UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The relationship between intelligence and attention in kindergarten children Carter, John D. 1992

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1992_spring_carter_john_dale.pdf [ 3.33MB ]
JSON: 831-1.0086629.json
JSON-LD: 831-1.0086629-ld.json
RDF/XML (Pretty): 831-1.0086629-rdf.xml
RDF/JSON: 831-1.0086629-rdf.json
Turtle: 831-1.0086629-turtle.txt
N-Triples: 831-1.0086629-rdf-ntriples.txt
Original Record: 831-1.0086629-source.json
Full Text

Full Text

THE RELATIONSHIP BETWEEN INTELLIGENCE AND ATTENTION INKINDERGARTEN CHILDRENbyJOHN DALE CARTERB.Sc. Washington State University, 1975M.A. The University of Alberta, 1982A THESIS SUBMITTED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF EDUCATIONinTHE FACULTY OF GRADUATE STUDIESEducational Psychology and Special EducationWe accept this thesis as conformingto the required standardCOLUMBIA7 January 1992© John Dale Carter, 1992In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.Department of____________________The University of British ColumbiaVancouver, CanadaDate___DE-6 (2/88)ABSTRACTThe purpose of this study was to compare two conflicting theoreticalperspectives on the relationship between intelligence and sustained attention. Thecognitive resources theory assumes that lower IQ subjects are required to allocategreater amounts of their limited attentional resources during information-processingtasks than higher IQ subjects. The arousal theory assumes that there is anoptimal level of arousal associated with task performance, and that an increaseor decrease in arousal produces impairment in performance. Additionally thearousal theory predicts that increased time on task leads to a decrement inarousal as a function of IQ levels.Signal detection theory applications were used to operationalize andcompare the two theories. Specifically, the signal detection parameters of sensoryacuity (d’), the decision criterion (a), correct detections, and false alarms wereused to determine subject performance across three time periods (2, 4, and 6mm.) on a visual continuous performance task.Twenty-nine teacher-nominated at-risk for learning difficulties andtwenty-nine normally achieving kindergarten students were adminstered theStanford-Binet:Fourth Edition (SB:FE) and the Wechsler Preschool and PrimaryScale of Intelligence-Revised (WPPSI.R), as well as the Gordon Diagnostic System(GDS) Vigilance Task. The GDS is a standardized behaviour-based measure ofsustained attention.UThe results of this study were interpreted as suggesting that abilitygroup differences reflect attentional capacity. Two findings were important in thisinterpretation. First, regardless of IQ, the groups varied on the signal detectiondiscrimination index. Second, these measures did not vary over time in eithergroup. Thus, the arousal theory was not supported.IQ and attention intercorrelation patterns were higher for the at-riskgroup compared to the normally achieving group. Exploratory maximum-likelihoodfactor analyses indicated that intelligence plays a greater role in relation tovigilance for the at-risk for learning difficulties group than the normal achievinggroup.‘UTABLE OF CONTENTSAbstract iiTable of Contents ivList of Tables viList of Figures viiAcknowlegernents viiiI. Introduction 1A. Purpose of the Study 1B. Rationale for Sample Selection 3C. Signal Detection Theory 6D. Definitions and Terms 8H. Literature Review 10A. Historical Overview 10B. The Arousal Theory 11C. Resource Limitation Theory 13D. Attention and Academic Abilities 16E. Signal Detection Applications 21F. Hypotheses 23ifi. Methodology 26A. Subjects 26B. Examiners 32C. Procedure 32D. Data Analyses 34IV. Measurement Scales 36A. Teacher Ratings of Ability and Behaviour 361. School Learning Profile 36B. Gordon Diagnostic System 381. The Vigilance Task 40a. Summary Scores Definitions 41b. Signal Detection Calculations 432. The Delay Task 45a. Summary Scores Definitions 47C. Wechsler Scales of Intelligence 491. The WPPSI 492. The WPPSI-R 51D. Stanford-Binet Intelligence Scales 531. Age-Scale and Point-Scale Formats 532. The Stanford-Binet: Fourth Edition 55E. Rationale for Two Intelligence Measures 58V. Results 60ivA. Intelligence Measures.60B. Vigilance Measures 65C. Intelligence and Vigilance Relationships 681. Hit Rate and Intelligence 692. False Alarms and Intelligence 733. d’ and Intelligence 764. and Intelligence 79D. Intercorrelational Patterns 821. Variable Selection Rationale 832. Matrices Comparisons 843. Average Correlation Comparisons 85E. Factor Analyses 871. Normality Assumptions 872. Maximum-Likelihood Factor Analyses 90F. Review of Hypotheses 96VI. Discussion 98A. Review of the Purpose of the Study 98B. Ability Group Differences in Vigilance 100C. Time on Task 101D. Intercorrelational Patterns 103E. Implications of the Study 104F. Limitations and Future Research 108References 112Appendix A: Letters to Teachers 126Appendix B: Student Learning Profile 129Appendix C: Parent Permission Form 133Appendix D: Intercorrelation Tables 137Appendix E: WPPSI-R and SB:FE Correlations 146VLIST OF TABLESTable 1: Occupational Comparisions 31Table 2: WPPSI Subtests 50Table 3: WPPSI-R Subtests 52Table 4: Theoretical Model of the SBFE 56Table 5: WPPSI-R Means & Standard Deviations by Group 62Table 6: SBFE Means & Standard Deviations by Group 63Table 7: Mean Percentage of Correct Detections,False Alarms, d’, and i3 by Group 64Table 8: At-Risk Group Hit Rate Correlations with IQ 71Table 9: Normal Group Hit Rate Correlations with IQ 72Table 10: At-Risk False Alarms Correlations with IQ 74Table 11: Normal False Alarms Correlations with IQ 75Table 12: At-Risk d’ Correlations with IQ 77Table 13: Normal d’ Correlations with IQ 78Table 14: At-Risk 3 Correlations with IQ 80Table 15: Normal j3 Correlations with IQ 81Table 16: Intercorrelations of Intelligence and Vigilance 86Table 17: Normality Statistics 89Table 18: Maximum Likelihood Factor Analyses 95viLIST OF FIGURESFigure 1: Student Learning Profile 28Figure 2: Signal Detection Theory 44viiACKNOWLEGEMENTSAs a marathon runner I am all too familiar with the time and toil ittakes to travel the distance. This dissertation represents a reasonable analogy.Although I cross the finish line as an individual, in this marathon, like all theothers, I am indebted to many coaches, supporters, and cheerleaders.My main mentor, Dr. H. Lee Swanson, has run the Denver Marathonwhich is a hard undulating race. Lee knows what it feels like when you wonderwhether “the wall” is approaching. Dr. Swanson’s genius as a scholar is onlysurpassed by his kindness and positive enthusiasm. Lee taught me that the racedoes not necessarily go to the swiftest, but to those that keep running. I respectLee’s intellect and value his friendship.Drs. Julianne Conry, Robert Conry, and William McKee served assupervisory committee members. Their support, effort, and encouragement helpedme see this project through to the end. Liz Harris helped calculate the and d’scores (we therefore know that the margin of error is negligible). Dr. PatriciaArlin, Head of Educational Psychology and Special Education refused to let mequit when I was too tired to travel any further. Her advice was well taken.My parents, William and Enid, have coached and supported all myvarious and ubiquitous endeavours since early childhood. Everyone knows that itis not easy to raise a Type A personality. Nonetheless, itty parents have alwaysbeen there to encourage and offer there unconditional devotion.My boys, Drew Alan and Kyle Brooks Carter, are my inspiration anddreams for tomorrow. I know the world is a better place because they are here.Finally, this work is dedicated to Lynda, my partner in life’s long run.Without her support none of this would matter very much.viiiI. INTRODUCTIONA. PURPOSE OF THE STUDYThorngate (1990) states that “We must pay attention to be informed.Attention is what we ‘pay’ to get messages into and out of our heads, toconvert information into knowledge and knowledge into information” (p.263).Success in mastering classroom instruction generally requires the abilityto sustain attention for processing information requisite to learning (Edley &Knopf, 1987; Kupietz & Richardson, 1978; Samuels & Turnure, 1974; Simon,1982). Although attention may be related to success in learning, its relationshipto intelligence is unclear. Tomporowski and Simpson (1990) indicate that duringthe past thirty-five years nearly one thousand studies have assessed factorsrelated to sustained attention with relatively few studies assessing the relationshipbetween intelligence and attention. Moreover, the findings of these studies areinconsistent. Some studies have found a significant relationship between intelligenceand attention (Carlson & Jensen, 1982; Herman, Kirchner, Streissguth, & Little,1980; Noland & Schuldt, 1971; Sen & Sachdev, 1977; Simon, 1982; Swanson &Cooney, 1989; Tomporowski & Simpson, 1990); while other studies report thatthere is a non-significant relationship between intelligence and attention (Gale &Lynn, 1972; Mackworth, 1969; Myors, St.ankov, & Oliphant, 1989; Stankov,1989; Wilkinson, 1961). Thus, the existing literature is unclear whether individualdifferences in sustained attentional resources relates to primary mental abilities orother cognitive processing variables.1Introduction / 2In the literature concerned with attention and intelligence two competingtheories are prominent: one is the limited cognitive resources theory, while theother is the arousal theory.The cognitive resource model assumes that individual differences inattentional resources underlie intelligence. That is, attentional resourcest underlie ageneral (g) factor (Hunt, 1980). Tomporowski and Simpson (1990) suggest thatthe availability of attentional resources during a sustained attention task isrelated to inteiligence and that “lower IQ individuals are required to allocategreater amounts of their limited attentional resources during effortfulinformation-processing activities than higher IQ individuals” (p.38). Empiricalsupport for the attentional resource model has been found in a few studies (e.g.,Carlson & Jensen, 1982; Noland & Schuldt, 1971; Sen & Sachdev, 1977;Swanson & Cooney 1989; Tomporowski & Simpson, 1990; also see Stankov,1983, for a review); while others report a lack of support for the relationshiplinking attention to intelligence (e.g., Gale & Lynn, 1972; Mackworth, 1969;Myors et al., 1989; Stankov, 1989; Wilkinson, 1961).The arousal model assumes that there is an optimum level of arousalassociated with optimum performance, and that either a decrease or increase inarousal produces impairment in performance (Loeb & Alluisi, 1984). According toSwanson and Cooney (1989), the arousal theory assumes that as more time isspent on a sustained attentional task, negative correlations emerge betweenattentional performance and intelligence. That is, arousal levels interact with timet In this study the terms “resource” and “capacity” wiU be used interchangably.Introduction I 3on task. Stankov (1983) suggests that the monontonous conditions of anattentional task causes people with higher scores on intelligence tests to performworse in the latter phase of the task because they become bored and thus losearousal. Therefore, increased time on task leads to a decrement in arousal as afunction of higher intelligence levels.B. RATIONALE FOR SAMPLE SELECTIONThe purpose of this present study is to compare these two competingtheoretical models (resource capacity versus arousal) as an explanation ofindividual differences in the relationship between sustained attention andintelligence. To operationalize these two competing theories, the present studyutilizes a signal detection methodology (the application of this methodology will bediscussed further in the methodology section). To address the individual differenceissue, two groups of kindergarten children (one group judged as “at-risk” forlearning difficulties and the other as normal nonhandicapped students) areassessed on measures of intelligence and sustained attention. Tomporowski andSimpson (1990) explain two approaches used to determine the relationshipbetween intelligence and sustained attention. One approach is to measure theperformance of observers within the normal IQ range; the second has been calledthe extreme group approach in which mentally retarded and nonretarded personsare compared.The present study uses a third approach comparing a teacher-nominatedat-risk for learning difficulties group with a normal group. The two groups,Introduction / 4normal nonhandicapped and teacher nominated at-risk for learning difficulties,were selected in order to test the hypothesis that there will be qualitativelydifferent correlational patterns between the groups intelligence and attention.There is indirect support for the assumption that differences incorrelations may exist. Although it is assumed that correlations among mentaltests would be about the same in a group of low IQ subjects as it would be ina high group of IQ subjects, Detterman and Daniel (1989) state that thisassumption is incorrect. Using the standardization data from the Wechsler AdultIntelligence Scale - Revised (WAIS-R) and Wechsler Intelligence Scale for Children- Revised (WISC-R), they found that correlations declined systematically withincreasing IQ. That is, the correlations among the WAIS-R and WISC-R subtestsare higher for low IQ subjects than for high IQ subjects. They further statethat:Higher correlations among low IQ subjects may also explain why IQtests seem to rind more uses at the low end of the IQ scale than atthe high end. The WISC-R and WAIS-R, and perhaps other tests, willbe more ‘g’-loaded at the low end of the distribution than at the highend. That is, a general factor should account for more of the totalvariation among low ability subjects than it does among high.Conversely, high ability subjects will show more subtest scatter(p.358).In the present study, teacher-nominated children at-risk for difficultieslearning school materials were compared with normally achieving children. TheIntroduction I 5decision to compare normally achieving subjects with at-risk for learningdifficulties allows one to test the generality of Detterman and Daniel’scorrelational pattern of retarded to a sample that contains learning disabled, anddevelopmentally delayed students. Thus, teachers were invited to nominate childrenwhose ability to learn school materials was judged as “high risk”. These teacherjudgements were based on their cumulative teaching experience and the teacherswere not required to nominate a specific number or percentage of students. Someteachers nominated many students, some none.A focus was also placed on Kindergarten children since the relationshipbetween attentional tasks and intelligence may be more important at youngerages (e.g., Edley & Knopf, 1987; Murphy-Berman, Rosell, & Wright, 1986;Simon, 1982; Swanson, 1983; Wickens, 1974). Levy (1980) reports that childrendevelop more efficient levels of attention as they get older: “the development ofmore efficient strategies of children’s attention with age has been shown by anumber of studies over the last ten years” (p.77). Previous research has found aweak relationship between intelligence and sustained attention for children betweenthe ages of seven and fifteen (e.g., Berch & Kanter, 1984). In contrast, Hermanet al. (1980) have provided some empirical evidence showing intelligence andsustained attention are positively related for preschool children. Moreover, Hermanet al. (1980) suggest that “perhaps intellectual and vigilance behavior aredevelopmentally, if not ultimately, related in some way” (p.867).Introduction I 6C. SIGNAL DETECTION THEORYThe purpose of this present study is to compare these two competingtheoretical models (resource capacity versus arousal) as an explanation ofindividual differences in the relationship between sustained attention andinteiligence. To operationalize these two competing theories, the present studyutilizes a signal detection methodology.Swets (1986b) notes that signal detection theory was originally developedas a mathematical theory for the process of detecting radar signals (Peterson,Birdsall, & Fox, 1954). However, signal detection theory was soon found usefulin understanding the behaviour of human observers of visual and auditory signals(Tanner & Swets, 1954). Hochhaus (1972) states that,At this point in the history of psychology the methodology of signaldetection theory has become established as a preferred technique forassessing a subject’s ability to discriminate the occurence of discretebinary events. The primary argument for its use has been theprovision for a discrimination index (d’) which is independent ofresponse bias (j3) factors. Bias refers to the fact that, independent ofthe stimulus, not all responses are equally likely, and as such biasshould not be confounded with d’. (p. 375).In 1990, Swets received the American Psychological Association’s Award forDistinguished Scientific Contributions:For his pioneering work on the theory of signal detection and itsapplication to psychophysics, work that caused a paradigm shift inIntroduction / 7sensory psychology in which the idea of a discrete threshold wasreplaced with that of a continuously-variable detection process.(American Psychologist, 1991, p. 294).The American Psychological Association commended Swets “For his role inestablishing clearly the importance of decision criterion as well as acuity as adeterminant of performance in any discrimination task”.The decision criterion ((3) “... specifies the optimal weighting of hitsrelative to false alarms” (Swets, Tanner, & Birdsall, 1961, p. 309). Thus, (3 isdetermined by the a priori probabilities of signal occurrence and the degree ofaction in a subject’s reporting an event as a signal. Acuity is represented by d’which is a measure of a subject’s attentiveness and sensory capabilities. Swets(1986a) describes d’ as a valid and reliable index of pure discrimination capacity.The ideal observer (cf. Davies & Parasuraman, 1982; Warm, 1984)would have a larger d’ measure relative to a smaller (3 measure. Thus, thelarger the d’ measure, the greater the discriminability of the signal, or observerattentiveness. Clearly, a subject with a high d’ scores shows evidence of highability in sustaining attention on an information-processing task. Whereas thehigher the value of (3, the observer shows a more conservative response criterion.Warm (1984) describes high (3 values as indicative of a conservative responsecriterion where the observer trades off correct detections (or misses) in order toavoid false alarms (or gain correct rejections). The observer’s response criterion((3) is determined by two kinds of correct decisions (correct detections and correctIntroduction / 8rejections) relative to the costs of the two possible errors (signal omissions andfalse alarms). A low 3 value indicates the subject is adept in sustainingattention and correctly deciding whether a signal is a correct response and not afalse alarm. Similarly, the low value indicates the subject is able to payattention with a low rate of omissions of correct signals and low false alarmreports.A further discussion of the mathematical formulas used in this study forcalculation signal detection measures is presented in the Measurement Scaleschapter.D. DEFINITIONS AND TERMSVigilance. N.H. Mackworth (1948, 1957) was one of the first of a longlist of researchers to study vigilant behaviour. Mackworth used the term“vigilance” a subject’s ability to direct attention and respond to stimuli over time.In the present study the vigilant behaviour of kindergarten students is measuredon a six-minute continuous performance task broken into three blocks of twominutes per block.Arousal. In this study arousal is operationalized similarly to the thesisdeveloped by Stankov (1983). That is, arousal mediates the correlation betweenintelligence and vigilance attention measures over time intervals. The arousalmodel assumes that the correlational patterns between intelligence and attentionbecome significantly stronger with increased time on task.Introduction / 9Resources. Navon (1984) explains, simply, that “the human processingsystem in toto is, by some loose definition, a resource, and because selectionamong stimuli, thoughts, or actions surely occurs, it may be viewed as thecommitment of that resource to those objects of processing” (p.2l7). That is,according to Navon, resource theory ascribes variability in performance of a taskto the amouiit of some limited internal input dedicated to the task. The presentstudy hypothesizes that resource-limited components of sustained attention(vigilance) are an important source of individual differences in kindergartenchildren’s intelligence. The resource model assumes that attention strength (d’)remains constant over time on task. Thus, correlations between intelligence and d’are hypothesized to be constant across time.Intelligence. In this study intelligence is operationalized as g and isequated with psychometric g (cf. Jensen, 1987). Intelligence is measured usingstandardized norm referenced tests. A further detailed discussion of these tests isdescribed in Chapter IV. However, at this point it is important to indicate thatintelligence is defined as psychometric g and relies on the face or content validtyof the standardized tests and their relationship with g.II. LITERATURE REVIEWA. HISTORICAL OVERVIEWWarm (1984) explains that theories of sustained attention or vigilancerange from psychophysiological accounts anchored in arousal or habituation notionsto accounts invoking motivational principles or operant conditioning principles tocognitive views featuring concepts such as expectancy formation, attentionalfiltering, decison making and automatic and control processing.The early beginnings of this area of attention research dates back toN.H. Mackworth (1948, 1957), who was the first to study the theoretical andpractical implications of sustained attention using controlled laboratory experimentsdesigned to measure vigilance behaviour. The British Royal Airforce, towards theend of 1943, asked Mackworth to conduct experiments to help determine theoptimum length of time a radar observer could maintain a watch without missinga detection or signaling a false alarm (for further historical reviews see:Broadbent, 1971; Davies & Parasuraman, 1982; Parasuraman & Davies, 1984;Warm, 1984).N.H. Mackworth used the term vigilance to describe a subject’s ability todirect attention and respond to stimuli over time. Mackworth advanced acomprehensive interpretation of vigilance relating it to principles of Pavlovianclassical conditioning. The analogy drawn with classical conditioning was that the10Literature Review / 11original conditioning occurred in the trial demonstration, and the subsequentdecline in performance represents the extinction period (Levy, 1980).Broadbent (1953) discussed classical conditioning and watch-keepingbehaviour similar to Mackworth, but instead of interpreting vigilance in terms ofclassical conditioning, he interpreted both in terms of attention. Broadbentexplained that stimulus selection was determined by the intensity of the stimulusand novelty. Broadbent accounted for decrement in accuracy by stimuluscompetition.B. THE AROUSAL THEORYFrankmann and Adams (1962) were among the First to theorize theconstruct of arousal as an explantion of vigilance behaviour. The arousal theorypredicts that there is an optimum