Presented at the 16th Annual Meeting of the Society for the Scientific Study of Reading, Boston, MA, 25–27 June 2009

Purpose: In the relatively transparent Greek orthography, vocabulary measures take up most of reading comprehension variance accounted for by print-related measures. Here we extend previous work by concurrent and longitudinal multiple linear regression analyses of reading comprehension. Method: Data from 516 children in Grades 3, 4, and 5 included measures of vocabulary (V: PPVT & WISC-III), word identification (W: accuracy and fluency), pseudoword decoding (D: accuracy and fluency), reading comprehension (RC) and auditory comprehension (A). Results: V accounted for 22% of total concurrent RC variance (after age, grade, and WISC blocks), more than A (11%), W (14%), or D (8%); V also accounted for more unique variance (entered last: 7.6%) than A (1.6%), W (1.6%), or D (.3%). In the longitudinal prediction of RC one year later (data from 482 children), V was the strongest predictor either without (22%) or with (6.3%) an RC autoregressor (compared to A: 14% and 4.9%; W: 13% and 3.2%; D: 8% and 2.1%), accounting for most unique variance (7.0% and 2.7%, respectively, compared to A: 2.9% and 2.0%; W: .8% and .3%; D: .2% and .1%). Including text fluency (words per minute) substantially increased the concurrent but not the longitudinal prediction of RC. Conclusions: These results are interpreted in the context of the lexical quality hypothesis, suggesting that the simple view may be too simple to capture the development of sight-word reading. Moreover, vocabulary measures seem to assess more than depth and breadth of word knowledge, perhaps due to experience with visual word recognition.