IMPAIRED VISUAL PERCEPTUAL CATEGORIZATION IN RIGHT BRAIN-DAMAGED PATIENTS: FAILURE TO REPLICATE Michael Bulla-Hellwigl, George Ettlinger\ Dierk Dommasch 2 , Eduard Ebel 3 and Wolfhard Skreczeck 1 CDept. of Psychology, Bielefeld University, Bielefeld; 2Dept. of Neurology, Bethesda Clinic, Bielefeld-Bethel; 3Dept. of Neurology, Ev. Johannes Hospital, Bielefeld)
INTRODUCTION
At present, our knowledge of agnosia is restricted to the enumeration of several selective disorders of recognition. These are largely descriptive, prov iding no general or reliable explanation of the processes involved. A recent neuropsychological model of object recognition relates visual ob ject agnosia to defective categorization (Warrington, 1982, 1985). Warrington and Taylor (1978) reported a double dissociation of disorders of visual cate gorization: they found a disorder of perceptual categorization (PC) specific to right-sided brain damage (RBD) and a disorder of semantic categorization (SC) specific to left-sided brain damage (LBD). PC was defined as the ability to clas sify as "same" in spite of physical differences, SC stood for the ability to clas sify according to the functional significance of objects. Based on this double dissociation Warrington and Taylor proposed a model of object recognition, which has been extended by Warrington (1985) and integrates Lissauer's dis tinction between apperceptive and associative agnosia. (1) Sensory analysis: At this stage visual features are processed, this pro cessing takes place in both hemispheres. Disorders at this stage may result in re cognition errors, but they only mimick agnosia. Therefore, Warrington terms such disorder as "pseudo-agnosia". (2) Perceptual categorization: The output of stage (1) is further processed to achieve PC, that is, to establish the perceptual equivalence of objects. This processing is performed by the right hemisphere (RH). A lesion there may re sult in PC disorders, reflecting apperceptive agnosia. (3) Semantic categorization: The output of ,stage (2) is further processed to achieve SC, that is categorization, to establish the functional significance of objects. This process is accomplished by the left hemisphere (LH). A lesion there may result in disorders of semantic categorization, reflecting associative agnosia. According to Warrington, these three stages are serially organized: Visual information is transferred from sensory analysis via perceptual categorization to semantic categorization. By means of modality-specific memory systems (not yet specified by Warrington) recognition is finally accomplished. Recently, fol lowing the observation that patients with PC disorders were not handicapped in everyday recognition, Warrington (Warrington and James, 1986) suggested that the PC-stage may be only an optional stage of object recognition. She has thereCortex, (1992) 28, 261-272
262
M. Bulla-Hellwig and Others
by weakened the overall significance of this stage for the recognition of objects (without specifying the conditions under which this "option" applies). In addition, the study by Sergent and Lorber (1983) casts doubt on the val idity of the suggestion that perceptual categorization is specific to the RH. Ser gent and Lorber demonstrated in normal subjects that both hemispheres are capable of PC, and that the RH is superior only under conditions precluding normal processing of visual input. Sergent and Lorber concluded that the dis orders of PC in RBD patients may in part be related to functional sensory de ficits. They also asserted that sensory deficits have to be ruled out by careful ex amination before denying their contribution to perceptual disorders. (Instead of assessing the sensory performance of their patients, Warrington and Taylor had relied on earlier studies that had failed to reveal a significant contribution of sensory deficits to object recognition.) The present study had the following goal: using material like that described by Warrington and Taylor (1978), we wished to ascertain the relationship be tween categorization performance and sensory performance. For PC we pre dicted, following the argument of Sergent and Lorber, that RBD patients hav ing disordered perceptual categorization would show associated sensory de fects, whereas RBD patients with intact perceptual categorization would show unimpaired sensory functioning. In LBO patients PC performance should be independent of sensory deficits, on the basis that their intact right hemisphere would allow them to perform the task. The relationship between SC and sens ory performance was to be established for the first time. MATERIALS AND METHODS
Subjects Subjects formed a consecutive series of patients from the Departments of Neurology of two local hospitals (Evang. Johanneskrankenhaus and Bethesda Clinic). Every patient hav ing a unilateral lesion (verified by CT/MRI) was included in the series, except that patients with (1) cerebral atrophy, (2) diffuse pathology, (3) metastases, (4) developmental brain dys function, (5) clinically or psychometrically determined severe intellectual deterioration, (6)
TABLE!
