Measuring the Dimensions of Alphabetic Principle on the Reading Development of First Graders Journal of Learning Disabilities Volume 41 Number March/April 2008 143-157 © 2008 Hammill Institute on Disabilities 10.1177/0022219407313585 http://journaloflearningdisabilities sagepub.com hosted at http://online.sagepub.com The Role of Automaticity and Unitization Beth A Harn Mike Stoolmiller University of Oregon, Eugene David J Chard Southern Methodist University, Dallas, Texas This article presents critical issues related to word reading development within Ehri’s theoretical context by focusing primarily on the relation of decoding skill (of nonwords) to word reading and the development of unitization Within this context, issues and considerations related to research, measurement, and reading development are presented from research and field-based perspectives Analyses examining the relation between a measure of alphabetic principle, Nonsense Word Fluency (NWF), and fluency with connected text, Oral Reading Fluency (ORF), across first grade demonstrated a linear relation between the measures, which was attenuated for students initially above criterion on the NWF measure A new scoring approach to the NWF measure is presented to capture initial unitization development and was found to account for unique variance in initial status and growth on ORF and provide instructionally relevant information on the nature of developing alphabetic principle skills Considerations for future research and school-based applications are provided Keywords: alphabetic principle; automaticity; measurement; reading development; unitization Importance of Word Reading Development Measuring and understanding reading development requires recognizing the multidimensional and dynamic nature of this critical constellation of skills The interplay of language, attention, word identification, vocabulary, comprehension, experience, phonology, intelligence, instruction, and fluency have all (as well as other domains/skills) been investigated and demonstrated to have an impact on the development of reading (Kame’enui, Good, & Harn, 2005) In addition, the importance of successful early reading development has been consistently demonstrated across numerous studies for the past 10 years (Francis, Shaywitz, Stuebing, Shaywitz, & Fletcher, 1996; Juel, 1988; Stanovich, 1988) The pivotal role of assessment stands at the intersection of early success and the developmental plasticity of skills in reading Research clearly demonstrates that to read and understand connected text is dependent on acquiring automaticity in the alphabetic principle, the understanding that letters of the alphabet and the phonemes to which they correspond can be used to read words (Adams, 1990; Foorman, Francis, Shaywitz, Shaywitz, & Fletcher, 1997; Torgesen, 2000) In contrast, students who struggle to acquire the alphabetic principle fail to develop early, successful word reading skills (Stanovich, 1986) In their research with struggling older readers, Torgesen, Rashotte, and Alexander (2001) found that the most common characteristic of poor readers was a difficulty reading words accurately and instantly, indicating poor or incomplete representation of words in memory (Rack, Snowling, & Olson, 1992) As they grow older, Authors’ Note: Preparation of this article was supported in part by Project CIRCUITS, Grant No H324X010014 This material does not necessarily represent the policy of the U.S Department of Education, and neither is the material necessarily endorsed by the federal government Please address correspondence to Beth Harn, Department of Special Education & Clinical Services, 5261 University of Oregon, Eugene, OR 97403; e-mail: bharn@uoregon.edu 143 144 Journal of Learning Disabilities these struggling readers are often identified as having concomitant problems with reading fluency, vocabulary, and comprehension The past several decades represent a watershed period in research on the alphabetic principle and word reading development (Duncan & Seymour, 2003; Perfetti & Bolger, 2004) However, many important questions remain regarding how students acquire initial skills in the alphabetic principle as well as how most students successfully transition to more advanced word reading proficiency and later to proficient reading We strongly believe the best way to accomplish these goals is through a theory-based program of research A good theory allows researchers to focus their efforts and not get distracted by irrelevant information; moreover, findings linked to useful theories are more likely to have practical implications for teaching (Carnine, 2000) Weakly theoretical research focused largely on risk factors, correlations, proficiency assessment, or predictions is crude, inefficient, and slow to progress the field As discussed by Vaughn and Dammann (2001), theories, facts, and findings are not interchangeable Rather, theories explain the relations between facts and findings and how they might be applied in contexts The purposes of this article are to (a) present a strong theoretical model of word reading development that focuses on the development of the alphabetic principle; (b) further articulate the measurement needs surrounding assessing decoding skill development; (c) provide analyses demonstrating the complex, nonlinear relation between the development of decoding skills and oral reading fluency; and (d) present a new scoring method to improve instructional decision making within the critical year of first grade A Theoretical Model of Word Reading Development One of the most widely accepted theories of reading development that demonstrates the complex and multidimensional nature of reading acquisition is Ehri’s phase theory (Ehri, 2002, 2005a) Ehri’s theory of word reading development postulates that students master multiple word reading phases on the path to proficient reading Ehri (2005a) specifically stated that students may use multiple approaches to reading words but that the reader’s phase is determined by the predominant approach the reader uses at a given time Ehri articulated how in each phase, bonds of specific word features (e.g., letter-sounds, syllables, words) are encoded in memory through experience, instruction, practice, and repetition A brief discussion of each phase is presented, followed by assessment needs and implications Children, typically preschoolers, in the prealphabetic phase have no appreciation of the alphabetic principle, and the phase is typified by children attempting to translate the unfamiliar visual forms of print into familiar oral language through visual clues in the print (e.g., the oo that look like eyes in the word look) As children learn the link between oral language and print, they transition to the partial alphabetic phase, in which readers have learned to focus on salient parts of a word using initial and later final letters as clues to a printed word’s pronunciation As they become more familiar with printed letters and sounds, they move into the full alphabetic phase Now, even without prior exposure, students can read short words through their learned strategy of using the common letter sounds and blending Students in this phase, typically in first grade, often display a “painstaking decoding” approach to reading unknown words (Ehri & Snowling, 2004) After multiple, successful word reading experiences, readers will soon identify simple word types without attending to the individual letter-sound associations, beginning the process of unitization, making word reading much more efficient In more advanced word reading phases, the reader develops and relies more on graphophonemic combinations that represent larger functional units (e.g., -ing, able, tion) connections Thus, unitization establishes connections across increasingly larger units of mastery, from letter combinations to syllables to morphemes, to assist and expand word reading development and assist in developing automaticity and word understanding (Ehri, 2005b; Ehri & Snowling, 2004) Readers who recognize whole words instantly have reached the consolidated alphabetic phase, typically in second grade and beyond It is in this phase that unitization becomes the common approach to reading unknown words In addition to storing known words as units for automatic recognition and understanding, repeated encounters with words allow a reader to store frequently encountered letter patterns across