What's the use of picture discrimination experiments?

Abstract

If we want to know whether it is useful to examine picture set discriminations, we need to ask which of these reasons lies behind the experiment in question. Of course, things are never as simple as that. Researchers are not always either clear or consistent about why we do experiments, and often more than one of these motivations may be detectable in the design and description of their research. Furthermore, those who read and cite research often attribute motives to the researchers that they did not in fact hold, or interpret results in ways that the original authors would not endorse. Nonetheless, the research techniques that are appropriate depend critically on which of these motivations are operative, and it follows that how we should assess the experiments and our results depends critically on which of these goals they are aiming at. In this commentary, I argue that the underlying motivation for most of the research that Weisman and Spetch (2010) discuss is (c) above, to investigate object representation. However, much of Weisman and Spetch’s (2010) critique assumes that it is (a), to investigate the discriminability of the real objects. They build a damning case against the use of much of the published research to answer that question. But it is not clear that there is anyone in the dock – and in any case, it will be argued below that it is not improper to argue from the discriminability of pictures to the discriminability of the objects depicted, though the converse argument does indeed fail. Weisman and Spetch (2010) also argue strongly for the use of playback experiments, which are indeed the most usual and useful method for answering questions driven by (b) above, discovering the releasers of natural responses, but of limited use if what we are really interested in is object representation.