FRAMES OF REFERENCE FOR ALLOCATING SPACE: EVIDENCE FROM THE NEGLECT
ATTENTION SYNDROME
TO
MARTHA J. FARAH,* JENNIFERL. BRUNK. ADRIENNE B. WONG. MARGIE A. WALLACE
and PATRICIA A. CARPENTER Departmentof Psycholog!. Carnegie-MellonUmversity.Pittsburgh.PA 15213. U.S.A
Abstract-With respect to what frames of reference. or spatial coordinate systems. is attention allocated to locations in space’? We posed this question about the spatial attention system that has been damaged in neglect patlen&. distinguishing among three possible types of spatial reference frame: viewer-centered. according to which locations are coded with respect to the vieHer. environment-centered. according to which locations are coded with respect lo the environment. and object-centered, according to which locations are coded with respect to an object. The three candidate frames of reference were decoupled from one another by rotating either the viewer or the stimulus object. Visual search performance suggested that the neglected hemifield was defined with respect to both viewer-centered and environment-centered frames ofreference. but not with respect to an objectcentered frame of reference. The role of objects in the allocation ofattention to space. and the relation between our findings and the “two cortical visual systems” hypothesis. are discussed.
INTRODUCTION PATIENTSwith neglect fail to detect stimuli
on the side of space contralateral to their lesions, despite having sensory and motor capabilities adequate to the task. The underlying nature of their deficit has been described both in terms of an impairment in the allocation of attention to the affected side of space [ 12. 15. 19. 201. and in terms of the inability to consciously represent that side of space [4. S]. Whichever type of explanation is invoked, the issue arises of how space is represented by the brain systems that have been damaged in neglect patients. This issue arises directly in the case of representational accounts, and indirectly in the case of attentional accounts, insofar as spatial attention is allocated to representations of locations in space and not. of course. to the locations themselves. A basic question about the representations of space involved in the neglect syndrome concerns the frames of reference. or spatial coordinate systems, that arc used to code location. In other words when a right hemisphere-damaged patient neglects the left side of space. with respect to what frame (or frames) of reference is “left” defined? In this article we consider three candidate frames of reference. Researchers in computational vision have found that these three types of reference frame are needed for performing the range of visualspatial tasks of which humans are capable. These frames of reference are: viewer-centered. environment-centered, and object-centered [ 2. I I. 13. 171. Each of these types of frames of reference will be described brietly here.
*Author
lo whom corrc\pondence
should
he addrescd.
336
M. J.
FAKAH rr al.
With a viewer-centered frame of reference. the locations of stimuli are represented relative to the viewer. This is the form of spatial representation in which stimuli are initially registered, specifically in a frame of reference centered on the retina. One disadvantage of a viewer-centered frame of reference is that if either the viewer moves or the stimuli move. the stimulus representation changes. In MARR and NISHIHARA’S [ 1S] terms, viewer-centered representations have poor stability over movements of the viewer or the object. An advantage of viewer-centered representations is that they require little computation to derive, as visual stimuli are automatically encoded in retinotopic coordinates. That is. they are easily uccessihle. If spatial attention operates over a viewer-centered frame of reference, then “left” is defined as to the left of the viewer’s midline, regardless ofthe position and orientation of the viewer with respect to objects and the environment. With an environment-centered frame of reference, the locations of stimuli are represented relative to the environment. If the viewer moves, the representation of stimuli does not change. The stability of environment-centered representations over viewer movements makes these representations useful for spatial functions such as navigation and the guidance ofeye-movements. They allow the viewer to move and explore the visual world without being confronted with a changed representation for that world after every step or every eye movement. However, this stability comes at the cost of some accessibility: The environmentcentered representation must be constructed from the initial viewer-centered representation using cues such as the locations of visual landmarks in the room, proprioceptive and vestibular information. If spatial attention operates over an environment-centered frame of reference, then “left” is defined as to the left of the environmental midline. regardless of the position and orientation of the viewer and the objects in the environment. With an object-centered frame of reference, the locations of objects (or parts of objects) are represented relative to an object. To see why such a representation is useful. note that neither of the two representations just discussed is stable over movements of the stimuli with respect to a stable viewer and environment. A representation of the spatial structure of an object, for example the location of a telephone’s receiver