Spatial Compression in Visual Neglect: A Case Study

NOTE SPATIAL COMPRESSION IN VISUAL NEGLECT: A CASE STUDY Peter W. Halligan and John C. Marshall (River mead Rehabilitation Centre, Oxford, and Neuropsychology Unit, University Department of Clinical Neurology, The Radcliffe Infirmary, Oxford)

INTRODUCTION

As the term suggests, many standard interpretations of visual neglect imply that a variable 'slice' of the spatial world is ignored ('neglected') in the contralateral neglect syndromes. The clearest illustration of this theoretical paradigm is to be found in Bisiach et al. (1983), although Bisiach and Vallar (1988) have also supported a contrary view in which there is a more or less continuous left-right gradient of neglect. With respect to the former view, consider the well-known fact that patients with left neglect (after right hemisphere injury) bisect horizontal lines to the right of true centre. Bisiach et al. (1983) depict this situation by doubling the line length to the right of the patient's (erroneous) transection; the new (and shorter) line so formed is termed the internally represented (as opposed to the objective) line. The extent to the left of the deduced left endpoint of the represented line is then the neglected portion of the line (and is appropriately shaded in the figures of Bisiach et aI., 1983). Although Bisiach et al. (1983) phrase their account more carefully than most (and draw attention to important differences, both quantitative and qualitative, between patients with neglect), the overall framework of their interpretation is not idiosyncratic. Most students of neglect, when looking at task performance on most tests (e.g. drawing, copying, or cancellation) have assumed that a particular spatial domain (typically to the left of the midsagittal plane in patients with right-sided lesion) is neglected. And that 'neglected' means 'unavailable to further cognitive processing' (but see Marshall and Halligan, 1988; Kartsounis and Warrington, 1989; and Seron, Coyette and Bruyer, 1989). There is, in other words, a pathological boundary to the spatial world, analogous to that produced by a frank visual field deficit, although no-one, of course, now believes that vi suo-spatial neglect is extensionally identical to a field cut (Halligan, Marshall and Wade, 1990). For example: the phenomenon of rightward displacement of details in the description of a place from memory immediately suggests an obvious dis analogy between a field cut and 'neglect' (Bisiach et aI., 1981). Nonetheless, the case for looking at neglect as if it were a 'high level' field cut has often been won by default. What else could possibly be happening (either descriptively or theoretically) other than a cognitive cut in the visuo-spatial domain? We wish to suggest that there is (at least conceptually) an alternative. And we shall illustrate this alternative by means of a little analogy. Think of a line or plane as a scarf. How could a scarf be made shorter? One possibility is to cut off a segment from one end. But another possibility is to put the scarf in the washing machine. If it shrinks, it becomes shorter, but no spatially defined segment has been lost (or 'neglected'). The study that follows is an attempt to see if a variant of this latter simile is empirically applicable to visuo-spatial neglect. A similar account of perceived auditory lateralization after right hemisphere lesions has been proposed by Bisiach et al. (1984). MATERIAL AND METHOD

Subjects P.P. is a 60 year old lady who sustained a stroke on 15 October 1987. Left hemiplegia, left hemianopia, and severe left vi suo-spatial neglect were the most prominant signs. CT-scan Cortex, (1991) 27, 623-629

624

P. W. Halligan and J.e. Marshall

indicated a large area of low density in the right temporo-parietal region. All relevant medical details have been reported in Marshall and Halligan, 1989a; that paper, plus Halligan and Marshall, 1989, and Marshall and Halligan, 1989b, contain clinical testing for neglect in P.P. and her performance in a variety of pertinent experimental investigations. The current data were collected between February and May 1988 when her neglect manifestations were fairly stable. As normal controls, 10 adults, 5 male and 5 female, without neurological disease or history thereof, were recruited. Their mean age was 37.7 (range 25 to 73) with a standard deviation of 14.9; all were right handed.

