- Email: [email protected]
Imitation learning of information-processing
OURNAL
OF EXPERIMENTAL
Imitation
CHILL)
Learning
PSYCHOLOGY
of
12, 223-227 (19711
Information-Processing1
PETER A. LAMAL” University
of North
Carolina
at Charlotte
Modified twentyquestions problems were solved by 72 8s in a study of the influence of an adult model on the information-processing strateg) used by children of various grade levels. The independent variables were: (a) information-processing of model (hypothesis-scanning, constraint-seeking, or control), (b) sex of model, (c) school grade (three, five, or seven), (d) sex of 8. Major results were: (a) fewer questions to solution with the constraint-seeking model than the hypothesis-testing model or control, which did not differ; (b) both a smaller percentage of constraint, questions and fewer items per question with the hypothesis-scanning model; (c) shorter time to solution for A’s who had observed a model than for control Ss; (d) a lower percentage of constraints for third graders, as well as fewer items per question for third and fifth graders; (e) shorter time to solution for fifth graders than third graders, with no difference between fifth and seventh graders and seventh and third graders; (1) no effects for model sex or S sex.
Laughlin, Moss, and Miller (1969) reported a study of imitation of information-processing among third, fifth, and seventh grade children. Their study combined the approach of Bruner, Olver, and Greenfield (1966) with that of Bandura and Walters (1963). Bruner et al. (1966) had used a modified game of “twenty questions” in studying the types of questions grade-school children asked. With this method, the child was presented with an array (consisting of words or pictures) of 42 common objects and the task was to determine which object t’he experimenter (E) had selected as the target object. It was found that children typically employed one or both of two basic information-processing strategies. The “hypothesis-scanning” strategy consisted of a series of unrelated specific questions, for example, “Is it the hammer?” or ‘(1s it the apple?” “Constraint-seeking,” the other information-processing strategy, is evidenced when the child asks questions which include at least two objects, for example, “Is it an animal?” or “Is it something which is worn?” Constraint-seeking is the
learning
‘This research was supported by Grant 30-7116,4 from the Research Foundation of State University of New York. ‘The author wishes to thank Dr. Clayt Farrell of the Congdon CampusSchool, SUNY-College at F’otsdam for his cooperation. as well as Kathy Sullivan and Rill Wilson for data collection. 223
more clfficient strategy iii that it yields more information per question. It ~a:: found that, in general, the use of constraint-seeking relative to liyl)otliesis-scanrling increased with age. Bandura and Walters ( 1963) have provided evidence that children can learn novel responses simply from observing another person perform the responses. Thus, as with the experiment of Laughlin et (II. (19691, the purpose of the present experiment was to relate the al)proaches of Bruner et al. (1966) and Bandura and Walters (1963 1 in a study of the information-processing behavior of children who had first observed an adult model solve a sample problem by using a strategy of hypothesis-scanning or constraint-seeking. For control subjects the adult model posed the problem to the subject (S), but did not first solve a problem. The present experiment differs from the Laughlin et al. (1969) study in that only a verbal display was used in the present experiment. Also in the present experiment the possible effect of systematic differences in intellectual ability among the groupings of Ss was controlled by using S’s’ I.Q. scores (Large-Thorndike) as the covariate in an analysis of covariance. A further difference from the previous study was the addition of time to solution as a dependent variable in the present experiment. METHOD
Design and Subjects The experimental design was a 3 X 2 X 3 X 2 repeated-measures factorial with the following variables: (a) information-processing of model (hypothesis-scanning, constraint-seeking, or control), (b) sex of model, (c) school grade (three, five or seven), (CE)sex of S. The Xs were 72 children from the campus school of the State University of New York-College at Potsdam. Two Ss were randomly assigned to each of the 36 cells of the design. Information
Display
The display consisted of a 6 X 7 matrix of lettered names of the 42 common objects used by Bruner et al. (1966) in their pictorial display. A smaller verbal array consisting of the lettered names of 16 objects in a 4 X 4 matrix was used to demonst,rate the problem during the initial instructions. Procedure and Instructions The instructions were the same as those used by Laughlin et al. (1969). In hypothesis-scanning conditions E asked a series of specific hypotheses (e.g., “Is it the apple?“) until he solved the problem. In con-
