5th international workshop on attention in cognitive systems
associated with icvs2008

12 May 2008,
Santorini, Greece

Invited Talks

"Learning to Attend: From Bottom-Up to Top-Down"

Jochen Triesch
Frankfurt Institute for Advanced Studies, Frankfurt am Main, Germany
and Department of Cognitive Science, Univ. of California, San Diego, USA


The control of overt visual attention relies on an interplay of "bottom-up" and "top-down" mechanisms. Purely "bottom-up" models cannot provide a satisfactory account of human attention orienting in many natural behaviors. But how do humans learn to incorporate "top-down" mechanisms into their control of attention? The phenomenon of "gaze following", i.e. the ability to infer where someone else is looking, offers an interesting window into this question. I review findings on the emergence of gaze following in human infants and present a computational model of the underlying learning processes. The model proposes an explanation for the gradual shift of emphasis from bottom-up to top-down cues in attention control. It explains this process in terms of generic reinforcement learning mechanisms and predicts a new class of "mirror neurons" specific for looking behaviors. Finally, it offers an explanation for deficits in gaze following in developmental disorders such as autism.

"Brain Mechanisms of Attentional Control"

Portrait of

Steve Yantis
Department of Psychological and Brain Sciences
The Johns Hopkins University


Perceptual tasks require attentional selection of relevant sensory input. We have investigated the neural basis of attentional control using human fMRI during tasks that require voluntary shifts of visual and auditory attention between locations, features, objects, or sensory modalities. In each domain, we observe domain-specific attentional modulation of sensory cortex and domain-independent transient shift signals in posterior parietal and superior prefrontal cortex that reflect the initiation of new attentive states.  We speculate that these cortical control signals are likely to be part of a corticostriatal loop that mirrors the well-known control circuits for eye and limb movements.