Comparison of faster and slower responders to early intervention in reading: differentiating features of their language profiles.

In the longitudinal research reported in this article we examine issues related to the nature and duration of early intervention of relatively low intensity. Specifically, we consider (a) how children who respond faster and slower differ at the end of a 24-lesson intervention spaced over four months in first grade, (b) whether the faster responders maintained their gains during second grade, (c) whether the slower responders could make relative gains with additional intervention, and (d) how the faster and slower responders during the first-grade intervention differed in second grade. In comparing these children who varied in rate of response to early intervention, we focus on differences in their language profiles for potential instructional cues for enhancing the reading of slower responders.

**********

A question driving much of the research on at-risk beginning readers is what conditions must be in place for these children to develop adequate reading skills (e.g., Lyon, Alexander, & Yaffe, 1997; Torgesen, 2000). Considerable research has confirmed that explicit instruction in phonological awareness and phonological decoding is an essential component of instruction in early intervention (e.g., Footman, Francis, Fletcher, Schatschneider, & Mehta, 1998; Torgesen, Wagner, & Rashotte, 1997; Torgesen, Wagner, Rashotte, & Herron, 1999; Tunmer & Nesdale, 1985; Vadasy, Jenkins, & Pool, 2000; Vellutino, Scanlon, Sipay, Small, Pratt, Chen, & Denckla, 1996). Less is known about why some at-risk readers respond rapidly to early intervention, while others do not respond or respond very slowly. For example, about 2% to 6% of beginning readers do not appear to respond to early intervention that emphasizes phonological awareness and phonological decoding in the context of balanced reading instruction (Torgesen, 2000). Also, little is known about the required intensity and duration of early intervention for at-risk beginning readers with well-defined characteristics.

In the longitudinal research reported in this article we examine issues related to the nature and duration of early intervention of relatively low intensity. (1) Specifically, we consider (a) how children who respond faster and slower (2) differ at the end of a 24-lesson intervention spaced over four months in first grade, (b) whether the faster responders maintained their gains during second grade, (c) whether the slower responders could make relative gains with additional intervention, and (d) how the faster and slower responders during the first-grade intervention differed in second grade. In comparing these children who varied in rate of response to early intervention, we focus on differences in their language profiles for potential instructional cues for enhancing the reading of slower responders.

Berninger, Abbott et al. (2000) reported comprehensive details about the first year of this longitudinal study: (a) the first-grade sample drawn from eight schools in four school systems (inclusion criteria, selection procedures, descriptive statistics); (b) the instructional procedures and materials for the first-grade intervention; (c) repeated measures at pretest, midtest, and posttest; and (d) the results as they pertain to application of linguistic principles described in Venezky (1970, 1995, 1999) and of connectionist models in cognitive neuroscience (e.g., Harm & Seidenberg, 1999; Seidenberg & McClelland, 1989) to instructional intervention. Briefly, the intervention, which was delivered 20 minutes a day, twice a week over a four-month interval to dyads as a supplement to the regular reading program, consisted of three segments. In the first 5-minute segment, all children received explicit training in the connections between spelling units (3) and phonemes in alphabetic principle. In the second l0-minute segment, children were randomly assigned to alternative treatments for modeling connections between units of written and spoken words, singly or in combination. In the third 5-minute segment, all children read engaging story books with the tutor providing assistance as necessary using the same kind of prompts as were modeled in the second segment. Discussion was used to create a set that the purpose of reading is to glean meaning and pleasure. For the sample as a whole, whose average Verbal IQ fell at the population mean, mean achievement on the Woodcock Reading Mastery Test-Revised (WRMT-R; Woodcock, 1987) increased 10 standard score points for age (2/3 standard deviation) on Word Identification (real word reading) and 14.2 standard score points for age (nearly 1 standard deviation) on Word Attack. The treated groups outperformed an untreated control group in the regular first-grade program on these measures (Stage, Abbott, Jenkins, & Berninger, in press). Student achievement was not statistically related to teacher-reported instructional variables in the regular reading programs, which tended to be balanced (Berninger, Abbott et al., 2000).

Four broad research questions are addressed in the second-year followup study. First, how did the faster and slower responders differ in their reading-related skills before, during, or after the first-grade intervention Second, did children who reached or exceeded grade level in word reading achievement by the end of the first-grade intervention maintain their gains at the beginning and end of second grade? Third, did the children who were still below grade level in word reading achievement at the end of the first-grade intervention continue to make reliable gains when intervention was sustained over time? If so, were there individual differences that predicted how well they responded to the second year of continued supplementary instruction? Fourth, how did the faster and slower responders differ in their reading profiles at the beginning and end of second grade? The method and results section below is organized by these four research questions.

METHODS AND RESULTS

Comparison of Faster and Slower Responders in Reading-Related Skills

By the end of first grade, half of the 128 children were achieving at or above grade level and half were still achieving below grade level. The faster and slower responders were compared at the fourth month of first grade prior to the first-grade intervention, the end of first grade, and from the fourth month to the end of first grade. The Verbal IQ on the Wechsler Intelligence Scale for Children, Third Edition, WISC III, (Psychological Corporation, 1991) was given before the first-grade intervention. Other measures given both before and after the first-grade intervention included the following: Word Identification and Word Attack of the Reading Mastery Test-Revised (WRMT-R) (Woodcock, 1987), phoneme deletion (Berninger, Thalberg, DeBruyn, & Smith, 1987), orthographic coding in short-term memory (Berninger, Yates, & Lester, 1991), adaptation of Olson, Kliegl, Davidson, and Foltz's (1985) orthographic choice for first graders, and rapid automatic naming of letters (Wolf, Bally, & Morris, 1986) and rapid automatic switching between letters and numbers (Wolf, 1986).

WISC-III Verbal IQ reliabilities range from .91 to .95, depending on age. WRMT-R Word Identification has an average reliability coefficient of .97, and WRMT-R Word Attack has an average reliability coefficient of .87. The phoneme deletion task used has a reliability of .85. Internal reliability coefficients of orthographic coding tasks range from .74 for letter cluster coding to .75 for letter coding and .84 for word coding. RAN (rapid naming of category-constant stimuli) test-retest over a 9-month interval was .65 for letters; RAS (rapid naming of category-switching stimuli) test-retest over the same interval was .85 for letters and numbers.

The groups differed significantly (but not substantially) in Verbal IQ; they did not differ in chronological age or gender (see Table 1). Prior to the first-grade intervention, the faster responding group (M = 86.3, SD = 9.5) was significantly higher than the slower responding group (M = 80.8, SD = 8.8) on WRMT-R Word Identification, t(90) = 2.86, p < .005, and the faster responding group (M = 85.8, SD = 11.6) was significantly higher than the slower responding group (M = 75.9, SD = 11.9) on WRMT-R Word Attack, t(90) = 4.16, p < .001. Thus, both groups were at the low end of the reading continuum, but the slower responders even more so than the faster responders. At the beginning of the first-grade intervention, the difference in reading skills between the two groups was greater for pseudowords than for real words.

At the end of the first-grade intervention, the faster responding group was significantly higher than the slower responding group not only on real word reading and pseudoword reading but also on phoneme deletion, orthographic coding in short-term memory of whole words, letters in a word and letter clusters in a word, word-specific orthographic representations of words in long-term memory (orthographic choice), rapid naming of category-constant stimuli (RAN), and rapid naming of category-switching stimuli (RAS); see Table ! for means and standard deviations. Thus, they differed on all reading-related skills assessed.

Additional insights were gleaned from growth-curve analyses from the beginning to the middle, to the end of the intervention. These analyses are important because they evaluate individual growth over the course of the intervention. Since the faster responders started out better in word reading, it is not surprising that they ended up better in reading than the slower responders. At issue is whether the faster responders showed more individual growth over the same period of time than did the slower responders. Hierarchical linear modeling (HLM) (Bryk, Raudenbush, & Congdon, 1996) was used to model growth on raw scores for WRMT-R Word Identification and Word Attack for each child in the Year 1 intervention. This technique, which allows researchers to model both growth in instructional interventions for individual children and group growth, has advantages not gleaned by comparing growth curves based simply on fitting linear regression models. For example, HLM employs empirical Bayes estimation and yields a reliability estimate for the slope; also, HLM allows analysis of cross-level relationships in an integrated statistical framework. The empirical Bayes procedure in HLM has an advantage over ordinary least-squares regression in slope estimation (Stage, in press). The individual growth curves enabled us to evaluate whether individuals, regardless of treatment group, showed significant growth (slope derived from HLM modeling was significantly different from chance at p < .05). As shown in Table 1, substantially more of the faster responders than the slower responders showed significant individual growth.

