by Stephen J. Ceci, Ph.D. and
Eduardus de Bruyn, M.A.
Children Today, U.S. Dept. of Health and Human Services, 1993,
Volume 22, No.1.
Stephen J. Ceci, Ph.D. is the Helen L. Carr Professor of Development
Psychology at Cornell University, Ithaca, N.Y. Eduardus de Bruyn, M.A., is a
doctoral student at Cornell. The first author has conducted research on
children's recollections for two decades, and has written over 100 articles,
chapters and books. He is the recipient of a Research Career Development Award
from N.I.H., and a senior Fullbright-Hayes Award. Currently, he serves as a
member of six editorial boards and is a fellow of both the American
Psychological Association and the American Psychological Society.
As a result of society's reaction to dramatic increases in reports of
abuse and neglect, children increasingly are being admitted as witnesses in
juvenile and criminal proceedings. Data from a 1989 survey of all 50 states
indicate there were 2.4 million reports of suspected children
maltreatment--i.e., 1 out of every 40 children below the age of 12. Of these,
almost 900,000 cases were "substantiated," and about a quarter of these were
sexual in nature. Although there are no reliable figures on the number of
children who end up participating in family court or criminal justice
proceedings, an extrapolation from some recent New York State data to the
entire nation suggests that this number could be in the vicinity of 100,000.
If one adds to this 100,000 figure the large number of non-abuse cases that
result in children participating in court proceedings (e.g., as witnesses to
domestic violence, road accidents, playground injuries), then their
participation in the legal system rises considerably. Thus, it has become
common to see young children providing testimony in a range of cases, from
custody disputes to felonious murders. And in the case of sexual abuse, it is
common to see children provide uncorroborated testimony.
From all that we know about jurors', judges' and attorneys' perceptions of
children's veracity and testimonial competence, however, there is cause for
concern. Young child witnesses appear to be disproportionately less
likely to be believed than older witnesses. Although crime statistics are not
broken down by age of complainant, surveys indicate that fewer preschoolers are
disproportionately more likely to be abused than older children, their
testimony may be less likely to be believed by juries. One implication of this
conclusion is that the trial testimonies of alleged preschool witnesses
resulted in fewer criminal convictions than for cases involving older child
In sum, America is faced with a problem: At the same time that more and more
children are becoming involved in the legal system, the latter has not yet
quite decided how much weight to give their testimony.
Some Past and Present Trends
In gainsaying children's testimony a valid response to their inferior
memories and greater suggestibility-proneness? Or have the credibility risks
of young witnesses been exaggerated by some legal participants, notably defense
attorneys and their hired experts? Answering these questions is not easy. The
task is made even more difficult by the extreme pronouncements of advocates
from the pro-defense and pro prosecution camps.
The scientific research on the suggestibility of children's recollections is
contradictory and confusing, even for researchers who work in the emotionally
charged and contentious area. Although the historical literature on children's
suggestibility, with only two exceptions, led to jaundiced portrayal of
children's suggestibility-proneness, the most recent studies are more equivocal
in their conclusions about children's testimonial competence. It is easy to
locate in the modern literature scores of studies claiming that young
children's testimonial competence is not substantially different from that of
adults, and it is also easy to find scores of studies claiming the opposite.
This has resulted in a confusing juxtaposition of newspaper headlines, such as
"Study shows children are credible as witnesses"' "Research shows child
witnesses easily coached."
A careful reading of the scientific literature indicates that there are
reliable age differences in suggestibility, with preschool aged children's
reports being somewhat more impeded by erroneous suggestions than older
children's reports. To be sure, three are many caveats that accompany this
assertion, lest it become as glib as the unqualified headlines. Elsewhere we
have provided an extensive analysis of the twentieth century research on
children's suggestibility in North America and Europe, but, due to space
constraints, we cannot recount that here. The interested reader is advised to
consult that source for evidentiary bases of the claims we shall make below,
and a complete description of the many necessary caveats.
During the 1980s all states dropped their corroboration requirement for
children involved in sex abuse cases, a crime that by its nature is private,
and often without corroborating evidence. As more children have been allowed
to give uncorroborated testimony, courts have turned to research to inform
their proceedings. And as children are increasingly admitted as witnesses in
court, criminal procedures have been modified. For example, most states have
evidentiary codes that permit "leading" a child sex abuse victim: "The Court
may, in the interest of justice, permit a leading question to be asked of a
child under the age of 10 in cases involving specified sex offenses"
(California Evidence Code, Sec. 767 (b). Shield laws (i.e. permitting a child
witness to testify behind a one-way screen that occludes the child's view of
the defendant but not the defendant's view of the child, or testifying over
one-way video) and hearsay exceptions (e.g. allowing therapists pediatricians,
and others, to describe what children said to them) have also been instituted
to assist child witnesses who otherwise might be psychologically "unavailable"
to testify in open court. In light of claims that many of these modifications
challenge the constitutional rights of defendants, it is important to obtain
evidence that such procedures do, in fact, facilitate the accuracy of
children's testimony. This is an important point to obtain evidence that such
procedures do, in fact, facilitate the accuracy of children's testimony. This
is an important point behind many recent court decisions.