level of performance related to an optimumlevel of arousal and that either an increase or decrease from the optimum levelof arousal produces impairment in performance.The theory is that a given stimulus will produce arousal, but withrepetition of the same stimulus or similar stimuli (e.g. non-signal stimuli differingonly slightly signal stimuli), these arousal responses habituate (Loeb & Alluisi,1984). J.F. Mackworth (1970) suggests that the vigilance decrement may beattributed to either a lowering of the arousal level or to habituation of thearousal reaction. Nonetheless, according to Richter, Senter, and Warm (1981) thisprogressive decline in performance (vigilance decrement) represents the mostLiterature Review I 12ubiquitous and consistent discovery in sustained attention research.Stankov (1983) used the arousal construct to explain vigilance decrementin relation to intellectual functioning levels. That is, Stankov (1983) predicted thatthere would be a differential performance level on an attention task where lowerIQ subjects will show less of a decrement than higher IQ subjects. Stankovexplains that this is due to the monotonous conditions of the attention taskwhich causes the people with higher intellectual functioning levels to performworse on the latter stages of the task because they become bored and thus losearousal more easily on a simple task than lower intellectually functioning people.Therefore, increased time on task leads to a decrement in arousal as a functionof intelligence levels and we can expect intelligence to correlate negatively withattention.With a view to deriving a prediction from the attentional resourcestheory, Myors et al. (1989) used two tasks known to be good measures of fluidintelligence as well as being able to be varied in such a way that theirattentional resources requirements could be controlled: Number Series and LetterSeries tests (cf. Holzman, Pellegrino, & Glaser, 1982, 1983). Myors et al. (1989)hypothesized that if subjects were divided into higher and lower groups withrespect to IQ scores, a two-way interaction between attentional performance andintelligence would emerge. That is, they expected the difference between the twoability groups would be small when the attentional load was small and shouldincrease linearly as the attentional load increased. It was expected that people inthe higher IQ group should be able to cope with an increased attentional loadLiterature Review I 13better that those in the lower IQ group.Myors et al. (1989) found the opposite results to their predictions of atwo-way interaction effect. The differences between the high and low abilitygroups were approximately the same regardless of the attentional load of thetasks. They concluded that in the absence of any significant interaction effectsthere was little support for the idea that a limited attent.ional capacity processingsystem underlies individual differences in intelligence. Thus, Stankov (1989)suggests in light of the evidence which questions the link between attentionalresources and intelligence, “our task should be to search for those cognitiveprocessing variables which affect a test’s correlation with intelligence” (p.967).C. RESOURCE LIMITATION THEORYResource theory, originally formulated as a theory of attention and ofthe limited capacity (e.g., Kahnemah, 1973; Navon, 1984; Navon & Gopher,1979; Wickens, 1984), has been used to explain individual differences in cognition(Myors et al., 1989). In this theoretical model it is assumed that individualdifferences in attentional capacity underlie intelligence (Stankov, 1989). The limitedresource theory assumes that people differ in the attentional resources theypossess and that these individual differences in capacity underlie intelligence(Stankov, 1989). Tomporowski and Simpson (1990) suggest that the availability ofattentional resources during a sustained attention task is related to intelligenceand that “lower IQ individuals are required to allocate greater amounts of theirlimited attentional resources during effortful information-processing activities thanLiterature Review / 14higher IQ individuals” (p.38).To test the proposition that the sustained attention of retarded andnonretarded subjects is related to the availability of attentional resources during avigilance task, Tomporowski and Simpson (1990) administered two sixty minutevigilance tests that differed in memory demand. These tests were designed ascontinuous matching tasks in which the subjects viewed a series of single digitspresented successively at the center of a CRT (memory set) and, following adelay, compared the memory set to a series of digits presented simultaneously(test set). The subjects were instructed to press the spacebar of a computerkeyboard when the two sets were identical.Tomporowski and Simpson (1990) predicted a significant interactionbetween IQ and time period. They expected that the two IQ groups would notdiffer during the early part of the vigilance task but, as predicted by Stankov(1983), differences would occur later in the task. Specifically, they expected thelower IQ group (retarded subjects) would miss more targets and commit morefalse alarms than the higher IQ group.As predicted, Tomporowski and Simpson (1990) found that neither targetdetection performance (hit rate), nor measures of detectability (d’), differedbetween the retarded and nonretarded subjects during the initial phase of thevigil; moreover, both groups evidenced similar vigilance decrements during thisperiod. However, again as predicted, during the final phase of the vigil, thedetectability of the retarded subjects declined more rapidly than the nonretardedLiterature Review 1 15subjects.Both IQ groups showed increased false alarm rates as the vigilprogressed. Tomporowski and Simpson (1990) explain that this suggests that theshifts in the sustained attention of subjects were attributable to a ehange indetectability (d’) rather than a change in response bias (a). Furthermore, theyindicate that “the increase in errors of commission during the vigil precludesneurological arousal or sensory habituation for the vigilance decrement, such asthose proposed by Broadbent (1971), Stankov (1983), and J. Mackworth (1969,1970)” (p.37).The purpose of this present study was to compare these two competingtheoretical models (resource capacity versus arousal) as an explanation for therelationship between sustained attention and intelligence. To address this issue,two groups of kindergarten children (one group judged as “at-risk” for learningdifficulties and the other as normal ‘nonhandicapped students) were assessed onmeasures of intelligence and sustained attention. Kindergarten children wereselected to test the theoretical issues since previous research has shown that thisage group is developmentally sensitive to attention (e.g., Edley & Knopf, 1987;Murphy-Berman, Resell, & Wright, 1986; Simon, 1982; Swanson, 1983; Wickens,1974).In the following section a review of studies that examine attention andacademic abilities is presented.Literature Review / 16D. A11ENTION AND ACADEMIC ABILITIESPsychologists commonly note that some children are more skillful insustaining attention to classroom instruction than others. That is, children whohave difficulty sustaining attention to classroom tasks and information processingare most likely to present as “at-risk” for learning difficulties. Moreover, Edleyand Knopf (1987) indicate that children who are attentive are more likely tosucceed academically than children who are distractible and that “the importanceof examining sustained attention for its potential as an early predictor oflearning problems in preschool youngsters should not be overlooked” (p. 341).This is particularly important considering the rmding of a significant relationshipbetween attention and achievement among students just beginning to read andwho have not established a history of academic failure (Samuels & Turnure,1974).In their study of attention and reading achievement, Samuels andTurnure (1974) constructed a behavior observation schedule to record theattentional behaviors during the reading hour of eighty-eight first graders (53boys and 35 girls) from four classrooms. An observer was assigned to each ofthe four classrooms to record task-relevant attentive behavior and inattentiveness.Attentive behaviors consisted of task-relevant acts such as orienting eyes to textor teacher, working on reading follow-up exercises, observing chalkboard oroverhead projection, or otherwise following the instructional directions of theteacher. Negative attentiveness consisted of nontask-orienting behaviors such asfailing to follow directions, closing eyes, and working or playing with nonassignedLiterature Review I 17materials. Interrater reliability was relatively high at 0.89 indicating observeraggreement on attentiveness and inattentiveness. Reading achievement wasmeasured by presenting 45 words, randomly selected from the Dolch (1956) listof basic sight words.Samuels and Turnure (1974) found that girls were significantly superiorin classroom attentiveness and reading word recognition. Also, they found thatincreasing degrees of attention were related to superior word recognition. Thus,although Dolch (1956) word recognition cannot be equated with readingachievement it does indicate that attentiveness stands as an important variablerelated to academic development in young students.Edley and Knopf (1987) examined sustained attention as a predictor oflow academic readiness in one hundred and sixty kindergarten students. Theirstudy was designed to determine the extent a continuous performance task (CPT),as a measure of sustained attention, could be used as an early predictor ofacademic achievement. The CPT is a visual task that requires the student tomonitor a series of letters presented one at a time at very brief intervals. Thestudents were asked to respond as quickly as possible to the appearance of aparticular letter (usually an X). Additionally, the interrelationships between theCPT and two other predictive measures were examined. The tests selected werethe Peabody Picture Vocabulary Test Revised (PPVT-R), a cognitive screeningmeasure, and the Metropolitan Readiness Test (MET), an academic readinessbattery. They expected the three predictor measures, the CPT, PPVT-R, andMET, would be positively correlated with each other as well as with theLiterature Review / 18teachers’ ratings of academic readiness.Edley and Knopf (1987) indeed found the three predictor measurescorrelated positively with each other as well as with the teachers’ ratings ofacademic readiness. Pearson correlations yielded the following correlations: ThePPVT-R and MRT (r = 0.52), the PPVT-R and CPT (r = 0.39), and the MRT andCPT (r 0.59), all of which were significant at the 0.01 level. Similarly, thePearson correlations between the three predictor measures and teacher ratings ofacademic development were also significant at the 0.01 level: The FPVT-R andTRAD (r=0.5), the MRT and TRAD (r0.58), and the CPT and TRAD(r=0.41). Edley and Knopf (1987) concluded that the three measures indicate“potential screening validity” (p. 349). Moreover, they explain that “the trueutility of a preschool screening tool comes not only from its validity as areadiness indicator, but from such factors as its quickness, ease of administration,and objectivity” (p.350). They report that the strength of the MRT as anacademic readiness battery was supported by the study. However, the CPT canbe objectively administered in as little as five minutes and that it assesses animportant school-related behavior which is missed by teacher ratings or readinesstests. Also, the CPT, as a measure of sustained attention, has the potential forapplication in developing intervention approaches for poor attenders and poorperformers.Simon (1982) used a vigilance task to determine school readiness ofpreschool children. This experiment used a visual monitoring task with thepurpose “as a diagnostic tool to identify children who are not yet ready to enterLiterature Review / 19the first grade” (p.102°). Using a sample of twenty (ten boys and ten girls) fiveyear od white children selected from a suburb of Atlanta, Georgia, Simon (1982)found that children above the mean age (5 years, 6 months) produced morecorrect detections than those children below the mean age. He also found thatvigilance performance (correct detections) was significantly predictive (r0.58, p<0.01) of scores on a psychometric test of school readiness (MetropolitanReadiness Test). Of course, although the internal and external validity of Simon’s(1982) experiment appears lacking due to sample size and selection, this “pilotwork” does nonetheless make a contribution to the literature in demonstrating thesignificance of age in relation to vigilance and scores on a school readiness test.Swanson (1983) studied vigilance in learning disabled and nondisabledchildren. A major purpose of the study was to test the proposition that learningdisabled children manifest a sustained attentional deficit. That is, the assumptionthat learning disabled children have the same attentiveness or sensitivity totransmitted information as nondisabled children, and that learning disabled childrenshow a decline in attentiveness as the time spent on the task increases wastested. Additionally, a second assumption, which stands as a contradiction to thefirst assumption, was also tested. That is, it may be that learning disabledchildren start a vigilance task at a lower level of sensitivity but without anygreater rate of decline in attentiveness than nondisabled children.The second major purpose of the Swanson (1983) study was to providea developmental perspective of sustained attention for learning disabled andnondisabled children. That is, this study questions the notion whetherLiterature Review / 20developmental constraints or limitations in sustained attention apply to youngerchildren. Swanson (1983) explains that an application of the maturationalhypothesis suggests that learning disabled children will exhibit similar thresholdsof sustained attention as younger nondisabled children.There were 72 children in the Swanson (1983) study separated intogroups of 12 by age level (8, 10, and 14.15 years) and diagnostic group(learning disabled or nondisabled). The children were selected from regular or LDclasses in the Greeley, Colorado, public schools and the University of NorthernColorado Laboratory School. All LD children met the federal definition, State ofColorado Standards, and the school review and appraisal definition. LI) childrenwere further selected on reading achievement measures below the 35th percentileon the reading subtest of the Metropolitan Achievement Test. All the childrenwere tested on a visual and auditory Continuous Performance Test (CPT).As expected, Swanson (1983). found that learning disabled children madefewer correct detections and more false responses to critical stimuli thannondisabled children at all ages. The results support the prediction that agedifferences in attentional capacities between learning disabled and nondisabledchildren reflect a reduced attentional capacity for visual and auditory information.Swanson emphasizes that “the notion the LI) and young children start avigilance task with the same capacity as nondisdabled older children but show adecline in attention as time on task increases is not supported” (p.428).In a subsequent study, Swanson and Cooney (1989) tested the hypothesisLiterature Review / 21that resource-limited components of vigilance are an important source of individualdifferences in verbal intelligence. “An important question to be considered in theanalysis of attention is whether individual differences in sustained attention relateto primary mental abilities” (p.141). To test this hypothesis Swanson and Cooney(1989) randomly selected sixty-three children (42 males and 21 females) fromgrades five to seven and adminstered the Wechsler Intelligence Scale forChildren-Revised (WISC-R), the Scott Foresman Achievement Test, and theContinuous Performance Test (CPT). Swanson and Cooney (1989) found that amoderate relationship (r0.54, p<.O5) exists between children’s verbal IQresources and their ability to sustain attention and discriminate continuously topresented stimuli-information. As a result, they explain that poor attention in theclassroom cannot be merely described in terms of problems of motivation,interest, or arousal. That is, the data in this study supports the notion thatvigilance and intelligence are related.E. SIGNAL DETECTION APPLICATIONSVIt is expected that the application of the theory of signal detectionmethodology to the analysis of CPT Vigilance Task performance and intelligencewill serve as a base to test the two competing theories of cognitive resources vsarousal levels as an explanation of whether individual differences in attentionalresources is related to primary mental abilities or other cognitive processingvariables.Tomporowski and Simpson’s (1990) study of intelligence and performanceLiterature Review I 22on a CPT measure of attention, compared retarded and nonretarded subjectsusing Signal Detection Analysis to provide independent measures of detectability(d’) and decision criterion (a). Swanson and Cooney (1989) explain that the useof signal detection measures, such as the dt index, allows one to assess astudent’s ability to discriminate the occurrence of discrete events. The principalargument for its use has been the provision of a stimulus detection index (d’)that is independent of response criterion (j3) factors (Broadbent, 1971). Previousresearch (e.g., Green & Swets, 1970) has demonstrated that separation of the dand 13 factors is necessary because variation of the response criterion inchildren’s performance suggests that all responses are not equally likely, and thatvariation in the response criterion should not be confounded with d’. In previousstudies (e.g., Grossberg & Grant, 1978; Sostek, Bucksbaum, & Rapport, 1980;Swanson, 1981, 1983; Swanson & Cooney, 1989; Tomporowski & Simpson, 1990),the d’ index has been associated with a subject’s resource limitations, such asverbal ability and intelligence. Cognitive monitoring strategies have been associatedwith 3 the response criterion.Several researchers (e.g., Levy, 1980; Stankov, 1983; Swanson, 1983;Swanson & Cooney, 1989) have found that children tend to vary their responsecriterion (13) cognitive strategies on a vigilance task. It appears that children tendto become more conservative in their responding the longer they engage in avigilance task. Consequently, some unbiased index besides hit rate and falsealarm rate should be used as a measure of vigilance performance. Hit rates andfalse alarm rates are insensitive to both variations in detection strength ( d’)and response criterion (13).Literature Review I 23Swanson and Cooney (1989) report that within the signal detectionframework an alternative explanation, the limited-resource hypothesis, wasformulated to account for the relationship between intelligence and sustainedattention on a CPT vigilance task. They hypothesized that resource limitations inthe child, as reflected in d, are an increasing function of intelligence.Additionally, Swanson and Cooney (1989) predicted that subjects will vary theirstrategy for monitoring their limited resources, as reflected in their scores. Inan earlier study, Swanson (1983) provided indirect support for the limited-resourcehypothesis by demonstrating that the decrement in vigilance over time was dueto variations in , rather than a decrement in d’. Consequently, it is likely thatcorrelations between d’ and intelligence will remain constant, but correlationsbetween and intelligence will vary across time.F. HYPOTHESESThe purpose of this study was to compare two competing theoreticalmodels’ (arousal versus resource limitation) explanations of the relationshipbetween attention and intelligence. The arousal model assumes that as time isspent on the vigilance task, negative correlations will emerge between vigilanceperformance and intelligence. Thus, the arousal model expects negative correlationsbetween measures of vigilance (hit rates, false alarm rates, d’, ) and IQ onlater time periods. The limited-resource model predicts that resource limitations,such as the child’s IQ, mediates vigilance-intelligence correlations. Consequently, itis expected that a measure most sensitive to these limitations (d’) will remainpositively correlated with IQ regardless of time on task. Moreover, becauseLiterature Review I 24students of lower IQ must effectively monitor their resources (e.g., they may bemore conservative than other students in their responses) the correlations betweenand IQ will be negative. These negative correlations will be maintained acrossthe duration of the task.To address this issue, and compare the two models, two groups ofkindergarten children (one group of teacher-nominated “at-risk” for learningdifficulties and the other as normal nonhandicapped randomly selected students)were assessed on measures of intelligence and sustained attention. Kindergartenchildren were selected to test whether individual differences are due to arousal orlimited resources as it appears that the developmental connection is morepronounced (higher correlations for younger children) at this age level forsustained attention and intelligence (e.g., Berch & Kanter, 1984; Herman et al.,1980; Levy, 1980).A signal detection methodology was used to test the two models as anexplanation of the relationship between sustained attention and intelligencemeasures. There were four major hypotheses in this study. The dependentvariables most important to the predictors are HR, FAR, d’ and t3 from thecontinuous performance task and Full Scale IQ from the WPPSI-R, and TestComposite SAS from the SBFE.• Hypothesis 1. The normal group’s mean level of vigilance performance willbe significantly higher than the corresponding means of the group withlearning difficulties.Literature Review / 25• Hypothesis 2. Measures of vigilance (HR., FAR, d’ and ) will besignificantly positively correlated with measures of intelligence, and notinfluenced by time on task.• Hypothesis 3. The correlations between measures of vigilance and measuresof intelligence will be significantly higher among subjects identified as havinglearning difficulties than among normally achieving subjects.