Assessment of Problem-Solving Abilities in Normal Adolescents

This study examined the problem-solving abilities of adolescents , using a newly designed test , RAPS (the Rapid Assessment of Problem-Solving) , by Marshall & Karow (2001). The tool is a modified version of the Twenty Questions Test that measures performance based on the number of questions asked to solve each problem , the percent of constraint-seeking questions used, and the efficiency or the amount of information gained from the first four questions asked. Participants included a total of 20 children with no history of neurological or psychiatric disorders who were categorized by age groups (10-11 , 12-13, 14-15, and 16-17) with five subjects in each group . ANOV A results revealed there were no statistical differences among the four age groups for any of the three RAPS measures. Although children did not solve RAPS problems optimally , where they eliminate half of the picture board with each question asked, they did primarily ask questions which targeted groups of pictures based on semantic category labels. Compared to previous RAPS studies (Marshall , Karow , Morelli, Iden , & Dixon , 2003) children performed similarly to adult normal subjects both in the efficiency with which they solved the problems and in the types of questions they asked. The youngest group of children did appear to perform differently than the other groups; however , these differences were not identified statistically , possibly due to the low number of subjects per group. Finally , normal children did not improve their performance on successive administrations , which further supports the developing methodologies and scoring system of RAPS. ACKNOWLEDGEMENTS The completion of this thesis brings a flood of feelings. Relief is high among them , as well as appreciation. Many people deserve recognition and my gratitude for their contributions to this thesis , including mentorship, encouragement , feedback , sympathy, infusion of humor, and friendship. I dedicate this thesis to all of the people who have made my goals a reality. First and foremost , I would like to thank Dr. Colleen Karow for her support and guidance. I am grateful for her professionalism , impeccable knowledge , and concern for quality though all stages of this project. She has made my graduate school experience rewarding , challenging, and a never-ending source of education. I would also like to express my appreciation to all of the members of my thesis committee, including Dr. Mikyong Kim, Dr . Jay Singer, Dr. Charles Collyer , and Dr. Susan Roush . Their dedication, time , and support helped mold this thesis. Additionally, I want to acknowledge all of the people who helped with the recruitment of subjects for this thesis. Kerry, Julie , Ellen, Diane, Meredith , Paulette, because of you, this project is complete. Finally , for putting up with limited accessibility and mood swings, and for always being there for moral support, my thanks go to the most important people in my life. This thesis might not have been started or finished without the support , enthusiasm, and patience of my parents. Also, I thank Matt for his love, understanding , and motivation throughout this project. Without all of your help, I would not have been able to accomplish my degree. Thank you.

Associates I and II, and the Verbal Paired Associates I and II from the Weschler Memory Scale-Revised (Weschler, 1987), and the Coloured Progressive Matrices (Raven , Raven , & Court, 1998)

Background of Problem-Solving
Problem-solving encompasses a vast range of normal cognitive activity. At its most basic, the ability to "problem solve" requires the modulation and control of more fundamental or routine cognitive skills (McCarthy & Warrington , 1990). The right skill has to be harnessed at the right time and changing between skills has to be flexible. Cognitive skills have to be integrated and adapted so that they comply with situational constraints and yet are optimally coordinated so that goals are achieved as efficiently as possible . Luria (1973) stated: Man not only reacts passively to incoming information but creates intentions , forms plans and programmes of his actions , inspects their performance and regulates his behaviour so that it conforms to these plans and programmes , finally he verifies his conscious activity comparing the effects of his actions with the original intentions and correcting any mistakes he has made. (p. 79) Problem-solving includes the ability to draw higher order inferences that require the individual to abstract the necessary information from the elements of the problem and to analyze how its properties may be related to those of others. The formulation of strategies is a necessary component of successful problem-solving and allows the individual to plan an action (McCarthy & Warrington, 1990). These processes develop through childhood and adolescence, and play an important role in a child's cognitive functioning, behavior, and social interaction.

