Perception, 1977, volume 6, pages 685-690
Discrimination by young infants of stimuli presented discontinuously t
Paul L Harris^, Elizabeth Bassett Department of Psychology, University of Lancaster, Lancaster, England Received 4 November 1976
Abstract. Four experiments are reported showing that young infants can pick up visual information from very brief displays. Pattern, edge orientation, and form can be discriminated when stimuli are presented for repeated exposures of only 300 ms. 1 Introduction It is now firmly established that young infants of three to four months can discriminate between stimuli differing in amount of contour, orientation and form (Salapatek 1975). However, there is little information currently available concerning the speed with which the young infant can pick up such visual information. Karmel et al (1974) studied the effect of tachistoscopic presentation of stimuli varying in amount of contour upon visually evoked potentials. However, there appear to be no published studies in which the overt looking preferences of the young infant, or alternatively his ability to identify familiar forms, have been examined under conditions of brief exposure. Most current methods of testing infant discrimination (e.g. preference and habituation methods) require a sustained exposure of paired stimuli so that selectivity of fixation can be examined over time. Such methods clearly preclude the use of a single tachistoscopic exposure. However, the effect of discontinuous exposure may be examined by presenting infants with the same pair of stimuli for a repeated series of brief exposures. Where each exposure is sufficiently short, no eye movements can be initiated and executed within any single exposure. In experiment 1, we asked whether young infants would spend more time looking at a patterned versus a homogeneous target when each was briefly but repeatedly presented. The two targets were simultaneously presented; one consisted of an alternation each 300 ms of a black and white vertical grating, with a homogeneous grey field. The second target consisted of an alternation of a homogeneous grey field with an identical homogeneous field. A fixation preference for the first target would indicate that pattern can be detected by the young infant despite brief exposure. If this discrimination were impossible for the young infant then form discrimination for briefly exposed stimuli would be unlikely. 2 Experiment 1 2.1 Method 2.1.1 Subjects. Subjects were twenty infants (mean age 112 days; s = 60), Subjects in experiments 1 - 4 were recruited by newspaper advertisement and were always tested in the infant laboratory of the psychology department. t An earlier version of this paper was contributed to the symposium: Infant Perception, XXIst International Congress of Psychology, Paris, 1976. % Current address: Department of Psychology, Free University, 22 Koningslaan, Amsterdam, Netherlands.
686
P L Harris, E Bassett
2.1.2 Procedure and apparatus. Subjects were presented simultaneously with two targets. One target consisted of a vertical grating and a grey field of equivalent brightness, which alternated with one another every 300 ms. The second target consisted of two grey fields of equivalent brightness, which alternated with one another every 300 ms. Both targets were simultaneously exposed for two 30 s trials with a 15 s intertrial interval. Starting position for the targets was counterbalanced across subjects; target positions were always switched between trials. Subjects faced a black cardboard screen 53 cm high x 80 cm long. Two portholes (d = 8-5 cm) were cut in the screen so that their midpoints were 24-5 cm apart. The screen of a TV monitor was visible through these two portholes. Targets were presented at tachistoscopic speeds by means of a prerecorded video film. The film was constructed as follows: two fields of a tachistoscope were alternated by a function generator. Both left and right halves of one field were homogeneous grey. One half of the second field was also grey, the other half contained a vertical grating. Hence an alternation between the two fields, produced a grey-grey alternation, visible during playback at one porthole, and a grey-grating alternation visible during playback at the other porthole. On the TV monitor the stripes of the grating appeared 2 cm wide and 2 cm apart (spatial frequency of 0-135 cycles per degree of visual angle). The overall illumination of each field was 86-1 lx. Infants were held by their mothers in an upright position between the two targets at a distance of approximately 30 cm from the screen. The midpoints of the two targets were separated by a visual angle of approximately 45°. Mothers were asked to hold their babies such that they could easily turn to either target; no information was given concerning experimental expectations until testing was completed. On later questioning, mothers were usually unaware of which target their infants had preferred to fixate. Fixation times were recorded on a Rustrak event recorder by an observer stationed at one side of the monitor. Shifts of fixation from one target to the other were easily identified, given the large angular separation of the two targets. Pretests with paired observers stationed at either side of the screen showed that observers agreed where the infant was looking for 93% of a 60 s target exposure. 2.2 Results Mean fixation (31 -6 s; s = 15-4) for the grating was significantly greater than the mean fixation (9-4 s; s = 9-4) for the homogeneous target (t = 4-51; d.f. = 19; p < 0-001; two-tailed test). Similarly, the percentage of fixation time devoted to the grating (75-3%) was significantly greater than 50% (t = 5-69; d.f. = 19; p< 0-001). 