Dissociation of local and global processing in visual agnosia

Yision Res. Vol. 34, No. 7, pp. 963-971, 1994 Pergamon Copyright 0 1994 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0042-6989/94 $6.00 + 0.00 Dissociation of Local and Global Processing in Visual Agnosia INGO RENTSCHLER,*t BERNHARD TREUTWEIN,* THEODOR LANDISS Received 10 October 1992; in revised form 20 August 1993 Subsequent to strokes in the right and left inferomedial occipito-temporal lobes, two patients became prosopagnosic and alexic, respectively. They also show a complementary dissociation of the analysis of handwritten text. The patient with the right posterior stroke can read it but not recognize whose handwriting it is; the patient with the left posterior stroke cannot read the text but knows who wrote it. The analysis of spatial vision revealed that the prosopagnosic patient has no problem with seeing texture elements when presented in isolation. Yet she performs poorly with Moire and texture perception, i.e. she sulfers from a selective loss of global visual perception. The alexic patient performs well with Moire patterns but neither with (complex) texture elements nor with textures. She seemingly can locally and globally process patterns composed of simple figural elements but fails with stimuli that require the integration of features. This finding of a concomitant dissociation of local and global visual processes in the two patients supports the view that prosopagnosia as well as alexia are the most conspicuous aspects of more general alterations of visual perception. Visual agnosia Prosopagnosia Alexia Contrast sensitivity Moire perception Texture perception INTRODUCTION Recently we have described the clinical conditions of two patients with remarkably complementary dis- The view of a domplementarity of function with respect sociations of face recognition and reading with respect to the two cerebral hemispheres (Wigan, 1844; Jackson, to the side of lesion (Landis & Regard, 1988). One of 1874) is older than the discovery of the dominance of the these patients, KD, has a right-sided infero-medial oc- left hemisphere for language (Broca, 1861). Neverthe- cipito-temporal lesion and cannot recognize familiar less, the concept of complementarity of visual disturb- faces, i.e. she suffers from prosopagnosia (Landis, ances, especially that of a complementary dissociation of Cummings, Christen, Bogen & Imhof, 1986, case 2). face recognition and reading with respect to the side of Similar to other prosopagnosic patients, she scores in the lesion, was not considered until 1937, when Hoff and lower range of normal controls in face matching exper- Poetzl published a report entitled “On anoptic-agnosic iments provided that the stimuli are presented for an disturbance of memory for physiognomies”. Not only unlimited time. When exposure duration is reduced, her did these authors consider for the first time the inability performance worsens dramatically (Christen, Landis & of identifying familiar faces, later called prosopagnosia Regard, 1985). KD also shows deficits in the recognition (Bodamer, 1947), to be a specific form of visual agnosia, of facial expressive gestures but can lipread speech but also specifically opposed the recognition of letters (Campbell, Landis & Regard, 1986). Most strikingly, and words to that of the recognition of physiognomies. this patient is unable to identify the author of handwrit- These two functions were viewed by Hoff and Poetzl as ing familiar to her, whereas she has no problem in complementary, and consequently prosopagnosia and reading it (Landis & Regard, 1988). Such an observation pure alexia were interpreted as reciprocally linked symp- has also been made earlier (Niessl von Mayendorf, toms. Recently the dissociative nature of prosopagnosia 1933). The other patient, MT, with a left-sided occipito- vs pure alexia has gained new interest, regarding their temporal lesion and a right hemianopia, shows no anatomical differences and possible differences between deficits in face recognition or the classification of facial the two syndromes in early visual processing (e.g. expressive gestures but she is alexic and impaired at Sergent, 1982; Levine & Calvanio, 1989; Farah, 1990; lipreading (Campbell et al., 1986). The latter patient can Griisser & Landis, 1991). identify the author of familiar handwriting but is unable to read the written words (Landis & Regard, 1988), again an observation made earlier by Alajouanine, *Instituteof