Probabilistic Cues to Grammatical Category in English Orthography and Their Influence During Reading

Abstract

We investigated probabilistic cues to grammatical category (noun vs. verb) in English orthography. These cues are located in both the beginnings and endings of words—as identified in our large-scale corpus analysis. Experiment 1 tested participants' sensitivity to beginning and ending cues while making speeded grammatical classifications. Experiment 2 tested sensitivity to these cues during lexical decisions. For both tasks, words with consistent ending cues (with respect to grammatical category) were processed more quickly and with lower error rates than words with inconsistent ending cues. However, for beginnings, consistent cues resulted in lower errors but no differences in response times. The data reported here point to the multifaceted nature of grammatical category representation and indicate that probabilistic orthographic cues relating to grammatical category have a clear influence on lexical processing particularly when these cues are located at the end of the word.

Notes

¹We make no special claim for orthography taking precedence over phonology, we only claim that there are multiple cues to grammatical category some of which are realised in orthography. Note, though, that Kelly (2004) did find effects of orthographic-only features of words that assisted in stress assignment in reading, so there is the possibility that orthography has an additional effect over phonological processing in reading.

²Variables that failed the tolerance test (too little variance) were omitted from the discriminant analyses, though in each analysis there was just one variable failing: zu- for the beginnings and -ytes for the endings.

³Lemma frequency was also not significantly different across the groups for both main effects and all interactions.

⁴Error rates most likely reflect a number of variables, such as the low frequency of some items, emphasis in task instructions to respond rapidly, and the fact that this study examined processing of multisyllabic words. Most previous studies of on-line single-word reading have examined the processing of monosyllables. Jared and Seidenberg (1990) investigated the naming of disyllabic words and reported error rates comparable to ours in some experiments.

⁵As in Experiment 1, these error rates most likely reflect a number of variables.