Paths to Reading Comprehension in At-Risk Second-Grade Readers Virginia W. Berninger, Robert D. Abbott, Karin Vermeulen, and Cynthia M. Fulton

Abstract Two studies of second graders at risk for reading disability, which were guided by levels of language and functional reading system theory, focused on reading comprehension in this population. In Study 1 (n = 96), confirmatory factor analysis of five comprehension measures loaded on one factor in both fall and spring of second grade. Phonological decoding predicted accuracy of real-word reading; automatic letter naming predicted rate of real-word reading; accuracy and rate of both real-word reading (more so than decoding of pseudowords) and text reading predicted reading comprehension; and Verbal IQ also predicted reading comprehension. In Study 2 (n = 98), the treatment group (before/after school clubs receiving an integrated instructional approach that was supplementary to the general reading program) improved significantly more in phonological decoding and state standards for reading fluency than the control group (general reading program that had some code instruction but emphasized comprehension). The rate of phonological decoding explained 60.3% of real-word reading. Both treatment and control children improved significantly in reading comprehension, but controlling for pretreatment individual differences in oral vocabulary or in phonological decoding eliminated this effect. Taken together, the results of the two studies support two paths to reading comprehension: one from vocabulary and verbal reasoning, and one from written language that has multiple links between subskills: (a) alphabetic principle → phonological decoding, (b) automatic phonological decoding → accurate real-word reading, (c) automatic letter coding → automatic word reading, and (d) automatic word reading → fluent text reading. Instructional implications of both paths and the links within the written language are discussed.

R

ecently, the United States National Reading Panel issued a summary of research evidence for effective instructional practices in the general education classroom (National Reading Panel, 2000). This report contained scientific information that is valuable in preventing reading disabilities and in optimizing the achievement of all students. Based on a review of the scientific evidence, the panel emphasized the importance of instruction aimed at phonological awareness, alphabetic principle and phonological decoding, fluency training, and reading comprehension. In exploring the application of the panel’s findings to practice, more attention has been given to phonological awareness and decoding (e.g., Ehri, Nunes, Stahl, & Willows, 2001) than to reading comprehension, but Carlisle and Rice (2002) provided a review of research-

supported instructional practices for vocabulary and reading comprehension. The purpose of this article is to investigate issues related to improving reading comprehension in second graders who are at risk because of problems in learning to read words. Longitudinal research has clearly documented that without early intervention, these students who have difficulty in reading single words out of sentence context are at substantial risk for remaining poor readers (e.g., Juel, 1988; Stanovich, 1986). However, longitudinal research on effective instructional approaches for children who struggled in learning to read words out of sentence context (128 children in first and second grade) showed that (a) some at-risk readers have only word reading impairment, whereas others have both word reading and reading

JOURNAL OF LEARNING DISABILITIES VOLUME 39, NUMBER 4, JULY/AUGUST 2006, PAGES 334–351

comprehension impairments; and (b) the faster responders tended to have only a word reading impairment, but the slower responders tended to have both word reading and reading comprehension impairments (Berninger et al., 2002). This longitudinal study was followed by a randomized, controlled experiment that compared three instructional approaches to a contact control for at-risk second-grade readers: (a) word decoding only, (b) reading comprehension only, and (c) combined word decoding and reading comprehension; the combined treatment was more effective than either of the first two treatments in improving word decoding, but all three treatments led to improved reading comprehension (Berninger et al., 2003). This article reports the results of two studies that further investigated reading comprehension and related

335

VOLUME 39, NUMBER 4, JULY/AUGUST 2006

skills in at-risk second-grade readers. Both studies were guided by a conceptual framework that designed assessment and instruction based on levels of language theory and functional systems theory. According to levels of language theory, aural language has multiple levels (e.g., phonology, morphology, syntax, discourse), and written language has these same levels as well as orthography. According to functional systems theory, reading draws on many component processes, which must be organized to work in concert in real time, creating a functional reading system in the learner’s mind to support efficient, fluent reading. For further information on either of these theories, see Berninger and Richards (2002). Both studies differentiated between automaticity (rapid, effortless, context-free retrieval, assessed by the rate of single-word reading) and fluency (fast, smooth, coordinated, contextual reading, assessed by the rate of text reading; Berninger, Abbott, Billingsley, & Nagy, 2001). The first study addressed three questions related to theory and measurement. First, are the distinctions between text-based and reader knowledge–based reading comprehension that have informed models of skilled comprehension useful in understanding the reading comprehension of struggling second-grade readers? Second, are some of the widely used approaches for assessing reading comprehension more authentic (i.e., more closely resembling real-world reading comprehension) than others for assessing reading comprehension in this population? Third, which component reading skills or related skills best explain the processes underlying reading comprehension in this population? We used factor analysis to answer the first two questions and multiple-regression to answer the third question. The second study compared all schools in a school district, which were randomly assigned to treatment or control groups, to evaluate whether the treatment groups that received instruction aimed at all levels of

language (subword, word, and text) improved more in word decoding, reading fluency, and reading comprehension than a control group that received only the general education program in reading for second graders. According to functional reading system theory, this integrated approach aimed at all levels of language should lead to coordinated, fluent reading. Participating children in both groups were identified by their teachers at the beginning of second grade on the basis of the state-mandated assessment of reading accuracy and fluency.

Study 1 Rationale The measures used to assess reading comprehension differed in the degree to which they emphasized text-based or situation-based comprehension (Kintsch, 1998). The representation of text in the reader’s mind is always textbased and situation-based to some degree, but one kind of representation tends to predominate. In text-based comprehension, information is derived directly from the text. In situationbased comprehension, text interpretation is based on the integration of the local text-derived microstructure in words and sentences and the constructed knowledge, which is based on domain and background knowledge activated in long-term memory, inferences that go beyond what is stated in the text, and emergent global structures that the reader generates as meaning is constructed. Both researchers and educational practitioners have strong views about which of the kinds of reading comprehension measures used in schools are better measures of reading comprehension. Some believe that asking children questions about whole texts is a more authentic assessment than asking them to supply a missing word (i.e., cloze procedure). In this study, we chose representative measures of normreferenced and criterion-referenced measures of reading comprehension used in the schools to evaluate empiri-

cally whether these measures are assessing different comprehension processes.

Method Participants. Second-grade teachers in eight schools serving diverse student populations were asked to refer their poorest readers. Project personnel administered a screening battery to the 179 referred children. This battery included the Wechsler Intelligence Scale for Children, third edition (WISC III; Psychological Corp., 1991), Vocabulary subtest; and the Woodcock Reading Mastery Test–Revised (WRMT-R; Woodcock, 1987) Word Identification and Word Attack subtests. Inclusion criteria were (a) a scaled score of 6 (equivalent to a Verbal IQ of 80) or more on the WISC III Vocabulary subtest; and (b) a score at or below 85 (1 SD below the mean) on WRMT-R Word Identification (realword reading) or Word Attack (pseudoword reading), with both subtests below the population mean. Of the 179 children tested, 96 met the inclusion criteria and their parents granted informed consent for their participation. The final sample of 96 children (56 girls, 40 boys) had a mean Verbal IQ in the average range, but their mean scores on WRMT-R real-word reading (Word Identification), pseudoword reading (Word Attack), and reading comprehension (Passage Comprehension); Reading Comprehension from the Wechsler Individual Achievement Test (WIAT; Psychological Corp., 1992); and Reading Comprehension from the Gray Oral Reading Test–Third Edition (GORT-3; Wiederholt & Bryant, 1992) fell in the low average range (see Table 1). The sample was representative of the U.S. public school population; the majority was European-American, but more than a third (37%) were of other ethnicities. Parental level of education was also diverse: About a third of the parents had a college education, but the majority had less than a college education. Nearly half of the parents reported a family history of reading problems (47.7%). Some children had a

336

JOURNAL OF LEARNING DISABILITIES

history of delay in producing their first word (11.5%), whereas others (not necessarily the same ones) were delayed in talking in sentences (11.4%). Some children were receiving speech services (13.6%), whereas others were receiving Chapter 1 services for reading (19.3%). Test Battery. Additional subtests of the WISC III (Information, Similarities, and Comprehension) were administered to yield a prorated Verbal IQ comparable to the Verbal Comprehension factor. This factor is a measure of verbal reasoning that does not require children to read items. The following measures were given both at the be-

ginning and end of second grade. See Table 1 for means and standard deviations. Accuracy and rate of oral word reading. The WRMT-R Word Identification (average reliability coefficient = .97) and Word Attack (average reliablity coefficient = .87) subtests, respectively, assessed the accuracy of reading real words, which may be familiar, and of pseudowords, which are novel, of increasing difficulty in lists (i.e., without context clues). The children also completed measures of the rate of reading single real words and single pseudowords in lists. These measures were prepublication versions of the subtests of the Test of Word Reading Efficiency

TABLE 1 Means and Standard Deviations for Screening and Comprehension Measures at the Beginning of Second Grade Study 1

Study 2

M

SD

WISC-III Verbal IQ

98.68

14.26

WRMT-R Word Identification Word Attack Passage Comprehension

88.50 88.33 88.5

9.85 9.39 11.5

2.17

0.91

Measure

DRA Fluency WIAT Reading Comprehension

87.8

10.00

GMG Reading Comprehension

34.17

16.19

7.04

2.51

.04

.82

QRI

5.33

6.50

GORT-3 Accuracy

7.14

1.92

TOWRE Real Words Pseudowords

−1.19 −0.73

0.61 0.41

Working Memory

−.00

.76

RAN Letters Letter Clusters

.00 .00

1.00 1.00

GORT-3 Comprehension PAL Sentence Sense

M

SD

94.92 95.55

12.61 11.31

Note. WISC-III = Wechsler Intelligence Scale for Children, 3rd ed. (Psychological Corp., 1991); WRMT-R = Woodcock Reading Mastery Test–Revised (Woodcock, 1987); DRA = Developmental Reading Assessment (Beaver, 1997); WIAT = Wechsler Individual Achievement Test (WIAT; Psychological Corp., 1992); GMG = Gates–MacGinitie Reading Comprehension Test (MacGinitie & MacGinitie, 1989); GORT-3 = Gray Oral Reading Test, 3rd ed. (Wiederholt & Bryant, 1992); PAL = Process Assessment of the Learner Test Battery for Reading and Writing (Berninger, 2001); QRI = Qualitative Reading Inventory (Leslie & Caldwell, 1990); TOWRE = Test of Word Reading Efficiency (Torgesen, Wagner, & Rashotte, 1999); RAN = rapid automatic naming.