• Hypothesis 4. When measures of intelligence and vigilance are included ina factor analysis, measures of both types of ability will load together ongeneral ability factor(s) in subjects identified as having learning difficulties,whereas among normally achieving subjects the factor pattern will showrelatively more specific factors.Ill. METHODOLOGYA. SUBJECTSFifty-eight students from twelve elementary schools (twenty-two teachers)participated in this study which was conducted in the Courtenay School, Districton Vancouver Island in a community with a population of 45,000. Two groupsof twenty-nine kindergarten students (one group consisted of teacher-nominatedat-risk for learning difficulties students and the other was normally achievingrandomly selected students) were selected for inclusion in the study in the springof their kindergarten year.Upon securing permission to conduct the study from the school district’ssuperintendent and the university’s Behavioural Sciences Screening Committee forResearch and Other Studies Involving Human Subjects, a letter was sent to allof the kindergarten teachers in the school district, explaining the nature of theproject and inviting their participation in the study (see Appendix A). Enclosedalong with the letter were copies of the Kindergarten School Learning Profile (seeAppendix B).tThis study was concerned with the differential performance of normalchildren and children rated as “at-risk’ for learning difficulties by theirkindergarten teachers. The Kindergarten School Learning Profile (SLP) was usedtAlthough the SLP consists of thirteen separate items, for the purposes of thisstudy, only item 1 was used to identify the at-risk group. Thus, the otheritems pertain to a parallel project, not the present study.26Methodology / 27to identify the “at-risk” group. In the letter to the school district’s kindergartenteachers, and in a meeting with the teachers, they were invited to refer anystudent whose ability, in their opinion, based on their teaching experience, tolearn school material falls into the lowest 10% of the normal curve as shown onthe first item on the SLP (see Figure 1). Thus, the SLP served as the methodwhereby teachers nominated students to be included in the “at-risk” group.The normal children were randomly selected in proportion to the totalnumber of males and females referred in the at-risk group. The total number ofkindergarten students registered in the school district, minus French immersionprogram students, cadre, multiple handicapped special needs kindergarten students,and one class where the teacher did not want to participate in the study, was396.The class lists from all of the schools were gathered and the names ofthe students who had been referred were noted. Then all of the boys notreferred were numbered, and similarly the girls. Two sets of computer generatedrandom numbers, one for boys, the other for girls, in proportion to the referredgroup, were used to randomly select the non-referred group of students. It isimportant to note that subject selection was not based on classroom or school“blocks”.Methodology / 28C_pped to other children you have observed, please rate this child’s overaliabili(5’ to learn school materia1.USE THE FWE POINT SCALE BELOW;Circle one number between I and 5.Figure 1SCHOOL LEARNING PROFILElowest 10% lower 30% middle 40% upper 30% highest 10%but not. but notlowest 10% highest 10%IMethodology / 29After the teacher referrals were complete, and the normal grouprandomly selected, parent permission letters (see Appendix C) were sent. Oncethe parent permission forms had been returned the two groups were establishedand schedules set for counterbalancing the tests.The at-risk for learning difficulties group consisted of 11 girls and 18boys. Attempts were made to obtain a stratified random sample of normallyachieving students in porportion to the number of boys and girls in the at-riskgroup. However, in the final analysis, the normal group of randomly selectedstudents consisted of 16 boys and 13 girls.The mean age of the entire sample, at the time the Stanford-Binet:Fourth Edition (SBFE) was administered, was 70 months (standard deviation of4.23 months). The teacher nominated at-risk group also had a mean age of 70months (standard deviation of 4.69 months). The normal children also had amean age of 70 months (standard deviation of 3.75 months).At the time the Wechsler Preschool and Primary Scale of Intelligence -Revised (WPPSI-R) was administered, the at-risk group had a mean age of 70months (standard deviation of 4.45 months). The normal group had a mean ageof 70 months (standard deviation of 4.2 months). Both groups had a mean ageof 70 months (standard deviation of 4.3 months).The ethnic composition of both groups is similar, with white childrencomprising 92% of the entire sample, 96% of the normal group, and 88% of theMethodology I 30nominated at-risk group. The only other ethnic groups represented in the sample,as reported by the parents on the permission form, were 1 Asian student and 3Native Canadian students in the teacher nominated group. The random selectednormal children had 28 white and 1 Native Canadian students.First language does not appear as a variable as 100% of both groupsreported that they were not bilingual.In terms of family composition, both groups are comparable. In therandomly selected group of normal children, 68% of the parents indicated thattheir child was living with both parents. The teacher-nominated at-risk group’sparents reported on the permission form that 66% of the children were livingwith both parents.In Canada, a major census occurs every ten years. The last majorcensus was in 1981. However, a minor census also occurs at staggered ten-yearintervals, with the most recent occurring in 1986. Table 1 shows a comparisonof the two groups in this study with the 1981 census report of labour force byindustry for the school district’s catchment area.Methodology / 31Table 1Occupational ComparisionsPercentagesCensus Normal ReferredAgriculture 2.9 3.4 6.4Forestry 7.9 6.8 25.8Fishing 1.1 10.3 3.2Mines and Oil 0.5 3.4 3.2Manufacturing 7.0 0.0 3.2Construction 9.1 6.8 3.2Transportation and Communications 5.8 3.4 3.2Trade 15.5 13.7 19.3Finance and Real Estate 4.3 3.4 0.0Service 27.7 27.5 19.3Public Administration and Defence 16.6 10.3 16.1Unspecified or Undefined 2.9 6.8 3.2Table 1 shows the occupational similarites and disparities between thetwo groups in this study compared with the census data. Forestry represents asignificant source of employment for the parents of the referred group. TheMethodology / 32normal group and census figures, however, are quite similar. PublicAdministration and Defence is seen as a significant employer for both groupswith the referred group having a similar percentage as the census figures.B. EXAMINERSThe team of examiners consisted of one professor of school psychology,two doctoral level school psychology students, two masters level students, and twoschool district special education personnel. All of the examiners had completedtraining on the administration of individual assessment measures.This study used the standardization version of the WPPSI-R. AU of thepractice protocols submitted by the examiners in this study were approved byThe Psychological Corporation. Additionally, all the WPPSI-R subject protocolswere scored by The Psychological Corporation.tC. PROCEDUREThe WPPSI-R, SBFE, Gordon Diagnostic System, and SLP wereadministered to all the children in the sample. The Psychological CorporationtThis study contracted with The Psychological Corporation to conduct a validitystudy using the Standardization Version of the Wechsler Preschool and PrimaryScale of Intelligence Revised (WPPSI-R). The Psychological Corporation requiresthe following statement in any published report, article, or version research: Astandardization edition of the Wechsler Preschool and Primary Scale of IntelligenceRevised has been used with the permission of The Psychological Corporation. Thisstandardization edition may differ in significant ways from the final publishededition.Methodology I 33required, in the administration of the standardiation version of the WPPSI-R, thata break of no less than 15 minutes occur in between the sixth subtest,Arithmetic, and the seventh subtest, Mazes. As a result of this requirement,most (8 7%) of the WPPSI-R administrations were carried out over a two dayperiod. Generally, the SBFE and the Gordon Diagnostic System were administeredin one session. All of the students’ teachers completed a Kindergarten StudentLearning Profile (SLP) prior to the administration of any of the above measures.The WPPSI-R and SBFE were administered in a counterbalanced order.That is, one half of the students were administered the WPPSI-R, while theother half were administered the SBFE. After a minimum one month intervalthe procedure was reversed with those students who had first been administeredthe WPPSI-R were then administered the SBFE. Likewise for those who hadfirst been administered the SBFE, one month later they were administered theWPPSI-R.As much as logistically possible, the test administration schedules wereset to counterbalance students as well as scales. That is, efforts were made toreduce any possible bias by balancing the groups in test administration. As aresult, approximately half of the normal group received the WPPSI-R first, asdid half of the teacher nominated at-risk group. The other half of the twogroups were administered the SBFE first. After a one month interval the processwas reversed to counterbalance the tests.Methodology I 34D. DATA ANALYSESThe purpose of this study is to compare two competing theoreticalmodels’ (arousal versus resource limitation) explanations of the relationshipbetween attention and intelligence, using the measurement scales described above.There are four major predictions in this study. The following points belowdescribe the data analyses used to test the predictions.1. It is expected that the two groups will differ in actual vigilanceperformance. These differences will persist even when IQ is partialled outin the analysis. An analysis of covariance, with IQ as the covariate will beused to compare groups on the four attention measures (BR, FAR, d’, ).The factorial design is a 2 (ability group) X 3 (time interval), withrepeated measures on the last factor.2. It is expected that the measures of vigilance (HR, FAR, d’, and ) will besignificantly positively correlated with measures of intelligence, and notinfluenced by time on task. To test this prediction, the Pearsonproduct-moment correlations between attention and intelligence are computedas a function of time interval on the vigilance measures.3. In the present study, it is expected that the teacher-nominated at-risk forlearning difficulties group will show higher correlations between theintelligence and vigilance measures in comparison to the normal group. Totest this prediction a correlation matrix comparison procedure, MULTICORR,(Steiger, 1979), is used to compare the intercorrelations between attentionand IQ between the groups. It is expected that the magnitude of theMethodology / 35correlations will be higher in the at-risk group. A z-score comparisonprocedure is used to statistically compare coefficients between groups.Additionally, it is also expected that there will be a difference in patternsbetween the intercorrelation matrices.4. It is expected that when measures of intelligence and vigilance are includedin a factor analysis, measures of both types of ability will load together ongeneral ability factor(s) in subjects identified as having learning difficulties,whereas among normally achieving subjects the factor pattern will showrelatively more specific factors. To test this prediction, exploratorymaximum-likelihood factor analyses will be used.IV. MEASUREMENT SCALESThis chapter presents the four measurement scales used in this study forassessing the relationship between intelligence and sustained attention inkindergarten students: teacher ratings of student learning abilities using theKindergarten School Learning Profile (SLP), the Gordon Diagnostic System (GDS),the Wechsler Preschool and Primary Scale of Intelligence-Revised, and theStanford-Binet Scale of Intelligence: Fourth Edition (SBFE).A. TEACHER RATINGS OF ABILITY AND BEHAVIOURIn this study Kindergarten teacher ratings of behaviour and schoollearning abilities were utilized for subject selection purposes. That is, the SchoolLearning Profile, described in greater detail below, was used as a teacher ratingreferral form for subject selection. In this study, teacher-nominated students as“at-risk” for learning difficulties were compared with normally achieving students.1. School Learning ProfileThis scale was constructed for the purposes of the present study using asimilar format, but different items than a scale used in previous research byConry and Conry (1985). In this present study, “at-risk” students are identifiedand classified by their kindergarten teachers on the basis of a rating scale:Kindergarten School Learning Profile (see Appendix B). Incorporating majoraspects integral to school readiness, this scale was designed to assess attention36Measurement Scales / 37and impulsivity (American Psychiatric Association, 1987; Anderson, Brubaker,Alleman-Brooks, and Duffy, 1985; Barkley, 1981; Gordon, 1979, 1986; Ross andRoss, 1982), spoken language skills (Wiig and Semmel, 1976, 1980), alphabetrecitation (Fletcher, Satz, and Morris, 1982; Satz and Fletcher, 1982), letteridentification (Woodcock, 1986), fine motor-coodination and printing skills (Simner,1983, 1987).The Kindergarten School Learning Profile (SLP) is an attitudinal andbehaviour rating scale with twenty-two items on a five point scale.t Thebehaviour rating items are similar to the diagnostic criteria in the Diagnostic andStatistical Manual of Mental Disorders (DSM ffl-R, American PsychiatricAssociation, 1987) for attention defict disorder with hyperactivity and, to theitems of the Conners Teacher Rating Scale (Conners, 1985).The first twelve items of the SLP ask the teacher to rate the student’sparticular level of ability on tasks related, to school readiness skills. Specifically,the teacher is asked to rate the student’s attention span and level ofdistractibility; the child’s ability to articulate and speak clearly; the child’s abilityto verbally describe a sequence of events; the child’s gross motor ability formovement in physical education; the child’s social participation and play behaviourwith the other children; can this child recite the alphabet; can this child correctlyname uppercase letters shown in random order; can this child correctly namenumbers between one and twenty shown in random order; can this child printtAlthough the SLP consists of 22 items, for the purposes of the present study,only item # 1 was used to identify the at-risk group. The other items arerelated to an associated project.Measurement Scales / 38his/her name correctly without reversals, deletions, additions, or misalignments;identify primary random colours shown in random order; and can this child usescissors to cut paper correctly?The SLP behaviour rating items ask the . teacher to rate the student’sbehaviour using measures designed for identifying a variety of behaviouralproblems in children. Specifically, the teacher is asked to “Please rate thebehaviour of this child, with reference to your observations of him/her in theclassroom, on the playground, or in other situations you have seen”. The teacheris next instructed: “To what degree does this child exhibit each behaviourbelow?” The range runs from behaviour frequency of: Never, Rare, Occasional,Frequent, and Constant for the following items: fidgets, difficulty staying seated,difficulty waiting turn in games or group activity, easily distracted, defiant anduncooperative, has temper tantrums, has difficulty listening, has difficulty playing,fails to finish things started (short attention span) and blurts out answers toquestions before the have been completed.In the present study, the SLP was used as the referral and screeningform for students classified as “at-risk” for learning difficulties.B. GORDON DIAGNOSTIC SYSTEMThe Gordon Diagnostic System (GDS) is a standardized behavior-basedassessment measure of attention (Gordon, 1986; Gordon & Mettelman, 1988). TheMeasurement Scales / 39GDS is a small portable microprocessor designed to assess impulse control andsustained attention in children (Gordon, DiNiro, Mettelman, & Tallmadge, 1989).The GDS tasks were standardized and produced normative data based on theprotocols of 1266 children between the ages of four to sixteen.Gordon and Mettelman (1988) report test-retest reliability data for ninetychildren randomly selected from the standardization sample who were retestedbetween 30 and 45 days and then again after one year of the initial testadministration. The test-retest reliability coefficients for the less that 45 daysinterval ranged from 0.68 to 0.85 and ranged from 0.54 to 0.94 for the oneyear interval.In this study the GDS scores, based upon standardization data, aredivided into three ranges: Abnormal, Borderline, and Normal. A score isconsidered Abnormal if it falls in the range of less than five percent of thenormal population (e.g., the 5th percentile or less). Borderline scores are thosewhich fall between the 6th and 25th percentile ranks. Normal classification scoresare those that fall above the 25th percentile (Gordon, 1986).There are two game-like tasks that are used with the GDS forkindergarten children: the Vigilance Task and the Delay Task which are discussedfurther below.Measurement Scales I 401. The Vigilance TaskThis task measures the child’s ability to focus attention on a task andto maintain this attention over time (Gordon, 1986). This is a portablecomputerized version of the Continuous Performance Test (Rosvold, Mirsky,Sarason, Bransone & Beck, 1956). The Continuous Performance Test (CPT) hasbeen used in previous research using microcomputers to collect attentional data(e.g., Klee & Garfinkel, 1983; Murphy-Berman et al., 1986).On the GDS Vigilance Task a series of digits from 1 to 9 are flashedfor 200 milhi-second and are presented at a rate of one every two seconds forsix minutes on the front display LED screen. The child is required to press thebig blue button every time the number “1” appears. The standardized instructionsare as follows:Now to play the next game, you need to know what the number “1”looks like. Can you show me a “1”? (Showing the card with theLED-stlye numbers.) Can you show me a “7”? Can you show me a“1” again? (The child is asked to point to the number “7” because itlooks similar to a “1”; it is necessary to establish that the child candifferentiate between the two numbers. Since some children don’t knowtheir numbers, you should try to teach them. If the child still cannotpick out the “1” then the task cannot be administered.)Now you are going to practice. In this game you will see numbersflash quickly on the screen, and I want you to press the blue buttonevery time you see a “1” and not when any other number flashes.Measurement Scales I 41Are you ready? Begin. (Administer the practice trial).After the trial say:Now you are going to play this game for a longer time. I want youto press the blue button only when you see a “1”. Only press theblue button if you see a “1”. I’ll sit here and wait until you’re done.You will know when the game is over because the green light willcome on and you will hear a beep. If you have any questions, orwant to talk about the game, I want you to wait until after thegame is over, and we can discuss it then (p.1°).The Vigilance Task measures the student’s ability to focus attention onthe task and to maintain that attention, without any feedback, for a six minutetime period. On this task, the GDS keeps track of the student’s total correctscore (hit. rate) over three time Blocks, at two minutes per Block. In each Blockthe GDS records (1) correct responses (hit rate), (2) omissions of correctresponses, and (3) commission responses where the student responds incorrectly(false alarms).a. Summary Scores Definitions-Following are the operational definitions of the summary scores recorded on theVigilance Task:1. Correct Responses (Hit Rate): This score represents the total number oftarget responses by the student. There are 29 signals presented over thesix minute time period which is divided into three blocks. Block 1 has 12possible correct responses; Block 2 has 9 possible correct responses; andMeasurement Scales I 42Block 3 has 8 possible correct responses. Gordon (1986) indicates that theCorrect Response (CR) score is an index of the student’s ability to achieveand maintain alert, vigilant responding. Students with CR scores in theNormal Range are able to sustain attention appropriately.2. Errors of Omission (EO): This score represents the number of times the1” was presented to which no response was made by the student. Asindicated above, the CR score indicates the number of correct responses.Thus, the CR and EO scores mirror each other. That is, if a studentachieved a CR score of 22, the EO score is 7 (22 correct out of apossible 29). As a result, the EO score is an index of vigilant lapses. Thelarger the EO score the more frequently the student was not attendingsufficiently to respond to the signal.3. Errors of Commission (EC): This score represents the number of times thestudent responded to incorrect stimuli other than the number aiM. That is,Errors of Commission (EC) are generally not related to either CorrectResponses (CR) or Errors of Omission (EO), and may stand as data on aseparate dimension. Gordon (1986) explains that while CR and EO reflectthe adequacy of sustained attentional capabilities, the Errors of Commission(EC) reflects the degree of impulsivity the student demonstrates under thestructure imposed by the cognitive demands of the Vigilance Task. The ECrepresents an index of the student’s inappropriate responding. These areresponses which are not lapses in sustained attention, but rather responsesto stimuli in a poorly controlled manner. Errors of Commission are anindication of alertness and the ability to respond correctly.Measurement Scales I 43b. Signal Detection CalculationsIn this study, similar to previous studies (e.g., Broadbent, 1971; Green& Swets, 1970; Grossberg & Grant, 1978; Sostek et aL, 1980, Swanson, 1980,1981, 1983; Swanson & Cooney, 1989; Tomporowski & Simpson, 1990), SignalDetection Analysis provides independent measures of stimulus detectability (d’) anddecision criterion j3. The measures are calculated from a mathematicalformulation that combines the probability of correct detections with the probabilityof false alarms. d’ is defined as attentional strength and as variations inattentional strategies.In the present study, formulas for the calculation of d’ and are takenfrom Hochhaus (1972):d’ = (ABS X HR) — (ABS X FAR), whereABS = the abscissa values of a standardized normal distribution,which is the distance from the mean to the point of the dichotomy inthe standard normal distribution.HR = hit rate (proportion of correct signals recognized).FAR = false alarm rate (proportion of times a signal was identifiedwhen it was not presented).ORD = the ordinate at the point of dichotomy in the standardnormal distribution.= ORD(HR)/ORDCFAR),In sum, the hit rate is the number of times the target stimulus was presentedMeasurement Scales I 44and was correctly identified. The false alarm rate was the number of times thesubject identifies a stimulus that was not the target stimulus. The abscissa andordinate values of a normal distribution that are necessary to calculate d’ andcan be obtained from Hochhaus (1972).Figure 2, used with permission from Warm and Jerrison (1984, p. 35)shows the hypothetical distribution of sensory events assumed by signal detectiontheory.