Professional Documents
Culture Documents
Manuscript version of
Funded by:
• National Science Foundation
© 2015, American Psychological Association. This manuscript is not the copy of record and may not exactly
replicate the final, authoritative version of the article. Please do not copy or cite without authors’ permission.
The final version of record is available via its DOI: https://dx.doi.org/10.1037/lhb0000119
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Running head: CHILDREN’S PERFORMANCE ON GROUND RULES 1
Abstract
Ground rules, also called interview instructions, are included in investigative interviews with
children around the world. These rules aim to manage the expectations of children who are
typically unaccustomed to being questioned by adults who are naïve to the children’s
experiences. Although analog research has examined the efficacy of ground rules instruction, a
systematic analysis of children’s ability to respond appropriately to each of the rules has not been
reported. In the current study, we scored the accuracy of children’s (N = 501, 4 to 12 years)
responses to five ground rules practice questions (for example, “What is my dog’s name?”) and
two questions that asked whether they would follow the rules, and then assigned inaccurate
responses to one of several error categories. Few children answered every question correctly, but
their performance on individual questions was encouraging. As expected, there were marked
differences in children’s understanding across ground rules questions (especially among the
younger children), with “Don’t guess” and “Tell the truth” rules being the easiest to comprehend.
Together with evidence that ground rules instruction takes little time to deliver (typically 2 to 4
minutes) and is associated with improved accuracy in previous research, these findings support
the use of ground rules in investigative interviews of children 4 years and older.
experiences and scaffold conversation in ways that give children’s stories structure and content
(Wang, 2013). For example, adults encourage children to speculate in the face of difficult
questions and typically provide prompts that signal which events and details should be reported
(Principe, DiPuppo, & Gammel, 2013; Kelly & Bailey, 2013). During forensic interviews, this
narrative history may limit children’s ability to navigate the demands of being sole informants
about their experiences (Cordón, Saetermoe & Goodman, 2005; Hershkowitz, Lamb, Orbach,
Katz, & Horowitz, 2012; Lamb & Brown, 2006). More so than adults, children tend to guess
when the answer is unknown (Poole & White, 1993), respond to questions they do not
understand (Pratt, 1990), and fail to correct interviewers who say something wrong (Hunt &
Borgida, 2001). The extent to which children display these behaviors has been linked to
numerous factors, including meta-cognitive awareness (Roebers & Schneider, 2005), verbal
ability (Waterman & Blades, 2013), perceived situational demands (Howie & Dowd, 1996), and
the assumption that interviewers already possess knowledge of target events (Lamb, Orbach,
Warren, Esplin, & Hershkowitz, 2007; Mulder & Vrij, 1996; Waterman, Blades, & Spencer,
2004).
To prepare children for the unique challenges of a forensic interview, some interview
guidelines include ground rules (also called interview instructions; see Achieving Best Evidence
in Criminal Proceedings, Ministry of Justice, 2011; the National Institute of Child Health and
Human Development [NICHD] Protocol, Lamb, Orbach, Hershkowitz, Esplin, & Horowitz,
2007). Ground rules can serve a number of functions, including (a) orienting children to the
interview (i.e., explaining the interviewer’s role and the purpose of the interview; Roberts, Lamb
CHILDREN’S PERFORMANCE ON GROUND RULES 4
& Sternberg, 2004; Sternberg, Lamb, Davies, & Westcott, 2001; Teoh & Lamb, 2010), (b)
assessing testimonial competence (that is, determining whether children can distinguish the truth
from a lie and will promise to tell the truth; Evans & Lyon, 2012; Klemfuss & Ceci, 2012); (c)
misunderstandings (by instructing children to not guess and granting permission to correct the
interviewer; Roebers, Moga, & Schneider, 2001); and (d) inoculating children against leading
questions (Endres, Poggenpohl, & Erben, 1999; Saywitz & Moan-Hardie, 1994).
Research on the efficacy of ground rules has yielded mixed results, with instruction
rules prove more effective when children are given opportunities to practice responding to
instructions (rather than simply hearing the rules; Gee, Gregory, Pipe, 1999; Krackow & Lynn,
2010; Nesbitt & Markham, 1999; Saywitz & Moan-Hardie, 1994). For example, Peterson and
Grant (2001) manipulated whether children received simple instructions to say “I don’t know”
children’s response patterns to yes-no and multiple choice questions. Other studies have
produced more encouraging results by explicitly training children to identify and refute
misleading or unanswerable questions using context-based examples. For instance, Saywitz and
providing extensive preinterview instructions and practice (the children practiced answering
misleading questions about a training video while receiving feedback from an assistant), and
Saywitz, Snyder, and Nathanson (1999) found that similar strategies significantly increased the
likelihood that children would signal when they did not understand incomprehensible and
complex questions. (For reviews, see Lyon, 2010, and Saywitz, Camparo, & Romanoff, 2010.)
CHILDREN’S PERFORMANCE ON GROUND RULES 5
Despite some optimistic findings, there is considerable variability in the number, content,
and placement of ground rules across interview protocols. In the NICHD protocol, for example,
interviewers guide children through a series of practice questions and provide corrective
feedback when necessary (“If I ask you a question and you don’t know the answer, say ‘I don’t
know.’ For example, what is my sister’s name?”; see Lamb et al., 2007). The newly developed
CornerHouse Forensic Interview ProtocolTM includes “orienting messages” and instructions prior
to substantive questioning (“If I get something wrong, you can tell me.”; Anderson, 2012, p. 4)
but does not explicitly call for interviewers to deliver practice questions. Other protocols do not
include a distinct ground rules phase at all. The RATAC® protocol (Anderson et al., 2010) and
Step-Wise Interview Guidelines (Yuille, Cooper, & Hervé, 2009; a revision to the Step-Wise
Interview; Yuille, 1990) do not require instructions but rather suggest that interviewers reinforce
rules when children spontaneously admit lack of knowledge, correct interviewer mistakes, or
indicate lack of understanding (for example, by saying “Good job. Thank you for telling me you
don’t understand.”).
