AUDITORY AND VISUAL PERCEPTION PROCESSES AND READING ABILITY: A QUANTITATIVE REANALYSIS AND HISTORICAL REINTERPRETATION.
Mann (1979) outlined the historical course of process concepts in education and described how the identification and training of processes has been an enduring theme. The seminal work of J. M. Itard and E. Seguin was based on training processes. Later, the influence of Gestalt psychology and principles of perceptual organization were significant influences on ideas about processes and their relation to learning (Henle, 1961).
The field of learning disabilities (LD) has historically been associated with process concepts, particularly the relationship between perceptual motor deficits and learning problems (see Hallahan & Cruickshank, 1973). Beginning with the work of K. Goldstein (1942) and continuing with the investigations of A. Strauss and H. Werner (see Strauss & Lehtinen, 1947), attention was focused on perceptual motor deficits as fundamental elements of learning disabilities. Additionally, more generalized theories relating perception and cognition were posited (e.g., Gibson, 1969; Hebb, 1949; Piaget, 1969).
After LD was established as a category of special education, a majority of its early theoretical conceptualizations postulated the presence of deficient perceptual processes as a primary deficit (see Kavale & Forness, 1995). The analyses offered by, for example, Ayres (1972), Getman (1965), Johnson and Myklebust (1967), Kephart (1960), and Wepman (1964), emphasized deficiencies in perception as the most prominent deficit associated with LD. The tenor of these ideas is illustrated by this statement from Frostig, Lefever, and Whittlesey (1966)
It is most important that a child's perceptual disabilities, if any exist, be discovered as early as possible. All research to date which has explored the child's general classroom behavior has confirmed the authors' original finding that kindergarten and first-grade children with visual perceptual disabilities are likely to be rated by their teachers as maladjusted in the classroom; not only do they frequently find academic learning difficult, but their ability to adjust to the social and emotional demands of classroom procedures is often impaired. (p. 6)
In practice, assumptions about the presence of process deficits meant that process training was a favored intervention and, although it did not prove to be efficacious (Kavale & Forness, 1987; Kavale & Mattson, 1983), theoretical questions remained about the nature of the relationship between processes and academic achievement (Torgesen, 1979). Of particular interest was the association between auditory and visual perception and reading achievement. Historically, perceptual skills were assumed to be important correlates of reading ability and, although training perceptual skills had limited effect on reading, their real association with reading remained open to question. The emerging LD field thus witnessed debate about one of its fundamental assumptions.
In attempting to resolve questions about the nature of the association between perceptual processes and reading ability, it was possible to find support for a positive relationship between auditory perception (e.g., Dykstra, 1966; McNinch, 1971; Sabatino, 1973), visual perception (e.g., Barrett, 1965; Frostig, 1972; Goins, 1958) and reading achievement as well as opposition to the assumption that auditory perception (Groff, 1975; Hammill & Larsen, 1974) and visual perception (Cohen, 1969; Larsen & Hammill, 1975) were important correlates of reading achievement. The variant interpretations were partially the result of difficulties in delineating facets of auditory and visual perception (Sabatino, 1979) and measuring perceptual skills (Coles, 1978), as well as methodological problems in the empirical literature (Samuels, 1973). Consequently, the nature of the relationship between perception and reading was subject to equivocal interpretation.
In an attempt to bring closure to the question of whether or not perceptual processes were significant correlates of LD, Kavale (1980, 1981, 1982) performed several meta-analyses. The goal was to compile and analyze correlation coefficients that were descriptive of the relationships. The findings from these meta-analyses suggested that auditory and visual perception were correlates of reading ability, but the relationships were complex and contingent upon the variables defining the association. Because auditory and visual perception were studied separately, questions about the contribution of both auditory and visual perceptual skills in predicting reading achievement were left unanswered. Since perceptual skills rarely operate in isolation, it is important to assess the nature of the association when both auditory and visual perception are considered in their relationship to reading ability. Thus, the purpose of this investigation was to reanalyze and reinterpret the findings from the previous meta-analyses to determine the relative importance of auditory and visual perception in predicting reading achievement.
Meta-analysis was proposed as a quantitative alternative to traditional means for summarizing research findings (Glass, 1976). The methods of meta-analysis have been well described (e.g., Glass, McGaw, & Smith, 1981; Hunter, Schmidt, & Jackson, 1982; Rosenthal, 1991) and have undergone a number of advances (e.g., Cooper & Hedges, 1993; Hunter & Schmidt, 1990; Hedges & Olkin, 1985). The process has become an accepted means for summarizing statistically a research domain to produce objective, verifiable, and replicable findings (e.g., Hunt, 1997; Kavale, 1984; Wachter & Straf, 1990).
The earlier meta-analyses used standard literature-search procedures to identify empirical investigations of the relationship of either auditory or visual perceptual skills to reading achievement. For the earlier metaanalyses, the search spanned the years 1950 to 1980, a time frame that permitted an evaluation of auditory and visual perception in their classic sense debated earlier in the literature. Since that time, perceptual research has deemphasized auditory and visual perception in the sense of them being particular modalities, instead examining processes in a more modular and cognitive framework (Vellutino, 1979). These more sophisticated analyses are exemplified in investigations of the role of phonological processing in reading ability (e.g., Bradley & Bryant, 1985; Stanovich, 1988; Wagner & Torgesen, 1987). Nevertheless, the relationship between perceptual factors and reading ability retains historic interest, and it would be useful to bring closure to a long-standing question. The goal of the search was to examine auditory and visual perceptual research as it was conceptualized in the history of LD.
A comprehensive search that included all available published and unpublished research investigating the relationship between auditory and visual perception and reading ability conducted during the specified time flame was undertaken (see Cooper, 1984). To achieve this goal, citations were obtained from empirical studies cited in previous reviews, from abstract searches (e.g., ERIC), from major journals, and from the references cited in the studies obtained through these search procedures. The search yielded 367 studies, but 100 were excluded primarily because they contained no usable data. The remaining 267 studies included 106 investigating auditory perception and 161 evaluating visual perception and reading.(1)
The next step in meta-analysis consists of describing quantitatively data from individual studies. The first stage of this process was to code information about variables and features included in individual studies in order to investigate relations among findings and study characteristics. The second stage involves calculating the basic effect size (ES) statistic common to metaanalysis. In the present case, the ES was represented by the Pearson product-moment correlation coefficient (r) descriptive of the magnitude of the relationship between auditory or visual perceptual skills and reading ability. When a study reported actual correlation coefficients, the process was straightforward but when other summary statistics (e.g., t, F, [chi square]) were reported to describe what was essentially a correlational finding, conversion to r was required. Guidelines for these conversions have been outlined (e.g., Glass et al., 1981; Hedges & Olkin, 1985; Wolf, 1986). They were required in 112 (44%) of the 267 studies to produce Pearson correlation coefficients.
To make the obtained correlation coefficients creditable (and conservative) estimates of the real magnitude of relationship between perceptual skills and reading ability, the rs were corrected for two potential sources of error: sampling and measurement (see Hunter et al., 1982, pp. 35-94). Measurement error produces variation among correlations because assessments differ in the extent to which they are affected by random error. Random measurement error produces an effect ("attenuation") that causes the correlation to be lower than it would have been with perfect measurement. The magnitude of measurement error is indicated by test reliability, and an individual correlation coefficient can be corrected for attenuation ([r.sub.c]).(2)
Sampling error is created when a mean correlation is obtained by averaging correlations across studies. With large samples, sampling error is minimal (i.e., if the mean correlation is the average across 50 studies with a total sample size of 5000, then sampling error is approximately the same as calculating a single correlation on a sample of 5000). The variance among correlations does not cancel out in summation, and it is likely that sampling error causes the variance across study correlations to be systematically larger than the variance found for population correlations. The sampling error variance can be calculated and then subtracted from the observed variance with the difference representing an estimate of the population correlation variance.(3) Besides amending individual rs, these corrections also make later analyses more robust.
The previous meta-analyses (i.e., Kavale, 1980, 1981, 1982) reported mean correlations (r), related descriptive statistics, and interpreted the magnitude of the relationship in terms of a coefficient of determination ([r.sup.2]) reflecting the percent of variance explained by the particular relationship, that is, the proportion of variance in reading ability that can be predicted with the particular perceptual skill.
