Cognitive Brain Research 10 (2001) 317–322 www.elsevier.com / locate / bres Short communication Shifting visuo-spatial attention in a virtual three-dimensional space Francesco Maringelli a , John McCarthy b , Anthony Steed c , Mel Slater c , Carlo Umilta` d , * a Cognitive Neuroscience Sector, International School for Advanced Studies, Trieste, Italy b Department of Psychology, University College London, London, UK c Department of Computer Science, University College London, London, UK d Department of General Psychology, University of Padova, Padova, Italy Accepted 13 June 2000 Abstract The present investigation, with a virtual reality set-up, aimed to study attentional orienting within a three-dimensional visual world. Near and far stimuli were used. Half of the subjects were provided with a virtual representation of their body, whereas half were not. Results showed a different distribution of attentional resources in the two conditions, suggesting a dissociation between attentional systems controlling the proximal and the distal visual space. In particular, attention was focused close to the subject’s body when a virtual representation of it was present, whereas attention was focused away from the body when a virtual representation of the body was not present. 2001 Elsevier Science B.V. All rights reserved. Theme: Neural basis of behavior Topic: Cognition Keywords: Spatial attention; Orienting; Visual space; Space representation; Three-dimensional; Virtual reality 1. Introduction During the last 50 years several paradigms were developed to examine the dynamics of visual attention. In particular, Posner [16] showed that it is possible to pay attention to a certain spatial location, independently of the direction of gaze. Subjects were instructed to fixate a central fixation point and to respond as quickly as possible to the appearance of a simple target, a light, in the periphery. Before the target was shown, a cue was presented, which predicted with a high probability where the subsequent target would be shown. There were three experimental conditions: a valid condition (cue and target were in the same spatial location), an invalid condition (cue and target were in different spatial locations), and a neutral condition (all target locations were cued). Posner showed faster reaction time (RT) in the valid condition than in the neutral condition (attentional ‘benefits’), and slower RT in the invalid condition than in the neutral *Corresponding author. condition (attentional ‘costs’). Because subjects were not allowed to move their eyes, he interpreted these differences as a consequence of a correct / incorrect orienting of attention. Posner ([16]; see also Ref. [11]) also showed that peripheral non-symbolic cues (e.g. flashing a light below a given spatial location) elicited an automatic (involuntary) orienting of attention, whereas centrally located symbolic cues (e.g. an arrowhead pointing to a given spatial location) elicited a controlled (voluntary) orienting of attention. What we know about spatial attention is mainly based on experiments that were conducted within a two-dimensional visual world. Only a few studies addressed the dynamics of attention in a three-dimensional (3D) space. Downing and Pinker [8] asked their subjects to respond as quickly as possible to a target, a light, which was shown at one of eight possible locations. The lights were arranged in two different depth planes, with the subjects’ fixation directed between them. Before appearance of the target, a centrally located cue was presented. They showed that attentional benefits and costs for attentional shifts in both the frontal plane and depth. Also, they found that, in depth, 0926-6410 / 01 / $ – see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S0926-6410( 00 )00039-2 318 F. Maringelli et al. / Cognitive Brain Research 10 (2001) 317 – 322 attentional costs were greater when subjects were invalidly cued to a near location then when they were invalidly cued to a far location. Similar results were obtained by Gawryszewski et al. [9], who too found greater costs for shifting attention from near to far locations than for shifting attention from far to near locations. Gawryszewski et al. proposed that attentional resources are allocated in depth in a viewer-centred fashion, with a peak close to the subject’s body. Other studies addressing visual orienting in 3D space were carried out by Andersen [1] and Andersen and Kramer [2]. Andersen used a task in which subjects were instructed to respond to a centrally located target while ignoring the surrounding distractors. The distractors were presented in seven different depth planes: three were between the subject and the plane of the target, one was at the plane of the target, and the other three were beyond the plane of the target. The displays were random dot stereograms, viewed through a stereoscopic prism. The main result was that the