Neural Correlates of Implied Motion

1MS: 2002-12021B

Neural Correlates of Implied Motion

Bart Krekelberg, Sabine Dannenberg, Klaus-Peter Hoffmann, Frank Bremmer, John Ross

MS: 2002-12021B

Supplementary Information

A. Local Motion Signals.

Introduction. Glass patterns do not contain globally coherent motion vectors1. It is possible, however, that accidental local correlations biased the subjects percept of motion enough to induce a global percept of motion. We investigated whether this was the case by presenting the same sequence of rotational Glass patterns twice, first in forward order, and then in reverse order. If the percept of coherent global rotation were based on accidental local correlations in the randomly chosen sequences, the reversal of the order of presentation should change the percept from clockwise to counter clockwise and vice-versa.

Method. A set of 20 rotational Glass patterns, each containing 200 pairs of points, was newly constructed for each experimental session. Paired dots (white on a gray background) were separated by 5 degrees of rotation. Luminance and dot size was the same as in the experiments reported in the paper. On each trial 5 patterns were selected at random from the available set of 20 and presented each once only in a random order at a rate of 24 Hz. In the Reverse condition (100 trials) the same 5 patterns were shown again in reverse order. On each presentation (initial and reverse) subjects judged whether the direction of rotational motion they had seen was clockwise (CW) or counter clockwise (CCW). The whole procedure was repeated with non-Glass patterns made up of single, unpaired points.

Results.Averaged over all trials, the two subjects reported 52% and 46% clockwise decisions. This confirms that on average the direction of motion was equally likely to be clockwise or counter clockwise, hence the direction of motion is truly ambiguous on average. If the subjects based their decisions on accidental local motion cues, a CW decision should be followed by a CCW decision in the following trial (in which the order of presentation and hence the local motion cues were reversed). Hence, the percentage of successive same-direction decisions (CCW/CCW or CW/CW) should be near zero. This was not the case. Two naive subjects reported 76% and 72% same-direction decisions.

Discussion. This experiment demonstrates that observers did not base their percept of motion direction on accidental local non-randomness. Taken together with the known global randomness1, this convincingly shows that the motion implied by the form of the Glass patterns drives subjects’ motion percepts. For completeness, these results also indicate, however, that subjects’ successive responses were not entirely independent. If they were, the same-direction percentages would have been ~50%. That this is not the case reflects the fact that these subjects typically clustered their decisions: a few trials CW, followed by some CCW. That these decisions were based on truly perceived motion can be inferred from the perceptual rating experiments discussed by Ross et al1. Moreover, the clustering itself is not peculiar to sequences of Glass patterns. When forced to report a CW or CCW motion direction for random noise, the same clustering was observed.

B. Ambiguous motion directions.

Because the direction of motion in Glass patterns is inherently ambiguous, one might expect that the cells’ responses alternated between the response to the preferred direction and the anti-preferred direction. Such an alternation may take place on a trial-by-trial basis, but also in a single trial. We investigated this for rotational Glass patterns by testing whether a cell’s Glass response was statistically indistinguishable from the mean of the counter- and clockwise rotations (the ‘motion-mean’). To exclude cells whose responses were simply noisy and therefore indistinguishable from the motion-mean, we excluded cells whose response to the noise stimulus was indistinguishable from the motion-mean. Of the remaining 33 cells, 25 cells responded to a Glass pattern with a rate that was indistinguishable from the motion-mean. On average, 3.6 Glass patterns per cell showed this behaviour. The distribution of these ‘switching’ Glass shifts was significantly peaked in central range of Glass shifts. This suggests that these cells responded to the two ambiguous implied directions of motion. Note that such an alternation of responses increases the variance of trial and time averages. This increased variance made it less likely that the null-hypothesis of ‘no-tuning’ could be rejected. Our estimate of the fraction of cells that were significantly tuned to the Glass-shift may therefore in fact be an underestimate.

C. Spiral Space.

Studies of judged motion direction were also conducted with spiral glass patterns. The apparent pitch of spiral motion was again a compromise between the actual pitch of motion and Glass pattern pitch. The influence of implied motion was greater than for linear Glass patterns in linear motion. Real and implied motion had equal influence when pattern signal strength was 25% and motion signal strength was 50%, the reverse of the values for linear patterns. Once again, implied motion exerted a strong attractive influence only when conflict between the implied motion in the pair orientation and motion direction was small. Further evidence of the influence of Glass patterns on motion, both linear and spiral, is provided by Burr and Ross.2 They show that varying the orientation of Glass pattern dipoles randomly from trial to trial has the same effect as motion noise: it raises thresholds for detecting deviations in the direction of global motion, provided differences between local motion direction and pair orientation are small.

References

1.Ross, J., Badcock, D. R. & Hayes, A. Coherent global motion in the absence of coherent velocity signals. Curr Biol10, 679-82. (2000).

2.Burr, D. C. & Ross, J. Direct evidence that "Speedlines" influence motion mechanisms. J Neurosci22 (2002).