Although this method of interpretation is the traditional way to discuss correlational analyses, it can lead to disagreement because predictions found to account for relatively small percentages of variance are often considered unimportant or not worth further study (Rosenthal & Rubin, 1979). For example, Gersten and Carnine (1984) responded to Kavale's (1982) metaanalysis by suggesting that the average correlations reported were "weak," primarily because they did not account for a substantial proportion of the variance in the relationship between auditory perceptual skills and reading achievement. Although Kavale (1985) demonstrated that the correlational data were not misinterpreted and were, in fact, useful, Gersten and Carnine (1985) were not convinced, suggesting that the situation surrounding interpretation was much like "two ships passing in the night" (p. 47).
The real dilemma may lie in the traditional interpretation of correlation coefficients, which renders them solely in the conceptually difficult "proportion of variance explained" notion. Rosenthal and Rubin (1979) suggested that the [r.sup.2] statistic may be misleading because it often results in a tendency to underestimate the importance of relationships because they are seen as associated with what are believed to be low values of [r.sup.2] (see Rosenthal & Rubin, 1979).
As an alternative, Rosenthal and Rubin (1982) suggested the "binomial effect size display" (BESD), which indicates the change in predictive accuracy attributable to the relationship in question and is computed from the formula .50 [+ or -] r/2. The BESD shows the extent to which prediction is enhanced (i.e., the percentage increase in correct decisions) with the use of, for example, perceptual variable X to forecast reading skill Y; relationships are made more intuitive and more informative by being rendered in real-world terms. Preece (1983) suggested that the median split (i.e., .50) may not be appropriate in all cases and provided formulas for determining where a distribution should be dichotomized that were used when necessary.
The BESD statistic addresses the question: What, if any, is the percentage increase in the number of correct predictions about reading ability with a particular measure of perceptual skill? Suppose some relationship reveals an r of .32; it is traditionally said to account for "only 10% of the variance in the relationship" but a BESD interpretation shows a relationship of this magnitude is equivalent to increasing the predictive accuracy about subsequent reading ability from 34% to 66% correct decisions, which conversely means reducing the number of incorrect predictions about reading ability from 66% to 34% (see Rosenthal & Rubin, 1982). The BESD interpretation thus possesses potential for adding a new perspective in trying to answer old questions about the relationship between perceptual skills and reading ability.
A total of 2,294 correlation coefficients were collected. From this total, 1,509 were descriptive of the relationship between auditory and visual perceptual skills and reading ability with 375 in the auditory domain, 1,017 in the visual domain, and 117 investigating auditory-visual integration. The remaining 568 correlations were descriptive of the relationship among IQ and perception, IQ and reading, and reading achievement subskills.
Across 267 studies, the average investigation included 190 subjects (total N = 50,000) in grade 3.3 (range = K-7) with an average age of 7.87 years (range = 4-11), and an average IQ of 104.5 (range = 88-129). The correlations ranged from .01 to .89 with a median of .369, suggesting a modest negative skew for the distribution. At the highest level of aggregation across all coefficients, the average correlation (r) was .355 with a standard deviation (SD) of .147 and a standard error (SE) of .006. Using the traditional interpretation, auditory and visual perception would be said to account for about 19% of the variance in reading achievement.
The 267 studies included two primary subject groups: students with average reading achievement (AA) and students with learning or reading disability (LD/RD). The average AA subject was in grade 3.1 with an average age of 7.21 years and an average IQ of 109.3. For the LD/RD group, the average subject was in grade 4.0 with an average age of 8.11 years and an average IQ of 99.6. No differences were found between groups for grade level, age, or IQ. Across all auditory and visual correlation coefficients, the average was .362 for the AA group and .378 for the LD/RD group. Using the traditional interpretation, auditory and visual perception accounts for about 13% of the variance in reading ability for AA subjects and 14% for LD/RD subjects. A comparison of the average correlations for AA and LD/RD groups showed no difference ([t.sub.(1157)] = 1.74, p [is less than] .10). Therefore, the analysis can proceed without any distinction between AA and LD/RD groups.
The Structure of Auditory and Visual Perception and Reading
The 267 studies investigated a variety of auditory and visual perceptual skills. Although difficulties exist with respect to precise definition of auditory and visual perception (see Sabatino, 1979), there is consensus about the fundamental nature of the major component processes (e.g., Bartley, 1969; Dember, 1960; Murch, 1973). Using consensus definitions, it was possible to identify common auditory and visual perceptual skills among the 267 studies.
For auditory perception, four skills were distinguished: (a) auditory discrimination (AD): ability to differentiate among auditorially presented stimuli; (b) auditory memory (AM): ability to recall a sequence of auditorially presented stimuli; (c) auditory blending (AB): ability to perceive separate auditory stimuli (phonemes) and to combine them into a whole unit (word); and (d) auditory comprehension (AC): ability to interpret and to understand auditorially presented material.
In the realm of visual perception, seven skills were distinguished: (a) visual discrimination (VD): ability to perceive dominant features in visual stimuli; (b) visual memory (VM): ability to recall a sequence of visually presented stimuli or ability to recall a dominant feature of a visual stimulus; (c) visual-motor integration (VMI): ability to integrate visual stimuli with body movements (i.e., copying); (d) visual closure (VC): ability to recognize a complete figure from fragmented visual stimuli; (e) visual association (VA): ability to relate conceptually visually presented stimuli; (f) visual spatial relationship (VS): ability to perceive position of objects in space; and (g) figure-ground discrimination (FG): ability to distinguish an object from irrelevant background visual stimuli.
The 267 studies also investigated a number of component reading skills. In almost all cases, reading was assessed with standardized tests, usually an achievement battery including reading subtests or an individual reading achievement test. Although possibly measured differently, little controversy surrounds the definition of reading skills. The four components common to investigations of auditory and visual perception included general reading ability (GR) (total reading score on a standardized test), word recognition (WR), reading comprehension (RC), and vocabulary (VO). The latter three reading skills were usually represented by scores on subtests.
To validate the proposed data analysis structure for investigating the relationship between auditory and visual perceptual skills and reading ability, a principal component solution was applied to a 16 x 16 correlation matrix where cell entries represent the average correlation for the interrelationships among perceptual and reading variables as well as IQ. The varimax rotated factor matrix revealed four factors. The first factor (I) was reading, the second factor (II) was auditory perception, the third (III) was a visual-cognitive factor, and the fourth (IV) was a visual differentiation factor; thus, visual perception did not emerge as a unitary phenomenon like auditory perception. Nevertheless, all variables were contributors and should be considered in analyzing the association between reading and perception.
Before correlations could be aggregated by individual perceptual skills, it was necessary to test the homogeneity of correlations across studies. Although the earlier correction for sampling error ensured equal variance in the total sample of correlations, the present test of homogeneity was used to determine whether smaller aggregations of correlations were reasonably consistent with the model of a single underlying population correlation; if they are not, it can be misleading to pool estimates into smaller aggregations. Hedges and Olkin (1985, pp. 235-236) provided a test of homogeneity (Q) based on Fisher's z-transformation and a critical value from the chi-square distribution.(4) Hypotheses about homogeneity of the population correlations were not rejected for either the four auditory perceptual skills (Q = 15.92), or the seven visual perceptual skills (Q = 26.87), suggesting that the correlations were homogeneous. Thus, the obtained rs can be aggregated by individual auditory and visual perceptual skills.
Auditory and Visual Perceptual Skills and Reading Achievement
The individual auditory and visual perceptual skills aggregated across reading skills are shown in Table 1. The display is in BESD format showing the increase in predictive accuracy associated with each perceptual skill in making decisions about reading achievement.
