Qualitatively Different Semantic Representations for Abstract and Concrete Words: Further Evidence from the Semantic Reading Errors of Deep Dyslexic Patients

Abstract

This paper presents an investigation of the hypothesis that conceptual knowledge for abstract and concrete items is underpinned by qualitatively different representational frameworks (Crutch and Warrington, 2005a). A re-analysis of the semantic reading errors of four deep dyslexic patients is presented, examining the incidence of semantically associated and semantically similar errors in response to abstract and concrete target words. The results demonstrate that abstract target words elicit a greater proportion of associative than similar errors, while concrete words show the reverse pattern. These findings provide evidence which converges with that previously documented for a semantic refractory access dysphasic to suggest that abstract concepts are represented in an associative network while concrete concepts are represented in a categorical framework.

The author would like to express his gratitude to Professor Elizabeth Warrington and Professor Max Coltheart for their helpful comments on an earlier version of this paper. He is also indebted to Mr. Chris Frost for his advice on the statistical analysis of the data, and to Professor Martin Rossor for his support of this work. The author is an Alzheimer's Research Trust Fellow.

Notes

The author would like to express his gratitude to Professor Elizabeth Warrington and Professor Max Coltheart for their helpful comments on an earlier version of this paper. He is also indebted to Mr. Chris Frost for his advice on the statistical analysis of the data, and to Professor Martin Rossor for his support of this work. The author is an Alzheimer's Research Trust Fellow.

¹Concreteness constitutes a continuous variable. As a consequence, the method for classifying concepts into ‘abstract’ and ‘concrete’ classes on the basis of normative data rather than qualitative definition or intuition is not clear-cut. In some previous studies, specification of high or low concreteness has been equated to whether the rating of a particular item falls above or below an arbitrary point in an established range of concreteness ratings such as that of Brown and Ure (1969; e.g., Warrington, 1981; Breedin et al., 1994). Other investigators have gathered their own concreteness ratings by asking subjects to rate items on a scale from 1 (highly abstract) to 7 (highly concrete) and thence defining words as abstract (<4) or concrete (>4) relative to the central limit of the range (e.g., Kroll and Merves, 1986). In the current study, the former technique was employed, with reference to normative data from the MRC Psycholinguistic Database which presents ratings ranging from 100–700. However, a strict central limit was not imposed following analyses suggesting that stimuli commonly used in neuropsychological investigations of concrete word naming and comprehension (e.g., Snodgrass and Vanderwart, 1980) typically fall within the range 550–660 on this measure (Crutch and Warrington, 2003b).