Child Witness: Cognitive and Social Factors Related to Memory and Testimony

Stephen J. Lepore*

ABSTRACT: This article reviews and critically evaluates research on social and cognitive factors related to children's witness abilities.  Several consistent findings emerged despite the manifold methodological problems noted.  First, children's free recall is generally accurate, but the amount of information recalled increases from preschool-age to adolescence, at which time it reaches adult levels.  Second, performance on cued recall tasks approaches adult levels by the age of 6 or 8 years.  Third, children start to distinguish real-pretend dimensions by the age of 1 to 5 years.  Fourth, suggestive questioning has little influence on children's free recall, though preschooler's recognition memory is easily corrupted by suggestive questioning.  By 7 to 10 years, children appear to be no more vulnerable to suggestions than adults.  Child witness research paradigms are summarized and recommendations are made for increasing the validity of future studies.

In the fall of 1988, Frank Tozzolina, a 42-year-old school teacher, was arrested on charges that he had fondled four students, ages 10 and 11, under their blouses, on their breasts, and on their buttocks (Wares, 1989; Wares & Grimaldi, 1989).  The girls claimed that Tozzolina touched them during class time in the presence of other students and a teacher's aide.  Supposedly he touched them while standing behind a large desk and lectern that hid his activities from view.  A fifth student claimed to have witnessed one of the alleged molestations.

The misdemeanor molestation counts of two of the children were dismissed before being tried because of insufficient evidence.  One count that was thrown out was a claim by a student that Tozzolina had touched her back.  In the other, a student changed her story, first stating Tozzolina had rubbed her breast and later claiming he touched her side.  The final two counts were dismissed when a jury pronounced Tozzolina innocent.

The main strategy of the defense was to tarnish the girls' credibility.  In addition to Tozzolina's colleagues, classmates of the girls testified that the accusers were chronic troublemakers.  The defense also produced a letter that Tozzolina intercepted from one of the girls before his arrest, stating, "I want him fired now!"  Presumably, the four students were upset with their teacher because they all flunked his science exam.

The drama of the Tozzolina case is unfortunate and highlights the myriad issues that surface when children become witnesses.  Is it possible that children younger than those in the Tozzolina case could contrive an incidence of sexual molestation or abuse?  Can children be led by others (e.g., parents or peers) to go along with a false allegation, or suggestion of abuse?  If so, are interrogators equipped to detect such fabrications, and possibly spare children (and those falsely accused) from court procedures?  Are children readily confused about events, or likely to forget more than adults?

The foregoing issues are more pressing now than ever.  The number of children being called upon by courts to give testimony has mushroomed in the past two decades (cf. Beach, 1983; Weinstein, 1978; Whitcomb, Shapiro, & Stellwagen, 1985).  So, too, has the number of empirical studies on eyewitness testimony burgeoned since the mid-1970s (cf. Wells & Loftus, 1984); an increasing proportion of studies focus on the special case of the child witness.  As a reflection of this trend, the journal Issues in Child Abuse Accusations was founded (1989), and a volume of the Journal of Social Issues (1984) was devoted to psycholegal issues and research relating to the child witness. In addition, symposia on the child witness are now a regular part of psychological conventions.

The growing need for children to testify either as witnesses or victim-witnesses demands a thorough understanding of the social and cognitive factors that influence children's competency to bear witness.  Expert knowledge is vital to know when children's testimony can be trusted and how to make it more trustworthy.  This article reviews studies of cognitive and social factors that may influence children's reports of witnessed events.  Throughout, I will highlight the differences in the reliability of children's and adults' testimony.  The article is divided into five main sections: (a) Problem of the child witness, (b) Cognitive factors influencing children's testimony, (c) Social factors influencing children's testimony, (d) Methodological approaches and problems, and (f) Conclusions and future directions.

Problem of the Child Witness

Research on child witnesses involves a complex of issues that cut across social, legal, and psychological areas of specialization.  The child witness has become a social issue because of the alarmingly high number of children who witness crimes or experience maltreatment (especially sexual abuse) (Finkelhor & Hotaling, 1984; Gelles & Cornell, 1985; Russell, 1983; Suski, 1986), and the expanding public and professional interest in child maltreatment (see Melton & Thompson, 1987; Sheridan, 1990) which has led to strengthened mandatory reporting laws.

A few statistics will illustrate the current situation.  Nearly one-fourth of assaults in America occur in or near residences where children may be present (Bureau of Justice Statistics, 1983).  The yearly incidence of child abuse in America may range from 150,000 to 200,000 (Finkelhor & Hotaling, 1984).  In a study of 930 randomly selected women, 28% of the sample reported at least one experience of intrafamilial and/or extrafamilial sexual abuse before the age of 14 years (Russell, 1983).  Even if the preceding statistics are double the true incidence and there is reason to believe they are underestimates social responsiveness to child victims requires that they be called upon in increased numbers to testify on behalf of themselves and others.

Although adult witnesses often provide dubious and unreliable testimony, criminal cases often hinge on such evidence (Buckhout, 1974).  The reliability of child witnesses, however, has been questioned for centuries on the grounds of supposed deficits in children's psychological faculties for recording, storing, and retrieving the details of a witnessed event.  There is substantial evidence that children omit more details in their spontaneous recollections of events than adults do.  Legal practitioners are more concerned, however, with the presumed errors of commission in children's testimony.  In some United States courts the testimony of young children (less than 10 to 14 years) will not stand without corroboration from another source, such as a medical doctor or an adult witness.  Furthermore, a judge may determine the competency of the preadolescent child on the basis of testimony or a pretrial examination.  Methods of determining competency are not standardized, nor have they been generated from a systematic examination of children's testimonial abilities.  Many presumptions regarding children's (in)competency to testify can be refuted.  Age, in particular, should not determine a court's opinion on competency.  Rather, the following questions should be considered when establishing competency:

a) Does the child have adequate cognitive skills for comprehending and accurately reporting the witnessed event?

b) Can the child successfully discriminate actual events from his or her internal thoughts, images, or fantasies?

c) Can the child resist suggestions from various sources to change his or her report or memory of an event?

Cognitive Factors Influencing Children's Testimony


The primary cognitive variable of interest in the child witness literature is memory.  Although there are at least a dozen commonly employed paradigms in experimental studies of memory (Perlmutter, 1984), these paradigms are not representative of situations a child witness would be asked to recall or of everyday situations that most people use memory for (cf. Baddeley & Wilkins, 1984; Neisser, 1982; Perlmutter, 1980).  As Perlmutter (1984) notes, "while most experimental memory tasks test for deliberate, short-term retention of discrete stimuli (e.g., word lists), most everyday use of memory involves nondeliberate, long-term retention of complex events" (p. 253).

The degree of similitude between experimental stimuli and events a child might witness reflects the ecological validity of the research design.  Studies in which children witness or participate in live events generally have greater ecological validity than those using films; studies that use narratives or audiotapes are among the least realistic.  This article emphasizes the modality of stimuli presentation as a criterion for reviewing studies.  However, content of stimuli used in witness studies is not ignored.  The psychological impact of an event and a child's recall ability may depend more upon the content of the event than on the medium of presentation.  Yet, in general, live events are likely to be the subject of interest in a court situation.  Therefore, the results of witness studies with live events are considered to be most relevant to understanding children's witness abilities in the real world.

Differences in stimuli and the many uncontrolled variables in ecologically valid studies make comparisons problematic.  The memory studies reviewed below vary in a number of ways, including age range of participants, interest level of stimuli, exposure time, delay between exposure and memory testing, format of test questions, frequency of questioning, and characteristics of interrogators.  In spite of these variations, some consistent findings have emerged.  Moreover, variations in methods occasionally indicate conditional and individual explanators of differences in children's performance across studies.

Two main forms of memory are examined in child witness research, free recall and cued recall or recognition.  The following discussion has been divided into these two areas of memory because they each include many studies and appear to be abilities that develop at different rates in children.

Free Recall

Since the late 1970s, an increasing number of researchers have examined children's capacity to recall complex events.  One of the most consistent findings is that young children tend to recall fewer details of events than do older children and adults (e.g., Goodman & Reed, 1986; Jones, Swift, & Johnson, 1988; King & Yuille, 1987; List, 1986; Marin, Holmes, Guth, & Kovac, 1979; Ratner, Smith, & Dion, 1986; Saywitz, 1987).  Nevertheless, young children are generally as accurate in what they do recall as adults (e.g., Goodman & Reed, 1986; King & Yuille, 1987; List, 1986; Marin et al., 1979).

In a study by Goodman and Reed (1986), 3- and & 6-year-old children and adults conversed with an unfamiliar adult male confederate and then played a game similar to Simon Says.  In a test of free recall administered four to five days after the event, the 3-year-olds recalled less about the event than did the 6-year-olds, who in turn recalled less than the adults.  However, children's recall perhaps by virtue of its brevity contained fewer errors of commission than adults'.  Marin et al. (1979) found a similar pattern using a briefer delay period in a sample of kindergarten- through college-aged students: Ten or 30 minutes after being exposed to an unexpected confrontation between two males, the number of items correctly and incorrectly recalled increased linearly with age.(1)  To the extent that adults recalled more incorrect details than young children, their recall may appear to be less accurate.  However, the proportion of incorrect to correct items in free recall revealed a similar level of accuracy across age groups.

Saywitz (198 has observed age differences in the types of recall errors that are made.  Children aged 8, 11, or 15 heard an audiotaped story about a theft, then recalled the story immediately after presentation and again 5 days later.  Consistent with other findings, the recall of the 8-year-olds was less detailed but as accurate as that of the 11- and 15-year-olds (there was no interaction with time).  However, the types of errors made by younger versus older children differed in some respects.  The youngest children were more likely to add extraneous information than were the two older groups, including exaggerations and contradictions of the facts of the original event.

The type of correctly recalled information also seems to vary by age.  Ratner and colleagues (1986) had kindergarten children and adults make clay and play with clay.  Half of the participants were immediately interviewed, and all participants were interviewed 7 to 10 days after the event.  In keeping with other studies, children recalled less than the adults.  However, the overall difference in amount of information reported by children and adults was attributed to adults' more complete reporting of "subordinate" information.  That is, adults were able to report more about the details of their clay productions.  Children were as capable as adults in reporting the superordinate, or main goal, of making clay or making shapes with the clay.  These findings corroborate those of other studies that have shown children's recall is best for the central aspects of an event or story (e.g., Mandler & Johnson, 1977; Smiley, Oakley, Worthen, Campione, & Brown, 1977).

Recent research has shown that the type of memory content to be recalled affects the extensiveness of recall in children of different ages.  For example, Jones and colleagues (1988) found developmental differences in memories of objects but not in memories of activities.  Compared with 3-year-olds, 4-year-olds recalled significantly more details about objects they saw during a field trip, although there was no age difference in the recall of activities on the trip.  The difference in object recall but not in activity recall between these young children suggests that recall ability, especially for activities, develops quite rapidly during the preschool years.