Table 1 Relationship Between Auditory and Visual Perceptual Skills and Reading Ability
Number of Mean Perceptual Correlation Correlation Skill Coefficients Coefficient Auditory Comprehension (AC) 26 .402 Memory (AM) 99 .383 Blending (AB) 67 .377 Discrimination (AD) 183 .371 Visual Memory (VM) 139 .472 Closure (VC) 77 .427 Discrimination (VD) 291 .385 Association (VA) 95 .377 Motor Integration (VMI) 305 .361 Spatial Relation (VS) 64 .326 Figure-Ground Discrimination (FG) 46 .251 Increase in Predictive Number of Perceptual Accuracy Correct Skill Increased Decisions From To Auditory Comprehension (AC) 30% 70% 40% Memory (AM) 29% 67% 38% Blending (AB) 33% 71% 38% Discrimination (AD) 31% 69% 38% Visual Memory (VM) 28% 76% 48% Closure (VC) 24% 66% 42% Discrimination (VD) 33% 71% 38% Association (VA) 31% 69% 38% Motor Integration (VMI) 28% 64% 36% Spatial Relation (VS) 34% 66% 32% Figure-Ground Discrimination (FG) 39% 65% 26%
On average, each auditory and each visual perceptual skill increased the accuracy of predicting reading ability by 40%. Little variability emerged among the auditory perceptual skills with the difference between the best and worst predictor amounting to only 2%. The greatest percentage increase was found for auditory comprehension (AC), while the largest actual percentage (71%) was found for auditory blending (AB), which amounted to the same percent increase (38%) found for auditory discrimination (AD) and auditory memory (AM). These findings suggest little difference in predictive accuracy among auditory perceptual skills.
Visual perceptual skills exhibited greater divergence between the best and worst predictor, amounting to 22% more correct decisions. This variability was accounted for primarily by the relatively small predictive increase (26%) found for figure-ground discrimination (FG) compared to the large increase in predictive accuracy for visual memory (VM) (48%). Although, on average, predictions about reading ability were approximately equal to those using auditory perceptual skills (69%), VM emerged as the best single predictor with successful determinations about subsequent reading ability increasing from 28% to 76%.
The average correlations were next compared statistically to determine whether differences existed among individual perceptual skills in their relationship to reading ability. These inferential tests were based on Fisher's z-transformation rather than the averaged correlations themselves. The weighted average of correlations presented earlier did not use Fisher's z-transformation because it has been found to possess a positive bias (Hunter et al., 1982). For inferential tests, however, the positive bias presents no difficulties because large correlations are expanded relative to smaller ones. This causes the confidence intervals around large correlations to be smaller than those around small correlations, resulting in greater robustness for inferential tests.
The comparisons revealed no significant differences (F(3,371) = 2.47 p [is less than] .10) among auditory perceptual skills and reading ability, suggesting approximately equal strength of association for all auditory perceptual skill relationships with reading. However, differences did emerge among visual perceptual skills (F(6,1010) = 17.53, p [is less than] .001) with significant Tukey-Kramer (T-K) comparisons (see Stoline, 1981) revealing that VM and visual discrimination (VD) increased predictive accuracy more than the other five visual perceptual skills.
Although auditory and visual perceptual skills are initiated by different stimuli, the fundamental processes involved share common elements that permit comparison. Discrimination skills (AD vs. VD) and memory skills (AM vs. VM) were found to be significantly different, with VD larger than AD (t(472) = 2.19, p [is less than] .025) and VM larger than AM (t(236) = 4.63, p [is less than] .001). Thus, VD and VM appear to provide a greater increase in predictive accuracy about reading achievement than their auditory counterparts (AD and AM).
Auditory and Visual Perceptual Skills and Reading Skills
Correlation coefficients were aggregated by auditory and visual perceptual skills for the individual reading skills; homogeneity tests on all aggregations showed them to be homogeneous. The findings are presented in Table 2.
Table 2 Relationship Between Auditory and Visual Perceptual Skills and Reading Skills
General Reading Ability (GR) PAI Perceptual Skill N M From To Auditory AC 11 .394 28% 68% AM 33 .368 32% 68% AB 22 .413 31% 73% AD 59 .382 31% 69% Visual VMem 52 .472 29% 77% VC 23 .376 29% 67% VD 143 .418 29% 71% VA 17 .410 29% 71% VMI 164 .409 28% 68% VS 11 .367 32% 68% FG 13 .262 37% 63% Word Recognition (WR) PAI Perceptual Skill N M From To Auditory AC 5 .277 34% 62% AM 24 .413 27% 69% AB 12 .307 39% 69% AD 46 .342 35% 69% Visual VMem 45 .463 27% 73% VC 18 .272 36% 64% VD 18 .477 26% 74% VA 24 .356 32% 68% VMI 16 .398 28% 68% VS 23 .348 31% 65% FG 6 .266 37% 63% Reading Comprehension (RC) PAI Perceptual Skill N M From To Auditory AC 6 .381 31% 69% AM 23 .404 30% 70% AB 14 .336 33% 67% AD 32 .361 32% 68% Visual VMem 49 .463 31% 77% VC 16 .363 32% 68% VD 25 .372 31% 69% VA 26 .301 35% 65% VMI 13 .361 30% 66% VS 20 .345 31% 65% FG 9 .244 38% 62% Vocabulary (VO) PAI Perceptual Skill N M From To Auditory AC 4 .267 39% 65% AM 9 .603 22% 82% AB 5 .242 35% 59% AD 19 .399 30% 70% Visual VMem 21 .488 28% 76% VC 9 .297 33% 63% VD 14 .315 34% 66% VA 20 .355 30% 66% VMI 14 .342 31% 65% VS 17 .401 28% 68% FG 8 .225 37% 59%
N = Number of Correlation Coefficients.
M = Mean Correlation Coefficient.
PAI = Predictive Accuracy Increased.
AC = Auditory Comprehension.
AM = Auditory Memory.
AB = Auditory Blending.
AD = Auditory Discrimination.
VMem = Visual Memory.
VC = Visual Closure.
VD = Visual Discrimination.
VA = Visual Association.
VMI = Visual Motor Integration.
VS = Visual Spatial Relation.
FG = Figure-Ground Discrimination.
For general reading ability (GR), auditory perceptual skills increased predictive accuracy, on average, by 39%, which was almost the same as the average 39% increase in predictive accuracy for visual perceptual skills. The greatest increase in predictive accuracy (from 31% to 73%) was found for AB among the auditory perceptual skills, while the best predictor of GR among the visual perceptual skills was VM, where successful prediction was increased from 29% to 77%. Although less variability was found among auditory perceptual skills compared to visual perceptual skills (10% vs. 21% between largest and smallest r), approximately equal predictive increases were noted across auditory and visual variables in determining GR when the variables AB, VM, and FG were not considered. No differences (F(3,121) = 1.55, p [is less than] .25) were found among auditory perceptual skills, while visual perceptual skills showed two significant T-K comparisons (F(6,416) = 4.02, p [is less than] .001), revealing that VM and VD were the best predictors of GR.
For the word recognition (WR) subskill, the auditory perceptual skills of AD, AB, and AM increased predictive accuracy to 69% while predictive accuracy was increased to about 73% with the visual skills of VD and VM. Although also showing increased predictive accuracy, the perceptual skills of AC, visual closure (VC), and FG appeared less predictive of WR, while the remaining visual skills (VMI, VS, VA) fell between the best and worst predictors with an average increase of 37% in predictive accuracy. As was the case with GR, no significant differences (F(3,83) = 1.12, p [is greater than] .25) were found among auditory perceptual variables, while significant T-K comparisons (F(6,143) = 4.43, p [is greater than] .001) showed VM and VD to be the best predictors among the visual perceptual skills for WR.
Measures of reading comprehension (RC) found approximately equal percentage increases (to about 69%) for all auditory perceptual skills. This was confirmed by the lack of significant differences (F(3, 71) = 1.12, p [is less than] .25) among the correlations for auditory perceptual skills; RC can be better predicted, by approximately 37% with the use of any auditory perceptual assessment. Greater variability was again noted among visual perceptual skills, ranging from 24% (from 38% to 62%) for FG to 46% (from 31% to 77%) for VM. The remaining visual variables led to an average predictive increase to 67%, which was only slightly less than that found for auditory perceptual skills. Differences (F(6,151) = 3.66, p [is less than] .01) were found among visual perceptual variables with T-K comparisons revealing VM to be a better predictor than the other visual perceptual skills. Thus, in contrast to auditory perceptual variables, which were found to be almost equal predictors of RC, only VM among visual perceptual variables appeared to significantly increase predictive accuracy.
