A potential confound does not a theory make: Comment on Salis and Edwards (2006)

Abstract

Background: This article responds to the abstract report of a study by Salis and Edwards (2006). Aims: The aim is to show how that article, and prior articles, have incorrectly interpreted the hypothesis of Cupples and Inglis (1993). A further purpose is to critique other weaknesses of the research paradigm of sentence-picture matching. Main Contribution: It is argued in this paper that it is not appropriate to rule out a potential confound merely on the basis of finding that it does not seem to influence certain patients. Potential confounds should be controlled in the design, for otherwise the result of any study is ambiguous; the effects of the confounded variables cannot be teased apart. Ruling out the confound per se could only come from lack of support for the construct by researchers in the field. In view of the fact that such support exists, the confound remains, and should be addressed by any researcher continuing to make use of sentence-picture matching. Conclusions: The assessment task of sentence-picture matching has been superseded by other tasks which control for the particular confound found by Cupples and Inglis (1993). Testing patients for the confound, even if it turns out not to be operative, is a sign of good experimental control rather than any reflection on the validity of the confound. The task still has perhaps a screening role, but should not be relied on as a valid measure of sentence comprehension. There are inconsistencies in how it is administered across studies, which make comparisons difficult. A range of offline sentence-comprehension tasks is needed in order to be able to triangulate on the nature of a particular patient's comprehension deficit when no single task is ideal.