A largely ignored issue in the psycholegal literature is whether long-term delays, or retention intervals, between witnessing and recalling an event have age-dependent effects on recall.  Three studies on the effect of delay on children's recall provide mixed results.  Jones et al. (1988) found that 3-year-olds' recall about a field trip was less complete than 4-year-olds', but there was strong retention of recalled material for both age groups over an 8-week period.  Hamond and Fivush, (1990, cited in Fivush, in press) interviewed 2- and 4-year-old children about a trip to Disneyworld at either 6 or 18 months after the trip.  Consistent with Jones et al., they found that younger children's recall was less detailed than the older children's but, there was no effect of age on retention.  In contrast, Dent and Stephenson (1979) found that a two-week or a two-month delay before the first recall session had an adverse effect on the completeness of recall of 10- and 11-year-old children.

Methodological differences between the studies could explain the discrepant findings.  Most significantly, children in the Jones et al. and Hamond and Fivush studies recalled a fun event that they had participated in (visiting a farmhouse or Disneyworld), whereas children in the Dent and Stephenson study recalled a film clip.  Children visiting a farmhouse or Disneyworld may have talked with family or friends about their trips; it is not likely that the participants in Dent and Stephenson's study talked about the film.  Thus rehearsal, higher interest level, and active participation, all may have contributed to the persistence of children's free recall in the studies including trips.

An additional factor that may influence recall is stress experienced while witnessing or participating in an event.  There is a great deal of evidence that children can be traumatized by witnessing a crime (e.g., Pynoos & Eth, 1984) or being victim-witnesses (e.g., Finkelhor & Browne, 1985; Terr, 1979, 1981, 1983).  Recently, a great deal of debate has focused on the effects of stress on memory (see Goodman, in press; Peters, in press a; Peters, in press b).  There are questions, for example, whether the heightened arousal that might accompany experiencing or witnessing a crime facilitates or interferes with the encoding and memory of that event.  This is a promising area of research that certainly will be the focus of many future studies.  To date, however, the results of these studies are mixed.

Goodman, Aman, and Hirschman (1987) reported that children 3 to 7 years old recalled similar levels of detail about stressful medical procedures, such as venipuncture or inoculations.  Unfortunately, methodological problems precluded the authors from generalizing about their findings.  The venipuncture study had too few children to provide the statistical power needed to uncover potential true differences.  The inoculation study did not have a control group (i.e., no- stress treatment).  A study reported by Peters (in press a), also found no differences in the free recall of children 6 to 9 years old who were asked to describe a female confederate whom they had observed entering their laboratory room during a stressful (fire-alarm sounding) or nonstressful (radio playing) event.  In contrast, Goodman, Rudy, Bottoms, and Aman (1990, cited in Goodman, in press) as well as Ochsner and Zaragoza (1988, cited in Goodman, in press) found higher levels of recall in stressed groups than in nonstressed groups.  Until these studies are presented in greater detail and in final form (i.e., peer-reviewed journals), little can be said about the validity of the findings.

In sum, the amount of details spontaneously recalled by children increases up to the age of 11 or 12 years, and then approaches the level of adults' recall.  The type of correctly and incorrectly recalled information varies with age, although children seem to accurately recall the most salient details of an event and do not distort information more than adults.  The effects of stress at time of event on children's recall are unclear, although the trends suggest that stress has no effect on, or improves, free recall.  Overall, it appears that young children can provide useful and accurate information about events they have witnessed, particularly live-action events they have seen or participated in.  Additional research is required to determine the exact nature of age differences in retention following a delay between an event and recall.

Cued Recall and Recognition

Free recall generally represents the most accurate form of memory in children and adults (see Loftus & Davies, 1984).  Specific cues, such as those used in cued recall and recognition tasks, enable both children and adults to retrieve additional information from memory.  However, developmental patterns in recognition memory are sometimes the opposite of those patterns that are found in recall memory.  For example, List (1986) found that children's free recall was less complete but as accurate as young adults', whereas children's recognition memory was as complete but less accurate than young adults'.

Deficiencies in young children's free recall may compound any deficiencies they have in cued recall and recognition tasks.  Because children provide relatively little spontaneous information in free recall (see, for example, Fivush, in press; Pillemer & White, 1989), they are typically asked more specific questions about events than would adults.  Even if children and adults make the same number and kind of errors in their cued recall and recognition memory, children will tend to make more errors because they are routinely asked more direct questions to compensate for their inadequate free recall.

A variety of cued recall and recognition tasks have been devised to assess children's memory for events.  The typical method is direct questioning, which has two main forms: suggestive and nonsuggestive.  This discussion will focus on memory performance on nonsuggestive, cued recall and recognition tasks.  The effects of suggestive questioning on memory will be discussed in a later section.  As with the studies of free recall, variations in experimental paradigms make comparisons among cued recall and recognition studies difficult.  Furthermore, cued recall studies present questions in a variety of forms, including a yes/no or true/false format, a multiple-choice format, or a short answer format.

Several researchers have observed no effect of age on subjects' recognition performance.  For example, in a test administered to 8-to 15-year-olds immediately after they heard an audiotaped theft, Saywitz (1987) found no age differences in percent of correct responses to true/false questions.  In a test administered to individuals from kindergarten to college age following a 10- or 30-minute and two-week delay, Marin et al. (1979) found no age differences in percent of correct responses to yes/no questions about a staged confrontation between two male confederates.  Participants answered on average 74% of the questions correctly.  Ceci, Ross, and Toglia (1987) found a similar pattern with subjects as young as 3 years.  They presented 3- to 12-year-old children with a story and illustrations, followed by suggestive and nonsuggestive questioning, and a forced-choice recognition task.  The results of the control group, which received nonsuggestive questions, will be reported here.  Children were assigned to four groups according to age: 3 to 4, 5 to 6, 7 to 9, and 10 to 12 years.  The percentage of correctly identified information was uniformly high, ranging from 84% for the youngest to 95% for the oldest age groups.  There were no statistically significant differences in recognition scores between the age groups.  In a second study, Ceci et al. (1987) used a similar paradigm with preschoolers only, and the younger children (M = 3.8 years) consistently performed as well as the older ones (M = 4.6 years) (71% vs. 74% correct).

In contrast, other researchers have found age differences in children's capacity to respond to direct questioning about past experiences.  Goodman and Reed (1986) had subjects interact with an adult for five minutes and questioned them about the experience four or five days later.  They found that the memory performance of 6-year-olds and adults was significantly superior to that of 3-year-olds, although the absolute difference in amount of information recalled was not great.  This closeness in performance can be seen by comparing the proportion of correct answers given by the three groups in ascending age order: 59%, 69%, and 74%.  In real numbers, adults correctly answered an average of two to three more questions than did the 3-year-olds.  Findings from other studies suggest that the weaker performance of the 3-year-olds might be attributed to the four- to five-day delay between stimulus exposure and testing.  In the study by Jones et al. (1988), for example, decline in recognition memory over time was greater for 3-year-olds than for 4-year-o;ds.

One study has revealed memory differences among older cohorts.  Cohen and Harnick (1980) showed 9- and 12-year-olds and adults a 12 minute film of two episodes of petty crime, followed by a mix of suggestive and open-ended, nonsuggestive questions.  Performance on the nonsuggestive questions revealed age differences in memory: the 9-year-olds, with an average of 51% correct, performed significantly poorer in the memory test than either the 12-year-olds or adults, who averaged 76% and 79% correct, respectively.  One week later, participants were asked 22 four-choice questions about the events they had previously been tested on.  Age had a significant effect on memory for objective questions; again, the youngest participants' memory performance was poorer than the two older groups'.(2)

Cohen and Harnick suggest that the poorer performance of the 9-year-olds in the first session signifies that these children did not encode the relevant events in the film as well as the older children.  However, there was no baseline measure of straightforward nonsuggestive memory because all participants received both suggestive and nonsuggestive questions in the first memory test.  The apparent age differences in memory may have resulted from the suggestive questions rather than a deficit in the younger children's encoding.  That is, the mix of suggestive and nonsuggestive questions may have confused and distorted the memory of only the younger children.  Without an adequate comparison group, it is unclear whether there is a memory deficit, an encoding deficit, or merely confusion arising from suggestive questioning in the 9-year-olds in this study.  Another problem with the Cohen and Harnick study is that the authors chose a rather liberal alpha for their post-hoc comparisons between the three age groups (Scheffe, using .10 significance level).  A 10% probability of committing a Type 1 error is rather high, and raises a question about the reliability of the difference between the 9-year-olds and the older age groups.

Several studies have shown that variations in the content of questions can reveal age differences in memory. For example, Parker, Haverfield, and Baker-Thomas (1986) presented a slide sequence of a mock theft to college students (M = 24 years) and elementary-school children (M = 8 years), followed by four central questions (e.g., hair color of thief) and five peripheral questions (e.g., color of a blanket in the slides).  All questions were multiple choice.  Results indicated that adults could answer more of the central questions than the peripheral questions, but there was no similar difference for children.  In a replication study, however, Parker and Carranza (1989) found that both children (M = 9 years) and adults (M = 21 years) answered central questions better than peripheral ones.  The replication study did not include central questions related to age and weight, which the authors suggest were questions in the Parker and Carranza study that could have suppressed children's correct responses to central questions.  Unfortunately, Parker and colleagues do not report the exact proportion of correctly answered questions for children or adults in either study.  If children do have a greater awareness than adults of peripheral details of an event, this would be an advantage in cases that hinge on a detail and not the central action (e.g., whether a purse snatcher had a tattoo on his arm).

Goodman et al. (1987) also found effects of stimuli and question content on recognition memory.  In their study, 3- to 4-year-olds and 5- to 6-year-olds were questioned about an inoculation they had received at a clinic three to nine days earlier.  Regardless of age, children accurately answered significantly more questions about central information than about peripheral information.  The length of delay between inoculation and questioning had a detrimental effect on some aspects of memory in the younger children (M = 3.4 years), but not in the older children (M = 5.6 years).  Specifically, the younger children answered 85% of the central questions correctly following a three- or four-day delay, yet only answered 66% of the central questions correctly following a seven- to nine-day delay.  The older children consistently answered 79% of the central questions correctly.

Gender also may factor into children's recognition performance.  After witnessing a mock argument between a male experimenter and a male confederate, female participants ranging in age from kindergarten through college correctly answered significantly more items than their male counterparts on the same objective test about the event (77% vs. 71% correct) (Marin et al., 1979).  List (1986) found that female participants ranging in age from 10 to 70 years recalled details of a videotaped theft more accurately than their male peers (87% versus 78% correct).  In a study by Parker et al. (1986), 8- to 24-year-old males correctly answered more central questions than peripheral questions, while the same aged females were not biased toward central or peripheral details.  Unfortunately, none of the above studies was explicitly designed to test hypotheses regarding gender, therefore many uncontrolled variables could explain the gender differences.

