You are on page 1of 8

Swiss Journal of Psychology 65 (3), 2006, 193200

During the last decades, research on human memory and


memory development has mainly focused on the quantita-
tive aspects of memory performance (i.e., how many items
of the list can the individual remember correctly?). How-
ever, it is now again widely recognized (Bartlett, 1932) that
in many everyday memory tasks (for example, self-regu-
lated learning situations, academic tests, forensic inter-
views) memory reports stem from an active reconstruction
of the past. The reconstructive nature of recollections ne-
cessitates an emphasis on quality (i.e., to what degree does
the original information and the memory report corre-
spond?) rather than quantity as the critical factor for which
memories are evaluated. As a result, the focus of experi-
mental memory research both in adults and children has
changed considerably during the last years. Increasingly
more research is investigating individuals competencies to
strategically regulate memory performance. Strategic reg-
ulation of memory accuracy involves, on the one hand, the
use of metacognitive monitoring and control processes
(Nelson & Narens, 1990) and, on the other hand, the con-
sideration of the social situation and the demand charac-
teristics in which a memory report is given (Koriat, Gold-
smith, & Pansky, 2000).
The ability to strategically regulate memory performance
comes into play in any memory (test) situation in which the
individual has the option to decide which information to
provide and which information to withhold, for example,
in situations in which there is uncertainty whether the an-
swer is going to be correct or because the test situation em-
phasizes memory accuracy rather than quantity. A test sit-
uation such as free recall (Tell me everything you
remember about it!) typically motivates powerful strate-
gic decisions about which information to provide, what de-
gree of detail or from which perspective to report
(Dudukovic, Marsh, & Tversky, 2004). Empirical studies
have repeatedly shown that although such free recall tasks
are difficult and yield only small amounts of information,
the accounts are highly accurate, independent of age (e.g.,
Bruck, Melnyk, & Ceci, 2000; Peterson & Whalen, 2001;
Roebers & Schneider, 2001). Thus, strategic regulation can,
under certain and facilitating conditions, occur from a rel-
atively early age on.
However, strategic memory decisions appear to be much
more complicated and difficult in cued recall, for example
when asked to answer specific questions. Koriat and Gold-
smith (1996) have outlined a theoretical model that speci-
Swiss J Psychol 65 (3), 2006 by Verlag Hans Huber, Hogrefe AG, Bern
Original Communication
Developmental Progression
in Childrens Strategic Memory
Regulation
Claudia M. Roebers
University of Berne, Switzerland
A developmental study was conducted that investigated 7- to 11- years-olds ability to strategically regulate their memory performance. The
study, based on Koriat and Goldsmiths (1996) theoretical framework, sought to stimulate strategic regulation processes. In order to do so, the
threshold to provide or withhold answers was manipulated and included bonuses for correct responding and penalties for incorrect answers.
Participants were shown a video concerning the production of sugar from beets and were individually interviewed a week later. Responses had
to be made to both answerable and unanswerable questions in both an open-ended and a yes/no question format. The results revealed that de-
pending on the question format, there were different effects of the threshold manipulations on the frequencies of correct, incorrect, and I dont
know responses. Although there were no differences in response behaviour between the 1:0 and 1:1 incentives conditions when the questions
were open-ended, children were able to differentially respond to the two bonus-to-penalty ratios when the questions were in yes/no format. Con-
sistently, there were no interactions between age and response threshold indicating that strategic regulation competencies develop slowly but
continuously during the primary school years showing the first signs of emerging competencies from an age of 7 onwards.
Keywords: memory development, metacognition, recall accuracy, monitoring, control processes
DOI 10.1024/1421-0185.65.3.193
194 C. M. Roebers: Strategic Regulation of Childrens Memory Reports
fies the conditions under which individuals can utilize mon-
itoring and control processes for the strategic regulation of
memory accuracy in cued recall tasks. According to this
model, asking for a specific detail will elicit several possi-
ble answers. Each piece of information is then assessed in
terms of the probability that it is correct. In the next step,
the individual decides to volunteer or withhold that infor-
mation (control process) given a previously set response cri-
terion. The response criterion is sensitive to many factors
such as the situational demands of an interview situation
(e.g., the interviewer; Jackson & Crockenberg, 1998; Good-
man, Sharma, Thomas, & Golden Considine, 1995) or in-
centives for correct answers and penalties for incorrect an-
swers (e.g., Koriat & Goldsmith, 1998; Roebers, Moga, &
Schneider, 2001).
Thus, memory accuracy is dependent not only on the un-
derlying memory trace but also on two metacognitive pa-
rameters: monitoring effectiveness (i.e., the validity of con-
fidence judgments) and control sensitivity (i.e., defining the
threshold with respect to the social situation). When mon-
itoring is precise, possibly incorrect answers are screened
out which leads to an increase in memory accuracy. When
monitoring is poor, the wrong answers are screened out,
without any benefit to memory accuracy. In control sensi-
tivity, a low response criterion should lead to more specif-
ic answers, some of which might turn out to be incorrect.