Additional analyses focused on the patterns of individual growth for the profile of real-word reading and pseudoword reading for the children in general, regardless of their overall level of rate of response. Growth groups were formed on the basis of significant individual growth curves during the first-grade intervention on (a) both real-word reading and pseudoword reading, (b) real-word reading only, (c) pseudoword reading only, and (d) neither. As shown in the top rows for Year 1 in Table 2, of the 128 children, 41% showed significant growth in real-word and pseudoword reading, 31% showed significant growth in real-word reading only, 3% showed significant growth in pseudoword reading only, and 24% showed no significant growth in either real-word or pseudoword reading during the first-grade intervention. Like students with reading disabilities studied by Lovett, Warren-Chaplin, Ransby, and Bordon (1990) and Lovett et al. (1994), these children tended to have more difficulty with phonological decoding (pseudoword) than word-specific (real word) reading. Size of IQ-achievement discrepancy did not predict slope of growth curves for either real word or pseudoword reading (Stage et al., in press).

Also at issue was which individual difference variables at the beginning and end of the first-grade intervention differentiated these growth groups. Thus, the four growth groups were compared on the individual difference measures (4) in Table 1, both at pretest and posttest, and Verbal IQ at pretest. Main effects in analysis of variance for the four growth groups were found on each of the following learner characteristics: letter cluster at pretest and posttest, orthographic choice at posttest, RAN letter at pretest and posttest, phoneme deletion at pretest and posttest, and Verbal IQ. We then used t-tests to analyze which of these learner characteristics differentiated (a) the group that did not grow on either real-word or pseudoword reading from the other groups, and (b) differentiated the group that grew in both real-word and pseudoword reading from the group that grew in real-word reading only.

The group lacking any growth in reading differed from the other groups on all measures except letter cluster coding at pretest, on which many children scored low. Thus, children showing no growth were at the bottom of the continuum for reading-related processes as well as reading; they had lower Verbal IQs, poorer phonological skills, poorer rapid naming skills, and poorer orthographic skills except for pretreatment letter cluster coding on which all children were initially poor. By the end of the intervention, the zero growth group that grew in neither word reading skill differed from the other groups on letter cluster coding as well.

The group that grew in both real-word and pseudoword reading differed from the group that grew in real-word reading only on letter cluster coding at pretest, t(91) = -2.56, p = .012, and posttest, t(90) = -3.06, p = .003, rapid naming of letters at posttest, t(88) - 2.35, p =.021, and prorated Verbal IQ t(89) = 3.35, p = .001; differences in degrees of freedom are due to missing data. Thus, the ability to code single letters and letter clusters, which is needed to apply the alphabet principle, was not as well developed in children who did not show significant growth in phonological decoding of pseudowords at the same time they showed significant growth in real-word reading. However, general verbal ability also contributed to a lack of significant growth in decoding at the same time children grew in real-word reading.

The findings for letter cluster coding and rapid naming of single letters in this comparison suggest that if these orthographic skills are underdeveloped, they may compromise early reading development for at-risk beginning readers in general, not just the slower responders, especially in application of alphabetic principle to unfamiliar words. Theories of orthographic processing should distinguish among (a) coding written words and their constituents into short-term memory where phonemes are mapped onto spelling units in words just read; (b) word-specific representations in long-term memory that are activated when familiar words are read; and (c) metaknowledge (e.g., judgments of legal spellings, positional information, and word-like properties). The first appears to compromise acquisition of alphabetic principle in at-risk beginning readers (results comparing growth groups; also Berninger et al., 1991). The second skill may compromise development of real-word reading skills (Stage et al., in press), but the third skill does not appear to differentiate good and poor readers (Siegel, Geva, & Share, 1995).

It was surprising, based on both empirical studies (e.g., Wagner & Torgesen, 1987) and computational simulations (e.g., Harm & Seidenberg, 1999), that phonological deletion did not differentiate the group that grew in real-word reading and pseudoword reading and the group that grew in real-word reading only. We suspect that our alphabetic principle training, by virtue of the decontextualization of the target phoneme and training in spelling-phoneme correspondences that enable phoneme awareness (Perfetti, 1985), facilitated development of phoneme skills in both growth groups. Because both phoneme manipulation and letter and letter cluster coding in short-term memory are skills needed to apply the alphabetic principle, future studies might include an instructional component for improving letter cluster coding in memory. The point is that both phoneme and orthographic (letter coding) skills contribute to phonological decoding of unknown words, as assessed by pseudoword reading, which is not purely a phonological task. Second-Grade Monitoring of Faster Responders

Not all 128 children were available for the Year 2 followup study due to the mobility of the families in schools in our area-either to schools outside a reasonable driving range or to other schools than the eight with which we had permission to work. The Year 2 study was confined to those schools that had participated in Year 1. Attrition rate was not systematically related to the treatment children had experienced during first grade. Of the 64 faster responders at or above grade level at the end of first grade, 44 (20 girls, 24 boys) still attended one of the eight participating schools and their parents granted informed consent for them to be monitored in fall and spring of second grade. Of the 44 in this Monitor Only Group at the beginning of second grade, 41 still attended the same school and completed the study at the end of second grade.

Three measures were given to both the Monitor Only Group (faster first-grade responders) and the Continue Tutor Group (slower first-grade responders) in October of second grade and again in May of second grade. The WRMT-R Word Identification and Word Attack subtests were administered individually. These tests require children to read lists of real words or pseudowords, respectively. The Gates MacGinitie Reading Comprehension Test (GM Comp) (MacGinitie & MacGinitie, 1989), which had not been given in Year 1, was group-administered.

This test requires children to read passages silently and answer multiple-choice questions.

Table 3 (first column) reports the second-grade fall and spring mean test scores and t-test results for the Monitor Only Group on two measures of word reading and one measure of reading comprehension. The second-grade fall scores were not significantly different from the first-grade spring scores. At each time point in second grade, the average score fell above the mean. Although the second mean test score was always higher than the first, the differences in means were not statistically significant. However, the relative gains made in first grade were maintained the following school year for the Monitor Only Group. The early boost seemed sufficient to give these faster responding at-risk readers, who began their journey to literacy at the low end of the reading continuum, a sufficient jumpstart to function consistently in the average range over a yearlong interval. Both at the beginning and the end of second grade, the Monitor Only Group was higher than the Continue Tutor Group on real-word reading, pseudoword reading, and reading comprehension (see last column in Table 3), indicating that the faster first grade responders retained their relative superiority over the slower responding first graders. Continuing Tutoring during Second Grade of the Slower First-Grade Responders

Of the 64 still below grade level, the first 48 (18 girls, 30 boys) whose parents granted informed consent were offered continuing tutoring during second grade. This Continue Tutor Group was also assessed in fall and spring of second grade. The Monitor Only Group (5) did not differ from the Continue Tutor Group (6) in distribution of ethnic groups, [x.sup.2](5) = 2.25, p = .81. Likewise, the groups did not differ in distribution of mother's level of education, ; [x.sup.2](5) = 6.39, p = .27.

The 48 children were randomly assigned to one cell of a four-condition design: alphabetic principle + question asking; alphabetic principle + summarization; combined alphabetic principle and syllable awareness + question asking; and combined alphabetic principle and syllable awareness + summarization. Application of alphabetic principle in English depends to some degree on knowledge of syllable structure; for example, whether the letter "a" should be decoded as a short or long vowel depends on whether it occurs in a closed or open syllable, respectively. Thus, we evaluated whether combining syllable awareness and alphabetic principle training was more beneficial than alphabetic principle training alone. In contrast to the Year 1 intervention that trained monosyllabic words, Year 2 intervention trained only polysyllabic words. Question answering and summarization are two techniques for developing reading comprehension in reciprocal teaching (e.g., Palincsar & Brown, 1984). We compared these two techniques in the context of teacher-led rather than student-led activities in implementation of reciprocal teaching in a classroom setting. In contrast to the Year 1 intervention that was delivered to dyads, Year 2 intervention was delivered to individuals so the potential for peer interaction was removed. The questions were all who, what, and where questions that focused on accurate recall of text-based information rather than inferential questions that would draw on students' background of knowledge, which could mask what they really comprehended in text.

