Psycholinguistic correlations of academic achievement

Psycholinguistic correlations of academic achievement

Journal of School Psychology 1975 • Vol. 13, No. 3 PSYCHOLINGUISTIC CORRELATIONS OF ACADEMIC ACHIEVEMENT DONALD HAMMILL Austin, Texas RANDALL PARKER ...

340KB Sizes 4 Downloads 107 Views

Journal of School Psychology 1975 • Vol. 13, No. 3

PSYCHOLINGUISTIC CORRELATIONS OF ACADEMIC ACHIEVEMENT DONALD HAMMILL Austin, Texas RANDALL PARKER and PHYLLIS NEWCOMER The University of Texas at Austin

Summary: This research investigated the relationship of the ITPA subtests

to

measures of academic performance (i.e., the California Achievement Test). Subjects were 137 9-year old children similar to those used in the standardization sample of the 1968 revision of the ITPA. Two kinds of data analyses were undert a k e n - l , correlation coefficients among the variables were derived, and 2. the subjects were divided into low, average, and high groups based on their CAT performance, and analyses of covariance were mn to determine the significance of ITPA differences. The results failed to support the hypothesis that psycholinguistic abilities, except those which contribute to the Grammatic Closure subtest, are related to academic proficiency. Many school aged children who experience difficulty in mastering academic subject matter are referred for psychoeducational evaluation. Frequently they are found to exhibit varying patterns of psycholinguistic strengths and weaknesses, as evidenced primarily49y their performance on the Illinois Test of Psycholinguistic Abilities (ITPA). The questions which arise at this j~ancture are: Is the psycholinguistic deficit in some way causing or contrfbuting appreciably to the pupil's difficulties in reading, spelling, arithmetic, etc.? Or do his psycholinguistic and academic problems merely co-exist in him without affecting each other in any significant manner? Generally, educators have assumed that the answers to these questions are "yes" and " n o " respectively. As a result they have used information pertaining to psycholinguistic abilities as a basis for formulating educational prescriptions, apparently confident that remediation of these deficits will remove blocks to learning and lead to increased competencies in academic areas. The ITPA, in particular, has inspired many remedial programs to be used for this purpose. Many individuals have studied these questions empirically. They have, for the most part, used one of two investigative approaches. The first involves testing groups of children on psycholinguistic and academic variables and correlating the two in order to determine their predictive relationships, i.e., the extent to which scores on one test can be used to predict performance on the other tests. The second approach involves identifying groups of children who differ in academic competence (such as good and poor readers) and measuring the extent to which they differ in psycholinguistic abilities, thereby demonstrating the diagnostic power of ITPA. 248

Psycholinguistic Correlations o f Academic Achievement

249

Newcomer and Hammill (Note 1, Note 2) have recently reviewed both types of pertinent research. The conclusions drawn from 20 correlational studies were that among the 12 ITPA subtests, only Grammatic Closure is a statistically significant predictor of reading proficiency, and even that relationship is so flight that it is of little practical value to educators. Results of 19 studies of the ITPA's diagnostic value were similarly discouraging. None of the subtests consistently differentiated between groups of proficient and deficient readers. Although the results of that review appeared conclusive, one should not accept them too hastily for the following reasons: (1) only one of the studies performed b o t h a correlative and a group-difference analysis using the same set of subjects, (2) few of the studies controlled for the influence of intelligence and several failed to control for the effects of age, (3) many authors used too few subjects, thus reducing the confidence one might have in their findings, and (4) while the relationship to reading was repeatedly investigated, the situation regarding arithmetic, spelling, and English mechanics was relatively unexplored. Consequently, the authors decided to design a study of the psycholinguistic-academic relationship which carefully avoided these difficulties. PROCEDURES