MedicalPsychology,Universityof Munich,Goethestrasse 31, 8000,Miinchen2, Germany. Lhermitte and de Ribaucour-Ducarne (1960). tTo whom all correspondenceshould be addressed. These findings cast doubt on the notion that proso- SDepartmentof Neurology,UniversityHospital,Ziirich,Switzerland. pagnosia and alexia are caused by the loss of highly 963 964 INGO RENTSCHLER et at. specialized visual functions. Rather it seems that these discriminate radial Moire: patterns with different degrees alterations of visual perception constitute the most con- of spatial decorrelation, as well as textures made up of spicuous aspects of more general problems that the right compound Gabor patterns. Local visual processing was and the left hemispheres have, due to their lesions, with analysed in terms of disc~mination sensitivities to iso- the processing of complex and meaningful spatial pat- lated compound Gabor patterns with different spatial terns. This view was, in recent years, surrounded by waveforms. Contrast sensitivity function and optotype considerable controversy for both prosopagnosia (see acuities of the prosopagnosic and the alexic patient were Sergent, 1982; Farah, 1990; DavidofT & Landis, 1990; measured for control purposes. Griisser & Landis, 1991) and alexia (see De Renzi, Zambolin & Crisi, 1987; Farah, 1990). With these issues CASE HISTORIES in mind, it occurred to us that recognition and reading of handwritten text can be thought of as tasks of global Since the case histories, including CT scans, of both and local visual processing, respectively. This led us to patients KD and MT have been published on several investigate in the two patients perceptual abilities for occasions in the frame of different experiments (Christen which the dichotomy of global vs local processing has et al., 1985; Campbell et al., 1986; Landis et al., 1986; been extensively discussed in the literature, namely Landis & Regard, 1988; Davidoff & Landis, 1990; MoirC: and texture perception. Griisser & Landis, 1991) their clinical picture is given The MoirC effect (Glass, 1969; Glass & Perez, 1973; here only briefly in table form. Glass & Switkes, 1976) can be observed in dot patterns, As Table 1 shows, both patients exhibit extremely or patterns made of a wider class of micropatterns, simiiar pictures of primary ophthalmolo~~al and neuro- consisting of two su~rimposed copies of a random-dot logical symptoms except for the side of lesion. However, pattern where one copy has been slightly rotated, trans- the clinical pictures strongly dissociate at the level of lated, or dilated. In such patterns one perceives locally higher visual associative processing. parallel structures, which globally reflect the invariance properties of the underlying geometric transformation, GENERAL METHOD i.e. concentric circles, parallel streaks, or radial streaks. Texture (Beck, 1972; Julesz, 1981; JuIesz & Bergen, 1983) is a global visual perception in the sense that, with Stimulus patterns were generated as digital images the awareness of an overall picture, one sees a large with 8-bit grey levels on a TV monitor (Barco TVM number of local figural elements, or micropatterns as a 3/3.2, P4 phosphor) linked to a Videograph image coherent region in the scene. This ability allows the processing system with 50 Hz frame rate (interlaced). human observer to segment images into parts and, The frame buffer was interfaced to a LSI 11/73 com- therefore, rapidly identify physical objects. puter. The mapping between frame buffer content and In the present study we investigated the two patients’ display luminance was linear (for details see Rentschler, abilities of global visual perception by having them Hiibner & Caelli, 1988). TABLE 1. Clinical and neuropsychological findings in the patients KD and MT KD MT Right-handed secretary, Right-handed housewife, known hypertension (1 yr) uneventful medical history Right posterior cerebral artery Left posterior cerebral artery stroke in 1981 at age 61 stroke in 1984 at age 64 Left homonymous hemianopia Right homon~ous h~ianopia (partial recovery of the lower left quadrant) Visual acuity (decimal notation) 0.74.8 Visual acuity 0.9-1.0 CT: large right medial temporo-occipital CT: Large left media1 occipito-temporal hypodensity including fusiform, lingual hypodensity including f&form, lingual and posterior para~p~oc~pal gyrus and posterior par~~pp~rn~ gyrus Dense and persistant prosopagnosia, Pure alexia, initially global, initially