For vocabulary (VO) measures, AM was by far the best predictor, increasing predictive accuracy by 60% (from 22% to 82%). The only visual perceptual skill approaching the magnitude of AM was VM, whose predictive success was increased from 28% to 76% for an improvement of 48% in predictive accuracy. With these two exceptions, the remaning auditory and visual perceptual skills revealed more modest increases in predictive accuracy to approximately 64%; an average percentage increase of 30% and 32% for auditory and visual variables, respectively. When the correlations were compared, both auditory (F(3,33) = 14.22, p [is less than] .001) and visual (F(6,103) = 6.63, p [is less than] .001) skills revealed differences, with T-K comparisons showing AM and AD in the auditory area and VM and visual spatial relationship (VS) in the visual area to be significantly better predictors than their companion skills.
Auditory and Visual Perceptual Measures and Reading Ability
The 267 studies analyzed included an assortment of instruments to assess auditory and visual perceptual ability. A number of these assessments represent historically important process measures. Table 3 presents data describing the relationship between specific perceptual measures and reading ability.
Table 3 Relationship Between Auditory and Visual Perceptual Measures and Reading Ability
Number of Perceptual Correlation Skill Measure Coefficients Auditory Wepman(1) 59 Discrimination (.371) Murphy-Durrell-SD(2) 35 Auditory Roswell-Chall(3) 29 Blending (.377) ITPA-SB(4) 16 Auditory Memory (.383) WISC-Digit Span(5) 34 ITPA-ASM(6) 27 Visual-Motor Bender(7) 81 Integration (.361) Frostig-Eye Motor(8) 31 Figure-Ground Frostig-Figure Ground(9) 32 Discrimination (.251) Mean Increase in Perceptual Correlation Predictive Skill Coefficients Accuracy Auditory .352 36% Discrimination (.371) .313 32% Auditory .465 46% Blending (.377) .597 60% Auditory Memory (.383) .329 32% .388 38% Visual-Motor .318 32% Integration (.361) .237 24% Figure-Ground .226 22% Discrimination (.251)
(1) Wepman Auditory Discrimination Test.
(2) Murphy-Durrell Diagnostic Reading Readiness Test (Sound Discrimination).
(3) Roswell-Chall Auditory Blending Test.
(4) Illinois Test of Psycholinguistic Abilities (Sound Blending).
(5) Wechsler Intelligence Scale for Children (Digit Span).
(6) Illinois Test of Psycholinguistic Abilities (Auditory Sequential Memory).
(7) Bender Visual-Motor Gestalt.
(8) Marianne Frostig Developmental Test of Visual Perception (Eye-Motor Coordination).
(9) Marianne Frostig Developmental Test of Visual Perception (Figure Ground Perception).
For five perceptual skills, meaningful integration was possible. The resulting aggregations could be compared to the average correlation for each perceptual skill to assess their relative predictability. In the case of AD, the two most popular measures were less predictive. Although there was a 38% increase in the number of correct decisions overall, the Wepman and Murphy-Durrell increased predictive accuracy by 36% and 32%, respectively. When compared, however, the correlations were not significantly different (F(2,274) = 2.96, p [is less than] .10). For AB, the most frequently used measures were far more associated with reading ability than the average measure (.377). Although the typical AB measure led to 38% more correct predictions, the Roswell-Chall and the Illinois Test of Psycholinguistic Ability-Sound Blending (ITPA-SB) produced 46% and 60% increases in predictive accuracy, respectively. Both were significantly larger than the AB average correlation (F(2,109) = 44.29, p [is less than] .001) with ITPA-SB larger than the Roswell-Chall. The 60% increase in predictive accuracy for the ITPA-SB suggests that the use of this test moves the number of correct decisions to 80%, indicating that for 8 out of 10 students there are accurate predictions about subsequent reading ability.
The recent emphasis on phonological awareness (see Stanovich, 1988), and specifically sound blending, supports the strong association with reading ability (Slocum, O'Connor, & Jenkins, 1993; Torgesen, Morgan, & Davis, 1992). For AM, the most frequently used measures were split with respect to their relation with the average measure (.383). The ITPA-auditory sequential memory (ASM) was about the same as the average AM measure, but the WISC-Digit Span was significantly less predictive (F(2,157) = 3.54, p [is less than] .05). The use of the WISC-Digit Span thus produces 6% fewer correct decisions about subsequent reading ability.
In the visual perceptual realm, the two most frequently used measures of visual-motor integration (VMI) were both poorer predictors than the average VMI measure (.361). Although the number of correct predictions increased 36% overall, the percentages for the Bender and the Frostig were 32% and 24%, respectively. For both popular measures, predictions were significantly smaller than the VMI average (.347), with the Frostig being significantly less predictive than the Bender (F(2,514) = 17.13, p [is less than] .001). Finally, the most frequently used assessment of FG (Frostig) showed about equal association with the average correlation (t(76) = [is less than] 1, p [is less than] .50).
Auditory and Visual Perceptual Skills and Subject Characterizations
The next analysis addressed the question: Are there differences in the relationship between auditory and visual perceptual skills and reading ability for different subject groups? Although there were no differences between average achieving (AA) and learning or reading disabled (LD/RD) groups across auditory and visual perception, it is useful to determine whether the relationship between individual auditory and visual perceptual skills and reading ability differed for the two groups. The findings aggregated by subject classification are shown in Table 4.
Table 4 Relationship Between Auditory and Visual Perceptual Skills and Subject Classification
Subject Classification Students with Average Reading Achievement PAI Perceptual Skill N M From To Auditory Comprehension 13 .545 25% 79% Memory 48 .352 34% 70% Blending 35 .363 32% 68% Discrimination 92 .381 31% 69% Visual Memory 67 .476 28% 76% Closure 47 .438 28% 72% Discrimination 191 .397 30% 70% Association 12 .383 31% 69% Motor Integration 193 .352 30% 66% Spatial Relations 49 .319 34% 66% Figure-ground Discrimination 36 .242 36% 60% Subject Classification Students with Learning or Reading Disabilities PAI Perceptual Skill N M From To Auditory Comprehension 10 .241 38% 62% Memory 31 .327 32% 64% Blending 13 .553 24% 80% Discrimination 41 .372 31% 69% Visual Memory 62 .465 29% 75% Closure 17 .431 28% 72% Discrimination 77 .427 29% 71% Association 13 .388 29% 67% Motor Integration 100 .363 32% 68% Spatial Relations 7 .379 31% 69% Figure-ground Discrimination 5 .292 35% 65%
N = Number of Correlation Coefficients.
M = Mean Correlation Coefficient.
PAI = Predictive Accuracy Increased.
Students with average reading ability exhibited approximately the same mean percentage increases in predictive accuracy across auditory (41%) and visual (38%) perceptual skills with both showing approximately the same amount of variability. In the auditory perceptual domain, AC revealed the greatest increase (from 25% to 79%) while the best predictor among the visual perceptual skills was VM, with an increase in predictive accuracy from 28% to 76%. Differences (F(3,184) = 4.46, p [is less than] .01) across auditory perceptual skills emerged, with T-K comparisons confirming the finding that AC was a significantly better predictor than all other auditory variables for students with average reading ability. In the visual realm, T-K comparisons revealed that VM, VD, and VC were significantly better predictors than FG (F(6,588) = 3.39, p [is less than] .01).
For students with LD/RD, visual perceptual skills increased predictive accuracy, on average, slightly more than auditory skills (39% vs. 37%). The greatest increase (from 24% to 80%) in predictive accuracy was found for AB, whereas VM was again the best predictor among visual perceptual skills as evidenced by the number of correct decisions about reading ability being increased from 29% to 75%. Comparison of the average correlations in the auditory realm confirmed the finding that AB was a significantly better predictor than each of the other auditory perceptual skills (F(3,91) = 7.61, p [is less than] .001). For the visual perceptual skills, the one significant T-K comparison (F(6,274) = 2.39, p [is less than] .05) revealed that VM was a better predictor than FG for students with LD/RD.
