NOTE COORDINATE FRAMES FOR NAMING MISORIENTED CHIMERICS: A CASE STUDY OF VISUO-SPATIAL NEGLECT Giuseppe di Pellegrino, Francesca Frassinetti and Gianpaolo Basso (Neurological Department, University of Modena, Italy) ABSTRACT A persistent line of inquiry for the students of visuo-spatial neglect has involved the perceptual frame of reference respect to which the neglected region of space is defined. On standard testing conditions viewer-centered and object-centered systems of coordinates are confounded. In order to disambiguate these two reference frames FB, a patient with severe left visual neglect consequent upon a right parieto-temporal haemorrhage, was asked to identify chimeric figures presented at different orientations. FB continued to recognize poorly the left side of chimeric figures even when the display was rotated 90° clockwise or anticlockwise so that the 'left' of the chimeric fell on the patient 's egocentric up or down, respectively. The result suggests that, at least under the present testing conditions, unilateral neglect is tied to the principal (top-bottom) axis of the object. Object-centered vs. (viewer­ centered) representational accounts of this finding are discussed. INTRODUCTION Any account of visuo-spatial neglect must specify the frame of reference or coordinate system respect to which the side of space showing neglect is defined. A reference frame can be viewed as an orthogonal grating used to assign directions (top, bottom, right or left) in space or to mental representation (Palmer, 1985). Thus, 'left' in left neglect may be interpretable in terms of a variety of different spatial coordinates: retinal, head- or body­ centered, gravitational and also in relation to coordinates intrinsic to the depicted object. Other than for neglect, this question has obvious implications for the current debate on space­ based and object-based models of visual attention (Egly, Driver and Rafal, 1994; Vecera and Farah, 1994). Recent investigations have examined the contribute of viewer-centered (egocentric) and environment-centered (allocentric) coordinates to visual neglect. An egocentric frame specifies positions relative to the viewer's midline, whereas an allocentric reference system codes information with respect to the gravitational vertical. For example, Calvanio, Petrone and Levine (1987) had patients report items arranged on spatial arrays, either while seated upright or while reclining on one side. In the latter condition the subjects' midline axis and the environmental vertical were out of alignment (roughly orthogonal), so that it became possible to examine in which frame unilateral neglect operated. The results revealed that neglect patients ignore information appearing on the left of their bodies, but also omit stimuli on the left of the display regardless of the posture they adopted. This finding led Calvanio et al. (1987) to propose that neglect affects both egocentric and allocentric representations of space. Similar experimental evidence has been also provided by other laboratories (Ladavas, 1987; Farah, Brunn, Wong et al., 1990). Viewer- and environment--centered coordinates do not exhaust, however, the range of reference frames within which visual information is described (Hinton, 1981; Feldman, 1985). Indeed, some studies reveal that, under some experimental circumstances, neglect can be based upon object-centered coordinates in which Cortex, (1995) 31, 767-777 768 G. di Pellegrino and Others measurements are referred to the principal axis of the perceptual object. Driver and Halligan (1991) provided the most striking evidence for object-centered neglect. They asked a right­ hemisphere patient with left neglect, PP, to judge whether two vertically elongated shapes, presented one above the other, were same or different. Stimuli were displayed either upright or rotated 45°. PP omitted details on the left of the objects' principal axis even when the shapes were tilted so that their left details fell on the patient's egocentric right. Further evidence supporting neglect in object-centered coordinates has been reported by Young, Hellawell and Welch (1992). Their patient failed in recognizing the left side of chimeric faces made by joining left and right halves of famous faces. When stimuli were rotated 90° to the right (or left), the patient continued to perform poorly on left-sided chimerics, defined in object-centered code, even though they actually appeared in the top half (or bottom half) of egocentric space. This result indicates that visuo-spatial neglect can operate within a reference frame defined by coordinates intrinsic to the object. In the reading domain object-based neglect has been documented by Caramazza and Hillis (1990a, 199Gb). They described a patient with damage to the left hemisphere who misread only the right half of words (e.g. humid --+ human). Interestingly, her problems in identifying the end letters were unaffected by the vertical or mirror-reversed orientation of the words. Caramazza and Hillis proposed that this patient had an impairment at the level in which a word-centered grapheme representation is processed. However, other studies failed to confirm the existence of object-centered neglect. For instance, in a pioneering study Farah eta!. (1990) asked 10 left neglect patients to perform a task which involved two distinct steps: first to identify drawings of familiar items and then to read letters randomly scattered over the space delimited by the outline of the drawings. To examine in which frame of reference neglect manifests itself, the authors compared the letter naming performance, when both viewer and stimulus display were upright, to that observed when the viewer or the stimulus were rotated 90°. Their findings showed that neglect arises within both viewer-centered and environment-centered coordinates, but refuted the