Finally, stress has been considered to have adverse effects on memory particularly recognition of confederates in photo and live lineups.  In a series of 5 studies reported by Peters (in press a), high arousal levels induced at the time of witnessing/experiencing an event generally had a detrimental effect on children's abilities to correctly identify a confederate or a room in target-present lineups.  In some of the studies, additional differences between stressed and nonstressed groups were observed on some measures of cued recall, but not all measures.  In the first study reported by Peters, 3- to 8-year-old (M = 7 years) children either visited a male dentist (stress) or were visited by a male confederate (nonstress) at their home.  Face recognition was poorer in photographs of target-present lineups for the stressed group (43% correct) than for the nonstressed group (71% correct).  In the second study, an adult, male confederate pretended to give children aged 3 to 6 years (no mean age reported) a medical checkup (i.e., pulse reading).  Half of the subjects also got their head "vigorously" rubbed by the confederate.  Stress was measured (observational rating of anxiety) rather than manipulated.  Face recognition was poorer in photographs of target-present lineups for the high-anxious group (37% correct) than for the low-anxious group (63% correct).  No associations were reported between anxiety level and memory for head rubbing.

In the third study, 3- to 9-year-olds (M =6.4 years) either went individually to an exam room and received a shot (stress) or went in groups to a university room and listened to a talk about immunization (no stress).  The nurse, who was the same in both conditions, had all children count to 10, and placed a toy mouse beside her.  Dependent measures were children's memory of counting to 10, presence of the toy mouse, cued recall of details about the nurse (e.g., hair color), and recognition of the nurse or the room in target-present and target-absent photographs.  The only significant effect of stress was on the identification of the room in target-present photographs: 94% correct for the nonstressed group, and only 38% correct for the stressed group.  Unfortunately, the differences between the groups might not be attributable to stress, primarily because the experiment did not control for type of room (university vs. exam room), quality of manipulation (shot vs. talk), or type of administration (individual vs. group).  It is possible, for example, that children sitting in a group lecture (no stress) were more likely to look around the room, or felt less compelled to attend to what the nurse was doing or saying than the children who were individually interacting with the nurse during their shot administration (stress group).  Finally, Peters noted that younger children answered a greater percentage of details about the nurse incorrectly than older children.

In the fourth study, 5- to 10-year-old (M =7 years) children observed the theft (stress) or retrieval (nonstress) of a moneybox by a male confederate, and then identified the confederate in a live lineup (stress) or photo lineup (nonstress).  The lineup condition also manipulated target presence/absence.  The worst response (33% correct) was among the group that was stressed during both the event (theft of money) and identification (live-target present); the best performance (83% correct) was among the group that observed both the innocuous event (retrieval of money) and identification (photo-target present). This study represents an important departure from the previous one because it adds stress at the time of identification.  Identification, and other court-related features of interrogation, can be very stressful to children (e.g., DeFrancis, 1969; Gibbens & Prince, 1963).

In the final study discussed by Peters, children aged 6 to 9 years (M =7.3) observed a female confederate enter a room they were in, go to a window, and drop something on her way out while either a fire-alarm was sounding (stress) or a radio was playing (nonstress).  In a series of 10 objective questions, children in the stress group made more errors (73% correct) than children in the nonstressed group (83% correct).  In addition, stressed children made more errors in identifying the confederate in target-present lineups (36% correct) than children in the nonstressed group (71% correct).  In this study, as in 3 of the previous 4, there were no age or gender differences in performance across stress groups.  In addition, there were no differences in errors in identification in target-absent lineups across stress groups (all groups performed poorly on this task) in any of Peter's 5 studies.

In sum, cued recall and recognition questions do not provide complete, error-free reports, but they do elicit more information than free recall questions.  However, the increase in information derived from objective questioning is an increase in both correct and incorrect information.  This tradeoff between completeness and accuracy might be minimized if (a) free recall questions are administered before cued recall questions, and (b) the cued recall questions are of a broad, general nature (e.g., what color was the man's hair?) rather than a specific one (e.g., was his hair black?) (cf. Dent & Stephenson, 1979; Saywitz, 1987).

The studies described above suggest that young children's (6 to 8 years) cued memory is usually on a par with older children's and adults'.  Cued recall is particularly helpful in eliciting memories regarding actions and central rather than peripheral aspects of an event.  Even children as young as 3 years old can provide some reliable information about a witnessed event.  In the recognition tasks, the average percent of correct recognition for all studies combined was approximately 70%, with a range of 51% to 95% correct, under normal conditions.  When events are witnessed, or tested, under stressful condition, performance may be worse.  Such performance is not wholly satisfying or adequate.  It is important to note, however, that reported age differences in recognition memory were seldom large, even under stressful conditions.

Moreover, it should be noted that children may be able to recognize a great deal more about the stimuli they observed, but have been unable to demonstrate this knowledge in testing situations because of the question parameters established by the investigators.  Children (and adults) can only attend to a limited number of aspects and actions in any situation.  The likelihood of a child correctly recognizing a particular aspect of a situation that the investigator deemed worthy to remember is limited by attentional capacities and interests of the child.  What is "central" to an adult observer may be peripheral to the child, and vice versa.  In addition, what is central to an event will not always be what is important or relevant, in a trial.

Contemporary researchers who have reviewed the literature on the development of memory in children suggest that young children's memory especially recall is limited for three reasons.  First, relative to older children and adults, young children have a more constrained knowledge base for efficiently organizing information.  Second, young children lack the cognitive strategies (e.g., rehearsal or generating images) that older children use when trying to remember events.  Third, children of diverse ages may attend to different environmental features and, hence, encode different information into memory.

The following section briefly discusses each of these cognitive factors knowledge, strategies, and encoding as they may bear on children's abilities as witnesses.  Unlike the foregoing section on memory, the following discussion relies heavily on laboratory research.  This limits the applicability of the findings to child witnesses, but provides a potential theoretical explanation for some of the child witness research findings.  Furthermore, the theoretical propositions that have emerged from the more basic laboratory studies suggest avenues of future research with child witnesses.

Knowledge: Representations and Memory


Current conceptualizations of memory suggest that it is a dynamic process of construction and reconstruction of facts.  The memorial construction of events is guided by contextual cues, inferences, new information, past experiences, and world knowledge (Chi, 1981; Loftus, 1979; Neisser, 1976).  Memories therefore have long lasting and modifiable properties.

Schema theory has been posited to explain how general knowledge is organized, how this organization influences subsequent information processing, and how new information is perceived and incorporated into existing knowledge structures.  Although schemas were initially used to describe the central cognitive structure in perception (Bartlett, 1932), more recently scripts have been used to explain problems of memory (e.g., Mandler, 1983; Nelson, Fivush, Hudson, & Lucariello, 1983).

Schemas are organized sets or structures of knowledge that develop from repeated experiences with objects, people, places, and events.  Schemas and related structures, such as scripts (Abelson, 1981; Shank & Abelson, 1977), represent elements of common experience, rather than specific details.  Scripts can be thought of as event schemas, or spatially-temporally organized representations of typical events, that produce expectations about the sequence of actions in a well understood situation or context (Abelson, 1981; Mandler, 1979).  According to schema theory, once a schema is activated for a familiar event, incoming information is selectively processed to match the schematic expectation.  Information that is inconsistent with the schema, or altogether missing, can be supplanted by schematic information.  Therefore, within the schema framework, perception of events is selective and recall is general and occasionally distorted.  Interestingly, the structure of schemas appears to be age invariant.  Nelson and colleagues (1983) argue that the basic structure of scripts, or the way that children organize their knowledge about events, is essentially the same for children aged 3 to 8 years.

Scripts and Memory

Schema-driven processes can both enhance and distort memory.  For example, Lucariello and Nelson (1983) found that 3- and 4-year-olds had better recall for words from a list derived from children's scripts (e.g., going to the zoo) rather than from adults' categories.  Additionally, using a script cue in questioning (e.g., "Tell me all of the things you could see at the zoo") helped children to recall more words.  List (1986) compared accuracy and completeness of memory for schema-consistent and schema-inconsistent information in children (M =10 years), young adults (M = 20 years), and older adults (M = 72 years).  All age groups demonstrated more complete memory for schema-consistent items in recall and recognition tests.  However, List also found distorting effects of schemas on memory.  When presented with schema-consistent information, subjects apparently filled in gaps in their memory about the real event.  Thus, memory performance was less accurate for schema-consistent items than for schema-inconsistent items.  In effect, schema-consistent items led to greater errors of commission, while schema-inconsistent items seemed to account for errors of omission.

The ways in which schemas contribute to age differences in memory have not been explored in great detail.  Mandler and Ritchey (1977) argue that age differences in memory are attenuated for schema-consistent information and are exacerbated for schema-inconsistent information.  In a partial test of this hypothesis, List (1986) failed to find age-related differences in memory for schema-consistent versus schema-inconsistent information.  She suggested that the expected deficits in schema-inconsistent information failed to emerge because level of knowledge was equivalent across age groups in her study.

Several researchers have documented age-related differences in children's scripts and reports for events.  Nelson et al. (1983) found that in comparison with 3- to 4-year-old children, 5- to 8- year-old children gave significantly longer and more complete scripts, and used more complex language.  Furthermore, schematic properties that distort incoming information and recall fusion, confusion, selection, and repair were more apparent in the younger children's performances.

In a study by Rabinowitz, Valentine, and Mandler (1981), fifth graders made more false recognitions of schema-consistent foils than did adults in a lunchtime script recognition task.  There were no age differences in recognizing schema-inconsistent foils or actually presented items.  Brown (1976) similarly found that in a recognition task for a pictorial event sequence, preschoolers were more likely to make false recognitions of schema-consistent items that were not actually presented than older children.  Thus, young children appear to be more dependent upon their schematic representations in recognition tasks than older children and adults.

Saywitz (1987) applied schema theory to explain why 8-year-olds in her study were more likely than the 11 - and 15-year-olds to produce recall errors by adding extraneous information:

These findings may be explained by the difficulties in distinguishing between what actually occurred and what might have occurred, based on one's expectations or schemata.  When subjects are exposed to a stimulus, other related pieces of information resident in memory are activated as well ... typically other aspects of the schema that have been frequently associated with the stimulus item in the person's experience. ... From the outset, younger children may have a less well-developed ability to discriminate original material from additionally activated pieces of information that might have been part of the original material (pp. 47-48).

Prior Knowledge

Age differences in knowledge base contribute to age differences in memory performance.  Bjorklund (1987) argues that increased knowledge about entities can increase the ease with which information in semantic memory (e.g., memory for stories) can be accessed and activated.  As less effort is required to access specific items in memory and the relations among sets of items, more mental effort is available for using cognitive strategies for processing information.  Knowledge base, which increases with age, can therefore enhance memory performance by facilitating the use of deliberate memory strategies.  In the child witness studies that test semantic memory, age differences in memory performance might be explained by differences in knowledge base.