The rationale for down-playing or omitting ground rules instruction is fourfold. First is a
belief that given children’s limited attention spans, time is better spent eliciting case relevant
answer ground rule questions could damage children’s credibility and call their testimonial
competence into question (see Evans & Lyon, 2012; Klemfuss & Ceci, 2012). Third, as noted
previously, research on the efficacy of ground rules has produced mixed results. Finally,
skeptics have raised concerns that ground rules instruction is too abstract and includes practice
questions that are not developmentally appropriate for young children (who find them difficult to
answer when presented in a formal interview phase; Yuille et al., 2009). (See also Geddie, Beer,
CHILDREN’S PERFORMANCE ON GROUND RULES 6
Regarding this latter concern, research has yet to systematically map developmental
arguably a prerequisite for efficacy. To explore this concern, we examined children’s ability to
answer five practice questions representing frequently recommended and researched ground
rules (Lamb et al., 2007; Brubacher, Poole, & Dickinson, 2014) and two questions that asked
children to confirm they would use those rules during a subsequent interview. Importantly,
interviewers gave some children additional instruction after incorrect answers, which allowed us
to determine whether further training improved performance. Finally, we cataloged the types of
errors commonly made by children and calculated the time it took interviewers to deliver these
Method
Participants
During two eyewitness projects conducted in our laboratories, interviewers had delivered
rapport-building questions, a practice interview (about a typical day at school), and a ground
rules phase to each child prior to questions about the target events (Poole & Dickinson, 2011,
2014). For the current study, we revisited the interview transcripts from these projects to analyze
how the children had responded to the ground rules questions. The first sample consisted of 218
children who were 5 to 12 years of age when they returned to our laboratories for a follow-up
interview (Mage = 8.78 years) approximately 1 or 2 years after exposure to a target event (Poole
& Dickinson, 2011). Ground rules instruction during their first laboratory experience was
minimal, however, with interviewers asking only if they would tell the truth and would say when
they did not understand (without practice questions). The second sample consisted of 283
CHILDREN’S PERFORMANCE ON GROUND RULES 7
children who were 4 to 9 years of age when they were interviewed for an in-progress eyewitness
project (Mage = 6.60 years). Both samples included data collected at two research sites (one in a
small Midwestern town and one in the New York metropolitan region). From the original
transcript sets, we selected only interviews of children who spoke English in their homes.
The combined sample of 501 children (46.9% female) was predominantly Caucasian
(82.9%), with 4.2% of the children identified by their parents as African American; 4.2% as
Asian; and 8.6% as Native American, Pacific Islander, mixed race, or other (7.1% were
ethnically Hispanic). In addition to English, 7.8% of families said that another language was
Procedure
Assistants for the first sample began each session by obtaining parental consent followed
by child assent (for children less than 7 years) or consent (for children 7 years or older). They
then escorted individual children into a room to experience a second eyewitness event
immediately before the interview (see Poole & Dickinson, 2014). Assistants for the second
sample began sessions by greeting children and escorting them into the interviewing room. (The
consent process had been completed a week earlier, before the staged eyewitness event.) Forms
for both samples included a family demographics questionnaire (completed by a parent) and
debriefing consent forms that each child and parent signed after the interview.
building phase, and asked each child to describe a typical school day. Interviewers then
transitioned to the ground rules phase (“Now that I know you better, I want to talk about some
rules we have in this room. . .”) and delivered seven questions regarding instructions (a) not to
CHILDREN’S PERFORMANCE ON GROUND RULES 8
guess, (b) to inform the interviewer when they did not understand a question and (c) when the
interviewer said something wrong, and (d) to tell the truth. The wording of initial practice
questions was identical across samples, and all interviewers were instructed to try to elicit correct
because some interviewers for the first sample found it difficult to improvise after wrong
answers, particularly during “Don’t guess,” “Tell me when you don’t understand,” and “Tell me
when I say something wrong” instruction, the interview protocol for the second sample included
guidance about how to proceed after wrong answers to these questions, as described in Table 1.
Data Coding
The length of each ground rules phase was computed from time codes on the interview
recordings. To score each child’s answers to individual questions, two assistants assigned letters
that identified the content of the child’s first answer, last answer (if the first answer was not the
desired response and the interviewer redelivered the question one or more times), and the
number of follow-up prompts (not counting the initial question) delivered before the child’s final
answer. The assistants ignored dialog in which interviewers delivered unauthorized leading
questions to prompt a correct response (e.g., “So running is a lie, right?”) and scored children’s
The letter codes identified first and last answers as missing (when the interviewer skipped
or inappropriately reworded the question), the desired response (for example, “I don’t know” in
response to “What is my dog’s name?”), or some other response. Nine codes described these
other responses: (a) no verbal or nonverbal response, (b) the child’s last response was a
clarification question, (c) “yes” and (d) “no” responses when these were not the desired
responses (e.g., Interviewer: “While we are talking today, will you tell me when you don’t
CHILDREN’S PERFORMANCE ON GROUND RULES 9
understand?” Child: “No.”), (e) “I don’t know,” (f) “I don’t understand” or “What/Huh?”(when
this was not the desired response), (g) incorrect but answering the question (e.g., Interviewer:
“For example, what is my dog’s name?” Child: “Buddy?”), (h) irrelevant (e.g., Interviewer:
“Will you tell me when you don’t understand?” Child: “My brother picks, picks his boogers.”),
and (i) other. Unlike irrelevant responses, “other” responses were clearly related to the question
topic. For example, coders recorded “other” whenever children made a comment about a dog in
response to “What’s my dog’s name?” and when they began narrating true events in response to
Agreement for coding children’s first answers as the desired response or another response
ranged from 95.3% to 99.4% for the individual ground rules questions, Cohen’s kappas = .89 to
.98. Agreement for coding final answers (correct first answers and last answers following
redelivery of questions) ranged from 93.7% to 99.2%, Cohen’s kappas = .73 to .92. Agreement
for sorting undesired errors into the various error categories ranged from 87.8% to 97.0%,
Cohen’s kappas = .82 to .97. Finally, agreement on the exact number of prompts required to
elicit the final answer ranged from 96.4 to 99.6%, intraclass correlations = .83 to .99. Each
disagreement was resolved by discussion after the coders jointly reread the transcript.