Executive Functioning and Problem-Solving
Executive function is comprised of a range of cognitive abilities that facilitate intentional , goal-directed, problem-solving (Gioia, Isquith , Retzlaff, & Espy, 2002).
Executive function is considered to be an umbrella term "that includes all supervisory or self-regulatory functions, which organize and direct cognitive activity " (Gioia, Isquith, Kenworthy, & Barton, 2002). While some authors regard executive functioning as a unitary system, most support the idea that the subdomains are separable (Diamond & Goldman-Rakie, 1989;Welsh , Pennington , & Grossier, 1991).
Commonly agreed upon executive subdomains include the ability to initiate and maintain behavior, select goals, strategize, and monitor and evaluate one ' s own behavior.
Executive function has been hard to define, however one component consistently included is problem-solving skills.  suggested that the problem-solving process requires interaction between components of executive function to achieve efficient multi-step performance. Certain executive functions (e.g. , inhibition) may be more important and allow for sustained strategic problemsolving.
Neuroanatomy of Executive Functioning Anderson (2002) proposes that the anterior regions of the brain mediate executive functioning. It is well established that the frontal lobes, and in particular the prefrontal cortex , is involved in the cognitive aspects of executive function. There are complex circuits that contain extensive connections to and from the frontal lobes providing an ideal system to allow information process ing, which is necessary to higher level cognitive functioning, such as executive function . The prefrontal region is an association region with extension connections to all areas of the neocortex via cortico-cortical projections, in addition to Links with limbic and subcortical structures such as the cingulate gyrus, hippocampus, basal ganglia, and thalamus . Damage to any of these areas may affect the efferent and/or afferent connections of the prefrontal cortex , and in turn , influence executive functioning. Therefore , executive dysfunction is not always associated with prefrontal pathology directly , but may be related to network disconnections such as white matter damage or impairment to other brain regions that interact with the frontal lobes. Hughes and Graham (2002) report executive impairments in a number of neurological disorders , including autism, attention-deficit/hyperactivity disorder (ADHD) , head injury, epilepsy , Gilles de la Tourette's syndrome and conduct disorder; although autism and ADHD are the two disorders in which impaired executive function is most evident.

Development of Problem-Sol ving
The development of different components of problem-solving in children is related to the development of the frontal systems in the brain. The frontal cortex is relatively slow to develop with some anatomical changes extending into adulthood (Anderson, 2002). Not surprisingly, problem-solving appears to have a prolonged developmental course, with evidence that some basic problem-solving skills emerge in the first year of life, and that various components of problem-solving continue to develop into adolescence.
Some researchers have employed Piagetian techniques to investigate early cognitive development and its relationship with cerebral development. Diamond and Goldman -Rakie (1989) used the classic Piagetian object permanence paradigm , as well as an object retrieval task , to investigate goal-directed behaviors in infants.
Object permanence and object retrieval are important aspects of problem-solving in which the infant demonstrates the ability to plan means-ends sequences. The understanding that objects continue to exist even when they are out of site for a period of time is apparent in children as young as 12 months of age. On object retrieval tasks , human infants showed age-related improvements in planning and self-control.
Maturation of the dorsolateral prefrontal cortex is necessary to intentionality, and has been demonstrated through animal research. Diamond and Goldman-Rakie (1989) found that adult rhesus monkeys exhibit object permanence by successfully locating hidden objects even when delays of ten seconds were in place. They tested two groups of adult rhesus monkeys , one with bilateral lesions to the prefrontal cortex and another with bilateral lesions to the parietal lobes on tests of object permanence.
Although the group with parietal lesions continued to demonstrate intact object permanence, the prefrontal lesio11-ed group performed like human infants that could only locate hidden objects as long as there was no delay between covering the object and allowing for retrieval of it. The ability to locate a hidden object without a delay is present in infants between 7.5 and 9 months of age (Diamond & Goldman-Rakie , 1989). Because of these findings the authors reported object permanence begins to develop around 7 .5 months in the infant and is dependent on maturation of the dorsolateral prefrontal cortex.
Other studies have attempted to map developmental paths for aspects of problem-solving skills in older children. Passler, Isaac, and Hynd (1985) have shown that children as young as 6 years are able to exhibit strategic and planful behavior on tasks of frontal lobe function. Their results suggest that the emergence of frontal-lobe function in children represents a multi-stage process with the period of greatest development occurring at the 6-8 year-old level, with mastery of behaviors associated with frontal lobe function occurring around the age of 12.
Standardized tests of executive function designed to measure problem-solving have been used to determine developmental levels in children. Levin et al. (1991) evaluated 52 normal children and adolescents in three age bands, 7-8 years, 9-12 years, and 13-15 years. They administered a range of measures and identified developmental gains, reflecting progress in concept formation, mental flexibility , and problem-solving through childhood. Specifically, they found major gains in mental flexibility between the 7-8 and 9-12 year-old groups on the Wisconsin Card Sorting Test and further advances were evident on concept formation and problem solving in the 13-15 year-old age range on the Twenty Question s and Tower of London Test.
Likewise , Welsh et al. (1991)  The Twenty Questions (20Qs) Test (Mosher & Hornsby, 1966) assesses a child's ability to utilize feedback and reevaluate goals to reach a correct response. The child is shown a card with 42 hand-drawn pictures , which may be grouped into various categories (e.g. animals , plants , utensils). The child is asked to identify which picture the examiner has in mind, and is able to ask 20 questions to do so. Only questions necessitating a yes or no response are allowed . Unfortunately , developmental norms are not available for this task, and administration protocols vary considerably.
The Wisconsin Card Sorting Test (WCST) (Chelune & Baer, 1986) is considered to tap the ability to form abstract concepts and shift and maintain set. The child is presented with four stimulus cards, a red triangle, two green stars, three yellow crosses , and four blue circles. The child is then directed to match each of the response cards , each with configurations similar to those on the stimulus cards, to one of the four stimulus cards and informed that they will be told whether or not each response is correct. Chelune and Baer (1986) provide normative data for children aged between 6-12 years on the WCST, indicating improvements in performance throughout childhood , reflecting increasing abilities in concept formation.
The Tower of London (TOL) (Culberson, & Zilimer, 1999) measures problemsolving aspects of executive functioning. The task involves 12 items, · with each requiring children to rearrange three colored balls to a configuration presented on a stimulus card, and in a prescribed number of moves. When a child fails to complete an item correctly , the balls are replaced in their original configuration , and the child has the opportunity to try again. This test taps planning speed , impulsivity , and flexibility, however , its clinical use has been restricted because it lacks standardized administration protocols and normative data.
Traditional measures typically use specially formulated problems and tasks that are unfamiliar to the client. Some researchers have questioned the accuracy of such unnatural tasks and their assessments of executive functioning, indicating a need for ecologically valid tests (Anderson, 2002;Marshall et al., 2003).

