Chi-Tai Huang a,*, Yi-Chun Chen b
a Department of Psychology and Research Center for Mind, Brain and Learning, National Chengchi University, Taipei, 11605, Taiwan
b Department of Human Development, Tzu Chi University, Hualien 97004, Taiwan
*Address correspondence to:
Department of Psychology National Chengchi University
No.64, Sec.2, Zhinan Rd., Wenshan District Taipei, 11605 Taiwan
Tel: +886-2-29393091 ext.62379 Fax: +886-2-29390644
E-mail: ucjtchu@nccu.edu.tw (C.-T. Huang)
38 Abstract
This study clarified the nature of prior experience during observational learning to facilitate the acquisition of tool use, as was documented by Carpenter, Call and Tomasello (2002). In order to rule out the possibility of affordance detection prior to seeing the main display, we devised two sets of stimuli to present the prior display and the main display, separately. Before the demonstration of how to retrieve a toy from the target box, children in the two Prior Intention conditions either saw the experimenter (E) retrieve a toy from a distinctively different box or saw the doors of this box already opened. In the two No Prior Intention conditions, children saw E approach this different box with irrelevant acts or directly proceed with the main display. Children opened the target box more efficiently in the two Prior Intention conditions than in the two No Prior Intention conditions. Children in the two Prior Intention conditions also showed a proclivity, promoted by the tendency to follow E’s sequence, to devise their own strategy to reproduce the end result. Thus, given the unavailability of task-relevant affordances during the prior display, children’s
understanding of others’ prior intention seems to play a significant role in addition to affordance learning.
Keywords: Observational learning; Imitation; Emulation; Intention; Affordances
39 Introduction
Imitation has long been seen as a mechanism of social cognitive abilities such as empathy, theory of mind, and action understanding (Iacoboni, 2009; Meltzoff, 2002).
Until recently, its complexity and diverse nature began to filter into developmental research. Concepts and methods from comparative sciences have been influential in bringing this major advance. Recent evidence showed that the scope of imitation typically studied by developmentalists may actually present different social learning mechanisms. Therefore, imitation takes on many manifestations. For example, after watching two adults take turns to retrieve a toy with a different strategy, preschoolers prefer to reproduce the strategy of the adult, who remains in the room to give them the apparatus (Nielsen & Blank, 2011). They could also be primed to mimic a model following ostracism experience (seeing one shape ostracized by a group of other shapes on the computer display; Over & Carpenter, 2009). Under certain
circumstances, however, they simplify the observed strategy, driven by a tendency to emulate the end result (Williamson & Markman, 2005; Nielsen, 2006). While these various manifestations suggest that imitation is a flexible learning mechanism, it is likely that we overlook the various other forms of social learning discussed within comparative psychology (Want & Harris, 2002).
Recently, there has been considerable interest in studying infants’ intention-reading ability capitalizing on their imitative tendencies. A growing consensus suggests that imitation is a viable tool for studying infants’ understanding of intention. Various lines of research have contributed to this. First, results from Meltzoff’s (1995) behavioral reenactment procedure show that 18-month-old infants imitate not what they observe the adult do (e.g. unsuccessful attempts to drape a loop over a prong) but rather what they infer the adult intended to do (i.e. draping the loop over the prong). Similarly, Carpenter, Akhtar and Tomasello (1998) show that 14- to 18-month-old infants are more likely, for example, to pull a ring, than to move a top, when they see E pull the ring on purpose (by saying “There!”) and move the top by accident (by saying
“Whoops!”) even if both acts lead to the same end result. Several recent models of imitation have also identified goal attribution as a key process underlying the
continuum of copying fidelity, stating that precision of imitation varies as a function of the clarity of the goal. Therefore, imitation tends to be less precise when the goal of a demonstration is clearly set out (Bekkering, Wohlschläger, & Gattis, 2000;
Carpenter, Call, & Tomasello, 2005; Williamson & Markman, 2006). A slightly different view, however, suggests that imitation is guided by evaluation of the efficiency of the means in relation to the goal and the actor’s situation constraints (Gergely & Csibra, 2006). Infants thus selectively reproduce the “head action” to switch on a light box when E’s hands are free placed on the table than when her hands tightly wrapped with a blanket (Gergely, Bekkering, & Király, 2002). These studies concern children’s understanding of a person’s intentions emerging from the
performance of a given action, what Searle (1983) called “intentions in action.”
