At each level of abstraction, the brain performs the same operation: transforming the incoming information by chunking and extracting patterns to obtain more adequate and efficient representations that can support its goals. Despite the prevalence of such processes across sensory modalities, investigations into the organizing principles of segmentation (chunking) typically focus on individual modalities separately. We hypothesize that at least some basic chunking principles and corresponding perceptual biases are akin across modalities with tractable neural correlates in the primary sensory cortical areas.
To test this hypothesis, we focused on the auditory segmentation principle called the Iambic-Trochaic law (ITL). Established in language processing, ITL posits that longer syllables in a sequence signal word-ends while an increase in intensity signals the beginning of a word. Moreover, as a behavioral relevance, such chunking leads to decreased accuracy in detecting perceptual changes at perceived chunk boundaries compared to that within the segment itself.
Importantly, ITL biases have been found unrelated to linguistic content and across multiple species, but never tested in other modalities, we implemented a stream segregation go/no-go paradigm for human participants in audition and vision to explore the generality of the phenomenon. Participants’ task was to identify unexpected gaps in a structured stream in an identical manner in the visual and auditory modalities. The stream had 3-element intensity or duration patterns to probe segmentation biases.
We found that variability in sensitivity to gap deviations showed similar, pattern-derived biases across the two modalities. This sensitivity bias could be explained neither by the repetition of individual elements nor by the absolute feature value (e.g. duration, intensity) of the individual elements alone. Instead, this bias depended on the internal repeating structure of the stream and it had an effect even when conscious recognition of the structure itself did not occur.
We also analyzed neuronal activity in the auditory cortex (AC) of awake, head-fixed mice passively exposed to similar acoustic stimuli to see how AC neurons respond to changes within a continuous stream. We found that AC activity significantly increased in response to stimuli featuring unexpected gaps, again, as a function of their position in the pattern.
By employing consistent paradigms across sensory modalities (auditory/visual) and experimental models (human/mouse), our results support the idea of domain-general non-linguistic grouping principles and raise well-testable further questions that have the capacity to lead to a domain-independent model of sensory processing.