with respect to the dial, in a viewer-centered frame of reference will change when either the viewer or the object moves. At one moment the dial might be closer to the viewer than the receiver, and after some relative movement between the object and the viewer (e.g. the viewer walking around to the other side of the telephone) the dial might be further away than the receiver. A representation of the spatial structure of an object in an environment-centered frame of reference is also unstable. The receiver might be above the dial in gravitational coordinates when the phone is in its normal orientation. and below the dial when it is turned upside-down. Object recognition requires that the spatial structure ofan object be recoverable at any of the different possible positions and orientations of the object with respect to the viewer and the environment. One way to accomplish this is to represent the spatial structure ofan object with respect to itself, using an object-centered frame of reference. Because the object-centered frame of reference moves with the object, spatial representations in object-centered coordinates are stable over changes of the object’s position and orientation with respect to the viewer and the environment. The receiver is always at the top of the telephone. even when the top of the telephone is below the bottom of the telephone in view-centered or environment-centered coordinates. Again. stability comes at the price of accessibility. In the case of object-centered representations. it is computationally very difficult to assign an object-centered frame of reference to an object (e.g decide where its top. bottom. left. right. front and back are) before one has recognized the object. If spatial attention operates over an object-centered frame of
FRAMES
OF REFERENCE
FOR ALLOCATING
ATTENTIOU
TO SPACE
337
reference, then “left” is defined as to the left of the object’s intrinsic midline. regardless of the position and orientation of the object with respect to the viewer and the environment. Two recent investigations of neglect have examined the roles of these different frames of reference in the allocation of spatial attention to space. CALVANIO er al. [6] assessed the distribution of spatial attention over a stimulus display while right hemisphere-damaged neglect patients viewed the display sitting up and reclining on their sides. When sitting up. the patient’s viewer-centered left, the array-centered left and the environment-centered left are all aligned. When reclining, the viewer-centered frame of reference is brought roughly 90~ out of alignment with the other two, and hence the viewer’s left corresponds to the bottom half or top half of the other two reference frames. depending upon whether the viewer is reclining to the left or right, respectively. The stimulus displays consisted of an array of small pictures and words spanning the four quadrants of the visual field, and the number of pictures or words named by the patient in each quadrant was taken as a measure of the depress of spatial attention allocated there. When patients were sitting up, they named fewer items on the left than on the right. When they reclined, they named fewest items in the quadrant that was both to their personal left and the to left of the other two reference frames, and most items in the quadrant that was both the their personal right and to the right of the other two reference frames. Performance in the remaining two quadrants was intermediate. This striking finding led to CALVANIO et al. to propose that spatial attention is allocated to two representations of space: one defined with respect to the viewer. and, they conjectured, one defined with respect to the environment. LADAVAS[16] presented right parietal-damaged patients with a lateraiized simple reaction time task, in which a visual target stimulus could occur in one of two locations, located to the upper right and upper left of a fixation point. When patients viewed the display with their heads upright, they were slower in responding to target stimuli on the left than on the right because of parietal damage. When patients tilted their heads to the left or the right. bringing the two target locations entirely into the right or left visual field, they continued to be slower responding to the stimulus that would have been left-most if subjects’ heads were upright. Ladavas interprets this result as implying that neglect is both retinal and gravitational. The findings Of CALVANIO ef al. [6] and LADAVAS [ 163 demonstrate that neglect involves the differential allocation of attention over space within more than one spatial frame of reference. However, neither of these results completely specifies what frames of reference are used in the allocation of attention. Although Calvanio et al. and Ladavas interpret their findings in terms of two particular frames for allocating attention, one viewer-centered and one environment-centered, their findings are equally compatible with the use of a viewercentered and an object-centered frame of reference, centered on the stimulus array. In addition, the Ladavas result is compatible with two viewer-centered frames alone: one centered on the retinae or on the head. and one centered on the trunk. The purpose of the present experiment was to simultaneously contrast the contributions of viewer-centred, environment-centered and object-centered frames of reference in the allocation of spatial attention.
