Reading and Writing

, Volume 25, Issue 1, pp 251–281

The reading performance of English learners in grades 1–3: the role of initial status and growth on reading fluency in Spanish and English

Authors

    • Center on Teaching & Learning5292 University of Oregon
  • Yonghan Park
    • Center on Teaching & Learning5292 University of Oregon
  • Scott K. Baker
    • Center on Teaching & Learning5292 University of Oregon
Article

DOI: 10.1007/s11145-010-9261-z

Cite this article as:
Baker, D.L., Park, Y. & Baker, S.K. Read Writ (2012) 25: 251. doi:10.1007/s11145-010-9261-z

Abstract

The purposes of this study were to (a) examine the developmental patterns in pseudoword reading and oral reading fluency in Spanish and English for Spanish-speaking English learners (ELs) in grades 1–3, and (b) investigate whether initial status and growth rates in reading fluency in Spanish and English, significantly predicted reading comprehension within languages and across languages. Participants were 173 Spanish-speaking ELs in first grade, 156 ELs in second grade, and 142 ELs in third grade across four schools providing a paired bilingual reading program. Results of hierarchical linear modeling indicated different patterns of reading growth in Spanish and English across measures and across grades. ELs at the beginning of first grade had higher scores on pseudoword reading in Spanish than in English and had a higher rate of growth on Spanish pseudoword reading. In second and third grades, initial scores on oral reading fluency were comparable in both languages, but oral reading fluency growth rates were higher in English than in Spanish. Results from regression and path analysis indicated that student initial scores and growth on reading fluency were strong and direct predictors of their reading comprehension within the same language, but not across different languages.

Keywords

BiliteracyEnglish learnersFluencyEarly readingComprehension

Introduction

Extensive research indicates that successful reading development in the early grades is the best assurance schools and parents have that children are on track to read at grade level by the end of third grade and beyond (Kame’enui, Simmons, & Coyne, 2000; Simmons, Kame’enui, Harn, Cole, & Braun, 2002). Risk factors for poor reading development include family poverty and children whose primary language is not English. Despite years of evidence demonstrating that language factors are associated with academic achievement and policies that focus on altering this association, English learners (ELs) in the United States, on average, continue to read well below grade level standards, and lag behind other students in their reading and academic performance. For example, according to the National Assessment of Educational Progress (NAEP), in 2009, ELs scored 36 points lower than non-ELs on the fourth grade national reading assessment (SD = 35, Cohen’s d = 1.03), and 47 points lower than non-ELs on the eighth grade reading assessment (SD = 34, Cohen’s d = 1.38; National Center for Education Statistics, 2010).

Given that approximately 80% of ELs in the United States are Spanish-speakers (Capps, Fix, Murray, Ost, & Passel, 2005), examining the reading trajectory of this group of ELs in the early grades is important in understanding how to effectively serve ELs generally, and Spanish-speaking ELs specifically. This focus is also important because Latinos are the fastest growing ethnic minority group in the United States and the majority of them speak Spanish as their native language (Hamilton, Martin, & Ventura, 2009).

One issue that has dominated how to best serve ELs is language of instruction, particularly in the early grades when students are just beginning school (August & Hakuta, 1997). Several syntheses have addressed whether academic outcomes are higher when ELs are instructed in their native language or in English (Francis, Lesaux, & August, 2006; Greene, 1997; Rolstad, Mahoney, & Glass, 2005; Slavin & Cheung, 2005). The general conclusion has been that native language instruction in an early transition model (i.e., where students learn to read in their native language first, followed by learning to read in the second language) or a paired bilingual model (i.e., where students learn to read in their native language and in English simultaneously) results in better outcomes than English-only instruction (Goldenberg, 2008). There is also general agreement, however, that important shortcomings in the language of instruction studies reduce the strength of this general conclusion. Shortcomings include the following: (a) there are a few studies overall on which to base this conclusion; (b) many of the studies were conducted some time ago, and educational conditions have changed appreciably since then; (c) the quality of the studies varies considerably, and (d) few details are provided in the studies to discern how instruction was organized and delivered (Baker & Baker, 2008; Francis, Lesaux, & August, 2006; Gersten & Baker, 2000a, b).