Characteristics of the Patients Variable Age Sex Education IQ VFD
Mean S.D. Female Male 10 years 13 years Mean S.D. Yes No
LH
RH
Controls
56.63 11.66 8 19 23 4 97.22 11.32 7 20
47.18 12.83
51.43 9.35 9 12 19
11
17 26 2 100.64 12.09 11 17
IQ assessed with the WIP (a standardized German short-form of the WAIS, Dahl, 1972).
2
94.38 10.30
263
Perceptual categorization TABLE II
Neurological Details of the Patients
Aetiology
Vascular Tumour Other
RH
21
1
21 3 4
5 0
2
5
Location F FT
FP FTP T
1 1
8 8 0
p
0 PT OP OPT Duration of illness
LH
1 2 1
< 1 month 1-6 months > 6 months
8 11 8
8
0 0 4
6
1
5
1
1 11
10 7
F =frontal; T =temporal; P =parietal; 0 =occipital.
alcoholism, or (7) patients older than 75 and younger than 20 years of age were excluded. Only right handed subjects (determined by the Edinburgh Handedness Inventory of Old field, 1971) participated. 28 RBD and 27 LBD patients participated; using the Token Test (Spellacy and Spreen, 1969), 18 LBD patients were classified as aphasic. The sample is des cribed in Table I. The aetiology, duration of illness, laterality and location of the lesions are presented in Table II. A control group of 21 patients without known brain disease was sim ilarly selected and tested.
Tests and Procedure Perceptual Categorization and Semantic Categorization ofPhotographs of Objects For both categorization tasks we used pictures of everyday objects, which were typical representatives of a given object class (cf. classification by Battig and Montague, 1969) and/ or had been used by other authors (see below). Perceptual categorization. Twenty photographs of common objects were printed with good contrast in black-on-white (paper size: 10 x 15 em; object size 5-6 em). Ten photo graphs showed an object from a conventional view, 10 photographs were taken from an un conventional view. The conventional view was taken according to the pictures of Snodgrass and Vandervart (1980) and Warrington and James (1986). The unconventional view was tak en following the examples of Humphreys and Riddoch (1984) and Warrington and James (1986). Illustrations are shown in Figure 1. Five pairs of photographs were of the same ob ject, five pairs were of different objects. Same and different pairs showed one object from a conventional view and one object from an unconventional view. Subjects were first asked to identify the unconventionally photographed object (naming, demonstrating its use). Then the conventionally photographed object was presented and the subjects were required to res pond whether the pair was "same" or a "different", that is, to judge the physical identity of the depicted objects. If the response was "different", subjects were asked to identify the con ventionally viewed object. The photographs were presented horizontally on a table (viewing distance: 35-40 em) and free inspection time was allowed. Semantic categorization. Twenty black-and-white photographs were prepared, showing another set of common objects taken only from a conventional view; Five "same" and five "different" pairs were made, sameness and difference being defined by the functional equi valence of the objects. "Same" object pairs fulfill similar functions (e.g. knife and scissors
264
M. Bulla-Hellwig and Others
Fig. I -Illustrations of the material used in the PC task with photographs of objects. The upper half illustrates a pair where the answer "same" is correct, the lower a pair where the answer "different" is cor rect.
are both used for cutting); "different" object pairs do not do so (e.g. toaster and TV -set). Examples are given in Figure 2. Subjects were first asked to identify a given object (naming; demonstrating its use). They were then shown the second photograph and were to decide whether the depicted object might serve the same function (response "same") or not (res ponse "different"). Finally, the second object was to be identified. The photographs were presented exactly as for PC.
