Effects of morphological families on English compound word recognition: A multitask investigation

Abstract

Three experiments examined the influence of first lexeme morphological family size on English compound word recognition. Concatenated compound words whose first lexemes were from large morphological families were responded to faster in word naming and lexical decision than compounds from small morphological families. In addition, an eye movement experiment showed that gaze durations were shorter on compounds from large morphological families during sentence reading. This was mainly due to more refixations on compounds from small morphological families. Posthoc analyses and re-analysis of past studies suggested that compounds with a larger number of higher frequency family members (HFFM) are read more slowly than compounds with fewer HFFM. Thus, while morphological family size is generally facilitative, the presence of HFFM has an inhibitory effect on eye movement behaviour. The time-course of these effects is discussed.

Keywords:

• Compound words
• Eye movements

Acknowledgements

These data were collected as part of a Departmental Senior Honors Thesis conducted by the second author at Wesleyan University. She was supported by a summer fellowship from the Howard Hughes Medical Institute. We would like to thank the other member of her thesis committee, Andrea Patalano, for helpful comments. In addition, we would like to thank Joanna Dicke and Elizabeth Litvina for their help collecting data for Experiment 3 and Alix Haber for her help calculating the morphological family size and number of HFFM for past studies. We would like to thank Raymond Bertram, Albrecht Inhoff, and Victor Kuperman for their helpful comments on an earlier version of the manuscript. The data were presented at the 6th Morphological Processing conference.

Notes

¹In the items analysis for the lexical decision task, it was noticed that one item (streambed, from the small family group) led to extremely long response latencies and high errors; when this item was deleted, the results remained significant.

²One item, streambed, did not appear in the English Lexicon Project database.

³The phonemes were distributed across the two conditions as follows: for large morphological families, 2-/b/, 2-/k/, 4-/h/, 6-/s/, and 1-/w/ and for small morphological families, 2-/b/, 4-/k/, 3-/h/, 2-/s/, and 6-/w/.

⁴Additional participants received credit for participating, but their data were not analysed because they were non-native speakers, had extensive track losses (20% or more), had inaccurate calibrations, or terminated the experiment early.

⁵Some pairs, for example, had a large morphological family compound composed of eight characters and a small morphological family compound composed of nine characters. These pairs were not included in the supplementary analyses.

⁶All compounds in the database were considered for the computation of morphological family size and HFFM in past experiments—compounds were not removed on the basis of low familiarity.

⁷Compounds with high and low frequency first lexeme were initially paired in Juhasz (2008) and two sentence frames were created for each pair. For these new analyses, data were averaged across both sentence frames for each compound. There were two compound words which were items in more than one experiment. These compounds were considered separate items in the present analysis due to differences in experimental set up and materials.

⁸One compound word did not appear in the English Lexicon Project database.