‘ICa’aI’I’.Figure 2. Hypothetical distributions of sensory events assumed by signal-detectiontheory. The magnitude of sensory excitations, x, from weak to strong, is arrayed fromleft to ngh along the horizontal axis, while the relative frequency of these excitationsis represented along the vertical axis. An observation samples an excitation, x. and theobserver must decide whether x4 is from distribution N or distribution SN. The decisionis based on whether x4 > or x, <0. The distribution of effects for noise alone andfor signal plus noise are the two normal probability distributions N and SN. Note thatthe presence of a signal actually shifts the N distribution to the right, transforming itinto the SN distribution, but leaves the shape of the distribution unchanged. The indicesd’ and 0 represent signal-detection parameters of sensitivity and response bias.I CViterson 9 SDecision: no signolContinuum of sensory mognitudeDecision: signolrespectively.Measurement Scales / 452. The Delay TaskThe second GDS task used in this study which is age appropriate forkindergarten children is the Delay Task. The Delay Task is designed to measurebehavioral suppression and impulse inhibition. Gordon and Mettelman (1988)explain that the Delay Task is based on a differential reinforcement of lowrates-4 second interval (DLR-4) behavior schedule. Unlike the Vigilance Task,described above, where the student’s attention is instrument paced requiringalertness, the Delay Task is sell-paced based on reinforcing feedback of responses.In this task the child is administered a set of standardized instructionswhich requires the child to inhibit responding in order to earn points in what isdescribed as a game. The child is seated in front of the GDS, which appears asa small plastic covered metal box with a small display screen and a big bluebutton on the front. Above the display screen is a red light and right next to itis a green light. The child is read the following instructions:You are going to play a game in which you will get a chance towin a lot of points, not just 1 or 2 but a whole bunch of points. Doyou see this light (Pointing to the small red light)? Every time youmake this light go on you’ll earn a point and this counter (pointing)will keep track of how many points you’ve won. At the end of thegame, we’ll see how many points you have earned. Now, to makethe light go on, all you have to do is push this blue button, andwait a little while before you can press it to get another point. But,if you push the button, wait a while, then push it again, you’ll getMeasurement Scales I 46a point every time (p. 9).The Delay Task requires the student to wait a set period of time, fourseconds, (called the Delay Intervai) before pressing the blue response button. Ifthe student has waited long enough (four seconds), the red light will shine andthe counter on the display screen will increment. If the child presses the buttonbefore the Delay Interval has elapsed then no point is earned, the red lightstays off and the delay timer resets. The entire task takes six minutes.Unlike the instrument-paced Vigilance Task, which requires alertness anda relatively high level of arousal, the Delay task is subject-paced in that themicroprocessor does nothing but record the student’s responses in relation the(DLR-4) behavior schedule. Thus, the Delay Task places different behavioraldemands and cognitive skills than the CPT Vigilance Task. On the Delay Tasksustained attention is not as important as behavioral suppression of the impulseto respond. Gordon, McClure, & Post (1986) explains that the Delay Taskrequires “cognitive mediation” where the student develops strategies, usuallyverbally-mediated behaviors like counting, before responding.Gordon (1979) demonstrated that the GDS Delay Task significantlydiscriminates hyperactive from nonhyperactive clinic-referred children and issignificantly correlated with teacher ratings from the Conners Teacher RatingScale (cf. Barkley, Fischer, Newby and Breen, 1988).On the Delay Task the microprocessor records the “hit rate”, the correctMeasurement Scales I 47number of times the student responds by pressing the button after the delayinterval lasped. Also, the total number of responses is recorded which is used tocalculate the Efficiency Ratio (ER). This ratio is obtained by dividing the hit rate(correct responses) by the total number of responses.The microprocessor records student responses totally as well as by four90 second blocks. These Block Scores enables the tracking of a student’sresponding performance over the six minute task and to measure changes fromone block to another.a. Summary Scores DefinitionsFollowing, are the operational definitions of the summary scores on the DelayTask:1. Efficiency Ratio (ER): This represents the percentage of times the childresponded after having waited the appropriate interval time. The ER is thetotal number of correct scores divided by the total number of responses.Gordon et al. (1986) indicate that the ER is the single best indicator ofthe level of impulsivity or behavior suppression shown by the student. Anormal child will generally achieve a Total ER of at least 0.62, while animpulsive child will generally achieve a Total ER of 0.42 or less. Inaddition to the Total ER, the microprocessor records the ER for each ofthe four Block Scores and calculates ER Block Variability.2. Total Number of Responses (R): This score represents the student’s overallrate of responding. The significance of R is that it determines the validityof the Efficiency Ratio. Gordon et al. (1986) explain that both low andMeasurement Scales I 48high R scores are indicative of students who are overly-cautious (low R)and who have an inability to suppress responding (high R). Also, asignificantly high R score may indicate the student’s motivation could beconsidered suspect.3. Total Number of Correct (COR): By itself the COR is not very meaningful.However, when the COR is interpreted in relation to the total R and thesubsequent Efficiency Ratio a great deal of significant information about thestudent’s responding is evident. The impact of the COR is subsumed underinterpretation of the ER.4. Block Score Analyses: The microprocessor is programmed to separate thetotal task into four 90 second blocks. This can generate a Block Variabilityscore which is useful in determining patterns of responding over time.Students with Normal Block Variability scores (0.13 or less) indicatesconsistent performance, while Borderline (0.14 to 0.19) and Abnormal (0.2or higher) indicates inconsistent performance and a level of impulsivity.Measurement Scales I 49C. WECHSLER SCALES OF INTELLIGENCEWechsler first published the Wechsler-Bellevue intelligence Scale in 1939.Sattler (1988) describes this scale as the forerunner to the Wechsler intelligenceScale for Children (WISC, Wechsler, 1949), Wechsler Intelligence Scale forChildren - Revised (WISC-R, Wechsler, 1974), Wechsler Preschool and PrimaryScale of Intelligence (WPPSI, Wechsler, 1967), the Wechsler Adult intelligenceScale (WAIS, Wechsler, 1955), and the Wechsler Adult Intelligence Scale-Revised(WAIS-R, Wechsler, 1981).Wechsler’s subtests, taken together, form a scale with a focus on theglobal nature of intelligence. The Full Scale IQ score serves as an index ofgeneral cognitive ability. And although Wechsler made no attempt to designsubtests to measure primary abilities of intelligence, all of the Wechsler scaleshave five or six subtests which measure verbal comprehension on the VerbalScale, and another five or six subtests which measure perceptual organization onthe Performance Scale (Anastasi, 1988; Sattler, 1988).1. The WPPSIThe Wechsler Preschool and Primary Scale of Intelligence (WPPSI) wasfirst published by The Psychological Corporation in 1967. Although essentiallysimilar, as a downward extension, of the 1949 WISC, the WPPSI has 3 of the11 subtests which are unique: Sentences, Animal House, and Geometric Design.The other 8 subtests are also found on the WISC and WISC-R. However,Measurement Scales I 50Wechsler (1967) emphasizes in the WPPSI manual that the WPPSI “has notbeen a mere pulling down of, or an addendum to, the WISC, but a separatescale, optimally suited for the mental examination of the 4 to 6.5 year oldchild” (p.2).Table 2WPPSI SubtestsVerbal Tests Performance TestsInformation Animal HouseVocabulary Picture CompletionSimilarities MazesComprehension Geometric Design(Sentences) Block Design(Animal House Retest)Source: The Psychological Corporation (1989).Table 2 shows the WPPSI’s 5 Verbal Tests with the sentences subtest,in brackets, as it is to be used as a supplementary test. Only 5 of the VerbalTests may be used to calculate the Verbal Score.Measurement Scales I 51Table 2 also shows the 5 Performance Tests with Animal House listedfirst and Animal House Retest, in brackets, listed last. Animal House wasadministered twice in the standardization sample to check for reliability. TheAnimal House Retest is not used in calculating a Performance IQ.2. The WPPSI-RThis study used the Standardization Version of the Wechsler Preschooland Primary Scale of Intelligence - Revised (WPPSI-R). The WPPSI-RStandardization Manual explains that after 20 years of usage, the WPPSI hasbeen revised:The principal goals of the revision are to update the norms and toextend the age range of the scale downward to age 3 and upward toage 7. Other objectives include adding an Object Assembly subtest tothe 11 original WPPSI subtests; deleting selected test items that arealso on the WISC and its revision, the WISC-R; eliminated or revisingany material that is considered dated or biased; and making the testmaterials more appealing to children (p.iii).Measurement Scales / 52Table 3WPPSI..R SubtestsVerbal Subtests Performance SubtestsInformation Object AssemblyComprehension Geometric DesignArithmetic Block DesignVocabulary MazesSentences Animal PegsSimilarities Picture CompletionSource: The Psychological Corporation (1989).Table 3 shows the 12 WPPSI-R subtests. When Table 3 is comparedwith Table 2 it shows that the WPPSI-R subtests are quite similar to those ofthe WPPSL The Standardization Manual explains that “in order to extend theapplicable age range, some new items have been added to the lower and upperlimits of some subtests” (p.v). Another change was the addition of the ObjectAssembly subtest which results in the WPPSI-R having 12 subtests compared tothe WPPSI’s 11.Measurement Scales / 53D. STANFORD-BINET INTELLIGENCE SCALESThe Binet-Simon and Stanford-Binet Intelligence Scales have a longhistory dating back to 1905 when the rwst scale was developed. Sattler (1988)presents an excellent discussion of the scale’s history, evolution, and development.1. Age-Scale and Point-Scale FormatsThe current point-scale format found on the Stanford-Binet IntelligenceScale: Fourth Edition (SBFE, Thorndike, Hagen, & Sattler 1986) represents amarked departure in its revision compared to earlier forms which used anage-scale format. This change was made nearly seventy years after Yerkes’(1917) article comparing the Binet age-scale format with the advantages of thepoint-scale format.Some fundamental differences exist between an age-scale and point-scaleformat. Standardization procedures, item selection and development, as well asscoring, all differ in principle and function. In an age-scale format, test itemsare standardized on children at particular age levels. That is, age categories arenormed with tests and tasks tied to developmental behaviour. Sattler (1988)states that “in age scales tests are selected on the assumption that importantforms of behavior appear at various point in development, whereas in pointscales tests are selected to measure specific functions” (p.43).The Wechsler scales are an example of a point-scale format. SubtestsMeasurement Scales / 54are designed to measure similar aspects or constructs of behaviour at every agein the standardization sample.Central to earlier editions of the age-scale format on the Stanford-Binetscales is the concept of Mental Age (MA) and the concomitant Intelligence Quotient(10). This ratio IQ was obtained by dividing an individual’s MA by his/herchronological age (CA) and multiplying by 100.An individual’s MA is an age-equivalent for the raw score. Essentially,mental age represents an indication of the amount of general mental or cognitiveability a child demonstates in relation to the average child at a CAcorresponding to the MA score. For example, a child scoring a MA of 5 isobserved as having the general mental ability of an average 5 year old (Sattler,1988).There are some significant limitations to the MA as a construct inmeasuring mental or cognitive abilities. MA measurements are not equal intervaldistance units. The distance between MAs of 2 and 3 is much greater than atolder ages of adolescence. It does not follow that someone with a MA of 10 istwice as intelligent as someone with a MA of 5. Moreover, MA units varythroughout a child’s development as does general cognitive development -which isnot seen as a linear function.Although the MA construct is fraught with limitations, its value rests inthe ability to stand as a developmental measure indicative of an individual’s levelMeasurement Scales I 55of cognitive functioning. Moreover, an individual’s MA may have psychoeducationalimplications. Jensen (1979) suggests that an individual’s MA may represent acertain level of achievement or readiness to learn.2. The Stanford-Binet Fourth EditionPrevious editions of the Stanford-Binet have been criticized for the heavyreliance on verbal abilities and the reporting of a single composite score. Inrevising the SBFE, Thorndike et al. (1986) explain in the Technical Manual (TM)that they decided to drop the age-scale format, and use new research incognitive psychology to design a scale that measures four areas of cognitiveability: Verbal Reasoning, Quantitative Reasoning, Abstract/Visual Reasoning, andShort-term Memory. Additionally, they decided that the revised scale wouldcontinue to provide a composite score that would represent g, or generalreasoning ability.Table 4 shows the theoretical model of the SBFE. The TM explains that“the selection of the four areas of cognitive abilites was based on a three-levelhierarchical model of the structure of cognitive abilites that was adopted to guidethe construction of the Fourth Edition” (p.9). The top level of the model indicatesg as the general reasoning factor. The second level shows three broad factors:crystallized abilities, fluid-analytic abilities, and short-term memory. The third levelconsists of the specific factors: verbal reasoning, quantitative reasoning, andabstract/visual reasoning. The fourth level indicates the respective subtests on theSBFE for each factor.Measurement Scales / 56Table 4Theoretical Model of the SBFEgCrystallized Fluid-Analytic Short-TermAbilities Abilities MemoryVerbal Quantitative AbstractlVisualReasoning Reasonmg ReasoningVocabulary Quantitative Pattern Analysis Bead MemoryComprehension Number Series Copying Memory for SentencesAbsurdities Equation Building Matrices Memory for DigitsVerbal Relations Paper Folding Memory for ObjectsSource: Adapted from Technical Manual SB:FE, Thorndike, Hagen, and Sattler(1986).Chapter one of the TM (Thorndike et al., 1986) is an adaptation ofThorndike’s (1986) invited address to the National Council on Measurement inEducation. In this address Thorndike outlines the historical and theoreticalperspectives of general ability models versus multidimensional models ofintelligence. The concept of g is central to his discussion, concluding with g’srelationship in the SBFE. Clearly, the purpose of the SBFE is to measure g.Measurement Scales / 57At the next conceptual level of the hierarchical model, the g is dividedinto a modification of Cattell and Horn’s theory of intelligence (Cattell, 1982;Horn and Cattell, 1966): Crystallized Abilities, Fluid-Analytic Abilities, andShort-Term Memory.The third level in the model consists of the specific factors: verbalreasoning, quantitative reasoning, and abstractlvisual reasoning. These factors,along with Short-Term Memory, are measured by the 15 subtests.Although, as previously mentioned, the fourth revision represents asignificant departure from earlier versions of the scale, the authors state in theTM that they decided to maintain continuity with previous editions by retainingas many item types as possible.The criteria used to choose item types for inclusion in the new editionwere the following: (1) they have to be shown to be acceptablemeasures of verbal reasoning, quantitative reasoning, abstractivisualreasoning, or short-term memory; (2) they could be scored reliably; (3)they were generally recognized as being free of ethnic and genderbias; and (4) they functioned adequately over a wide range of agegroups. The following item types from Form L-M were retained forpossible inclusion in the Fourth Edition: Vocabulary (which includeddefining abstract words), Comprehension (which included giving reasons),Verbal Absurdities, Picture Absurdities, Opposite Analogies, PaperFolding and Cutting, Copying (which included copying figures andbuilding blocks), Ingenuity, Repeating Digits (which included repeatedMeasurement Scales / 58reversed digits), Memory for Sentences, Copying a Bead Chain fromMemory, Similarities, Identifying Parts of the Body, Form Board, andvarious quantitative item types (p.10).Although the revison dropped the age-scale format, the adaptive-testingformat was retained. That is, no individual takes all the items at a particularage level of a subtest. Moreover, not all of the 15 subtests are used at everyage level. Thus, the authors state in the TM that every examinee is tested witha range of tasks best suited to his or her ability level. Therefore, theoretically,as the standardization ages range from 2 to adult, the scale should provide goodfloor and ceiling measures.E. RATIONALE FOR TWO INTELLIGENCE MEASURESThis study used two inteffigence measures: the WPPSI-R and the SB:FE.Both of these tests have been recently restandardized. Both of these testsrepresent the most frequently used and age appropriate measures to assessintelligence in kindergarten children. However, the literature to date has notindicated any empirical evidence in favour of one measure over the other forassessing the cognitive abilities of kindergarten children. Therefore, one reason forusing both measures was to examine the concurrent validity of the two testswith this study’s population.It is also important to point out that the data from the two intelligencetests were collected as part of a parallel project contracted with TheMeasurement Scales I 59Psychological Corporation to determine the concurrent validity of the WPPSI-Rand SB:FE. Thus, the present study used the two intelligence measures for somecomparison purposes in relation to the attention measures.V. RESULTSA. INTELLIGENCE MEASURESThe means, standard deviations, and ANOVA F-ratio values for the twointelligence measures as a function of the two groups are shown in Table 5 andTable 6. As shown, significantly lower scores emerged for the at-risk group whencompared to the normal group. Overall, a comparison of the WPPSI-R Full ScaleIQ and the SB:FE Test Composite SAS show the at-risk group (mean = 82.62,s.d. = 10.9 and mean = 82.24, s.d. = 8.9) was lower compared to the normalgroup (mean = 110.65, s.d.= 13.9 and 108.51, s.d. 12.6).In terms of concurrent validity, in addition to the similarities betweenthe WPPSI-R Full Scale IQ and SB:FE Composite SAS means and standarddeviations, the correlations are significant (both groups: r = 0.885, p< .001;at-risk group: r = 0.725, p< .001; normal group: r = 0.729, p< .001).However, one way to interpret the correlation coefficients of r = 0.725 and r =0.729, is to consider the amount of variance that is “shared” by the two tests.That is, the coefficient of determination, r2, conceptually represents the amountof variance in one measure with the other. The coefficient of determinationindicates the degree to which the two tests are measuring the same construct. Acorrelation coefficient of 0.72 squared equals 0.51 and can be interpreted asshowing 50% shared variance between the two tests which are purported tomeasure intelligence. If they only share 50% of their variance, then the other50% is unique to the measures themselves. Therefore, in terms of concurrent60Results I 61validity, and the construct of intelligence, it does not appear that these two testsshould be considered as interchangeable instruments based on the data in thepresent study.A MANOVA analysis between groups on the WPPSI-R across twelvesubtests was significant, Wilk’s F(11,46) = 8.64, p<.001. One-way ANOVA’sindicated that all twelve WPPSI-R subtests showed significant differences (p<.001)between groups.Similarly, a MANOVA analysis between groups across eight subtests onthe SB:FE was significant, Wilk’s F(8,49) = 9.98, p<.OO1. One-way ANOVA’sindicated that all eight of the relevant SB:FE subtests showed significantdifferences (p <.001) between groups.Results / 62Table 5. WPPSI-R Means and Standard Deviations by GroupSubtests At-Risk Normal ANOVAPerformance Scale Mean sd Mean sd F(l57)Object Assembly 8.9 2.8 12.7 3.1 24.14*Geometric Design 8.1 3.2 11.9 3.3 19.05*Block Design 6.7 2.3 11.4 2.4 56.65*Mazes 7.4 2.6 9.8 2.9 10.42*Picture Completion 7.6 2.7 10.7 2.9 16.04*Animal Pegs 6.5 3.2 10.1 3.0 19.84*Verbal ScaleInformation 7.7 2.1 12.9 2.9 63.46*Comprehension 7.0 2.2 11.3 2.9 39.36*Arithmetic 6.8 2.6 11.3 2.8 40.12*Vocabulary 7.5 2.3 11.8 2.7 45.23*Similarities 7.3 1.9 10.9 2.3 39.83*Sentences 7.4 2.1 11.1 3.4 41.36*Performance IQ 85.27 12.5 108.13 13.8 43.64*Verbal IQ 83.48 9.4 110.51 14.3 72.69*Full Scale IQ 82.62 10.9 110.65 13.9 73.01** p<.001Results / 63Table 6. SB:FE Means and Standard Deviations by GroupAbstract/VisualReasoning SASPattern AnalysisCopyingQuantitative AreaStandard Age ScoreQuantitative85.72 10.843.89 8.2110.06 14.7 51.41*55.03 7.4 29.65*Test Composite 82.24 8.9 108.51 12.6 84.40*Subtests At-Risk Normal ANOVAMean sd Mean sd F(l,57)Verbal ReasoningStandard Age Score 93.06 7.8 110.83 10.6 48.23*Vocabulary 46.03 4.6 54.55 6.2 35•44*Comprehension 47.27 4.0 54.82 6.0 31.42*Absurdities 47.86 5.1 53.96 5.6 18.95*82.79 10.8 105.13 12.7 51.88*45.20 7.1 55.93 6.0 38.90*39.79 4.5 47.48 7.1 24.35*Short-Term MemoryStandard Age Score 79.93 7.4 103.00 13.6 70.44*Bead Memory 38.65 4.6 49.93 7.6 47.52*Memory for Sent. 43.34 4.5 53.03 7.8 33.83** p<.001Results I 64Table 7. Mean Percentage of Correct Detections,False Alarms, Mean dand Beta Values byGroup and Time Periods.Time Periods1 2 3Mean Sd Mean sd Mean sdVigilance measureAt-Risk Group% Correct detections 83.74 0.20 68.03 0.24 64.10 0.27% False Alarms 16.81 0.19 8.61 0.17 8.59 0.15d’ 2.45 1.17 1.52 2.12 1.70 1.93Beta 1.37 1.34 4.23 4.57 4.28 4.35Normal Group% Correct detections 91.72 0.19 89.83 0.20 83.86 0.24% False Alarms 6.77 0.13 1.81 0.03 1.52 0.04d’ 3.68 0.81 3.64 1.44 3.61 1.11Beta 1.88 2.87 3.23 3.55 4.63 4.98Note: At-Risk N = 29Normal N = 29Results I 65B. VIGILANCE MEASURESThe percentage of correct signal detections (response within 2 seconds,following a stimulus presentation) and false alarms (extraneous responses orcommission errors) made by each child were recorded for the three time periodson the six minute task. Table 7 shows the mean percentage of correct detections(hit rate, HR), false alarms (FAR), d’, and 13 values for each time period.t Thefour vigilance measures (HR, FAR, d’, 13) were analyzed in a 2(at-risk vs normalgroup) X 3(time periods) design, via an Analysis of Variance and Covariance,with repeated measures on the last factor. Intelligence served as the covariate inthe ANCOVA analysis. It should be noted that in the ANCOVA analyses theassumption of parallel regression lines was not violated (Elashoff, 1969; Glass &Hopkins, 1984). That is, the two groups had equal slopes, and when IQ wascovaried, and the subjects scores adjusted, both groups’ scores were adjustedequally when EQ was removed. The normal group moved down and the at-riskgroup moved up in an equally linear homogeneous fashion.The ANOVA on the percentage of correct detections (hit rate) yieldedsignificant main effects for group, indicating the normal