2.3 Discussion Experiment 1 shows that infants can detect the presence of contour within 300 ms since the grating is discriminable from and preferred to the homogeneous grey target. The results provide no support for the hypothesis that the preference for contour depends upon eye movements along or across an edge, assuming that such movements cannot be executed within 300 ms. Such an assumption appears reasonable in the light of results obtained by Aslin and Salapatek (1975). For a target 10° from fixation point, two-month-old infants initiated a saccadic shift with a median latency of 480 ms. Latencies increased to 1280 ms for 30° targets. Two cautionary points should be made, however. First the above experiment establishes a preference for patterned versus homogeneous stimuli despite discontinuous presentation. Such a preference does not establish that edge orientation is discriminable. For example, the young infant might perceive a speckled pattern and a striped pattern as equivalent. Second, since the vertical stripes were presented
Discrimination by young infants of stimuli presented discontinuously
687
repeatedly, it is possible that random eye movements would have resulted in the traversal of a contour during its brief exposure. Moreover, an eye movement initiated during exposure n might have sometimes reached its target during exposure n + 1. Hence it cannot be concluded that eye movements are irrelevant to the preference for pattern. A second experiment was designed to check these possibilities. Subjects were presented with a choice of fixating either an alternation of horizontal and vertical gratings or alternatively a steady exposure of a vertical grating. While the latter target would permit sustained scanning, the former should remain more novel if the change of orientation could be detected (Fantz 1964), and should therefore elicit longer fixation times. 3 Experiment 2 3.1 Method 3.1.1 Subjects. Subjects were twenty-four infants (mean age 108 days: s = 57). 3.1.2 Procedure and apparatus. Subjects were presented with two targets with the use of the same apparatus and general design as in experiment 1. At one porthole a vertical grating alternated each 300 ms with an identical vertical grating. Hence an adult viewer perceived a steady vertical grating. At the other porthole, a vertical grating alternated each 300 ms with a horizontal grating. The overall illumination of each field and the spatial frequency of the gratings were equivalent to those used in experiment 1. 3.2 Results Infants spent a mean of 25-4 s (s = 14-3) fixating the alternating vertical-horizontal grating and a mean of 15-6 s (s = 12-3) fixating the steady vertical grating. Twotailed t tests confirmed that these two means differed (7 = 2-17, d.f. = 23, p < 0-05) and also that the mean percentage of fixation time devoted to the alternating pattern (62-2%) was greater than 50% (t = 2-84, d.f. = 23, p < 0-01). 3.3 Discussion Since both targets contained an equal amount of contour, and only differed in the constancy of contour orientation, experiment 2 demonstrates that not only the presence of an edge but the orientation of an edge can be detected by young infants within 300 ms. This experiment also indicates that eye movements are irrelevant to the discrimination of contour. The steady vertical grating, unlike the alternating pattern, permitted continuous scanning of its edges. Thus, if edge scanning leads to longer fixation time, the steady vertical grating should have been preferred. Experiments 1 and 2 indicate that the basic pattern-detecting abilities of the human infant operate for stimuli which are not continuously exposed. In experiment 3, we asked whether a familiar form would be preferred despite brief, repeated exposure. Young infants are able to discriminate between a regular and a scrambled outline face, despite their equivalence in amount of contour (Haaf and Brown 1975). Hence these stimuli were used in experiment 3. 4 Experiment 3 4.1 Method 4.1.1 Subjects. Subjects were twenty infants (mean age 103 days; s = 58). 4.1.2 Procedure and apparatus. Subjects were again presented with two targets. One target consisted of a line-representation of a human face which alternated with a grey field. The second target consisted of a line representation of a human face, with the features misarranged, which alternated with a grey field. The two faces
688
P L Harris, E Bassett
were modelled after the high-complexity stimuli used by Haaf and Brown (1975, experiments 3 and 4). Both faces appeared on the film 22 cm high and 15 cm wide; they were bilaterally symmetrical and contained the same amount of contour. The illumination of each field was 64-6 lx. The procedure differed from experiments 1 and 2 in that the black screen was removed so that the full dimensions of each face were visible to subjects on the TV monitor. The'midpoints of the two faces were 24-5 cm apart. Targets were exposed for four 30 s periods with a 15 s intertrial interval. Target positions were switched between trials. Starting positions for targets on the first trial were systematically varied across subjects. 4.2 Results Mean fixation time (52-1 s, s = 19-9) of the regular face was significantly greater than mean fixation (25-2 s, s = 12-8) for the misarranged face (t = 5-43, d.f. = 19, p < 0-001, two-tailed test). Similarly, the mean percentage of fixation time devoted to the regular face (65-4%) was significantly greater than 50% (t = 3-74, d.f. = 19, p< 0-001). 