The role of interword spaces in the processing of English compound words

Abstract

Four experiments are reported which examined the role of interword spaces in the processing of English compound words. Normally nonspaced compounds (e.g., softball) as well as normally spaced compounds (e.g., front door) were presented with either their correct spatial layout (softball, front door) or with an incorrect spatial layout (soft ball, frontdoor). Lexical decisions and first fixations on the compounds showed an advantage for interword spaces. However, when refixations on the compounds were taken into account, inserting a space into a normally nonspaced compound significantly disrupted processing. This disruption was larger for adjective-noun compounds than for noun-noun compounds. The results indicate that spatially segmenting compounds facilitates access to the constituent lexemes while spatial unification of compounds benefits the specification of full compound meaning.

Portions of the data were presented at the 12th European Conference of Eye Movements and the 3rd International Conference on Morphology.

Portions of the data were presented at the 12th European Conference of Eye Movements and the 3rd International Conference on Morphology.

Acknowledgments

This research was supported by a pre-doctoral fellowship on Grant MH16745 to the first author, and by grants HD17246 and HD26765. We thank Debra Machacek, Anna Doran, Kadian Leslie, Matt Starr, and Lars Placke with their help on various aspects of these experiments, and Kathy Rastle and two anonymous reviewers for their helpful comments on an earlier version of this paper.

Notes

Portions of the data were presented at the 12th European Conference of Eye Movements and the 3rd International Conference on Morphology.

¹Single fixation duration (the duration of the first fixation on the compound if it receives only a single fixation) was also analysed. It showed a significant crossover interaction that was in the same direction as first fixation duration for both the noun-noun and adjective-noun compounds.

²Correct spatial presentation for these items was defined as how the compounds appeared in the Celex English database (Baayen et al., 1995). However, the results of this study could be compromised if the participants did not agree with how the compounds are supposed to be spaced. Therefore, 13 University of Massachusetts students who did not participate in any aspect of any of the experiments reported were provided with each compound, both with and without a space, and asked to circle how the compound was supposed to spaced. Items in which 50% or more of the participants did not agree with how correct spatial layout was defined in this study were removed. In addition to this, some compounds were removed due to a familiarity rating of less than 4 on a 7 point scale. In all, this led to the removal of 28 items (20 nonspaced and 8 spaced). Participants and items analyses were then rerun on the subset of stimuli. Importantly, the main result (a crossover interaction between compound type and presentation) was still significant for this restricted subset of items.

³A separate group 20 University of Massachusetts undergraduates were asked to circle the correct spacing for the 96 adjective-noun compounds used in Experiments 2 and 4. Using the same criteria as for the noun-noun compounds, 16 items were removed (9 nonspaced and 7 spaced). As with the noun-noun stimuli, the crossover interactions were still significant for this restricted set of stimuli.

⁴We would like to thank an anonymous reviewer for suggesting this interpretation.