Each of these four themes is clearly relevant to an understanding of young children’s abilities to provide accurate testimony in a forensic context. In- formation may not be revealed in an interview because it was not entered into memory in the first place (Theme 1), or because the resulting memory representation was too weak or fragmented (Theme 2), or because the status of the representation changed over the course of an extended delay interval (Theme 3), or because of difficulties with retrieval (Theme 4). Of these themes, however, perhaps the most fundamental are Themes 1 and 3. Indeed, a thorough consideration of young children’s testimony must rely on an understanding of the status of the information in memory that is potentially available for retrieval.
Accordingly, we now focus in greater depth on the encoding and estab- lishment of representations in memory, as well as on their fate over long delay intervals. In the following sections, we emphasize the dramatic im- pact of knowledge on encoding and the establishment of a representation,
as well as its continuing influence on the representation over time, as rele- vant new knowledge is added into memory. We recognize that some knowledge is endogenous in that it is brought to the situation by the child. Indeed, what is already known enables a child to make inferences about an ongoing experience that facilitate comprehension and the establish- ment of a coherent representation in memory. Other knowledge can be characterized as exogenous in that it is either provided for the child by adults (or older children) or jointly constructed in the context of interac- tion. These external forces that guide children’s understanding and en- coding are often critical in shaping comprehension, particularly when an event is novel or ambiguous. Importantly, we also recognize that knowl- edge can be a double-edged sword in that it can both facilitate and distort memory.
This perspective leads us to argue that the processes involved in encod- ing an experience must be viewed as being extended in time, often beyond the duration of the event itself. Events that are personally meaningful do not have a finite ending point, as children (and adults, for that matter) often ruminate for an extended period of time on their meaning and conse- quences. Moreover, depending on the nature of an experience, adult– child dialogues often continue for hours, days, and even years. As we see it, this extended within-the-child rumination and adult–child conversation must have serious implications for the status of the memory representation. Initial Encoding
In this section, we consider knowledge-driven processes that operate as an event is experienced to determine initial encoding and the establish- ment of a representation in memory. We first illustrate the impact of prior knowledge that the child brings to the situation, making the assumption that this knowledge affects comprehension of the experience, influences the deployment of attention as the event unfolds, and contributes to the initial encoding in memory. We then consider cases in which the child does not have much a priori knowledge of the events being experienced but nonetheless acquires some understanding on the basis of information provided by adults. Finally, we explore the related situation in which knowledge about the event being experienced is derived from ongoing in- teractions with an adult.
Prior Knowledge. Abundant evidence attests to the impact of knowl- edge on encoding and subsequent memory (Bjorklund, 1985; Chi & Ceci, 1987; Ornstein & Naus, 1985). As already indicated, prior knowledge en- ables individuals to make sense out of the events that they experience, and
this comprehension is essential for effective registration in memory. For example, studies that examine the development of expertise in specific do- mains (e.g., chess, soccer) have indicated repeatedly that the highly organ- ized and accessible knowledge of experts enables them to encode and re- member domain-related information more effectively than novices (e.g., Chi, 1978; Schneider, Körkel, & Weinert, 1989). In a similar manner, chil- dren’s generalized event representations or “scripts” (Nelson, 1986) for frequently occurring events (e.g., bedtime rituals, trips to a restaurant) af- fect their understanding and later memory of specific instances of these events (e.g., Farrar & Goodman, 1990). More generally, a considerable body of evidence suggests that children’s initial expectations about events, generated in part by their scripts, can affect perception and interpretation and consequently influence what gets into memory (Nelson, 1986).
One illustration of the impact of prior knowledge can be seen in our program of work on children’s memory for the details of specific pediatric checkups (see, e.g., Baker-Ward et al., 1993; Ornstein, 1995). In our studies, we routinely observe age-related changes between 3 and 7 years of age in children’s initial recall and retention over time. Recognizing that varia- tions in children’s knowledge about aspects of the physical examination could have affected their recall of the experience, Clubb et al. (1993) car- ried out a study to gather normative data about what children know about routine visits to the doctor. Clubb et al. used an interview protocol that fo- cused on general knowledge and interviewed 5-year-olds about their un- derstanding of what goes on during a regular checkup. They then pooled the children’s responses into open-ended questions about the physical ex- amination and formed “knowledge” scores for each feature of the office visit (e.g., heart check, urine specimen). These knowledge scores were based on the proportion of children in the normative sample who nomi- nated each component of the checkup in response to the interviewer’s general probes.
