Purchase this article with an account.
Jan Drewes, Frederic Barthelemy, Guillaume S. Masson; Human ocular following and natural scene statistics. Journal of Vision 2008;8(6):383. doi: 10.1167/8.6.383.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
It is commonly assumed that the visual system is optimized to process naturalistic inputs for both low and high level processing. Here we search for an advantageous effect of natural scene statistics when estimating speed. Ocular following responses (OFRs) are reflexive eye movements known to reflect many properties of low-level motion processing. Using the scleral search coil technique, we recorded human OFRs to drifting sinusoidal gratings (1D) as well as narrow bandpass noise images (2D).
For sinusoidal gratings, it was previously shown that OFRs are best elicited with low spatial frequency stimuli ([[lt]]1cpd) moving at optimal speed (20–40°/s). We were able to confirm this for 2D noise stimuli as well. However, we found a systematic difference in the acceleration profiles: OFRs to 2D noise stimuli consistently showed longer latencies, yet stronger overall responses than the 1D gratings. When combining two or more spatial frequencies, we found a gain in response strength mostly in the higher spatial frequency range. Also, we found evidence that stimuli consisting of the normalized sum of several spatial frequencies can create stronger OFRs than the normalized sum of the OFRs to the individual frequencies. When combining spatial frequencies, the weighting (mix ratio) of the individual frequencies influences the response gain. The optimum weighting with multiple bandpass noises appeared to be similar to the spectral shape of natural scenes (1/f).
These results show a systematic difference between OFRs evoked by 1D gratings and 2D noises, and provide a first behavioral evidence that speed is best estimated by combining information across different channels, with weighting based on natural scene statistics.
This PDF is available to Subscribers Only