Past research (e.g. Daelemans et al. 1994, Krott et al. 2001, Skousen 1989) has shown effects of statistical analogy in language production. Language users tend to create morphologically complex forms after the examples of phonologically, morphologically or semantically similar words, instead of on the basis of a rule. As a consequence, they may come up with forms which violate the general rules (or generalizations). For instance, they may create the past-tense form "rept" after the examples "slept" and "kept".
The question arises whether analogy also affects language comprehension. We carried out two self-paced reading experiments, in order to investigate
In Experiment I, we focussed on the past-tense forms in Dutch, which consist of the verbal stem and the suffix "-te" or "-de". In a production experiment, Ernestus & Baayen (2001) showed that for a verb that takes "-de" participants are more likely to incorrectly select the past-tense suffix "-te" when phonologically similar words take "-te" (incongruous verbs) than when similar words take "-de" (congruous verbs). Similarly, participants are more likely to incorrectly select "-de" when phonologically similar words take "-de".
If analogy affects reading, we expect that misspelled past-tense forms slow down reading times more if the verb is congruous than if it is incongruous. In Experiment I, we found that this is indeed the case. We conclude that analogy affects reading times.
Since there is a near one-to-one relation between sounds and graphemes in Dutch, the analogy effect attested in Experiment I can either be due to sound similarity (phonological analogy) or to grapheme similarity (orthographic analogy). In order to investigate which type of analogy affects reading, we carried out a second experiment, in which we investigated the effects of misspelled past participles. In Dutch, all regular past participles are realized with a final [t]. Nevertheless, the past participles of verbs with past-tense forms ending in "-de" are spelled with "d" instead of "t". If the analogy is phonological, we do not expect a difference between the mispelled past participles of congruous and incongruous verbs. If, on the other hand, the analogy is orthographic, we expect more problems with the misspelled past participles of congruous verbs than with the misspelled past-participles of incongruous verbs. The results of Experiment II show that the latter is the case. The analogy that affects the reading of Dutch verb forms is orthographic in nature.
References
Daelemans, W. & S. Gilles, G. Durieux (1994). The acquisition of stress, a data-oriented approach. Computational Linguistics 20 (3), 421-451.
Ernestus, M. & H. Baayen (2001). Choosing between the past-tense suffixes -te and -de. Linguistics in the Netherlands 2001. In press.
Krott, A. & R.H. Baayen, R. Schreuder (2001). Analogy in morphology: modeling the choice of the linking morphemes in Dutch. Linguistics 39 (1), 51-93.
Skousen, R. (1989). Analogical Modeling of Language, Kluwer, Dordrecht.