- Email: [email protected]
Preference for rotating objects in 5-month-old infants
INFANT
BEHAVIOR
AND
DEVELOPMENT
10,
365369
(1987)
BRIEF REPORT
Preference for Rotating Objects in SMonth-Old Infants HOLLY A.RUFF Albert Einstein College of Medicine Five-month-olds were habituated to several objects undergoing a variety of rotations or nonrotations. The test trials presented a new obiect undergoing examples of the old and new motions. Without exception, infants preferred the rotating object on the test trials, regardless of their previous experience.
perception
motion
of objects
preference
Previous work (Ruff, 1985) has shown that Smonth-olds are capable of discriminating rotating from nonrotating objects in several different contexts. The purpose of the present study was to determine if infants of 5 months would respond to rotating objects (those moving around their own axes) as equivalent to each other and as different from objects moving but not rotating. Therefore, infants were habituated to several different objects undergoing either several kinds of rotation or several kinds of nonrotation; they were then tested for their responsiveness to a novel object undergoing a novel rotation and a nonrotation, or a novel nonrotation and a rotation. A significant differentiation between novel and familiar motions in this case would suggest that discrimination and recognition of rotation or nonrotation was highly general and not based on any particular pattern of stimulation. The subjects for this study were 20 infants (11 males and 9 females) with a mean age of 22 weeks and 4 days. Another 11 infants were seen but were not included either because they were too fussy (9) or because of experimenter error (2). There were five different objects which were rectilinear solids of different shapes, some asymmetrical. For each infant, four objects were used for habituation and the fifth was used for the test trials. For every rotation, the object remained in the same place while it moved around one of its own axes; all rotations, therefore, involved changes in orientation of the objects with respect to an objective three-dimensional framework. The five motions were: (1) oscillating rotations of 180” around the vertical axis; (2) similar oscillations around The research reported here was supported by Grant BNS 80-13064 from the National Science Foundation. I wish to thank John Ruff and Gerald Turkewitz for reading and discussing an earlier version of this paper with me. Correspondence and requests for reprints should be addressed to Holly A. Ruff, 222 Kennedy Center, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, NY 10461.
365
366
RUFF
the horizontal axis; (3) similar oscillations around a line toward the infant;! (4) full rotations around the vertical axis; and (5) full rotatations around a line toward the infant. In contrast to the rotations, the five nonrotations involved no changes in orientation with respect to an objective framework. These motions, each of which covered about 10 cm in different directions from the center position, were: (1) moving laterally in alternate directions in the frontal plane; (2) moving alternately toward and away from the infant; (3) moving up and down; (4) moving so that a circle was described in the frontal plane; and (5) moving so that a square was described in the frontal plane. Every infant was habituated to four of the motions in one category and the fifth was used as the “familiar” motion on the test trials. Each infant was seated on the mother’s lap at a table; the objects were presented at the edge of the table opposite the infant (76.2 cm away). A large backdrop excluded the experimenter and observer from the infant’s view. Each object was rigidly fixed to a dowel 1.3 cm in diameter and 30.5 cm in length; the dowel allowed the object to be moved by hand without the hands being visible to the infant. All motions were well-practiced and smooth; there was a set number of cycles per trial, and no differences among the motions in the degree of variability were observed. For the habituation periods of both the rotation and nonrotation conditions, the four chosen motions were combined with the four objects, so that there were 16 possible combinations of object and motion. Every trial was 30 s long, and an observer recorded the duration of the infant’s looking at the object on a digital stopclock. An infant control procedure was used; that is, trials were continued until the duration of the infant’s looking in two consecutive trials was less than or equal to half of the mean duration of looking on the first two trials. A maximum of 20 trials was presented.* HaIf of the infants were habituated to rotation and half to nonrotation, leading to two conditions defined by the type of motion presented during habituation. When the criterion was reached, the infant was presented with two test trials, one in which the fifth and novel object rotated and the other in which the object moved without rotating. The order in which familiar and novel motions were presented was completely counterbalanced. Considering the motions presented successively in the test trials as a pair, five pairs of rotation and nonrotation were used. Both members of the pair involved either continuous motion (e.g., an object rotating fully in the frontal plane vs. a nonrotating object moving in a circular path in the frontal plane) or changes in direction (oscillating around the vertical axis vs. translating laterally back and forth in the frontal plane). Each pair was presented to four infants. ’ The phrase “a line toward the infant” is used rather than the “line of sight” because active observers with binocular perception can never see a three-dimensional object rotating exactly around the line of sight. Even slight deviations from the line of sight mean that rotation of the object
will lead to perspective changes and new points z Three infants in the rotation condition and two
reach criterion in 20 trials; 75% of the initial level.
