Mechanisms of visual development

These behavioral data on vision and visual development immediately raise questions about neural mechanisms. What and where are the immaturities in the young infant's visual system that allow temporal resolution to be almost adult-like, grating acuity to be reduced by an order of magnitude, and the response to binocular disparity not to happen at all?

As discussed by Rakic (chapter 1 of this volume) and Bourgeois, Goldman-Rakic, and Rakic (chapter 4 of this volume), many postnatal anatomical changes are well documented in the visual system of infant primates; consider, for example, the period of intense synaptogenesis and the segregation of ocular dominance columns during the first two to three postnatal months in infant monkeys, the long slow development of the fovea (Yuodelis and Hendrickson, 1986), and the late maturation of local cortical circuits (Burkhalter, Bernardo, and Charles, 1993) documented in human infant tissue. Physiological development is much less well documented, with only a dozen or so studies in print at present.

Moreover, causal relationships between behavior and substrate are notoriously difficult to establish. Visual functions mature because the neural substrate matures, and the causes of functional maturation undoubtedly lie in neural maturation. But the length of the big toe matures too, yet we do not see it as causal in relation to the development of grating acuity. The puzzle is, which of the many immaturities of the neural substrate provide the critical immaturities (Brown, 1990) that limit a particular visual capacity at a particular age? Temporal coincidences become convincing causal stories only as a result of quantitative theoretical argument. The following paragraphs provide brief vignettes of some of the models and speculations that currently attempt to provide bridges between anatomy, physiology, and behavior.

Optics and Accommodation The optical quality of the primate eye is excellent at birth, and places no major limitations on visual acuity or other visual functions (see Banks and Bennett, 1988, for a review). Moreover, the development of accommodation (the capacity of the lens of the eye to change its focal length in order to focus on objects at varying distances) provides an example of an immaturity that is probably not a critical immaturity. Newborn infants typically do not accommodate differentially for objects at different distances (Haynes, White, and Held, 1965), and it might be thought that the resulting blur in the retinal image would contribute to their reduced grating acuity. However, the opposite direction of causality is more likely. That is, if the infant's visual system cannot process high spatial frequencies (see below), the infant has no need to accommodate, and no basis for generating signals that would lead to changes in accommodation.

Foveal Immaturities and Spatial Vision The profound and long-lasting anatomical immaturities of the primate fovea (Yuodelis and Hendrickson, 1986) provide a more likely critical immaturity for limiting infant acuity and contrast sensitivity. At birth, infants' foveal cones are coarsely packed, having very short outer segments and probably inferior waveguide properties.

These characteristics drastically reduce the foveal quan-tal catch in infants compared to adults, and should therefore (due to quantal fluctuations) reduce the overall sensitivity of infants' vision by tenfold or more (Banks and Bennett, 1988). In addition, the change in cone packing density (in combination with changes in eye size) affords a predicted change in spatial scale of about 1:4 between birth and adulthood, in remarkably good correspondence with the spatial scale shift seen in infant CSFs (see figure 6.6; Kelly, Borchert, and Teller, 1997). several exemplary quantitative models of the effects of foveal immaturity on infant acuity and contrast sensitivity have been developed (Banks and Bennett, 1988; Banks and Crowell, 1993; Brown, Dobson, and Maier, 1987; Jacobs and Blakemore, 1988; Movshon and Kior-pes, 1993; Wilson, 1988, 1993).

single unit recordings from monkey lateral geniculate nucleus (LGN) show that the LGN neurons with the best spatial resolution have acuities of only about a factor of 2 higher than the behaviorally measured acuities of infant monkeys (Blakemore and Vital-Durand, 1986; Hawken, Blakemore, and Morley, 1997). Thus most, but not all, of the limit on infant grating acuity and the reductions in sensitivity and spatial scale seen in infants' CSFs are probably caused at the earliest stages of visual processing.

Moreover, these losses of sensitivity and spatial scale will doubtless play through importantly to influence the development of many other visual functions. For example, Banks and Bennett (1988) have argued quantitatively that the known foveal immaturities are sufficient to predict the development of vernier acuity in infants. This conclusion is initially surprising, because the visual processing required to analyze vernier offsets in adults is often attributed to visual cortex. However, a function limited by a particular critical locus in the adult can nonetheless be limited by a critical immaturity at a different locus in the infant. An adequate model for the development of each visual function will have to incorporate the effects of foveal immaturities before an argument for more central critical immaturities can be maintained.