(TOWRE; Torgesen, Wagner, & Rashotte, 1999; test–retest reliability = .98). Accuracy and rate of oral reading of text. Both standardized and criterionreferenced measures of the accuracy and rate of oral reading of text (i.e., words in context) were administered. The Gray Oral Reading Test, third edition (GORT-3; Wiederholt & Bryant, 1992) has separate scores for the accuracy (reliability = .90) and rate (reliability = .92) of oral reading of graded passages. Rate on this measure was used as an index of fluency, because it correlated .71 (p < .001) with the National Assessment of Educational Progress (NAEP) fluency rating (on a 4-point scale) in another sample of elementary school students (Berninger, unpublished data). Accuracy of oral reading was assessed with graded passages on the Qualitative Reading Inventory (QRI; Leslie & Caldwell, 1990) until an instructional level (based on accuracy) was achieved. The reading rate (words per minute) was also measured for the passage at the child’s oral reading instructional level. Sentence-level comprehension. The sentence-level comprehension measure was also referred to as meaning judgment or silent reading fluency. In the Process Assessment of the Learner Test Battery for Reading and Writing (PAL) Sentence Sense subtest (Berninger, 2001), the child has to choose the one sentence from a set of three that is meaningful. The three sentences contain only real words, but two contain one word (a foil) that renders the whole sentence meaningless. Correct answers on this timed test require close coordination of word-level decoding and sentence-level comprehension of syntax during silent reading (Berninger, Abbott, & Alsdorf, 1997). This test assesses primarily text-based comprehension (Kintsch, 1998), but at the sentence level rather than discourse level. Predicting a missing word in text. The WRMT-R Passage Comprehension subtest (Woodcock, 1987) employs the cloze procedure, in which the child supplies a missing word that makes sense

VOLUME 39, NUMBER 4, JULY/AUGUST 2006

in the text (reliability = .92). This task requires close coordination of word reading with comprehension of unfolding text—especially comprehension related to creating local cohesion and coherence among sentences that are close together in the text. Correct answers require close attention to what is stated in the text and depend as much on metaknowledge of language (e.g., rules for local coherence and cohesion) as on world knowledge; thus, the cloze test primarily assesses textbased comprehension. Answering questions from memory. The GORT-3 Comprehension subtest (Wiederholt & Bryant, 1992; test–retest reliability = .87) was administered in a multiple choice format following oral reading of graded passages of increasing difficulty. The examiner read each question and the multiple choice alternatives to the child, who then chose one of the multiple choice options. Although this format eliminates decoding requirements for the comprehension questions, some of the answers to the comprehension questions may be guessed correctly on the basis of background knowledge, even if the text was not read accurately. In fact, standardized administration procedures allow continuing administration of the comprehension questions until a ceiling is reached, even if the ceiling has already been reached for accuracy and rate of passage reading. This test has memory requirements, in that the child is not allowed to refer to the passage when answering. In contrast to the cloze procedure, correct answers for this comprehension measure rely more on world knowledge and situationbased comprehension than on metaknowledge of language or text-based comprehension (Kintsch, 1998) and on the comprehension of discourse rather than on relationships among sentences. Answering questions without a memory component. The Wechsler Individual Achievement Test Comprehension subtest (Psychological Corp., 1992; reliability = .90) requires the child to answer open-ended questions following silent reading. However, unlike the

GORT-3 Comprehension subtest, but like the PAL and WRMT-R cloze measures, the text can be consulted in answering. Both the GORT-3 and the WIAT Comprehension subtests include questions that require inferential thinking as well as factual recall. Correct responding on the WIAT may be more influenced by the participant’s world knowledge (extracting informational content from text) than by metaknowledge of language and requires both text-based and situation-based comprehension (Kintsch, 1998) for discourse. Criterion-referenced comprehension. The criterion-referenced Qualitative Reading Inventory (QRI; Leslie & Caldwell, 1990) represents authentic assessment, in that its format is similar to the way in which teachers assess comprehension during instruction—by asking questions. Five comprehension questions were asked about the passage at the child’s oral reading instructional level to determine whether the child was independent, instructional, or frustrated in comprehension at the level where the text could be decoded. The first two questions were especially sensitive to text-based comprehension (Kintsch, 1998), but the last three questions were relatively more sensitive to situation-based comprehension (Kintsch, 1998). Time did not permit giving enough passages to determine an instructional level for reading comprehension independent of the ability to read orally. Working memory. An experimenter-designed measure of working memory span for orally presented text was administered because of the documented relationship between working memory and reading comprehension (Oakhill & Yull, 1996; Swanson & Berninger, 1995). Children listened to sets of sentences, which increased in the number of sentences included in each set. They were required to answer questions about the sentences to show that they had processed them and understood them. They were also required to recall the designated parts of the sentences. Scoring on this measure of

337

working memory span took into account both processing (accuracy of question answering) and storage (accuracy of recall). Rapid automatic naming. Children were asked to name 10 single letters. If they could identify all letters, they were then asked to name the same letters arranged as single letters or as letter clusters in five rows. Their score was the time required to name all the single letters or all the letter clusters (Berninger, 2001). Data Analyses. Confirmatory factor analysis. A confirmatory factor analysis using EQS Structural Equation Modeling (Bentler & Wu, 2000) was performed on the five measures of reading comprehension. These analyses were repeated at Time 1 (fall) and Time 2 (spring) of second grade, because reading comprehension improved significantly from the beginning to the end of second grade on each of the five measures. Correlations. Concurrent Pearson product-moment correlations were computed between nine predictor variables (accuracy of real-word reading, accuracy of phonological decoding, rate of real-word reading, rate of phonological decoding, accuracy of oral reading, rate of oral reading on a norm-referenced measure, rate of oral reading on a criterion-referenced measure, working memory span, and Verbal IQ), and each of the five comprehension outcomes was evaluated in the confirmatory factor analysis. Multiple regressions. Theory-based, hierarchical multiple regression analysis was used to test theoretical models of which reading skills contributed to reading comprehension in at-risk second-grade readers. For the first theoretical model, in the first step, the accuracy of oral word reading was entered, and in the second step, the rate (efficiency) of oral word reading was entered for each comprehension measure. For the second theoretical model, in the first step, the accuracy of oral text reading was entered, and in the second step, the rate of oral text read-

338

JOURNAL OF LEARNING DISABILITIES

ing was entered for selected reading comprehension measures. Additional multiple regressions evaluated (a) which of a set of predictors (real-word rate, phonological decoding rate, working memory, and Verbal IQ) uniquely predicted reading comprehension, and (b) whether automatic naming of single letters or letter groups uniquely explained the accuracy and rate of real-word reading in and out of context. Consistent discrepancies between word decoding and reading comprehension over time. Children were identified who—both at the beginning and at the end of second grade—had higher word reading scores (based on oral reading accuracy for real words) than reading comprehension scores (based on the cloze procedure) or higher reading comprehension than word reading. Two criteria were used to define higher performance on one skill than the

other: (a) a difference of slightly more than 1⁄2 SD (8 standard score points), or (b) a full 1 SD or more (15 standard score points or more).

Results Relationship Among Comprehension Measures. Because analysis of variance confirmed that reading comprehension improved significantly from fall to spring of second grade on each of the five measures of reading comprehension, separate confirmatory factor analyses based on all five comprehension measures were performed at Time 1 (beginning of second grade) and Time 2 (end of second grade). As shown in the covariance matrix in Table 2, the standardized factor loadings in the confirmatory factor analysis indicated that at both Time 1 and Time 2, a one-factor solution fit the co-

variances well. Meaning judgment had the most shared variance with the cloze procedure; both were more dependent on text-based comprehension than on situation-based comprehension (Kintsch, 1998). Of all five comprehension measures, the cloze and open-ended measures, which are the most widely used individually administered psychometric measures of reading comprehension in the schools, shared the most common variance. Concurrent Correlates of Reading Comprehension at Time 1 and Time 2. Table 3 summarizes the correlations between the measures of word reading, decoding, text reading, working memory, and Verbal IQ and the five measures of reading comprehension either in the fall (Time 1) or the spring (Time 2). At Time 1, all predictors were correlated with most or all reading comprehension measures (r = .2618,

TABLE 2 Results of Confirmatory Factor Analyses, Correlations, and Standard Deviations for Five Comprehension Measures Correlation Measure

1.

2.

3.

4.

5.