In contrast, little research has been done to examine imitative learning in relation to a different kind of intention that specifies knowledge about a person’s reasons for action before he acts (Searle, 1983). This concept of “prior intention” is exemplified by the study of Carpenter, Call and Tomasello (2002), who explored whether
2-year-olds would benefit from prior experience when it could be related to the agent’s reason for proceeding further. Participants observed E pull out a pin and then open the door of a box, in order to retrieve a toy from inside. Some children were informed about E’s motive before E gave this demonstration: they saw E attempt but
40
fail to open the box (Prior: Failed Attempt), or open three other different containers (Prior: Context), or simply watched the already-opened box (Prior: End State).These children observationally learned about the solution more efficiently than those who saw E directly demonstrate the solution (No Prior: No Pre-Demo) or firstly approach the box with irrelevant actions (No Prior: Irrelevant Action).
In Carpenter et al.’s study, E’s prior intentions in the Prior: Failed Attempt and Prior: End State conditions were delivered with the same apparatus that was used for subsequent demonstration of the target acts. Then it is possible that the
target-act-relevant parts of the apparatus were made more noticeable prior to the demonstration. A number of studies have suggested that infants’ and preschoolers’
reenactment of a person’s intended but unfulfilled acts is due, not to intention reading, but to a range of emulation learning possibilities (Huang, Heyes, & Charman, 2002, 2006). For example, they can emulate based on the object movements critical to producing a desirable outcome (Huang & Charman, 2005; Slaughter & Corbett, 2007;
Thompson & Russell, 2004). They can also emulate by exposure to the initial and end states of a combinatory object set (Huang, in press; Huang et al., 2002, 2006). The evidence of imitation through “ghost” and outcome-based conditions raises the possibility that affordance detection may influence children’s interpretation of the prior event in the Prior: Failed Attempt and Prior: End State conditions. Although the performance in the Prior: Context condition (E interacted with different containers) may independently indicate sensitivity to prior intention, it did not preclude the possibility that affordance learning facilitated a similar level of performance in the two other Prior Intention conditions.
In the present study, we modified three conditions in the Carpenter et al. study (Prior: Context, Prior: End State, and No Prior: Irrelevant Action) by using two
different tasks to present the prior event and the target solutions, respectively. The two tasks were similar in the number of components to be dealt with and manual dexterity involved, but differed in color, shape and methods of unblocking the components. The first modification was to examine whether task-relevant affordances in the prior event influence children’s performance in the Prior: End State condition. We illuminated the role of E’s prior intention by eliminating information about target-relevant
affordances during observation of the prior acts. If, as Carpenter et al. suggest, children’s success in the Prior: End State condition depends on understanding of E’s prior intention (rather than affordance detection), the lack of target-relevant
affordances would not detract from successful tool use. Alternatively, if children’s success in the Prior: End State condition is sensitive to both E’s prior intention and target-relevant affordances, the children would perform more poorly. A second modification was the amount of context information in the Prior: Context condition.
In the Carpenter et al. study, contextual cues about prior intention were displayed by E opening a series of three containers. In contrast, in the present study, E opened only one box (the same box used to display end-state information). It is interesting to see whether children’s detection of prior intention in this condition is susceptible to the amount of contextual cues. This would help further to clarify factors related to observational experience before the demonstration.
Method
Participants
Eighty children (43 boys and 37 girls) who were 2.5 years old (M = 31 months, SD
41
= 1.5 months, range= 28-33 months) participated in this study. All were ethnic
Chinese living in Hualien county of Taiwan. Nineteen additional children were tested but excluded from the final sample due to experimental error (7), fussiness (1), noncooperation (4), non-response (5), or shyness (2).
Materials
The test stimuli (Fig.1) consisted of a square box (30 × 30 × 30 cm) and a
birdhouse (30 × 30 × 45 cm). They were visually distinct in terms of shape, color and ways to unblock the components. The square box was used for displaying E’s
intentional stance before the demonstration of how to open the birdhouse (referred to as the prior task and the target task, respectively). Both boxes consisted of a front door that opened from outside, and an inside door that divided the interior of the box into two compartments. A small toy was placed behind the inside door. To retrieve the toy, the two doors had to be unblocked. The stimuli were designed so that the doors comprising each box could be opened independently.
