P280 - Cognitive and sensorimotor distance coding
Effect of structuring the workspace on cognitive and sensorimotor
distance estimation : no dissociation between perception and action
Yann COELLO & Orianne IWANOW
Unité de Recherche sur l'Evolution des Comportements et l'Apprentissage
UPRES EA 1059
Université Charles de Gaulle. Lille, France.
Running title: Cognitive and sensorimotor distance coding
Key words: Vision, perception, action, context information, distance coding, reachability, pointing movement.
Mailing Address: Pr. Yann COELLO
Laboratoire URECA
Université Charles de Gaulle - Lille
BP 60149
F.59653 Villeneuve d'Ascq cedex
Tel: +33.3.20.41.64.46 Fax: +33.3.20.41.63.24
Email:
Abstract
Independent processing of visual information for perception and action is supported by studies about visual illusions, which showed that context information influences overt judgement but not reaching attempts. It was however objected that performances are not directly comparable as they generally focus on different properties of the visual input. The aim of the present study was to quantify the influence of context information (in the form of a textured background) on the cognitive and sensorimotor processing of egocentric distance. We found that the subjective area comprising reachable objects (cognitive task) decreased whereas the amplitude of reaching movement (sensorimotor task) increased in presence of the textured background when viewing binocularly or monocularly. Directional motor performance was not affected by the experimental conditions but there was a tendency for the kinematic parameters to mimic trajectory variations. The similar but opposite effect of the textured background in the cognitive and sensorimotor task suggested that, in both tasks, the visual targets were perceived closer in sparse environment. A common explanation for the opposite effect was confirmed by the percentage of background influence, which highly correlated in the two tasks. We conclude that visual processing for perception and action cannot be dissociated from context influence, since it does not differ when the tasks entail the processing of similar spatial characteristics.
Introduction
In order to modify the disposition of our sense organs and motor effectors appropriately in relation to the environment, the central nervous system must select just those inputs that are currently relevant, while simultaneously suppressing irrelevant inputs (Allport 1989, Tipper et al. 1992, Desimone & Duncan 1995, Mattingley & Driver 1997). A fascinating debate has been in the scientific foreground the last decade about the way such selection occurs, in particular when the situation focuses on the perceptual or the motor aspect of the behaviour. In this respect, neurophysiological, neuropsychological and psychophysical data have provided converging evidences for distinct processing of spatial information within the visual system depending on whether the task underlies object recognition, description or visually guided reaching movement (Schneider 1969, Trevarthen 1968, Ungerleider & Miskin 1982, Milner & Goodale 1995, Pagano & Bingham 1998). Determining psychophysical arguments for such dissociation came with visual illusions, where metric judgement of the central part of a visual pattern (e.g. the Müller-Lyer, Ponzo or Titchner illusion) or its location (e.g. the induced Roelofs effect) was influenced by context information when responding verbally (cognitive task) but not when performing manual reaching (sensorimotor task, Rossetti 1998). Considering the latter visual illusion more precisely, healthy adults were requested to estimate the position of a luminous target appearing inside a surrounding frame being centred according to the egocentric straight-ahead direction or with a lateral offset of +/- 5 degrees in the fronto-parallel plane (Bridgeman 1991). When the frame was presented with a lateral offset, target location was misperceived in the opposite direction when estimated through verbal responses (probing the cognitive processing), but not when estimated with a manual reaching (probing the sensorimotor processing). This specific influence of context information was thought to result from the fact that the visual system selects spatial characteristics in different ways depending on the output (Bridgeman 1991, 2000). Dealing with relative positions, visual processing for cognitive purpose aims at elaborating an explicit qualitative encoding of the visual space, which includes information relating to the whole visual scene, even when the influence of contextual elements leads to localisation errors. This encoding serves as the basis for categorisation and verbal report (for instance, determining if one object is circular or above another one, Kosslyn 1994). Conversely, dealing with absolute positions, visual processing for sensorimotor propose aims at elaborating an implicit quantitative (metric) encoding of the visual space that is insensitive to context information and serves for the guiding of goal-directed behaviour. The dissociation observed with the induced Roelofs effect was taken as evidence for distinct processing of visual information depending on the cognitive or sensorimotor purpose of the task, since only the verbal account of target location was sensitive to the perturbing contextual frame (Paillard 1987, Bridgeman 1991, Jeannerod & Rossetti 1993, Bridgeman 2000). The observation that the visual system is anatomically organised in two main streams projecting from the visual cortex (V1) to the inferotemporal cortex (the ventral stream) and to the posterior parietal cortex (the dorsal stream) has naturally served as a general framework that supported well the distinct use of spatial information depending on the constraints of the task (for a review: Milner & Goodale 1995, Rossetti 1998).