topographagnosia, impaired recovered to unreliabb single letter memory for non-verbal material but reading, initially associative object relatively spared memory for verbal agnosia for line drawings and photographs, material colour anomia, impaired memory for verbal material, relatively spared memory for non-verbal material Recognition of familiar handwriting impaired Intact recognition of familiar handwriting Lip reading intact Lip reading impaired Episodes of visual pseudo-hallucinations Mood depressed Mood unchanged  LOCAL AND GLOBAL PROCESSING IN AGNOSIA 965 perception was studied by measuring discrimination sensitivities for prototype/distorted-copy stimulus pairs. Stimuli and procedure The Moir& stimuli (Fig. 2) were generated in a 512 x 512 pixels format using one random distribution of dots. Seven copies of this pattern were perturbed by randomly displacing each dot within a square area whose side was n = 1, 5, 9, 13, 17, 21, or 25 pixels (with the original corresponding to n = 1). The resulting maxi- mum vertical and horizontal displacement was (n - I)/2 pixels. The Moire patterns were obtained by superimpos- ing a copy of the original pattern and one of the seven I I perturbed copies with a relative rotation angle of 3 deg I I I 1 2 4 8 16 about the centre of the patterns. Dot luminance was 80 cd/m*; background luminance was 18 cd/m*. Stimulus Spatial frequency (c/deg) size was 4.5 x 4.5 deg at a viewing distance of 220 cm. FIGURE 1. Contrast sensitivity functions of the two patients, KD and Exposure duration was 120 msec. MT, to vertical sinusoidal gratings. Binocular viewing conditions. Exposure duration, 1 set, space average luminance, 60 cd/m*. The 95% The Moirt patterns were shown one at a time. Sub- statistical confidence interval of threshold estimates was set at 0.15 log jects were instructed to press one of six buttons, depend- units. ing on whether they believed that a pattern with pronounced (buttons l-3) or weak (buttons 4-6) struc- ture was being displayed. From these rating responses, Experiment I: Contrast Sensitivities for Sinusoidal signal detection parameters were obtained for each of Gratings the perturbed dot patterns (see Green & Swets, 1975, Stimuli and procedure chap. 4). A suitable measure for discrimination sensi- Six sinusoidal gratings and one blank field were tivity is the area under the receiver operating character- generated as 128 x 128 pixel digital images. Stimulus size istic (ROC). This measure is equivalent to the percentage was 2 x 2 deg at a viewing distance of 123 cm; spatial of correct responses in a 2-AFC situation. It was deter- frequencies were 1, 2, 4, 8, and 16 c/deg. Space average mined by least-squares fit (Program RSCORE, L. 0. luminance of patterns was 60 cd/m*; background lumi- Harvey Jr; original version by Dorfman & Alf, 1969). nance was 5 cd/m*. Threshold contrasts were found by The reliability of this parameter is given as SE which can means of an adaptive psychophysical procedure with be used for statistical hypothesis testing in the same way maximum likelihood estimation (Harvey, 1986). A tem- that the SEM is used [95% confidence interval for poral two-alternative forced-choice (ZAFC) procedure a parameter value is this value + 1.96 * SE (Harvey, was used for experimental trials. The latter continued unpublished observation)]. until a criterion of 82% correct had been reached. The Results were derived from four (MT) and two (KD) 95% statistical confidence interval of the corresponding experimental runs totalling 80 and 40 exposures per threshold value was set at 0.15 log units. To achieve this pattern, respectively. degree of accuracy, between 20 and 40 trials were required for each stimulus pair. Stimuli were binocularly Results and discussion tested with 1 set exposure duration and 700 msec ISI. Figure 3 shows the results for the experiment on MoirC perception. The alexic patient, MT, had no Results and discussion problem with detecting the gradually increasing pertur- The results are presented in Fig. 1, where contrast bations of the original MoirC pattern. Indeed, her per- sensitivities (i.e. the logarithms of the inverse detection formance was virtually the same as that of two younger thresholds) are plotted vs spatial frequency. The two control subjects (aged 26 and 27 yr; data not shown). patients’ data are virtually identical and display the The performance of the prosopagnosic patient, however, usual inverted-U shape. Clearly, there is no way to was much