METHODS Malrrials
Twenty line
drawings of common oblects and ammals that have a canonical orientation were selected from the and VANDERWAHT 1211collection of drawings.They were: ape, cake, camel, cow, elephant, frog, house. iron, kettle, rabbit. racoon. record player. rhino, sailboat. squirrel, telephone, tree, turtle, wagon, well. These were SNOUGRASS
33x
M. J.
FAKAH
et al
printed on sheets of X.5 by I1 in. paper. one per page. at a sire that filled the page. Five of the 10 animals were depicted facing left. and five Pacing right. File copies of the drawings were made, resulting in fibe sets of the 20 drawings, a total of 100 pages of drawings. Sixteen randomly chosen letters of the alphabet were printed on each of these lOOdrawings, scattered pseudo-randomly over the ptcture with theconstraint that each ofthe four quandrants of the picture contain four letters, and these letters were no closer than a half-mch to the quandrant boundary (so that quadrant membership was unambiguous for irregularly shaped objects). No letter occurred more that once on a copy of a drawing. Figure 1 shows an example of a stimulus.
Fig. 1. Sample stimulus.
Subjects were tested while in their beds. The stimuli were presented hehind an upright clear plastic cool\-hooh stand. The stand was located on an adjustable height table that could be moved to either side of the bed or placed over the bed. directly in front of the subject. The distance between the subjects and the stimuli was approxtmately 2 ft. The experiment was carried out in two sessions ofapproximately equal length. The subjects’ task was to name the object or antma) tn the ptcture. and then to read all ofthe letters they could see. Responses wcrc tape-recorded for later transcrtption. Suhtects viewsed the stimuli under five different conditions: (I) subject rotated 90 clockwtse (i.e. lying on their right side) with the picture upright. (2) subject rotated 9W counterclockwtse (i.e. lying on thetr left side) with the ptcture upright. (3) subject upright (i.e. sitting up), with the picture rotated 90 clockwtse. (3) subject upright, with the ptcturc rotated 90 counterclockwise. and (5) subject upright and picture uprtght. Each subject viewed ten ptcturcs at a ttme undcrcach cundttion. and thcrcfore had to perform in each conditton twice tn order to have vicwcd each ofthe tacnty stnnulus ttcms under each condition. The order ofcondttions was varied over subjects so that each condttton occurred equally often in each serial position. in order to prevent possthle confoundtngs between condttton and practtcc or fatigue. In additton. each of the five verstons of the ptcturcs was asstgned equally often to each of the live condittons to prevent confoundings between particular vcraions (which might vary in difhculty dcpendtns upon the placement oflcttcrs) and viewing conditions. Care was taken to center the pictures with respect to the subjects‘natural lincofgazc tn each p~)sition.Twocspcrimcnterh were present. and thcsolc tash ofone ofthem ha\ to monttor suhjccts closely to prevent them from changtng thetr head position (e.g. ttltmg tt when biewtng rotated ptcturcs or ratstng tt tov.ard an upright position when lymg on thctr stde).
Suhjccts Hcrc screened for neglect ustnp an uprtght sample stunulus vtcucd from a normal uprtght positton. Ten rtght tlcmlsphcrc-dama8cd pattcnts neglected more than tv+o Icttcra on the left side of the screenmg stimulus and \*cre recruttcd t”rthe e\pertment. All subjects were right-handed: nine had suRered infarcttons and one (subject
m*bmm
no. 5) a hemorrhage. Table and neurological signs.
