When do combinatorial mechanisms apply in the production of inflected words?

Abstract

A central question for theories of inflected word processing is to determine under what circumstances compositional procedures apply. Some accounts (e.g., the dual-mechanism model; Clahsen, 1999) propose that compositional processes only apply to verbs that take productive affixes. For all other verbs, inflected forms are assumed to be stored in the lexicon in a nondecomposed manner. This account makes clear predictions about the consequences of disruption to the lexical access mechanisms involved in the spoken production of inflected forms. Briefly, it predicts that nonproductive forms (which require lexical access) should be more affected than productive forms (which, depending on the language task, may not). We tested these predictions through the detailed analysis of the spoken production of a German-speaking individual with an acquired lexical impairment resulting from a stroke. Analyses of response accuracy, error types, and frequency effects revealed that combinatorial processes are not restricted to verbs that take productive inflections. On this basis, we propose an alternative account, the stem-based assembly model (SAM), which posits that combinatorial processes may be available to all stems and not only to those that combine with productive affixes.

Keywords:

• Morphology
• Inflections
• Productivity
• Affixation
• Speech production
• Neuropsychology

We are especially grateful to Jeremiah Bertz for his help with test preparation and administration. We are deeply thankful to WRG and his wife for all the efforts they made to allow us the opportunity to understand the nature of his language impairment. This research was supported by National Institutes of Health (NIH) Grants DC006242 (to Michele Miozzo) and DC006740 (to Brenda Rapp), and by the Spanish Science and Innovation Ministry, MICINN, RYC200903917 (to Joana Cholin).

Notes

¹ The AAA past participles are clearly predominant in large corpora such as CELEX (Marcus et al., 1995). In smaller corpora that only include the most frequent words, past participles taking the suffixes –t and –n have a more balanced distribution, in terms of both type and token frequencies (Clahsen, 1999; Marcus et al., 1995).

² It is not necessary to assume inhibitory links in German since the activation received by the –en suffix from ABA verbs and stem-change verbs overrides the default activation reaching the –t suffix.

³ Penke et al. (1999) and Penke and Westermann (2006) examined the production of regular and irregular past participle forms in a group of German-speaking aphasics using the same elicitation task as the one that we used in our study. While most of the participants of the two studies (10/13) responded less accurately to irregular forms, 2 participants were equally impaired with both verb forms. Unfortunately, their results cannot help us to adjudicate between the contrasting predictions that DMM and SAM make about accuracy for ABA verbs, error types, and frequency effects. These limitations are due to the fact that ABA verbs were not tested specifically, and the critical errors for evaluating the models were not reported. Moreover, it is difficult to draw conclusions on the effect of surface frequency reported by Penke et al. (1999) given that a control of stem frequency, a variable typically correlated to surface frequency, was not included.

⁴ Some ABA verbs can take different stems in the (indicative) present tense (e.g., ich laufe [I run], first person singular vs. du läufst [you run], second person singular) and past participle (gelaufen, [run]). These ABA verbs were not used to elicit the responses we analysed here (but they were used in a task described in Footnote 5).

⁵ Similar percentages were found when we examined the entire corpus of 1,280 responses (75% vs. 35%), χ²(1) = 216.11, p < .001. Another argument against this interpretation comes from an elicitation task in which the prompts were second person singular ABA verbs that do not bear the infinitive stem. An example is trittst [(you) kick]; its stem tritt– differs from the infinitive stem tret–. In this task, WRG only reused the form given in the prompt in 3/29 times (10%), indicating that prompt use was not a common response for WRG.

⁶ It should be noted that these results were obtained with nonwords. It is presently unclear whether similar effects of analogy will be observed with the inflections of real words.

⁷ But it is worth pointing out that linguists have also described productive forms that do not involve stem + affix combinations, though these combinations exist in the language. Examples include the subtractive process observed with the nominative nouns in Lardil and the perfect verbs in Papago, and the incomplete phase in Rotuman (see Blevins, 1999, for a discussion of these cases).