When subject groups were compared across auditory (.373 vs .341) and visual (.378 vs .405) perceptual skills, there were no differences between average and LD/RD readers for auditory perceptual abilities (t(281) = 1.78, p [is less than] .10) but differences favoring LD/RD readers in the visual perceptual realm (t(874) = 2.45 p [is less than] .025); thus, visual perception may be more associated with reading ability for subjects with LD/RD. Although visual perception revealed a larger relationship overall, comparisons between groups for individual perceptual skills showed differences only among auditory perceptual skills. Significant increases in predictive accuracy were found for AC (t(21) = 4.47, p [is less than] .001) favoring students with average reading ability while AB (t(36) = 2.94, p [is less than] .01) favored the students with LD/RD.
The Case of Auditory-Visual Integration
During the mid-1960s, attention shifted from individual auditory or visual perceptual skills to the ability to integrate auditory and visual perceptual stimuli as the most important skill for achieving success in reading (see Jones, 1972). Major proponents of this view were Birch and Belmont (1964, 1965), who used a paradigm involving a pattern-matching task to test the ability to match a visual spatial pattern to an auditory temporal pattern. Generally, findings suggested that the ability to perform this task was significantly correlated with reading achievement (e.g., Beery, 1967; Muehl & Kremanek, 1966; Reilly, 1971).
Among the 2,294 correlation coefficients, 117 investigated the association between auditory-visual integration (AVI) and reading ability. The average correlation was .331 with an SD of .151 and an SE of .015. The range of correlations was .025 to .617 with a median of .328. For the 4,400 subjects studied, the average age was 7.94 years and average IQ was 103.7. With a BESD interpretation, an assessment of AVI yields an increase in predictive accuracy from 33% to 67% resulting in 34% more correct decisions.
The ability to integrate perceptual stimuli appears no more associated with reading ability than individual auditory or visual skills. However, problems were noted with the nature of the assessments. The basic Birch and Belmont (1964) task was refined and extended to include (a) longer versions of the basic task (presumably to increase reliability), (b) increased test ceiling, and (c) more precision in presenting auditory stimuli. Additionally, the Birch and Belmont assessments were criticized as "impure" measures of AVI that confounded integrative abilities (Rudnick, Martin, & Sterritt, 1972; Rudnick, Sterritt, & Flax, 1967; Sterritt, Martin, & Rudnick, 1971). Consequently, procedures that did not confound the basic integration task were developed and were considered "pure" tests of AVI. Table 5 shows AVI data integrated by assessment procedure.
Table 5 Relationship Between Auditory-Visual Integration Assessment Procedures and Reading Ability
Predictive Accuracy Number of Mean Increased Correlation Correlation Assessment Coefficients Coefficient From To Birch & Belmont Original 27 .329 34% 66% Modified 32 .333 33% 67% Longer 26 .341 33% 67% Sterritt-Type 32 .299 36% 66% Increase in Number of Correct Assessment Decisions Birch & Belmont Original 32% Modified 34% Longer 34% Sterritt-Type 30%
Little difference was found among different assessment procedures; on average, each increased the number of correct decisions by 33%. Comparisons among the assessment procedures yielded no significant differences (F(3,113) = 1.09, p [is less than] .50). Although it remains unclear exactly what is being assessed when AVI ability is tested, some form of integrative process appears related to reading ability but no more so than individual auditory or visual perceptual skills.
The relationship between AVI and reading skills is shown in Table 6.
Table 6 Relationship Between Auditory-Visual Integration Assessment Procedures and Reading Ability
Number of Mean Correlation Correlation Reading Skill Coefficients Coefficient General Reading Ability (GR) 33 .321 Word Recognition (WR) 27 .342 Reading Comprehension (RC) 19 .347 Vocabulary (VO) 14 .208 Predictive Accuracy Increase in Increased Number of Correct Reading Skill From To Decisions General Reading Ability (GR) 34% 66% 32% Word Recognition (WR) 34% 68% 34% Reading Comprehension (RC) 33% 67% 34% Vocabulary (VO) 40% 60% 20%
For GR, WR, and RC, AVI demonstrated about the same association as any other auditory or visual perceptual skill. On average, an assessment of AVI resulted in 33% more correct decisions about GE, WR, or RC. For VO, however, the number of correct decisions was increased only 20%, suggesting that AVI is less predictive of word knowledge; the obtained correlation (.208) was the lowest among all other auditory or visual perceptual skills. However, there were no significant differences in the comparison of AVI ability across reading skills (F(3,89) = 2.23, p [is less than] .10).
With respect to subject classification, AVI appeared to be more associated with reading ability for students with average reading achievement (r = .356) than for students with LD/RD (r = .209). In terms of prediction, there was a 36% increase in the number of correct decisions (33% to 69%) for students with average reading achievements compared to a 20% increase (40% to 60%) in correct decisions for students with LD/RD. The 9% difference in number of correct decisions represents a significant difference between correlations (t(91) = 4.90, p [is less than] .001) and the conclusion that AVI is a better predictor for students with average reading ability and may not be a significant LD/RD correlate.
Relationship Among Auditory Perceptual Skills, Visual Perceptual Skills, and Reading Skills
Thus far, this analysis has presented a description of how auditory and visual perceptual skills individually enhance the prediction of reading ability. The fact that auditory and visual perceptual skills themselves were related to reading variables was established with a canonical correlation analysis. In the auditory perceptual realm, four significant canonical correlations were found to account for 2% to 38% of the variance in the relationship between perceptual and reading skills. Examination of the vector weights revealed AC and AM as major contributors to the predictor (i.e., perceptual) set, while GR, RC, and VO were primary in the criterion (i.e., reading) set. Five significant canonical correlations were found among visual perceptual and reading skills, which accounted for 4% to 52% of the variance in the relationship between predictor and criterion. The vector weights showed that VM, VD, and VMI were major contributors to the predictor set, while all reading skills were equally distributed in the criterion set.
Although auditory and visual perception are independently related to reading ability, they do not operate in isolation. Left unanswered,therefore, are important questions about the extent and significance of each auditory and visual perceptual skill in predicting reading ability. For example, what is the best combination of auditory and visual perceptual variables for predicting facets of reading achievement? This question may be answered with a stepwise multipleregression analysis that indicates both the order and proportion of variance explained by each auditory and visual perceptual variable in predicting reading skills. The results are shown in Table 7.
Table 7 Stepwise Multiple-Regression Analysis Using Auditory and Visual Perceptual Skills to Predict Reading Abilities
Reading Skill General Word Reading Ability Recognition (GR) (WR) Perceptual Perceptual Skill [R.sup.2] Skill [R.sup.2] VM .203 VM .212 AB .317 AM .314 VC .373 VD .365 VD .400 VC .399 VMI .419 AC .427 AD .431 AB .461 AM .434 VMI .483 AC .437 AD .497 VA .439 AVI .502 AVI .440 VA Not Entered Reading Skill Reading Comprehension Vocabulary (RC) (VO) Perceptual Perceptual Skill [R.sup.2] Skill [R.sup.2] VM .212 VM .240 AM .314 AM .360 VD .351 AD .396 AB .373 AVI .447 AC .395 VMI .493 VC .420 VA .520 AD .437 AC .539 VMI .446 AB .547 AVI .451 VC .547 VA .452 VD .548
[R.sup.2] = Percent of variance accounted for.
AB = Auditory Blending.
AC = Auditory Comprehension.
AD = Auditory Discrimination.
AM = Auditory Memory.
AVI = Auditory-Visual Integration.
VA = Visual Association.
VC = Visual Closure.
VD = Visual Discrimination.
VM = Visual Memory.
VMI = Visual Motor Integration.
VS = Visual Spatial Relation.
When 10 auditory and visual perceptual variables were considered, VMem emerged as the first step, accounting for approximately 22% of the variance in the relationship with each of the four reading skills. The second step to enter for each reading skill was an auditory variable (AM for WR, RC, and VO; AB for GR), which added approximately 10% to the predicted variance. The next two variables to be entered were visual perceptual skills for GR (VC and VD) and WR (VD and VC); two auditory variables (AD and AVI) were entered for VO, making three auditory perceptual skills in succession. For RC, a visual and auditory variable (VD and AB) next were entered, increasing the predicted variance by 5% compared to 8% (GR) and 9% (WR and VO) increases, respectively, for the remaining reading variables. The remaining perceptual variables added anywhere from 4% (GR) to 10% (WR and VO) to the predicted variance with the totals reaching 44% (GR) to 55% (VO) of the explained variance in the relationship. These findings suggest that VM is the most important perceptual skill in the prediction of reading ability. The next most important variable was an auditory perceptual skill (AM or AB) with no discernable pattern of auditory or visual skills entering beyond the second step. In summary, prediction of reading skills appears to be best accomplished with a VM measure and an auditory measure (AM or AB).
