Summary: New research shows that human olfaction relies heavily on predictive coding—more so than vision—reframing smell as a sophisticated, proactive sense rather than a primitive, reactive one. Behavioral tests and fMRI scans reveal that odors that violate expectations trigger not only olfactory regions but also visual brain areas, highlighting smell’s multisensory and anticipatory processing.
These findings emphasize how smell depends on prior expectations and input from other senses to identify odors accurately, demonstrating a distinct neural strategy for evaluating unexpected smells.
Key Facts:
- Smell depends more heavily on predictive cues than vision, indicating it uses anticipatory mechanisms for perception.
- Unexpected odors activate both olfactory and visual cortex regions, suggesting cross-sensory engagement when predictions fail.
- The conclusions are supported by two behavioral experiments and one fMRI study, demonstrating that predictive coding applies strongly to olfactory perception.
Source: Stockholm University
Background: A prominent framework in neuroscience proposes that the brain’s main function is to predict incoming sensory events and react primarily to prediction errors. Most research testing this predictive coding framework has concentrated on vision. Until now, it has been unclear whether other senses—especially olfaction—use similar predictive mechanisms and how their cortical processing differs.
To investigate how smell uses predictive information, researchers at Stockholm University designed three experiments: two behavioral studies and one fMRI experiment conducted at the Stockholm University Brain Imaging Centre (SUBIC).
“Our primary finding is that olfaction is far more dependent on predictions than vision,” explains Stephen Pierzchajlo, PhD student in the Department of Psychology and lead author. “Many people assume smell is a basic, reactive sense. Our data show it is actually complex and proactive, relying strongly on expectation to guide perception.”
The experiments demonstrate the value of accurate cross-modal cues—such as words or images—when the brain classifies sensory inputs.
“We’ve all noticed how an unexpected smell can grab our attention, for instance when entering a room and encountering an unfamiliar odor,” says Jonas Olofsson, professor of psychology and co-author. “The sense of smell is highly shaped by cues from other senses, much more so than sight or hearing, which appear less dependent on such predictive information.”
Using spoken word cues and subsequent targets, the researchers found that unexpected smells engaged both olfactory and visual brain regions even when no visual input was presented. This cross-sensory activation suggests the brain recruits visual processing to help identify surprising odors—perhaps to locate or recognize their source.
“When an odor violates expectation, the olfactory system triggers a distinct processing route: it not only signals the mismatch but also recruits visual areas, likely to assist in resolving what the smell indicates. This makes sense because humans struggle to recognize odors without contextual clues,” Olofsson adds.
In the tasks, participants heard spoken labels like “lemon,” then encountered either a picture or an odor. They judged quickly whether the target matched the cue (for example, a lemon picture or scent) or did not match (for example, a rose image or scent). Response times and neural activity were compared across congruent and incongruent trials to evaluate the role of prediction.
“As expected from predictive coding, congruent items produced faster responses overall. We used the delay in responses to unexpected stimuli as a measure of how much a sense relies on prediction—a larger delay indicates greater dependence,” says Pierzchajlo. The difference in delay between olfactory and visual targets showed that smell is particularly reliant on prior cues.
This work represents the first completed portion of Pierzchajlo’s doctoral research into olfactory prediction and categorization.
“Our results support the view that the human sense of smell is proactive: it evaluates odors against expectations and employs a specialized, cross-modal cortical strategy to resolve unexpected smells,” he summarizes.
Study details
- Three experiments were conducted: two behavioral studies and one pre-registered fMRI experiment at SUBIC.
- Behavioral Experiment 1 included 69 participants; Behavioral Experiment 2 included 50 participants.
- For the fMRI portion, initial pilot data were collected from 15 participants, followed by a main sample of 32 healthy volunteers.
- Across experiments, the same set of four familiar stimuli—lavender, lilac, lemon, and pear—were repeatedly presented as odors, pictures, or spoken words. This design maintained high accuracy and comparable conditions for measuring response times without bias.
About this olfaction research news
Author: Gunilla Nordin
Source: Stockholm University
Contact: Gunilla Nordin – Stockholm University
Image: The image is credited to Neuroscience News
Original Research: Open access. “Olfactory categorization is shaped by a transmodal cortical network for evaluating perceptual predictions” by Stephen Pierzchajlo et al., Journal of Neuroscience.
Abstract
Olfactory categorization is shaped by a transmodal cortical network for evaluating perceptual predictions
Predicting sensory input and evaluating prediction errors are central to perception, but little is known about how different senses rely on predictions and which cortical circuits they engage. This study tested whether olfaction depends on non-olfactory predictive cues and whether it recruits distinct cortical resources compared with vision.
Participants heard spoken cues (for example, “lilac”) and judged whether subsequent targets—odors or pictures—matched the cue. Across two behavioral experiments (total n = 113; 72 female), the response-time difference between congruent and incongruent trials was significantly larger for olfactory targets than for visual ones, indicating a stronger reliance on verbal predictions when identifying smells.
A pre-registered fMRI study (n = 30; 19 female) found that incongruent odors elicited greater activation in the anterior cingulate cortex, a region known for error detection, suggesting its involvement in olfactory prediction error processing. In addition, incongruent olfactory trials activated both primary olfactory and visual cortices, pointing to cross-sensory processing during olfactory prediction errors—an effect not observed for visual prediction errors.
These results suggest olfaction is uniquely characterized by a high dependency on predictive, non-olfactory cues, and that odors are evaluated through a transmodal cortical network that integrates predictions and cross-sensory information during rapid decision-making.