size of the interference effect was directly related to the distance in depth between targets and distractors. However, contrary to the result of Downing and Pinker [8] and Gawryszewski et al. [9], the interference effect was greater for distant distractors than for close distractors. Attention seemed to be allocated in a viewercentred fashion, as shown by previous investigations, but with its peak concentrated at locations beyond fixation. However, as Andersen pointed out, the prism stereoscope and close viewing conditions (21 cm from the subject’s eyes) might have led subjects to perceive the far distractors as being larger. This difference in perceived size may account for the greater interference from far distractors. This hypothesis was investigated by Andersen and Kramer [2]. They reported results consistent with those reported by previous studies [8,9], that is greater costs were observed for near than for far distractors. Because the computer generated stereograms were perceived as being at a greater distance than in the previous study (140 cm vs. 21 cm), and subjects might have mapped them on to distant, extrapersonal space, Andersen and Kramer argued that this factor was responsible for the inversion of the attentional asymmetry. The studies described above suggest a viewer-centred (egocentric) system of co-ordinates for orienting attention in visual space. However, whereas such a co-ordinate system is needed to carry out some tasks, such as movement towards a target, there are converging lines of evidence in favour of another, probably parallel, co-ordinate system, based on an environment-centred (allocentric) reference frame. The latter is thought to be necessary for object identification. For example, a recent study [15] showed that the luminance centroid of the display serves as centre of gravity for attention allocation during pattern recognition. The luminance centroid appears to serve, in turn, as point of origin for the frame of reference within which this process is carried out. Several neurophysiological and neuropsychological ob- servations (e.g. Refs. [3–7,10,13,14,17,18]) are in favour of the existence of a functional and anatomical distinction between mechanisms controlling the near and the far space, and of the fact that information in near space might be processed separately and independently from information in far space. Moreover, near and far space seem to be characterised by different ‘attributes’. Near space is mainly motoric, and is internally represented in egocentric co-ordinates. In contrast, far space is mainly perceptual and is internally represented through allocentric co-ordinates. The present investigation aims to demonstrate in normal subjects a dissociation between two attentional systems operating in these two reference systems. To this purpose, we made use of virtual reality (VR) scenarios. 2. Experiment In this experiment the stimuli were presented at different depth planes within the same display, using the binocular disparity as an index of depth. The characteristics of the cues and targets were kept constant at the two depth planes. To use binocular disparity as an index of depth, the stimuli were presented in the so-called ‘overlapping region’ of the visual field. In addition, we gave half of the subjects a ‘virtual’ experience of their body, trunk and hand, whereas the others were not provided with such a virtual body. We thought that the group provided with the virtual body would be aware of the relative distance of near and far stimuli with respect to their body, whereas the other group would be aware of the differences between the two depth planes, but not (or less) aware of the relative distance of the stimuli from their body. Although we did not explicitly ask our subjects to map space and / or to make any distance judgement about space, we invited them to take a look at the virtual environment and to explore it ‘walking’ inside it for a while. It is conceivable that this exploration produced an internal representation of virtual space, based on distance evaluation among relevant spots. Our prediction was that, in the body absent condition, subjects would focus their attention further away than in the body present condition, in which attention should be naturally focused closer to the body. In particular, we thought that the presence of the body should automatically lead to a ‘normal’ distribution of attention to the virtual limbs and trunk, whereas subjects not provided with the virtual body would not be able to anchor their attention to a representation of their body inside the virtual environment. This different allocation of attention to the visual field should, in turn, lead to a difference in performance between the two experimental groups. 