Currently, approximately 50% of ELs in the United States receive reading instruction in their native language, either in a paired bilingual program or a transition program (Goldenberg, 2008). Cummins’ (1981) interdependence hypothesis has been highly influential in providing the theoretical rationale for native language instruction. This hypothesis suggests there is an interaction between the development of a second language and the type of competence in the native language (L1) the child has developed when intensive exposure to the second language (L2) begins. In other words, the level of proficiency an individual attains in their first language influences how easy it is for that person to acquire proficiency in a second language. This linguistic interdependence, as Cummins termed it, is reflected in both written and spoken language. Of interest in the current study is the interdependence between L1 and L2 when the learning task is related to specific reading objectives (e.g., pseudoword reading, reading connected text, answering multiple-choice comprehension questions) in a context-reduced communication where the reader has to rely entirely on linguistic cues to interpret the meaning and logic of words, as opposed to a context-enriched communication where there is the opportunity to negotiate the meaning of words with others (Cummins, 1991; Verhoeven, 1994).

Most empirical studies of the interdependence hypothesis have addressed the impact of level of performance of L1 reading skills on the level of performance of L2 reading skills, either at one point in time or at two points in time, focusing on the difference between time one and time two (Cárdenas-Hagan, Carson, & Pollard-Durodola, 2007; Cirino et al., 2009; Dickinson, McCabe, Clark-Chiarelli, & Wolf, 2004; Durgunoglu, Nagy, & Hancin, 1993; Verhoeven, 1994). These studies have addressed the interdependence of L1 and L2 in areas related to phonological awareness, syntactic awareness, decoding, vocabulary, and comprehension. To date, few studies have examined the interdependence of reading fluency in L1 and L2 (Dominguez De Ramírez & Shapiro, 2007), which is the emphasis in the current study.

We are also not aware of studies that have examined whether the relation between reading skills in L1 and L2 is affected by the amount of growth students make on pseudoword reading and oral reading fluency in both languages. In the next section we consider studies that have examined the effect of growth on pseudoword reading and oral reading fluency on end-of-year reading outcomes in one language only, either in English or in Spanish. These studies provide a rationale for studying whether fluency growth in one language influences growth in the second language.

Measuring reading growth over time

Formative reading assessments in English and Spanish currently being investigated in the early academic years include pseudoword reading in kindergarten and first grade, and oral reading fluency from the middle of first grade on (Baker, Park, & Baker, in press; Baker et al., 2008; Crosson & Lesaux, 2009; Fien et al., 2008; Good, Baker, & Peyton, 2009; Leafstedt, Richards, & Gerber, 2004; Dominguez De Ramírez & Shapiro, 2006, 2007). A pseudoword reading fluency measure contains non-words (e.g., “lut” in an English measure and “panu” in a Spanish measure) and students must apply the alphabetic principle to sound out or read the pseudowords correctly. An oral reading fluency measure assesses a student’s proficiency in reading connected text with speed and accuracy. In this study we use the general term “fluency” to refer to students being able to read pseudowords in first grade, and real words in connected text in second and third grades with speed and accuracy in both, Spanish and English. Although this definition of fluency is somewhat narrow, it has considerable standing in educational research (Crosson & Lesaux, 2009; Deno, Mirkin, & Chiang, 1982; Fuchs, Deno, & Mirkin, 1984; Fuchs, Fuchs, Hosp, & Jenkins, 2001; Klauda & Guthrie, 2007; Shinn, Good, Knutson, Tilly, & Collins, 1992; Wolf & Katzir-Cohen, 2001).

As measures of important dimensions of reading proficiency, pseudoword reading and oral reading fluency are backed by strong theoretical support (LaBerge & Samuels, 1974; Perfetti, 1985; Posner & Snyder, 1975). The most direct connection is LaBerge and Samuels’ (1974) automaticity theory which suggests that the reader’s ability to read words automatically is closely related to reading comprehension because effortless word reading frees up mental resources that can be devoted to reading comprehension. Perfetti (1985) makes a similar case in his Verbal Efficiency Theory. Readers who are highly efficient in basic reading skills, including phonemic awareness, and understanding the alphabetic principle, can direct their processing resources to comprehension. Posner and Snyder (1975) also posit a model of word reading that links directly to the concept of fluency as we use it in the current study. They propose that two context-based expectancy processes facilitate word recognition. The first process operates when the context of the passage activates the mental location of word meaning that automatically spreads to related semantic memory locations, privileging the retrieval of some words over others. The second process requires the reader to use conscious processes to identify words. These two processes account for context facilitation of word recognition, implying that rate and accuracy increases when reading is done in the context of connected text versus contexts divorced from meaning such as reading words in isolation.

