ERIC EJ941696: Effects of Interviewer Behavior on Accuracy of Children's Responses

2011·0.5 MB·English

by  ERIC

JOURNALOFAPPLIEDBEHAVIORANALYSIS 2011, 44, 587–592 NUMBER3 (FALL2011)
EFFECTS OF INTERVIEWER BEHAVIOR ON ACCURACY OF
CHILDREN’S RESPONSES
JESSICA SPARLING, DAVID A. WILDER, JENNIFER KONDASH,
MEGAN BOYLE, AND MEGAN COMPTON
FLORIDAINSTITUTEOFTECHNOLOGY
Previousresearchhasshownthatcertaininterviewerbehaviorscanevokeinaccurateanswersby
children.Inthecurrentstudy,weexaminedtheeffectsofapprovinganddisapprovingstatements
on the accuracy of 3 children’s answers to questions in an interview (Experiment 1). We then
evaluated3questioningtechniquesthatmaybeusedbyinterviewersduringaforensicinterview
inwhichachildprovideseyewitnesstestimony(Experiment2).Allparticipantsrespondedwith
more inaccurate answers when approving statements followed inaccurate information and
disapprovingstatementsfollowedaccurateinformationinExperiment1.DuringExperiment2,
1participantrespondedmostinaccuratelywhenshewasrequestionedafterprovidinganinitial
answer,whereastheremaining2participantsrespondedmostinaccuratelywhentheinterviewer
providedcowitness information andsuggestive questions.
Key words: accuracy of children’s responses, eyewitness testimony, forensic interviews,
interviewerbehavior
_______________________________________________________________________________
Evaluation of the accuracy of children’s Garven et al. (1998) examined the effects of
reportshasbecomeanimportanttopic(Saywitz suggestive questions, cowitness information,
& Camparo, 2009). Some research on child approving and disapproving statements, and
interviews has focused on how the behavior of requestioning (i.e., re-presenting a question
the interviewer influences children’s accuracy. after an initial answer) on children’s accuracy
Garven, Wood, Malpass, and Shaw (1998) in interviews. One week after a classroom visit
suggested that improper interviews have the by a graduate student, children were exposed to
potential to evoke false information from all five of the above techniques or only to
children and described four broad categories suggestivequestionsduringinterviewsaboutthe
of improper interview techniques: suggestive visit. Children exposed to the five interview
questioning (e.g., asking ‘‘Did he touch her on techniquesweremorelikelytoanswerquestions
the bottom?’’ when a child has acknowledged inaccurately; however, it was not possible to
touching but has not mentioned any inappro- determine which of the five techniques had the
priate touching), influence (e.g., telling a child greatesteffectonthechildren’sresponding.Ina
what a cowitness has said or has seen about a follow-up study, Garven, Wood, and Malpass
topic), reinforcement and punishment (i.e., (2000) examined the effects of approving
delivery of approving and disapproving state- statements, cowitness information, and sugges-
ments contingent on a child’s report), and tive questioning on children’s accuracy. The
removal from direct experience (e.g., inviting a children responded inaccurately most often
child to speculate about what might have when the interviewer delivered approving
occurred). statements contingent on the child providing
inaccurate information.
Despite the importance of this topic, few
Correspondence concerning this article should be
addressed toJessica Sparlingor David A.Wilder, Florida studies have examined these interviewing tech-
Institute of Technology, School of Psychology, 150 W. niques individually, and only one study (which
University Blvd., Melbourne, Florida 32901 (e-mail:
focused on suggestive questioning) has been
[email protected] [email protected]).