different words For example, once a student learns the word light, if he or she develop an instant recognition of the unit -ight, he or she can use this knowledge to quickly decode new words such as flight, tight, and so on Upon encountering the word flight for the first time, a consolidated alphabetic reader would need to connect only two units: fl and -ight Similarly, students with this skill would be able to decode nonwords with the same approach If the nonword were symight, the student would connect the two units sym and -ight Perfetti and Hogaboam (1975) found that the most prominent characteristic of proficient third- and fifth-grade readers was their skill in reading multisyllabic nonwords and not their skills in reading sight words or more complex real words Harn et al / Development of Alphabetic Principle 145 Although using this unitization approach to reading a word is quicker than blending the individual phonemes, it is still not the same as having automaticity in word recognition Building automaticity in word recognition skills is dependent upon repeated successful exposures to word reading for meaning, which is highly dependent on the nature of the instruction and text reading experiences the student receives (Ehri & Snowling, 2004) Ehri (2005a) stated that automaticity “may be a separate phase that follows the consolidated phase during development and characterizes mature readers who recognize most words automatically by sight and who are facile if not automatic in decoding unfamiliar words” (p 151) However, these inferences have limited empirical demonstration across grades and skill levels of students (Ehri, 2005a) One reason for the lack of empirical support may have to with available measures of word reading development (Wagner et al., 1997) Measuring Alphabetic Principle A central argument in Ehri’s theory is that subtle changes in word reading skill carry important instructional implications, which we think is not only accurate but also represents a missed opportunity in the reading instruction most students receive Our inability to detect and reliably measure these subtle changes in an efficient manner underscores a missed opportunity for researchers and educators Furthermore, the theory’s emphasis on the development of unitization and automaticity has direct implications on instruction and assessment Schools need data that can be readily gathered and are instructionally relevant to support more efficient instruction Researchers need standardized and validated measures that elucidate critical changes in word reading development to better understand reading proficiency, which have yet to be identified or agreed upon (Ehri, 2005a; Fuchs, Fuchs, & Compton, 2004; Verhoeven & Perfetti, 2003) Measuring the Development of Word Reading When developing technically sound measures of reading proficiency to identify predictors of success, it is not necessary to base decisions on a theoretical model However, without a guiding theory, this endeavor does not necessarily provide valid and reliable data to inform instruction (Salvia & Ysseldyke, 2004; Samuels, 2006; Shinn, 1998) In addition, without a guiding theory that specifies the developmental details, measures adequate for identifying eventual success may be inadequate for capturing important developmental change (Wagner et al., 1997) Finally, a guiding theory helps anticipate and explain how relations among measures of reading skills and subskills change as skill level progresses within and across students (Gersten et al., 2005) Understanding and anticipating differences in learning based on developmental theory is particularly critical in the field of special education (Vaughn & Dammen, 2001) The changing, or temporal, relation of measures in the development of subcomponents of reading across time has been previously demonstrated For example, Mehta, Foorman, Branum-Martin, and Taylor (2005) demonstrated the changing role of phonological awareness across Grades through in the development of a range of readers; however, its predictiveness for the lowest performers in 1st grade remained strong Jenkins, Fuchs, van den Brock, Espin, and Deno (2003) also demonstrated differential relationships of measures based on the skill level of students in a sample of 4th-grade readers In this study, using a range of measures, researchers examined the relationship across skills/measures (word level, fluency, and comprehension) and found a strong relationship between a student’s performance on a measure of fluency with connected text and comprehension skills However, this relationship was moderated by a student’s initial skill level in text fluency For students with lower initial fluency skills, they found that word list reading skill had a significant and unique relationship to comprehension beyond a student’s fluency score For typical and high-performing readers, word list reading added no significant variance to predicting comprehension performance beyond text reading Understanding that relations between measures can vary both by skill level of the student and temporal/developmental phase is another area particularly relevant to special education research and practice (O’Connor & Jenkins, 1995) As we further refine and improve our interventions, we need to appreciate the role of instruction in impacting the relationships between measures (Wagner et al., 1997) One measure designed to be an efficient screener of beginning alphabetic principle that may also have a variable relationship with other measures depending on the skill level of the students (high vs low performers) as well as time of year is the Nonsense Word Fluency (NWF; Good & Kaminski, 2003) measure NWF is an individually administered, standardized measure developed within the framework of CurriculumBased Measurement (CBM; Deno, 1985; Shinn, 1989) Developed as an “indicator skill” of developing alphabetic principle, the examiner presents a student with a randomly ordered page of VC and CVC nonwords/pseudowords (e.g., vic, mej), and the student is given minute to produce as many letter sounds as possible Students can approach 146 Journal of Learning Disabilities Figure Scoring the Nonsense Word Fluency (NWF) Measure, Two Students With Different Skills in Unitization reading the nonwords at multiple levels of unitization Students can produce the individual sounds in the nonwords (/m/ /e/ /j/) or read the sounds as a unit (/mej/) or a combination of partial blending (e.g., /m/ /ej/) The traditional scoring approach has the examiner underline the letter sounds produced correctly and slash sounds produced incorrectly; the final score is the number of correct letter sounds produced within the minute This means that two students may receive the same score (i.e., 32) but receive that score in qualitatively different ways, which may provide meaningful information for instruction For example, Figure shows two students who received the same score but display different levels of unitization in their performance Student A displays a partial blending approach and can actually recode the whole word (/w/ /ub/ /wub/) yet is not automatic with the recoding or fully unitizing (/wub/) This student first must break the nonword into smaller units, an approach that Ehri (2005b) would hypothesize may negatively impact semantic activation Compared to Student B however, Student A appears to have a more advanced word reading strategy Notice that Student B also receives a score of 32 yet only produces individual letter sounds and made many errors The progressive benchmarks developed for this measure and the scoring methods were empirically derived without linkage to reading theory, much less Ehri’s model of word reading development and the role of unitization As a screening device, NWF is efficient in identifying students who are on track and those who are in need of additional support (Good & Kaminski, 2003) Even with strong psychometric properties, concerns have been raised regarding the instructional relevance of the measure, the limited time frame for when it is used, the possibility of ceiling effects or performance asymptotes due to the timed nature of the task, and the lack of discrimination between using individual letter sounds or whole words (Fuchs et al., 2004; Paris, 2005) The measure was designed, as are most screening and progress monitoring devices, to reliably identify who is at risk for reading difficulties, not to explain what