Temporal Processing The anatomical and physiological elements that control infants' temporal processing capabilities (CFFs and tCSFs) remain obscure, and data are scarce. The only available temporal response functions of single neurons in early infancy, obtained in neonatal vervet monkey LGN, peak at low temporal frequencies (Hawken, Blakemore, and Morley, 1997), and these authors conclude that the temporal resolution of single LGN cells increases considerably postnatally. These data would lead one to expect changes in temporal scale-rightward shifts of the tCSF-with develop

78 development ment, but no such shift is seen in the behavioral data (figure 6.7). The discrepancy between the physiologically based prediction and the behavioral facts remains to be reconciled.

M vs. P Pathways in Early Visual Processing Modern anatomical and physiological studies suggest the presence of (at least) two distinct pathways from the retina to cortical area V1 (for a review see Merigan and Maunsell, 1993). These are the M or magnocellular pathway, and the P or parvocellular pathway. Global models of visual development have occasionally been proposed in which the P pathway precedes the M pathway in development or vice versa. It seems more likely that each pathway has its own course of development, with different functional capabilities of each pathway developing at different rates.

in adults, M-initiated signals are thought to dominate the tCSF for luminance-modulated stimuli, while P-initiated signals are thought to dominate the tCSF for chromatic stimuli. on the basis of their observation of band-pass tCSFs for red/green chromatic modulation in infants, Dobkins, Lia, and Teller (1997) speculated that in the immature visual system, temporally modulated red/green gratings might be detected via M- rather than P-initiated signals. such an outcome suggests either that temporal resolution in the M pathway matures earlier than temporal resolution in the P pathway, or that the tCsF of P cells changes from band-pass to low-pass during development. in any case, the change in shape of the chromatic tCsF between infancy and adulthood is unique in the infant vision literature, and clearly invites study and modeling at the physiological level.

Stereopsis The development of stereopsis provides a final example. Anatomically, ocular dominance columns in primates are not fully segregated at birth, and continue to segregate over the early postnatal weeks or months (Hubel, Wiesel, and LeVay, 1977; Horton and Hocking, 1996). Thus, there is a general temporal coincidence between the end of anatomical segregation of ocular dominance columns and the onset of responses to large binocular disparities.

On this basis, Held (1985) proposed a model of the development of stereovision. in the adult, neurons in layer 4 of cortical area V1 are monocular, with some neurons driven by the left eye and others by the right. Held hypothesized that in very young infants, left-eye and right-eye inputs might converge too early, on the neurons in layer 4, making them binocular rather than monocular. This convergence was hypothesized to cause a loss of eye-of-origin information, and a conse quent disabling of disparity selectivity of neurons in the upper cortical layers. in this view, the onset of stereop-sis could be caused by the sorting out of left-eye and right-eye inputs in layer 4, enabling disparity detectors to work.

Surprisingly, it took 12 years before physiological data bearing on this hypothesis became available. using a binocular phase paradigm, Chino et al. (1997) showed recently that in infant monkeys, in contradiction to the hypothesis, both the ocular dominance properties and the disparity selectivity of neurons in V1 are adult-like at birth. The Chino study also confirmed earlier findings (Blakemore and Vital-Durand, 1981) that the sensitivity and spatial tuning properties of V1 neurons are immature (as they are at the LGN and presumably at the retina). Chino and colleagues argue that changes in sensitivity and spatial scale alone are sufficient in principle to produce changes in the infant's behavioral response to binocular disparity. For example, sensitivity to a fixed interocular phase difference, in combination with a shift of the CSFs of neurons toward higher spatial frequencies, will yield sensitivity to smaller absolute disparities.

A quantitative theoretical account of how the early changes in sensitivity and spatial scale should play through to influence stereoacuity remains to be developed. Such a theory might suggest that foveal immaturities should yield quantitatively different developmental time courses for grating acuity and disparity sensitivity (as Banks and Bennett, 1988, argue for grating acuity and vernier acuity). But even so, it is difficult to see how the same changes in sensitivity and scale could allow both the negligible changes of grating acuity and the very rapid improvement of stereoacuity seen in postnatal months 3-6 (figures 14.2, 14.3, and 14.5).

it seems likely that a second critical immaturity will be needed to account for the rapid development of ste-reopsis. This mechanism did not reveal itself in Chino's binocular-phase-based study of V1 neurons. The rapid development of stereoacuity (like the change in the shape of the chromatic tCSF) cries out for further theoretical and experimental analysis, particularly with stimuli that more closely resemble the stimuli used in the behavioral studies.

in sum, in a curious modern repetition of the history of visual science at the adult level, we know much more about the development of visual functions than we do about the development of visual physiology. it seems certain that the tentative causal stories presented here will be rewritten as more extensive physiological data become available, and as more and more sophisticated experimental paradigms are used (e.g., Chino et al., 1996; Movshon et al., 1997; Hatta et al., 1998).

Was this article helpful?

0 0

Post a comment