IMITATION
225
LEARNING
straint-seeking conditions he solved the problem as efficiently as possible by a series of constraint questions. After observing E solve the problem, S solved two problems selected by drawing pieces of paper from a new set of 42 pieces corresponding to the verbal array. In control conditions S solved one problem from the sample 4 X 4 array and two from the full array. In all conditions the problem was solved when S asked the correct hypothesis. Two Es (one male and one female) ran equal numbers of male and female 8s under all conditions at all grade levels. RESULTS
Number of questions to solution. The effect of the model was significant, F = 5.30, df = 2/43, (p < ,011. The means of this and all other significant differences are given in Table 1. Duncan’s multiple-range test indicated that &who had observed a constraint-seeking mndel asked a significantly fewer number of questions (p < .Ol) than did Ss who had observed an hypothesis-scanning model or control Ss. The latter two groups did not significantly differ from each other. Testing of all other main effects and first-order interactions yielded no significant results. Percentage of constraints. Each question asked by each S was scored as either a specific hypothesis or a constraint. The question was scored as a specific hypothesis if it referred to only one object (e.g., l‘Is it the airplane?“), and as a constraint if it referred to at least two objects (e.g., “Is it used for transportation?“). The number of constraint questions was divided by the total number of questions for a problem to obtain the percentage of const’raints. The effect of the model was significant, F = 14.45, df = 2/43 (p < .Ol). Duncan’s test indicated that 5s who had observed an hypothesis-scanning model asked a significantly smaller TABLE 1 MEAN NUMBER OF QUESTIONS TO SOLUTION, PERCENTAGE OF CONSTRAINTS, AVERAGE NUMBER OF ITEMS INCLUDED PER QUESTION, AND TIME T O SOLUTION Number of questions Control ASS Ss observing hypothesisscanning model Ss observing constraintseeking model Third-grade Ss Fifth-grade Ss Seventh-grade Ss
Percentage constraints
of
Items per question
Time to solution
36.46 37.13
0.41 0.31
2.37
8:376 4:45b
22
33
0.51
4 31
.5: 39”
3x. 04”
0.26
2.1c5
7
3.5.5w
0.39
33. z9a
0.48
3.07 4.56
a No significant differences. * Expressed in minut,es and seconds.
Xl1
: 32” 4: 00” 6 : 44”
226
I’ETEH
A.
LhMr\L
of constraint questions than Ss Who hd 0lX3XWtl :1 constraint-seeking model (p < .Ol) or control Ss (p < .05). The latter two groups did not significantly differ from each other. The effect, of grade was significant, F = 4.72, cJf = 2/43, ip < .05). Duncan’s test. revealed that third-grade Ss asked a significantly lower percentage of constraint quest,ions (p < .05) than did fifth or seventh graders, who did not significantly differ from each other. Testing of all other main effects and first-order interactions yielded no significant results. Items per question. This was a measure of the average number of items included per question. By definition, a specific hypothesis included one item, while a constraint question included two or more. The total number of items for all questions asked by an S was divided by the number of questions the S asked to obtain the average number of items included per question. The effect of the model was significant, F = 8.46, df = 2/43, (p < .Ol). Duncan’s test indicated that’ Ss who had observed an hypothesis-scanning model had significantly fewer items per question (p < .Ol) than Ss who had observed a constraint-seeking model. The latter did not significantly differ from control Ss, nor did control Ss significant’ly differ from S’s who had observed an hypothesis-scanning model. The effect of grade was significant, F = 9.12, df = 2/43, ip < .Ol). Duncan’s test’ revealed that third-grade S’s had significantly fewer items per question (p < .Ol) than did seventh-grade Ss. Fifth-grade Ss also had significantly fewer items per question (p < .05) than did seventh-grade Ss. Testing of all other main effects and first-order interactions yielded no significant results. Time to solution. The effect of the model was significant, F = 5.97. rlf = 2/43, (p < .Ol). Duncan’s test. showed that Ss who had observed an hypothesis-scanning model had a significantly shorter time to soluCon (p < .Ol) than did control Ss. Ss who had observed a constraintseeking model also had significantly shorter time to solution (p < .05) t’han cont’rol Ss. The effect of grade was significant, F = 5.14, df = 2/43, (p < .05). Duncan’s test indicated that, fifth-grade Ss had a significantly shorter time to solution (p < .05) than third-grade Ss. The fifth-grade S’s did not significantly differ from the sevent,h-grade Ss, nor did the seventh- and third-grade Ss significantly differ from each other. Testing of all other main effects and first -order interactions yielded no significant results.
