Anthony M Norcia

The visual system of the human, like other altricial species, undergoes an extensive period of postnatal development. Development is both quantitative (e.g., visual performance improves substantially with age) and qualitative (e.g., the visual system of the neonate displays a number of attributes that are absent in the adult). This chapter selectively reviews major aspects of quantitative and qualitative change. The emphasis is on functions that are thought to be the province of the early stages of the visual pathway, where the neural image is formed and primitive features of visual objects are first extracted. Large areas of activity in the field of infant visual development, such as object permanence, attention, and the recognition of objects and faces, are not covered. As with most reviews of human visual development, this review starts with a consideration of methodological issues unique to the field.

Research methods

Methodological considerations are of paramount importance in studies of human visual development. Because the behavioral repertoire is so limited (and constantly changing with development) and because infant observers cannot be instructed, specialized behavioral techniques have been developed (Dobson, 1990; Hainline, 1993; Hamer and Mayer, 1994). Electrophysiological measures, such as the visual evoked potential (VEP) require no overt behavioral responses other than fixation and accommodation. Because the VEP is generated in visual cortex, it is ideally suited for studying the early elements of the visual pathway, the development of which is the focus of this review.

The VEP is a far-field potential that is time- or phase-locked to the visual stimulus. It is this synchronization of the response to the stimulus that allows the investigator to design filters that can select out the visually driven activity from the background electroencephalographic (EEG) and other sources of biological and industrial noise. VEPs were first detected in the early 1950s by averaging the EEG record over fixed time intervals that were precisely aligned with the stimulus onset. The logic of averaging is that all activity that is not synchronized with the stimulus onset is random; thus, the average of these random signals will tend to zero in the limit of an infinitely large number of averaged records. The time-locked activity, by contrast, is not diminished by the averaging process. As the number of records in the average increases, the time-locked activity represents an ever-increasing proportion of the variance in the record.

The steady-state VEP When the stimulus is presented at periodic time intervals, the VEP can also be isolated using spectral analysis procedures. Periodic inputs (stimuli) lead to periodic outputs (responses), and the location of these responses in the spectrum is under the control of the investigator. This prior knowledge of what frequencies in the EEG are relevant (phase-locked at a frequency related to the input frequency) and which are irrelevant (non— phase-locked or of the wrong frequency) makes it possible to design frequency-specific filters that are highly effective in separating the signal from the noise.

Figure 13.1 A shows an example of a typical steady-state VEP (SSVEP) spectrum obtained in response to a low spatial frequency grating that was reversed in contrast at 7.25 Hz (14.5 pattern reversals per second). This spectrum is compared to that obtained with the subject viewing a blank screen (lower spectrum in Fig. 13.1A). The SSVEP is composed of several response peaks that occur at exact integer multiples of the stimulus frequency. In the case of the pattern reversal stimulus, the response is symmetrical for each reversal of the pattern, and the spectrum comprises even harmonics (2F, 4F, 6F. . . where F is the stimulus frequency). Because the stimulus is periodic, the response waveform is also periodic, as illustrated in Figure 13.15. The waveform is not sinusoidal and has two large positive and negative peaks corresponding to the frequency doubling of the response that is apparent in the spectrum.

The SSVEP differs critically from the EEG in that the phase of the SSVEP is constant over repeated presentations, whereas that of the EEG is random. This is illustrated in Figure 13.1 C, which plots the SSVEP in polar coordinates. The components of the SSVEP response are complex numbers; that is, the response has both an amplitude and a phase value. The phase of the response is nonzero owing to the presence of neural delays and integration processes. The fact that the response phases from individual trials cluster around a value of 180° in the figure (filled symbols) indicates that the response is synchronized with the stimulus. A phase angle of 180° is consistent with relative delays with respect to the stimulus of 34.5, 103.5, 148.0 . . . , ms for the 14.5-Hz response frequency. This ambiguity is due to phase being a modulo 2-pi variable. The phase of the EEG is random (see Figure 13.1 C, open symbols). If one averages the x values (real part) and the y values (imaginary part) and computes the magnitude of the resulting vector, the SSVEP has an amplitude that is not consistent with zero, given the error, but the EEG is consistent with zero (see

larger and smaller filled and open symbols in Figure 13.1 C, respectively).

The Swept-parameter VEP Spectral analysis of the SSVEP often results in signal-to-noise ratios that are sufficiently high to detect the response and to measure its amplitude and phase using relatively short data records; as the stimulus value is systematically varied (Regan, 1977). This technique, known as the swept-parameter or sweep VEP, is useful for measuring sensory thresholds-based extrapolation of VEP response functions (Campbell and Maffei, 1970).

