Part III Theory and Practice Interdependent In his search for what Margaret Rawson calls "the constructive reconciliation of opposites," John Dewey postulated the existence of continua at the ends of which are located seeming opposites (freedom and authority, theory and practice, mind and body). Dewey saw these, not as mutually exclusive concepts, but as points of variation, which themselves, contain elements of their apparent opposites. Thus, Dewey denies the existence of absolutes, claiming that freedom, for example, could not exist in the absence of all authority. For Annals readers, Dewey's belief that neither theory nor practice can survive in isolation, that each achieves its highest fulfillment only when it incorporates aspects of its opposite, has special significance. This journal serves as a major arena for the researchers in dyslexia, while also offering practical material for field workers. The research studies in this section have many implications for teachers and clinicians. We encourage the educators among our readers to make use of these studies as they work to raise the quality of education for all children. A growing body of data indicates a strong correlation between language development in the preschool years and later academic performance. School districts now routinely provide systematic preschool screening for children age three and older and increasingly districts are offering programs for children considered at risk for academic problems. Three papers in this section focus on early childhood education: "Screening for Reading Problems: The Utility of SEARCH" by Delmont Morrison, Panayota Mantzicopoulos, and Elizabeth Stone; "A ClassroomBased Model of Language Intervention for Preschool Language-Impaired Children: Principles and Procedures" by Diane Paul-Brown; and "Phoneme Segmentation Training: Effect on Reading Readiness" by Eileen W. Ball and Benita A. Blachman. Morrison and his colleagues conducted a study to determine the level of False Positives and False Negatives among 1,309 kindergarten children screened with SEARCH, an instrument used to identify children at risk for reading problems. Their procedures and findings regarding the validity of screening predictions should be reviewed by those selecting and using preschool screening batteries. t79
180
THE INTERDEPENDENCE OF THEORY AND PRACTICE
Paul-Brown reviews the literature which links oral language development in early childhood and later achievement in reading, writing, and spelling. She presents six language development principles used as the foundation for an intervention program for preschoolers with deficient language skills and describes the program and its results. Ball and Blachman provide convincing evidence that instructing kindergarten children in phoneme segmentation has a positive effect on later reading performance. The authors review several major studies which support this conclusion before describing their training project, conducted in the Syracuse Public Schools in 1987. Ball and Blachman give examples of segmentation lessons which can be presented by kindergarten teachers. Teachers of older children will be interested in papers by Aylett R. Cox and Lenox Hutcheson, Robert Calfee and Marilyn Chambliss, and Marcia K. Henry. Cox and Hutcheson consider syllable division in "Syllable Division: Prerequisite to Dyslexics" Literacy." They report on a tenyear study of more than 1,000 dyslexics who were taught how to tackle multisyUabic words of the type often encountered in text books at the upper elementary and secondary levels. In "Beyond Decoding: Pictures of Expository Prose." Calfee and Chambliss examine the reading problems presented by exposition, particularly in content-area textbooks. They suggest ways to empower students to "decode" such texts, using prereading discussion on the text topic, reading guided by that discussion, and follow-up activities based on the material read. "Beyond Phonics: Integrated Decoding and Spelling Instruction Based on Word Origin and Structure" describes Henry's efforts to make the structure of the language meaningful. Students learn to identify, compare, and contrast letter-sound correspondences, syllable patterns, and morpheme patterns and to call on the origins of words, roots, and affixes to enhance their reading and spelling skills. "Dyslexic/Learning Disabled Students at Dartmouth College" by Nancy W. Pompian and Carl P. Thum provides data about the slowly emerging awareness of the dyslexic college student and describes the accommodations and programs available. The authors surveyed 20 selective colleges in 1986 and 41 in 1988 to ascertain the level of sensitivity and opportunity. They report some progress over the two years, but note that real progress will be evident when colleges and universities implement a series of actions derived from the intent of Section 504 of the Rehabilitation Act of 1973. From preschool to college and beyond, implicit in the papers in this section is the interdependence of theory and practice, a key to the quality of education too often misplaced by both academicians and practitioners.