Bjorklund's theory might further explain why some studies find no differences in memory performance: knowledge base might be similar for younger and older subjects on the particular task.  Several laboratory studies have confirmed that children and adults have similar recall when prior knowledge is equivalent.  For example, researchers have found that word lists composed of age-familiar words are remembered better (Barrett & Wright, 1981; Lindberg, 1980) and learned more rapidly (Richman, Nida, & Pittman, 1976) by different age groups.  Chi (1978) has shown that adults do not have a memory advantage over 10-year-olds when familiarity with the situation is controlled.  Chi had subjects perform two memory tasks: recall a list of ten digits, and reproduce a pattern of chess pieces from memory.  Adults outperformed children in the digit-span test; but children's recall exceeded adults' by more than 50% on the chess pattern test.  These results were attributed to the knowledge advantage of the better performing age group in the respective conditions. In the former test, adults had the advantage of familiarity with digits and recall of digits (e.g., remembering phone numbers, social security numbers, etc.).  In the latter condition, the advantage went to the children, who were more experienced chess players.

In sum, schemas of older children are more complex, detailed, and varied than those of younger children.  These differences may account for advantages older children and adults have over young children in recall tasks.  The fuller schemas of older individuals enable them to encode more information during typical events, and to absorb information more rapidly.  In comparison with young children, older individuals have more numerous and flexible schemas.  Older individuals are therefore less apt to distort incoming information and recall.

Schemas also may account for other age differences in memory.  Depending upon the nature of the memory task, older children by virtue of their experience may be more familiar with the stimuli and therefore surpass younger children in testing.  However, if young children are questioned about repeated experiences, such as chronic physical or sexual abuse, they can be expected to recall a significant amount of information.  Children's knowledge of the sequence of a repeated event should be particularly reliable (Myles-Worsley, Cromer, & Dodd, 1986).  Schemas develop from experience and knowledge accretion.  The studies reviewed above suggest that advantages in memory stem from advantages in knowledge.  When children and adults have the same knowledge about an event, their recall should be equal.  In most child witness studies, the presentation of stimuli to subjects of different ages for brief periods of time may result in advantages for older groups who may have prior knowledge about the stimuli.  Many more studies are needed to examine young children's memory abilities for familiar events.  Finally, it should be noted that schemas for typical events can also result in distortions in children's (and adults') memories for details.

Cognitive Strategies

Mnemonic Strategies

Older children and adults are more likely than young children to spontaneously employ mnemonic devices, or purposeful strategies, to improve their memory (see reviews by Guttentag, 1985; Kail, 1984).  Such strategies are typically used at two levels: storage and retrieval.  Individuals may rehearse information or group related bits of incoming information to facilitate memory storage.  In laboratory studies, simple rehearsal strategies are apparent at the age of 7 or 8 years.  By the age of 10 years, rehearsal strategies increase in complexity as children begin to rehearse and organize stimuli according to its properties.  Individuals also may use grouping, or category, information at the retrieval level.  For example, in an attempt to retrieve information about furniture, individuals may use particular categories of furniture (e.g., bedroom furniture) to guide their retrieval.  The ability to use stimulus properties, or categorical information, improves between the ages of 10 years and adulthood.

Mnemonic and memory researchers almost exclusively focus on purposeful memory.  That is, subjects are told that they will be tested on the stimuli.  In everyday life, however, individuals spontaneously use mnemonic strategies.  More research is needed to estimate the age at which children are likely to use cognitive strategies to store and retrieve information about witnessed events.  More importantly, current knowledge could be applied to teach children strategies for maintaining and retrieving information about events related to a case they are testifying on.


Metamemory is knowledge of memory.  Metamnemonics is knowledge about one's memory skills and strategies, and when to activate various strategies.  Metamemory and metamnemonics may partially explain age differences in memory.  Very young children (e.g., preschoolers) may not be able to adequately detect erroneous information, or to monitor and avoid erroneous responses to questions.  Detection and monitoring of erroneous information are skills derived from "metamnemonic knowledge" that appear to follow a developmental trajectory (Kail, 1984).

Children's responses to direct questions might be more reliable if they answered only questions they were certain about, and if they forestalled answering questions that they were uncertain about.  In free recall tasks children seem to monitor themselves this way, and to provide relatively accurate information.  However, in recognition tasks, children seem to be less aware of their own uncertainty, or less capable of forestalling choice.

Awareness of uncertainty and the ability to forestall choice are subject to developmental change.  In a study by Scholnick and Wing (1988), third- through fifth-grade children and adults played a fantasy board game involving situations that either could be solved with additional information or were unsolvable.  Participants had to pay for information with play money.  On solvable tasks, additional information would reveal clues to locate and rescue a unicorn from a castle, which was the goal of the game.  Participants were informed that they should conserve their money in order to afford a ransom for the unicorn.  Therefore, players who determined that a situation was unsolvable should have moved to their next turn rather than guess at a solution.  Results indicated that the youngest children were more likely to guess than the older children and adults.  A shift in performance was evident for fifth graders, who, like adults, were less likely to guess on the unsolvable situations.

The Scholnick and Wing (1988) findings may generalize to memory tasks.  It is possible that younger children feel more compelled than older children to guess during recognition tasks than to say that they do not know a solution.  In a study by Hughes and Grieve, 5- and 7-year-old children would typically give answers to very bizarre questions, such as "Is red heavier than water?"  The emphasis in witness studies to give some form of answer (Dent & Stephenson, 1979), and to discourage respondents from "don't know" responses (Warnick & Sanders, 1980), may result in demand characteristics to which children have little resistance.  When an adult presents children with an array of possible responses, they are likely to surmise that one of those responses must be correct and proffer a response.

In sum, young children make less spontaneous use of rehearsal and organizational strategies than do older children and adults.  Laboratory studies have shown that young children are capable of using strategies effectively, although they generally do not do so spontaneously.  Knowledge about memory and memory strategies also plays a role in memory performance particularly on recognition tasks, at which young children display their greatest deficits.  Young children appear less adept at detecting erroneous information, knowing what they know, and monitoring erroneous responses.  Furthermore, when faced with unsolvable tasks, young children tend to guess while older children and adults are more likely to forestall an answer on the basis of insufficient information.

Attention and Encoding

Cognitive structures are required to assimilate and remember an event, but memory also depends upon attentional capacity.  Age differences in memory performance may stem from age differences in how much, or what kind of, information is initially attended to and encoded (cf. Case, 1984).  In general, young children attend to a greater variety of stimuli in their environments, and are easily distracted by "irrelevant" stimuli.  For example, in a study of children watching television, 2-year-olds were more likely to be distracted, play with toys, look around the room, and talk with other people than 4-year-olds (Anderson & Levin, 1976).  Occasionally, younger children's scattered attention may enable them to encode information that is relevant to an interrogator and missed by older children and adults.

In addition to developmental factors, situational factors influence children's attention and eventual recall of stimuli.  One such situational factor is the interest value of the stimuli.  For example, Renninger and Wozniak (1985) found that children's attention, recognition, and free recall was better for toys that had high interest value than for toys with low interest value.  Another situational factor is the setting in which encoding and remembering occur.  In a study of prospective memory conducted in a laboratory or home setting, 10- and 14-year-old children were instructed to either remove cupcakes from an oven or cables from a battery every 30 minutes (Ceci & Bronfenbrenner, 1985).  Children played a video game while the cupcakes were baking or the battery was charging.  The experimenters measured children's time monitoring (checking a wall clock).  Certain patterns of time monitoring are more adaptive for not forgetting to complete the task than other patterns.  Time monitoring was described by the authors as a "strategy."  However, time monitoring can also be considered a form of attention to the environment that facilitates children's prospective memory.  The results indicated a significant effect of setting on performance: strategic time-monitoring was more common in the home than in the laboratory.

Stress is another situational, or contextual, factor that may influence attention and encoding.  Among adults, heightened arousal during stressful events appears to decrease later eyewitness accuracy (see review by Deffenbacher, 1983).  As noted above (e.g., Goodman et al., 1987; Peters, in press a), the effects of stress on children's memory performance are mixed.  Child witness studies have yet to systematically examine the effects of stress on children's allocation of attention (e.g., eye fixation during task) or ability to encode information.

In sum, age and situational factors influence what individuals attend to in their environments, and therefore partially determine what information is encoded and stored in memory.  When left to their own devices, young children (e.g., preschoolers) are more likely than older children and adults to scan environments rather than focus on a single stimuli.  Younger children also appear more distractible than older children and adults.  Studies on context and memory suggest that laboratory settings may interfere with children's attentional capacities and use of effective memory strategies.  The effects of stress on children's attention, encoding, and memory are only beginning to receive empirical attention.

Distinguishing Fact from Fantasy and Appearance

Although most of the research on child witnesses has focused on memory, other cognitive variables are associated with children's competency as witnesses.  This section discusses children's cognitive abilities to discriminate the sources of their experiences and memories.  In particular, the discussion focuses on two questions: Can children adequately discriminate between real, or perceived, events and imagined events?  Can children adequately discriminate between different external sources of memory for an event (e.g., eyewitness memory versus memory of something someone told you, or hearsay)?

Data from several studies suggest that the ability to separate fact and fantasy develops with age (Lottan, 1967; Morison & Gardner, 1978; Scarlett & Wolf, 1979; Taylor & Howell, 1973).  Morison and Gardner(1978) found that the use of a fantasy-reality classification scheme is apparent in kindergarten children, and improves steadily through sixth grade.  Children in their study were given two sorting tasks.  First, a three-card sorting task was used to determine the extent to which children use reality-fantasy as a classification scheme.  Two of the three cards depicted fantasy figures (e.g., dragon, elf) and the third depicted a real entity (e.g., frog).  Children were given two opportunities to make pairings with the cards and to explain their pairings.  The incidence of fantasy pairing (e.g., dragon and elf) and fantasy explanations (e.g. "both are fake") increased steadily with age.  In the second task, children sorted a series of pictures into piles of "real" and "pretend" figures.  Children made very few classification errors, although errors decreased as a function of age.  Analysis of the classification errors revealed that more mistakes were made by calling fantasy characters real than by calling real characters fantastic.

Morison and Gardner's (1978) findings indicate that awareness of the pretend-real distinction develops at least by the time children attend kindergarten.  Other researchers suggest that the pretend-real distinction develops by the age of 3 years, and prior to the development of the theoretically related apparent-real distinction (Flavell, Flavell, & Green, 1987).  The ability to distinguish an object from its appearance (e. g, to know a candle that looks like an apple is really a candle) is a (meta)cognitive skill (Flavell et al., 1987) that has been ignored in the child witness literature.  However, the implications of deficiencies in making apparent-real classifications are obvious.  For example, a witness's reliability would be dubious if he could not distinguish a real gun from toy.

