تشخیص چهره لمسی و پروزوپاگنوزیا

Abstract Cases of cross-modal influence have been observed since the beginning of psychological science. Yet some abilities like face recognition are traditionally only investigated in the visual domain. People with normal visual face-recognition capacities identify inverted faces more poorly than upright faces. An abnormal pattern of performance with inverted faces by prosopagnosic individuals is characteristically interpreted as evidence for a deficit in configural processing essential for normal face recognition. We investigated whether such problems are unique to vision by examining face processing by hand in a prosopagnosic individual. We used the haptic equivalent of the visual-inversion paradigm to investigate haptic face recognition. If face processing is specific to vision, our participant should not show difficulty processing faces haptically and should perform with the same ease as normal controls. Instead, we show that a prosopagnosic individual cannot haptically recognize faces. Moreover, he shows similar abnormal inversion effects by hand and eye. These results suggest that face-processing deficits can be found across different input modalities. Our findings also extend the notion of configural processing to haptic face and object recognition.

1. Introduction Cases of cross-modal influence have been noted since the beginning of psychological science. In 1839, Brewster reported that observers who saw indented objects (e.g., engraved seals) through an optical device that inverted apparent concavity, also experienced a haptic inversion effect when they explored these objects simultaneously by touch (Brewster, 1839). The corresponding question—is failure to recognize what one sees also associated with a failure to recognize what one touches—has rarely been raised. In light of the ongoing debate on face specificity and the importance of prosopagnosia to this discussion, it appears highly relevant to ask whether a deficit in face recognition by vision might be associated with a deficit in face recognition by touch (i.e., the haptic system). Neurologically intact individuals process faces more by their overall configuration than by their local features (de Gelder & Rouw, 2000a; Freire, Lee, & Symons, 2000). To investigate this configural or holistic (Tanaka & Farah, 1993) recognition strategy, researchers have predominantly used the inversion effect, which is defined as a decrease in performance when recognizing inverted as oppose to upright faces (Valentine, 1988 and Yin, 1969). Results that show a relatively stronger inversion effect for faces than for other mono-oriented objects have also been interpreted as evidence that faces occupy a special status (Diamond & Carey, 1986) among visually apprehended objects. This weaker inversion effect for non-face objects is presumably due to recognition that is more strongly based on features and less disrupted by inversion (Leder & Bruce, 2000). The inversion effect plays an important role in understanding the visual deficits of patients with a category-specific recognition deficit for faces (prosopagnosia). Some prosopagnosic individuals do not demonstrate the typical inversion effect, while others process inverted faces better than upright faces (de Gelder & Rouw, 2000b; Farah, Wilson, Drain, & Tanaka, 1998). The paradoxical inversion effect (de Gelder & Rouw, 2000b) indicates that configural processing is disrupted but not totally absent. When the need for configural processing is removed (by inverting the face), a feature-based analysis can be performed more easily. Previous studies have been confined to investigating face recognition and its deficits in the visual modality only. Yet there is no intrinsic link between vision and face recognition or prosopagnosia. In fact, intact lower-level visual abilities figure prominently among the diagnostic criteria for prosopagnosia. And disorders of higher cognition can either be limited to a single sensory modality or occur across more than one modality (Feinberg, Gonzalez-Rothi, & Heilman, 1986), depending on whether the information is available to more than one sensory system. Haptic face recognition has recently been demonstrated in normal individuals (Kilgour & Lederman, 2002); however, it has never been studied in prosopagnosics. Can a prosopagnosic individual recognize faces solely by touch? We investigated this question using a haptic inversion paradigm in which our prosopagnosic participant, LH, was required to decide whether two faces (or two non-faces) were the same or different from one another. To date, LH’s sense of touch has never been assessed formally. We therefore also evaluated his sensorimotor hand function.

