The fallacy of the cross-modal Stroop effect: A rejoinder to Cowan (1989)

Abstract

An Attempt at Further Replication of Cowan and Barron (1987) Cowan (1989) pointed to a number of methodological differences between Cowan and Barron's (1987) study and the attempted replications reported by Miles et al. (1989). In particular, the studies differed in (1) the rate of presentation of auditory stimuli (1.66 words/sec and 1 word/sec, respectively), and (2) the size of the potential response set (five items and four items, respectively). It is acknowledged that these methodological differences may plausibly account for the observed discrepancy in the data. Because of this possibility, the experiment reported here adopted a procedure identical in all respects to that of Cowan and Barron (1987), with the exception that the 30 subjects (19 women and 11 men) completed the task in only two auditory conditions: silence and spoken color words. Auditory material (a male voice) was presented in digitized form over headphones at a rate of 1.6 words/sec and at an intensity of 75 dBA. This intensity level is some 10 dBA below Cowan and Barron's reported intensity of 85 dBA (see Cowan, 1989, Note 2). However, as Cowan argued, if the comparison is with silence, then intensity is unimportant. For details of the visual materials, see Cowan and Barron (1987). Both reading times and error rates were recorded. A 2 x 2 analysis of variance (ANDVA) was performed on the reading times with visual condition (color words and Xs) and auditory condition (spoken color words and silence) as within-subjects factors. The main effect of visual materials [F(I,29) = 496.07,p .89, sign test). With respect to reading errors, a 2 x 2 ANDVA revealed a main effect of visual materials [F(I,29) = 37.33, p < .001]. Again, neither the main effect of auditory condition [F(1,29) = 0.22] nor the interaction effect approached significance [F(1,29) = 0.09; see Table IJ. Clearly, these data unequivocally failed to confirm the results presented by Cowan and Barron (1987). Reading times in this study for the spoken color-word condition are comparable across visual materials to those reported by Cowan and Barron; however, the error rate reported here is below that reported by Cowan and Barron (1987; see Figure 2). In the present study, the subjects encountered spoken color-word interference on two out offour trials (50%), compared with 2 out of 10 trials (20%) in Cowan and Barron. However, in absolute terms, subjects in both studies received the same number of spoken color-word trials. Therefore, Cowan's (1989) suggestion that the crossmodal Stroop interference effect might attenuate with increased experience cannot account for our failure to replicate Cowan and Barron's original findings. In sum, the present study, with only minor procedural differences from Cowan and Barron's study, produced data completely consistent with our earlier findings, and in direct contradiction to the data of those authors.