Characterizing perception as intrinsic meaning: a quantitative foundation for the science of the Perception Box
Giulio Tononi, Andrew M. Haun, Chen Song, Matteo Grasso, Will Mayner
August 13, 2026 DOI: 10.17605/osf.io/eazmx (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Theoretical or philosophical paper |
|---|---|
| Key points | Argues that perception is subjective and shaped by learning history, and proposes that Integrated Information Theory can account for the intrinsic quality of percepts. Plans experiments to show that implicit learning in different environments leads to systematically different perceptions of the same stimulus, transforming a 'jumble' into a 'unitized' and 'categorized' object. |
Abstract
Everyone lives within a Perception Box (PB). What we see and hear—how we interpret it and what it means to us—depends on our brain and learning history. Recognizing that perception is subjective, and understanding why, are key steps to open our minds, counteract bias, and reduce suffering. This project seeks to establish a rigorous theoretical and empirical foundation for the science of the PB. A key requirement is understanding what makes a perception feel the way it does, and why it may feel differently in different people. Integrated Information Theory (IIT) is unique in providing a principled, testable account of the quality of experience. According to IIT, a percept corresponds to a cause–effect structure specified by the substrate of consciousness in its current state. Although triggered by an external stimulus, a percept’s feeling/meaning is always intrinsic—consistent with the PB notion. IIT already provides a principled account of the feeling of spatial extendedness and temporal flow based on the kinds of cause-effect structures specified by different brain regions. Ongoing work aims to explain what it takes to perceive a stimulus as an object—that is, to experience it as a unitized configuration of features (perceptual unitization) and as a specific instance of a general category (perceptual categorization). This project uses simple model systems (animats) to illustrate how different animats come to perceive the same stimulus in different ways and as different objects. We will expose two groups of animats to two different visual environments, characterized by distinct stimulus statistics. Through this exposure (implicit learning), units within each animat’s brain will develop unique connectivity patterns, shaping the formation of intrinsic meanings. We will show how the cause-effect structure specified by the stimulus-evoked brain activity can account for the percept’s intrinsic meaning, including perceptual unitization, categorization, and their binding. Importantly, we will show that the two groups of animats will come to form systematically different intrinsic meanings—perceiving different objects within the same stimulus (for example, as ‘segment’ versus ‘centered odd’). We will also show how similarities and dissimilarities of cause–effect structures account for similarities and differences in intrinsic meanings. Parallel experiments in human subjects will employ the paradigm of implicit perceptual learning to provide a firsthand demonstration of the PB, illustrating how different people come to perceive the same stimulus in different ways and as different objects. Two groups of subjects will be exposed to two different visual environments—streams of visual stimuli—that embed distinct statistical regularities unbeknownst to them. Through this exposure (implicit learning), plasticity mechanisms will drive changes in neural connectivity in each subject, shaping the formation of intrinsic feelings/meanings. We will show how this process transforms the percept of a stimulus from a “jumble” of random strokes to a “unitized” configuration (perceptual unitization) and then to a “categorized” object (perceptual categorization), such as a character. Importantly, we will show that the two groups of subjects will come to form systematically different intrinsic feelings/meanings—perceiving different objects within the same stimulus (for example, 令 vs. 瓜). Crucially, these protocols will assess the way percepts feel and how their intrinsic feeling/meaning changes with learning, emphasizing phenomenology (perceptual unitization and categorization), rather than behavior or function (speed and accuracy of responses). We will demonstrate that what changes in these experiments is perception itself—not judgements about perception. The firsthand demonstration that the feeling/meaning of the same stimuli can change with implicit perceptual learning, depending on factors one is unaware of, will help the public recognize that perception is interpretation. This realization can start a journey towards understanding one’s PB and commitment to expanding its borders.