Sensory Tests Sensory tests were designed to achieve formal comparability with the PC and SC tasks. Subjects were to judge again the sameness or difference of a stimulus pair, but as a function of physical identity. Each task consisted of 12 same and 12 different pairs, given in a room without daylight. Same and different pairs were presented in randomized order in the centre of the visual field. The members of each pair were presented successively for 1 sec each (1 1.5 sec interstimulus interval). The sensory tasks were performed monocularly with the pre ferred eye (operationally defined as the eye used for 2 out of 3 tasks: looking through a tube, looking through a hole, aiming). The responses were given verbally ("same" /"different");
Perceptual categorization
265
Fig. 2 -As for Figure 1 but for the SC task.
and were transformed according to the principles of signal detection theory. Two tasks were presented with a Tektronics Oscilloscope (RM 656 Dual Beam, viewing distance 76 em). The stimulus field (the vertical screen of the oscilloscope, colour P2 phos phor) was 3 x 3 degree of visual angle and was surrounded by a cardboard field of 23 X 23 degree, matched for colour and average luminance (14 cd/m 2). Subjects were required to judge whether the orientation (horizontal/vertical) of two sinewave gratings was identical (response "same") or different (response "different"). (a) Sensitivity for spatial frequency (FREQUENCY): Sinewave gratings (contrast=0.17) were presented horizontally or vertically, varying spatial frequency: 1, 3, 5 and 7 cycles per degree of visual angle, respectively. (b) Contrast sensitivity (CONTRAST): sinewave gratings (at 5 cycles per degree) were presented again horizontally or vertically, varying contrast (c= .06, .17, .28, .38 respective ly). Contrast was defined as c = (Lmax- Lmin)/(Lmax + Lmin); Lmax: maximum lumi nance; Lmin: minimum luminance.· Four tasks were given on a Tiibinger Perimeter (Oculus: viewing distance 35 em). The background luminance of the perimeter was set at 3.182 cd/m 2 (photopic vision) and the tar get luminance was set at 318.3 cd/m 2 •
266
M. Bulla-Hellwig and Others
Fig. 3 - PC performance and SC per formance (Index A ') with photographs of objects in relation to the laterality and the site of the lesion.
Cate•oriEation l0e8r
p•r~ormanc•
.05. Task F= 13.11; d.f. = 1, 71; p < .05. GroupxTask F= .21; d.f. =4, 71; p > .05.
M. Bulla-Hellwig and Others
268
than PC performance. In addition, the PC and SC scores were found to be equally and significantly correlated in RBD and LBD patiens (Pearson r for RH = .63; for LH = .62; p < .01). Since no significant differences between an terior and posterior patients were obtained, these were combined to RBD and LBD groups for further analyses.
Relationship between Perceptual Categorization ofPhotographs of Objects and Sensory Performance This relationship was analysed by way of multiple regression, with the A scores of the six sensory tests as independent variables and the A -score of the PC task as the dependent measure, see Figure 4. The regression functions for the two lesion groups differed significantly (F = 3.74; d.f.= 7, 41; p < .01). RBD and LBD patients were therefore analysed separately (see Table V). In the RH group PC performance was significantly correlated with the overall sensory performance, but no single loading proved to be meaningful by itself. In the LH group no significant correlation between PC performance and overall sensory · performance was found (see Table V). 1
1
Relationship between Semantic Categorization and Sensory Performance This relationship was analysed in the same way as for PC performance, see Figure 5. The regression coefficients for the two lesion groups did not differ
Sensor~
.. .. . .. . ....-. ...
