The development of vernier acuity in human infants

The development of vernier acuity in human infants

0042-6989/92 $5.00 + 0.00 Copyright 0 1992 Pergamon Press Ltd Vision Res. Vol. 32, No. 8, pp. 1557-1564, 1992 Printed in Great Britain. All rights re...

1MB Sizes 1 Downloads 35 Views

0042-6989/92 $5.00 + 0.00 Copyright 0 1992 Pergamon Press Ltd

Vision Res. Vol. 32, No. 8, pp. 1557-1564, 1992 Printed in Great Britain. All rights reserved

Research Note The Development

of Vernier Acuity in Human Infants

JOHANNES ZANKER,* GESINE MOHN,ta URSULA KARIN ZEITLER-DRIES&t MANFRED FAHLEt Received 29 August 1991; in revised form 20 November

WEBER,?

1991

Vernier acuity, i.e. the detection of a small misalignment between lines, is about one order of magnitude finer than the resolution of periodic gratings in adult humans. This hyperacuity is generally attributed to cortical mechanisms, and the time-course of its development seems to differ from the development of grating resolution that probably is limited by retinal factors. We investigated 271 human infants and children between 2 months and 8 yr of age with essentially identical stimuli and experimental procedures. Vernier thresholds for Vernier targets were compared to grating resolution. The preferential looking experiments led to the following results: (i) Vernier acuity starts below grating resolution. (ii) Like grating resolution, Vernier acuity develops gradually, but more rapidly and longer; at the age of 5 yr performance becomes comparable to that of adults. (iii) Flanking borders without offset, added to the Vernier targets at various distances, did not affect thresholds consistently across distances and age groups. Development

Vernier acuity

Grating resolution

INTRODUCTION Since the pioneering work of Fantz (1961) our knowledge about the development of human visual perception has expanded considerably. It is clear now that newborns have a rather poor visual resolution; but the overall performance of the visual system develops rapidly, apparently in parallel with the maturation of eye and brain (reviewed by Mohn & van Hof-van Duin, 1991). Much effort was put into answering the question of which specific visual function can be related to which subcortical or cortical part of the developing CNS (for reviews, see Atkinson, 1984; van Sluyters, Atkinson, Banks, Held, Hoffmann & Shatz, 1989). One spatial vision task of particular interest is hyperacuity because there is evidence that it is specifically linked to cortical processing (McKee & Levi, 1987). For instance, the threshold for the detection of small misalignments between two lines in Vernier stimuli is below the diameter and spacing between fovea1 receptors (Westheimer, 1979). Such precision for positional estimates can be understood considering the light distribution of Vernier stimuli generated by diffraction on the retina. The relative activation of neighbouring receptors, looking at the flanks of the almost Gaussian intensity distribution, depends strongly on the relative position of the stimulus. This information could be used by the brain to calculate

*Max-Planck-Institute for Biological Cybernetics, Spemannstrak 38, D-7400 Tiibingen 1, Fed. Rep. Germany. TDepartment of Neuroophthalmology, University Eye Hospital, 7400 Tiibingen, Fed. Rep. Germany. “R

32,8-G

Preferential looking

Crowding

the position with high precision, limited by the signalto-noise ratio of the neuronal signal (Barlow, 1979; Westheimer, 1979; Wilson, 1986). Despite the general relevance of hyperacuity for our understanding of vision, relatively little is known about the development of Vernier acuity. In a longitudinal study of infants between 0.5 and 6 months of age, Manny and Klein (1984) found consistently higher Vernier acuity for static Vernier offsets in a grating, as compared to grating acuity. However, the grating acuity they reported was much lower than the values usually observed. This, as well as the fact that they modified the stimulus conditions between the age groups, makes their results difficult to interpret. Later, the same authors (Manny & Klein, 1985) using a three-alternative tracking paradigm, and moving Vernier offsets, found higher Vernier acuity, without measuring grating acuity in the same infants. Shimojo, Birch, Gwiazda and Held (1984) presented, in a forced-choice preferential looking (“FPL”) paradigm, vertical stripe patterns with a small horizontal offset, moving up and down, versus gratings without offset. They found similar thresholds for grating and Vernier acuity at the age of 2 months but with increasing age, Vernier acuity improved over grating acuity. At 9 months, Vernier acuity reached a displacement threshold of 2.5’, whereas grating acuity was confined to a bar width of 5’. Thus the rate of Vernier acuity development resembled that of stereo acuity (Birch, Gwiazda & Held, 1982). In a subsequent study, however, using an improved method allowing for larger Vernier displacements, Shimojo and Held (1987) found Vernier acuity being worse than grating acuity for 1557