4.3 Discussion Since the forms were equivalent in amount of contour and symmetry, the preference for the regular face can be attributed to recognition of the face as a familiar gestalt. Fixation times for normally presented stimuli are known to vary in terms of facial resemblance by ten to fifteen weeks of age (Haaf and Brown 1975). Before drawing any firm conclusions, a final experiment was designed, to counter possible methodological objections to the procedures used in experiments 1-3. Two television monitors were placed side by side, and the same film was simultaneously replayed on each. The targets differed only in orientation (i.e. one of the monitors was inverted). Since the same film was replayed through both monitors, the targets were unequivocally identical in brightness, amount of contour, and symmetry. The use of two monitors also permitted the observer to be stationed directly facing the infant by looking through a peephole located between the two monitors. 5 Experiment 4 5.1 Method 5.1.1 Subjects. Subjects were twelve infants (mean age 61 days; s = 22). 5.1.2 Procedure and apparatus. Subjects faced a black cardboard screen 53 cm high x 80 cm long. Two portholes (d = 15 cm) were cut into the screen so that their midpoints were 31 cm apart. A TV monitor was located directly behind each porthole. A film, which consisted of a grey field that alternated with a schematic face every 300 ms, was replayed simultaneously through both monitors. The face appeared in an upright position at one porthole, but in an inverted position at the other porthole, because of the inversion of one TV monitor. Targets were exposed for four 30 s trials with a 15 s intertrial interval. Starting position for the upright target was counterbalanced across subjects; target positions were always switched between trials. (This was achieved by filming an upright face and an inverted face on successive trials.) The facial stimulus was identical to the regular schematic face used in experiment 3, but its dimensions were reduced (11 cm high x 7-5 cm wide). The illumination of each field was again 64-6 lx. Observations were made through a peephole in the screen, located midway between the two portholes. Observations were made blind in that the observer did not know which orientation appeared at each of the two monitors.
Discrimination by young infants of stimuli presented discontinuously
689
5.2 Results Mean fixation (48-2 s; s = 13-3) for the upright target was significantly greater than mean fixation (31 -6 s; s = 14-6) for the inverted target (t = 4-471, d.f. = 11, p < 0*001, two-tailed test). Similarly, the percentage of fixation time devoted to the upright target (62-4%) was significantly greater than 50% {t = 9-722, d.f. = 11, p< 0-001). 6 General discussion Experiment 4 demonstrates that infants of approximately two months can discriminate between an upright and an inverted schematic face, despite discontinuous exposure. This preference cannot be attributed to variation between the faces in brightness, symmetry, or contour. The preference for an upright face is consonant with recent findings (Goren 1975) although it runs counter to earlier reports in the literature (McGurk 1970). The mode of presentation may explain the selectivity of the infant's response. In experiments 3 and 4 the repeated appearance and disappearance of the facial stimulus appeared to produce an illusion of animacy which would be absent for continuously exposed targets, as employed by McGurk (1970). The infants' behaviour during testing supports this interpretation. Many subjects (particularly the older subjects in experiment 4) responded with smiles or vocalisation. Such apparently social responses were not noticed in experiments 1 and 2. The results of the four experiments considered together, indicate that continuous presentation of a stimulus is unnecessary for the detection of pattern, edge orientation, or even the presence of a familiar form by the young infant. How can the infant's successful discrimination despite discontinuous exposure be explained? One might argue that the infant never actually perceives the discontinuous nature of the stimulus. For example, in experiments 3 and 4, the 300 ms interval between successive presentations of the facial stimulus might be too brief to be detected by the infant, so that what appears to an adult as a discontinuously presented stimulus appears to the infant as a continuously presented stimulus. Such an argument is difficult to sustain, however, since it applies equally to the homogeneous target interdigitated with the facial target; if infants effectively perceived a steady grey target at both portholes in experiments 3 and 4, they should have exhibited no reliable preference, contrary to the results actually obtained. Accepting, then, the assumption that subjects did detect the discontinuous nature of the stimulus presentations in all four experiments, might it not be argued that such input differs very little from that experienced during normal vision even during infancy? It is known that newborn infants execute relatively accurate saccadic eye movements (Harris and Macfarlane 1974; Macfarlane et al 1976). Moreover such saccades are normally executed during the examination of extended figures (Salapatek 1975). Hence, the infant should be familiar with the discontinuous nature of normal visual input. However, the conditions of discontinuous exposure employed in the present experiments clearly differ from those that would be experienced during normal vision. Previously visible features would be absent whenever the infant fixated during exposure of the homogeneous target. Even on the unlikely hypothesis that the infant could time-lock his fixation plus saccade pattern to the duration of exposure of the alternating face and homogeneous target, the facial stimulus would still completely disappear during saccades; it would not be simply blurred as is the case during normal exposure. Thus it does not seem reasonable to explain the results by arguing that the discontinuous exposure went undetected or that it was perceived as quite normal.
