How Tangrams Reveal the Secrets of Our Vision?

Our brain is a true visual wizard: it can recognize thousands of objects, even when they are highly abstract. To study this power, scientists designed a new “tangram paradigm,” a set of geometric shapes that can be combined in countless ways. Each tangram was analyzed for three key dimensions: whether it depicts something alive (animacy), how abstract it is, and how many small features it contains.

Using this data, researchers built a hierarchical model describing how the brain organizes visual information. They then showed 85 tangram images to 24 volunteers while recording brain activity with electroencephalography (EEG). The results reveal that the three dimensions are processed at different moments: first the brain distinguishes animacy, then it evaluates abstraction level, and finally it analyzes the tiniest details.

More advanced analyses, such as representational similarity and temporal generalization, showed that during “mind‑change” moments the brain can activate or suppress local feature processing. In practice, our visual system shifts from a global view to a finer one, depending on what is most useful at the time. This study proves that tangrams, thanks to their flexible structure, are a powerful tool for exploring the dynamic stages of human visual cognition.

The tangram paradigm introduced in this study relies on a cognitive‑associative encoding approach, allowing formal mapping of relationships between visual attributes and semantic categories. Practically, each tangram is represented as a multidimensional vector whose components correspond to variables such as animacy (binary value), abstraction level (continuous scale), and local feature density (count of angles or lines). This representation enables the computation of similarity matrices between stimuli using Euclidean or cosine metrics, facilitating the application of Representational Similarity Analysis (RSA).

EEG, with its millisecond temporal resolution, allowed tracing the dynamic evolution of neural representations. Data showed an early activity burst (≈100‑150 ms) linked to animacy, typical of occipito‑temporal areas involved in biological form recognition. A second burst (200‑300 ms) reflected abstraction processing, engaging parietal and prefrontal regions associated with abstract reasoning. Finally, between 300‑500 ms, a response related to local feature density emerged, involving primary and secondary visual cortex.

Temporal generalization analyses revealed that higher‑level representations (abstraction) can re‑emerge in later phases, suggesting a "mind‑change" mechanism where the brain recalibrates attention from a global to a more detailed coding. This aligns with top‑down attention models, where prefrontal areas modulate sensory region activity based on cognitive context. The tangram paradigm, thanks to its capacity to simultaneously manipulate multiple dimensions in a controlled way, provides an ideal platform for testing hypotheses on how the brain integrates and reorganizes visual information in real time. Moreover, the public dataset of over 900 tangrams enables replication and extension of experiments, fostering standardization of neuro‑cognitive measures in this field.