JOURNAL

OF EXPERIMENTAL

CHILD

PSYCHOLOGY

50,

287-304 (19%)

Scripts or Scraps: Reconsidering the Development of Sequential Understanding PATRICIA J. BAUER University of Minnesota AND DONNA J. THAL university of California, San Diego, and San Diego State University A growing literature attests to temporally ordered recall of events by children under 2 years of age. Other data suggest a developmental sequence wherein the ability to reproduce unfamiliar and/or arbitrarily ordered events, and familiar events in other than canonical order develops well after the first ordered productions of events. Early ordering is thus argued to be dependent upon familiarity, rather than upon general temporal principles. This suggestion was investigated by using elicited imitation to assess 21-month-olds’ recall of familiar-canonical, familiar-reversed, novel-causal, and novel-arbitrary event sequences. Subjects reproduced canonical and both types of novel sequences in modeled order. On reversed sequences they vacillated between reproducing the events as modeled and “correcting” them to canonical order. The results suggest that temporal organization is not imposed upon an existing unordered event representation, but rather, is an integral aspect of the representation from its initial construction. It is suggested that young children’s difficulty with reversed sequences may be attributed to a reluctance to reorganize existing representations, rather than to the absence of applicable temporal principles. o 1990 Academic PKSS. IK. This research was supported in part by the John D. and Catherine T. MacArthur Foundation Network on the Transition from Infancy to Early Childhood, by a First Investigator award from the National Institute of Mental Health (grant PHS-N526107) to Donna J. Thal. and by National Science Foundation grants BNS-8109657 and BNS-8510218 to Jean M. Mandler. We thank the children and parents who participated in the study, and Heather Boles, Lana Renneberg, Mark Seamans, Karen Sheriff, and Elisabeth Vangsness for their help at various stages of the project. We also thank Jean M. Mandler, Elizabeth Bates, and Virginia Marchman for their helpful comments on an earlier version of this manuscript. Correspondence concerning this article should be sent to Patricia J. Bauer. Institute of Child Development, 51 East River Road, University of Minnesota, Minneapolis, MN 554550345. 287 0622-0965&X1 $3 .OO Copyright 0 1990 by Academic Press. Inc. All rights of reproduction in any form reserved.

288

BAUERANDTHAL

A consistent finding in the literature on children’s knowledge of events is that even preschoolers represent past events in an organized fashion. For example, 3-year-olds recount familiar events in the correct temporal order, and they tend to “correct” temporal violations introduced by reversed or scrambled models (Nelson & Gruendel, 1981, 1986). Kindergartners (Smith, Ratner, & Hobart. 1987) and 3- to 7-year-olds (Hudson, 1986) provide well-ordered accounts of an event after only one experience of it. Preschoolers also recall aspects of events with causal and temporal invariances more frequently and at a younger age than they do aspects of events lacking such structure (Hudson & Nelson, 1983; Slackman, Hudson, & Fivush, 1986). In short, like older children and adults, children as young as age 3 organize their recall of events around causal and temporal invariances (Schank & Abelson, 1977).’ There is growing evidence that similar patterns of event representation and recall also obtain for children younger than 3 years of age. Although the earliest age of onset of temporal ordering ability is not known (see Bauer 1989a for a report of correct temporal sequencing by 13-montholds), children 2 years of age and younger include temporal order in their representations of familiar events (e.g., Bauer & Mandler, 1989; Bauer & Shore, 1987; Ungerer, 1985). Further, like older preschoolers, I- to 2-year-olds organize their recall around causal and temporal invariances (Bauer. 3989b; Bauer & Mandler, 1989). However, another body of data suggests that, regardless of the presence of temporal sequencing ability, important changes in the ability to ltse temporal organization occur between the ages of 24 and 36 months. For example, Fenson and Ramsay (1980) noted changes in the tendency to spontaneously combine actions into ordered sequences at 24 months, and Shore, O’Connell, and Bates (1984) reported large proportional increases in the number of ordered actions produced in elicited play between 20 and 28 months. Additionally, O’Connell and Gerard (1985) reported that children do not use temporal information as a general orgunizing principle until 36 months of age. These data suggest that early sequencing skills, such as those demonstrated by Ungerer and by Bauer and her colleagues, might yet undergo qualitative change before the preschool years. The purpose of the present ’ Items in a sequence arc arbitrarily ordered when there is nothing inherent in the items to dictate their position in the sequence. For example, you may put on your coat either before or after putting on your hat. Causal relations exist when one item in a sequence rn~t be performed before another in the same sequence. For example, you must open the door before you can go through it. In this example. as in the others to be described. the relations really would be more appropriately described as enabling is, opening the door does not c~111se one to go through it. However. step possible. We will adopt the vocabulary of those working in these relations as “causal.” Regardless of whether these relations merely enabling. they imp/y temporal invariance.

than as causal. That it does make the next this area and refer to are truly causal or

