A Thesis
entitled
Children’s Memory for a Dyadic Conversation after a One-Week or a Three-Week Delay
by
Monica L. Rohrabaugh
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the
Master of Arts Degree in Psychology
_________________________________________
Kamala London, PhD, Committee Chair
_________________________________________
Stephen Christman, PhD, Committee Member
_________________________________________
A. John McSweeny, JD, PhD, Committee Member
_________________________________________
Patricia R. Komuniecki, PhD, Dean
College of Graduate Studies
The University of Toledo
December 2014
Copyright 2014, Monica Lynn Rohrabaugh
This document is copyrighted material. Under copyright law, no parts of this document
may be reproduced without the expressed permission of the author.
An Abstract of
Children’s Memory for a Dyadic Conversation after a One-Week or a Three-Week Delay
by
Monica L. Rohrabaugh
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the
Master of Arts Degree in
Psychology
The University of Toledo
December 2014
A growing body of literature demonstrates that child witnesses are frequently asked to
recall conversations during forensic investigations. The present study examined
children’s (n = 90) ability to recall a dyadic conversation after a one-week or three-week
delay. Children were questioned about the target conversation using free recall and
recognition style questioning during the memory test. Children’s overall accuracy and
characteristics of their memory reports were examined. Children in the one-week delay
condition accurately reported significantly more of the conversation than children in the
three-week delay condition, but free recall of the conversation was low for children in
both conditions. The majority of what children recalled from the conversation was
accurate. Additionally, when asked to recall the conversation in its entirety, children had
a strong tendency to recall what they said from the target conversation and rarely recalled
utterances said by their conversational partner. Memory for self-generated and partnergenerated utterances did not differ during recognition testing. Memory for the structure
of the conversation was low in both delay conditions and across all question types.
Forensic implications and future directions are discussed.
iii
Table of Contents
Abstract ………………………………………………………………………………
iii
Table of Contents …………………………………………………………………….
iv
List of Tables ………………………………………………………………………… vii
List of Figures ……………………………………………………………………….
viii
I. Introduction ……………………………………………………………………….
1
1.1 Hearsay Testimony ……………………………………………………..
7
1.2 Conversation Memory and Memory for Events ………………………..
8
1.3 Existing Memory for Conversation Literature ………………………….
10
1.3.1 Characteristics of Conversational Memory ………………….
12
1.4 Theoretical Considerations ……………………………………………..
16
1.5 Summary ………………………………………………………………..
18
II. The Present Study ……………………………………………………………….
20
2.1 Hypotheses ……...……………………………………………………..
20
2.1.1 Conversational Memory across Delays ………………………
20
2.1.2 Characteristics of Recall ……………………………………..
20
2.1.3 Accuracy of Reports by Question Format ……………………
21
2.2 Methodology ……………………………………………………………
21
2.2.1 Participants …………………………………………………...
21
2.2.2 Materials ………………………………………………………
22
2.2.3 Procedures …………………………………………………….
22
2.2.3.1 Session I: The Target Conversation ………………...
22
2.2.3.2 Session II: The Memory Test ……………………….
24
iv
2.2.3.3 Creation of the Recognition Test …………………… 25
III. Results …………………………………………………………………………..
28
3.1 Data Coding …………………………………………………………….
28
3.1.1 Coding of the Target Conversation …………………………..
28
3.1.2 Coding of the Memory Test ………………………………….
28
3.1.3 Inter-rater Reliability …………………………………………
30
3.1.4 Recognition Test Scoring …………………………………….
31
3.2 Characteristics of the Target Conversation …………………………….
31
3.3 Memory Test: Overall Recall of the Target Conversation ……………..
33
3.4 Memory Test: Accuracy of Memory Reports during Free Recall ……...
37
3.5 Memory Test: Characteristics of Memory Reports during Free Recall ...
39
3.6 Memory Test: Recall of the Storyteller’s Utterances and Verbal Exchanges
in Cued Recall …………………………………………………………
40
3.7 Memory Test: Recognition Test Performance ………………………….
43
3.8 Memory Test: Accuracy of Recall by Question Format ………………..
46
IV. Discussion ……………………………………………………………………..
47
4.1 Conversation Memory across Delays …………………………………...
47
4.2 Characteristics of Recall ………………………………………………..
49
4.3 Accuracy of Reports by Question Format ………………………………
51
4.4 Forensic Implications ……………………………………………………
52
4.5 Limitations and Future Directions ………………………………………. 54
References …………………………………………………………………………..
59
Appendices …………………………………………………………………………
66
v
A. Pre-Recorded Story Script ………………………………………………………
66
B. Target Conversation Script ………………………………………………………
67
C. Memory Test Script ……………………………………………………………...
71
D. Example Recognition Test ……………………………………………………….
75
vi
List of Tables
Table 1
Outline of Procedures for Session I and Session II. ........................................27
Table 2
Means and Standard Deviations of Recognition Test Performance by
Question Type and Delay Condition. ..............................................................45
Table 3
Means and Standard Deviations of Report Accuracy by Question Format
and Delay Condition. ......................................................................................46
vii
List of Figures
Figure 1
Interaction of Recognition Test Performance by Question Type and Delay
Condition. ........................................................................................................45
viii
Chapter One
Introduction
Witnesses are often asked to recall prior conversations during criminal and civil
court proceedings. Testimony regarding prior conversations can form the foundation of a
trial. For example, in United States v. Libby (2006), Vice-Presidential Chief of Staff, I.
Lewis “Scooter” Libby was accused of leaking the identity of a Central Intelligence
Agent officer and exposing confidential information to reporters and other government
agents. Various witnesses testified on the conversations, but Libby allegedly failed to
remember the conversations and defended that they did not occur. The jury believed that
Libby could have forgotten the specific content of the conversations, but did not believe
that Libby completely forgot that the seminal conversations occurred. Libby was found
guilty of obstruction of justice, making false statements, and perjury (Kassam, Gilbert,
Swencionis, & Wilson, 2009).
Conversational testimony also played a key role in the Watergate investigation
during the Nixon administration. John Dean, former counsel to President Nixon,
provided testimony regarding conversations he had with President Nixon during the days
leading up to the Watergate scandal. While Dean acknowledged that his recollection of
the conversations were not verbatim, he insisted that his testimony accurately depicted
the gist of the conversations. After Dean provided testimony, the Senate Committee
discovered that all of the conversations in the Oval Office were secretly recorded. An
analysis of Dean’s testimony revealed that Dean accurately reported the gist of the
conversations, but did not have verbatim memory for the conversations (Neisser, 1981).
1
John Dean’s testimony indicates that the gist of a conversation can be remembered when
the verbatim content of the conversation is forgotten.
Libby and Dean’s testimony demonstrates that conversational memory often lacks
verbatim-like recollection. However, conversational testimony remains common in civil
and criminal trials. Conversational testimony includes testimony regarding what was said
in prior conversations, who said a specific utterance, as well as identifying the receiver of
the utterance (Davis, Kemmelmeier, & Follette, 2005). Witnesses providing
conversational testimony are often requested to identify others who could have been
present during the conversation as well as provide information regarding the setting of
the particular conversation. This form of testimony is nearly unavoidable during criminal
or civil investigations and later during trial. Duke, Lee, and Pager (2007) noted that the
frequency of conversational testimony is not surprising due to the fact-finding nature of
investigations, as testimony regarding prior conversations allows witnesses to provide
insight regarding the perpetrators’ and/or victims’ mental state when the event occurred.
However, researchers have largely neglected to systematically evaluate the
reliability of conversational testimony provided by witnesses (Davis & Friedman, 2007;
Davis et al., 2005; Duke et al., 2007). Davis and Friedman (2007) coined conversational
memory as the “orphan child” of the eyewitness literature and recent calls for research
evaluating conversational memory have been made (Lyon & Stolzenberg, 2014). A
special case of importance in evaluating the reliability of conversational testimony is
when child witnesses are asked to provide reports of prior conversations. Approximately
3.5 million investigations are completed annually in response to child maltreatment
allegation in the United States (Malloy, La Rooy, Lamb, & Katz, 2011). In 2012, 62,936
2
child sexual abuse (CSA) allegations were substantiated in the United States (U.S.
Department of Health and Human Services, 2013). Child sexual abuse cases are
particularly interesting because they typically lack corroborative evidence and the child is
usually the only witness for the alleged abuse (Malloy et al., 2011). Testimony provided
by child witnesses in CSA cases often acts as pivotal evidence that abuse did or did not
occur.
The reliability of testimony provided by child witnesses is of the upmost
importance in CSA cases. Researchers have evaluated methods that bolster and/or
impede child witnesses’ ability to provide accurate and full event reports. A central
theme of this research is that conversations, which introduce elements of misinformation,
suggestive questioning, stereotypes, and reinforcement, can distort the reliability of
children’s event reports (for a review see, Ceci & Bruck, 1993). This research suggests
that discussions with child witnesses regarding the allegation can potentially alter their
memory reports for the alleged event.
Suspected victims of CSA typically engage in multiple conversations regarding
the allegation. During the forensic investigation, child witnesses usually endure multiple
formal (e.g., forensic interviews, social workers, lawyers) and informal (e.g. parents,
peers) interviews (Quas & Sumaroka, 2011). Forensic interviewing protocols have been
developed as a means to extract the most reliable testimony from child witnesses by
avoiding harmful interviewing techniques. For example, forensic interviewers can be
trained to follow the National Institute of Child Health and Human Development
(NICHD) protocol; a highly scripted and empirically supported forensic interviewing
protocol specifically for child witnesses (Lamb, Sternberg, Orbach, Hershowitz,
3
Horowitz, & Esplin, 2002). Empirical evidence shows that forensic interviews following
the NICHD protocol are high in interview quality and information obtained from the
child witness, largely by reducing the amount of harmful questioning techniques
employed (Lamb, Orbach, Hershkowitz, Esplin, & Horowitz, 2007). However, no safe
guard exists to protect child witnesses from exposure to harmful questioning techniques
during informal interviews.
Recent work by Principe and colleagues demonstrated that naturally occurring
conversations with peers (Principe, Cherson, DiPuppo, & Schindewolf, 2012) and
mothers (Principe, DiPuppo, & Gammel, 2013) can distort children’s event reports. For
example, children who did not experience an event, but discussed the event with a peer
who did, falsely reported experiencing details from the events that they only heard.
Likewise, children incorporated misinformation that they only talked about in
conversation with their mother as event details that actually happened during their
memory report for the event. Principe and colleagues found that children who heard, but
did not experience an event tended to have richer memory reports for the event compared
to children who actually participated. This finding was especially strong for older
children. These studies show that information discussed during conversations regarding
an event are often integrated into children’s event reports (for a review, see Principe &
Schindewolf, 2012).
Alleged victims of CSA report discussing the alleged abuse with multiple
disclosure recipients (Malloy, Brubacher, & Lamb, 2013). The most common disclosure
recipients reported by child witnesses in Malloy et al. (2013) were mothers, peers, and
other family members. These previous conversations run the risk of contaminating the
4
reliability of child witnesses’ reports if they include aspects of misinformation or employ
harmful questioning techniques. Information about these conversations can provide a
context in which to interpret the veracity of testimony provided by child witnesses.
As such, child witnesses are often asked to recount prior conversations during
forensic interviews. The main goals of forensic interviewing is to find out what
happened according to the child’s perspective as well as test alternative hypotheses
regarding the unfolding of the allegation (Lamb, Sternberg, & Esplin, 1998). A common
alternative hypothesis to explore is that the child’s testimony was influenced by prior
conversations. Malloy et al. (2013) found that 82% of forensic interviews of alleged
CSA victims following the NICHD protocol included a “disclosure phase” in which
children were asked to report prior disclosure conversations. Additionally, forensic
interviewers often ask child witnesses if the suspected perpetrator said anything to them
(e.g., threats, promises for secrecy) in order to evaluate the dynamic of the alleged abuse
(Lyon & Stolzenberg, 2014).
Defense and prosecuting attorneys often ask child witnesses to recall
conversations during child sexual abuse trials (Stolzenberg & Lyon, 2014). Prosecutors
are motivated to display that the child delayed disclosure due to threats or request for
secrecy from the suspected perpetrator. Defense attorney are motived to display the
potential of misinformation or suggestibility effects transmitted to the child during
conversations regarding the alleged event. In an examination of CSA trial transcripts,
Stolzenberg and Lyon (2014) found that child witnesses were asked 47 questions
regarding conversations while providing testimony, on average. Further analysis of these
questions showed that child witnesses are most often questioned about their disclosure
5
and child-suspect conversations. Questions regarding prior conversations are often
linguistically complex and require the child witness to discuss specific elements of a
particular conversation (e.g., “Did you tell [your father] when he came into your room at
Thanksgiving, that the suspect dragged you out of your room?”) (Stolzenberg & Lyon,
2014, p. 24). Currently, it is unknown if child witnesses understand complex questions
regarding conversations (Lyon & Stolzenberg, 2014).
Furthermore, the extent to which children can accurately recall conversations
remains unknown. No previous studies have evaluated children’s memory for dyadic
conversations (i.e., a conversation between two people) as well as the accuracy of their
reports after a delay. Child maltreatment investigations are typically long in duration.
Victims of CSA often delay disclosure (London, Bruck, Ceci, & Shuman, 2005) and
child maltreatment investigations (not including trials) are typically completed 30- to 60days after child protective services receive a report (Child Welfare Information Gateway,
2013). Evaluating children’s ability to recall conversations after delay is imperative in
discussing the reliability of child witnesses’ conversational testimony.
