Action-specific extrapolation of target motion in human visual system

Abstract

Neuropsychological studies have indicated two distinct visual pathways in our brain, one dedicated to conscious perception and one to visuomotor control. Some psychophysical results support this idea with normal subjects, but they are still controversial. This study provides new psychophysical evidence for the dissociation by showing action-specific extrapolation of the visual target trajectory. When a moving target disappears, the perceived final position is liable to be shifted forward (representational momentum). In experiment 1, larger and more robust forward shifts were found when the position was directly touched without seeing the screen (open-loop pointing) than when the position was judged perceptually. The most striking dissociation was that fixation did not affect the forward shift in open-loop pointing while it almost abolished the shifts in perceptual judgements. In experiment 2, this action-specific result was found to disappear after a response delay of 4000 ms. Experiments 3 and 4 confirmed that the results were not affected by the external reference frames. The specific forward shifts found in open-loop pointing suggest that the visuomotor system compensates for the neural delays by extrapolating the target motion. The results, together with earlier findings, lead to a psychophysical double dissociation of the two visual pathways.

Introduction

Neuropsychological studies on brain damage patients have indicated that the human brain has distinct visual pathways for action control and for detailed visual perception, a conclusion which is also supported by monkey neurophysiology (Milner, 1999; Milner & Goodale, 1995). For example, a patient with ‘visual form agnosia’ was able to post a plaque into a slot without being able to report its orientation. Together with the knowledge that damage in the parietal cortex can lead to visuomotor disorders (Bálint syndrome), there seems to be a double dissociation in terms of both function and physical pathways. Psychophysical studies on intact observers, however, have not shown such a clear picture. One line of evidence for separate functional pathways comes from results where the grasping action resists visual geometrical illusions such as the size contrast effect in Titchner–Ebbinghaus circles (Aglioti, DeSouza, & Goodale, 1995). However, there are possible artefacts (Franz, 2001), and dissociation in the results might not be evidence of two pathways (Bruno, 2001; Franz, Bulthoff, & Fahle, 2003). Given that a dissociation between perception and action has been reported in many other cases (e.g. Bridgeman et al., 1997, Burr et al., 2001, Dyde and Milner, 2002), the weakness seems to be that action is always more immune to illusion, a claim which has no a priori justification.

Yamagishi, Anderson, and Ashida (2001) provided the counterpart to complete the double dissociation. Using drifting Gabor patches, they demonstrated that the positional bias due to carrier motion (Anstis and Ramachandran, 1995, De Valois and De Valois, 1991) is more pronounced in an immediate open-loop reaching task than in perceptual judgement, showing that action can be more prone to illusion. The time taken for visual processing is a problem when determining the location of a moving target, and one possibility is that our visual system anticipates the target’s true location on the basis of the motion signal (Anstis & Ramachandran, 1995). If so, it is not surprising that a larger extrapolation should be found in a real-time reaching action in order to avoid missing the target. This idea, however, remains speculative because anticipation is not a real requirement in the case of motion-related positional biases because the envelope pattern does not move. We need to test the anticipatory effect for a moving target in a more straightforward situation.

While anticipatory mechanisms in the retina of rabbits and salamanders have been reported (Berry, Brivanlou, Jordan, & Meister, 1999), it is unclear if primates or humans, with their more complex visual systems, have a similar function. Discussion on anticipatory visual coding has been active in psychophysics since the rediscovery of the flash-lag illusion (Nijhawan, 1994), in which a continuously moving target is perceived ahead of a flashed stimulus when the two are physically aligned in space and time. Nijhawan (1994) argued that the visual system extrapolates the moving object’s instantaneous location to compensate for the processing delay. Although it is an intriguing idea, later results are more favourable to other factors as the cause of the flash-lag illusion, for example, processing delay for the flashed target (Whitney & Murakami, 1998) or time averaging of the position of the moving object (Eagleman & Sejnowski, 2000), and discussion on the flash-lag continues (see Nijhawan, 2002). Since anticipation is particularly important in direct action (Nijhawan, 1994, Nijhawan, 2002), it is notable that Nijhawan and Kirschfeld (2003) found a similar flash-lag phenomenon between visual and motor domains. They showed that perception of a flashed visual target lagged behind the position of an unseen rod that was manually controlled. While they argued for analogous delay-compensating mechanisms in visual and motor processing, they did not provide clear evidence either against the other theories of visual flash-lag or against the possibility that all compensation is accomplished in the motor system.

Here we address two main questions: firstly, whether the brain operates purely visual extrapolation to compensate for the neural delays, and secondly whether such extrapolation selectively affects direct visuomotor co-ordination. To answer these questions, this study investigated the judged final position of a linearly moving target after it suddenly disappeared. The perceived final position is apt to be shifted forward, an effect referred to as “representational momentum” (RM), because it seems as if the inner representation has momentum (Freyd, 1983). There are two major advantages to using this phenomenon. First, the target actually moves and the forward shift is more straightforwardly explained in terms of extrapolation than in the case of the Gabor stimuli. Second, since observers are not asked to point to the target itself, compensation for motor delays is not required. A substantial shift in the motion direction would therefore be a signature of visual extrapolation along the target path. The main interest is in the way in which forward shift occurs in perceptual and motor responses.

While RM has been found for several kinds of motion, a linear and smooth motion as tested by Hubbard and Bharucha (1988) is the most desirable for the current purpose. Two main and two control experiments revealed a distinct pattern of results for open-loop action, evidence of specific visual extrapolation for the immediate control of action. The results also provide another piece of evidence for the distinct visual processing for perception and action.