Procedure
All participants were tested individually in a university laboratory. They were asked to sit down on the floor with E. The square box and birdhouse covered with cloth were placed in front of them. When the child appeared to be comfortable, E removed the cover and started the task presentation specified for the experimental condition.
Sessions were videotaped for later coding.
There were five conditions (one Baseline, two Prior Intention, and two No Prior Intention conditions). In the latter four conditions all children saw E demonstrate how to retrieve the toy inside the birdhouse, but were given different information about the prior task before this target demonstration. In the two Prior Intention conditions, children saw either the solutions to opening the square box (Prior: Context) or the box already opened (Prior: End State). In the two No Prior Intention conditions, the children saw either no prior event (No Prior: Target-Demo only) or acts unrelated to opening the box (No Prior: Irrelevant Action).
No Prior: Target-Demo only. Children in this condition saw the target
demonstration only. E removed the cloth over the birdhouse. The front door faced the child. “Look over here!’’ she said and then started the demonstration. The sequence in which she firstly dealt with the front door or the inside door was counterbalanced:
half the children saw E push the knob, pull up the door, then pull out the pin, and finally take out the toy; half saw E pull out the pin, push the knob, pull up the door, and then take out the toy. In half of each of these trials, the front door bounced open following E pushing the knob and in half the front door remained closed. Note that when the front door was firstly opened, this gave the child a chance to witness the inside door slide aside while watching E pull out the pin. After a brief display of the toy, E put it back to the birdhouse. E covered the box with cloth so that its initial state was restored unseen by the child. At last, E removed the cover, presented the child with the box, and said “Now, it’s your turn.”
Prior: Context. Children in this condition saw the prior demonstration and then the target demonstration. The prior demonstration served to suggest E’s intention to retrieve a toy. E removed the cloth over the square box with its front door facing the child. She said “Look over here!’’ and then started the demonstration. As in the target demonstration, the sequence in which two doors were unblocked was counterbalanced:
42
half the children saw E pull the front door downward (with hands grasping the notches), pull the handle upward; half saw the two solutions in reverse order. Similar to the birdhouse, when the front door of the square box was firstly opened, the child had a chance to watch the inside door move vertically while seeing E pull up the handle. E displayed the toy briefly, put it back to the box, and covered the box with cloth. She went on to the birdhouse, and then the modeling procedure was identical to that involved in the No Prior: Target-Demo Only condition.
Prior: End State. Children in this condition observed the end states of the square box and then the solution to the birdhouse. First, E removed the cloth over the square box, which was set in the initial state. She said “Look at this,” and led the child to look at the box from different angles. Next, she covered the box with cloth, unblocked one door out of the child’s view, and then removed the cloth to show the door already opened. After that, E covered the box with cloth again, unblocked the other door, and removed the cloth. This time the end states of both doors were displayed. E took out the toy, displayed it briefly, and put it back to the box. After putting the cloth over the square box, she went on to uncover the birdhouse, and started the target
demonstration as described in the procedure for the No Prior: Target-Demo Only condition. Half the children saw the front door firstly opened, and half saw the inside door firstly opened (protruding from the top surface of the box).
No Prior: Irrelevant Action. Children in this condition watched a demonstration unrelated to toy retrieval and then the target demonstration. This condition controlled for the possibility that seeing E manipulate the box facilitated subsequent social learning without knowing E’s prior intention. E uncovered the square box, said “Look over here!’’ and then showed the two unrelated actions. Half the children observed E knock on the front door three times and then twist the handle three times; half saw these acts in reverse order. E then covered the box with cloth, went on to uncover the birdhouse, and started the target demonstration as described in the procedure for the No Prior: Target-Demo Only condition.
Baseline. Children in this condition did not see any action on each set of stimuli.
The Baseline was included to assess the spontaneous production of the solutions to the target task. E uncovered the birdhouse, placed it in front of the child, and said
“You can play with this.”