In sum, conversational testimony plays an important role in cases involving child
witnesses. Child witnesses are often asked to recall conversations during forensic
interviews and the following trial. Generally, questions asked to child witnesses about
conversations focus on disclosure and child-suspect conversations (Stolzenberg & Lyon,
2014). Child witnesses’ ability to provide accurate conversational testimony relies on
their memory for the conversation in question. Children’s memory for conversation has
received very little empirical attention. The current study aimed to address this issue by
6
evaluating children’s memory for a dyadic conversation after a one-week or three-week
delay.
1.1
Hearsay Testimony
Conversational testimony is often admissible in court. Some forms of
conversational testimony are inadmissible under hearsay conventions, but it is a
misconception that all testimony regarding outside conversations constitutes as hearsay
testimony. Hearsay refers to testimony of an out-of-court statement that is “offered into
evidence to prove the truth of the matter asserted” (U.S. House, 2010, p. 16). Testimony
regarding out-of-court statements is often inadmissible under hearsay conventions when
the declarant of the out-of-court statement is unable to undergo cross-examination as the
reliability of the testimony cannot be challenged by opposing counsel (Davis &
Friedman, 2007). Out-of-court statements provided as testimony by the declarant of the
out-of-court utterance is also considered hearsay when the testimony is “to prove the
truth of the matter asserted” (U.S. House, 2010, p. 16). Out-of-court statements that are
not meant “to prove the truth of the matter asserted” (U.S House, 2010, p.16) are not
considered hearsay are admissible in court.
Furthermore, there are over 30 exceptions to the hearsay rule (Myers, 2011).
First, out-of-court statements with probative value (i.e., information that is useful for the
jury during decision making but not meant to prove guilt) are considered admissible as
testimony (Davis & Friedman, 2007). Additionally, Myers (2011) discussed three
hearsay exceptions that are important in child maltreatment cases. The excited utterance
exception to the hearsay rule admits testimony regarding out-of-court statements
produced while the declarant of the utterance was under extreme emotional distress as a
7
reaction to a traumatic event. Additionally, various professionals who provided medical
assessments or treatment services for child witnesses can provide hearsay testimony
regarding various symptoms that the child was experiencing (e.g., pain, stress). Lastly,
the residual hearsay exception acts as a catchall exemption to the hearsay rule, which
permits out-of-court statements as testimony if the statement is considered reliable. In
CSA allegations, out-of-court statements are an imperative form of evidence and are
often considered admissible under various hearsay exceptions (Lyon & Stolzenberg,
2014; Myers, 2011). Therefore, in most case, conversation testimony provided by child
witnesses will be admissible in court (Lyon & Stolzenberg, 2014).
1.2
Conversation Memory and Memory for Events
Children’s conversation memory needs to be systematically evaluated
independently from children’s memory for events (Lyon & Stolzenberg, 2014). The
existing child eyewitness memory literature emphasizes children’s reliability in providing
testimony regarding what they saw or did during an event rather than what they heard
during an event. No studies have systematically compared the reliability of children’s
reports for events versus conversations.
There is reason to believe that conversational memory and event memory are two
distinct entities. First, event memory and conversational memory utilize different
perceptual modalities in the earliest stages of memory acquisition. Visual stimuli are
processed by iconic memory (Sperling, 1960) and auditory stimuli are processed by
echoic memory (Sams, Hari, Rif, & Knuutila, 1993). Iconic and echoic sensory
memories retain incoming stimuli prior to further cognitive processing. Auditory and
8
visual stimuli are perceived and momentarily retained by different systems during the
earliest stages of memory acquisition.
Additionally, remembering a conversation may be especially difficult due to the
fluidity, complexity, and frequency of conversations. Based on this assumption, Duke et
al. (2007) argued that conversational memory is more susceptible to errors than event
memory during encoding, storage, and retrieval. For example, conversations can easily
be interpreted incorrectly due to the variety of semantics and pragmatics associated with
a particular conversation (Davis & Friedman, 2007; Duke et al., 2007). Inaccurate
interpretations can lead to erroneous encoding. Additionally, conversational memory
traces might deteriorate faster than event memory traces, as decay and interference
effects are stronger for complex stimuli (Duke et al., 2007). Lastly, source misattribution
errors might occur more often while retrieving conversational memory versus event
memory (Davis & Friedman, 2007). We can expect that children’s’ ability to accurately
recall prior conversations differs from their ability to recall events.
Furthermore, children’s memory for conversations might be particularly poor due
to their developing language and communicative abilities. Conversations with children
are often incoherent, as children tend to abruptly switch topics and make statements that
are irrelevant to the conversation (Poole & Lamb, 1998). In these ways, conversations
can be extremely complicated and ambiguous to children. We can expect that children’s
memory for conversations will be poor, as ambiguous events (versus non-ambiguous
events) tend to produce weak memory traces (La Rooy, Malloy, & Lamb, 2011).
Taken together, the cognitive demands in providing memory reports for prior
conversations cannot be assumed as equivalent to the cognitive processes involved in
9
remembering experienced events. The reliability of conversational testimony needs to be
evaluated as a separate entity from event testimony. As the study of conversational
testimony is neglected by the current eyewitness literature, the small literature on
memory for conversations will be discussed to provide insight regarding the reliability of
conversational testimony.
1.3
Existing Memory for Conversation Literature
The systematic evaluation of conversational testimony has been largely neglected
(Davis & Friedman, 2007). However, there are a handful of studies that evaluated
conversational memory in adults. No studies have evaluated children’s memory for
dyadic conversation. Empirically evaluating children’s memory for dyadic conversations
is forensically important as child witnesses are often asked to recount conversations that
occurred between themselves and another conversational partner (e.g., suspect, mother,
peers) (Stolzenberg & Lyon, 2014). Hedelin and Hjelmquist (1998) inadvertently
evaluated children’s ability to immediate recall one sentence in their study concerning
children’s distinction of sentence form. Children in Hedlein and Hjelmquist (1998)
engaged in an individual play session with their schoolteacher. The teacher asked the
child to relay a message, for example, “The hippopotami are hungry” to another teacher
who was coming into the classroom to replace the current teacher. Once the child told
the new teacher the message, the new teacher immediately asked the child to identify
what they just told them. The new teacher asked the child either a verbatim repetition
(e.g., “Did you say the hippopotami are hungry?”), a true paraphrase (e.g., “Did you say
give the hippopotami food?”), or a false sentence (e.g., “Did you say the hippopotami are
big?” and “Did you say the cows are hungry?” (Hedelin & Hjelmquist, 1998, p. 427).
10
This cycle of teacher-new-teacher message carrying occurred four times for each child
with various statements. Eighty percent of the children recalled the verbatim statement at
some point, either they said the verbatim statement and/or correctly identified the
verbatim prompt from the new teacher. This study indicates that most children are able
to immediately recall a verbatim sentence, but the findings are not generalizable to
children’s memory for a rich dyadic conversation. The adult memory for conversation
literature will be discussed to review the quantity of conversational utterances that are
accurately recalled from dyadic conversations.
Undergraduate participants in Stafford and Daly (1984) engaged in a sevenminute dyadic conversation concerning a topic of their choice with a stranger.
Participants watched a five-minute distractor video after the conversation and then
completed a free recall memory test for the conversation. Recalled statements were
considered accurate if they contained at least the gist of a statement from the original
conversation. On average, participants accurately recalled approximately 10% of the
conversation after a five-minute delay. This approximation is widely cited by researchers
interested in memory for conversations (Davis & Friedman, 2007; Davis et al., 2005;
Duke et al., 2007). The highest performing participant accurately recalled 40% of the
conversation after the five-minute delay. Accurate recall of the target conversation
declined to four percent a month after the target conversation occurred (Stafford,
Burggraf, & Sharkey, 1987). Stafford et al. (1987) concluded that dissolving memory
traces for the target conversations were filled with inferences, elaborations, and false
utterances after the extended delay. These results indicate that memory for conversation
drastically declines after a short delay and continues to decay over time.
11
1.3.1
Characteristics of Conversational Memory
Understanding the quality of children’s conversational memory reports is
forensically relevant. Child witnesses are often asked to report what they told during
disclosure conversations as well as what others said to them regarding the allegation
(e.g., suspect statements) (Stolzenberg & Lyon, 2014). Evaluating children’s memory for
statements they said (i.e. self-generated) and statements their conversational partner said
(i.e., partner-generated) has forensic importance.
Researchers have put forth two different effects that make opposite predictions
regarding memory for self-generated versus partner-generated utterances. The
generation effect posed by Miller, Winstanely, and Carey (1996) describes a phenomenon
in which memory is more robust for utterances that were self-generated than utterances
that were stated by the conversational partner. The idea behind the generation effect is
that self-generated utterances are remembered more than partner-generated utterances
due to additional cognitive processing involved in producing utterances during a
conversation (Miller et al., 1996).
On the other hand, the familiarity effect posited by Stafford and Daly (1984)
makes the prediction that participants should recall more partner-generated than selfgenerated utterances from a dyadic conversation. The rationale for the familiarity effect
holds that participants are highly familiar with self-generated utterances and lack the
necessity to remember their own contributions to conversations. Therefore, declarants
allocate their cognitive capacity towards remembering partner-generated utterances rather
than their own contributions to the conversation (Stafford & Daly, 1984).
12
There is empirical support for both the generation effect and familiarity effect in
the adult memory for conversation literature (Miller et al., 1996; Stafford & Daly, 1984).
However, these are the only studies that have tested memory for self-generated versus
partner-generated utterances. Children often have greater memory for self-generated
actions versus partner-generated actions during events (Baker-Ward, Hess, & Flannagan,
1990). However, it is unknown if children will display greater memory for self-generated
utterances versus partner-generated utterances from a dyadic conversation.
Furthermore, forensic interviewers and attorneys are often interested in how the
child’s statements were brought about during a conversation. Reporting both the eliciting
utterance and the child’s response is needed to form an accurate interpretation of the
conversational context in which the child’s statements were produced. Knowledge of the
conversational context is forensically imperative, “Conversations are fluid interactions
between two or more participants in which the crucial units include not individual
utterances, but the verbal exchanges in which these utterances are couched” (Bruck, Ceci,
and Francoeur, 1999, p. 91). Ideal conversational testimony provides an accurate account
of verbal exchanges by describing the eliciting utterance as well as the child’s response.
An important aspect of conversational testimony involving child witnesses
regards knowing what the child said and how their statements were elicited during the
conversation. For example, research has shown that children’s statements elicited via
open-ended questioning are typically more accurate than statements elicited by forcechoice questioning (Lamb et al., 2007). However, repeated open-ended prompts such as
“tell me what happened” can lead children to provide inaccurate information regarding an
event (Lamb et al., 2007). Knowledge regarding what the child said as well as how their
13
statements were elicited can provide insight to the reliability of child witnesses’
testimony.
Few studies have evaluated adults’ memory for verbal exchanges from a dyadic
conversation. Bruck and colleagues (1999) evaluated mother’s memory for a
conversation they had with their child regarding activates that occurred during a play
session. Mothers’ memory for the conversation was tested three or four days after the
conversation occurred. During a free recall test, mothers accurately recalled
approximately five percent of the total utterances said during the conversation. Bruck et
al. (1999) concluded that this finding is a conservative estimate of mothers’ memory for
the conversation as aspects of the conversation that are not likely to be recalled were
included in this analysis. For example, empty utterances (e.g., “okay” “good”) were
included in the transcripts used to judge the mothers’ reports. Additionally, mothers
typically rephrased and re-asked the same question during the conversation, but only
recalled asking repeated questions once during the memory test. Furthermore,
approximately 25% of the conversations were off-topic and were not expected to be
recalled, but were included in this initial estimate of mothers’ memory. In a second
analysis, Bruck et al. (1999) found that mothers accurately recalled approximately 35%
of the on-topic details (as opposed to utterances) from the original conversation. Details
were classified as information about the play activity that the mother learned about
during the conversation. Details were not synonymous with utterances as one detail
could contain multiple utterances.
After free recall testing, mothers were presented with a transcript that included
altered verbal exchanges from the original conversation among other prompts. Mothers
14
were instructed to correct errors that appeared in the altered transcripts. Promptedspontaneous alterations changed the way in which the child’s statements were elicited in
the original conversation. The following is a passage from a mother-child conversation
in Bruck et al. (1999):
Mom: What did you do? Tell me one of what you did.
Child: I know some of what I did.
Mom: What was it?
Child: Making a horsy drawing.
Mom: A horsy?
Child: Ya.
Mom: You made a horsy?
Child: Ya. We drawed a horsy.
Mom: You drew it? What did you draw it with?
Child: My crayon.
Mom: But were they yours or were they Lynne’s?
Child: Lynne.
Mom: They were Lynne’s crayons. Were they colored?
Child: They have sparkles on them. (p. 95)
The following is an example of a prompt-spontaneous alteration of this passage:
Mom: What did you do? Tell me one of what you did.
Child: We drawed a horsy. With Lynne’s crayon. One with sparkles. (p.95)
In this example, the child’s statements regarding their drawing appeared to come
spontaneously from the child, but in fact the child’s statements were highly prompted
during the conversation. Additionally, prompted-spontaneous alterations included cases
in which the child’s spontaneous utterances in the original conversation were altered to
appear highly prompted in the testing transcript. Mothers correctly identified
approximately 31% of the prompted-spontaneous alterations in the transcripts.