RESEARCII

15%

younger infants and better at the age of 5 months. due to a significantly higher speed of development for Vernier acuity. Taken together, the results of these studies are puzzling in that they differ from each other in the absolute levels of Vernier thresholds and in the rate of development. as compared to the development of grating acuity. Unfortunately these behavioural data cannot be compared directly to VEP recordings, the other major source for data from nonverbal infants. This is mainly because, in contrast to adults, the specificity of VEPs to Vernier stimuli is not very clear in infants up to 13 months (Levi, Manny, Klein & Steinman, 1983; Manny, 1988). Animal data are sparse, though the development of Vernier acuity in infant monkeys (Kiorpes & Movshon, 1987; Kiorpes & Movshon, 1988) appears to resemble that observed by Shimojo and Held (1987). One of the problems with the experiments described in the literature investigating development of Vernier acuity is that they all used repetitive’ Vernier targets increasing the stimulus efficiency. It is not clear how the lateral interaction between the stripes of such gratings with misalignments can affect the results. Indeed, some deterioration was found for Vernier gratings in adults (Levi, Klein & Aitsebaomo, 1985). This can be interpreted as a crowding effect, which is typically described as decreased acuity for closely spaced symbols as compared to isolated targets. In adults, crowding effects are maximum at a contour separation of three times minimum angle of separation (Flom, Weymouth & Kahneman, 1963; Jacobs, 1979), but the area of interaction is larger for additional lines flanking Vernier targets (Westheimer & Hauske, 1975; Levi ef al., 1985). It is not clear whether crowding in children of 3 yr or older is comparable to the effects observed in adults; it seems to depend strongly on the stimulus actually used (Haase & Hohmann, 1982; Fern, Manny, Davis dc Gibson, 1986). For younger children no data are

SOTE

available. Thus we do not know how lateral interactions influence the detection of Vernier displacements in young infants, and how the critical distance for such a spatial interaction develops. In our report on the development of Vernier acuity three experimental aspects are emphasised. (i) In a cross-sectional study it was attempted to test a broad range of ages leading to a comprehensive developmental curve of Vernier acuity. (ii) Isolated Vernier targets were used to measure Vernier thresholds without any further spatial interaction. (iii) Flanking bars were added to the Vernier targets 10 investigate lateral interactions between neighbouring stimulus contours. MATERIALS

AND METHODS

Subjects

271 infants in the age between 2 months and 9 yr, and 5 adult control subjects (between 28 and 37 yr old, average 31 yr), were investigated in the present study. The infants were pooled into 13 age groups. For the first 8 groups a deviation of 10% from the average age (2, 3, 4, 5, 6, 8, 12 and 18 months) was tolerated, whereas the older children were collected into adjacent age intervals (2-3, 3--4, 4-5, 5-7, 7-9 yr). Some of the infants did not finish the experiments and were excluded from the results because they became inattentive or dizzy or started crying, and thus did not approach a threshold plateau in the staircase procedure (see below). The number of children investigated in each group and the corresponding success rates are given in Table 1. The children were recruited from the birth ward of the University Hospital, from friends, and by means of leaflets distributed in local kindergartens. Experimental setup

The infants were seated on the accompanying parent’s lap in front of a grey metal wall. Two windows of

TABLE I. Age groups for the experiments, number of subjects tested in each group, and number of subjects who completed at least 2 runs of the staircase procedure. Also, the number of subjects of each age group which were averaged for the data plots for the single bar Vernier stimulus, as well as for the multiple edges Vernier stimulus with 2, 4 and 8 mm bar distance, and grating acuity is given Number Distance Age 2 months 3 months 4 months 5 months 6 months 8 months I2 months I8 months 2 -3 yr 3 4yr 4-5 yr 5-7 yr 7 -9 yr Adults

Procedure

(cm)

FPL FPL FPL FPL FPL FPL FPL OPL OPL OPL OPL 2AFC 2AFC 2AFC

47 47 47 41 40 40 40 100 100 100 100 200 200 200

The viewing distance looking paradigm and older subkts

._-Overall 23 22 22 22 23 22 23 24 22 I6 IS I9 IX 5

Successful

(% )

Single

20 20 20 20 I4 13 9 IO I2 I3 I3 I6

87.0 90.9 90.9 90.9 60.9 59. I 39. I 41.7 54.5 81.3 86.7 X4.2 83.3 100.0

IO IO IO IO IO IO IO IO IO IO IO

IS 5

of subjects 2mm

4mm IO

IO IO 5 5 5 4 7 3 6

I0

8

II 5

10 5

IO IO IO 10 IO 7 9 7 I2 II 12 Y 5

8mm IO IO IO IO 7 5 4 7 3 3 3 x II 5

Acuity

I5 I5 I5 I8 I3 Y I6 14

and the experimental procedure is listed: whereas young infants were tested in a forced-choice preferential (FPL), an “operant” preferential looking paradigm (OPL) was used for infants between I8 months and 5 yr, made their own decisions in a two-alternative forced-choice paradigm (ZAFC).

