Dominance-driven morphological decomposition

The term dominance refers to the relative frequency of a given derived or inflected word form with respect to its stem word form. For example, a plural form such as worlds is labeled as singular-dominant because the inflected form worlds is less frequent than the stem form world; conversely, a plural form such as windows is labeled as plural-dominant because the inflected form windows is more frequent than the stem form window. In traditional lexical decision tasks, plural-dominant forms ( windows ) has been shown to be recognized faster than singular-dominant forms ( worlds ). In light of these results, Baayen et al. (1997) made a strong claim in support of Schreuder & Baayen (1995)’s dual-route race model, in which word recognition goes through both a parsing (decomposition) route and a storage (whole-form search) route. The route that reaches the recognition point first “wins” the race, in the sense that it is responsible for recognition. In singular-dominant plural forms ( worlds ), the whole form is low-frequency, so the storage route takes time to search for it in the lexicon; this allows the form to go through decomposition and therefore be recognized in the parsing route. In plural-dominant plural forms ( windows ), the whole form is high-frequency, so the storage route recognizes it faster than the parsing route would; hence, the fast response to windows is due to activation of the whole-form unit in the storage route.

Notice that in the race model, the two routes run independently of one another. This means that the decomposition procedures in the parsing route and the whole-word lexical search in the storage route occur regardless of which route ends up being responsible for recognition. We tested this hypothesis in a visual masked priming experiment, in which we elicited the priming response to (i) singular-dominant plural prime forms (sgdom-sgdom condition; e.g., worlds-HEAVENS); and (ii) plural-dominant plural forms (pldom-sgdom condition; e.g., windows-GODS). Results showed that priming arose in the sgdom-sgdom condition, but not in the pldom-sgdom condition, thus suggesting that dominance asymmetries (and therefore word frequency) impinge on decomposition. This seems to suggest a model of decomposition in which activation of morpho-orthographic units at early stages of processing may be affected by their resting activation level, which is generally assumed to correlate with their frequency. In such a model, at early stages both the whole-form and stem-form units compete with one another for activation. Competition is solve through the mechanism of lateral inhibition, in which a unit with a higher resting activation level activates faster and therefore inhibits the other competing, compatible units.

The model just outlined predicts the priming response to be systematically affected by dominance asymmetries between the affixed and the stem forms, regardless of the affix being used and of the priming type being elicited (i.e., root/affix priming). A series of experiments are currently being designed and carried out to test this prediction, in which the masked priming response is elicited to stem- and affixed-dominant words sharing the same affix.