Screening for Reading Problems: The Utility of SEARCH Delmont Morrison
University of California,San Francisco Panayota Mantzicopoulos
University of California,Berkeley Elizabeth Stone
CHILDCenter, Kentfield,California
To evaluate the accuracy of SEARCH as a screen for identifying children at risk for developing learning problems, 1107 kindergarten children were evaluated with SEARCH and 284 (26 percent) were classified as at risk. At-risk children were of average intelligence and SEARCH scores were significantly correlated with sequential and simultaneous information processing skills. Children whose group preacademic achievement scores were at or below the third stanine at the end of kindergarten were classified as having inadequate skills for learning how to read; those scoring above the third stanine formed the adequate skills group. Using this criterion, SEARCH predicted children who had adequate or inadequate preacademic reading skills with 77 percent accuracy. However, approximately half of the children identified as at risk by SEARCH performed adequately. A reading test was individually administered to a group of at-risk children at the end of first (N = 49) and second (N =35) grade. Of those children identified at risk by SEARCH at kindergarten, 39 percent at first grade and 66 percent at second grade performed at grade level. Fewer children from the upper SES were identified by SEARCH as being at risk.
Introduction The early identification of kindergarten children w h o are at risk for developing learning disabilities requires a procedure that m u s t have Annals of Dyslexia,Vol. 38, 1988. Copyright ©1988by The Orton DyslexiaSociety ISSN 0474-7534 181
182
THE INTERDEPENDENCE OF THEORY AND PRACTICE
certain basic features. Because a large number of children have to be screened, the procedure should not take a great deal of time. Although this means that the sample of behavior evaluated is limited, the screening should be accurate in its prediction. This is particularly true in two categories of predicted performance: False Negatives and False Positives (Mercer, Algozzine and Trifiletti 1979). In the False Negative category, the prediction indicated that there was no risk for learning problems, but the child later performed inadequately on a criterion measure. In the case of the False Positive category, the child was identified as being at risk and requiring extra intervention, but, in fact, would have performed adequately on the criterion measure without intervention. A child falling in the False Negative category needed intervention but failed to receive it because of the inaccuracy of the screening procedure. The child falling in the False Positive category received unneeded intervention, and resources were used needlessly. There is considerable research indicating that many screening procedures have significant error in terms of predicting performance in either the False Negative or False Positive categories (Lindsay and Wedell, 1982; Mercer, Algozzine, and Trifiletti 1979). The exception to this is SEARCH which appears to make accurate predictions regarding children w h o need or do not need intervention (Silver and Hagin 1981). The present study represents a systematic replication of earlier research with SEARCH demonstrating its concurrent validity and False Positive prediction errors. A frequent assumption in early identification is that the screening instrument will identify individual children who lack the basic preacademic abilities necessary to acquire cognitive skills such as reading. There are major complications in this assumption. Screening instruments are typically used in kindergarten or early first grade and evaluate abilities that are assumed to be related to the yet-to-be-developed academic skill. The behavior measured by the procedure is not the same behavior that will be evaluated on the criterion measure (Lindsay and Wedel11982). The occurrence of various preacademic perceptual and cognitive abilities may correlate with later academic performance (Horn and Packard 1985), but the causal link with the acquisition of academic skill may not be strong enough to predict failure in acquisition except in extreme cases (Aldrich and Holliday 1971). Another major complication in predicting later performance from an earlier evaluation of kindergarten-aged children is the growing evidence for the pronounced variation in the development of perceptual and cognitive skills in this age range (Hindley and Owen 1978; McCall, Applebaum, and Hogarty 1973). All screening instruments must employ some method for establishing cut-off scores to determine if an individual child's score is in the at-risk range. This procedure implies that the child must have a "critical level" of the preacademic ability, that the scanning instrument itself is reliable, and that the variation in the child's test performance is clinically significant. Recent research with a well standardized test, the Kaufman ABC (Kaufman and Kaufman 1983), has es-
THE UTILITY OF S E A R C H
183