worse. predict their behavioural deficiencies from the contrast To assess the implications of these findings, we note sensitivity function (or from visual letter acuities; see that MoirC perception has at least two components. The Table 1). one is the detection of locally parallel structure, and the other is the appreciation of the properties of the under- lying global transform of the random-dot distribution, Experiment II: Moir& Perception i.e. of concentric circles in the present stimulus patterns. This experiment measured the perception of Moirt What can be inferred from the data shown in Fig. 3 is, structure in two-tone dot patterns such as discussed by therefore, that KD has a problem with recovering either Glass et al. (Glass, 1969; Glass & Perez, 1973; Glass & the local or the global component of Moirt structures or Switkes, 1976). A prototype Moir& pattern and a number both, whereas these functions are seemingly unimpaired of distorted copies thereof were generated and MoirC in MT. 966 INGO RENTSCHLER er al. Experiment III: Micropattern and Texture Discrimination Rizzolatti and Buchtel (1977) in the context of face The aim of the following experiment was to decide recognition. whether the deficiency of KD in seeing global structure stems from a weakness in global or local visual process- stimuli and procedure ing. To examine this issue, we compared the discrimi- Stimuli were generated as 256 x 256 pixels images. nation of micropatterns with that of textures made up Stimulus size was 6.2 x 6.2 deg at a viewing distance of from such micropatterns. The rationale of this exper- 80cm; space average luminance was 60cd/m2. Com- iment is the following. We can assume that the discrimi- pound Gabor signals, or compound grating patches (see nation of micropatterns depends on local processing if also Lawden, Hess & Campbell, 1982; Lawden, 1983), they are presented successively to the same retinal lo- were used as micropatterns. They consisted of small cation, whereas the study of global visual processing sub-images (32 x 32 pixels) which were generated by requires the grouping of such micropatterns in textures superimposing two Gabor signals (see Marcelja, 1980) of (see Beck, 1972; Julesz, 1981). same location and same parameters of the Gaussian It would be erroneous, however, to assume that a window [see Fig. 4(a)]. The modulation frequencies of task of texture discrimination necessarily involves the superimposed Gabor signals were kept fixed at global visual processing. Unless a suitable exper- f, = 2 c/deg, andf, = 6 c/deg. The respective (Michelson) imental paradigm is used, the discrimination of two contrasts were also kept fixed at 70 and 23%. The successively displayed textures might be simply achieved fundamental& was in cosine phase, the third harmonic by monitoring a fixed location in the stimulus display. If fX either in 0 or 180 deg phase. Hence the two types of there were a difference in micropatte~s detected, this mi~ropatterns differed in spatial phase by 180 deg. The could then be the clue to the existence of a texture resulting waveforms are shown in Fig. 4(a). difference. To prevent our subjects from using such a In case of micropattern discrimination, test patterns local strategy, we employed for measuring texture dis- were centered on an otherwise empty image set at space crimination an experimental paradigm introduced by average luminance [Fig. 4(b)]. Stimulus pairs were *. _- -0.. l . . ..* FIGURE 2. MoirC test patterns. Four of the seven stimulus patterns with zero, intermediate and maximum spatial distortion (from top left to bottom right’) are shown.  LOCAL AND GLOBAL PROCESSING IN AGNOSIA 967 The fact that KD had no problem with distinguishing the micropatterns per se (i.e. when seen in isolation, at the same retinal location, and at sufficiently long ex- posure duration) is consistent with her relatively good visual acuity and contrast sensitivity (Expt I). Her poor performance with the texture discrimination task could simply mean that this task is very difficult. However, 8 o.8 cr; normal subjects have no problem with discriminating z such textures even when the phase separation between x , 0.7 - the two types of micropatterns is only 120deg 2 [Rentschler et al., 1988, Fig. 6(a)]. Thus we conclude that a KD, is selectively impaired in texture discrimination. 