It is a general property of sensory systems that evoked response amplitude increases as a function of stimulus intensity over a range of stimulus values. Figure 13.2 shows examples of response functions obtained in experiments where either the spatial frequency (leftpanel) or contrast (rightpanel) of a grating pattern was systematically varied over presentations lasting 10 seconds. In each case, the spectrum analysis was carried out at a series of 1-second intervals in order to map out the shape of the spatial frequency tuning or contrast response functions. Repeated measurements of the response functions were averaged as in Figure 13.1 (bottom), with all repetitions from the first second of the sweep being averaged together, all repetitions from the second being averaged together, and so on.

The SSVEP response is a roughly linear function of log-contrast (see Figure 13.2, right panel), suggesting that a contrast threshold can be estimated by extrapolation of this function to zero amplitude. This threshold extrapolation technique yields estimates of contrast threshold that are generally within a factor of 2 or 3 of adult psychophysical contrast thresholds (Campbell and Kulikowski, 1971; Cannon,

Figure 13.1. A, Spectra for steady-state VEP (data plotted above the x axis) and background EEG (data plotted below the x axis). The spectrum of the 7.25-Hz pattern reversal response comprises narrow peaks at frequencies that are even multiples of the stimulus frequency (e.g., 2F = 14.5, 4F = 29, 6F = 43.5 Hz). The response (signal) is larger than the EEG (noise) at each of these frequencies in the spontaneous EEG and at nearby frequencies in the record obtained during visual stimulation. B, Time averages for the steady-state VEP shown in A and for the spontaneous EEG. The time average is periodic, with a base period equal to twice the stimulus period of 138 msec. The averaged response to a blank screen is also shown, indicating the experimental noise level (line near zero). C, Amplitude and phase of the steady-state VEP for each of 16 10-second trials composing the averages for stimulus-present (.filled symbols) and EEG (open symbols) conditions previously shown in figure parts A and B. The VEP responses from individual trials cluster around a phase of 175°, indicating that they are synchronized with the visual stimulus. The spontaneous EEG shows a random phase relationship, with the resulting vector average (open large dot) being only a fraction of a microvolt.

Consistent with the emergence of photopigment in the outer segments, visual responses are first recordable at around 24 weeks in the VEP (Kos-Pietro et al., 1997; Taylor et al., 1987).

Each photoreceptor class has a characteristic spectral sensitivity; thus, measurements of spectral sensitivity and receptor isolation conditions have been used to determine when each class of photoreceptor becomes functional. Spectral sensitivity measurements with the scotopic VEP indicate that rods are functioning at 4 weeks of age (Werner, 1982). Similarly, the SWS cones show an adult spectral sensitivity by 4 weeks of age (Volbrecht and Werner, 1987). Photopic spectral sensitivity, measured with the VEP, is similar to that of the adult in 8-week-olds, suggesting that LWS cones are functional by this age (Bieber et al., 1995; Dobson, 1976). Knoblauch et al. (1998) found VEP responses to MWS- and LWS-cone isolating stimuli as early as 4 weeks of age, and Bieber et al. (1998) found adult-like action spectra for both MWS- and LWS-cone classes in 8- to 12-week-olds. Thus, all major photoreceptor classes are known to be functional by no later that 4 weeks of age.

Although all major photoreceptor classes are functional early in infancy, foveal cones are dramatically immature relative to more peripheral cones (Hendrickson and Drucker, 1992; Hendrickson and Yuodelis, 1984). At birth, foveal cones lack outer segments, whereas those in the parafovea are 30% to 50% of adult length. In the midperiphery cone, outer segments are 50% longer than they are in the parafovea, making them relatively adult-like. Despite the relative immaturity of the central retina, throughout development, visual acuity appears to be highest in the central visual field, compared to the peripheral field (Allen et al., 1996; Spinelli et al., 1983).

Spatiotemporal contrast sensitivity development

Perception of the visual world requires that the visual system must transduce light incident on the retina into neural signals. At the earliest stages, neural responses are dependent on the number of absorbed quanta. At the level of the inner plexiform layer, responses are recoded in terms of contrast, that is, spatial and temporal variations in light intensity. Spatiotemporal contrast forms the basis of all subsequent form processing, and contrast sensitivity forms a fundamental limit on what information is made available to visual cognition.

There is a long tradition, based on the theory of linear systems, of using sinusoidal gratings to measure contrast sensitivity (Campbell and Robson, 1968). In this framework, thresholds are measured as a function of spatial frequency (e.g., the number of grating cycles per degree of visual angle). Figure 13.3 shows the development of contrast sensitivity in human infants during the first 8 months of

0 0

Post a comment