hypothesis that an object-based frame is used to code the location of the neglected region of space. Behrmann and Moscovitch (1994) adopted a paradigm similar to that described in Farah et al.'s study, except that, instead of reading letters, neglect patients had to report the colors drawn around the boundaries of line drawings. A significant object-centered neglect was observed only when stimuli were drawings of asymmetrical letters (e.g. letters with an intrinsic handedness, such as the letter B), not when drawings of common objects or symmetrical letters were used. Both Farah et a!. (1990) and Behrmann and Moscovitch (1994) investigated object­ centered neglect with a visual search task. Since the nature of the visual representation required by the task may be crucial in bringing out space-based vs. object-based attentional selection (Vecera and Farah, 1994), we employed the chimeric figures recognition task, previously adopted by Young, Hellawell and Wekh (1992), which may be more suited to engage an object-based representation than the visual search task of previous studies. Young et a!.' s observation about object-centered coding in visual neglect was based on a small set of faces which are stimuli with a high level of intra-class similarity. There is ample evidence that the processing mechanisms leading to the visual recognition of faces differ from those used for object recognition (see Bruce and Humphreys, 1994, for a review). For example, object recognition seems to involve a parts-based representation, whilst faces are more likely to be encoded holistically (Tanaka and Farah, 1993). It remains therefore possible that the object­ centered effect reported by Young et a!. (1992) is tied to the nature of stimuli (Behrmann and Moscovitch, 1994). Recall that negative studies of object-centered neglect employed drawings of familar objects. The goal of the present research is to assess the generality of Young et al.'s finding with line drawings of objects belonging to diverse categories. Object-centered coordinates were decoupled from viewer-centered coordinates through a rotation of the stimulus display while the patient remained upright. The task required to identify chimeric figures presented in the upright or tilted conditions. Our assumption was that if neglect mechanisms affect an object-centered representation of space, movements of the object should move the location of the neglected area. Thus, neglect of the left side of the chimeric figures, which is expected when the stimulus is upright, should become neglect Naming misoriented chimerics and ULN 769 of the upper side when figures are rotated 90° clockwise, but should change to neglect of the lower side when figures are rotated 90° anticlockwise. CASE REPORT FB was a 65-year-old right-handed man with high-school education, who had recently retired from an administrative job. He was admitted at the Modena University Ho~ital on 13 May 1994 with complaints of intense right-sided headache and tendency to bump into things on his left-hand side, which had abruptly appeared on that morning. On neurological examination, the patient was found alert and cooperative, well oriented in time and space. A complete left homonymous hemianopia was revealed on testing to confrontation; clinical inspection showed that head and eye movements to visual targets were full and symmetric. Tactile stimuli were perceived well when given singly, but a marked extinction on the left was observed with double simultaneous stimulation (DSS). Sensory findings also included grossly defective appreciation of posture in the left arm and leg and auditory extinction. Left arm strength was only slightly diminished. No dysphasia, dysarthria or dyspraxia were observed. Analysis of the CT scan (performed on 13 May 1994), using the method described in Damasio and Damasio (1989), revealed a large area of high density involving the right angular gyrus (area 39) and the posterior sector of the middle temporal gyrus (area 37). The supramarginal gyrus (area 40) was partially damaged. The lesion was attributed to a haemorrhage in the vascular territory of the right middle cerebral artery (Figure 1). Examination of higher cognitive functions started two weeks after admission; at this time the visual field defect had recovered, leaving in its wake a left visual extinction on DSS. FB obtained a verbal IQ of 115 on the WAIS-R. Language comprehension and production were normal. The ability to imitate with the right hand gestures made by the experimenter was normal. His digit span was five and no error was scored on simple arithmetics. A florid neglect could be elicited by various behavioral tests which included line and figure cancellation, line bisection, figure copying/drawing by memory and text reading. Thus, on the Bell Cancellation Test (Gauthier, Dehaut and Joannette, 1989), FB failed to cross any of 17 left-sided targets and omitted 4/18 on the right side. Horizontal lines (6 lines for each of 5 lengths), presented individually across the centre of an A4 paper, were bisected significantly to the right of the true center. On drawing a daisy from memory, petals on the left side were completely omitted; a bicycle was drawn (from memory) with its left wheel smaller than the right wheel; the spokes on the left of each wheel were left out. In oral reading of text, the patient omitted the first 4-7 words on each line, with no apparent insight that the resulting content often lacked of coherence. Neglect was not restricted to the visual modality. FB showed a similar pattern of left omissions during blindfolded tactile search of tokens scattered on the desktop in front of him: he scored 8110 on the right side and 0110 on the left side. On the contrary, there was no evidence of personal neglect (tested following Bisiach, Vallar, Perani et al., 1986). MATERIALS AND METHODS Stimuli