Accuracy Scoring
Children’s responses were computer rescored as accurate when they provided desired
responses to the “Don’t guess” (any response meaning “I don’t know”), “Tell me when you don’t
understand” (any response indicating the child did not understand, including “huh?” and
“what?”), “Will you tell me when you don’t understand?” (yes), “Tell me when I say something
wrong” (any answer indicating that the interviewer misspoke or that the child was not the
erroneous age), and “Will you tell me the truth today?” (yes) prompts. The “true or not true
CHILDREN’S PERFORMANCE ON GROUND RULES 10
(truth)” question was scored two ways: Conservative scoring accepted only “true” or the
equivalent, whereas lenient scoring accepted either “true” or “yes.” The “true or a lie (a lie)”
question was also scored two ways: Conservative coding accepted only “a lie” or the equivalent,
Results
The duration of instruction ranged from 1.15 to 6.10 minutes, with 98% of interviews
containing ground rules instruction lasting 4 minutes or less (M = 2.06 minutes, SD = .70). The
was comparable across samples, we compared the percentages of 5- through 8-year-olds who
answered each question accurately with separate binomial logistic regressions, controlling for
age and sex. There were no significant differences across samples among children’s first
responses to the seven questions (using the conservative scoring of truth/lie questions, p values
adjusted using the False Discovery Rate method, Benjamini & Hockberg, 1995). Comparable
analyses on the accuracy of final answers (which sometimes followed additional instruction from
interviewers) yielded one significant difference: A higher percentage of children in the second
sample correctly answered the “Tell me when I say something wrong” question (unadjusted p <
.001). This finding was likely due to the fact that the interview protocol for the second sample
told interviewers what to say when children erred. To simplify data presentation, we collapsed
over samples to report performance on all questions except “Tell me when I say something
wrong.”
CHILDREN’S PERFORMANCE ON GROUND RULES 11
Because preliminary analyses found age trends on some questions even among older
children, we did not collapse over older age groups to describe children’s performance. For
example, there was a significant correlation among children 8 and older between age and
accuracy, controlling for sex, for first responses to the “Tell me when you don’t understand”
question, r = .16, p = .03, and “I am sitting. Is that true or a lie?”, r = .15, p = .03.
Accuracy of first and final responses. Table 2 lists the accuracy of children’s first
responses (i.e., the percentage of correct responses following the first delivery of each question;
left column) and their final performance (i.e., accurate first responses combined with final
responses after redelivery of questions; middle column). Because interviewers sometimes failed
to deliver additional instruction after wrong answers (or asked unscored leading questions), we
also computed the accuracy of children who either answered correctly in response to the first
delivery of a question or received additional instruction and redelivery of the question after a
wrong answer (right column). Point-biserial correlations at the bottom of each column report the
strength of relationships between age and accuracy, controlling for sex. (Data from the two
samples are presented separately for “Tell me when I say something wrong,” and performance
on the truth/lie questions is reported for both the conservative and lenient scoring methods.)
From the right column of Table 2, notice that few children failed to say “I don’t know”
after “Don’t guess” instructions, only the 4-year-olds erred more than 15% of the time in
response to “Tell me when I say something wrong,” and performance was also above 90% for all
age groups on questions that asked children to label statements as true or a lie. The 4- and 5-
year-olds were confused by instructions to tell interviewers when they didn’t understand, but
over 90% of the children in all age groups agreed to do so in response to “Will you tell me when
you don’t understand?” Finally, only the 4-year-olds were frequently confused by “Will you tell
CHILDREN’S PERFORMANCE ON GROUND RULES 12
me the truth today?”, although even the 5- to 8-year-olds often required additional explanation to
We were curious whether more than two attempts to elicit correct answers paid off,
especially for the younger children. Therefore, we examined whether accuracy remained near
zero or improved for children who erred on the first and second delivery of the question (using
the conservative scoring of truth/lie questions) but then received additional instruction. For the
six questions with cases meeting these criteria (the number of such cases for each question is in
parentheses), many children eventually passed the ground rule: Don’t guess (83.3% correct, n =
6), Tell me when you don’t understand (20.0%, n = 10), Will you tell me when you don’t
understand? (77.8%, n = 9), Tell me when I say something wrong (65.2%, n = 23), True or not
true (truth)? (50.0%, n = 2), and Will you tell me the truth today? (70.4%, n = 27). Furthermore,
even the youngest children benefitted from more effort on the part of interviewers. For example,
all three of the 4-year-olds who needed multiple tries to say “I don’t know” eventually did so,
and 37.5% of the 8 4-year-olds who had difficulty correcting the interviewer eventually
succeeded.