Johnson and Foley (1984) present data from four experiments that suggest children have difficulty discriminating real and imagined events in some situations, but not all.  For example, 6-year-olds did not confuse imagined events with perceived events, or things they had done with things they had perceived.  However, compared with 9- and 17-year-olds, 6-year-olds had more difficulty distinguishing what they had said from what they had only imagined themselves saying aloud (also see Foley, Johnson, & Rahe, 1983).  Both 6- and 9-year-olds had more difficulty than college students in distinguishing what they had only thought of doing from what they actually had done (also see Foley & Johnson, 1985).  Thus, children's abilities to distinguish memories for real versus imagined events are fairly intact from about the age of 6 years.  However, the memories of children this young can be muddled when they are distinguishing actions they did from those they imagined doing.

By focusing on children's abilities to distinguish the real from the imagined or fantastic in self-generated activities, Johnson and Foley (1984) have followed a long and distinguished tradition in this area (e.g., Freud, 1922; Kohlberg, 1969; Piaget, 1928, 1929; Werner, 1961).  However, this research line must turn in the direction of exploring children's abilities to distinguish pretend-real, or apparent-real, in other-generated activities.  The accuracy of children's testimony is likely to depend more on these latter abilities than on those explored by Johnson and Foley (1984).  For example, the reliable child witness should be able to distinguish what a person said they would do from what they actually did (e.g., did the man verbally threaten the woman, or physically assault her?).

Both children and adults seem to distinguish self-generated events from other-generated events better than they distinguish between different other-generated events (Lindsay & Johnson, 1987).  Therefore, individuals are probably better at separating what they thought about an event from what someone else told them about an event than they are at separating what two different people told them about an event.  This deficiency in separating different other-generated events can lead to inaccuracies in testimony, and is discussed in terms of suggestibility (see. below).

In sum, these studies suggest that the use of fantasy classifications and explanations increases with age.  Children's ability to separate real-pretend dimensions of events is apparent by the age of 3 years, and seems to develop before the ability to separate apparent-real dimensions.  Johnson and Foley's (1984) research has indicated that (a) children as young as 6 years of age perform as well as adults when asked to determine whether an event was perceived or imagined, (b) 6-year-olds do not perform as well as older individuals when distinguishing imagined from spoken words, and (c) 6-and 9-year-olds have more difficulty than adults in distinguishing what they had thought of doing from what they actually had done.  In general, children's deficits appear to result from confusions between self-generated behaviors and imaginings of self-generated behaviors.  Much more research needs to be conducted on children's capacity to distinguish between memory sources of other-generated events.

A general problem with the studies in this area is that they do not apply directly to the types of situations children might be asked to testify on in court.  The stimuli are often artificial and not as rich as the stimuli found in the context of a real event.  Moreover, the stimuli in these studies often are chosen for their age relevancy (e.g., elves and frogs).  Using age-relevant stimuli probably increases the interest in the stimuli, but this also may increase children's ability to make real-imagined and apparent-real distinctions.  In the real world, children may have to testify about ambiguous or novel events, and this could have a strong impact on their discriminatory abilities.  Another problem with the above studies is that experimental instructions to imagine an event may produce mental representations that are qualitatively different from those spontaneously generated by subjects, and such instructions may lead to less stable memories (Lindsay & Johnson, 1987).

Social Factors Influencing Children's Testimony


Children are generally held to be more suggestible than adults (Leippe, Brigham, Cousins, & Romanczyk, 1989; Yarmey & Jones, 1983), despite a lack of substantial empirical evidence to support this notion.  Suggestibility in children is an especially important research topic because children are asked more leading questions in courts than adults (Thomas, 1956), and many jurisdictions permit child witnesses to be asked leading questions.  As noted above, children are asked leading questions because they often provide spontaneous reports that are poor in content, and the examiner needs to probe for more details.  In addition to leading questions, children may at times be persuaded to interpret others' behaviors in incriminating or exculpating ways.  For example, one parent may encourage a child to support a false allegation of abuse against the other parent in order to get custody in a divorce (Green, 1986; Wakefield & Underwager, 1990).  Alternatively, one parent may encourage a child to recast an incestuous relationship with the other parent as an ordinary relationship in order to avoid "publicity" or to "keep the family together" (cf. Wideman, 1990).

Suggestibility is defined here as the extent to which individuals' reports and representations about events vary from their original form as a result of information obtained subsequent to the event.  This definition allows for both cognitive and social explanations of suggestibility effects, or an interaction of such factors (cf. Baxter, 1990; Cole & Loftus, 1987; Dodd & Bradshaw, 1980; Loftus & Davies, 1984; Moston, 1990; Weinberg, Wadsworth, & Baron, 1983).

The mechanisms of suggestibility effects are currently the subject of much debate.  Some researchers hold that misleading information impairs memory by altering the original information in memory so that it is lost from memory (Loftus, 1979; Loftus & Loftus, 1980); others argue that misleading information merely renders the original information inaccessible, or impossible to retrieve (Bekerian & Bowers, 1983; Christiaansen & Ochalek, 1983).  Still others argue that memory impairment is only one of many factors that contribute to the "misinformation effect" not the least of which are demand characteristics (cf. Weinberg et al., 1983), forgetting of original information (cf. McCloskey & Zaragoza, 1985), and source credibility (cf. Ceci et al., 1987).  These various theoretical explanations of suggestibility effects will be discussed following a presentation of the recent empirical findings in this area.(3)

Recent Studies of Suggestibility

In 1984, Loftus and Davies reviewed four published studies (Cohen & Harnick, 1989; Dale, Loftus, & Rathbun, 1978; Duncan, Whitney, & Kunen, 1982; Marin, Holmes, Guth, & Kovac, 1979) and one unpublished dissertation (i.e., Murray, 1983) on the topic of suggestibility in children.  The basic paradigm of each study was the same: (a) participants observe a brief live event, slide sequence or film, or hear a story; (b) participants are asked questions that contain embellishments or distortions of the originally witnessed event; and (c) participants answer another series of questions about the observed event, in order to elicit any memory content that has been distorted by the intervening questions.

Loftus and Davies did not find a clear developmental trend in the studies they reviewed.  Dale et al. (1978) found that 4- and 5-year-old children's reports were influenced by leading questions.  Similarly, Marin et al. (1979) found that leading questions biased responses in subjects aged 5 to 22 years, and younger children were no more susceptible to leading questions than adults.  In contrast, Cohen and Harnick (1980) found that the youngest cohort in their sample (9 years) was significantly more likely to agree with leading questions than the two older groups (12 years and college-aged).  However, contrary to Cohen and Harnick's results, Duncan et al. (1982) did not find developmental differences in suggestibility in subjects ranging in age from 6 years to adulthood.  Moreover, Duncan et al. noted that the young children with strong memories may have been less suggestible than older children and adults.  Finally, Murray (1983) found that leading questions significantly reduced accuracy in a recognition task performed by 7- and 1 1-year-olds, but there was no age difference in degree of suggestibility.

The main conclusion following Loftus and Davies' review was that "these studies support the conclusion that adults spontaneously recall more about events they have witnessed than do children, but not the simple notion that children are always more suggestible than adults" (p.62).  Four factors may explain the inconsistencies across studies: (a) different age groups; (b) different delay intervals between the initial event, the suggestive information, and the final test; (c) different stimuli (e.g., film, live event, cartoons); and (d) different dependent measures, or final test formats (e.g., free recall, recognition).  Methodological problems, such as inadequate control of memory effects that may supersede suggestibility effects (cf. Zaragoza, 1987), also could explain the differential outcomes.  Children who appear more suggestible may simply pick the wrong solution or choice on a memory task more frequently because they have forgotten the correct choice, and not because they had a suggestion.

More sophisticated studies of suggestibility in children have emerged since Loftus and Davies' review.  Several studies have found no age differences in the suggestibility of children tested with free recall methods.  Saywitz (1987) presented 8-, 11-, and 15-year-old children with an audiotaped story about a theft.  Immediately after hearing the audiotaped story, children received three written tasks, including six direct questions about the thief three of which contained suggestive misleading information that did not appear in the original story.  For example, the children were asked about the color of the "old" man's hair, which is suggestive because age was not mentioned in the story.  Five days later, children were asked to describe the thief in detail; their free-recall was then examined for intrusion of the suggested information.  Suggestibility effects were minimal: only 14 out of 72 subjects spontaneously "recalled" any of the misleading propositions in their descriptions.  The mean number of intrusions for all age groups was quite small, ranging from .04 to .33.  There was a marginal effect of age on suggestibility, with younger children being slightly less susceptible than older children.  The trend was not an artifact of the amount of recall produced, because number of words per description was similar for all ages.

Saywitz's findings bolster the argument that young children's free recall is reliable even following suggestive questioning.  However, other researchers have shown that accuracy in free recall following suggestive questioning does not guarantee accuracy in recognition tasks.  For example, Goodman and Reed (1986) had 3- and 6-year-old children and adults briefly interact with an adult.  Four to five days later participants were asked suggestive and objective questions, followed by a free recall task.  Similar to Saywitz's findings, these results indicated that children and adults seldom assimilated misleading information into their free recall.  Only four members of the two older groups incorrectly recalled items from the suggestive questions, and none of the 3-year-olds incorporated misleading information into their reports.  Unlike Saywitz, however, Goodman and Reed suggest that the lack of errors from young children resulted from the overall dearth of information they provided.  Goodman and Reed also were able to compare suggestibility effects on a recognition task.  In response to four suggestive questions, adults were more resistant to the suggestions than 6-year-olds and 3-year-olds, and 6-year-olds were more likely than 3-year-olds to answer suggestive questions correctly.  Thus, developmental differences in suggestibility effects appear to be underestimated when free recall methods are used.

Two other studies reported by Goodman et al. (1987) support the hypothesis that suggestibility diminishes with age.  In contrast to Goodman and Reed's (1986) investigation of children's witness abilities following an innocuous event, Goodman et al. (1987) examined children's witness abilities in the context of a stressful event.  In the first study, children ranging from 3 to 7 years old had blood drawn (stress) or had designs rubbed on their arms (control).  Three to four days later they were asked to recall what happened and to answer objective and suggestive questions.  The stressed children did not differ from control children in their degree of suggestibility.  However, there was a significant positive correlation between age and resistance to suggestion.

In the second study, 3-to 6-year-old children received inoculations at a clinic, and were asked questions similar to those asked in the blood drawing study: free recall was elicited, and objective and subjective questions were asked.  However, there was no control group in this study, and a seven- to nine-day delay was added to testing.  No analyses were reported on intrusion of misinformation in free recall, and the authors did not clarify whether free recall preceded or followed suggestive questioning.  Suggestive questions included leading information about a person, room, or action associated with the inoculation.  The 5- to 6-year-olds were more resistant to the misleading information than the 3- to 4-year-olds.  Interestingly, the older children appeared to be more resistant to misinformation about actions (75% correct) or a person (76% correct) than about the room (57% correct).  The authors did not report an effect of the seven- to nine-day delay versus the three- to four-day delay on suggestibility.

