Abstract
Human studies of speed-accuracy tradeoff (SAT) with non-invasive measures have offered insights but highlight uncertainty about fundamental questions regarding how the brain accomplishes SAT. Different tasks and analysis pipelines across studies identify or emphasize different cortical and subcortical regions. Several laboratories propose that medial frontal cortex sets the excursion between baseline and threshold levels of the evidence accumulation process through interactions with striatum and other regions. Using neural spiking data measured in supplementary eye field (SEF) as well as frontal eye field (FEF) and superior colliculus (SC), we investigated whether spike rate correlations across areas contribute to SAT. Two macaque monkeys performed a visual search task to locate a target (T/L) presented amongst seven distractors (L/T). Trials began when monkeys fixated a central stimulus, the color of which (green or red) cued emphasis on response speed (Fast condition) or accuracy (Accurate condition). Search contingency was fixed as either T amongst L’s (less difficult) or L amongst T’s (more difficult). Response time was shorter, and choice error rate, higher, in the Fast relative to the Accurate condition. We measured changes in spike count correlation (rsc) between neurons recorded in SEF and those recorded simultaneously in FEF or SC as a function of task condition, trial epoch, and trial outcome. On correct trials, we observed no effect of SAT on rsc between SEF and FEF/SC. However, on choice error trials, rsc between SEF and FEF/SC was significantly elevated in the Accurate relative to the Fast condition. In contrast, on timing error trials, rsc between SEF and FEF/SC was elevated in the Fast relative to the Accurate condition. These correlations were significantly larger after relative to before saccades. These new observations complement descriptions of medial frontal influence on SAT in humans and offer new constraints on neuro-computational mechanisms of SAT.