The aim of the present research was to extend the earlier work of Donald Broadbent on the relationship between cognitive failures scores and the difference between performance of categoric search and focused attention tasks. Results from four experiments, involving over 400 subjects, replicated Broadbent's finding that there was a small negative correlation between level of cognitive failure and the difference in speed of responding in categoric search and focused attention choice reaction time tasks. However, consideration of correlated attributes of cognitive failure showed that this effect could be explained in terms of differences in trait anxiety. It was, however, possible to derive another measure from the tasks, which was correlated with cognitive failure but independent of anxiety. In contrast to the relationship initially described by Broadbent, the correlation between the new measure and cognitive failure proved less reliable, being modified by the different contextual factors in the various experiments.Two of Donald Broadbent's major interests in the 1980s were the measurement and nature of cognitive failures and dimensions of selective attention. These two areas of research were brought together in studies which examined the relationship between cognitive failures, measured by the cognitive failures questionnaire (CFQ), and aspects of selective attention measured using choice reaction time tasks Jones. 1986, 1989). Broadbent et al. (1986) demonstrated that CFQ scores were related to the difference between performance of focused attention and categoric search tasks. A later study (Broadbent et al., 1989) also showed that CFQ scores were related to other aspects of attention (e.g. the Eriksen effect, where distracting stimuli impair performance if they are less than 1 degree away from the target; Eriksen and Eriksen, 1974), but that these effects also depended on background conditions, such as the time of day at which testing took place. The aim of the present research was to extend Broadbent's work in three ways. First, the relationship between CFQ and performance of focused attention and categoric search tasks was re-examined controlling for correlated attributes of cognitive failure. Second, analyses were conducted to determine whether other measures from the two attention tasks were more strongly associated with CFQ than the scores considered by Broadbent. Finally, these issues were examined across several large-scale studies, which varied in background factors, to determine whether any effects were robust enough to be used in studies of human error outside of the laboratory. Correspondence and requests for reprints should be sent to Professor A. P. Smith, Department of Psychology, 8 Woodland Road, Bristol BS8 ITN, UK CCC 0888-4080/95/SIOI 15-12 0 1995 by John Wiley & Sons, Ltd.