Measuring pseudoword reading

Reading pseudowords is a strong indicator of student understanding of the alphabetic principle, that is, how well one is able to use letter-sound correspondence to read words (Ehri, 2006). For example, of the 38 studies included in the National Reading Panel report (NRP, 2000) on English interventions to teach reading using phonics (i.e., the instructional approach associated with explicit instruction in the alphabetic principle), 18 included a measure of pseudoword reading to determine intervention impact, and the association between mastery of the alphabetic principle and reading comprehension at the end of first grade (Ehri, 2006; NRP, 2000).

More recent studies have focused on whether a measure of pseudoword reading that emphasizes automaticity or fluency provides useful information on overall reading development. The idea is whether growth on how well students are learning to read pseudowords—for example, students being able to read more pseudowords in the middle of first grade compared to the beginning of first grade—is associated with the amount of growth students are making in terms of overall reading proficiency.

Good et al. (2009) conducted the first study on the relation between growth on pseudoword reading and end of year performance on oral reading fluency for first grade students attending schools participating in a large-scale reading initiative (i.e., Reading First) in which students in the early grades received extensive explicit instruction in the alphabetic principle (i.e., decoding). Results indicated that growth on pseudoword reading during the first semester of first grade accounted for 27% of the variance on oral reading fluency at the end of first grade for students at high risk of reading difficulties (e.g., poor decoders), and 9% of the variance for students at low risk for reading difficulties (e.g., good decoders). In other words, the growth poorer readers made on the alphabetic principle accounted for more of the variance on an important reading outcome measure (i.e., oral reading fluency) than the growth students made who were already relatively strong readers, that is, proficient in the alphabetic principle.

In a similar study that included a measure of reading comprehension as an outcome, Fien, Park, Baker, Smith, Stoolmiller, and Kame’enui (in press) examined the relation between initial status and growth across Nonsense Word Fluency (NWF) (i.e., a measure of pseudoword reading, Good & Kaminski, 2002) in first grade on oral reading fluency and reading comprehension. Participants were 3,506 first graders attending 50 schools participating in the large-scale reading initiative (i.e., Reading First). Findings indicated that for students who started the year with low to relatively high decoding skills, initial status and gains on NWF were strongly associated with oral reading fluency and reading comprehension scores at the end of the year. Also, the findings replicated Good et al. (2009), by demonstrating that the value of NWF growth on the two outcome measures, above and beyond initial status on NWF, was strongest for those students who began the year with low decoding skills.

Finally, a study by Baker, Park, and Baker (in press) examined the effect of initial status and growth on a Spanish pseudoword reading measure in kindergarten and first grade on Spanish reading comprehension at the end of first grade. Participants were 165 Spanish-speaking ELs who were learning to read in Spanish and English in schools participating in Reading First. Results indicated that 53% of the variance in Spanish reading comprehension at the end of first grade was explained by initial status on pseudoword reading in Spanish in the middle of kindergarten, scores at the end of kindergarten on a summative assessment (i.e., the Aprenda-3 Preprimario 2, Harcourt, 2005), and growth in pseudoword reading in Spanish from the middle of kindergarten to the end of first grade.

In summary, a small number of studies demonstrate that initial status and growth on measures of pseudoword in English and Spanish in kindergarten and first grade predict important outcomes of reading proficiency at the end of first grade, including reading comprehension and oral reading fluency. It also appears that the growth made by struggling readers on pseudoword reading in the early grades (i.e., kindergarten and first grade) is a stronger predictor of general reading outcomes than the growth made by students who decode proficiently and are on track for successful reading outcomes.