REFERENCEFOR
ALLOCATING
ATTESTlOU
TOSPACE
1shows then age. sex. number of weeks since onset. lesion localization
339 from CT report
RESULTS Subjects rarely showed any difficulty naming the stimulus drawing. and occasional errors were quickly corrected by the subjects after being given a general semantic cue (e.g. “lives near a pond” for “frog”). Two initial measures of neglect were obtained from subjects’ performance in the task: the number of letters read by the subject in each quadrant of each stimulus, and the quadrant in which the subject started reading the letters for each of the stimuli. These initial measures were used to calculate two corresponding measures of the distribution of attention relative to different frames of reference. as shown in Fig. 2 and explained in the following paragraph. In the condition in which both the stimuli and the subject were upright (Fig. 2a). quadrants 1 and 4 constitute the left of an environment-centered. viewer-centered, and object-centered frame of reference. In the condition in which the subect was lying on his or her side (Fig. 2b), quadrants 1 and 4 constitute the left and quandrants 2 and 3 the right ofan environment-centered and an object-centered frame of reference. When the subject is lying on his or her left side, quadrants 3 and 4 constitute the left and quadrants I and 2 the right of a viewer-centered frame of reference. When the subject is lying on his or her right side. quandrants 1 and 2 constitute the left and quandrants 3 and 4 the right side of a viewercentered frame of reference. When the stimulus if rotated (Fig. 2~). quandrants I and 4 constitute the left and quandrants 2 and 3 the right in an object-centered frame of reference. When the stimulus is rotated clockwise, quadrants 3 and 4 constitute the left and quandrants 1 and 2 the right of an environment-centered and viewer-centered frame of reference. When the stimulus is rotated counterclockwise quandrants 1 and 2 constitute the left and 3 and 4 the right of an environment-centered and a viewer-centered frame of reference. Table 2 shows the two measures of the distribution of attention with respect to each frame of reference for each subject in each condition. When all frames are aligned (part a). fewer letters are named on the left than on the right side. and this is significant by Wilcoxon test. T= 0,n= 10, P< 0.01. There is also a trend of borderline significance for subjects to begin reading on the right side, T= 6.5. II = 9.0.1 > P> 0.05. These results confirm that the subjects do indeed have left neglect. but they do not distinguish among the candidate frames of reference. When the viewer-centered frame of reference is decoupled from the environmmentcentered and object-centered frames of reference by having the subjects lie down on one side (Table 2b), there is evidence that more than one frame of reference is used to allocate attention. Subjects named fewer letters on the left than on the right with respect to a viewercentered frame of reference. T.=0,II = IO,P < 0.01. They also started reading more often on the right than on the left with respect to a view-centered frame of reference, T=O. II=IO, P~0.01.In addition. subjects named fewer letters on the left than on the right with respect to the environment-centered and:or object-centered frames of reference, T=O,II = IO,P 0.I These results indicate that a viewer-centered frame of reference is used in allocating attention to space. They also indicate that either an environment-centered or an object-centered frame of reference. or
340
M. _I. FAHAH
et
al
Table I. Age. sex. onset-to-test Interval. CT report and neurological signs of the ten patients who served as subjects in the experiment. All subjects were right-handed: nine had suffered infarctions and one (subject no. 5) a hemorrhage.
Age
Sex
Weeks since CVA
62 19 45 77
F M M F
1II I 4 8
41
M
9
6X
F
4
13 56
F F
8
14 71
F M
8 4
Location (CT scan) Partetal, Parletal. Parietal. Parietal, temporal, Parietal.
frontal frontal frontal frontal. occipital basal gangh a
L hemiparesis. L homonymous L hemiparcsls. L hemisensory L hemiplegia, L hemisensory L hemiparesis. L hemisensor! L homonymous hemianopsia L hemtplegla. L facial droop. hemianopsia. L hemlsensory L hemiplegla. L hemisensory L inferior quadrantopsia L hemtplegia. L hemisensorq L hemiplegia. L hemtsensory L homonyous hemtanopsia L hemtparesis. L hemisensory L hemiparesis. L hononymous
Parletal. frontal. temporal Parietal. frontal Basal ganglia Basal ganglia Parietal. frontal, occipital
I
I
.1, H IA /
Neurologtcal stgns to left neglect
ltn addition
I
hemianopsla defictt extinction delicit L homonymous deficit deficit. deficit delicit, defclt hemianopsla
I
I
1
:, ;-
Fig. 2 Rclattons between stimulus quadrants and the orientations or the vwwcr. cn\lronment and object in the three conditions: (a) object and viewer upright. (hl ohjcct upri_eht. viewer rotated. (cl object rotated. viewer upright.
