Abstract
The macaque middle temporal (MT) area is well known for its visual motion selectivity and relevance to motion perception, but the possibility of it also reflecting higher-level cognitive functions has largely been ignored. We tested for effects of task performance distinct from sensory encoding by manipulating subjects' temporal evidence-weighting strategy during a direction discrimination task while performing electrophysiological recordings from groups of MT neurons in rhesus macaques. This revealed multiple components of MT responses that were, surprisingly, not interpretable as behaviorally relevant modulations of motion encoding, or as bottom-up consequences of the readout of motion direction from MT. The time-varying motion-driven responses of MT were strongly affected by our strategic manipulation—but with time courses opposite the subjects' temporal weighting strategies. Furthermore, large choice-correlated signals were represented in population activity distinct from its motion responses, with multiple phases that lagged psychophysical readout and even continued after the stimulus (but which preceded motor responses). Together, these approaches revealed multiple cognitive contributions to MT responses that are task related but not functionally relevant to encoding or decoding of motion for psychophysical direction discrimination, providing a new perspective on the assumed status of MT as a simple sensory area.
Introduction
In the hierarchy of the visual system, the Middle Temporal (MT) area has long been considered the "gold standard" for sensory encoding. For decades, neuroscientists have used MT to study how sensory noise influences behavior, often quantified as "Choice Probability." The prevailing assumption is that MT provides the feedforward evidence that higher-level areas "read out" to make a decision. However, it has remained unclear whether the choice-related signals observed in MT are truly the cause of a decision or merely a consequence of feedback. This paper investigates whether MT activity is strictly tied to sensory encoding or if it reflects higher-level cognitive variables like task strategy and choice maintenance.
Methods and Experimental Design
To dissociate sensory encoding from cognitive strategy, researchers trained two rhesus macaques to perform a direction-discrimination task using a novel temporal evidence-weighting manipulation. The stimulus consisted of seven motion pulses, and the average motion strength was varied across three conditions:
- FLAT: Motion strength was constant on average throughout the trial.
- LATE: Motion was weak early and strong late.
- EARLY: Motion was strong early and weak late.
Electrophysiological recordings were obtained from 583 MT units across 71 sessions. To analyze population-level activity, the team moved beyond single-neuron metrics and employed variational Latent Gaussian Process (vLGP) analysis. This allowed for the extraction of low-dimensional latent factors, effectively denoising the signals and separating stimulus-driven activity from choice-driven signals.
Key Results
Paradoxical Sensory Gain
Standard models of attention suggest that if a subject ignores a part of a stimulus, the neural response to that part should decrease. Surprisingly, the researchers found the opposite: in the "EARLY" condition (where monkeys behaviorally ignored late pulses), MT actually showed increased sensitivity to those late pulses. This "paradoxical gain" suggests that MT might be compensating for a mismatch between expected and actual stimulus statistics, rather than simply reflecting behavioral weighting.
Choice Signals and Feedback
By analyzing the "psychophysical kernel"—the time course of how the stimulus influenced the monkey's choice—the study found that choice-related signals in MT peaked after the most influential stimulus pulses. Furthermore, these choice signals persisted in MT during the 500ms post-stimulus delay period, long after the motion stimulus had vanished. This suggests that MT maintains a representation of the chosen action, likely via feedback from higher-level areas, rather than acting as the primary driver of the decision.
Population-Level Insights
The use of population decoders revealed significantly stronger and cleaner choice signals than traditional single-neuron analyses. Choice Probability (CP) reached values >0.6 during the stimulus and >0.7 during the delay period, substantially higher than the ~0.55 typically reported in single-neuron MT studies.
Discussion and Conclusion
These findings push MT out of the "simple sensory" box and into the realm of cognitive processing. The dissociation between behavioral weighting and neural sensitivity, combined with the persistence of choice signals during delay periods, indicates that even our most basic sensory areas are deeply integrated with internal strategies and decisions. While the specific source of the feedback (e.g., LIP or Prefrontal Cortex) remains to be identified, the results provide a new perspective on MT as a site where sensory information meets high-level cognitive modulation.
Summary updated