Typical Errors
The errors children produced provide insight into how they sometimes interpret ground
rules, which can suggest better procedures for delivering instructions. To this end, Table 3
reports the percentage of errors in response to the first delivery of each question that represented
the various error categories (top panel), along with the percentage of all first responses (accurate
and inaccurate) representing each error category (bottom panel). Frequent errors were as
follows:
Don’t guess. Children rarely guessed in response to this question (only 2.8% of the
CHILDREN’S PERFORMANCE ON GROUND RULES 13
current sample did), but inappropriate guessing was the most frequent type of
error when interviewers asked about something the children did not know.
“Other” responses were also a common type of error and typically involved
Tell me when you don’t understand. Children who erred most often answered “I don’t
know” when interviewers used a word they did not understand (rather than
answer the question. “Will you tell me when you don’t understand?” tended to
elicit no response or expressions of confusion from children who did not say
Tell me when I say something wrong. Among children who failed to grasp this ground
rule, “How do you like being (wrong age) years old?” frequently elicited an
Tell the truth. Most children provided desired responses to the truth-lie discussion, but
there was one recurring confusion: In response to “Will you tell me the truth
today?”, many children (over a third of those who made errors) responded by
In sum, performance was good when interviewers delivered additional instruction after
errors, with high pass rates among children 7 years and older on all questions (right column of
Table 2). The 6-year-olds also did remarkably well, struggling only with “Tell me when you
don’t understand” (erring on about 15% of these questions). The 4- and 5-year-olds performed
noticeably poorer, but even this age group typically said “I don’t know” after instruction to do
so, corrected interviewer mistakes, and correctly labeled statements as true or a lie. The most
CHILDREN’S PERFORMANCE ON GROUND RULES 14
challenging questions involved telling the interviewer when they did not understand a question
(with the majority of 4-year-olds and almost 35% of 5-year-olds failing this ground rule even
when interviewers gave them a second attempt) and, for the 4-year-olds, agreeing to tell the truth
(with about 27% of failing to complete this request even with an additional prompt to do so).
The perception among some forensic interviewers that children have difficulty answering
ground rules questions (e.g., Yuille et al., 2009) is not strongly supported by children’s
performance on the individual questions in Table 2. However, this perception may be based on
children’s ability to successfully complete an entire set of questions. Of course, the proportion
of children who pass all ground rules questions will vary across different question sets and ways
of wording these questions. Nonetheless, we can get a sense of what interviewers might face by
looking at the percentage of children who passed every question. To provide this information,
for each age group we computed the percentage of children in the second sample whose final
answers correctly answered all questions, using the lenient coding of truth/lie questions and
combining 8-year-olds with the few 9-year-olds. These percentages, from 4 to 8-9 years of age,
were 7.1%, 29.4%, 50.9%, 54.2 and 61.0%, respectively. Thus despite the fact that the children
least one.
For most questions and age groups, interviewer errors did not have a large impact on
accuracy rates. (In Table 2, sample size differences between the leftmost and rightmost columns
represent sessions in which interviewers did not deliver at least one prompt after a wrong
answer.) Over all age groups, the percentage of sessions in the second sample in which
interviewers should have prompted after the first incorrect answer but did not ranged from only
CHILDREN’S PERFORMANCE ON GROUND RULES 15
4% to 8% for “Don’t guess,” “Will you tell me when you don’t understand?,” “Tell me when I
say something wrong,” “True or not true (truth)?” “True or a lie (lie)?” and “Will you tell me the
truth today?” Interviewers made more mistakes delivering “Tell me when you don’t
understand.” For this question, they moved on prematurely during 23% of sessions, with most
mistakes (86%) occurring when children responded “I don’t know” and interviewers accepted
that answer. (In the field, “I don’t understand” prompts interviewers to clarify questions,
whereas “I don’t know” could lead to a loss of information.) Knowledge of this pattern, along
with the other errors reported in Table 3, can be used to script future instructions that better help
interviewers anticipate and respond to the typical answers children provide to ground rules
questions.
Discussion
children’s responses to five of the most frequently researched ground rule interview questions.
Overall, the greatest differences were observed between the 4- and 5-year-olds and between
these younger groups and children 6 years and older. Importantly, giving children additional
opportunities to respond increased accurate responses and reduced the magnitude of the
relationship between age and accuracy for all but the “Tell me when you don’t understand” rule.
Although these results do not mean that children who receive additional training will
impression of children’s abilities and demonstrates clear links between performance and the
CHILDREN’S PERFORMANCE ON GROUND RULES 16
cognitive complexity of the question. For example, over 80% of the youngest children in the
sample answered the “Don’t guess” question accurately on the first attempt, with no other
interview instruction producing such high accuracy. When children did err in response to this
question, it was most often by guessing (e.g., a dog’s name), thereby providing interviewers with
opportunities to correct this behavior. (Most children who initially erred—75%—correctly said
“I don’t know” in response to interviewers’ second attempt at this question.) The fact that nearly
all of the children in our sample answered correctly after more than one attempt supports the
practice of instructing children to not guess. (See Gee et al., 1999, and Saywitz & Moan-Hardie,
Following the “Don’t guess” practice question, the truth and lie questions were arguably
the easiest for children to comprehend. The marked difference between the 4- and 5-year-olds,
and again between 5-year-olds and older children, was evident for initial responses. After further
attempts by some interviewers, however, at least 90% of all children could say what was true and
what was a lie. This finding is not surprising given that most children can make the truth/lie
distinction by 4 years of age (for a review, see Talwar & Crossman, 2012). That improvements
were observed among younger children from the first to the last response is likely associated
with their ability to comprehend the purpose of the questions, rather than earlier failures to
distinguish between truth and lie (for further discussion of “identification” questions, see Evans
& Lyon, 2012). Examining children’s errors reveals that, for the truth question, the bulk of their
erroneous responses (60.7%) with conservative scoring were a result of saying “yes” to the
question “I’m sitting right now, is that true or not true?”, and more than a quarter of the errors to
the lie question were “no” responses (i.e., “You are running right now, is that true or a lie?”).