RESEARCH

12 x 16 cm, with their centres 28 cm apart, had been cut into the wall and covered with milk glass screens on which stimulus patterns were projected from behind by means of two slide projectors. Rotatable mirrors placed in the projection paths allowed to oscillate both stimulus images in synchrony up and down (i.e. orthogonally to the Vernier offset, see below) with an average speed of 3 cmjsec and a period of about 1 sec. The viewing distance was 40-47 cm for the youngest infants, and had to be increased up to 200 cm for the older subjects, in order to produce sufficiently small Vernier displacements (cf. Table 1). In some experiments with children beyond 5 yr of age, identical angular stimulus dimensions were achieved by adjustment of magnification of the slide projectors, while keeping the viewing distance below 200 cm. These changes in distance and projection scale had minor effects on the mean luminance of the stimuli which were partly corrected by neutral density filters inserted into the ray paths. The remaining variability in mean luminance was < 15% and may be neglected because infants’ spatial processing appears to be little affected by luminance increases above lOcd/m’ (reviewed in Mohn & van Hof-van Duin, 1991). The infants were allowed to watch the screens freely with both eyes, whereas the parent’s view was occluded by means of a curtain. The experimenter observed the subjects on a monitor connected to an i.r. video camera looking through a small hole below the two stimulus screens. The experimenter evaluated the fixational response of the infant and entered the result with the mouse into the computer which controlled the experiment. The room was illuminated just by the stimuli, the video- and computer-monitor, and a dim light illuminating the keyboard and the face of the subject. Stimuli

The basic stimulus, the “single bar Vernier”, was a bright vertical bar (bar width s was 8 mm on the screen, corresponding to 1” at 47 cm distance) on dark background in which two segments (each 30 mm long) just above and below its centre (30 mm apart) were displaced laterally. For the staircase procedure (see below), the lateral displacement was decreased in one octave steps between 16 and 0.03125 mm (see Fig. 1, top row, for displacements between 16 and 0.125 mm). The choice alternative stimulus (see below) was a straight bar without displaced segments. In order to investigate possible lateral interactions between the Vernier stimulus and flanking contours, parallel bars were added on each side of the Vernier bar, limiting the maximum Vernier displacement to the size of the dark gap between the bright bars. The inner edges of these additional bars followed the displacement of the central bar, whereas the outer edges were kept straight. Thus “multiple edges Vernier” stimuli consisted of four contours with Vernier displacements flanked by two contours without Vernier displacements (see Fig. 1, bottom rows). The distance d between the central bar and the flanking bars, i.e. the width of the black gaps between the bright bars, was set to 8, 4 or 2 mm corresponding to the range between

NOTE

1559

d = s and d = s/4. For 2 and 3 month old babies the flanking distance was increased to d = 16 mm, because

this allowed for bigger Vernier displacements. Three straight vertical bars without any lateral displacement were used as control stimuli for the multiple edges Vernier targets. The luminance of the bright bars was 570 cd/m2 and of the dark background was 6 cd/m’, the Michelson contrast thus amounted to 98%. Procedure