tablished that a normal sample of children between the ages of five and seven demonstrates considerable variation in cognitive and perceptual skills that reach statistical significance but may not be an indication of a child's strengths or weaknesses in the acquisition of information and academic performance (Chatman, Reynolds, and Willson 1984). SEARCH is based on the concept that antecedents for learning failure may be found in delay in the acquisition of spatial and temporal information. These delays are evaluated on SEARCH by ten subtests that assess a child's visual and auditory performance, body image, and a range of perceptual modalities (Silver and Hagin 1981). For each of the component subtests, a cut-off score, set empirically at the lowest one-third of the distribution of scores for each subtest, is determined. A child who scores above this point is not considered at risk on that particular test and if a child scores below that point on five or more of the subtests he/she is considered at risk. These scoring procedures result in approximately 25 percent to 30 percent of the samples that have been evaluated by SEARCH being classified at risk. SEARCH appears not to miss children who are really at risk, or to classify children as being at risk when they are not, at a rate that is better than most other procedures (Silver, Hagin, and Beecher 1978). For example, when using the criterion of reading at grade level, SEARCH on the average, has a False Negative prediction error rate of 8 percent at the end of first grade, and 4 percent at the end of second and third grade. The data for the False Positive prediction rate is equally impressive. Using the same criterion of reading at grade level, SEARCH typically predicts that a child will not be at grade level at the end of first grade in the range of 87 percent to 100 percent accuracy. A review of the predictionperformance accuracy of screening procedures suggests that SEARCH is significantly more accurate than most available single-instrument procedures (Mercer, Algozzine, and Trifiletti, Table I, 1979). There are two interrelated major issues that must be addressed before the prediction-performance accuracy of SEARCH is clearly established. The first issue is the possible influence of socioeconomic status (SES) on the accuracy of SEARCH. There are no published studies investigating the relationship between a child's SES and his/her performance on SEARCH. There is evidence that there are relationships between SES and poor perceptual-motor performance. For example, comparisons between Caucasian children from lower versus upper SES demonstrated poorer intellectual function in the low SES group, but also impaired performance on tests of visual perception (Amante 1975). Since SEARCH is designed to evaluate perceptual delays, the research by Amante (1975) suggests that the procedure may have a low False Positive rate with lower SES samples but a higher False Negative rate with upper SES samples. The second issue relates to the evidence that the four discrete perceptual areas thought to underlie SEARCH (Silver and Hagin 1981) may not be valid. In a recent factorial study of SEARCH, principal component analyses supported a two-factor solution with orthogonal factors entitled: Ver-
184
THE INTERDEPENDENCE OF THEORY AND PRACTICE
bal/Language and Visual Perception (Hinshaw et al. 1986). The Visual Perception factor consisted of only three subtests: Lamb Chop Matching, Lamb Chop Recall, and Designs, while the Verbal/Language factor was formed by five subtests. The Finger Schema subtest was not consistently loaded on either factor, and the Pencil Grip test was not entered in the analysis. Preliminary analysis of follow-up achievement data revealed superior criterion-related validity for the Verbal/Language factor. This study suggests that SEARCH may actually be more of a screening procedure for language delays, rather than perceptual delays. Language ability and fluency is known to be influenced by SES and cultural variables, with lower SES children typically doing less well (Labov 1970). This would again suggest that SEARCH may have a low False Positive rate with lower SES samples, but a higher False Negative rate w h e n used with higher SES samples. The purpose of the present research was to investigate the error rate of SEARCH at the end of kindergarten for potential False Negative and False Positive categories, as well as the long-term rate in the False Positive category in a sample of children representing the full range of SES.
Method Subjects Four school districts in Marin County, California, with a total kindergarten enrollment of 1,309 children, participated in this study. The parents of 1,107 (84 percent) agreed to participate. SEARCH was individually administered to this sample in February, 1984 (n=444), and 1985 (n = 663). In these two cohorts, 284 (25.7 percent) were designated at risk on the basis of scores below the 33rd percentile (local norms) on at least four of the nine valid SEARCH subtests (Silver and Hagin 1981). In order to select a control group of non-at-risk children for the study of the effects of SES, the number, gender ratio, and racial status of the atrisk children in each of the classrooms were noted. An equivalent number of non-at-risk children from each classroom were randomly selected, subject to the constraint that the groups had the same ratio of boys to girls and of Caucasian to non-Caucasian children as the at-risk sample within that classroom. The control group consisted of 299 children. A slightly higher number of control children were selected in order to allow for anticipated attrition. These procedures produced a total sample of 583 at-risk and control children. Because of children being retained or moving, the analysis of SES is based on a sample of 385 (82 percent) children on w h o m achievement test scores and SES were available. Of these children there were 45 (12 percent) in Class I, professional; 75 (19 percent in Class II, upper middle; 134 (35 percent) in Class III, middle; 113 (29 percent) in Class IV, lower middle; and 30 (5 percent) in Class V, unskilled labor and unemployed. This sample had a mean age of 68.9 months (5.7 years) with a range of 6380 months. There were 205 boys (53.2 percent) and 180 girls (46.8 per-