Problem-Solving in Learning Disabled Children
Learning to ask questions effectively is an important achievement with considerable practical application. Such information seeking enables a child to acquire knowledge, to clarify ambiguity, and to solve problems. It is also an important aspect of children's developing communicative ability.
Recent studies have shown that many of the academic problems of students with learning disabilities are tied to problem-solving difficulties (Norris & Foxcroft , 1996). Researchers suggest that difficulties in the application of efficient task strategies may be characteristic of some students with learning disabilities , and that poor academic performance is in part due to the lack of, or failure to utilize specific goal-directed strategies. Norris and Foxcroft (1996) found that boys with learning disabilities were as adept as their normally achieving peers in the induction of equivalence concepts , but were unable to utilize these concepts "as the basis for the development of an effective questioning strategy." As a result , they were less able than their normally achieving peers to formulate effective questioning strategies for an information-seeking task.

Purpose
Although executive skills appear to be crucial to successful functioning of children in school and in society at large , there is a lack of measures that (a) are designed for use with children and adolescents and (b) yield information that accurately reflects their day-to-day behavior. Most clinical measures of problemsolving have been designed primarily for adult populations. As a result, many assessment tools are irrelevant for children and possess limited supporting normative data .
Present research using RAPS, a modified version of the 20Qs Test , focused solely on the problem-solving abilities of adults , both young and old (Marshall et al., 2003) . However, past research has been conducted not only on the decline of problem-solving in the elderly, but also on the overall development of the executive 8 functioning skills beginning in childhood (Denny, 1985). Mosher and Hornsby (1966) examined six to eleven year-old children and determined that older children perform better than younger children on the 20Qs Test. Likewise, Levin et al. (1991) utilized the 20Qs Test in their study of executive function, and found that older children needed to ask fewer questions to identify target pictures, suggesting better capacity to form concepts and utilize feedback. Garth, Anderson, and Wrennall (1997) employed the 20Qs Test to investigate the problem-solving skills of children who had sustained frontal lobe damage. They found no differences between a clinical group and controls on summary measures of total questions asked and time to complete the task. Using a qualitative analysis of the nature of questions posed, they identified less efficient performance by children with frontal lesions. This group exhibited higher frequencies in guessing and less frequency in more efficient ways of solving problems such as constraining a large number of items by the formulated question.
There is a need to devise valid and well-standardized assessment measures for children that are based on current understanding of the nature of both cerebral and cognitive development through childhood. Most available or commonly utilized tests proported to measure executive function in children have been developed for use with adults. These tests may be of little interest or relevance to young children, and frequently lack normative information with respect to developmental expectations.
The purpose of this study is to evaluate how normal children solve RAPS problems and to determine if this measure can be used to successfully evaluate problem-solving skills. It is predicted that children's performance will improve as they increase in age and overall they will be less efficient than adults reported in previous studies. This study is the first to obtain information on how children complete RAPS problems and will contribute to the growing normative database on performance across the lifespan.