Armed with these knowledge scores, which can be viewed as reflecting variability in the 5-year-olds’ understanding of the specific examination features, Clubb et al. (1993) then reanalyzed the recall protocols of the 5- year-olds who had been studied by Baker-Ward et al. (1993). In contrast to Baker-Ward et al., who reported their data in terms of the mean levels of performance for the 3-, 5-, and 7-year-olds in their sample, Clubb et al. changed the unit of analysis from the individual child to the individual ex- amination feature. In this way, they were able to compute “memory scores” for the 5-year-olds for each component of the office visit at each recall as- sessment. Paralleling the knowledge scores, these new memory scores rep- resented the proportion of 5-year-olds in the Baker-Ward et al. study who recalled each component of the checkup in response to open-ended probes.
Figure 2.1 illustrates the considerable variability in both knowledge and immediate recall for specific features of the checkup. For example, when asked open-ended questions about what happens during a check- up, receiving a prize and getting a shot were features most nominated by the sample, and checking the wrist and walking forward were features nominated least often. Similarly, receiving a prize and having a blood test (with finger sticks) were well remembered, whereas checks of the feet and elbows were remembered so poorly that one could question whether these features had been encoded and represented in memory. Most inter- estingly, the data displayed in Fig. 2.1 also indicate a linear relation (r = .68, p< .01) between the knowledge and recall scores, with increases in knowledge being associated with corresponding increases in memory performance. These data and analyses reported by Clubb et al. (1993) and Ornstein et al. (1997b) suggest strongly that what a child knows about a medical exam can seriously affect the extent to which component features of the checkup are coded and placed in memory.1
It is also important to note that the effects of prior knowledge on encod- ing are not restricted to language-based assessments of memory. Indeed, the use of elicited imitation paradigms indicates that by the middle of the second year of life infants’ nonverbal recall of action sequences is already guided by their knowledge of the temporal and causal structure of events in the world (Bauer, 1995). In elicited imitation, a researcher models an ac- tion or sequence of actions that are associated with a specific object or ob- jects. After a delay interval, the props are given to the child, and the spon- taneous production of the previously modeled actions is understood to be indicative of recall of the event. Using this procedure, Meltzoff (1988a, 1998b) has shown that single, novel, object-specified actions can be repro- duced by 9-month-olds after a delay of 24 hours, and that 14-month-olds can remember the actions for up to 1 week. In addition, Bauer and her col- leagues (see Bauer, 1995, for a review) have demonstrated that 13-month- olds are capable of ordered recall of multiple-action sequences over delays of 1 week, and that older toddlers can retain temporal order information for up to a year (see also McDonough & Mandler, 1994; Meltzoff, 1995).
1 1We recognize that there is no direct readout of the status of an underlying representa-
tion, but we suggest that memory performance immediately after an event has been experi- enced can be taken as a proxy measure of what has been encoded (Baker-Ward et al., 1997). Moreover, our intent in presenting the Clubb et al. (1993) data is to suggest that prior knowl- edge can have a strong impact on encoding and therefore on the nature of the representation that is constructed. We also recognize, of course, that knowledge is important at all stages of information processing. As such, we discuss its subsequent influence on the status of the rep- resentation in memory storage and on processes associated with retrieval and reporting. However, even though knowledge has a continuing effect on memory, extending beyond the interpretation and encoding of an ongoing event, it may be particularly significant in the construction of the initial representation.