0.6 - This answers the question that we were left with from Expt II, in that it is the global and not the local component of texture perception which is deficient in 0.5 this patient. t The data obtained from MT prove that she suffers I I I J- I I I from a weakness in local visual processing. Given her 0 2 4 6 8 10 12 good performance in Moire perception (Expt II), this Pattern distortion suggests that her weakness in discriminating compound FIGURE 3. Discrimination of Moire. patterns. Abscissa, maximum Gabor textures is simply a consequence of her inability vertical and horizontal displacement of corresponding dots, i.e. to properly see their micropatterns. This raises the amount of spatial decorrelation in stimulus patterns. Ordinate, area under ROC as obtained from a signal detection rating procedure. question why MT is impaired in seeing compound Exposure duration, 120 msec. Solid symbols, prosopagnosic patient Gabor patterns but not in dots (i.e. the elements of KD (40 exposures per condition); open symbols, alexic patient MT (80 Moire patterns). The answer might be that there exist exposures per condition). Error bars: + 1 SE. visual neurons specialized for the detection of dots [“dot responsive cells” (see Zetzsche 8z Barth, 1990)], whereas sequentially displayed for 120 msec with an IS1 of the discrimination of compound Gabor patches involves 700 msec. Subjects were instructed to press one of six the (nonlinear) integration of the outputs of at least two buttons, depending on whether they believed that a pair types of detectors (Caelli, Rentschler & Scheidler, 1987). with same (buttons l-3) or different (buttons 4-6) In other words, it seems that MT’s problem with local patterns was being displayed. Discrimination perform- visual processing is not located at the level of visual ance was evaluated as area under ROC as in Expt II. resolution but at some integrative neural function of In case of texture discrimination, two types of non- local pattern analysis. This would explain why her visual targets (NT) and two types of targets (T) were used. The deficit does not interfere with visual acuity and contrast non-targets consisted of displaying textures made up sensitivities. from 8 x 8 identical micropatterns (nontarget NT, con- sisting only of micropatterns with 0 deg phase; non- GENERAL DISCUSSION target NT, consisting only of micropatterns with 180 deg phase). The targets consisted of 8 x 8 textures made up We studied Moire and texture perception in two from a central array of 4 x 4 micropatterns with 0 deg patients with ischemic infarctions in the territory of the phase surrounded by micropatterns with 180 deg phase right and left posterior cerebral arteries, respectively. (T,), or vice versa (T,). The four non-target and target They not only dissociated with respect to the recognition textures are shown in Fig. 4(b). of faces and words but also in the meaningful analysis Texture stimuli were presented one at a time with of handwritten text. The patient with the right posterior 120 msec exposure duration. Subjects were presented at lesion could read the text but could not recognize who each experimental condition with 100 non-targets and 50 had written it. The patient with the left posterior stroke targets. They had to rate their confidence of whether could not read but recognized the author of the text. they saw a non-target (l-3) or a target texture (46). We found that the prosopagnosic patient performed Discrimination performance was evaluated as area poorly with Moire (Expt II) and texture perception under ROC as in Expt II. (Expt III). This weakness may be the result of a defi- ciency in either the local or the global or both modes of visual processing. Yet the patient had relatively good Results and discussion visual acuity and contrast sensitivity, and she had no Figure 5 shows the results of Expt III. The proso- problem with the discrimination of micropatterns (i.e. pagnosic patient, KD, displayed perfect discrimination compound Gabor patches, Expt III) which apparently sensitivities for micropatterns, whereas with textures her require more complex local processing. We conclude performance virtually dropped to chance level. By con- that she must be deficient at some aspect of global trast, the alexic patient, MT was about equally impaired processing, a result which implies the existence of a with micropatterns and textures. disassociation of local and global processing. 