Stimuli and Procedure Stimulus configurations were displayed on an Acorn BBC Master micro-computer connected to a high resolution colour visual display unit (VDU). The dimensions of the VDU were 280 by 210 mm. On each trial, the display consisted of: (a) a row (or column) of the numerals I to 15 (the target array) and (b) an arrow whose head pointed towards the numerals. The horizontal (or vertical) extent of the entire row (or column) was 140 mm, and the distance between each numeral and the next was 10 mm. The arrow was 15 mm long, and each numeral (approximately) 4 mm high. The distance from the arroW to the numeral that was aligned with it (either horizontally or vertically) was 140 mm. There were 4 spatial conditions: (I) The numeral row was 20 mm from the top of the VDU and the arrow 140 mm below the row; (2) The numeral row was 20 mm from the bottom of the VDU and the arrow was 140 mm above the row; (3) The numeral column was 20 mm from the left edge of the VDU screen and the arrow 140 mm to the right of the row; (4) The numeral column was 20 mm from the right edge of VDU and the arrow 140 mm to the left of the row. The four array configurations are shown in Figure I, with, in each case, one illustrative arrow position. On any given trial, the arrow appeared opposite to (and pointing directly towards) one of the numbers in the row or column. The subject's task is to estimate visually which spatial position on the target array the arrow is pointing to. The body of the arrow is always at right angles to the target array and the response mode is verbal report of a number in the target array. All numerical positions were sampled 30 times (in pseudo-random order), leading to a full data set for P.P. of 4 Spatial Configurations (SC) x 15 Numerical Positions (NP) x 30 trials = 1800 readings. Data collection was extended over a four-month period with randomization of all spatial configurations over that time. Control subjects performed the task once each, to give a data set of 4 SC x 15 NP x I trial x 10 subjects = 600 readings . Spatial configurations were blocked; the row (or column) remained on the VDU throughout a session of 15 trials, and on each trial the arrow appeared opposite one of the numerals. The arrow remained until the subject reported (verbally) which of the numbers it was opposite. Subjects sat at a comfortable distance (approximately 400 mm) from the VDU screen; the midpoint of the VDU was aligned with the subject's midsagittal plane. This design enables us to map the (subjective) Cartesian coordinates of a two-dimensional Euclidian space against the (objective) Euclidian geometry thereof. RESULTS

With respect to the 10 control subjects, analysis is unnecessary; the 'grain' of the task is such that perceptual discrimination is essentially errorless for normal subjects. In 300 judgments with horizontal target arrays (top and bottom of the screen), there are only 4 errors (3 veer one position to the right, 1 veers one position to the left). In 300 judgments with vertical target arrays (left and right of the screen), there are 6 errors (4 veer one position to the top, 2 veer one position to the bottom). We accordingly turn to the data from P.P .. Collapsed over all arrow positions (I to 15), for horizontal orientation targets (top and bottom = 900 trials), there are 204 correct responses (22.70/0) and 696 incorrect responses (77.3%). Of the incorrect responses, 4 (0.6%) deviate to the left, and 692 (99.4%) deviate to the right. With vertical targets (left and right = 900 trials), there are 764 correct responses (84.9%) and 136 incorrect responses (15.1 %). Of the incorrect responses 68 (50%) deviate to the top, and 68 (50%) to the bottom. Table

625

Spatial compression in visual neglect

! i

Fig. 1 - Schematic representation of thefour stimulus configurations, with, in each case, one illustrative arrow position. See text for quantitative description of the arrays.

--+---

TABLE!

A - Deviations in Judgment (P.P.) for Horizontal Target Arrays at the Top and Bottom of the VDU Screen. The notation for extent of deviation is the number of instances of a particular deviation, followed by the numerical magnitude of that deviation (in brackets)

Trials No. correct No. incorrect Deviation: Left Right Extent of deviation: Left Right

Horizontal top

Horizontal bottom

450 (100ll1o) 131 (29. 1ll1o) 319 (70.9Il1o) 1 (0,3Il1o) 318 (99.7Il1o)

450 (100ll1o) 73 (16.2Il1o) 377 (83.8Il1o) 3 (0.8Il1o) 374 (99.2Il1o)

1 (1)

246 (1), 54 (2), 11 (3), 3 (4)

