The role of the pSTS in gaze following and joint attention

Abstract

In humans, eye gaze of another person is a powerful stimulus, drawing the observer’s attention precisely to places and objects of interest to the other one. Monkeys also follow gaze, but rather than using eye-gaze, they rely more on peer head orientation to shift attention. Are the human and the monkey gaze following systems functionally related and do they depend on the same anatomical substrates, eventually sharing a common phylogenetic background? How can we voluntarily control our gaze following behavior and eventually suppress it when it is not suitable in a certain context? How does the brain decide which object to attend to in a situation in which the other’s gaze direction seems to point towards more than one object at the same time? What are the properties of single neurons in the putative gaze following region of the brain and which specific computations are they capable of? My thesis tried to address these intimately related questions. In order to address the first question, we asked human subjects to make saccades to distinct spatial targets, either identified by the eye-gaze of a demonstrator portrait or, alternatively, by associating the demonstrators’ iris color to four spatial targets (gaze following vs. color matching). Using functional magnetic resonance imaging (fMRI) we identified a highly specific region, namely the “gaze following patch (GFP)”, in the posterior superior temporal sulcus (pSTS) activated by gaze following, well separated from those parts of the STS known to process visual information on faces and heads, a finding very similar to previous experiments on rhesus monkeys. In order to answer the second question, we looked at the brain activity while subjects were required to suppress the gaze following response. In this experiment, we could show that the cognitive control of gaze following was based on the activation of two regions in the frontal cortex, the dorsolateral prefrontal cortex and the anterior cingulate cortex. In a subsequent experiment, we integrated the need to extract the gaze vector direction and provided a-priori information in order to allow the viewer to select only one object out of several met by the same gaze vector. Using fMRI we demonstrated that the disambiguation of the potential object was mainly confined to a region in the inferior frontal junction. Finally, we used monkeys head gaze following as a model for human gaze following and recorded the activity of single neurons from the GFP of two rhesus monkeys8 engaged in a battery of tasks in an attempt to understand how the other’s gaze might guide spatial attention to a target shared by the two agents. We establish that the properties of neurons in the pSTS are indeed able to explain the monkeys’ ability to follow gaze. The fMRI work on the relationship of gaze following and face processing-related activity supports the notion that monkeys’ head gaze following might be homologous to the human eye-gaze following. The fMRI studies on cognitive control and the role of context reveal important features of the network centering on the GFP and the single neuron work on the monkey GFP is able to explain how gaze-related information is translated into shifts of attention by distinct sets of single neurons.