DataAnalysis

The data collected from the assessment and the questionnaire were quantitatively analysed. The Pearson's correlation coefficients were

calculated for nine contributing factors of intelligibility in order to find out which factors were eligible for multiple regression analysis which would in

turn show the relationship between intelligibility and its contributing

factors. As the criteria for the further analysis, the contributing factors should have at least a modest correlation to intelligibility (r ) .40). All the statistical analyses were performed using SPSS 13.0 for Windows.

5.4.2 Results

5.4.2.1 Descriptive Statistics and Correlations

Table 5.5 presents the mean scores (mo and standard deviations (SD)

Table 5.5

Deseni tive Statisties ofJntelligr'in7ity andlts Contributing Faetors in

ExperimentB

Assessment Items

ENLa ^ESLb

M ^SD M ^SD

ENL-ESL

Mean

Difference

Overall Intelligibility 3.06 .75

Inte]ligibility Factors:

Sound aecuracy Word stress Sentence stress

Adjustments in connected speech Intonation

Rhythm

FIuency

Lexical accuracy

Grammatical aceuracy

2.75 3.19 3.47 3.72 3.14 3.25 3.50 3.e3

3. 19

.69 .86 .65

.57 .83 .81

.78 .85 .74

3.00

3.61 3.56 3.28 2.94 3.17 3.31 3.39 2.75 3.14

.79

.60 .65 .85 .86 .78 .79 .80 .69 .76

.06

.86 .37 .19 .78 .es

.06 .11

.28 .05 Note. Maximum score = 5.00.

an = 36; bn = 36.

of the ENL speakers and of the ESL speakers, and the mean differences between the scQres by the ENLtspeaker assessors and those by the ESL'

speaker assessors (ENL - ESL Mean Difference).

Table 5.6 shows the correlation coefficients of the factors determining

intel!igibility of EFL learners' speech and their ranks in strength of

correlation for the native English speaker assessors and the ESL speaker

assessors. Conceming the rank order of the intelligibility factors, the

Spearman's correlation test revealed that there was no significant

relationship between the assessment of the ENL'speaker assessors and that ofthe ESL'speaker assessors, rs = .22, p= .58. It means that the data of the five ENLtspeaker assessors do not correlate with the data of the five ESLt speaker assessors.

Table 5.6

Correlations ofFactors Determining Intelhl ibility in ExperimentB

Intelligibility Factors

ENLa

r Rank

ESLb

r Rank

Sound accuracy Word stress Sentence stress

Adjustments in connected speech Intonation

Rhythm

Fluency

Lexical accuracy

Grammatical accuracy

.08 -.06

.Ol .51de*

.17

.12

.39de

.49dede

-.02

6 7

9

1

4 5 3 2 8

.66*de

-.11

=17

.55de*

.14 .05 .45**

.05 .05

1

6 4 2 5 7 3 8 9 Note. an = 36; bn = 36.

dep < .05; dedep < .Ol.

In the case of the data of the ENL'speaker assessors, three of the nine contributing factors had a modest correlation to intelligibility: adjustments in connected speech, fluency, and lexical accuracy. As for the ESL"speaker assessors, there were also three factors which had a moderate correlation to

intelligibility: sQund aceuracy, adjustments in connected speech, and fluency.

Although these correlation coefficients show the relationship of each factor to intelligibility, they cannot tell us much about the predictive power

of the independent variables. In order to show the relationship between

intelligibility and its contributing factors, the enter'method regression

analyses were subsequently performed te each set of data of the ENL'

speaker assessqrs and the ESL-speaker assessors, which were significantly

correlated to intelligibility.

5.4.2.2 Factors Determining Intelligibility

The summary of the regression analyses is presented in Table 5.7 for the ENL'speaker assessors and in Table 5.8 for the ESL'speaker assessors.

For the ENL'speaker assessors, it was found that the value of the multiple

coefficient of determination was R2 = .37, p < .Ol. This means that the

investigated contributing factors (i.e., adjustments in connected speech, fluency, and accuracy) might account significantly for 370/o of the variation in intelligibility of EFL leamers' utterances and 630/o ean be explained by other than the investigated factors. However, a closer examination of the result has revealed that only the standardised coefficient ts) ef adjustments in connected speech was statistically significant, 6 = 51, p < .Ol, implying that only adjustments in connected speech might have a positive effect on the intelligibility of EFL learners' speech.

As for the ESL-speaker assessors, it was found that the value of the multiple coeflicient of determination was R2 = .53, p < .05. This means that the investigated contributing factors (i.e., sound accuracy, adjustments in

connected speech, and fluency) might account significantly fer 530/o of the variation in intelligibility of EFL learners' utterances and 470/o can be

explained by other than the investigated facters. However, a careful eonsideration of the result has disclosed that only the standardised

coefficient ts) of word stress was statistically signifieant, 6 = 38, p < .e5.

This implies that only sound accuracy might have a positive effect on the intelligibility ofEFL leamers' speech.

Table 5.7

Malti; fe Regression Analysis of Contributing Faetors for Intelhgi'bih'ty

Assessed by the ENL'speakerAssessors in Experiznent B

Independent Variables (Contributing Factors)

Dependent Variables

(Intelligibility)

B SEB ⁶

Adjustments in connected speeeh

Fluency

Lexical accuracy

. b )

Intelligibility

Intelligibility

Inte}ligibMty

O.67 O.24 O.30

e.2o O.16

O. 15

.51*k .25 .34 Note. R2 = .37, p<.Ol; se"p<.Ol.

Table 5.8

Maltil Je Regression Analysis of Contributing Faetors for

Assessed by the ESL 'speaker Assessors in Experinient B

IntellZgn'bility

Independent Variables (Contributing Factors)

Dependent Variables

(Intelligibility)

B SEB ⁶

Sound accuracy

Adjustments in connected speech

Fluency

) ) .

Intelhgibihty Intelligibility

Intelligibility

O.51 O.28 O.24

O.22 O.14 O.14

.38*

.30 .25 IVrote. R2 = .53, p< .05; *p < .05.

As Experiment A, the current study also attempts to explore the three same research questions: M7/hieh faetors ofinteth' gi'bility are important for native Engfish speakers.?, MZhieh faetors ofintellign' 'bdity are i2nportant for

ESL speakers2, and What will emerge as eommon priniaivfaetors ofg7obal

intelligr'bility2

In the case of the ENL'speaker assessors, the result of the study has

revealed that there are three factors which may have substantial

relationship to intelligibility: adjustments in connected speech, fluency, and

lexical accuracy. Among these factors, the component of adjustments in connected speech has shown to be the most significant factor for

intelligibility. As far as the ESL'speaker assessors are concerned, there are

also three factors which may predict intelligibility: sound accuracy,

adjustments in connected speech, and fiuency. Of these three factors, sound accuracy appears to be the most significant factor determining intelligibility.

ドキュメント内 A Study on Factors Determining Global Intelligibility of EFL Learners' Speech (ページ 112-117)