The primacy of behavioral research for understanding the brain

Introduction

Yael Niv argues that to understand the brain we must answer what it does and how it does it, and that behavior often provides stronger, earlier constraints for the first question than neural measurements alone. He contends that purely behavioral research is essential for understanding cognitive functions and that neuroscience should complement, not replace, behavioral inquiry.

What has behavior taught us about the brain?

Behavioral data have predicted neural and computational principles long before invasive or high-resolution neural methods. For example, color vision was inferred from psychophysics (three cone types) before cone-level physiology was measured, and high-level attention and cognitive control concepts emerged from simple behavioral paradigms such as visual search. In navigation research, a rat in a T-maze revealed that decision strategies shift with training: egocentric versus allocentric strategies can be distinguished by turning the maze around and observing the choice pattern. This behavioral manipulation helped generate computational models of learning and strategy switching, and later neural studies refined those ideas but did not supplant the behavioral findings.

Blockage experiments in classical conditioning revealed that learning requires a prediction error rather than mere co-occurence of stimuli. The Rescorla–Wagner model formalizes this idea, and later reinforcement-learning frameworks showed how prediction errors guide learning, with dopaminergic signals often linked to these errors. Retrieval-induced forgetting demonstrated how competition during retrieval shapes memory traces, constraining how memory networks are organized and updated. Across domains—from memory to attention and perception—behavioral data have historically driven the formulation of core computational and neural hypotheses.

In newer work, behavioral paradigms have begun to dissect distinctions such as how surprising rewards affect memory encoding and how attention operates as a rhythmic sampling process at around 8\ \text{Hz}, constraining where in the brain these processes must be implemented. These findings illustrate how rich behavioral data can reveal the structure of memories, perception, and cognitive control without requiring invasive neural measurements.

The synergy of behavior, computation, and neural data

Neuroscience can specify where in the brain a function is localized, but understanding what computation a process embodies often benefits from carefully crafted behavioral tasks and computational modeling. The history of the field shows that many deep insights—such as the distinction between goal-directed and habitual control and the multiple memory systems—originated in behavioral studies paired with theoretical accounts. When neural data are integrated with precisely designed behavioral experiments and formal models, they tend to confirm or refine computational hypotheses rather than replace them. In perception and higher cognition, pure neural measures have been most informative when tied to hypotheses that emerge from behavioral data and computational theory.

Clever behavioral experiments can yield causal inferences beyond correlational neural data

A key strength of behavioral paradigms is their ability to induce causal engagement of cognitive processes without relying on neural perturbations. For example, the N-back task can turn working-memory processes on or off by adjusting task demands, allowing causal inferences about the role of maintained representations in behavior. Similarly, dual-task manipulations can reveal whether different modalities share or separate working-memory resources. These designs show that clever behavioral control can isolate specific cognitive operations noninvasively, providing tight tests of computational accounts even before, or instead of, neural perturbations. While neural measurements remain valuable, these behavioral approaches often yield clearer causal constraints on the mechanisms at play.

Why behavioral work is undervalued and why that is a problem

Niv argues that there is a persistent bias in neuroscience funding and publishing that elevates neural data over behavioral data, sometimes treating behavior as auxiliary to neural explanations. Misperceptions include viewing neural data as inherently objective while interpreting behavior through theory as subjective, and assuming that mind-level questions are solved once neural correlates are found. He notes real-world pressures from funding agencies and journals that may discourage non-neural approaches, despite the fact that many foundational insights into perception, memory, attention, and higher cognition arose from behavioral research. He also cites cases where neural data provided limited additional constraint beyond what behavior had already revealed, underscoring the value of behavioral paradigms in guiding efficient and cost-effective scientific progress.

Conclusion

The central message is a call for restoring behavior to its rightful place as the foundational basis for understanding cognition and, in tandem with neuroscience, for explaining how mental processes are implemented in the brain. By prioritizing well-crafted behavioral questions, supported by computational modeling, and using neural data to constrain and test competing theories, we can accelerate progress in understanding the mind and its neural substrates. Behavior is not merely subsidiary to neural data; it is the robust engine that has historically driven, and should continue to drive, advances in cognitive neuroscience.