What's in a sentence? The crucial role of lexical content in sentence production in nonfluent aphasia

Abstract

This study investigated the effect of lexical content on sentence production in nonfluent aphasia. Five participants with nonfluent aphasia, four with fluent aphasia, and eight controls were asked to describe pictured events in subject–verb–object sentences. Experiment 1 manipulated speed of lexical retrieval by varying the frequency of sentence nouns. Nonfluent participants' accuracy was consistently higher for sentences commencing with a high- than with a low-frequency subject noun, even when errors on those nouns were themselves excluded. This was not the case for the fluent participants. Experiment 2 manipulated the semantic relationship between subject and object nouns. The nonfluent participants produced sentences less accurately when they contained related than when they contained unrelated lexical items. The fluent participants exhibited the opposite trend. We propose that individuals with nonfluent aphasia are disproportionately reliant on activated conceptual–lexical representations to drive the sentence generation process, an idea we call the content drives structure (COST) hypothesis.

• Aphasia
• Nonfluent aphasia
• Broca's aphasia
• Sentence production
• Lexical retrieval
• Lexical competition.

We thank all our participants for their time spent to support our research. Thank you to Jacob Cameron and Angie Slocombe for help and advice regarding the magnetic resonance imaging (MRI) scans. We further thank Bridget Burmester and Josh Faulkner for help with the data collection. Part of this research was presented at the 50th Annual Meeting of the Academy of Aphasia, San Francisco, CA, USA, 2012.

This research was supported by a Victoria University Doctoral Scholarship to P. Speer; and by a Marsden Project grant to C.E. Wilshire [grant number VUW0505].

Notes

1 The individual analyses were performed using repeated measures logistic regression, which included the predictor variables noun position (subject vs. object), subject noun frequency, and object noun frequency. The number of errors made in each error position on each sentence item served as the dependent variable, which ranged from zero (no errors) to three (errors in all three sessions). The distribution of the model was set to multinomial in order to capture the nonbinomial nature of the dependent variable.

2 All errors made by control participants were semantic/visual substitutions, out of which 34% were self-corrections [e.g., the dog (target: goat) oh goat is biting the cat].

3 An additional 2.8% of data points (across all groups) were lost due to technical problems or other disturbances (like background noise).

4 When collapsed across all individuals, this led to a removal of an average of 2.0% of data for the aphasic groups and 1.0% for the control group. Due to technical problems and other disturbances, an additional 2.2% of data points (across all groups) were lost.

5 Conversely, the lesioning of semantic input led to a fluent anomic speech pattern, represented by a good production of determiners and semantically impoverished verbs, but a lack of nouns and semantically rich verbs (Gordon & Dell, 2003).