Measuring oral reading fluency

Several studies have looked at oral reading fluency as a direct index of reading growth over time (Baker & Good, 1995; Baker et al., 2008; Deno, Fuchs, Marston, & Shin, 2001; Dominguez De Ramírez & Shapiro, 2006; Francis et al., 2008; Fuchs, Fuchs, Hamlett, Walz, & Germann, 1993; Jenkins, Graff, & Miglioretti, 2009; Schilling, Carlisle, Scott, & Zeng, 2007). We will review the three studies that reported having more than 30% ELs in their sample given our interest in examining specifically the role of initial status and growth of oral reading fluency on reading comprehension for Spanish-speaking ELs. Baker et al. (2008) investigated the use of oral reading fluency growth as a predictor of reading proficiency for all students in grades 1 through 3 in low-performing, high poverty schools. Results indicated that oral reading fluency growth accounted for a statistically significant percentage of the unique variance explained in reading comprehension at the end of the school year in all three grades. Although approximately 32% of the students in the sample were ELs, this study did not address specifically whether these findings were generalizable to this group of students.

In the first published study on the use of oral reading fluency with ELs, Baker and Good (1995) examined whether growth rates on oral reading fluency in second grade were comparable for ELs and non-ELs. Fifty ELs who spoke English and Spanish and 26 English-only students were assessed on oral reading fluency twice a week for 10 weeks. Findings indicated that (a) oral reading fluency was highly reliable for both groups (above 90%), (b) the aggregate of 5–10 assessments was a strong predictor of reading proficiency at the end of second grade, and (c) oral reading fluency was a more sensitive measure of growth for ELs than non-ELs.

Dominguez De Ramírez and Shapiro (2006) built on the study by Baker and Good (1995) by examining the oral reading fluency growth rates among Spanish-speaking ELs across grades 1 through 5 in both English and Spanish. Students attended an early transition program in which the transition from Spanish to English reading instruction increased as students moved up in grade. Results indicated that ELs in first grade had higher gains on oral reading fluency in Spanish than in English, but from second to fifth grade ELs had higher gains on oral reading fluency in English than in Spanish. The pattern of results is aligned with the instructional focus of the transition program, supporting the conclusion that oral reading fluency is sensitive to instruction. In first grade, instruction focused predominantly on Spanish reading acquisition, while in the subsequent grades the focus was on English reading acquisition. This study did not include an overall measure of reading proficiency, however, so it is not clear whether growth on oral reading fluency was associated with overall improvements in reading development. The studies by Baker and Good (1995) and Dominguez De Ramírez and Shapiro (2006) do suggest that oral reading fluency growth may provide a useful component for evaluating the impact of bilingual programs for ELs.

Purpose

Our current study has two main purposes. The first is to build on earlier studies by examining growth rates on pseudoword reading fluency in first grade and oral reading fluency in second and third grades for ELs attending a paired bilingual program. The fact that instruction in the paired bilingual program focused simultaneously on the acquisition of reading skills in both Spanish and English in all three grades is very important in this study conceptually and methodologically. Because an instructional emphasis was placed on reading development in both languages simultaneously, the context of reading acquisition is quite different from transition models where the instructional emphasis is primarily sequential (first one language, then the other).

The second purpose is to examine whether growth on pseudoword reading and oral reading fluency in Spanish and English have a significant effect on reading comprehension within and across languages. We examine the value of growth above and beyond the association between initial status on pseudoword reading and oral reading fluency and performance on reading comprehension. Specifically, we address the following questions:
  1. 1.

    What are the initial status levels and growth rates in English and Spanish reading fluency (pseudoword reading in first grade, oral reading fluency in second and third grades) for Spanish-speaking ELs receiving reading instruction in both English and Spanish? Do these performance levels and growth rates differ by language and by grade?

     
  2. 2.

    What is the contribution of pseudoword reading and oral reading fluency growth in English and Spanish to the prediction of reading comprehension within and across languages?

     

Method

Participants

Participants were 471 Spanish-speaking ELs in grade 1 through grade 3 enrolled in four schools during the 2006–2007 and 2007–2008 school years. By grade, 173 students were in first grade, 156 in second grade, and 142 in third grade across the four schools. Fifty-three percent were girls and approximately 4.5% were in special education.