968 INGO RENTSCHLER et al. The alexic patient performed well with Moire patterns the textures have more energy at lower spatial frequen- but not with compound Gabor micropatterns and tex- cies. tures. That is, she can locally and globally process the Thus we shall proceed to considering more elaborate dots of which Moire patterns consist but has difficulties concepts of local and global modes of visual perception. to do so with stimuli that require the integration of To begin with, the original explanation of the Moire features. We have no definite proof that this difficulty effect by Glass (1969) had two components. One was the arises only at the level of local processing but the findings stimulation of oriented receptive fields of cortical simple with Moire patterns suggest that this is the case. If this cells by pairs of dots. These cells were expected to evoke were true, we could assume that the alexic patient also a relatively strong signal due to the locally parallel displays a dissociation of local and global visual percep- orientation of correlated pairs (i.e. dots and their trans- tions. formed images). The other was the recovery of such One may ask whether these findings are consistent signals from noise resulting from random stimulation of with the hypothesis that the intact right cerebral hemi- cortical units by uncorrelated pairs. For obvious sphere is faster and more accurate in processing low reasons, the latter process involved integration over spatial frequencies and the left hemisphere in processing limited regions of the random pattern. This concept has high spatial frequencies (Sergent, 1982). In the present been examined by Stevens (1979) who computed locally experiments, the spectra of the texture elements (Gabor parallel structure from the orientation statistics of “vir- patches) contained more energy at lower frequencies tual lines” connecting pairs of dots (see also Marr, 1982). than those of the Moire elements (dots). From this one Yet the same author noted that “the global structure is would expect that the alexic patient, due to her intact derived from the local pairings and constitutes a later, right hemisphere, would perform better than the proso- distinct computational problem” (Stevens, 1979, p. 23). pagnosic one in micropattern discrimination. This, how- The situation is the same with Encke’s (1990) study, who ever, was not the case. In much the same way, the analysed Moire perception in terms of local autocorrela- argument fails with the global patterns of which, again, tion functions. Thus we can assume that global processes (a) + FIGURE 4(a). Caption on facing page.  LOCAL AND GLOBAL PROCESSING IN AGNOSIA 969 Tl T2 FIGURE 4. Compound Gabor micropatterns and textures. (a) The micropatterns consisted of two superimposed grating patches with 2 and 6c/deg modulation frequencies. The two types of micropatterns used differed in spatial phase of third harmonic by 180deg [peaks add, (a) bottom left; peaks subtract, (a), bottom right]. (b) For measuring micropattem discrimination, these two stimuli were presented in isolation on a background of space average luminance (top). For measuring texture discrimination, four types of texture pairs were used. Two of them were non-targets [NT,, NT,, (centre)], consisting of identical micropattems in the centre and surround areas. Two of them were targets [T,, T, (bottom)] consisting of the two different types of micropatterns in the centre and the surround areas. act on the results of local processes in Moirk perception, the extraction of features by means of spatial filters of although we are ignorant about the nature of the global the centre-surround type and the evaluation of the process itself. spatial density of such features (“textons”, e.g. elongated Theories of texture perception are also based on the blobs, line crossings, and line ends). Computational assumption of local and (more) global processes of models of texture perception addressed the problem of visual pattern analysis. Psychologically oriented theories how these texture boundaries are formed (Caelli, 1985, (Julesz, 198 1, 1986; Beck, 1983) characterized texture via 1988; Malik & Perona, 1990; Vorhees & Poggio, 1988). 