The four tutors were balanced across the four treatment conditions; each tutored three children in each condition. As in Year 1, intervention commenced in January of the school year and concluded by the end of April; again, children were given twenty-four 20-minute individual tutorial sessions which, on average, occurred twice a week except for school closures.

During the first 3 minutes, children used student cards in which each spelling unit was paired with a picture of a word containing the target phoneme associated with it to review the alphabetic principle to which they were introduced in Year 1. Thirteen correspondences not covered in Year 1 were added, mostly related to polysyllabic words; for example, spellings for the schwa or reduced vowel in unaccented syllables. The 109 correspondences were divided into three roughly equal groups, which were alternated across the 24 lessons so that each third was practiced every three lessons for a total of eight times. During the first 12 lessons, the tutor modeled the associations between spelling unit picture and phoneme, as in Year 1. The tutor named the letter or letters in the spelling unit, said the name of the pictured word with the target phoneme, and made the phoneme in isolation. The child then repeated the auditory sequence -- naming the letter(s) in the spelling unit, naming the word, and producing the target phoneme while looking and pointing at the visual stimuli on their student cards. During the last 12 lessons the tutor pointed to the spelling unit, and the child alone produced the sequence of naming the spelling unit and producing the associated phoneme, unless assistance was required and was provided; the intermediate step of naming a word with the target phoneme was omitted during the last 12 lessons.

For the next 10 minutes, instruction focused on specific words. For each lesson, 14 polysyllabic words were selected from the text to be read later in the session. Four sessions were required to complete each of the six texts, (7) so each list was practiced four times. Of these, at least two exemplar words contained a syllable representing each of the seven syllable types in English: closed, open, silent e, vowel team, r-controlled, -le, and schwa (see the Appendix for sample word list). In each word the one- and two- letter spelling units taught and reviewed in the first segment were color coded to draw students' attention to the spelling units that correspond to phonemes. Time was equated across the treatment conditions by including a filler activity (listening to a story) in the alphabetic principle training only condition during the portion of time the other group spent on syllable awareness training.

Children who received only alphabetic principle training practiced the words in the second segment in the following way. First, the tutor named the word and spelled it, spelling unit by spelling unit. The child repeated the name of the word and the sequence of spelling units. Then, in left-to-right sequence, the tutor pointed to each color-coded spelling unit (one or two letters) and modeled its corresponding phoneme. Finally, the child pointed to each spelling unit and produced the associated phoneme. Following this procedure with all 14 words, the child was shown a list with all 14 words and was asked to name them. This approach integrates lexical feedback as in connectionism (e.g., Seidenberg & McClelland, 1989) and sublexical feedback (Thompson, 1996) following alphabetic principle training. For the remaining time in the second segment, the tutor read part of a continuing story to the child.

Children who received both syllable awareness training and alphabetic principle training always received the following training geared to syllables before practicing the words in the way just described for the alphabetic principle training only. First, the tutor said the word. Next, the tutor asked the child to say the word syllable by syllable. Then, the tutor said the word syllable by syllable and asked the child to place one colored disc on a "desktop" card for each phoneme in each syllable. Because syllable boundaries in English may vary with the speed with which a word is pronounced, the phoneme segmentation step was adapted to how the child had segmented the word into syllables. (For example, depending on how fast a word is said, rabbit may be segmented orally as rab/bit or rabb/it.) Then, the tutor showed the child a card with the written form of each of the words, with slashes that segmented the word into syllables. Children were asked to classify each syllable into its type using speaking syllables cards, with a visual schematic of a closed, open, silent e, vowel team, r-controlled, -le, or schwa syllable using the symbols C for consonant, V for vowel, dia-critical marks for short and long vowels, and x over silent letters (reproduced in Berninger, 1998). Children were told that sometimes the visual segmentation of the word into syllables may seem different from the syllables they perceived in the spoken form of the word, which may differ depending on rate of speech. However, when relevant, tutors pointed out how the orthography represents the spoken form; for example, the "b" in rabbit is doubled to mark the first syllable as closed and not open. Children were not taught canned rules for segmenting visual words into syllables; they were always encouraged to think about segmentation both in the spoken and the visual forms of the word.

For the last 7 minutes, children read engaging texts orally. If they could not pronounce a word, the tutor encouraged them to use the student cards for alphabetic principle and/or for speaking syllables to decode unknown words, depending on what was taught in their treatment condition. Altogether, children read six texts (7) divided into four sections over the course of the 24-tutorials. When the selection for the day was completed, the tutor posed a set of scripted questions or asked the child to summarize the selection in his or her own words,a

A four-pronged approach was used to ensure that the intervention was delivered in a standard, consistent manner. First, tutors used timers to ensure compliance with time limits. Second, weekly meetings were held with tutors to address any instructional or management problems that arose and deal with them in a consistent manner. Third, tutors were required to read and periodically review the instructional manual with all procedures and to complete a self-monitoring checklist at the end of each day. Fourth, tutors audiotaped at least 10 sessions across the various treatments. Assessment of audiotapes of each treatment resulted in overall mean ratings of 9.9 (SD = 0.32), where 10 is perfect.

Separate two-way ANOVAs (two levels of word training -- alphabetic principle only or combined alphabetic principle and syllable awareness; and two levels of comprehension training -- question asking or summarization) performed on each of the three measures in Table 3 generated the following results for the Continue Tutor Group. Time effects were statistically significant only for the measures of word reading. Continued tutoring resulted in improved real-word reading, F(2, 84) = 5.63, p < .005, and pseudoword reading, F(2, 84) = 23.33, p < .0001, but there were no treatment-specific main effects. Adding syllable awareness training to alphabet principle training did not appear to have an advantage over alphabetic principle training alone for these slower responding at-risk beginning readers. No main effects or interactions were found for the two kinds of comprehension training, showing that it is easier to increase word reading skills in the slower responders than it is to increase their reading comprehension. However, taking into account specific combinations of word reading training and reading comprehension training, the contrast between the treatment that paired alphabetic principle training and summarization (mean gain of 13.5 normal curve equivalent points) and that paired the combined syllable awareness and alphabetic principle training with question answering (mean loss of 1.7 normal curve equivalent points) was statistically significant, F(1, 44) = 4.04, p = .05. Additional research is needed to evaluate whether this finding, which suggests that pairing alphabetic principle training with summarization is especially effective with the slower responding at-risk readers, replicates on other samples of slower responders.

Additional analyses examined individual differences among the slower responders in their response to continued tutoring. Before, midway through, and immediately after the continued tutoring, each of the 48 children had completed a pseudoword reading task that contained 89 of the spelling-phoneme correspondences taught. This task was scored in terms of whether specific spelling-phoneme correspondences within the word were correctly decoded rather than on the basis of whether the whole word was decoded correctly. For example, in a pseudoword probe like vor, the "v" would be scored as a one-letter unit and the "or" (r-controlled vowel) would be scored as a two-letter unit. The one-letter and two-letter spelling units were analyzed separately. As means and F-tests reported in Table 4 show, over the course of the tutorial, children improved significantly in phonologically decoding both the one- and two- letter spelling units. However, as shown in Table 5, after treatment only the two-letter spelling units on our transfer measure of pseudoword reading contributed unique variance to reading pseudowords and real words on a standardized test; but before treatment only one-letter units on our test contributed unique variance to reading pseudowords on a standardized test and only two-letter units on our measure contributed unique variance to reading real words on a standardized test.

Considering that ability to learn phonemes associated with two-letter spelling units may be critical to continued word reading growth of the slower responders, we then examined which individual difference measures, which were available only at the end of first grade, might predict these children's initial abilities in decoding two-letter spelling units at the beginning of second grade prior to ongoing intervention. Table 6 summarizes the findings. Phoneme deletion contributed unique variance to both one- and two- letter spelling units. Rapid naming of letters contributed unique variance to knowledge of phoneme-single letter correspondences and marginally significant unique variance to phoneme-two letter correspondences. Letter cluster coding contributed marginally significant unique variance to phoneme correspondences with spelling units of either one- or two- letter size. One interpretation of these data is that phoneme skills are very important in learning to decode words but later in reading acquisition orthographic coding of two-letter spelling units becomes important in applying alphabetic principle.