Factor 1 standardized loading

Time 1 (Fall of second grade) 1. PAL Sentence Sense 2. WRMT-R Passage Comprehension 3. WIAT Reading Comprehension 4. GORT-3 Comprehension 5. QRI Comprehension Standard deviation

— .2847*** .2368* .1944 .2916*** 1.3259

— .7494*** .4674*** .5637*** 6.3902

— .4506*** .4615*** 4.3165

— .5491*** 5.5589

— 6.5748

.316 .925 .808 .520 .605

Time 2 (Spring of second grade) 1. PAL Sentence Sense 2. WRMT-R Passage Comprehension 3. WIAT Reading Comprehension 4. GORT-3 Comprehension 5. QRI Comprehension Standard deviation

— .2336* .0525 .1762 .1641 1.7800

— .4719*** .7466*** .3980*** 6.8587

— .4696*** .2992*** 4.4382

— .3140*** 8.3735

— 1.9406

.240 .905 .539 .823 .431

Note. N = 96. All comprehension measures loaded significantly on a single factor. Comparative fit index at Time 1 was .981 and at Time 2 was 1.00. PAL = Process Assessment of the Learner Test Battery for Reading and Writing (Berninger, 2001); fluency measure of timed silent reading comprehension of sentences; multiple choice response format; yields a norm-referenced, age- or grade-corrected score. WRMT-R = Woodcock Reading Mastery Test–Revised (Woodcock, 1987); silent reading cloze procedure; open-ended response format; yields a norm-referenced, age- or grade-corrected score. WIAT = Wechsler Individual Achievement Test (Psychological Corp., 1992); answering questions following silent reading of text that can be inspected; open-ended response format; yields a norm-referenced, age- or grade-corrected score. GORT-3 = Gray Oral Reading Test, 3rd ed. (Wiederholt & Bryant, 1992); answering questions from memory following oral reading of text; multiple choice response format; memory requirement (cannot inspect text while responding); yields a norm-referenced, age- or grade-corrected score. QRI = Qualitative Reading Inventory (Leslie & Caldwell, 1990), percentage correctly answered questions for passage at grade level based on oral reading accuracy; open-ended response format; memory requirement; yields a criterion-referenced instructional level. *p < .05. **p < .01. ***p < .001.

339

VOLUME 39, NUMBER 4, JULY/AUGUST 2006

p < .001): accuracy of single-word reading (four measures), rate of singleword reading (five measures), accuracy of phonological decoding (three measures), rate of phonological decoding (three measures), accuracy of oral text reading (four measures), rate of normed oral text reading (four measures), rate of criterion-referenced oral reading text reading (four measures), working memory (four measures), and Verbal IQ (four measures). At Time 2, all the predictors were still correlated with at least one comprehension outcome, and most were still good predictors of most comprehension outcomes

(see Table 3): accuracy of real-word reading (five measures), accuracy of phonological decoding (three measures), rate of real-word reading (five measures), rate of phonological decoding (three measures), accuracy of oral text reading (three measures), rate of normed oral text reading (four measures), rate of criterion-referenced oral text reading (five measures), working memory (one measure), and Verbal IQ (four measures). In sum, the comparison of the correlations across the school year showed that although phonological decoding was significantly correlated with reading compre-

hension, the measures of accuracy and rate of real-word reading—both in and out of context—were consistently stronger predictors of reading comprehension, in terms of the size of the correlations and the number of comprehension measures with which they were correlated, both in the fall and in the spring of second grade. When the magnitude of the correlation was used to identify the three best predictors for each comprehension measure (see subscripts a, b, and c in Table 3), there was further evidence that the predictors involving real words consistently showed stronger

TABLE 3 Correlations Among Predictor and Reading Comprehension Measures at Beginning and End of Second Grade Comprehension measure Predictor

PAL SS

WRMT-R PC

WIAT-2 RC

GORT-3 Comp

QRI Comp

.7638***a .3616*** .6729***c .2954*** .6320*** .7301***b .6502*** .2834*** .3609***

.5010*** .2462** .4782*** .2390** .5068***b .5043***c .5097***a .2665*** .3280***

.2162* −.0227 .2406**c .0129 .2989***a .1852 .2294* .2374** .2867***b

.7140***b .4764*** .6913*** .5149*** .7104***c .7302***a .5747*** .2955*** .4012***

.4890***a .2866*** .4172*** .3314*** .4620***b .4301***c .3960*** .1528 .3602***

.2944***b .0778 .3499***< .0895 .2121* .2448**c .4142***a .1398 .3316***

Time 1 WRMT-R WI WRMT-R WA TOWRE SW TOWRE PR GORT-3 Acc GORT-3 Rate QRI Working Memory Recall VIQ

.1920 .1514 .3137***b .0624 .1413 .2765***c .3906***a .1115 .1678

.8392***a .4877*** .6556***b .3344*** .5829*** .6298***c .5401*** .2393** .2681*** Time 2

WRMT-R WI WRMT-R WA TOWRE SW TOWRE PR GORT-3 Acc GORT-3 Rate QRI Working Memory Recall VIQ

.3046***c .1364 .2745*** .0911 .2498** .3404***b .3850***a .0263 .1263

.7769***a .5060*** .6359*** .4507*** .7409***b .7134***c .5840*** .2186* .3490***

Note. N = 96. For each outcome (comprehension measure), the predictor that has the highest correlation coefficient is subscripted a, the predictor that has the second highest correlation coefficient is subscripted b, and the predictor that has the third highest correlation coefficient is subscripted c. PAL SS = Process Assessment of the Learner Test Battery for Reading and Writing (Berninger, 2001) Sentence Sense subtest; fluency measure of silent reading comprehension of sentences based on 3-min limit. WRMT-R PC = Woodcock Reading Mastery Test–Revised (Woodcock, 1987) Passage Comprehension subtest; silent reading cloze procedure. WIAT-2 RC = Wechsler Individual Achievement Test, 2nd ed. (Psychological Corp., 2001), Reading Comprehension subtest; answering questions following silent reading of text that can be inspected. GORT-3 Comp = Gray Oral Reading Test, 3rd ed. (Wiederholt & Bryant, 1992), Comprehension subtest; answering questions from memory following oral reading of text. QRI Comp = Qualitative Reading Inventory (Leslie & Caldwell, 1990) Comprehension subtest; percentage correctly answered questions for passage at grade level based on oral reading accuracy. WRMT-R WI = Woodcock Reading Mastery Test–Revised Word Identification (real words) subtest. WRMT-R WA = Woodcock Reading Mastery Test–Revised Word Attack (pseudowords) subtest. TOWRE SW = Test of Word Reading Efficiency (Torgesen, Wagner, & Rashotte, 1999) Sight Word Efficiency (real words) subtest. TOWRE PR = Test of Word Reading Efficiency Phonemic Reading (pseudowords) subtest. GORT-3 Acc = Gray Oral Reading Test, 3rd ed., oral reading accuracy. GORT-3 Rate= Gray Oral Reading Test, 3rd ed., oral reading rate. QRI = Qualitative Reading Inventory, words read per minute in oral reading at grade level. VIQ = Wechsler Intelligence Scale for Children, 3rd ed. (Psychological Corp., 1991), Verbal IQ; prorated score based on Information, Similarities, Vocabulary, and Comprehension subtests. *p < .05. **p < .01. ***p < .001.

340

correlations with comprehension than did predictors involving pseudowords (i.e., phonological decoding). In the fall (Time 1), the best predictors included the rate of oral text reading and realword efficiency for silent reading comprehension fluency; the accuracy and rate of real-word reading and the rate of oral reading for the cloze and openended comprehension measures; the accuracy of single-word reading and the accuracy and rate of oral text reading for multiple choice comprehension; and the accuracy of oral text reading, rate of real word efficiency, and Verbal IQ for criterion-referenced question asking. In the spring (Time 2), the best predictors included the accuracy of single-word reading and the rate of oral text reading for silent reading comprehension fluency and criterionreferenced question answering; and the accuracy of single-word reading and the accuracy and rate of oral text reading for cloze, open-ended, and multiple-choice comprehension measures (see Table 3). The most interesting change was that in the fall, the accuracy of real-word reading was among the best predictors for only two of the comprehension measures, but in the spring, it was among the best predictors for all five comprehension measures. Thus, phonological decoding appears to be a bridge to real-word reading, which, as it improves, increasingly becomes a bridge to reading comprehension. Although phonological decoding may not have as strong or as direct a link with reading comprehension, it appears to enable real-word reading, which in turn enables reading comprehension. Unique Predictors of Reading Comprehension at Time 1 and Time 2. Although all comprehension measures had loaded on the same factor, each of the comprehension measures differed in how strongly it correlated with the factor (see Table 2). Thus, we wondered whether a different set of predictors might uniquely explain each of the comprehension measures. To eval-

JOURNAL OF LEARNING DISABILITIES

uate which of these predictors might contribute uniquely, a series of theorydriven, stepwise regressions that took into account levels of language (Berninger & Richards, 2002) and verbal efficiency theory (Perfetti, 1985) were performed. In keeping with levels of language theory, the unit of language was kept constant (i.e., either word or text) for each comprehension outcome to be predicted. In keeping with the view that word recognition needs to be not only accurate but also automatic, in each set of regressions, accuracy was always entered in the first step, and rate was always entered in the second step. The underlying logic is that fast but inaccurate reading is not automatic. Automatic word reading is both accurate and fast. Fluent text reading is both accurate and fast. The criterionreferenced question-answering measure was not used in these analyses because the dependent measure was categorical rather than quantitative. As summarized in Table 4, both in the fall and in the spring, all regressions were statistically significant for predicting reading comprehension from single real-word reading. In the fall, only the single-word reading rate contributed uniquely to meaning judgment (the only comprehension measure with time limits); only the accuracy of single-word reading contributed uniquely to cloze comprehension. However, both accuracy and rate of single-word reading contributed uniquely to multiple choice and openended comprehension (see Table 4). In the spring, neither accuracy nor rate of single-word reading contributed uniquely to silent reading fluency; only the accuracy of single-word reading contributed uniquely to multiple choice comprehension; and both accuracy and rate of single-word reading contributed uniquely to cloze and open-ended comprehension (see Table 4). In sum, at Time 1 and Time 2, both accuracy and rate of single-word reading contributed to reading comprehension, but the results depended somewhat on the measure of reading comprehension used.