Approximately 23% of the detected prompted-spontaneous alterations were accurately
edited to match verbal exchanges from the original conversation. Although mothers
displayed some ability to recall verbal exchanges from conversations with their child,
15
Bruck et al. (1999) concluded that mothers displayed poor memory for verbal exchanges
and poor memory for the conversation overall.
Similarly, professional interviewers have a hard time immediately reporting the
components of an interview with child witnesses. Professional interviewers in Lamb,
Orbach, Sternberg, Hershkowitz, & Horowitz (2000) were instructed to take
contemporaneous notes while interviewing alleged CSA victims. The majority (57%) of
the statements said during the interviews were not reported in the interviewers’ notes.
Furthermore, interviewers significantly underreported the amount of questions they asked
the child witness during the interview. Less than half of the child witnesses’ statements
were recorded with the correct eliciting question. These results indicate that even socalled contemporaneous notes from professional interviewers are not verbatim and do not
reflect the interview in its entirety.
In sum, memory for verbal exchanges is often incomplete. Children’s memory
for verbal exchanges is expected to be poor as mothers (Bruck et al., 1999) and
professional interviewers (Lamb et al., 2000) have poor memory for verbal exchanges.
However, examining children’s ability to report verbal exchanges allows legal personnel
to better understand children’s limitations in recalling verbal exchanges from prior
conversations.
1.4 Theoretical Considerations
The Fuzzy-Trace Theory (FTT) (Brainerd & Reyna, 2004) provides a theoretical
explanation for the high frequency of gist conversational recall as opposed to verbatim
recall of conversations. According to the FTT, episodic memories are independently
encoded into surface content (i.e., verbatim traces) and semantic content (i.e., gist traces).
16
Verbatim traces and gist traces are independently stored, encoded, and retrieved.
Verbatim traces retain the actual experience whereas gist traces represent the
understanding of an experience. Generally, verbatim traces decay faster than gist traces,
which causes greater dependency on gist memory traces across delays. Fuzzy-trace
theory suggests that adult’s verbatim memory for conversation is poor due to rapidly
decaying verbatim memory traces and increased dependence on gist traces.
The verbatim-gist retrieval dissociation principle of FTT holds that various
retrieval cues activate verbatim and gist traces. According to this principle of FTT,
verbatim recall of conversations is possible if cued to retrieve verbatim traces. This
principle of FTT has been confirmed in studies of children’s memory for sentences. For
example, Reyna and Kiernan (1994) evaluated 6- and 9-year-olds gist and verbatim
memory for sentences via recognition testing. Children were instructed to identify to-beremembered sentences if the sentence during the memory test was exactly the same as the
sentence they heard or if the sentence had the same meaning as a sentence they heard. In
this study, 6- and 9- year olds were able to differentiate between verbatim and gist
recognition items of previously heard sentences.
These findings supported FTT, as Reyna and Kiernan (1994) posited that gist and
verbatim memory traces are retrieved independently. Much of the adult conversational
memory studies ask participants to broadly recall everything that was said during the
conversation. These instructions might be too weak to activate verbatim memory traces.
However, as time passes, dependency of gist traces will overshadow the ability to retrieve
verbatim traces no matter how strong the verbatim trace is cued (Brainerd & Reyna,
2004). Fuzzy-trace theory explains why conversational memory reflects the gist of
17
conversations rather than verbatim recall for conversations. However, this theory does
not explain the generally recall of conversations.
A central tenet of memory is that memory for meaning is far superior to memory
for details. Therefore, it is not surprising that memory for conversation is far from
verbatim and complete. Kellermann’s (1995) conversational MOP (memory organization
packet) theory provides some theoretical background as to why conversational memory
reports tend to omit large portions of the conversation. The conversational MOP theory
is based on Schank’s (1982) model of memory. According to Schank (1982), the basic
building blocks of memories are scenes, which are a broad collection of common event
features. Within each scene are scripts that provide specific details or actions associated
with a particular scene. MOPs, are a combination of scenes that commonly occur
together. For example, a MOP for “airplane” might include scenes such as “check-in”
and “waiting area” which are each associated with specific scripts (Humphreys, 2003).
Kellermann (1995) posited that conversations are divided and separated into
scenes. As such, multiple memory structures (i.e. scenes) are used to store one
conversation. Poor conversational memory is reflected by the inability to activate all of
the memory structures that contain pieces of a conversation. Furthermore, conversations
can be completely forgotten if none of the specific conversational scenes are activated.
Consistent with the conversational MOP theory, adults display incomplete recall of
conversations (Stafford & Daly, 1984).
1.5 Summary
Child witnesses are often asked to provide conversational testimony during
forensic interviews and the following trial. The reliability of child witnesses’
18
conversational testimony hinges on their ability to remember the conversation in
question. As child witnesses are often asked to report conversations that occurred
between themselves and a conversational partner, the systematic evaluation of children’s
memory for a dyadic conversation (as opposed to memory for a sentence) is of the
upmost forensic importance. The current study is the first empirical examination of
children’s memory for a dyadic conversation. Evaluating children’s memory for a dyadic
conversation will provide a foundation for evaluating the reliability of conversational
testimony provided by child witnesses. The main goals of the current study were to 1)
quantify the overall accuracy of children’s memory for a dyadic conversation, 2)
understand the quality of children’s conversational recall in terms of self-generated and
partner-generated utterances as well as verbal exchanges and 3) evaluate the accuracy of
children’s conversational memory elicited by free recall and recognition style
questioning, all as a function of delay.
19
Chapter Two
The Present Study
The purpose of the present study was to evaluate children’s memory for a dyadic
conversation after a one-week or three-week delay. Second grade children engaged in a
to-be-remembered conversational event. The conversational event was conceptualized as
a storytelling game between the experimenter (hereafter referred to as the storyteller) and
the child. During the storytelling game, the storyteller and child worked together to make
up a story about a monkey and a baseball. This conversation served as the target
conversation in which children’s memory was tested following either a one-week or
three-week delay.
2.1
Hypotheses
2.1.1
Conversational Memory across Delay
A common principle of memory is that memory decays over time (Ebbinghaus,
1885/1964). Accordingly, children assigned to the short delay condition (one-week)
were expected to have superior memory for the target conversation compared to children
assigned to the long delay condition (three-weeks).
2.1.2
Characteristics of Recall
Children’s memory for self-generated and partner-generated utterances was
expected to be superior to their memory for verbal exchanges. Self-generated utterances
were utterances that the child said during the target conversation. Partner-generated
utterances were utterances stated by the storyteller during the target conversation. Verbal
exchanges were composed of an eliciting utterance from the storyteller and the child’s
response. Importantly, verbal exchanges were two consecutive utterances. Children’s
20
memory for verbal exchanges was expected to be poor, as memory for verbal exchanges
requires memory for two consecutive utterances and depends on verbatim-like
recollection. However, it was unknown if children would demonstrate superior recall for
self-generated or partner-generated utterances as the generation effect (Miller et al.,
1996) and the familiarity effect (Stafford & Daly, 1984) predict opposing outcomes.
2.1.3
Accuracy of Recall by Question Format
Children were expected to demonstrate superior conversational memory accuracy
during free recall versus recognition testing. Lamb and his colleagues (2007)
demonstrated that children’s statements elicited via free recall questioning are typically
more accurate than statements elicited by forced-choice questioning. Additionally,
adult’s free recall of conversations tends to be accurate, although incomplete (Lamb et
al., 2000). Accordingly, children’s free recall of the conversation was expected to be
more accurate than their memory elicited by recognition testing.
2.2
Methodology
2.2.1
Participants
The initial sample was comprised of 92 children. Two children were excluded
from the analyses, one due to a prolonged school absence and one due to a digital
recording error. The following analyses were computed with a total of 90 participants.
Ninety (n = 47 male) 7- and 8-year-olds (range = 7.46 – 8.84 years of age) were recruited
from elementary schools in the Toledo, Ohio area. On average, participants were 8.07years-old (SD = .31 years) on the day of memory testing. The sample was largely
Caucasian (n = 81), followed by African American (n = 7), and other (n = 2).
21
All participants were randomly assigned to a delay condition. Forty-nine subjects
(n = 23 male) were assigned to the one-week delay condition and 41 subjects (n = 24
male) were assigned to the three-week delay condition. On average, the memory test
occurred seven days (SD = .83 days) after the target conversation for the one-week delay
condition and 23 days (SD = 2.60 days) after the target conversation for the three-week
delay condition.
2.2.2
Materials
A digital audio recorder was used to record the target conversation and memory
test. The storyteller wore a tall yellow hat during the storytelling event. Additionally,
during the storytelling event, participants were presented with a storytelling box
containing two toys (i.e., a monkey plush and a baseball). These toys acted as props for
the target conversation.
2.2.3
Procedure
This study was a two-session study. Children individually engaged in the target
conversation during the first session. The second session was the memory tests for the
target conversation which occurred either one-week or three-weeks after the target
conversation depending on delay condition assignment.
2.2.3.1
Session I: The Target Conversation
Children were tested individually at their elementary school. An unfamiliar
experimenter (i.e., the storyteller) obtained the child’s assent and ushered the child to the
testing room. The storyteller built rapport with the child then proceeded with the
storytelling event. The storyteller put on a tall yellow hat and explained to the child that
s(he) likes to wear the hat while telling stories. The purpose of the yellow hat was to help
22
the children distinguish between the target conversation and other conversations that that
child engages in on a daily basis during the memory test (at session two).
The storyteller introduced the storytelling game to the child. Children were told,
“(Child’s name) I’m here to talk to you today and see how big of an imagination you
have.” Then the storyteller told the child, “(Child’s name) I have a really fun game to
play with you today. It’s about storytelling and making up stories.” The storyteller
explained to the child that they were going to create a story based on two objects selected
from the storytelling box. The story that the storyteller and child created together was the
target conversation.
Before engaging in the target conversation, the storyteller played a prerecorded
story which served as an example of how to make a story based on the two objects
selected form the storytelling box. This recording lasted approximately two minutes.
Please see Appendix A for the prerecorded story script.
After listening to the prerecorded story, children were instructed to select two
objects from the storytelling box. All children selected a baseball and monkey plush
because they were the only items in the storytelling box. The experimenter was trained to
provide scripted prompts to guide the flow of the target conversation. The script
consisted of 25 questions that the storyteller asked the child. The storytellers’ were
trained to ask additional open-ended prompts such as, “Can you tell me more about that?”
when the child’s responses to the scripted questions were limited.
The purpose of the storytelling event was to create a novel and unique dyadic
conversation between the storyteller and the child. This semi-structured conversation
was strategically designed to maintain as much control as possible while allowing the
23
child and storyteller to participate in the target conversation. As this event was entirely
based on conversations, except for the storytelling box and the storyteller’s yellow hat, it
can be assumed that any memories that children have of the target conversation is based
on their ability to remember what was said. The entire storytelling event was recorded
with a digital tape recorder. Please see Appendix B for the target conversation script.
2.2.3.2
Session II: The Memory Test
Children’s memory for the target conversation was assessed during the second
testing session which occurred either one-week or three-weeks after the target
conversation took place. All children were randomly assigned to a delay condition. On
the day of memory testing, an unfamiliar experimenter obtained assent from the child and
ushered them to the testing room. The experimenter built rapport with the child then
proceeded into the memory test.
There were two main sections of the memory test. First, children were asked a
series of open-ended questions. Children were prompted to recall the prerecorded story,
the target conversation, and questions the storyteller asked during the target conversation.
Of particular interest for this study was children’s report of the target conversation and
questions the storyteller asked during the target conversation. After free recall
questioning, children’s memory for the target conversation was assessed with recognition
testing. The recognition test evaluated children’s memory for self-generated utterances,
partner-generated utterances, and verbal exchanges from the target conversation. The
recognition tests were structured in a way to maximize the standardization among all
recognition test. The standardized structure of the recognition test is discussed below.
Please see Appendix C for the memory test script.
24
2.2.3.3
Creation of the Recognition Test
Every recognition test was tailored to the individualized target conversation.
However, every recognition test was constructed from a standardized template in order to
maintain as much standardization as possible among the recognition tests. There were
three sections in every recognition test: self-generated recognition items, partnergenerated recognition items, and verbal exchange recognition items. Each section
consisted of five false and five true recognition items.
False self-generated recognition items were standardized across all recognition
tests. False self-generated items were utterances that were plausible with the theme of the
target conversation, but were not expected to be said by the child. It was impossible to
create standardized true self-generated recognition items because every child contributed
unique information to the target conversation. As such, the selection of true selfgenerated recognition items was standardized based on the storytellers’ eliciting utterance
(which were standard).
True and false partner-generated recognition items were completely standardized
in every recognition test as there was experimental control over what the storyteller said
and did not say during the target conversation. True partner-generated items were
utterances stated by the storyteller during the target conversation. False partnergenerated utterances were plausible with the target conversation, but were not actually
said by the storyteller during the target conversation.
Verbal exchange recognition items could not be held constant in every
recognition tests because children responded differently during the target conversation.