HLM growth curves (Bryk et al., 1996) were computed for individuals who received the Year 2 continued tutoring and evaluated for significance above chance. Growth groups were then formed, according to whether significant growth occurred in (a) both real-word and pseudoword reading, (b) real-word reading only, (c) pseudoword reading only, or (d) neither; results are reported in Table 7. As in Year 1, children were more likely to show significant growth in real word than in pseudoword reading (all but 15 of the 48 subjects). Children who grew significantly in both real-word and pseudoword reading had the highest overall achievement in real-word reading. The six children who grew only in pseudoword reading had higher overall achievement in real-word and pseudoword reading than those who grew only in real-word reading. This finding suggests that children who do not begin to develop phonological decoding skills by the end of second grade, even though they develop word-specific reading skills, may be at continued risk in further development of their word recognition skills.

Consistency of growth groups from first to second grade was also examined. Although there was some consistency across grades (see diagonals in Table 7), there was also flexibility in growth across the years (see off-diagonals in Table 7), suggesting that these growth groups reflect normal fluctuations in developmental processes and not fixed subtypes. Manis, Seidenberg, Doi, McBride-Chang, and Petersen (1996) reported comparable findings for developmental fluctuations in lexical and sublexical reading processes. The distributions of first-grade and second-grade growth groups for the continuing tutoring group did not belong to the same underlying distribution, as evidenced by the nonsignificant [X.sup.2](9) of 3.65, p = .93. That is, for the slower responding children, type of growth in first grade did not determine how they grew in second grade. Five of the nine children who grew in neither measure in second grade had grown in either both or in real-word reading in first grade. Conversely, seven of the children who grew in neither in first grade grew in both in second grade. Eight who grew in neither in first grade grew in real-word reading in second grade. Three who grew in neither in first grade grew in pseudoword reading in second grade. Of 48 children, only four grew on neither measure in first grade or second grade. Thus, developmental flux and discontinuities are likely to be found in slower responding at-risk readers, who should be carefully monitored for growth in real-word and pseudoword reading, during both first and second grade.

Reasons for the lack of growth of four children in two consecutive years were explored. One child with an above-average IQ and no family history of reading problems had experienced psychological trauma in the home. One child who met the WISC III vocabulary cut-off of a scaled score of 6 ended up having a prorated Verbal IQ in the mildly retarded range, which was consistent with parent and teacher observations of her general lack of responsiveness to any kind of instruction. A child from a family riddled with dyslexia had a history of language delay in the preschool years, attention deficit disorder for which he took ritalin, and had received every kind of service available in the school (speech and language, Chapter 1, and counseling) except special education for which he did not qualify because the discrepancy between his Verbal IQ and reading achievement was not large enough. The fourth child was from a family with a strong history of reading problems but the school was not aware of her extreme difficulty and had told the parent she would outgrow the problems with maturation. As one of six of the slower responders with better comprehension than word reading, her strong comprehension skills were masking her inordinate difficulty in learning word recognition skills, as happens in some cases of dyslexia (see Berninger, 2001).

Comparison of Second-Grade Reading Profiles of Faster and Slower Responders

Also of instructional significance is the relationship between word reading and reading comprehension in the two groups both at the beginning and the end of second grade. To compare the relationship between these two reading skills, we first rescaled the comprehension measure to the same scale as the word recognition measures with a mean of 100 and a standard deviation of 15. For this purpose we used scaled scores for grade on WRMT-R Word Identification and Word Attack because the GM Comp normal curve equivalents (nce's) are based on grade norms. Then we coded the profiles that were consistent at the beginning and end of the school year -- that is, the same skill was always relatively higher (either Word Identification or Reading Comprehension) and the same skill was always relatively lower. Such profiles of uneven development of word and text level skills were also found in normally developing second, fourth, and sixth graders (Berninger, 1994) and in third and fifth graders (Berninger, Abbott, & Alsdorf, 1997). As shown in Table 8, about 75% of these profiles in the total sample were consistently uneven across second grade for these at-risk beginning readers. However, the groups showed a contrasting pattern in the relationship between word reading and reading comprehension skills. The Monitor Only Group tended to be relatively better at comprehension than word reading, and the Continue Tutoring Group tended to be relatively better at word reading than comprehension. As shown in Table 8, 46% of the Monitor Only Group, compared to 13% of the Continue Tutor Group, had substantially higher reading comprehension than word reading skills, and 60% of the Continue Tutor Group, compared to 29% of the Monitor Only Group, had substantially higher word reading skills. This pattern of association between fast` and slow-responding groups during early intervention and the relationship between word reading and reading comprehension achievement was statistically significant, [X.sup.2] (1), with continuity correction = 11.80, p < .001.

For GM Comp, a two-way ANOVA for group (two levels) and size of the difference between word reading and comprehension (see Ms and SDs in Table 8) revealed no main effect for group, but there was a main effect for the language profile (whether word reading or reading comprehension was relatively more developed), F(1, 62) = 104.9, p < .001. The size of the difference was slightly greater when reading comprehension was better developed. However, this main effect was qualified by an interaction between group and language profile, F(1, 62) = 8.96, p < .004. As the mean differences between word reading and reading comprehension in Table 8 show, the slower responders who had better reading comprehension demonstrated even larger differences between their reading comprehension and word reading skills than the slower responders who had better word reading. In contrast, the size of the difference was about the same for the faster responders whether or not their reading comprehension or word reading was better developed. Figure 1 plots results as a function of the word reading-reading comprehension skill difference and illustrates that the groups do not differ when word reading is better developed (white bars above the x-axis are approximately the same height for each group), but do when reading comprehension is better developed (the shaded bars below the x-axis are different for the groups). The difference is relatively greater for individuals in the Continue Tutor Group (whose word reading skills are less well developed than their reading comprehension skills) than in the Monitor Only Group.

[FIGURE 1 OMITTED]

Taken together, these results indicate that aiming instruction solely at word reading skills in the slower responding at-risk beginning readers may not be sufficient. Many (not all) need explicit, intensive, sustained comprehension instruction in addition to explicit instruction in word reading. At the same time, helping many (not all) faster responders develop better word reading skills may remove a bottleneck in the reading system, allowing their reading comprehension skills to develop normally.

The results also suggest that four subtypes, defined on the basis of component reading skills, may emerge in response to early intervention in a sample of at-risk beginning readers, selected because they are at the lower end of the continuum of word reading skills in mid first grade. The first subtype, which is the most prevalent in the faster responders, has relatively less well developed word reading than reading comprehension skills; their relative strength in reading comprehension may be due to their relatively well developed verbal reasoning skills. With an early boost in developing word reading skills, their reading takes off and falls, on average, in the average range and in some cases even above the mean. The bottleneck in their word reading is not severe and is responsive to early intervention. However, a smaller percentage of these faster responders may develop relatively better word reading skills, the focus of the intervention, than reading comprehension skills and thus represent a second subtype. The less frequent second subtype among the faster responders, whose word reading skills are relatively better developed, should be monitored during third and fourth grade to make certain they do not develop specific problems in comprehension, such as those studied by Oakhill (e.g., Oakhill & Yuill, 1996), Scarborough (e.g., Leach & Scarborough, 1999; Scarborough & Leach, 2000), or Swanson (e.g., Swanson & Alexander, 1997), for which they will need specially tailored comprehension instruction. Both of these emergent sub-types probably reflect normal variation in relative development of two reading skills, both of which fall in the acceptable range of reading skills for grade.

Among the slower responders there are also two sub-types. Slower responders whose word reading skills are relatively better developed than their reading comprehension skills represent a third subtype that may need the most sustained intervention of any of the at-risk beginning readers. They have significant instructional needs in both word reading and reading comprehension. The smaller group within the slower responders is a fourth subtype who should be closely monitored for dyslexia, a reading disorder in which word reading is underdeveloped compared to text-level reading comprehension and verbal comprehension skills (Berninger, 2001). Their relatively strong comprehension skills may mask the difficulties they are having with word reading early in the reading acquisition process. Often their reading disability does not become apparent until the intermediate grades when they are no longer able to keep up with reading assignments, and instruction geared to their word reading problems is no longer a standard part of the reading curriculum. Even if they make strong gains in reading comprehension during early intervention, their progress in word reading should be carefully monitored until it reaches at least an average level for grade and remains consistently at that level for at least two years.