A strict response criterion, conversely, should lead to only
a few specific answers, all of which the individual is very
sure of, and also to many answers being withheld, some of
which would in fact be correct (i.e., quantity-quality trade
off). Thus, based on the theoretical framework proposed by
Koriat and Goldsmith (1996), the unique contributions of
each different factor can be isolated from each other, that is
retention (memory), monitoring (validity of confidence
judgements), and control processes (as a function of the re-
sponse criterion).
Developmental research on the validity of the theoreti-
cal framework is still at a very early stage and to date lim-
ited to event recall tasks. Applying the theoretical model
and experimental design developed by Koriat and Gold-
smith (1996), Roebers and Schneider (2001) showed that
when trying to remember details of a movie children as
young as 6 years were able to significantly increase their
overall recall accuracy when answering unbiased, specific
questions under a relative strict response criterion (get one
and lose one token for every correct and incorrect answer,
respectively; 1:1 bonus-to-penalty). These responses were
obtained compared to children interviewed under a more
liberal response criterion (no bonuses). However, in a fol-
low-up study (Roebers & Fernandez, 2002) cases in which
children were no longer taken away a token for incorrect
responding, the benefit observed for open-ended questions
was due to both, a significant increase in correct responses
and also to fewer incorrect responses. Although the model
predicts fewer incorrect and more I dont know respons-
es (indicating adequate metacognitive control processes)
for a high threshold, in this study children achieved an in-
crease in memory accuracy by other means when faced with
the 1:0 bonus-to-penalty ratio (Roebers & Fernandez,
2002). This pattern of results for a strict response criterion
was confirmed in another recent study using the same ex-
perimental manipulations and memory task (Roebers &
Schneider, 2005). Thus, there is now growing evidence that
the theoretically assumed independent contribution of
memory and metacognitive processes for accurate remem-
bering may not hold for children. Although, Koriat and
Goldsmith (1996), have shown that monitoring and control
processes are as important as memory factors for the strate-
gic regulation of memory accuracy, studies including chil-
dren of various ages underline the importance of memory
retrieval as the base for monitoring and control processes
(Roebers & Schneider, 2005).
The methodological problems associated with memory
factors (confounding age differences in encoding, retriev-
ing and monitoring) can be eliminated when prompting sub-
jects to report about information that was never encoded,
that is, when including unanswerable questions in an inter-
view or a memory test. Unanswerable questions cannot be
answered correctly by any participant, simply because the
information in question was never provided. In order to
identify an unanswerable question, monitoring and control
processes are necessary, which include realizing that con-
fidence judgments for any upcoming answer are low and
consequently giving an appropriate I dont know re-
sponse. Age-related differences in the occurrences of ap-
propriate I dont know responses to unanswerable ques-
tions thus mirror age-related differences in metacognitive
competencies. So far, there is only one study that applied
the theoretical framework of Koriat and Goldsmith (1996),
and also included unanswerable questions. Roebers and
Fernandez (2002), showed that, when rewarded for correct
answers, childrens ability to correctly identify unanswer-
able questions and to appropriately respond with I dont
know increased significantly, independent of age (I dont
know responses were not rewarded).
Stemming from these findings, the present study ad-
dresses some open questions regarding childrens ability to
strategically regulate memory accuracy. First, two different
bonus-to-penalty ratios will be contrasted to investigate
whether only pronounced manipulations of the response
threshold increases the dont know responses in a young
sample. This can be interpreted as a unique contribution of
control processes for increases in memory accuracy. Sec-
ond, the present study aims to expand the application of the
theoretical model to other stimulus materials. While previ-
ous studies have exclusively used purely episodic stimuli,
in the current study more semantic information is present-
ed. Although this is done in an episodic context (presenting
children an educational television program), the cued recall
task taps mainly semantic knowledge (on sugar production)
and thus resembles more everyday testing situations in
school than previous interviews did.
A third so far unanswered question concerns childrens
strategic memory competencies that will be addressed in
Swiss J Psychol 65 (3), 2006 by Verlag Hans Huber, Hogrefe AG, Bern
C. M. Roebers: Strategic Regulation of Childrens Memory Reports 195
the current study by focusing on the test format. Cued re-
call implies the option to withhold answers and thus allows
strategic regulation processes. Within this test format, how-
ever, two different question formats need to be considered
separately: open-ended wh-style and the yes/no question
format. These two question formats can be seen to repre-
sent two extremes of social impact that provide a particu-
lar answer resulting, in terms of the underlying theoretical
model, in two different a priori response thresholds. While
open questions do not suggest a specific or preferred an-
swer and thus result in a stricter response criterion, yes or
no questions to which the correct answer is no may im-
ply an invitation to acquiesce to the wrong answer and there-
fore result in a more liberal response criterion. Many stud-
ies have shown that depending on a childs age, memory
accuracy varies significantly across the two question for-
mats (e.g., Bruck & Ceci, 1999 for a review). Consequent-
ly, the model predicts different patterns of results indicat-
ing differences in the strategic regulation of memory
performance in response to different question formats, as a
function of the individuals age.