970 INGO RENTSCHLER ef al. They have again in common that they assume filtering of dissociations of local and global visual processing in by receptive fields (possibly completed by further local our patients. processing) and global evaluation of resulting neural This conclusion implies an affirmative answer to the activities via associative networks. question of whether KD’s prosopagnosic condition is These similarities between models of texture and one aspect of a more general alteration of perception and Moir& perception may not be by coincidence. On the one the situation seems to be the same for the alexic patient hand, as has been conjectured by Glass (1969) and MT. Yet this does not necessarily imply that there is a proven by Encke (1990), the locally parallel structure of causal relationship between the present findings and the Moirt patterns may be recovered by evaluating local main behavioural impairments of the patients, i.e. proso- autocorrelation functions. On the other hand, the pagnosia and alexia. However it is noteworthy that the (global) autocorrelation of a binary texture is nothing prosopagnosic patient, KD, is well able to discriminate else but its “second-order statistics”, and a sufficient textures which are shown long enough [Fig. Z(top right)]. difference in this property renders two textures discrim- This adds to the observation that, at unlimited exposure inable (Julesz, 1981). Thus one may say that MoirC duration, she is in the lower range of normal controls in structure is brought about by an ordered distribution of face matching (Christen et al., 1985). We conclude that, (local) texture differences. as a substitute for global perception, the prosopagnosic To summarize then, it seems that models of global patient employs a compensatory strategy by means of visual perceptions cannot avoid the assumption of local scanning eye movements-if only she has enough time to receptive-field type of filtering and subsequent global do so. evaluation of filter responses. It is conceivable that such The perceptual alteration of the alexic patient is quite processing strategies can be disturbed either on the level different. On the one hand, she has problems with of local filtering or of global interactions, or both. This rapidly discriminating local luminance modulations as concept is consistent with our conjecture of the existence well as textures composed of such signals. On the other hand, she has virtually perfect visual resolution and is able to rapidly recover structure from dot patterns. Her deficit of global perception results, therefore, most likely KD from a weakness to rapidly build up and use an internal 1.0 visual representation of more complex items. Such a weakness would explain why she has difficulties to recognize patterns but not why these difficulties occur mainly with verbal material. 0.9 REFERENCES m: Micropattern Alajouanine, T., Lhetmitte, F. & de Ribaucour-Ducarne, B. (1960). t: Texture Les alexies agnosique et aphasiques. In Les grundes ucfivit& du lobe occipital. Paris: Masson. Beck, J. (1972). Similarity grouping and peripheral discriminability 0.8 MT under uncertainty. American Journal of Psychology, 85, I-19. Beck, J. (1983). Textural segmentation, second order statistics, and I textural elements. Biological Cybernetics, 48, 125-l 30. Bodamer, J. (1947). Ueber die Prosop-Agnosie (Die Agnosie des Physiognomieerkenne). 179, 6-54. Archiu. Psychiatric. Neruenkrankheiten, Broca, P. (1861). Remarques sur le si&ge de la facultk du langage articult, suivies d’une observation d’aphkmie. Bulletin de la Soci’ere d’anatomie. Paris, 6, 33&357. Caelli, T. (1985). Three processing characteristics of visual texture segmentation. Spatial Vision, I, 19-30. Caelli, T. M. (1988). An adaptive computational model for texture 0.6 segmentation. IEEE Transactions on Systems, Man, and Cybernetics, 18, 9-17. Caelli, T., Rentschler, I. & Scheidler, W. (1987). Visual pattern recognition in humans I. Evidence for adaptive filtering. Biological Cybernetics, 57, 233-240. Campbell, R., Landis, T. & Regard, M. (1986). Face recognition and 0.5 31 t lip reading-a neurological dissociation. Brain, 109, 509-521. I I -L_.._l Christen, L., Landis, T. & Regard, M. (1985). Left hemispheric m t m t functional compensation in prosopagnosia? A tachistoscopic study FIGURE 5. Discrimination of micropatterns and textures. Ordinate, with unilaterally lesioned patients. Human Neurobiology, 4, g-14. discrimination sensitivities for compound Gabor micropatterns and Davidoff, J. & Landis, T. (1990). Recognition of unfamiliar faCeS in textures given as area under ROC. Exposure duration, 120msec. prosopagnosia. Neuropsychologia, 28, 1143-l 161. Prosopagnosic patient, KD, left; alexic patient, MT, right. 100 trials De Renzi, E., Zambolin, A. BE Crisi, G. (1987). The pattern of with non-targets and 50 trials with targets were used at each exper- neuropsychological impairment associated with left posterior cer- imental condition and for each subject. Error bars: f 1 SE. ebral artery infarcts. Brain, IIO, 1099-1116.  