Specifically, the court's decisions regarding the treatment of child witnesses
are predicated not solely or even primarily on humanity issues (i.e., reducing
the child witness' stress by permitting them to testify out of the defendant's
line of sight), but on the presumption that greater testimonial accuracy will
result from modifications in courtroom interviewing. Yet, without exception
this presumption has never been empirically satisfied. We simply do not have
any scientifically adequate data base to determine the testimonial implications
of testifying while shielded. It may prove to be the case that children's
testimony is more accurate when shielded, but it also might prove less
accurate. Several investigators are now studying this and we ought to be in a
better position to know the answers in a few years.
Having said this, there are some new studies that bear on related issues,
particularly the effect of prolonged erroneous suggestions on young children's
recollections and the influence of interviewers' biases on children's reports.
In what follows, we confine our discussion mainly to two studies that were
recently conducted in our laboratory at Cornell University, as they were
designed specifically to address these issues.
Two Recent Studies of Interviews
Misled Preschoolers. In one study, preschoolers were exposed to
a visit to their daycare center by a stranger named Sam Stone. Following Sam
Stone's visit, they were asked about its details four different times over a 10
week period. On these 4 occasions the interviewer refrained from using
suggestive questions. She simply encouraged children to describe Sam Stone's
visit in as much detail as possible.
One month following the fourth interview, the children were interviewed a fifth
by a new interviewer, who used forensic procedures (e.g., first acclimating the
child, then eliciting a free narrative, then using probes, urging the children
to say when they do not recall, taking breaks). This interviewer probed about
two non-events--that is, two behaviors that never happened during Sam
Stone's visit. These involved his doing something (unspecified) to a teddy
bear and a book. Sam Stone never touched either item, and there was no reason
to imagine that children would claim that he had done so, even when probed:
"Did Sam Stone do anything to a book or a teddy bear?"
As predicted, children resisted claiming these non-events occurred. Only 10%
of the 3-to 4-year olds' answers contained claims that Sam Stone did anything
to a book or teddy bear. And when asked if they actually saw him to do
anything to the book or teddy bear, as opposed to "thinking they saw him do
something, or hearing he did something," only 5% of their answers contained
claims that anything occurred. Finally, when these 5% of their answers
contained claims that anything occurred. Finally, when these 5% of children
who erred were gently challenged ("You didn't really see him do
anything to the book/the teddy bear, did you?") , only 1 of the 40 or .025% of
the answers was that these children supplied insisted they did. None of the
5-to 6-year-olds claimed to have actually seen Sam Stone do either thing. This
condition can be considered as a control against which we can assess the
effects of the use of repeated suggestive questioning, especially about
characters who are the object of children's stereotypes.
A second group of preschoolers was presented a stereotype about Sam Stone
before he ever visited their school. Each week, beginning a month prior to his
visit, the children were told a new Sam Stone story, in which he was depicted
as very clumsy. For example:
You'll never guess who visited me last night. (pause). That's right. Sam
Stone! And guess what he did this time? He asked to borrow my Barbie and when
he was carrying her down the stairs, he tripped and fell and broke her arm.
That Sam Stone is always getting into accidents and breaking things!
Following Sam Stone's visit, these children were treated identically to control
group children, i.e., they were interviewed four times, avoiding all
suggestions. Then they were given the same forensic interview by a new
interviewer, starting with free narrative, then proceeding to probes about
anything happening to the book/teddy bear.
The stereotyping had an effect, particularly for the 3-to 4-year-olds, 42% of
whom claimed Sam Stone ripped the book or soiled the teddy bear in response to
suggestive probes. (However, none of these false allegations were part of the
children's free narratives-all of them were in response to specific questions).
Of the youngest children, 19% claimed they actually saw Sam Stone do
these misdeeds. But after being challenged gently, only 11% continued to hold
to this claim. In contrast, 5-to 6-year-olds were significantly more resistant
to the influence of the stereotype, with roughly half the rate of errors that
the younger children exhibited.