Method
Participants
A total of 142 participants took part in the study. There were
47 7-year-olds (23 girls, 24 boys, M = 92 months, SD = 5
months), 47 9-year-olds (23 girls, 24 boys, M= 118 months,
SD = 6 months), and 48 11-year-olds (24 girls, 24 boys, M
= 142 months, SD= 7 months). Within each age group, chil-
dren were randomly assigned to one of three experimental
conditions, with the only constraint that girls and boys were
approximately equally distributed within each condition.
All children lived in the region of Berne, Switzerland, were
recruited from three different schools and came from di-
verse socioeconomic backgrounds. Parents gave informed
written consent prior to the study.
Procedure
In the first phase of the experiment, small groups of 5 to 10
children from a single age group were shown a short video
(7 minutes) about the production of sugar from sugar beets.
This video was shown in an empty multi-purpose classroom
using a DVD player and a projector producing a 12 me-
ter image on a silver screen. Once children were seated, the
experimenter told them he/she was interested in their opin-
ion of the film. This was done to prevent any intentional re-
hearsal strategies. After viewing the video, participants
were asked if they liked the video, and if they considered it
appropriate for other children. After some very brief and
general comments, children were thanked for their cooper-
ation and brought back to their classroom.
In the second phase of the experiment, approximately one
week later (range 68 days), all the children were individ-
ually questioned about the video in a small private room at
their school. After a brief period of establishing rapport, the
experimenter directed childrens attention to the film, which
all of them remembered watching. Children in the Free Re-
port condition were given the following instructions: Well,
now Im going to ask you some questions about the film,
and I want you to try the best you can to give me many and
only correct answers. If you are not sure what the right an-
swer is or if you dont remember it, thats o.k.! Nobody can
remember everything. Just tell me you dont remember or
youre not sure, o.k?. When the child indicated to have un-
derstood the instructions, the interview started. Children
were asked a set of 52 questions about the video, which will
be described in more detail in the Materials subsection.
Children in the other two conditions were first given the
same general accuracy instructions described above for the
Free Report condition, but they were additionally given a
complementary set of instructions. Children in the 1:0 In-
centives condition were told: for every correct answer, Ill
give you one of these coins (moving towards the child one
of the mock gold coins placed on the table halfway between
the experimenter and the child); for every incorrect answer,
Ill take a coin for myself. If you answer with dont know
or dont rememberthe coin remains in the middle. Then,
childrens attention was directed to a variety of toys placed
on a separate table next to the experimenters table, and they
were told that the more gold coins they earn, the better the
toy that they could buy later on. We used across all age
groups a set of four age appropriate toys: two of higher val-
ue and two small-sized toys of lower value. The attractive-
ness and the age-appropriateness of the toys were piloted
with a small sample of children from each age group. In any
case, all children were allowed to select whichever toy they
wanted at the end of the session.
Children assigned to the 1:1 Incentives condition were
additionally instructed as follows: Let me give you 10 gold
coins as a start, because in a moment, when we start with
the interview, youll get one more coin for every correct an-
swer; for every incorrect answer, however, Ill take a coin
away from you and take it for myself. If you answer with
dont know or dont remember all the coins stay where
they are. As in the 1:0 Incentives condition, childrens at-
tention was then directed to the variety of toys placed on a
separate table next to the experimenters table, and the same
explanations were given to the children (see above). Thus,
the critical difference between the 1:0 and 1:1 Incentives
conditions is, that with the 1:0 bonus-to-penalty ratio there
is no penalty for incorrect responding, while in the 1: 1 con-
dition children are taken one token away for every incor-
rect answer.
Materials
The video was approximately 7 minutes long and depicts
the production of sugar in a sugar factory. A speaker ex-
Swiss J Psychol 65 (3), 2006 by Verlag Hans Huber, Hogrefe AG, Bern
196 C. M. Roebers: Strategic Regulation of Childrens Memory Reports
plains the single production steps and machines in simple
words: The film starts by showing how a farmer harvests
the sugar beets and who brings them to a factory. Inside the
factory, they are first cleaned, then cut into pieces and
cooked. The resulting sugar juice is filtered, cooked again
and is finally centrifuged. This film is part of a widely seen
German speaking federal broadcasting program for chil-
dren from 3 to 13 years called Die Sendung mit der Maus.
The video and most of the questions used in this study were
already used in previous experiments (e.g., Beuscher, Roe-
bers, & Schneider, 2005; Beuscher & Roebers, 2005).
Each participant was asked the same set of 52 questions
about the video by following the general structure of the
film. Within that interview, there were two question types,
answerable and unanswerable questions. Altogether, twen-
ty-two questions were answerable and the relevant infor-
mation was presented in the video. Twenty questions were
unanswerable and covered information that was not pre-
sented in the video. Within these two different question
types, the question format was also varied: half of the ques-
tions of one type (answerable or unanswerable) were asked
using an open-ended question format (requested a one or
two word answer) while the other half was asked using a
yes or no question format. An example for an answerable
question in an open-ended question format is What colour
are the sugar beets?. An answerable question in a yes or
no question format is Is the liquid sugar white?. An unan-
swerable question in an open-ended question format is
Where is the water coming from with which the sugar beets
are washed? (not shown in the video), and an example for
an unanswerable question in a yes/no question format is
Do the sugar beets cuts have to be cooked for 3 hours?