GENERAL DISCUSSION

Early intervention in addition to the regular program can be expensive. Schools need to consider cost effectiveness. It is cost effective to provide early intervention only as long as it is needed. Failure to provide sufficient early intervention when the probability of prevention is the highest is not cost effective in the long run. The results of this longitudinal study suggest that some beginning at-risk readers only need intervention of brief duration whereas others need more sustained intervention.

The faster responding at-risk readers maintained their relative gains throughout second grade. These faster responding children differed from the slower responding children on the level of their initial reading skill and Verbal IQ. That is, the faster responders tended to be at about a standard deviation below the mean, whereas the slower responders tended to be about 1 1/3 to 1 2/3 standard deviations below the mean in their initial word reading skills. Further research is needed to evaluate whether -1 standard deviation in word reading is a valid, cost-effective cutoff for predicting who will need early intervention of shorter as opposed to longer duration. The faster responders also tended to differ from the slower responders on all the reading-related language skills shown to be the best predictors of response to early intervention, namely, phonological, orthographic, rapid naming, and verbal reasoning skills (also see, Stage et al., in press). That is, the slower responders tended to have a language system in which all the component skills were less well developed than was the case for the faster responders. Both groups had language skills within the normal range; none of the children had primary language disability. Thus, the slower responders may be children whose language skills (not necessarily other cognitive skills) fall in the low end of the normal distribution of language skills and will therefore require more intense, explicit language-based instruction over a longer period of time to acquire adequate levels of reading skill. In addition, the faster responders and the slower responders tended to differ in the relative patterning of their component reading skills. The faster responders tended to have relatively better developed reading comprehension, whereas the slower responders tended to have relatively better developed word reading skills. Early intervention that focuses solely on word reading may not be sufficient for most slower responders.

However, there were exceptions within both the faster and the slower responding groups. Nearly a third of the faster responders demonstrated a persisting pattern in which their word reading skills were relatively better developed than their reading comprehension. These children should be monitored during the third-to-fourth-grade transition to make sure that they do not exhibit specific comprehension disabilities as curriculum demands increase in the area of comprehension (cf. Oakhill & Yuill, 1996; Scarborough & Leach, 2000; Swanson & Alexander, 1997). Some of the slower responders had relative strengths in reading comprehension that might mask their continuing difficulties in word reading during the primary grades. These students might need continuing special assistance in learning word reading skills throughout elementary schooling even though the curriculum often does not provide such instruction in the intermediate grades. The point to emphasize is that early intervention may not be sufficient to eliminate all kinds of specific reading disabilities, and further research is needed on progress monitoring for specific at-risk profiles throughout elementary school.

The slower responding at-risk beginning readers continued to make reliable gains during the second year of continued tutoring. This finding demonstrates that with intervention of sufficient duration, at-risk readers will continue to make progress. However, it also raises an issue related to decision rules, which are needed not only for identifying who needs early intervention but also for terminating early intervention. On what basis can educators determine that a child has reached a reasonable level of achievement and that the regular program of reading instruction will be sufficient? Like other domains of human ability, reading is characterized by normal variation along a continuum of ability. Just as not everyone is expected to be a world-class athlete, musician, or chess player, likewise it is unreasonable, despite the Lake Wobegon longing for everyone to be above-average that is fueling the current educational reform movement, to expect everyone to be an above-average reader, a statistical impossibility. As shown in Table 3, the Continue Tutor Group attained word reading scores in the average range (but below the mean) by the end of the second year of intervention, and their average level of achievement was commensurate with what would be expected on the basis of their verbal ability (in the average range but below the mean) (see Table 1). These results suggest that they are ready, as a group, to graduate from early intervention. At the same time, those with the pattern of relatively better word reading skills would probably benefit from continued explicit instruction in comprehension in the regular reading program and those with the pattern of relatively better reading comprehension would probably benefit from continued explicit instruction in word reading in the regular reading program. In both cases this explicit instruction should be delivered in the context of a balanced reading program so that all reading skills continue to develop.

Level of reading achievement may not be the only yardstick to consider when evaluating whether reading achievement continues to be acceptable. Growth curves that evaluate whether change in a skill is reliable over time are also important in evaluating whether an individual student has higher skill at the end of the school year than at the beginning. The individual growth-curve analyses highlighted the uneven development of word-specific reading (real words) and phonological decoding (pseudowords) in some beginning readers. Those who grew in both were the better readers (see Table 2), but failure to grow in one did not rule out growth in that skill the following year (see Table 7). It is important not to give up on slower responders. With sufficient instruction (not merely maturation as some teachers believe), children may grow in each skill, although not necessarily at the same time.

This longitudinal study also suggested instructional components that might be added to phonological awareness and phonological decoding for the slower responders. Specifically, these children may benefit from explicit instruction in how phonemes correspond to two-letter as well as one-letter spelling units. Ability to learn correspondences for the two-letter spelling units uniquely predicted their reading achievement outcome after continued intervention. Both ability to code two-letter units in words represented in short-term memory and ability to manipulate phonemes in spoken words in short-term memory may affect ability to learn spelling-phoneme correspondences in alphabetic principle. Much of the predictability of English spelling involves spelling units larger than the single letter (Venezky, 1970, 1995), so it makes pedagogical sense to make explicit to beginning readers that the size of the spelling unit corresponding to a phoneme may vary (Berninger, Abbott et al., 2000). Initially, phonics instruction should focus on the correspondences between phonemes and single-letter spelling units, but eventually the correspondences involving two-letter spelling units should be introduced. The Talking Letters Manual (Berninger, 1998) outlines a sequence of how to teach them that was validated in prior intervention studies. Further research is needed on which learner characteristics predict who will benefit from instruction in which the most frequent spelling-phoneme correspondences are introduced within a relatively short time interval (e.g., four months) versus over a relatively longer period (e.g., two years).

Although we did not find a beneficial effect of adding syllable awareness training to alphabetic principle training, children improved in word learning when syllable awareness was included. The syllable awareness training did not hinder their word reading development. It may be that simply teaching words (see the Appendix) that contained exemplars of the seven syllable types in each lesson was sufficient to develop syllable awareness -- at least at the level of implicit knowledge. It is also possible that the beneficial effect of syllable awareness training depends on skill domain and developmental level. For teaching spelling to third graders at-risk for writing disability, syllable awareness training had some beneficial effects (Berninger, Vaughan et al., 2000). Further research is needed to evaluate whether the observed beneficial effect of pairing summarization with alphabetic principle training replicates.

This longitudinal research adds to a growing body of research that questions the validity of diagnosing reading disability solely on the basis of IQ-achievement discrepancy based on one-time administration of IQ and achievement tests (e.g., Berninger & Abbott, 1994; Fletcher et al., 1994; Siegel, 1989; Vellutino, Scanlon, & Lyon, 2000; Vellutino, Scanlon, & Tanzman, 1994, 1998; Vellutino et al., 1996). The longitudinal study contributes to that line of research by demonstrating that four subtypes of readers may emerge in response to early intervention: (a) faster responders who have a persisting relative strength in reading comprehension, (b) faster responders who have a persisting relative strength in word reading, (c) slower responders who have a persisting relative strength in word reading, and (d) slower responders who have a persisting relative strength in reading comprehension. We speculate that (a) the first subtype has the best prognosis; (b) the second subtype should be monitored for possible specific comprehension disability in the later grades; (c) the third subtype may need varying degrees of ongoing support throughout schooling if expected to achieve at grade level as they fall below the mean but in the normal range in verbal ability; and (d) the fourth subtype should be monitored for dyslexia throughout elementary school. The fourth subtype is the one that is most likely, if not given early intervention of appropriate intensity and duration, to show genetically constrained IQ-achievement discrepancy (see Olson, Datta, Gayan, & DeFries, 1999). Educational policy and federally mandated services should address the needs of all four subtypes, not just those who exhibit IQ-achievement discrepancies of an arbitrary size.

