Why a single behavioral report cannot tell you whether sensory coding changed
When you manipulate expectations—say, by making a target more likely—any change in what people report could come from at least two places: the sensory signal itself might be sharper, or the decision stage might be biased toward the expected answer. A behavioral report alone cannot distinguish these. The solution is to add a second, independent measure that tracks sensory processing more directly, such as an electrophysiological signal that reflects early attentional guidance or neural firing that follows stimulus properties rather than the animal's choice [4][5].
In one visual search study, researchers probabilistically manipulated expectations about whether a target would be a digit or a letter. Because the target was embedded among distractors from both categories, an expected-category target should not capture attention more than an unexpected one. Indeed, behavioral and electrophysiological indices of attentional guidance showed no effect of expectation, yet targets from the expected category were still more likely to be identified. That pattern points to a later encoding stage—after attention has done its work—rather than a change in early sensory selection [4].
What independent neural measures reveal about where expectations act
A second strategy is to record from neurons at different stages of a sensory pathway and see which ones change with expectation or task context. In the somatosensory system, neurons in secondary somatosensory cortex (S2) show a mix: some have invariant sensory responses, others correlate with the animal's perceptual report, and most fall along a continuum between the two. When the task became non-demanding, sensory responses stayed the same while perceptual responses and categorical information decreased—evidence that sensory coding and perceptual coding can be pulled apart by task context [5].
This kind of dissociation is powerful because it does not rely on the participant's report alone. If a neuron's firing tracks the physical stimulus regardless of what the animal reports, it is a candidate for sensory coding. If another neuron's firing tracks the report regardless of the stimulus, it is a candidate for perceptual or decision-related coding. Finding both types in the same region, and seeing them shift differently with task demands, supports the idea that the brain converts sensory inputs into perception through intermediate stages [5].
How adaptation can shift perception while leaving a neural code intact
Sensory adaptation offers another window. In honeybees, researchers adapted one odor in a mixture and then measured both neural responses and behavior. At the level of olfactory receptor neurons and projection neurons, adaptation reduced the adapted odor's contribution but preserved the relative strength of the non-adapted component. Behaviorally, however, the bees' perceptual quality of the mixture shifted toward the non-adapted odor. That is a dissociation: the neural representation of the mixture's relative composition stayed stable, yet the reported perceptual quality changed [2].
The honeybee study also used a conditioning protocol to manipulate expectations, which makes it relevant to the question of how expectations interact with sensory coding. But the key lesson is methodological: by measuring neural coding and perceptual quality separately, the experiment showed that a change in perception does not always require a change in the sensory code. The authors suggest that adaptation enhances mixture segregation, helping detect low-concentration components that might otherwise be masked—a possible function, not a proven purpose [2].
What these approaches can and cannot settle
These designs share a common logic: manipulate expectations, measure at least two things—a neural or electrophysiological index of sensory processing and a behavioral report—and look for a pattern where one changes while the other does not. When that happens, it is evidence that the expectation acted at a stage downstream of sensory coding. The visual search study ruled out repetition priming and response bias as alternative explanations, strengthening the case that the effect was on encoding rather than on decision criteria [4].
Still, the evidence has limits. The honeybee work shows a dissociation between neural coding and perceptual quality, but it does not directly manipulate expectations in a way that isolates their effect on awareness [2]. The somatosensory study demonstrates a continuum of sensory and perceptual coding, but it does not manipulate expectations—it manipulates task demand [5]. And the visual search study provides strong evidence for attention-independent expectation effects on encoding, but it measures identification accuracy, not a direct readout of conscious awareness [4]. Together, these studies suggest that separating sensory coding from reported awareness is possible when experiments combine independent neural measures with careful behavioral designs, but no single study fully solves the problem for all cases.
Sources used in this answer
Expectations change the signatures and timing of electrophysiological correlates of perceptual awareness
Expectations modulate the timing and properties of electrophysiological correlates of perceptual awareness, but the study does not explicitly separate sensory coding from reported awareness.
Sensory adaptation modulates coding and the perceptual quality of odor mixtures
In honeybees, sensory adaptation reduced the adapted odor's neural contribution while preserving the non-adapted component's relative strength, yet shifted the perceptual quality of odor mixtures toward the non-adapted odor, demonstrating a dissociation between sensory coding and perceptual quality.
Integration of sensory evidence and reward expectation in mouse perceptual decision-making task with various sensory uncertainties
Mice integrated sensory inputs and reward expectations by updating a single choice threshold with a simple reinforcement-learning rule rather than using separate optimal thresholds for different sensory uncertainties.
A dissociation between the effects of expectations and attention in selective visual processing
Probabilistic category expectations improved target identification without modulating attentional guidance or engagement, affecting only later encoding-related stages, and ruling out repetition priming and response bias.
A continuum of invariant sensory and behavioral-context perceptual coding in secondary somatosensory cortex
In secondary somatosensory cortex, neurons with invariant sensory responses coexist with neurons correlating with perceptual behavior along a continuum, and during a non-demanding task sensory responses remained unaltered while perceptual responses decreased.