantic similarity measures to evaluating L2 collocation congruency and considered the properness of the resulting classification of congruency. In the third section, effects of learners’

proficiency levels and collocation congruency on learners’ collocation performance were investigated. Based on the empirical results in the quantitative sub-studies, a new model that represented the interaction between collocation congruency and learners’ collocation production was proposed and its utility in characterizing learning behaviors on L2 collocations was discussed in the fourth section. In the fifth section, learners’ response to collocation processing and congruency was elicited with a questionnaire survey and primary factors were derived by a quantitative analysis.

Lastly, the think-aloud protocol was conducted to extract cross-linguistic collocational processing patterns in the sixth section.