Subjects
Twenty children between the ages of 10 and 17 participated in the study.
Subjects were assigned to each of four age groups (10-11 , 12-13, 14-15 , and 16-17) with five participants per group. Groups were not balanced for gender , howe ver, every attempt was made to include both males and females (see Table 1 for subject demographics). The purpose of the study was to investigate normally developing problem-solving skills in children , therefore , any potential subject with a history of poor academic performance , psychiatric or neurological disorders was excluded (see Appendix A for screening criteria). Center . An initial screening process via telephone was conducted with a parent in order to obtain information regarding the exclusion criteria . No subjects were excluded on the basis of race or ethnic background. Written informed consent was obtained from one parent of all subjects (see Appendix B), and written informed assent was obtained from all subjects (see Appendix C).

Experimental Design
Prior to participation in the study, each subject was screened to test whether I 1 they met the inclusion criterio n via a telephone screening. Once the screening was passed, each subject part icipated in a single, individua l session that lasted between 90 to 120 minutes in length. Testing took place in quiet, private rooms located at the subject's home or in the Department of Communicative Disorders. During the session, only the subject and the investigator were present.
Standardized Tests: All participants comp leted a batter y of standardized cognitive tests prior to the experimental task on prob lem-solving. Each testing se_ ssion was initiated by the comp letion of a case history interview to gather data on the subject 's educational background, medica l history , and social behavior (see Appendix D). This was followed by four subtests from the Wechsler Memory Scale-Revised (Wechsler , 1987) including the Visual Paired Associates I and II, and the Verbal Paired Associates I and II, and the 36-item Rav en Coloured Progressiv e Matri ces (Raven, Raven, & Court , 1998). Comp lete descriptions of each measure of cognitive ability is availab le in Appendix E. Mean scores for each test are reported in Table 2.  (We schler, 1987), and the Coloured Progressive Matric es (Raven, Raven , & Court, 1998 The following instructions were read by the examiner preceding the initial game and were only repeated prior to games two and three if requested by the subject: "We are going to play a question-asking game. I am thinking of one of these pictures (gesturing to the board) and your job is to figure out which one it is. You can ask any question you want so long as I can answer it "yes" or "no." Try to ask as FEW questions as possible.
When you are ready, go ahead and ask your fust question. There is no time limit for this test." As subjects asked each question , the examiner responded "yes" or "no" accordingly. Each time one or multiple pictures were eliminated, the examiner covered the targeted pictures with small black cards. For example, if the target picture was the chair and the question was "Is it a living thing?," the examiner would respond "No" and cover all the picture s of live objects. Conversely, if the target picture was the grasshopper, and the same question was asked , the examiner would respond "Yes" and cover all the pictures of inanimate objects . If a question was asked that could not be answered "yes" or "no," the examiner informed the subject to rephrase the question and be sure it can be answered "yes" or "no." Additionally , if the examiner was unsure which pictures were targeted by a specific question , the subject was prompted to identify the intended picture(s). The subject was allowed to correct the examiner at any time if a picture was covered or not covered correctly . If during the administration of RAPS a subject asked only questions that were guesses (e.g., "Is it the cat?"), a maximum of IO questions was permitted before the examiner responded "Yes" to the final guess in order to end the problem. Throughout the administration of RAPS , the examiner recorded all questions and eliminated pictures.

Scoring
The problem-solving skills required to solve RAPS are calculated based on three separate scores: 1. the total number of questions needed to solve the problem; 2.
the percentage of constraint-seeking questions used; and 3. the question-asking efficiency scores for questions 1 to 4. Each scoring procedure is described below. See sample score sheet in Appendix H.
Total number of questions-Each problem was considered solved when the game board was reduced to two pictures. It was from this point that the number of questions asked was tallied. Only questions that could be answered "yes" or "no" were calculated in the total.
Percentage of constraint-seeldng (CS) questions-Each question asked was coded as a constraint-seeking question (CS), or a guess. CS questions refer to those that eliminate more than one picture and target categories or groups of pictures on the game board (e.g., "Is it an instrument?" or "Is it a living thing?"). CS questions are considered to be more efficient than guesses. There are two type of guesses, frank guesses (G), and pseudo-constraint seeking questions (PC). PC questions are questions that only eliminate one picture, but are phrased like CS questions (e.g., pictures, the number 4 was used to calculate the QE score. It did not matter if the answer was "yes" (which would eliminate only 4 pictures) or "no" (which would be lucky and eliminate 12 pictures) , the smaller of the two numbers was used. In this example, the QE score would be calculated by the following equation: (4/16) 2 = 50%.
The most efficient question asked would have resulted in a QE score of 100% (e.g., if the question targeted 8 of the 16 pictures , the QE score would be (8/ 16) 2 = 100%).
In previous normative studies using RAPS , an intra-scorer and inter-scorer reliability testing for RAPS was conducted on 53 solved RAPS problems. Reliability on the scoring of the number of questions asked , question clarification , percent of CS questions , and question-asking efficiency was considered. Intra-scorer and interscorer reliability ranged from 95.2-99% (Marshall et al., 2003) for all of the measures .

