Abstract
Single-unit electrophysiological studies have demonstrated that attention could result in several different modulations of the contrast response function (CRF) measured in visually responsive neurons: contrast gain (the CRF shifts leftward), response gain (the multiplicative gain factor increases) and additive shift (the CRF moves upwards by a constant amount). Yet, decisions made about visual stimuli are likely based on population-level responses. Attentional modulation of CRFs, measured using population-level techniques such as fMRI or steady-state visually evoked potentials (SSVEP) in EEG, have uncovered conflicting patterns of gain modulation. When measured using fMRI, attention results in a purely additive increase in CRFs (Buracas & Boynton, 2007; Murray, 2008; Pestilli et al, 2011). However, SSVEP CRFs exhibit either response gain or contrast gain (Kim et al, 2007; Lauritzen et al, 2010; Itthipuripat et al, in press). So far, it is unclear whether the difference in gain pattern between fMRI and EEG is due to the differing modulation of the vascular and electrophysiological responses or due to the fact that results were obtained across different participants and tasks with differing stimulus properties. Here, we conducted a study where CRFs were measured using fMRI and SSVEP from the same participants using an identical spatial attention task. We found that attention primarily enhances the additive gain in the hemodynamic CRF but results in either response or contrast gain in the SSVEP CRF, depending on the participant. Furthermore, we used a multivariate spatial encoding model analysis with fMRI data (Sprague & Serences, 2013) to reconstruct spatial representations of the stimulus display and found a spatially selective enhancement of stimulus representations with attention. Together, these results suggest that local pooling reflected in hemodynamic responses masks response or contrast gain patterns with attention that can be well-characterized using large-scale electrophysiology which constrains inferences that can be made using each method.
Meeting abstract presented at VSS 2014