pppyntnge
DISCUSSION
In general, the results were in agreement with the findings of Laughlin et al. (1969). Thus Ss who had observed a constraint-seeking model had fewer questions to solution than did control S’sor those who had observed
IMITATION
LEARNING
227
an hypothesis-scanning model. Also, Ss who had observed an hypothesisscanning model asked a smaller percentage of constraint questions than control Ss or those who had observed a constraint-seeking model. The model’s influence was again apparent in that Ss who had observed an hypothesis-scanning model had fewer items per question than Ss who had observed a constraint-seeking model. Furthermore, Ss who observed either type of model had shorter times to solution than control Ss. The idea that a model can facilitate or hinder performance (Laughlin et al., 1969), depending upon what is modeled, was thus supported by the present results. With respect to age differences in performance, t’he results were mixed. There were no significant differences among the age (school grade) groupings for number of questions to solution. A facilitation of performance with increasing school grade was found for two of the three remaining dependent variables. Third-grade Ss asked a lower percentage of constraint questions than did fifth or seventh graders, who did not differ from each other. This increased use of the constraint-seeking strategy from third to fifth grade is consistent with the finding of Bruner et a2. (1966). The beneficial effect of school grade was also indirectly indicated by the finding that third- and fifth-grade Ss had fewer items per question than seventh-grade Ss. The positive effect of school grade was not unequivocal, however, when time to solution was considered. Here it was found that fifth-grade Ss had shorter time to solution than thirdgrade Ss. Third-grade and seventh-grade Ss did not differ from each other, however. The significant Model X Grade interaction of the Laughlin et al. (1969) study, indicating that. the effect of the model was more pronounced for fifth graders than for third or seventh graders, was not found in the present study. Finally, the lack of a significant effect of model’s sex is consistent with the findings of some investigators, but contrary to those of others (Flanders, 1968). The lack of a significant difference between male and female Ss is consistent with the Laughlin et al. (1969) study. REFERENCES A.. AND WALTERS, R. H. Social learning and personality developme&. New York: Holt, R.inehart & Winston, 1963. BRUNER. J. S., OLVER, R. R., AND GREENFIELD. P. M. Studies irk cognitizle growth. NW York: Wiley, 1966. FLANDERS. J. P. A review of research on imitation behavior. Psychologicnl BTLUetirc, 1968, 69, 316-337. LAIEHLIN, P. R.. MOSS, I. L.. ,4ND MILLER, S. M, Information-p~occssing in &i&en as a function of adult model, stimulus display. school grade, and sex. Jourrln~ of Educational Psychology, 1969. 60, 188-193.
BANDURA,
OF EXPERIMENTAL
Imitation
CHILL)
Learning
PSYCHOLOGY
of
12, 223-227 (19711
Information-Processing1
PETER A. LAMAL” University
of North
Carolina
at Charlotte
Modified twentyquestions problems were solved by 72 8s in a study of the influence of an adult model on the information-processing strateg) used by children of various grade levels. The independent variables were: (a) information-processing of model (hypothesis-scanning, constraint-seeking, or control), (b) sex of model, (c) school grade (three, five, or seven), (d) sex of 8. Major results were: (a) fewer questions to solution with the constraint-seeking model than the hypothesis-testing model or control, which did not differ; (b) both a smaller percentage of constraint, questions and fewer items per question with the hypothesis-scanning model; (c) shorter time to solution for A’s who had observed a model than for control Ss; (d) a lower percentage of constraints for third graders, as well as fewer items per question for third and fifth graders; (e) shorter time to solution for fifth graders than third graders, with no difference between fifth and seventh graders and seventh and third graders; (1) no effects for model sex or S sex.