(not mentioned during the film). Reliability of the resulting
four subgroups of question was generally high (Cronbach
= .71.79) and did not systematically vary between the
three different age groups).
The remaining ten questions served as filler questions,
that is, there were ten very easy questions asked in a posi-
tive leading format. These were mixed under the other crit-
ical questions in order to guarantee some correct respons-
es for every participant, to maintain the interviewers
credibility and to avoid uniformity of the questioning. The
filler questions were not included in the analysis reported
below.
Results
Data analysis was first performed on the answerable ques-
tions, and secondly on the unanswerable questions. In both
cases, analyses were separately run for the two question for-
mats (open-ended questions and yes or no questions) be-
cause they constitute two distinct test formats and pose dif-
ferent cognitive demands on the participants. Preliminary
data analysis yielded no systematic differences between
girls and boys and no significant interactions between sex
and age, condition, question type and question format.
Therefore, data was collapsed across gender. All results are
reported at the p< .05 level unless otherwise specified. New-
man-Keuls post-hoc tests were performed for single group
comparisons. In order to be able to directly compare the ef-
fect sizes across dependent variables, partial Eta
2
-values are
reported for significant effects.
Open-Ended Answerable Questions
Table 1 presents the mean percentages of correct, incorrect
and I dont know responses to the open-ended questions
as a function of age and experimental condition. The 33
ANOVA with age (7-, 9- and 11-year-olds) and condition
(1:1, 1:0, Free Report) as between-participants factors on
the percentages of correct answers revealed main effects of
age, F(2, 142) = 15.61,
2
= 0.19, and condition, F(2, 142)
= 3,39,
2
= 0.05, but no interaction (F < 1). Post-hoc tests
on the main effect of age indicated that 7- and 9-year-olds
(28% and 30%, respectively), who did not differ from one
another, gave significantly fewer correct responses than did
11-year-olds (46%). The main effect of condition was due
to a significant difference between the Free Report (31%)
and the 1:0 condition (40%) with the 1:1 condition falling
in between the two other conditions (34%).
Swiss J Psychol 65 (3), 2006 by Verlag Hans Huber, Hogrefe AG, Bern
Table 1
Mean percentages of Correct, Incorrect, and I dont know An-
swers in Response to the Open-ended Answerable Questions as a
Function of Age and Experimental Condition (Standard Deviations
in Parentheses)
Free Report 1:0 Incentives 1:1 Incentives
7-year-olds
% Correct 30.1 (13.9) 28.5 (20.2) 26.7 (17.1)
% Incorrect 18.2 (10.5) 9.1 (13.3) 9.1 (11.5)
% I dont know 51.7 (11.8) 62.4 (25.7) 64.2 (23.3)
9-year-olds
% Correct 23.9 (10.9) 40.6 (19.1) 25.6 (11.6)
% Incorrect 26.7 (14.3) 11.5 ( 9.3) 11.3 ( 9.1)
% I dont know 49.4 (20.7) 47.9 (19.9) 63.1 (18.0)
11-year-olds
% Correct 38.6 (15.0) 49.4 (18.2) 48.3 (19.3)
% Incorrect 24.4 (15.5) 11.4 (13.1) 8.5 ( 8.4)
% I dont know 36.9 (20.3) 39.2 (22.4) 43.2 (20.9)
An ANOVA on the percentages of incorrect answers to
the open-ended answerable questions with both age and
condition as between-participants factors revealed a main
effect of condition, F(2, 142) = 18.9,
2
= 0.22, but no main
effect of age (F < 1) and no interaction (F < 1). Collapsed
across the three age groups, participants in the Free Report
condition gave significantly more incorrect answers (23%)
than in the two other experimental conditions that did not
differ from one another (10% and 11%, for the 1:1 and 1:0
condition, respectively).
The corresponding ANOVA on the percentages of I
dont know responses with age and condition as between-
participants factors revealed main effects of age F(2, 142)
C. M. Roebers: Strategic Regulation of Childrens Memory Reports 197
= 11.36,
2
= 0.15, and condition, F(2, 142) = 3.38,
2
=
0.05, but no interaction (F < 1 ). Post-hoc tests on the main
effect of age revealed that 11-year-olds (40%) gave signif-
icantly fewer I dont know responses than 9- (54%) and
7-year-olds (60%) that did not differ from one another. The
main effect of condition was due to a reliable difference be-
tween the Free Report (46%) and the 1:1 condition (57%),
with the 1:0 condition (50%) falling in between the other
two conditions and not differing from either of them.
Answerable yes or no Questions
Table 2 presents the mean percentage of correct, incorrect,
and I dont know responses to the answerable yes/no ques-
tions as a function of age and experimental condition. Again,
multivariate 34 ANOVAs with age and condition as be-
tween-participants factors were performed on the percent-
ages of correct, incorrect, and I dont know answers.