Comparable multiple regression analyses were conducted to predict reading comprehension from text-level accuracy of oral reading entered in the first step and text-level rate of oral reading entered in the second step. The multiple regressions were significant, either at Time 1 or at Time 2, except for the silent reading sentence fluency measure. In the fall, only oral text reading rate contributed uniquely to cloze and open-ended comprehension; no text-level reading skills contributed uniquely to multiple choice comprehension. In the spring, both accuracy and rate of text-level reading contributed uniquely to cloze and openended comprehension, but only the accuracy of text-level reading contributed uniquely to multiple choice comprehension. In sum, in the spring, the accuracy of text-level reading contributed uniquely to reading comprehension, whereas in the fall, it had not. Accuracy and rate of single-word reading may be necessary but not sufficient for reading comprehension; the accuracy of larger units of text than the single word are necessary as comprehension comes to depend increasingly on processing larger chunks of text than the single word. To assess the unique contribution of real-word reading when other potential factors are considered, another set of multiple regressions entered four measures (real-word reading rate, phonological decoding rate, working memory, and Verbal IQ), without a specified order, into multiple regressions for three comprehension outcomes (cloze, multiple choice, and open-ended comprehension). The results, which are summarized in Table 5, showed that in the fall, the realword reading rate uniquely explained all three comprehension outcomes, but in the spring, not only the real-word reading rate but also Verbal IQ uniquely explained all three comprehension outcomes. Moreover, in the spring, the verbal working memory measure also contributed uniquely to one of the comprehension measures. The improved word reading rate from

341

VOLUME 39, NUMBER 4, JULY/AUGUST 2006

TABLE 4 Contribution of Accuracy and Rate of Reading Single Words or Text to Reading Comprehension Item

R2

F

df

p

β

t

p

−.048459 .347246

−0.332 2.376

.7411 .0199

.7392 .1446

0.651 1.888

.0001 .0621

.325845 .248209

2.613 1.990

.0105 .0495

.571943 .277551

6.497 3.153

.0001 .0022

.2189461 .127698

0.652 0.964

.1018 .3377

.636976 .208669

7.466 2.446

.0001 .0163

.380251 .162189

3.148 1.343

.0022 .1826

.456439 .385178

5.115 4.316

.0001 .0001

.187827 .472927

1.286 3.238

.2016 .0017

.283017 .267826

1.756 1.662

.0824 .0999

.072871 .669238

0.563 5.171

.5747 .0000

.476046 .323017

4.069 2.761

.0001 .0069

.333844 .156287

2.088 0.978

.0395 .3308

.340620 .450874

2.871 3.800

.0051 .0003

Stepwise Regressions at Time 1 (Word Level) Meaning Judgment (PAL) WRMT-R Word Identification TOWRE Word Efficiency

.10

Cloze (WRMT-R) WRMT-R Word Identification TOWRE Word Efficiency

.72

Multiple Choice (GORT-3) WRMT-R Word Identification TOWRE Word Efficiency

.28

Open Ended (WIAT) WRMT-R Word Identification TOWRE Word Efficiency

.62

4.48

116.73

18.06

77.04

2, 81

2, 93

2, 92

2, 93

.01

.0001

.0001

.0001

Stepwise Regressions at Time 2 (Word Level) Meaning Judgment (PAL) WRMT-R Word Identification TOWRE Word Efficiency

.10

Cloze (WRMT-R) WRMT-R Word Identification TOWRE Word Efficiency

.63

Multiple Choice (GORT-3) WRMT-R Word Identification TOWRE Word Efficiency

.25

Open Ended (WIAT) WRMT-R Word Identification TOWRE Word Efficiency

.59

5.27

78.35

15.80

67.62

2, 93

2, 93

2, 93

2, 93

.007

.0001

.0001

.0001

Stepwise Regressions at Time 1 (Text Level) Cloze (WRMT-R) GORT-3 Oral Text Reading Accuracy GORT-3 Oral Text Reading Rate

.34

Multiple Choice (GORT-3) GORT-3 Oral Text Reading Accuracy GORT-3 Oral Text Reading Rate

.26

Open Ended (WIAT) GORT-3 Oral Text Reading Accuracy GORT-3 Oral Text Reading Rate

.39

47.87

32.14

61.84

2, 92

2, 92

2, 92

.0001

.0001

.0001

Stepwise Regressions at Time 2 (Text Level) Cloze (WRMT-R) GORT-3 Oral Text Reading Accuracy GORT-3 Oral Text Reading Rate

.55

Multiple Choice (GORT-3) GORT-3 Oral Text Reading Accuracy GORT-3 Oral Text Reading Rate

.21

Open Ended (WIAT) GORT-3 Oral Text Reading Accuracy GORT-3 Oral Text Reading Rate

.50

114.42

25.51

95.75

2, 93

2, 93

2, 93

.0001

.0001

.0001

Note. Accuracy was entered first and rate entered second throughout. PAL = Process Assessment of the Learner Test Battery for Reading and Writing (Berninger, 2001) Sentence Sense subtest; WRMT-R = Woodcock Reading Mastery Test–Revised (Woodcock, 1987); TOWRE = Test of Word Reading Efficiency (Torgesen, Wagner, & Rashotte, 1999); GORT-3 = Gray Oral Reading Test, 3rd ed. (Wiederholt & Bryant, 1992); WIAT = Wechsler Individual Achievement Test (Psychological Corp., 1992).

342

JOURNAL OF LEARNING DISABILITIES

the beginning to the end of second grade may allow individual differences in verbal reasoning ability (and, in some cases, verbal working memory span) to begin to exert their influence on reading comprehension. Neither in the fall nor in the spring did phonological decoding contribute uniquely to any of the reading comprehension outcomes. Its contribution thus appears to be indirect, by mediating realword reading. To evaluate the potential unique contribution of automatic verbal coding of letters and letter clusters to real-

word reading, a final set of multiple regressions entered measures of the rate of naming single letters or letter clusters as predictors of the various measures of real-word reading (accuracy and rate in and out of context). The results, which are summarized in Table 6, show that both at the beginning and at the end of second grade, automatic verbal coding of letters and letter clusters uniquely explained realword reading but not phonological decoding accuracy. In the fall, only the automatic verbal coding of single letters uniquely explained the rate of

reading or decoding single words or the accuracy or rate of real-word reading in context; however, in the spring, automatic verbal coding of both single letters and letter clusters uniquely explained each of those four skills. Profiles of Relative Strengths or Weaknesses in Word Reading and Comprehension. Of the 90 children for whom complete data were available, in fall and spring of second grade, based on the criterion of at least 1⁄2 SD of difference, 9 (10%) showed consistently better cloze reading comprehen-

TABLE 5 Predicting Different Kinds of Comprehension from Phonological Decoding Rate, Word Reading Rate, Verbal Working Memory, and Verbal IQ in Fall and Spring of Second Grade Unique predictors Type of reading comprehension

Fall

Spring

Sentence-level meaning judgment (PAL)

Word reading rate (TOWRE)

Word reading rate (TOWRE)

Cloze (WRMT-R) IQ

Word reading rate (TOWRE)

Word reading rate (TOWRE) and Verbal (WISC-III)

Multiple choice (GORT-3) IQ

Word reading rate (TOWRE)

Word reading rate (TOWRE) and Verbal (WISC-III)

Open-Ended Word Reading (WIAT)

Word reading rate (TOWRE) and Verbal IQ (WISC-III)

Word reading rate (TOWRE), Verbal IQ (WISC-III), and verbal working memory

Note. Results of the multiple regressions are available from the first or second author. PAL = Process Assessment of the Learner Test Battery for Reading and Writing (Berninger, 2001); TOWRE = Test of Word Reading Efficiency (Torgesen, Wagner, & Rashotte, 1999); WRMT-R = Woodcock Reading Mastery Test–Revised (Woodcock, 1987); WISC-III = Wechsler Intelligence Scale for Children, 3rd ed. (Psychological Corp., 1991); GORT-3 = Gray Oral Reading Test, 3rd ed. (Wiederholt & Bryant, 1992); WIAT = Wechsler Individual Achievement Test (Psychological Corp., 1992).

TABLE 6 Predicting Single Word Reading or Text Reading from Rapid Automatic Naming (RAN) of Single Letters or Letter Clusters in Fall and Spring of Second Grade RAN Letters Oral reading skill

RAN Letter Clusters

Fall

Spring

Fall

Spring

Accuracy of Single Real-Word Reading (WRMT-R)

yes

yes

yes

yes

Accuracy of Single-Word Decoding (WRMT-R)

no

no

no

no

Rate of Single Real-Word Reading (TOWRE)

yes

yes

no

yes

Rate of Single-Word Decoding (TOWRE)

yes

yes

no

yes

Accuracy of Text Reading (GORT-3)

yes

yes

no

yes

Rate of Text Reading (GORT-3)

yes

yes

no

yes

Note. Results of the multiple regressions are available from the first or second author. RAN = Process Assessment of the Learner Test Battery for Reading and Writing (Berninger, 2001) Rapid Automatic Naming subtests; WRMT-R = Woodcock Reading Mastery Test–Revised (Woodcock, 1987); TOWRE = Test of Word Reading Efficiency (Torgesen, Wagner, & Rashotte, 1999); GORT-3 = Gray Oral Reading Test, 3rd ed. (Wiederholt & Bryant, 1992).

343

VOLUME 39, NUMBER 4, JULY/AUGUST 2006

sion than oral word identification, and 7 (7.8%) showed consistently better oral word identification than cloze reading comprehension. When the criterion for discrepancy was increased to at least 1 full standard deviation, in the fall and spring of second grade, 2 students (2.2%) showed consistently better cloze reading comprehension than oral word identification, and 3 (3.3%) showed consistently better oral word identification than cloze reading comprehension.

ers, including the accuracy and rate of oral reading of real words, in and out of context; general language comprehension; and, to some extent, working memory. The automaticity of verbal coding of single letters or letter clusters may also contribute to the real-word reading rate, which in turn contributes to reading comprehension in these atrisk readers.

