The holistic listening comprehension of 112 Kyoto University students, operationalized as TOEFL iBT listening (long listening), was investigated with a battery of 12 tests, including a phoneme and word recognition test, a test of short 10-second listening, a test of long 3- to 5-minute listening, a reading comprehension of listening scripts test, listening and reading cloze tests, a gap-filling test designed to assess syntactic awareness, a grammatical error detection test, and the Vocabulary Size Test.
Rasch analyses were employed to yield person ability measures; these measures were used for correlation studies, a series of linear regression analyses, principal components analysis, and structural equation modeling.
Long listening correlated most strongly with the reading comprehension test (.756) and the listening cloze test (.705), and these two variables explained as much variance in long listening as all the variables combined in a linear regression (68%). Of the two prominent components yielded by a principal components analysis, capturing sounds and processing for meaning, long listening loaded significantly only on processing for meaning (.727) and showed no notable loading on capturing sounds.
When long listening comprehension was viewed as a two-stage activity consisting of capturing input and processing that input for meaning, the participants were found to rely mainly on processing for meaning. As a result, long/holistic listening had more in common with reading comprehension than with short listening, for which the first stage of input capture was more important.
As a part of this study, long listening was expressed as a product of aural word recognition and processing for meaning as in the Simple View of Reading, where reading comprehension is regarded as a product of decoding and linguistic comprehension. While the Simple View of Reading typically accounts for 48% of the variance in reading comprehension, its listening counterpart in this study explained up to 58% of the variance; as much as an improved version of the Simple View of Reading named the Component Model of Reading.
The identification of the structural equation models required an additional component for a total of three latent variables; availability of written text, aural activities, and processing for meaning. The three-latent-variable model for long listening incorporated all the variables as indicators except for the grammatical error detection due to its insignificant contribution to holistic understanding. Generally speaking, structural equation approach produced models which were in good qualitative agreements with correlation studies, principal components analysis, and multiple regression; thus, providing an integrative view and a unified treatment of the participants' proficiency with a focus on long listening.
Overall, the results highlighted the importance of processing for meaning, a skill largely shared with reading comprehension, for the long listening comprehension of Kyoto University students. This finding indicates a transfer of meaning formation skill from L1 and L2 reading to L2 listening.