Coexistence of prosopagnosia and object agnosia - a novel approach to understanding how faces are represented in the brain

One of the longstanding controversies in the neuropsychology literature concerns the functional organization of high-level vision, and the extent to which the recognition of different classes of visual stimuli engage modular or distributed representations. The conventional view of object recognition is that high-level visual areas contain a number of functionally specialized modules (Kanwisher, 2010). For example, neuroimaging studies have consistently found a region in the fusiform gyrus - the fusiform face area (FFA) that responds to faces more than to non-face objects (Kanwisher, 2010). Consistent with focal brain damage to this region of the brain leads to a syndrome known as prosopagnosia in which there is an impairment in the perception and recognition of faces, but not other objects (Damasio, Damasio & Van Hoesen, 1982).

The modular view of high-level vision been challenged, recently, by studies that suggest a more distributed representation. For example, although many individuals with prosopagnosia have face-selective deficits, more recent studies have shown an impairment for some non-face objects (Behrmann et al., 2018). It remains unclear, however, why the impairment in face recognition is selective for particular objects. One explanation for these findings is that higher-level visual areas are not selective for categories of objects but rather have a distributed organization that is based on more basic properties of the stimulus (Andrews et al., 2015). From this perspective, the reason that regions respond to faces is because the neural representation in these regions is selective to these properties that are typically found in faces. Support for this has been demonstrated by the positive correlation that is found between low-level visual properties (such as the distribution of orientation and spatial frequency information) of the stimulus and patterns of neural response in the high-level visual areas (Rice, Watson, Hartley & Andrews, 2014). These results suggest that the appearance of category-selective regions may be explained by the systematic convergence of responses to low-level features that are characteristic of each category.

The hypothesis of this proposal is that the objects affected in prosopagnosia will have similar properties to those found in faces. One way to address this issue is to test patients with developmental prosopagnosia (DP) on their ability to recognise and discriminate objects with similar and dissimilar low-level properties to faces. For example, one way to behaviourally test prosopagnosic patients is to use the Cambridge Face Memory Test which is a Delayed Match to Sample (DMS) task. In this task prosopagnosic patients will be shown faces as well as non-face images with similar and dissimilar low-level visual properties to faces. After a certain delay, participants will complete a forced-choice recognition phase where they will have to identify the image that matches the sample one among an array of images. Accuracy and reaction time will be measured as dependent variables. The prediction is that prosopagnosia patients will exhibit selective deficits for visual stimuli that have similar low-level properties to faces but their performance will be normal for images with dissimilar properties to faces.

A second approach is to use fMRI to measure whether the selectivity of the FFA can be explained by the sensitivity to low-level image properties. To address this question, we will compare both the neural pattern of response and the magnitude of this response to objects that have similar basic properties to faces compared to objects that have dissimilar properties to faces in healthy participants. Our prediction is that both the magnitude of response and the neural pattern of response are likely to be more similar to faces for objects that are have similar low-level properties, reflecting selectivity to image properties rather than image category.