However, recent psychophysical data cast some doubt on the validity of this widespread view by showing that motor performance can be severely influenced by visual context when various spatial dimensions can be differentiated. In particular, when pointing to a visual target in absence of dynamic visual feedback, distance performance was found to be much more accurate when the target was presented on a textured rather than a non-perceptible background (Foley & Held 1972, Conti & Beaubaton 1980, Treisilian et al. 1999, Coello et al. 2000) within a large rather than a restricted visual field (Bingham 1993, Coello & Grealy 1997). The location of context information in relation to the self and target played also a crucial role in determining reaching accuracy, with elements placed in the space through which the reach occurs conferring the most benefit (Coello 2002, Grealy et al. 2003). We recently reported that the unexpected provision of a textured background in the action space has an instantaneous concomitant effect on movement amplitude and early kinematic characteristics such as peak velocity (Magne & Coello 2002), suggesting that improvement of the motor performance was mainly the consequence of a more accurate visual system. Interestingly, the conspicuous effect of structuring the visual space on distance accuracy usually left direction performance unaffected (Magne & Coello 2002).
This parametric framework has proved to have great implication in the debate relating to the perception-action dichotomy (Milner & Goodale 1995). Using the induced Roelofs effect (Bridgeman 1991) but in slightly different experimental conditions, we found that the presence of the off-centre frame can influence motor responses in a similar way as perceptual reports depending on the spatial dimension tested (Coello et al. 2003). When the frame was displaced along the fronto-parallel axis, the visual target was perceived shifted in the opposite direction than the off-centre frame but the manual capture of the target was not affected by the illusion, in agreement with the original studies (Bridgeman, 1991, 1997, 2000). However contrasting with the previous finding, the induced Roelofs effect interfered with perceptual and motor responses in identical ways when the frame was displaced along the sagittal axis. Our interpretation for the dissociated effect on motor performance depending on the dimension tested was that target location for action is not always immune from contextual influence, which really depends on the type of visual information that needs to be processed according to the task constraints. Because distance coding is strongly influenced by the visual cues available on a large part of the retina (Magne & Coello 2002), especially those lying within the effector-to-target gap that delimits action space (Coello 2002), it was not surprising to observe that context information can influence motor production when performed in the near-far rather than the right-left dimension (Coello et al. 2003). The fact that similar pattern of results was subsequently obtained with the patient (I.G.) suffering from optic ataxia due to bilateral damage of posterior parietal cortex indicated that the integrity of the dorsal pathway, in particular the superior parietal lobe, is not a necessary condition to obtain an influence of visual context on motor acts (Coello & Rossetti 2004).
Though these findings directly challenge a radical separation between perception and action based on the distinct influence of context information, one must considered them cautiously and cannot conclude straight away that the visual system can be influenced by context information in a similar way for perception and action. Indeed, just like in many previous studies dealing with visual illusions (e.g. see Franz et al. 2000), the perceptual and the motor tasks in the adapted version of the induced Roelofs effect required the processing of different kinds of spatial information. In the perceptual task, participants had to estimate the relative position of the visual target with regard to another one presented 500ms ahead. The first target was always visible within a centre frame (duration 400ms), whereas the second target was presented within an off-centre frame (duration 400ms) or without any frame. By contrast, the motor performance consisted of a pointing movement towards the second target with no need to explicitly process the previous one (see Coello et al., 2003 for a detailed description). Estimating the egocentric absolute location of the last visible target was thus enough to achieve the task accurately. Consequently, one cannot firmly conclude from these data that context information can influence in a similar manner the perceptual and motor responses when the task emphasised the processing of distance parameter.