Against this background of impressive nonverbal memory perform- ance, it is particularly interesting to observe that children as young as 13 months old can make use of their knowledge of the world to better re- member structured, as opposed to unstructured action sequences. Con- sider, for example, studies in which Bauer and her collaborators (Bauer & Dow, 1994; Bauer & Hertsgaard, 1993) compared children’s abilities to re- member three different types of action sequences: familiar sequences (e.g., putting a bear in a bed, covering it with a blanket, and reading it a story), novel sequences that involved some built-in “enabling” relation (e.g., making a rattle by putting a ball into one cup, covering it with another cup, and shaking), and novel–arbitrary sequences that do not require an invariant temporal order (e.g., making a party hat by putting a balloon on the top, putting a headband around the cone, and putting a sticker on the front). The children’s superior production of the familiar and novel–en- abling action sequences, in contrast to the novel–arbitrary sequences, rep- resents an early demonstration of the impact of prior knowledge on event comprehension, encoding, and remembering.
Newly Acquired Knowledge. Long-standing prior knowledge may not be necessary for encoding and subsequent retention, as long as “new”
FIG. 2.1. Scatter plot of 5-year-olds’ knowledge and immediate memory scores. Reprinted from Cognitive Development, 8, P. A. Clubb, R. Nida, K. Merritt, and P. A. Ornstein, “Visiting the Doctor: Children’s Knowledge and Memory,” pp. 361–372. CopyrightÓ 1993, with permission of Elsevier Science.
knowledge can be provided to facilitate understanding of an event that is being experienced. Consider, for example, Bransford and Johnson’s (1972) classic demonstration of the dramatic impact of new knowledge on com- prehension and encoding. In their study, adults who were presented with a complex passage that made little sense without an accompanying pic- ture were essentially unable to understand and recall it, but control partic- ipants who were provided with the picture had no difficulty understand- ing and remembering the passage.
In our laboratory, we recently carried out a similar exploration by studying the impact of medically relevant information on children’s recall of a stressful and unfamiliar medical procedure. More specifically, Prin- cipe et al. (1996) reanalyzed data collected by Merritt et al. (1994) in their VCUG study to examine the extent to which naturally occurring variation in the information provided was associated with recall. As previously dis- cussed, the VCUG involved urinary bladder catheterization, and Principe et al. (1996) classified the radiological technologist’s talk during each child’s procedure as representing a “procedural narrative” if it provided the child with an account containing three central components of the VCUG. The three components were (a) a description of the catheter and its insertion, (b) mention of contrast fluid going through the catheter, and (c) a description of the filling of the child’s bladder with this fluid. From our perspective, children receiving the procedural narrative would be better able than others to understand what was happening to them in this diffi- cult radiological procedure.
Principe et al. had access to the videotaped records of 21 of the 3- to 7- year-olds who took part in the Merritt et al. (1994) study. Analysis of these tapes indicated that 13 of the children met the procedural narrative crite- ria. That is, these children were judged to have been given sufficient infor- mation about the VCUG procedure to be included in a Procedural Narra- tive Provided group. Admittedly, among these children, some received the narrative just prior to catheterization, whereas others were given the information as the event was in progress. Nonetheless, it seems likely that all of the children in this group were given enough information to under- stand the VCUG as it was unfolding. In contrast, the remaining eight chil- dren did not receive this information about the VCUG procedure and were placed in a Procedural Narrative Omitted group. Interestingly, the groups differed substantially in their initial and delayed recall. Immedi- ately after the VCUG, in contrast to the children in the Procedural Narra- tive Omitted group, those in the Procedural Narrative Provided group displayed higher levels of total recall (.97 vs. .81 proportion of the features recalled) and recall in response to open-ended questions (.60 vs. .22). The difference between the two conditions increased somewhat over a 6-week
delay interval in terms of both total recall (.97 vs. .67) and open-ended re- call (.51 vs. .13).2
Although the assignment of children to these two conditions was clearly done on a post hoc basis, subsequent analyses revealed that the dif- ferences in recall could not be attributed to either age or levels of stress. As such, the group differences in recall are consistent with the view that the greater knowledge conveyed by the procedural narrative enhanced the encoding of this stressful and unfamiliar radiological procedure. Obtaining Knowledge Through Conversational
Interactions
Children can also obtain knowledge about unfamiliar events through con- versations with adults as these experiences are taking place. Indeed, a consensus view holds that narration contributes in an important manner to children’s understanding of their personal experiences (e.g., Fivush & Haden, 1997; Nelson, 1996). From our perspective (see also Fivush, Pipe, Murachver, & Reese, 1997; Nelson, 1996), language-based interactions during an event have the potential to influence the encoding and subse- quent recovery of information. Although adult–child narratives about on- going experiences have not been explored as thoroughly as conversations about past events (e.g., Fivush & Haden, 1997; Reese, Haden, & Fivush, 1993), it seems likely that talk during an event serves to focus children’s attention on salient components and to increase their understanding of it. These conversations are thought to result in the establishment of a richly detailed and organized representation of the experience.