The basic experimental procedure was a forced-choice preferential looking paradigm (FPL). On each of the two screens a stimulus image appeared, one with the Vernier target and the other with the control stimulus, as described above. The position of the target was randomised by the computer and not known to the experimenter who was observing the infant. Large Vernier offsets elicited clear preference of fixation, thus the experimenter could conclude solely from the eye and head movements of the infant on which side the target was presented. This decision was entered into the computer and compared with the actual position, and the observer was informed by an acoustical feedback signal whether (s)he was right. To get a fast estimate of the displacement threshold, a rapid staircase procedure was applied (cf. Rose, Teller & Rendleman, 1970). Initially, four correct decisions were required to proceed with the experiment, in order to reduce the probability that the staircase was entered just by chance. Thereafter, the Vernier displacement was halved, when two correct decisions were made in a sequence. After each false decision the Vernier displacement was doubled. Thus an equilibrium corresponding to 7 1% correct responses was approached which was estimated as the average displacement presented in the stimuli between the second and fourth change from going up to going down in the staircase procedure. Individual thresholds were averaged for several subjects and are plotted as means with their 95% confidence intervals. For the four youngest age groups, there were doubts on the reliability of some individual threshold estimates, because the subjects remained close to the base level of the staircase. Since it was taken care that the poor performance of these individuals was not due to reduced awareness to the stimuli, the data were pooled for all individuals of each of these age groups. The 75% threshold was then estimated from the psychometric curve by means of Probit analysis (Finney, 1962). For strong stimuli, these estimates were very close to the averaged thresholds which may justify a cautious usage of all data from very young infants. In order to account for the increasing visual and cognitive capacity of the infants, the basic procedure had to be varied for older infants. Besides the increase of viewing distance mentioned above, the FPL procedure was extended to an operant-like procedure (OPL) for children between 18 months and 5 yr. In order to stimulate and keep the infants’ attention, colour slides of popular toys appeared on both windows after two consecutive correct responses. Older infants were

encouraged to point at the target, though the experimenter’s decision was still based on the eye movements, which appeared to be more reliable than the children’s direct reports. From 5 yr of age on, the children were allowed to press the mouse buttons by themselves. This increased their motivation and concentration considerably. If possible, at least two thresholds for two stimulus conditions were obtained for each subject. Some older children and adult controls allowed us to measure thresholds for all four sets of stimuli. Typically, about 25- 35 stimulus pair presentations were required to measure one threshold, which usually took < 15 min of time. Not all of the infants finished the staircase procedure properly; especially children around 1 yr of age were difficult to test, mainly due to their booming motoric activity. Threshold estimates were excluded from averaging when babies had become fussy or inattentive. and

correspondingly their staircase didn’t approach an equilibrium of going up and going down. In two cases no reliable threshold was found because ofapparatus failure. The success rate for each age group is shown in Table I. After the Vernier experiment, grating acuity was assessed using a derivative of Teller Acuity Cards (Mohn. van Hof-van Duin, Fetter, de Groot & Hage. 1988). if the subjects were still cooperative. For the adult controls and for children older than 4 yr, conventional optotype charts were used to measure a grating acuity equivalent. RESULTS

The development of Vernier acuity for single targets, and for grating acuity is plotted in Fig. 2. The ordinate is scaled in min of arc with lower values, better thresholds, up. and the right ordinate in c/deg gratings, with a 30c/deg threshold, which means a

FIGURE I. In “single bar Vernier” stimuli, two segments were cut out of a bright vertical bar and displaced laterally. The width s of the bar on the screen was 8 mm, i.e. 1’ at a distance of 47 cm. A straight vertical bar was used as a control stimulus in the other screen window. In the “multiple edges Vernier” stimuli, additional bars were introduced on either side of the centre bar carrying the Vernier displacement. The outer edges of the additional bars were straight whereas their inner edges followed the lateral displacement of the centre bar. The distance d between the central and the flanking bars (dark area) was varied between 2s and s/4. The stimuli with d of 8, 4 and 2 mm are shown in the three bottom rows. For these tests. the control stimulus consisted of three straight vertical bars.

bar left i.e. for bar

RESEARCH

1561

NOTE

-60

G

1 .o

-30

8

2.0

-15

2 Y

1 5.0 --- .---___-10.0 1

5

2

x

*

x

.__A--__

2

______ .____----___

5

10

20 age

1+

single bar

50 [months]

-X

gratings

,-6

.c

-3

s

1 0.6 t

100

FIGURE 2. The development of the single bar Vernier acuity (squares) is compared to grating acuity (crosses). The error bars indicate 95% confidence intervals. The broken horizontal lines indicate the acuity range reported by Shimojo and Held (1987) for infants between 2 and 5 months of age. During the first months of life, grating acuity is higher than Vernier acuity, whereas from 54 months on it is only half as fine. Adult values of Vernier acuity (rightmost data points) are reached within about 5 yr, considerably later than the maturation of grating acuity. The development curves for grating and Vernier acuity intersect at about 15 months, comparatively late when compared to literature data that were collected for repetitive Vernier stimuli.