True verbal exchange recognition items consisted of an eliciting utterance and the child’s
25
response from the target conversation. Similar to prompted-spontaneous alterations in
Bruck and colleagues (1999), false verbal exchange recognition items reflected incorrect
eliciting from the storyteller. Highly prompted questions during the target conversation
appeared open-ended in the recognition test and open-ended questions in the target
conversation appeared highly prompted in the recognition test. The child’s responses in
all of the verbal exchange recognition items always represented their true response during
the actual conversation. As such, the purpose of this analysis was to examine children’s
verbatim-like memory for eliciting utterances and their responses from the target
conversation.
In total, children were asked 30 recognition questions during the recognition test.
Each recognition item followed a yes/no question format and pertained to the target
conversation. The ordering of the 10 recognition items in each section was determined
with a random number generator. The ordering of the items within each section was
consistent in every recognition test. However, the presentation of the three sections of
the recognition test (i.e., self-generated, partner-generated, and verbal exchanges) was
fully counterbalanced. In total, there were six recognition test formats. Please see
Appendix D for an example recognition test.
26
Table 1
Outline of Procedures for Session I and Session II.
Session I:
Target
Conversation
1.) Pre-recorded story
2.) The target conversation
Delay:
One-week or three-weeks
Session II:
Memory Test
1.) Free recall
a.) Pre-recorded story
b.) The target conversation
c.) Storytellers’ questions
2.) Recognition Test (counterbalanced)
a.) Self-generated utterances
b.) Partner-generated
utterances
c.) Verbal exchanges
27
Chapter Three
Results
3.1
3.1.1
Data Coding
Coding of the Target Conversation
Undergraduate research assistants transcribed audio recordings of the target
conversations. Another research assistant verified the transcripts for accuracy.
Transcripts of the target conversation were coded in a series of steps. Frist, false starts
(e.g., um, ah) were eliminated from the transcripts. Then, the transcripts were divided
into utterances. An utterance was operationally defined as a verb bounded by a pause
(Bruck et al., 1999). For example a statement such as, “The monkey ran and he jumped”
would have been divided into two utterances. Utterances were designated as either selfgenerated (i.e., said by the child) or partner-generated (i.e., said by the storyteller). Each
transcript contained a breakdown of verbal exchanges. Every eliciting utterance from the
storyteller and the child’s response were identified in transcripts of the target
conversation. The total number of utterances, self-generated utterances, partnergenerated utterances, and verbal exchanges were tailed and verified.
3.1.2
Coding of the Memory Test
During the memory test, children were asked three free-recall style questions.
Children were prompted to individually recall the prerecorded story, the target
conversation, and questions the storyteller asked the child during the target conversation.
Of particular interest for this study was children’s recall of the target conversation and
questions asked by the storyteller during the target conversation. Undergraduate research
28
assistants transcribed the digital recordings of the memory tests. All memory test
transcripts were verified.
There were multiple steps in coding the memory test transcripts. First, false starts
(i.e., um, uh) and questions for clarification on the free recall task were eliminated from
the transcript. Second, transcripts were divided into utterances (see above for operational
definition). Utterances stated during the memory test that represent a semantic summary
of multiple utterances from the target conversation were coded as recalled themes. For
example, a broad recall of, “We talked about a monkey and a baseball” was coded as a
recalled theme. Recalled themes were judged for accuracy by assessing overall themes
within the subject’s target conversation. Recalled themes were analyzed separately from
children’s memory for utterances from the target conversation.
All utterances that were reported during the memory test were compared to the
target conversation transcript to determine accuracy of recall. Recalled utterances during
the memory test were considered accurate if the reported utterance contained at least the
gist of an original utterance from the target conversation. Gist was operationally defined
as a semantic recall of an utterance from the target conversation (adapted from Brainerd
& Reyna, 2004). In other words, accurate recalled utterance during the memory test
could have represented a paraphrase of an utterance from the target conversation as the
original meaning is maintained. Accurately recalled utterances during the memory test
were identified as utterances said by the child (i.e., self-generated) or the storyteller (i.e.,
partner-generated) during the target conversation in order to evaluate the quality of
children’s conversational reports. Recalled utterances were considered inaccurate if the
29
reported utterance during the memory test did not reflect the meaning of an utterances
from the target conversation or was an elaboration of the target conversation.
Accurately recalled verbal exchanges consisted of an accurately recalled electing
utterance and the child’s response from the target conversation. Recalled verbal
exchanges that did not reflect two consecutive utterances (one utterance from the
storyteller and one utterance from the child) were considered inaccurate. The amount of
accurate and inaccurate recalled themes, recalled utterances, and verbal exchanges were
tailed and verified.
3.1.3
Inter-Rater Reliability
Guidelines provided by Hallgren (2012) were used to determine the inter-rater
reliability (IRR) of the target conversation and memory test coded transcripts. Two
highly trained research assistants coded transcripts of the target conversation into
utterances and verbal exchanges. Each transcript was coded by both research assistants
and compared for accuracy. Any inconsistencies were discussed and rectified between
the researchers. Inter-rater reliability was calculated on the coded transcripts from both
researchers before comparing transcripts and discussing inconsistencies. As both
research assistants independently coded all transcripts, the IRR was calculated using a
two-way IRR measure. Specifically, the inter-class correlation (ICC) was selected as the
IRR measure because it is suitable for studies in which every subject was coded by
multiple coders (all transcripts were coded twice) and ICC is an appropriate measure of
IRR with continuous data (Hallgren, 2012).
The resulting ICC was in the excellent range, ICC = .99 (Cicchetti, 1994). This
indicates that the researchers had a very high degree of agreement when coding the target
30
conversation transcripts into utterances. The high ICC value also indicates that the
amount of measurement error in coding the transcripts used for the following analyses
was low. Additionally, the IRR for verbal exchanges in the target conversation
transcripts was excellent, ICC = .99. Therefore, the coded target conversation transcripts
were considered suitable for hypotheses testing.
Two different research assistants were trained on coding the memory test
transcripts for utterances and accuracy of recall. Again, the ICC measure was appropriate
for the following IRR analyses as two researchers coded all transcripts and the data were
continuous. The ICC scores were excellent (ICC = .99) between both researchers for the
total number of utterances stated during the memory test, as well as the total number of
accurately and inaccurately recalled utterances. The ICC scores of .99 indicate low
measurement error between coders. The coded memory test transcripts were suitable for
hypotheses testing.
3.1.4
Recognition Item Scoring
Recognition items were scored as accurate or inaccurate following a key from the
recognition template. Accepting true recognition items and correctly rejecting false
recognition items was coded as accurate. Accepting false recognition items and
incorrectly rejecting true recognition items was coded as inaccurate. Accurate and
inaccurate response rates for each section were tallied and verified.
3.2
Characteristics of the Target Conversation
First, I examined the characteristics of the target conversation. The duration and
number of utterances of the target conversation should not significantly differ between
the one-week and three-week delay conditions as the target conversation was the initial
31
event and occurred prior to delay. The one-week and three-week delays occurred after
the target conversation. The duration of the average conversation did not significantly
differ between delay conditions (M = 5.30 minutes, SD = 1.70 for the one-week and M =
4.94 minutes, SD = 1.20 for the three-week delay conditions), t(89) = 1.16, p = .25.
Additionally, the average amount of utterances (M = 102.62, SD = 28.53) did not
significantly differ between the one-week and three-week delay conditions (M = 105.86
utterances, SD = 32.66 and M = 98.76 utterances, SD = 22.42 respectively), t(88) = 1.18,
p = .24. On average, the target conversations were approximately five minutes long (SD
= 1.50) and consisted of 102.62 utterances (SD = 28.53).
On average, children contributed 51.20 utterances (SD = 22.61) to the target
conversation. Children in the one-week delay condition did not talk significantly more
(M = 54.31 utterances, SD = 26.60) than children in the three-week delay condition (M =
47.49 utterances, SD = 16.19) during the target conversation, t(80.89) = 1.49, p = .14.
The storytellers spoke, on average, 51.54 utterances (SD = 12.70 utterances) during the
target conversation. This average includes questions asked by the storyteller as well as
any additional utterances that were stated by the storyteller during the target conversation
(e.g., “I like that name” “That’s an awesome color”). The amount of utterances said by
the storytellers’ during the target conversation did not significantly differ across delay
conditions (M = 51.55 utterances, SD = 13.50 and M = 51.51 utterances, SD = 11.84 for
the one-week and three-week delay conditions) t(88) = .01, p = .99. Furthermore, a
paired samples t-test showed that the storyteller and children equally contributed
utterances to the average target conversation, t(89) = -.14, p = .49.
32
The storytellers asked approximately 28 questions (SD = 2.80) regarding the
monkey and/or asking the child to expand on previously provided information during the
target conversation (e.g., “Can you tell me more about the monkey’s school?”). The
number of questions the storytellers asked did not significantly differ between delay
conditions, (M = 27.84 questions, SD = 3.23 for the one-week and M =28.12 questions,
SD = 2.20 for the three-week delay conditions), t(84.79) = -.50, p = .62. Additionally, the
average number of verbal exchanges in the conversations did not significantly differ
between the one-week (M = 32.49, SD = 6.84) and the three-week delay conditions (M =
31.88, SD = 4.85), t(88) = .48, p = .63. On average, there were 32.21 verbal exchanges
(SD = 5.99) between the storyteller and child during the target conversation.
This analysis demonstrates that the target conversations between the one-week
and three-week delay conditions were similar in terms of utterance composition and
duration. Additionally, the average conversation was equally composed of utterances
from the child and storyteller, indicating that the target conversations were truly dyadic.
The average conversation lasted approximately five minutes and was composed of
approximately 103 utterances.
3.3
Memory Test: Overall Recall of the Target Conversation
Virtually all participants (n = 87) remembered the target conversation and did not
require additional prompting during the memory test (i.e., “Do you remember your story
about the monkey and the baseball?”). The vast majority of participants (n = 74)
accurately recalled themes from the target conversation (e.g., “We talked about a monkey
and a baseball” “We talked about my monkey’s family”). A theme was defined as a
semantic summary of multiple utterances from the target conversation. There was a
33
limitless amount of recalled themes in every target conversation as there are multiple
variations in how several utterances could be summarized. Every theme (100%) that was
recalled from the target conversation was accurate. However, recalled themes were
limited in both the one-week (M = 1.15 themes, SD = .43) and three-week (M = 1.37
themes, SD = .69) delay conditions and did not significantly differ between conditions,
t(55.92) = -1.61, p = .11. On average, children accurately reported 1.26 (SD = .57)
themes from the target conversation during the memory test. The maximum number of
reported themes was four.
The following analyses regarding overall recall of the target conversations were
computed with logarithm transformed data as the raw data violated the parametric
assumption of normality. Multiple techniques were utilized to screen the data for
normality, including histograms, p-p plots, Kolmogorov-Smirnov test, Shapiro-Wilks
test, as well as skewness and kurtosis z-scores. The findings from all of these approaches
were taken into consideration when assessing the parametric assumption of normality.
Ultimately, many of the dependent variables violated the parametric assumption of
normality as the distributions were positively skewed. The following analyses regarding
the number of utterances recalled form the target conversation, the proportion of the
conversation accurately reported, and the proportion of the child’s initial utterances that
were accurately recalled were conducted on logarithmic transformed data as the raw data
violated the parametric assumption of normality.
Using transformations to alleviate issues caused by violating assumptions has
been debated in the statistical literature (Osborne, 2002). Opponents of data transforming
highlight that transformations inherently reduce the scale of measurement of the
34
dependent variable (often from interval/ratio to ordinal) by eliminating equidistant
intervals between data points. Additionally, transformed data is often difficult to interpret
as the unit of measurement is in transformed units rather than raw units. For these
reasons, opponents of data transformations will often suggest deleting subjects or using
non-parametric statistics, which have more liberal assumptions. However, Zimmerman
(1995) demonstrates that the non-parametric, “assumption free,” tests often are hindered
when the assumption of normality is violated. Additionally, the deletion of outliers in an
analysis has also been criticized when the outlier represents a true data point (Howell,
2013). Ultimately, the uniqueness of the data and analyses will dictate which approach is
appropriate in correcting for violations.
The following analyses were conducted on logarithmic transformed data
following the guidelines provided by Field (2013), Howell (2013), and Osborne (2002).
Square root transformations, non-parametric testing, and deletion of outliers were
entertained as potential methods to rectify the normality violation. Square root
transformations did rectify the normality violation. However, square root transformations
are considered less desirable than logarithmic transformations as they are more difficult
to interpret (explained in more detail below, also see Bland & Altman, 1996a).
Additionally, non-parametric testing and the deletion of subjects resulted in the same
statistical outcomes reported below. As the same statistical story was being told by the
data, regardless of the analysis technique, dependent variables that violated the normality
assumption were transformed to maintain the increased power associated with parametric
testing and all values were used in the analyses to capture the essence of the sample.
35
Following Osborne (2002), all dependent variables were anchored at a value of
one. Negative values are undefined under logarithmic transformations and the log of a
value less than one will produce a negative number. The present data did not contain
negative values, but did contain values less than one. As such, a constant was added to
all dependent variables that contained values less than one to avoid negative values under
the logarithmic transformation.
Additionally, the following reported means and confidence intervals were
computed by back transforming the logarithmic values into raw units. As per Blan and
Altman (1996a) and Howell (2013), the geometric mean (GM) was reported in the
following analyses. The geometric mean was computed by calculating the arithmetic
mean on the transformed data and back transforming that value. In this case, the antilog
of the arithmetic mean of the transformed data was computed to calculate the geometric
mean.
