Asymmetries in perception of 3D orientation

Abstract

Visual scene interpretation depends on assumptions based on the statistical regularities of the world. People have some preference for seeing ambiguously oriented objects (Necker cubes) as if tilted down or viewed from above. This bias is a near certainty in the first instant (approximately 1 s) of viewing and declines over the course of many seconds. In addition, we found that there is modulation of perceived orientation that varies with position--for example objects on the left are more likely to be interpreted as viewed from the right. Therefore there is both a viewed-from-above prior and a scene position-dependent modulation of perceived 3-D orientation. These results are consistent with the idea that ambiguously oriented objects are initially assigned an orientation consistent with our experience of an asymmetric world in which objects most probably sit on surfaces below eye level.

Figure 1. A Necker cube and its two interpretations.

A. A Necker cube is a 2D projection of a 3D wireframe cube (left). People experience stochastic switching between the two interpretations of orientation shown in the solid cubes. B. An array of ambiguous cubes. Observers are very likely to see the cubes as if viewed from above. Rotating the display by 90 degrees eliminates the orientation-based grouping. This is based on one frame of an animation from the first experiment with rotating cubes in which one cube after another is highlighted (green). In the dynamic display (Expt. 1), rotating each cube about a similarly oriented axis links the orientational ambiguity to the rotational ambiguity. For example, a particular cube can appear to be tilted down and rotating to the left, or, tilted up and rotating to the right. Orientation and rotation were randomly coupled throughout the array. If all the cubes were perceived to spin in the same direction, half would appear to be tilted up and half tilted down. Alternatively, if all the cubes were perceived to have the same orientation, half would appear to spin leftward and half rightward. (An animated version of the cube array can be seen at: http://www.vsrc.uab.edu/adobbins.htm).

Figure 2. Initial perception bias in a single cube and an array.

In a five by five array of cubes, one cube at a time is highlighted (Fig. 1B). In other trials only one cube is present and it appears successively at different positions. A. Vertical axis (ambiguous above/below) and horizontal axis (ambiguous left/right). For both the highlighted cube in the array and the single cube there is a profound viewed-from-above bias in the initial percept. However there is no bias in the horizontal axis trials. Colored crosses represent the mean proportion of the individual observers (N = 6). Observe the tighter clustering in the array condition (left column). B. Persistence of interpretation from one presentation to the next. The left half of the plot shows orientation/viewpoint persistence and the right half plot shows rotational persistence. Only the tilted up/down interpretation has high persistence. (Mean ± s.e.m.).

Figure 3. Time evolution of perception for a single rotating cube.

A. Fraction of time a single rotating cube is seen from above in vertical axis trials (left bar) and from the right in horizontal axis trials (right bar). There is a substantial tilted down/viewed from above bias, but no corresponding bias for tilted left/viewed from right cubes (N = 6). B. Peristimulus time histograms for the vertical axis trials for two observers: one of the most biased observers (top) and the least biased observer (bottom). The strongly biased observer shows a gradual decline in the probability of seeing the cube from above over the course of the trial. The neutral observer is as strongly biased as the strongly biased observer over the first second or so of the trial but not thereafter. Pre-Choice (red) represents the time before an initial choice is reported.

Figure 4. Variation of viewpoint bias with array position.

The top two curves show that there is weak modulation of the probability of viewed-from-above with elevation in the array. In contrast, the lower two curves show a very strong modulation of perceived left/right orientation with left/right position – a cube in the leftmost column is much more likely to be seen as if tilted left/viewed from the right. (Mean ± s.e.m.) Least squares linear fits for the horizontal axis conditions: array (y = −1.97 θ+51.37, R² = 0.9672, p = 0.0025) and single cube condition (y = −1.20 θ+47.08, R² = 0.9317, p = 0.0077), where θ is in degrees and y is in percent. Fitting the mean of the paired differences for each observer in the horizontal axis conditions yielded: y = −0.57 θ+3.97, R² = 0.9678, p = 0.0025.