The Transition From Sublexical to Lexical Processing in a Consistent Orthography: An Eye-Tracking Study

Abstract

We studied the transition in predominant reading strategy from serial sublexical processing to more parallel lexical processing as a function of word familiarity in German children of Grades 2, 3, 4, and adults. High-frequency words, low-frequency words, and nonwords of differing length were embedded in sentences and presented in an eye-tracking paradigm. The size of the word length effect was used as an indicator of serial sublexical decoding. When controlling for the generally higher processing times in younger readers, the effect of length over reading development was not direct but modulated by familiarity: Length effects were comparable between items of differing familiarity for Grade 2, whereas from Grade 3, length effects increased with decreasing familiarity. These findings suggest that Grade 2 children apply serial sublexical decoding as a default reading strategy to most items, whereas reading by direct lexical access is increasingly dominant in more experienced readers.

View correction statement:

Corrigenda

ACKNOWLEDGMENTS

Many thanks to Jessica Geier and Sandra Hillian-Tress for helping with data acquisition and preparation.

Notes

¹ Gaze duration was highly correlated with the number of fixations in first-pass reading for participants of all groups; correlation coefficients ranged from .86 for long nonwords through .96 for short low-frequency words (all ps < .01). We therefore refrained from reporting number of fixations/saccades.

² We did not experimentally control for words immediately preceding and following the targets. However, both length and frequency of words n–1 and n+1 were fairly comparable between different experimental conditions. Between seven and 10 of 12 words n–1 and n+1 per condition were very high frequency in terms of CELEX frequency counts (occurrences > 1000/million; these were words such as direct and indirect articles, conjunctions such as und [and], conjugated verb forms of high-frequency verbs such as geht [goes], sagt [says], hat [has]). The only condition with a lesser amount of very high-frequency words was the word immediately preceding short nonwords. Because many of the nonwords were used as proper names in the sentences, the words preceding short nonwords were often familiar first names (e. g., Paul, Linda), which are not listed in CELEX. Similarly, word length of words n–1 and n+1 was mostly comparable for most conditions (namely, approximately four letters/word) apart from the words preceding long high-frequency words and short nonwords (5.1 and 5.3 letters/word, respectively). On the whole, words n-1 and n+1 were fairly comparable in terms of frequency and length between all conditions with the exception of n–1 for short nonwords. However, because there was nothing peculiar about processing times for short nonword targets, we do not think that the longer and less frequent words in position n–1 necessarily exerted a detrimental influence on processing times of these words.

³ Stimuli are available on request.