A similar procedure was followed to calculate confidence intervals (Bland &
Altman, 1996b). Ninety-five percent confidence intervals were calculated by taking the
antilog of the confidence intervals values computed from the transformed data. Bland
and Altman (1996b) reported that back transforming confidence intervals for square root
transformed data is undesirable and often leads to invalid conclusions. However, back
transforming confidence intervals for logarithmic transformations is sound, a
characteristic that makes logarithmic transformations preferable to others (Bland &
Altman, 1996b). While back transforming logarithmic confidence intervals is sound,
they have to be appreciated as the geometric confidence intervals rather than confidence
36
intervals of the raw data. Raw means, medians, and standard deviations were provided
for ease of interpretation. Medians are reported due to the positive skew in the raw data.
Proportion scores were created for each participant to reflect their accurate recall
of the target conversation proportionate to the total number of initial utterances stated
during the target conversation. For example, if there were 100 utterances in the target
conversation and the child correctly recalled 10 utterances during the memory test, this
child would be given a proportion score of 10%.
As hypothesized, participants in the one-week delay condition accurately recalled
significantly more of the initial target conversation (GM = 6.85%, M = 7.72%, Mdn =
6.82%, SD= 4.27%) compared to participants in the three-week delay condition (GM =
4.46%, M = 5.88%, Mdn = 4.85%, SD = 4.44%) during the memory test. The geometric
mean difference, 1.44%, 95% CI [1.10%, 1.88%], was significant and had a medium to
large effect, t(64.78) = 2.75, p < .01, d = .59.
3.4
Memory Test: Accuracy of Memory Reports during Free Recall
Regardless of accuracy, participants in the one-week delay condition recalled
more utterances during the memory test (GM = 10.48 utterances, M = 12.71, Mdn =
11.00, SD = 8.43) than participants in the three-week delay condition (GM = 6.60
utterances, M = 8.63, Mdn = 7.00, SD = 6.28). This geometric difference, 1.51, 95%CI
[1.32, 2.01], was significant and had a medium to large effect size, t(88) = 2.85, p < .01, d
= .60. Children in the one-week delay condition said significantly more (regardless of
accuracy) while recalling the target conversation compared to children in the three-week
delay condition.
37
The assumption of normality was not violated in the following analysis and thus,
raw data was utilized in the following computation. Only two participants in the threeweek delay condition did not recall any utterances from the target conversation during the
memory test. Of the participants that recalled at least one utterance from the target
conversation (n = 88), participants in the one-week condition were more accurate in their
recall (M = 70.73% accurate, Mdn = 72.73%, SD = 22.74%) than the participants in the
three-week condition (M = 65.43% accurate, Mdn = 66.67%, SD = 22.38%). However,
this difference was not significant, t(86) = 1.09, p = .29. On average, the entire sample
was approximately 68% accurate in their free recall of the target conversation (Mdn =
66.67%, SD = 22.61%) during the memory test. In other words, on average,
approximately a third of the utterances that were recalled during the memory test were
inaccurate.
In sum, the vast majority of participants recalled that the conversation occurred
and many children accurately recalled themes of the target conversation during the
memory test when prompted with free recall questioning. However, recalling what was
actually said during the conversation proved difficult a week after the conversation
occurred and continued to be problematic for the three-week delay condition.
Nonetheless, the majority of children’s recall was accurate (68% accurate on average),
even after a three-week delay. While children were mostly accurate in there recall, they
also reported a high frequency of inaccurate utterances from the target conversation. The
accuracy of children’s free recall reports of the target conversation did not vary across
delay conditions. Participants in the three-week delay condition recalled less of the target
conversation, but did not incorporate inaccurate information above and beyond their one-
38
week delay counterparts to compensate for their dwindling memory for the conversation
during the memory test.
3.5
Memory Test: Characteristics of Memory Reports during Free Recall
During the memory test, children were asked to recall the target conversation
exactly as it was created. Recall that the average target conversation was equally
composed of utterances said by the child and the storyteller. However, when recalling
the target conversation during the memory test, children were strongly inclined to recall
their initial utterances versus utterances said by the storyteller during the target
conversation. Virtually, all recalled utterances (98.36% of recalled utterances) were
utterances declared by the child during the target conversation. This pattern was
consistent across delay conditions (99.03% and 97.17% for the one-week and three-week
delay conditions). Participants in the one-week delay condition accurately recalled a
significantly higher proportion of their initial utterances from the target conversation
during the memory test compared to participants in the three-week delay condition (GM
= 13.92% accurately recalled, M = 15.84%, Mdn = 14.29%, SD = 8.60% and GM =
8.01% accurately recalled, M = 11.30%, Mdn = 9.10%, SD = 8.36% respectively),
t(59.87) = 3.14, p < .01, d = .68. The geometric mean difference in percent accurate was
1.66%, 95% CI [1.20%, 2.28%]. This was a medium to large effect.
Only six participants (four in the one-week delay condition) spontaneously
recalled any statements said by the storyteller during target conversation. The four
participants in the one-week delay condition recalled one utterance said by the storyteller
during the conversation, on average (two maximum), and were 50% accurate in their
recall of the storytellers utterances during the memory test. Two participants in the three-
39
week delay condition each recalled five utterances said by the storyteller from the target
conversation with 100% accuracy. As so few participants spontaneously recalled the
storyteller’s utterances from the target conversation during the memory test, further
statistical analyses were not warranted.
Furthermore, spontaneous recall of verbal exchanges was extremely limited.
Only five children (four in the one-week delay condition) spontaneously recalled any
verbal exchanges from the target conversation. Of the participants in the one-week delay
condition that recalled verbal exchanges (n =4), the maximum number of recalled verbal
exchanges was two and recalled verbal exchanges were 50% accurate. One subject in the
three-week delay condition recalled four verbal exchanges with 75% accuracy.
Overall, the vast majority of children’s reports during the memory test were
comprised of utterances said by the child during the target conversation. Children in the
one-week condition accurately recalled a higher proportion of their own initial utterances
from the target conversation than children in the three-week condition during the memory
test. Very few participants spontaneously recalled any statements (inaccurate or
accurate) said by the storyteller during the target conversation when asked to recall the
target conversation in its entirety. Furthermore, recall of the storyteller’s utterances and
verbal exchanges for children who spontaneously (and accurately) recalled these
utterances remained minimal.
3.6
Memory Test: Recall of the Storyteller’s Utterances and Verbal Exchanges in
Cued Recall
Perhaps spontaneous recall of the storyteller’s utterances and verbal exchanges
from the target conversation was limited due to the wording of the free recall question
40
during the memory test. In the memory test, children were asked to recall the target
conversation exactly the same way it occurred, starting at the very beginning to the end.
During the target conversation, the storyteller’s role was to ask the child questions to help
them create their story rather than provide information about the monkey and the
baseball. As such, perhaps participants perceived the free recall question during the
memory test as asking them to recall information about the monkey and the baseball
rather than the interchange of utterances between themselves and the storyteller. The
storyteller’s utterances would have been excluded while recalling the story if participants
perceived the free recall questions in this manner. As this possible confusion was
anticipated, children were specifically asked to recall the storyteller’s utterances after
their initial free recall of the entire conversation during the memory test. One subject in
the three-week delay condition was not asked this question during the memory test and
was excluded from the following analyses.
The majority of participants (n = 77) correctly reported that the storyteller asked
questions during the target conversation. However, actual recall of the storyteller’s
utterances remained limited. Approximately a third of the sample (n = 28) did not report
any of the storyteller’s questions from the target conversation, even when cued to provide
this information during the memory test. Reporting the storyteller’s questions from the
target conversation remained limited even for participants who recalled at least one of the
storytellers’ questions during the memory test (n = 61). On average, these participants
recalled approximately three of the storyteller’s questions, regardless of accuracy (M =
2.81, SD = 1.58 and M = 2.80, SD = 1.68 in the one-week and three-week delay
conditions, respectively). Participants in the three-week delay condition were more
41
accurate in their report of the storyteller’s questions (M = 67.29% accurate, SD =
29.77%) compared to participants in the one-week delay condition (M = 62.05%
accurate, SD = 38.24%) when prompted to retell this information during the memory test.
However, this difference was not significant, t(59) = -.57, p = .57. On average,
participants were 64.20% accurate (SD = 34.84%) in their recall of the storyteller’s
questions when cued to provide this information. Thus, while the majority of participants
recalled that the storyteller asked questions during the target conversation, actually
recalling the questions proved challenging even when specifically cued to recall this
information during the memory test.
Additionally, recall of verbal exchanges form the target conversation was low for
both delay conditions even when specifically cued to recall this information during the
memory test. Over a third of the sample (n = 32) did not report any verbal exchanges
from the target conversation when prompted during the memory test. Additionally, recall
of verbal exchanges from the target conversation remained limited even for participants
who reported at least one verbal exchange during the memory test (n = 57). Of
participants that recalled at least one verbal exchange, participants in the one-week and
three-week delay condition recalled approximately two verbal exchanges, regardless of
accuracy (M = 2.57, SD = 1.69 and M = 2.41, SD = 1.10 respectively). Participants in the
three-week delay condition were more accurate in their report of the verbal exchanges (M
= 60.23% accurate, SD = 38.83) compared to participants in the one-week delay
condition (M = 55.97% accurate, SD = 39.64%) when prompted to recall this
information. However, this difference was not significant, t(55) = -.40, p = .69. On
42
average, participants were 57.61% (SD = 39.03) accurate in their recall of verbal
exchanges when specifically asked to recall this information during the memory test.
These results indicate that recall of the storyteller’s utterance and the structure of
the conversation (i.e., verbal exchanges) remained extremely limited even when
specifically cued to report these utterances at during the memory test. These results show
that the structure of the target conversation was not reported while recalling the
conversation at both time points.
3.7
Memory Test: Recognition Test Performance
The recognition test was constructed so that each participant would be asked a
total of 30 questions (10 questions about each of self-generated utterances, partnergenerated utterances, and verbal exchanges). In some instances, statements used in the
standardized recognition template were not said during the target conversation due to
researcher error. When this error occurred, the recognition item was not asked to the
child and the missing item was treated as missing data during the subsequent analysis.
Fifteen subjects had recognition test composed of 29 items, one subject was asked 28
questions, and one subject was asked 26 questions. For these participants, their scores on
the three sections (i.e., self-generated, partner-generated, verbal exchanges) were
computed by calculating the proportion of correctly answered items over the total number
of items asked per section. For example, if a participant scored five items correct out of
nine items in a given section, they received an adjusted score of 5.56. The total score was
computed by adding the adjusted scores from all three sections of the recognition test to
create one summed score.
43
The following analyses were conducted on raw data as parametric assumptions
were not violated. On average, children in the one-week delay condition performed
better on the recognition test (M = 22.28 correct, SD = 1.81) than children in the threeweek delay condition (M = 20.42 correct, SD = 2.25). This difference, 1.86, 95% CI
[1.01, 2.71], was significant and had a large effect, t(88) = 4.35, p < .001, d = .91.
To further explore performance on question type between delay conditions, a 2
(delay condition: one-week, three-week) x 3 (question type: partner-generated, selfgenerated, verbal exchange) mixed factorial ANOVA with repeated measures on the last
factor was conducted. There were significant main effect for the question type, F(2, 176)
= 237.38, p < .001, p2 = .73 and the delay condition, F(1, 88) = 18.94, p < .001, p2 =
.18. However, these main effects were qualified by a significant interaction between
delay condition and question type, F(2, 356) = 4.47, p = .01, p2 = .05. Please see Figure
1. Bonferroni adjusted t-tests showed that children in the one-week condition performed
better on the self-generated recognition items (M = 8.43, SD = 1.04) than children in the
three-week delay condition (M = 7.57, SD = 1.48). This difference, .86, 95%CI [.31,
1.41], was significant and had a medium to large effect, t(69.93) = 3.13, p = .003, d = .67.
Additionally, children in the one-week condition performed better on the partnergenerated recognition items (M = 8.78, SD = 1.01) compared to children in the threeweek delay condition (M = 783, SD = 1.18). This difference, .95, 95%CI [.49, 1.40] was
significant and had a large effect, t(88) = 4.11, p < .001, d = .86. Performance on the
verbal exchange recognition items did not vary across delay conditions, t(88) = .25, p =
.80. Please see Table 2 for means and standard deviations.
44
Figure 1. Interaction of Recognition Test Performance by Question Type and Delay
Condition.
Table 2
Means and Standard Deviations of Recognition Test Performance by Question Type and
Delay Condition
Delay Condition
Question Type
One-Week
Three-Weeks
Row Means
Self-Generated
8.43 (SD = 1.04)
7.57 (SD = 1.48)
8.00 (SE = .13)
Partner-Generated
8.78 (SD = 1.01)
7.83 (SD = 1.18)
8.30 (SE = .12)
Verbal Exchange
5.08 (SD = 1.04)
5.02 (SD = .96)
5.05 (SE = .11)
Column Means
7.43 (SE = .10)
6.81 (SE = .11)
SD = standard deviation; SE = estimated standard error
45
3.8
Memory Test: Accuracy of Recall by Question Format
Next, I examined whether accuracy differences emerged for report produced in
free recall versus recognition testing for the one-week and three-week delay conditions.
