The retina may help the brain predict what it will see next

A computer model suggests the eye’s first stage of vision prepares the brain for visual scenes about 100 milliseconds ahead.

Editorial illustration — not from the study.

Researchers built a model of the retina that processes visual information as patterns of nerve-cell spikes. They trained it on natural movies under limits intended to resemble the retina’s energy constraints, comparing a model that encoded the present with one that predicted upcoming scenes.

The prediction-focused model produced retina-like response features, including sensitivity to timing, anticipation of motion, different responses to different kinds of movement and activity when an expected stimulus was omitted. It also predicted responses of retinal ganglion cells—the retina’s output cells—to natural images and movies across animal species.

How the retina model predicted

When optimized to predict visual scenes about 100 milliseconds ahead, the model developed response patterns resembling those seen in the retina. These included receptive fields, which describe how cells respond to different parts of an image, arranged in mosaic-like patterns; timing-based signals; anticipation of moving objects; different tuning for different types of motion; and responses when an expected stimulus was left out.

The prediction-focused model also accurately predicted how retinal ganglion cells from different animal species responded to natural images and movies. Together, the results suggest that the retina may be organized not only to compress incoming visual information, but also to provide the brain with information about what is likely to appear next.

Why prediction matters

The findings support a view of the retina as an active prediction system rather than only a camera-like compressor of visual input. That interpretation could help explain why the retina produces a range of timing- and motion-related signals before information reaches the brain.

Evidence and caveats

This is evidence from a computational spiking-neural-network model trained on natural movies, combined with comparisons between the model’s outputs and retinal responses across animal species. The model reproduced several established response features and predicted responses to natural images and movies, but the abstract does not establish that the biological retina uses exactly the same predictive mechanism. The study’s conclusion is therefore a model-based explanation of retinal function, rather than a direct demonstration of the mechanism in living retinas.

Twenty years of no-till farming improved Central Valley soil structure and water holding

A 20-year comparison in California’s Central Valley found that reduced tillage with cover crops produced more stable soil aggregates, greater water retention and different forms of carbon protection than standard tillage without cover crops. The conservation system also retained room for additional carbon storage, although the study measured potential rather than future gains.

1 hour ago·1 views

Biology

Human brain organoids classified Braille letters with up to 83% accuracy

Human forebrain organoids classified artificial tactile Braille signals after a sensor converted the touch data into electrical stimulation. Accuracy averaged 61% with one organoid and rose to 83% when responses from three organoids were combined.

2 hours ago·2 views

Biology

A peptide reduced persistent traumatic memories in mice

A study in mice found that a peptide targeting the memory-related protein CPEB3 reduced the persistence of traumatic memories and anhedonia, a loss of pleasure. The findings are an early genetic preclinical test of a possible approach for PTSD, not evidence of a treatment in people.

8 hours ago·5 views