2.1. Method 2.1.1. Subjects Twenty students at the University College London F. Maringelli et al. / Cognitive Brain Research 10 (2001) 317 – 322 319 Fig. 1. Display size and stimuli eccentricities in the Near space (A) and Far space (B) condition. served as subjects in the experiment. They all had normal ¨ as to the or corrected-to-normal vision and were naıve purpose of the experiment. They were paid UK£4 for their participation. Half of the subjects were randomly assigned to the ‘body present’ condition, whereas the other half were assigned to the ‘body absent’ condition. 2.1.2. Apparatus and display The scenarios were implemented using dVS / dVISE 4.01 software from Division Limited. This ran on a Silicon Graphics Onyx with twin 196 MHz R10000 processors, Infinite Reality Graphics and 64M main memory. The tracking system was a Polhemus Fastrak, with two receivers. One receiver was attached to the head-mounted display (HMD), and the second to a standard three-button mouse that was held in the right hand. HMD was a Virtual Research VR4 that has a resolution of 7423230 pixels for each eye, 170 660 colour elements, and a field of view of 678 diagonal at 85% overlap. The scene consisted of a garden of dimension 50390 m. The garden, comprising a boundary wall, a path and some trees, was made by a few hundred polygons, which rendered at 60 Hz, 30 Hz per eye. Latency between the participant’s head movements and the corresponding change inside the HMD was approximately 120 ms. Half of the subjects were provided with a virtual body. The torso hung vertically under the position of the neck, and rotated, in yaw only, to point along the direction of gaze. The arm was modelled with a simple inverse-kinematics procedure so that the elbow flexed as the participants moved their hand. Apart from the virtual garden and the virtual body, the stimuli consisted of a classical spatial-attention experimental paradigm (e.g. Ref. [16]). The cues were yellow, bright arrows (4.48 high31.58 wide) and were shown at 168 of eccentricity (see Fig. 1). The target was a red, 3D modelled ball (3.18 diameter), which appeared either just over the cue or in the corresponding position in the opposite visual field, depending on the type of trial. There was also a fixation cross (2.932.98), which moved consistently with the subject’s head. Subjects did not move through the environment, but were free to turn around. Half of the trials were presented in the near space, at a distance of about 50 cm from the subject’s head, within his / her reachable space. The remaining trials were shown in the far space, at a distance of about 20 m far from the subject’s head, well beyond his / her reachable space. 2.1.3. Procedure In this experiment we used a 23232 factorial design. One of the factors, namely the Body / No-Body condition, was treated as a between-subjects factor, whereas Type of Space (near / far), and Type of Trial (valid / invalid) were manipulated within subjects. Within the main factor Type of Trial, we manipulated both the ‘X’ dimension (Side) and the ‘Z’ dimension (Depth). In other words, valid cued trials were situations in which the target appeared on the same side and at the same depth as the cue, whereas invalid cued trials were situations in which the target was shown at the diagonally opposite location with respect to the cue. We also used two filling conditions in which we showed valid side / invalid depth cued stimuli and vice versa. This manipulation was intended to ensure that subjects had a clear impression of both side and depth main manipulations. The Type of Space condition was manipulated using different binocular disparity for the near and far cues / targets. The Type of Trial manipulation was designed to explore benefits / costs due to the correct / incorrect orient- 320 F. Maringelli et al. / Cognitive Brain Research 10 (2001) 317 – 322 ing of attention.1 The proportion of valid / invalid trials was 2:1. Subjects responded to 60 valid trials and to 30 invalid trials. About 10% of the trials (10) were ‘catch trials’, on which no target was presented after the cue, and subjects were instructed to refrain from responding. Valid, invalid and catch trials were presented in random order. The fixation crosses and the cues were turned on at the beginning of the experiment and remained on until the end of it. The sequence of the event for each trial was as follows. First, either the left or the right cue was flashed for 35 ms.2 After 90 ms, the target appeared either in the same or in the opposite hemifield with respect to the cue, depending on type of trial. The target was turned off after 300 ms and the next trial began after an interval randomly distributed between 1500 and 2500 ms. Subjects were instructed to fixate the central fixation cross and to press as fast as possible one of the mouse keys when they saw the target. There were four fixation crosses in the environment, located on a circle centred at the subjects position and separated by intervals of 908. Subjects completed a block of 100 trials at each of the four positions. When