Analyses on the percentages of correct answers produced
significant main effects of age, F(2, 142) = 6.44,
2
= 0.09,
and condition, F(3, 142) = 6.20,
2
= 0.08, but no interac-
tion (F < 1 ). Post-hoc tests on the age effect indicated that
the oldest age group (57%) gave significant more correct
answers than the two younger age groups who did not dif-
fer from each other (7-year-olds: 47% = 9-year-olds: 46%).
Post-hoc tests on the main effect of condition revealed sig-
nificantly higher percentages of correct answers in the two
incentives conditions (1:0 condition: 52% = 1:0 condition:
55%) in comparison to the Free Report condition (43%).
Finally, an ANOVA on the percentages of I dont know
responses to the answerable yes or no questions revealed
significant main effects of age, F(2, 142) = 3.67,
2
= 0.05,
and condition, F(2, 142) = 5.99,
2
= 0.08, but no interac-
tion (F < 1 ). With respect to the frequencies of I dont
know responses, the 1:1 condition (19%) yielded higher
frequencies than the two other conditions that did not dif-
fer from one another (10% and 11%, for the Free Report
and 1:0 condition, respectively).
Unanswerable Questions in Open-Ended
and yes/no Question Format
Table 3 presents the mean percentage of appropriate dont
know responses to the open-ended and yes or no unan-
swerable questions as a function of age and experimental
condition. Again and for the reasons outlined above, the two
question formats were analyzed separately. A two-way 34
ANOVA with age and condition as between-subject factors
applied to the percentages of dont knowresponses to the
unanswerable open-ended questions revealed a main effect
of condition, F(2, 142) = 15.08,
2
= 0.19, but no main ef-
fect for age (F< 1) and no interaction (F< 1). Post-hoc tests
showed that the main effect of condition was due to signif-
icantly fewer appropriate I dont know responses in the
Free Report condition (63%) than in the two incentives con-
ditions that did not differ from one another (78% and 84%,
for the 1:0 and 1:1 condition, respectively).
Swiss J Psychol 65 (3), 2006 by Verlag Hans Huber, Hogrefe AG, Bern
Table 2
Mean Percentages of Correct, Incorrect, and I dont know An-
swers in Response to the Yes/No Answerable Questions as a Func-
tion of Age and Experimental Condition (Standard Deviations in
Parentheses)
Free Report 1:0 Incentives 1:1 Incentives
7-year-olds
% Correct 34.6 (16.7) 48.5 (13.6) 59.1 (14.8)
% Incorrect 51.8 (21.9) 41.8 (14.9) 24.4 ( 9.2)
% I dont know 13.6 (16.2) 9.7 ( 9.4) 16.5 (14.5)
9-year-olds
% Correct 42.6 (18.7) 47.9 (14.4) 47.2 (20.5)
% Incorrect 48.3 (23.2) 35.7 (17.0) 27.8 (15.8)
% I dont know 9.1 (14.1) 16.4 (10.4) 25.0 (18.3)
11-year-olds
% Correct 52.8 (10.1) 60.3 (20.4) 59.1 (20.2)
% Incorrect 41.5 (13.3) 31.8 (17.2) 26.7 (18.3)
% I dont know 5.7 ( 9.8) 7.9 ( 8.0) 14.2 (16.6)
An ANOVA on the percentages of incorrect answers to
the answerable yes/no questions yielded a strong main ef-
fect of condition, F(2, 142) = 17.52,
2
= 0.21, but no main
effect of age (F < 1) and no interaction (F < 1). Post-hoc
tests revealed that all three experimental conditions differed
significantly from each other (1:1 condition: 26% < 1:0 con-
dition: 36% < Free Report condition: 47%).
Table 3
Mean Percentages of Adequate I dont know Answers in Re-
sponse to the Unanswerable Questions as a Function of Age, Ex-
perimental Condition, and Question Format (Standard Deviations
in Parentheses)
Free Report 1:0 Incentives 1:1 Incentives
Open-ended
7-year-olds 59.6 (20.9) 76.9 (16.1) 80.7 (10.9)
9-year-olds 64.2 (31.0) 71.5 (20.6) 85.8 (13.7)
11-year-olds 64.2 (26.0) 83.5 (13.8) 85.2 (10.4)
Yes/No
7-year-olds 26.3 (28.5) 36.7 (26.1) 48.7 (18.9)
9-year-olds 25.6 (31.6) 50.0 (27.5) 65.0 (25.0)
11-year-olds 28.1 (35.4) 59.4 (31.3) 76.2 (16.7)
The two-way ANOVA on the unanswerable questions in
the yes or no question format revealed main effects of age,
F(2, 142) = 4.78,
2
= 0.07 and condition, F(2, 142) = 21.78,

2
= 0.25. Concerning the age differences, post-hoc tests
showed that only the difference between 7- (37%) and 11-
year-olds (55%) was significant, while the 9-year-olds fell
in the middle (47%). The experimental manipulation yield-
ed significant differences between all three conditions in
the expected order (Free Report: 27% < 1:0 Incentives: 49%
< 1:1 Incentives: 63%).
198 C. M. Roebers: Strategic Regulation of Childrens Memory Reports
Discussion
The present study aimed to expand the empirical evidence
on the applicability of the theoretical framework from Ko-
riat and Goldsmith (1996) for childrens memory reports.