doi:10.1901/jaba.2011.44-587 conducted at the individual level using single-
587
588 JESSICA SPARLING et al.
subject methodology (Doepke, Henderson, & took place over several days. Accurate responses
Critchfield, 2003). Repeated measurement of were defined as a ‘‘yes’’ or ‘‘no’’ response to a
the dependent variable, which is characteristic question that required the corresponding an-
of single-subject methodology, may be partic- swer. Inaccurate responses were defined as a
ularly relevant for this topic; children who ‘‘yes’’ or ‘‘no’’ response to a question for which
participateaseyewitnessesinlegalinvestigations the opposite answer would be appropriate. If a
may be asked to recount events multiple times participant answered ‘‘I don’t know’’ to a
(Zaragoza, Graham, Hall, Hirschman, & Ben- question, that question and answer were
Porath,1995).Thus,thepurposeofthecurrent excluded from the analysis (one to two
study was to examine further the effect of questions per session were excluded for each
interviewing techniques on children’s accuracy. participant). In Experiment 1, reinforcement
In the first experiment, we evaluated the effects was defined as saying ‘‘good’’ and smiling.
of approving and disapproving statements. The Punishment was defined as saying ‘‘that’s not
effects of suggestive questions, cowitness infor- right’’ and frowning. A neutral response was
mation, and requestioning were evaluated in a defined as a quick nod and saying ‘‘okay’’ with
second experiment. no vocal inflection. General questions, which
were used in both experiments, were defined as
METHOD yes or no questions that contained information
abouteventsinthevideos(e.g.,askingifScooby
Participants, Setting, and Materials
fell, if he had fallen). In Experiment 2,
Threechildrenparticipated:Matthew(4years
suggestive questions were defined as yes or no
old), Owen (5 years old), and Abby (8 years
questions that contained information about
old).Allchildrenhadage-appropriatelanguage
events that did not occur in the videos (e.g.,
skills, had no disabilities or psychiatric diagno-
asking ‘‘Didn’t Scooby fall?’’ whenhe had not).
ses,andwere ingeneral educationclassroomsat
Cowitness information was defined as a ques-
their schools. All sessions took place at a
tion preceded by the statement ‘‘Someone told
university in a room that contained a computer
me that …’’ Requestioning was defined as an
(to watch videos) and a table and two chairs.
interviewer repeating a question twice after an
Videos were obtained from various online
answer had been given. In this condition, the
resources and consisted of clips of children’s
participant’s answer to the final question was
cartoons (e.g., Scooby Doo) that were 3 to
scored.
5 min in length. Up to 20 questions and
Trained graduate students scored responses
corresponding correct answers based on events
(either accurate or inaccurate) on data sheets.
in each video were prepared in advance by a
Interobserver agreement data were collected for
graduate student. Videos were randomly as-
atleast33%ofresponsesforeachconditionper
signed to the various conditions of the studies;
participant by having a second observer collect
thatis, each video hadan equal chanceof being
used in each condition of each study. Tto data from video or audio recordings of
ensurethatrepeatedexposuretoavideodidnot interviews. Exact agreement was calculated on
affect responding, participants watched each a trial-by-trial basis by dividing the number of
video only once. agreements by the number of agreements plus
disagreements and multiplying by 100%.
Response Measurement Agreement ranged from 95% to 100% in
The dependent variable was the accuracy of Experiment 1 and 94% to 98% in Experiment
participants’ responses to interviewer questions. 2. Independent variable integrity was deter-
Participants were interviewed in sessions that mined by recording the number of correct and
EFFECTS OF INTERVIEWER BEHAVIOR 589
incorrect interviewer responses or deliveries answers were followed by punishment, as
(i.e., delivery of reinforcement, punishment, definedabove;inaccurateanswerswerefollowed
or a neutral response in Experiment 1 and by a neutral response, as defined above. In the
delivery of the question in the general, reinforcement plus punishment conditions,
suggestive, cowitness, or requestioning format reinforcement was delivered contingent on
in Experiment 2), dividing correct responses by inaccurateanswers and punishment wasapplied
total responses, and multiplying by 100%. to accurate answers.
Independent variable integrity ranged from Experiment 2. Procedures for Experiment 2
94% to 98% during Experiment 1 and 97% were identical to those of Experiment 1, with
to 100% for Experiment 2. the exception of the techniques evaluated. That
is, participants were shown the video and were
Procedure
subsequently interviewed by the experimenter.