DEVELOPMENT

OF

SEQUENTIAL

UNDERSTANDING

289

study is to provide information essential to evaluation of the status of early sequencing ability. Since O’Connell and Gerard provided the most detailed and, we believe, most controversial, account of a developmental sequence in the use of temporal organization, we summarize their data below. O’Connell and Gerard (1985) used elicited imitation to assess 20- to 36month-olds’ use of temporal organization in the representation of event sequences. Elicited imitation involves using props to model an action or sequence of actions and encouraging the child to imitate. Since the component actions are suggested by props, the methodology does not lend itself to investigation of the content of early representations of events. However, the order in which the components are reproduced does suggest the structure of the underlying representation. To test children’s use of temporal principles to organize their representations of events O’Connell and Gerard presented subjects with sequences of familiar events in canonical order (e.g., giving a bear a bath modeled as: put the bear in the bathtub, wash the bear, dry the bear) and in reverse order (e.g., dry the bear, wash the bear, put the bear in the bathtub). Twentymonth-olds produced the individual components of both sequence types, but they failed to produce them in any recognizable order. Twenty-fourmonth-olds showed a fledgling ability to reproduce canonical sequences. However, like the 20-month-olds, they showed no reliable ordering on the reversed sequences. Twenty-eight-month-olds produced both canonical and reversed sequences in the canonical order. That is, rather than producing the reversed sequences in the modeled order, they “corrected” them to canonical order. On the basis of these data, O’Connell and Gerard (1985) concluded that “At 20 months, children are unable to reproduce the order of a sequence regardless of the meaningfulness of the modeled event. . .” (p. 680), and that between 24 and 28 months, the ability to reproduce a modeled sequence is dependent upon familiarity with the event. As suggested above, O’Connell and Gerard (1985) observed important changes at 36 months. At that age the children accurately reproduced the order of the canonical sequences. On the reversed sequences they reliably produced the events in recognizable order, but not necessarily in modeled order. Instead, they vacillated between reproducing the events as modeled, and “correcting” them to canonical order. O’Connell and Gerard argued that in contrast to the 28-month-olds’ rigid adherence to canonical order, the 36-month-olds’ pattern of vacillation represented the beginning of reversibility, i.e., the ability to use order information in more than one direction. In addition, they argued that the oldest subjects’ performance on a third sequence type, namely, scrambled sequences of unrelated but familiar, meaningful, actions (e.g., wipe a bear’s mouth, the bear pays money, cover the bear with a blanket). also signaled

290

BAUER

AND

THAL

important changes in sequencing ability. On scrambled sequences the younger subjects produced few of the components, and even fewer of them in order. In contrast, the 36-month-olds were relatively accurate at reproducing these sequences. The above data have important implications for our understanding of children’s use of temporal principles in the organization of event representations. From them, O’Connell and Gerard concluded that “. . .temporal organization is imposed upon primitive representations of familiar events” (p. 671), and that “. . .only later is the principle of temporal organization generalized to novel events” (p. 681). The implication is that early event representations resemble an unorganized group of snapshots of individal components. The temporal organization necessary to combine the components into a coherent motion-picture is imposed later, with repeated experience of the event and/or with development (Bauer & Mandler, in press). Further, it is only after temporal principles have been applied to familiar events that they can be extended to novel ones. This conceptualization stands in sharp contrast to dominant views of early event representation. Both Nelson (1986) and Mandler (1986) argue that in their earliest event representations, children include the components of the event, as well as information about their order. This maxim is seen to apply to both familiar and novel events. Thus, temporal principles are not imposed upon an existing representation. but rather, are an integral aspect from the time of its initial construction. O’Connell and Gerard’s data also have important implications for expectations of developmental continuity or discontinuity in the use of temporal organization. They suggest a qualitative change in the use of temporal principles around 36 months. Before that time, temporal organization is dependent upon familiarity; after that time, it is operational, i.e., temporal principles can be used to reorganize existing event representations (i.e., reverse-order sequences) and to organize even unfamiliar and/or arbitrarily ordered events (i.e., scrambled sequences). The notion of stage-specific development in temporal organization is of course not new. Piaget (1926, 1969) argued that construction of ordered representations is dependent upon reversible operations not available until the concrete-operational period. It is now apparent that Piaget underestimated both the sequencing ability of young children and their ability to operate on existing representations; even O’Connell and Gerard’s 3year-olds showed some ability to reproduce sequences in unconventional orders. Thus, whether the capacity to reorganize existing representations is associated with a particular developmental event (e.g., construction of reversible mental operations) is questionable. Similarly, whether the ability to temporally organize unfamiliar and arbitrarily ordered sequences is associated with a particular developmental event also is ques-

DEVELOPMENT

OF

SEQUENTIAL

UNDERSTANDING

291

tionable. For 4- to 7-year-olds (Brown & Murphy, 1975; Fivush & Mandler, 1985) and adults (Mandler, 1986), recalling unfamiliar and/or arbitrary sequences is more difficult than recalling familiar ones. Thus, difficulty with such sequences does not disappear with development. Why might O’Connell and Gerard’s data have led them to conclude that younger children’s use of temporal organization is qualitatively different from that of older children? First, there is evidence that they underestimated the sequencing abilities of their subjects. Using elicited imitation of three- and four-element event sequences, studies have shown that while older children are more skilled at reproducing events than are younger children, subjects aged 16 to 30 months nevertheless recall familiar events in canonical order (Bauer & Mandler, 1989; Bauer & Shore, 1987; Ungerer, 1985). Even in spontaneous play, children’s earliest combinations of actions (around 24 months) are correctly ordered (Fenson & Ramsay, 1980). It is generally agreed that O’Connell and Gerard’s underestimation of sequencing ability is due to aspects of their testing procedure which contributed to low overall levels of performance (see Bauer & Mandler, 1989, in press; Bauer & Shore, 1987; Gerard, 1984; Mills, Mandler, Schreibman, & Oke, 1988; and Ungerer, 1985, for discussion). Second, O’Connell and Gerard used the “scrambled” sequences described above to test application of temporal principles to organization of unfamiliar and/or arbitrarily ordered events. However, since the scrambled sequences were concatenations of elements from several fumilk-w events, they were not truly novel. In fact, because each of the component acts was associated with a different familiar event, three different event representations might have been suggested by the modeled series. This could have resulted in interference from existing representations and thereby have had a detrimental effect on performance. A genuine test of whether order information is included in initial event representations requires that subjects be tested on events that truly are novel, and for which they have no existing representation, and therefore, no prior expectations. Further, a test of response to arbitrarily ordered events requires sequences that also have no causal or conventional constraints on their order. Sequences of this type were used by Bauer and Mandler (1989) and Bauer and Shore (1987) to test 16- to 23-month-olds’ response to unfamiliar events. They used novel sequences containing causal relations (e.g., making a rattle of two nesting cups and a rubber ball), and novel sequences that are arbitrarily ordered (e.g., making a picture with chalk and stickers). In both studies, subjects preserved the modeled order of the unfamiliar causally related event sequences. Thus, at least for causally connected events, familiarity with the sequence is not a pre-