.10
.80
.64
< .Ol
significantly (F = .70; d.f. = 7, 41; p > .25), so both groups were combined for further analysis. Only a trend for a correlation between SC and overall sensory performance was obtained (see Table VI). DISCUSSION
In the present study RBD and LBO patients did not differ significantly at PC or SC. Thus, the present findings fail to support a double dissociation of cate-
Sensorw
INTRODUCTION
At present, our knowledge of agnosia is restricted to the enumeration of several selective disorders of recognition. These are largely descriptive, prov iding no general or reliable explanation of the processes involved. A recent neuropsychological model of object recognition relates visual ob ject agnosia to defective categorization (Warrington, 1982, 1985). Warrington and Taylor (1978) reported a double dissociation of disorders of visual cate gorization: they found a disorder of perceptual categorization (PC) specific to right-sided brain damage (RBD) and a disorder of semantic categorization (SC) specific to left-sided brain damage (LBD). PC was defined as the ability to clas sify as "same" in spite of physical differences, SC stood for the ability to clas sify according to the functional significance of objects. Based on this double dissociation Warrington and Taylor proposed a model of object recognition, which has been extended by Warrington (1985) and integrates Lissauer's dis tinction between apperceptive and associative agnosia. (1) Sensory analysis: At this stage visual features are processed, this pro cessing takes place in both hemispheres. Disorders at this stage may result in re cognition errors, but they only mimick agnosia. Therefore, Warrington terms such disorder as "pseudo-agnosia". (2) Perceptual categorization: The output of stage (1) is further processed to achieve PC, that is, to establish the perceptual equivalence of objects. This processing is performed by the right hemisphere (RH). A lesion there may re sult in PC disorders, reflecting apperceptive agnosia. (3) Semantic categorization: The output of ,stage (2) is further processed to achieve SC, that is categorization, to establish the functional significance of objects. This process is accomplished by the left hemisphere (LH). A lesion there may result in disorders of semantic categorization, reflecting associative agnosia. According to Warrington, these three stages are serially organized: Visual information is transferred from sensory analysis via perceptual categorization to semantic categorization. By means of modality-specific memory systems (not yet specified by Warrington) recognition is finally accomplished. Recently, fol lowing the observation that patients with PC disorders were not handicapped in everyday recognition, Warrington (Warrington and James, 1986) suggested that the PC-stage may be only an optional stage of object recognition. She has thereCortex, (1992) 28, 261-272
262
M. Bulla-Hellwig and Others
by weakened the overall significance of this stage for the recognition of objects (without specifying the conditions under which this "option" applies). In addition, the study by Sergent and Lorber (1983) casts doubt on the val idity of the suggestion that perceptual categorization is specific to the RH. Ser gent and Lorber demonstrated in normal subjects that both hemispheres are capable of PC, and that the RH is superior only under conditions precluding normal processing of visual input. Sergent and Lorber concluded that the dis orders of PC in RBD patients may in part be related to functional sensory de ficits. They also asserted that sensory deficits have to be ruled out by careful ex amination before denying their contribution to perceptual disorders. (Instead of assessing the sensory performance of their patients, Warrington and Taylor had relied on earlier studies that had failed to reveal a significant contribution of sensory deficits to object recognition.) The present study had the following goal: using material like that described by Warrington and Taylor (1978), we wished to ascertain the relationship be tween categorization performance and sensory performance. For PC we pre dicted, following the argument of Sergent and Lorber, that RBD patients hav ing disordered perceptual categorization would show associated sensory de fects, whereas RBD patients with intact perceptual categorization would show unimpaired sensory functioning. In LBO patients PC performance should be independent of sensory deficits, on the basis that their intact right hemisphere would allow them to perform the task. The relationship between SC and sens ory performance was to be established for the first time. MATERIALS AND METHODS
Subjects Subjects formed a consecutive series of patients from the Departments of Neurology of two local hospitals (Evang. Johanneskrankenhaus and Bethesda Clinic). Every patient hav ing a unilateral lesion (verified by CT/MRI) was included in the series, except that patients with (1) cerebral atrophy, (2) diffuse pathology, (3) metastases, (4) developmental brain dys function, (5) clinically or psychometrically determined severe intellectual deterioration, (6)
TABLE!