Sample. The sample consisted of 137 children who were similar to those used in the standardization sample for the 1968 revision of the ITPA. The children belonged to the middle socioeconomic class and did not display behaviors suggestive of sensory, perceptual-motor, or social-emotional disturbance. They were selected from 12 fourth-grade classes in five schools located in Chester County, Pennsylvania, and ranged in age from 105 to 118 months (mean age: 112 months). The California Short Form Intelligence Test was used to measure the children's general mental ability; the IQs ranged from 87 to 111 (the mean was 97). Testing. Four subtests of the ITPA were used intact (Auditory and Visual Sequential Memory, Sound Bleding, Visual Reception); the remaining eight subtests were administered in shortened form, i.e., only the even-numbered items were scored. The reliability coefficients associated with the abbreviated subtests were determined by using either the Kuder-Richardson (#20), the Spearman-Brown split-half, or the Hoyt procedure, as appropriate, and were found to be approximately equal to, or greater than, those reported in the ITPA test manual, with the exception of that for Visual Association which was somewhat lower (Newcomer, Note 3). When using this test with fourthgrade pupils, it is necessary to demonstrate that ceiling effects are not present in the data. We inspected the raw score distributions for each subtest and found that they were more or less normal and that no more than the expected number of children reached the test's ceiling. For example, on Auditory Sequential Memory, only two of the 137 subjects scored in the 50s; the ceiling for this subtest is 56. Academic achievement was measured by the California Achievement Test (CAT), which yields information concerning reading, spelling, arithmetic, and language (grammar and expression). The ITPA and CAT were administered in April, May, September, or October, 1972.

250

Journal of School Psychology

Data Analysis. Two kinds of data analysis were deemed necessary to explore fully the relationships among the variables of interest in this study. First, a Pearson Product-Moment correlation matrix was computed with the effects of IQ partiaUed out using the formula suggested by Guilford (1956, p. 316). Second, the 137 children were divided into three groups, i.e., low, average, and high, based on their performance on each of the five academic variables sl~udied, and 60 one-way analyses of covariance were run using IQ as the single covariate. Planned orthogonal comparisons were employed to test the differences among means for the trichotomized groups. RESULTS AND DISCUSSION

Predictive Relationships. A correlation matrix comprised of coefficients depicting the relationships between the psycholinguistic and the achievement variables is presented in Table 1. Because of the comparatively large sample used in this study, coefficients o f . 17 or greater are statistically significant at beyond the .05 level of confidence. Of the 60 coefficients in the matrix, 45 are significant. In interpreting significant coefficients, Guilford (1956) suggests that r's which range from .3 to .8 represent "the level of validity coefficients usually found for useful predictive instruments in psychology and educational practice" (p. 378). On the other hand, Garrett (1954) suggests that only coefficients of .4 or above are useful, as lesser values denote negligible or at best slight relationships. For the purpose of this paper, .35 was used as the cut-off point between coefficients with practical predictive ability and those without. Using this criterion, only 10 of the 60 ~coefficients evidence predictive

Table 1 Correlation Coefficients between ITPA Subtests and Five Measures of Academic Achievementa Academic Abilities Psycholinguistic Abilities

Reading

Spelling

Arithmetic

Lan. Mech

Lan. Exp.

Aud. Rec. Vis. Rec. Verb. Exp. Man. Exp. Aud. Seq. Mem. Vis. Seq. Mem. Aud. Assoc. Vis. Assoc. Gram. Clos. Vis. Clos. Aud. Clos. Sd. Blend.

34 (16) 32 (06) 28 (09) 18 (12) 29 (21) 09 (-02) 45 (22) 25 (06) 57 (38) 34 (18) 22 (11) 29 (10)

23 (05) 15 (-11) 23 (07) 16 (10) 07 (-06) 12 (03) 34 (12) 27 (13) 38 (15) 16(-02) 03 (-11) 18 (00)

33 (16) 32 (10) 24 (07) 21 (16) 22 (13) 08 (-03) 43 (22) 23 (06) 51 (31) 32 (16) 05 (-10) 24 (06)

36(21) 32 (09) 20(00) 26(23) 16 (03) 12 (03) 37(13) 31 (16) 50(29) 31(16) 16 (05) 32(16)

25 (04) 22 (-07) 24 (05) 11 (03) 21 (09) 05 (-08) 39 (14) 25 (06) 55 (35) 25 (06) 16 (04) 33 (16)

aCoefficients with IQ partialled out appear in parentheses. All data are two-place decimals. N = 137, df = 135, P05 = .17.