on the 19th
and 20th
trials,
coming infants
into view. in the nonrotation
however,
:heir
looking
condition ranged
did not from
52 to
PREFERENCE
FOR ROTATION
367
It should be noted that though the study involved an habituation period, there were no postcriterion trials; in this case, the most appropriate approach to the data is to consider the habituation period as a period of familiarization, and to compare the test trials directly to each other rather than to the criterion trials. Descriptive data for the two conditions are presented in Table 1. The data show that there was not much difference in the looking at novel and familiar motions on the test trials, but there was a large difference in the looking, between rotating and nonrotating objects. In fact, every infant looked longer at the rotating object than at the nonrotating object on the test trials, a highly significant difference, t(19) = 7.40, p < .OOl. The subsequent analyses, therefore, focus mainly on the difference between rotating and nonrotating objects rather than on the distinction between novel and familiar motions. A 2 (Condition) x 4 (Trials: initial, criterion, rotation, nonrotation) analysis of variance revealed no Condition effect, F< 1, but a highly significant Trials effect, F(3,54) =42.60, p< .OOl; and a significant Condition by Trial interaction, F(3,54) = 6.70, p< .005. A Neuman-Keuls post hoc analysis showed that looking on the criterion trials was significantly lower than looking on the initial trials, and that looking at the rotating objects was significantly higher than looking on either the criterion trials or the test trials with nonrotating objects. An examination of the means in Table 1 shows that the interaction is caused by the fact that the infants in the nonrotation condition looked less during the habituation phase and more during the test phase than the infants in the rotation condition. There is no evidence that the two conditions differed in the degree of decrement during habituation, t(18) = .88, or in the extent of the preference for rotation on the test trials, ~(18) = 1.39. In order to provide each infant with a single score for the two test trials and to test possible effects of order and novelty, the percent of total time spent looking at the rotating object was entered into a 2 (rotation first or second) x 2 (rotation novel or familiar) analysis of variance. There were no effects of TABLE Descriptive
Data
for
the
Rotation
1 and
Nonrototion
Conditions
Motion
used
Rotation Mean Looking Time Trials in Seconds Mean Looking Time Number
of Trials
in First in Two
Two
Habituation
Criterion
to Criterion
Trials
during
Habituation Nonrotation
27.0
20.4
(2.9) 12.9
(6.0)
(5.2) 14.9
w-3) (4.5) 25.1
0.3 13.4
Looking
Time
to Test
Rotation
(5.6) 19.1 (familiar)
Time
to Test
Nonrotation
(8.3) 9.4 (novel)
(6.1)
Looking
(6.7)
(4.8)
10.9
368
RUFF
order, F(1,16)=0; or of novelty, F(1,16)= .Ol, and no interaction, F(1,16)= .02. These results demonstrate an overwhelming preference for rotation that generalizes across five pairs of rotation and nonrotation and that occurs regardless of the nature of the habituation phase (see Slater, Earle, Morison, & Rose, 1985). The preference is consistent with the results of other studies (Gibson, Owsley, & Johnston, 1978; Ruff, 1985). The fact that Gibson et al. found only one significant difference among three possible differences between three rotations and one translation is probably related to the much smaller extent of rotation used in that study. Because the infants in both rotation and nonrotation conditions showed habituation and equivalent number of trials to criterion, it seemed reasonable to compare the responses to specific rotations and specific nonrotations in the habituation period using the mean looking time to each particular motion during the habituation period. Because each motion appeared once in each set of four trials, the mean was calculated on the basis of complete blocks of four trials; if, for example, an infant had 18 trials, a mean for each of the different motions was derived from the first 16 trials. Although there were five motions altogether, only four were used for any one infant; because of this, each infant contributed data for only four of the five means. The mean duration of looking per trial ranged from 22.3 to 25.4 