Experiment 1. At the start of Experiment 1, Interviewswereprecededbythedirection‘‘Iam
participantsweretoldtowatchallvideosclosely going to ask you some questions. Please answer
because they would be asked some questions. ‘yes’ or ‘no,’ and do the best you can.’’ Three
For each session, participants were shown the different interviewer techniques were evaluated
video (described above) and were subsequently in Experiment 2: (a) suggestive questions, (b)
interviewed by the experimenter. Interviews cowitness information, and (c) requestioning.
were preceded by the direction ‘‘I am going to During baseline, the interviewer asked general
ask you some questions. Please answer ‘yes’ or questions and responded neutrally to all of the
‘no,’ and do the best you can.’’ Four different child’sanswers.Duringthesuggestivequestions
conditions were implemented: control, rein- condition and cowitness information condi-
forcement for inaccurate responding, punish- tions, the interviewer asked eight suggestive
ment for accurate responding, and reinforce- questions per session for Matthew and Abby
ment plus punishment. The interviewer, who (due to time constraints) and 14 suggestive
was the same person throughout both experi- questions per session for Owen. In the baseline
ments, wore a different colored shirt for each and requestioning conditions, the interviewer
condition to assist participants to discriminate asked 12 questions per session for Abby and
between conditions. The effect of the different Matthew and 20 for Owen. Effects of interview
consequences on response accuracy was com- techniques were evaluated using an ABACA-
pared using a multielement design. DABACAD reversal design for Owen and
General questions were asked during all Matthew and an ABCDABCDdesign for Abby
conditions, with a total of 12 questions asked (due to time constraints), in which A was
per session for Matthew and Abby (due to time baseline, B was the suggestive questions condi-
constraints), and 20 questions per session for tion, C was the cowitness information condi-
Owen. In the control condition, reinforcement tion, and D was the requestioning condition.
was delivered for each response, independent of
the accuracy of the response. Thus, one would
RESULTS AND DISCUSSION
expect to see lower (but not nonexistent) levels
ofinaccurateresponsesinthisconditionthanin Results of Experiment 1 are depicted in
the other conditions. In the reinforcement for Figure 1 (left). The highest levels of inaccuracy
inaccurate responding condition, all accurate occurred during the reinforcement plus punish-
answerswerefollowedbyaneutralresponseand ment conditions for all participants. In addi-
inaccurate answers were followed by reinforce- tion, the control condition produced relatively
ment, as defined above. In the punishment for low levels of inaccuracy for all participants.
accurate responding condition, all accurate Matthew responded inaccurately to a mean of
590 JESSICA SPARLING et al.
33% (range, 8% to 50%) of questions in the techniques. This is important, because the
reinforcementpluspunishmentconditionanda majority of prior studies on this topic have
meanof8%(range,0%to17%)ofquestionsin not been conducted in a manner that allows
the control condition. Owen responded inac- such analysis. Finally, in previous studies
curatelytoameanof53%(range,37%to65%) neither suggestive questioning (Garven et al.,
of questions in the reinforcement plus punish- 1998) nor providing cowitness information
ment conditions, and a mean of 20% (range, (Garven et al., 2000) were found to increase
10% to 39%) of questions in the control inaccuracy substantially. In contrast, results of
condition. Abby responded inaccurately to a the current study suggest that both of these
meanof58%(range,25%to92%)ofquestions techniques may increase inaccuracy for some
in the reinforcement plus punishment condi- children. The methodological differences be-
tionandameanof17%(range,8%to25%)of tween the current study and previous research
questions in the control condition. may account for this finding.
Results of Experiment 2 are depicted in Of course, the methodology employed in the
Figure 1 (right). Matthew responded most current study is different from a forensic
inaccurately in the suggestive questioning interview in which a child provides an eyewit-
condition (M 5 81%; range, 50% to 100%) ness testimony. In the current study, children
and cowitness information condition (M 5 were asked about events they observed in
69%; range, 38% to 100%). Matthew’s per- cartoons. In actual forensic interviews, children
centages of inaccurate answers in the reques- are often asked questions about potentially
tioningandbaselineconditionswere4%(range, traumatic events they observed in real life.