Characteristics of the Patients Variable Age Sex Education IQ VFD
Mean S.D. Female Male 10 years 13 years Mean S.D. Yes No
LH
RH
Controls
56.63 11.66 8 19 23 4 97.22 11.32 7 20
47.18 12.83
51.43 9.35 9 12 19
11
17 26 2 100.64 12.09 11 17
IQ assessed with the WIP (a standardized German short-form of the WAIS, Dahl, 1972).
2
94.38 10.30
263
Perceptual categorization TABLE II
Neurological Details of the Patients
Aetiology
Vascular Tumour Other
RH
21
1
21 3 4
5 0
2
5
Location F FT
FP FTP T
1 1
8 8 0
p
0 PT OP OPT Duration of illness
LH
1 2 1
< 1 month 1-6 months > 6 months
8 11 8
8
0 0 4
6
1
5
1
1 11
10 7
F =frontal; T =temporal; P =parietal; 0 =occipital.
alcoholism, or (7) patients older than 75 and younger than 20 years of age were excluded. Only right handed subjects (determined by the Edinburgh Handedness Inventory of Old field, 1971) participated. 28 RBD and 27 LBD patients participated; using the Token Test (Spellacy and Spreen, 1969), 18 LBD patients were classified as aphasic. The sample is des cribed in Table I. The aetiology, duration of illness, laterality and location of the lesions are presented in Table II. A control group of 21 patients without known brain disease was sim ilarly selected and tested.
Tests and Procedure Perceptual Categorization and Semantic Categorization ofPhotographs of Objects For both categorization tasks we used pictures of everyday objects, which were typical representatives of a given object class (cf. classification by Battig and Montague, 1969) and/ or had been used by other authors (see below). Perceptual categorization. Twenty photographs of common objects were printed with good contrast in black-on-white (paper size: 10 x 15 em; object size 5-6 em). Ten photo graphs showed an object from a conventional view, 10 photographs were taken from an un conventional view. The conventional view was taken according to the pictures of Snodgrass and Vandervart (1980) and Warrington and James (1986). The unconventional view was tak en following the examples of Humphreys and Riddoch (1984) and Warrington and James (1986). Illustrations are shown in Figure 1. Five pairs of photographs were of the same ob ject, five pairs were of different objects. Same and different pairs showed one object from a conventional view and one object from an unconventional view. Subjects were first asked to identify the unconventionally photographed object (naming, demonstrating its use). Then the conventionally photographed object was presented and the subjects were required to res pond whether the pair was "same" or a "different", that is, to judge the physical identity of the depicted objects. If the response was "different", subjects were asked to identify the con ventionally viewed object. The photographs were presented horizontally on a table (viewing distance: 35-40 em) and free inspection time was allowed. Semantic categorization. Twenty black-and-white photographs were prepared, showing another set of common objects taken only from a conventional view; Five "same" and five "different" pairs were made, sameness and difference being defined by the functional equi valence of the objects. "Same" object pairs fulfill similar functions (e.g. knife and scissors
264
M. Bulla-Hellwig and Others
Fig. I -Illustrations of the material used in the PC task with photographs of objects. The upper half illustrates a pair where the answer "same" is correct, the lower a pair where the answer "different" is cor rect.
are both used for cutting); "different" object pairs do not do so (e.g. toaster and TV -set). Examples are given in Figure 2. Subjects were first asked to identify a given object (naming; demonstrating its use). They were then shown the second photograph and were to decide whether the depicted object might serve the same function (response "same") or not (res ponse "different"). Finally, the second object was to be identified. The photographs were presented exactly as for PC.