292

BAUER

AND

THAI

requisite for successful ordered reproduction. In Bauer and Mandler, but not in Bauer and Shore, subjects also correctly ordered the unfamiliar arbitrary sequences. Finally, O’Connell and Gerard (1985) tested the ability to operate on existing representations by presenting familiar events in reverse order. However, they presented both canonical and reversed orders of the same events in the same session. As discussed elsewhere (Bauer & Mandler, 1989, in press; Bauer & Shore, 1987), this resulted in presentation of conflicting order information across trials, and is likely to have contributed to an overall lower level of performance. Additionally, it could have contributed to either of the patterned responses to the reversed sequences observed by O’Connell and Gerard (i.e., the tendency to “correct” the order, and to vacillate between production in canonical and modeled order). As a result, the data cannot be used to draw conclusions about young children’s response to reverse-order events. In light of the theoretical implications of O’Connell and Gerard’s work, and of the methodological concerns discussed above, it is important to reevaluate the pattern of development in sequencing performance they described. While the data provided by Ungerer (1985), Bauer and Shore (1987), and Bauer and Mandler (1989) indicate sequencing ability in children aged 2 and younger, these studies do not allow an adequate evaluation of all of O’Connell and Gerard’s claims. First, indications of children’s ability to use temporal principles to order familiar events have been based on a limited number of sequences. To ensure that sequencing ability is not unique to these sequences, a larger pool of events should be tested. Second, to provide an adequate foundation for evaluation of subjects’ general ordering abilities, data on reversed and unfamiliar and/or arbitrary-order sequences are necessary. Ungerer and Bauer and her colleagues did not include reverse order sequences in their studies. In addition, while in two studies subjects reproduced novel sequences characterized by causal relations (Bauer & Mandler, 1989; Bauer & Shore, 1987). in only one (Bauer & Mandler, 1989) did they reproduce arbitrarily ordered sequences. To ensure that subjects are able to use temporal principles on unfamiliar events with and without causal relations, it is desirable to replicate the effect observed in Bauer and Mandler. To pursue the above issues we used elicited imitation to assess 21- to 23-month-olds’ use of temporal information to organize recall of familiar and novel events. The age group was selected because it overlaps with the ages tested by Bauer and her colleagues and falls midway between the youngest two age groups tested by O’Connell and Gerard (i.e., 20 and 24 months). A single age group was selected so that we could sample a larger number of children. Subjects were tested on four different sequence types: (I) familiar events in canonical order; (2) familiar events in reverse order; (3) novel events characterized by causal or enabling

DEVELOPMENT

OF

SEQUENTIAL

relations among the components; enabling relations.

UNDERSTANDING

293

and (4) novel events lacking causal or

METHOD Subjects

Twenty-four children with a mean age of 21 months; 26 days (range 21;2 to 22;17) participated. An equal number of females and males was included. All subjects were seen for one 45 to 60-min session. Two additional subjects (one female and one male) were excluded from the final sample because they produced none of the target actions after modeling on three or more of the eight sequences. Procedure

Subjects were seen individually in a laboratory set up as a play room. After a warm-up period, they were seated in a booster-seat or on their parents’ lap across an adult-size table from the experimenter (E). The parents were asked not to suggest behaviors to their children and not to “remind” them of the next step in the sequence. Two practice sequences were used to acquaint the children with the imitation procedure. The practice sequences were: (1) roll a ball across the table and place it on a box; and (2) “drink” from a toy cup and place it on a saucer. For practice sequences, subjects first were allowed to manipulate the props prior to modeling. Next, the sequence was modeled by E, two times in succession, with narration. The props then were returned to the subjects and they were encouraged to imitate. If the subjects failed to produce both of the modeled behaviors, they were encouraged to do so with specific prompts, such as “Can you roll the ball and put it on the box, just like I did?” The procedure for the test sequences was identical to that in the practice sequences except that no prompts were used and the sequences contained three actions. Eight sequences were tested for each subject, two of each type: familiar-canonical, familiar-reversed, novel-causal, and novel-arbitrary. A pool of four familiar sequences was developed, such that, with the exception of a teddy bear used as the patient in every sequence, they each involved three distinctive props and actions. For each subject, two canonical and two reverse-order sequences were drawn from the pool of four familiar events. Each subject thus saw each sequence in only one order of presentation. Each familiar sequence was presented equally often in its canonical and reverse order. This allowed us to test several different canonical and reverse-order sequences without duplicating props or action. The novel-causal and novel-arbitrary sequences were drawn from a pool of three events of each type. For each subject, two of the three events were presented. Thus, each of the novel