3 (1) 249 (1), 116 (2), 7 (3),

2 (4)

B - Deviations in Judgment (P.P.) for Vertical Target Arrays at the Left and Right of the VDU Screen. The notation for extent of deviation is the number of instances of a particular deviation, followed by the numerical magnitude of that deviation (in brackets)

Trials No. correct No. incorrect Deviation: Top Bottom Extent of deviation: Top Bottom

Vertical left

Vertical right

450 (100ll1o) 388 (86.2Il1o) 62 (13.8Il1o) 57 (92Il1o) 5 (8%)

450 376 74 11 63

57 (1) 3 (1), 2 (2)

(100ll1o) (83.5Il1o) (16.5Il1o) (14.9Il1o) (85. 1ll1o)

11 (1) 62 (1), 1 (6)

IA divides the data from the horizontal into target array at top and bottom, respectively; Table IB divides the data from the vertical orientation into target array at left and right, respectively. Performance as a function of arrow positions is shown in Table II (Horizontal arrays) and Table III (Vertical arrays).

626

P. W. Halligan and J.C. Marshall TABLE II

Accuracy of Perceptual Judgment (P.P.) for Horizontal Target Array at the Top and Bottom of the VDU Screen Top

SP M 2.5 (SD) (0.7) 1.5 D

2

3

4

5

6

7

3.4 (1.0) 1.4

(1.2)

4.7

5.0 (0.8) 1.0

5.9 (0.7) 0.9

6.6 (0.6) 0.6

8.2 (0.7)

1.7

1.2

8

9

10

11

12

13

14

15

9.4 10.3 10.7 11.7 12.9 13.5 14.2 14.9 (1.1) (0.8) (0.4) (0.5) (0.3) (0.5) (0.4) (0.2) 1.4 1.3 0.7 0.7 0.9 0.5 0.2 -0.1

Bottom

SP M 3.0 (SD) (0.7) D 2.0

2

3

4

5

6

7

3.7 (0.6) 1.7

4.6 (0.6) 1.6

5.4 (0.6) 1.4

6.5 (0.5) 1.5

7.4 (0.8) 1.4

8.1 (0.4) 1.1

8

9

10

11

12

13

14

15

9.3 10.0 11.1 12.0 12.8 13.6 14.3 15 (0.5) (0.5) (0.5) (0.5) (0.4) (0.5) (0.6) (0.2) 1.3 1.0 1.1 1.0 0.8 0.6 0.3 0.0

TABLE III

Accuracy of Perceptual Judgment (P.P.) for Vertical Target Arrays at the Left and Right of the VDU Screen Right

Left SP

M

(SD)

D

SP

M

(SD)

D

1 2 3 4 5 6 7

1.03 2.03 3.03 3.93 5.0 6.0 6.73 7.9 8.86 9.73 10.67 11.87 12.87 13.5 14.6

(0.2) (0.2) (0.4) (0.3) (0) (0.3) (0.5) (0.3) (0.3) (0.5) (0.4) (0.4) (0.4) (1.1) (0.5)

0.03 0.03 0.03 -0.07 0 0 -0.27 -0.1 -0.14 -0.27 -0.33 -0.13 -0.13 -0.5 -0.4

1 2 3 4 5 6 7 8 9 10

1.13 2.13 3.23 4.40 5.27 6.10 7.07 7.97 9.1 10.03 11.17 12.0 13.17 14.07 14.97

(0.3) (0.3) (0.4) (0.5) (0.4) (0.6) (0.3) (0.2) (0.3) (0.5) (0.4) (0) (0.4) (0.3) (0.2)