they need to get back on track As an indicator of the alphabetic principle, it is not designed to determine which letter sounds or words the student can read but rather to determine if the student has an effective strategy for reading any novel word For this reason, the stimuli during testing are nonwords (i.e., not real words) to ensure that students are not reading memorized words, another important yet distinctly different reading skill (Ehri, 2005a) The issue of ceiling effects or performance asymptotes is implicitly acknowledged in the Dynamic Indicators of Basic Early Literacy Skills (DIBELS) system of focusing on critical benchmarks rather than the absolute level of performance Across DIBELS measures, what is important is if a student is above the benchmark by a specified time, not whether his or her score was 50 or 150, both of which are well above the benchmark for the beginning of first grade (i.e., 24) For research purposes however, when the NWF measure is used as a quantitative predictor of later reading skill, the distinction between 50 and 150 can matter greatly if the NWF measure is assumed to be linearly related to reading skill For this reason, we will take a closer look at how NWF scores relate to later oral reading fluency (ORF) during first grade The purpose of this study is to investigate the development of alphabetic principle in students first learning to read (i.e., first grade) First, we use a large sample of first-grade students to investigate the differential relationship of first-grade slope on the NWF measure to endof-year ORF based on beginning of the year alphabetic Harn et al / Development of Alphabetic Principle 147 principle skills More specifically, we hypothesize an interaction between initial NWF (fall) performance and NWF slope across first grade in predicting end-of-year ORF We expect this interaction effect to be negative, indicating that the higher the initial NWF performance, the weaker the relation of NWF slope to end-of-year ORF Second, we present a differential scoring approach designed to capture the development of unitization and examine its relation to a student’s ability to read connected text fluently We hypothesize that students with more developed unitization skills will read more fluently (i.e., higher ORF) at the end of first grade Method Participants First-grade students across two Pacific Northwest school districts (N = 938) were used to examine the relationship of NWF to ORF across the school year Both districts administered the NWF measure in the fall (approximately the third week of September), winter (approximately the second week of January), and spring (approximately the second week of May) to all first-grade students as part of their schoolwide approach to reading success (Kame’enui et al., 2005) In addition, they administered the ORF measure at the same time as the NWF but only in the winter and spring Five schools from District participated District is a predominantly White (71% White, 17% Latino), rural school district with a free and reduced lunch rate of 45% Seven schools from District participated District is a predominantly White (81% White, 13% Latino), suburban school district with a free and reduced lunch rate of 24% To examine the role of unitization in first grade, a second independent group of 109 students across the participating schools in District were randomly selected Measures DIBELS Nonsense Word Fluency (Good & Kaminski, 2003) The NWF task is a standardized, 1-minute, individually administered measure that assesses a student’s knowledge of the alphabetic principle and was described in detail earlier Published alternate form reliability for NWF ranges from 67 to 87, and concurrent validity with the readiness subtests of the Woodcock-Johnson PsychoEducational Test ranges from 35 to 66 In our sample, test-retest reliability for fall of first grade was 94 Modified scoring procedures for NWF The administration of the measure followed standard procedures, and examiners were instructed to explicitly mark how the student approached each nonword as previously discussed in Figure After administration, trained data entry staff then categorized each nonword on the test into one of four strategies related to the level of unitization for the nonword: sound only (/f/ /e/ /k/), sound by sound and then recode (/f/ /e/ /k/ /fek/), partial blend (/f/ /ek/), or unit (/fek/) We then computed for each student a proportion score for each strategy as the total frequency of using a particular strategy divided by the total number of nonwords attempted For example, if a student attempted 10 nonwords and used sounds times, partial blends times, and units times, the student would get proportion scores of 20, 40, 40, and for sound, partial blend, unit, and recode, respectively DIBELS Oral Reading Fluency (Good & Kaminski, 2003) The DIBELS Oral Reading Fluency task is a standardized, individually administered test of accuracy and fluency with connected text The ORF passages and procedures are based on the program of research and development of Curriculum-Based Measurement of reading described in Shinn (1989) The student is presented with a grade-level passage and asked to read the passage aloud The final score is the number of correct words read in minute Test-retest reliabilities for elementary students ranged from 92 to 97 while alternate form reliability of different reading passages drawn from the same level ranged from 89 to 94 (Tindal, Marston, & Deno, 1983) Criterion-related validity coefficients range from 72 to 91 on the Scholastic Aptitude Test and the Woodcock Reading Mastery Test Passage Comprehension measures (Good & Kaminski, 2002) Results Relation of Initial (Fall) NWF and Growth (Slope) on NWF to End-of-Year ORF Correlations and descriptive statistics for the firstgrade measures are shown in Table We started by using the established benchmarks for fall of first-grade NWF to break the sample into separate strata, at risk (0–10), some risk (11–24), and low risk (25 or more) as reported on the DIBELS Web site Within each of these larger strata (subgroups of performance) however, there was still considerable variation in fall NWF, so we further subdivided each into smaller substrata to examine this variability We plotted the NWF slope against spring ORF, conditional on (within) the substrata defined by fall NWF performance (see Note 1) Such plots are known as conditioning scatterplots or coplots for short (Cleveland, 1992) Figure displays the coplots with a fitted linear regression line (dashed line) and a fitted scatterplot smoother (lowess, solid line) added to each subplot The 148 Journal of Learning Disabilities Table Correlations and Descriptive Statistics for First Grade Dynamic Indicators of Basic Early Literacy Skills (DIBELS) Measures, Nonsense Word Fluency (NWF) and Oral Reading Fluency (ORF) Variable Fall NWF Fall NWF Winter NWF Spring NWF Winter ORF Spring ORF M (SD) 0.75 0.62 0.82 0.73 31.70 (25.11) Winter NWF Spring NWF Winter ORF Spring ORF 0.75 0.62 0.76 0.82 0.77 0.69 0.73 0.74 0.78 0.91 0.76 0.77 0.74 61.58 (31.22) 0.69 0.78 83.10 (37.92) 0.91 35.25 (33.52) 61.10 (37.77) Figure Subplots by Fall Nonsense Word Fluency (NWF) Performance Displaying Relation of Slope of NWF to Spring Oral Reading Fluency (ORF) 50 100 150 Fall NWF = 25-30 Fall NWF = 21-24 Fall NWF = 31-36 50 100 150 Fall NWF = 37-48 Fall NWF = 49+ 150 Spring ORF 100 50 Fall NWF = 2-6 Fall NWF = 0-1 Fall NWF = 7-10 Fall NWF = 11-15 Fall NWF = 16-20 150 100 50 0 50 100 150 50 100 150 50 100 150 NWF Fall to Spring Slope range of fall NWF is indicated in the top margin of each subplot and the subplots are arranged so that the fall NWF ranges increase moving from left to right and up the page (e.g., students with NWF scores between and in the fall are in the bottom left corner, students with scores between 21 and 24 are in the top left corner) Three things are apparent in Figure First, the relation between NWF slope and spring ORF is reasonably well represented by a linear regression because the smoother fit (solid line) approximates the linear regression fit (dashed line) Second, the only dramatic shift in the linear effect of the NWF slope on spring ORF is a substantial reduction in effect in the very highest substratum, when fall NWF is 49 or higher (right most plot in the top row) Finally, the residual variation about the regression line is not constant but rather increases as