Although IQ-achievement discrepancy did not predict individual growth and is irrelevant, IQ was not completely irrelevant. It did differentiate the faster and the slower responders, but was only one of several variables that differentiated them. Prior to intervention, these groups differed on all the language variables in Table 1, which included orthographic, phonological, and rapid naming skills. In response to intervention, substantially more of the faster than the slower responders showed individual growth in both real-word and pseudoword reading (see Table 1). Variables that differentiated the groups showing individual growth in word and/or pseudoword reading compared to the group showing zero growth in either kind of word reading included Verbal IQ and orthographic, phonological, and rapid naming skills. Variables that differentiated those who grew in only real-word reading from those who grew in both pseudoword and real-word reading included rapid naming of letters and letter cluster coding in short-term memory, both of which are relevant to applying orthographic skills to alphabetic principle. Verbal codes (phonemes) must be applied to letter codes quickly because of time constraints in short-term memory, and sometimes two-letter functional spelling units are involved. In fact, among the slower responders who received continued tutoring, skill with two-letter spelling units uniquely predicted pseudoword reading outcome after (not before) treatment in second grade (see Table 5), which in turn was unqiuely predicted by phoneme deletion skills at the beginning of first grade (see Table 6). Therefore, not only phonological but also orthographic skills are important in applying alphabetic principle to phonological decoding.

Appendix

Sample List of Words (a) Taught That Appeared in Storybook Read Later

Word on List   Syllable   Syllable Type   Syllable   Syllable Type

pencil         pen        closed          cil        closed
landed         land       closed          ed         closed
paper          pa         open            per        r-controlled
began          be         open            gan        closed
middle         midd       closed          le         le
uncle          unc        closed          le         le
story          stor       r-controlled    y          open
mother         moth       closed          er         r-controlled
outside        out        vowel team      side       silent e
nicer          nice       silent e        er         r-controlled
playing        play       vowel team      ing        closed
ready          read       closed          y          open
again          a          schwad          gain       closed
upon           u          schwa           pon        closed

(a) Each word list, which was taught and reviewed over the four
sessions in which the storybook that contained these words was
read, contained at least two exemplars of each of the syllable types.
Children saw only the word list (alphabetic principle training only)
or the word list and seven Speaking Syllable Cards (reproduced in
Berninger, 1998) that represented each syllable type schematically,
with c for constant and V for vowel, diacritical marks for long and
short vowels, and  x over silent letters.
Table 1

Comparison of Groups at the End of 24-Lesson, 4-Month Early Reading
Intervention in the Spring of First Grade

                     Faster           Slower
                     Responders (a)   Responders (b)   t(87)

Word
Identification (c)
  % significant
  growth (a)             95.4%            50.0%         5.54 ***
  M                      102.3             86.9         7.25 ***
  SD                       9.4             10.9

Word Attack (c)
  % significant
  growth (d)             77.3%            20.8%         6.49 ***
  M                      103.0             90.7         7.60 ***
  SD                       8.3              7.2

Prorated Verbal
IQ (e,k)
  M                       97.2             91.4         2.47 *
  SD                      12.4             10.1

Orthographic
Coding (f)
  Letter
  M (%)                   84.4             74.8         4.75 ***
  SD                      10.2              9.1
  Letter
  Cluster
  M (%)                   71.3             65.1         2.56 *
  SD                      12.3              8.4
  Whole Word
  M (%)                   89.2             79.6         5.10 ***
  SD                       7.7             10.1

Orthographic
Choice (g)
  M (%)                   82.7             71.7         4.47 ***
  SD                       9.8             13.4

Phoneme
Deletion (h)
  M (of 20)               11.2              8.7         3.13 **
  SD                       3.5              4.2

RAN/RAS
  RAN
  Letters (i)
  M (sec)                 39.3             52.5        -4.29 ***
  SD                      11.3             17.3
  RAN Digits (i)
  M (sec)                 38.6             46.6        -3.27 **
  SD                      10.4             12.5
  RAS Letters
  and Digits (j)
  M (sec)                 51.2             67.0        -4.16 ***
  SD                      13.1             21.6

Chronological
Age
  M (mos)                 92.5             91.9         0.60
  SD                       4.4              4.6

Gender
(% female)
  M                       0.45             0.38         0.77
  SD                      0.50             0.49

*p < .05.

**p < .01.

***p < .001.

(a) n = 44.

(b) n = 48.

(c) WRMT-R (Woodcock, 1987); M = 100, SD = 15.

(d) % with significant growth curves during first grade intervention.

(e) WISC III (Wechsler, 1991); M = 100, SD = 15.

(f) Berninger et al. (1991).

(g) Olson et al. (1994).

(h) Modified Rosner (Berninger et al., 1987).

(i) RAN (Wolf et al., 1986).

(j) RAS (Wolf, 1986).
Table 2

Mean Age-Corrected Standard Scores at Posttest
for Growth Groups Defined on the Basis
of Significant Individual Growth Curves in Wave
I Intervention in First Grade (N = 128)
and Wave 2 Intervention in Second Grade (N = 48)

           Both Word                  Word        Word Attack   Neither
           Identification        Identification   Only
           and Word Attack (b)   Only

Year 1            n=53               n = 40          n = 4       n = 31
  WI (a)
  M               101.8                94.5           87.8         80.1
  SD               10.1                 8.8            8.4         10.2
  WA (b)
  M               104.2                90.9           97.3         86.8
  SD                6.6                10.1            3.9          7.9

Year 2            n= 14              n = 19          n = 6        n = 9
  WI (a)
  M                94.1                89.3           92.5         86.3
  SD               11.2                12.7           12.6         12.8
  WA (b)
  M                98.8                90.9          102.5         91.6
  SD                9.8                 9.5           11.1          8.7

(a) Real-word reading.

(b) Pseudoword reading.
Table 3
Second-Grade Followup in Fall and
Spring for Students in First-Grade Intervention

                          Monitor    Continue      Monitor vs.
                          Only (a)   Tutor (b)     Continue

Word
Identification (c)
  Fall                                             t(87) = 5.06 ***
  M                        105.2       85.3
  SD                        12.2       11.6
  Spring                                           t(87) = 6.92 ***
  M                        107.4       90.6
  SD                        10.5       12.2
  Fall vs. Spring t(40)     2.27        3.91 ***

Word Attack (c)
  Fall                                             t(87) = 7.79 ***
  M                        103.5       88.2
  SD                         9.4        9.2
  Spring                                           t(87) = 4.01 ***
  M                        105.5       94.8
  SD                        14.8       10.4
  Fall vs. Spring t(40)     1.11        6.16 ***

Reading
Comprehension (a)
  Fall
  M                         54.9       33.4                6.04 ***
  SD                        13.7       18.9
  Spring
  M                         56.6       37.8                5.06 ***
  SD                        13.7       20.2
  Fall vs Spring t(40)      1.03       1.61

*** p < .001.

(a) n = 44 beginning 2nd, 41 end 2nd; faster responders in first grade.

(b) n 48; slower responders in first grade.

(c) WRMT-R (Woodcock, 1987), M = 100, SD = 15.

(d) GM Comp (MacGinitie & MacGinitie, 1989), M = 50, SD = 20.11.
Table 4

Descriptive Statistics for Phonologically
Decoding Taught Single-Letter and Two-Letter
Spelling Units in Pseudoword Transfer Words
Over the Course of the Second-Grade Tutorial
for the Continue Tutor Group

                 Beginning      Middle      End (c)      F(2,     p
                   (a)           (b)                     44)

                M      SD      M      SD     M     SD
1-letter (d)   22.2    3.8   24.9    4.8   27.5   4.3   47.25   .0001
2-letter (e)   25.7   10.6   29.2   11.3   33.9   9.9   32.07   .0001

(a) Prior to first tutorial session.

(b) After 12 tutorial sessions.

(b) After 24 tutorial sessions.

(d) 33 possible in transfer words.