Tn~ UTILITYOFSEARCH
185
cent). The racial composition of the sample was as follows: 307 (79.7 percent) Caucasian, 11 (2.9 percent) Black, 25 (6.5 percent) Hispanic, 23 (6 percent) Asian, and 19 (4.9 percent) Other. Measures
SEARCH. This screening instrument is designed for children between 63 and 80 months of age (Silver and Hagin 1981). The individual administration takes approximately 20 minutes and is to be given prior to the entry into first grade. There are ten subtests that measure visual discrimination, immediate visual recall, visual-motor copying, rote sequencing, auditory discrimination, articulation, associating sounds with their visual symbols, directional orientation, finger schema, and pencil grip. Local norms for SEARCH were established and the procedures for administration and scoring recommended by Silver and Hagin (1981) were followed. However, 48 percent of the children in the first cohort demonstrated an appropriate pencil grip. This is in contrast to the 75-79 percent found in the Silver and Hagin (1981) sample and suggests that in the Marin County kindergarten population the lack of an appropriate pencil grip does not discriminate between vulnerable and nonvulnerable children. The pencil grip subtest data were not used in the data analysis and the test was not given to the second cohort. SEARCH was scored by the number of subtests passed with scores of five or less indicating at-risk performance. This means that at-risk children scored at or below the vulnerable level on four or more of the nine subtests. Kaufman Assessment Battery for Children (K-ABC). The K-ABC was administered to the at-risk children in both cohorts in March. This test was selected for use in this research because: (1) it is appropriate for children from ages 21/2 to 12; (2) it provides a quantitative measure of the cognitive level of this sample of children; (3) it emphasizes simultaneous and sequential information processing skills that have been shown to be associated with the development of reading skills (Kaufman and Kaufman 1983). The K-ABC yields a Mental Processing Composite scale (MPC) which is subdivided into Sequential and Simultaneous Processing scales. All scales provide a standard score with M = 100 and SD = 15 (Kaufman and Kaufman 1983). Socioeconomic Status (SES). SES in both the at-risk and control children was assessed using Hollingshead's (1957) Two-Factor Index of Social Position. Information on parent occupation and highest education level attained was weighed, summed, and converted to one of five socioeconomic status ranks. Academic Achievement Testing. Group academic achievement scores from the kindergarten year were used for the concurrent validity study of SEARCH, and the study of the relationship between SES and SEARCH. Data from the group academic achievement testing in kindergarten and first grade and individual academic testing during the first and second grades were used in the analysis of the False Positive error rate.
186
THE INTERDEPENDENCE OF THEORY AND PRACTICE
Group Academic Achievement Tests. The public schools routinely administer these tests in April-May of the school year. In this sample, kindergarten children, depending on their district, were tested with one of three tests: the Stanford Early School Achievement Test (SESAT), Comprehensive Test of Basic Skills (CTBS) and the California Achievement Test (CAT). From these tests, the following subtests were used as criteria of reading performance: Word Reading (SESAT): scores on this subtest provide assessment of the child's ability to identify a correctly spelled word and match it with its picture (Madden, Gardner, and Collins 1983). Total Reading (CTBS): scores on this subscale are derived from subtests of Visual Recognition, Sound Recognition, Oral Comprehension and Vocabulary (Comprehensive Test of Basic Skills 1982). Total Pre-Reading (CAT): scores on this subscale are derived from the following subtests: (a) Listening for Information, (b) Total Alphabet Skills, (c) Letter Sounds and (d) Total Visual and Auditory Discrimination (California Academic Achievement Test 1977). Individual Academic Achievement Test. Stanford Diagnostic Reading Test (SDRT). The Red Level of the SDRT was individually administered by a qualified examiner to the children in the study of the False Positive prediction rate in April and May of the first grade; the Green Level was administered in April and May of second grade. The examiner was blind in terms of the group membership of the child. The SDRT was chosen as a control for the possible effects of group or individual administration on the level of reading performance. The total reading comprehension score was used as the dependent variable on the data analysis and to establish a child's grade level (Karlsen, Madden, and Gardner 1976). Summarizing the sequence of evaluations: (1) the SEARCH screen was administered in February; (2) the K-ABC was administered in March and the Hollingshead Index of Social Position mailed to parents; (3) the academic achievement tests were administered in the months of April and May; (4) children in the False Positive prediction study were individually evaluated with the SDRT in April and May of the first and second grades.