294

BAUER

AND

THAL

sequences was used for 18 of the 24 subjects. This allowed us to test a wider variety of novel events, without requiring that any one subject be tested on more than four of them. A complete description of each sequence is provided in the Appendix. For each sequence, the subjects first were allowed to manipulate the props. The sequence then was modeled, with narration, two times in succession. Immediately after modeling, E returned the props to the subjects and encouraged exact imitation with statements such as “Can you give the bear a bath, just like I did‘?” The sequencing task was one of two tasks administered during the session. To lessen fatigue. four of the sequences (one of each type) were presented, then an unrelated task, and then the remaining four sequences were presented. The sequences were administered in counterbalanced order, with the exception that testing never began with a reverse-order event. With this constraint, each sequence was presented equally often in each serial position. The sessions were videotaped for later analysis. Scoring

Two independent raters trained together until they reached reliability of 90% or above on three successive subjects from an existing corpus of data. Each rater then coded approximately 50% of the sample. They watched the videotapes of the sessions and recorded all occurrences of target behaviors. Rater 1 then recoded four of the subjects originally coded by Rater 2. Overall agreement on both occurrence and order of target behaviors was 97% (range: 93 to 100%). For each sequence, the total number of different target actions produced and the number of different pairs of target actions produced in the modeled order were calculated. For the latter measure, only the first occurrence of each target action was included. For example, on the “giving teddy breakfast” sequence, if a subject produced all three components in the modeled order, she would receive credit for three different target actions, and for two different pairs of actions in the modeled order. She would receive one point for the pair pour in the milk (1) and stir in the bowl (2), and one point for the pair stir in the bowl (2) and feed teddy (3). If she produced feed teddy (3), stir in the bowl (2), pour in the milk (I), stir in the bowl (2), she would still receive credit for production of three different target actions. However, she would not be credited with any pairs of actions in the modeled order, since the first production of each of the behaviors was not in the correct order. This scoring procedure reduces the likelihood of a subject receiving credit for production of a sequence by chance or by trial and error. It should be noted that the number of different actions produced affects production of pairs of actions in the modeled order. Thus, the dependent measures are not independent of one another.

DEVELOPMENT

OF SEQUENTIAL TABLE

NUMBER

295

UNDERSTANDING

I

OF DIFFERENT TARGET ACTIONS, NUMBER OF PAIRS OF ACTIONS IN ORDER, AND NUMBER OF PAIRS OF ACTIONS IN BACKWARD ORDER

Familiar

Different actions Mean (SD) Pairs in forward order Mean (SD) Pairs in backward order Mean (SD)

FORWARD

Novel

Canonical

Reverse

2.50 (.61)h

2.31 (.59)h

Causal

Arbitrary

(Max. possible = 3) I .29 c.62)’ .25 (.29)

2.88 (.22)"

(Max. possible = 2) .73 (S3) I.81 (.29)” (Max, possible = 2) .69 (.46) .06 (.25)

2.60 (.53p

I .oo (.5l)h.’ .42 (.32)

Note. Means with the same alpha superscript do not differ significantly from one another.

Additionally, for each sequence type, we also calculated a backward sequencing score which represents production of the target actions in reverse of the modeled order. As discussed by O’Connell and Gerard (1985), the backward sequencing score actually is relevant only for the reverse-order sequences. For those sequences it represents “corrections” to the modeled order. Nevertheless, we calculated the backward sequencing score for all four sequence types, to control for the possibility that children might tend to reverse the order of the sequence, regardless of the sequence type (see O’Connell & Gerard for related dscussion). RESULTS

Descriptive statistics for the mean number of different target actions, the mean number of pairs of actions produced in the forward order, and the mean number of pairs of target actions produced in the backward order for all four sequence types are presented in Table 1. For the first two dependent variables, 2 (Gender) by 4 (Sequence Type: canonical, reverse-order, causal, arbitrary) mixed analyses of variance were conducted (Sequence Type is a within-subjects factor). Forward and backward sequencing scores were compared in a separate analysis. Where appropriate, Tukey tests of significant difference (p < .05) were used to further examine the effects. There were no reliable gender effects. The analyses are discussed in turn below. Number of different target actions produced. As is evident from Table 1, the subjects produced most of the components of the sequences. A significant main effect, F(3, 66) = 6.07, p < .OOl, and subsequent Tukey tests revealed that they produced the greatest number of different target actions on causal sequences. They produced an approximately equal number of different target actions in the reverse-order and canonical sequence conditions. The number of actions produced in the two “fa-