s for the five/rotations and from 17.1 to 19.7 s for the five nonrotations. The mean for each rotation was, therefore, higher than the mean for any nonrotation. When the 25 possible combinations of rotations and nonrotations were compared with t tests for independent means, 22 of the 25 achieved the .l level; of these, 14 were significant beyond the .05 level. Although theoretically some of these tests could be significant by chance, the probability of all of the 25 possible comparisons favoring rotation by chance is < .OOl. In sum, these data also demonstrate a very strong preference for rotation, in this case, across infants. In contrast, when the five motions within each category were compared with one-way repeated measures analyses of variance,’ there was no evidence of differentiation among rotations, F(4,36) = 1.98, or among nonrotations, F(4,36) = .98. These data suggest that the motions in each category were equivalent in some respect. There could be several reasons for the failure to demonstrate a preference for the novel motion. Because other studies (Ruff, 1985) have shown that infants habitutated to one kind of rotation will generalize to that motion made with a novel object and recover to a nonrotation of that object, the most likely interpretation is that the infants in this study were overwhelmed by the variety of motions and objects involved and consequently responded mainly on the ’ The group mean was used to fill in the fifth motion for each infant so that the repeated measures analysis of variance could be conducted. Since such a procedure increases the degrees of freedom, as well as taki.lg advantage of the correlation among scores, the power of the test for detecting differences is increased. The analysis could, therefore, be considered a conservative test of the hypothesis that there were no differences among motions within conditions.
PREFERENCE
FOR ROTATION
369
basis of a spontaneous preference. The fact that the preference was in evidence across infants in the habituation period would support this possibility. There was no evidence, therefore, that infants at this age generalize from their experience with various rotations or nonrotations to a novel rotation or a novel nonrotation. On the other hand, because the preference for rotation occurred with a variety of different examples in each category and because there was more evidence for discriminations between categories than within categories, one must conclude then that infants of 5 months discriminate between rotations and nonrotations in a very general way. The preference for rotating objects may exist because rotations of an object generally provide more information about it than do nonrotations. That is, the changes that take place during rotation are more likely to reveal previously unprojected parts of the object than would the generally slighter changes involved in viewing an object that is moving but not rotating. In turn, the more extensive changes may increase the likelihood that the viewer will detect invariants specifying the structure of the object. Rotation may also be generally more stimulating or arousing than nonrotation. Both of these possibilities would suggest that if the changes produced by a rotating object were reduced by decreasing the angle of rotation and the changes produced by a nonrotating object were increased, there would be a reversal of the preference. In contrast, it is possible that the infants were responding to a qualitative difference between the two kinds of motion, for example, the presence or absence of orientation changes with respect to an objective framework. The data from this study are not definitive regarding the causes of the preference. Because of its strength and consistency, however, the preference could serve as a useful basis for the exploration of the stimulus factors that are involved in the perception of motion. REFERENCES Gibson, Ruff, Slater,
E.J., fants:
Owsley, C.J., Differentiation
H.A. (1985). old infants.
& Johnston, J. (1978). Perception of invariants by five-month-old inof two types of motion. Developmental Psychology, 14, 407-415.
Detection
Developmenrol
of information
Ps.vcholog.v,
A., Earle, D.C., Morison, V., teraction with habituation-induced
Psychology,
specifying
the motion
of objects
by 3- and 5-month-
21, 295-305.