they were introduced into a silent room, they were fitted with the HMD and given details about the environment in which they were to be immersed, and the task they were asked to perform. The entire experiment consisted of a unique session of 400 trials (240 valid, 120 invalid, and 40 catch trials) and lasted about 20 min. Most of the time subjects remained unaware of their virtual body, virtual arm and hand, because of the relatively limited field of view. 3. Results and discussion Errors were rare (less than 1%) and were not analysed. RTs comprised between 200 ms and 750 ms, which accounted for about 97% of the total number of responses, were entered into a three-way repeated-measures MANOVA with Virtual Body (present vs. absent) as a betweensubjects factor, and Type of Space (near vs. far), and Type of Trial (valid vs. invalid) as within-subjects factors. The two filling conditions (i.e. valid side / invalid depth and valid depth / invalid side) were not analysed. The analysis showed significant main effects of Type of 1 Although the classical paradigm consists of valid trials, invalid trials, and neutral trials, in this experiment we used only valid and invalid trials to shorten the experiment. This was due to an intrinsic characteristic of VR scenarios, namely that subjects may feel a sense of sickness when they spend more than half an hour in a virtual environment. Therefore, our benefits / costs analysis does not correspond to the classical one, in which benefits are estimated by the RT difference between valid and neutral conditions and costs are estimated by the RT difference between invalid and neutral conditions. 2 All timing had a variation of 65 ms, because of general characteristics of computer’s and keyboard / mouse updating frequencies. Fig. 2. Mean reaction time in ms for the Body present condition (group 1) and the Body absent condition (group 2). Space (F(1,18)57.64, P,0.013), and Type of Trial (F(1,18)523.69, P,0.0001). Subjects were about 8 ms faster to respond to far than to near targets (453 vs. 461 ms), and about 24 ms faster to respond to validly cued targets than to invalidly cued targets (445 vs. 469 ms). The between-subjects factor (Virtual Body) did not reach significance. Two interactions were significant, namely that between Virtual Body and Type of Space (F(1,18)57.29, P, 0.015), and that between Virtual Body, Type of Space, and Type of Trial (F(1,18)58.4, P,0.01). Because of the significant three-way interaction, we split our group into two sub-groups (simple main effect analysis; see Ref. [12]), on the basis of the Body / No-Body manipulation. Then two separate repeated-measures MANOVAs were run (see Fig. 2). 3.1. Group 1: Virtual Body present The MANOVA run on this condition (see Table 1) showed a significant main effect of Validity (F(1,9)5 12.47, P,0.006), and the interaction between Type of Space and Validity (F(1,9)55.72, P,0.04), indicating that, when the Virtual Body was provided, RTs were only marginally affected by orienting of attention in the near space in comparison to the far space. A Newman–Keuls post hoc analysis showed significant differences (P,0.01) between valid and invalid trials in both near and far space, but the difference in the far space was almost twice as big as the difference in near space. Table 1 Mean reaction time in ms for the Body present condition Invalid trials Valid trials Near space Far space 482 462 491 453 F. Maringelli et al. / Cognitive Brain Research 10 (2001) 317 – 322 Table 2 Mean reaction time in ms for the Body absent condition (group 2) Invalid trials Valid trials Near space Far space 462 436 440 429 These subjects seem to allocate attention more densely to the near space than to the far one. That is attested by the fact that smaller attentional benefits due to attentional orienting were present when targets were shown in the near space, in comparison to the far space. Note that this asymmetry in attentional allocation is congruent with the asymmetry found by Andersen and Kramer [2], Downing and Pinker [8], and Gawryszewski et al. [9]. 3.2. Group 2: Virtual Body absent The MANOVA run on data from this condition (see Table 2) showed significant main effects of Type of Space (F(1,9)512.31, P,0.007) and Type of Trial (F(1,9)5 12.27, P,0.007), whereas the interaction was not significant. Subjects were about 15 ms faster to respond to targets shown in the far space (434 vs. 449 ms), and about 18 ms faster to respond to validly cued targets (433 vs. 451 ms). In addition, attentional benefits in near space were more than twice greater than attentional benefits in far space. This asymmetry is opposite to the one found in the body present condition and is in accord with Andersen’s [1] results. 