King and Yuille (1987) report that in a series of three studies they conducted with colleagues, two showed younger children to be more susceptible to accepting misleading information, and the third showed no differences in suggestibility among 8- to 14-year-olds.  In the first study by King and Yuille (1986, cited in King & Yuille, 1987), children witnessed a person tending to plants, and were subsequently asked two misleading questions, one about a watch that the confederate supposedly had worn, but did not, and the other about removing a leaf with scissors that were never actually used.  Later, when asked about the watch and the scissors, the 6-year-olds were significantly more likely to agree with the misleading suggestions than were the older children (9 to 17 years).

It is not clear whether children's memory for the plant-care event was distorted, or whether children ever had a representation of the confederate with a watch or using the scissors.  Anecdotal information provided by King and Yuille suggests that demand characteristics may have influenced children's reports.  Specifically, several of the participants who agreed with the misleading suggestion admitted that they had simply "gone along" with the suggestion even though they had no specific recollection of the event detail in question.

In the second study, conducted by Yuille, Cutshall, and King (1986, cited in King & Yuille, 1987), 8- to 14-year-old children witnessed a staged bicycle theft.  Following the event, children were asked objective and misleading questions.  The misleading questions did not affect witnesses' reports.  Children from all age groups were very accurate in their responses, although the specific number of intrusions for each age group were not presented.  King and Yuille concluded that the misleading questions referred to items that were salient to children (e.g., the thief's shoes), and therefore they were less susceptible to suggestibility effects.  As noted above, Goodman et al. (1987) also concluded that salient aspects of events (e.g., actions) were more memorable and less influenced by suggestions than peripheral details.

Several recent studies have provided compelling evidence that demand factors account for some of the decrement in memory performance in the presence of suggested information.  Zaragoza (1987) presented 3- to 6-year-old children with a narrated slide sequence of a children's story, and subsequently provided a verbal synopsis of the story in order to introduce misleading information about two of the slides (e.g., a boy playing "basketball" in the story was said to be playing "football" in the synopsis).  Following the synopsis the children were shown pairs of pictures and asked to select the one corresponding to the slide they originally saw.  Two different final memory tests were employed in this study.  In the "original" test condition, original slides were matched with slides depicting suggested information.  In the "modified" test condition, children had to choose between the original slide and one that depicted a unique detail that had not been present in the original or synopsis story.  For example, a child who originally saw toy plane and was misled in the synopsis with toy boat, would be offered a modified test choice of toy plane (original item) and toy train (new item).  The modified design is used to examine the extent to which subjects' post-suggestion reports vary as a function of memory distortion versus other factors (cf. McCloskey & Zaragoza, 1985; Zaragoza, McCloskey, & Jamis, 1987).  If misleading information does not impair subjects' memory for original events, then misled subjects will perform as well as controls; however, if memory was distorted by misinformation then misled subjects will have to guess at the correct answer on the final test (e.g., 50% correct).  Subjects who remember the original stimulus will likely choose it over the new stimulus in a recognition task, because the new stimulus has no demand characteristics associated with it.  Thus, the modified design attempts to control for social demand effects and tests whether postevent information distorts memory.

Performance in the modified test group indicated that children's memory for the original event was not impaired by exposure to misleading information: the misled group's performance was 69% correct and the control group's performance was 70% correct.  However, in the original test group, the misled group's performance was only 54% and the control group's performance was 77% correct.  As Zaragoza notes, the latter finding confirms that the children actually attended to and encoded the misleading information.  In a replication study, Zaragoza (1987) presented misinformation twice during the synopsis and found comparable results.  The data from these two studies suggest that children are resistant to forgetting information that is followed by suggestive information at least when the stimulus is familiar and interesting stories.  However, there was an effect of suggestibility in the original test, and Zaragoza argues that demand factors accounted for some portion of the effect.  Unfortunately, age was not considered as a factor in any of the analyses.

In a series of developmental studies, some of which employed the modified test procedure, Ceci et al. (1987) provide more direct evidence of the importance of demand characteristics in the questioning of children.  In Experiment 1, the original test paradigm was used and young children were found to be particularly vulnerable to suggestion.  Children from four age groups were sampled: 3 to 4, 5 to 6, 7 to 9, and 10 to 12 years.  Each group was told a brief story that was accompanied by line drawings.  Following a one-day delay, suggestive questions were put to the biased group, while the unbiased group heard objective questions.  The questions pertained to relatively peripheral details, such as the kind of food shown in one picture or how the character in the story was feeling.  Three days later children were tested with a two-alternative, forced-choice picture recognition task.  There was an effect of age on children's suggestibility: the 3- to 4-year-old biased group remembered significantly less than the two oldest biased groups.  The authors reported that misleading information affected the 3- to 4-year-olds most strongly, but failed to note that the 5- to 6-year-olds did not perform significantly better than the 3- to 4-year-olds (nor significantly worse than the other two age groups).  These results are consistent with those of Goodman and Reed (1986) and the inoculation study by Goodman et al. (1987).

In Experiment 2, preschool children's (M = 4.6 years) susceptibility to misleading information was reduced when another child, as opposed to an adult, provided the misleading information.  Using the same paradigm as in Experiment 1, Ceci and his colleagues found an 18% difference in performance in favor of the unbiased group.  Thus, there is still a suggestibility effect, even when a less authoritative source is providing the misinformation.  However, the 18% difference in performance is insignificant relative to the 42% difference in biased and nonbiased preschoolers who were given information by an adult in Experiment 1.  Thus, these results indicate that demands characteristics associated with an adult giving suggestions to a young child partially account for the suggestibility of young children in Experiment 1.

In Experiment 3, Ceci and his associates examined whether the method of assessing memory in the first two studies accounted for the differences in performance, as opposed to the effect of the misleading information on children's memory.  The procedures were similar to those used in the first two experiments: a story was read to children, one day later a child introduced misleading information, two days later memory was assessed.  However, a modified test condition was added.  In the modified test condition, children's final recognition task consisted of a choice between a picture depicting original story information (e.g., boy wearing a red cap) and new information (e.g., boy wearing a green cap).  Performance in this condition was contrasted with performance in the original test condition, which presented children with a choice between original information (e.g., boy wearing a red cap) and suggested information (e.g., boy wearing a blue cap).

The performance of children in the original test condition and control groups replicated the results of the first two studies.  Moreover, children in the modified test condition (71% correct) performed better than children in the original test condition (52% correct).  These results nearly replicated Zaragoza's (1987) results (modified = 69% correct versus original = 54% correct).  However, the modified group in the Ceci et al. study was still more likely to choose the new information (green cap) than the control group who had received no misinformation.  Although the accuracy rate of the modified group was 71%, the control group's accuracy rate was significantly higher at 87%. These findings are important because they suggest that distortion can occur following post-event suggestions even when demand characteristics are controlled for.

Although the modified procedure attenuated the misinformation effects, the misled group still did not perform as well as the control group.  These latter findings of Ceci and colleagues counter Zaragoza's (1987) finding that the performance of children in the modified and control groups was equivalent, as well as the finding with adults by McCloskey and Zaragoza (1985).

Ceci et al. provided additional support for their findings in Experiment 3 by conducting a fourth experiment.  Experiment 4 added a sample of preschoolers and adults and included a few changes to the basic procedure used in Experiment 3: critical stimuli were counterbalanced across testing conditions, the child misleader provided two pieces of accurate information (his name and home town), and a new story was presented.  Results for the children and adults were analyzed separately because the adults performed at ceiling levels for all three conditions.  The authors noted, however, that although adults performed better than children in the original and modified test conditions, performance was comparable between young and older control subjects both performed at ceiling levels.  In addition, children were somewhat better than adults at accurately recognizing the misleader's name and home town.  The results for the preschoolers were similar to the earlier studies: the control group (no misinformation) performed better than the modified test group (88% versus 72% correct) which, in turn, performed better than the original test group (46% correct).

As noted above, there are inconsistencies across studies using the modified procedure.  In addition to problems of replication, there are larger questions regarding the application of this work to children's testimony in real life.  For example, a child may be asked to testify about a fight she witnessed between her parents.  If the child's memory is faint, an interrogator's suggestion that the father started the fight when in fact the mother had provoked the fight might confuse the child's memory.  If the child is interrogated again, at a later date, it is likely that she will be asked to determine whether the mother or the father provoked the fight, and not whether the father or a next-door neighbor started the fight.  If, in fact, this latter, modified procedure were to be used, the results would not get the interrogator closer to the truth.  Given the choice between the father or the neighbor, the child would probably choose the father because she would have no recollection whatsoever for the neighbor being involved.  At least with the original procedure, there would still be some ambiguity in the child's mind about whether the mother or the father provoked the fight, and the truth might emerge

Other procedural factors in suggestibility studies, such as strength of the suggestion, as well as the nature of the outcome, appear to be an important in estimating children's suggestibility.  Clarke-Stewart, Thompson, and Lepore (1989) found that following strong and evaluative suggestions, children were able to recall accurate details about events that happened to them, but their interpretations of events were so distorted and colored by suggestions that the facts recalled did not reveal the truth of the situation.  Their study departs from prior ones in this area by focusing primarily on children's interpretations of events, rather than on their accuracy in recall of event details.  This is an important distinction, because how children and their interrogators interpret events (e.g., abusive or innocent) may be more important than the type of details recalled.

In the study by Clarke-Stewart and colleagues(1989), 5- and 6-year-old children saw and interacted with a confederate ("janitor") in a laboratory designed as a "playroom."  The janitor did some perfunctory cleaning of the playroom and then either cleaned or played with a doll in front of the child.  In each condition, the child was entreated to get involved.  The cleaning and playing actions were parallel, but had different meanings depending upon the condition.  For example, the cleaning janitor noted that the "doll is dirty" and sprayed the face to wipe it clean; the playing janitor noted that he likes to "play with dolls and spray them in the face just for fun."

Approximately one hour later, an interrogator ("Janitor's boss") questioned the child about the janitor's actions.  After the boss left, a second interrogator re-interviewed the child.  Interrogations followed one of three scripts: neutral, incriminating, or exculpating.  Nonsuggestive statements and questions were administered in the neutral condition; suggestive statements and questions were administered in the other conditions.  In the incriminating condition, the interrogator said such things as: "You know he (janitor) sometimes stops working and plays with dolls ... he sometimes does naughty things with the dolls like take off their clothes."  In the exculpating condition, the interrogator said such things as: "You know the janitor is supposed to clean the dolls ... clean them all over, even under their clothes."  In both interrogations conditions, suggestions started subtly and became more forceful.  Also, questioning started in a broad, open-ended fashion and became increasingly explicit and close-ended.  Global ratings were made on whether the open-ended questions were colored toward cleaning or playing; the close-ended questions consisted of factual items (e.g., "did he spray the dolls face?") and interpretive items (e.g., "was he cleaning the doll or playing with it?").  Shortly after the second interview, the child was united with his or her parent, who reiterated the questions about the janitor.  Finally, one week later, the parents again quizzed the child about the janitor's activities.