According to the authors, strategic regulation of memory
accuracy involves independent contributions of memory re-
trieval, metacognitive monitoring and control processes. Ef-
ficient regulation results in a highly accurate memory re-
port because uncertain answers are withheld (decreases in
incorrect answers) and replaced with I dont know re-
sponses. Since previous studies have produced mixed re-
sults concerning the question under which task demands or
to what degree response threshold manipulations are nec-
essary to stimulate childrens strategic regulation abilities,
the present study directly compared different threshold ma-
nipulations with a control condition (Free Report) using two
different test formats and involving three different age
groups.
With respect to age differences, the expected advantage
of older children in comparison to younger children was
found. Older children gave more correct answers to open-
ended and to yes/no questions than younger children there-
by confirming general findings from memory development
(Schneider & Bjorklund, 1998). Since a pilot study with 5
to 6 year olds and 7 to 8 year olds had shown that using this
specific film on sugar production and its questions result-
ed in floor effects for correct responding and an overcau-
tious answering behaviour in the youngest age group, the
present study involved somewhat older children compared
to previous studies (Roebers & Fernandez, 2002; Roebers
& Schneider, 2005). Consequently, direct comparisons in
the level of performance across studies are difficult. Nev-
ertheless, the age differences and the differences in perfor-
mance for the two question formats certainly emphasise the
reliability of the data.
One of the major aims of the present study was to ex-
plore childrens ability to strategically regulate their mem-
ory performance for to-be-remembered information that is
less episodically structured and therefore represented in a
less script-like format. Comparisons between the three ex-
perimental conditions revealed that with respect to all de-
pendent measures and the different question formats chil-
dren did engage in strategic regulation activities. Although
the extent to which these mental activities were successful
varied considerably in all conditions, the Free Report con-
dition differed consistently from the two conditions in
which the response threshold had been augmented. Both in-
centives conditions yielded either more correct and I dont
know and/or fewer incorrect answers than the Free Report
condition. Thus, the present study shows that children are
able to strategically regulate their memory reports when
more semantic knowledge is tested. The emergence of the
necessary skills, however, seems to take place later in de-
velopment. Children as young as 6 years may show first
signs of these emerging abilities when asked to report on
episodic or autobiographical information, however, weak-
er underlying memory traces and/or different underlying
memory representations appear to complicate the process-
es involved (Roebers et al., 2001). The existing literature
on the general development of episodic and autobiograph-
ical memory supports this assumption (Fivush & Haden,
2003).
Although a wide age range of children was studied, no
age-dependent effects of the threshold manipulations
reached significance. The consistent lack of interactions be-
tween age and the experimental manipulations was also
found in previous studies (Roebers et al., 2001; Roebers &
Fernandez, 2002; Roebers & Schneider, 2005) and appears
to indicate relatively slow and continuous improvements in
the strategic abilities rather than a sudden on-set of these
processes. The literature on the development of metacog-
nitive monitoring and control competencies across the pri-
mary school years also suggests slow but continuous de-
velopmental pathways (Schneider, 1998). In so far, the
development of strategic memory abilities fits well into the
general picture of memory development (Schneider et al.,
1998).
Among the major issues addressed in the present study
were (a) the investigation of the specific means by which
children achieve better memory performance and (b)
whether children are able to differentially respond to the
two different bonus-to-penalty ratios. While the model pre-
dicts decreases in incorrect and increases in I dont know
responses for any augmentation of the response threshold,
it is only through extreme manipulations of the response
threshold that effects that conform to the theory have been
observed in children. Similarly, moderate manipulations
lead to increases in correct responding (Koriat, Goldsmith,
Schneider, & Nakash-Dura, 2001; Roebers et al., 2001;
2005). The present study revealed important results with re-
gard to these research questions. For both question formats,
there were fewer correct answers in the non-incentives con-
dition than in either incentives condition. In line with this
finding, there were also more incorrect answers in the non-
incentives condition than in either incentives condition,
again independent of question format. Thus, the increase in
response threshold through incentives positively affected
memory retrieval processes.
Furthermore, the study identified question format as an
additional factor in strategic control. The effects of the two
incentives conditions differed significantly as a function of
question format. For the open-ended questions, both in-
centives conditions yielded the same pattern of results in
terms of incorrect and I dont know answers. The simi-
larities in response patterns here do, however, not neces-
sarily indicate that children are not sensitive to the differ-
ences between the two bonus-to-penalty ratios. For the
yes/no questions, in contrast, childrens answering behav-
iour differed significantly among the two incentives condi-
tions with, as predicted by the model, fewer incorrect and
more I dont know answers in the 1:1 than in the 1:0 in-
centives condition. Possibly, the results for the open-ended
questions are due to the fact that in the 1:0 condition, the
Swiss J Psychol 65 (3), 2006 by Verlag Hans Huber, Hogrefe AG, Bern
C. M. Roebers: Strategic Regulation of Childrens Memory Reports 199
mean percentages of I dont know responses are relatively
high and the percentages of incorrect responses are low,
leaving relatively little scope for further improvement in the
1:1 condition. In the yes/no questions, the demand charac-
teristics of this question format when questioned normally,
may be ambiguous for children with respect to interview-
ers expectations.