width of 1 min, corresponding to a displacement threshold of the same size. This way of plotted the data is somehow deliberate because it combines different performances which are obtained with different psychophysical techniques. But it follows common way (cf. Shimojo & Held, 1987) to compare the thresholds for a resolution task and a positional task. The error bars indicate the 95% confidence intervals for each data point. Despite the changes of some details in the experimental procedure and of the viewing distance, a continuous, gradual development of Vernier acuity can be observed throughout the first 5 yr of life. Vernier thresholds start off with about 24’ at the age of 2 months and develop rather steadily for about 5 yr, reaching roughly 0.2’. This threshold is close to that of adult controls under the same experimental conditions (0.14’). For comparison grating acuity was measured in the same subjects if possible. Since the attention of the youngest infants often faded out after the Vernier tests, the first four values of grating acuity were taken from other studies of our group which were obtained with the same acuity cards (Mohn et al., 1988, and unpublished data). These values are close to those published by other groups (for review, see Mohn & van Hof-van Duin, 1991), but the small discontinuity between 5 and 6 months indicates that the actual grating acuity for 2-5 months of age might by slightly higher. Thus they can be treated as a conservative estimate for the comparison with the even lower Vernier acuity. It could be argued more generally that grating resolution was measured under different experimental conditions than Vernier acuity. For instance, Vernier stimuli were oscillated, whereas acuity cards were presented by the experimenter at a fixed position. However, the experimenter does not stabilise the cards perfectly, nor the baby its eyes, and thus there is a considerable amount of relative motion of the grating image the retina. Also the psychophysical procedure of Teller Acuity Cards (McDonald, Dobson, Sebris, Baitch, Varner & Teller, 1985) differs

from the FPL procedure used to determine Vernier acuity, but acuity cards give estimates well within the range obtained by FPL (e.g. see Mohn, van Hof-van Duin, Groenendaal, Fetter de Groot & Hage, 1990). Furthermore, it should be noted that the older children were tested with optotypes and many of them reached maximum values on the cards from 5 yr on. Hence, the last two values, as well as the adult control, might be a slight underestimation of the true acuity due to the restricted viewing distance. Despite all these reasons for caution, the grating acuity development curve plotted in Fig. 2 can be used as a rough estimate for comparison with Vernier acuity, because the major effects are obvious and much larger than possible deviations caused by these experimental limitations. The data show that the development of grating acuity is not as fast as that of Vernier acuity. This is reflected by the fact that the slope of a linear regression line through the grating data points plotted on a double log scale as in Fig. 2 [0.74 log (arcmin-‘)/log (month), 95% confidence interval 0.56-0.771 differs significantly from that through the Vernier data points [1.30 log (arcmin-‘)/log (month), 95% confidence interval 1.13-l .47]. Furthermore, thresholds for grating resolution are better during the first year of life (15’ at the age of 2 months, for instance), but are worse than Vernier acuity after about 4 yr (l.l’-0.8’ between 4 and 8 yr). The last two measures (5-9 yr) are very close to adult levels (0.8’). Most important, in consequence of these different developmental speeds, the two curves intersect. The regression lines fitted to the two data sets on a double log scale intersect at an age of 15 months, with a broad 95% confidence interval (estimated graphically) from 6 to 30 months. It should be remembered that this estimate of the age at which the two acuities are about equal, is influenced by the choice of the experimental techniques and depends on the conventions of plotting the different sets of data mentioned above, and thus should not be treated as quantitative measure.

RESEARCH

I St2

There is one age group, namely the infants of I yr, which differs from the general pattern of continuously increasing Vernier acuity. Toddlers between i and 2 yr of age are very difficult to investigate, because they prefer exploring their surroundings actively rather than watching boring stripes. Performing well in the Vernier experiments requires a considerable amount of motivation which is very limited in these subjects, thus the deterioration of Vernier acuity in 1 yr olds should not be treated as reflecting the true visual capacity of this age group. Interestingly, during this ““difficult” period, female subjects tend to perform better than male subjects, perhaps due to gender differences in concentration versus motor activity. With 18 month olds the difference is strongest, when the threshold of girls is 1.1’ ( f 0.2 SEM) whereas boys just reach 3.4’ (+ 1.1 SEM). This resembles an observation of Held et al. (1984) that females showed significantly higher levels of Vernier acuity in early development. However, big differences cannot be observed in other groups of our sample (data not shown), and compared over the whole range of ages no significant effects can be found between genders (paired l-test, P ,> 0.1). When flanking bars are added to the Vernier targets the basic development resembles that for the single bar Vernier stimulus (Fig. 3), Vernier acuity develops rapidly, starting at values worse than grating acuity, and leading to values better than grating acuity. It is obvious from Fig. 3 that the comparison between the acuity for single bar and multiple edges Vernier stimuli is problematic because thresholds for multiple edges Vernier stimuli scatter around those for single bar Vernier stimuli. For the first age groups some of the thresholds estimated directly from the experimental data are higher than the maximum displacement which was possible for the respective flanking distance s, and the corresponding data points in Fig. 3 are surrounded by brackets. This happens when the Probit analysis extrapolates a threshold (and leads to huge 95% confidence intervals) because the frequency of correct responses is ~75%

