Aphasias and theories of linguistic representation: representing frequency, hierarchy, constructions, and sequential structure
Article first published online: 3 OCT 2013
© 2013 John Wiley & Sons, Ltd.
Wiley Interdisciplinary Reviews: Cognitive Science
Volume 4, Issue 6, pages 651–663, November/December 2013
How to Cite
Menn, L. and Duffield, C. J. (2013), Aphasias and theories of linguistic representation: representing frequency, hierarchy, constructions, and sequential structure. WIREs Cogn Sci, 4: 651–663. doi: 10.1002/wcs.1257
- Issue published online: 11 OCT 2013
- Article first published online: 3 OCT 2013
- Manuscript Accepted: 18 AUG 2013
- Manuscript Revised: 12 AUG 2013
- Manuscript Received: 18 DEC 2012
Error and preservation patterns in aphasic speech show that the brain makes use of the frequencies of words, constructions, and collocations, as well as category membership and hierarchical structure, during language processing. Frequency effects are evident along two quasi-independent axes: syntagmatic (the sequential context, e.g., deploying correct functors, categories, and utterance-level intonation) and paradigmatic (the choice at any given linguistic level, e.g., selecting content words and modifying structures). Frequency along the syntagmatic axis is shown to play a role in errors involving idioms, constructions, and collocations that cross major phrasal boundaries. Along the paradigmatic axis, frequency affects errors involving lexical selection, competing functors and inflected forms (e.g., using plural where singular is required). An account of language representation and processing that encompasses frequency as well as categorization and structure is compatible with what we know about how the brain works: increased experience with a linguistic structure results in increased activation—and strengthening—of the neural networks involved in processing that structure. These claims are supported by the literature on experimental work in normal speakers. Parsimony, plus the unexamined assumption that mental representation is like a written record (entries either present or absent, structure displayable in two dimensions), has been a misleading guide to modeling language representation. The substantial redundancy in representations and processing that is introduced by incorporating both frequency-based and hierarchy-based information is in fact appropriate for the brain as a fast, reliable, massively parallel error-correcting network with very large storage capacity and gradient representation strength. WIREs Cogn Sci 2013, 4:651–663. doi: 10.1002/wcs.1257
Conflict of interest: The authors have declared no conflicts of interest for this article.
For further resources related to this article, please visit the WIREs website.