Two of the schools were rural and two were urban, and the total number of students in the schools ranged from 188 students to 629 students. The percent of Hispanic students in the schools ranged from 44 to 73%. The range of students who received free or reduced rate meals, which is the primary index used to determine student poverty status ranged from 74 to 88%. Two of the schools received schoolwide Title 1 funding.

These four schools were part of a state reading initiative designed to improve reading instruction and student reading outcomes in grades K to 3. They all provided reading instruction for Spanish-speaking ELs in both English and Spanish 5 days a week for the entire school year. Schools provided 90 min of Spanish reading instruction in grades 1 and 2, and 60 min of Spanish reading instruction in grade 3. In addition, schools provided between 30 and 60 min of English reading instruction daily. All schools used the same core reading program in Spanish and English published by Houghton Mifflin (2003a, b).

Within the 90 min of Spanish reading instruction in first and second grades, each school provided at least 30 min of small group instruction daily. During small group instruction students were grouped homogenously by ability, determined on the basis of regular reading assessments. Groups were flexible and students moved among groups based on progress monitoring assessments. For students who were struggling to read in Spanish, small group instruction focused on reteaching lesson content taught during whole-group instruction. For students at grade level, small group instruction consisted of primarily independent reading or reading core program selections with an assistant. Two of the schools also used the Spanish version of Read Naturally (1998) to increase student fluency skills.

English reading instruction was also provided in homogenous reading groups determined by regular reading assessments. Students who were below benchmark according to the DIBELS assessment (see a description of the measure in the measures section) at the beginning and middle of the year, received English small group instruction using a research based explicit reading program such as Horizons (Science Research Associates, 1981), Reading Mastery (Engelmann & Bruner, 1995), or Read Well (Sprick, Howard, & Fidanque, 1998). ELs who were reading at benchmark level or above in English received small group reading instruction with English only students.

Spanish and English teachers and instructional assistants received extensive professional development on how to use the core reading program and supplemental programs with fidelity. In addition, expert researchers and trained teachers visited the schools to observe, model, and provide feedback to teachers on the delivery of instruction in both languages. Researchers and expert teachers also participated in the monthly data meetings where student reading performance in Spanish and English was analyzed using formative assessments. Small group instruction was modified using results of these assessments. In addition to English reading instruction, all schools also provided 30 min of English language development (ELD) using a language program approved by the state Department of Education.

English measures

Nonsense word fluency (NWF)

NWF is a standardized, individually administered test of the alphabetic principle. It is a subtest of the Dynamic Indicators of Beginning Early Literacy Skills (DIBELS, Good & Kaminski, 2002). Successful performance on NWF indicates knowledge of (a) letter-sound correspondences, in which letters represent their most common sounds, and (b) how to blend letter-sounds into whole units (i.e., pseudowords). According to Good and Kaminski (2002), alternate-form reliability coefficients for NWF ranged from 0.67 to 0.87, and concurrent validity coefficients with the readiness subtests of the Woodcock-Johnson Psycho-Educational Test ranged from 0.35 to 0.55. Recent studies indicated moderate correlations (r = 0.56) between NWF at the end of kindergarten and the SAT-10 reading comprehension subtest at the end of first grade (Good et al., 2009; Fien et al., 2008). In this study, we used student NWF scores in the beginning, the middle, and the end of first grade.

Oral reading fluency (ORF)

ORF is a standardized, timed, individually administered test of accuracy and fluency with reading connected text. We used DIBELS ORF, a subtest of DIBELS (Good & Kaminski, 2002). Oral reading fluency is designed to (a) identify children who may need additional instructional support, and (b) monitor progress toward instructional goals. Reading passages are calibrated for each relevant grade level, and the median number of words students read correctly across three different passages is reported. Students read each passage for 1 min. Words omitted, substituted, and hesitations of more than 3-s are scored as errors. Words self-corrected within 2 s are scored as accurate. In previous studies, alternate-form reliability coefficients of different reading passages from the same level of difficulty have ranged from 0.89 to 0.94 (Good & Kaminski, 2002). In Oregon, the correlation between ORF and the Oregon Assessment of Knowledge and Skills (OAKS) reading measure at the end of third grade was reported as 0.67 (Good, Simmons, & Kame’enui, 2001). In this study, we used ORF scores in the beginning, the middle, and the end of the year in second and third grades.

