Lexical Quality and Reading Skill: Bottom-Up and Top-Down Contributions to Sentence Processing

Abstract

This research investigated whether spelling ability, an index of precise lexical representations, predicts the balance between bottom-up and top-down processing in online sentence processing among skilled readers, over and above contributions of reading ability, vocabulary, and working memory. The results showed that the combination of superior reading and spelling was associated with more accurate report of rapidly presented sentences and that spelling uniquely predicted reduced reliance on context to identify words. These findings support the lexical quality hypothesis by demonstrating that individual differences in the reading strategies used by skilled readers reflect differences in the precision of their lexical representations.

ACKNOWLEDGMENTS

The research reported in this article was supported by an Australian Research Council Discovery grant to Sally Andrews. This work was presented at the Australian Language and Speech Conference held at the University of New South Wales in Sydney, Australia, in December 2009.

Notes

¹The gender imbalance is typical of the Psychology 1 participant pool (71% female). The sample included seven early bilinguals who did not differ from the rest of the sample in terms of performance on any of the individual differences measures or experimental task variables.

²To provide some insight into the normative level of reading and vocabulary performance demonstrated by the University of Sydney students composing our sample, we recently administered the Nelson-Denny Reading Test (Brown, Fishco, & Hanna, 1993) to a sample of 36 participants recruited from the same 1st-year participant pool as those tested in this experiment. Relative to Brown et al.'s (1993) normative data for 852 U.S. students in the 1st year of 4-year university programs, the median score for this sample fell at the 85th, 79th, and 67th percentile for vocabulary, reading comprehension, and reading rate, respectively. These data suggest that the present participants are generally above average in written language proficiency, even relative to a university student sample. The Nelson-Denny test scores were also generally strongly positively skewed, confirming that the available standardized tests of reading do not discriminate as effectively within the population of skilled readers as our specially developed tests, which all yield relatively normal distributions in our university student samples.

³This penalty was incorporated into the scoring method as a simple means of ensuring that participants were actually reading the material—in practice it was applied rarely.

⁴Although the distractor was always a poorer fit with the sentence than the target word, there was some variation in semantic and syntactic fit such that the sentence produced by including the distractor could be grammatically incorrect, semantically implausible or merely rather strange. The complete list of sentences can be obtained from the authors.

⁵Summaries of the full analyses can be obtained from the authors.

⁶This does not reflect ceiling performance for identifying the target—average report of upper-position targets was 86%—but the average bias ratio for upper-position targets was very high (0.95) because the lower distractor was rarely intruded, whereas the upper distractor was reported about half the time (average bias = 0.49) when the target was in the lower position.