both. is also used. To distinguish the contributions of these latter two frames, we must examine the distribution of attention when they are decoupled. When the stimulus is rotated (Table 3~). the object-centered frame of reference is decoupled from the environment-centered frame. Because subjects were upright in this condition. the environment-centered frame is then aligned with the viewer-centered frame. There was a small and nonsignificant trend for fewer letters to be read on the left than on the right with respect to the object-centered frame (T= 12,II=10.P>O.l).and this was balanced by an opposite and also nonsignificant trend in starting quadrant: Subjects started more often on the left than on the right with respect to an object-centered frame. T=30, r1=8, P> 0.I As would be expected on the basis of the results of the other two conditions, subjects read fewer letters on the left than on the right with respect to the viewer-centered and/or environment-centered frames, T= 0.H = IO,P
ATTENTION SYNDROME
TO
MARTHA J. FARAH,* JENNIFERL. BRUNK. ADRIENNE B. WONG. MARGIE A. WALLACE
and PATRICIA A. CARPENTER Departmentof Psycholog!. Carnegie-MellonUmversity.Pittsburgh.PA 15213. U.S.A
Abstract-With respect to what frames of reference. or spatial coordinate systems. is attention allocated to locations in space’? We posed this question about the spatial attention system that has been damaged in neglect patlen&. distinguishing among three possible types of spatial reference frame: viewer-centered. according to which locations are coded with respect to the vieHer. environment-centered. according to which locations are coded with respect lo the environment. and object-centered, according to which locations are coded with respect to an object. The three candidate frames of reference were decoupled from one another by rotating either the viewer or the stimulus object. Visual search performance suggested that the neglected hemifield was defined with respect to both viewer-centered and environment-centered frames ofreference. but not with respect to an objectcentered frame of reference. The role of objects in the allocation ofattention to space. and the relation between our findings and the “two cortical visual systems” hypothesis. are discussed.
INTRODUCTION PATIENTSwith neglect fail to detect stimuli
on the side of space contralateral to their lesions, despite having sensory and motor capabilities adequate to the task. The underlying nature of their deficit has been described both in terms of an impairment in the allocation of attention to the affected side of space [ 12. 15. 19. 201. and in terms of the inability to consciously represent that side of space [4. S]. Whichever type of explanation is invoked, the issue arises of how space is represented by the brain systems that have been damaged in neglect patients. This issue arises directly in the case of representational accounts, and indirectly in the case of attentional accounts, insofar as spatial attention is allocated to representations of locations in space and not. of course. to the locations themselves. A basic question about the representations of space involved in the neglect syndrome concerns the frames of reference. or spatial coordinate systems, that arc used to code location. In other words when a right hemisphere-damaged patient neglects the left side of space. with respect to what frame (or frames) of reference is “left” defined? In this article we consider three candidate frames of reference. Researchers in computational vision have found that these three types of reference frame are needed for performing the range of visualspatial tasks of which humans are capable. These frames of reference are: viewer-centered. environment-centered, and object-centered [ 2. I I. 13. 171. Each of these types of frames of reference will be described brietly here.
*Author
lo whom corrc\pondence
should
he addrescd.
336
M. J.
FAKAH rr al.
With a viewer-centered frame of reference. the locations of stimuli are represented relative to the viewer. This is the form of spatial representation in which stimuli are initially registered, specifically in a frame of reference centered on the retina. One disadvantage of a viewer-centered frame of reference is that if either the viewer moves or the stimuli move. the stimulus representation changes. In MARR and NISHIHARA’S [ 1S] terms, viewer-centered representations have poor stability over movements of the viewer or the object. An advantage of viewer-centered representations is that they require little computation to derive, as visual stimuli are automatically encoded in retinotopic coordinates. That is. they are easily uccessihle. If spatial attention operates over a viewer-centered frame of reference, then “left” is defined as to the left of the viewer’s midline, regardless ofthe position and orientation of the viewer with respect to objects and the environment. With an environment-centered frame of reference, the locations of stimuli are represented relative to the environment. If the viewer moves, the representation of stimuli does not change. The stability of environment-centered representations over viewer movements makes these representations useful for spatial functions such as navigation and the guidance ofeye-movements. They allow the viewer to move and explore the visual world without being confronted with a changed representation for that world after every step or every eye movement. However, this stability comes at the cost of some accessibility: The environmentcentered representation must be constructed from the initial viewer-centered representation using cues such as the locations of visual landmarks in the room, proprioceptive and vestibular information. If spatial attention operates over an environment-centered frame of reference, then “left” is defined as to the left of the environmental midline. regardless of the position and orientation of the