Stanford achievement test: tenth edition (SAT-10)

The SAT-10 is a group administered, multiple-choice standardized test of reading achievement (Harcourt Educational Measurement, 2002). The measure is not timed, although guidelines with flexible time recommendations are given. The internal consistency reliability coefficients for the total reading score were 0.97 at grade 1 and 0.95 at grade 2. The correlations between the SAT-10 total reading scale and the Otis-Lennon School Ability Test ranged from 0.61 to 0.74. In this study, we used the reading comprehension subtest to measure comprehension of reading connected text. The SAT-10 reading comprehension subtest assesses student comprehension of three types of material: (a) literary, (b) informational, and (c) functional. Students are asked to read ten different types of material and answer multiple choice comprehension questions. In this study, we used the reading comprehension subtest as an English reading performance outcome at the end of first, and second grades.

Oregon assessment of knowledge and skills (OAKS)

Student performance on the OAKS was used to determine grade level reading proficiency in third grade. The OAKS is an untimed, multiple-choice computerized test administered yearly to all students in Oregon beginning in third grade. Reading passages representing literary, informative, and practical selections are included in the third grade test. These passages are intended to represent selections that students might encounter in both school settings and in other daily reading activities. The OAKS assesses seven essential reading skills including (a) the understanding of word meanings in the context of a selection, (b) locating information in common resources, (c) answering literal comprehension questions, (d) answering inferential comprehension questions, (e) answering evaluative comprehension questions, (f) recognizing common literary forms such as novels, short stories, poetry, and folk tales, and (g) analyzing the use of literary elements and devices such as plot, setting, personification, and metaphor. The Oregon Department of Education reports that the correlation between OAKS and the California Achievement Test was 0.75 and the correlation between the OAKS and the Iowa Test of Basic Skills was 0.78. The four alternate forms used in the OAKS demonstrated the internal consistency reliability of 0.95 (Oregon Department of Education, 2008). In this study, we used the third grade OAKS assessment as an English reading outcome at the end of third grade.

Spanish Measures

IDEL Fluidez en las Palabras sin Sentido (FPS)

FPS is a subtest of the Indicadores Dinámicos del Éxito en la Lectura (IDEL; Baker, Good, Knutson, & Watson, 2006). It is a standardized, individually administered test of the alphabetic principle and it is similar in structure to the DIBELS NWF in English. An important noticeable difference between NWF and FPS is that on the FPS, CV and CVCV nonsense words were used (e.g., lu, mosi), whereas on the NWF task, VC and CVC nonsense words were used (e.g., ug, lut). The reason is because, in Spanish, the most frequent words have a CV or CVCV pattern (e.g., la and sala; Guirao & de Manrique, 1972; Honig, Diamond, & Gutlohn, 2000). In a pilot study, the 3-week, alternate-form reliability of FPS in the middle of first grade was 0.76 (Baker, Good, Peyton, & Watson, 2004). The concurrent validity of FPS with the Woodcock-Muñoz Pruebas de Aprovechamiento subtest of Análisis de Palabras was 0.72 at the end of first grade (Watson, 2004). In this study, we used student FPS scores in the beginning, the middle, and the end of first grade.

IDEL Fluidez en la Lectura Oral (FLO)

FLO is a standardized, timed, individually administered test of accuracy and fluency with reading connected text in Spanish. It is a subtest of IDEL (Baker, Good, Knutson, & Watson, 2006). Passages were written taking into account sentence length, number of high frequency words, and number of letters and syllables in words. Administration and scoring of the measure is the same as those of the DIBELS ORF measure. Alternate-form reliability of different reading passages from the same level of difficulty ranged from 0.88 to 0.94 (Baker, 2005). Criterion-related validity with the Woodcock-Muñoz average score was 0.75 (Watson, 2004). Three scores on this measure across the school year (beginning, middle, and end) were included in the analysis for second and third grades.

La prueba de logros en español, Tercera edición (Aprenda-3)

The Aprenda-3 is the Spanish version of the SAT-10 (Harcourt Educational Measurement, 2005). The internal consistency reliability coefficients for this test were 0.96 in first grade, 0.93 in second grade, and 0.94 in third grade. The correlations between the Aprenda-3 and the Naglieri Nonverbal Ability Test ranged from 0.75 to 0.96 in first to third grades. In this study, we used the Aprenda reading comprehension subtest scores as a Spanish reading outcome at the end of first, second, and third grades. This subtest assesses student reading comprehension in Spanish and its administration and scoring procedures are similar to those for the SAT-10.