Psycholinguistic Correlations o f Academic Achievement

251

usefulness. Of the 12 ITPA subtests, Auditory Association and Grammatic Closure alone seem to consistently and adequately predict academic achievement variables. However, when the influence of IQ is controlled for, the significant coefficients are reduced from 45 to nine, only two of w h i c h , those associated with Grammatic Closure and Reading (.38) and Grammatic Closure and Language Expression ( . 3 5 ) - c a n be considered of practical value in prediction. Diagnostic Relationships. In light of the findings which dealt with correlational relationships, one is not surprised to learn that after controlling for IQ few of the ITPA subtests discriminated among the groups of children who differed in five types of academic performance. Of the ITPA subtests, only Grammatic Closure, Auditory Reception, Auditory Association, Manual Expression, and Sound Blending demonstrated any diagnostic validity. Of all the subtests, Grammatic Closure evidenced by far the most significant discriminating power, distinguishing among the achievement groups on reading, arithmetic, and both language variables. Of the other subtests, Sound Blendhag discriminated significantly on both language variables; Auditory Reception and Manual Expression on the language-grammar variable;and Auditory Association on the arithmetic variable. None of the ITPA subtests discriminated among the low-middle-high groups on spelling. The remaining subtests, Visual Reception, Verbal Expression, Auditory Sequential Memory, Visual Sequential Memory, Visual Association, Visual Closure, and Auditory Closure, did not discriminate among the groups on any of the five academic measures. Of the 60 covariance analyses only nine were statistically significant at the .05 level of confidence and three of these could be chance findings (60 analyses times .05 level of confidence equals three possible chance values). The adjusted means and associated F ratios are presented in Table 2. Application of planned orthogonal multiple comparisons of the nine significant analyses revealed that one of the differences between the middle and low groups was not significant. Those children in this sample who were poor in language grammar performed the same on Sound Blending as those children who were average in that skill (i.e., the middle group). Since it is the ability to differentiate between poor and average performance which denotes the diagnostic value of a test, this result largely renders Sound Blending useless. All the other multiple comparisons were significant. Only Grammatic Closure, the most linguistic of the ITPA subtests, consistently predicted academic achievement in that it evidenced significant predictive and diagnostic relationships in this study. Actually, four of the nine significant findings reported in Table 2 pertain to Grammatic Closure and only Grammatic Closure correlated at a practically useful level with academic achievement after the effects of IQ were controlled for (Table 1). This finding is compatible with the conclusions of other investigations which we reviewed previously. Therefore, if there are chance findings in our study, they are likely to be associated with the other ITPA subtests. Before continuing, we feel obligated to reiterate that when speaking of psycholinguistic abilities, we mean the constructs measured by the ITPA (i.e., the operational definition for this paper). We do not generalize the findings to psycholinguistic abilities in general or to samples of children other than those

252

Journal of School Psychology Table 2 Adjusted Means and F Ratios on the ITPA Subjects for Groups Trichotomized on Each of Five Academic Ability Variables Academic Abilities Reading Aud. Ree. Vis. Rec. Verb. Exp. Man. Exp. Aud. Seq. Mem. Vis. Seq. Mem. Aud. Assoc. Vis. Assoc. Gram. Clos. Vis. Clos. Aud. Clos. Sd. Blend. Lan. Mech Aud. Rec. Vis. Rec. Verb. Exp. Man. Exp. Aud. Seq. Mem. Vis. Seq. Mem. Aud. Assoc. Vis. Assoc. Gram. Clos. Vis. Clos. Aud. Clos. Sd. Blend. Lan. Exp Aud. Rec. Vis. Rec. Verb. Exp. Man. Exp. Aud. Seq. Mem. Vis. Seq. Mem.

X1

Xm

~Ch

F

20.97 27.50 11.93 11.38 27.24 20.06 14.82 13.52 13.02 20.44 10.58 11.47

21.77 27.93 12.27 11.93 29.40 20.84 15.48 13.94 14.01 21.01 11.06 11.66

21.73 29.22 13.41 11.97 31.35 20.25 15.84 13.84 14.73 22.47 10.92 12.19

2.2 1.8 1.3 0.7 2.4 0.5 1.7 0.5 7.9** 1.9 0.8 0.6

20.75 27.70 12.51 10.89 28.76 20.44 14.90 13.22 13.20 20.78 10.80 11.50

21.67 28.37 I2.56 11.96 29.64 20.63 15.60 13.97 14.14 20.66. 10.69 11.19

21.91 28.56 12.54 12.26 29.50 20.11 15.56 14.02 14.29 j~2.31 11.05 12.52

4.0* 0.6 0.0 3.2* 0.2 0.2 1.3 1.8 t 4.5* 1.9 0.5 3.2*

21.42 28.75 11.88 11.72 28.38 20.92

21.37 27.82 12.13 11.61 29.43 20.18

21.71 28.22 13.73 11.98 30.23 20.04

0.3 0.8 2.2 0.2 0.4 0.4

who served as the sample for this study. Considering these limitations, t h e results of this investigation fail to support the hypothesis that psycholinguistic abilities, except those which contribute to the Grammatic Closure subtest, are related to academic proficiency. Therefore, the assumption that these particular abilities are essential to, or play a significant role in, attaining academic success seems questionable. Furthermore, the results permit one to speculate that-questions concerning the effects of psycholinguistic training for the purpose of improving academic performance are rendered moot, but of course this is a researchable topic. The conclusions emerging from this study assume added credence in that they are consistent with those obtained in the vast majority of other studies which have investigated this particular issue (Newcomer & Hammill, Note 1, Note 2).