In partial support of this point of view, consider the groundbreaking work by Tessler and Nelson (1994). In one study, a small sample of moth- ers and their 3-year-old children were observed as they visited the Ameri- can Museum of Natural History in New York. When interviewed about the features of the museum trip 1 week later, only the objects that had been jointly talked about by both the mother and child during the event were recalled, suggesting the important role of adult–child talk during an event in focusing children’s attention. In a second study, 4-year-olds were recorded during a picture-taking walk with their mothers through an un- familiar neighborhood. Here again, the children did not recall aspects of their experience that had not been talked about during the event. Even more interesting, however, was Tessler and Nelson’s observation that
2 2It should be noted that overall levels of recall are lower in Principe et al.’s (1996)
reanalysis of the Merritt et al. (1994) data than in the original report because of the use of a subset of the sample and a decision to recode the data to conform to scoring procedures that are currently employed in our laboratory.
mothers who frequently connected the ongoing event to previous experi- ences had children who later recalled more of the pictures they had taken, and remembered more about the activity, in contrast to children of mothers who did not adopt this narrative “style” during the encoding of the event.
In our laboratory, we have also explored the linkages between mother– child narrative interaction during specified events and children’s subse- quent memory of these experiences. In the context of a short-term longitu- dinal study of children from 2.5 to 3.5 years of age, Haden et al. (in press) observed mothers and their children interact as they engaged in a spe- cially constructed experience at three time points across the year. Within the confines of each family’s living room, the mother–child dyads took part in a pretend camping adventure at 30 months, a birdwatching activ- ity at 36 months, and the “opening” of an ice cream store at 42 months. To illustrate the nature of these events, the camping activity began with the mothers and children loading backpacks with various food items (e.g., hotdogs, hamburgers, sodas) in preparation for their trip. They then hiked to a fishing pond where there was a fishing rod and net to use in catching the fish. After fishing, the dyads continued on to a campsite where there was a sleeping bag, along with a grill, pots, and utensils to use in cooking and eating the food. The birdwatching and ice cream store events were similarly structured, so that each activity was composed of a number of component features.
Because these interactions were videotaped, we had a precise record of how each mother–child pair nonverbally and verbally interacted with each feature as the events unfolded. Moreover, we could link mother– child talk about the events as they were ongoing to assessments of remembering that were obtained in interviews carried out by examiners after delays of 1 day and 3 weeks. These interviews were hierarchically structured and included general open-ended questions (e.g., “Tell me about the camping trip that you had with your mom.”), followed by more specific open-ended ques- tions (e.g., “What kind of food did you pack up?”, “What did you put the food in?”), and, finally, yes–no type probes (e.g., “Did you pack hotdogs?”). Given our interest in relating interaction during these activities to chil- dren’s event recall, a critical first step was to characterize mother–child behaviors directed toward specific features of each event. To do so, we de- veloped a coding system to reliably score both mother and child nonver- bal and verbal behaviors directed toward prespecified features. Nonver- bal behaviors included pointing to a feature, touching a feature (e.g., patting, tossing, dumping), manipulating a feature (e.g., manually explor- ing, showing), and functionally using a feature (e.g., putting the pan on the grill). Verbal behaviors included calling attention to a feature, request-