& Rose, D. (1985). Pattern novelty preferences.
preferences
Journal
at birth
and their
of Experimenral
in-
Child
39, 37-54. 6 May
1986;
Revised
27 February
1907
W
BEHAVIOR
AND
DEVELOPMENT
10,
365369
(1987)
BRIEF REPORT
Preference for Rotating Objects in SMonth-Old Infants HOLLY A.RUFF Albert Einstein College of Medicine Five-month-olds were habituated to several objects undergoing a variety of rotations or nonrotations. The test trials presented a new obiect undergoing examples of the old and new motions. Without exception, infants preferred the rotating object on the test trials, regardless of their previous experience.
perception
motion
of objects
preference
Previous work (Ruff, 1985) has shown that Smonth-olds are capable of discriminating rotating from nonrotating objects in several different contexts. The purpose of the present study was to determine if infants of 5 months would respond to rotating objects (those moving around their own axes) as equivalent to each other and as different from objects moving but not rotating. Therefore, infants were habituated to several different objects undergoing either several kinds of rotation or several kinds of nonrotation; they were then tested for their responsiveness to a novel object undergoing a novel rotation and a nonrotation, or a novel nonrotation and a rotation. A significant differentiation between novel and familiar motions in this case would suggest that discrimination and recognition of rotation or nonrotation was highly general and not based on any particular pattern of stimulation. The subjects for this study were 20 infants (11 males and 9 females) with a mean age of 22 weeks and 4 days. Another 11 infants were seen but were not included either because they were too fussy (9) or because of experimenter error (2). There were five different objects which were rectilinear solids of different shapes, some asymmetrical. For each infant, four objects were used for habituation and the fifth was used for the test trials. For every rotation, the object remained in the same place while it moved around one of its own axes; all rotations, therefore, involved changes in orientation of the objects with respect to an objective three-dimensional framework. The five motions were: (1) oscillating rotations of 180” around the vertical axis; (2) similar oscillations around The research reported here was supported by Grant BNS 80-13064 from the National Science Foundation. I wish to thank John Ruff and Gerald Turkewitz for reading and discussing an earlier version of this paper with me. Correspondence and requests for reprints should be addressed to Holly A. Ruff, 222 Kennedy Center, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, NY 10461.
365
366
RUFF
the horizontal axis; (3) similar oscillations around a line toward the infant;! (4) full rotations around the vertical axis; and (5) full rotatations around a line toward the infant. In contrast to the rotations, the five nonrotations involved no changes in orientation with respect to an objective framework. These motions, each of which covered about 10 cm in different directions from the center position, were: (1) moving laterally in alternate directions in the frontal plane; (2) moving alternately toward and away from the infant; (3) moving up and down; (4) moving so that a circle was described in the frontal plane; and (5) moving so that a square was described in the frontal plane. Every infant was habituated to four of the motions in one category and the fifth was used as the “familiar” motion on the test trials. Each infant was seated on the mother’s lap at a table; the objects were presented at the edge of the table opposite the infant (76.2 cm away). A large backdrop excluded the experimenter and observer from the infant’s view. Each object was rigidly fixed to a dowel 1.3 cm in diameter and 30.5 cm in length; the dowel allowed the object to be moved by hand without the hands being visible to the infant. All motions were well-practiced and smooth; there was a set number of cycles per trial, and no differences among the motions in the degree of variability were observed. For the habituation periods of both the rotation and nonrotation conditions, the four chosen motions were combined with the four objects, so that there were 16 possible combinations of object and motion. Every trial was 30 s long, and an observer recorded the duration of the infant’s looking at the object on a digital stopclock. An infant control procedure was used; that is, trials were continued until the duration of the infant’s looking in two consecutive trials was less than or equal to half of the mean duration of looking on the first two trials. A maximum of 20 trials was presented.* HaIf of the infants were habituated to rotation and half to nonrotation, leading to two conditions defined by the type of motion presented during habituation. When the criterion was reached, the infant was presented with two test trials, one in which the fifth and novel object rotated and the other in which the object moved without rotating. The order in which familiar and novel motions were presented was completely counterbalanced. Considering the motions presented successively in the test trials as a pair, five pairs of rotation and nonrotation were used. Both members of the pair involved either continuous motion (e.g., an object rotating fully in the frontal plane vs. a nonrotating object moving in a circular path in the frontal plane) or changes in direction (oscillating around the vertical axis vs. translating laterally back and forth in the frontal plane). Each pair was presented to four infants. ’ The phrase “a line toward the infant” is used rather than the “line of sight” because active observers with binocular perception can never see a three-dimensional object rotating exactly around the line of sight. Even slight deviations from the line of sight mean that rotation of the object
will lead to perspective changes and new points z Three infants in the rotation condition and two
reach criterion in 20 trials; 75% of the initial level.