the range of fall NWF substrata increases (i.e., as student score on fall NWF goes up) To more formally model these features, we fit a regression model in which the substrata were coded as a series of nine dummy variables using the lowest substratum as the omitted reference group and contrasting the remaining nine substrata against the lowest substratum The effect of the NWF slope on spring ORF was allowed to vary across the substrata by including a set of interaction terms of the dummy variable contrasts and the NWF slope, and the residual variation was allowed to be different across substrata Results are shown in Table Only one of the first eight interaction terms is significant (the effect of NWF slope in the second substratum is significantly higher, t = 2.59, p < 01, than the first substratum), but there is no discernable pattern to the variation Harn et al / Development of Alphabetic Principle 149 in slope in the first eight strata, consistent with the coplot in Figure As a follow-up test, we tested the first eight interaction terms against the hypothesis that they were all zero Despite the significant interaction term for the second versus first substratum, the omnibus test generated a nonsignificant chi-square, χ2 = 11.61, df = 8, p = 17 We also tested the hypothesis that the effect of NWF slope on spring ORF changed linearly across the first eight strata (a linear by linear interaction), and this omnibus test generated a nonsignificant chi-square as well, χ2 = 2.61, df = 1, p = 11 Finally, we reestimated the model without one highly influential student in the second substratum (the highest point in the subplot in the bottom row, column of Figure 2), and the significant interaction effect in this substratum disappeared completely (t = 0.99, p = 32) The ninth interaction term in Table contrasting the highest substratum against the lowest substratum is highly significant (t = –3.58, p < 001) The omnibus test that all nine interaction terms are zero generates a highly significant chi-square, χ2 = 45.67, df = 9, p < 001 These tests suggest that the only meaningful difference in the effect of NWF slope on spring ORF is when the highest substratum is compared to the rest Thus, it would appear that the effect of the NWF slope is essentially the same for students with a fall score between and 48 and significantly reduced for students with a score of 49 or higher We also used a likelihood ratio test comparing a model with different residual variance by substrata to a model with constant residual variance and obtained a highly significant chi-square, χ2 = 117.49, df = 9, p < 001 The residual variance increased almost perfectly monotonically across the substrata (the correlation between the substrata order, 1–10, and the magnitude of the residual variance was 95) Thus, the modeling results verify the patterns in Figure 2, where only in the highest substratum does NWF slope appear to have a significantly different and reduced effect on spring ORF and the residual variation tends to increase across the substrata The estimated effect of NWF slope for the first nine substrata is 52, indicating that spring ORF scores go up on average about half a word for each additional unit gain in correct sounds on the NWF slope variable In other words, for two students who start in the same fall substratum, a student who gains 60 points on NWF over first grade would have a spring ORF score 20 points higher than a student who only gains 20 points on NWF For the highest substratum, the estimated effect of NWF slope is obtained by adding the main effect of NWF slope, 52, with the interaction effect, –.32, to obtain 20, which although drastically reduced is still significantly different from zero (t = 3.14, p = 002) Spring ORF Table Effects of Fall and Full Year Nonsense Word Fluency (NWF) Slope on Spring Oral Reading Fluency (ORF) Effect on Spring ORF Value SE t p (Intercept) NWF slope Fall NWF = to Fall NWF = to 10 Fall NWF = 11 to 15 Fall NWF = 16 to 20 Fall NWF = 21 to 24 Fall NWF = 25 to 30 Fall NWF = 31 to 36 Fall NWF = 37 to 48 Fall NWF = 49+ NWF slope by fall NWF = to NWF slope by fall NWF = to 10 NWF slope by fall NWF = 11 to 15 NWF slope by fall NWF = 16 to 20 NWF slope by fall NWF = 21 to 24 NWF slope by fall NWF = 25 to 30 NWF slope by fall NWF = 31 to 36 NWF slope by fall NWF = 37 to 48 NWF slope by fall NWF = 49+ –4.67 52 –7.89 5.61 16.36 18.50 20.04 27.77 39.95 56.01 105.27 3.86 06 5.33 6.20 6.31 6.02 5.53 6.63 5.30 6.59 5.67 –1.21 8.32 –1.48 90 2.59 3.08 3.63 4.19 7.54 8.49 18.58 23 00 14 37 01 00 00 00 00 00 00 23 09 2.59 01 11 10 1.16 24 00 11 00 1.00 06 10 62 53 10 09 1.08 28 17 11 1.51 13 01 09 12 91 00 11 00 1.00 –.32 09 –3.58 00 scores go up on average about a fifth of a word for each additional unit gain in correct sounds on the NWF slope variable If the same two students described previously started in the highest substratum in the fall, their spring ORF scores would on average differ by only eight words These results demonstrate that slope on NWF predicts spring ORF; however, this relation is moderated for students with higher initial NWF skills Examining the Role of Unitization to ORF in First Grade Descriptive statistics and correlations for the winter strategy (proportion) variables (sound, partial blend, recode, and unit) and NWF and ORF are shown in Table Several things are worth mentioning about the winter NWF test First, the number of nonwords attempted is correlated 99 with the traditional NWF score In our sample, if students attempted a nonword, they almost 150 Journal of Learning Disabilities Table Descriptive Statistics and Correlations Among Nonsense Word Fluency (NWF), Oral Reading Fluency (ORF), and Strategy Types Attempts Winter NWF Winter ORF Unit Partial blends Recode Sound only Spring NWF Spring ORF Fall NWF Winter Attempts Winter NWF Winter ORF Units Partial Blends Recode Sounds Spring NWF 63 63 73 40 01 –.27 –.28 43 56 99 76 63 –.03 –.40 –.44 73 71 77 61 –.04 –.38 –.43 72 72 57 –.06 –.35 –.40 62 85 –.27 –.46 –.74 55 66 –.08 –.03 04 06 –.16 –.39 –.43 –.38 –.49 74 always got it right (i.e., minimal errors) Second, the correlations of fall NWF with the winter strategy variables indicate that students with high fall NWF scores tended to use a unit strategy and not a recode or sound strategy Fall NWF was not related to partial blends These results not support the suggestion (Fuchs et al., 2004) that students get high scores on the NWF assessment by using a sound-by-sound strategy Finally, the correlations of the winter strategy variable with spring ORF are higher compared to winter ORF Correlations for the strategies of units and partial blends go up 11 and 12, respectively, and correlations for the strategies of recodes and sounds go down –.08 and –.09, respectively This suggests that the strategy variables predict change in ORF from winter to spring In fact, the correlations of units, partial blends, recodes, and sounds with simple ORF gain are 28, 20, –.21, and –.25, respectively, all significant at p ≤ 05 The pattern of results was similar for the partial correlations of the strategy variables with residualized ORF gain scores, but the p values were somewhat lower except for partial blends, which went up slightly to p = 056 To determine if the strategy variables added anything to the prediction of ORF over and above the traditional NWF score and over and above winter ORF, we ran each strategy variable against fall and winter NWF and their interaction as predictors of winter and spring ORF and ORF residualized gain scores We used a similar approach as before, allowing the residual ORF variation to increase with the fitted value of ORF (see Note 2) The pattern of results was the same across the three outcomes; units had the strongest effect (positive and significant in all three cases) followed by sounds (negative and significant for spring ORF and residualized ORF gain) The strategies of partial blends and recodes did not have significant effects in any of the models although the Spring ORF effect of partial blends changed from negative on winter ORF to positive on spring and residualized ORF gain These results indicate that the strategy variables carry useful information concerning reading progress in first grade beyond the traditional scores on NWF and ORF As a final step, we used three of the four strategy variables as simultaneous predictors of winter, spring, and residualized ORF gain Because the