690
P L Harris, E Bassett
There appear to be two possible interpretations of the results. On the one hand subjects may have been able to identify the stimuli, including the more complex stimuli presented in experiments 3 and 4, during a single exposure of 300 ms. Subsequent exposures of the stimulus would therefore alter its familiarity but not its perceived identity. Alternatively subjects may have built up an integrated percept from a successive sampling of different features of the stimulus on successive exposures. The perceived identity of the stimulus would therefore alter across exposures. The present results do not permit a firm choice between these two alternatives. However, two points may be made about the latter alternative. First, it admits that young infants can extract visual features of some kind despite a brief exposure of 300 ms. Second, as noted above, successive sampling during discontinuous exposure would be different from successive sampling during continuous exposure. Thus it seems unlikely, albeit not impossible, that the identity of the facial stimulus emerged across stimulus exposures rather than within a single exposure. Nonetheless further research is required to establish firmly whether or not successive sampling does occur. The techniques adopted in experiments 1-4 could be readily adapted for this purpose. For example, lengthening of the interval and/or the introduction of random-dot patterns between successive target exposures should disrupt a successive sampling strategy but not a simultaneous strategy. In conclusion the present results show that infants as young as two months are able to identify complex stimuli despite their discontinuous exposure. Although much recent experimentation on form discrimination in early infancy has attempted to analyse the possible contribution of eye movements, the above results and techniques suggest a different approach in which the infant's ability to register and integrate features across a single fixation is studied. Older children exhibit marked changes in this ability (Liss and Haith 1970). It seems plausible to expect that major developments will also occur in early infancy. Acknowledgements. This work was carried out while the senior author was in receipt of an MRC grant. References Aslin R N, Salapatek, P, 1975 "Saccadic localization of peripheral targets by the very young human infant" Perception and Psychophysics 17 293-302 Fantz R L, 1964 "Visual experience in infants: decreased attention to familiar patterns relative to novel ones" Science 146 668-670 Goren C, 1975 "Form perception, innate form preferences and visually predicted head-turning in the human newborn" paper presented at meeting of Society for Research in Child Development, Denver, Col., USA Haaf R A, Brown C J, 1975 "Developmental changes in infants' response to complex facelike patterns" paper presented at meeting of Society for Research in Child Development, Denver, Col., USA Harris P L, Macfarlane A, 1974 "The growth of the effective visual field from birth to seven weeks" Journal of Experimental Child Psychology 18 340-348 Karmel B Z, Hoffman R F, Fegy M J, 1974 "Processing of contour information by human infants evidenced by pattern-dependent evoked potentials" Child Development 45 39-48 Liss P H, Haith M M, 1970 "The speed of visual processing in children and adults: effects of backward and forward masking" Perception and Psychophysics 8 396-398 Macfarlane A, Harris P L, Barnes I, 1976 "Central and peripheral vision in the neonate" Journal of Experimental Child Psychology 21 532-538 McGurk H, 1970 "The role of object orientation in infant perception" Journal of Experimental Child Psychology 9 363-373 Salapatek P, 1975 "Pattern perception in early infancy" in Infant Perception: From Sensation to Cognition Volume 1 Eds L B Cohen, P Salapatek (New York: Academic Press) © 1977 a Pion publication printed in Great Britain
Discrimination by young infants of stimuli presented discontinuously t
Paul L Harris^, Elizabeth Bassett Department of Psychology, University of Lancaster, Lancaster, England Received 4 November 1976
Abstract. Four experiments are reported showing that young infants can pick up visual information from very brief displays. Pattern, edge orientation, and form can be discriminated when stimuli are presented for repeated exposures of only 300 ms. 1 Introduction It is now firmly established that young infants of three to four months can discriminate between stimuli differing in amount of contour, orientation and form (Salapatek 1975). However, there is little information currently available concerning the speed with which the young infant can pick up such visual information. Karmel et al (1974) studied the effect of tachistoscopic presentation of stimuli varying in amount of contour upon visually evoked potentials. However, there appear to be no published studies in which the overt looking preferences of the young infant, or alternatively his ability to identify familiar forms, have been examined under conditions of brief exposure. Most current methods of testing infant discrimination (e.g. preference and habituation methods) require a sustained exposure of paired stimuli so that selectivity of fixation can be examined over time. Such methods clearly preclude the use of a single