Results Intelligence of M-Risk Children Standard Scores for MPC and the Sequential and Simultaneous Processing Scales were available on 216 at-risk children. The mean MPC Score for the at-risk children was 98.76 (SD = 11.16), with a mean Simultaneous Processing Score of 102.06 (SD =11.66) and mean Sequential Processing Score of 95.04 (SD = 11.97). The observed 7.02 point discrepancy between the Simultaneous and Sequential Processing Scores was not statistically significant. A correlation of .20 was established between reading level and
ThE UTILITY OF S E A R C H
187
both the Sequential and Simultaneous Processing Scores. SEARCH Scores correlated .42 with MPC, .38 with Sequential Processing Scores and .34 with the Simultaneous Processing Score. All correlations are Pearsons and significant at the
180
THE INTERDEPENDENCE OF THEORY AND PRACTICE
Paul-Brown reviews the literature which links oral language development in early childhood and later achievement in reading, writing, and spelling. She presents six language development principles used as the foundation for an intervention program for preschoolers with deficient language skills and describes the program and its results. Ball and Blachman provide convincing evidence that instructing kindergarten children in phoneme segmentation has a positive effect on later reading performance. The authors review several major studies which support this conclusion before describing their training project, conducted in the Syracuse Public Schools in 1987. Ball and Blachman give examples of segmentation lessons which can be presented by kindergarten teachers. Teachers of older children will be interested in papers by Aylett R. Cox and Lenox Hutcheson, Robert Calfee and Marilyn Chambliss, and Marcia K. Henry. Cox and Hutcheson consider syllable division in "Syllable Division: Prerequisite to Dyslexics" Literacy." They report on a tenyear study of more than 1,000 dyslexics who were taught how to tackle multisyUabic words of the type often encountered in text books at the upper elementary and secondary levels. In "Beyond Decoding: Pictures of Expository Prose." Calfee and Chambliss examine the reading problems presented by exposition, particularly in content-area textbooks. They suggest ways to empower students to "decode" such texts, using prereading discussion on the text topic, reading guided by that discussion, and follow-up activities based on the material read. "Beyond Phonics: Integrated Decoding and Spelling Instruction Based on Word Origin and Structure" describes Henry's efforts to make the structure of the language meaningful. Students learn to identify, compare, and contrast letter-sound correspondences, syllable patterns, and morpheme patterns and to call on the origins of words, roots, and affixes to enhance their reading and spelling skills. "Dyslexic/Learning Disabled Students at Dartmouth College" by Nancy W. Pompian and Carl P. Thum provides data about the slowly emerging awareness of the dyslexic college student and describes the accommodations and programs available. The authors surveyed 20 selective colleges in 1986 and 41 in 1988 to ascertain the level of sensitivity and opportunity. They report some progress over the two years, but note that real progress will be evident when colleges and universities implement a series of actions derived from the intent of Section 504 of the Rehabilitation Act of 1973. From preschool to college and beyond, implicit in the papers in this section is the interdependence of theory and practice, a key to the quality of education too often misplaced by both academicians and practitioners.
Screening for Reading Problems: The Utility of SEARCH Delmont Morrison
University of California,San Francisco Panayota Mantzicopoulos
University of California,Berkeley Elizabeth Stone
CHILDCenter, Kentfield,California
To evaluate the accuracy of SEARCH as a screen for identifying children at risk for developing learning problems, 1107 kindergarten children were evaluated with SEARCH and 284 (26 percent) were classified as at risk. At-risk children were of average intelligence and SEARCH scores were significantly correlated with sequential and simultaneous information processing skills. Children whose group preacademic achievement scores were at or below the third stanine at the end of kindergarten were classified as having inadequate skills for learning how to read; those scoring above the third stanine formed the adequate skills group. Using this criterion, SEARCH predicted children who had adequate or inadequate preacademic reading skills with 77 percent accuracy. However, approximately half of the children identified as at risk by SEARCH performed adequately. A reading test was individually administered to a group of at-risk children at the end of first (N = 49) and second (N =35) grade. Of those children identified at risk by SEARCH at kindergarten, 39 percent at first grade and 66 percent at second grade performed at grade level. Fewer children from the upper SES were identified by SEARCH as being at risk.