50.0

___----

NOTE

at all displacements. In all these cases. thresholds are definitely higher than the maximum possible displacement which thus can be treated as a conservative Vernier acuity estimate. The resulting values always are worse than Vernier acuity for single bars (2 months: 120’. 60’ and 30’, for instance). Thus the flanking bars seem to deteriorate Vernier acuity in young babies by an amount which cannot be estimated correctly from the present experiments. On the other hand, for instance with 779 yr olds, the closely flanking bars enhance Vernier acuity to 0.19’, as compared to 0.24’ for the single bar stimulus. This result resembles the one for adult control subjects, and might be related to their reports that in the multiple edges Vernier stimuli two thin black bars with Vernier displacement were perceived in front of a white field, especially for the very close bars (d = 2 mm; cf. Fig. 1, bottom row). In the intermediate age groups, thresholds for multiple Verniers are inconsistent, sometimes above and sometimes below those for single bar Verniers. Regarding the confidence intervals of single bar Vernier acuity shown in Fig. 2, it is clear that all differences between the thresholds for the two types of stimuli are not statistically significant. In consequence, and sadly despite the big amount of data collected in this study, it is premature to speculate about the development of lateral interactions. DISCUSSlON In the present paper, the development of grating acuity was compared to that of Vernier acuity by means of preferential looking. In FL experiments, it cannot be excluded that stimuli are processed by the visual system but that the infants do not respond in a way the experimenter can read properly. Therefore, the results give lower estimates of the infants’ true performance, and con~quently all conclusions have to be treated carefully. Besides this general argument, there are some problems with the details of the experiments which were partly discussed before. One difficulty is that the stimuli

5% d=2*s-ff

(xl ______ - ___________

-*i”Q’e

a

(j_s,2

d=l*s

x-_ d_/4

-1.5

-0.5

FIGURE 3. Using multiple edges Vernier stimuli (broken lines and open symbols, parameter bar distance d is given as fraction of bar width s), the development of displacement thresholds basically resembles that of single bar Vernier stimuli (solid lines and black squares). In some cases, the flanking bars appear to deteriorate acuity, in other cases single bar Vernier acuity seems to be worse than that for mul~ple*edges Vernier.

RESEARCH

were oscillated up and down. This might have influenced the results because it cannot be excluded that the attracting effect depends on the development of the infants’ sensitivity to motion or flicker. However, one should recall that Vernier displacements and the direction of motion were orthogonal in our study, and thus some neural representation of the Vernier offset is a precondition for the detection of the moving contour (vertical stripes cannot be seen moving vertically). Indeed, large offsets are discovered even by very young infants although the speed used here might be close to threshold (cf. Dannemiller & Freedland, 1989). Thus motion could have a modulating influence on the response to the Vernier offset (as discussed in the next paragraph). This might cause problems for the comparison with acuity card measurements which, apart from unvoluntary movements of the experimenter and the subject’s eyes, do not exploit the effects of motion. For this and other reasons, the difference in developmental speed between Vernier and grating acuity has to be treated carefully, but it will be compared in the following to data from the literature which partly suffer from similar drawbacks. It turns out that our results are not in good accordance with all observations on Vernier acuity development published so far. In agreement with Shimojo and Held (1987), Vernier acuity starts from a considerably lower value than grating acuity in young babies, and then develops faster. However, equal performance appears much later, namely at 15 months in our study, as compared to the intersection at 4 months in the former study. This is even outside the broad 95% confidence interval we estimated for the crossover point between the two developmental curves. It has to be recalled that the crossover point will depend, however, on the actual experimental conditions. Vernier acuity is consistently lower in our data than in that of Shimojo and Held, while grating acuities are similar in both studies, and in accordance with the other published data (for review, see Mohn & van Hof-van Duin, 1991). This discrepancy could be explained by the fact that Shimojo and Held used repetitive Vernier stimuli, whereas we tested single bars with Vernier displacements. If the repetition of the displacement generally increases the sensitivity for the displacement, one would expect the whole developmental curve to be shifted to higher acuity values, and correspondingly the intersection point with the grating acuity curve would be shifted to the left. This view seems to be further supported by the results of Manny and Klein (1984). By using repetitive, but static instead of oscillating Vernier targets, stimulus attraction might be attenuated as compared to those of Shimojo and Held. Correspondingly, Vernier acuities in the Manny and Klein study are closer to ours for single moving targets. Taking these speculations together, one might suggest that moving the Vernier targets or repeating the stimuli can increase acuity for such stimuli, leading to age shifts of the intersection point between the developmental curves of grating and Vernier acuity. In consequence, the