strategy variables sum to 1, the fourth variable is linearly dependent on the other three and only three can be used in the regression models We arbitrarily decided to omit the sounds variable from initial models Fall NWF and NWF change from fall to winter were mean centered about zero to increase interpretability Results for predicting winter ORF are shown in Table 4, Model 1, where only the unit strategy made a significant contribution and the effects of partial blends and recodes were not significantly different from zero and not significantly different from each other and indeed almost identical This suggests that these two effects could be dropped from the model in favor of sounds When this model is estimated (Model 2, right side of Table 4), only the unit strategy makes a significant positive contribution to predicting winter ORF The effect of the sound strategy is positive but very small and not significantly different from zero The estimated coefficient for the unit strategy suggests that students who exclusively used a unit strategy on NWF scored about 11 words higher on winter ORF than students who did not unitize at all Results for predicting spring ORF by strategy are shown in Table 5, Model 1, where only the unit strategy was significantly different from zero (positive) although the effect of the partial blends strategy was positive, marginally significant (p = 059) and not significantly different from the effect for the unit strategy; indeed, it was almost identical The recode strategy was negative but Harn et al / Development of Alphabetic Principle 151 Table Results for Prediction of Winter Oral Reading Fluency (ORF) With Nonsense Word Fluency (NWF) and NWF Strategy Variables Model Model Model Effect on Winter ORF Value p Value p (Intercept) Fall to winter NWF gain Fall NWF Unit Partial blends Recodes Sound only Fall to winter NWF gain by fall NWF 34.11 0.40 0.96 9.06 –3.38 –1.61 000 000 000 033 728 697 32.11 0.40 0.95 10.95 000 000 000 025 1.84 635 –0.01 136 –0.01 Table Results for Prediction of Spring Oral Reading Fluency (ORF) With Nonsense Word Fluency (NWF) and NWF Strategy Variables 131 Model Effect on Spring ORF Value p Value p (Intercept) Fall to winter NWF gain Fall NWF Unit Partial blends Recodes (Units + partial blends) Fall to winter NWF gain by fall NWF 50.65 0.57 1.00 29.19 29.28 –3.33 000 000 000 000 059 630 49.89 0.57 1.02 000 000 000 29.92 000 –0.01 050 –0.01 051 Note: Fall and fall to winter NWF gain centered about their respective means Note: Fall and fall to winter NWF gain centered about their respective means not significantly different from zero The shift in the effect for partial blends for winter ORF compared to spring ORF is very interesting, and the results suggest that the strategies of unit and partial blend could be combined and the strategies of recodes and sounds could be combined This results in an elegant, simple model with one strategy variable, the proportion of nonwords attempted using at least some level of unitization (i.e., unit or partial blend strategies) When this is done (see Table 5, Model 2), the effect for combined unit and partial blend strategy is significant and indicates that students who exclusively used these strategies (i.e., displayed some level of unitization) read on average 29 more words on spring ORF over students who exclusively used the sound or recode strategies after controlling for fall and winter NWF and their interaction Results for predicting spring ORF controlling for winter ORF, fall and winter NWF, and their interaction are shown in Table 6, Model 1, and where both units and partial blends were significant, recodes was not Interestingly, the introduction of winter ORF into the model substantially lowered the effect of units (B = 21.43) compared to the model without winter ORF (B = 29.19) but slightly increased the effect of partial blends (B = 31.91) compared to the model without winter ORF (B = 29.28) and also decreased the estimated standard error resulting in the significance of the partial blend effect (z = 2.62, p = 010) Model in Table shows similar results when sounds is used instead of recodes The effect of sounds is not significantly different from zero and small Evidently, using units on the NWF assessment is somewhat redundant with current ORF, but this is not true of partial blends, which appears to be an emerging Table Results for Prediction of Spring Oral Reading Fluency (ORF) With Winter ORF, Nonsense Word Fluency (NWF), and NWF Strategy Variables Model Model Effect on Spring ORF Value p Value p (Intercept) Fall to winter NWF gain Fall NWF Winter ORF Units Partial blends Recodes Sound only Fall to winter NWF gain by fall NWF 54.54 0.21 0.09 0.83 21.43 31.91 –2.74 000 064 545 000 000 010 638 51.80 0.21 0.09 0.83 24.17 34.65 000 064 545 000 001 007 2.74 638 0.00 365 0.00 365 Note: Fall, fall to winter NWF gain, and winter ORF centered about their respective means skill that is prognostic of future ORF progress Including winter ORF and the strategy variables also completely eliminated the effects of fall and winter NWF and their interaction compared to the model without winter ORF Thus, all of the effects of traditional winter and fall NWF scores on spring ORF appear to be mediated through winter ORF and the strategy variables Another potentially useful way to examine the utility of the strategy variables is to cluster students according to the strategies they used An empirical cluster analysis (k-means procedure; Hartigan & Wong, 1979) suggests that clusters is the optimal number of clusters The 152 Journal of Learning Disabilities Table Means of Strategy Variables by Cluster and Cluster Sizes Strategy Variable Cluster Cluster Label Sounds Recodes Partial Blends Units n Recode Partial blend Sound-unit Sound only Unit 13 29 11 86 02 77 07 13 04 01 02 39 13 02 00 09 25 63 08 97 13 12 13 24 47 means on the strategy variables for the clusters are shown in Table 7, and as is apparent, the largest clusters (4 and 5) were characterized by students who exclusively used the strategy of sounds (n = 24) or units (n = 47), respectively The other clusters included Cluster 3, which was composed of students (n = 13) who predominantly used units (.63) but also used sounds (.11), recodes (.13), and partial blends (.13) about equally; Cluster 2, which was composed of students who predominantly used partial blends (.39) but also used sounds (.29) and units (.25); and Cluster 1, which was composed of students who predominantly used recodes (.77) but also used some sounds (.13) and units (.09) Figure shows box plots of spring ORF by cluster and results of predicting spring ORF using the same model as described previously for the strategy variables as continuous predictors are shown in Table The sounds cluster (Cluster 4) was used as the omitted reference group and four dummy variables were used to contrast the remaining clusters against the sounds cluster Both the partial blends (3) and units (5) clusters were significantly higher than the sounds cluster; the recode (1) and sounds-words (2) clusters were both nonsignificantly different from the sounds cluster In a follow-up test, the sounds-words cluster was marginally higher than the recode cluster (p = 07) Discussion Results from this study provide a greater understanding of the nature of the NWF measure and its relation with ORF and preliminary information on a new, potentially more instructionally relevant scoring approach for the NWF measure The discussion will focus on the nature of the NWF measure and its scoring and how the measure and findings relate to Ehri’s theory of word reading development with implications for both researchers and practitioners Nature of the NWF Measure Using a large sample of first graders, we examined the role of (a) initial level (fall) and (b) linear change (slope) on the NWF measure across first grade (fall, winter, spring) to spring performance on ORF Not surprisingly, the NWF measure demonstrated moderate to strong correlations across assessment points (fall, winter, spring) to ORF in spring In addition, a linear relation adequately depicted the relation between NWF and ORF; as students’ scores on NWF improved across the year, there was a concomitant improvement in their scores on ORF However, for students who began first grade with established skills in alphabetic principle (i.e., NWF > 49), this linear effect dramatically attenuated This differential relationship