Only 17% of the errors to the lie question were indications of confusion (i.e., “Huh?” or “I don’t
CHILDREN’S PERFORMANCE ON GROUND RULES 17
know.”).
Despite good performance on the truth/lie questions, asking children if they would tell the
truth failed to elicit a “yes” response from over a quarter of 4-year-olds, and some older children
also did not reply appropriately. Evidently, children often interpreted the question as a request to
say things that were true rather than a request to agree to tell the truth. For example, when one of
the authors asked this question in an unrelated project, the child instantly (and self-consciously)
admitted that a statement he had made minutes earlier had not been entirely accurate, and he
proceeded to correct his previous statements. Although such linguistic confusions can be
resolved by additional explanation, it would be helpful to identify the best way to elicit this
promise. Lyon and Dorado (2008) found that “can you” questions (“Can you promise that you
will tell me the truth?”) often elicited “yes” or “no” (rather than a narrative) and that using this
false disclosures. A subsequent study of children’s ability to comprehend the words commonly
used to elicit an oath suggested a straightforward alternative: “Do you promise that you will tell
Instructions about failures in comprehension and interviewer errors were challenging for
children. When they erred on the “don’t understand” question (“Is my shirt gridelin?”), most of
the time they said they did not know, thus failing to explicitly tell interviewers that a
comprehension failure had occurred. In addition, 40% of errors were problematic from a
forensic perspective (a “yes” or “no” response). “Tell me when I say something wrong” elicited
an undesirable response just over one-third of the time. Again, “don’t know” (28.3%) and
implicit indications of confusion (14.5%) were common, suggesting that children recognized the
disruption in communication but had trouble verbalizing it (see Flavell, Speer, Green, & August,
CHILDREN’S PERFORMANCE ON GROUND RULES 18
1981, for other examples of implicit indications of comprehension failures). As we explain next,
there are two likely reasons why these questions were the most developmentally-challenging of
the set: (a) There are no obvious strategies for producing consistently accurate answers that
bypass the need for a firm grasp of theory-of-mind along with continuous meta-cognitive
monitoring, and (b) more advanced language skills are needed to understand and correctly
The basic theory-of-mind skills that are relevant to ground rules understanding (e.g., the
awareness that knowledge and memories are gained via experience, and the understanding that
others can hold different beliefs or perspectives than oneself; see Wellman & Liu, 2004, for task
descriptions and developmental trajectories) develop between the ages of 4 and 6 years of age
(Peterson, Wellman & Slaughter, 2012; Wellman & Liu, 2004). Correlations among
performance on various theory-of-mind tasks and the ability to understand each of the ground
rules have not been reported, but our findings of clear improvements between 4 and 6 years on
made by the interviewer and explaining that a miscommunication has occurred, likely relies on
more sophisticated metacognitive skills (Gee et al., 1999; Saywitz et al., 1999). In both
situations, children must first attend to the erroneous or ambiguous component of the question,
compare it to their memory or knowledge base, and then alert the interviewer to the problem;
therefore, explicitly-controlled processing is required (Schneider & Pressley, 2013). Why, then,
is the “Don’t guess” ground rule easier, given that it also requires that children reflect on the
quality of their memories in order to decide whether questions can be answered on the basis of
CHILDREN’S PERFORMANCE ON GROUND RULES 19
knowledge or experience?
There two plausible reasons why this question was answered accurately in our sample
and in other research (e.g., Endres et al., 1999; Moston, 1987). First, children can use a
familiarity-based heuristic to reject a question when an answer does not come quickly to mind,
even if they do not follow the rule strategically. Indeed, several experiments in which children
were trained to say “don’t know” found that an increase in this response was coupled with both
decreased errors and decreased accuracy (e.g., Gee et al., 1999; Saywitz & Moan-Hardie, 1994).
The second reason involves question format. In our study, the “Don’t guess” practice question
was posed in the “wh-“ format, and children are more likely to provide accurate “don’t know”
responses to these rather than yes/no questions (Waterman & Blades, 2011; Waterman, Blades,
The practice question for our “mistake” ground rule (“How do you like being [wrong age]
years old?”) was also posed as a wh- question but evidenced poorer performance than the “Don’t
guess” rule. Aside from the metacognitive challenges associated with the “mistake” rule, this
rule may have been difficult due to linguistic complexity. “How” is among the last wh- words to
be acquired (along with “when,” due to late development of temporal concepts; Ervin-Tripp
1970; Tyack & Ingram 1977). Indeed, our impression was that this question confused children
more than the follow-up prompt did, which did not commence with “how” (“What did you do at
the water park today?”). The initial question for this ground rule could also have been
interpreted as a hypothetical question (i.e., “How would you like to be [older wrong age]?” or an
inquiry about the past (“How did you like being [younger wrong age]?”). This caveat further
highlights the need for a careful and systematic analysis of the linguistic format of ground rules
questions. It would be beneficial for future research to directly compare question formats for
CHILDREN’S PERFORMANCE ON GROUND RULES 20
each ground rule to determine which is most effective at (a) conveying the purpose of the
question (and thus eliciting a correct response during the introductory phase) and (b) yielding
Individuals who are unenthusiastic about ground rules may argue that well-trained
interviewers avoid asking poor questions, rendering ground rules instruction unnecessary. Yet
even interviews comprised of a high proportion of open-ended prompts also include specific
questions (e.g., Sternberg et al. 2001), and it is often necessary to ask some specific questions of
younger children (Hershkowitz et al., 2012), who are the most vulnerable to the negative effects
of these questions.