1563

NOTE

intersection point between the two development curves measured under different experimental conditions has to be interpreted with caution. The comparison of a resolution task to a task in which the performance is limited by the precision of relative position estimation was originally planned to contribute to the question of whether the two tasks are based on different neural mechanisms. The many experimental problems make it difficult to relate the development of grating and Vernier acuity to the current ideas on the development of the optical system, the retinal receptor mosaic, and the neuronal processing of visual information. However, the decrease of Vernier thresholds from 28’ for the youngest age tested by us to 0.14’ for adults in the same setup means a factor of about 200 for the increase of Vernier acuity. This would require major changes in the spatial grain of the visual system following the considerations of Shimojo and Held (1987); and it is more than might be expected from the ideal observer theory of the development of spatial vision on the basis of “preneural” factors (Banks & Bennett, 1988), mainly the maturation of receptor sensitivity and density (Yuodelis & Hendrickson, 1986). Since the predictions depend on the actual stimulus configuration, this should only be understood as a hint on neural maturation as a possible limiting factor, until careful model simulations are performed. Interestingly, the proliferation of synapses in the cortex which are thought to play a prominent role in the maturation of cortical processing, reaches a maximum at the age of 8 months (Huttenlocher, de Courten, Garey & van der Loos, 1982); and a model for the detection of Vernier displacements by some sort of nonlinear interaction predicted an intersection between the two development curves during this period (Wilson, 1988). Taking into account the large confidence interval for the actual point of intersection, this consideration is in accordance with our data. In summary, however, we are still far away from settling the question of whether the development of Vernier acuity is limited by retinal or cortical factors.

REFERENCES Atkinson, J. (1984). Human visual development over the first 6 months of life. A review and a hypothesis, Human Neurobiology, 3, 61-74. Banks, M. S. & Bennett, P. J. (1988). Optical and photoreceptor immaturities limit the spatial and chromatic vision of human neonates. Journal of the Optical Society of America A, 5,2059-2079. Barlow, H. B. (1979). Reconstructing the visual image in space and time. Nature, 279, 189-190. Birch, E. E., Gwiazda, J. & Held, R. M. (1982). Stereoacuity development for crossed and uncrossed disparities in human infants, Vision Research, 22, 507-S 13. Dannemiller, J. L. & Freedland, R. L. (1989). The detection of slow stimulus movement in 2-5 month-olds. Journal of Experimental Child Psychology, 47, 337-355. Fantz, R. L. (1961). The origin of form perception. ScientiJic American, 204, 66-72. Fern, K. D., Manny, R. E., Davis, J. R. & Gibson, R. R. (1986). Contour interaction in the preschool child. American Journal of Optometry and Physiological Optics, 63, 313-318. Finney, D. J. (1962). Probit analysis. Cambridge: Cambridge University Press.

I.564

RESEAR<‘H

Flom, M. C., Weymouth, F. W. & Kahneman, D. (1963). Visual resolution and contour interaction. Journal o/the Optical Sopier? of America.

53, 1026

.1032.

Haase, W. & Hohmann, A. (1982). Ein neuer Test (C-Test) cur quantitativen PriXung der Trennschwierigkeiten (“crowding”)Ergebnisse bei Amblyopie und Ametropie. Klinische Monarsbliifter fir Augenheilkunde, 180, 210 -215. Held, R. M., Shimojo, S. & Gwiazda, J. (1984). Gender differences in the early visual development of the human visual resolution. Inuestigatice

Ophthalmology

and Visual

Science,

25.

Huttenlocher. P. R., de Courten, C., Garey, L. J. & van der Loos. H. (I 982). Synaptogenesis in human visual cortex-vidence for synapse elimination during normal development. Neuroscience titters. 33, 247 252. Jacobs, R. J. (1979). Visual resolution and contour interaction in the fovea and periphery. Vision Research, 19. 1187. 1195. Kiorpes. L. & Movshon, J. A. (1987). Vernier acuity and spatial resolution in infant monkeys. Investigative Ophthalmology and Visual Science,

Visual

Science,

of Vernier

and

Ophthalmology

29. 9.