Statistical Analyses
To determine if there are differences in performance between the three administrations of RAPS , the data was analyzed on a problem -by-problem basis . If subjects performed differently between administrations ( e.g., improved in performance between the first problem and the third problem) , then this may indicate that a learning effect has occurred across the administrations and the test itself may not be a stable indicator of the skills (i.e. problem-solving) it is proposed to measure . Therefore , a one-way analyses of variance (ANOVA) was conducted between each administration of the RAPS to determine if there were learning effects for the total number of questions used to solve the problem, the percentage of questions that are constraint-seeking, and the question-asking efficiency. Since there were no learning effects (which was predicted based on the Marshall et al. study), all three administrations of RAPS was used to analyze performance. Performance was analyzed as follows: Total number of questions-The total number of questions asked for each administration of RAPS was summed. Therefore , the score used for analyses ranged from 3 (if the subject guessed the correct answer on the first try for each of the three problems administered) to 30 (if the subject asked 10 questions and the problem was ended on all three administrations).
Percentage of constraint-seeking questions-The percentage of constraintseeking questions was obtained by dividing the total number of questions asked across the three problems by the total number of CS questions across the three problems.
The range of scores was 0% (if the subject did not ask any CS questions for each of the test administrations) to 100% (if the subject asked all CS questions for each of the test administrations).
Question-asking efficiency-The question asking-efficiency scores for the first four questions of each RAPS administration was averaged. If any subject solved a problem (i.e. the RAPS game) with less than four questions, then the remaining questions (from the remaining problems) was used for analyses.
After completing these calculations , a series of ANOVAs were conducted for each measure across the age groups to determine if subjects demonstrated differences in problem-solving ability as a function of age.

Descriptive Analyses
Types of Questions: To further analyze the RAPS problems , each question was analyzed according to the type of question asked. Category -focused questions are questions that targeted an entire category on the board, part of a category , more than one entire category , or those that narrowed the field after a category was targeted. Each category-focused question was labeled as a multi-category, whole-category, part-category , or narrowing question. Narrowing questions occured after a question identified the category containing the target picture or when there was only one category of pictures remaining. Each non-category focused question was coded as color-based (those that asked if the items were black and white or colored) , idiosyncratic (those that were non-categorical, targeting a large number of pictures such as "Is it bigger than a bread box?" or "Does it move?") , or novel (those that pull together items based on descriptors rather than labels such as "Is it round?" or "Is it furry?"). Each guess was labeled as a frank guess ( e.g., "Is it the cow?") or a pseudo-constraint question , which sounded like it would constrain more than one item, but ultimately only targeted or eliminated one item, therefore it was a guess ( e.g., "Is it an instrument that you beat on?").

Reliability
Interexaminer reliability checks were completed on 15 of 60 solved problems (70 questions) to determine measurement repeatability for the number of questions asked metric and the calculated QE scores. Two members of the research team counted the number of questions needed to solve the problem and calculated questionasking efficiency scores for each of the first four questions approximately 2 months apart. Interexaminer reliability checks were made by comparing point-to-point agreement for question counts and for the calculated QE scores for the two examiners.
The percent of interexaminer agreement was 98.57% and 97% respectively.

Learning Effects
To examine iflearning occurred across administrations of the three RAPS problems , the data was first analyzed on a problem-by-problem basis. Table 3a and 4.95 questions for problems 1, 2, and 3, respectively (see Table 4). ANOVA calculations determined that there was no significant differences in the number of questions asked for each of the three problems [F(2,57) = 1.249, p = 0.295]. The mean percent of CS questions asked across problems 1, 2, and 3 were 92% , 88.5%, and 87.51 %, respectively (see Table 4). ANOVA results revealed that there was no significant differences for percent of CS questions across the problems [F(2,57) = 0.333, p = 0.718]. The QE scores were calculated for subjects for the first four questions asked for each of the three problems . Mean QE scores for problems 1, 2, and 3 were 58.83% , 60.20% , and 60.24% , respectively (see Table 4). A single factor ANOVA for the four questions for problems 1, 2, and 3 was conducted to examine differences in QE scores across problems. A total of eleven subjects solved at least one of the problems with fewer than four questions. In these instances, the QE score used for analysis was calculated by averaging the remaining data points . ANOV A results revealed no significant difference in QE scores across games [F(2,57) = 0.041 , p = 0.960].    In summary , results for all measures obtained across the three administrations of RAPS revealed no significant differences , suggesting that subjects did not improve in their performance with the repeated trials. Therefore, each of the RAPS measures were analyzed for all three problems to determine if differences occurred between age groups.