0.13 0.13 0.23 0.40 0.27 0.10 0.07 -0.03 0.1 0.03 0.17 0 0.17 0.07 -0.03

8

9 10 11

12 13 14 15

11

12 13 14 15

Both conditions in which the (numerical) target array is horizontally-aligned show 'left neglece. That is, estimates of the position to which the arrow points are right wards displaced. When the target array is above the arrow, the linear regression of subjective position on objective position is + 1.63 + (0.911 x objective position). The equation accounts for 99.411,10 of the variance. When the difference score is regressed on objective position, the equation is + 1.63 - (0.089 x objective position), accounting for 61.4% of the variance. The two comparable equations when the target array is below the arrow are + 2.045 + (0.8846 x objective position), and + 2.045 - (0.1154 x objective position). These regressions account for 99.8% and 90.1 % of the variance, respectively. In short, for both horizontal target positions, subjective displacement is linearly proportional to the spatial position of the arrow, with magnitude of the displacement decreasing from left to right. When the target array (vertical orientation) is to the right of the arrow, the linear regression of subjective position on objective position is + 0.229 + (0.9865 x objective position), with 99.9511,10 of the variance accounted for. When the difference score is regressed on objective position, the equation is + 0.229 - (0.0135 x objective position), accounting thereby for 28.5% of the variance. The subjective displacement is thus towards the bottom of the array. This displacement increases from top to bottom, although overall accuracy is good (and hence the difference equation accounts for only a small proportion of variance).

Spatial compression in visual neglect

627

When the target array is to the left of the arrow, the linear regression of subjective position on objective position is + 0.0966 + (0.9692 x objective position) with 99.950/0 of the variance accounted for. When the difference score is regressed on objective position, the equation is + 0.0966 - (0.0308 x objective position), accounting for 66.8% of the variance. The subjective displacement is thus towards the top of the array. The magnitude of the displacement increases linearly from bottom to top of the spatial array. DISCUSSION

The results from the two horizontally-aligned target arrays are a manifestation of left neglect, albeit observed under novel experimental conditions. An 'attentionallperceptual' process that should construct an imaginary trajectory from arrow to array at right angles to that array is systematically deflected rightwards. This deflection is roughly equivalent in magnitude irrespective of whether the trajectory is from top to bottom (arrow above array) or bottom to top (arrow below array). There is no sharp boundary at the midsagittal plane - position 8 (Bisiach et aI., 1984); rather, the magnitude of the displacement decreases linearly as the arrow is positioned from left (position 1) to right (position 1~). In terms of our initial simile, target space has indeed been shrunk equivalently throughou~ the target range, as when, for example, a spring is compressed. The 'spring' (= Numerical target array) is (attentionally) fixed on the right hand side of the array and pushed in from the left. Hence when the difference regression is calculated, it follows that the magnitude of the displacement is proportional to the distance from the right hand side of the array. Or, to change the direction of the conceit, it is as if a force was located in far left space which blew from west (left) to east (right). The magnitude of this force, as assessed by its power to push the arrow's trajectory eastwards (right), then decreases as a function of the distance of the arrow from the force-source (Kinsbourne, 1970a). With the addition of one extra postulate, these same similes can also account for the pattern of results when the target arrays are vertically-aligned. Assume that, on some trials, mental rotation (or deitic perspective-taking) takes place (Bisiach and Luzzatti, 1978; Robertson and Lamb, 1988). Iii the condition where the arrow is to the right of the vertical target array, an observer whose line of sight was along the direction of the arrow's pointing would maintain that the bottom of the target array was to her left. Left neglect, seen in terms of our model, would accordingly force estimates of target position upwards, propoxtional to a bottom-to-top (or deitic left-to-right) gradient. By contrast, when the arrow is to the left of the verticattarget array, an observer following the direction of the arrowhead would find that the top of the target array was to her left. Left neglect would now force estimates of target position downwards, proportional to a topto-bottom gradient (which is again left-to-right in terms of Qeixis). These patterns are precisely what the results demonstrate. The magnitudes of the displacements with vertical arrays are smaller than with horizontal arrays; this would seem to indicate either that perspective-taking is only employed on a proportion of trials, or that its effects are held in check by the objectively vertical orientation of the target array. That a deitic approach is not always taken is shown by the results from the condition in which the horizontal target array is below the arrow. In this case the vertical (and gravitational) orientation of observer and screen take precedence over any tendency for 'attention' to 'follow behind' the arrow and reverse the leftright orientation of the target array. The observer, in other words, does not 'stand on her head' in this task (Ladavas, 1987; Robertson and Lamb, 1988). We are, of course, well-aware that our account of these results in terms of differential spatial compression or of (mysterious) vectorial forces (Kinsbourne, 1970b; De Renzi et aI., 1989) is only metaphorical. Nonetheless, the data are themselves extremely lawful and they have no obvious interpretation within any traditional framework that assumes a 'cognitive cut' in the spatial world of the patient with neglect. Similarly, it is unclear how models that focus on the 'disengagement' component of covert attention in parietal neglect (Posner, 1988), would account for our data. In this task, the problem for P.P. lies not in disengaging attention from the stimulus arrow. Rather she seems to be systematically 'blown off course' after disengaging from the arrow and when moving in the direction of the target; thus it is the second component of Posner's theory ('shifting') that appears to be impaired in P.P.'s