Data collection procedure

IDEL and DIBELS measures were administered by examiners selected by the school for their proficiency in test administration and their knowledge of English and Spanish. Teachers or assistants who had an advanced level of Spanish proficiency or who were native Spanish-speakers collected all Spanish data. Test administrators received a full day of training on administration and scoring before pretesting. Prior to the winter and spring benchmark testing, test administrators received a 2-h refresher training on test administration and scoring. Trainings were conducted by either the first author or trained examiners. Students were assessed on either DIBELS or IDEL first with a minimum of 30 min between the English and Spanish assessments to avoid confusion between languages.

The SAT-10 and Aprenda-3 assessments were conducted by classroom teachers and reading coaches at each individual school. Coaches received a 2-h training on how to administer these tests. Then they trained classroom teachers and instructional assistants in test administration. Students were assessed on the Aprenda-3 and the SAT-10 (or the OAKS for third grade) on different days given the length of each of the assessments, and to avoid confusion between languages.

Data analysis

The analysis for the present study was conducted separately for each grade. We first used hierarchical multivariate linear models (HMLM; Raudenbush & Bryk, 2002) to estimate student fluency growth in both English and Spanish for each grade. In this multi-level approach, the two growth models, one for Spanish and one for English, are conceived as a multivariate repeated measures analysis allowing different assumptions about the variance and covariance structure of the Spanish and English growth processes. Moreover, hierarchical multivariate linear modeling makes it possible to consider both multiple fixed time points and multiple measures effectively within one analytic model.

In first grade, each student had a total of six measures of decoding fluency related to the alphabetic principle (three FPS scores in Spanish and three NWF scores in English). Measurement points occurred at the beginning of the year, in the middle of the year, and at the end of the year. In second and third grades, each student also had a total of six measures on oral reading fluency per grade (three FLO scores in Spanish and three ORF scores in English). These measures were also administered at the beginning, middle, and end of the year.

As presented in the “Appendix”, the specific model for this study was constructed to estimate student initial status and growth rate on reading fluency in English and Spanish (i.e., NWF vs. FPS for first grade, ORF vs. FLO for second and third grades). With English fluency scores as the reference in the model, the intercept (π0i) indicates the estimated initial status of English fluency. The effect of ‘Spanish’ (π1i) refers to the gap in the initial fluency status for Spanish over English. The other two effects relate to the fluency growth rates across a year. The effect of ‘Time’ (π2i) represents the estimated growth rate of English fluency, and the effect of the ‘Time by Spanish’ interaction term (π3i) refers to the gap in the growth rate for Spanish over English. Because fluency was assessed three times across a school year, the growth rates estimated from this model represent the average fluency growth among these three time points.

We did not include a level-2, between-person covariate, because the main objective of our study was to estimate the average initial status and growth rate of Spanish and English fluency within participants, not to compare the initial levels or growth rates between participants. However, the level-2 variance structure was specified here to have the level-1 intercept (π0i) and two effects (π1i and π2i) randomly varying between students, after the examination of competing variance structures. Thus, the final model for our analysis included heterogeneous level-1 variance (eti) across different observation occasions and three randomly varying level-2 components (u0i, u1i, and u2i), as shown in the “Appendix”.

We also used multiple regression models with initial scores and gain scores over a year in Spanish and English as independent variables to determine their relative prediction of Spanish and English reading comprehension. In addition, we built path analytic models to test how student fluency scores predict reading comprehension within the same language and across languages. The directionality of the effect of fluency in one language on reading comprehension on the second language was not specified, but correlations between fluency measures in Spanish and English were considered in the models (see Figs. 1, 2, 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs11145-010-9261-z/MediaObjects/11145_2010_9261_Fig1_HTML.gif
Fig. 1