296

BAUER

AND

THAL

miliar” sequence conditions was significantly lower than that in the novelcausal condition. Subjects’ production of different target actions on the novel-arbitrary sequences was equivalent to that on all other sequence types. Number of pairs of actions produced in the modeled order. This measure indicates the degree to which subjects adhered to the modeled order. A significant main effect, F(3, 66) = 22.27, p < .OOl, and subsequent Tukey tests revealed that subjects produced a greater number of pairs of target actions in the causal condition, relative to all other conditions. They produced a greater number of pairs of actions in the modeled order in the canonical relative to the reverse-order condition. Adherence to the modeled order in the arbitrary condition was intermediate between that in the reverse and canonical conditions, and did not differ significantly from either of them. Forward and backward sequencing scores compared. We also compared subjects’ forward and backward sequencing scores on each of the sequence types in an analysis of variance. We did not include the causal condition in the analysis because only three subjects produced any pairs of actions in reverse of the modeled order; the group mean backward sequencing score on the causal sequences was .06. For the other three sequence types, we conducted a 3 (Sequence Type: canonical, reverseorder, arbitrary) by 2 (Order: forward, backward) within-subjects analysis of variance. The main effect for Sequence Type was not statistically significant: F(2, 46) = .61, p = S47. Subjects did produce a significantly greater number of pairs of target actions in the forward order (M = 1.Ol) than in the backward order (M = .45): F(1, 23) = 30.46, p < .OOl. However, a significant Sequence Type by Order interaction indicated that the pattern of difference between forward and backward sequencing scores differed across sequence types: F(2, 46) = 10.65, p < .OOl. Separate analyses by sequence type revealed that, in the canonical condition, subjects’ forward sequencing scores (M = 1.29) were significantly greater than their backward sequencing scores (M = .25): F(1, 23) = 38.75, p < .OOl . In the reverse-order condition, forward sequencing scores (M = .73) were not significantly greater than their backward sequencing scores (M = .69): F(1, 23) = .06, p = .814. In other words. in the canonical condition, children were unlikely to “make a mistake,” and produce the sequences in reverse order. However, in the reverseorder condition, children vacillated between reproducing the sequences in the modeled (reverse) order and “correcting” them to canonical order. On 34% of the reverse-order trials, subjects produced all or part of the sequence in the modeled order. On another 34% of these trials, they produced a complete or partial reversal or “correction” to canonical order. In the arbitrary condition, subjects were significantly more likely to

DEVELOPMENT

OF SEQUENTIAL

UNDERSTANDING

297

produce the sequences in forward order (M = 1.00) than in backward order (M = .42): F(1, 23) = 18.18, p < .OOl. Thus, although the sequences in this condition were novel and arbitrarily ordered, subjects demonstrated reliable temporal ordering of them. As mentioned above, in the causal condition, subjects rarely produced any pairs of actions in other than the modeled order. Their forward sequencing scores (M = 1.81) were far greater than their backward sequencing scores (A4 = .06). Comparison of sequencing performance against chance. Comparison of forward and backward sequencing scores provides one measure of the degree to which children adhere to the modeled order of the sequences. However, it does not indicate whether the amount of sequencing was significantly different from that expected by chance. The scoring procedure used reduces the likelihood that credit would be given for production of a pair of correctly ordered components by chance or by trial and error (only the first occurrence of each behavior is considered). Nevertheless, a formal test of this possibility is desirable. A single “chance sequencing score” could not be created because different chance scores would be expected for subjects who produced only two components than for those who produced all three components (production of pairs of components in the target order is dependent upon production of individual components). As a result, the number of possible combinations of two or three actions which would result in a correctly ordered pair of components was calculated and used to determine chance performance. One-half of all possible combinations of two components and one-half of all possible combinations of three components result in one correctly ordered pair of components. Thus, by chance alone. onehalf of all productions should include one correctly ordered pair of target actions. For each sequence type, a x’ statistic was calculated to compare the number of subjects who produced an average of at least one correctly ordered pair to the number who would be expected to do so by chance alone. For the canonical, causal, and arbitrary sequence types, a larger number of subjects than would be expected by chance met this criterion: canonical, x2 (1, N = 24) = 10.66, p < .005; causal, x2 (1, N = 24) = 24.00, p < .OOl; arbitrary, xz (1, N = 24) = 6.00, p < .05. For the reverse-order sequences, the number of subjects producing at least one correctly ordered pair of components did not exceed that expected by chance: x2 (1, N = 24) = 1.50, ns. DISCUSSION The results of this study provide clear evidence that children under 24 months of age include temporal order information in their representations of both familiar and novel events. This is most apparent in comparison of forward and backward sequencing scores for each condition. Virtually all subjects reproduced novel-causal sequences in the forward

298

BAUER

AND

THAL

order (only 3 out of 24 produced any reversals). Reproduction of the modeled order also was significantly higher than reversal of that order in the familiar-canonical and novel-arbitrary conditions. Only in the reverse-order condition was there no significant difference. In that condition, subjects produced the events in recognizable order, but not necessarily in modeled order. Instead, they were equally likely to produce the sequences in canonical as in modeled order. In all, the behavior of these 21-month-olds was very similar to that of O’Connell and Gerard’s (1985) 36-month-olds. The pattern of performance that they associated with a particular developmental event (i.e., the beginning of reversibility) was evident in our significantly younger subjects. We believe that the major source of difference between our findings (as well as those of Bauer & Mandler, 1989; Bauer & Shore, 1987; and Ungerer, 1985) and those of O’Connell and Gerard (1985) is methodological. The strong effects of methodology on children in this age period are clearly illustrated in Brownell (1988). In her study of combinatorial ability in 18- to 30-month-olds across six behavioral domains, she found a number of significant effects which were the result of the demands of the task. In the present study we eliminated several sources of task difficulty which we believe to have interfered with successful performance in O’Connell and Gerard but which have nothing to do with the ability to use temporal information in recall. For example, to reduce fatigue, we tested subjects on only eight sequences as compared to O’Connell and Gerard’s 18 sequences. Additionally, presentation of the sequences was interspersed with another unrelated task. With the exception of the teddy bear which was used in the familiar events, there was no duplication of either props or actions across sequences. Finally, for each child, each sequence was performed in only one order. These steps remove sources of action, prop, and order interference that could lower sequencing performance, despite the ability to use order information in recall. What implications do the present data have for the developmental sequence in the use of temporal organization described by O’Connell and Gerard? They argued that the first ordered reproductions of events, evident at 24 to 28 months, are dependent upon familiarity with the event. They also contended that the ability to reproduce unfamiliar and/or arbitrary sequences, and the ability to produce familiar events in other than canonical order are two components of sequencing ability that do not emerge until after the first ordered reproductions of familiarcanonical sequences. We have already noted that in the present study, children correctly reproduced novel-causal and novel-arbitrary sequences. Thus, at least by 21 months, familiarity with a sequence is not a n~cessar?/ condition for correct reproduction of it. This finding is consistent with Mandler’s

