Engineers use neural networks to control systems too complex for conventional engineering solutions. To examine the behavior of individual hidden units would defeat the purpose of this approach because it would be largely uninterpretable. Yet neurophysiologists spend their careers doing just that! Hidden units contain bits and scraps of signals that yield only arcane hints about network function and no information about how its individual units process signals. Most literature on single-unit recordings attests to this grim fact. On the other hand, knowing a system's function and describing it with elegant mathematics tell one very little about what to expect of interneuronal behavior. Examples of simple networks based on neurophysiology are taken from the oculomotor literature to suggest how single-unit interpretability might decrease with increasing task complexity. It is argued that trying to explain how any real neural network works on a cell-by-cell, reductionist basis is futile and we may have to be content with trying to understand the brain at higher levels of organization.

One of the most intriguing and controversial observations in oculomotor research in recent years is the phenomenon of express saccades in monkeys and man. These are saccades with such short reaction times (100 msec in man, 70 msec in monkeys) that some experts on eye movements still regard them as artifacts or as anticipatory reactions that do not need any further explanation. On the other hand, some research groups consider them not only authentic but also a valuable means of investigating the mechanisms of saccade generation, the coordination of vision and eye movements, and the mechanisms of visual attention.

This target article puts together pieces of experimental evidence in oculomotor and related research – with special emphasis on the express saccade – to enhance our present understanding of the coordination of vision, visual attention, and the eye movements subserving visual perception and cognition.

We hypothesize that an optomotor reflex is responsible for the occurrence of express saccades, one that is controlled by higher brain functions involved in disengaged visual attention and decision making. We propose a neural network as the basis for more elaborate mathematical models or computer simulations of the optomotor system in primates.

Eye movements were recorded with the scleral search coil method while striate cortex (area 17) was stimulated in alert cats with their heads fixed. Regardless of where stimulation was applied in the retinotopic map, eye position at the onset of stimulation strongly affected the amplitudes of evoked saccades, but had much less influence on their directions. Application of long stimulus trains evoked repeated saccades at all sites tested. Highly convergent or goal-directed saccades were not observed. Cortically evoked saccades appeared to habituate with repeated stimulation and had higher thresholds and longer latencies that those reported for saccades evoked from the superior colliculus. The directions of cortically evoked saccades generally agreed with those predicted from the retinotopic coordinates of the stimulus sites, but saccade amplitudes were usually lower than expected. It is suggested that these findings are consistent with certain characteristics of eye-head coordination in the cat's normal visual orienting behavior. The results are difficult to reconcile with the hypothesis that goal-directed saccades are a normal response to targets outside the cat's oculomotor range.

The activity of parietal cortex neurones primarily related to eye position (EP neurones) was studied in macaque monkeys with the aim of precisely defining the neurones' gaze fields (GF) and comparing them in two functionally different areas, MSTd and 7a. Discharge rates of single neurones in the inferior parietal lobule and in the underlying cortex of the superior temporal sulcus were recorded in two Java monkeys while the animals fixated a steady visual target positioned at several different points on a video screen. The GFs were then drawn as a regression surface fitting the mean discharge rates. Cells tonically influenced by the angle of gaze were found in both areas. The GFs most often took the form of a nearly planar surface best characterized as a ramp tilted towards a hemifield or quadrant of the visual field, shifted eccentrically with respect to the straight ahead (primary position), and with a midpoint centred between 0 deg and 20 deg of gaze eccentricity and saturation between 10 deg and 35 deg. In a minority of cases, the discharge rate was nearly maximal at the primary position and decreased to a minimum within 35 deg of eccentricity. In other instances, the GFs were peaked surfaces, limited to a restricted part of visual space. EP neurones, while showing similar gaze fields in areas MST and 7a, were found intermingled with functionally different types of cells. The results suggest that EP neurones similar to those already described in several areas of the monkey parietal cortex are present also in area MST. These cells, by signalling the degree of gaze eccentricity from the primary position, encode gaze position in an orbito-centered frame extending up to 30–35 deg from the straight-ahead. The role of EP neurones might be to supply contiguous elements with a gaze eccentricity signal required for visuo-motor processes such as the control of tracking movements.