LOCAL AND GLOBAL PROCESSING IN AGNOSIA 971 Dorfman, D. D. t Alf, E. Jr (1969). Maximum likelihood estimation Lawden, M. C., Hess, R. F. & Campbell, F. W. (1982). The discrim- of parameters of signal detection theory and determination of inability of spatial phase relationships in amblyopia. Vision Research, comidence intervals-rating method data. _iournal of Mathematical 22, 1005-1016. Psy~hoiogy, 6, 487-496. Levine, D. & Calvanio, R. (1989). Prosopagnosia: A defect in visual Encke, W. (1990). Moiremuster. In E~~ri~nte~~e und theoret~che configural processing. Bruin and Co~jfjon, 10, 149-170. Analyse eines Wahrnehmungsph&mmens. Munich: Kyrill & Method. Malik, J. & Perona, P. (1990). Preattentive texture discrimination with Farah, M. J. (1990). Visual agnosia. Cambridge, Mass.: MIT Press. early vision mechanisms. Journal of the Optical Society of America, Glass, L. (1969). Moire effect from random dots. Nature, 223,578-580. 7, 923-932. Glass, L. & Perez, R. (1973). Perception of random dot interference Marcelja, S. J. (1980). Mathematical description of the responses of patterns. Nature, 246, 36&362. simple cortical cells. Journal of the Optical Society of America, 70, Glass, L. & Switkes, E. (1976). Pattern recognition in humans: 1297- 1300. Correlations which cannot be perceived. Perception, 5, 67-72. Marr, D. (1982). Vision. San Francisco, Calif.: Freeman. Green, D. M. & Swets, J. A. (1974). SignaZ detection theory and Niessl Von Mayendorf, E. (1933). Beitrage zur Lehre von der Seelen- psychophysics. Huntington, N.Y.: Krieger. blindheit. Zeitschrifrfiir die gesamte Neurologie und Psychiatric, 149, Griisser, 0. J. & Landis, T. (1991). Visual agnosias and other 68-102. disturbances of visual perception and cognition. In Cronly-Dillon, Rentschler, I., Hiibner, M. & Caelli, T. (1988). On the discrimination J. R. (Ed.), Vision and visual dysfunction (Vol. 12) London: Macmil- of compound Gabor signals and textures, Vision Research, 28, lan Press. 279-29 1. Harvey, L. 0. Jr (1986). Efficient estimation of sensory thresholds. Rizzolatti, G. & Buchtel, H. A. (1977). Hemispheric superiority in Behavorial Research, Mefhorfs, and Instrumentation, 18, 623-632. reaction time to faces: A sex difference. Cortex, 13, 300-305. Hoff, H. & Poet&, 0. (1937). Ueber eine opt&h-agnostische Stijrung Sergent, J. (1982). The cerebral balance of power: Confrontation or des “Physiognomie-Gedlchtnisses”. Zeitschrift fur die gesamte Neu- cooperation. Journal of Experimental Psychology: Human Perception rologie und Psychiatric, 195, 367-395. and Performance, 8, 253-272. Jackson, J. H. (1874). On the nature of the duality of the brain. Medical Stevens, K. A. (1979). Computation of locally parallel structure. Press and Circular, I, 19-41. [Reprinted in Brain (1915), 38,8t3103.] Biologicul Cybernetics, 29, 19-28. Julesz, B. (1981). Textons, the elements of texture perception, and their Voorhees, H. & Poggio, T. (1988). Computing texture boundaries from interactions. Nature, 290, 91-97. images. Nature, 333, 364-367. Julesz, B. (1986). Texton gradients: The texton theory revisited. Wigan, A. L. (1844). A new view of insanity: The duality of the mind. Biological Cybernetics, 54, 245-25 1. London: Longman. Julesz, B. & Bergen, J. R. (1983). Textons, the fundamental elements Zetzsche, C. % Barth, E. (1990). Fundamental limits of linear filters in in preattentive vision and perception of textures. Beil Systems the visual processing of two-dimensional signals. Vision Research, 30, ~echn~ca2 Jour~~, 62, 16191645. 1111-1117. Landis, T. & Regard, M. (1988). Hemianopsie und Agnosie. Khnische Monatsbli’tter fir Augenheilkunde, 192, 525-528. Landis, T., Cummings, J. L., Christen, L., Bogen, J. E. & Imhof, H. G. (1986). Are unilateral right posterior cerebral lesions sufficient to Acknowledgements-This study was supported by Deutsche cause prosopagnosia? Clinical and radiological findings in six Forschungsgemeinschaft, DFG, grant PO 121/13, Project 5, to additional patients. Cortex, 22, 243-252. I.R. and the Swiss National Science Foundation grant No. 32-31260- Lawden, M. C. (1983). An inv~tigation of the ability of the human 91 to T.L. We thank Walter Encke for help with experiments, and visual system to encode spatial phase relationships. Vision Research, Terry Caelli and Norman Cook for critical remarks on the 23, 1451-1463. manuscript.