group made significantlymore correct detections than the at-risk group, F(1,56) = 13.86, pC.OOl, MSe =.08. There was also a significant main effect for time, F(2,112) = 8.65,t Raw scores for correct responses (hit rate) and false alarms were analyzedseparately, and transformed into d’ and 13 using a computational table fromHochhaus (1972). Consequently, the Gordon Diagnostic System norms were notused. Nonetheless, it bears noting that the normal group’s mean score in thepresent study falls in the Normal Range, in comparison to the standardizationsample (Gordon 1986), for Total Correct Responses and Total Commission Errors.The at-risk group falls in the Borderline Range for both measures.Results I 66p<.OOl, MSe = .03. There was not, however, a significant interaction effect forgroup X time, F(2,112) = 2.48, p > .05. A post hoc Newman-Keuls analysis(scores collapsed across diagnostic groups and time periods) indicated significantdifferences (p<.05) between all three time periods. When the groups werecontrasted separately, Newman-Keuls analysis indicated significant differences(p<.05) for the at-risk group for all three time periods; the normal groupshowed hit rates that were significantly lower (p<.05) for the last time periodthan for the first. No significant differences in hit rates were found between themiddle time period and the first.In the Analysis of Covariance for correct detection scores, no significantmain effect occurred for group, F(2,55) = 0.86, p >.05; time, F(2,55) =0.10, p>.05; or the two-way interaction, F(2,55) = 1.76, p >05. Thus, whenintelligence is taken out, there is no significant effect related to group (at-risk vsnormal) or time.The ANOVA computed on the percentage of false alarms yielded asignificant main effect for group, F(1,56) = 6.47, p<.001, MSe = .04, whichindicates that the at-risk group made significantly more false alarms than thenormal group. There was also a significant main effect for time, F(2, 112) =15.72, p<.001, MSe = .08. A Newman-Keuls analysis found false alarm ratesfor both the at-risk and normal groups were significantly different (p<.O5)between the first and third time period, and between the first and second timeperiod, but not significant (p > .05) between the second and third time period.Thus, more false alarms occur initially in the task and more conservativeResults / 67responding in the latter time period. There was no significant interaction effectfor group X time, F(2,112) = 0.86, p >05. In the Analysis of Covariance forfalse alarms, there were no significant main or interaction effects.For d’, the parameter of interest in relation to sustained attention andintelligence, a significant main effect occurred for group, F(1,56) = 35.73,p<.OO1, MSe = 3.74, which indicates the normal group was more sensitive tocritical stimuli (d’) than the at-risk group. However, no significant effects occurredfor time, F(2,112) = 2.65, p >.05; or for the interaction of group X timeperiod, F(2,112) = 2.09, p >.05.In the Analysis of Covariance for d’, significant main effects occurred forgroup, F(2,55) = 10.03, p<.O1, MSe = 3.73. No significant main effectsoccurred for time, F(2,110) = 0.07, p >.05; or interaction effects for group Xtime, F(2,110) = 1.05, p >.05.The ANOVA on j3 values did not show a significant main effect forgroup, F(1,56) = 0.09, p >05; but did show a significant main effect for time,F(2,112) = 16.92, p <.001, MSe = 1.21. There was not, however, a significantinteraction between group X time, F(2,112) = 0.62, p >.05.In the Analysis of Covariance for i3 values, there were no significantmain or interaction effects.Results I 68C. INTELLIGENCE AND VIGILANCE RELATIONSHIPSThe at-risk group and the normal group’s correlations between vigilanceand intelligence are shown in Tables 8 to 16. The intelligence measures arerepresented by the three WPPSI-R scales: Verbal, Performance and Full ScaleIQ’s. Additionally, the four SB:FE Area Standard Age Scores, (Verbal Reasoning,AbstractlVisual Reasoning, Quantitative Reasoning and Short-Term Memory), areused for comparision purposes with the vigilance measures. Also, the WPPSI-Rfactors and the SB:FE area scores are correlated with the vigilance measures(HR, FAR, d’, j3).tIt should be noted that in the following correlational analyses no attemptto correct for a restriction of range in IQ was conducted. That is, although itwas the purpose of the following correlational analyses to measure therelationship between IQ and attention to discover whether, as a function of groupmembership (and hence IQ), differences existed as a function of time, and thescattergrams were nonlinear, skewed, or asymmetrical in shape, no attempt wasmade to project one correlational pattern onto the other. That is, for example, ifthe correlation between IQ and attention is significantly different for at-riskchildren versus normal children, then this implies that the correlation for thegeneral population may not be linear and symmetric. If that is the case, atA cautionary note regarding the interpretation of the correlation coefficientsshown in Tables 8 to 16: Some readers may appropriately raise concerns abouttesting many correlations with a small sample size, as in the present study.However, due to the fact that very few of the correlation coefficients weresignificant, the null hypothesis was not rejected. Thus, the risk of Type I erroris reduced. In terms of Type H error, and the problem of failing to reject thenull hypothesis, again, due to the fact that few correlations were significant it isunlikely that a larger sample size would have made a difference.Results I 69correction for restriction of range would project one correlational pattern over intoan area in which the pattern did not fit, and in fact, a different pattern waspresent. Thus, corrections for restriction of range make assumptions which aredirectly contrary to the hypotheses in this study. Jarman (1980a) explains thatthe problem with corrections for restrictions is that any group-specific patternsare collapsed into the averages, and conclusions on variations in cognitiveprocesses among the groups are completely precluded.1. Hit Rate and IntelligenceAt-risk GroupTable 8 shows that the correlations between hit rate and IQ for theat-risk group are positive, but low in magnitude (range: r = 0.071 to 0.244).None of the correlation coefficients are significant (p > .05) in the first timeperiod. The second time period, however, indicates the at-risk group’s hit ratehas a moderately low, but significant, correlation coefficient (r = 0.394, p<.Ol)with the WPPSI-R Full Scale IQ. Also, the WPPSI-R Performance Scale, in thesecond time period, has a significant correlation coefficient (r = 0.485, p< .01)with hit rate. In the third time period, only the WPPSI-R Performance Scalecontinues to have a significant, but moderately low correlation coefficient (r =0.3 16, p<.Ol). The WPPSI-R Verbal Scale indicates a significant correlationcoefficient only for total hit rate (r = 0.420, p<.O5), but not for any of thethree time periods.Results / 70For the SB:FE, the at-risk group shows positive, but low level (range:0.002 to 0.207) correlation coefficients, for hit rate in all three time periods.None of the correlations are statistically significant.Table 8 shows positive correlations between intelligence measures and hitrate, these correlations are low to moderate in magnitude and the majority arenot statistically significant. No clear pattern emerges between time periods. Thatis, the correlations do not systematically increase or decline over time on eitherthe WPPSI-R or SB:FE.Normal GroupThe normal group’s hit rate correlations, as shown in Table 9, revealsthat the first time period was the only time period with significant correlations.Similar to the at-risk group, the WPPSI-R Full Scale IQ (r = 0.378, p<.O5)and the WPPSI-R Performance Scale. Cr = 0.451, p<.05) were significantlycorrelated with hit rate in the first time period.The normal group only has two significant correlations on the WPPSI.Rfor the first time period and none on the SB:FE. All the correlations arepositive, neither intelligence scale shows a pattern between time periods wherethe coefficients systematically increase or decline in magnitude.Results I 71Table 8. At-Risk Group Hit Rate Correlations with IQ.Time 1 Time 2 Time 3 TotalHit Rate Hit Rate Hit Rate Hit RateWPPSI—R .174 394* .186 .307Full ScaleWPPSI-RVerbal .071 .236 .010 .420*WPPSI-RPerformance .244 .485* .316 .128SB: FEComposite .063 .094 .035 .044SB:FEVerbal .200 .114 .207 .099SB:FEAbstract/Vis .170 .094 .048 .034SB:FEQuantitative .002 .056 .177 .110SB: FEMemory .174 .152 .050 .167*p<.05Note: N = 29Results / 72Table 9. Normal Group Hit Rate Correlations with IQ.Time 1 Time 2 Time 3 TotalHit Rate Hit Rate Hit Rate Hit RateWPPSI—R .378* .042 .124 .004Full ScaleWPPSI—RVerbal .214 .233 .056 .271WPPSI -RPerformance .451* .321 .145 .283SB:FEComposite .307 .109 .266 .040SB:FEVerbal .289 .293 .177 .165SB: FEAbstract/Vis .281 .196 .381 .181SB:FEQuantitative .110 .105 .255 .122SB: FEMemory .155 .319 .001 .305*p<.05Note: N = 29Results I 732. False Alarms and IntelligenceAt-risk GroupTable 10 shows correlations between false alarms and IQ for the at-riskgroup. All the correlation coefficients for the WPPSI-R factors are consistentlynegative throughout the three time periods, and are nonsignificant and low inmagnitude ( range: r = -0.121 to -0.235, p >.05).The four SB:FE area scores and Composite all appear as negative valuesconsistently ( range: r = -0.140 to -0.402, p<.O5). The SB:FE Composite SASshows a significant correlation coefficient for the first time period (r = -0.358,p<.O5) and total time period (r = -0.350, p<.05). Similarly, the SB:FEAbstractlVisuai Reasoning SAS shows a significant correlation coefficient in thefIrst time Cr = -0.402, p<.O5) and total time period (r = -0.359, p<.O5). Theother three area scores (Verbal, Quantitative, and Short-Term Memory) are notstatistically significant, and their correlation coefficients are low in magnitude(range: r = -0.140 to -0.311, p > .05).Normal GroupThe normal group does not show any significant correlation coefficientsbetween false alarms and the WPPSI.R as seen on Table 11. On the SB:FE,only one significant correlation coefficient emerged on the third time period forthe AbstractJVisual Reasoning Area.Results I 74Table 10. At—Risk Group False Alarm Correlations with IQ.Time 1False Alarms—.196—.121—.235Time 2False Alarms—.172—.153—.163WPPSI-RFull ScaleWPPSI-RVerbalWPPSI-RPerformanceSB:FECompositeSB:FEVerbalSB:FEAbstract/VisSB:FEQuantitativeSB:FEMemoryTime 3False Alarms—.182—.204—.146—.276—.196—.223—.234—.226Total—.226—.216—.204—.350—.220—.359—.293—.202—.358—.210—.402*—.274—.200—.289—.162—.311—.264—.140*p<.05Note: N = 29Results I 75Table 11. Normal Group False Alarm Correlations with IQ.WPPSI-RFull ScaleWPPSI-RVerbalWPPSI-RPerformance—.226—.332—.212—.022—.195—.020—.126—.078—.102.071—.258—.219—. 355—.087—.188—.257—.327—.258—.061—.165Time 1 Time 2 Time 3 TotalFalse Alarms False Alarms False Alarms—.237 .023 —.198 —.202—.224 —.095 —.116 —.209—.189 .181 —.228 —.128SB:FECompositeSB: FEVerbalSB:FEAbstract/VisSB:FEQuantitativeSB:FEMemory*p<.05Note: N = 29Results / 763. d’ and IntelligenceAt-risk GroupTable 12 shows the at-risk group’s d’ correlations with the intelligencemeasures. On the WPPSI-R, the first time period was the only time period toindicate significant correlation coefficients. These were found on the WPPSI-R FullScale IQ (r = 0.354, p<.05) and WPPSI-R Performance Scale IQ Cr = 0.434,p<.O5). The SB:FE did not have any significant correlations for the testcomposite or any of the four area scores.Normal GroupThe normal group’s d’ correlations with the intelligence measures isshown on Table 13. None of the time periods for any of the three WPPSI-Rfactors show a significant correlation coefficient. On the SB:FE, the Short-TermMemory SAS in the second time period was the only area score to show asignificant correlation coefficient (r = -0.391, p<.05). Although the at-risk group’sd’ correlations with the intelligence measures shows a decline in magnitudebetween the first, second and third time periods, no such pattern emerges onTable 13 for the normal group with the exception of the WPPSI-R PerformanceScale where the correlation coefficients increase over time. However, thesecorrelations are low in magnitude and not statistically significant, (p >.05).Results I 77Table 12. At-Risk Group d’ Correlations with IQ.Time 1 Time 2 Time 3 TotalWPPSI—R •354* .205 .051 .233Full ScaleWPPSI-RVerbal .219 .173—.088 .069WPPSI-RPerformance •434* .206 .160 .345SB:FEComposite .225 .223—.073 .130SB: FEVerbal —.049 .181—.273—.043SB:FEAbstract/Vis .195 .185—.015 .085SB:FEQuantitative .231 .175 .081 .157SB:FEMemory .305 .190—.087 .197*p<.05Note: N 29Results / 78Table 13. Normal Group c1 Correlations with IQ.Time 1 Time 2 Time 3 Totald’WPPSI—R —.006 —.016 .097 .217Full ScaleWPPSI-RVerbal—.033 —.253—.126 .018WPPSI-RPerformance .014 .226 .296 .341SB:FEComposite .084 —.086 .049 .248SB:FEVerbal .148 —.152—.073 .207SB:FEAbstract/Vis .144 .128 .230 .351SB:FEQuantitative .035 .148—.225 —.093SB:FEMemory —.071 —.391*—.026 —.025*p(.05Note: N = 29Results 1 794. and IntelligenceAt-risk GroupTables 14, and 15 indicate that on the WPPSI-R, and SB:FE nosignificant correlation coefficients emerged for any of the three time periods onthe three WPPSI-R factors for the values.Normal GroupThe normai group shows in Table 15, a significant correlation betweenthe third time period and the Quantitative Area SAS (r = -0.391, p<.O5). Thisrepresents the only significant correlation coefficient. Table 15 also shows asignificant correlation between the first time period and Short-Term Memory SAS(r = 0.480, p<.05) and second time period (r 0.433, p<.O5).Results I 80Table 14. At-Risk Group Beta Correlations with IQ.Time 1 Time 2 Time 3 TotalBeta Beta Beta BetaWPPSI—R .032—.205 .085 .002Full ScaleWPPSI -RVerbal—.008—.101 .084 .058WPPSI-RPerformance .072—.266 .067 —.043SB:FEComposite—.049—.268 .186 —.076SB:FEVerbal—.133 .012 .164 —.002SB: FEAbstract/Visual .091—.271 .219 .047SB:FEQuantitative—.133—.274 .065 —.071SB:FEMemory—.001—.271 .155 —.231Note: N = 29Results / 81Table 15. Normal Group Beta Correlations with IQ.Time 1 Time 2 Time 3 TotalBeta Beta Beta BetaWPPSI—R .161 .263—.111 —.101Full ScaleWPPSI-RVerbal .298 .318—.200—.321WPPSI-RPerformance—.005 .164 .006 .158SB:FEComposite .260 .327 —.016 —.078SB:FEVerbal .176 .185 —.015—.161SB:FEAbstract/Visual .110 .227 —.124 .018SB:FEQuantitative .052.074 —.391*—.115SB:FEMemory .480* •433*—.031—.313*p<.05Note: N = 29Results / 82D. INTERCORRELATIONAL PAVERNSIn order to determine if different correlational patterns emerged betweengroups, a computer program, MTJLTICORR, developed by Steiger (1979), wasused to analyze intercorrelations. MULTICORR was designed to perform accuratesmall sample tests between intercorrelational patterns. Steiger (1979, 1980a,1980b) explains that a correlational pattern hypothesis specifies that certaingroups of elements in a correlation matrix are equal to each other, andlor to aspecified numerical value.MULTICORR’s statistical rationale is as follows: First, sample correlationsare normalized using the Fisher r-to-z- transformation. The variance-covariancematrix of the transformed correlations is then estimated from the samplecorrelation matrix through the use of asymptotic theory. Steiger (1979) furtherexplains that the dependencies among the correlations are compensated in aquadratic form chi-square statistic, which is similar in structure to Hotefling’s T2(Hotelling, 1933, 1940, 1953).Monte Carlo studies (Steiger, 1980a,b) have empirically demonstrated thatMULTICORR controls Type I errors at nominal levels for sample sizes as smallas 20 subjects. Moreover, Steiger (1979, 1980a,b) reports that MULTICORR isnotably superior to traditional likelihood ratio approaches which tend to reject thenull hypothesis too often when sample sizes are small.In this study, it was hypothesized that the at-risk and normal group’sResults / 83intercorrelations between intelligence and attention measures would yielddistinctively different patterns.1. Variable Selection RationaleA matrix of the intercorrelations of all the vigilance measures (HR, FAR,d’, ) and the WPPSI-R Full Scale IQ was computed separately for both groupsand is shown in Table 16 (see appendices for additional intercorrelation tables).These particular variables were selected for theoretical reasons and for purposesof parsimony. That is, rather than include all the variables available foranalysis, the number of variables was reduced to include nonredundant measuresas well as measures of theoretical interest. The justification for reducing the totalnumber of variables was as follows. First, intelligence variables were reduced toinclude only the WPPSI-R Full Scale IQ. The SB:FE Composite SAS wasconsidered redundant because of the correlation with the WPPSI-R Full Scale IQ(both groups: r = 0.885, p < .001; at-risk group: r = 0.725, p <.001; normalgroup: r = 0.729, p <.001). The WPPSI-R Full Scale IQ has the highestreliability coefficient (r = 0.96) compared to the Verbal (r = 0.95) andPerformance (r = 0.92) scales (Wechsler, 1989).t Also, the global measure ofintelligence (WPPSI-R Full Scale IQ) was of the greatest theoretical interestbecause of its use in previous research (e.g., Detterman & Daniel, 1989).t The reliability coefficient for the SB:FE Composite SAS for all four areascores for five year olds is r = 0.97, compared to the Verbal Reasoning SAS (r= 0.91), AbstractlVisual Reasoning (r = 0.93), Quantitative Reasoning (r =0.88), and Short-Term Memory (r = 0.92) Standard Age Score (Thorndike et al.,1986).Results / 84Second, coefficients related to time intervals were not included in theanalyses because the earlier reported Pearson product-moment correlations werenot meaningful andlor significant.2. Matrices ComparisonsIntercorrelational matrices for the normal and the at-risk group is shownin Table 16. The results indicate that there are significant differences betweenthe two group’s intercorrelational patterns, (x2 (10) = 188.789, p <.001). Asshown in Table 16, a trade-off occurs in d’ and 3 scores for the at-risk group(r = -0.788). In contrast, the normal group’s d’ and are positively related (r= 0.4 19). Thus, the positive and negative correlation signs indicate the measuresdiffer in direction. Similarly, Table 16 shows the at-risk group’s correlationsbetween j3 and hit rate (r = -0.465) differs in magnitude and direction from thenormal group (r = 0.702). Table 16 also shows that the only meaningfulcorrelationt between IQ and vigilance occurs for the at-risk group. A moderatecorrelation occurred between IQ and hit rate (r = 0.3 18).Moreover, IQ notwithstanding, in interpreting the magnitude of theintercorrelations shown in Table 16, it is evident that the at-risk group hadthree correlation coefficients (d’ & hit rate, a & hit rate, 13 & d’) above the .40level in comparison to the normal group which had four correlation coefficients(d’ & hit rate, 13 & hit rate, j3 & d’, false alarms & d’) above the .40 level.t In interpreting the correlations, only coefficients above 0.30 were consideredmeaningful (cf., Swanson & Cooney, 1989).Results / 853. Average Correlation ComparisonsIn order to determine if the groups varied in terms of mean correlationsize, the average intercorrelation of each group was computed. This was doneusing Kaiser’s (1968) method of tabulating and comparing differences incorrelations. In this method the largest eigenvalue minus one divided by thenumber of variables minus one, yields an estimate of the average correlation inthe matrix. Thus, the average correlation for the at-risk group was r = 0.395,and the normal group was r = 0.301. However, the difference between themagnitude of these two correlations was not found to be significant using Fisher’sz transformation of the correlation coefficients (Glass & Hopkins, 1984).Results I 86Table 16.Intercorrelations of Intelligence and Vigilance.At-Risk Group1 2 3 4 51. WPPSI—R 1.0002. Hit Rate .318 1.0003. FAR —.196 —.022 1.0004. d .235 .834 —.363 1.0005. Beta .002 —.465 .262 —.788 1.000Normal Group1 2 3 4 51. WPPSI—R 1.0002. Hit Rate .017 1.0003. FAR —.201 .047 1.0004. d’ .211 .792 —.508 1.0005. Beta —.101 .702 .341 .419 1.000Results I 87E. FACTOR ANALYSESFactor analyses were conducted after Steiger’s (1979) MULTICORRanalysis. These factor analyses were used in order to gain a better interpretationof the respective group’s correlational patterns. Thus, although both Steiger’s(1979) MULTICORR program, and Kaiser’s (1968) correlation matrix comparisonsindicated differences in intercorrelation between the normal and at-risk groups inthis study, factor analyses were used to further interpret the correlationalpatterns and to test the hypothesis that intelligence and attention share acommon factor. This was done using a maximum-likelihood model testingprocedure (e.g., Gorsuch, 1983; Helwig & Council, 1979; SAS Institute, 1985).1. Normality AssumptionsMaximum likelihood estimation is known to be sensitive to violations ofthe multivariate normality assumption (Gorsuch, 1983). In order for thisassumption to be considered reasonable, a variable’s skewness and kurtosis mustapproach zero (Kyllonen & Christal, 1990). These values are shown in Table 17.Additionally, the intelligence measure (WPPSI-R Full Scale IQ) and the vigilancemeasures (hit rate, false alarms, d’, and 13) were tested to determine whetherthey met the assumptions for normality (Shapiro & Wilk, 1965, 1968). In thistest the null hypothesis predicts a normal distribution with an unspecified meanand variance. The null hypothesis is rejected when the probability of a is lessthan the critical value. Thus, for both the at-risk and normal groups, aShapiro-Wilks test statistic was computed for the intelligence measure (WPPSI-RResults I 88Full Scale IQ) and the four vigilance measures (d’, j3, I{R, FAR). Table 17shows that all of these variables met the W test criteria for the critical level,and as a result the null hypothesis was not rejected.Further analyses were conducted as the Shapiro-Wilk test indicated themeasures met the assumptions for normality. Geweke and Singleton (1980) statethat maximum-likelihood estimates of parameters are acceptable under theassumption that the subjects are drawn from a population that is multivariatenormally distributed in the variables being studied.Geweke and Singleton (1980) found that although the goodness-of-fit teststatistic is distributed asymptotically as chi-square, its use in small samples isreliable.The suggestion of Lawley and Maxwell (1971, p.36) that Bartlett’sversion of the test can be trusted only if n—p 50 is probablytoo pessimistic. The fewer factors being fit, the sooner asymptoticdistribution theory becomes appropriate as sample size is increased, thethreshold being approximately 10 observations for one factor andperhaps 25 for two ... the likelihood ratio test has considerable powereven when sample size is only 10. (p.136).Moreover, to the extent that there may be a small sample bias, it is likely tobe a conservative one which would lead to an increased probability of falserejection of the factor model. As a result, Geweke and Singleton’s (1980) MonteCarlo study found that testing factor models with small samples is reliable. Asfew as ten observations are adequate with five variables and one common factor.Results I 89Table 17. Normality Statistics.At-Risk GroupM SD Kurt Skew Mm Max W:NormalIQ 