Abstract
When rapid orienting to brief visual targets is reliably rewarded, associations between targets (cues) and orienting responses become established via learning processes. An important question is whether such learning is limited to the specific stimulus-response associations acquired during learning or is more abstract, allowing modulation of other oculomotor behaviours not specifically reinforced during learning. To investigate, we briefly (250 ms) presented three different Japanese letters (hiragana) in pairs and differentially reinforced choice; different items led to a monetary win, loss, or nothing. Half of participants (n = 24) were asked to choose the optimal hiragana; others actively rejected the non-optimal item, thus learning different cue-response associations but the same cue-value associations. After learning, each hiragana served as a fixation stimulus in a simple saccadic reaction time (SRT) task requiring speeded saccades to a left or right peripheral (4 deg) dot target (no rewards provided). Fixation stimuli either remained visible for 200 ms after the target appeared (overlap condition) or offset 200 ms prior to the target’s appearance (gap condition). SRTs are typically slower in the overlap versus gap conditions because fixation stimuli activate inhibitory ‘fixation neurons’ in the superior colliculus that prevent reflexive saccades during fixation. If learning builds cue-value associations (not just cue-response associations) and thus makes items more attractive, then SRTs should be slower in the overlap condition when the fixation stimulus is win (versus loss or zero) associated regardless of the task used during learning. Indeed, SRTs for both groups were significantly slowed (by 27 ms) with win versus zero (or loss) fixation stimuli in the overlap condition; no value effects in the gap condition were found, ruling out a strategic account. These results show that fixation neurons can be modulated by prior value-learning, and that cue-value associations can influence oculomotor behaviours not specifically reinforced during learning.
Meeting abstract presented at VSS 2015