Parafoveal processing within and between words

Abstract

Parafoveal preview was examined within and between words in two eye movement experiments. In Experiment 1, unspaced and spaced English compound words were used (e.g., basketball, tennis ball). Prior to fixating the second lexeme, either a correct or a partial parafoveal preview (e.g., ball or badk) was provided using the boundary paradigm (Rayner, 1975). There was a larger effect of parafoveal preview on unspaced compound words than on spaced compound words. However, the parafoveal preview effect on spaced compound words was larger than would be predicted on the basis of prior research. Experiment 2 examined whether this large effect was due to spaced compounds forming a larger linguistic unit by pairing spaced compounds with nonlexicalized adjective–noun pairs. There were no significant interactions between item type and parafoveal preview, suggesting that it is the syntactic predictability of the noun that is driving the large preview effect.

Keywords:

• Reading
• Eye movements
• Parafoveal processing
• Compound words

View correction statement:

Corrigendum

Acknowledgments

This research was initiated when all of the authors were at the University of Massachusetts, Amherst. The research was supported by Grant HD26765 from the National Institute of Child Health and Human Development. Jukka Hyönä acknowledges the financial support of Suomen Akatemia (the Academy of Finland). Denis Drieghe is a postdoctoral fellow of the Fund for Scientific Research (Flanders, Belgium). We thank Martin Fischer and two reviewers for their helpful comments.

Notes

¹ Similar to Experiment 1, second-lexeme durations were analysed separately for low-frequency and higher frequency first words (according to a median split on the first-word frequency). No significant interactions between item type and second-lexeme preview were observed. All viewing duration measures indicated an effect of parafoveal preview, although this effect failed to reach significance in subgaze2 and gaze duration for items with a low-frequency first word (ps > .2). There were also marginal effects of item type on subgaze2 and first fixation on Lexeme 2, mirroring the pattern from the overall analysis.

² To be a bit more precise, a prediction of a 6-ms difference was generated contrasting a whole-word preview with a preview where none of the letters were preserved assuming (a) that the launch site of the saccade to the second lexeme was always at the average fixation location and (b) that the eyes would always land in the optimal location on the second lexeme. Adding in variability to assumption (a) and adjusting for the fact that the fixation location on the second lexeme is rarely optimal would each make the estimate a bit larger. We should also note that current versions of E-Z Reader do not have a complete enough model of word encoding to make predictions for partial-preview conditions. (The values for the parameters were obtained from the corpus used to fit the model in Pollatsek et al., 2006.)