tachistoscopic exposure. However, the effect of discontinuous exposure may be examined by presenting infants with the same pair of stimuli for a repeated series of brief exposures. Where each exposure is sufficiently short, no eye movements can be initiated and executed within any single exposure. In experiment 1, we asked whether young infants would spend more time looking at a patterned versus a homogeneous target when each was briefly but repeatedly presented. The two targets were simultaneously presented; one consisted of an alternation each 300 ms of a black and white vertical grating, with a homogeneous grey field. The second target consisted of an alternation of a homogeneous grey field with an identical homogeneous field. A fixation preference for the first target would indicate that pattern can be detected by the young infant despite brief exposure. If this discrimination were impossible for the young infant then form discrimination for briefly exposed stimuli would be unlikely. 2 Experiment 1 2.1 Method 2.1.1 Subjects. Subjects were twenty infants (mean age 112 days; s = 60), Subjects in experiments 1 - 4 were recruited by newspaper advertisement and were always tested in the infant laboratory of the psychology department. t An earlier version of this paper was contributed to the symposium: Infant Perception, XXIst International Congress of Psychology, Paris, 1976. % Current address: Department of Psychology, Free University, 22 Koningslaan, Amsterdam, Netherlands.
686
P L Harris, E Bassett
2.1.2 Procedure and apparatus. Subjects were presented simultaneously with two targets. One target consisted of a vertical grating and a grey field of equivalent brightness, which alternated with one another every 300 ms. The second target consisted of two grey fields of equivalent brightness, which alternated with one another every 300 ms. Both targets were simultaneously exposed for two 30 s trials with a 15 s intertrial interval. Starting position for the targets was counterbalanced across subjects; target positions were always switched between trials. Subjects faced a black cardboard screen 53 cm high x 80 cm long. Two portholes (d = 8-5 cm) were cut in the screen so that their midpoints were 24-5 cm apart. The screen of a TV monitor was visible through these two portholes. Targets were presented at tachistoscopic speeds by means of a prerecorded video film. The film was constructed as follows: two fields of a tachistoscope were alternated by a function generator. Both left and right halves of one field were homogeneous grey. One half of the second field was also grey, the other half contained a vertical grating. Hence an alternation between the two fields, produced a grey-grey alternation, visible during playback at one porthole, and a grey-grating alternation visible during playback at the other porthole. On the TV monitor the stripes of the grating appeared 2 cm wide and 2 cm apart (spatial frequency of 0-135 cycles per degree of visual angle). The overall illumination of each field was 86-1 lx. Infants were held by their mothers in an upright position between the two targets at a distance of approximately 30 cm from the screen. The midpoints of the two targets were separated by a visual angle of approximately 45°. Mothers were asked to hold their babies such that they could easily turn to either target; no information was given concerning experimental expectations until testing was completed. On later questioning, mothers were usually unaware of which target their infants had preferred to fixate. Fixation times were recorded on a Rustrak event recorder by an observer stationed at one side of the monitor. Shifts of fixation from one target to the other were easily identified, given the large angular separation of the two targets. Pretests with paired observers stationed at either side of the screen showed that observers agreed where the infant was looking for 93% of a 60 s target exposure. 2.2 Results Mean fixation (31 -6 s; s = 15-4) for the grating was significantly greater than the mean fixation (9-4 s; s = 9-4) for the homogeneous target (t = 4-51; d.f. = 19; p < 0-001; two-tailed test). Similarly, the percentage of fixation time devoted to the grating (75-3%) was significantly greater than 50% (t = 5-69; d.f. = 19; p< 0-001). 2.3 Discussion Experiment 1 shows that infants can detect the presence of contour within 300 ms since the grating is discriminable from and preferred to the homogeneous grey target. The results provide no support for the hypothesis that the preference for contour depends upon eye movements along or across an edge, assuming that such movements cannot be executed within 300 ms. Such an assumption appears reasonable in the light of results obtained by Aslin and Salapatek (1975). For a target 10° from fixation point, two-month-old infants initiated a saccadic shift with a median latency of 480 ms. Latencies increased to 1280 ms for 30° targets. Two cautionary points should be made, however. First the above experiment establishes a preference for patterned versus homogeneous stimuli despite discontinuous presentation. Such a preference does not establish that edge orientation is discriminable. For example, the young infant might perceive a speckled pattern and a striped pattern as equivalent. Second, since the vertical stripes were presented