Another possible interpretation for these findings is that
monitoring effectiveness varies as a function of question
format. Convincing evidence for that assumption comes
from studies that directly investigate monitoring processes
(Roebers, 2002; Roebers & Howie, 2003; Roebers, von der
Linden, & Howie, in press). Consistently, childrens abili-
ty to monitor the correctness of their given answer was
shown to be more adequate and sophisticated when an-
swering open-ended compared to yes/no questions. Conse-
quently, control processes necessary for screening out in-
correct answers may be the result of imprecise monitoring.
Moreover, control sensitivity may be better for the yes/no
question format because the unequivocal rules of bonuses
and especially of penalties for incorrect responding may re-
solve childrens problems how to deal with the demands of
the interview. While the yes/no questions somehow imply
that the interviewer knows the correct answers, penalties
for yielding (incorrect) answers (only given in the 1:1 con-
dition) prove the opposite to be the case. This can explain
why the 1:0 and the 1:1 conditions differed significantly in
terms of strategic regulation processes only when consid-
ering the yes/no questions.
The inclusion of both answerable and unanswerable
questions within one interview allowed the memory factor
and its effects on strategic memory competencies to be ex-
amined. Again, depending on the question format, different
patterns of results emerged. For the open-ended questions,
increases in response threshold led to significant increases
in adequate I dont know responses to the unanswerable
questions while I dont know responses to answerable
questions were not affected. Thus, children from the age of
7 onwards proved to be able to metacognitively differenti-
ate between filling ones memory gaps (i.e., confabulating
an answer to an unanswerable question) and providing un-
certain answers to what were in principle answerable ques-
tions. On the other hand in the yes/no questions, the mem-
ory factor does not seem to play an equally important role.
For answerable and unanswerable questions likewise, the
three experimental conditions differed significantly from
each other. The interpretation of unambiguous task de-
mands in the incentives conditions, helping children to cope
with the social demands of the specific yes/no question for-
mat and the bonuses received for resisting suggestions can
here again be considered.
It is obvious that the chosen procedure does not allow a
full generalization to academic testing situations in school.
The procedure involved an individual interview with im-
mediate and explicit on-line feedback about memory accu-
racy. In school, however, students receive immediate feed-
back only in oral test situations. In the more important
written tests, however, feedback is typically given days lat-
er. Thus, it remains to be seen whether, to what extent and
from what age children can strategically regulate their mem-
ory behavior under different assessments modes. Despite
this shortcoming, the present study should be seen as a first
step to explore childrens metacognitive and control com-
petencies in a school setting. The evidence for childrens
ability to strategically regulate memory reports about do-
main-specific knowledge in the current study, even though
still open for improvement, underlines that in principle, chil-
dren are sensitive to the demands of a test situation and are
potentially able to adjust their answering behaviour ac-
cordingly. Based on these and future findings, methods to
train children in strategic regulation and to include instruc-
tions for this cognitive domain in class need to be devel-
oped and evaluated.
In sum, the present study offers intriguing substantiation
that the use of incentives when answering open-ended ques-
tions result in increased retrieval efforts in children and
these means can improve their memory reports. When an-
swering yes/no questions, in contrast, childrens improve-
ments in memory accuracy are achieved by theory-conform
increases of withholding uncertain answers when the re-
sponse threshold is increased. Although children practice
answering open-ended questions from an early age (Orn-
stein, Haden, & Hedrick, 2004), responding correctly to
yes/no questions, especially those that suggest a specific an-
swer, seems to be less familiar, more difficult to cope with
and appears to require explicit communicative rules. Thus,
there is a slow and continuous development of the neces-
sary skills that take place later in life when less familiar or
more difficult tasks are chosen. Such a context/task depen-
dency in childrens emerging cognitive abilities has also
been shown in many other areas (e.g., Bjorklund & Rosen-
blum, 2001; 2002) and encourages developmental progres-
sion in a more general way.
Author Note
I would like to thank Rahel Devenoge and Claudia Graf for
their help with the data collection, as well as the children,
parents and staff of the schools for their participation and
cooperation.
References
Bartlett, F. C. (1932). Remembering: A study in experimental and
social psychology. Cambridge: Cambridge University Press.
Beuscher, E., & Roebers, C. M. (2005). Does a warning help chil-
dren to more accurately remember an event, to resist mis-
leading questions, and to identify unanswerable questions?
Experimental Psychology, 52, 232241.
Beuscher, E., Roebers, C. M., & Schneider, W. (2005). Was erin-
nern Kinder von Lernfilmen? [What do children recall from
educational movies]. Psychologie in Erziehung und Unter-
richt, 52, 5165.
Swiss J Psychol 65 (3), 2006 by Verlag Hans Huber, Hogrefe AG, Bern
200 C. M. Roebers: Strategic Regulation of Childrens Memory Reports
Bjorklund, D. F., & Rosenblum, K. E. (2001). Childrens use of
multiple and variable addition strategies in a game context.
Developmental Science, 4, 184194.