82.5 10.8 —.32 .12 63 104 .974Hit Rate 73.3 .19 —.35 —.51 28 100 .944False Al 11.3 .114 1.38 .99 00 84 .5261.91 1.51 .64 —.91 .19 4.2 .925Beta 3.29 1.72 3.53 1.1 3.9 8.1 .905Normal GroupM SD Kurt Skew Mm Max W:NormalIQ 110.5 13.6 —.16 —.11 81 135 .962Hit Rate 92.0 .08 —.17 —.31 66 100 .919False Al 3.3 .04 1.88 .99 00 17 .6853.56 .71 .59 —.52 2.0 4.6 .935Beta 3.24 1.72 .55 2.1 .09 13 .761Note: Kurt = KurtosisW:Normal = Wilk’s NormalityResults / 902. Maximum-Likelihood Factor AnalysesPrevious research (e.g., Swanson & Cooney, 1989), found that amaximum-likelihood factor analysis of vigilance and intelligence indicated athree-factor model best fit the pattern of correlations for an older group ofchildren (grades 5 to 7; mean chronological age of 11.9 years, with a standarddeviation of 2.38) on similar measures of vigilance as in the present study, andthe Wechsler Intelligence Scale for Children - Revised (WISC-R). Thus, it washypothesized that a similar model would extend downward to kindergartenchildren (mean chronological age of 70 months, with a standard deviation of 4.23months) using the four vigilance measures (HR. FAR, d’, 13) and the WechslerPreschool and Primary Scale of Intelligence - Revised (WPPSI-R).In order to provide a preliminary test of the hypothesis that vigilancemeasures share a common factor with intelligence, a series of exploratorymaximum-likelihood factor models were fit to the correlations of the two groupsreported in Table 16. The maximum-likelihood analysis makes it possible to testhypotheses about the number of common (hypothetical) factors by a procedurethat extracts a simple factor and then adds factors that increase the fit of themodel to the data (Bentler, 1980; Bentler & Bonett, 1980; Bentler & Weeks,1980). As explained by Bentler and Bonett (1980), in any model-fitting procedureit is useful to supplement the statistical test with a coefficient or index thatreflects, with a percentage, how far a particular model is from perfect fit. Theydeveloped an index that is independent of the sample size and that gauges inproportional terms the degree to which the model departs from a perfectResults I 91accounting of the data. This coefficient of fit is referred to as . It is asimple function of the chi-square test for the specific theoretical model underconsideration and the chi-square test for a model that hypothesizes that thevariables are uncorrelated in the population. This goodness-of-fit index wascomputed in the present study to assess the amount of variance accounted for inthe maximum-likelihood analyses.The maximum-likelihood solution for the at-risk and normal groups isshown in Table 18. As indicated by the chi-square test, a two-factor model fitthe data quite well for the at-risk group and the normal group. Additionally,Schwarz’s Bayesian criterion (Schwarz, 1978) was used to determine the “coz4e t”number of factors to represent the variables (cf. Hakstian & Muller, 1973;Hakstian, Rogers, & Cattell, 1982). This criterion is used for estimating the bestnumber of factors to include in a model when maximum-likelihood estimation isused. This criterion assumes that the number of factors that yields the smallestSchwarz Bayesian value is considered best.For the at-risk group, the two-factor model fit the data so well thatadditional factors were not required. However, a model with one and two factorswere tested to see if three factors were necessary. However, a three-factor modelwas inappropriate as the ratio of the five variables (WPPSI-R FSIQ, Hit Rate,False Alarms, d’, 13) to the three factors yields negative degrees of freedom(Kim & Mueller, 1978, p. 50).The two-factor likelihood-ratio chi-square test yielded x2 (1) 9.891, pResults I 92<.001. Bentler and Bonnett’s (1980) goodness-of-fit index is computed from thenull model x2 (10) = 91.067, p <.001, and the current at-risk group two-factormodel, x2 (1) = 9.891, p <.001. Consequently, as = (91.067- 9.891)!9 1.067 = .89 1. Therefore, the model is 89% of the way to a perfect fit.For the normal group, the two-factor likelihood-ratio chi-square testyielded x2 (1) = 2.453, p = .117. Thus, as = (94.871 - 2.453)! 94.871 =.974. Therefore, the model is 97% of the way to a perfect fit.In both the at-risk and normal groups the Schwarz’s Bayesian criterioncontinued to decrease from the one-factor (at-risk = 34.382 and normal =48.617) to two-factor (at-risk = 29.505 and normal = 25.042) solutions.Consequently, a two-factor solution best fit the data for the at-risk group and forthe normal group. That is, according to Schwarz (1978), the number of factorsthat yields the smallest value for the Bayesian Criterion is considered best indetermining the number of factors.Gorsuch (1983) differentitates between exploratory maximum-likelihoodanalysis and confirmatory maximum-likelihood factor analysis. The former fmdsthose factors that best reproduce the variables under the maximum-likelihoodconditions and the latter tests specific hypotheses regarding the nature of thefactors and predefined characteristics of the extracted factors. The present studyuses exploratory maximum-likelihood factor analyses. These analyses, according toGorsuch (1983) are unrestricted solutions. An unrestricted solution can be obtainedby rotating from another maximum likelihood solution. In this study, orthogonalResults I 93varirnax rotations (Kaiser, 1958; SAS Institute, 1985) were used to simplify thefactor structure by maximizing the variance of the columns of the patternmatrix. tIn interpreting the maximum-likelihood factors, only coefficients above .30were considered meaningful (Gorsuch, 1983; Kim & Mueller, 1978). Table 18shows that the first factor extracted for the at-risk group is anintelligence-vigilance factor. That is, the WPPSI-R FSIQ (.323) and the vigilancemeasures of Hit Rate (.984), d’ (.726), and (-.332) show meaningful loadingson the first factor. The normal group, however, does not show a meaningfulfactor loading for IQ (.03 2). The first factor shows meaningful loadings on thevigilance measures of Hit Rate (.959), d’ (.738), and (.766).Also, Factor 1 shows meaningful loadings for (at-risk = -.332; normal= .766). Although 3, a signal detection measure of decision making or cognitivemonitoring of responses, shares a common factor for both groups, they arerelated in opposite directions at different ends of the factor continuum. That is,the at-risk group is at the negative end (-.332) and the normal group is at thepositive end (.766). This indicates, unlike the other signal dection measures, d’and hit rate, which load in the same positive direction, the two groups arepolarized in terms of conservative decision making or cognitive monitoring ofresponses. Thus, j3 as an independent decision criterion measure shows that thetPrior to factor analyzing the data, the scores were normalized using a z-scoretransformation. The justification for using transformed scores was due to the factthat maximum-likelihood estimation assumes normally distributed variables in orderto achieve an optimal solution. The normally distributed transformed z-score isthe best estimate of a score that is linear (i.e., perfectly correlated) with the“true” underlying trait score (Kyllonen & Christal, 1990).Results I 94at-risk group is less conservative or less risk taking the the normal group. Thisis consistent with previous research (e.g., Swanson, 1981, 1983) that comparedlearning disabled children with nondisabled.For both groups, Factor 2 shows an inaccuracies in responding commonfactor (FAR) with meaningful loadings on False Alarms (at-risk = .62 1; normal= .941). The at-risk group also shows a meaningful loading for (.795) onFactor 2 which also reflects inaccuracies in responding.Results / 95Table 18. Maximum Likelihood Factor Analysis.At-Risk Group Varimaz Rotated FactorVigilance/IQ False AlarmsWPPSI—R FSIQ .323 .003Hit Rate .984 -.173False Alarms .086 .621.726 —.686Beta—.332 .795% of total variance 46.8 53.1Normal Group Varimax Rotated FactorVigilance False AlarmsWPPSI—R FSIQ .032 —.279Hit Rate .959 —.124False Alarms .171 .941d .738 —.674Beta .766 .218% of total variance 41.5 58.4Note: N = 29 at-risk groupN 29 normal groupResults / 96F. REVIEW OF HYPOTHESESA summary of the results in relation to the hypotheses follows:• Hypothesis 1. The normal group’s mean level of vigilance performance willbe significantly higher than the corresponding means of the group withlearning difficulties.The ANOVA results indicated that the normal group’s mean level ofvigilance was significantly higher than the at-risk group. The at-risk groupmade more false alarm reports and made significantly fewer correctdetections than the normal group. There were no interaction effects.The ANCOVA results indicated that d’ was the only significant measurediffering between the groups. This indicates that regardless of differences inIQ, the d’ scores differed between the groups.• Hypothesis 2. Measures of vigilance (HR, FAR, d’ and ) will besignificantly positively correlated with measures of intelligence, arid notinfluenced by time on task.The results of the correlational analyses did show that the vigilancemeasures were not influenced by time on task. However, it is important tonote that only the WPPSI-R showed a few significant correlations with thevigilance measures. The SB:FE showed even fewer (hardly any) significantcorrelations with the vigilance measures.Results / 97• Hypothesis 3. The correlations between measures of vigilance and measures ofintelligence will be significantly higher among subjects identified as havinglearning difficulties than among normally achieving subjects.Intercorrelational matrices for the normal and the at-risk group indicate thatthere are significant differences between the two group’s WPPSI-R FSIQand vigilance intercorrelational patterns, (x2 (10) = 188.789, p <.001).• Hypothesis 4. When measures of intelligence and vigilance are included in afactor analysis, measures of both types of ability will load together on generalability factor(s) in subjects identified as having learning difficulties, whereasamong normally achieving subjects the factor pattern will show relatively morespecific factors.On the exploratory maximum-likelihood factor analyses, only the at-riskgroup had a meaningful factor loading of intelligence and vigilance. Thenormal group, on the other hand, showed no such loading. That is, thefirst factor for the at-risk group showed a factor loading of vigilance andIQ on a similar factor. Whereas the first factor for the normal group wasa vigilance factor independent of intelligence and the second factor reflecteda inaccuracies in responding (false alarms), which was also independent ofintelligence.VI. DISCUSSIONA. REVIEW OF THE PURPOSE OF THE STUDYThe purpose of this study was to compare two theoretical models(arousal versus resource-limitation) as an explanation for the relationship betweenintelligence and attention in kindergarten children. The major assumptions of thesetwo models are as follows.First, the arousal model assumes that there is an optimum level ofarousal associated with performance, arid that either a excessive decrease orincrease in arousal produces impairments on task performance (Leob & Alluisi,1984). Within the context of vigilance, the arousal theory assumes that as moretime is spent on a sustained attentional task, negative correlations betweenattentional performance and intelligence become more pronounced. Stankov (1983)suggests that the monotonous conditions of an attentional task causes people withhigher scores on intelligence tests to perform worse in the latter phases of atask because they become bored and lose an optimum level of arousal. Therefore,the arousal model assumes that increased time on task leads to a decrement inarousal levels as a function of intelligence.Second, the limited-resource model assumes that individual differences inattentional capacity underlie intelligence (Hunt, 1978, 1980; Swanson & Cooney,1989; Tomporowski & Simpson, 1990). Furthermore, this model assumes that98Discussion / 99people differ in the attentional resources they possess and these individualdifferences persist across time periods. Thus, individuals with lower attentionalresources are more likely to have lower intelligence and/or learning ability(Stankov, 1989). Previous research (e.g., Swanson, 1981, 1983; Swanson &Cooney, 1989; Tomporowski & Simpson, 1990), found that stimulus detection (d’)is reflective of a limited-resource capacity system. The limited-resource modelassumes that stimulus detectability, or attentional capacity remains independent oftime on task. Further, individuals with lower intelligence andlor learning abilityhave lower d’ scores than their counterparts.Overall, the results were interpreted as suggesting that ability groupdifferences reflect attentional capacity. Two findings were of importance in thisinterpretation. First, ability groups varied on the d’ measure. Previous researchliterature (e.g., Sostek et al., 1980) note that children with known impairedattentional capacity due to minimal brain dysfunction have a low sensitivity (d’scores) to continuously presented signals. Similarly, signal detection deficits havebeen related to individual differences in verbal ability (Sen & Sachdev, 1977),intelligence (Herman, Streissguth, & Little, 1980) and kindergarten schoolreadiness (Edley & Knopf, 1987; Simon, 1982). Based on this literature, it seemsplausible that high stimulus detectability (d’) reflects a subject’s capacity toprocess information. The notion of a limited processing capacity system isconsistent and central to a number of theories of information processing (Hasher& Zacks, 1979; Hunt, 1978; Kahneman, 1973; Schneider & Shiffrin, 1977;Shiffrin & Schneider, 1977).Discussion I 100Second, d’ measures did not vary over time. From a signal detectionperspective, the invariance of d’ over time is assumed to represent “truedetectability” in an individual’s capacity to discriminate continuously presentedinformation. The present results indicate that in the normal achieving group, d’remained stable over time. This finding is consistent with Swanson’s (1981, 1983)earlier studies, which indicated that learning disabled students suffer a from asmaller processing capacity rather than any loss of sustained attention over time.Similarly, in the present study, the at-risk group reflects a smaller capacity toprocess successive informational stimuli.The specific results to each of the four research questions will now beaddressed.B. ABILITY GROUP DIFFERENCES IN VIGILANCEAn important question addressed in this study was whether abilitygroups differ on vigilance measures. To that extent, one of the important findingsof this study, was that regardless of differences in IQ, the d’ scores differedbetween groups. This result is consistent with the notion that both groups differin actual attentional resources and that these differences persist even whenintelligence is controlled. What this finding suggests is that attentional differencesmay reflect fundamental capacity differences between the groups that cannot beimproved upon with variation in IQ. That is, when the influence of IQ iscovaried with attention, the d’ measure showed differences between the groups.Thus, in this study, actual attentional resources were evident and distinguishedDiscussion / 101by d’. Regardless of the influence of IQ, d scores differed, which suggests afundamental capacity difference between the groups. As stated by Das, Kirby,and Jarman (1979) state that “Capacity indicates something that is given, andperhaps cannot be altered” (p.25).Two measures of attention (Hit Rate and False Alarms), were a functionof IQ. Significant differences between the groups occurred for both hit rate andfalse alarms. That is, the normal group made significantly more correctdectections and fewer false alarms than the at-risk group. However, thesedifferences between the groups did not occur when intelligence was controlled.Therefore, on these two vfgilance measures, level of intelligence makes adifference. In contrast, the 13 signal detection measures did not show anydifferences between groups when intelligence was in the analyses, or whenintelligence was controlled. More discussion of the 13 measure will follow in alater section in this chapter.C. TIME ON TASKA second question addressed in this study was the relationship betweenIQ and time on task. Although the present study’s results indicate only arelative few significant correlations between the two IQ tests and attention, therewas also no support for the notion that intelligence and attention are influencedby time on task. The results do not support the notion that correlations betweenintelligence and vigilance measures are related to arousal. That is, as the timeintervals progressed, significant changes did not occur between the vigilance andDiscussion I 102intelligence correlations. Thus, this finding is inconsistent with Stankov (1983): “Itcan be hypothesized that individuals with higher scores on intelligence testsperform worse in the latter part of the working period because they becomebored (lose arousal) more easily with a simple task (italics added)” (p.481).The time periods in the present study differ from those reviewed byStankov (1983). The present study used a six minute continuous performancetask with kindergarten children. Stankov (1983) reviewed simple attentiontaskst with a working period of one hour using university students as subjects.Although the time periods differed greatly (six minutes versus one hour), theimportant point of relevance is Stankov’s (1983) notion of a simple task wherepeople with higher general ability show lower arousal levels at the end of thetask. The present study did not find that higher IQ subjects had lost arousal atthe end of the task.The present results are consistent with Tomporowski and Simpson’s(1990) study which did not find any empirical support for Stankov’s (1983)“arousal construct”. Tomporowski and Simpson (1990) found that individuals withlower intelligence levels are required to allocate greater amounts of their limitedattentional resources during information-processing activities than individuals withhigher intelligence levels. Thus, the information-processing demands placed on thesubjects throughout the vigilance task resulted in a continuous expenditure ofattentional resources. Therefore, both Tomporowski and Simpson’s (1990) studyand the present results suggest that the differences between the two groups ontStankov describes these tasks as the Dominoes test, Flexibility of Closure, twoPerceptual Speed and Search Tests, and simple reaction time.Discussion / 103the vigilance task is reflected in attentional resource availability rather thanarousal.D. INTERCORRELATIONAL PATTERNSA final question this study addressed was related to the intercorrelationalpatterns among the IQ and vigilance measures. The present results indicated thatthe intercorrelational patterns between intelligence and attention measures of thetwo groups were significantly different. The results indicated that only the at-riskgroup showed a moderate intercorrelational relationship between intelligence andthe vigilance measures of hit rate and stimulus detectability (d’), whereas thenormal achieving group did not.In order to gain a better interpretation of the respective group’sintercorrelational patterns, factor analyses were used to test a model thatintelligence and vigilance share a common factor in the at-risk but not thenormal achieving group.On the maximum-likelihood factor analysis, only the at-risk group had ameaningful factor loading of intelligence and vigilance. The normal group, on theother hand, showed no such loading. That is, the first factor for the at-riskgroup showed a factor loading of vigilance and IQ on a similar factor. Whereasthe first factor for the normal group was a vigilance factor independent ofintelligence and the second factor reflected a inaccuracies in responding (falsealarms), which was also independent of intelligence.Discussion / 104Jarman (1980a) notes that factor analyses of intelligence strata makes itpossible to hypothesize conditions which would produce different results fordifferent intelligence levels.In terms of factor analyses by intelligence strata, therefore, one wouldexpect increasing variation across groups due to the effects of differentstrategies, with the use of more highly strategic or heterogeneoustasks. Conversely, less variation or factor invariance should be foundin studies employing tasks which minimize cognitive strategies, that is,homogeneous tasks. (p. 81).In summary, the results in the present study indicate that intelligenceand vigilance share a common factor and correlate higher for the at-risk groupthat the normal group. Thus, intelligence plays a greater role in relation tovigilance for the at-risk group. This finding is consistent with Detterman andDaniel’s (1989) conclusion that cognitive tasks correlate more highly withintelligence at lower ability levels than at higher ability levels.E. IMPLICATIONS OF THE STUDYDetterman and Daniel (1989) suggest that correlations between intelligencetests vary systematically by ability levels. If this is found to be a generalfinding, and not specific to just certain cognitive ability tests, then theimplications for intelligence theory and research are substantial. However, asDetterman and Daniel (1989) caution, “Without further verification and replicationDiscussion / 105of this phenomenon, extensive theoretical speculation is premature” (p.357). Aplausible explanation is offered in the example Detterman and Daniel provide asthe .30 correlation barrier, as first discussed by Sternberg and Salter (1982).That is, correlations between cognitive tasks and IQ have often been under the.30 level. Detterman and Daniel (1989) suggest that the reason for the .30correlation barrier is probably due to the fact that most researchers investigatingthe relationship between cognitive abilities and intelligence generally do not includelow IQ subjects in their samples. Thus, research studies conducted with normalstudents alone will not be generalizable to the full spectrum of intellectual ability.The present study shows that intelligence and sustained attention share afactor for the at-risk group but not the normal group. Clearly, these resultssuggest that the role of intelligence is different for the at-risk students than thenormals. What causes these differences between the two groups may be one ofthe most interesting implications of the present study for future research. It maybe, as suggested by Detterman and Daniel (1989), cognitive tasks and intelligencecorrelate higher for lower ability groups. This may be because the at-risk groupsuffer information-processing capacity limitations which permeate IQ and vigilancemeasures. An analogy Detterman and Daniel (1989) suggest is that highcorrelations are due to executive function differences (cf. Butterfield, 1981;Swanson, 1989) between high and low ability groups. Detterman’s (1987)theoretical notions of inteffigence and mental retardation may serve as a potentialexplanation for what causes these differences in correlation at difference abilitylevels. As Detterman (1987) stated:Discussion / 106.intelligence is a system made up of a small number of independentprocesses. Mental retardation is caused by deficits in central processes,meaning processes which most heavily affect all other processes in thesystem. If these central processes are deficient, they limit