Consistent with prior research, children in the nonsuggestive condition were able to respond with accurate, albeit limited, information about the janitor.  These children also answered an average of 82% of the factual questions correctly, and an average of 83% of the interpretive questions correctly over all of the interrogations.

However, children in the suggestive conditions quickly, and repeatedly over the three interrogations, gave interpretations that were consistent with suggestion and inconsistent with what they had witnessed.  Following the first, subtle suggestion and question, 25% of the children had distorted what they had seen to be consistent with suggestion.  The error rate on the factual questions was similar to that found in the neutral group.  However, 90% of the children answered the interpretive questions incorrectly, following the interrogator's suggestion.  By the second interrogation, 83% of the children gave an open-ended description of the janitor's actions that was consistent with suggestion and inconsistent with the witnessed actions.  Again, the error rate on factual questions was similar to that in the neutral group, and 90% of the children answered the interpretive questions incorrectly.

Most striking, perhaps, was that during the parent interview and the one-week follow-up questionnaire, children in the suggestive situations continued to present an interpretation of events that was consistent with the interrogators' suggestions.  Thus, the young children's interpretations of an event were easily and strongly affected by suggestive adults, although their recollection of the facts was not strongly affected by the interpretive suggestions.

In sum, the studies reviewed above show that children as young as 3 years old can give accurate albeit limited free recall of witnessed events following suggestive questioning (Saywitz, 1987; Goodman & Reed, 1986).  Starting at approximately 8 years of age, misled children's free recall tends to be as accurate and complete as misled adults' (Saywitz, 1987).  Researchers who compared free recall and recognition performance found that free recall methods may underestimate the effects of suggestive questioning (Goodman & Reed, 1986).  Studies that used objective questioning techniques (e.g., recognition tasks, true/false, yes/no) found that resistance to suggestibility increases with age.  Preschoolers' performance is especially damaged by leading information (Ceci et al., 1987; Goodman et al., 1987; Goodman & Reed, 1986; Zaragoza, 1987), but by the age of 7 to 10 years, children appear to be no more vulnerable to suggestion than adults (King & Yuille, 1987; Ceci et al., 1987).

A variety of factors contribute to the degree of children's suggestibility.  Memory for the original event seems particularly important.  Studies that showed more complete recall with increasing age found that suggestibility effects diminish with age (e.g., King & Yuille, 1987; Cohen & Harnick, 1980).  However, the inferential link between memory and suggestibility must be drawn with caution.  The procedure of testing memory in most of the cited studies was the original testing procedure which tends to overestimate misinformation effects.

Studies that employed the modified testing procedure revealed that demand characteristics in the original testing procedure account for some of the variation in performance between misled subjects (poor performers) and control subjects (normal performers) (Saywitz, 1987; Ceci et al, 1987).  Other studies showed that the experimenter exerts a certain demand effect on the subject.  When the source of misinformation is a person with marginal authority (e.g., a child), the effects of misinformation are reduced (Ceci et al, 1987).  Finally, suggestibility effects can be inflated if young children "go along" with the misleader (King & Yuille, 1987).

In addition to age, memory, and social demand characteristics, aspects of the testing stimuli may influence children's suggestibility.  For example, several studies showed that children effectively resist suggestions about matters that are salient and memorable to them, such as actions, or characteristics of a main character in the original event (Goodman et al., 1987; King & Yuille, 1987).  The time delay between the initial event, the suggestive information, and the final test also may influence suggestibility.  However, the results are mixed on this factor.  Some studies demonstrated age differences in memory performance as a function of delay (e.g., Jones et al., 1988) whereas others did not (e.g., Saywitz, 1987).  Moreover, Goodman et al. (1987) reported that memory for a live event declined significantly after a seven- to nine-day delay for 3- to 4-year-olds, but was consistent for 5- to 6-year-olds.

A final, general caution is warranted in interpreting the above studies.  With the exception of the study by Clarke-Stewart and colleagues (1989), the suggestions made in the reviewed studies are very weak, and do not refer to personally meaningful events.  In the real world, it is likely that the people making suggestions to children (e.g., parents, prosecutors, principals) will be making evaluative suggestions, such as "the man was bad to touch you there" or "the mean woman hurt you when she grabbed you arm, didn't she?"  These types of suggestions repeated several times by an adult may have a strong effect on children, and may color their interpretations of an event.


In addition to suggestive statements and questions, witnesses may be socially motivated to distort their testimony.  Most studies on children's memory for events assume that children will voluntarily report all of the information that they can remember.  This is probably a fair assumption given the impersonal nature of the events and consequences of reporting than in typical witness studies.  However, when the events and consequences of a child's testimony are personally meaningful, testimony may be influenced by motives.  For example, a child whose parents are divorcing may be influenced by one parent to accuse the other of child abuse.  Media stories on such cases seem to keep pace with the rising rate of divorce (e.g., Laurino, 1988).

Fear also might motivate children to lie or withhold information.  In child sexual abuse cases, the adult typically attempts to bribe or coerce the child to secrecy (e.g., deYoung, 1986).  A child may fear retaliation for implicating a sexual or physical violator in a trial especially if the child is familiar with the accused.  Alternatively, an abused child may attempt to protect the abuser if he or she is a familiar and trusted adult which is not uncommon (see Finkelhor, 1984).  Unfortunately, no published studies have systematically investigated the role of motives in children's recall or testimony of a witnessed event.

Other Social Influences

Individual differences in social background and socialization by parents may influence children's capabilities as witnesses.  For example, parents' values related to child punishment, conformity to authority, and truth-telling may predict children's suggestibility or compliance with adults.  Unfortunately, to my knowledge, data on these indirect social influences on children's testimony have not been published.  Other social factors that influence children's reports, such as characteristics of interviewers (e.g., Dent, 1978), also have been largely ignored in the psycholegal literature.

Methodological Approaches

Typical Paradigms

Child witness studies share common methodological features.  One common characteristic is a focus on age differences in memory for events.  Samples typically range in age from middle school age to college age.   Recent studies have included preschool (e.g., Ceci et al., 1987) and elderly samples (e.g., List, 1986).  Child witness studies generally examine two broad forms of memory: free recall and cued recall/recognition.  In order to test memory, children are presented with an event and then, following a delay period, are asked questions about the event.  This approach can be simply referred to as the stimulus-delay-test paradigm.

The stimulus event is usually a narrative (e.g., Ceci et al., 1987), a series of slides (e.g., List, 1986), or a film (e.g., Cohen & Harnick, 1980).  Less frequently, the event is a live, staged-event, which the subjects either observe (e.g., King & Yuille, 1987) or participate in (e.g., Goodman & Reed, 1986).  The delay between the stimulus presentation and memory testing ranges from a few minutes to several days after exposure to the stimulus.  A few studies have employed delays as long as 8 weeks (e.g., Jones et al., 1988).  After witnessing the event, memory is measured by asking open-ended, free recall questions (e.g., "Tell me everything you remember about the man who came in here") or more direct, cued recall (e.g., "Tell me the color of the man's hair") and recognition questions (e.g., "Was the man's hair blonde or brunette?").

Studies on the effects of suggestibility, or misinformation, on children's memory and testimony are becoming more common.  These studies employ a variation of the above stimulus-delay-test paradigm, to be referred to as the stimulus-suggestion-delay-test paradigm.  After the stimulus, subjects are given misleading information, usually in the form of a question.  Memory for the event is tested following a delay.  The most common test is the original or standard form, in which the participants must recall whether they observed the original or the suggested stimulus.  The alternative test is a modification of the original test, in which participants must recall whether they observed the original stimulus or a unique stimulus that was neither observed in the original event nor suggested following the event.  The study by Clarke-Stewart et al. (1989) departs from other suggestibility studies by using a stimulus-delay-suggestion-test approach.  This approach may reflect more typical life situations.  That is, a child is not likely to he questioned about an event until some time has expired, at which time suggestions are also likely to start.

Interestingly, similar studies employing the stimulus-delay-test paradigm, or the stimulus-suggestion-delay-test paradigm, often result in different outcomes.  Seemingly slight procedural differences between studies may cause substantial differences in outcomes.  For example, the type of stimulus event seems to determine memory strength for that event.  Most noticeably, memory for narrated or recorded events may be less accurate than memory for a live event.  It is possible that active involvement in an event leaves a stronger impression than passive observation.  Children (and adults) probably pay closer attention to live events and therefore encode more information about such events.  Thus studies using live events may not be directly comparable with studies using recorded events.

Naturalistic Approaches

Studies using live events are usually more ecologically valid than those using recorded events because child witnesses are typically asked to recall live rather than recorded events.  However, ecological validity refers to both the form (e.g., live versus recorded) and the content (e.g., crime versus fairy tale) of the stimulus.  The events that child witnesses are asked to recall in real life are probably not of the neutral or innocuous variety found in most child witness studies.  Therefore, simply measuring memory for live events does not guarantee that the findings will generalize to the real world.

There are two major obstacles to examining children's witness abilities under realistic and compelling conditions.  First, ethical concerns related to the protection of human subjects are particularly stringent when children are involved.  Children typically cannot be exposed to crime scenes or other potentially harmful stimuli.  Second, the degree of control over variables particularly contaminating or confounding variables decreases as researchers move from the laboratory to the real world.  This last issue has led to a heated exchange over the value of ecologically valid research on memory (a.k.a. everyday memory, autobiographical memory, naturalistic memory research).  The central concern appears to be whether or not ecologically valid research produces generalizable results and theories (see Banaji & Crowder, 1989; Ceci & Bronfenbrenner, 1991; Conway, 1991; Roediger, 1991; Tulving 1991).  However, as noted above, laboratory results are not always replicable in real-world settings (e.g., Ceci & Bronfenbrenner, 1985).  Thus, the pragmatic constraints in the laboratory are matched by the ethical and methodological constraints in the real world.  A challenge to future researchers is to maximize both ecological validity and generalizability through controlling crucial parameters of "naturalistic" or "life-like" studies.

In addition to the form and content of stimuli, other characteristics contribute to the degree of ecological validity of child witness studies.  First, the number and duration of exposures to stimuli is notably limited in child witness studies.  Children seldom observe events more than once, or for more than a few seconds or minutes.  There are no reliable statistics available on the exact kinds of crimes children witness and report on, but it is likely that they testify about crimes they have experienced or seen more than once, such as chronic sexual or physical abuse.  This is an important area for future research because repeated experiences may be more memorable to children as they develop schemas and a knowledge base for these experiences.

Second, children also are likely to be familiar with the perpetrator of the crime, as when a teacher, neighbor, or relative is witnessed in a crime.  Future research would benefit by using adult confederates and interrogators familiar to the child.  This approach may provide data relevant to child-victim witnesses, because children often know their victimizers.  What happens, for example, when children are questioned about dubious activities involving their parents?  Or, how do children respond to suggestive questioning by their school teacher or principal?