4. General discussion Our investigation has shown that VR is not only an efficient way of controlling input variables, but may provide unique information. In particular, we found that having a virtual body framed our subjects’ central representation in a viewer-centred way, leading to a dense distribution of attentional resources in the near space, close to the subject’s body. In contrast, when the representation of the body was lacking, attention was allocated in a rather different way. It is interesting to note that, although the level of ‘immersion’ 3 rendered by the VR set-up is rather high, subjects were directly exposed to their virtual body for a few minutes only at the beginning of the experiment. Because of the relatively limited field of view, during the experiment subjects remained mostly unaware of their virtual body. Nevertheless, our results clearly showed an effect of the virtual body on the distribution of attention: attention was largely focused on the representation of the 3 Immersion refers to the extent to which a system generates streams of sensory output that deliver an extensive, surrounding, vivid, and inclusive virtual environment to a participant. 321 body, leading to an asymmetry. Conversely, the presence in a virtual environment without a representation of the body may have led to a relatively passive attitude toward the experimental set-up. In this condition, subjects may have felt as if they were observers rather than actors. Pavlovskaya et al. [15] proposed that object recognition is a dual-step process, in which the selection of the frame of reference precedes identification / discrimination. In their view, under simple conditions (i.e. single object analysis), attention is allocated to the centroid of the display. In contrast, when more complex visual configurations are shown, the point of origin of the reference frame is chosen separately for each discrete object. This process implies frequent shifts of attention on the visual display. Their results are in accord with those we reported here. Our subjects seem to centre their attention at different locations in relation to the way the depth plane and the body are defined. Our results are also in accord with those of previous studies of 3D attentional orienting. The group provided with a virtual body showed an attentional asymmetry in depth that is similar to the one found by Andersen and Kramer [2], Downing and Pinker [8], and Gawryszewski et al. [9]. This asymmetry indicates that attention was allocated in a viewer-centred fashion, with its resource peak close to the subject’s body. In this condition, it was more difficult to disengage attention from the near position, and to engage it on the far position, than the other way round. Conversely, when subjects were not provided with a virtual body, the attention asymmetry in depth favoured far stimuli. As reported by Andersen [1], the resource peak was further from the subject’s body, leading to higher costs to switch from far to near positions. Andersen [1] proposed that the direction of the asymmetry he found was not congruent with the one reported by Downing and Pinker [8] and by Gawryszewski et al. [9] because of differences in the experimental procedures. Later, Andersen and Kramer [2] pointed out that showing stimuli at different depth planes through a prism stereoscope may have led to a systematic overestimation of the size of the far distractors, and, in turn, to a greater interference due to far distractors than to near ones. They proposed that, when viewing the stereo images through such a device, the optical axis of the eyes are aligned in parallel and the convergence angle is zero. Under this viewing condition, small variations in apparent distance would have a relatively large effect on the perceived size of the stimuli. Although this explanation may be applicable to the Andersen [1] study, our results suggest a different interpretation of the asymmetry he reported. In our experiment, the viewing conditions were the same in either group: subjects viewed the virtual world through a direct projection of it in their eyes. The distance between the HMD and the subject’s eyes was about 3 cm, actually much less than the one employed by Andersen [1]. That is to say, in both the 322 F. Maringelli et al. / Cognitive Brain Research 10 (2001) 317 – 322 body present and the body absent conditions the subject’s eyes were aligned in parallel and information about distance had to be derived from binocular disparity only. It is worth noting that, in both conditions, subjects were generally faster in responding to far targets than to near targets. That may have been due to the perceptual factor identified by Andersen. However, we showed that both attentional asymmetries manifested themselves, depending on the presence / absence of the virtual body. Our results also support and extend several neuropsychological observations in favour of a dissociation between representations of the visual world. It has been proposed that different areas of the brain are involved in processing the near, egocentric space and the far, allocentric