Orientation Perception of Occluded Objects Is Based on Perceptually Completed Objects

Abstract

The natural world is full of objects, and clearly, the human visual system is good at determining their location as well as their orientation. However, this process is not as simple as it seems. This is especially true when parts of objects are occluded, which results in objects being only partly represented at the retinal level. It seems even more complicated when objects are constructed at only a higher level of visual information processing, as is the case for subjective contour objects, such as Kanizsa figures (e.g., Gold, Murray, Bennett, & Sekuler, 2000).

In the study reported here, we wanted to find out whether the visual system uses the same strategy to determine the orientation of both “real” objects and “subjective” objects. In other words, is an object’s perceived orientation determined by information available at the retinal level or at a higher level of information processing: the object’s representation? How can this be measured?

The Oblique Effect

A classic study by Olson and Attneave (1970) illustrated how object axes of symmetry can influence behavior in a visual search task. Recently, we designed a parsimonious paradigm to show that features of an object, such as axes and sides, can predict the accuracy with which the orientation of the entire object is determined (Borra, Hooge, & Verstraten, 2007). It is interesting that these features can be luminance defined (e.g., sides of objects) but can be axes within the object itself as well—that is, not projected on the retina.

The visual system preferentially selects object features with a horizontal-vertical orientation (Fig. 1a). This follows directly from the oblique effect, which holds that the visual system is more sensitive to horizontal-vertical orientations than to oblique ones (e.g., Appelle, 1972). When perceiving the orientation of objects, the visual system acts like an ideal observer: always selecting object features, allowing for the highest precision in orientation perception.

Fig. 1.

Experimental paradigm and results of the study. An optimal strategy for selecting object features (such as sides or axes; a) allows for the highest orientation-discrimination precision. For the orientation of each square, features are selected that are closest to horizontal-vertical (highlighted in boldface). This explains the “trough” in thresholds at the 45° square orientation, because the square contains horizontal-vertical axes of symmetry and the visual system is more sensitive to horizontal-vertical information (the oblique effect). The experimental paradigm is shown in (b). The upper left panel shows a real square with sides of 6.5° of visual angle; the bottom left panel shows a Kanizsa square—an illusory, subjective square in which the sides are not defined by luminance contrast—with the same dimensions. The upper right panel shows local contours for the Kanizsa square; the lower right panel shows sides and axes of symmetry of the perceptually completed square. The graphs in (c) show orientation-discrimination thresholds as a function of the object’s orientation. Thresholds based on the orientation of retinal projections are shown in the left panel, and thresholds based on the orientation of perceptually completed object representation are shown in the right panel. The graphs in (d) show data from the experiment reported here: Mean orientation-discrimination thresholds for both the real square (left panel) and the Kanizsa square (right panel) are shown as a function of the square’s orientation. The error bars in (d) represent standard errors of the mean.

Now, what happens when the projection on the retina is oblique (e.g., 45°) but what is perceived is horizontal-vertical? Take, for example, a Kanizsa square (Fig. 1b). When a Kanizsa square has a 45° orientation, all parts of the sides that are retinally projected are 45° away from the horizontal-vertical axes. In contrast, the perceived square contains axes of symmetry that are horizontal-vertical.

Following from the oblique effect, if orientation discrimination for a 45° Kanizsa square depends on what is projected on the retina (i.e., 45° sides), then one would predict low orientation sensitivity, resulting in high discrimination thresholds. However, if orientation discrimination is instead dependent on what is perceived (i.e., a perceptually completed square containing horizontal-vertical axes of symmetry), one would predict high orientation sensitivity, resulting in low discrimination thresholds, similar to the pattern shown in Figure 1a. This is depicted in Figure 1c, which shows predictions based on the retinally projected orientations or on a perceptually completed representation.

In the current study, we used the oblique effect to compare how well observers judged the orientation of real and subjective objects.

Method

Participants

Four observers (age = 28–42) who were experienced but naive to the purpose of the study took part in Conditions 1 and 2.

Stimuli and procedure

In Condition 1, we presented observers with a square (luminance = ~25 cd/m⁻²) with sides measuring 6.5° of visual angle on a dark background (~0.2 cd/m⁻²), with all edges clearly visible (Fig. 1b, upper left panel). For Condition 2, the stimulus was only partly represented on the retina: the square was identical in size to the square presented in Condition 1, but it was presented equiluminant to the background (~0.2 cd/m⁻²) and positioned on top of randomly arranged circles (~25 cd/m⁻²; diameter varying from 0.5° to 2°; Fig. 1b, lower left panel).