DEVELOPMENT

OF

SEQUENTIAL

UNDERSTANDING

299

(1986) argument that the ability to recall an event sequence is influenced not by familiarity, per se, but by how well the representation of the event is organized. Organization is in turn influenced by familiarity and by the type of relations among items in the sequence. Repeated experience of an event contributes to the development of a stable core around which its representation is organized. As a result, script-like events, such as those tested in the present experiment, are well-organized. The type of relations among the components also is important, perhaps especially when a sequence is experienced for the first time. Causal relations imply temporal invariance, even when the event is experienced only once. Thus, causal relations provide an invariant core around which a new representation can be organized (Bauer & Mandler, 1989). It is interesting that in the present study, as well as in Bauer and Mandler (1989), children’s ordered reproduction of novel-causal sequences was superior to that of familiar-canonical sequences. This effect might be associated with the higher level of production of the components of the novel-causal relative to the familiar-canonical events (reproduction of order is dependent upon production of components). Additionally, it is possible that performance on the familiar-canonical sequences actually was depressed by inclusion of the teddy bear. It often has been noted that self-directed acts (e.g., drinking from a toy cup) are an earlier development in symbolic play than are other-directed acts (e.g., giving a stuffed animal a drink) (see McCune-Nicolich, 1981, for a review). However. in pilot testing subjects were more likely to perform the target behaviors when they were modeled with the teddy bear than when they were not. It is unlikely then that superior performance on novel-causal relative to familiar-canonical sequences can be attributed to difficulties associated with using the teddy bear. Additional investigation of the relative strength of the organization of causal as compared to familiar events is indicated. Representations of novel-arbitrary event sequences would not be expected to be as well-organized as those of novel-causal and familiar ones. Even adults recall unfamiliar and arbitrarily ordered events less accurately than familiar and/or causally connected ones (see Mandler, 1986). Nevertheless, our 21-month-olds reproduced novel-arbitrary sequences relatively accurately. This effect is a replication of Bauer and Mandler (1989), and stands in contrast to Bauer and Shore (1987) and O’Connell and Gerard (1985). Bauer and Shore tested an arbitrary event that resulted in very low production of the individual components, thus presenting few opportunities for ordered reproduction of the sequence. O’Connell and Gerard tested arbitrary events which were scrambled sequences of meaningful actions. As discussed in the introduction, since the sequences were composed of elements of several different familiar events, (a) they were not truly novel. and therefore, are not an appropriate test of re-

300

BAUER

AND

THAL

sponse to unfamiliar sequences, and (b) they may have produced interference effects which lowered overall performance.* It is clear that when tested with novel sequences about which they have no expectations, children as young as 21 months (and as young as 16 months in Bauer & Mandler, 1989) can reproduce unfamiliar event sequences, even when they are arbitrarily ordered. The finding of 21-month-olds’ ordered reproduction of novel-causal and novel-arbitrary sequences leaves open the question of whether correct sequencing of familiar-canonical events precedes that of unfamiliar events. It is possible that at some point early in development, children would correctly sequence familiar events but not novel ones. Nevertheless, the findings from the present study, as well as those from Bauer and Mandler (1989) and Bauer and Shore (1987) make it clear that by the middle of the second year of life, children can correctly reproduce novel event sequences, especially if they contain causal connections. Recent findings of Bauer (1989a) suggest that this ability extends to 12to I4-month-olds as well. Finally, O’Connell and Gerard argued that the ability to reproduce a familiar sequence in other than canonical order develops after the ability to reproduce familiar, canonically ordered sequences. They found that only 36-month-olds showed any tendency to reproduce reverse-order events in the modeled order. They interpreted their oldest subjects’ vacillation between producing the events as modeled and “correcting” them to canonical order as the beginning of reversibility. However, the data from our 21-month-olds on reverse-order sequence production look very similar to those from O’Connell and Gerard’s 36-month-olds. They too vacillated between correct ordered production and correction to canonical order. The ISmonth difference in the appearance of this pattern is a strong argument against an age-specific onset of the ability to reproduce reverse-order event sequences. At this point we can only speculate about what makes reordering and reorganization difficult for young children. One possible explanation of

’ It is also the case that scrambled sequences do not have a clear goal or end-state, i.e.. nothing is accomplished in the sequence wipe a bear’s mouth. the bear pays money, cover the beur rcsith a blanket. The goal of an event is presumed to provide a source of organization for its representation, at least for adults (Schank & Abelson, 1977). Thus, the absence of a clear goal or end-state might have contributed to lower performance on the scrambled sequences in O’Connell and Gerard (1985). However, Bauer (unpublished data) has found equivalent levels of performance on novel-arbitrary sequences with a clear end-state or final product (e.g., three components combined to “make a bell”) and novelarbitrary sequences without a clear goal or end-state (e.g., “marching in a parade”: ride rc horse, shake a pompom. play c1 drum). This suggests that the absence of a clear goal state is not the crucial factor in accounting for the low level of performance on scrambled sequences.