A 2 (delay condition: one-week vs. three-week) x 2 (question format: recognition vs. free
recall) mixed factorial ANOVA with repeated measures on the last factor was conducted
to further examine the accuracy of children’s reports. There was a significant main effect
of delay condition, F(1, 89) = 4.46, p = .04, p2 = .05. Approximately five percent of the
variance in recall accuracy was attributed to delay condition. Performance was superior
in the one-week delay condition compared to the three-week delay condition regardless
of question format during the memory test. Please see Table 3. The main effect of
question format was not significant, F(1, 88) = 1.87, p = .18. Additionally, the delay
condition and question format interaction was not significant, F(1, 176) = .01, p = .92.
These results indicate that accuracy of performance significantly differed between delay
conditions, but did not interact with question format.
Table 3
Means and Standard Deviations of Report Accuracy by Question Format and Delay
Condition
Delay Condition
Question Format
One-Week
Three-Weeks
Recognition
74.28% (SD = 6.03%)
68.49% (SD = 7.30%)
Free Recall
70.73% (SD = 22.74%)
65.43% (SD = 22.38%)
46
Chapter Four
Discussion
A growing body of psycho-legal literature highlights the need to understand
children’s ability to recall dyadic conversations. Child witnesses are often asked to recall
conversations during forensic investigations. Additionally, defense and prosecuting
attorneys often ask child witnesses to provide conversational testimony in CSA trials
(Stolzenberg & Lyon, 2014). Child witnesses’ ability to provide accurate and
forensically relevant conversational testimony relies on their memory for the
conversation in question. Prior research has failed to examine children’s memory for
dyadic conversation and children’s memory for conversation across delay. The present
study was designed to address this gap by systematically evaluating children’s memory
for a dyadic conversation following either a one-week or three-week delay. The mains
goals of the study were to 1) examine children’s overall recall of the target conversation,
2) evaluate the quality of children’s reports in terms of self-generated and partnergarnered utterances as well as verbal exchanges, and 3) examine the accuracy of
children’s reports elicited by free recall and recognition style questioning, all as a
function of delay.
4.1
Conversation Memory across Delay
In general, children recalled the overall theme of the conversation (i.e., that they
talked about monkey and a baseball), but had difficultly reporting what was actually said
in the target conversation during the memory test, regardless of delay condition. While
children in the one-week delay condition significantly reported a higher proportion of the
target conversation (approximately 7%, on average) than children in the three-week delay
47
condition (approximately 4%, on average), the amount of the target conversation
accurately recalled was low for children in both conditions. In accordance with
Kellermann’s (1995) conversational MOP (memory organization packet) theory, children
omitted a large proportion of the conversation in their memory reports. Kellermann’s
(1995) theory suggest that children’s poor memory for the conversation reflects their
inability to access all of the various MOPs that contained information regarding the
conversation during memory testing.
While children’s recall of the conversation was limited, children were mostly
accurate, 68% on average, in their reports of the target conversation when asked to recall
the target conversation in its entirety during the memory test. In other words,
approximately one-third of children’s reports were inaccurate. This is not a trivial
finding and should not be overlooked. Children were mostly accurate in their free recall
of the target conversation, but children also incorporated a substantial amount of
incorrect information while recalling the conversation during the memory test. Accuracy
of children’s memory reports during free recall testing did not vary across delay
conditions.
Overall, these finding suggests that children have poor memory for dyadic
conversations after a delay as short as one-week, but the majority of their reports were
accurate up to three-weeks after the conversation occurred. The hypothesis that children
in the one-week condition would have superior reports for the target conversation
compared to children in the three-week delay condition was partially supported. Children
in the one-week delay condition reported a significantly greater proportion of the target
48
conversation compared to children in the three-week delay condition, but accuracy of
recall did not vary as a function of delay.
These findings need to be appreciated as children’s gist memory for utterances
rather than verbatim memory for utterances from the target conversation. An utterance
that was recalled during the memory test was considered accurate if it contained the
semantic meaning of an utterance said during the target conversation. Verbatim recall
was not a requirement for accuracy. Thus, the finding that approximately one-third of
children’s reports were inaccurate is concerning as accuracy was measured by gist rather
than verbatim recall.
4.2
Characteristics of Recall
All children had a strong tendency to recall utterances that they said during the
target conversation opposed to utterances stated by their conversational partner when
prompted to recall the conversation in its entirety during the memory test. Children were
not apt to report the storyteller’s utterances or verbal exchange from the target
conversation, regardless of delay condition. Only six children spontaneously recalled any
statements said by the storyteller and five children recalled verbal exchanges from the
target conversation when promoted to recall the entire conversation.
After the initial free recall of the target conversation, children were specifically
prompted to recall any questions the storyteller asked during the target conversation.
Approximately one-third of all children did not report any of the storytellers’ utterances
(and verbal exchanges) even when cued to recall this information. Of children who
recalled at least one of the storyteller’s questions from the target conversation, children
49
accurately recalled approximately two questions said by the storyteller and one verbal
exchange from the target conversation, on average. These findings suggests that children
had a strong tendency to report their own initial utterances from the target conversation
and recall of the storyteller’s utterances and verbal exchanged from the target
conversation was low, even when specifically prompted to recall this information. Recall
of the storyteller’s utterances and verbal exchanges was extremely low with free recall
and cued recalled questioning during the memory test.
This pattern provides strong support for the generation effect, which postulates
that memory for self-generated utterances should be stronger than memory for partnergenerated utterances from a dyadic conversation due to the additional cognitive processes
involved in producing utterances in a conversation (Miller et al., 1996). In the current
study, children had a strong tendency to recall utterances they said during the target
conversation rather than utterances said by their conversational partner when prompted
with free recall and cued recall questioning. Memory for verbal exchanges was
extensively low during free recall and cued recall questioning.
The significant interaction between delay condition and recognition test
performance provides further insight to children’s memory for self-generated utterances,
partner-generated utterances, and verbal exchanges. Children in the one-week delay
condition performed significantly better on the self-generated and partner-generated
sections of the recognition test compared to children in the three-week delay condition.
However, performance on the verbal exchange section of the recognition test did not
significantly vary between delay conditions. Children’s memory for verbal exchanges
was low when prompted with recognition style questioning regardless of delay condition.
50
In accordance with fuzzy-trace theory (FTT), performance on gist recognition
items (i.e., self-generated and partner-generated recognition items) was better than
performance on verbatim-like recognition items (i.e., verbal exchanges). Children’s
memory for verbal exchanges was extremely low, even when given salient cues to recall
this information during the recognition test. Fuzzy-trace theory holds that verbatim
traces deteriorate over time regardless of the strength of the prompt to recall this
information (Brainerd & Reyna, 2004). In this study, children in both the one-week and
three-week delay conditions performed poorly on the verbal exchange recognition test
section.
The hypothesis that children would have poor memory for verbal exchanges was
supported. Recall of verbal exchanges from the target conversation was low when
children were prompted to provide this information during free recall, cued recall, and
recognition questioning. Additionally, children had a strong preference to report their
own initial utterances from the target conversation rather than utterances said by the
storyteller when prompted to recall the target conversation during the memory test.
4.3
Accuracy of Reports by Question Format
Lastly, children were expected to be more accurate in their report of the target
conversation when allowed to freely recall what was said versus being prompted with
recognition style questions. There was insufficient evidence to support this hypothesis.
Accuracy of recall did not vary as a function of question format. However, similarly to
event reports (Lamb et al., 2007), we would expect that children’s reports for a
conversation elicited with recognition style questions would be less accurate than their
reports elicited with free recall questions if children are exposed to misinformation
51
regarding the conversation. The target conversation in the current study did not contain
any misinformation, which could explain why the accuracy of children’s report did not
vary by question format. Another possibility is that the false items in the recognition test
were too obviously false to the children, which could have inflated children’s
performance on the recognition test.
4.4
Forensic Implications
The overall findings from the current study suggests that children are able to
remember the theme of a prior conversation, but recalling what was actually said during
the conversation (with gist accuracy) is very difficult for children, even one-week after
the conversation occurred. Furthermore, children’s reports of a previous conversation,
while mostly accurate, also contain a high proportion of inaccurate information. In the
forensic context, this is problematic as distinguishing between true and false information
in children’s reports is very challenging and often unsuccessful (Ceci, Loftus, Leichtman,
& Bruck, 1994). Additionally, in the current study, the accuracy of children’s memory
reports for the target conversation was judged by gist rather than verbatim recollection.
Recalled utterances in the memory test were considered accurate if they accurately
depicted the gist of an utterance in the target conversation. In terms of conversational
testimony, recalling the gist of the conversation (rather than verbatim recall) might be
problematic.
The findings from the current study also suggest that children’s memory for
eliciting utterances is very poor when prompted with free recall, cued recall, and
recognition style questioning. This is problematic as legal personnel are often interested
in knowing what the child said in a prior conversation as well as how the child’s
52
statement was elicited during the conversation. Knowledge of how the child witnesses’
statements were elicited in prior conversations can provide a context in which to interpret
the reliability of testimony provided by child witnesses. Recalling individual utterances
outside of their conversational context can be problematic for the interpretation of the
conversation and for determining of the veracity of child witnesses’ reports. Findings
from the current study suggest that children’s memory for what they said and how their
statement was elicited is very poor. Children’s memory for the structure of the
conversation drastically deteriorates as short as one-week after the conversation occurs.
Child witnesses cannot be expected to provide accurate verbatim-like testimony
regarding the structure of previous conversation, as children’s memory for verbal
exchanges is extensively low.
The current study was the first study to systematically evaluate children’s
memory for a dyadic conversation, in general, and also across delay. Further
examination of children’s memory for conversation is necessary before final
recommendations can be made to legal professionals on the reliability of conversational
testimony provided by child witnesses. However, the findings from the current study
suggest that children’s memory for conversation sharply declines after short delays of one
and week threes. Furthermore, children’s reports of prior conversations will often
represent the gist of what was said during the conversation (as opposed to verbatim) and
has a high risk of incorporating substantial amounts of incorrect information that never
occurred during the conversation.
53
4.5
Limitations and Future Directions
The current study was the first study to examine children’s memory for a dyadic
conversation. As this was the first study to systemically evaluate children’s
conversational memory, there were limitations to the current study that need to be
addressed. Limitations of the current study and future directions are discussed below.
First, the current study did not examine developmental trends in children’s
conversational memory. The current study only evaluated second grade children’s
memory for conversation in order to maintain a clean design, as this was the first study to
systematically evaluate children’s conversation memory. Future studies need to evaluate
developmental trends in children’s memory for conversation. There is reason to believe
that there would be developmental differences in children’s reports of prior conversations
due to their developing understanding of conversations and cognitive capacities. Future
studies are needed to explore this area and evaluate developmental trends in children’s
memory for conversation in terms of quantity and quality of reports.
Second, while beyond the scope of the current study, evaluating children’s
memory for dyadic conversations after an extended delay is forensically relevant. Child
victims of CSA often delay disclosure (London, Bruck, Ceci, & Shuman, 2005) and
forensic investigations are often long in duration. In the current study, children’s
memory for the target conversation was assessed either one-week or three-weeks after the
target conversation occurred. These delay times were selected for feasibility. Future
studies need to systematically evaluate children’s memory for dyadic conversations after
an extended delay.
54
Third, the finding that children had a strong preference in reporting their own
initial utterances from the target conversation as opposed to utterances said by their
conversational partner should be generalized within the confines of the dyadic
conversation that occurred in the current study. In the current study, children participated
in a semi-structured dyadic conversation where their conversational partner was provided
scripted lines to guide the flow of the conversation. The use of a semi-structured
conversation was a major strength of the research design as it allowed for more control
and internal validity. The storyteller’s role in the conversation was to ask questions
rather than contribute substantive information to the conversation. Perhaps, children did
not recall the storyteller’s utterances (and thus verbal exchanges) because the researcher
did not provide direct information about the monkey and the baseball. Characteristics of
children’s conversational memory need to be appreciated in light of the target
conversation that occurred in the current study.
Future studies should examine children’s memory for dyadic conversations within
varying contexts. The target conversation in this study did not contain contextual
components that are often a part of conversations regarding abuse allegations (e.g.,
suggestibility, misinformation, inclusion of sexual subject matter). Children’s memory
for conversation could vary according to the content of the conversation, the context in
which the conversation occurred, and the context in which children are asked to recall the
conversation. Future studies should explore the relationship between the content of the
conversation and recall context on children’s reports for prior conversations.
Additionally, future studies need to evaluate children’s memory for conversations
with various conversational partners. In the present study, the storyteller and child were
55
complete strangers. However, child witnesses are often asked to recall conversations that
occurred between themselves and a familiar conversational partner. Child witnesses are
often asked to recount their disclosure conversations as well as conversations between
themselves and the perpetrator (Stolzenbergh & Lyon, 2014). Child sexual abuse victims
often know the perpetrator (U.S. Department of Health and Human Services, 2013) and
they typically disclose abuse to their mother, peers, and/or other family members (Malloy
et al., 2013). Thus, evaluating children’s memory for a particular conversation that
occurred with a familiar conversational partner is of the utmost forensic importance.
Defense attorneys are especially interested in learning about prior conversations
to gain insight to the reliability of the child witnesses’ report. Principe and colleagues
(for a review see, Principe & Schindewolf, 2012) demonstrated that natural occurring
conversations with a peer or the child’s mother about an event can distort children’s event
reports. Children often incorporate information from conversations into their memory
reports for an event. This is problematic when the conversation includes erroneous
information regarding the event, potentially jeopardizing the reliability of the children’s
event reports. As conversations with peers and parents can influence memory reports,
children’s memory for conversations with these familiar conversation partners ought to
be explored.