Discrimination by young infants of stimuli presented discontinuously
687
repeatedly, it is possible that random eye movements would have resulted in the traversal of a contour during its brief exposure. Moreover, an eye movement initiated during exposure n might have sometimes reached its target during exposure n + 1. Hence it cannot be concluded that eye movements are irrelevant to the preference for pattern. A second experiment was designed to check these possibilities. Subjects were presented with a choice of fixating either an alternation of horizontal and vertical gratings or alternatively a steady exposure of a vertical grating. While the latter target would permit sustained scanning, the former should remain more novel if the change of orientation could be detected (Fantz 1964), and should therefore elicit longer fixation times. 3 Experiment 2 3.1 Method 3.1.1 Subjects. Subjects were twenty-four infants (mean age 108 days: s = 57). 3.1.2 Procedure and apparatus. Subjects were presented with two targets with the use of the same apparatus and general design as in experiment 1. At one porthole a vertical grating alternated each 300 ms with an identical vertical grating. Hence an adult viewer perceived a steady vertical grating. At the other porthole, a vertical grating alternated each 300 ms with a horizontal grating. The overall illumination of each field and the spatial frequency of the gratings were equivalent to those used in experiment 1. 3.2 Results Infants spent a mean of 25-4 s (s = 14-3) fixating the alternating vertical-horizontal grating and a mean of 15-6 s (s = 12-3) fixating the steady vertical grating. Twotailed t tests confirmed that these two means differed (7 = 2-17, d.f. = 23, p < 0-05) and also that the mean percentage of fixation time devoted to the alternating pattern (62-2%) was greater than 50% (t = 2-84, d.f. = 23, p < 0-01). 3.3 Discussion Since both targets contained an equal amount of contour, and only differed in the constancy of contour orientation, experiment 2 demonstrates that not only the presence of an edge but the orientation of an edge can be detected by young infants within 300 ms. This experiment also indicates that eye movements are irrelevant to the discrimination of contour. The steady vertical grating, unlike the alternating pattern, permitted continuous scanning of its edges. Thus, if edge scanning leads to longer fixation time, the steady vertical grating should have been preferred. Experiments 1 and 2 indicate that the basic pattern-detecting abilities of the human infant operate for stimuli which are not continuously exposed. In experiment 3, we asked whether a familiar form would be preferred despite brief, repeated exposure. Young infants are able to discriminate between a regular and a scrambled outline face, despite their equivalence in amount of contour (Haaf and Brown 1975). Hence these stimuli were used in experiment 3. 4 Experiment 3 4.1 Method 4.1.1 Subjects. Subjects were twenty infants (mean age 103 days; s = 58). 4.1.2 Procedure and apparatus. Subjects were again presented with two targets. One target consisted of a line-representation of a human face which alternated with a grey field. The second target consisted of a line representation of a human face, with the features misarranged, which alternated with a grey field. The two faces
688
P L Harris, E Bassett
were modelled after the high-complexity stimuli used by Haaf and Brown (1975, experiments 3 and 4). Both faces appeared on the film 22 cm high and 15 cm wide; they were bilaterally symmetrical and contained the same amount of contour. The illumination of each field was 64-6 lx. The procedure differed from experiments 1 and 2 in that the black screen was removed so that the full dimensions of each face were visible to subjects on the TV monitor. The'midpoints of the two faces were 24-5 cm apart. Targets were exposed for four 30 s periods with a 15 s intertrial interval. Target positions were switched between trials. Starting positions for targets on the first trial were systematically varied across subjects. 4.2 Results Mean fixation time (52-1 s, s = 19-9) of the regular face was significantly greater than mean fixation (25-2 s, s = 12-8) for the misarranged face (t = 5-43, d.f. = 19, p < 0-001, two-tailed test). Similarly, the mean percentage of fixation time devoted to the regular face (65-4%) was significantly greater than 50% (t = 3-74, d.f. = 19, p< 0-001). 4.3 Discussion Since the forms were equivalent in amount of contour and symmetry, the preference for the regular face can be attributed to recognition of the face as a familiar gestalt. Fixation times for normally presented stimuli are known to vary in terms of facial resemblance by ten to fifteen weeks of age (Haaf and Brown 1975). Before drawing any firm conclusions, a final experiment was designed, to counter possible methodological objections to the procedures used in experiments 1-3. Two television monitors were placed side by side, and the same film was simultaneously replayed on each. The targets differed only in orientation (i.e. one of the monitors was inverted). Since the same film was replayed through both monitors, the targets were unequivocally identical in brightness, amount of contour, and symmetry. The use of two monitors also permitted the observer to be stationed directly facing the infant by looking through a peephole located between the two monitors. 