The stimuli consisted of chimeric figures and whole figures. We selected from the Snodgrass and Vanderwart (1980) collection of drawings several figures of animals (n = 19) and common objects (n = 21) that have a canonical upright orientation (for the list of figures used see Appendix). All but 1 drawing (butterfly) depicted an asymmetric view of the items. They were cut along the vertical midline into right and left halves and showed to 3 normal subjects who were asked to choose the half side of each object containing essential defining features. These more informative half-drawings were then employed to assemble a set of 20 chimeric figures by juxtaposing 20 right halves and 20 non-matching left halves. The mirror­ imaged versions of the same chimerics were also used so that the features of each half­ 770 G. di Pellegrino and Others Fig. 1 - Reconstructed computerized scan of FB"s brain damage. The figure shows the lesion (indicated as a shaded area) plotted onto standardized transverse sections following to the method described in Damasio and Damasio (1989). drawing were equally present on either side of the chimerics. A total of 40 chimeric figures was employed. For the whole figures, we took the same 40 drawings from which the chimerics were generated. For 20 of them the more informative half (used in the chimerics) was printed on the left side, and for the other 20 on the right side. The purpose of using whole figures was to assess the integrity of FB's object processing abilities. Chimeric and whole figures ranged in size from 4.5 X 3 em to 9 X 7 em and were printed across the centre of a white, vertically oriented, A4 page (27.9 X 21 em), one figure per sheet. The experiment was carried out 22 days after the stroke, in two sessions separate by a day; in the first session half of the stimuli were used (1 0 normal chimerics accompanied by 10 reversed versions and the corresponding 20 whole figures), whereas in the second session all 80 figures served as stimuli, yielding 120 stimuli in total. Procedure Each experimental session proceeded as follows: FB was seated in front of a table, in a quiet room, two experimenters were present. Each stimulus sheet was individually displayed Naming misoriented chimerics and ULN 771 Upright Tilted to the right Tilted to the lett Fig. 2 - Example of chimeric figures used for each pair of stimuli in different orientation con­ ditions. in front of the patient and always centered on the midsagittal plane of his head and trunk. Viewing distance was approximately 45 em. FB' s task was simply to name the stimuli; he was informed that some drawings could depict anomalous objects or animals. The patient was allowed to examine the stimuli for as long as he liked, and any time he failed to report both halves of the chimerics, he received the prompt: "Do you see anything else?". No other feedback was provided. Responses and any other patient's comments were written down by one experimenter for later analysis. Stimuli were displayed under 3 different conditions (Figure 2): (1) Upright, viewer and stimulus upright, (2) Tilted to the Right, viewer upright and stimulus rotated in the frontal plane from the environmental vertical to 90° clockwise, and (3) Tilted to the Left, viewer upright and stimulus rotated 90° anticlockwise. Conditions were given in separate blocks. The upright condition was presented first in both experimental sessions; the tilted to the right condition was second in the first session and third in the second session. Within each condition, chimeric and whole figures were presented in random sequences. Breaks of at least 5 minutes were inserted between blocks. FB was requested to maintain an upright head posture while inspecting the stimulus displays. The second experimenter observed the patient closely to ensure that he was complying with the instruction. Eye movements were not restricted. G. di Pellegrino and Others 772 TABLE I Number (and Percentage) of Correct Naming for the Left and Right Side of the Chimeric Figures in each Orientation Condition Orientation Upright Tilted to the right Tilted to the left Left side Right side 4/60 (7%) 22/60 (37%) 20/60 (33%) 59160 (98%) 48/60 (80%) 51/60 (85%) RESULTS FB's naming performance on 60 whole figures was flawless, irrespective of their orientation. The results for the chimeric figures are presented in Table I; correct naming scores (and percentages) for the left and right side of the chimerics (defined in an object­ centered frame of reference) are provided for each orientation condition. FB's performance was scored correct only if he identified the stimulus (e.g. "It is a horse") and not if he simply recognized the category to which the stimulus belonged ("It is an animal"). We distinguished between 3 kinds of errors. Omissions, if the stimulus went completely unnoticed. Non-recognition errors, if FB failed to report the name of the stimulus, but remarked that something was present. Category responses (infrequent), if only the category of the stimulus, and not the specific exemplar, was identified. In the Upright condition, when all frames were aligned, very few stimuli were reported on the left side, while practically all of those lying on the right side of the chimerics were reported (chi-square [1} = 101, p<0.0001), a finding that confirms the severity of FB's neglect for contralesional stimuli. More interestingly, FB's detected significantly less left-sided than TABLE II Number (and Percentage) of Left-sided Errors in Each Orientation Condition Left side Orientation Omission Non-recognition Category Upright Tilted to the right Tilted to the left 51/56 (91 %) 34/38 (90%) 33/40 (82%) 3/56 (5%) 4/38 (10%) 6/40 (15%) 2/56 (3%) 0/38 (0%) 1/40 (3%) right-sided stimuli, as defined by the chimeric's principal (top-bottom) axis, also in the Tilted to the Right condition (chi-square [1} = 23.1, p