Study 2 Rationale

Discussion Single Underlying Factor in Reading Comprehension. A single underlying factor explained the covariances among the measures of reading comprehension, which varied along many dimensions: (a) the degree to which they tapped text-based and situation-based comprehension (Kintsch, 1998); (b) the nature of the comprehension tasks, some of which are thought to be more authentic than others; and (c) the response requirements. The distinction between text-based and situation-based comprehension (Kintsch, 1998) is useful in understanding skilled comprehension, but it may be less useful in understanding individual differences in reading comprehension among second graders with word decoding, word reading, or fluency problems. The measures seemed to measure reading comprehension regardless of how authentic they were thought to be. The five comprehension measures sometimes differed as to which predictors were unique for each measure (see Tables 3, 4, and 5). However, the single factor underlying the five measures in the confirmatory factor analysis did not change from the beginning to the end of second grade. Unique Predictors of Reading Comprehension. Taken together, the results indicate that although phonological decoding is necessary for learning to read real words, other variables may contribute to the reading comprehension of at-risk second-grade read-

Preschool children acquire much of their early knowledge of aural and oral language through playful interactions with adults, older children, and peers, and they gain their later metalinguistic awareness of oral language through play with words (Garvey, 1990). Playing with language could, therefore, facilitate metalinguistic awareness of written language and early literacy learning (Stahl & Nagy, 2005). Humor is created through “play on words,” and children’s ability to understand humor is a reflection of their verbal comprehension of both spoken and written language (Mahoney & Mann, 1992). Mastering the alphabetic principle requires more than memorizing the rules of correspondence between letters and sounds; it also requires applying the correspondences across many word contexts. Accurate knowledge of the alphabetic principle may be necessary, but it may not be sufficient. Automatic phonological decoding, which is assessed by the rate of phonological decoding, may also be necessary. Reading fluency for text is a product of both (a) automaticity of decoding and single real-word reading, and (b) rapid, smooth coordination of words in context that reflects the prosody or melody of syntax (Kuhn & Stahl, 2003). Reading fluency is important because it influences reading comprehension (Jenkins, Fuchs, van den Broek, Espin, & Deno, 2003; Perfetti, 1985), but reading comprehension can also help to create reading fluency

(Dowhower, 1987). Children who move beyond accurate reading to automatic word reading and fluent text reading are more likely to develop the kinds of comprehension skills needed to meet high-stakes standards in reading that emphasize reading for meaning. To become fluent readers, children may benefit from repeated practice with the same words in many contexts, although this is not the only way to create fluent reading (Kuhn & Stahl, 2003). Creating a variety of ways of practicing the same words repeatedly in different contexts increases the probability that children will remain attentive and engaged—and will not “tune out” due to the repetition. Reading instruction that integrates teaching the alphabetic principle, applying it to word context, engaging in word play, and reading the same words repeatedly in many different contexts may promote reading fluency in a way that benefits reading comprehension.

Method Participants. Teachers in all second-grade classes in 10 elementary schools in a school district administered the Developmental Reading Assessment (Beaver, 1997) at the beginning of second grade to meet state requirements to screen children in order to identify those at risk for failing to meet state standards in reading. This instrument uses graded passages to test the accuracy, rate, and phrasing of oral text reading. Children received a score of 1, 2, 3, or 4 points for each of three categories: accuracy, rate, and phrasing. District criteria were used to classify children as passing, borderline, or failing. Only children who failed the district screening for second-grade reading standards were referred for participation. If their parents gave informed consent, these children completed the battery of tests described later. Altogether, 98 second-grade children (43 girls, 55 boys), whose mean age was 95.6 months (SD = 5.4), were identified as being at risk for reading difficulties. Their WISC-III prorated

344

Verbal IQ fell in the average range (M = 95.1, SD = 13.9). Their mean WRMT-R Word Identification (M = 95.6, SD = 11.3) and Word Attack (M = 94.9, SD = 12.6) scores also fell in the average range. These children would not have been flagged for special education assessment, because their verbal reasoning and reading achievement scores (based on accuracy of singleword reading and single-word decoding) were not discrepant; nevertheless, they were flagged in terms of state standards that recognize the importance of reading fluency in secondgrade readers. Nearly two thirds (65%) of the students were European American; slightly more than one third were from other ethnicities (11% Asian American, 5% Hispanic, 3% African American, 4% Native American, and 11% other). One third of their mothers had completed 4 years or more of college, and another one third had attended community college or vocational school after high school; one fifth had completed only high school, and 11% had completed less than high school. All children had adequate English to participate in the instructional activities, and teachers did not regard them as English language learners (ELLs). Language milestones reported by parents varied as to whether they were achieved early (n = 18), at a typical age (n = 61), or late (n = 14), or were not reported (n = 5). Languages spoken at home varied: English only (n = 68), English and another language (n = 18), no English (n = 5), or not reported (n = 7). Chisquare analysis was used to evaluate a potential relationship between the language milestones and the language spoken at home but was not statistically significant. Children were receiving a variety of services, including Chapter 1 (n = 18), special education (n = 14), speech therapy (n = 12), and physical therapy (n = 3), but no child received more than one of these kinds of services. Experimental Design and Instructional Procedures. Two schools did not participate further because

JOURNAL OF LEARNING DISABILITIES

only one child qualified at each school. The remaining eight schools were randomly assigned to treatment or control groups. The four treatment schools provided reading clubs for children before (two schools) or after school (two schools) twice a week for 1 hour each time from January to early June. The other four schools served as control schools, in which 46 children received only the general reading program and were assessed at the same times as the treatment group children. This general program used the district curriculum, which included some emphasis on phonological awareness and decoding but placed the greatest emphasis on meaning. Of the 47 treatment group children, 8, 11, 13, and 15 children at each of the four schools had complete data before and after the instructional treatment. Of the 46 control children, 3, 7, 16, and 20 children at each of the four schools had complete data at the first and second testing sessions. Teachers for the reading clubs included one teacher from the host school and at least two graduate students from the university research team. The intervention was supplementary (i.e., in addition to the general reading program) and was provided in an extended day model (either before or after school). To motivate the children to spend extra time on reading before or after school, the intervention was presented to the children as a club. Children had to whisper the secret password chosen by each club and have their hand stamped with a special club stamp to gain entrance each time the club met. The schedule for each club included initial word play (with riddles and jokes, sounds, and letters), word work (accuracy and automaticity of the alphabetic principle and its application to word context), story reading, and final word play (bingo for structure words and search for long words). Taken together, the word play, word work, and story reading included instructional components directed to subword, word, and text levels of language necessary for creating coordinated functional reading systems in the learner’s mind. Thus, it in-

tegrated several levels of language and many components of reading. Word Play. Five kinds of activities contributed to word play. First, at the beginning of each session, children selected riddles and jokes from a collection provided by the teachers and read them to each other. This activity was called You Got to Laugh. Children were also encouraged to collect their own jokes and riddles. Once a month, the children voted on their favorite jokes of the month by giving them scores on a “laugh-o-meter.” Second, they played Sound Games, in which they listened for target sounds, omitted target sounds, said the words without target sounds, and substituted sounds for target sounds in long, complex, interesting spoken words. Third, they played Looking Games, in which they carefully examined all the letters in a word and then wrote target letters in specific positions designated by the teacher. Fourth, at the end of each session, children played bingo with the structure words of the language (i.e., prepositions, pronouns, conjunctions, and articles that are the glue words that contribute to sentence meaning), which are very frequent in the language but are very difficult for at-risk readers (possibly because they have no meaning of their own). Fifth, at the end of each session, children discussed the long words (Mommy Long Words) that they were encouraged to find outside the club sessions and keep in their work folder. They discussed how these were like the daddy longlegs arachnid (many words could spin off from one body or base, just like the arachnid’s legs do) and different from the arachnid (sometimes, the word parts were not spinoffs of the same main body or meaning but, rather, new sound units). At the end of each session, the children were given homework assignments to search for a Mommy Long Word with specific characteristics (e.g., the longest word you can find that has an -ing or -ion at the end, or the longest word you can find that has five syllables). From time to time, one of the teachers dressed up in a Dr. Mrs. Seuss-Goose costume (Cat-in-the-Hat top hat and

345

VOLUME 39, NUMBER 4, JULY/AUGUST 2006

bow tie for Dr. Seuss, and feather mask and wings for Mother Goose) and gave out treats to children who had found words to fit each assignment and to those who found the longest words. All together, these activities provided word play with the meaning, sound, and letters in words. Word Work. During the word work part of the session, children participated in either Zip or Zap groups (to which they had been randomly assigned). The Zip group used accuracy criteria for (a) learning the correspondences between spelling units (oneand two-letter units) and phonemes in the alphabetic principle, and (b) assessing the application of these correspondences to word context. The Zap group used rate criteria for learning and assessing the correspondences. Pseudowords were used for the probes to evaluate whether children were transferring the application of the correspondences in alphabetic principle across different pseudoword contexts rather than relying on a representation of a familiar word in memory. While teachers worked with individual children to collect probes and provide individualized feedback, other children engaged in independent writing and illustrating activities, which they shared with other children in their small working groups. To introduce pseudowords (nonwords that have spellings and sounds but no conventional meaning), teachers read Dr. Seuss’s If I Ran the Zoo (Geisel, 1950) and discussed how novel names or words can be created by playful recombining of the sounds of the language. For example, some of the names of the animals in this zoo were flustard, joat, gootch, nerkle, mulligatawny, and fizza-ma-wizza-ma-dill. Children created their own “pseudoname,” which was written on their work folder. In this way, pseudowords, which require children to attend to letters and sounds without meaning cues, were used to encourage focused (but playful) work on the relationships between the sounds and letters in words. Story Reading. Early Success Readers Level 2 (Cooper, Pikulski, & Au,

1996), which contained high-interest, intellectually engaging texts for second graders, were used for story reading. In each session, the children read two stories. One was a new story that the children and teacher read in parallel like a choral reading. One was the story introduced in the prior session that each child had practiced reading and rereading with a partner during club time. Children were given a summary of this story to take home and read again and again with a parent. The following week, each child had a chance to reread that same story to the teacher, who recorded the child’s reading rate for a sample of 100 words in the story. Test Battery. Accuracy of phonological decoding and real-word reading. The WRMT-R Word Attack subtest was used as a measure of phonological decoding, as in Study 1. Similarly, the WRMT-R Word Identification subtest was used as a measure of real-word reading accuracy out of context. Reading fluency. The Developmental Reading Assessment (DRA; Beaver, 1997) was administered at the beginning and the end of second grade using graded passages. A score of 1, 2, 3, or 4 points was given for each of three skills contributing to fluency—accuracy, rate, and phrasing—and, on the basis of district criteria, children were classified as passing (2), borderline (1), or failing (0). Reading comprehension. The Gates– MacGinitie Reading Comprehension Test (GMG; MacGinitie & MacGinitie, 1989) was administered, in which children read graded paragraphs and answered multiple choice questions about the passages. Questions assessed both factual understanding of the text and inferential thinking (i.e., going beyond what was stated in the text). Accuracy scores for grade are based on normal curve equivalents, with a mean of 50 and a standard deviation of approximately 22 standard score points. Data Analyses. Comparing club and control children. Repeated-measures ANOVA was used to evaluate group

effects, time effects, and Group × Time interactions for each of the dependent measures in the test battery. Scores for individual children across schools were entered into these analyses instead of scores for instructional groups, because schools (treatment vs. control) and instructional groups were confounded. Consistent discrepancies between word reading and reading comprehension. Profiles were examined for WRMT-R Word Identification and Gates-MacGinitie Reading Comprehension (transformed to the same standard score scale as the WRMT-R). Children who showed higher word reading than comprehension or higher comprehension than word reading were identified on the basis of two criteria: (a) allowing the difference to vary freely and then describing the mean difference between the two reading skills, and (b) identifying those cases in which the discrepancy fell at or above a cutoff (0.8 SD).