Levi, D. M., Klein, S. A. & Aitsebaomo, A. P. (1985). Vernier acuity, crowding and cortical magnification. Vision Research, 25, 963 --997. Levi. D. M., Manny, R. E., Klein, S. A. & Steinman, S. B. (1983). Electrophysiological correlates of hyperacuity in the human visual cortex. Narure, 306, 468-470. Manny, R. E. (1988). The visually evoked potential in response to Vernier offsets in infants. Human Neurobiology, 6, 272-279. Manny, R. E. & Klein, S. A. (1984). The development of Vernier acuity in infants. Currenr Eye Research, 3, 453.-462. Manny, R. E. & Klein, S. A. (1985). A three alternative tracking paradigm to measure Vernier acuity ofolder infants. Vision Research. 25. 1245

Mohn, G.. van Hof-van Duin, J., Fetter. W. I’. I... de Groat, L. & Hage, M. (1988). Acuity assessment of non-verbal infants and children: Clinical experience with the acuity card procedure. Derelopmenral Medicine and Child ,Veurolog~. .?O, 232 244.

Mohn. G.. van Hof-van Duin, J.. Groenendaal. F.. Fetter, W. P. F . de Groot, L. & Hage. M. (1990). Klinische Erfahrungen mit Acuity Cards: SehschHrfebestimmungen bei SBuglingen und mehrfach behinderten Kindern. In Miihlendyck, H. & Riissmann, W (Eds), Augenhewegungen und Csuelle Wahrnehmung (pp. I51 .I S?). Stuttgart: Enke. Rose. R. M., Teller. D. Y. 8c Rendleman, P. (1970). Statistical properties of staircase estimates. ferceprion and P.~.vchophy.sics. 8. 199-204. Shimojo, S. & Held, R. M. (1987). Vernier acuity IS less than gratmg acuity in 2- and 3-months-olds. Vision Research, 27. 77 86. Shimojo. S., Birch. E. E.. Gwiazda. J. & Held, R. M. (1984). Development of Vernier acuity in infants. Vision Research. 24. 721 728.

28, 359.

Kiorpes. L. & Movshon. J. A. (1988). Development grating acuity in strabismic monkeys. Invesrigafiue and

NOTE

1252.

McDonald, M. A., Dobson, V., Sebris, S. L., Baitch, L., Vamer, D. & Teller, D. Y. (1985). The acuity card procedure: A rapid test of infant acuity. Inoestigative Ophthalmology and Visual Science. 26, 1158-1162. McKee, S. P. & Levi, D. M. (1987). Dichoptic hyperacuity: The precision of nonius alignment. Journal of the Oprical Society of America A, 4. 1104-1108. Mohn, G. & van Hof-van Duin, J. (1991). Development of spatial vision. In Regan, D. M. (Ed.), VGon and visual dysficfion 10. Spatial cision (pp. 179-211). London: Macmillan.

van Sluyters, R. C., Atkinson, J., Banks, M. S., Held, R. M., Hoffmann, K. P. & Shatz. C. J. (1989). The development of vision and visual perception. In Spillmann, L. & Warner. J. S. (Eds). The neurophysiological foundarions q/ cisual perception. London: Academic Press Westheimer, G. (1979). The spatial Sense of the eye. lncesligatice Ophthalmology

and V&al

Science,

18, 893 -912.

Westheimer, G. & Hauske. G. (1975). Temporal and spatial Interference with Vernier acuity. Vision Research, 15, 1137-l 141. Wilson, H. R. (1986). Responses of spatial mechanisms can explain hyperdcuity. Vision Research, 26, 453-469. Wilson. H. R. (1988). Development of spatiotemporal mechanisms m infant vision. Vision Research, 28, 61 l-628. Yuodelis. C. & Hendrickson, A. (1986). A qualitative and quantitative analysis of the human fovea during development. Gion Research. 26. 847.-855.

Acknowledgemenfs-First of all, we would like to thank V. Barth who assisted the experiments with great competence, and kept the baby-lab running through hard times. Also, we are indebted to 0. v. Nieuwenhuizen who took part in some early experiments. The stimulus setup was built with great skill by the mechanical workshop of the University Eye Clinic. This study was supported by grants of the Deutsche Forschungsgemeinschaft to G.M. (Fa I l9/42) and M.F. (Fa Il9/5-2, Heisenberg-Programm).