To unravel this issue, it was necessary to compare the cognitive and sensorimotor visual system in a situation requiring the processing of same kind of spatial information, as well as to quantify the respective influence of context information. To fulfil this requirement, the influence of visual context (in the form of a textured background) was investigated in two tasks implying the processing of the absolute egocentric distance of visual target. In the cognitive task, participants had to provide an overt perceptual judgement about whether the visual target was reachable with the hand or not, but with no actual movement being allowed. Several studies have shown in the past that people are quite accurate in estimating perceptually the critical limit of what is reachable. The general agreement is that estimation of one's own reaching capabilities slightly overestimates actual arm length by about 10% (Carello et al. 1989, Bootsma et al. 1992, Rochat & Wraga 1997). In the sensorimotor task, participants had to perform pointing movements in a dynamic open loop condition towards a visual target located at various distances along the sagittal axis. Assuming that the cognitive and the sensorimotor visual systems are similarly influenced by context information, one expected the effect of the textured background to be of same magnitude on verbal and motor responses, but opposite in direction. Indeed, taking for granted that reaching movement are less hypometric in presence than in absence of a textured background because visual objects appears nearer in sparsely structured environment (Foley 1980, Watt et al. 2000, Coello et al. 2003), the farthest target judged as reachable in darkness should be judged as not reachable in the presence of a textured background.
Methods
Participants
The present study involved 8 normal right-handed participants (3 males and 5 females). Their ages ranged between 16 and 35 years and they were self-declared volunteers to take part in an experiment relating to visual perception and visuomotor control. They all had normal or corrected-to-normal vision (the right eye was systematically the dominant eye) and were naive as to the purpose of the study.
Apparatus
The experimental device consisted of a rectangular box (60cm high, 100cm wide and 70cm deep) with one side left open (see Figure 1). The inside of the box was divided horizontally by an upward-facing reflecting mirror. With the head resting on the upper part of the box in front of the open side, only the top half of the box was visible to the participant who was anyway able to move the right arm into the bottom half when requested to point to a visual target. A computer monitor (20" Trinitron by Philips) was placed upside-down on the top surface of the apparatus, so that the image generated by the computer was reflected in the mirror. Due to optical geometry, the image of the computer screen projected on the bottom surface of the box, i.e. the workspace. The mirror was positioned so that the hand at starting position was visible, but was concealed by the mirror following movement onset. Several visual targets (green dots of 8mm diameter) were visible along the sagittal plane, each of them being presented alone (background: 0cd/m2, measured with the optical photometer by Cambridge Research System) or on a textured background made with grey dots of 5mm (4cd/m2) randomly positioned over the whole workspace (30cm x 39cm). To determine the random position of the texture dots, the following procedure was used. Inter-dots distance was initially fixed (one dot every 15mm in the X and Y direction), but before displaying the visual information, a random coefficient was assigned by the computer to each of the targets so that their (x,y) position corresponded to a random value comprised between ± 0% and ±50% of the fixed inter-dots distance. Dots could not overlap and their position was recomputed at the beginning of each trial. Targets' number and location varied as a function of the experimental condition, as described below.
Two targets displays were used according to whether the cognitive task or the sensorimotor task was considered. In the cognitive task, 21 visual targets were selected along the sagittal axis at distances corresponding to the individual maximum reachable distance with the hand (target N° 0 in Figure 1) and at ± 5mm, 10mm, 15mm, 20mm, 25mm, 30mm, 35mm, 40mm, 45mm, 50mm from that position (target N° +10 to target N°-10). The maximum reachable distance (critical boundary) was initially estimated within the experimental box for each participant by measuring the distance that could be reached with the right arm fully stretched out. In the pointing task, only three targets were used along the sagittal axis. They were respectively positioned 45mm, 75mm and 105mm nearer than the maximum reachable distance (target number -9, -15 and –21 see Figure 1). These distances were chosen so that they were easily reachable but accounted for significant variability along the sagittal axis. In order to minimise the participant’s ability to use visual cues in the form of reflections from the upper half or the side of the box, the internal surfaces of the box were smooth and painted matt black. No visual information from the external environment was available during the entire experimental session.