on the 19th
and 20th
trials,
coming infants
into view. in the nonrotation
however,
:heir
looking
condition ranged
did not from
52 to
PREFERENCE
FOR ROTATION
367
It should be noted that though the study involved an habituation period, there were no postcriterion trials; in this case, the most appropriate approach to the data is to consider the habituation period as a period of familiarization, and to compare the test trials directly to each other rather than to the criterion trials. Descriptive data for the two conditions are presented in Table 1. The data show that there was not much difference in the looking at novel and familiar motions on the test trials, but there was a large difference in the looking, between rotating and nonrotating objects. In fact, every infant looked longer at the rotating object than at the nonrotating object on the test trials, a highly significant difference, t(19) = 7.40, p < .OOl. The subsequent analyses, therefore, focus mainly on the difference between rotating and nonrotating objects rather than on the distinction between novel and familiar motions. A 2 (Condition) x 4 (Trials: initial, criterion, rotation, nonrotation) analysis of variance revealed no Condition effect, F< 1, but a highly significant Trials effect, F(3,54) =42.60, p< .OOl; and a significant Condition by Trial interaction, F(3,54) = 6.70, p< .005. A Neuman-Keuls post hoc analysis showed that looking on the criterion trials was significantly lower than looking on the initial trials, and that looking at the rotating objects was significantly higher than looking on either the criterion trials or the test trials with nonrotating objects. An examination of the means in Table 1 shows that the interaction is caused by the fact that the infants in the nonrotation condition looked less during the habituation phase and more during the test phase than the infants in the rotation condition. There is no evidence that the two conditions differed in the degree of decrement during habituation, t(18) = .88, or in the extent of the preference for rotation on the test trials, ~(18) = 1.39. In order to provide each infant with a single score for the two test trials and to test possible effects of order and novelty, the percent of total time spent looking at the rotating object was entered into a 2 (rotation first or second) x 2 (rotation novel or familiar) analysis of variance. There were no effects of TABLE Descriptive
Data
for
the
Rotation
1 and
Nonrototion
Conditions
Motion
used
Rotation Mean Looking Time Trials in Seconds Mean Looking Time Number
of Trials
in First in Two
Two
Habituation
Criterion
to Criterion
Trials
during
Habituation Nonrotation
27.0
20.4
(2.9) 12.9
(6.0)
(5.2) 14.9
w-3) (4.5) 25.1
0.3 13.4
Looking
Time
to Test
Rotation
(5.6) 19.1 (familiar)
Time
to Test
Nonrotation
(8.3) 9.4 (novel)
(6.1)
Looking
(6.7)
(4.8)
10.9
368
RUFF
order, F(1,16)=0; or of novelty, F(1,16)= .Ol, and no interaction, F(1,16)= .02. These results demonstrate an overwhelming preference for rotation that generalizes across five pairs of rotation and nonrotation and that occurs regardless of the nature of the habituation phase (see Slater, Earle, Morison, & Rose, 1985). The preference is consistent with the results of other studies (Gibson, Owsley, & Johnston, 1978; Ruff, 1985). The fact that Gibson et al. found only one significant difference among three possible differences between three rotations and one translation is probably related to the much smaller extent of rotation used in that study. Because the infants in both rotation and nonrotation conditions showed habituation and equivalent number of trials to criterion, it seemed reasonable to compare the responses to specific rotations and specific nonrotations in the habituation period using the mean looking time to each particular motion during the habituation period. Because each motion appeared once in each set of four trials, the mean was calculated on the basis of complete blocks of four trials; if, for example, an infant had 18 trials, a mean for each of the different motions was derived from the first 16 trials. Although there were five motions altogether, only four were used for any one infant; because of this, each infant contributed data for only four of the five means. The mean duration of looking per trial ranged from 22.3 to 25.4 s for the five/rotations and from 17.1 to 19.7 s for the five nonrotations. The