Scoring
Children were allowed 60 s to solve the target task. We scored whether E’s acts were reproduced in the response period: (1) pulling the pin out (the other end of the pin revealed had to include at least two yellow stripes); (2) pulling the knob upward (the front door moved all the way up and the door elevation angle had to be greater than 45 degrees); (3) pushing the knob (the door did not have to bounce up). Because unlocking the door catch was not required for opening the front door, the third act was nonfunctional. Therefore, success at opening the birdhouse required only the first two components1
Children’s responses were coded from digitized videotaped records by the second author. Inter-rater reliability was assessed for 40% of the data (8 children per
condition) independently coded by an undergraduate who was familiarized with the scoring criteria. Agreements were 100% for open birdhouse, pull pin, pull knob, and
. Latency to open the birdhouse (defined as both doors in unlocked states) was timed when the child first touched the apparatus.
1
43 push knob.
Results
Task completion Success
Fig. 2 shows the percentage of children who completed the target task within 60 s in each condition. There were significant differences between conditions,χ2 (4, N = 80) = 18.71, p = .001. Only one child (6%) in the Baseline condition opened the birdhouse, compared with 69% and 75% of the children who made it in the Prior: End State and Prior: Context conditions (hereafter referred to as the Prior Intention
conditions), respectively. In the No Prior: Target-Demo Only and No Prior: Irrelevant Action conditions (hereafter referred to as the No Prior Intention conditions), 43%
and 50% of the trials resulted in success, respectively. In addition to a social learning effect, a prior intention effect was also evident, with more children being successful in the Prior Intention conditions versus the No Prior Intention conditions, χ2 (1, N = 64)
= 4.15，p = .042.
Latency
Because the numbers of children who opened the birdhouse varied greatly across conditions, we ranked their latencies in descending order. The top 25 percent of the children were classified as fast learners (M = 7.94 s, SD = 1.56) and the bottom 25 percent as slow learners (M = 42.00 s, SD = 10.30), with 10 in each group. It turns out that while all fast learners were from the Prior Intention conditions, slow learners consisted of 5 Prior Intention children, 4 No Prior Intention children and 1 Baseline child. Children appear to perform more efficiently in the Prior Intention conditions versus the other three conditions, Fisher test, p < .04.
Success and E’s Action sequence
When the demonstration started with the front door, 31% and 16% of the children in the Prior Intention and No Prior Intention conditions completed the task,
respectively. When the demonstration started with the inside door, 41% and 31% of the children in the above conditions completed the task, respectively. In either case, there was no significant difference between sequences, Fisher test, ns. The
percentages of the children who followed E’s sequence for door opening were 87%
and 73% in the Prior Intention and No Prior Intention conditions, respectively. The tendency to follow E’s sequence was reliable in the Prior Intention conditions, Binominal test (P = .5), p < .001, but not in the No Prior Intention conditions, p > .1.
Methods used to open doors
Task completion could be achieved by reproducing demonstrated solutions or devising alternative strategies their own. We observed that children opened the inside door by pulling out the pin, as was demonstrated, or directly sliding it aside by inserting their fingers into the gap between the door panel and the door frame. The latter strategy worked only after the front door had already been opened. Similarly, they opened the front door by pushing and then pulling up the knob, or simply neglecting the unnecessary push2
2
.
44
On the whole, children tended to follow E’s solution to opening the inside door.
The percentages of children who pulled out the pin were 40% and 27% in the Prior Intention and No Prior Intention conditions, respectively, compared to 22% and 11%
using their fingers to slide the door, respectively, Fisher test, ns. On the other hand, there was an overall tendency to neglect the stylistic unnecessary act when dealing with the front door. The percentages of children who pushed the knob before pulling it up were 14% and 5% in the Prior Intention and No Prior Intention conditions,
respectively, compared to 49% and 32% who directly pulled up the knob, respectively, Fisher test, ns.
Components of the demonstration Object parts first touched
Fig. 3 shows the percentages of each object part children first touched. Although children appear ready to approach the pin and knob after seeing the demonstration, there was a significant difference between the Prior Intention and No Prior Intention conditions in the percentage of first-touched object parts that were demonstrator consistent, χ2 (1, N = 64) = 4.06, p = .044. The percentages of first-touched object parts consistent with the demonstration were: 56% in the Prior Intention conditions;
and 31% in the No Prior Intention conditions.
Reproduction of each demonstrated component
Fig. 4 presents the percentages of children who reproduced each of three
demonstrated components regardless of success. There were marginally significant
demonstrated components regardless of success. There were marginally significant