Inferring the future target trajectory from visual context: is visual background structure used for anticipatory...

Anticipatory pursuit is not exclusively based on the recent history of target motion ("temporal anticipation"): it can also use static visual cues ("non-temporal anticipation"). Large non-temporal anticipatory changes of the direction of smooth pursuit are observed when the future trajectory of the target can be inferred from a visual "path cue" (Ladda et al. Exp Brain Res 182:343-356, 2007). It is not known whether these anticipatory responses can be considered an example of volitional pursuit or whether a more automatic, fast mechanism exists that associates the visual shape of the path cue with a future change of target motion. We therefore compared anticipatory direction changes induced by path cues with those induced by more symbolic visual cues. Further, we measured the processing time of path cues by keeping them ambiguous until 300 ms before the target entered the curve. The cues became unambiguous either when a missing part of the path cue suddenly appeared or when a misleading (invalid) path cue disappeared. Five main results suggest that the non-temporal smooth pursuit anticipation induced by path cues is due to a particular visual processing of the cue, rather than to a general volitional smooth pursuit component: (1) A curved static band providing detailed spatial information about the target's trajectory was much more efficient than more symbolic visual cues. (2) The latency for processing the path cue was less than 200 ms. (3) The effect of the path cue did not depend on a visual transient in the retinal periphery. (4) It critically depended on the spatial relation between path cue and target trajectory. (5) The anticipatory response decreased, but was still highly significant when the path cue was non-informative with respect to the future target trajectory. These findings indicate that anticipatory modifications of ongoing pursuit do not rely exclusively on the processing of motion signals, but can directly interact with low-level processes analyzing background structures.