Number of Questions Asked Per Problem
The total number of questions asked for each of the four age groups (10-11, 12-13, 14-15, and 16-17) were summed across the three RAPS administrations. Mean scores were 15.8, 13.6, 14.4, and 14.4 for each age group, respectively (see Table 5).
A single factor ANOV A comparing groups by number of questions asked, revealed that there were no significant differences between the total number of questions asked across the four age groups [F(3,16) = 0.402, p = 0.753]. Figure 1 shows the mean scores for each group noting that subjects aged 12-13 asked the fewest number of questions while subjects in the 10-11 year-old age group asked the greatest number of questions to solve RAPS problems.

Percent of Constraint-Seeking Questions Asked Per Problem
The total number of constraint-seeking questions asked per problem were calculated by dividing the total number of questions asked across the three problem s by the total number of CS questions across the three problems. Mean scores were 79.66%, 90.54% , 90.03% , and 93.48% for each of the four age groups, respectively (see Table 5). A single factor ANOV A revealed no significant differences between age groups and the percentage of CS questions asked [F(3,16) = 0.700, p = 0.566].
Although the percent of CS que stions did not increase incrementally as the children got older, the 16-17 year-olds had the highest percentage (93 .48% ), while the 10-11 year-olds had the lowe st percentage (79.66%) . CS percenta ges across age groups are illustrate d in Figure 2.

Question-Asking Efficiency Scores
The question-asking efficiency scores were calculated for each of the age groups by averaging all the QE scores together for the first four questions of each RAPS administration. Mean QE scores were 51.29% , 59.98% , 66.56%, and 62.56% , respectively (see Table 5). Again , a single factor ANOVA on QE scores found no differences between age groups [F(3, 16)= 1.369, p= 0.288]. Figure 3 shows the total QE scores for each age group.  Figu re 3. Mea n questio n-aski n g efficiency scor es on RAPS across age groups .
In conclusion, the main findings from this study were: 1. Subject s do not learn the task during repeated administrations , therefore there is no difference in performance on the first trial of RAPS compared to subsequent trials; 2. There were no differences in performance as a function of age, and 3. There were no significant differences in the efficiency of questions asked as subjects proceed through the task ; performance did not improve in a linear fashion. This suggests that the mean amount of information gained (i.e. the efficiency of questions asked) did not improve as subjects increase in age.

Test-Retest Reliability
Ten subjects participated in a repeat administration of three RAPS problems two weeks after the initial administration. A paired two sample for means t-test was conducted between the first and second administration of RAPS on the total number of questions asked, percentage of constraint-seeking questions, and QE scores to determine if performance improved between the test and the retest administration .

Question Type and Form
All of the questions from the participants were yes/no questions and no reinstruction was required. In solving 60 RAPS problem s, participants asked a total of 291 questions. Each of the questions were identified as category-focused (CF), noncategory focused (NCF), or a guess. CF questions included those that targeted an entire category on the board , part of a category, more than one entire category , or those that narrowed the field after a category was targeted. NCF questions included those that asked if the items were black and white or colored , were idiosyncratic, or were novel (meaning tho se that pull together items based on descriptors rather than labels such as "Is it round or Is it furry?") . Of the 291 que stions, 175 (60 .14 %) were clas sified as CF questions , 75 (25.77 %) as NCF questions , and 41 (14.09 %) as guesses (see Figure 4). The 75 NCF questions often incorporate d pictures from more than one category. There were 37 (49.33 %) novel que stions which are que stions that tar get a group of items by describing a characteri stic (e.g., "Is it round ?" or "Is it furry ?"), location (e.g., "Is in found outside ?"), or movement (e.g., "Doe s it fly?" or "Does it spin?") . There were 22 (29 .33 %) NCF questions based on color (e.g., "Is it black and white?") .
Finally, there were 16 (21.33 %) NCF question s that were idiosyncratic ("Is it alive?" or "Is it bigger than a breadbox ?") .
Guessing Figure 4 shows that 41 of the 291 (14.09 %) questions asked were guesses.
Although guessing accounted for a relativ ely small portion of all the questions , the point in the question-asking sequences where guessing occurs appears to be important.
The reason that is most participants used a CF strategy to solve RAPS problems. In such cases , the participant will ultimately receive a "yes " answer that will inform him or her of the tar get picture 's category (e.g., " Is it an animal ?") . In thes e instance s, the participant has two choices. The more efficient strategy is to ask a narrowing question to further reduce the number of picture s under consideration (e.g., "Doe s it live on the farm?"). The less efficient strategy is to try to guess the target picture by asking a frank question or a pseudo constraint-seeking question (a question that sounds like it will constrain more than one item, but ultimately only targets or eliminates one item ; therefore , it is a guess). It appears that when there is a choice to make between asking a narrowing question and guessing , the likelihood of a participant asking a narrowin g question (69.57 %) was hi gher than that of guessing (30.43 %).

