Consistency and regularity in past-tense verb generation in healthy ageing, Alzheimer's disease, and semantic dementia

Abstract

Older adults, individuals with dementia of the Alzheimer's type (DAT), and individuals with semantic dementia (SD) produced the past tense of verbs based on present-tense carrier sentences (e.g., Everyday I ding the bell. Yesterday I_____the bell). Both regularity (i.e., whether or not -ed is used for the past tense) and consistency (i.e., the degree to which verbs of similar orthography and phonology in the present tense have similar past tenses to the target) were manipulated. Participants received regular consistent (e.g., land–landed), regular inconsistent (e.g., weed–weeded), irregular consistent (e.g., sting–stung), and irregular inconsistent (e.g., light–lit) verbs. The dependent measures were overall accuracy rates and error rate types (e.g., regularizations, analogies, and other errors). Both consistency and regularity influenced performance. In addition, individuals with DAT showed a disproportionate deficit for inconsistent verbs associated with a high summed frequency of enemies, whereas SD individuals produced disproportionate breakdowns in performance on regular inconsistent, irregular consistent, and irregular inconsistent verbs. These results are consistent with the perspective that semantic/lexical processes are involved in processing the past tense of both irregular verbs and regular inconsistent verbs, and that attention is used to select appropriate responses and control inappropriate responses.

This work was supported by NIA PO1 AGO3991 and NIA A605681. We thank Martha Storandt for providing the psychometric test support, John Morris and the Washington University Alzheimers Disease Research Center clinicians for their careful recruitment and description of the various participant groups. We thank Matt Lambon Ralph for helpful discussions regarding SD. In addition, we thank Avi Snyder for software involved in preprocessing MR images. Finally, we thank Karalyn Patterson and one anonymous reviewer for their comments on an earlier draft of this article.

Notes

¹ In contrast, individuals with nonfluent aphasia (NA, e.g., Ullman et al., 1997) have displayed the opposite pattern. However, a recent study by Bird et al. (Bird, Lambon Ralph, McClelland, Seidenberg, & Patterson, 2003) demonstrated that when the phonological complexity of regular and irregular verbs is controlled, there is no difference in the accuracy rates for irregular verbs and regular verbs in NA.

² Matched is defined as regular because according to the lexicon and rules (L&R) model (Pinker, 1999), phonological rules are applied after affixation. Specifically, the affix -ed is added to the verb stem, and the pronunciation of that affix is then determined according to the final phoneme of the stem. For stems ending in //, the pronunciation of the affix is /t/ (Pinker, 1999).

³ For evidence of independence between lexical syntax and semantics in SD, see Garrard, Carroll, Vinson, and Vigliocco (2004).

⁴ In all simple effects analyses involving group, we employed a modified Scheffe test for mixed designs. In these analyses, the mean square error from the omnibus analysis involving all three groups was used for each specific computation, and the degrees of freedom were based on the two groups combined.

⁵ SD1 received the regular inconsistent verbs king (e.g., he kinged his checkers) and breeze rather than ding and sneeze, and he did not receive the irregular inconsistent verbs slay and fly.

⁶ Although the PDP perspective would appear to predict the graded effects, as indicated, it is possible that differences in the strength of the semantic and phonological representations and/or connections could capture these results, and so, ultimately, quantitative predictions from this framework are necessary to directly test the predictions.