DEVELOPMENT

OF SEQUENTIAL

UNDERSTANDING

301

the difficulty in O’Connell and Gerard (1985) can be eliminated on the basis of the present experiment. We suggested that presentation of conflicting order information might have made the task confusing to their subjects. In the present study, the subjects saw only one order of presentation of each sequence, yet they still did not consistently reproduce the reverse-order sequences as modeled. It is apparent then that a procedural explanation for the difficulty young children have with this task is not sufficient. A second possibility is that the ability to reorganize an existing temporally ordered representation may require recognition that the event in question could occur in a different order in the world. In other words, it may require that the child understand that, in the absence of causal constraints, it is cultural convention that dictates the canonical order of an event. In fact, Bates, O’Connell, and Shore (1987) have proposed that recognition of the conventional basis for order underlies increases in nonlinguistic sequencing ability and lingustic grammaticization observed between 20 and 28 months (Shore et al., 1984). Since many of the routine events in which young children participate usually occur in an invariant order, it would not be surprising to find a delay in recognition of the arbitrary nature of that order. The extent of children’s sensitivity to temporal invariance is illustrated by a recent study by Bauer (1989b). Twenty-one-month-olds were significantly better at reproducing familiar temporally invariant event sequences than familiar yet temporally variant ones; performance on the latter sequence type was equivalent to performance on novel-arbitrary sequences (of the type used in the present experiment). That consistencies in the temporal order of an event facilitate ordered recall suggests that children are highly sensitive to temporal organization. Their high degree of sensitivity may in turn contribute to a reluctance to produce an event in other than its usual order. Although an incomplete understanding of the nature of the structure underlying familiar events may contribute to young children’s difficulty in reproducing reverse-order sequences, at best it can provide only a partial explanation of the phenomenon. As noted, older children (Brown & Murphy, 1975; Fivush & Mandler, 1985) and adults (Mandler, 1986), who presumably have a better understanding of the nature of the temporal connections underlying events, nevertheless find reverse-order sequence production difficult. Thus, while children become more proficient at reproducing reverse-order sequences with age, it is apparent that reorganizing existing temporally ordered representations remains a difficult task. Regardless of age, the task is particularly difficult when the event is unfamiliar (cf. Fivush & Mandler, 1985; Mandler, 1986). Increased familiarity with an event, which in turn implies increased organization of its representation, promotes flexibility in recall, for older as well as younger subjects.

BAUERANDTHAL

302

APPENDIX

I.

Familiar

Event Sequences-Canonical

Order

(Each subject received two of the following four sequences; for each sequence. a teddy bear and the following proportionally sized props were given to the subject [the props given to the subject are indicated in parentheses].) I. Giving teddy a bath (small bath tub, sponge, towel). E modeled: the

putting

the

bear

towel. 2. Putting teddy to bed sick (crib, small blanket, toy thermometer). E modeled:

the

tub,

washing

bear

in the

the

&cur

crib.

with

covering

the

the

sponge,

bear

drying

with

the

the

bear

blanket,

with

taking

the

3. Giving teddy breakfast (small plastic pitcher, bowl, spoon). “milk’

‘from

the pitcher

to the bowl,

stirring

in the

bowl.

in

the

feeding

putting

bear’s

temperature. E modeled: pouring the the bear from the spoon.

4. Brushing teddy’s teeth (toothbrush, empty tube of toothpaste, bathroom cup). E modeled: putting the toothpaste on the toothbrush, brushing the bear’s teeth, giring the bear

a drink,from

II.

Familiar

the

cup.

Event Sequences-Reverse

Order

(Each subject received two of the above familiar event sequences in reverse order; the same props as those listed above were used.) I. Giving teddy a bath-E with

the

sponge,

putting

modeled:

the

bear

2. Putting teddy to bed sick-E bear

with

the

blanket,

putting

pouring

the

“milk”

,from

4. Brushing teddy’s teeth-E the

beur’s

III.

teeth,

putting

the

Novel-Causal

the

bear

with

the

towel,

washing

the

bear

tub.

modeled:

the

3. Giving teddy breakfast-E bowl,

drying

in the

taking

the

bear’s

temperature,

in the crib. modeled: feeding the bear the pitcher to the bowl.

from

modeled:

a drink

covering

the

beur

toothpaste

on

giving the

the bear toothbrush.

the

spoon,

from

the

stirring cup,

in the brushing

Sequences

(Each subject received two of the following three sequences.) I. Making a rattle (two graduated nesting or stacking cups, small rubber ball). E modeled: putting

the

ball

in

the

larger

cup,

inrzerting

the

smaller

cup

into

the

larger,

shaking

the

cups.

2. Making a frog jump (small wooden board, wedge-shaped block, toy frog). E modeled: the bourd ON the wedge-shaped block (to form a teeter-totter). pluring the frog on the end of the board, hitting the board (thereby causing the frog to “jump”). 3. Making a spinner (cone-shaped base with a small doll perched on the top, small helicopter propeller). E modeled: taking the doll off the top of the &use (thereby revealing a stick protruding from the top of the cone). putting the propeller on the stick, hitting the

putting

propeller

IV.

to make

it spin.

Novel-Arbitrary

Sequences

(Each subject received two of the following three sequences.) I. Going for a train ride (two train cars that could be attached with Velcro, small doll to fit in the cars, piece of train track). E modeled: putting the driver in one of the cars, uttaching

the

cars

to one

onother.

putting

the

cars

on

the

truck.