between NWF and ORF based on time of year and skill level is important for two reasons First, effects of reading growth on some measures, such as the NWF measure, will be moderated by initial skill status, and thus typical growth analysis that does not account for initial skill level may be misleading in interpretation Students who begin first grade with an NWF score over 49 are well above the fall benchmark established for this measure (i.e., > 25) of alphabetic principle, and additional growth on NWF has much less value for further growth on ORF or reading development in general Our results may explain why some have found a limited relation between growth on NWF and later outcomes (Fuchs et al., 2004) and thus have questioned the utility of the NWF measure For example, the sample within the Fuchs et al (2004) study of “at-risk students” was small and had a mean fall NWF performance of 31.29, well above the risk criterion of 25 for the measure at the beginning of first grade Their sample mean is also quite similar to the 31.29 obtained on our random sample of fall first graders across multiple schools In addition, they reported that the mean performance on the Woodcock Reading Mastery Test was also in the average range compared to the normative sample Thus, it is Harn et al / Development of Alphabetic Principle 153 50 Spring ORF 100 150 Figure Box Plots of Spring Oral Reading Fluency (ORF) by Unitization Strategy on Winter Nonsense Word Fluency (NWF) N= 13 Recode 20 Sound 11 12 Sound-Unit Partial Blend 45 Unit Predominant Strategy Approach Table Cluster Membership Effects on Spring Oral Reading Fluency (ORF) Effect on Spring ORF Value SE t p Intercept Fall to winter Nonsense Word Fluency (NWF) gain Fall NWF Winter ORF Recode cluster Partial blend cluster Sound-unit cluster Unit cluster Fall to winter NWF gain by fall NWF 58.31 3.72 15.67 00 0.24 0.06 0.83 –2.68 16.81 9.19 15.91 0.12 0.16 0.12 5.14 6.47 5.85 5.46 2.02 0.36 6.94 –0.52 2.60 1.57 2.92 05 72 00 60 01 12 00 0.00 0.00 –1.14 26 Note: Fall, fall to winter NWF gain, and winter ORF centered about their respective means possible that the weak relations reported by Fuchs et al are in line with our results and reflect the more modest relation of NWF with ORF among initially higher performing readers Their results beg for reexamining subgroup differences and general model adequacy (e.g., linearity, normality, outliers, etc.) It would be interesting to examine the Fuchs et al results with this differential growth based on initial status (intercept) approach to see if the current findings are replicated within their intervention group Others report similar interactions but with different constructs, suggesting that differential growth by initial status may not be limited to just the NWF and ORF measures For example, Mehta et al (2005) found a changing role across time of phonological awareness in reading development, with the exception of the lowest performers where it remained highly predictive Second is to understand what the NWF measure actually assesses and how it fits within the DIBELS system of assessments The NWF measure was designed to be an efficient screener and indicator of beginning alphabetic principle development for use with most students starting from the middle of kindergarten and extending to the middle of first grade This developmental window is purposeful because the natures of the nonword types on 154 Journal of Learning Disabilities the test are simple (i.e., VC and CVC), the most common word types students of this age will be expected to read In addition, the criterion nature of the NWF device impacts how scores should be interpreted For example, if a student has a score of 60 on the NWF measure in fall of first grade, that student is well above the benchmark and most likely has well established alphabetic principle skills Trying to increase this student’s score on this measure is not in line with the nature of this system of measures or with our results, which suggest a point of diminishing returns for higher scores on initial NWF Students performing this well at the beginning of first grade most likely have mastered initial word reading and are using their more developed skills in phonological processing and unitization to read for understanding and thereby teachers should not expect their score to improve Understanding the criterion-referenced nature of these measures is critical for researchers when conducting analyses and educators when properly interpreting student performance Another measurement issue related to brief, timed screening measures is the potential issue of a plateau or ceiling effect Paris (2005) stressed that in measures of early literacy skills, such as phonological awareness or alphabetic principle, ceiling effects indicate a constrained skill and therefore focusing on measuring or intensively teaching skills of this type are a waste of time He gave an example from the nonword subtest of the Comprehensive Test of Phonological Processing (CTOPP; Torgesen, Wagner, & Rashotte, 1999), where according to the normative tables, 80% of students have mastered this skill by age years months We would argue that understanding the different phases of reading development is important In addition, most reading theories purposely imply that mastery of certain skills changes the relation of that subskill to higher order skills (Logan, 1988, 1997) A ceiling effect for a measure could also be a good thing from an instructional perspective For example, Ehri, as well as other reading theorists, acknowledge the separate and unique importance of developing skills to a level of automaticity It is not enough for students to simply reach a level of accuracy; certain skills need to become automatic to free cognitive resources to enable understanding and comprehension (Meyer & Felton, 1999; Nathan & Stanovich, 1991) A typical example is in word reading If a student knows the letter sounds /m/ /a/ /n/ but because of a lack of automaticity with lettersound skills struggles to retrieve them, it decreases the likelihood the student will read the word as a unit, which will negatively impact the student’s ability to access the meaning of the word and comprehension Measures such as NWF provide feedback not only on the student’s accuracy with letter sounds but more important, with his or her facility with blending sounds and unitizing In addition, understanding how subgroups (lower performers) learn is the critical charge of reading researchers and special education (Vaughn & Dammann, 2001) Continuing with Paris’s (2005) example of the CTOPP, the 20% who have not mastered the skills are most likely students with learning disabilities Paris indulged in a certain carelessness about a sizeable percentage of the population not at mastery that is disconcerting If we compared learning to read with the universally administered test for phenylketonuria (PKU, an inborn metabolic disorder) we believe a key point will be clear The percentage of the population effected by PKU is tiny (less than 1%), yet the condition produces lifelong intellectual devastation that is totally preventable through early detection and intervention Do we forget about PKU testing just because most of the population has mastered the ability to metabolize phenylalanine by birth? Clearly we need measures that are highly sensitive to identifying students at risk for long-term reading difficulties (i.e., measures of phonological processing) even if this is only a small portion of readers The consequences of letting these students go on without intervention are just too great Instructional Relevance and Implications for Unitization A study of a new scoring approach to the NWF measure was completed to determine if unitization could be measured but, more important, if it added unique information to understanding and predicting reading performance There were four strategy categories with varying levels of unitization skills that are related to Ehri’s discussion of unitization Students who used the sound-bysound strategy only provided the letter sounds for the nonword and did not attempt recoding or blending Students who used the recoding strategy initially said the sounds of each letter and then recoded the word (e.g., /t/ /e/ /m/ /tem/) Students who used the partial blending strategy displayed some level of unitization in that they would produce a larger unit of sound within the word beyond letter