Another future area of direction is exploring children’s source monitoring abilities
as a function of their memory for conversation. In source monitoring tasks, children are
asked to source the knowledge of their information as something that actually occurred
during an event or something that they only heard (Johnson, Hashtroudi, & Lindsay,
1993). Children often perform poorly on source monitoring tasks and have the tendency
56
to say that non-experienced but heard details of an event actually happened (Principe &
Schindewolf, 2012). Source monitoring tasks inherently assume that children remember
the target conversation and are able to utilize it as a source of their knowledge. The
current study suggests that children are able to remember the general topic of a
conversation, but have difficulty reporting what was actually said during the conversation
up to three-weeks after the conversation occurred. Future research conducted to explore
the relationship between source monitoring performance and children’s memory for the
conversation they are being asked to source is a warranted.
The reliability of child witnesses’ testimony has been subjected to years of
research and scrutiny in the courts. As conversations are a very influential force in
children’s memory reports, an empirically supported forensic interviewing protocol (i.e.,
the NICHD) was developed to minimize misinformation and suggestive statements said
to the child during forensic interviewing in efforts to avoid contaminating their testimony
(Lamb et al., 2007). However, there is no safeguard to protecting children from exposure
to misinformation and suggestions during informal interviews and conversations
regarding the allegations. Asking witnesses to recall prior conversation is often the only
way to learn what was discussed during informal conversations. Physical evidence of
informal conversations (e.g., recording, notes) is extensively rare. As such, child
witnesses are often asked to recall prior conversations regarding the alleged event. Child
witnesses’ ability to provide conversational testimony relies on their memory for the
conversation in question.
Children’s ability to recall conversations has long been overlooked in the
developmental literature. The current study was the first study to systematically evaluate
57
children’s memory for a dyadic conversation and, specifically, children’s memory for a
dyadic conversation after delay. Generally, children remembered the conversation
occurred and accurately reported the theme of the conversation, but recalling what was
said during the conversation was very difficult even after a short delay of one-week.
Further evaluation of children’s memory for conversation is a fruitful area for future
research.
58
References
Baker-Ward, L., Hess, T. M., & Flannagan, D. A. (1990). The effects of involvement on
children’s memory for events. Cognitive development, 5(1), 55-69.
Bland, J. M., & Altman, D. G. (1996a). The use of transformation when comparing two
means. British Medical Journal, 312(7039), 1153.
Bland, J. M., & Altman, D. G. (1996b). Transformations, means, and confidence
intervals. British Medical Journal, 312(7038), 1079.
Brainerd, C. J., & Reyna, V. F. (2004). Fuzzy-trace theory and memory development.
Developmental Review, 24(4), 396-439.
Bruck, M., Ceci, S. J., & Francoeur, E. (1999). The accuracy of mothers’ memories of
conversations with their preschool children. Journal of Experimental Psychology:
Applied, 5(1), 89-106.
Ceci, S. J., & Bruck, M. (1993). Suggestibility of the child witness: A historical review
and synthesis. Psychological Bulletin, 113(3), 403-439.
Ceci, S. J., Loftus, E. F., Leichtman, M. D., & Bruck, M. (1994). The possible role of
source misattributions in the creation of false beliefs among preschoolers. The
International Journal of Clinical and Experimental Hypnosis, 42(4), 304-320.
Child Welfare Information Gateway (2013). Making and screening reports of child abuse
and neglect. Washington: Government Printing Office.
Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb for evaluating normed
and standardized assessment instruments in psychology. Psychological
Assessment, 6(4), 284-290.
59
Davis, D., & Friedman, R. D. (2007). Memory for conversation: The orphan child of
witness memory researchers. In M. P. Toglia, J. D. Read, D. F. Ross, & R. C. L.
Lindsay (Eds.), Handbook of eyewitness psychology: Memory for events (pp. 352). Mahwah, NJ: Lawrence Erlbaum Associates.
Davis, D., Kemmelmeier, M., & Follette, W. C. (2005). Memory for conversation on
trial. In Y. I. Noy & W. Karwowski (Eds.), Handbook of human factors in
litigation (pp. 1-29). Boca Raton, FL: CRC Press.
Duke, S. B., Lee, A. S., & Pager, C. K. W. (2007). A picture’s worth a thousand words:
Conversational versus eyewitness testimony in criminal convictions. American
Criminal Law Review, 44(1), 1-52.
Ebbinghaus, H. (1885/1964). Memory (H. A. Ruger & C. E. Bussenius, Trans.). New
York, NY: Dover. (Original work published 1885).
Field, A. (2013). Discovering statistics using IBM SPSS statistics (4th ed.). Thousand
Oaks, CA: Sage Publications Inc.
Hallgren, K. A. (2012). Computing inter-rater reliability for observational data: An
overview and tutorial. Tutorials in Quantitative Methods for Psychology, 8(1), 2324.
Hedelin, L., & Hjelmquist, E. (1998). Preschool children’s memory of the from/content
distinction in communicative tasks. Journal of Psycholinguistic Research, 27(3),
421-452.
Howell, D. C. (2013). Statistical methods for psychology (8th ed.). Belmont, CA:
Wadsworth.
60
Humphreys, L. R. (2003). Understanding differences in conversational memory: A
comparison of close friends and strangers. (Unpublished master’s thesis). The
University of Georgia, Athens, GA.
Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Source monitoring.
Psychological Bulletin, 114(1), 3-28.
Kassam, K. S., Gilbert, D. T., Swencionis, J. K., & Wilson, T. D. (2009). Misconceptions
of memory: The Scooter Libby effect. Psychological Science, 20(5), 551-552.
Kellermann, K. (1995). The conversational MOP: A model of patterned and pliable
behavior. In D. E. Hewes (Ed.), The Cognitive Bases of Interpersonal
Communication (pp. 181-221). Hillside, NJ: Lawrence Erlbaum Associates.
La Rooy, D. J., Malloy, L. C., & Lamb, M. E. (2011). The development of memory in
childhood. In M. E. Lamb, D. J. La Rooy, L C. Malloy, & C. Katz, (Eds.),
Children’s testimony: A handbook of psychological research and forensic
practice (2nd ed) (pp. 49-68). New Delhi, India: John Wiley & Sons, Ltd.
Lamb, M. E., Orbach, Y., Hershkowitz, I., Esplin, P. W., & Horowitz, D. (2007).
Structured forensic interview protocols improve the quality and informativeness
of investigative interviews with children: A review of research using the NICHD
Investigative Interview Protocol. Child Abuse and Neglect, 31(11-12), 1201-1231.
Lamb, M. E., Orbach, Y., Sternberg, K. J., Hershkowitz, I., & Horowitz, D. (2000).
Accuracy of investigators’ verbatim notes of their forensic interviews with alleged
child abuse victims. Law and Human Behavior, 24(6), 699-708.
61
Lamb, M. E., Sternberg, K. J., & Esplin, P. W. (1998). Conducting investigative
interviews of alleged sexual abuse victims. Child Abuse and Neglect, 22(8), 812823.
Lamb, M. E., Sternberg, K. J., Orbach, Y., Hershkowitz, I., Horowitz, D., & Esplin, P.
W. (2002). Supervision on the quality of investigative interviews with alleged sex
abuse victims. Applied Developmental Science, 6(3),
London, K., Bruck, M., Ceci, S.J., & Shuman, D. (2005). Children’s disclosure of sexual
abuse: What does the research tell us about the ways that children tell?
Psychology, Public Policy, & the Law, 11(1), 194-226.
Lyon, T. D., & Stolzenberg, S. N. (2014). Children’s memory for conversations about
sexual abuse: Legal and psychological implication. Roger Williams University
Law Review, 19(2), 411-450.
Malloy, L. C., Brubacher, S. P., & Lamb, M. E. (2013). “Because she’s one who listens”:
Children discuss disclosure recipients in forensic interviews. Child Maltreatment,
18(4), 245-251.
Malloy, L. C., La Rooy, D. J., Lamb, M. E., & Katz, C. (2011). Developmentally
sensitive interviewing for legal purposes. In M. E. Lamb, D. J. La Rooy, L C.
Malloy, & C. Katz, (Eds.), Children’s testimony: A handbook of psychological
research and forensic practice (2nd ed) (pp. 1-13). New Delhi, India: John Wiley
& Sons, Ltd.
Miller, J. B., Winstanely, P. A., & Carey, P. (1996). Memory for conversation. Memory,
4(6), 615-631.
62
Myers, J. E. B. (2011). Children’s disclosure statements as evidence in the United States
legal system. In M. E. Lamb, D. J. La Rooy, L C. Malloy, & C. Katz, (Eds.),
Children’s testimony: A handbook of psychological research and forensic
practice (2nd ed) (pp. 1-13). New Delhi, India: John Wiley & Sons, Ltd.
Neisser, U. (1981). John Dean’s memory: A case study. Cognition, 9(1), 102-115.
Osborne, J. W. (2002). Notes on the use of data transformations. Practical Assessment,
Research & Evaluation, 8(6). Retrieved November 10, 2014 from
http://PAREonline.net/getvn.asp?v=8&n=6
Poole, D. A., & Lamb, M. E. (1998). Talking to children. In D.A. Poole, & M. E. Lamb
(Eds.), Investigative Interviews of Children: A Guide to Helping Professionals
(pp. 153-180). Washington, D. C., U.S.: American Psychological Association.
Principe, G. F., Cherson, M., DiPuppo, J., & Schindewolf, E. (2012). Children’s natural
conversations following exposure to rumor: Linkages to later false reports.
Journal of Experimental Child Psychology, 113(3), 383-400.
Principe, G. F., Dipuppo, J., & Gammel, J. (2013). Effects of mothers’ conversation style
and receipt of misinformation on children’s event reports. Cognitive
Development, 28(3), 260-271.
Principe, G. F., & Schindewolf, E. (2012). Natural conversations as a source of false
memories in children: Implications for the testimony of young witnesses.
Developmental Review, 32(3), 205-223.
63
Quas, J. A., & Sumaroka, M. (2011). Consequences of legal involvement on child victims
of maltreatment. In M. E. Lamb, D. J. La Rooy, L C. Malloy, & C. Katz, (Eds.),
Children’s testimony: A handbook of psychological research and forensic
practice (2nd ed) (pp. 1-13). New Delhi, India: John Wiley & Sons, Ltd.
Reyna, V. F., & Kiernan, B. (1994). Development of gist versus verbatim memory in
sentence recognition: Effects of lexical familiarity, semantic content, encoding
instructions, and retention interval. Developmental Psychology, 30(2), 178-191.
Sams, M., Hari, R., Rif, J., & Knuutilla, J. (1993). The human auditory sensory memory
trace persists about 10s: Neuromagnetic evidence. Journal of Cognitive
Neuroscience, 5(3), 363-370.
Schank, R. C. (1982). Reminding and memory organization: An introduction of MOPs.
In W.G. Lehnert & M. H. Ringle (eds.), Strategies for natural language
processing (pp. 455-493). Hillside, NJ: Lawrence Erlbaum Associates.
Sperling, G. A. (1960). The information available in brief visual presentation.
Psychological Monographs: General and Applied, 74(11), 1-29.
Stafford, L., Burggraf, C. S., Sharkey, W. F. (1987). Conversational memory: The effects
of time, recall, mode, and memory expectancies on remembrances of natural
conversations. Human Communication Research, 14(2), 203-229.
Stafford, L., & Daly, J. A. (1984). Conversational memory: The effects of recall mode
and memory expectancies on remembrances of natural conversations. Human
Communication Research, 10(3), 379-402.
64
Stolzenberg, S. N., & Lyon, T. D. (2014). How attorneys question children about the
dynamics of sexual abuse and disclosure in criminal trials. Psychology, Public
Policy, and Law, 20(1), 19-30.
United States v. Libby, 461 F. Supp. 2d.12 (D.D.C. 2006).
U.S. Department of Health and Human Services. (2013). Child Maltreatment 2012.
Washington, DC: U.S. Government Printing Office.
U.S. House. Committee on the Judiciary House of Representatives. (2010). Federal Rules
of Evidence. Washington: Government Printing Office.
Zimmerman, D. W. (1995). Increasing the power of nonparametric tests by detecting and
downweighting outliers. Journal of Experimental Education, 64(1), 71-78.
65
Appendix A
Prerecorded Story Script
My dinosaur is a girl named Sally. Sally the dinosaur is 10 years old. She lives with her
brother, Max, and her sister, Tabatha, in a big blue house in the city. Her house is huge
because the dinosaurs are so big!
Sally, the dinosaur, is in high school. When she’s asked what she wants to do when she
grows up, she always says she wants to be a racecar driver. People normally laugh at
Sally when she tells them this because Sally is a dinosaur and dinosaurs don’t know how
to drive racecars. But Sally keeps telling people that she will drive a racecar someday and
when she does, she will win.
One day, when Sally the dinosaur was watching T.V., she saw an advertisement to win a
chance to drive a racecar. The 10th person to call the phone number listed at the bottom of
the screen would win. Sally instantly called the phone number that was listed on the T.V.
When she called, a lady picked up and said, “Congratulations! You are the 10th caller and
our winner!” Sally, the dinosaur, was so happy! She was finally going to get her change
to drive a racecar.
The next day, Sally went to the racetrack to drive the racecar. Sally the dinosaur put on a
jumpsuit, a helmet, and jumped right into the racecar. Sally couldn’t believe her dream
was finally coming true. When the light turned green, Sally took off as fast as she could.