5 Experiment 4 5.1 Method 5.1.1 Subjects. Subjects were twelve infants (mean age 61 days; s = 22). 5.1.2 Procedure and apparatus. Subjects faced a black cardboard screen 53 cm high x 80 cm long. Two portholes (d = 15 cm) were cut into the screen so that their midpoints were 31 cm apart. A TV monitor was located directly behind each porthole. A film, which consisted of a grey field that alternated with a schematic face every 300 ms, was replayed simultaneously through both monitors. The face appeared in an upright position at one porthole, but in an inverted position at the other porthole, because of the inversion of one TV monitor. Targets were exposed for four 30 s trials with a 15 s intertrial interval. Starting position for the upright target was counterbalanced across subjects; target positions were always switched between trials. (This was achieved by filming an upright face and an inverted face on successive trials.) The facial stimulus was identical to the regular schematic face used in experiment 3, but its dimensions were reduced (11 cm high x 7-5 cm wide). The illumination of each field was again 64-6 lx. Observations were made through a peephole in the screen, located midway between the two portholes. Observations were made blind in that the observer did not know which orientation appeared at each of the two monitors.
Discrimination by young infants of stimuli presented discontinuously
689
5.2 Results Mean fixation (48-2 s; s = 13-3) for the upright target was significantly greater than mean fixation (31 -6 s; s = 14-6) for the inverted target (t = 4-471, d.f. = 11, p < 0*001, two-tailed test). Similarly, the percentage of fixation time devoted to the upright target (62-4%) was significantly greater than 50% {t = 9-722, d.f. = 11, p< 0-001). 6 General discussion Experiment 4 demonstrates that infants of approximately two months can discriminate between an upright and an inverted schematic face, despite discontinuous exposure. This preference cannot be attributed to variation between the faces in brightness, symmetry, or contour. The preference for an upright face is consonant with recent findings (Goren 1975) although it runs counter to earlier reports in the literature (McGurk 1970). The mode of presentation may explain the selectivity of the infant's response. In experiments 3 and 4 the repeated appearance and disappearance of the facial stimulus appeared to produce an illusion of animacy which would be absent for continuously exposed targets, as employed by McGurk (1970). The infants' behaviour during testing supports this interpretation. Many subjects (particularly the older subjects in experiment 4) responded with smiles or vocalisation. Such apparently social responses were not noticed in experiments 1 and 2. The results of the four experiments considered together, indicate that continuous presentation of a stimulus is unnecessary for the detection of pattern, edge orientation, or even the presence of a familiar form by the young infant. How can the infant's successful discrimination despite discontinuous exposure be explained? One might argue that the infant never actually perceives the discontinuous nature of the stimulus. For example, in experiments 3 and 4, the 300 ms interval between successive presentations of the facial stimulus might be too brief to be detected by the infant, so that what appears to an adult as a discontinuously presented stimulus appears to the infant as a continuously presented stimulus. Such an argument is difficult to sustain, however, since it applies equally to the homogeneous target interdigitated with the facial target; if infants effectively perceived a steady grey target at both portholes in experiments 3 and 4, they should have exhibited no reliable preference, contrary to the results actually obtained. Accepting, then, the assumption that subjects did detect the discontinuous nature of the stimulus presentations in all four experiments, might it not be argued that such input differs very little from that experienced during normal vision even during infancy? It is known that newborn infants execute relatively accurate saccadic eye movements (Harris and Macfarlane 1974; Macfarlane et al 1976). Moreover such saccades are normally executed during the examination of extended figures (Salapatek 1975). Hence, the infant should be familiar with the discontinuous nature of normal visual input. However, the conditions of discontinuous exposure employed in the present experiments clearly differ from those that would be experienced during normal vision. Previously visible features would be absent whenever the infant fixated during exposure of the homogeneous target. Even on the unlikely hypothesis that the infant could time-lock his fixation plus saccade pattern to the duration of exposure of the alternating face and homogeneous target, the facial stimulus would still completely disappear during saccades; it would not be simply blurred as is the case during normal exposure. Thus it does not seem reasonable to explain the results by arguing that the discontinuous exposure went undetected or that it was perceived as quite normal.