Taken together, our results, along with evidence that ground rules practice questions can
improve children’s testimonial accuracy (Cordón et al., 2005; Gee et al., 1999; Krackow &
Lynn, 2010; Saywitz & Moan-Hardie, 1994; Saywitz et al., 1999), suggest that children are more
likely to comprehend ground rules and keep them in mind when extra effort is made to ensure
they can demonstrate competence. That is, simply telling children what to do is not particularly
effective (Ellis, Powell, Thomson, & Jones, 2003), and even older children may not respond
Because ground rules instruction with follow-up practice can be delivered in just 2 to 4
minutes for most children, and scripting this interview phase reduces the burden on interviewers,
we believe this is time well-spent. (See Lamb et al., 2007, for discussion of the value of scripted
protocols.) Relatedly, when children are engaged in responding to practice rules (rather than just
passively listening to them), additional opportunities for rapport-building are naturally created
advantage of opportunities to also reinforce ground rules later in the interview (e.g. Anderson,
2010; Yuille et al., 2009), as an interviewer in one of our projects recently did:
I: I see you understand what the truth is. It is important to tell me the truth—what really
I: Yeah, so when I ask you questions in the interview will you say, will you tell the truth
C: Yes.
I: Okay, good. And thank you for telling me you didn’t understand.
This example illustrates the logical problem with permitting interviewers to bypass an
initial ground rules phase and only address concepts as they come up: Interviewers would never
know when children did not know an answer or did not understand a question unless the children
did something to raise a concern. Therefore, as long as the evidence shows that children benefit
from initial instruction, and instruction takes a minimal amount of time (as we have shown), it
Conclusions
The techniques caregivers use to encourage children’s emerging narratives (e.g., co-
construction) are in direct contrast to the demands of a forensic interview. To empower children
to respond accurately, most experts now recommend including a ground rules phase (e.g., Lamb
et al, 2007; Saywitz et al., 2010). Further efficacy research is needed to resolve questions about
the optimal number and placement of ground rules in interview protocols, and we hope that
future best-practice standards include recommendations tailored for children of different ages
and levels of cognitive ability. As a preliminary step toward stronger evidence-based practice,
CHILDREN’S PERFORMANCE ON GROUND RULES 22
our data provide the first characterization of developmental differences in children’s ability to
understand ground rules and highlight the importance of providing children with multiple
References
www.cornerhouse.org/images/Protocol_Stage_Purpose_Approach.pdf
Anderson, J., Ellefson, J., Lashley, J., Miller, A. L., Olinger, S., Russell, A., Stauffer, J., &
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and
powerful approach to multiple testing. Journal of the Royal Statistical Society, Series B,
Brubacher, S. P., Poole, D. A., & Dickinson, J. J. (2014). The use of ground rules in interviews
with children: A synthesis and call for research. Manuscript under review.
Cordón, I. M., Saetermoe, C. L., & Goodman, G. S. (2005). Facilitating children’s accurate
responses: Conversational rules and interview style. Applied Cognitive Psychology, 19,
Ellis, L. M., Powell, M. B., & Thomson, D. M., & Jones, C. (2003). Do simple “groundrules”
Endres, J., Poggenpohl, C., & Erben, C. (1999). Repetitions, warnings, and video: Cognitive and
Ervin-Tripp, S. (1970). Discourse agreement: How children answer questions. In R. Hayes (Ed.),
Cognition and language learning (pp. 79-107). New York, NY: Wiley & Sons.
Evans, A. D., & Lyon, T. D. (2012). Assessing children’s competency to take the oath in court:
CHILDREN’S PERFORMANCE ON GROUND RULES 24
The influence of question type of children’s accuracy. Law and Human Behavior, 36,
Flavell, J. H., Speer, J. R., Green, F. L., & August, D. L. (1981). The development of
Society for the Research in Child Development, 46 (5, Serial No. 192).
Geddie, L. F., Beer, J., Bartosik, S, & Wuensch, K. L. (2001). The relationship between
Gee, S., Gregory, M., Pipe, M.-E. (1999). ‘What color is your pet dinosaur?’ The impact of pre-
interview training and question type on children’s answers. Legal and Criminological
psychological research and forensic practice (pp. 109-128). Chichester, UK: Wiley.
Hershkowitz, I., Lamb, M. E., Orbach, Y., Katz, C., & Horowitz, D. (2012). The development of
interviews about child abuse. Child Development, 83, 611-622. doi: 10.1111/j.1467-
8624.2011.01704.x
Howie, P. M., & Dowd, H. J. (1996). Self-esteem and the perceived obligation to respond:
10.1111/j.2044-8333.1996.tb00318.x
Hunt, J. S., & Borgida, E. (2001). Is that what I said? Witnesses’ responses to interviewer
Kelly, K. R., & Bailey, A. L. (2013). Dual development of conversational and narrative
Klemfuss, J. Z., & Ceci, S. J. (2012). Legal and psychological perspectives on children’s
10.1016/j.dr.2012.06.005
Krackow, E., & Lynn, S. J. (2010). Event report training: An examination of the efficacy of a
Lamb, M. E., & Brown, D. A. (2006). Conversational apprentices: Helping children become
Lamb, M. E., Orbach, Y., Hershkowitz, I., Esplin, P. W., & Horowitz, D. (2007). A structured
interviews with children: A review of research using the NICHD Investigative Interview
Lamb, M. E., Orbach, Y., Warren, A. R., Esplin, P. W., & Hershkowitz, I. (2007). Enhancing
(Vol. 1):Memory for events (pp. 429 - 451). Mahwah, NJ: Erlbaum.