mean for each rotation was, therefore, higher than the mean for any nonrotation. When the 25 possible combinations of rotations and nonrotations were compared with t tests for independent means, 22 of the 25 achieved the .l level; of these, 14 were significant beyond the .05 level. Although theoretically some of these tests could be significant by chance, the probability of all of the 25 possible comparisons favoring rotation by chance is < .OOl. In sum, these data also demonstrate a very strong preference for rotation, in this case, across infants. In contrast, when the five motions within each category were compared with one-way repeated measures analyses of variance,’ there was no evidence of differentiation among rotations, F(4,36) = 1.98, or among nonrotations, F(4,36) = .98. These data suggest that the motions in each category were equivalent in some respect. There could be several reasons for the failure to demonstrate a preference for the novel motion. Because other studies (Ruff, 1985) have shown that infants habitutated to one kind of rotation will generalize to that motion made with a novel object and recover to a nonrotation of that object, the most likely interpretation is that the infants in this study were overwhelmed by the variety of motions and objects involved and consequently responded mainly on the ’ The group mean was used to fill in the fifth motion for each infant so that the repeated measures analysis of variance could be conducted. Since such a procedure increases the degrees of freedom, as well as taki.lg advantage of the correlation among scores, the power of the test for detecting differences is increased. The analysis could, therefore, be considered a conservative test of the hypothesis that there were no differences among motions within conditions.
PREFERENCE
FOR ROTATION
369
basis of a spontaneous preference. The fact that the preference was in evidence across infants in the habituation period would support this possibility. There was no evidence, therefore, that infants at this age generalize from their experience with various rotations or nonrotations to a novel rotation or a novel nonrotation. On the other hand, because the preference for rotation occurred with a variety of different examples in each category and because there was more evidence for discriminations between categories than within categories, one must conclude then that infants of 5 months discriminate between rotations and nonrotations in a very general way. The preference for rotating objects may exist because rotations of an object generally provide more information about it than do nonrotations. That is, the changes that take place during rotation are more likely to reveal previously unprojected parts of the object than would the generally slighter changes involved in viewing an object that is moving but not rotating. In turn, the more extensive changes may increase the likelihood that the viewer will detect invariants specifying the structure of the object. Rotation may also be generally more stimulating or arousing than nonrotation. Both of these possibilities would suggest that if the changes produced by a rotating object were reduced by decreasing the angle of rotation and the changes produced by a nonrotating object were increased, there would be a reversal of the preference. In contrast, it is possible that the infants were responding to a qualitative difference between the two kinds of motion, for example, the presence or absence of orientation changes with respect to an objective framework. The data from this study are not definitive regarding the causes of the preference. Because of its strength and consistency, however, the preference could serve as a useful basis for the exploration of the stimulus factors that are involved in the perception of motion. REFERENCES Gibson, Ruff, Slater,
E.J., fants:
Owsley, C.J., Differentiation
H.A. (1985). old infants.
& Johnston, J. (1978). Perception of invariants by five-month-old inof two types of motion. Developmental Psychology, 14, 407-415.
Detection
Developmenrol
of information
Ps.vcholog.v,
A., Earle, D.C., Morison, V., teraction with habituation-induced
Psychology,
specifying
the motion
of objects
by 3- and 5-month-
21, 295-305.
& Rose, D. (1985). Pattern novelty preferences.
preferences
Journal
at birth
and their
of Experimenral
in-
Child
39, 37-54. 6 May
1986;
Revised
27 February
1907
W