Results Accuracy of Phonological Decoding. Only the Group × Time interaction was statistically significant, F(1, 81) = 7.89, p = .006. In the fall, the club (M = 93.65, SD = 11.58) and control children (M = 97. 58, SD = 10.19) were not significantly different, but in the spring, the club children (M = 96.35, SD = 10.29) had improved significantly more on average (by 2.7 standard score points) than the control children (M = 96.74, SD = 9.79) on average (by −0.8 standard score points) on WRMT-R Word Attack. Accuracy of Real-Word Reading. The club children did not improve significantly more than the control children on WRMT-R Word Identification. Transfer of Rate-Based Training in Alphabetic Principle and Phonological Decoding to Real-Word Reading. Children meeting the criteria for dyslexia (i.e., word reading or decoding unexpectedly low based on Verbal IQ and age) who received training and feedback using rate-based criteria im-

346

proved significantly more in real-word reading than those receiving training and feedback using accuracy-based criteria, F(1, 12) = 7.5, p = .018, even though both groups improved significantly over time in real-word reading, F(1, 12) = 9.39, p = .01. For the sample as a whole, after treatment, the rate of phonological decoding alone on TOWRE (which was available only after treatment) explained 60.3% of the variance in accuracy of real word reading (p < .001). Reading Fluency. Both the club and control children improved significantly over time in the total score on the DRA, F(1, 83) = 139.91, p = .0001, but the club children improved significantly more than the control children, as evidenced by the significant Group × Time interaction, F(1, 83) = 4.06, p = .0472. Although control children started out significantly higher on this measure, on which 12 points would be a perfect score (M = 8.31, SD = 2.52), than the club children (M = 5.63, SD = 2.25), the club children had increased significantly more after treatment (M = 9.70, SD = 2.17) than did the control children (M = 11.20, SD = 2.36). However, when the children were evaluated on the basis of whether individual students met criteria for satisfactory performance (2), below standard (1), or borderline (0), there were no pretreatment differences, but a significant time effect for all children (from M = 0.54, SD = 0.59 to M = 1.52, SD = 0.81), F(1, 84) = 142.07, p = .0001, and a significant Time × Group interaction, F(1, 84) = 4.34, p = .0402, occurred. Club children increased more after treatment (M = 1.68, SD = 0.61) than did the control children (M = 1.38, SD = 0.58). Reading Comprehension. Only the time effect was significant, F(1, 78) = 6.249, p = .015. Gates-MacGinitie reading comprehension increased for the club and control children (from M = 33.59, SD = 1.70, to M = 37.40, SD = 1.79). However, if WISC III oral vocabulary or TOWRE pseudoword

JOURNAL OF LEARNING DISABILITIES

reading were used as a covariate in this analysis, none of the effects— including the time effect—was significant for reading comprehension. These results show that individual differences in oral vocabulary or rate of phonological decoding were influencing reading comprehension achievement until they were statistically controlled to eliminate their effects. Consistent Relative Weaknesses in Reading Comprehension or Word Decoding. Overall, 22.6% of these atrisk second-grade readers had a consistent profile near the beginning and at the end of second grade, in which their reading comprehension was less well developed than their word reading. On average, the difference was 11.3 standard score points early in the school year and 12.7 points at the end of the school year. However, if the size of the persisting discrepancy was set at 0.8 SD or more, 9.5% had average discrepancies of 17.8 (SD = 6.9) standard score points early in the school year, and 19.9 (SD = 5.6) standard score points later in the school year. In contrast, 45.2% had a consistent profile both at the beginning and at the end of the year, in which word decoding was less developed than reading comprehension, by 12.8 standard score points at the beginning and by 11.5 standard score points at the end of the school year. However, if the size of the persisting discrepancy was set at 0.8 SD or more, 14.3% had mean discrepancies of 18.2 (SD = 5.6) early in the school year, and 19.3 (SD = 6.2) later in the school year, when their reading comprehension was better developed. These discrepancies for the two components of the simple view of reading (Gough & Hillinger, 1980)— word reading and comprehension (see Note 1)—may have practical implications for reading instruction in the classroom. Supplementary training in oral vocabulary meaning may be beneficial when comprehension lags behind (as in nearly one out of four at-risk second-grade readers). Supplementary training in accuracy and rate

of phonological decoding may be beneficial when word reading lags behind (as in nearly one out of two at-risk second-grade readers).

Discussion Both treated (club) and control children improved significantly in reading comprehension, but when statistical controls for pretreatment differences in oral vocabulary knowledge were introduced, statistical effects for improved reading comprehension disappeared. This finding suggests that for at-risk readers at this developmental stage, individual differences in oral vocabulary could interfere directly with the development of either word reading or reading comprehension and may influence whether and how children respond to reading comprehension instruction. Both the results for individual differences (Study 1) and those for instruction (Study 2) support a model in which sequential steps in learning written language could contribute to developing reading comprehension. These steps progress from the accuracy of the alphabet principle and phonological decoding, which are steppingstones to the rate of phonological decoding, which in turn is a steppingstone to the rate of real-word reading, which in turn is a stepping-stone to the rate of text reading. This model is consistent with the findings of Uhry and Shepherd (1997), who showed that accuracy of phonological skills is necessary for achieving the next instructional goal for oral reading—accuracy of single real-word reading. These steps may be literacy milestones (much as infants and toddlers achieve milestones for oral language and motor skills), but they do not imply that each of the bridging skills should be taught until it is mastered before moving on to the next steppingstone. Integrated instruction aimed at all reading skills should be provided all along the journey to skilled reading comprehension. The lesson from Study 2 is that instruction that integrated decoding and reading comprehension im-

VOLUME 39, NUMBER 4, JULY/AUGUST 2006

proved the accuracy of phonological decoding significantly more than the general instruction in the control group, replicating the findings of another instructional study with at-risk second-grade readers (Berninger et al., 2003). However, these stepping-stones may represent the typical milestones that must be achieved for the journey to reading comprehension to proceed smoothly. These milestones, along with vocabulary knowledge, may offer diagnostic and treatment clues for children who struggle with reading comprehension. For example, when statistical controls for pretreatment differences in phonological decoding were used, statistical effects for improved reading comprehension disappeared. Phonological decoding may exert its constraints early in the sequence but nevertheless generate substantial effects downstream on reading comprehension. For some children, phonological decoding may be the bottleneck to treat in eliminating their reading comprehension problems. Training phonological decoding with ratebased criteria transferred to improved real-word reading, which may be a link to reading comprehension, because it draws on vocabulary meaning as well as phonological decoding. Integrated instruction aimed at both automatic word recognition and reading comprehension skills also significantly improved text reading fluency (based on total score and district criteria for state standards) in treated (club) children as compared to control children. Fluency has been shown in much previous research to be a bridge to reading comprehension (e.g., Kuhn & Stahl, 2003; Stahl, Heubach, & Crammond, 1997).

General Discussion Instructional Paths to Reading Comprehension Path From Vocabulary to Reading Comprehension. Study 1 pro-

vided direct evidence and Study 2 provided indirect evidence for a path from vocabulary to reading comprehension. In Study 1, Verbal IQ was significantly correlated with four of the reading comprehension measures at the beginning and at the end of second grade (see Table 3) and uniquely predicted reading comprehension on one measure at the beginning of second grade and on all measures analyzed at the end of second grade (see Table 5). In Study 2, using the WISC-III Vocabulary subtest, which is highly correlated with overall Verbal IQ (Psychological Corp., 1991), as a covariate, eliminated the treatment effect for improved reading comprehension; individual differences in vocabulary knowledge appeared to be mediating reading comprehension achievement. Verbal IQ may be a predictor of early reading comprehension because the WISC-III subtests that are used to assess this construct also assess the key components of vocabulary learning (see Note 2)—the retrieval of world knowledge or background knowledge (Information subtest); the ability to use decontextualized language to explain word meaning (Vocabulary subtest); the ability to detect similarities and differences among lexical items in a semantic network (Similarities subtest); and the ability to infer meaning from real-world contexts (Comprehension subtest). Thus, a weakness in oral vocabulary knowledge may compromise the development of reading comprehension. If so, the instructional program should also assess and teach oral vocabulary and reading vocabulary for meaning and not just for pronunciation of written words (see Carlisle & Rice, 2002, and Stahl & Nagy, 2005, for research-supported approaches to teaching vocabulary). Path From Written Language to Reading Comprehension. Combining the findings of Studies 1 and 2 suggests that there may be bridging links between steps in learning written language to reading comprehension that