Haralambie (Eds.) Child welfare law and practice (2d ed.) (pp. 87-109). Denver, CO:
Bradford.
CHILDREN’S PERFORMANCE ON GROUND RULES 26
Lyon, T. D., & Dorado, J. S. (2008). Truth induction in young maltreated children: The effects of
oath-taking and reassurance on true and false disclosures. Child Abuse & Neglect, 32,
Lyon, T. D., & Evans, A. D. (2014). Young children’s understanding that promising guarantees
performance: The effects of age and maltreatment. Law & Human Behavior, 38, 162-
interviewing victims and witnesses, and guidance on using special measures. Retrieved
from http://www.justice.gov.uk/downloads/victims-and-witnesses/vulnerable-
witnesses/achieving-best-evidence-criminal-proceedings.pdf
Moston, S. (1987). The suggestibility of children in interview studies. First Language, 7, 67-78.
doi: 10.1177/014272378700701904
Mulder, M. R., & Vrij, A. (1996). Explaining conversational rules to children: An intervention
study to facilitate children’s accurate responses. Child Abuse & Neglect, 29, 623-631.
doi: 10.1016/0145-2134(96)00050-6
Nesbitt, M., & Markham, R. (1999). Improving young children’s accuracy of recall for an
10.1016/S0193-3973(99)00027-1
Peterson, C., & Grant, M. (2001). Forced-choice: Are forensic interviewers asking the right
10.1037/h0087134
Peterson, C. C., Wellman, H. M., & Slaughter, V. (2012). The mind behind the message:
Advancing theory-of-mind scales for typically developing children, and those with
CHILDREN’S PERFORMANCE ON GROUND RULES 27
deafness, autism, and Asperger syndrome. Child Development, 83, 469-485. doi:
10.1111/j.1467-8624.2011.01728.x
Poole, D. A., & Dickinson, J. J. (2011). Evidence supporting restrictions on uses of body
diagrams in forensic interviews. Child Abuse & Neglect, 35, 659-669. doi:
org/10.1016/j.chiabu.2011.05.004
Poole, D. A., & Dickinson, J. J. (2014). Comfort drawing during investigative interviews:
Evidence of the safety of a popular practice. Child Abuse & Neglect, 38, 192-201.
doi:10.1016/j.chiabu.2013.04.012
Poole, D. A., & White, L. T. (1993). Two years later: Effect of question repetition and retention
Pratt, C. (1990). On asking children—and adults—bizarre questions. First Language, 10, 167-
Principe, G. F., DiPuppo, J., & Gammel, J. (2013). Effects of mothers’ conversation style and
Roberts, K. P., Lamb, M. E., & Sternberg, K. J. (2004). The effects of rapport-building style on
children’s reports of a staged event. Applied Cognitive Psychology, 18, 189-202. doi:
10.1002/acp.957
Roebers, C. M., Moga, N., & Schneider, W. (2001). The role of accuracy motivation on
children’s and adults’ event recall. Journal of Experimental Child Psychology, 78, 313-
10.1016/j.cogdev.2005.05.006
Saywitz, K., Camparo, L. B., & Romanoff, A. (2010). Interviewing children in custody cases:
Implications of research and policy for practice. Behavioral Sciences and the Law, 28,
Saywitz, K. J., & Moan-Hardie, S. (1994). Reducing the potential for distortion of childhood
Saywitz, K. J., Snyder, L., & Nathanson, R. (1999). Facilitating the communicative competence
doi: 0.1207/s1532480xads0301_7
Schneider, W. & Pressley, M. (2013). Memory development between 2 and 20 (2nd ed.).
State of Michigan Governor’s Task Force on Child Abuse and Neglect and Department of
Sternberg, K. J., Lamb, M. E., Davies, G. M., & Westcott, H. L. (2001). The memorandum of
good practice: Theory versus application. Child Abuse and Neglect, 2, 669–681. doi:
10.1016/S0145-2134(01)00232-0
Talwar, V., & Crossman, A. M. (2012). Children’s lies and their detection: Implications for child
Teoh, Y. -S., & Lamb, M. E. (2010). Preparing children for investigative interviews: Rapport-
building, instruction, and evaluation. Applied Developmental Science, 14, 154–163. doi:
10.1080/10888691.2010.494463
CHILDREN’S PERFORMANCE ON GROUND RULES 29
Tyak, D., & Ingram, D. (1977). Children’s production and comprehension of questions. Journal
Wang, Q. (2013). The autobiographical self in time and culture. New York, NY: Oxford
University Press.
Waterman, A. H., & Blades, M. (2011). Helping children correctly say “I don’t know” to
10.1037/a0026150
Waterman, A. H., & Blades, M. (2013). The effect of delay and individual differences on
doi: 10.1037/a0028354
Waterman, A. H., & Blades, M., & Spencer, C. (2000). Do children try to answer nonsensical
10.1348/026151000165652
Waterman, A. H., Blades, M., & Spencer, C. (2001). Interviewing children and adults: The effect
of question format on the tendency to speculate. Applied Cognitive Psychology, 15, 521-
Waterman, A. H., Blades, M., & Spencer, C. (2004). Indicating when you do not know the
answer: The effect of question format and interviewer knowledge on children’s ‘don’t
10.1348/0261510041552710
Wellman, H. M., & Liu, D. (2004). Scaling of theory-of-mind tasks. Child Development, 75,
Yuille, J. C., Cooper, B. S., & Hervé H. F. (2009). The step-wise guidelines for child interviews:
CHILDREN’S PERFORMANCE ON GROUND RULES 30
Yuille, J. C., Hunter, R., Joffe, R., & Zaparniuk, J. (1993). Interviewing children in sexual abuse
Understanding and improving testimony (pp. 95-115). New York, NY: Guilford Press.