On each trial, a reference stimulus was presented for 500 ms; this stimulus was randomly chosen from five orientations (0°, 22.5°, 45°, 67.5°, 90°). The reference stimulus was followed by a 750-ms circular mask (diameter = 20°, luminance ~25 cd/m⁻²). The mask was followed by a randomly rotated test stimulus, randomly displaced for both the x and y positions (where 3° ≤ displacement ≤ 5°). The observers’ task was to indicate whether the test stimulus was rotated clockwise (by pressing 6 on a numerical keypad) or counterclockwise (by pressing 5 on a numerical keypad) with respect to the reference stimulus. The amount of rotational difference between the test and reference was controlled by a Quest procedure for each reference orientation. A black cloth masked the edges of the display, and the room was dark.

Results and Discussion

Orientation-discrimination thresholds for the real square and the Kanizsa square are shown in Figure 1d. Each data point represents 40 trials for each observer. The patterns of discrimination thresholds for the two stimuli are basically identical: lower thresholds for reference orientations of 0°, 45°, and 90° and higher thresholds for 22.5° and 67.5°.

The lower thresholds for the 45° real square and 45° Kanizsa square orientations indicate that the visual system selects horizontal-vertical object features to determine object orientation. It is interesting to note that horizontal-vertical object features are not present in the retinal projection of the modified Kanizsa square, as all contours are oriented 45° away from the horizontal-vertical. The perceptually completed square, however, does contain horizontal-vertical features: the axes of symmetry. So, the thresholds that we obtained provide clear evidence that the visual system determines object orientation based on features of a perceptually completed object representation.

Current physiological measures seem to indicate that the oblique effect possibly originates at the level of visual area V1. Furmanski and Engel (2000) reported populations of neurons in human visual cortex that display asymmetries in their preferred orientations, with more neurons tuned to horizontal-vertical orientations (see also Yacoub, Harel, & Ugurbil, 2008). This asymmetry might reflect the oblique effect, but no such asymmetry was found in higher extrastriate areas. This implies that our results, which are based on axes that arise from completed objects, can be explained only through feedback from higher areas. This is because the object representation supplying the axes is present only at levels beyond V1. This finding is consistent with results reported in the rapidly growing body of research on feedback mechanisms in the human brain (e.g., Juan & Walsh, 2003; Lee, Mumford, Romero, & Lamme, 1998; Zipser, Lamme, & Schiller, 1996). This is, however, a topic for future research.

In sum, using a classic visual effect, we showed how the visual system reconstructs information and uses it to determine the orientation of occluded objects in the absence of direct retinal stimulation.

Footnotes

Declaration of Conflicting Interests

The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.

Funding

This research was supported by a PIONIER Grant of the Netherlands Organisation for Scientific Research (NWO-MagW).

References

Appelle

(1972). Perception and discrimination as a function of stimulus orientation: The “oblique effect” in man and animals. Psychological Bulletin, 78, 266–278.

Borra

Hooge

I. T. C.

Verstraten

F. A. J.

(2007). The use of optimal object information in fronto-parallel orientation discrimination. Vision Research, 47, 3307–3314.

Furmanski

C. S.

Engel

S. A.

(2000). An oblique effect in human primary visual cortex. Nature Neuroscience, 3, 535–536.

Gold

J. M.

Murray

R. F.

Bennett

P. J.

Sekuler

A. B.

(2000). Deriving behavioural receptive fields for visually completed contours. Current Biology, 10, 663–666.

Juan

C. H.

Walsh

(2003). Feedback to V1: A reverse hierarchy in vision. Experimental Brain Research, 150, 259–263.

Lee

T. S.

Mumford

D. B.

Romero

Lamme

V. A. F.

(1998). The role of the primary visual cortex in higher level vision. Vision Research, 38, 2429–2454.

Olson

R. K.

Attneave

(1970). What variables produce similarity grouping? The American Journal of Psychology, 83, 1–21.

Yacoub

Harel

Ugurbil

(2008). High-field fMRI reveals orientation columns in humans. Proceedings of the National Academy of Sciences, USA, 105, 10607–10612.

Zipser

Lamme

V. A. F.

Schiller

P. H.

(1996). Contextual modulation in primary visual cortex. Journal of Neuroscience, 16, 7376–7389.