CHAPTER IV. DISCUSSION
This study was designed to assess the problem-solving abilities of adolescents, ages 10-17, by measuring performance using the newly designed RAPS (the Rapid Assessment of Problem Solving, Marshall and Karow , 2001) . The tool is a modified version of the 20Qs Test (Mosher and Hornsby, 1966) and requires subjects to determine a pre-selected target picture from a total of 32 pictures by only asking questions that can be answered "yes" or "no." Examinees are instructed to use as few questions as possible in order to solve each problem. The test design and format restrict individuals in such a way that the task is one of problem-solving , a component of executive functioning, placing demands on skills such as organization, goal formulation , and working memory.
Normal children between the ages of 10 and 17 solved RAPS problems with approximately 4.86 questions per problem . Approximately 60% of the questions asked were category-focused. Children do not solve RAPS problems optimally , where they eliminate half of the picture board with each question asked. However , they do perform similarly to adult normal subjects and eliminate on average a little less than one-third of the remaining picture s with each subsequent question. The RAPS QE measure has quantified this skill and the subjects from this study were considered 60% efficient in completing RAPS problems. The efficiency of solving a RAPS problem can also be measured by determining the percentage of questions that target more than one picture. The children in this study asked constraint-seeking questions approximately 88% of the time. This is somewhat higher than adults in previous RAPS studies (Marshall et al., 2003) that asked CS questions only 80% of the time .
Finally , children do guess while completing a RAPS problem; however , this occurs infrequently, approximately 14% of the time.
An important finding to support the methodology and scoring of RAPS is that normal children do not improve their performance on successive administrations. The results demonstrating no learning effects were similar to previous studies (Marshall et al ., 2003), and therefore data from all three administrations of RAPS could be analyzed collectively.
Compared to previous research that found age related differences in the performance of children completing the 20Qs Test, this study found no differences among the four age groups for any of the RAPS measures. This finding may be interpreted in a number of ways. The first possibility is that there are differences between age groups that were not identified because the number of subjects per group was small (n=5). There was a tendency for the youngest group to have less similar scores on RAPS compared to the other three groups collectively, particularly for the number of questions asked and the percent of constraint-seeking questions. Perhaps, performance was reflective of the fact that most of the children had reached adult level performance on the skills necessary to complete the task. Problem-solving skills are dependent on executive functions. Welsh et al. (1991) reported speeded responding developed by age 6, hypothesis testing and impulse control reaching adult levels by age 10, and planning ability by age 12. Additionally, Chelune and Baer (1986) reported concept formation to improve significantly between the ages of 6 and 12. Passler et al. (1985) reported that strategic and planful behavior is mastered by age 12.
Therefore , RAPS performance may mirror these skill level s and most of the adolescent subjects may already have developed them , accounting for the lack of differences in the groups.
Another possible interpretation is that RAPS differs from the 20Qs Test in several dimensions, which could lead to the different findings. Half of the RAPS pictures are colored and half are black and white, rather than strictly black and white she has already asked.
The manner in which the subjects went about solving the RAPS problems provided was analyzed descriptively. Although children were not "optimally" efficient (i.e. eliminating half of the pictures on the game board-rated as 100%), eight of the 20 subjects tested eliminated 16 out of the possible 32 pictures with the first question asked. It was also noted that 3 of the questions asked were multi-category labeled questions that targeted 2 or more categories (e.g., "Is it an animal or a bird?") and six targeted multiple items without labeling the categories in the question , for example "Is it used for entertainment?" or "Can you play with it?", which included all the sports and all the musical instruments. Perhaps this question strategy was rarely used because it involved identifying the existence of more than one category of pictures at the same time, requiring more complex cognitive skills.
Further examination of the types of questions children asked revealed several other findings. There were two factors that appeared to influence question-asking efficiency scores of many of the participants. The first was to initiate problem-solving effort by asking an optimal question, namely one that would eliminate half of the pictures from consideration. To do this , the individual needed to use information available from the problem-solving board in planning . This could involve recognizing that the pictures are arranged in rows and columns and that half of the pictures are colored and half are not. Moreover, initial question-asking efficiency could be improved by asking a multiple category question ( e.g., "Is it an animal , clothing, or dessert?"). As mentioned, participants rarely targeted more than a single category ( only 9 of 291 questions). Another important piece of information available to the participants is that the problem-solving boards contain more pictures of some categories than others. Thus , efficiency can be improved by first asking question s targeting the larger categories. Thirty-seven percent of the questions asked targeted more than 8 items. Twenty-six percent targeted eight items, 23% six items, and 6% four items. This demonstrated that many of the questions did target larger categories with 73% of the questions targeting eight or more items. This ability to ask "efficient question s" can be attributed to the manner in which participants use information from the problem-solving board to plan how to solve a RAPS problem.
Participants asked primarily constraint-seeking questions in solving RAPS problems and most of these were category-focused. Ultimately, a CF approach will identify the target picture's category. Once this occurs, the individual needs to make a strategy shift and ask a narrowing question to avoid using extra questions. Normal participants varied in their ability to shift to a narrowing question. Two possible reasons for this come to mind. One is that once the target picture category is known , the participant becomes impulsive and is prone to gamble with a guess. A second is that the individual is unable to shift to a question that focuses on features common to members of a category and render a guess. Examination on where guessing occurred in the question-asking sequence revealed that 41 guesses were made and only 10 occurred after the category was known. This result differed from performance for adults completing RAPS and suggests that children were better able to make a strategy shift, thereby narrowing the targeted category.