Introduction The early identification of kindergarten children w h o are at risk for developing learning disabilities requires a procedure that m u s t have Annals of Dyslexia,Vol. 38, 1988. Copyright ©1988by The Orton DyslexiaSociety ISSN 0474-7534 181
182
THE INTERDEPENDENCE OF THEORY AND PRACTICE
certain basic features. Because a large number of children have to be screened, the procedure should not take a great deal of time. Although this means that the sample of behavior evaluated is limited, the screening should be accurate in its prediction. This is particularly true in two categories of predicted performance: False Negatives and False Positives (Mercer, Algozzine and Trifiletti 1979). In the False Negative category, the prediction indicated that there was no risk for learning problems, but the child later performed inadequately on a criterion measure. In the case of the False Positive category, the child was identified as being at risk and requiring extra intervention, but, in fact, would have performed adequately on the criterion measure without intervention. A child falling in the False Negative category needed intervention but failed to receive it because of the inaccuracy of the screening procedure. The child falling in the False Positive category received unneeded intervention, and resources were used needlessly. There is considerable research indicating that many screening procedures have significant error in terms of predicting performance in either the False Negative or False Positive categories (Lindsay and Wedell, 1982; Mercer, Algozzine, and Trifiletti 1979). The exception to this is SEARCH which appears to make accurate predictions regarding children w h o need or do not need intervention (Silver and Hagin 1981). The present study represents a systematic replication of earlier research with SEARCH demonstrating its concurrent validity and False Positive prediction errors. A frequent assumption in early identification is that the screening instrument will identify individual children who lack the basic preacademic abilities necessary to acquire cognitive skills such as reading. There are major complications in this assumption. Screening instruments are typically used in kindergarten or early first grade and evaluate abilities that are assumed to be related to the yet-to-be-developed academic skill. The behavior measured by the procedure is not the same behavior that will be evaluated on the criterion measure (Lindsay and Wedel11982). The occurrence of various preacademic perceptual and cognitive abilities may correlate with later academic performance (Horn and Packard 1985), but the causal link with the acquisition of academic skill may not be strong enough to predict failure in acquisition except in extreme cases (Aldrich and Holliday 1971). Another major complication in predicting later performance from an earlier evaluation of kindergarten-aged children is the growing evidence for the pronounced variation in the development of perceptual and cognitive skills in this age range (Hindley and Owen 1978; McCall, Applebaum, and Hogarty 1973). All screening instruments must employ some method for establishing cut-off scores to determine if an individual child's score is in the at-risk range. This procedure implies that the child must have a "critical level" of the preacademic ability, that the scanning instrument itself is reliable, and that the variation in the child's test performance is clinically significant. Recent research with a well standardized test, the Kaufman ABC (Kaufman and Kaufman 1983), has es-
THE UTILITY OF S E A R C H
183
tablished that a normal sample of children between the ages of five and seven demonstrates considerable variation in cognitive and perceptual skills that reach statistical significance but may not be an indication of a child's strengths or weaknesses in the acquisition of information and academic performance (Chatman, Reynolds, and Willson 1984). SEARCH is based on the concept that antecedents for learning failure may be found in delay in the acquisition of spatial and temporal information. These delays are evaluated on SEARCH by ten subtests that assess a child's visual and auditory performance, body image, and a range of perceptual modalities (Silver and Hagin 1981). For each of the component subtests, a cut-off score, set empirically at the lowest one-third of the distribution of scores for each subtest, is determined. A child who scores above this point is not considered at risk on that particular test and if a child scores below that point on five or more of the subtests he/she is considered at risk. These scoring procedures result in approximately 25 percent to 30 percent of the samples that have been evaluated by SEARCH being classified at risk. SEARCH appears not to miss children who are really at risk, or to classify children as being at risk when they are not, at a rate that is better than most other procedures (Silver, Hagin, and Beecher 1978). For example, when using the criterion of reading at grade level, SEARCH on the average, has a False Negative prediction error rate of 8 percent at the end of first grade, and 4 percent at the end of second and third grade. The data for the False Positive prediction rate is equally impressive. Using the same criterion of reading at grade