A weakness of the fusiform face area hypothesis is that research has provided conflicting evidence for this region being associated exclusively with faces. For instance, using functional magnetic resonance imaging, Gauthier et al (2000) found the area to be active when participants identified the type of bird or car presented and thus concluded the fusiform area is not face specific. Another weakness is that although the fusiform face area is associated with face memory, this does not mean it is responsible for facial memory. It is quite possible that the area is a neurological relay point which passes information from one part of the brain to another and so damage to the fusiform gyrus simply appears to have a casual role in prosopagnosia. On the other hand, a strength of the fusiform hypothesis is that the area is clearly linked to face processing because of abnormal activity in this region leads to hallucinations of faces. The fusiforms anterior portion represents the teeth, eyes and if activated can lead to deformed hallucinations of faces with large eyes or teeth, Sacks (2009). Until there is more definitive evidence that the fusiform area is dedicated to faces alone and that it is more than just a relay point, the most prudent option is to leave it as an open question as to whether prosopagnosia is explained by damage to this region.
Another explanation is prosopagnosia as an impairment of a special face recognition system. This explanation proposes that face recognition and object recognition are subserved by different processes and prosopagnosia arises due to the former being impaired. Of course, the plausibility of this explanation rests upon the extent and quality of the evidence which indicates that face and object recognition can be impaired independently of each other, i.e. that there is a double dissociation. This double dissociation is shown by these examples, the first example shows impaired face processing with intact object processing; Duchaine and Nakayama (2005) examined seven developmental prosopagnosics' memory for faces, horses, cars, guns, tools, houses and natural landscapes by requiring recognition of exemplars within each category. Some of the Ps were within normal range on all, or nearly all of the non-face tasks but were impaired on the face tasks. Similarly, Duchaine et al. (2006) reported another developmental prosopagnosic, Edward, who was poor at face processing, performing no better with upright than inverted stimuli. However, he executed most memory tasks involving non-face items slightly better than healthy controls, even when he was required to recognise within-category exemplars. The second example is that of intact face processing with impaired object processing; Moscovitch et al. (1997) described the case of CK who had impaired object recognition but performed as well as controls on face tasks irrespective of whether the face appeared as a photograph, caricature or cartoon, as long as it was upright and the internal features were in the appropriate locations. Likewise, McMullen et al. (2000) discussed HH whose face recognition was good whereas his object recognition was severely impaired. This strongly indicates that face recognition and object recognition can be impaired independently of each other and that they are in fact separate systems, this is a strength as it indicates that a real qualitative difference is present in the nature of the cognitive deficits.
Duchaine and Nakayama's (2005) and Duchaine et al. (2006) studies are impressive too because of their careful stimulus design, in both cases P's were required to recognise within-category exemplars with objects. This design feature is important because it equated the difficulty of object and face recognition by ensuring that judgements were made according to the shared characteristics of individual stimulus types. Despite the above evidence for a double association, as a weakness this is still not unequivocal confirmation that faces and objects are dealt with separately. Whilst on balance it is increasingly likely that this is the case, until more patients are reported who confirm the existence of a classic double dissociation, it is wise to exercise caution and regard the question as not yet fully answered.
The second explanation of prosopagnosia is the associative and apperceptive prosopagnosia: prosopagnosia as the result of impaired access to facial stored knowledge or perceptual impairment. Evidence exists that prosopagnosia can sometimes result from an inability to access stored knowledge about a face (associative prosopagnosia) or perceptual problems (apperceptive prosopagnosia). Etcoff et al. (1991) described LH who performed well on perceptual tasks that required configural processing but who was nevertheless prosopagnosic. Of course, faces possess a particular configuration and as LH could deal with that kind of processing, it can be assumed his prosopagnosia was more associative in nature, he experienced difficulty in accessing stored knowledge about a face. DeRenzi et al. (1991) reported two patients GD and VA from whom they argued for a separation between a primarily perceptual prosopagnosia and an associative form of the disorder. GD performed much more poorly on perceptual tests than those involving memory for famous faces. VA on the other hand, had the reverse deficit, performing more poorly on famous face than perceptual tasks. VA's case can be explained similarly to LH: perceptual processing was comparatively intact, so VA's difficulty with faces cannot be ascribed to perceptual issues and thus can be accounted for in terms of an associative deficit. With GD, however, there is a perceptual problem which means perceptual difficulties may underlie the facial deficit.
A weakness of DeRenzi et al.'s results is that they do not show a classic double dissociation (Shallice 1988) and instead form a relative dissociation in that the patients were deficient on both tasks, hence the evidence for the two different types of prosopagnosia could be artefactual: uncontrolled differences between the patients may be responsible for their differential impairment rather than real, qualitative differences in cognitive deficit (Parkin, 2002).