. Results 3.1. Assessment of sensorimotor hand function LH’s sensory thresholds for tactile pressure sensitivity and tactile spatial acuity were both within normal limits (2.69 mg, and 3.2 mm, respectively). LH showed some impairment on the Grooved Pegboard Test (Lafayette Instrument, 1970) of fine motor control inasmuch as his performance was slower than that of an age-matched normative sample. He performed the test with the dominant and non-dominant hands in 217 and 236 s, respectively. The data for the normative sample are 68 s (S.D.=9.4) and 75 s (S.D.=10.5), respectively. However, like normals, LH made no errors. LH successfully identified 19 of 24 (79%) common household objects by touch alone. For two of the five incorrectly named objects, LH described their function. Moreover, the three remaining unidentified objects were also identified relatively poorly by a neurologically intact sample (Jessel, 1985): 52, 43, and 18% correct. Thus, LH’s accuracy was comparable to that of the neurologically intact sample (overall 88% accuracy). However, he was relatively impaired in terms of his response times for naming the common objects: the mean response time for the neurologically intact sample (n=23) was 3.3 s (S.D.=1.9), whereas LH’s mean response time was 5.6 s (S.D.=6.5). With these preliminary tests we conclude that the accuracy of LH’s haptic performance on the face-processing task was not due to impairments in sensory-motor hand function. His sensory thresholds fell within normal limits. However, we expected LH’s performance with the faces to be slower than the control group. 3.2. Discrimination of face and non-face objects We tested LH’s ability to haptically discriminate upright faces from a set of upright faces, inverted faces, upright non-face objects (teapots), and inverted non-face objects (teapots). Of 64 presentations, LH made one error when he failed to identify an upright facemask as such. 3.3. Haptic face-matching accuracy Fig. 2 (panel A) shows how accurate LH and the control group were when required to judge members of face pairs presented in the same orientation as being the same or different. Given that each trial produced a response that was either correct or incorrect, the data were analyzed using the binomial distribution. Analysis of variance could not be used to compare means as our dichotomous data violated the assumptions of that statistical technique. Of the four conditions—upright faces, inverted faces, upright teapots, and inverted teapots—LH’s ability to determine whether two stimuli were the same or different was at chance level only for the upright faces (P=0.14). His performance with the inverted faces was not statistically different from that with the upright faces (P=0.33), but it was better than chance (P=0.04). LH’s accuracy in discriminating between the upright teapots was significantly above chance (P=0.04). The inverted-teapot condition was statistically better than the upright-face condition (P=0.01) in addition to being significantly better than chance (P=0.001). The superior performance with inverted non-face objects is consistent with earlier findings that LH was significantly better when visually matching non-face objects ( de Gelder & Rouw, 2000b). LH did not show any advantage when touching as opposed to seeing upright facemasks: his ability to haptically discriminate between two upright faces was at chance, as it was when he looked at them. Mean accuracy (%; panel A) and response time (s; panel B) as a function of ... Fig. 2. Mean accuracy (%; panel A) and response time (s; panel B) as a function of stimulus type by participant. Error bars represent S.E.M. Figure options The control group performed above chance in all four conditions: upright facemasks (t(6)=4.7, P=0.005); inverted facemasks (t(6)=11.2, P<0.0001); upright teapots (t(6)=12.1, P<0.0001); inverted teapots (t(6)=23.0, P<0.0001). However, the control group showed a different pattern of results than LH: they demonstrated the inversion effect that is typically found with visual face stimuli. That is, the control group performed better with upright faces than with inverted faces (t(6)=2.6, P=0.04). However, there was no inversion effect with the teapots: the accuracy of discriminating between teapots did not differ with orientation, (t=−1.08, P=0.32). We therefore note that the control group generally had no difficulty discriminating between two upright faces, but their performance deteriorated when the faces were inverted. LH, on the other hand, was unable to discriminate between faces above chance level when they were presented in the upright position; his performance was 1.5 standard deviations below that of the mean of the control group. Furthermore, contrary to the control group, LH improved slightly (i.e., relative to chance) when the faces were presented in the inverted position, and his performance was closer to that of the control group (z=−0.9). Although LH had difficulty discriminating between upright teapots (z=−2.0, relative to controls), he had no difficulty when they were inverted and his performance was similar to that of the control group (z=−0.2). 3.4. Haptic face-matching response time LH’s deficit in haptic-face processing is clearly apparent in the data shown in Fig. 2 (panel B), which presents the response times corresponding to the accuracy results (panel A). Response time was calculated as the total time spent haptically exploring the first and second stimulus objects prior to responding whether they were the same or different. We entered these data into a 2×2×2 analysis of variance, with Participant (LH versus controls) as the between-subjects factor, and stimulus (faces versus teapots) and orientation (upright versus inverted) as the within-subjects factors. Averaged across participants and conditions, mean response time for incorrect trials (29.8 s, S.D.=16.2) and correct trials (22.8 s, S.D.=11.5) were not statistically different from one another (t(31)=1.91, P>0.05); therefore, all trials were analyzed. Although not statistically significant, there was a trend for incorrect trials to be slower than correct trials. Table 1 shows a breakdown of the response times by condition and participant. There was a main effect of Participant, F(1,6)=22.34, P=0.003: overall, LH was slower than the control group. There was also a main effect of Stimulus, F(1,6)=18.04, P=0.005: all participants (LH and controls) were faster responding to teapots than to faces. However, the stimulus×participant interaction term was also statistically significant, F(1,6)=14.05, P=0.01: LH was notably and statistically slower than the control group when responding to the faces but statistically equivalent when responding to the teapots. Although there was no inversion effect apparent in the response–time data for either LH or the control group, our data strongly support a face-specific processing deficit in LH. One reviewer suggested that LH might have been slower with faces than teapots simply because he might have experienced more hesitancy when touching faces as opposed to teapots. We do not believe this to be the case for two reasons. First, it is not clear why this would occur inasmuch as the faces were inanimate masks, as opposed to real faces. Second, the videotapes indicated that LH began exploring faces immediately and without interruption. It is difficult to interpret these response–time data further. Ideally, we would have found a face-inversion effect for the control group consistent with the accuracy data for the face-inversion effect. Similarly, we would have found a paradoxical teapot-inversion effect for LH corresponding to his accuracy data. The inherent variability of response–time data and the limited number of participants may have contributed to the absence of these effects. Table 1. Mean response times (s) and standard deviations (in parentheses) for correct and incorrect trials as a function of condition and participant Condition LH Control group Correct Incorrect Correct Incorrect Facemasks Upright 40.5 (15.1) 43.8 (17.5) 19.77 (5.5) 22.22 (6.4) Inverted 40.7 (11.5) 46.8 (15.0) 19.34 (6.2) 19.35 (6.5) Teapots Upright 29.0 (7.5) 30.0 (3.8) 19.1 (5.1) 20.5 (4.4) Inverted 29.7 (6.9) 36.4 (23.4)a 16.7 (4.8) 20.7 (6.7) a The large S.D. in this condition is attributable to the fact that only two trials contributed to this mean, one of which was anomalously large (52.9s). Table options 3.4.1. Visual face-matching To ensure that LH was processing the facemasks as faces per se, as opposed to non-face objects, we also tested LH’s ability to perform the same/different task with the facemasks visually. His visual accuracy was similar to his haptic performance. When the faces were presented in the upright orientation, his accuracy was only 55% (chance=50%). He also performed at chance level when the faces were inverted (61% correct). The control group was not tested visually with the facemasks for two reasons. First, they were tested with a standardized visual face recognition test (Benton et al., 1994). Second, previous research (Kilgour & Lederman, 2002) on a visual match-to-sample task with the same set of facemasks showed a ceiling effect.