Two discrete areas in frontal cortex are involved in generating saccadic eye movements—the frontal eye field (FEF) and the supplementary eye field (SEF). Whereas FEF represents saccades in a topographic retinotopic map, recent evidence indicates that saccades may be represented craniotopically in SEF. To further investigate the relationship between these areas, the topographic organization of afferents to FEF from SEF in Macaco mulatto was examined by placing injections of distinct retrograde tracers into different parts of FEF that represented saccades of different amplitudes. Central FEF (lateral area 8A), which represents saccades of intermediate amplitudes, received afferents from a larger portion of SEF than did lateral FEF (area 45), which represents shorter saccades, or medial FEF (medial area 8A), which represents the longest saccades in addition to pinna movements. Moreover, in every case the zone in SEF that innervated lateral FEF (area 45) also projected to medial FEF (area 8A). In one case, a zone in rostral SEF projected to both lateral area 8A from which eye movements were evoked by microstimulation as well as medial area 8A from which pinna movements were elicited by microstimulation. This pattern of afferent convergence and divergence from SEF onto the retinotopic saccade map in FEF is indicative of some sort of map transformation between SEF and FEF. Such a transformation would be necessary to interconnect a topographic craniotopic saccade representation in SEF with a topographic retinotopic saccade representation in FEF.

Ciliary ganglia from the pigeon, cat, and monkey were investigated for the presence of NADPH-diaphorase reactivity by use of a standard histochemical method. In the pigeon, where the ganglion is known to control lens and pupil function, and the choroidal vasculature, about one-third of the ganglion cells were densely stained and most other somata were lightly stained. In some cases, preganglionic terminals with a cap-like morphology were also darkly stained. The pattern of NADPH-diaphorase staining in mammals was very different from that seen in pigeons. In both mammalian species, where the ganglion is known to control lens and pupil function, a small number (less than 2%) of the ganglion cells were shown to be densely NADPH-diaphorase positive, revealing their neuronal processes. The presence of NADPH-diaphorase positive cells in pigeon, cat, and monkey ciliary ganglia suggests that nitric oxide may be used for intercellular communication in this ganglion, or in light of the known importance of nitric oxide in vascular control, some of these positive neurons may participate in the control of choroidal vasodilation.

A countermanding paradigm was utilized to investigate the regulation of saccade initiation. Two rhesus monkeys were instructed to generate a saccade to a peripheral target; however, on a fraction of trials after a delay, the monkeys were signaled to inhibit saccade initiation. With short delays between the presentation of the target and the signal to inhibit saccade generation, monkeys withheld saccades to the peripheral target. As the delay of the stop signal increased, monkeys increasingly failed to withhold the saccade. The hypothesis that the generation of the saccade is determined by a race between a go and a stop process provides three explicit means of estimating the covert latency of response to the stop signal. This latency, known as stop signal reaction time, was estimated to be on average 82 ms for both monkeys. Because the stop signal latency represents the time required to exert inhibitory control over saccade production, the countermanding paradigm will be useful for studying neural mechanisms that regulate saccade initiation.

Nitric oxide (NO) is a neuronal messenger molecule that mediates pathway refinement in some brain regions. We used nicotinamide adenine dinucleotide phosphate diaphorase (NADPHd) histochemistry to examine the development of NO expression in the superior colliculus (SC) of kittens aged E28–E58 and P2–P57 and adults in order to determine if NO expression is correlated with pathway refinement. At E28, labeled cells were seen only within the subventricular zone (SVZ). At E36–E41, labeled cells were also found within the deep gray layer (DGL) of SC. At E51 and E58, a few labeled neurons were also present in the intermediate gray layer (IGL). These neurons already had extensive dendritic fields and well-developed morphologies at the time that they first expressed nitric oxide synthase (NOS). The number of neurons labeled in the DGL and IGL increased postnatally, reaching a peak density between P14 and P35. Neurons within the optic (OL) and superficial gray layers (SGL) were first visible at P7 and increased slightly in number until adulthood. However, SGL-labeled neurons were relatively limited in number and lightly labeled at all ages examined. We conclude that (1) NADPHd expression occurs in SC beginning in the second trimester in kittens and progresses in a ventral to dorsal pattern between E36–P35; (2) few neurons in kitten SGL are labeled by NADPHd and these appear relatively late in postnatal development; and (3) there is no correlation between NOS expression and retinocollicular pathway refinement in kittens, a result different from that seen in rodents.