Summary of Physiological Findings

The purpose of this section is to review central results on the anatomy and development of lateral connections in the cortex. Some of these results are still controversial, and the details will undoubtedly be refined as more experimental results are obtained. However, a fairly good understanding of the area already exists and can serve as a basis for computational modeling.

The essential properties of long-range lateral connections are outlined in the Computational and ... section. Although most of the data comes from the low-level visual areas, it is important to note that patterned lateral connections are not just a feature of the primary visual cortex: They are found at all levels of the cortical hierarchy and appear to be a general feature of cortical organization [2].

Extent:

Long-range lateral connections form a dense network within the cortex. Each connection extends over several millimeters and gives rise to clusters of axon endings at regular intervals ([12,14,37]; figure 1). The higher the area, the longer the range of lateral connections. For example in the primary visual cortex, where the receptive fields of neurons are very small and local, the average extent of long-range connections is approximately 1.5 mm, or 10% of the length of the cortical surface. In contrast in the inferotemporal cortex (area 7a), which is one of the highest visual centers and has large receptive fields, these connections can range up to 10 mm, and the average extent ranges between 44% and 90% of the cortical length [2]. Lateral connections appear to be recurrent, that is, if area A connects to B, then B connects back to A, although the evidence on this aspect is rather fragmentary at present.

  

Figure 1: Long-range lateral connections of a typical neuron in the visual cortex. Lateral connections, also sometimes called horizontal or intrinsic connections, run parallel to the cortical surface. In the visual cortex they run over distances covering several degrees of the visual field, and sprout branches at intervals (marked by arrows). The branches form a local cluster of connections to other cells in the region. Such clusters occur only in regions with similar functional properties as the parent cell. Adapted from [15].

Synapses:

The size of the dendritic arbor of pyramidal cells in layers 2 and 3 matches the axonal patches made by each horizontal connection in a cluster. Horizontal connection patterns in different layers (2, 3, 5, and 6) are aligned. The long-range horizontal connections have primarily excitatory synapses. About 70-80% synapse on excitatory pyramidal cells, while the remaining 20-30% seem to synapse on inhibitory interneurons [13,33]. Imaging studies and other measurements indicate a substantial amount of long-range inhibition in the cortex, more than predicted by the above 80-20 rule [17,18,19].

Patterns:

Long-range connections are clustered in patches whose distribution corresponds closely to the organization of receptive fields in the sensory map. For example in the mature visual cortex, lateral connections primarily run between areas with similar receptive field properties, such as iso-orientation columns [13,15]. The connections are elongated along the orientation axis of each neuron. In the immediate vicinity of each neuron, the patterns are relatively unspecific to orientation, but over larger distances, they are closely matched to orientation preference. To a lesser degree, the patterns are also shaped by other perceptual features such as ocular dominance and spatial frequency [3,9,27,28,43]. Under special conditions such as strabismus (squint-eye), the dependency on ocular dominance can become quite strong. In the strabismic cat, long-range connections run between ocular dominance columns, and their clustering is not influenced by orientation column patterns.

Development:

Immediately after birth, lateral connections grow exuberantly and to long distances in a short period. In the coming weeks, they get slowly pruned into well-defined clusters [7,8,26,29]. What process drives such pruning? Enormous amounts of genetic information would be required to specify each connection and each synaptic weight of the neurons in a cortical map. Instead, it appears that lateral connections develop in an activity-dependent manner. Several observations support this view:

1. When the primary visual cortex (of the cat) is deprived of visual input during early development, lateral connectivity remains crude and unrefined [8].

2. The patterns of lateral connection clusters can be altered by changing the input to the developing cortex. The resulting patterns reflect correlations in the input. For example, when a kitten is made strabismic, thereby removing correlations between the visual inputs in the two eyes, the lateral connections in the primary visual cortex organize differently, linking only the regions responding to the same eye [28].

3. In the mouse somatosensory barrel cortex, sensory deprivation (by sectioning the input nerve) causes drastic decreases in the extent and density of lateral connections in comparison to a normally reared animal [32].

These observations suggest that the development of lateral connections, like that of afferent connections, depends on cortical activity caused by external input, and represents correlations in the input. Moreover, the development of the two kinds of connections appears to be strongly related. The development of lateral connections happens at approximately the same time as the afferent connections organize into topographic maps [6,26]. Although each individual lateral connection is weak, their total effect on neural activity can be substantial [13], and thereby affect the development of afferent connections. Changes in afferent connections then change the activity patterns on the cortex, which in turn influences the organization of lateral connections. The development of both sets of connections thus appears to proceed synergetically and simultaneously, eventually evolving to a state of equilibrium in the adult animal.

Some of the above findings are still preliminary and will require further confirmation and more detailed quantification. However, several questions about the computational mechanisms and functional role of lateral connections can already be addressed, as will be discussed below.