0% to 10%) and 3% (range, 0% to 10%), Nevertheless, for ethical and practical reasons,
respectively. Owen was most inaccurate in the questions about fictional characters and events
cowitness information condition (M 5 52%; arestandardinresearchoneyewitnesstestimony
range, 14% to 71%) and the suggestive (Ceci & Bruck, 1995). Also, although the
questioning condition (M 5 49%; range, dichotomous questions and the interview tech-
21% to 100%). Owen’s percentages of inaccu- niques employedin this study are not typical of
rate answers in the requestioning and baseline forensic interviews, they are sometimes used
conditions were 6% (range, 0% to 11%) and (Garven et al., 1998; Saywitz & Camparo,
15% (range, 0% to 61%), respectively. Abby’s 2009). An additional limitation is the small
highest levels of inaccurate responding occurred sample size; future research should employ
in the requestioning condition (M 5 87%; more participants.
range, 57% to 100%), relative to suggestive Children in this study ranged in age from 4
questioning (M 5 20%; range, 0% to 50%), to 8 years. It is likely that some developmental
co-witnessinformation(M525%;range,12% differences in language skills exist among this
to 63%), and baseline conditions (M 5 53%; age range that might affect accuracy. Indeed,
range, 17% to 100%). forensic interview guidelines reflect this; recom-
These results suggest that both antecedents mendations for younger children call for more
(i.e., the structure of questions) and conse- specific questions than for older children
quences (i.e., reinforcement and punishment) (Saywitz & Camparo, 2009). In Experiment 2
may influence the accuracy of children’s of the current study, the oldest participant
responding to an interviewer’s questions. Fur- responded most inaccurately when the inter-
ther, the differential responding observed across viewer re-questioned her after an initial answer,
participants suggests that some children may be whereas the younger children responded most
more or less sensitive to specific interviewer inaccurately in other conditions. Future re-
EFFECTS OF INTERVIEWER BEHAVIOR 591
Figure 1. Percentage of inaccurate responses during control, reinforcement for inaccurate responding (Sr),
punishmentforaccurateresponding(Sp),andreinforcementforinaccuratepluspunishmentforaccurateresponding(Sr
+ Sp) conditions for Matthew, Owen, and Abby in Experiment 1(left). Percentage of inaccurate responses during
baseline(BL),suggestivequestions(SQ),cowitnessinformation(CW),andrequestioning(RQ)conditionsforMatthew,
Owen,andAbbyinExperiment2 (right).
592 JESSICA SPARLING et al.
search should investigate additional age-related Garven, S., Wood, J. M., Malpass, R. S., & Shaw, J. S.
(1998). More than suggestion: The effects of
differences in sensitivity to interviewer tech-
interviewing techniques from the McMartin pre-
niques. school case. Journal of Applied Psychology, 83,
347–359.
Saywitz, K. J., & Camparo, L. B. (2009). Contemporary
REFERENCES child forensic interviewing: Evolving consensus and
innovation over 25 years. In B. Bottoms, C.
Ceci, S., & Bruck, B. (1995). Jeopardy in the
Najdowski, & G. Goodman (Eds.), Children as
courtroom: A scientific analysis of children’s testimony.
victims, witnesses, and offenders: Psychological science
Washington, DC: American Psychological Associa-
andthelaw (pp. 102–127). NewYork: Guilford.
tion.
Zaragoza, M. S., Graham, J. R., Hall, G. C. N.,
Doepke, K. J., Henderson, A. L., & Critchfield, T. S.
Hirschman,R.,&Ben-Porath,Y.S.(1995).Memory
(2003). Social antecedents of children’s eyewitness
and testimony in the child witness. Thousand Oaks,
testimony: A single-subject experimental analysis.
CA:Sage.
JournalofAppliedBehavior Analysis,36,459–463.
Garven, S., Wood, J. M., & Malpass, R. S. (2000).
Allegations of wrongdoing: The effects of reinforce- Received May 25,2010
ment on children’s mundane and fantastic claims. Final acceptance January3, 2011
JournalofAppliedPsychology,83,38–49. Action Editor,MaryLouise Kerwin