Laughlin, Moss, and Miller (1969) reported a study of imitation of information-processing among third, fifth, and seventh grade children. Their study combined the approach of Bruner, Olver, and Greenfield (1966) with that of Bandura and Walters (1963). Bruner et al. (1966) had used a modified game of “twenty questions” in studying the types of questions grade-school children asked. With this method, the child was presented with an array (consisting of words or pictures) of 42 common objects and the task was to determine which object t’he experimenter (E) had selected as the target object. It was found that children typically employed one or both of two basic information-processing strategies. The “hypothesis-scanning” strategy consisted of a series of unrelated specific questions, for example, “Is it the hammer?” or ‘(1s it the apple?” “Constraint-seeking,” the other information-processing strategy, is evidenced when the child asks questions which include at least two objects, for example, “Is it an animal?” or “Is it something which is worn?” Constraint-seeking is the
learning
‘This research was supported by Grant 30-7116,4 from the Research Foundation of State University of New York. ‘The author wishes to thank Dr. Clayt Farrell of the Congdon CampusSchool, SUNY-College at F’otsdam for his cooperation. as well as Kathy Sullivan and Rill Wilson for data collection. 223
more clfficient strategy iii that it yields more information per question. It ~a:: found that, in general, the use of constraint-seeking relative to liyl)otliesis-scanrling increased with age. Bandura and Walters ( 1963) have provided evidence that children can learn novel responses simply from observing another person perform the responses. Thus, as with the experiment of Laughlin et (II. (19691, the purpose of the present experiment was to relate the al)proaches of Bruner et al. (1966) and Bandura and Walters (1963 1 in a study of the information-processing behavior of children who had first observed an adult model solve a sample problem by using a strategy of hypothesis-scanning or constraint-seeking. For control subjects the adult model posed the problem to the subject (S), but did not first solve a problem. The present experiment differs from the Laughlin et al. (1969) study in that only a verbal display was used in the present experiment. Also in the present experiment the possible effect of systematic differences in intellectual ability among the groupings of Ss was controlled by using S’s’ I.Q. scores (Large-Thorndike) as the covariate in an analysis of covariance. A further difference from the previous study was the addition of time to solution as a dependent variable in the present experiment. METHOD
Design and Subjects The experimental design was a 3 X 2 X 3 X 2 repeated-measures factorial with the following variables: (a) information-processing of model (hypothesis-scanning, constraint-seeking, or control), (b) sex of model, (c) school grade (three, five or seven), (CE)sex of S. The Xs were 72 children from the campus school of the State University of New York-College at Potsdam. Two Ss were randomly assigned to each of the 36 cells of the design. Information
Display
The display consisted of a 6 X 7 matrix of lettered names of the 42 common objects used by Bruner et al. (1966) in their pictorial display. A smaller verbal array consisting of the lettered names of 16 objects in a 4 X 4 matrix was used to demonst,rate the problem during the initial instructions. Procedure and Instructions The instructions were the same as those used by Laughlin et al. (1969). In hypothesis-scanning conditions E asked a series of specific hypotheses (e.g., “Is it the apple?“) until he solved the problem. In con-
IMITATION
225
LEARNING
straint-seeking conditions he solved the problem as efficiently as possible by a series of constraint questions. After observing E solve the problem, S solved two problems selected by drawing pieces of paper from a new set of 42 pieces corresponding to the verbal array. In control conditions S solved one problem from the sample 4 X 4 array and two from the full array. In all conditions the problem was solved when S asked the correct hypothesis. Two Es (one male and one female) ran equal numbers of male and female 8s under all conditions at all grade levels. RESULTS
Number of questions to solution. The effect of the model was significant, F = 5.30, df = 2/43, (p < ,011. The means of this and all other significant differences are given in Table 1. Duncan’s multiple-range test indicated that &who had observed a constraint-seeking mndel asked a significantly fewer number of questions (p < .Ol) than did Ss who had observed an hypothesis-scanning model or control Ss. The latter two groups did not significantly differ from each other. Testing of all other main effects and first-order interactions yielded no significant results. Percentage of constraints. Each question asked by each S was scored as either a specific hypothesis or a constraint. The question was scored as a specific hypothesis if it referred to only one object (e.g., l‘Is it the airplane?“), and as a constraint if it referred to at least two objects (e.g., “Is it used for transportation?“). The number of constraint questions was divided by the total number of questions for a problem to obtain the percentage of const’raints. The effect of the model was significant, F = 14.45, df = 2/43 (p < .Ol). Duncan’s test indicated that 5s who had observed an hypothesis-scanning model asked a significantly smaller TABLE 1 MEAN NUMBER OF QUESTIONS TO SOLUTION, PERCENTAGE OF CONSTRAINTS, AVERAGE NUMBER OF ITEMS INCLUDED PER QUESTION, AND TIME T O SOLUTION Number of questions Control ASS Ss observing hypothesisscanning model Ss observing constraintseeking model Third-grade Ss Fifth-grade Ss Seventh-grade Ss
Percentage constraints
of
Items per question
Time to solution
36.46 37.13
0.41 0.31
2.37
8:376 4:45b
22
33
0.51
4 31
.5: 39”
3x. 04”
0.26
2.1c5
7
3.5.5w
0.39
33. z9a
0.48
3.07 4.56
a No significant differences. * Expressed in minut,es and seconds.
Xl1
: 32” 4: 00” 6 : 44”
226
I’ETEH
A.