2. Making a picture (small chalkboard, easel, piece of chalk, sticker). E modeled: the sticker on the with the chalk.

chalkboard,

leaning

the

board

against

the easel.

scribbling

on the

putting board

DEVELOPMENT

OF SEQUENTIAL

303

UNDERSTANDING

3. Making a bell (small metal bell on a stand, plastic mallet, magnetic sticker). E modeled: turning the handle on the top of the bell, striking attaching the magnetic sticker to the bell.

the bell Njith the mallet

to make

it ring,

REFERENCES Bates, E., O’Connell, B., & Shore, C. M. (1987). Language and communication in infancy. In J. Osofsky (Ed.), Handbook of infant development (pp. 149-203). New York: Wiley 82 Sons. Bauer. P. J. (1989a. April). Putting the horse before the cart: The use of temporal order in recah of events by /2- to 14month-olds. Paper presented at the biennial meeting of the Society for Research in Child Development, Kansas City, MO. Bauer, P. J. (1989b. April). Ho/ding it all together: Effects of causal and temporal invariance on young children’s event recall. Paper presented at the biennial meeting of the Society for Research in Child Development, Kansas City, MO. Bauer, P. J.. & Mandler, J. M. (1989). One thing follows another: Effects of temporal structure on I- to 2-year-olds’ recall of events. Developmental Psychology, 25, 197206. Bauer, P. J.. & Mandler, J. M. (in press). Remembering what happened next: Very young children’s recall of event sequences. In R. Fivush & J. A. Hudson (Eds.), Knowing and remembering in young children-Emory Symposia in Cognition. New York: Cambridge University Press. Bauer, P. J.. & Shore, C. M. (1987). Making a memorable event: Effects of familiarity and organization on young children’s recall of action sequences. Cognitir,e Development,

2, 327-338.

Brown, A. L., & Murphy, M. (1975). Reconstruction of arbitrary versus meaningful sequences by preschool children. Journal of Experimental Child Psychology, 20, 307326. Brownell, C. A. (1988). Combinatorial skills: Converging developments over the second year. Child Development, 59, 675-685. Fenson, L., & Ramsay, D. S. (1980). Decentration and integration of the child’s play in the second year. Child Development, 51, 171-178. Fivush. R., & Mandler, J. M. (1985). Developmental changes in the understanding of temporal sequence. Child Development, 56, 1437-1446. Gerard, A. B. (1984). Imitation and sequencing in ear/y childhood. Unpublished doctoral dissertation, University of California, San Diego. Hudson. J. A. (1986). Memories are made of this: General event knowledge and development of autobiographic memory. In K. Nelson (Ed.), Event know/edge: Structure and function in development (pp. 97-l 18). Hillsdale, NJ: Erlbaum. Hudson, J. A., & Nelson, K. (1983). Effects of script structure on children’s story recall. Developmental

Psychology,

19, 625-635.

Mandler, J. M. (1986). The development of event memory. In F. Klix & H. Hagendorf (Eds.), Human memor?, and cognitive capabilities-Mechanisms and performance (pp. 459-467).North-Holland: Elsevier Science Publishers B. V. McCune-Nicolich. L. (1981). Toward symbolic functioning: Structure of early pretend games and potential parallels with language. Child Development, 52, 785-797. Mills, D. L., Mandler, J. M., Schreibman, L., & Oke. J. N. (1988). Order in the script: Temporal organization and event knowledge in I-year-olds. Unpublished manuscript. Nelson, K. (1986). Event knowledge and cognitive development. In K. Nelson (Ed.), Event knowledge: Structure andfunction in development (pp. 1-19). Hillsdale, NJ: Erlbaum. Nelson, K., & Gruendel, J. (1986). Children’s scripts. In K. Nelson (Ed.). Eventknowledge: Structure and function in development (pp. 21-46). Hillsdale, NJ: Erlbaum.

BAUER AND THAL

304

Nelson, K., & Gruendel, J. (1981). Generalized event representations: Basic building blocks of cognitive development. In M. E. Lamb & A. L. Brown (Eds.), Advances in dev~/opmenral psychology (Vol. I, pp. 131-158). Hillsdale, NJ: Erlbaum. O’Connell, B. G., & Gerard, A. B. (1985). Scripts and scraps: The development of sequential understanding. Child Development, 56, 671-681. Piaget, J. (1969). The child’s conception of time. New York: Ballantine Books, Piaget. J. (1926). The lungugr and thought of the child. New York: Harcourt Brace. &hank. R. C., & Abelson, R. P. (1977). Scripts, plans, goals and understanding. Hillsdale, NJ: Erlbaum. Shore, C. M., O’Connell, B. G., & Bates, E. (1984). First sentences in language and symbolic play. Developmental Psychology, 20, 872-880. Slackman. E. A.. Hudson, J. A., & Fivush, R. (1986). Actions. actors, links, and goals: The structure of children’s event representations. In K. Nelson (Ed.), Event knowledge: Structure and function in development (pp 47-69). Hillsdale. NJ: Erlbaum. Smith, B. S., Ratner, H. H., & Hobart, C. J. (1987). The role of cueing and organization in children’s memory for events. Journal qf Experimenfal Child Psychology, 44, I24. Ungerer. J. (1985, April). The development of script knowledge in children from 18 to 30 months. Paper presented at the biennial meeting of the Society for Research in Child Development, Toronto. Canada. RECEIVED:

September 8. 1989:

REVISED:

December 15, 1989; March 19, 1990.