347

have diagnostic and instructional implications. These stepping-stones, listed in sequential order of mastery, all involve the creating of mapping relationships between spoken and written language: • knowledge of the alphabet principle of phoneme–spelling correspondences → • accurate phonological decoding (applying the alphabetic principle to word context) → • automatic phonological decoding → • accuracy of real-word reading → • automatic real-word reading → • accurate oral reading of text → • fluent oral reading of text. Although group instruction should be aimed at all levels of language close in time, teachers may have to individualize instruction, as necessary, for individual children with difficulties in accuracy of phonological decoding, rate of phonological decoding, accuracy of real-word reading, rate of realword reading, accuracy of reading text, or rate of reading text. Such individualization of instruction may benefit from assessment information that pinpoints why individual children do not show reasonable annual growth in reading comprehension. The goal of this assessment is to pinpoint which of several potential processes contributing to reading comprehension is causing problems for specific children: (a) vocabulary knowledge (and, if so, whether background knowledge, metalinguistic knowledge, conceptual structures linking words, or inferencing ability is the bottleneck); (b) automatic verbal coding of single letters and letter clusters; (c) accuracy or rate of phonological decoding of pseudowords and of real-word reading; or (d) accuracy or rate of text reading. Just because there was no treatment-specific effect for reading comprehension outcomes in Study 2, one should not conclude that any instructional approach to reading comprehen-

348

sion will be effective. Once the bottleneck is identified, one may refer to Henry (2003) for word decoding and word reading strategies, Carlisle and Rice (2002) for research-supported instructional activities to develop comprehension skills, and Stahl and Nagy (2005) for research-supported instructional activities to teach vocabulary meaning. Both phonological training (e.g., Foorman, Francis, Fletcher, Schatschneider, & Mehta, 1998; Torgesen, Wagner, & Rashotte, 1997; Vellutino et al., 1996) and explicit vocabulary (Stahl & Nagy, 2005) and comprehension strategies (e.g., Carlisle & Rice, 2002; Englert, Tarrant, Mariage, & Over, 1994; Pressley & Wharton-MacDonald, 1997) may improve reading comprehension.

Relationship of Reading Fluency and Comprehension In contrast to earlier findings in adults, for whom reading efficiency (rate) was unrelated to reading comprehension (Walczyk, 2000), in Study 1, both accuracy and rate of word-level and textlevel oral reading were correlated significantly with (see Table 3) and contributed unique variance to (see Table 4) reading comprehension in atrisk second-grade readers. In contrast to fourth graders, for whom reading fluency in context was more related to reading comprehension than reading fluency on lists (Jenkins et al., 2003), for the at-risk second graders in Study 1, the efficiency of reading words both in and out of context was significantly and uniquely related to reading comprehension. Study 2 also demonstrated that instruction that integrated phonological decoding, real-word reading, text reading, and reading comprehension improved reading fluency in atrisk second-grade readers. A substantial literature now demonstrates that automatic single-word reading and fast, efficient, smooth, and coordinated text reading that reflects the prosody (musical melody) of the language enables reading comprehension (Biemiller, 1977–1978; Biemiller & Siegel,

JOURNAL OF LEARNING DISABILITIES

1997; Breznitz, 1997a, 1997b; Downhower, 1987; Kuhn & Stahl, 2000, 2003; Levy, Abello, & Lysynchuk, 1997; Perfetti, 1995; Rasinski, Padak, Linek, & Sturtevant, 1994; Reutzel & Hollingsworth, 1993; Samuels, 1985; Stahl et al., 1997; Tan & Nicholson, 1997).

Importance of Language Profiles in At-Risk Readers Relative weaknesses in word-level reading compared to text-level comprehension or relative weaknesses in text-level comprehension compared to word-level reading may both influence the ease with which children acquire literacy skills. The average reading skills in word reading and comprehension that were reported for pretreatment children in Study 2 masked the fact that some children had profiles in which one of these skills was substantially better developed than the other, and this unevenness may have influenced how they responded to instruction designed to improve their reading comprehension. The two components of the simple view of reading (Gough & Hillinger, 1980)—word reading and comprehension—may exert their influences jointly, as a function of the profile in which they are embedded, rather than as single, independent variables.

Developmental Changes in Reading Comprehension Knowledge of the world, knowledge of language, and understanding of test requirements are all factors that contribute to the assessment of beginning reading comprehension. Widely used measures of reading comprehension do appear to tap a general comprehension factor, at least in at-risk secondgrade readers. However, although the various reading comprehension measures shared a common factor, they were not perfectly correlated. Different comprehension tasks do require different kinds of comprehension skills, and these differences are likely to increase across schooling, as the nature of reading comprehension and curriculum re-

quirements changes qualitatively and dramatically as children journey toward skilled reading. The results of Study 1 showed that reading fluency is necessary for developing comprehension skills in atrisk second graders, but fluency alone may not be sufficient for developing all the kinds of comprehension skills that will be necessary in the curriculum beyond second grade (Carlisle & Rice, 2002). The growth trajectory for many academic skills transcends a single school year. The slowly evolving nature of reading comprehension between novice and skilled reading needs to be charted throughout schooling. Whether a single underlying factor (as observed in Study 1) explains reading comprehension throughout this journey and across all tests of reading comprehension is a question that requires further research.

Future Research Perspectives Research is also needed on whether reading comprehension develops differently in students who do and do not have significant discrepancies between their word reading and reading comprehension skills (Berninger & O’Donnell, 2004). In contrast to studies that have observed such discrepancies post hoc (e.g., Study 2), research is needed in which such groups are preselected and matched on other relevant variables before the instructional intervention begins. Likewise, research is needed on how the two paths to reading comprehension—vocabulary and written language (each with its own stepping stones)—may converge in learning to read real words. Although written words are single units, they point to mental representations that are connected to many other representations. How children’s phonological, morphological, and orthographic representations underlying written word learning relate to the rich semantic network of word meaning and world knowledge is far from fully understood, but it probably has important instructional

349

VOLUME 39, NUMBER 4, JULY/AUGUST 2006

implications. Some children have more difficulty with learning to recognize real words than pseudowords, and our research group is trying to understand why. The results for the at-risk second graders we studied may not generalize to able readers whose reading comprehension is developing typically in second grade. Many of the debates and outright disagreements within the reading community have been the result of researchers or practitioners talking about different populations. These populations differ in the degree to which they require explicit reading instruction (Berninger & Richards, 2002) and are able to deal with decontextualized language. Explicit instruction refers to the degree to which instruction creates conscious awareness and manipulation of the units of language; most of our knowledge of language is implicit, on the unconscious channel. Language is first learned in contextualized conversation (Garvey, 1990), but the transition to school requires children to learn to use decontextualized language to understand teachers’ instruction and discover the relationships between the aural and oral language that children already know and the written language they are learning (see Note 3). The studies reported here to do not address how typically developing second-grade readers develop their reading comprehension skills. These studies were aimed at a population of second graders who struggled in reading comprehension, and the results should be generalized only to this population. These are the children who are the most in need of researchsupported approaches—approaches based not on the belief systems of reading professionals but, rather, on the results of scientific research.

Conclusion Struggling readers have to cross many bridges in their path to overcome risk factors and achieve strong reading comprehension skills (Biemiller &

Siegel, 1997). One bridge helps them proceed along the path from vocabulary knowledge to reading comprehension. Another bridge helps them proceed along the path from written language to reading comprehension. To ensure that as many at-risk secondgrade readers as possible meet the high-stakes standards in reading comprehension, it is important that their instruction focus on vocabulary development, knowledge of the alphabetic principle, accuracy and rate of phonological decoding, accuracy and rate of real-word reading, and accuracy and rate of text reading as well as specific comprehension strategies. This is the real whole language. ABOUT THE AUTHORS

Virginia W. Berninger, PhD, is a professor of educational psychology at the University of Washington, where she teaches in the school psychology program. She is principal investigator of NICHD-funded research on normal writing development and its connections with reading and oral language; early intervention for reading and writing; and multidisciplinary learning disabilities—their genetic and brain basis, behavioral diagnoses, and effective treatment. Robert D. Abbott, PhD, is a professor of educational psychology at the University of Washington where he teaches in the measurement, statistics, and research design program. He is Principal Investigator of the Statistics Core of NICHD-funded research on multidisciplinary learning disabilities and autism. Karin Vermeulen, MEd, is currently employed by Fairfax County Public Schools in Virginia as a school psychologist. Her interests include early intervention programs for reading and writing, the response to intervention eligibility for specific learning disorders, and culturally competent assessment. Cynthia Fulton, following 14-year career as an elementary teacher and literacy specialist, recently made a transition to the field of holistic healthcare, where she educates clients about health and wellness in her clinic in Shelburne, Vermont. Address: Virginia W. Berninger, 322 Miller, Box 353600, University of Washington, Seattle, WA 98195-3600; e-mail: [email protected] AUTHORS’ NOTES

1. This research was supported by P50 3381205, a Multidisciplinary Learning Disabili-

ties Center Grant from the National Institute of Child Health and Human Development. 2. We are grateful to the participating secondgrade children in Study 1 at Adams, Montlake, Olympic View, and Sacajawea in the Seattle Public Schools; Briarcrest, Meridian Park, and Ridgecrest in the Shoreline Public Schools; and St. Luke’s in the Archdiocese of Western Washington. We are also grateful to the participating second-grade children and their teachers in Study 2 at Briarcrest, Echo Lake, Lake Forest Park, Meridian Park, North City, Parkwood, Ridgecrest, Sunset, and Syre in the Shoreline Public Schools. We also acknowledge the contribution of Dr. Mark Jewell, Chief Academic Officer, Federal Way School District, for facilitating the study design for Study 2 when he was in the Shoreline School District. We also thank Joseph Torgesen, Richard Wagner, and Carol Rashotte for the use of the prepublication version of the TOWRE real-word and pseudoword reading rate measures; Doug Hacker for recommending a study of text-based and situation-based comprehension in young atrisk readers; and Joanne Carlisle, William Nagy, and Richard Wagner for helpful discussions on the role of vocabulary in learning to read. NOTES