Perceptual Skills and Intelligence
Most of the 267 studies also included an assessment of intelligence (IQ). It was found that IQ represented a major factor influencing the nature of the relationship between perceptual skills and reading ability. In the present investigation, the relationship between IQ and each auditory perceptual skill, each visual perceptual skill, and each reading skill was significant (p [is less than] .05, at least). Consequently, a complete description of the relationship between auditory and visual perceptual skills and reading ability requires an assessment of the role of IQ in order to assess the influence of the cognitive component underlying perceptual skills.
The influence of intelligence was examined with a stepwise multiple-regression analysis that included IQ as a variable in addition to the perceptual variables used earlier in predicting reading skills. The results are shown in Table 8.
Table 8 Stepwise Multiple-Regression Analysis Using Auditory Perceptual Skills, Visual Perceptual Skills, and Intelligence (IQ) to Predict Reading Abilities
Reading Skill General Word Reading Ability Recognition (GR) (WR) Variable [R.sup.2] Variable [R.sup.2] IQ .580 IQ .512 VD .612 AM .566 VC .634 VD .597 VM .646 VM .616 AB .658 VC .624 VMI .666 AC .629 AD .670 AB .632 AC .675 VMI .634 VA .676 AD .635 AM .676 AVI .636 AVI .676 VA .636 Reading Skill Reading Comprehension Vocabulary (RC) (VO) Variable [R.sup.2] Variable [R.sup.2] IQ .578 IQ .449 AC .657 AC .508 VD .693 AD .558 AD .707 VM .586 AVI .715 AVI .616 VC .717 VMI .634 VA .720 VA .648 AM .722 AM .661 VMI .723 AB .661 AB .724 VD .662 VM .724 VC .662
[R.sup.2] = Percent of variance accounted for.
When IQ was included, it became the first variable to enter for each reading skill and accounted for anywhere from 45% (VO) to 58% (GR and RC) of the variance in the relationships. Thus, the proportion of variance explained by IQ exceeded the total variance explained by only perceptual variables in their relationships with GR, WR, and RC and was better than 80% of the total explained variance for VO.
For GR, the next three variables were visual skills, (VD, VC, and VM), which increased the amount of predicted variance by 7%, while the seven remaining variables added only an additional 3% to the total predicted variance. The second variable for WR was an auditory skill (AM) followed by two visual skills (VD and VM), which increased the predicted variance by 10%; the remaining seven skills produced only 2% more explained variance. In the case of RC and VO, AC was the second step, accounting for an 8% and 6% increase in predicted variance, respectively. The remaining variables increased the explained variance by 7% for RC but by more than twice as much (15%) for VO.
With IQ entered in the regression analysis, the amount of predicted variance increased, on average, by 19% making the total explained variance account for better than two-thirds of the variance in the relationship between IQ, perceptual skills, and reading skills. A comparison of the two regression analyses revealed clearly the significant influence of IQ on the association between perception and reading.
Auditory Perceptual Skills, Visual Perceptual Skills, Reading Abilities, and Intelligence
With IQ found to make such a significant contribution, it is important to assess its effect on individual relationships between perceptual skills and reading ability. The primary concern is to determine the extent to which perceptual variables remain independent predictors when IQ is considered. To define the role of intelligence, partial correlations between auditory and visual perceptual skills and reading skills were computed with IQ being held constant. The findings are shown in Table 9.
Table 9 First-Order Correlations for the Relationship Between Auditory and Visual Perceptual Skills and Reading Skills with Intelligence Constant
Reading Skill General Word Perceptual Skill Reading Ability Recognition [.597] [.461] Auditory Discrimination [.441] .226 (12%) .233 (10%) Blending [.563] .250 (14%) .171 (12%) Memory [.627] .194 (11%) .228 (20%) Comprehension [.679] .190 (20%) .172 (10%) Visual Integration [.455] .208 (10%) .218 (8%) Visual Discrimination [.388] .244 (18%) .359 (12%) Memory [.534] .192 (28%) .303 (16%) Motor Integration [.396] .288 (12%) .270 (12%) Closure [.423] .201 (18%) .143 (14%) Spatial Relation [.477] .183 (18%) .163 (18%) Association [.493] .105 (32%) .180 (18%) Figure-Ground Discrimination [.373] .061 (20%) .128 (12%) Reading Skill Reading Perceptual Skill Comprehension Vocabulary [.792] [.696] Auditory Discrimination [.441] .246 (12%) .272 (12%) Blending [.563] .188 (16%) .125 (12%) Memory [.627] .216 (18%) .472 (12%) Comprehension [.679] .132 (14%) .128 (14%) Visual Integration [.455] .237 (12%) .103 (6%) Visual Discrimination [.388] .120 (26%) .247 (8%) Memory [.534] -.017 (46%) .075 (40%) Motor Integration [.396] .018 (34%) .132 (20%) Closure [.423] .004 (36%) .020 (28%) Spatial Relation [.477] -.073 (34%) .117 (32%) Association [.493] -.167 (30%) .096 (26%) Figure-Ground Discrimination [.373] -.133 (24%) -.020 (22%)
[ ] = Correlation between intelligence and either perceptual or reading skill.
() = Loss in predictive accuracy with intelligence constant.
When IQ was partialed out, the magnitude of the relationship between both auditory and visual perceptual skills and reading abilities declined. The auditory perceptual skills declined in predictive accuracy by an average 13% (range from 6% to 20%) while the predictive accuracy for visual perceptual skills declined by an average 23% (range from 8% to 46%). Contributing to the greater average suppression for visual perceptual skills were the eight instances where the predictive accuracy was reduced to zero when IQ was held constant. Individual auditory perceptual skills revealed average decreases in predictive accuracy from 9% (AVI) to 17% (AM) across reading skills, while visual perceptual skills showed predictive accuracy declining by an average 16% (VD) to 33% (VM) across reading skills. With IQ constant, the greatest suppression in predictive accuracy was found for the memory skills (AM and VM) while the discrimination abilities (AD and VD) exhibited the greatest independence from intelligence. Thus, the two memory skills (AM and VM) as well as AC and AB in the auditory perceptual area, and VC, VS, and visual association (VA) among visual perceptual skills appear to share a large common variance with IQ in their relationship to reading ability.
The primary results of this investigation indicated that auditory and visual perceptual skills were correlates of reading ability. The magnitude of the associations suggested that consideration of perceptual variables (especially auditory or visual memory, auditory or visual discrimination, auditory comprehension, auditory blending, and visual closure) increased the accuracy with which reading ability is predicted. Although the percentage of accurate predictions about individual reading skills was increased generally, the magnitude of that increase was contingent upon the particular subset of perceptual and reading variables considered. The nature of the association between perceptual and reading skills suggested that perceptual processes possess modest value in predicting subsequent reading performance.
The present findings are more positive than conclusions reached in previous analyses (Hammill & Larsen, 1974; Hammill & McNutt, 1981; Larsen & Hammill, 1975), which suggested that auditory and visual perceptual skills were not sufficiently related to reading achievement to be useful indicators of subsequent reading performance. The differing conclusions may be partially attributable to two factors: (a) previous analyses maintained a static criterion (r = .35) for indicating a level of practical predictive significance and (b) mean correlation coefficients were interpreted solely in terms of the percent of variance explained. With these two factors, there is probably a tendency to underestimate the importance of a correlational relationship; they do not demonstrate the extent to which auditory or visual perceptual skill may be used to increase the accuracy with which reading abilities are predicted. Since the magnitude of association was found contingent upon the combination of perceptual and reading variables studied, defining a single standard for judging practical significance is probably not warranted. The essential question that needs to be addressed is: What is the best combination of auditory and visual perceptual skills that might be assessed to provide the greatest increase in predictive accuracy about subsequent reading skills?