level, SEARCH typically predicts that a child will not be at grade level at the end of first grade in the range of 87 percent to 100 percent accuracy. A review of the predictionperformance accuracy of screening procedures suggests that SEARCH is significantly more accurate than most available single-instrument procedures (Mercer, Algozzine, and Trifiletti, Table I, 1979). There are two interrelated major issues that must be addressed before the prediction-performance accuracy of SEARCH is clearly established. The first issue is the possible influence of socioeconomic status (SES) on the accuracy of SEARCH. There are no published studies investigating the relationship between a child's SES and his/her performance on SEARCH. There is evidence that there are relationships between SES and poor perceptual-motor performance. For example, comparisons between Caucasian children from lower versus upper SES demonstrated poorer intellectual function in the low SES group, but also impaired performance on tests of visual perception (Amante 1975). Since SEARCH is designed to evaluate perceptual delays, the research by Amante (1975) suggests that the procedure may have a low False Positive rate with lower SES samples but a higher False Negative rate with upper SES samples. The second issue relates to the evidence that the four discrete perceptual areas thought to underlie SEARCH (Silver and Hagin 1981) may not be valid. In a recent factorial study of SEARCH, principal component analyses supported a two-factor solution with orthogonal factors entitled: Ver-
184
THE INTERDEPENDENCE OF THEORY AND PRACTICE
bal/Language and Visual Perception (Hinshaw et al. 1986). The Visual Perception factor consisted of only three subtests: Lamb Chop Matching, Lamb Chop Recall, and Designs, while the Verbal/Language factor was formed by five subtests. The Finger Schema subtest was not consistently loaded on either factor, and the Pencil Grip test was not entered in the analysis. Preliminary analysis of follow-up achievement data revealed superior criterion-related validity for the Verbal/Language factor. This study suggests that SEARCH may actually be more of a screening procedure for language delays, rather than perceptual delays. Language ability and fluency is known to be influenced by SES and cultural variables, with lower SES children typically doing less well (Labov 1970). This would again suggest that SEARCH may have a low False Positive rate with lower SES samples, but a higher False Negative rate w h e n used with higher SES samples. The purpose of the present research was to investigate the error rate of SEARCH at the end of kindergarten for potential False Negative and False Positive categories, as well as the long-term rate in the False Positive category in a sample of children representing the full range of SES.
Method Subjects Four school districts in Marin County, California, with a total kindergarten enrollment of 1,309 children, participated in this study. The parents of 1,107 (84 percent) agreed to participate. SEARCH was individually administered to this sample in February, 1984 (n=444), and 1985 (n = 663). In these two cohorts, 284 (25.7 percent) were designated at risk on the basis of scores below the 33rd percentile (local norms) on at least four of the nine valid SEARCH subtests (Silver and Hagin 1981). In order to select a control group of non-at-risk children for the study of the effects of SES, the number, gender ratio, and racial status of the atrisk children in each of the classrooms were noted. An equivalent number of non-at-risk children from each classroom were randomly selected, subject to the constraint that the groups had the same ratio of boys to girls and of Caucasian to non-Caucasian children as the at-risk sample within that classroom. The control group consisted of 299 children. A slightly higher number of control children were selected in order to allow for anticipated attrition. These procedures produced a total sample of 583 at-risk and control children. Because of children being retained or moving, the analysis of SES is based on a sample of 385 (82 percent) children on w h o m achievement test scores and SES were available. Of these children there were 45 (12 percent) in Class I, professional; 75 (19 percent in Class II, upper middle; 134 (35 percent) in Class III, middle; 113 (29 percent) in Class IV, lower middle; and 30 (5 percent) in Class V, unskilled labor and unemployed. This sample had a mean age of 68.9 months (5.7 years) with a range of 6380 months. There were 205 boys (53.2 percent) and 180 girls (46.8 per-
Tn~ UTILITYOFSEARCH
185
cent). The racial composition of the sample was as follows: 307 (79.7 percent) Caucasian, 11 (2.9 percent) Black, 25 (6.5 percent) Hispanic, 23 (6 percent) Asian, and 19 (4.9 percent) Other. Measures
