- P1-126 Time: -
Monkeys perceive reversed depth in anti-correlated random dot stereograms
- 小賀 智文 / Tomofumi Oga:1 中谷 光宏 / Mitsuhiro Nakatani:1 藤田 一郎 / Ichiro Fujita:1,2
- 1:大阪大院生命機能認知脳科学 / Grad Sch Frontier Biosci, Osaka Univ, Suita, Japan 2:脳情報通信融合研究センター / Center for Information and Neural Networks
Human observers perceive depth in the opposite direction to geometrically defined depth, i.e., near for uncrossed disparity and far for crossed disparity, in binocularly contrast-reversed random dot stereograms (anti-correlated random dot stereograms; aRDSs). This reversed depth perception suggests that disparity signals derived from computation of cross-correlation between left and right retinal images contribute directly, i.e., without being fully transformed to binocular match signals, to depth perception. It is unclear whether monkeys, an animal model for searching for the neuronal mechanism of stereoscopic depth perception, also perceive reversed depth in aRDSs. Previous studies reported negative answers to this question, but the tests were performed under conditions even human observers cannot perceive reversed depth. Here, we revisited the question by manipulating the relevant visual stimulus parameters.
We trained a Japanese monkey (Macaca fuscata) to discriminate depth of a center disk relative to its surround in concentric-bipartite RDSs. In the training sessions, both center disk and annulus were correlated random dot stereograms (cRDSs). The center disk had ±0.3°, while the annulus always had zero disparity. The stimulus refresh rate was set at 30 Hz. The monkey reported the perceived depth of the disk by making a saccade towards either one of two targets shown after stimulus offset. When the monkey's correct performance rate with cRDSs reached over 90%, we inserted trials of aRDSs every 5 consecutive success of cRDS trials. The monkey received liquid reward for all aRDS trials regardless of her answer. The correct rate for aRDSs was under 30% below the chance level of 50% (binomial test, p< 0.05). She tended to choose `far' for aRDSs with crossed disparity and `near' for aRDSs with uncrossed disparity, indicating that she perceived reversed depth. When we decreased the disparity magnitude to ±0.06 or ±0.12°, or when we lowered the stimulus refresh rate to 7.5 Hz, the monkey's choice was at chance level.
The monkey thus perceived reversed depth in aRDSs when stimuli were refreshed at a high temporal rate and had a large disparity magnitude. These perceptual properties were similar to those reported for human observers. In both humans and monkeys the contribution of binocular correlation computation to depth perception is greater for rapidly changing stimuli with coarse binocular disparity.
研究助成：Research funds : 15H01437