space. Rizzolatti et al. [17], for example, showed a double dissociation between impairment of near and far visual information processing in monkeys, due respectively to posterior or anterior brain damages. Observations on human neglect (e.g. Refs. [3,6,7,10]) favour a functional and anatomical specialisation of the human brain in processing either near (personal and peripersonal) or far (extrapersonal) stimuli. We reported similar asymmetry in healthy subjects: the distribution of attentional resources is related to the ‘suggested’ co-ordinates frame. In addition, our results seem to favour a parallel organisation of two largely independent attentional systems. Information falling within a near, reachable space seems to be represented in terms of egocentric co-ordinates and selected by an attentional system devoted to the analysis of this portion of the visual space. Conversely, when information falls outside the reachable space, its representation appears to be largely allocentric and selection is carried on by a second attentional system, working mainly on the far portion of the visual space. Finally, it is worth noting that our results are consistent with a parallel organisation of two largely independent attentional systems, although we did not provide direct evidence about the neural organisation. That means that the alternative hypothesis of a single neural system with attention being placed at varying depth planes depending on the display cannot be ruled out. However, because we showed a different distribution of attentional resources in the two experimental conditions, it would seem likely that two mechanisms rather than a single mechanism are involved. Information falling within a near, reachable space seems to be represented in terms of egocentric coordinates and selected by an attentional system devoted to the analysis of this portion of visual space. Conversely, when information falls outside the reachable space, its representation appears to be largely allocentric and selection is carried on by a second attentional system, working mainly on the far portion of visual space. Acknowledgements C.U. was supported by grants from CNR and MURST. References [1] G.J. Andersen, Focused attention in three-dimensional space, Percept. Psychophys. 47 (1990) 112–120. [2] G.J. Andersen, A.F. Kramer, Limits of focused attention in threedimensional space, Percept. Psychophys. 53 (1993) 658–667. [3] E. Bisiach, D. Perani, G. Vallar, A. Berti, Unilateral neglect: personal and extrapersonal, Neuropsychologia 24 (1986) 759–767. [4] E. Bisiach, The spatial features of unilateral neglect, in: P. Their, H.-O. Karnath (Eds.), Parietal Lobe Contributions to Orientation in 3D Space, Springer, Berlin / Heidelberg, 1997, pp. 465–495. [5] W.R. Brain, Visual disorientation with special reference to lesions of the right cerebral hemisphere, Brain 64 (1941) 244–272. [6] A. Cowey, M. Small, S. Ellis, Left visuo-spatial neglect can be worse in far than in near space, Neuropsychologia 32 (1994) 275–288. [7] D. Denny-Brown, Discussion fourth session, A, in: V.B. Mountcastle (Ed.), Interhemispheric Relations and Cerebral Dominance, Johns Hopkins Press, Baltimore, MD, 1962, pp. 244–252. [8] C. Downing, S. Pinker, The spatial structure of visual attention, in: M. Posner, O. Martin (Eds.), Attention and Performance, Vol. XI, Erlbaum, Hillsdale, NJ, 1985, pp. 171–187. ` Move[9] L.D.G. Gawryszewski, L. Riggio, G. Rizzolatti, C. Umilta, ments of attention in the three dimensions and the meaning of neutral cues, Neuropsychologia 25 (1987) 19–29. [10] P.W. Halligan, J.C. Marshall, Left neglect for near but not for far space in man, Nature 350 (1991) 498–500. [11] J. Jonides, Towards a model of the mind’s eye’s movement, Can. J. Psychol. 34 (1980) 103–112. [12] G. Keppel, Design and Analysis, Prentice Hall, Englewood Cliffs, NJ, 1991. [13] A.D. Milner, Neglect, extinction, and the cortical streams of visual processing, in: P. Their, H.-O. Karnath (Eds.), Parietal Lobe Contributions to Orientation in 3D Space, Springer, Berlin / Heidelberg, 1997, pp. 3–21. [14] A. Paterson, O.L. Zangwill, Disorders of visual space perception associated with lesions of the right cerebral hemisphere, Brain 67 (1944) 331–358. [15] M. Pavlovskaya, I. Glass, N. Soroker, B. Blum, Z. Groswasser, Coordinate frame for pattern recognition in unilateral spatial neglect, J. Cogn. Neurosci. 9 (1997) 824–834. [16] M. Posner, Orienting of attention, Q. J. Exp. Psychol. 32 (1980) 3–25. [17] G. Rizzolatti, M. Matelli, G. Pavesi, Deficit in attention and movement following the removal of postarcuate (area 6) and prearcuate (area 8) cortex in macaque monkeys, Brain 106 (1983) 655–673. [18] G. Rizzolatti, M. Gentilucci, M. Matelli, Selective spatial attention: one center, one circuit, or many circuits, in: M. Posner, O. Martin (Eds.), Attention and Performance, Vol. XI, Erlbaum, Hillsdale, NJ, 1985, pp. 251–265.
© Copyright 2026 Paperzz