viewer and the objects in the environment. With an object-centered frame of reference, the locations of objects (or parts of objects) are represented relative to an object. To see why such a representation is useful. note that neither of the two representations just discussed is stable over movements of the stimuli with respect to a stable viewer and environment. A representation of the spatial structure of an object, for example the location of a telephone’s receiver with respect to the dial, in a viewer-centered frame of reference will change when either the viewer or the object moves. At one moment the dial might be closer to the viewer than the receiver, and after some relative movement between the object and the viewer (e.g. the viewer walking around to the other side of the telephone) the dial might be further away than the receiver. A representation of the spatial structure of an object in an environment-centered frame of reference is also unstable. The receiver might be above the dial in gravitational coordinates when the phone is in its normal orientation. and below the dial when it is turned upside-down. Object recognition requires that the spatial structure ofan object be recoverable at any of the different possible positions and orientations of the object with respect to the viewer and the environment. One way to accomplish this is to represent the spatial structure ofan object with respect to itself, using an object-centered frame of reference. Because the object-centered frame of reference moves with the object, spatial representations in object-centered coordinates are stable over changes of the object’s position and orientation with respect to the viewer and the environment. The receiver is always at the top of the telephone. even when the top of the telephone is below the bottom of the telephone in view-centered or environment-centered coordinates. Again. stability comes at the price of accessibility. In the case of object-centered representations. it is computationally very difficult to assign an object-centered frame of reference to an object (e.g decide where its top. bottom. left. right. front and back are) before one has recognized the object. If spatial attention operates over an object-centered frame of
FRAMES
OF REFERENCE
FOR ALLOCATING
ATTENTIOU
TO SPACE
337
reference, then “left” is defined as to the left of the object’s intrinsic midline. regardless of the position and orientation of the object with respect to the viewer and the environment. Two recent investigations of neglect have examined the roles of these different frames of reference in the allocation of spatial attention to space. CALVANIO er al. [6] assessed the distribution of spatial attention over a stimulus display while right hemisphere-damaged neglect patients viewed the display sitting up and reclining on their sides. When sitting up. the patient’s viewer-centered left, the array-centered left and the environment-centered left are all aligned. When reclining, the viewer-centered frame of reference is brought roughly 90~ out of alignment with the other two, and hence the viewer’s left corresponds to the bottom half or top half of the other two reference frames. depending upon whether the viewer is reclining to the left or right, respectively. The stimulus displays consisted of an array of small pictures and words spanning the four quadrants of the visual field, and the number of pictures or words named by the patient in each quadrant was taken as a measure of the depress of spatial attention allocated there. When patients were sitting up, they named fewer items on the left than on the right. When they reclined, they named fewest items in the quadrant that was both to their personal left and the to left of the other two reference frames, and most items in the quadrant that was both the their personal right and to the right of the other two reference frames. Performance in the remaining two quadrants was intermediate. This striking finding led to CALVANIO et al. to propose that spatial attention is allocated to two representations of space: one defined with respect to the viewer. and, they conjectured, one defined with respect to the environment. LADAVAS[16] presented right parietal-damaged patients with a lateraiized simple reaction time task, in which a visual target stimulus could occur in one of two locations, located to the upper right and upper left of a fixation point. When patients viewed the display with their heads upright, they were slower in responding to target stimuli on the left than on the right because of parietal damage. When patients tilted their heads to the left or the right. bringing the two target locations entirely into the right or left visual field, they continued to be slower responding to the stimulus that would have been left-most if subjects’ heads were upright. Ladavas interprets this result as implying that neglect is both retinal and gravitational. The findings Of CALVANIO ef al. [6] and LADAVAS [ 163 demonstrate that neglect involves the differential allocation of attention over space within more than one spatial frame of reference. However, neither of these results completely specifies what frames of reference are used in the allocation of attention. Although Calvanio et al. and Ladavas interpret their findings in terms of two particular frames for allocating attention, one viewer-centered and one environment-centered, their findings are equally compatible with the use of a viewercentered and an object-centered frame of reference, centered on the stimulus array. In addition, the Ladavas result is compatible with two viewer-centered frames alone: one centered on the retinae or on the head. and one centered on the trunk. The purpose of the present experiment was to simultaneously contrast the contributions of viewer-centred, environment-centered and object-centered frames of reference in the allocation of spatial attention.
