Why do expectations change activity in early sensory areas?
Early sensory areas are often described as the first stop for raw input, but they do not operate in isolation. Higher regions send predictions downward, and these top-down signals adjust how strongly sensory neurons respond to what arrives. In the auditory system, signals from the medial prefrontal cortex appear to govern prediction errors in primary auditory cortex, so the mismatch between an internal prediction and the actual sound is registered with help from a higher control area [2].
The same principle appears in vision. Prior experience modulates top-down predictive processing in ventral visual areas, meaning that what a person has recently learned or attended to changes how predicted versus unpredicted visual stimuli are processed [3]. In both cases the expectation is not a separate add-on to perception; it is part of the machinery that decides how much a sensory area responds.
What does a prediction-error signal actually look like?
A prediction error is the neural response to a mismatch between what was expected and what actually happened. It is tempting to assume that unexpected stimuli always produce a bigger response, but the evidence is more nuanced. In six-month-old infants, an audiovisual omission paradigm showed that unexpected stimuli failed to evoke neural responses, which the authors discuss as a possible prediction-error signal rather than a simple surprise effect [1]. That is a striking result: the unexpected event produced less activity, not more.
The interpretation matters here. The same paper notes that an enhanced response to an infrequent stimulus could reflect a prediction error, a nonspecific surprise effect, or recovery from greater repetition of the frequent stimulus [1]. So a large response to something rare is not by itself proof of prediction-error signaling; it could be a general novelty reaction. Distinguishing these possibilities requires designs that separate expectation from mere rarity.
When does the prediction-error signal change?
Prediction errors are not fixed reactions; they shift as the brain's model of the world is updated. In a study of context-dependent sensory representations, reversing a learned rule caused the lateral orbitofrontal cortex to signal a context-prediction error, and that signal inhibited top-down input to sensory areas [4]. In other words, a higher-level mismatch can actively dial down the influence of expectations on early sensory cortex, which changes what those areas encode next.
This suggests a hierarchy of prediction errors, where errors at one level reshape the predictions passed to the level below [4]. The practical consequence is that the same unexpected stimulus may generate a strong error signal in one context and a weak or absent one in another, depending on what the higher areas currently expect and how recently those expectations were violated.
Is it prediction, or just attention?
A central difficulty is that attention and prediction can produce similar-looking changes in early sensory responses. A study of speech perception used EEG to try to separate prediction-error responses from attentional effects, examining how word surprisal and predictive context influence early auditory encoding [5]. The goal was to distinguish between a genuine mismatch signal and the boost that comes simply from focusing on a stimulus.
This distinction matters for interpreting any result about unexpected stimuli. If an unexpected sound or word draws more attention, the resulting response may look like a prediction error without being one. The infant omission result, where unexpected stimuli evoked no response, is a useful counterexample because it does not fit a simple attention-boost pattern [1]. Taken together, the evidence is consistent with top-down expectations shaping early sensory cortex and with prediction errors being real but context-dependent signals, not a uniform response to anything surprising [2][4][5].
Sources used in this answer
Top-down modulation in the infant brain: Learning-induced expectations rapidly affect the sensory cortex at 6 months
In six-month-old infants, learning-induced expectations rapidly affected sensory cortex responses, and unexpected stimuli in an audiovisual omission paradigm failed to evoke neural responses, which the authors discuss as a possible prediction-error signal.
Top-down prediction signals from the medial prefrontal cortex govern auditory cortex prediction errors
Top-down prediction signals from the medial prefrontal cortex govern prediction errors in primary auditory cortex when sensory input does not match internal predictions.
Prior experience modulates top-down predictive processing in the ventral visual areas
Prior experience modulates top-down predictive processing in ventral visual areas, with short-term learning and attention changing responses to predicted versus unpredicted stimuli.
Hierarchy of prediction errors shapes the learning of context-dependent sensory representations
Upon rule reversal, lateral orbitofrontal cortex signals a context-prediction error that inhibits top-down input to sensory areas, supporting a hierarchy of prediction errors that shapes context-dependent sensory representations.
Attention or prediction? Characterizing the top-down influence of predictive context on speech encoding
EEG during speech perception was used to try to distinguish prediction-error responses driven by word surprisal and predictive context from attentional effects on early auditory encoding.