628

P. W. Halligan and J. C. Marshall

performance here. We accordingly suggest that fresh metaphors of so-called 'neglect' will playa valuable heuristic role in generating new data. And that such data may eventually force a radical revision in our theories of the neglect syndromes. As far as we know, the notion of spatial compression has not been previously applied to data from patients with visual neglect. However, Werth and Poppe I (1988) have shown that there is an apparent subjective compression of space when normal subjects are required to bisect imaginary (horizontal) lines that extend to the left or right of a (real) vertical mark. In both imagined hemispaces, their subjects "overestimated the distance between the mark and the midpoint of the shortest (imagined) line, and they underestimated the distance between the mark and the midpoint of the longest (imagined) line". This result is reminiscent of our finding that neglect patients (Halligan and Marshall,1988) and normal controls (Manning, Halligan and Marshall, 1990) show 'crossover' effects on (real) line bisection as a function of stimulus length; leftwards transection displacements at short line lengths give way to rightwards displacements at longer lengths. Crossover phenomena, and the characteristic form of drawing or copying performance have no immediate interpretation in terms of compression. Nonetheless, our current data, and those of Werth and Poppe! (1988) do, at very least, suggest that the metaphor of spatial compression should be explored more systematically in both normal subjects and in patients with 'neglect' disorders. And that the time may be fast approaching when neglect data can be deployed to constrain genuine computational models of vi suo spatial attention (see Bisiach and Berti, 1987, and Kosslyn et aI., 1990). ABSTRACT

In the standard account of left neglect, some manner of attentional boundary is postulated such that elements to the left of that boundary are cognitive!y neglected. We propose an alternative model in which space is distorted ('compressed') in neglect. A new task is devised whereby the subject must follow 'in imagination' the direction of an arrowhead across 'empty' space to its corresponding position in a numerical target array. The two-dimensional array is Euclidian and all four arrow/array relationships are incorporated (arrow to the left/ right of a vertical array, arrow to the top/bottom of a horizontal array). Normal subjects perform at ceiling, with excellent accuracy in all orientations and positions. A patient with severe left neglect (consequent upon lesion of the right temporo-parietal region) shows systematic deflections in her judgement of target positions. These distortions are fully consistent with a model whereby points in 'left space' are compressed rightwards; the compression function is linearly proportional to the coordinates of Euclidian space.

Acknowledgements. This work was supported by the Chest, Heart, and Stroke Association (P.W.H.), and by the Medical Research Council (J.C.M.). We wish to thank Dr. George Burnett-Steuart (Oxford), and Professor Roger Wales (Melbourne) for some extremely helpful discussions. The incisive comments of Dr. Edoardo Bisiach (Milano) are also gratefully acknow ledged. REFERENCES