Bjorklund, D. F., & Rosenblum, K. E. (2002). Context effects in
childrens selection and use of simple arithmetic strategies.
Journal of Cognition and Development, 3, 225242.
Bruck, M., & Ceci, S. J. (1999). The suggestibility of childrens
memory. Annual Review Psychology, 50, 419439.
Bruck, M., Melnyk, L., & Ceci, S. J. (2000). Draw it again Sam:
The effect of drawing on childrens suggestibility and source
monitoring ability. Journal of Experimental Child Psycholo-
gy, 77, 169196.
Dudukovic, N. M., Marsh, E. J., & Tversky, B. (2004). Telling a
story or telling it straight: The effects of entertaining versus
accurate retellings on memory. Applied Cognitive Psycholo-
gy, 18, 125144.
Fivush, R., & Haden, C. A. (Eds.) (2003). Autobiographical mem-
ory and the construction of a narrative self: Developmental
and cultural perspectives. Mahwah, NJ: Erlbaum.
Goodman, G. S., Sharma, A., Thomas, S. F., & Golden Considine,
M. (1995). Mother knows best: Effects of relationship status
and interviewer bias on childrens memory. Journal of Ex-
perimental Child Psychology, 60, 195228.
Jackson, S., & Crockenberg, S. (1998). A comparison of sug-
gestibility in 4-year-old girls in response to parental or stranger
misinformation. Journal of Applied Developmental Psychol-
ogy, 19, 527542.
Koriat, A., & Goldsmith, M. (1996). Monitoring and control
processes in the strategic regulation of memory accuracy. Psy-
chological Review, 103, 490517.
Koriat, A., & Goldsmith, M. (1998). The role of metacognitive
processes in the regulation of memory performance. In G.
Mazzoni & T. O. Nelson (Eds.), Metacognitive and cognitive
neuropsychology: Monitoring and control processes (pp.
97118). Mahwah, NJ: Erlbaum.
Koriat, A., Goldsmith, M., & Pansky, A. (2000). Toward a psy-
chology of memory accuracy. Annual Review of Psychology,
51, 481537.
Koriat, A., Goldsmith, M., Schneider, W., & Nakash-Dura, M.
(2001). The credibility of childrens testimony: Can children
control the accuracy of their memory reports? Journal of Ex-
perimental Child Psychology, 79, 405437.
Nelson, T. O., & Narens, L. (1990). Metamemory: a theoretical
framework and new findings. In G. Bower (Ed.), The psy-
chology of learning and motivation: Advances in research and
theory (Vol. 26, pp. 125173). New York: Academic Press.
Peterson, C., & Whalen, N. (2001). Five years later: Childrens
memory for medical emergencies. Applied Cognitive Psy-
chology, 15, S7S24.
Ornstein, P. A., Haden, C. A., & Hedrick, A. M. (2004). Learning
to remember: Social-communicative exchanges and the de-
velopment of childrens memory skills. Developmental Re-
view, 24, 374395.
Roebers, C.M. (2002). Confidence judgments in childrens and
adults event recall and suggestibility. Developmental Psy-
chology, 38, 10521067.
Roebers, C. M., & Fernandez, O. (2002). The effects of accuracy
motivation on childrens and adults event recall, suggestibil-
ity and their answers to unanswerable questions. Journal of
Cognition and Development, 3, 415443.
Roebers, C. M., & Howie, P. (2003). Confidence judgments in
event recall: Developmental progression in the impact of ques-
tion format. Journal of Experimental Child Psychology, 85,
352371.
Roebers, C. M., Moga, N., & Schneider, W. (2001). The role of
accuracy motivation on childrens and adults event recall.
Journal of Experimental Child Psychology, 78, 313329.
Roebers, C. M., & Schneider, W. (2001). Memory for an observed
event in the presence of prior misinformation: Developmen-
tal patterns in free recall and identification accuracy. British
Journal of Developmental Psychology, 19, 507524.
Roebers, C. M., & Schneider, W. (2005). The strategic regulation
of childrens memory performance and suggestibility. Jour-
nal of Experimental Child Psychology, 91, 2444.
Roebers, C. M., von der Linden, N., & Howie, P. (in press).
Favourable and unfavourable conditions for childrens confi-
dence judgments. British Journal of Developmental Psychol-
ogy.
Schneider, W. (1998). The development of procedural metamem-
ory in childhood and adolescence. In G. Mazzoni & T. O. Nel-
son (Eds.), Metacognitive and cognitive neuropsychology:
Monitoring and control processes (pp. 121). Mahwah, NJ:
Erlbaum.
Schneider, W., & Bjorklund, D. F. (1998). Memory. In W. Damon
(Series Ed.), D. Kuhn & R. S. Siegler (Vol. Eds.), Handbook
of child psychology: Vol. 2. Cognition, perception, and lan-
guage (5th Ed. pp. 467521). New York: Wiley.
Claudia M. Roebers
Institute of Psychology
University of Berne
Muesmattstrasse 45
CH-3000 Berne
Roebers@psy.unibe.ch
Swiss J Psychol 65 (3), 2006 by Verlag Hans Huber, Hogrefe AG, Bern

You might also like