690
P L Harris, E Bassett
There appear to be two possible interpretations of the results. On the one hand subjects may have been able to identify the stimuli, including the more complex stimuli presented in experiments 3 and 4, during a single exposure of 300 ms. Subsequent exposures of the stimulus would therefore alter its familiarity but not its perceived identity. Alternatively subjects may have built up an integrated percept from a successive sampling of different features of the stimulus on successive exposures. The perceived identity of the stimulus would therefore alter across exposures. The present results do not permit a firm choice between these two alternatives. However, two points may be made about the latter alternative. First, it admits that young infants can extract visual features of some kind despite a brief exposure of 300 ms. Second, as noted above, successive sampling during discontinuous exposure would be different from successive sampling during continuous exposure. Thus it seems unlikely, albeit not impossible, that the identity of the facial stimulus emerged across stimulus exposures rather than within a single exposure. Nonetheless further research is required to establish firmly whether or not successive sampling does occur. The techniques adopted in experiments 1-4 could be readily adapted for this purpose. For example, lengthening of the interval and/or the introduction of random-dot patterns between successive target exposures should disrupt a successive sampling strategy but not a simultaneous strategy. In conclusion the present results show that infants as young as two months are able to identify complex stimuli despite their discontinuous exposure. Although much recent experimentation on form discrimination in early infancy has attempted to analyse the possible contribution of eye movements, the above results and techniques suggest a different approach in which the infant's ability to register and integrate features across a single fixation is studied. Older children exhibit marked changes in this ability (Liss and Haith 1970). It seems plausible to expect that major developments will also occur in early infancy. Acknowledgements. This work was carried out while the senior author was in receipt of an MRC grant. References Aslin R N, Salapatek, P, 1975 "Saccadic localization of peripheral targets by the very young human infant" Perception and Psychophysics 17 293-302 Fantz R L, 1964 "Visual experience in infants: decreased attention to familiar patterns relative to novel ones" Science 146 668-670 Goren C, 1975 "Form perception, innate form preferences and visually predicted head-turning in the human newborn" paper presented at meeting of Society for Research in Child Development, Denver, Col., USA Haaf R A, Brown C J, 1975 "Developmental changes in infants' response to complex facelike patterns" paper presented at meeting of Society for Research in Child Development, Denver, Col., USA Harris P L, Macfarlane A, 1974 "The growth of the effective visual field from birth to seven weeks" Journal of Experimental Child Psychology 18 340-348 Karmel B Z, Hoffman R F, Fegy M J, 1974 "Processing of contour information by human infants evidenced by pattern-dependent evoked potentials" Child Development 45 39-48 Liss P H, Haith M M, 1970 "The speed of visual processing in children and adults: effects of backward and forward masking" Perception and Psychophysics 8 396-398 Macfarlane A, Harris P L, Barnes I, 1976 "Central and peripheral vision in the neonate" Journal of Experimental Child Psychology 21 532-538 McGurk H, 1970 "The role of object orientation in infant perception" Journal of Experimental Child Psychology 9 363-373 Salapatek P, 1975 "Pattern perception in early infancy" in Infant Perception: From Sensation to Cognition Volume 1 Eds L B Cohen, P Salapatek (New York: Academic Press) © 1977 a Pion publication printed in Great Britain
