Dependence by any other name smells just as sweet: reply to van der Velde and van der Heijden (1997)

The hypothesis that people selectively attend to entire objects predicts that all attributes of an object will be reported either very accurately (if the object was attended) or very inaccurately (if it was unattended). Hence, reports of object attributes should show positive dependence. M. Monheit and J. Johnston (1994) have confirmed this prediction. F. van der Velde and A. H. C. van der Heijden (1997), however, have argued that dependence in the overall data is spurious. They advocate a model that partitions the data into 2 subsets, 1 for perception trials and 1 for guessing trials, each of which separately exhibits independence. Here, the authors argue that this treatment of guessing is misguided because, in effect, guesses are discarded rather than treated as failures of perception. The Monheit and Johnston analysis, on the other hand, is fundamentally sound and demonstrates precisely the kind of dependence predicted by the spatial attention hypothesis.