theefficiency of all other processes in the system. Because of the deficitin the central process, the entire system is brought to a uniform lowlevel of operation. So all processes in subjects with deficits tend tooperate at the same uniform level. However, subjects without deficitsshow much more variability across processes because they do not havedeficits in important central processes. This causes high correlationsamong mental measures in low IQ subjects and low correlations inhigh IQ subjects (p.358).As shown in the present study the vigilance measure d’ (detectability)covaries with IQ in terms of a central processing function. Likewise, in terms ofother explanations, Jarman and Das (1977) note that information processing speedalso varies with IQ in terms of a central intellective function. Wickens (1974)notes that a number of experimental paradigms have shown that,developmentally, “...central processing is a function of age or maturation, and italso covaries with a number of nonprocessing variables such as practice,motivation, incentive, and attentiveness” (p.166). Also, Holden’s (1970) research onretarded and normal children indicated that central processing differences betweenthe groups was not specific to any single modality. Thus, when viewedcollectively, Wicken’s (1974) review of developmental trends, Jarrnan and Das’(1977) research on speed of successive information processing, Detterman andDiscussion / 107Daniel’s (1989) study of cognitive tasks and IQ ability group correlations, andthe present study all indicate that a variety of cognitive tasks covary with IQin relation to a central processing function. The correlation patterns between IQand vigilance may be mediated by central processing capacity. This centralprocessing capacity may be reflected in the d’ measures.If the implications of this study can be generalized, it would appear thatattention and intelligence are closely linked in at-risk kindergarten students. Thus,instruction may involve linking attention sustaining activities with learning. Thisis consistent with Das (1973) description that school learning in the lower gradessimply consists of two basic processes: the registration of information and theprocessing of information. This registration and processing will be moreconstrained in some children, and therefore compensatory means of enhancing alimited capacity will have to be explored in future studies.The present study moves the field along by demonstrating the differentialabilities of at-risk kindergarten students compared to normally achieving studentsin the registration and processing of information. The results of the presentstudy contribute to the limited-resources theory by demonstrating that limitedcapacity to register and process information affects the relationship of attentionand intelligence. The results of this study did not find similar support for thearousal model.Discussion / 108F. LIMITATIONS AND FUTURE RESEARCHThe relatively small sample size serves as a limitation in terms ofstatistical power.The measures in this study may have limited generalizability toexisting studies. This may be due to the short six minute vigilance task.Parasuraman and Davies (1984) report that many studies have shown thattime-on-task effects should be seen in rather than d’. Williges (1971) suggeststhat if observers are given sufficient time and exposure to a task, they willeventually show response criteria () that are nearly optimal. Moray, Fitter,Ostry, Favreau, and Nagy (1976) found a ten hour signal detection task hadresponse criteria () very close to the optimal value. However, most vigilancestudies use a fifteen to thirty minute task with school children. Nevertheless, onemay argue with some justification, that the six minute vigilance task used in thepresent study is reasonably representative of an intructional time period in akindergarten classroom.This study was about the relationship between two standardized IQ testsand a computerized behaviour-based measure of sustained attention. The attentiontest, The Gordon Diagnostic System’s Vigilance Task, was the only directmeasure of attention used in this study. This in and of itself may serve as alimitation. It is suggested that future research use other simple, but directmeasures of attention to examine the relationship between the cognitive processesand the attentional tasks. In addition to simple sustained attention tasks, futureDiscussion I 109research could examine the relationship of intelligence and complex competingattention tasks (e.g., Stankov, 1989) and selective attention with both receptiveand expressive attentional tasks (e.g., Naglieri, Das, & Jarman, 1990).The construct of intelligence, in this study, was conceptualized andequated with psychometric g (cf. Jensen, 1987). On the SB:FE, g represents theTest Composite or overall score indicating general reasoning ability. The SB:FE gfactor accounts for 42% of the overall variance (Thorndike et al. 1986).Similarly, the on the WPPSI-R the overall proportion of variance due to g was40% (Roid & Gyurke, 1991). Thus, these tests are heavily g dependent in theconceptualization of intelligence. Thorndike et al. (1986) explain that the use of gin intelligence testing has a controversial and long history. However, “the generalability factor, g, refuses to die. Like a phoenix, it keeps rising from its ashesand will no doubt continue to be an enduring part of our psychometric theoryand practice” (p.6).On the other hand, Das (1989) suggests:No one doubts that a measure of ‘general’ intellectual functioning isuseful; in this sense IQ may be compared to the general depth of ariver near a village. However, if I am a new swimmer, I ought toknow which parts are just too deep and hence risky for me.Similarly, if I am a fisherman, I need to know the depth of theriverbed in specific parts so that I may expect shrimps to live in oneand carps in another. Knowledge about the general depth of the riverhas limited use. In the same sense, IQs are not useful forDiscussion / 110understanding the cognitve processes of general or specific populations,but specific cognitive measures relating to coding, planning, andattention are (italics added, p.541).Clearly, the IQ tests used in the present study •measure a generalability factor. However, some readers may appropriately ask to what extent didthe IQ measures in this study incorporate aspects of attention, and how couldthis have affected the results?Gordon, Thomason, and Cooper (1990) stated that tests of intellectualability place high demands upon a child to maintain attention. However, recentresearch on the WPPSI-R (e.g., Blaha & Wallbrown, 1991; Gyurke, Stone, &Beyer, 1990; Stone, Gridley, & Gyurke, 1991) report a lack of support for the“Third” factor solution found on the WISC-R. That is, the WPPSI-R AnimalPegs, Arithmetic, and Sentences Subtests which have been associated withattention and distractibility are not appropriately interpreted as a separate thirdattentional factor. Thus, this research suggests the WPPSI-R does not directlymeasure attention.According to Delaney and Hopkins (1987), performance on the SB:FEBead Memory and Memory for Sentences Subtests “may be subject to theinfluence of attention” (p.84). The present sudy, however, did not find anysignificant correlational evidence to suggest the memory subtests measure or areinfluenced by attention. Naglieri et al. (1990) indicate that current intelligencetests do not have any provisions for a proper and detailed analysis of attention.Discussion / 111Future research might examine how other IQ measures appropriate tothis age group may produce different results than the present study. Gordon etal. (1990) found significant interrelationships among the Kaufman AssessmentBattery for Children (Kaufman & Kaufman, 1983) and the Gordon DiagnosticSystem Vigilance Task.Future research might depart from a psychometric g perspective ofintelligence and use a chronometric approach such as the Das-Naglieri CognitiveAssessment System’s attention tasks to examine their relationship acrossdiagnostic groups (Naglieri et aL, 1990).REFERENCESAmerican Psychiatric Association. (1987). Diagnostic and statistical manual ofmental disorders (3rd ed.). Washington, DC: Author.American Psychological Association. (1991). Awards for distinguished scientificcontributions: 1990. Washington, DC: Author.Anastasi, A. (1988). Psychological testing. (6th ed.). New York: MacMillan.Anderson, L.M., Brubaker, N.L., Alleman-Brooks, J., & Duffy, G.G. (1985). Aqualitative study of seatwork in first-grade classrooms. Elementary SchoolJournal, 86, 123-140.Barkley, R.A. (1981). Hyperactive children: A handbook for diagnosis andtreatment. New York: Guilford Press.Barkley, R.A., Fischer, M., Newby, R.F., & Breen, M.J. (1988). Development ofa multimethod clinical protocol for assessing stimulant drug response inchildren with attention deficit disorder. Journal of Clinical Child Psychology,17, 14-24.Beery, K.E. (1989). The Developmental Test of Visual-Motor Integration (3rd rev.).Cleveland: Modern Curriculum Press.Bentler, P.M. (1980). Multivariate analysis with latent variables: Causal modelling.Annual Review of Psychology, 31, 419-456.Bentler, P.M., & Bonett, D.G. (1980). Significance tests and goodness of fit inthe analysis of covariance structures. Psychological Bulletin, 88, 5 88-606.Bentler, P.M., & Weeks, D.G. (1980). Linear structural equations with latentvariables. Psychometrika, 45, 289-308.Berch, D.B., & Kanter, D.R. (1984). Individual differences. In J.S. Warm (Ed.).112I 113Sustained attention in human performance (pp. 143-170). New York: Wiley.Blaha, J., & Walibrown, F.H. (1991). Hierarchial factor structure of the WechslerPreschool and Primary Scale of Intelligence - Revised. PsychologicalAssessment, 3, 455-463.Bracken, B.A. (1988). Ten psychological reasons why similar tests producedissimilar results. Journal of School Psychology, 26, 155-166.Broadbent, D.E. (1953). Classical conditioning and human watchkeeping.Psychological Review, 66, 33 1-339.Broadbent, D.E. (1971). Decision and stress. London: Academic Press.Butterfield, E. (1981). Testing process theories of intelligence. In M. Friedman, J.Das, & N. O’Connor (Eds.), Intelligence and learning. (pp. 277-296). NewYork: Plenum.Carison, J.S., & Jensen, C.M. (1982). Reaction time, movement time, andintelligence: A replication and extension. Intelligence, 6, 265-274.Cattell, R.B. (1982). The inheritance of personality and ability: Research methodsand findings. New York: Academic Press.Conry, R.F., & Conry, J.L. (1985). Special needs survey: Tsartlip Project.Unpublished manuscript.Conners, C.K. (1985). The Conners rating scales: Instruments for the assessment ofchildhood psychopathology. Unpublished manuscript. Children’s HospitalNational Medical Center, Washington, DC.Das, J.P. (1973). The uses of attention. Alberta Journal of Educational Research,19, 99-108.Das, J.P. (1989). A system of cognitive assessment and its advantage over IQ.In D. Vickers & P.L. Smith (Eds.), Human information processing: Measures,mechanisms, and models (pp. 535-546). North-Holland: Elsevier Science/ 114Publishers.Das, J.P., Kirby, J.R., & Jarman, R.F. (1979). Simultaneous and successivecognitive processes. New York: Academic.Davies, D.R., & Parasuraman, R. (1982). The psychology of vigilance. New York:Academic.Delaney, E.A., & Hopkins, T.F. (1987). Examiners handbook: An expanded guidefor fourth edition users. Stanford-Binet Intelligence Scale. Chicago: RiversidePublishing Company.Detterman, D.K. (1987). Theoretical notions of intelligence and mental retardation.American Journal of Mental Deficiency, 92, 2-11.Detterman, D.K., & Daniel, M.H. (1989). Correlations of mental tests with eachother and with cognitive variables are highest for low IQ groups.Intelligence, 13, 349-359.Doich, E.W. (1956). School research in reading. Elementary English, 33, 76-80.Dunn, O.J., & Clark, V.A. (1969). Correlation coefficients measured on the sameindividuals. Journal of American Statistical Association, 64, 366-377.Edley, R.S., & Knopf, I.J. (1987). Sustained attention as a predictor of lowacademic readiness in a preschool population. Journal of PsychoeducationalAssessment, 4, 340-352.Elashoff, J.D. (1969). Analysis of covariance: A delicate instrument. AmericanEducational Research Journal, 6, 383-401.Fletcher, J.M., Satz, P., & Morris, R. (1984). The Florida longitudinal project: Areview. In S.A. Mednick & M.A. Harway (Eds.), Handbook of longitudinalresearch (Vol. 1, pp. 280-304). New York: Praeger.Frankmann, J.P., & Adams, J.A. (1962). Theories of vigilance. PsychologicalBulletin, 59, 257-272./ 115Gale, A., & Lynn, R. (1972). A developmental study of attention. Journal ofEducational Psychology, 42, 260.266.Gerber, M.M., & Semmel, M.I. (1984). Teacher as imperfect test:Reconceptualizing the referral process. Educational Psychologist, 19, 137-148.Geweke, J.F., & Singleton, K.J. (1980). Interpreting the likelihood ratio statisticin factor models when sample size is small. Journal of the AmericanStatistical Association, 75, 133.137.Glass, G.V, & Hopkins, K.D. (1984). Statisical methods in education andpsychology (2nd ed.). Englewoodcliffs, NJ: Prentice-Hall.Gordon, M. (1979). The assessment of impulsivity and mediating behaviors inhyperactive and non-hyperactive boys. Journal of Abnormal Child Psychology,7, 317-326.Gordon, M. (1986). Microprocessor-based assessment of attention deficit disorders.Psychopharmacology Bulletin, 22, 288-290.Gordon, M., DiNiro, D., Mettelman B.B., & Tallmadge, J. (1989). Observations oftest behavior, quantitative scores, and teacher ratings. Journal ofPsychoeducationcil Assessment, 7, 14 1-147.Gordon, M., McClure, D.F., & Post, E.M. (1986). Interpretive guide to the GordonDiagnostic System. Syracuse: Gordon Systems, Inc.Gordon, M., & Mettelman, B.B. (1988). The assessment of attention: I.Standardization and reliability of a behavior-based measure. Journal ofClinical Psychology, 44, 682-690.Gordon, M., Thomason, D., & Cooper, S. (1990). To what extent does attentionaffect K-ABC scores? Psychology in the Schools, 27, 144-147.Gorsuch, R.L. (1983). Factor Analysis (2nd ed.). Hillsdale, NJ: Erlbaum.Green, D.M., & Swets, J.A. (1970). Signal detection theory and psychophysics.I 116New York: Wiley.Gresham, F.M., Reschly, D.J., & Carey, M.P. (1987). Teachers as “tests”:Classification accuracy and concurrent validation in the identification oflearning disabled children. School Psychology Review, 16, 543-553.Grossberg, J.M., & Grant, B.F. (1978). Clinical psychophysics: Application of ratioscaling and signal detection method to research on pain, fear, drugs, andmedical decision making. Psychological Bulletin, 85, 11-54.Gyurke, J.S., Stone, B.J., & Beyer, M. (1990). A confirmatory factor analysis ofthe WPPSI-R. Journal of Psychoeducational Assessment, 8, 15-21.Hakstian, A.R., Rogers, W.T., & Cattell, R.B. (1982). The behavior ofnumber-of-factors rules with simulated data. Multivariate Behavioral Research,17, 193-219.Hakstian, A.R., & Muller, V.J. (1973). Some notes on the number of factorsproblem. Multivariate Behavioral Research, 8, 461-475.Hasher, L., & Zacks, R.T. (1979). Automatic and efTortful processes in memory.Journal of Experimental Psychology: General, 138, 356-388.Helwig, J.T., & Council, K.A. (1979). SAS User’s Guide. Cary, NC: SASInstitute.Herman, C.S., Kirchner, G.L., Streissguth, A.P., & Little, E. (1980). Vigilanceparadigm from preschool children used to relate vigilance behavior to IQand prenatal exposure to alcohol. Perceptual and Motor Skills, 50, 863-867.Hochhaus, L. (1972). A table for the calculation of d’ and 3. PsychologicalBulletin, 77, 375-376.Holden, E.A. (1970). Unimodal and multimodal sequential information processingin educable retardates. Journal of Experimental Psychology, 86, 181-185.Holzman, T.G., Pellegrino, J.W., & Glaser, R. (1982). Cognitive dimensions ofI 117numerical rule induction. Journal of Educational Psychology, 74, 360-373.Holzman, T.G., Peflegrino, J.W., & Glaser, R. (1983). Cognitive variables inseries completion. Journal of Educational Psychology, 75, 603-618.Horn, J.L. & Cattell, R.B. (1966). Refinement and test of the theory of fluidand crystallized intelligence. Journal of Educational Psychology, 57, 253-276.Hotelling, H. (1933). Analysis of a complex of statistical variables into principalcomponents. Journal of Educational Psychology, 24, 417-441.Hotelling, H. (1940). The selection of variates for use in preduction with somecomments on the general problem of nuisance parameters. Annals ofMathematical Statistics, 11, 27 1-283.Hotelling, H. (1953). New light on the correlation coefficient and its transforms.Journal of the Royal Statistical Society, Series B, 15, 193.Hunt, E. (1978). Mechanics of verbal ability. Psychological Bulletin, 85, 109-130.Hunt, E. (1980). Intelligence as an information processing concept. British Journalof Psychology, 71, 449-474.Jarman, R.F. (1980a). Comments on John B. Carroll’s “How shall we studyindividual differences in cognitive abilities? -Methodological and theoreticalperspectives”. Intelligence, 4, 73-82.Jarman, R.F. (1980b). Modality-specific information processing and intellectualability. Intelligence, 4, 201-217.Jarman, R.F., & Das, J.P. (1977). Simultaneous and successive syntheses andintelligence. Intelligence, 1, 151-169.Jarman, R.F., & Das, J.P. (1991). Cognitive integration: Alternative model forintelligence. In H.A.H. Howe (Ed.). Intelligence: Reconceptualization andmeasurement. (pp. 163-181). Hillsdale, NJ: Erlbaum.Jensen, A.R. (1979). The nature of intelligence and its relation to learning./ 118Journal of Research and Development in Education, 12, 79-95.Jensen, A.R. (1987). Psychometric g as a focus of concerted research effort.Intelligence, 11, 193-199.Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ: Prentice-HallKaiser, H.F. (1958). The varimax criterion for analytic rotation in factor analysis.Psychometrika, 22, 187-200.Kaiser, H.F. (1968). A measure of the average intercorrelation. Educational andPsychological Measurement, 28, 245-257.Kaufman, A.S., & Kaufman, N.L. (1983). K-ABC: Kaufman Assessment Battery forChildren. Circle Pines, MN: American Guidance Service.Keith, T.Z., Cool, V.A., Novak, C.G., White, L.J., & Pottebaum, S.M. (1988).Confirmatory factor analysis of the Stanford-Binet Fourth Edition: Testingthe theory-test mafth. Journal of School Psychology, 26, 253-274.Kim, J.O., & Mueller, C.W. (1978). Factor analysis: Statistical methods andpractical issues. Sage University Paper Series on Quantitative Applications inthe Social Sciences, series no. 07-0 14, Beverly Hills: Sage.Klee, S.E., & Garfinkel, B.D. (1983). The computerized continuous performancetask: A new measure of inattention. Journal of Abnormal Child Psychology,11, 489-496.Kupietz, S.S., & Richardson, E. (I -78). Children’s vigilance performance andinattentiveness in the classroom. Journal of Child Psychology and Psychiatry,19, 145-154.Kyilonen, P.C., & Christal, R.E. (1990). Reasoning ability is (little more than)working memory capacity?!. Intelligence, 14, 389-433.Lawley, D.N., & Maxwell, A.E. (1971). Factor Analysis as a Statistical Method.New York: Macmillan.I 119Levy, F. (1980). The development of sustained attention (vigilance) and inhibitionin children: Some normative data. Journal of Child Psychology andPsychiatry, 21, 77-84.Loeb, M., & Alluisi E.A. (1984). Theories of vigilance. In J.S. Warm (Ed.),Sustained attention in human pei’formance. (pp. 179-201). New York: Wiley.Mackworth, J.F. (1969). Vigilance and habituation. Middlesex, England: Penguin.Mackworth, J.F. (1970). Vigilance and attention: A signal detection approach.Middlesex, England: Penguin.Mackworth, N.H. (1948). The breakdown of vigilance during prolonged visualsearch. Quarterly Journal of Experimental Psychology, 1, 6-2 1.Mackworth, N.H. (1957). Some factors affecting vigilance. Advancement of science,53, 389-393.Maxwell, A.E. (1972). The WPPSI: A marked discrepancy in the correlations ofthe subtests for good and poor readers. The British Journal of Mathematicaland Statistical Psychology, 25, 283-29 1.Moray, N., Fitter, M., Ostry, D., Favreau, D., & Nagy, V. (1976). Attention topure tones. Quarterly Journal of Experimental Psychology, 28, 27 1-283.Murphy-Berman, V., Rosell, J., & Wright, G. (1986). Measuring children’sattention span: A microcomputer assessment technique. Journal ofEducational Research, 80, 24-28.Myors, B., Stankov, L., & Oliphant, G. (1989). Competing tasks, workingmemory and intelligence. Australian Journal of Psychology, 41, 1-16.Naglieri, J.A., Das, J.P., & Jarman, R.F. (1990). Planning, attention,simultaneous, and successive cognitive processes as a model for assessment.School Psychology Review, 19, 423-442.Navon, D. (1984). Resources - A theoretical soup stone? Psychological Review, 2,I 120216-234.Navon, D., & Gopher, D. (1979). On the economy of the human processingsystem. Psychological Review, 86, 214-255.Noland, F.C., & Schuldt, W.J. (1971). Sustained attention and readingretardation. Journal of Experimental Education, 40, 73-76.Parasuraman, R. (1976). Consistency of individual differences in human vigilanceperformance: An abilities classification analysis. Journal of AppliedPsychology, 61, 486-492.Parasuraman, R., & Davies, D.R. (1984). Varieties of attention. New York:Academic Press.Peterson, W.W., Birdsall, T.G., & Fox, W.C. (1954). The theory of signaldetectability. Transactions of the IRE Professional Group on InformationTheory, 4, 171-212. (Reprinted in RD. Luce, R.R. Bush, & E. Galanter(Eds.). (1963). Readings in mathematical psychology (pp. 167-211). NewYork: Wiley.The Psychological Corporation. (1989). Wechsler Preschool and Primary Scale ofIntelligence - Revised: A Technical Report. Volume 1, Number 1. SanAntonio, TX: Psychological Corporation.Reynolds, C.R. (1988). Straw men and the practice of the ancients: Comment.National Association of School Psychologists Communique, 17 (4) 6.Richter, D.O., Senter, R.J., & Warm, J.S. (1981). Effects of the rate andregularity of background events on sustained attention. Bulletin of thePsychonomic Society, 18, 207.2 10.Roid, G.H., & Gyurke, J. (1991). General-factor and specific variance in theWPPSI-R. Journal of Psychoeducational Assessment, 9, 209-223.Ross, D.M., & Ross, S.A. (1982). Hyperactivity: Current issues, research, andtheory. New York: John Wiley and Sons.I 121Rosvold, H., Mirsky, A., Sarason, I., Bransone, E., & Beck, A. (1956). Acontinuous performance test of brain damage. Journal of ConsultingPsychology, 20, 343-352.Samuels, S.J., & Turnure, M. (1974). Attention and reading achievement infirst-grade boys and girls. Journal of Educational Psychology, 6, 29-32.SAS Institute. (1985). SAS User’s Guide: Statistics Version 5. Cary, NC:SASInstitute.Sattler, J.M. (1988). Assessment of Children. San Diego: J. Sattler, Publisher.Satz, P., & Fletcher, J.M. (1982). Florida Kindergarten Screening Battery. Odessa,FL: Psychological Assessment Resources.Schneider, W., & Shriffrin, R.M. (1977). Controlled and automatic humaninformation processing: I. Detection, search, and attention. PsychologicalReview, 84, 1-66.Schwarz, G. (1978). Estimating the dimensions of a model. Annals of Statistics,6, 461-464.Sen, A., & Sachdev, U. (1977). A study on vigilance in relation to efficiency inverbal learning. Psychological Studies, 22, 36-42.Shapiro, S.S., & Wilk, M.B. (1965). An analysis of variance test for normalities.Biometrika, 52, 591-611.Shapiro, S.S., & Wilk, M.B. (1968). Approximations for the null distribution ofthe W statistic. Technometrics, 10, 86 1-866.Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic humaninformation processing: II. Perceptual learning, automatic attending and ageneral theory. Psychological Review, 84, 127- 190.Simner, M.L. (1983). The warning signs of school failure: An updated profile ofthe at-risk kindergarten child. Topics in Early Childhood Special Education,/ 12210, 17-27.Simner, M.L. (1987). Predictive validity of the Teachers School ReadinessInventory. Canadian Journal of School Psychology, 3, 21-32.Simon, M.J. (1982). Use of a vigilance task to determine school readiness ofpreschool children. Perceptual and Motor Skills, 54, 1020.Sostek, A.J., Bucksbaum, M.S., & Rapport, J.L. (1980). Effects of amphetamineon vigilance performance in normal and hyperactive children. Journal ofAbnormal Child Psychology, 3, 49 1-500.Stankov, L. (1983). Attention and intelligence. Journal of Educational Psychology,75, 47 1-490.Stankov, L. (1989). Attentional resources and intelligence: A disappearing link.Personality and Individual Differences, 10, 957-968.Steiger, J.H. (1979). MULTICORR: A computer program for fast, accurate,small-sample testing of correlational pattern hypotheses. Educational &Psychological Measurement, 39, 677-680.Steiger, J.H. (1980a). Testing pattern hypotheses on correlational matrices:alternative statistics and some• empirical results. Multivariate BehavioralResearch, 15, 335-352.Steiger, J. H. (1980b) Tests for comparing elements of a correlation matrix.Psychological Bulletin, 37, 245-25 1.Sternberg, R.J., & Salter, W. (1982). Conceptions of intelligence. In R.J.Sternberg (Ed.), Handbook of human intelligence (pp. 3-28). New York:Cambridge University Press.Stone, B.J., Gridley, B.E., & Gyurke, J.S. (1991). Confirmatory factor analysis ofthe WPPSI-R at the extreme end of the age range. Journal ofPsychoeducational Assessment, 9, 263-270.I 123Swanson, H.L. (1980). Auditory and visual vigilance in normal and learningdisabled readers. Learning Disability Quarterly, 3, 7 1-78.Swanson, H.L. (1981). Vigilance deficit in learning disabled children: A signaldetection analysis. Journal of Child Psychology and Psychiatry, 22, 393-399.Swanson, H.L. (1983). A developmental study of vigilance in learning disabledand nondisabled children. Journal of Abnormal Child Psychology, 11,415-429.Swanson, H.L. (1989). The effects of central processing strategies on learningdisabled, mildly retarded, average, and gifted children’s elaborative encodingabilities. Journal of Experimental Child Psychology, 47, 370-397.Swanson, H.L., & Cooney, J.B. (1989). Relationship between intelligence andvigilance in children. Journal of School Psychology, 27, 141-153.Swets, J.A. (1986a). Indices of discrimination or diagnostic accuracy: Their ROCsand applied models. Psychological Bulletin, 99, 100-117.Swets, J.A. (1986b). Form of empirical ROCs in discrimination and diagnostictasks: Implications for theory and measurement of performance. PsychologicalBulletin, 99, 181-198.Swets, J.A., Tanner, W.P., Jr., & Birdsall, T.G. (1961) Decision processes inperception. Psychological Review, 68, 301-340.Tanner, W.P., Jr., & Swets, J.A.. (1954). A decision-making theory of visualdetection. Psychological Review, 61, 401-409.Thorndike, R.L., Hagen, E.P., and Sattler, J.M. (1986). Stanford-Binet IntelligenceScale: Fourth Edition. Chicago: Riverside.Tomporowski, P.D., & Simpson, R.G. (1990). Sustained attention and intelligence.Intelligence, 14, 3 1-42.Thorngate, W. (1990). The economy of attention and the development ofI 124psychology. Canadian Psychology, 31, 262-271.Warm, J.S. (1984). Sustained attention in human performance. New York: Wiley.Warm, J.S., & Jerison, H.J. (1984). The psychophysics of vigilance. In J.S.Warm, (Ed.). Sustained attention in human performance (pp. 15-51). NewYork: Wiley.Wechsler, D. (1939). The measurement of adult intelligence. Baltimore: Williams &Wilkins.Wechsler, D. (1949). Manual for the Wechsler Intelligence Scale for Children. NY:Psychological Corporation.Wechsler, D. (1955). Manual for the Adult Intelligence Scale. NY: PsychologicalCorporation.Wechsler, D. (1967). Manual for the Wechsler Preschool and Primary Scale ofIntelligence. NY: Psychological Corporation.Wechsler, D. (1974). Manual for the Wechsler Intelligence Scale for Children-Revised. NY: Psychological Corporation.Wechsler, D. (1981). Manual for the Wechsler Adult Intelligence Scale Revised.NY: Psychological Corporation.Wechsler, D. (1989). Manual for the Wechsler Preschool and Primary Scale ofIntelligence-Revised. San Antonio, TX: Psychological Corporation.Wickens, C.D. (1974). Temporal limits of human information processing.Psychological Bulletin, 81, 739-755.Wickens, C.D. (1984). Processing resources in attention. In R. Parasurman &Davies, D.R., Varieties of Attention. New York: Academic Press.Wiig, E.H., & Semel, E.M. (1976). Language Disabilities in Children andAdolescents. Columbus: C.E. Merrill./ 125Wiig, E.H., & Semel, E.M. (1980). Language Assessment and Intervention for theLearning Disabled. Columbus: C.E. Merrill.Wilkinson, R.T. (1961). Comparison of paced, unpaced, irregular and continuousdisplay in watchkeeping. Ergonomics, 4, 259-267.Williges, R.C. (1971). The role of payoffs and signal ratios on criterion changesduring a monitoring task. Human Factors, 13, 261-267.Woodcock, R.W. (1986). Woodcock Reading Mastery Tests- Revised. Circle Pines,MN: American Guidance Service.Yerkes, R.M. (1917). The Binet versus the point scale method of measuringintelligence. Journal of Applied Psychology, .1, 111-122.APPENDIX A: LEVI’ERS TO TEACHERS126I 127The University of British ColumbiaThe Education ClinicFaculty of Education2125 Main MallVancouver, B.C.V6T 1Z5March 7, 1988Dear Kindergarten Teacher:RE: Study of Standardized Tests and Kindergarten Teacher Ratings of Readiness.We are conducting a research study that examines measures of “schoolreadiness” in kindergarten students. Although we do have permission from the universityand the school district to begin this project, it really is your help and cooperation as ateacher that is most important in developing the project’s foundation.In this study we are interested in comparing children who display “high risk”for school learning with “normal” children. We will be sampling kindergarten studentsthroughout the district using a variety of standardized assessment measures. At thistime, however, we would appreciate a few minutes or your time to help us identifystudents who may possibly be appropriate candidates to be included in the project.Enclosed you will find copies of the School Learning Profile. Please notice thenormal curve on question number one. In particular, we would like to know whichstudents in your class, in your opinion, fail into the lowest 10% of the curve. Pleasecomplete a School Learning Profile for all the children in your class who you feel fallinto the lowest 10% category. Please feel free’ to xerox more copies if needed.As soon as possible, please send the completed profiles to John Carter, atCourtenay Elementary School. We need the completed profiles before we can begin thenext phase of the research project.Thank you for your assistance. We will be corresponding with you shortly as weplan to begin formal assessments immediately after spring break. However, we arefollowing strict time limitations and we need completed Student Learning Profiles beforewe can move further. If you have any questions or concerns, please do not hesitate tocall.Sincerely,Dr. Julianne Conry, Ph.D. John D. Carter, M.A.Dept. of Educational Psychology Special Counsellorand Special Education Courtenay Elementary SchoolUniversity of British Columbia School District No. 71Vancouver, B.C. Courtenay, B.C.Telephone: 228.5260 338-5396 (School)Message: 228-5384 338-0275 (Home)I 128The University of British ColumbiaThe Education ClinicFaculty of Education2125 Main MailVancouver, B.C.V6T 1Z5April 5, 1988Dear Kindergarten Teacher:RE: Study of Standardized Tests and Kindergarten Teacher Ratings of Readiness.Thank you for your assistance in helping us get the project started. We havereceived the Student Learning Profiles that you seat for those students judged as “highrisk” for learning difficulties. Enclosed you will find parent permission forms for thesestudents as well as those students in your class selected through systematic randomselection procedures from class lists.In this study we are comparing two groups of students. One group judged as“high risk” for learning difficulties and the other group of “normal” students. Thesecond group was selected through random sampling procedures. It is important,however, that we do not include any students in this group who have any knownneurological deficits, diagnosed learning disabilities, significant emotional problems, orE.S.L (English Second Language) students. We expect a range of students as a resultof random selection, but we need to exclude those students from the “normal” groupwho fall in the forementioned categories.We will begin testing students when we have received parent permission. Pleaseforward completed parent permission forms to John Carter at Courtenay ElementarySchool. If you or any of the parents have any questions or concerns, please do nothesitate to call.Again, thank you for your assistance, and we look forward to visiting yourclassroom.Sincerely,Dr. Julianne Conry, Ph.D. John D. Carter, M.A.Dept. of Educational Psychology Special Counsellorand Special Education Courtenay Elementary SchoolUniversity of British Columbia School District No. 71Vancouver, B.C. Courtenay, B.C.Telephone: 228.5260 338-5396 (School)Message: 228-5384 338-0275 (Home)APPENDIX B: STUDENT LEARNING PROFILE129/ 130Student’s NameSCHOOL LEARNING PROFILETeacher’s Name1. Cpjred to other children you have observed, please rate this child’s overallabiZii5 to learn school material.USE THE FIVE POINT SCALE BELOW;Circle one number between 1 and 5.,4’N%%11121 314 51lowest 10% lower 30%but notlowest 10%middle 40%2. Attention span and distractibility: In class, doesdirections?Rating:Poor Attention1 2 3this child have difficulty following4Very good attention53. Spoken language skills: Is this child able to articulate and speak clearly?Rating:Poor1 2 3 4Very good articulation54. Verbal sequences: Can this this child verbally describe a sequence of events?Rating:Poor Very good skillsupper 30%but nothighest 10%highest 10%1 2 3 4I 1315. Alphabet recitation: Can this child recite the alphabet?Rating:Poor2Very good6. Letter identification skills: Can this child correctly name uppercase letters shown inrandom order?None AU correct1 2 3 4 57. Number identification: Can this child correctly name numbers between one andtwenty shown in random order?Rating.None AU correct1 2 3 4 58. Printing. Can this child print his! her name correctly without reversals, deletions,additions, or misalignment.s?Rating:Poor Perfect printing1 2 3 4 59. Fine-motor skills: Can this child use scissors to cut paper correctly?Rating:Poor Very good1 2 3 4 510. Gross motor skills: Rate this child’s ability for movement in physical education andsports abilities.Rating:Poor Very good1 2 3 4 511. Social participation: Rate this child’s play behavior with the other children.Rating:Aggressive Not aggressive1 2 3 4 512. Colour discrimination: Can this child correctly identify by name primary coloursshown in random order?Rating:None All correct1 2 3 4 5I 13213. Please rate the behaviour of this child, with reference to your observations ofhim/her in the classroom, on the playground, or in other situations you have seen.To what degree does this child exhibit each behaviour below?Circle one number on the scale for each item.Never Rare OccasionaiFrequent ConstantA.Fidgets 0 1 2 3 4B. Difficulty staying seated 0 1 2 3 4C. Difficulty waiting turnin games or group activity 0 1 2 3 4D. Easily distracted 0 1 2 3 4E. Defiant & uncooperative 0 1 2 3 4F. Has temper tantrums(explosive & unpredictable) 0 1 2 3 4G. Has difficulty listening 0 1 2 3 4H. Has difficulty playing quietly 0 1 2 3 4I. Fails to finish thingsstarted (short attention span) 0 1 2 3 4J. Blurts out answers toquestions before they havebeen completed 0 1 2 3 4APPENDIX C: PARENT PERMISSION FORM133I 134The University of British ColumbiaThe Education ClinicFaculty of Education2125 Main MaUVancouver, B.C.V6T 1Z5April 28, 1988Dear ParentfGuardian:RE: Study of Standardized Tests and Kindergarten Teacher Ratings of Readiness.This is to request your permission to allow your child,________________________to participate in a research project which is planned for May, 1988, in School DistrictNo. 71 (Courtenay). This project has been approved by the schooL district.The purpose of the project is to compare several newly developed standardizedability tests and kindergarten teacher ratings of school readiness. In this way we maydetermine how effective the tests are in diagnosing learning difficulties early in a child’sschool career, so that appropriate intervention may occur. It is hoped that follow-upassesment will be possible in one to two years time. At this time, however, permissionfor your child’s participation is requested for the kindergarten screening.The tests we will be using involve some thinking tasks, problem solving,vocabulary and drawing skills. Based on past experience, it has been found thatchildren enjoy working with the test materials.Approximately 15% of kindergarten students in School District No. 71 (Courtenay)will be selected for testing which requires approximately two hours in total.The results of the project will be used for research purposes. Your child’s scoreson the tests, therefore, will not become part of your child’s school record. In the eventthat the tests do indicate some specific difficulties, we would contact you to ask forpermission to consult with your child’s school and discuss the possibilities of additionaltesting or recommendations for special education programs.We would be pleased to answer any questions you may have regarding theproject. Please contact us at either of the telephone numbers or addresses below.It is important to note that your child’s participation in this project is completelyvoluntary. If you decide that your child should not participate in the project, or wish towithdraw at any time, this decision will not affect your child’s progress or status inschool in any way.Please see the Parent Permission Form on the following page...J2I 135Page 2Parent Permission FormI do or do not (circle one) grant permission for my child to participate in thisproject, and I acknowlege receipt of a copy of this letter. I understand that my childwill be tested by a qualified examiner in the child’s school, and that my child’s teachermay be asked to complete some brief rating forms about him/her. I also understandthat my child’s individual results will be kept strictly confidential.I am this child’s parent or legal guardian and I am completing this form on thechild’s behalf.Name: (please print)_________________________________________Signature:_________Relationship to child:Address:______Telephone:____________________________** ***s **** **** **** **** **** **** **** **** **** **** **** **** **** **** **** **If you consent for your child to participate, please complete the following confidentialbackground information.Child’s full name:______________________________Sex:_____Child’s Age:_____ Birthdate: Year Month_____DayParent’s education (please check one in each column):Years of Education Completed Mother or Father orFemale Guardian Male GuardianUp to Grade 8Grade 9 to 11High school diploma or equivalent1 3 years of college or technical schoolFour years oC university or moreDoes the child live with this person?/ 136Mother or female guardian’s occupation (please be specific):Father or male guardian’s occupation (please be specific):Page 3Check the one category that best describes each parent’sMother orFemale GuardianManagerial, professionalTechnical, sales, administrative supportServiceFarming, forestry, fishingPrecision production, craft, repairOperator, fabrication, labourerHomemakerNot currently in labour forceoccupation:Father orMale GuardianIs your child bilingual? Yes No —If yes, what other language does your child speak besides English?What language is spoken most of the time in your home?Your child’s race and ethnicity (please check one):Asian — Black — Native Indian WhiteOther (please specify)______________________Name of child’s school:____________________________Morning or Afternoon class (circle one)Teacher’s Name:__________ ___Thank you for your consideration.Sincerely,Dr. Julianne Conry, Ph.D.Dept. of Educational Psychologyand Special EducationUniversity of British ColumbiaVancouver, B.C.Telephone: 228-5260Message: 228-5384John D. Carter, M.A.Special CounsellorCourtenay Elementary SchoolSchool District No. 71Courtenay, B.C.338-5396 (School)338-0275 (Home)Hispanic-White—Hispanic-BlackAPPENDIX D: INTERCORRELATION TABLES137I 138At-Risk Group WPPSI-R Subtest Intercorrelationson the Performance Scale.OBJ GEO BLOCK MAZE PlC APEG PIQ VIQFSIQ .764* .638* .756* .698* .590* .491* .938* .903*OBJ —— .389* .581* .363 •457* •444* .760* .630*GEO —— •473* •477*.303 •457* 747* .404*BLOCK.351 .303 .319 .756* .660*MAZE—— .520* .280 .724* .542*PlC—— .341 .701* .363*APEG—— .512* .404**p<.05Note: N = 29I 139Normal Group WPPSI-R Subtest Intercorrelationson the Performance Scale.OBJ GEO BLOCK MAZE PlC APEG PIQ VIQFSIQ .432* .651* .656* .142 .705* •597* .822* .868*OBJ —— .432* .264 .057 .193 .351 .501* .246GEO .368* •373* .422* .386* .763* •377*BLOCK—— .130 .438* 539* .700* .427*MAZE—— .299 .181 533* .242PlC—— .488* .738* .481*APEG—— .620* .401**p<.05Note: N = 29/ 140At-Risk Group WPPSI-R Subtest Intercorrelationson the Verbal Scale.INF COMP ARITH VOCAB SIM SENT PIQ VIQFSIQ .679* .560* .701* .722* .564* .515* 939* .903*INFORMATION .749* .465* .591* •379* .423* .432* .521*COMPREHENSION —— .338 .332 .057 •433* .472* .587*ARITHMETIC —— .525* .426* .312 .560* .768*VOCABULARY —— .719* 379* .590* .858*SIMILARTITIES—— .187 .382* .691*SENTENCES—— .432* .521**p<.05Note: N = 29I 141Normal Group WPPSI-R Subtest Intercorrelationson the Verbal Scale.INF COMP ARITH VOCAB SIM SENT PIQ VIQFSIQ .706* •599* .811* .705* .520* .719* .823* .868*INFORMATION .562* .624* .514* .608* •599*.150 .851*COMPREHENSION —— .377* .352 .164 .558* .360 .625*ARITHMETIC —— .760* .546* .701* .505* .855*VOCABULARY—— .550* .628* .367* 793*SIMILARTITIES —— •475*.108 .742*SENTENCES —— 457*.748**p<.05Note: N = 29/ 142At-Risk Group SB:FE Subtest Intercorrelations1 Vocab --2 Comp .2523 Absurd .2744 PatternAnalys .1185 Copy .2946 QuantSubtes7 Bead8 Sent1 2 3 4 5 6 7 8Note: N = 291 Vocab = Vocabulary2 Comp = Comprehension3 Absurd = Absurdities4 Pattern = PatternAnalys Analysis5 Copy = Copyin96 Quant = Quantitative7 Bead = Bead Memory8 Sent = Sentence Memory.568*.354 .313.201 .343 .258.361 .317 .288 .397* .558*.481* .236 .105 .064 .212 .296.143 .244 .360 .400* .074 .298 .029*p<.05/ 143Normal Group SB:FE Subtest Intercorrelations1 2 3 4 5 6 7 81 Vocab2 Comp .504*3 Absurd .329 .507*4 PatternAnalys .407* •55Q* .391* ——5 copy .404* .517* .401* .3606 QuantSubtes .423* .382* 575* .584*7 Bead .239 .314 .104 .2688 Sent •449* .566* .414* .301Note: N = 291 Vocab = Vocabulary2 Comp = Comprehension3 Absurd = Absurdities4 Pattern = PatternAnalys Analysis.551*.451* .296.337 .308 .2715 Copy = Copyin96 Quant = Quantitative7 Bead = Bead Memory8 Sent = Sentence Memory*p<.05/ 144At-Risk Group SB:FE Area SAS Intercorrelations1 2 3 41 VerbalReasoning2 Abstract/VisualReasoning .487*3 QuantitativeReasoning •533* .761*4 Short-TermMemory •553* .284 .476* ——5 Composite •773* .837* .894* .673**p<05Note: N = 29/ 145Normal Group SB:FE Area SAS Intercorrelations1 2 3 41 VerbalReasoning2 Abstract/VisualReasoning .662*3 QuantitativeReasoning .556* .676*4 Short-TermMemory .572* .421* .371* ——5 Composite .838* .849* .825* .725**p<.05Note: N = 29APPENDIX E: WPPSI-R AND SB:FE CORRELATIONS146I 147At-Risk Group Verbal WPPSI-R and SB:FECorrelations.WPPSI-R Verbal SubtestsINF COMP ARITH VOCAB SIM SENT PIQ VIQ FSIQSB:FECompos .549* .446* .624* .467* •434* .561* .656* .673* .725*SB: FEVerbal .609* .506* .427* .584* .531* .471* .525* .723* .675*Vocab •534* .114 .368* .562* .517* .487* .225 .561* .414*Comp .365 .696* .277 .369* .161 .451* .423* •5Q4* .502*Absurd .446* .518* .338 •37Q* .391* .161 .478* .548* .563*AbstractVisual .391* .385* .320 .279 .323 .471* .607* .462* .592*PaternAnalys .191 .236 .200 .183 .136 .140 .486* .255 .421*Copy .501* .424* .346 .282 .453* .294 .483* .551* •555*QuantArea .428* .257 •545* 347 .307 .515* .518* .525* .568*QuantSubtes .255 .291 •435* .265 •395* .381* 354 •453* •433*MemoryArea .245 .234 •459* .297 .212 .567* .367* .423* .428*Bead •434* .210 .277 .531* •373* .502* .324 •494* .441*Sent .011 .115 .146 .107 .081 .277 .184 .094 .154*p<.05 N = 29I 148Normal Group Verbal WPPSI-R and SB:FECorrelations.WPPSI-R Verbal SubtestsINF COMP ARITH VOCAB SIM SENT PIQ VIQ FSIQSB:FECompos •74Q* .413* .704* .761* .588* .582* .404* .803* .730*SB: FEVerbal .043 •443* .565* .581* •453* .521* .206 .715* .564*Vocab .093 .292 .411* .492* .248 .628* .052 •495* .286Comp .716* .293 .516* .517* •44Q* .583* .274 .611* .530*Absurd .586* .488* •433* .438* .383* .398* .280 .604* •537*AbstractVisual .653* .438* .488* .568* .407* .512* •399* .632* .622*PatternAnalys .628* .349 .445* .446* .409* .392* .482* .579* .633*Copy .259 .350 .287 .497* .267 .378* .193 .396* .363QuantArea .124 .018 .172 .139 .140 .582* .404* .802* •73Q*QuantSubtes •453* .458* .600* .727* .410* •439* .511* .653* .697*MemoryArea .572* .020 .629* .580* .621* •444* .180 .617* .483*Bead .293 .078 .366* .329 .407* .031 .124 .326 .275Sent .602* .158 .676* .618* •533* .682* .235 .672* .548**p(.05 N = 29/ 149At-Risk Group Performance WPPSI-R and SB:FECorrelations.SB:FEVerbal .511* .193 .541* •439* .270Vocab .414* .064 .212 .184 .032Comp •435* .105 .484* .381* .168Absurd .432* .128 .529* •475* .271AbstractVisual .343 .512* .695* .458* .203PatternAnalys .286 .387* .672* .232 .182Copy .250 .432* .390* .571* .133QuantArea .381* .386* .421* .504* .171QuantSubtes .267 .225 .309 .424* .087MemoryArea .245 .194 .234 .430* .254Bead .362 .209 .179 .329 .145Sent .011 .067 .145 .260 .183.396* .607* .463* .592*.116 .486* .255 .421*.133 .484* .551* 555*.261 .518* .524* .568*.238 .354 •453* •433*.307 .367* .423* .428*.139 .324 •494* .441*.281 .184 .093 .154WPPSI-R Performance SubtestsOBJ GEOM BLOCK MAZES PlC PEGS PIQSB:FECompos .496* .428* .624* .611* .270VIQ FSIQ.332 .656* .673* .725*.396* .525* .723* .674*.271 .225 .561* .414*.346 .424* .504* .502*.247 .478* .584* .563**p<.05 N = 29I 150Normal Group Performance WPPSI—R and SB:FECorrelations.WPPSI-R Performance SubtestsOBJ GEOM BLOCK MAZE PlC PEGS PIQ VIQ FSIQSB:FECompos .069 .387* .342 .111 .498* .336 .404 .802* .730*SB:FEVerbal .043 .193 .315 .328 .324 .279 .206 .715* .564*Vocab .093 .088 .221 .438 .138 .003 .093 •495* .286Comp .078 .356* •373* .188 .241 .339 .274 .611* .530*Absurd .073 .208 .173 .093 .410* •397* .280 .603* 537*AbstractVisual .018 .452* .329 .087 .424* .305 •399* .623* .622*PatternAnalys .072 .465* .402* .012 .503* .359 .483* .579* .633*copy .045 .154 .183 .037 .263 .317 .193 .396* .363QuantArea .097 .165 .004 .109 .133 .336 .044 .071 .012QuantSubtes .192 .392* •375* .119 .468* .365 .510* .652* .697*MemoryArea .029 .208 .104 .107 .392* .163 .180 .617* .482*Bead .171 .076 .078 .065 .338 .241 .124 .326 .275Sent .133 .253 .145 .135 .349 .138 .235 .627* .548**p<.05 N = 29


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items