Stimulus-Related Neuroimaging in Task-Engaged Subjects Is Best Predicted by Concurrent Spiking

The implicit goal of functional magnetic resonance imaging is to infer local neural activity. There is considerable debate, however, as to whether imaging correlates most linearly with local spiking or some local field potential (LFP) measurement. Through simultaneous neuroimaging (intrinsic-signal optical imaging) and electrode recordings from alert, task-engaged macaque monkeys, we showed previously that local electrophysiology correlates with only a specific stimulus-related imaging component. Here we show that this stimulus-related component—obtained by subtracting a substantial task-related component—is particularly linear with local spiking over a comprehensive range of response strengths. Matches to concurrent LFP measurements are, to varying degrees, poorer. As a control, we also tried matching the full imaging signal to local electrophysiology without subtracting task-related components. These control matches were consistently worse; they were, however, slightly better for gamma LFP than spiking, potentially resolving discrepancies between our findings and earlier reports favoring LFP.