1. The word reading component is variously interpreted as decoding or automatic word recognition, depending on whether a student is at a beginning or a later stage of word reading, and the comprehension component is variously interpreted as listening comprehension or reading comprehension, depending on whether a student is at a beginning or a later stage of decoding and word recognition. 2. These key requirements for vocabulary learning are based on ideas presented at an international conference on vocabulary organized by Richard Wagner in February 2004 (Wagner, Muse, & Tannenbaum, in press) and discussed by Joanne Carlisle, William Nagy, and Richard Wagner in a phone conference in March 2004, with the first author, who was unable to attend. 3. The authors thank William Nagy, who joined our research center as a visiting scientist after the studies reported in this article were conducted, for helping us understand how the two issues of explicitness and decontextualization, which are axiomatic principles in child language and linguistic research, are the sources of continuing mis-

350

understandings within the community of reading professionals. REFERENCES

Beaver, J. (1997). Developmental reading assessment. Parsippany, NJ: Pearson Learning Group. Bentler, P., & Wu, E. (2000). EQS program. Los Angeles: Multivariate Software. Berninger, V. (2001). Process assessment of the learner (PAL) test of reading and writing. San Antonio, TX: Psychological Corp. Berninger, V., Abbott, R., & Alsdorf, B. (1997). Lexical- and sentence-level processes in comprehension of written sentences. Reading and Writing: An Interdisciplinary Journal, 9, 135–162. Berninger, V., Abbott, R., Billingsley, F., & Nagy, W. (2001). Processes underlying timing and fluency: Efficiency, automaticity, coordination, and morphological awareness. In M. Wolf (Ed.), Dyslexia, fluency, and the brain (pp. 383–414). Baltimore: York Press. Berninger, V., Abbott, R., Vermeulen, K., Ogier, S., Brooksher, R., Zook, D., et al. (2002). Comparison of faster and slower responders: Implications for the nature and duration of early reading intervention. Learning Disability Quarterly, 25, 59– 76. Berninger, V., & O’Donnell, L. (2004). Research-supported differential diagnosis of specific learning disabilities. In A. Prifitera, D. Saklofske, L. Weiss, & E. Rolfhus (Eds.), WISC-IV Clinical use and interpretation (pp. 189–233). San Diego: Academic Press. Berninger, V., & Richards, T. (2002). Brain literacy for educators and psychologists. San Diego: Academic Press/Elsevier. Berninger, V., Vermeulen, K., Abbott, R., McCutchen, D., Cotton, S., Cude, J., et al. (2003). Comparison of three approaches to supplementary reading instruction for low achieving second grade readers. Language, Speech, and Hearing Services in Schools, 34, 101–115. Biemiller, A. (1977–1978). Relationship between oral reading rates for letters, words, and simple text in the development of reading achievement. Reading Research Quarterly, 13, 223–253. Biemiller, A., & Siegel, L. (1997). A longitudinal study of the effects of the Bridge Reading Program for children at-risk for reading failure. Learning Disability Quarterly, 20, 83–92.

JOURNAL OF LEARNING DISABILITIES

Breznitz, Z. (1997a). Enhancing the reading of dyslexics by reading acceleration and auditory masking. Journal of Educational Psychology, 89, 103–113. Breznitz, Z. (1997b). The effect of accelerated reading rate on memory for text among dyslexic readers. Journal of Educational Psychology, 89, 287–299. Carlisle, J., & Rice, M. (2002). Improving reading comprehension. Research-based principles and practices. Timonium, MD: York Press. Cooper, J., Pikulski, J., & Au, K. (1996). Early success. An intervention program (Level 2). Boston: Houghton Mifflin. Dowhower, S. (1987). Effects of repeated reading on second-grade transitional readers’ fluency and comprehension. Reading Research Quarterly, 22, 389–406. Ehri, L., Nunes, S., Stahl, S., & Willows, D. (2001). Systematic phonics instruction helps students learn to read: Evidence from the National Reading Panel’s metaanalysis. Review of Educational Research, 71, 393–447. Englert, C., Tarrant, K., Mariage, T., & Over, T. (1994). Lesson talk as the work of reading groups: The effectiveness of two interventions. Journal of Learning Disabilities, 27, 165–185. Foorman, B., Francis, D., Fletcher, J., Schatschneider, C., & Mehta, P. (1998). The role of instruction in learning to read: Preventing reading failure in at-risk children. Journal of Educational Psychology, 90, 1–15. Garvey, C. (1990). Play (Developing Child Series). Cambridge, MA: Harvard University Press. Geisel, T. S. (Dr. Seuss). (1950). If I ran the zoo. New York: Random House. Gough, P., & Hillinger, M. (1980). Learning to read: An unnatural act. Bulletin of the Orton Society, 30, 179–196. Henry, M. (2003). Unlocking literacy: Effective decoding and spelling instruction. Baltimore: Brookes. Jenkins, J., Fuchs, L., van den Broek, P., Espin, C., & Deno, S. (2003). Sources of individual differences in reading comprehension and reading fluency. Journal of Educational Psychology, 95, 719–729. Juel, C. (1988). Learning to read and write: A longitudinal study of 54 children from first through fourth grades. Journal of Educational Psychology, 80, 437–447. Kintsch, W. (1998). Comprehension. A paradigm for cognition. Cambridge, UK: Cambridge University Press.

Kuhn, M., & Stahl, S. (2000, March). Fluency. A review of developmental and remedial practices (CIERA Report # 2-008). Athens: University of Georgia, Center for the Improvement of Early Reading Achievement. Kuhn, M., & Stahl, S. (2003). Fluency: A review of developmental and remedial practices. Journal of Educational Psychology, 95, 3–21. Leslie, L., & Caldwell, J. (1990). Qualitative reading inventory. Glenview, IL: Scott Foresman/Little, Brown. Levy, B., Abello, B., & Lysynchuk, L. (1997). Transfer from word training to reading in context: Gains in reading fluency and comprehension. Learning Disability Quarterly, 20, 173–188. MacGinitie, W., & MacGinitie, R. (1989). Gates-MacGinitie reading tests. Itasca, IL: Riverside. Mahoney, D., & Mann, V. (1992). Using children’s humor to clarify the relationship between linguistic awareness and early reading ability. Cognition, 45, 163–186. Nagy, W. (in press). Metalinguistic awareness and the vocabulary–comprehension connection. In R. K. Wagner, A. Muse, & K. Tannenbaum (Eds.), Vocabulary acquisition and its implications for reading comprehension. New York: Guilford. National Reading Panel. (2000). Teaching children to read: An evidence-based assessment of the scientific research literature on reading and its implications for reading instruction. Washington, DC: National Institute of Child Health and Human Development. Oakhill, J., & Yull, N. (1996). Higher order factors in comprehension disability: Processes and remediation. In C. Cornoldi & J. Oakland (Eds.), Reading comprehension difficulties: Processes and intervention (pp. 69–92). Mahwah, NJ: Erlbaum. Perfetti, C. (1985). Reading ability. New York: Oxford University Press. Pressley, M., & Wharton-MacDonald, R. (1997). Skilled comprehension and its development through instruction. School Psychology Review, 26, 448–466. Psychological Corp. (1991). Wechsler intelligence scale for children (3rd ed.). San Antonio, TX: Author. Psychological Corp. (1992). Wechsler individual achievement test. San Antonio, TX: Author. Rasinski, T., Padak, N., Linek, W., & Sturtevant, E. (1994). Effects of fluency devel-

VOLUME 39, NUMBER 4, JULY/AUGUST 2006

opment on urban second-grade readers. The Journal of Educational Research, 87, 158–165. Reutzel, D., & Hollingsworth, P. (1993). Effects of fluency training on second graders’ reading comprehension. The Journal of Educational Research, 86, 325–331. Samuels, S. (1985). Automaticity and repeated reading. In J. Osborn, P. Wilson, & R. Andeson (Eds.), Reading education: Foundations for a literate America (pp. 215– 230). Lexington, MA: Lexington Books. Stahl, S., Heubach, K., & Crammond, B. (1997, Winter). Fluency-oriented reading instruction (Reading Research Report No. 79). Athens, GA: National Reading Research Center Project of the University of Georgia and University of Maryland. Stahl, S., & Nagy, W. (2005). Teaching word meaning. Mahwah, NJ: Erlbaum. Stanovich, K. (1986). Matthew effects in reading: Some consequences of individual differences in the acquisition of liter-

acy. Reading Research Quarterly, 21, 360– 407. Swanson, H. L., & Berninger, V. (1995). The role of working memory in learning disabled readers’ comprehension. Intelligence, 21, 83–108. Tan, A., & Nicholson, T. (1997). Flashcards revisited: Training poor readers to read words faster improves their comprehension of text. Journal of Educational Psychology, 89, 276–288. Torgesen, J., Wagner, R., & Rashotte, C. (1997). Prevention and remediation of severe reading disabilities. Scientific Studies of Reading, 1, 217–234. Torgesen, J., Wagner, R., & Rashotte, C. (1999). Test of word reading efficiency. Austin, TX: PRO-ED. Uhry, J., & Shepherd, M. (1997). Teaching phonological recoding to young children with phonological processing deficits: The effect on sight-vocabulary acquisition. Learning Disability Quarterly, 20, 104–125.

351

Vellutino, F., Scanlon, D., Sipay, E., Small, S., Pratt, A., Chen, R., et al. (1996). Cognitive profiles of difficult-to-remediate and readily remediated poor readers: Early intervention as a vehicle for distinguishing between cognitive and experiential deficits as basic causes of specific reading disability. Journal of Educational Psychology, 88, 601–638. Wagner, R. K., Muse, A., & Tannenbaum, K. (Eds.). (in press). Vocabulary acquisition and its implications for reading comprehension. New York: Guilford. Walczyk, J. (2000). The interplay between automatic and control processes in reading. Reading Research Quarterly, 35, 554– 566. Wiederholt, J., & Bryant, B. (1992). Gray oral reading test (3rd ed.). Odessa, FL: Psychological Assessment Resources. Woodcock, R. (1987). Woodcock reading mastery tests–Revised. Circle Pines, MN: American Guidance Service.