(e) 56 possible in transfer words.
Table 5

Unique Contribution of Single- and Two-Letter
Spelling Units (a) to Word Reading for the
Continue Tutor Group

                      r     R (2)  F (b)    p      beta    t      p
Before Treatment
Word
Identification (c)         .64     38.82   .001
  1-letter unit      .38                          .1900   1.15   .2593
  2-letter unit      .35                          .6366   3.84   .0004
Word Attack (c)            .44     17.04   .001
  1-letter unit      .65                          .4530   2.19   .034
  2-letter unit      .61                          .2380   1.15   .257
After Treatment
Word
Identification (c)         .49    21.74    .001
  1-letter unit      .61                          .2297   1.43   .1596
  2-letter unit      .68                          .5121   3.19   .0026
Word Attack (c)            .65    16.30    .001
  1-letter unit      .58                          .2746   1.60   .1164
  2-letter unit      .62                          .4162   2.43   .0193

(a) Taught spelling units that were on transfertest of
pseudoword reading, (b) df = 2, 43 before treatment;
2, 45 after treatment because
of missing data before treatment. (c) WRMT-R (Woodcock, 1987).
Table 6

Predicting Single- and Two-Letter Decoding in Transfer Pseudowords at
the Beginning of Second Grade From Individual Differences in
Phonological, Rapid Automatic Naming (RAN), and Orthographic Skills at
the `End of First Grade

                                 r       [R.sup.2]  F(4, 42)    P
Decoding Single-Letter Units             .39          6.71    .0003
  Phoneme Deletion             .45 ***
  RAN Letters                 -.40 **
  Letter Coding                .31 *
  Letter Cluster Coding        .39 **

Decoding Two-Letter Units              .64          7.42    .0001
  Phoneme Deletion             .48 ***
  RAN Letters                 -.37 **
  Letter Coding                .37 **
  Letter Cluster Coding        .41 **

                                 beta       t      p

Decoding Single-Letter Units    .356215   2.880  .0063
  Phoneme Deletion             -.307044  -2.313  .0063
  RAN Letters                   .000324   0.002  .9983
  Letter Coding                 .273353   1.982  .0541
  Letter Cluster Coding

Decoding Two-Letter Units     .382007   3.15   .0030
  Phoneme Deletion             -.245439  -1.89   .0661
  RAN Letters                   .094854   0.655  .5161
  Letter Coding                 .252224   1.866  .0690
  Letter Cluster Coding

* p < .05.

** p < .01.

*** p < .001.
Table 7

Consistency of Growth Groups From Wave 1 in First Grade to Wave 2 in
Second Grade
                      Year 2
          Both   WI Only   WA Only   Neither
Year 1
Both       3        3         0         2
WI only    3        7         3         3
WA only    1        1         0         0
Neither    7        8         3         4

Note. WI = word Identification; WA = word Attack.
Table 8

Relationships Between Word Identification (WI) and Comprehension (Comp)
in Second Graders Who Received Early Intervention in Reading During
First Grade (a)

                                           Mean Difference
                                             Beginning 2nd
                       n    % Consistent        M      SD
Monitor Only Group
  (b) (n = 41)
Consistent
  (d) WI > Comp        12         29           14.2   10.4
Consistent
  (d) Comp > WI        19         46           11.5    5.6

Continue Tutoring
  Group (c) (n = 48)
Consistent
  (d) WI > Comp        29         60           19.6   15.6
Consistent
  (d) Comp > WI         6         13           27.3   12.0

                       Mean Difference
                         End 2nd
                          M      SD
Monitor Only Group
  (b) (n = 41)
Consistent
  (d) WI > Comp          16.7   10.3
Consistent
  (d) Comp > WI          14.0    7.9

Continue Tutoring
  Group (c) (n = 48)
Consistent
  (d) WI > Comp          23.3   15.0
Consistent
  (d) Comp > WI          28.7    8.1

(a) wl = WRMT-R Word Identification standard score for grade
(M = 100, SD = 15); Comp = GM Reading Comprehension (normal curve
equivalents rescaled to M = 100, SD = 15).

(b) Intervention only in first grade.

(c) Intervention in first and second grade.

(d) Same relative pattern at beginning and end of second grade.

ACKNOWLEDGMENTS

This research was supported by a Multidisciplinary Learning Disabilities Center Grant from the National Institute of Child Health and Human Development, P50 33812-05.

We thank the participating children and staff at Briarcrest, Ridgecrest, and Meridian Park in the Shoreline Schools, at Enatai Elementary in the Bellevue Schools, at Island Park and West

Mercer in the Mercer Island Schools, and at St. Benedict's and St. Luke's in the Archdiocese of Seattle for their willingness to work with us on a continuing basis during a two-year longitudinal study of early intervention for at-risk reading.

Please address correspondence to the first author at 322 Miller, Box 353600, University of Washington, Seattle, WA 98195-3600, <[e-mail]>, or 206/616.6311 (fax).

NOTES

(1) Examples of high-intensity intervention are the 340 hours of specialized instruction (Brown & Felton, 1990), the 35 to 65 hours of individual tutoring (Vellutino et al., 1996), the 174 hours of class-room-based instruction (Footman et al., 1998), the 80 minutes of tutoring a week for two years (Torgesen et al., 1997), and the 200 minutes a week of instruction for 8 months (Torgesen et al., 1999) that at-risk beginning readers received. Reading Recovery (see Shanahan & Barr, 1995) is also a relatively intensive individualized early intervention.

(2) We use the term "slower responder" rather than "treatment resister" to emphasize that some children can learn to read adequately if given intervention of sufficient duration and that there are individual differences in rate of reading acquisition that fall in the normal range in learning to read.

(3) We use the term "spelling unit" to refer to functional spelling unit, a term Venezky (1970) introduced to emphasize that a phoneme may map onto either a one- or two-letter unit. Although the term "grapheme" used by linguists includes both single and multiletter units, we prefer the term "spelling unit" to clarify that the number of letters in the spelling unit that corresponds to a phoneme may vary. Many teachers and children assume mistakenly that English words can be read letter by letter going left to right and do not realize that for purposes of phonological decoding the written word must be parsed into one, two or more letters before recoding the spelling into phonemes.

(4) Of the RAN and RAS measures, only RAN letters were entered into these analyses.

(5) Monitor Only Group, Asian American, 6; Black American, 1; European American, 34; Hispanic, 2; unknown, 1. Mother's level of education, < high school, 2; high school, 10; community college /vocational, 12; college, 16; graduate degree, 2; unknown, 2.

(6) Continue Tutor Group, Asian American, 6; Black American, 3; European American, 35; Hispanic, 1; Other or Unknown, 3. Mother's level of education, < high school, 0; high school, 16; community college/vocational, 17; college, 13; graduate degree, 2.

(7) Texts included Kermit and Robin's Scary Story by Michaela Muntean, 1995, New York: Penguin; Small Pig by Arnold Lobel, 1988, New York: Harper Trophy; She Sells Seashells by Grace Kim, 1995, New York: Scholastic; The Missing Tooth by Joanna Cole, 1988, New York: Random House; The Best Teacher in the World by Bernice Chardiet and Grace Maccarone, 1990, New York: Scholastic; The Blind Men and the Elephant by Karen Backstein, 1992, New York: Scholastic.

(8) A complete set of the lesson plans is available from the first author.

REFERENCES

Berninger, V. (2001). Understanding the lexia in dyslexia. Annals of Dyslexia, 51, 23-48.

Berninger, V. (1998). Process assessment of the learner (PAL): Guides for intervention and PAL intervention kit. San Antonio, TX: The Psychological Corporation.

Berninger, V. (1994). Intraindividual differences in levels of language in comprehension of written sentences. Learning and Individual Differences, 6, 433-457.

Berninger, V., & Abbott, R. (1994). Redefining learning disabilities: Moving beyond aptitude-achievement discrepancies to failure to respond to validated treatment protocols. In G.R. Lyon (Ed.), Frames of reference for the assessment of learning disabilities: New views on measurement issues (pp. 163-202).

Baltimore: Paul H. Brookes.

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., Brooksher, R., Lemos, Z., Ogier, S., Zook, D., & Mostafapour, E. (2000). Making the predictability of English orthography explicit to at-risk beginning readers: Evidence for alternative, effective strategies. Developmental Neuropsychology, 17, 241-271.

Berninger, V., Thalberg, S., DeBruyn, I., & Smith, R. (1987). Preventing reading disabilities by assessing and remediating phonemic skills. School Psychology Review, 16, 554-565.

Berninger, V., Vaughan, K., Abbott, R., Brooks, A., Begay, K., Curtin, G., Byrd, K., & Graham, S. (2000). Language-based spelling instruction: Teaching children to make multiple connections between spoken and written words. Learning Disability Quarterly, 23, 117-135.