LhMr\L
of constraint questions than Ss Who hd 0lX3XWtl :1 constraint-seeking model (p < .Ol) or control Ss (p < .05). The latter two groups did not significantly differ from each other. The effect, of grade was significant, F = 4.72, cJf = 2/43, ip < .05). Duncan’s test. revealed that third-grade Ss asked a significantly lower percentage of constraint quest,ions (p < .05) than did fifth or seventh graders, who did not significantly differ from each other. Testing of all other main effects and first-order interactions yielded no significant results. Items per question. This was a measure of the average number of items included per question. By definition, a specific hypothesis included one item, while a constraint question included two or more. The total number of items for all questions asked by an S was divided by the number of questions the S asked to obtain the average number of items included per question. The effect of the model was significant, F = 8.46, df = 2/43, (p < .Ol). Duncan’s test indicated that’ Ss who had observed an hypothesis-scanning model had significantly fewer items per question (p < .Ol) than Ss who had observed a constraint-seeking model. The latter did not significantly differ from control Ss, nor did control Ss significant’ly differ from S’s who had observed an hypothesis-scanning model. The effect of grade was significant, F = 9.12, df = 2/43, ip < .Ol). Duncan’s test’ revealed that third-grade S’s had significantly fewer items per question (p < .Ol) than did seventh-grade Ss. Fifth-grade Ss also had significantly fewer items per question (p < .05) than did seventh-grade Ss. Testing of all other main effects and first-order interactions yielded no significant results. Time to solution. The effect of the model was significant, F = 5.97. rlf = 2/43, (p < .Ol). Duncan’s test. showed that Ss who had observed an hypothesis-scanning model had a significantly shorter time to soluCon (p < .Ol) than did control Ss. Ss who had observed a constraintseeking model also had significantly shorter time to solution (p < .05) t’han cont’rol Ss. The effect of grade was significant, F = 5.14, df = 2/43, (p < .05). Duncan’s test indicated that, fifth-grade Ss had a significantly shorter time to solution (p < .05) than third-grade Ss. The fifth-grade S’s did not significantly differ from the sevent,h-grade Ss, nor did the seventh- and third-grade Ss significantly differ from each other. Testing of all other main effects and first -order interactions yielded no significant results.
pppyntnge
DISCUSSION
In general, the results were in agreement with the findings of Laughlin et al. (1969). Thus Ss who had observed a constraint-seeking model had fewer questions to solution than did control S’sor those who had observed
IMITATION
LEARNING
227
an hypothesis-scanning model. Also, Ss who had observed an hypothesisscanning model asked a smaller percentage of constraint questions than control Ss or those who had observed a constraint-seeking model. The model’s influence was again apparent in that Ss who had observed an hypothesis-scanning model had fewer items per question than Ss who had observed a constraint-seeking model. Furthermore, Ss who observed either type of model had shorter times to solution than control Ss. The idea that a model can facilitate or hinder performance (Laughlin et al., 1969), depending upon what is modeled, was thus supported by the present results. With respect to age differences in performance, t’he results were mixed. There were no significant differences among the age (school grade) groupings for number of questions to solution. A facilitation of performance with increasing school grade was found for two of the three remaining dependent variables. Third-grade Ss asked a lower percentage of constraint questions than did fifth or seventh graders, who did not differ from each other. This increased use of the constraint-seeking strategy from third to fifth grade is consistent with the finding of Bruner et a2. (1966). The beneficial effect of school grade was also indirectly indicated by the finding that third- and fifth-grade Ss had fewer items per question than seventh-grade Ss. The positive effect of school grade was not unequivocal, however, when time to solution was considered. Here it was found that fifth-grade Ss had shorter time to solution than thirdgrade Ss. Third-grade and seventh-grade Ss did not differ from each other, however. The significant Model X Grade interaction of the Laughlin et al. (1969) study, indicating that. the effect of the model was more pronounced for fifth graders than for third or seventh graders, was not found in the present study. Finally, the lack of a significant effect of model’s sex is consistent with the findings of some investigators, but contrary to those of others (Flanders, 1968). The lack of a significant difference between male and female Ss is consistent with the Laughlin et al. (1969) study. REFERENCES A.. AND WALTERS, R. H. Social learning and personality developme&. New York: Holt, R.inehart & Winston, 1963. BRUNER. J. S., OLVER, R. R., AND GREENFIELD. P. M. Studies irk cognitizle growth. NW York: Wiley, 1966. FLANDERS. J. P. A review of research on imitation behavior. Psychologicnl BTLUetirc, 1968, 69, 316-337. LAIEHLIN, P. R.. MOSS, I. L.. ,4ND MILLER, S. M, Information-p~occssing in &i&en as a function of adult model, stimulus display. school grade, and sex. Jourrln~ of Educational Psychology, 1969. 60, 188-193.
BANDURA,