BISIACH, E., and BERTI, A. Dyschiria: An attempt at its systematic explanation. In M. leannerod (Ed.), Neurophysiological and Neuropsychological Aspects of Spatial Neglect. Amsterdam: North-HoIland, 1987. BISIACH, E., and LUZZATTI, C. Unilateral neglect of representational space. Cortex, 14: 129-133, 1978. BISIACH, E., and VALLAR, G. Hemineglect in humans. In F. Boller and 1. Grafman (Eds.), Handbook of Neuropsychology (Volume 1). Amsterdam: Elsevier, 1988. BISIACH, E., BULGARELLI, C., STERZI, R., and VALLAR, G. Line bisection and cognitive plasticity of unilateral neglect of space. Brain and Cognition, 2: 32-39, 1983. BISIACH, E., CAPITANI, E., LUZZATTI, C., and PERANI, D. Brain and conscious representation of outside reality. Neuropsychologia, 19: 543-551, 1981. BISIACH, E., CORNACCHIA, L., STERZI, R., and VALLAR, G. Disorders of perceived auditory lateralization after lesions of the right hemisphere. Brain, 107: 37-52, 1984.

Spatial compression in visual neglect

629

DE RENZI, E., GENTILINI, M., FAGLIONI, P., and BARBIERI, C. Attentional shift towards the rightmost stimuli in patients with left visual neglect. Cortex, 25: 231-237, 1989. HALLIGAN, P.W., and MARSHALL, J.C. How long is a piece of string? A study of line bisection in a case of visual neglect. Cortex, 24: 321-328, 1988. HALLIGAN, P.W., and MARSHALL, J.C. Perceptual cueing and perceptuo-motor compatibility in vi suospatial neglect: A single case study. Cognitive Neuropsychology, 6: 423-435, 1989. HALLIGAN, P.W., MARSHALL, J.C., and WADE, D.T. Do visual field deficits exacerbate vi suo-spatial neglect? Journal of Neurology, Neurosurgery and Psychiatry, 53: 487-491, 1990. KARTSOUNIS, L.D., and WARRINGTON, E.K. Unilateral visual neglect overcome by cues implicit in stimulus arrays. Journal of Neurology, Neurosurgery and Psychiatry, 52: 1253-1259, 1989. KINSBOURNE, M. A model for the mechanism of unilateral neglect of space. Transactions of the American Neurological Association, 95: 143-146, 1970a. KINSBOURNE, M. The cerebral basis of lateral asymmetries in attention. Acta Psychologica, 33: 193-201, 1970b. KOSSLYN, S.M., FLYNN, R.A., AMSTERDAM, J.B., and WANG, G. Components of high level vision: A cognitive neuroscience analysis and accounts of neurological syndromes. Cognition, 34: 203-277, 1990. LADAVAS, E. Is the hemispatial deficit produced by right parietal lobe damage associated with retinal or gravitational co-ordinates? Brain, 110: 167-180, 1987. MANNING, L., HALLIGAN, P.W., and MARSHALL, J.C. Individual variation in line bisection: A study of normal subjects with application to the interpretation of visual neglect. Neuropsychologia, 28: 647655, 1990. MARSHALL, J.C., and HALLIGAN, P.W. Blindsight and insight in visuo-spatial neglect. Nature, 336: 766767, 1988. MARSHALL, J .C., and HALLIGAN, P.W. When right goes left: An investigation of line bisection in a case of visual neglect. Cortex, 25: 503-515, 1989a. MARSHALL, J.C., and HALLIGAN, P.W. Does the midsagittal plane play any privileged role in 'left' neglect? Cognitive Neuropsychology, 6: 403-422, 1989b. POSNER, M.I. Structures and functions of selective attention. In T. Boll and B.K. Bryant (Eds.), Clinical Neuropsychology and Brain Function. Washington, DC: American Psychological Association, 1988. ROBERTSON, L.C., and LAMB, M.R. The role of perceptual reference frames in visual field asymmetries. Neuropsychologia,26: 145-152, 1988. SERON, X., COYETTE, F., and. BRUYER, R. Ipsilateral influences on contralateral processing in neglect patients. Cognitive Neuropsychology, 6: 475-498, 1989. WERTH, R., and POPPEL, E. Compression and lateral shift of mental coordinate systems in a line bisection task. Neuropsychologia, 26: 741-745, 1988. Dr. Peter Halligan, MRC Neuropsychology Unit, University Dept. of Clinical Neurology Rivermead Rehabilitation Centre Abingdon Road, Oxford, OXl 4XD.