Third, the delay between observing an event and reporting about it is usually briefer in experiments (e.g., a few days or minutes) than what one would expect in real life (e.g., several days or weeks).

Finally, the number of times children are questioned and the number and familiarity of different interrogators used in child witness studies probably do not approximate real life situations.  Rather than receiving one interrogation from an unfamiliar person such as an experimenter in a child witness study child witnesses in real life are subject to multiple interrogations and answer to many interrogators, who range from the familiar (parents and teachers) to the unfamiliar (social workers, police, lawyers, and judges).

Laboratory Approaches

Although witness researchers should strive for greater ecological validity, some traditional laboratory studies produce robust findings that can generalize to everyday life.  For example, results from both traditional developmental laboratory studies and more naturalistic studies have shown that (a) free recall memory tends to be more accurate and less complete than recognition memory (e.g., Dent & Stephenson, 1979; Goodman et al, 1987; Todd & Perlmutter, 1980), (b) young children generally have greater difficulty retrieving information from long-term memory than do older children and adults (e.g., Brown, 1979; Chi, 1976; Goodman et al., 1987; Johnson & Foley, 1984), and (c) under certain conditions, young children can perform as well as older children and adults on cued recall or recognition tasks (e.g., Brown, 1973; Ceci et al., 1987; Chi, 1978; Kobasigawa, 1974; Perlmutter & Meyers, 1974, 1975, 1970; Ritter, Kaprove, Fitch, & Flavell, 1973; Saywitz, 1987).

Analytic Approaches

A final issue that child witness researchers should attend to is their method of data analysis.  Most of the studies discussed in this article, particularly the developmental ones, have examined group differences in witness abilities.  In real life situations, individual differences in memory, expressiveness, and resistance to suggestions will influence the reliability of their testimony.  Moreover, competence is likely to be determined on an individual, or case-by-case, basis.  Analyses based on average, or group, performances tend to hide individual differences.  Indeed, researchers interested in differences between groups purposely design studies to minimize within group differences (e.g., use wide age margins).  There is a great need to develop analytical methods that are sensitive to individual differences in witnesses' abilities.  A greater emphasis should be placed on determining "developmental effects" rather than "age effects" on children's memory and testimony.  That is, researchers should examine developmental factors such as prior experience (e.g., parenting styles, education) and maturity (e.g., social competence, moral and cognitive levels of development).  This approach could potentially be valuable in courtroom procedures for identifying those children who are likely to be very suggestible or have great difficulty in accurately recalling an event.  This approach also would shift the focus of child witness research from description to explanation of differences in child witness capabilities.

Conclusions and Future Directions

Systematic psychological research on the child witness has emerged only in the past two decades, despite the fact that there has been interest in this area for nearly a century (e.g., Binet, 1900, 1911; Marple, 1933; McCarty, 1929; Messerschmidt, 1933; Pear & Wyatt, 1914; Stern, 1910, 1939; Varendonck, 1911; Whipple, 1909, 1911, 1912).  The rising tide of child witness research is primarily due to legal and social concerns stemming from children's increasing contact with the courts.  Two issues are particularly acute: (a) describing children's capabilities as witnesses, as well as the factors that influence those capabilities, and (b) determining how child witnesses should be treated and questioned in order to obtain accurate testimony.  This article highlighted the complexity of these issues by discussing the research literature on cognitive and social factors related to children's witness capabilities.  Three central questions guided the discussion: (a) Do children have the cognitive skills to comprehend and accurately report a witnessed event?, (b) Can children discriminate memories of real events from imagined or apparent events?, and (c) How susceptible are children to suggestions?

The central cognitive variable in child witness research is memory.  Most of the studies reviewed indicate that free recall increases until early adolescence, and then approaches adult levels.  Free recall is generally very accurate, even among preschoolers.  Cued recall and recognition techniques of questioning elicit more details from memory than free recall approaches, yet the details are less accurate.  Children's performance on cued recall tests seems to approach that of adults' by the age of 6 or 8 years considerably earlier than when free recall skills mature.  Research on the effects of stress on free recall, cued recall, and recognition is only now beginning to emerge.  Laboratory research on stress and memory suggest that memory for focal features of an event (e.g., things that experimental subjects are asked to focus on) may be facilitated by stress, but memory of incidental features and more complex meanings of an event may be degraded by stress (see review in Cohen, Evans, Stokols, & Krantz, 1986, pp.145-150).

Child witness research on stress effects could also benefit by attending to the many factors known to influence stress reactions (e.g., chronicity of and control over stressors).  For example, stress researchers have found many cognitive deficits among children exposed to chronic environmental stressors (e.g., noise, crowding) (Cohen et al.,1986).  These findings suggest that certain groups of children may have greater difficulty with encoding and recalling events than other groups.  In addition, the focus on physiological arousal should he expanded to include characteristics of stress appraisals (e.g., Lazarus & Folkman, 1984), or the meanings that children attribute to their testimony.  For example, children that perceive something bad will happen to them, or a defendant, may be more threatened by an interrogator than children with more benign appraisals.  One result of negative appraisals of event outcomes could be the purposeful withholding of information, or distortion in memory resulting from fear and nervousness.

As an applied science, child witness research is primarily descriptive.  Yet, in an attempt to explain developmental differences in memory, a few child witness researchers have turned to developmental theory (e.g., List, 1986; Saywitz, 1987).  Schema theory and the constructs of metamemory and metamnemonics are particularly promising explanators of developmental differences in memory.  Older children and adults generally have more complex schemas, more knowledge, and better command of encoding and retrieval strategies than young children.  With increasing age, children also develop capabilities to focus and maintain attention on relevant environmental features.  Naturally, age differences in attention and encoding contribute to differences in memory.  Theoretical models describing the interplay of these cognitive variables schemas, knowledge, strategies, attention, and encoding are potentially useful for explaining individual and group differences in witness capabilities.  The descriptive results of child witness studies may, in turn, serve as tests of the generalizability of developmental models in the real world.

Another important cognitive skill for witnesses is the ability to distinguish the various sources of their memories for events.  Use of fantasy classifications increases with age, although children start to distinguish between real-pretend dimensions at an early age.  The ability to distinguish self-generated behaviors from imaginings of self-generated behaviors takes longer to develop, as does the ability to make apparent-real discriminations.  Researchers in this area need to further investigate children's abilities to distinguish sources of memory in other-generated activities, because children's competency as witnesses may depend upon such capabilities.

The section on social factors related to children's testimony focused on suggestibility.  Even preschoolers can give fairly accurate free recall of events following some forms of suggestive questioning.  However, when suggestions are repeated and forceful, preschoolers will readily reinterpret events to be consistent with the suggestions.  Additional research is needed to see whether strong suggestions can lead to distorted interpretations of events in older cohorts as well.  Although preschoolers' performance on free recall is fairly reliable, their recognition task performance has many errors resulting from suggestive questions.  By the age of 7 to 10 years, children appear to be no more vulnerable than are adults are suggestions.  Suggestibility effects result from an interaction of suggestive questioning with other factors, such as memory, social demand characteristics, duration and frequency of suggestions, and salience of the stimuli.  Thus, a child who observes a personally meaningful event, has strong recall abilities, and is tested in a low social demand situation would probably not be highly vulnerable to suggestive questioning.

However, other social factors, such as social motives, fear, and socialization, can influence children's testimony.  Given that many cases, such as custody or child-abuse cases, can pull children between competing loyalties, these other factors can have a strong influence on children's testimony.  Therefore, special attention should be given to recognizing these forces during interviews with children.

Two basic paradigms are used in child witness studies: (a) stimulus-delay-test, and (b) stimulus-suggestion-delay-test.  The former is used primarily to examine unbiased recall, the latter is used to examine the biasing effects of postevent information on recall.  The following procedural and methodological variations may explain the differential outcomes of reviewed studies that used essentially the same paradigm: (a) form (live versus recorded) and content (personally meaningful versus trivial) of stimulus, (b) method of testing memory (free recall versus cued recall and recognition), (c) method of testing for suggestibility effects (original versus modified), (d) length and number of exposures to stimuli, (e) delay between stimulus exposure and memory testing, and (f) number of memory tests and testers.  These factors, to varying degrees, influence children's memory and ability to retrieve information.  Researchers attempting to design ecologically valid child witness studies should carefully manipulate and examine the effects of all of these variables.  Researchers should also attend to individual differences in children's performances and attempt to model these differences.  In addition, there is a great need for replication of many of the extant studies, perhaps with variations in the context of the witnessed event and testing conditions (e.g., stressful vs. nonstressful conditions).

In order to adequately examine the issues discussed in this article, researchers must be sensitive to the real world contingencies affecting child witnesses.  Until that time, child witness research will have a limited impact on the treatment of children in the legal system and relevant child welfare arenas.  Conducting socially useful and scientifically sound research is, of course, no easy feat.  For child witness researchers, this may involve considering a new paradigm.  That paradigm should focus on individual differences rather than group (age) differences, in order to increase predictability of who will be most reliable and in what conditions.  Stimulus events should be repeated, extended in duration, ambiguous, and personally involving so that they reflect real world situations.  Perhaps multiple interrogators should be used, some familiar and others less familiar with the witness.  Finally, more behavioral observations should be included in research and less reliance on tests and children's self-report.  Such an approach to child witness research might provide practical information to those who must make decisions about children's testimony every day.


(1) Interestingly, Marin et al. (1979) also found that the amount of information correctly recalled was greater after a 30-minute delay in questioning than it was after a 10-minute delay.  [Back]

(2) Notably, all three groups in the Cohen and Harnick (1980) study performed better on the multiple-choice task than on the open-ended recognition task given a week prior.  The 9-year-olds were performing at chance levels (51% correct) in the open-ended recognition task, but during the multiple-choice task they correctly responded to an average of 76% of the items.  The performance of the two older groups was also much higher on the multiple-choice task, and actually showed a ceiling effect 8% to 93% correct, on average.  However, multiple-choice questions are not likely to he used in a real life situation, mainly because the interrogator typically would not know the "correct response" to put in an array of choices for a witness.  [Back]

(3) Some researchers, such as Dent and Stephenson (1979), have suggested that any form of questioning is suggestive regardless of whether there is misinformation or objective information. That is, by asking an individual a question about an event (e.g., "What color was the thief's hair?"), the interrogator is tipping off the witness that hair color is an important detail to remember.  The witness has, as it were, received information about what the interrogator is interested in, even though no misinformation was communicated.  The demand may lead the witness to provide an answer about hair color, even though he or she may have no recollection whatsoever about the thief's hair.  [Back]



l am grateful to Alison Clarke-Stewart, Gary Evans and Wendy Kliewer for their useful comments on this article.

* Stephen Lepore is an assistant professor in the Department of Psychology, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213.  [Back]

[Back to Volume 3, Number 2]  [Other Articles by this Author]

Copyright 1989-2014 by the Institute for Psychological Therapies.
This website last revised on April 15, 2014.
Found a non-working link?  Please notify the Webmaster.