It must be stressed that the present findings possess value for the sole purpose of predicting reading ability particularly for students in the early elementary school years with average intelligence. In this synthesis, the average subject was in grade 3, and no generalizations should be drawn about the relationship between perception and reading beyond that level. Similarly, subject characteristics had little influence; the nature of the relationships appears relatively unchanged by the presence or absence of reading disability.
Although it was generally concluded that auditory and visual perceptual assessments can successfully increase the accuracy of predicting reading achievement, their inclusion in a psychoeducational battery should not be routine. With the exception of measures of auditory (sound) blending, the classic tests used to assess perceptual skills proved to be less efficient measures. When the poor reliability and validity of perceptual measures is also considered (see Coles, 1978), individual perceptual assessment does not appear warranted. Additionally, the significant influence of intelligence on the relationship between perceptual skills and reading ability suggests that when an IQ score is available, no individual perceptual assessment is either necessary or desirable.
In fact, questions about the relationship between auditory and visual perception and reading may be moot. For example, in the visual realm, the difficulties may not be with perceptual processes but rather with the speed of information processing (e.g., Blackwell, McIntyre, & Murray, 1983). Similarly, slow processing of auditory stimuli may be a significant factor in reading ability (e.g., Tallal, 1980). Thus, perceptual ability per se may not be as important as believed (Vellutino, Steger, & Kandel, 1972). The ability to analyze, sequence, and remember auditory stimuli may be the more critical skill for reading (Liberman & Shankweiler, 1985). Reading difficulties may be more associated with the inability to break down sentences into words, words into syllables, and syllables into sequences of individual sounds rather than the ability to perveive those elements (e.g., Bradley & Bryant, 1978; Cermak, 1983; Snowling, 1981). With findings showing that this phonemic awareness is even a better predictor of reading ability than IQ, assessment of perceptual processes is probably not necessary (Adams, 1990; MacDonald & Cornwall, 1995; Tummer & Nesdale, 1985).
Although the LD field was originally dominated by conceptualizations emphasizing the importance of perceptual-motor processes, alternative ideas appear to hold more promise for providing insight into the basic nature of LD. More than 20 years ago, Vellutino, Steger, Moyer, Harding, and Niles (1977) asked the question, "Has the perceptual deficit hypothesis led us astray?" Although the answer is not entirely affirmative, perceptual processes, in their traditional sense, appear to play only a modest role in reading ability and clearly have been supplanted by more modern and sophisticated analyses of processes. It, therefore, appears that questions about the nature of the relationship between auditory and visual perception and reading ability may be laid to rest.
(1) A complete list of the 267 studies is available from Kenneth A. Kavale, N235 Lindquist Center, The University of Iowa, Iowa City, IA 52242.
(2) The correction for attenuation is provided by the formula
[r.sub.c] = r/[square root of [r.sub.xx]] [square root of [r.sub.yy]]
where r = obtained correlation [r.sub.xx] = reliability of measure #1, and [r.sub.yy] = reliability of measure #2
Each correlation was corrected for attenuation and the corrected correlation ([r.sub.c]) represented estimates with the influence of sampling error eliminated.
(3) Sampling error is eliminated by transforming the distribution (i.e., mean and standard deviation) of observed correlations into a distribution of population correlations. The best estimate of the mean population correlation is the weighted mean of the sample correlations (r), given by
r = [Sigma]([N.sub.i][r.sub.i]) [Sigma][N.sub.i]
where [r.sub.i] = correlation in study i and [N.sub.i] = sample size in study i
The sampling error variance is then computed by
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
while the variance in the population correlation is estimated by
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where k = number of studies
The population correlation variance is then obtained by
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
The difference is the correction factor for sampling error.
(4) The homogeneity test is computed by
Q = ([n.sub.i] - 3) ([z.sub.i] - [[z.sub.+]).sup.2]
where [n.sub.i] = sample size [z.sub.i] = Fisher's z transformation for [r.sub.i], and [z.sub.+] = weighted average transformed correlations
Adams, M. J. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA: MIT Press.
Ayres, A. J. (1972). Sensory integration and learning disorders. Los Angeles, CA: Western Psychological Services.
Bartley, S. H. (1969). Principles of perception (2nd ed.). New York: Harper & Row.
Barrett, T. C. (1965). The relationship between measures of pre-reading visual discrimination and first-grade reading achievement. Reading Research Quarterly, 1, 51-76.
Beery, J. W. (1967). Matching of auditory and visual stimuli by average and retarded readers. Child Development, 38, 827-833.
Birch, H. G., & Belmont, L. (1964). Auditory-visual integration in retarded readers. American Journal of Orthopsychiatry, 35, 842-861.
Birch, H. G., & Belmont, L. (1965). Auditory-visual integration, intelligence, and reading ability in school children. Perceptual and Motor Skills, 20, 295-305.
Blackwell, S. L., McIntyre, C. W., & Murray, M. E. (1983). Information processed from brief visual displays by learning-disabled boys. Child Development, 54, 927-940.
Bradley, L., & Bryant, P. E. (1978). Difficulties in auditory organization as a possible cause of reading backwardness. Nature, 301, 419-421.
Bradley, L., & Bryant, P. E. (1985). Rhyme and reason in reading and spelling. Ann Arbor: University of Michigan Press.
Cermak, L. S. (1983). Information processing deficits in children with learning disabilities. Journal of Learning Disabilities, 16, 599-605.
Cohen, S. A. (1969). Studies in visual perception and reading in disadvantaged children. Journal of Learning Disabilities, 2, 498-503.
Coles, G. S. (1978). The learning-disability test battery: Empirical and social issues. Harvard Educational Review, 48, 313-340.
Cooper, H. M. (1984). The integrative research review: A systematic approach. Beverly Hills, CA: Sage.
Cooper, H. M., & Hedges, L. V. (1993). The handbook of research synthesis. New York: Russell Sage Foundation.
Dember, W. N. (1960). The psychology of perception. New York: Holt, Rinehart, & Winston.
Dykstra, R. (1966). Auditory discrimination abilities and beginning reading achievement. Reading Research Quarterly, 1, 5-33.
Frostig, M. (1972). Visual perception, integrative functions, and academic learning Journal of Learning Disabilities, 5, 1-15.
Frostig, M., Lefever, W., & Whittlesey, J. (1966). Marianne Frostig Developmental Test of Visual Perception: Administration and scoring manual. Palo Alto, CA: Consulting Psychologists Press.
Gersten, R., & Carnine, D. (1984). Auditory-perceptual skills and reading: A response to Kavale's meta-analysis. Remedial and Special Education, 5, 16-19.
Gersten, R., & Carnine, D. (1985). Two ships passing in the night. Remedial and Special Education, 6, 46-47.
Getman, G. N. (1965). The visuomotor complex in the acquisition of learning skills. In J. Hellmuth (Ed.), Learning disorders (Vol. 1, pp. 49-76). Seattle, WA: Special Child Publications.
Gibson, E. J. (1969). Principles of perceptual learning and development. New York: Appleton-Century-Crofts.
Glass, G. V. (1976). Primary, secondary, and meta-analysis of research. Educational Researcher, 5, 3-8.
Glass, G. V., McGaw, B., & Smith, M. L. (1981). Meta-analysis in social research. Beverly Hills, CA: Sage.
Goins, J. T. (1958). Visual perceptual abilities and early reading progress. Supplementary Educational Monographs, No. 87. Chicago: University of Chicago Press.
Goldstein, K. (1942). After-effects of brain injuries in war. New York: Grune & Stratton.
Groff, P. (1975). Reading ability and auditory discrimination: Are they related? The Reading Teacher, 28, 742-747.
Hallahan, D. P., & Cruickshank, W. M. (1973). Psychoeducational foundations of learning disabilities. Englewood Cliffs, NJ: Prentice-Hall.
Hammill, D. D., & Larsen, S. C. (1974). The relationship of selected auditory perceptual skills and reading ability. Journal of Learning Disabilities, 7, 429-435.
Hammill, D. D., & McNutt, G. (1981). The correlates of reading. Pro-Ed Monograph #1. Austin, TX: Pro-Ed.
Hebb, D. O. (1949). The organization of behavior. New York: Wiley.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. New York: Academic Press.
Henle, M. (1961). Documents of Gestalt psychology. Berkeley: University of California Press.
Hunt, M. (1997). How science takes stock: The story of meta-analysis. New York: Russell Sage Foundation.
Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Correcting error and bias in research findings. Newbury Park, CA: Sage.
Hunter, J. E., Schmidt, F. L., & Jackson, G. B. (1982). Meta-analysis: Cumulating research findings across studies. Beverly Hills, CA: Sage.
Johnson, D. J., & Myklebust, H. R. (1967). Learning disabilities: Educational principles and practices. New York: Grune & Stratton.
Jones, J. P. (1972). Intersensory transfer, perceptual shifting, modal preference, and reading. Newark, DE: International Reading Association.
Kavale, K. A. (1980). Auditory-visual integration and its relationship to reading achievement: A meta-analysis. Perceptual and Motor Skills, 51,947- 955.
Kavale, K. A. (1981). The relationship between auditory perceptual skills and reading ability: A meta-analysis. Journal of Learning Disabilities, 14, 539- 546.
Kavale, K. A. (1982). Meta-analysis of the relationship between visual perceptual skills and reading achievement. Journal of Learning Disabilities, 15, 42-51.
Kavale, K.A. (1984). Potential advantages of the meta-analysis technique for special education. Journal of Special Education, 18, 61-72.
Kavale, K. A. (1985). Auditory perceptual skills and reading: A rejoinder to Gersten and Carnine about what Kavale did. Remedial and Special Education, 6, 43-45.
Kavale, K. A., & Forness, S. R. (1987). A matter of substance over style: A quantitative synthesis assessing the efficacy of modality testing and teaching. Exceptional Children, 54, 228-234.
Kavale, K. A., & Forness, S. R. (1995). The nature of learning disabilities: Critical elements of diagnosis and classification. Mahwah, NJ: Erlbaum.
Kavale, K. A., & Mattson, P. D. (1983). "One jumped off the balance beam": A meta-analysis of perceptual-motor training. Journal of Learning Disabilities, 16, 165-173.
Kephart, N. C. (1960). The slow learner in the classroom. Columbus, OH: Merrill.
Larsen, S. C., & Hammill, D. D. (1975). The relationship of selected visual-perceptual abilities to school learning. Journal of Special Education, 9, 281-291.
Liberman, I. Y., & Shankweiler, D. (1985). Phonology and the problems of learning to read and write. Remedial and Special Education, 6, 8-17.
MacDonald, G. W., & Cornwall, A. (1995). The relationship between phonological awareness and reading and spelling achievement eleven years later. Journal of Learning Disabilities, 28, 523-527.
Mann, L. (1979). On the trail of process. New York: Grune & Stratton.
McNinch, G. (1971). Auditory perceptual factors and measured first- grade reading achievement. Reading Research Quarterly, 6, 472-492.
Muehl, S., & Kremanek, S. (1966). Ability to match information within and between auditory and visual sense modalities and subsequent reading achievement. Journal of Educational Psychology, 57, 230-239.
Murch, G. M. (1973). Visual and auditory perception. Indianapolis, IN: Bobbs-Merrill
Piaget, J. (1969). The mechanisms of perception. London: Routledge & Kegan Paul.
Preece, P. F. (1983). A measure of experimental effect size based on success rates. Educational and Psychological Measurement, 43, 763-766.
Reilly, D. H. (1971). Auditory-visual integration, sex, and reading achievement. Journal of Educational Psychology, 62, 482-486.
Rosenthal, R. (1991). Meta-analytic procedures for social research. Beverly Hills, CA: Sage.
Rosenthal, R., & Rubin, D. B. (1979). A note on percent variance explained as a measure of the importance of effects. Journal of Applied Social Psychology, 9, 395-396.
Rosenthal, R., & Rubin, D. B. (1982). A simple, general purpose display of magnitude of experimental effect. Journal of Educational Psychology, 74, 166- 169.
Rudnick, M., Martin, V., & Sterritt, G. M. (1972). On the relative difficulty of auditory and visual, temporal and spatial, integrative and nonintegrative sequential pattern comparisons. Psychonomic Science, 27, 207- 210.
Rudnick, M., Sterritt, G. M., & Flax, M. (1967). Auditory and visual rhythm perception and reading ability. Child Development, 38, 581-588.
Sabatino, D. A. (1973). Auditory perception: Development, assessment, and intervention. In L. Mann & D. Sabatino (Eds.), The first review of special education (pp. 49-86). New York: Grune & Stratton.
Sabatino, D. A. (1979). The definition and assessment of visual and auditory perception. Journal of Clinical Psychology, 8, 188-194.
Samuels, S. J. (1973). Success and failure in learning to read: A critique of the research. Reading Research Quarterly, 8, 200-239.
Slocum, T. A., O'Connor, R. E., & Jenkins, J. R. (1993). Transfer among phonological manipulation skills. Journal of Educational Psychology, 85, 618- 630.
Snowling, M. J. (1981). Phonemic deficits in developmental dyslexia. Psychological Research, 43, 219-234.
Stanovich, K. E. (1988). Explaning the differences between the dyslexic and garden-variety poor reader: The phonological-core variable-difference model. Journal of Learning Disabilities, 21, 590-612.
Sterritt, G. M., Martin, V., & Rudnick, M. (1971). Auditory-visual and temporal-spatial integration as determinants of test difficulty. Psychonomic Science, 23, 289-291.
Stoline, M. R. (1981) The status of multiple comparisons: Simultaneous estimation of all pairwise comparisons in one-way ANOVA designs. The American Statistician, 35, 134-141.
Strauss, A. A., & Lehtinen, L. E. (1947). Psychopathology and education of the brain-injured child. New York: Grune & Stratton.
Tallal, P. (1980). Auditory temporal perception, phonics, and reading disabilities in children. Brain and Language, 9, 182-198.
Torgesen, J. K. (1979). What shall we do with psychological processes? Journal of Learning Disabilities, 12, 514-522.
Torgesen, J. K., Morgan, S. T., & Davis, C. (1992). Effects of two types of phonological awareness training on word learning in kindergarten children. Journal of Educational Psychology, 84, 364-370.
Tummer, W. E., & Nesdale, A. R. (1985). Phonemic segmentation skill and beginning reading. Journal of Educational Psychology, 77, 417-427.
Vellutino, F. R. (1979). Dyslexia: Theory and research. Cambridge, MA: MIT Press.
Vellutino, F. R., Steger, J. A., & Kandel, G. (1972). Reading disability: An investigation of the perceptual deficit hypothesis. Cortex, 8, 106-118.
Vellutino, F. R., Steger, J. A., Moyer, S. C., Harding, C. J., & Niles, J. A. (1977). Has the perceptual deficit hypothesis led us astray? Journal of Learning Disabilities, 10, 375-385.
Wachter, K. M., & Straf, M. L. (1990). The future of meta-analysis. New York: Russell Sage Foundation.
Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological processing and its causal role in the acquisition of reading skills. Psychological Bulletin, 101, 192-212.
Wepman, J. M. (1964). The perceptual basis for learning. In H. A. Robinson (Ed.), Meeting individual differences in reading (pp. 25-33). Chicago: University of Chicago Press.
Wolf, F. M. (1986). Meta-analysis: Quantitative methods for research synthesis. Beverly Hills, CA: Sage.
Please address correspondence to: Kenneth A. Kavale, The University of Iowa, College of Education, N235 Lindquist Center, Iowa City, IA 52242.
KENNETH A. KAVALE, Ph.D., is professor of special education, University of Iowa.
STEVEN R. FORNESS, Ed.D., is professor and chief of psychological services, Neuropsychiatric Institute, University of California, Los Angeles.
|Printer friendly Cite/link Email Feedback|
|Author:||Forness, Steven R.|
|Publication:||Learning Disability Quarterly|
|Article Type:||Statistical Data Included|
|Date:||Sep 22, 2000|
|Previous Article:||READING OUTCOMES FOR STUDENTS WITH AND WITHOUT READING DISABILITIES IN GENERAL EDUCATION MIDDLE-SCHOOL CONTENT AREA CLASSES.|
|Next Article:||EFFECTS OF COMPUTER-BASED TEST ACCOMMODATIONS ON MATHEMATICS PERFORMANCE ASSESSMENTS FOR SECONDARY STUDENTS WITH LEARNING DISABILITIES.|