METHODS Malrrials
Twenty line
drawings of common oblects and ammals that have a canonical orientation were selected from the and VANDERWAHT 1211collection of drawings.They were: ape, cake, camel, cow, elephant, frog, house. iron, kettle, rabbit. racoon. record player. rhino, sailboat. squirrel, telephone, tree, turtle, wagon, well. These were SNOUGRASS
33x
M. J.
FAKAH
et al
printed on sheets of X.5 by I1 in. paper. one per page. at a sire that filled the page. Five of the 10 animals were depicted facing left. and five Pacing right. File copies of the drawings were made, resulting in fibe sets of the 20 drawings, a total of 100 pages of drawings. Sixteen randomly chosen letters of the alphabet were printed on each of these lOOdrawings, scattered pseudo-randomly over the ptcture with theconstraint that each ofthe four quandrants of the picture contain four letters, and these letters were no closer than a half-mch to the quandrant boundary (so that quadrant membership was unambiguous for irregularly shaped objects). No letter occurred more that once on a copy of a drawing. Figure 1 shows an example of a stimulus.
Fig. 1. Sample stimulus.
Subjects were tested while in their beds. The stimuli were presented hehind an upright clear plastic cool\-hooh stand. The stand was located on an adjustable height table that could be moved to either side of the bed or placed over the bed. directly in front of the subject. The distance between the subjects and the stimuli was approxtmately 2 ft. The experiment was carried out in two sessions ofapproximately equal length. The subjects’ task was to name the object or antma) tn the ptcture. and then to read all ofthe letters they could see. Responses wcrc tape-recorded for later transcrtption. Suhtects viewsed the stimuli under five different conditions: (I) subject rotated 90 clockwtse (i.e. lying on their right side) with the picture upright. (2) subject rotated 9W counterclockwtse (i.e. lying on thetr left side) with the ptcture upright. (3) subject upright (i.e. sitting up), with the picture rotated 90 clockwtse. (3) subject upright, with the ptcturc rotated 90 counterclockwise. and (5) subject upright and picture uprtght. Each subject viewed ten ptcturcs at a ttme undcrcach cundttion. and thcrcfore had to perform in each conditton twice tn order to have vicwcd each ofthe tacnty stnnulus ttcms under each condition. The order ofcondttions was varied over subjects so that each condttton occurred equally often in each serial position. in order to prevent possthle confoundtngs between condttton and practtcc or fatigue. In additton. each of the five verstons of the ptcturcs was asstgned equally often to each of the live condittons to prevent confoundings between particular vcraions (which might vary in difhculty dcpendtns upon the placement oflcttcrs) and viewing conditions. Care was taken to center the pictures with respect to the subjects‘natural lincofgazc tn each p~)sition.Twocspcrimcnterh were present. and thcsolc tash ofone ofthem ha\ to monttor suhjccts closely to prevent them from changtng thetr head position (e.g. ttltmg tt when biewtng rotated ptcturcs or ratstng tt tov.ard an upright position when lymg on thctr stde).
Suhjccts Hcrc screened for neglect ustnp an uprtght sample stunulus vtcucd from a normal uprtght positton. Ten rtght tlcmlsphcrc-dama8cd pattcnts neglected more than tv+o Icttcra on the left side of the screenmg stimulus and \*cre recruttcd t”rthe e\pertment. All subjects were right-handed: nine had suRered infarcttons and one (subject
m*bmm
no. 5) a hemorrhage. Table and neurological signs.