Sensory Tests Sensory tests were designed to achieve formal comparability with the PC and SC tasks. Subjects were to judge again the sameness or difference of a stimulus pair, but as a function of physical identity. Each task consisted of 12 same and 12 different pairs, given in a room without daylight. Same and different pairs were presented in randomized order in the centre of the visual field. The members of each pair were presented successively for 1 sec each (1 1.5 sec interstimulus interval). The sensory tasks were performed monocularly with the pre ferred eye (operationally defined as the eye used for 2 out of 3 tasks: looking through a tube, looking through a hole, aiming). The responses were given verbally ("same" /"different");
Perceptual categorization
265
Fig. 2 -As for Figure 1 but for the SC task.
and were transformed according to the principles of signal detection theory. Two tasks were presented with a Tektronics Oscilloscope (RM 656 Dual Beam, viewing distance 76 em). The stimulus field (the vertical screen of the oscilloscope, colour P2 phos phor) was 3 x 3 degree of visual angle and was surrounded by a cardboard field of 23 X 23 degree, matched for colour and average luminance (14 cd/m 2). Subjects were required to judge whether the orientation (horizontal/vertical) of two sinewave gratings was identical (response "same") or different (response "different"). (a) Sensitivity for spatial frequency (FREQUENCY): Sinewave gratings (contrast=0.17) were presented horizontally or vertically, varying spatial frequency: 1, 3, 5 and 7 cycles per degree of visual angle, respectively. (b) Contrast sensitivity (CONTRAST): sinewave gratings (at 5 cycles per degree) were presented again horizontally or vertically, varying contrast (c= .06, .17, .28, .38 respective ly). Contrast was defined as c = (Lmax- Lmin)/(Lmax + Lmin); Lmax: maximum lumi nance; Lmin: minimum luminance.· Four tasks were given on a Tiibinger Perimeter (Oculus: viewing distance 35 em). The background luminance of the perimeter was set at 3.182 cd/m 2 (photopic vision) and the tar get luminance was set at 318.3 cd/m 2 •
266
M. Bulla-Hellwig and Others
Fig. 3 - PC performance and SC per formance (Index A ') with photographs of objects in relation to the laterality and the site of the lesion.
Cate•oriEation l0e8r
p•r~ormanc•
.05. Task F= 13.11; d.f. = 1, 71; p < .05. GroupxTask F= .21; d.f. =4, 71; p > .05.
M. Bulla-Hellwig and Others
268
than PC performance. In addition, the PC and SC scores were found to be equally and significantly correlated in RBD and LBD patiens (Pearson r for RH = .63; for LH = .62; p < .01). Since no significant differences between an terior and posterior patients were obtained, these were combined to RBD and LBD groups for further analyses.
Relationship between Perceptual Categorization ofPhotographs of Objects and Sensory Performance This relationship was analysed by way of multiple regression, with the A scores of the six sensory tests as independent variables and the A -score of the PC task as the dependent measure, see Figure 4. The regression functions for the two lesion groups differed significantly (F = 3.74; d.f.= 7, 41; p < .01). RBD and LBD patients were therefore analysed separately (see Table V). In the RH group PC performance was significantly correlated with the overall sensory performance, but no single loading proved to be meaningful by itself. In the LH group no significant correlation between PC performance and overall sensory · performance was found (see Table V). 1
1
Relationship between Semantic Categorization and Sensory Performance This relationship was analysed in the same way as for PC performance, see Figure 5. The regression coefficients for the two lesion groups did not differ
Sensor~
.. .. . .. . ....-. ...
.10
.80
.64
< .Ol
significantly (F = .70; d.f. = 7, 41; p > .25), so both groups were combined for further analysis. Only a trend for a correlation between SC and overall sensory performance was obtained (see Table VI). DISCUSSION
In the present study RBD and LBO patients did not differ significantly at PC or SC. Thus, the present findings fail to support a double dissociation of cate-
Sensorw
