• No results found

2. Chapter 2 Subjective contours

2.1. Introduction

The slalom illusion has robustly produced a potent visual effect in the case of strongly- defined contours (Cesàro & Agostini, 1998). In section 1.7.1 of the general introduction, the broader context was provided for Cesàro and Agostini's proposed explanation of the slalom illusion, according to which the local distortions in motion direction signals at the points of intersection between the horizontal dot trajectory and the tilted lines integrate into a global percept of a sinusoidal trajectory. In particular, these local distortions could be rooted in the V1 mechanisms proposed by Blakemore et al. (1970), who reported that neurons encoding edge orientation and motion direction signals in primary visual cortex are biased towards perpendicular angles of intersection.

However, not all contours that are subjectively perceived are also locally present in the image. As discussed in section 1.4.2 of the general introduction, illusory contours such as those induced by the Kanizsa triangle (Kanizsa, 1976) preserve many of the

properties of real contours, both phenomenologically and functionally. In the neuronal responses of V1 and V2, illusory (or subjective) contours are eventually encoded similarly to real contours, following feedback signals from higher-level visual areas (von der Heydt et al., 1984; Lee & Nguyen, 2001). This brings forward an interesting question: does the slalom illusion still occur if the inducing tilted lines are only subjectively present, similar to the contours of the Kanizsa triangle? That is, when the physical intersections between the oriented lines and the dot trajectory – proposed by Cesàro and Agostini (1998) to give rise to the local distortions in motion direction underlying the slalom illusion – are therefore absent?

43

The general question of whether illusory contours can induce the perceptual phenomena apparent with real contours has previously been addressed in a number of papers on static geometric illusions. The Poggendorff illusion, in particular, has in a number of studies been adapted to a version where the real contours of the inducing rectangle are replaced with Kanizsa-like illusory ones (Figure 2.1). The misalignment effect typical of the illusion was robustly replicated in a number of studies (Gregory, 1972; Meyer & Garges, 1979; Westheimer & Wehrhahn, 1997; Tibber et al., 2008).

Figure 2.1. Illustrations of the Poggendorff effect in the original first-order contours version (A), and in the illusory lines adaptation (B).

One study (Day, Dickinson, & Jory, 1977) did not confirm the effectiveness of illusory contours as inducers to the Poggendorff illusion. Their data showed no difference between illusory contour conditions and a control condition consisting of only an oblique line (without the rectangle). However, the authors themselves point out that the oblique line of their display interfered with the Kanizsa elements inducing the illusory contours and might therefore have negated the expected effect. Given the robust effect

44

subsequently found in the more carefully controlled study of Meyer and Garges (1979), the importance of the negative finding of Day et al. (1977) should therefore not be overstated due to this confound.

It can therefore be concluded that Kanizsa-like subjective contours are capable of driving the Poggendorff effect. However, despite the repeated replication of this phenomenon, it has also been found that the magnitude of the misalignment bias is smaller in subjective-contour Poggendorff displays, when compared to the bias that can be observed using a classical display with real contours.Tibber et al. (2008) suggest that the attenuation in the Poggendorff effect when driven by subjective contours could be explained by the lower salience of the subjective contours. That is, that subjective contours are detected by a smaller subset of cortical neurons in the early visual areas than real contours and that, consequently, lateral inhibition between orientations columns in V1 is weaker than when the entire stimulus is defined by real contours. In the study of Westheimer & Wehrhahn, 1997), the authors additionally included

experimental conditions where the contrast of the display was reduced, both for the real- contour and the subjective-contour variants of the Poggendorff illusion. Although the subjective contours elicited a weaker misalignment bias than the real contours, the full size of the effect was in both cases reached at a very low level of stimulus contrast. The response strength of V1 neurons is however strongly dependent on stimulus contrast (Carandini, 2007), as are psychophysical tasks relying on low-level orientation discrimination mechanisms (Wehrhahn & Westheimer, 1990). This then suggests that the magnitude of the misalignment bias does not simply depend on the response strength of V1 neurons. Therefore, the weakened misalignment bias in subjective contour conditions similarly cannot simply be reduced to a weaker V1 response to subjective contours, as compared to real contours.

45

The literature also registers that the kinetic version of the Poggendorff illusion, where the oblique line is replaced by an oblique dot trajectory, produces the misalignment bias typically found in the static Poggendorff display (Fineman & Melingonis, 1977;

Wenderoth & Johnson, 1983). However, a recent search of published papers revealed that no combination of the two factors (kinetic instead of static and subjective/illusory instead of real contours) has been tested with the Poggendorff effect, or any other illusion of direction. It remains therefore an outstanding question whether subjective contours are capable of maintaining kinetic illusions of direction.

In the current study, the kinetic nature of the slalom illusion will therefore be combined with the illusory type of contours. In addition, the contrast of the real-contour tilted lines will be manipulated. If the mechanisms of the slalom illusion are similar to those of the Poggendorff illusion, it can be expected that the slalom illusion can be replicated using subjective-contour displays, but possibly at a decreased magnitude of the illusion. In the real-contour conditions, the contrast of the tilted lines should then also not affect the magnitude of the illusion.

This would suggest that the slalom illusion is not simply rooted in momentary local distortions of motion direction early in the visual processing stream, but that higher- level contour representations interact with the signals of motion direction. If, on the other hand, subjective contours fail to elicit the slalom illusion and contrast

manipulations strongly affect its magnitude when using real contours for the tilted lines, then it can be concluded that the slalom illusion is rooted in low-level interactions between the motion direction of the dot and the orientation of the tilted line edges – as originally suggested by Cesàro and Agostini (1998).

46