Psycholinguistic Correlations o f Academic Achievement

253

Table 2

(Continuea/

Aud. Assoc. Vis. Assoc. Gram. Clos. Vis. Clos. Aud. Clos. Sd. Blend.

15.16 13.63 12.82 21.14 10.61 10.82

15.16 13.67 14.14 21.28 11.05 12.02

15.87 14.03 14.84 21.52 10.87 12.50

1.1 0.4 11.4"* 0.0 0.7 3.6*

21.58 28.31 12.16 11.87 30.77 20.65 15.11 13.43 13.54 21.77 11.17 11.90

21.74 28.84 12.92 11.55 28.52 19.44 15.80 13.80 14.16 21.07 10.66 11.61

21.11 27.53 12.54 11.86 28.65 21.12 15.22 14.09 14.08 21.06 10.72 11.81

1.4 1.8 0.5 0.3 1.2 2.3 1.4 1.0 1.6 0.4 1.1 0.2

20.97 27.98 12.19 11.40 28.47 20.80 14.60 13.43 13.03 20.57 11.18 11.59

21.69 28.49 12.53 11.85 28.73 20.44 15.49 14.01 14.17 21.89 10.67 11.73

21.77 28.23 12.90 12.01 30.89 19.88 16.02 13.81 14.51 21.36 10.74 12.01

1.9 0.3 0.3 0.5 1.1 0.4 3.4* 0.9 7.0** 1.2 1.0 0.2

Spelling Aud. Rec. Vis. Rec. Verb. Exp. Man. Exp. Aud. Seq. Mem. Vis. Seq. Mere. Aud. Assoc. Vis. Assoc. Gram. Clos. Vis. Clos. Aud. Clos. Sd. Blend.

Arithmetic Aud. Rec. Vis. Rec. Verb. Exp. Man. Exp. Aud. Seq. Mem. Vis. Seq. Mere. Aud. Assoc. Vis. Assoc. Gram. Clos. Vis. Clos. Aud. Clos. Sd. Blend.

1Assumption of homogeneity of regression was not met. *p < .05 **p < .01 Proponents of the psycholinguistic approach could point out with some justification that there " m i g h t " exist within any group of children who have both academic and psycholinguistic problems a subgroup for whom the psycholinguistic deficits are adversely affecting school performance. If true, this subgroup should have specific characteristics which would permit its identification. Research could then be designed to demonstrate that, on this subsample at least, there was a relationship between psycholinguistic and academic abilities. If this proved true, studies which probed the effects of systematic training on such subgroups would assume considerable importance. Until such demonstration is available, however, one must conclude that statements which presume a positive relationship between psycholinguistic and academic performance lack a conclusive data-base and are in fact incompatible with the greater part of the existing relevant research literature.

254

Journal of School Psychology

REFERENCES Garrett, H. E. Statistics in psychology and education. New York: Longmans Green, 1954. Guilford, J. P. Fundamental statistics in psychology and education. New York: McGrawHill, 1956.

Reference N o t e s 1. Newcomer, P., & Hammill, D. The relationship of the ITPA subtests to academic achievement: A review of pertinent research. The Reading Teacher, 1975, in press. 2. Newcomer, P., & Hammill, D. Psycholinguistics in the schools. Columbus, Ohio: Charles E. Merrill, 1975, in press. 3. Newcomer, P. The construct validity of the Illinois Test of Psycholinguistic Abilities. Unpublished doctoral dissertation, Temple University, 1973. Donald Hammill 1505 Sunny Vale #217 Austin, Texas 78741

Phyllis Newcomer Assistant Professor of Special Education Sutton Hall The University of Texas Austin, Texas 78712 Received: May 14, 1974 Revision Received: September 15, 1974

Randall Parker Associate Professor of Special Education Sutton Hall 327 The University of Texas Austin, Texas 78712