Another group of preschoolers were assigned to a "suggestion condition" that
involved the provision of suggestive questions during the 4 interviews
following Sam Stone's visit. These children were not given any stereotype
about Sam Stone prior to his visit. Each suggestive interview contained 2
erroneous suggestions, one having to do with ripping a book and the other with
soiling a teddy bear. (e.g., "Remember that time Sam Stone visited your
classroom and spilled chocolate on that white teddy bear? Did he do it on
purpose or was it an accident?" "When Sam Stone ripped that book, was he being
silly or was he angry?")
Ten weeks later, when the forensic interviewer probed these children about
these events ("Did anything happen to a book?"), 52% of the 3- to 4- year-olds'
answers(31 out of 58) and 38% of the 5- to 6-year-olds' answers contained
claims that Sam Stone was responsible for one or both misdeeds. Thirty-five
percent of the youngest children's answers contained the claim that they
actually witnessed him do these things, as opposed to just being told that he
did them. Even after being gently challenged, 12% of their answers indicated
that they continued to claim they saw him do one or both misdeeds. Older
children were also susceptible to these suggestive interviews, though at a
reduced level: Only 2 out of 20 continued to maintain that they saw him do the
misdeeds when gently challenged.
Finally, 39 children were provided a pre-visit stereotype and
post-visit suggestive interviews. During the forensic interview, 72% of the 3-
to 4-year-olds claimed that Sam Stone did one or both misdeeds, a figure that
dropped to 44% when asked if they actually saw him do these things.
Importantly, 21% continued to insist that they saw him do these things, even
when gently challenged. Though better, the situation was still cause for
concern for the older preschoolers.
Some researchers have opined that the presence of perceptual details is
indicative of true memories, as opposed to confabulated reports. In this
study, however, perceptual details were no assurance that the report was
accurate. In fact, it was surprising to see the number of false perceptual
details children in the stereotype and suggestion condition provided to
embellish the non-events(e.g. claiming that Sam Stone took the teddy bear into
a bathroom and soaked it in hot water before smearing it with a crayon). So
strikingly believable were their reports that we presented videotapes of these
interviews to researchers and clinicians who work in the area of children's
testimonial competence to see if they could discriminate erroneous reports from
We did this at two recent conferences (the American Psychology/Law Society
Biennial Meeting, San Diego, CA 3/15/92; and a NATO Advanced Study Institute,
Italy, 5/19/92). The results were the same at both conferences: The majority
of the audience got it wrong--very wrong. Experts who do research on
children's testimonial competence, provide therapy to children suspected of
having been abused, and conduct law enforcement interviews of child victims,
all failed to detect which children which were accurate and which were not,
despite being confident in their opinions. The children's reports would fool
anyone who thinks that it is easy to detect a young child's false report,
contrary to claims from some quarters. The reason is that, unlike the typical
study in which a child is presented a single erroneous suggestion, these
children received persistent and intense suggestions that built on a prior set
of expectations (i.e., stereotype).
We believe that this ingredient is more similar to what transpires in some
actual cases; it is common for child witnesses to be interviewed many times
prior to being given a formal, videotaped interview with anatomical dolls, or
to testifying in court. This is the first research to examine over long
periods of time the effect of persistent, erroneous suggestions that are
consonant with children's expectations.
Thus, the procedures we employed occur, albeit in altered form, in actual
therapy sessions and law enforcement/CPS interviews. We patterned our
experimental manipulations after materials that we have collected over the past
decade from court transcripts and therapeutic interviews.
Misled Interviewers. In this study we were interested in whether
interviewers attempted to falsify their hypotheses, in conjunction with
Popper's well-known prescription, or whether they exhibited a confirmatory
bias. There is some evidence that interviewers' hypotheses can have
detrimental effects on the outcome of the interview, but no direct evidence as
to how this happens.
Preschoolers were exposed to a gamelike event, and then interviewed one month
later. The interviewer, and experienced social worker, was given a one-page
report that contained event that the interviewer was told might have
occurred. The interviewer was asked to examine the child to determine how much
of the information they could, in fact, still recall. The only instruction
given to the interviewer was that she should begin by asking the child for a
free narrative, avoiding all forms of suggestions and leading questions.
Following this, the interviewer was instructed to use whatever props (e.g.
dolls) or strategies she felt were necessary to elicit the most factually
accurate recall from the child.
The one-page report given to the interviewer contained three types of
information about the event that the children allegedly had experienced: a)
accurate information, b) erroneous information, and c) neutral (control)
information. For example, if the game entailed one child touching another
child's nose and the latter child rubbing her own tummy, the interviewer might
be told that the first child touched the toe of the other child, who in turn
rubbed her own tummy (i.e., some information was accurate, some was not).
