ABSTRACT
To investigate whether feature correlations can account for the seven category-level trends described above, the 37 categories were sorted in descending order in terms of the mean percentage of correlated feature pairs per concept for each category (Table 8.2). To assist in understanding this analysis, the categories are labelled by seven sets: creatures, non-living things, fruit/ vegetables, musical instruments, jewellery, food, and trees (the importance of which will become apparent). The most noticeable aspect of Table 8.2 is that there exists a great deal of variance within each domain, so that the sets do not group cleanly. For example, the creatures and non-living things are separated for the most part, but insects, amphibians, and reptiles have a low percentage of correlated feature pairs, whereas fashion accessories, clothing, and buildings have a high percentage. Furthermore, three of the fruit/vegetable categories are dense in terms of correlated feature pairs (as are trees and food), whereas plant is ranked 20th of 37. Finally, Table 8.2 might provide some insight into two counterintuitive results. Both musical instruments and jewellery are firmly embedded within the creatures and fruit/vegetables in terms of percentage of correlated feature pairs. Therefore, this might be one reason why these categories have been found to pattern with living things in terms of patients’ deficits. In summary, although ranking categories in terms of feature correlations mirrors some of the key phenomena in the category-specific deficits literature, it fails to mirror others.