REFERENCEFOR
ALLOCATING
ATTESTlOU
TOSPACE
1shows then age. sex. number of weeks since onset. lesion localization
339 from CT report
RESULTS Subjects rarely showed any difficulty naming the stimulus drawing. and occasional errors were quickly corrected by the subjects after being given a general semantic cue (e.g. “lives near a pond” for “frog”). Two initial measures of neglect were obtained from subjects’ performance in the task: the number of letters read by the subject in each quadrant of each stimulus, and the quadrant in which the subject started reading the letters for each of the stimuli. These initial measures were used to calculate two corresponding measures of the distribution of attention relative to different frames of reference. as shown in Fig. 2 and explained in the following paragraph. In the condition in which both the stimuli and the subject were upright (Fig. 2a). quadrants 1 and 4 constitute the left of an environment-centered. viewer-centered, and object-centered frame of reference. In the condition in which the subect was lying on his or her side (Fig. 2b), quadrants 1 and 4 constitute the left and quandrants 2 and 3 the right ofan environment-centered and an object-centered frame of reference. When the subject is lying on his or her left side, quadrants 3 and 4 constitute the left and quadrants I and 2 the right of a viewer-centered frame of reference. When the subject is lying on his or her right side. quandrants 1 and 2 constitute the left and quandrants 3 and 4 the right side of a viewercentered frame of reference. When the stimulus if rotated (Fig. 2~). quandrants I and 4 constitute the left and quandrants 2 and 3 the right in an object-centered frame of reference. When the stimulus is rotated clockwise, quadrants 3 and 4 constitute the left and quandrants 1 and 2 the right of an environment-centered and viewer-centered frame of reference. When the stimulus is rotated counterclockwise quandrants 1 and 2 constitute the left and 3 and 4 the right of an environment-centered and a viewer-centered frame of reference. Table 2 shows the two measures of the distribution of attention with respect to each frame of reference for each subject in each condition. When all frames are aligned (part a). fewer letters are named on the left than on the right side. and this is significant by Wilcoxon test. T= 0,n= 10, P< 0.01. There is also a trend of borderline significance for subjects to begin reading on the right side, T= 6.5. II = 9.0.1 > P> 0.05. These results confirm that the subjects do indeed have left neglect. but they do not distinguish among the candidate frames of reference. When the viewer-centered frame of reference is decoupled from the environmmentcentered and object-centered frames of reference by having the subjects lie down on one side (Table 2b), there is evidence that more than one frame of reference is used to allocate attention. Subjects named fewer letters on the left than on the right with respect to a viewercentered frame of reference. T.=0,II = IO,P < 0.01. They also started reading more often on the right than on the left with respect to a view-centered frame of reference, T=O. II=IO, P~0.01.In addition. subjects named fewer letters on the left than on the right with respect to the environment-centered and:or object-centered frames of reference, T=O,II = IO,P 0.I These results indicate that a viewer-centered frame of reference is used in allocating attention to space. They also indicate that either an environment-centered or an object-centered frame of reference. or
340
M. _I. FAHAH
et
al
Table I. Age. sex. onset-to-test Interval. CT report and neurological signs of the ten patients who served as subjects in the experiment. All subjects were right-handed: nine had suffered infarctions and one (subject no. 5) a hemorrhage.
Age
Sex
Weeks since CVA
62 19 45 77
F M M F
1II I 4 8
41
M
9
6X
F
4
13 56
F F
8
14 71
F M
8 4
Location (CT scan) Partetal, Parletal. Parietal. Parietal, temporal, Parietal.
frontal frontal frontal frontal. occipital basal gangh a
L hemiparesis. L homonymous L hemiparcsls. L hemisensory L hemiplegia, L hemisensory L hemiparesis. L hemisensor! L homonymous hemianopsia L hemtplegla. L facial droop. hemianopsia. L hemlsensory L hemiplegla. L hemisensory L inferior quadrantopsia L hemtplegia. L hemisensorq L hemiplegia. L hemtsensory L homonyous hemtanopsia L hemtparesis. L hemisensory L hemiparesis. L hononymous
Parletal. frontal. temporal Parietal. frontal Basal ganglia Basal ganglia Parietal. frontal, occipital
I
I
.1, H IA /
Neurologtcal stgns to left neglect
ltn addition
I
hemianopsla defictt extinction delicit L homonymous deficit deficit. deficit delicit, defclt hemianopsla
I
I
1
:, ;-
Fig. 2 Rclattons between stimulus quadrants and the orientations or the vwwcr. cn\lronment and object in the three conditions: (a) object and viewer upright. (hl ohjcct upri_eht. viewer rotated. (cl object rotated. viewer upright.
both. is also used. To distinguish the contributions of these latter two frames, we must examine the distribution of attention when they are decoupled. When the stimulus is rotated (Table 3~). the object-centered frame of reference is decoupled from the environment-centered frame. Because subjects were upright in this condition. the environment-centered frame is then aligned with the viewer-centered frame. There was a small and nonsignificant trend for fewer letters to be read on the left than on the right with respect to the object-centered frame (T= 12,II=10.P>O.l).and this was balanced by an opposite and also nonsignificant trend in starting quadrant: Subjects started more often on the left than on the right with respect to an object-centered frame. T=30, r1=8, P> 0.I As would be expected on the basis of the results of the other two conditions, subjects read fewer letters on the left than on the right with respect to the viewer-centered and/or environment-centered frames, T= 0.H = IO,P