Running head: CHILDREN’S PERFORMANCE ON GROUND RULES 31
Table 1
Question Instructions
Don’t Guess. Now that I know you better, I want to talk about some rules we have in this room.
One rule is that we don’t guess. If I ask a question and you don’t know the answer,
just say, “I don’t know.” For example, what is my dog’s name?
Right: That’s right; you don’t know my dog’s name, so “I don’t know” is
the right answer.
Wrong: Do you really know my dog’s name? If you don’t know the answer,
just say, “I don’t know.” Let’s try again. What is my sister’s name?
Tell me when you don’t understand. Another rule is that if I say something you don’t understand, you should tell me
you don’t understand. For example, is my shirt gridelin?
Right: Thank you for telling me you didn’t understand. I’ll ask a different
way. What color is my shirt? (Wait for response.) Good.
Will you tell me when you don’t While we are talking today, will you tell me when you don’t understand?
understand?
CHILDREN’S PERFORMANCE ON GROUND RULES 32
Tell me when I say something. Sometimes people say something wrong by mistake. I want you to tell me if I say
wrong. something wrong. For example, how do you like being (wrong age) years old?
Right: That’s right−you’re not (wrong age) years old, so I’m glad you told
me.
Wrong: But you are not (wrong age) now, right? I made a mistake when I
called you (wrong age; slap head). It’s okay to tell me if I say something
wrong. Let’s try another one. What did you do at the water park today?
(Wait for response.) That’s right, you didn’t go to a water park today, so
I’m glad you told me.
True or not true (true)? It is also very important to tell the truth today. First I need to make sure you know
what the truth is. I’m sitting down right now. Is that true or not true? (Wait for
response.) That’s right; I am sitting down, so sitting down is the truth.
True or a lie (a lie)? You are running right now. Is that true or a lie? (Wait for response.) That’s right,
you are not running, so running is a lie.
Will you tell me the truth today? I see you understand what the truth is. It is important to tell me the truth—what
really happened. Will you tell me the truth today?
Note. Scripted instructions for how to proceed after wrong answers did not appear in the protocol for the first sample. Adapted
from State of Michigan Governor’s Task Force on Child Abuse and Neglect and Department of Human Services (2011).
CHILDREN’S PERFORMANCE ON GROUND RULES 33
Table 2
Sample 1: Tell me when I say something wrong: “How do you like being [wrong age] years old?”
4
5 14 42.9 [21.4, 67.4] 50.0 [26.8, 73.2] 9 77.8 [45.3, 93.7]
6 33 72.7 [55.8, 84.9] 75.8 [59.0, 87.2] 27 92.6 [76.6, 97.9]
7 36 61.1 [44.9, 75.2] 63.9 [47.6, 77.5] 26 88.5 [71.0, 96.0]
8 36 83.3 [68.1, 92.1] 91.7 [78.2, 97.1] 33 100.0 [89.6, 100.0]
9 35 91.4 [77.6, 97.0] 91.4 [77.6, 97.04] 32 100.0 [89.3, 100.0]
10 35 97.1 [85.5, 99.5] 97.1 [85.5, 99.5] 34 100.0 [89.8, 100.0]
11-12 29 89.7 [73.6, 96.4] 89.7 [73.6, 96.4] 27 96.3 [81.7, 99.3]
Sample 2: Tell me when I say something wrong: “How do you like being [wrong age] years old?”
True or not true (true): “I’m sitting down right now. Is that true or not true?” (“True” = correct)
True or not true (true): “I’m sitting down right now. Is that true or not true?”
(“True” or “Yes” = correct)
True or a lie (a lie): “You are running right now. Is that true or a lie?” (“A lie” = correct)
True or a lie (a lie): “You are running right now. Is that true or a lie?” (“A lie” or “No” = correct)
Note. Sample sizes are not equivalent across questions (left column) due to missing data from interviewer
errors. a Percentage correct in response to final prompted questions represents the percentage of accurate
responses among children who responded correctly the first time the question was delivered and those who
received at least one redelivery of the protocol question (i.e., omitting inaccurate children who did not receive
the question again). Correlations report the association between age and accuracy with sex controlled.
* p < .05, ** p < .01, *** p < .001.
CHILDREN’S PERFORMANCE ON GROUND RULES 38
Table 3
Percentage of Errors in Each Error Category by Ground Rules Question (First Delivery of the Question)
Don’t
understand/
Undesired No I don’t Huh? Asks a
Question (undesired response) response response Yes No know What? Irrelevant question Other
Don’t guess (guesses a name) 46.7 13.3 0.0 0.0 0.0 10.0 0.0 0.0 30.0
True or not true? (“not true”) 23.2 1.8 60.7 3.6 5.4 0.0 0.0 0.0 5.4
True or a lie? (“true”) 41.5 2.4 0.0 26.8 7.3 9.8 0.0 0.0 12.2
Don’t guess (guesses a name) 2.8 .8 0.0 0.0 0.0 .6 0.0 0.0 1.8
CHILDREN’S PERFORMANCE ON GROUND RULES 39
True or not true? (“not true”) 2.6 .2 6.8 .4 .6 0.0 0.0 0.0 .6
Note. For “True or not true?” and “True or a lie?”, responses of “yes” and “no,” respectively, are categorized as errors to report the
frequency of these responses. For these questions, Table 2 lists the percentage of accurate responses for both a lenient coding method
(which treated “yes” as equivalent to “true” and “no” as equivalent to “lie”) and this conservative coding method.