"In the study of visual pattern recognition, most people's initial assumption is that we have a single, general purpose system for recognizing all the different types of stimuli in our visual world." But this assumption is wrong -- vision is highly modular. This will be illustrated by focussing on face recognition. I.E., face recognition does not use general purpose object recognition processes. The point of the chapter is to provide evidence of this.
What kind of evidence could support the claim that face recognition and object recognition use different mechanisms? (1) Functional independence. (2) Physical distinctiveness. (3) Qualitatively different types of process.ing. Farah will explore these three kinds of evidence as they relate to face perception.
Visual agnosia is "an impairment of ojbect recognition that is not attributable to a loss of gerneral intellectual ability or to an impairment in such elementary visual perceptual processes as brightness, acuity, depth, and color." In associative agnosia, patient can copy an ojbect, but cannot recognize it -- but this is still a visual problem "Associative visual agnosia is currently viewed by most neuropsychologists as an impariment at the highest levels of visual representation, rather than as an inability to associate normal visual representations with other types of knowledge."
Prosopagnosia is the inability to recognize faces after brain damage. "Prosopagnosics cannot recognize familiar people by their faces alone, and must rely on other cues for recognition, such as a person's voice or distinctive clothing or hairstyles. The disorder can be so severe that the patient will not even recognize close friends and family members."
Most basic view -- high level vision is modular, and prosopagnosics have lost the fast recognition module, but have otherwise normal object recognition. But...maybe faces are just really difficult for general processes to deal with. "Prospagnosia could then be explained as a mild form of agnosia in which the impariment is detectable only on the most tasking form of recognition task." To see if this the case, studies must take into account stimulus complexity -- controlling complexity of faces and nonfaces. E.g., for one prosopagnosic (LH), a recognition memory task was used in which later recognition memory for faces and nonfaces was equated on the basis of studies with normal subjects. "LH showed a significantly larger performance disparity for the two stimulus sets than the normals, achieving only 62 percent correct for faces and 92 percent correct for objects." Similar results were found when all the nonfaces came from one object category (eyeglass frames).
Finally, would LH show a face inversion effect, because for normal subjects "inverting a face makes it much harder for normal subjects to recognize." A same-different judgement task was used. Normal subjects performed better with upright faces than with inverted faces. "LH's results were more surprising. He was significantly more accurate with inverted faces, achieving 58 percent correct for upright and 72 percent correct for inverted faces!" (NB: This was a surprise, because the prediction was no difference, under the assumption that LH would not be using any face recognition processors at all.) The surprise suggests that "LH's specialized face-perception system was contributing to his performance, even though it was impaired and clearly maladaptive." I.e, even when impaired, the action of the face recognition module is mandatory!
Bottom line: face recogntion involves specialized mechanisms.
Even more specialized impariments than LH's can be found. "CT is imparied at learning new faces, but his ability to recognize previously familiar faces and to learn other nonface visual objects is relatively intact. This pattern of performance is consistent with a disconnection between intact face representations and an intact medial-temporal memory system."
"Some associative agnosics appear to have more difficulty with object recognition than with face recognition, presenting us with the miror image of the prosopagnosic's impaired and spared abilities." I.e, a double dissociation. Issue: are modules arranged as serial processing stages (object recognition first, which then feeds into a later face recognition system) or do both work simultaneously in parallel?
Existence of patients with object agnosia, but normal face perception, indicates that the two systems must work in parallel -- otherwise, with the serial view, face perception would be affected by object recognition damage.
What different kinds of visual processing might be carried out by these distinct object recognition systems? Examination of a range of impairments suggests the following hypothesis: "There are two systems, one of which is essential for face recognition, useful for common object recognition, and not at all needed for printed word recognition, and the other of which is essential for printed-word recognition, usefule for common object recognition, and not at all needed for face recognition." I.E., nonface system finds printed words more difficult to process than common objects. Words are composed of numerous individually recognizable parts (letters). "The conjecture being put forth here is that word recognition involves extensive part decomposition and, therefore, requires the ability to represent a large number of parts; face recogntiion, on the other hand, is holistic in that it invovles virtually no part decomposition, and hence requires the ability to represent complex parts." In other words, two systems, each with complementary (but, importantly, different) systems of shape representation.
Test of holistic representation -- how well can an object be recognized if subject is only presented a part of it? If representation is holistic, recognition should be impaired. Indeed, there are poor results for face recognition when only parts are presented. Similarly, "part masks" mask nonface objects better than faces; this effect is only true for upright faces! Finally, LH was studied -- and showed no difference for recognizing faces from parts vs. whole faces, in contrast to normals. "He can no longer benefit from seeing faces as wholes."
Basic points:
Pearl Street | "An Invitation To Cognitive Science" Home Page | Dawson Home Page |