The information we provided influenced the interviewer's hypotheses, and
appeared to exercise an influence on the dynamics of the interview, with the
interviewer sometimes shaping children's reports to be consistent with her
hypothesis about what happened. When the interviewer was accurately informed,
she got children to correctly recall between 70% and 100% of such events.
However, when she was misinformed, 34% of the 3- to 4- year-olds corroborated
one or more events that the interviewer falsely believed had transpired. This
finding would seem to have relevance for forensic interviewers, and possibly
even for sexual abuse interviews, though the latter context is so unlike the
one we employed that we are reluctant to press this generalization.
Finally, the children seemingly became more credible as the interview unfolded.
Many children initially stated details inconsistently, or with reluctance or
even denial, but as the interviewer persisted in asking about non-events, some
children abandoned their denials and hesitancy. A month later, we asked a
different interviewer to follow up the first interview. She had access to the
notes of the first interviewer, and appeared to base her hypotheses on what
that interviewer had "found." This resulted in some additional errors and
increased confidence on the child's part.
An interview should be a form of scientifically-guided inquiry, because science
is the most powerful way of knowing. Like any scientific endeavor, a forensic
interview aims at ascertaining the truth, and it ought to do so in the same
manner that science attempts to determine the truth. But what exactly does
this mean, and why is it important to state this assumption explicitly?
At its core, science is "proof by disproof," meaning that a common hallmark of
all scientific endeavors is that scientists attempt to disprove a hypothesis by
giving alternative hypotheses a fair chance of confirmation while
simultaneously giving their pet hypothesis a fair chance at being falsified.
Although not all scientists attempt to test alternative hypotheses all the
time, they nevertheless acknowledge the benefit of doing so, and recognize that
the occasional lapse into a confirmatory bias may result in weaker science.
Interviewers, like scientists, should also attempt to rule out rival hypotheses
instead of exclusively striving to confirm their favored hypothesis. Because
of the needs of child protective service interviewers, however, it is not
feasible or even desirable to insist that they generate and test all
conceivable hypotheses or, conversely, be "blinded" from all relevant
information that pertains to the State's main hypothesis. Doing so could
result in missed opportunities whenever an interviewer does not recognize the
relevance of a given piece of information. But, as we demonstrated in this
study, the failure to test arrival hypothesis can result in various types of
How do CPS workers, clinicians, and law enforcement professionals interview
children? Do they attempt to disprove alternative hypotheses? Do they
generate hypotheses that rival their "pet" hypothesis, or do they operate
primarily within a confirmatory mode? Does it really matter whether they
attempt to disprove pet hypotheses?
We do not know the answers to the above questions, at least not in any adequate
manner. But there is some indication that interviewers do not follow training
instructions, and that they frequently pursue a single hypothesis
single-mindedly. For example, one researcher trained interviewers to avoid
leading questions. Then they provided the interviewers with accurate and
inaccurate details about an event at the children's school. The researchers
discovered that despite instructions, 34% of interviewers' questions could be
categorized as leading; most importantly for the purposes of this article, it
appeared as if they pursued a single hypothesis. Interviewers who were misled
about the event elicited significantly more errors from children than did
non-misled interviewers. So children often ended up corroborating
interviewers' erroneous hypotheses.
In closing, we wish to point out a difference between the conditions in the two
studies just described and those that occur in actual social work, therapy, and
law enforcement interviews. The latter events are seldom as sanitized of
affect and absent of motives as those we used here. They usually entail high
levels of stress, assaults to one's own body, and loss of control. Neither of
these studies contained those ingredients. So, before anyone leaps to the
conclusion that children or interviewers will behave in exactly the same way
when the critical event involves bodily abuse, extra factors need to be
considered. We are still far from being able to provide a definitive
conclusion other than this : Regrettably, pronouncements by extremist
advocates on both side of the bench are simply unfounded. Children are neither
as hyper suggestible and coachable as some pro-defense advocates have alleged,
not are they as resistance to suggestions about their own bodies as some
pro-prosecution advocates have claimed.
They can be led to incorporate false suggestions into their accounts of
intimate bodily touching if these suggestions are made by powerful adult
authority figures, delivered over prolonged periods. On the other hand, they
can be amazingly resistant to false suggestions, and able to provide highly
detailed and accurate reports of events that transpired weeks or months ago.
This underscores the need for great care in accepting the claims of those who
are eager to put pro-defense or pro-prosecution "spins" on children's