Abstract

A neural network for responding looming objects: An fMRI study

For most animals a looming object elicits a defensive or escape response; such a behaviour is perhaps optimal for species survival, and thus intuitively rooted in evolutionary archaic areas of the brain.  A more sophisticated response to a looming object (e.g. catching an approaching ball) may require a relatively more accurate judgement of time to collision (TTC), thus reside, or interact, with higher order brain regions.Evidence for looming detection in embryologically older brain regions has been found in several species.  For example, deep layers of the superior colliculous (SC) respond to looming stimuli (1), and electrical stimulation of SC results in an avoidance response in rats (2).  Research using looming stimuli with pigeons implicates neurons in the optic tectum and nucleus rotundas (mammalian homolog: SC and pulvinar nucleus thalamus respectively) in detecting both approximations of TTC and looming (3, 4). In humans, a recent time perception task (5) found that a region of the anterior insula cortex (AIC), the inferior parietal lobe (IPL) and the anterior putamen were involved in making duration judgements. The AIC may be of particular importance as it is functionally connected to the ventral striatum and a recent review implicated the AIC in higher order human awareness; possibly representing “global emotional moments” across time (6). This study focuses on both sub-cortical and cortical regions in humans which may be involved in detecting looming stimulus and calculating approximations of TTC.

Looming detection within natural scenes and potential errors in roadside judgments

Detection of looming is a critical for successful collision avoidance. Regan et al (1979 – 96) has been foremost in documenting humans sensitivity to looming and MiD (motion in depth), but there are various methodological factors that make it difficult to extrapolate from these measures to performance in natural settings.  The current study was concerned with looming thresholds in the context of roadside behaviour. Adaptive (BEST-PEST) staircase procedures were run using photo-realistic images of a motorbike or car presented for 200ms, in order to determine sensitivity to looming of vehicles in central or peripheral vision, under monocular viewing conditions and against a neutral grey background or a realistic static road scene. Two critical TTC arrival time values were simulated:  5s (sufficient time to cross) & 3s (critical decision point).  Vehicle images changed in size and expansion to simulate approach at different speeds, within a display configuration that ensured sufficient pixel resolution for all trials/steps. The participant’s task was simple detection of looming (opponent edge motion) for a vehicle image when there was also lateral translation of the image. Thresholds for looming in these conditions were substantially higher than those reported by Regan et al under more constrained psychophysical conditions.  We also found a significant increase in thresholds when stimuli were presented only 6deg in the periphery.  The results suggest that, in displays that contain the contrast and edge-detail of natural scenes, and where other motion information may be present, the detection of looming may be significantly poorer than previously reported. This still allows for accurate detection if the object is foveated, but if in a cluttered scene the observer glances slightly off-target they may fail to detect fast approaching vehicles.  This may be particularly a problem for smaller profile vehicles such as motorcycles and may explain driver errors with respect to these.

Research supported by the UK ESRC ES/F017650/1

Department of Psychology, Royal Holloway, University of London, Egham, Surrey TW20 0EX
Tel/Fax : +44 (0)1784 443526/434347