Future Implications
There are few clinical measures available to assess problem-solving skills in children. Anderson (2002) reports that there are both theoretical and practical challenges in evaluating executive functions. One problem is the lack of normative data on commonly used tests, which were developed for adults. Second, many assessment measures are very structured tests and do not allow for flexibility and evaluation of outcomes that are foundation skills in problem-solving . RAPS has been found useful for assessment of clinical populations in adults . It is fast, easy to administer and motivating to the examinee because of its "game-like" structure (Marshall et al., 2003). Because it appears that adolescents demonstrate adult-like Your child has been invited to take part in a study that will examine his or her use of problem-solving strategies when presented with a "game-like " activity that is very similar to the Twenty Questions game. The purpose of this study is to learn more about how children solve problems.
If you decide that your child will take part in this study, here is what will happen: your child will be asked to participate in one or two individual sessions that will last for approximately one hour. During the sessions, your child will be presented with four different activities that include looking at pictures, sorting cards and designs , remembering words, and answering brief questions.
There is no foreseeable risk to your child. However, if your child experiences any discomfort during any part of the study, he or she should let the researcher know immediately.
Although there will be no direct benefit to your child for taking part in this study, the researcher hopes to learn more about cognitive functions that are related to problemsolving skills. By helping the researcher learn more about problem-solving , a better understanding of this cognitive function may lead to the development of better assessment tools of problem-solving ability for adolescents. If interested, the results of your child's performance will be shared with you, and any publications generated from this study will be available at your request.
Your child's participation in this study is confidential. None of the information will identify him or her by name . All records of your child ' s participation will be kept with a number or letter code , and any publications resulting from this study will identify your child only with this code. All paper records will be stored in locked cabinets.
The decision for your child to take part in this study is up to you. Your child does not have to participate. If you decide that your child will take part in the study, he or she may quit at any time without any penalty. If your child wishes to quit , you or your child can simply inform Dr . Colleen Karow (401)   What type of hobbies/activities does the child participate in?
What type of physical activities does the child participate in?

Prenatal and Birth History
Briefly describe mother 's general health during pregnancy (illness, accidents, medications , etc.)

Length of Pregnancy _________ Length of Labor _______ _ _ General Condition
Birth Weight_~~-------Type of Delivery □ Head First D Feet First D Breech D Cicerian Section Were there any unusual conditions during pregnancy or birth? If yes, please describe .

D Yes D No
Provide the approximate age at which the child began to do the following: Crawl _______ Walk Sit _______ Stand Feed Self _______ Dress Self Single Words ______ _ Combine Words -------Briefly describe the child's speech and language skills Briefly describe the child's motor skills (gross and fine motor skills) Briefly describe the child's social skills

Educational History
School -------------------------   (1987) The WMS is a comprehensive test of cognition that was designed as a diagnostic and screening measure to be used as part of a neuropsychological examination. This test looks at short-term learning and delayed recall of verbal and figural information or items. Four sub-tests will be used: The CPM is a test of problem-solving abilities. It was developed to measure eductive ability , which involves the ability to make meaning out of confusion and going beyond the given to perceive that which is not immediatel y obvious. Subjects are presented with three sets of 12 patterns with missing pieces. For each pattern , subjects are to select the matching piece from six options shown below the patt ern .
Each set involves different types of patterns : "Set A" uses continuous patterns , "Set Ab" requires subjects to distin guish discreet figure s as spatial ly related wholes to complete the pattern , and "Set B" uses figural analogies. Mosher and Hornsby (1966