sounds (e.g., /t/ /em/, /wa/ /f/) Students who used the unit strategy simply read the nonword as a unit (e.g., /tem/, /waf/) Ehri’s theory would predict that students displaying a predominant approach to word reading using a higher level of unitization would be more proficient readers Ehri (2005b) would also predict that students at this age may be inconsistent or use multiple approaches of word reading (e.g., partial, full alphabetic) because they have not yet mastered the previous phase Harn et al / Development of Alphabetic Principle 155 This prediction is partially supported by the finding that many students (53%) used multiple strategies within the 1-minute assessment, but each student had a predominant strategy as well, as demonstrated in the cluster analysis Students who displayed unitization not only scored higher on NWF but more important, read significantly more fluently both in the winter (11 points higher on ORF) and spring (29 points higher on ORF) of first grade than students without this strategy It could be that these students have reached that “touching off” point in instant, accurate word recognition to facilitate successful text reading discussed long ago (1908) by Huey (as cited in Samuels, 2006) In addition, although students who used a partial blending strategy did not read more fluently in the winter, they grew more fluent by spring (i.e., greater ORF growth) compared to students with more limited skills in alphabetic principle or unitization (sounds or sound-recodes) That this new scoring approach adds unique variance above the traditional NWF scoring approach both in the fall and winter is meaningful from both a measurement and an intervention perspective Students with automaticity in unitizing may be in the consolidated alphabetic phase of Ehri’s word reading theory, where students have skills to automatically recognize chunks, letter combinations, or units within words, making them more efficient in word reading This unitizing strategy in winter of first grade accounted for unique variance in both winter and spring level of performance on ORF and in growth on ORF across winter and spring Instructionally, this could mean that students displaying unitization in the fall and winter of first grade are more advanced than classmates without unitization Teachers can use this information to assist in differentiating instruction to advance these students’ reading skills as well as strategically supporting developing recoding skills for other students (Ehri, 2005a) A related finding was that the unitizing strategy variable (when clustering students who used a unit or partial blending approach) was partially redundant with winter ORF performance When adding students’ winter ORF performance into the model to predict spring ORF, the effect of the unit strategy approach decreased In addition, students using a partial blending approach grew the most on ORF, more than students using the unit strategy, but did not necessarily end first grade with a higher score on ORF This could be related to a possible reciprocal relationship between alphabetic principle and fluency in connected text such that as students’ skill in reading connected text improves, their facility in alphabetic principle or decoding also improves (Chard, Pikulski, & McDonagh, 2006; Samuels, 2006) An additional reason could be related to the nature of the stimuli, or simple nonword types, on the NWF measure itself Students may be transitioning to a level of automaticity in these word types, thereby minimizing additional contribution when controlling for overall reading proficiency as measured by ORF Cluster analyses demonstrated that students who approached the nonwords using a nonunitizing strategy (i.e., sounds) scored significantly lower than unitizers on spring ORF Perhaps more interesting, however, were the sound-recoders, who went sound by sound and then recoded the nonword correctly These students were also significantly lower than unitizers on spring ORF and in fact had the lowest mean of all the clusters These students are unitizing, yet they seem to lack automaticity These may be readers Ehri discussed in the full alphabetic principle phase who are taking a “painstaking” approach to word reading (Ehri & Snowling, 2004) Despite the fact that they can recode the nonword, these students may not be attending to the larger unit that they are producing, even within these simple word types Within this sample, approximately one third of the first graders were clustered into the sounds or sound-recode groups, with 70% of these students not reaching the spring ORF benchmark In contrast, a little more than half of the students (54%) were in the unitizing clusters (unit, partial blends), where only 5% did not meet the spring ORF benchmark, about a sixfold difference in relative risk compared to the nonunitizers Definitive reasons why this happens are beyond the scope of the current study, but obviously the nature and consistency of the instruction these students have received, individual language differences, and other areas need to be considered Limitations The limitations of this preliminary study revolve around the sole use of DIBELS measures for two reasons First, these measures are designed to be efficient screening and progress monitoring tools; hence, they are brief, timed, fluency-based measures The efficiency effect of these measures obviously constrains the sample of student performance that we attempt to interpret for decision making In future studies, the use of nontimed measures would be critical to rule out the conclusion being biased by measurement type Second, the measures used within this study were designed to be indicators of, in this instance, alphabetic principle and fluency with connected text Needless to say, there are many other reading skills that are vital within this time frame that we did not collect Furthermore, using a range of measures that vary in the breadth and depth of literacy skill assessment is necessary to more fully investigate the 156 Journal of Learning Disabilities role of unitization as discussed by Ehri In addition, when we examined the utility of scoring using a strategy approach, the random sample (N = 109) was modest, and prior to wide scale use they should be replicated on a larger sample with a range of measures Unitization Although these results will need to be replicated with an independent sample and a range of measures, the results in general support Ehri’s discussion of unitization as a critical developmental process in word reading development Contrary to what some have implied, students who approach the NWF task using the sound only strategy not perform well on ORF, and growth on NWF was related to level and growth on ORF (Fuchs et al., 2004) Students who approached the NWF task at a more advanced unit level (whole word) may be categorized as in the full alphabetic phase and were quantitatively and qualitatively better readers in the middle and the end of the year on a measure highly predictive of future reading proficiency, which is a finding Ehri’s theory would have predicted Notes For simplicity, we used the Nonsense Word Fluency (NWF) gain score, spring NWF minus fall NWF, which is the same except for a scaling factor as the ordinary least squares slope based on fall, winter, and spring The NWF gain score is scaled in terms of full year change Because the sample size was much smaller, instead of using strata based on fall NWF, we used an exponential variance function to allow the residual variance to increase with the fitted value of spring Oral Reading Fluency (ORF) according to an exponential: var(eij) = exp(2δµij), where eij are the residuals, µij are the fitted spring ORF values, and δ is the estimated variance parameter See Pinheiro and Bates (2000) for more details on variance functions References Adams, M J (1990) Beginning to read: Thinking and learning about print Cambridge: Massachusetts Institute of Technology Carnine, D (2000) Why education experts resist effective practices (and what it would take to make education more like medicine) Washington, DC: Thomas B Fordham Foundation Chard, D J., Pikulski, J J., & McDonagh, S H (2006) The link between decoding and comprehension for struggling readers In T Rasinski, C L Z Blachowicz, & K Lems (Eds.), Fluency instruction: Research-based best practices (pp 39–61) New York: Guilford Cleveland, W S (1992) A science of graphical data