SEARCH. This screening instrument is designed for children between 63 and 80 months of age (Silver and Hagin 1981). The individual administration takes approximately 20 minutes and is to be given prior to the entry into first grade. There are ten subtests that measure visual discrimination, immediate visual recall, visual-motor copying, rote sequencing, auditory discrimination, articulation, associating sounds with their visual symbols, directional orientation, finger schema, and pencil grip. Local norms for SEARCH were established and the procedures for administration and scoring recommended by Silver and Hagin (1981) were followed. However, 48 percent of the children in the first cohort demonstrated an appropriate pencil grip. This is in contrast to the 75-79 percent found in the Silver and Hagin (1981) sample and suggests that in the Marin County kindergarten population the lack of an appropriate pencil grip does not discriminate between vulnerable and nonvulnerable children. The pencil grip subtest data were not used in the data analysis and the test was not given to the second cohort. SEARCH was scored by the number of subtests passed with scores of five or less indicating at-risk performance. This means that at-risk children scored at or below the vulnerable level on four or more of the nine subtests. Kaufman Assessment Battery for Children (K-ABC). The K-ABC was administered to the at-risk children in both cohorts in March. This test was selected for use in this research because: (1) it is appropriate for children from ages 21/2 to 12; (2) it provides a quantitative measure of the cognitive level of this sample of children; (3) it emphasizes simultaneous and sequential information processing skills that have been shown to be associated with the development of reading skills (Kaufman and Kaufman 1983). The K-ABC yields a Mental Processing Composite scale (MPC) which is subdivided into Sequential and Simultaneous Processing scales. All scales provide a standard score with M = 100 and SD = 15 (Kaufman and Kaufman 1983). Socioeconomic Status (SES). SES in both the at-risk and control children was assessed using Hollingshead's (1957) Two-Factor Index of Social Position. Information on parent occupation and highest education level attained was weighed, summed, and converted to one of five socioeconomic status ranks. Academic Achievement Testing. Group academic achievement scores from the kindergarten year were used for the concurrent validity study of SEARCH, and the study of the relationship between SES and SEARCH. Data from the group academic achievement testing in kindergarten and first grade and individual academic testing during the first and second grades were used in the analysis of the False Positive error rate.
186
THE INTERDEPENDENCE OF THEORY AND PRACTICE
Group Academic Achievement Tests. The public schools routinely administer these tests in April-May of the school year. In this sample, kindergarten children, depending on their district, were tested with one of three tests: the Stanford Early School Achievement Test (SESAT), Comprehensive Test of Basic Skills (CTBS) and the California Achievement Test (CAT). From these tests, the following subtests were used as criteria of reading performance: Word Reading (SESAT): scores on this subtest provide assessment of the child's ability to identify a correctly spelled word and match it with its picture (Madden, Gardner, and Collins 1983). Total Reading (CTBS): scores on this subscale are derived from subtests of Visual Recognition, Sound Recognition, Oral Comprehension and Vocabulary (Comprehensive Test of Basic Skills 1982). Total Pre-Reading (CAT): scores on this subscale are derived from the following subtests: (a) Listening for Information, (b) Total Alphabet Skills, (c) Letter Sounds and (d) Total Visual and Auditory Discrimination (California Academic Achievement Test 1977). Individual Academic Achievement Test. Stanford Diagnostic Reading Test (SDRT). The Red Level of the SDRT was individually administered by a qualified examiner to the children in the study of the False Positive prediction rate in April and May of the first grade; the Green Level was administered in April and May of second grade. The examiner was blind in terms of the group membership of the child. The SDRT was chosen as a control for the possible effects of group or individual administration on the level of reading performance. The total reading comprehension score was used as the dependent variable on the data analysis and to establish a child's grade level (Karlsen, Madden, and Gardner 1976). Summarizing the sequence of evaluations: (1) the SEARCH screen was administered in February; (2) the K-ABC was administered in March and the Hollingshead Index of Social Position mailed to parents; (3) the academic achievement tests were administered in the months of April and May; (4) children in the False Positive prediction study were individually evaluated with the SDRT in April and May of the first and second grades.
Results Intelligence of M-Risk Children Standard Scores for MPC and the Sequential and Simultaneous Processing Scales were available on 216 at-risk children. The mean MPC Score for the at-risk children was 98.76 (SD = 11.16), with a mean Simultaneous Processing Score of 102.06 (SD =11.66) and mean Sequential Processing Score of 95.04 (SD = 11.97). The observed 7.02 point discrepancy between the Simultaneous and Sequential Processing Scores was not statistically significant. A correlation of .20 was established between reading level and
ThE UTILITY OF S E A R C H
187
both the Sequential and Simultaneous Processing Scores. SEARCH Scores correlated .42 with MPC, .38 with Sequential Processing Scores and .34 with the Simultaneous Processing Score. All correlations are Pearsons and significant at the
