ABSTRACT
The functional imaging techniques that will be considered in this chapter are positron emission tomography (PET) and functional magnetic resonance imaging (fMRI). Although there are some subtle differences between these techniques (PET measures blood flow, fMRI measures changes in the concentration of deoxyhaemoglobin), they both measure neural activity by detecting locally specific changes in blood composition and flow that accompany neural activity. The response variable in functional neuroimaging is therefore the haemodynamic response. As in the use of reaction times (RTs), the characterisation of the haemodynamic response involves either categorical comparisons (e.g. animal relative to tool naming) or a correlation with a continuous psychological variable (e.g. increasing word frequency). The former approach is referred to as “subtraction” and the second is referred to as “parametric”. Parametric designs are limited to those variables that offer a continuous measure, and they have not so far been applied to functional imaging studies of category specificity. In fact, all functional imaging studies of category specificity to date have used cognitive subtraction and this is the technique we will therefore focus on. There are two pitfalls in subtraction that deserve consideration. The first relates to the selection of conditions that differ only in the process of interest. This requires tasks that are matched for every other possible variable (word frequency, word familiarity, auditory or visual features, etc.). Obviously, a complete match is impossible because different categories have, by definition, variable orthographic, structural, semantic, and phonological content. The difficulty matching these variables is particularly relevant to functional neuroimaging studies because even if subtle task differences have little effect on RTs, they can have profound effects on the distribution of brain activity. Additionally, condition specific brain activity can be elicited that is unrelated to the task performance and not predicted by the task analysis. For instance, face naming relative to face viewing could elicit emotional responses that change brain activity but do not affect naming RTs. In brief, functional neuroimaging data might be far more sensitive to subtle changes in conditions that relate to different aspects of processing, particularly those components that might be implicit or indeed incidental.