display In R Penman & D Sless (Eds.), Designing information for people (pp 111–121) Canberra: Communication Research Institute of Australia Deno, S L (1985) Curriculum-Based Measurement: The emerging alternative Exceptional Children, 52, 219-232 Duncan, L G., & Seymour, P H K (2003) How children read multisyllabic words? Some preliminary observations Journal of Research in Reading, 26, 101-120 Ehri, L C (2002) Phases of acquisition in learning to read words and implications for teaching In R Stainthorp & P Tomlinson (Eds.), Learning and teaching reading (pp 7–28) London: British Journal of Educational Psychology Monograph Series II Ehri, L C (2005a) Development of sight word reading: Phases and findings In M J Snowling & C Hulme (Eds.), The science of reading: A handbook (pp 135–154) Cambridge, MA: Blackwell Ehri, L C (2005b) Learning to read words: Theory, findings, and issues Scientific Studies of Reading, 9, 167-188 Ehri, L., & Snowling, M J (2004) Development variation in word recognition In C A Stone, E R Silliman, B J Ehren, & K Apel (Eds.), Handbook of language and literacy: Development and disorders (pp 433–460) New York: Guilford Foorman, B R., Francis, D J., Shaywitz, S E., Shaywitz, B., & Fletcher, J M (1997) The case for early reading intervention In B Blachman (Ed.), Foundations of reading acquisition: Implications for intervention and dyslexia (pp 103–115) Mahwah, NJ: Lawrence Erlbaum Francis, D J., Shaywitz, S E., Stuebing, K K., Shaywitz, B A., & Fletcher, J M (1996) Developmental lag versus deficit models of reading disability: A longitudinal, individual growth curves analysis Journal of Educational Psychology, 88, 3-17 Fuchs, L S., Fuchs, D., & Compton, D L (2004) Monitoring early reading development in first grade: Word identification fluency versus nonsense word fluency Exceptional Children, 71, 7-21 Gersten, R., Fuchs, L., Compton, D., Coyne, M., Greenwood, C., & Innocenti, M (2005) Quality indicators for group experimental and quasi-experimental research in special education Exceptional Children, 71, 149-164 Good, R H., & Kaminski, R A (Eds.) (2002) Dynamic indicators of basic early literacy skills (6th ed.) Eugene, OR: Institute for the Development of Educational Achievement Good, R H., & Kaminski, R A (2003) DIBELS™: Dynamic indicators of basic early literacy skills (6th ed.) Longmont, CO: Sopris West Hartigan, J A., & Wong, M A (1979) A k-means clustering algorithm Applied Statistics, 28, 100-108 Jenkins, J R., Fuchs, L S., van den Broek, P., Espin, C., & Deno, S L (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 Kame’enui, E J., Good, R H., III, & Harn, B A (2005) Beginning reading failure and the quantification of risk In H L Heward, T E Heron, N A Neef, S M Peterson, D M Sainato, G Cartledge, et al (Eds.), Focus on behavior analysis in education: Achievements, challenges and opportunities (pp 69–89) Upper Saddle River, NJ: Pearson Education, Inc Logan, G D (1988) Toward an instance theory of automatization Psychological Review, 95, 492-527 Logan, G D (1997) Automaticity and reading: Perspectives from the instance theory of automatization Reading and Writing Quarterly: Overcoming Learning Difficulties, 13, 123-146 Mehta, P D., Foorman, B R., Branum-Martin, L., & Taylor, W P (2005) Literacy as a unidimensional multilevel construct: Harn et al / Development of Alphabetic Principle 157 Validation, sources of influence, and implications in a longitudinal study in grades to Scientific Studies of Reading, 9, 85-116 Meyer, M., & Felton, R (1999) Repeated reading to enhance fluency: Old approaches and new directions Annals of Dyslexia, 49, 283-306 Nathan, R G., & Stanovich, K E (1991) The causes and consequences of differences in reading fluency Theory into Practice, 30, 176-184 O’Connor, R E., & Jenkins, J R (1995) Improving the generalization of sound/symbol knowledge: Teaching spelling to kindergarten children with disabilities Journal of Special Education, 29, 255-275 Paris, S G (2005) Reinterpreting the development of reading skills Reading Research Quarterly, 40, 184-202 Perfetti, C A., & Bolger, D J (2004) The brain might read that way Scientific Studies of Reading, 8, 293-304 Perfetti, C A., & Hogaboam, T (1975) Relationship between single word decoding and reading comprehension skill Journal of Educational Psychology, 67, 461-469 Pinheiro, J C., & Bates, D M (2000) Mixed-effects models in S and S-PLUS New York: Springer Rack, J P., Snowling, M J., & Olson, R K (1992) The nonword reading deficit in developmental dyslexia: A review Reading Research Quarterly, 27, 28-53 Salvia, J., & Ysseldyke, J E (2004) Assessment (9th ed.) Boston: Houghton Mifflin Samuels, S J (2006) Reading fluency: Its past, present, and future In T Rasinski, C Blachowicz, & K Lems (Eds.) Fluency instruction: Research-based practices (pp.7–20) New York: Guilford Shinn, M R (1989) Curriculum-based measurement: Assessing special children New York: Guilford Shinn, M R (1998) Advanced applications of curriculum-based measurement New York: Guilford Stanovich, K E (1986) Cognitive processes and the reading problems of learning-disabled children: Evaluating the assumption of specificity In J K Torgesen & B Y L Wong (Eds.), Psychological and educational perspectives on LD (pp 75–103) San Diego, CA: Academic Press Stanovich, K E (1988) Explaining the differences between the dyslexic and the garden-variety poor reader: The phonologicalcore variable-difference model Journal of Learning Disabilities, 21, 590-604, 612 Tindal, G., Marston, D., & Deno, S L (1983) The reliability of direct and repeated measurement Minneapolis: University of Minnesota Institute for Research on Learning Disability Torgesen, J K (2000) Individual differences in response to early interventions in reading: The lingering problem of treatment resisters Learning Disabilities Research & Practice, 15, 55-64 Torgesen, J K., Rashotte, C A., & Alexander, A (2001) Principles of fluency instruction in reading: Relationships with established empirical outcomes In M Wolf (Ed.), Dyslexia, fluency, and the brain (pp 335–355) Parkton, MD: York Press Torgesen, J K., Wagner, R K., & Rashotte, C A (1999) Preventing reading failure in young children with phonological processing disabilities: Group and individual responses to instruction Journal of Educational Psychology, 91, 579-593 Vaughn, S., & Dammann, J E (2001) Science and sanity in special education Behavioral Disorders, 27, 21-29 Verhoeven, L., & Perfetti, C (2003) Introduction to this special issue: The role of morphology in learning to read Scientific Studies of Reading, 7, 209-217 Wagner, R., Torgesen, J., Rashotte, C., Hecht, S., Barker, T., Burgess, S., et al (1997) Changing relations between phonological processing abilities and word-level reading as children develop from beginning to skilled readers: A five-year longitudinal study Developmental Psychology, 33, 468-479 Beth A Harn, PhD, is an assistant professor in the Department of Special Education at the University of Oregon She has expertise in instructional design, assessment for instructional decision making, prevention and early intervention, and systemlevel coordination of instructional decision making in the area of elementary reading Mike Stoolmiller, PhD, is a research associate at the Center on Teaching Learning at the University of Oregon Dr Stoolmiller is nationally recognized for his expertise in latent growth modeling, structural equation modeling, and methodology His publications include conceptual and quantitative articles investigating the development of antisocial behavior and reading development David J Chard, PhD, is dean of the School of Education and Human Development at Southern Methodist University His research and teaching interests are in the area of early literacy and mathematics development for struggling students and students with learning disabilities ... 24 Fall NWF = 25 to 30 Fall NWF = 31 to 36 Fall NWF = 37 to 48 Fall NWF = 49+ NWF slope by fall NWF = to NWF slope by fall NWF = to 10 NWF slope by fall NWF = 11 to 15 NWF slope by fall NWF = 16... fall NWF = 16 to 20 NWF slope by fall NWF = 21 to 24 NWF slope by fall NWF = 25 to 30 NWF slope by fall NWF = 31 to 36 NWF slope by fall NWF = 37 to 48 NWF slope by fall NWF = 49+ –4.67 52 –7.89... ORF 100 50 Fall NWF = 2-6 Fall NWF = 0-1 Fall NWF = 7-10 Fall NWF = 11-15 Fall NWF = 16-20 150 100 50 0 50 100 150 50 100 150 50 100 150 NWF Fall to Spring Slope range of fall NWF is indicated