Visual units in the central nervous system of the frog

A Midbrain Inspired Recurrent Neural Network Model for Robust Change Detection

We present a biologically inspired recurrent neural network (RNN) that efficiently detects changes in natural images. The model features sparse, topographic connectivity (st-RNN), closely modeled on the circuit architecture of a "midbrain attention network." We deployed the st-RNN in a challenging change blindness task, in which changes must be detected in a discontinuous sequence of images. Compared with a conventional RNN, the st-RNN learned 9x faster and achieved state-of-the-art performance with 15x fewer connections. An analysis of low-dimensional dynamics revealed putative circuit mechanisms, including a critical role for a global inhibitory (GI) motif, for successful change detection. The model reproduced key experimental phenomena, including midbrain neurons' sensitivity to dynamic stimuli, neural signatures of stimulus competition, as well as hallmark behavioral effects of midbrain microstimulation. Finally, the model accurately predicted human gaze fixations in a change blindness experiment, surpassing state-of-the-art saliency-based methods. The st-RNN provides a novel deep learning model for linking neural computations underlying change detection with psychophysical mechanisms.SIGNIFICANCE STATEMENT For adaptive survival, our brains must be able to accurately and rapidly detect changing aspects of our visual world. We present a novel deep learning model, a sparse, topographic recurrent neural network (st-RNN), that mimics the neuroanatomy of an evolutionarily conserved "midbrain attention network." The st-RNN achieved robust change detection in challenging change blindness tasks, outperforming conventional RNN architectures. The model also reproduced hallmark experimental phenomena, both neural and behavioral, reported in seminal midbrain studies. Lastly, the st-RNN outperformed state-of-the-art models at predicting human gaze fixations in a laboratory change blindness experiment. Our deep learning model may provide important clues about key mechanisms by which the brain efficiently detects changes.

Selective attention without a neocortex

Selective attention refers to the ability to restrict neural processing and behavioral responses to a relevant subset of available stimuli, while simultaneously excluding other valid stimuli from consideration. In primates and other mammals, descriptions of this ability typically emphasize the neural processing that takes place in the cerebral neocortex. However, non-mammals such as birds, reptiles, amphibians and fish, which completely lack a neocortex, also have the ability to selectively attend. In this article, we survey the behavioral evidence for selective attention in non-mammals, and review the midbrain and forebrain structures that are responsible. The ancestral forms of selective attention are presumably selective orienting behaviors, such as prey-catching and predator avoidance. These behaviors depend critically on a set of subcortical structures, including the optic tectum (OT), thalamus and striatum, that are highly conserved across vertebrate evolution. In contrast, the contributions of different pallial regions in the forebrain to selective attention have been subject to more substantial changes and reorganization. This evolutionary perspective makes plain that selective attention is not a function achieved de novo with the emergence of the neocortex, but instead is implemented by circuits accrued and modified over hundreds of millions of years, beginning well before the forebrain contained a neocortex. Determining how older subcortical circuits interact with the more recently evolved components in the neocortex will likely be crucial for understanding the complex properties of selective attention in primates and other mammals, and for identifying the etiology of attention disorders.

Color-Change Detection Activity in the Primate Superior Colliculus

The primate superior colliculus (SC) is a midbrain structure that participates in the control of spatial attention. Previous studies examining the role of the SC in attention have mostly used luminance-based visual features (e.g., motion, contrast) as the stimuli and saccadic eye movements as the behavioral response, both of which are known to modulate the activity of SC neurons. To explore the limits of the SC’s involvement in the control of spatial attention, we recorded SC neuronal activity during a task using color, a visual feature dimension not traditionally associated with the SC, and required monkeys to detect threshold-level changes in the saturation of a cued stimulus by releasing a joystick during maintained fixation. Using this color-based spatial attention task, we found substantial cue-related modulation in all categories of visually responsive neurons in the intermediate layers of the SC. Notably, near-threshold changes in color saturation, both increases and decreases, evoked phasic bursts of activity with magnitudes as large as those evoked by stimulus onset. This change-detection activity had two distinctive features: activity for hits was larger than for misses, and the timing of change-detection activity accounted for 67% of joystick release latency, even though it preceded the release by at least 200 ms. We conclude that during attention tasks, SC activity denotes the behavioral relevance of the stimulus regardless of feature dimension and that phasic event-related SC activity is suitable to guide the selection of manual responses as well as saccadic eye movements.

Frogs and toads in front of a mirror: lateralisation of response to social stimuli in tadpoles of five anuran species

Tadpoles of five anuran species were tested for preferences in the use of the eyes during inspection of their own visual image in a mirror. When tested in a tank with several small mirrors, tadpoles of five different species (Bufo bufo, Bufo viridis, Rana temporaria, Rana esculenta, Bombina variegata) preferentially approached and positioned themselves with the mirror located on their left side, thus looking at the image with the monocular field of their left eye. Similar results were obtained with tadpoles of R. temporaria tested in a simple task in which they had to choose approaching one or other of two large mirrors located on their left and right side. Control experiment showed that the behavioural asymmetry was not due to motor preferences and that it was independent of morphological asymmetries in the positions of the spiracles. This is the first demonstration of a functional visual lateralisation among juvenile amphibia before metamorphosis.