Berninger, V., Yates, C., & Lester, K. (1991). Multiple orthographic codes in reading and writing acquisition: Reading and Writing: An Interdisciplinary Journal, 3, 115-149.

Brown, I., & Felton, R. (1990). Effects of instruction on beginning reading skills in children at risk for reading disability. Reading and Writing: An Interdisciplinary Journal, 2, 223-241.

Bryk, A., Raudenbush, S., & Congdon, R. (1996). Hierarchical linear and nonlinear modeling with the HLM/2L and HLM/3L program. Chicago: Scientific Software.

Fletcher, J., Shaywitz, S., Shankweiler, D., Katz, L., Liberman, I., Steubing, K., Francis, D., Fowler, A., & Shaywitz, B. (1994). Cognitive profiles of reading disability: Comparisons of discrepancy and low achievement definitions. Journal of Educational Psychology, 86, 6-23.

Footman, 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, 37-55.

Harm, M., & Seidenberg, M. (1999). Phonology, reading acquisition, and dyslexia: Insights from connectionist models. Psychological Review, 106, 491-528.

Leach, J., & Scarborough, H. (1999, April). The "second" wave of reading disabilities: What kinds of reading problems are not identified until grade 4 or later? Presented at the annual Conference of the Society for Scientific Study of Reading, Montreal.

Lovett, M., Bordon, S., Deluca, T., Lacerenza, L., Bensen, N., & Brackstone, D. (1994). Treating the core deficits of dyslexia: Evidence of transfer of learning after phonologically-based and strategy-based reading training programs. Developmental Psychology, 30, 805-822.

Lovett, M., Warren-Chaplin, P., Ransby, M., & Bordon, S. (1990). Training the word recognition skills of reading disabled children: Treatment and transfer effects. Journal of Educational Psychology, 82, 769-780.

Lyon, G. R., Alexander, D., & Yaffe, S. (1997). Progress and promise in research in learning disabilities. Learning Disabilities: A Multidisciplinary Journal, 8, 1-6.

MacGinitie, W., & MacGinitie, R. (1989). Gates-MacGinitie. Reading Tests. Third Edition. Manual for scoring and interpretation. Itasca, IL: The Riverside Publishing Company.

Manis, F., Seidenberg, M., Doi, L., McBride-Chang, C., & Petersen, A. (1996). On the base of two subtypes of developmental dyslexia. Cognition, 58, 157-195.

Oakhill, J., & Yuill, N. (1996). Higher order factors in comprehension disability: Process and remediation. In C. Cornoldi & J. Oakhill (Eds.), Reading comprehension difficulties: Processes and intervention (pp. 69-92). Mahwah, N.J.

Olson, R., Datta, H., Gayan, J., & DeFries, J. (1999). A behavioral-genetic analysis of reading disabilities and component processes. In R. M. Klein & P.A. McMullen (Eds.), Converging methods for understanding reading and dyslexia (pp. 133-151). Cambridge, MA: MIT Press.

Olson, R., Kliegl, R., Davidson, B., & Foltz, G. (1985). Individual and developmental differences in reading disability. In G. E. MacKinnon & T. G. Waller (Eds.), Reading research: Advances in theory and practice (pp. 1-64). New York: Academic Press.

Palincsar, A., & Brown, A. (1984). Reciprocal teaching of comprehension-fostering and monitoring activities. Cognition and Instruction, 1, 117-175.

Perfetti, C. (1985). Reading ability. New York: Oxford University Press.

Scarborough, H., & Leach, J. (2000, November). Late-emerging reading disabilities. New research on the "fourth-grade slump." Presented at the annual conference of the International Dyslexia Association, Washington, DC.

The Psychological Corporation. (1991). The Wechsler Intelligence Scale for Children, Third Edition, WISC Iii. San Antonio, TX: Author.

Seidenberg, M., & McClelland, J. (1989). A distributed, developmental model of word recognition and naming. Psychological Review, 96, 523-568.

Shanahan, T., & Bart, R. (1995). Reading Recovery. An independent evaluation of the effects of an early instructional intervention for at-risk learners. Reading Research Quarterly, 30, 958-995.

Siegel, L. (1989). IQ is irrelevant to the definition of learning disabilities. Journal of Learning Disabilities, 22, 469-478.

Siegel, L., Geva, E., & Share, D. (1995). Evidence of superior orthographic skills in dyslexics. Psychological Science, 6, 250-254.

Stage, S. (in press). Program evaluation using hierarchical linear modeling with curriculum-based measurement reading probes. School Psychology Quarterly.

Stage, S., Abbott, R., Jenkins, J., & Berninger, V. (in press). Predicting response to early reading intervention using Verbal IQ, reading-related language abilities, attention ratings, and Verbal IQ-word reading discrepancy. Journal of Learning Disabilities.

Swanson, H. L., & Alexander, J. (1997). Cognitive processes as predictors of word recognition and reading comprehension in learning-disabled and skilled readers: Revisiting the specificity hypothesis. Journal of Educational Psychology, 89, 128-158.

Thompson, B. (1996). Sublexical orthographic-phonological relations early in the acquisition of reading: The knowledge sources account. Journal of Experimental Child Psychology, 62, 190-222.

Torgesen, J. K. (2000). Individual differences in response to early intervention in reading: The lingering problem of treatment resisters. Learning Disabilities Research and Practice, 15, 55-64.

Torgesen, J., Wagner, R., & Rashotte, C. (1997). The prevention and remediation of severe reading disabilities: Keeping the end in mind. Scientific Studies in Reading, 1,217-234.

Torgesen, J., Wagner, R., Rashotte, C., & Herron, J. (1999). A comparison of two computer assisted approaches to the prevention of reading disabilities in young children. Manuscript in preparation.

Tunmer, W., & Nesdale, A. (1985). Phonemic segmentation skill and beginning reading. Journal of Educational Psychology, 77, 417-427.

Vadasy, P., Jenkins, J., & Pool, K. (2000). Effects of tutoring in phonological and early reading skills on students at-risk for reading disabilities. Journal of Learning Disabilities, 33, 505-600.

Vellutino., F., Scanlon, D., & Lyon, G. R. (2000). IQ scores do not differentiate between difficult to remediate and readily remediated poor readers: More evidence against the IQ achievement discrepancy definition of reading disability. Journal of Learning Disabilities, 33, 223-238.

Vellutino, F., Scanlon, D., Sipay, E., Small, S., Pratt, A., Chen, R., & Denckla, M. (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.

Vellutino, F., Scanlon, D., & Tanzman, M. (1998). The case for early intervention in diagnosing specific reading disability. Journal of School Psychology, 36, 367-397.

Venezky, R. (1970). The structure of English orthography. The Hague: Mouton.

Venezky, R. (1995). The varieties of orthographic knowledge II: Relationships to phonology, reading, and writing. In V. W. Berninger (Ed.), The varieties of orthographic knowledge II: Relationships to phonology, reading, and writing (pp. 23-46). Dordrecht, The Netherlands: Kluwer Academic Publishers.

Venezky, R. (1999). The American way of spelling. New York: Guilford.

Wagner, R., & Torgesen, J. (1987). The nature of phonological processing and its causal role in the acquisition of reading skills. Psychological Bulletin, 101, 192-212.

Wolf, M. (1986). Rapid alternating stimulus naming in the developmental dyslexias. Brain and Language, 27, 360-379.

Wolf, M., Bally, H., & Morris, R. (1986). Automaticity, retrieval processes, and reading: A longitudinal investigation. Child Development, 57, 98-100.

Woodcock, R. (1987). Woodcock Reading Mastery Test-Revised. Circle Pines, MN: American Guidance Service.

VIRGINIA W. BERNINGER, Ph.D., is professor of educational psychology and director of Learning Disabilities Center, Literacy Trek Project, and the Write Stuff Intervention Project, University of Washington.

ROBERT D. ABBOTT, Ph.D., is area chair, educational psychology, and statistician, Learning Disabilities Center, University of Washington.

KARIN VERMEULEN, M.Ed., is a school psychologist, Tennessee.

STACY OGIER, M.Ed., is a doctoral student in school psychology, University of Washington.

REBECCA BROOKSHER, M.Ed., is a school psychologist, Washington.

DORI ZOOK, M.Ed., is a school psychologist, Washington.

ZENIA LEMOS, M.Ed., is a school psychologist, California.