Alan Saul Alan B Saul, Ph.D.

721-0695 (office)

721-0273 (lab)


Personal web pages

We investigate how the brain lets us see and learn, and how it gets so smart. We're particularly interested in how the brain processes time, with an emphasis on relatively slow changes. Our main technique involves recording from the retina (electroretinography, ERG) and from single neurons. We work with humans and other animals, asking questions like:

Other techniques we use include psychophysics, and modeling.


Our work has focused on the processing of time in the geniculocortical visual system. The best example comes from our work on the generation of visual cortical direction selectivity. Motion consists of changes in space over time. It's easy to obtain the needed spatial differences, since, starting at the retinal photoreceptors, different cells look at different parts of space. How does the brain obtain the temporal differences that are equally necessary? It turns out that cat visual cortex receives temporally diverse inputs from the lateral geniculate nucleus (LGN), and uses these to create direction selectivity. Some of this temporal diversity is created in retina, in the form of sustained and transient ganglion cells. However, LGN performs a radical transformation of its retinal inputs, as discovered by David Mastronarde. In nonlagged LGN cells, the retinal input is relayed to cortex largely unaltered, but in lagged cells inhibitory interneurons shift the phase of the input. This generates a full range of timing on the way to cortex.

We showed that the lagged and nonlagged inputs to cortex (40% and 60%, respectively, of the total X-type inputs) do in fact produce cortical direction selectivity. The characteristic way that phase changes with temporal frequency in the LGN cells shows up as a tendency for direction selectivity to be most prominent at low temporal frequencies. The lagged input targets lower layer 4, whereas the nonlagged input projects throughout layer 4.

Direction selectivity is present in kittens at the time of eye-opening, and a large majority of cat cortical cells are direction selective. Direction selectivity can be almost completely abolished, however, if kittens are reared in an environment illuminated solely by a strobe light flashing at 8 Hz. In these strobe-reared cats, lagged and nonlagged cells in LGN and cortical timing are seemingly normal, but the normal convergence of inputs with different timing onto single neurons is lost. We speculated that the synchronized stimulation provided by the strobe light, together with the differing latencies of lagged and nonlagged cells, leads to a decorrelation of activity in these cells, and thereby a dissociation of their synapses due to Hebbian mechanisms. We also showed that in normal kittens, timing matures quite slowly, and the immature timing produces a different pattern of the temporal frequency tuning of direction selectivity. In kittens, many cells are direction selective at high temporal frequencies. Our modeling efforts on this problem led to a general analysis of how poorly correlated inputs can be associated onto cells based on postsynaptic activity. The development of timing in LGN cells suggested that a class of large smooth neurons that develops late might be the substrate of the deficit in about 5% of the population of kindergarten-aged children who might be diagnosed with specific learning impairment due to a basic temporal processing defect. We intend to pursue this idea first by recording in thalamic slices.

The monkey visual system differs from the cat's in several important ways. The retina and lateral geniculate nucleus contain 3 classes of cells, called koniocellular, parvocellular, and magnocellular. DeValois and colleagues proposed that cortical direction selectivity in monkey primary visual cortex is generated by combining magnocellular and parvocellular inputs. We showed that, while this might happen in some cells, many direction selective cells appear to be dominated by magnocellular inputs. Monkey cortical cells obtain a range of timing from LGN, but the characteristic phase behavior as a function of temporal frequency differs markedly from that seen in cats, and monkey cells tend to be direction selective across a broad range of frequencies, or only at high frequencies. We have recorded from lagged cells in both parvocellular and magnocellular compartments, and confirmed many earlier findings (which have been sadly neglected) that magnocellular LGN has a diversity of timing.

Psychophysicists have classically used adaptation to dissect mechanisms underlying visual behaviors. Adaptation is a ubiquitous phenomenon in cortical cells, and we showed that adaptation aftereffects occur with meaningful specificity in both spatial and temporal domains. The most spectacular specificity shows up in the timing of aftereffects, rather than in the commonly-studied amplitudes. The response to the onset of a stimulus is delayed, but the offset of the response is unaffected. These observations provided evidence that inhibition is strongest between cells that prefer the same direction of motion. We proposed a model where excitatory and inhibitory inputs with different timing produce direction selectivity. Intracellular recording experiments by Ferster and colleagues and by Hirsch and colleagues provided further evidence for this model.

Self-motion, eye and head movements, does not prevent us from maintaining a stable view of the world. We speculate that general mechanisms contribute to this stability, as cells that respond to movement of the whole visual scene (produced by self-motion) give negative feedback to cells that respond to real movement of objects. These general mechanisms may be important throughout the brain for segregating figure from ground and meaning from context, and disturbances of these processes could occur in syndromes such as autism. One of the mechanisms for compensating for our movements could involve feedforward processes that depend on timing changes in the early visual system (in particular, LGN) evoked by eye movements.

At the core of all of our behaviors is sequencing in time. Even, or especially, behaviors that evolve over long periods such as years depend on the orders in which their components occur. Our brains create proper sequencing over these long periods. Just like visual cortical direction selectivity, this requires inputs that differ in timing by approximately quarter cycles. Our current work will test hypotheses about whether these differing timings exist, where they are found in the brain, and how behavioral direction specificity arises. We ordinarily assign the term memory to explain how we know what to do next, and these studies will hopefully flesh out this term with actual mechanisms.