Sally, the dinosaur, went extremely fast around the track. Before Sally knew it, the race
was over and Sally won! Sally has never been happier. Till this day, Sally the dinosaur
still drives racecars.
66
Appendix B
Target Conversation Script
Subject #:
Birthday:
Date:
Recorder Folder:
Age:
Time:
School:
Gender:
Ethnicity:
Interviewer Name:
Rapport Building
Get the child from their classroom. We want to make the child feel comfortable
interacting with you.
Hi (child’s name), my name is (our name). I’m from the University of Toledo and
I’m here to talk to you today is that okay? Great, we’re going to (describe location
of testing room).
Engage in small talk with the child while walking to the testing room. For example: I
really like your hairband, is pink your favorite color?
When in the testing room, transition into…
So, how is school going today?
That’s great! What is your favorite part of school and why?
Cool. Let’s talk a bit more about your day. Can you tell everything that you’ve done
so far today? Tell me what you did from the moment you woke up till right now.
Tell me everything you can remember.
Is there anything more you can tell me about your day so far?
Great! Can you tell me what you’re doing this afternoon?
That sounds like a lot of fun!
***Follow-up with “can you tell me more about that” if the child’s response is limited.
67
Introduction
So, (Child’s name) I’m here to talk to you today and see how big of an imagination
you have. Okay?
I want to make sure I remember everything we talk about today. So I’m going to use
this (point out recorder) to record everything we talk about. Okay?
(Child’s name), I have a really fun game to play with you today. It’s about
storytelling and making up stories. I’m going to put on my storytelling hat (put on
hat). Do you like my hat?
I like to wear this hat when I play the storytelling game.
Do you want to play the game with me?
Okay, let me tell you how to play the game. Here is my storytelling box (place box on
table). Inside my storytelling box is a bunch of objects to help us makeup stories.
What you do is close your eyes and pick two objects from the box. Then, we’ll work
together to create a story based on the two objects you pick. Doesn’t this sound like
fun!
Pre-Recorded Story
I played this game yesterday with my friend and she picked a dinosaur and a
racecar from my storytelling box. She created a really fun story about Sally, the
racecar driving dinosaur. Her story was so much fun that I want you to hear it! I
recorded her story and I’m going to play it for you. Okay?
Turn on Pre-recorded story on other recorder
Wow, I just love that story! Did you like it?
Child’s Story
Okay, let’s play the storytelling game together! Go ahead, close your eyes and pick
two objects from my storytelling box.
The child will always select a monkey plush and baseball.
Okay, let’s see what you picked. Wow, a monkey and a baseball! Do you think we
can make-up a story about a monkey and a baseball? I’ll help you get started, okay?
68
I’m going to ask you a bunch of questions to help you create the story. When I ask
you a question, you can create any answer you want. Remember, storytelling is most
fun when you use your imagination! Let’s get started!
First, is your monkey a girl or a boy?
What is your monkeys’ name?
(Repeated monkey’s name), okay. I really like that name! How old is your monkey?
Does your monkey go to school?
Can you tell me more about that?
How many brothers and sisters does your monkey have?
Can you tell me more about them?
Most monkeys live in the jungle. Does your monkey live in the jungle?
Does your monkey live a house?
Can you tell me more about the monkeys’ house?
Now that we know more about your monkey, let’s talk about the monkey and the
baseball.
Did your monkeys’ dad teach him(her) how to play baseball?
Is your monkey good at baseball?
Who does your monkey play baseball with?
What color is their team uniform?
How old was your monkey when s(he) started playing baseball?
Can you tell me more about that?
I bet your monkey can hit the baseball really far. Has your monkey hit more than
five homeruns?
Can you tell me more about that?
69
How does your monkey feel when s(he) hits homeruns?
How many games did the monkeys’ team win?
How does your monkey feel after winning a game?
Does your monkey get ice cream after each game to celebrate?
Does your monkey like ice cream?
What other sports does your monkey like to play?
Is there anything else you can tell me about your monkey?
**Follow up with “can you tell me more about that” when the child’s response is very
limited
Conclusion
Awesome! That was a really great story! I really liked playing my storytelling game
with you today! Did you have fun? I think it’s time to take you back to class.
Turn off tape recorder after the child answers this question.
Take the child back to their classroom.
70
Appendix C
Memory Test Script
Subject #:
Birthday:
Date:
Recorder Folder:
Age:
Time:
School:
Gender:
Interviewer Name:
Ethnicity:
Get the child from their classroom. Make the child feel comfortable on their way to the
testing room by build rapport by engaging in casual conversation with the child.
Rapport Building
Get the child from their classroom. We want to make the child feel comfortable
interacting with you.
Hi (child’s name), my name is (our name). I’m from the University of Toledo and
I’m here to talk to you today is that okay? Great, we’re going to (describe location
of testing room).
Engage in small talk with the child while walking to the testing room. For example: I
really like your hairband, is pink your favorite color?
When in the testing room, transition into…
So, how is school going today?
That’s great! What is your favorite part of school and why?
Cool. Let’s talk a bit more about your day. Can you tell everything that you’ve done
so far today? Tell me what you did from the moment you woke up till right now.
Tell me everything you can remember.
Is there anything more you can tell me about your day so far?
Great! Can you tell me what you’re doing this afternoon?
That sounds like a lot of fun!
***Follow-up with “can you tell me more about that” if the child’s response is limited.
71
1.) Addressing the Event
(Child’s name), did you talk to a storyteller last week (a while ago)? My friend is a
storyteller and s(he) likes to play storytelling games with a lot of people. S(he) wears
a big colorful hat when s(he) plays the storytelling game. Did you talk to my friend,
the storyteller?
Yes
No
DK
That’s great! We’re going to talk about what you did when the storyteller came to
visit you, but first, I want to learn a little more about you. I’m going to read you a
list of objects and I want you to tell me which hand you use with the object. Okay?
Give Handedness Measure
Great! (Child’s name), I’m going to use this (reference to the recorder) to record
everything you tell me about the time the storyteller came to visit you, okay?
Turn on the recorder.
Did you talk to the storyteller?
Do you remember what you did when s(he) visited?
That sounds like a lot of fun! I bet you guys talked about a lot of fun things. I’m
here today to learn about what happened during the storytelling game. Okay?
2.) Open Ended Question
I heard that the storyteller played a story for you that her friend made up. The
storyteller played it for you on her tape recorder. Do you remember this story?
Can you tell me the story exactly the same way you heard it? Start at the very
beginning and tell me everything you can, but don’t guess or make anything up, just
tell me the story exactly the same way you heard it.
Let the child retell the story.
Provide the child with up to four additional open-ended prompts such as…
72
Can you tell me anything else?
What else did s(he) say?
After the child is done retelling the story…
**If the child can’t remember anything after 2 prompts…say…
The storyteller told me it was a story about a dinosaur and a racecar, do you remember
that story? Can you tell me more about it?
Wow, that sounds like a fun story! Thank you for telling me about it!
Did you tell the storyteller a story?
Can you retell me your story exactly the same way you told the storyteller? Start at
the very beginning and tell me everything you can, but don’t guess or make
anything up, just tell me the same exact story you told the storyteller.
Let the child retell their story.
Provide the child with up to four additional open-ended prompts such as…
Can you tell me anything else?
What else happened?
After the child is done retelling the story…
**If the child can’t remember anything after 2 prompts…say…
The storyteller told me it you made up a story about a monkey and a baseball, do you
remember that story? Can you tell me more about it?
Wow, that's another great story! Thank you for telling it to me!
Did the storyteller ask you any questions when you were creating your story?
Do you remember any of the questions she asked you?
For every question the child recalls, say…
73
What did you tell the storyteller?
Provide the child with up to four additional open-ended prompts such as…
Can you tell me anything else?
What else happened?
3.) Recognition Test
That’s great! Thank you for telling me about the storytellers’ visit. Now, I have a
few more questions about the storytellers’ visit. I just want to make sure that I
know everything that happened. I need you to help me understand exactly what you
guys talked about during the storytelling game. Okay?
Get Recognition Test
4.) Ending the Interview
It sounds like you had a lot of fun when the storyteller came to visit! (Child’s name)
that’s all the questions I have for you today. Thank you for telling me about the
storytellers’ visit. I really enjoyed hearing about your stories!
End the recording.
Take the child back to their classroom.
74
Appendix D
Example Recognition Test1
Partner-Generated Utterances
(Child’s name), I want to make sure I understand everything the storyteller said to you. If
I say something that the storyteller said to you, I want you to say yes. If I say something
that the storyteller did not say to you, I want you to say no. Okay?
1.) Did the storyteller ask you, “What is the name of the monkeys’ baseball coach?”
[False]
Yes
No
DK
2.) Did the storyteller ask you, “How tall is your monkey?” [False]
Yes
No
DK
3.) Did the storyteller ask you, “Does your monkey like ice cream?” [True]
Yes
No
DK
4.) Did the storyteller ask you, “What is the color of your monkey’s hair?” [False]
Yes
No
DK
5.) Did the storyteller ask you, “Where does your monkey play baseball?” [False]
Yes
No
DK
6.) Did the storyteller ask you, “Does your monkey play any other sports?” [True]
Yes
No
DK
7.) Did the storyteller ask you, “Is your monkey good at baseball?” [True]
Yes
No
DK
8.) Did the storyteller ask you, “Does your monkey go to school?” [True]
Yes
No
DK
9.) Did the storyteller ask you, “What is your monkeys’ name?” [True]
Yes
No
DK
10.) Did the storyteller ask you, “Does you monkey like to pitch or hit?” [False]
Yes
No
DK
75
Self-Generated Utterances
Now, (child’s name) I have a few questions about things that you told the storyteller. If I
say something that you told the storyteller, I want to you say yes. But if I say something
that you did not tell the storyteller, then I want you to say no. Okay?
1.) Did you tell the storyteller your monkey is a [gist response to, Is your monkey a girl
or a boy]? [True]
Yes
No
DK
2.) Did you tell the storyteller your monkey feels [gist response to, How does your
monkey feel after winning a game]? [True]
Yes
No
DK
3.) Did you tell the storyteller that your monkeys’ “family goes to the baseball games?”
[False]
Yes
No
DK
4.) Did you tell the storyteller that your monkey “teaches other monkeys how to play
baseball?” [False]
Yes
No
DK
5.) Did you tell the storyteller that your monkey “plays baseball every day after school?”
[False]
Yes
No
DK
6.) Did you tell the storyteller your monkey plays baseball with [gist response to, Who
does your monkey play baseball with]? [True]
Yes
No
DK
7.) Did you tell the storyteller that your monkey “gets candy at the concession stand?”
[False]
Yes
No
DK
8.) Did you tell the storyteller that your monkey “had an arm injury?” [False]
Yes
No
DK
9.) Did you tell the storyteller your monkeys’ uniform is [gist response to, What color is
their team uniform]? [True]
Yes
No
DK
76
10.) Did you tell the storyteller your monkeys’ house is [gist response to, Can you tell me
more about the monkeys’ house]?” [True]
Yes
No
DK
Verbal Exchanges
(Child’s name), for this last part, I want to know exactly what the storyteller asked you
and what you told her (him). I’m going to ask you a few questions about things the
storyteller asked you and what you told her (him). I want you to say yes, if what I say is
exactly what happened. I want to you say no, if what I say did not happen. Okay?
1.) Did the storyteller ask you, “How old is your monkey?” and you told the storyteller,
“[child’s gist response].” Is that right? [True]
Yes
No
DK
2.) Did the storyteller ask you, “How old was your monkey when s(he) started playing
baseball?” and you told the storyteller, “[child’s gist response].” Is that right? [True]
Yes
No
DK
3.) Did the storyteller ask you, “How many homeruns has your monkey hit?” and you
told the storyteller, “[child’s response to Has your monkey hit more than five
homeruns?].” Is that right? [False]
Yes
No
DK
4.) Did the storyteller ask you, “Where does your monkey live?” and you told the
storyteller, “[child’s response to Does your monkey live in the jungle?]?” Is that right?
[False]
Yes
No
DK
5.) Did the storyteller ask you, “Does your monkey live a house?” and you told the
storyteller, “[child’s gist response].” Is that right? [True]
Yes
No
DK
6.) Did the storyteller ask you, “What does your monkey do after games to celebrate?”
and you told the storyteller, “[child’s response to Does your monkey get ice cream after
each game to celebrate].” Is that right? [False]
Yes
No
DK
7.) Did the storyteller ask you, “Did your monkeys’ dad teach him(her) how to play
baseball?” and you told the storyteller, “[child’s gist response].” Is that right? [True]
Yes
No
DK
77
8.) Did the storyteller ask you, “Has the team won more than 10 games?” and you told
the storyteller, “[child’s response to How many games did the monkeys’ team win?].” Is
that right? [False]
Yes
No
DK
9.) Did the storyteller ask you, “Does your monkey have one brother and two sisters?”
and you told the storyteller, “[child’s response to How many brothers and sisters does
your monkey have]. Is that right?
[False]
Yes
No
DK
10.) Did the storyteller ask you, “How does your monkey feel when s(he) hits
homeruns?” and you told the storyteller, “[child’s gist response].” Is that right? [True]
Yes
No
DK
1
Indications of true and false recognition items are labeled here for demonstration
purposes. True and false labeling did not appear on the actual recognition test for this
study.
78