Path model between reading fluency and reading comprehension in English and Spanish for grade 1. Note: Standardized coefficients are presented (↔: correlation coefficient, →: standardized regression coefficient). Unstandardized weights are in parenthesis. *** p < 0.001

https://static-content.springer.com/image/art%3A10.1007%2Fs11145-010-9261-z/MediaObjects/11145_2010_9261_Fig2_HTML.gif
Fig. 2

Path model between reading fluency and reading comprehension in English and Spanish for grade 2. Note: Standardized coefficients are presented (↔: correlation coefficient, →: standardized regression coefficient). Unstandardized weights are in parenthesis. ** p < 0.01; *** p < 0.001

https://static-content.springer.com/image/art%3A10.1007%2Fs11145-010-9261-z/MediaObjects/11145_2010_9261_Fig3_HTML.gif
Fig. 3

Path model between reading fluency and reading comprehension in English and Spanish for grade 3. Note: Standardized coefficients are presented (↔: correlation coefficient; →: standardized regression coefficient). Unstandardized weights are in parenthesis. * p < 0.05; ** p < 0.01; *** p < 0.001

Results

Results from our analysis are presented in three sections. First, we compared English and Spanish initial status and growth rates by testing hierarchical multivariate linear models. Second, we used sequential multiple regression to examine the intra- and inter-language relations between reading fluency and reading comprehension to determine how initial fluency and growth predicted reading comprehension across different languages. Third, we explored the relations between fluency and reading comprehension using path models that included Spanish and English measures simultaneously. All analyses were conducted separately for each grade, meaning that we statistically examined initial status and growth of fluency and their relations to reading comprehension within grade only, not across grades. Table 1 presents the means and standard deviations for all the measures used in this study across all grades.
Table 1

Descriptive statistics of English/Spanish test scores across grades

Grade

Statistics

English

Spanish

Fall fluency

Winter fluency

Spring fluency

Spring reading comp.

Fall fluency

Winter fluency

Spring fluency

Spring reading comp.

1st

 

NWF

SAT-10

FPS

Aprenda-3

N = 173

Mean

17.14

45.75

72.37

514.09

30.30

73.74

111.46

552.08

 

(SD)

(18.45)

(25.39)

(32.04)

(36.64)

(29.42)

(36.76)

(43.67)

(33.06)

2nd

 

ORF

SAT-10

FLO

Aprenda-3

N = 156

Mean

28.22

55.37

74.06

569.51

31.50

53.03

67.26

586.22

 

(SD)

(20.12)

(30.94)

(34.40)

(30.61)

(20.00)

(23.68)

(25.41)

(27.87)

3rd

 

ORF

OAKS

FLO

Aprenda-3

N = 142

Mean

53.94

77.08

99.67

203.33

52.44

67.66

75.43

591.09

 

(SD)

(29.16)

(35.58)

(34.12)

(10.51)

(22.95)

(23.46)

(23.78)

(32.05)

Initial status and growth rates of reading fluency in Spanish and English

We first estimated the average initial score and growth rate of Spanish and English reading fluency for each grade, using HMLM for growth modeling. English fluency scores were used as the reference point, and the gap for Spanish fluency was specified in the model to examine the differences in initial status and growth rates between English fluency and Spanish fluency. Results are presented for each grade in Table 2.
Table 2

Results from the hierarchical multivariate linear modeling for English and Spanish fluency growth

Fixed effect

Grade 1 (NWF/FPS)

Grade 2 (ORF/FLO)

Grade 3 (ORF/FLO)

Coefficient

se

t

Coefficient

se

t

Coefficient

se

t

Initial status

 Intercept, β00

17.09

1.40

12.25***

29.70

1.98

15.00***

53.57

2.64

20.27***

 Spanish, β10

13.41

1.36

9.83***

1.80

1.10

1.64

−0.29

1.56

−0.19

Growth rate

 Time, β20

27.91

1.03

27.00***

22.85

0.75

30.55***

23.10

0.66

34.81***

 Time × Spanish, β30

13.39

1.07

12.50***

−4.25

0.61

−7.01***

−11.26

0.66

−17.06***

Variancecovariance component

Level-2

\( \tau = \left[ {\begin{array}{*{20}c} {316.22} \hfill & {124.54} \hfill & { - 13.39} \hfill \\ {} \hfill & {231.38} \hfill & {18.34} \hfill \\ {} \hfill & {} \hfill & {120.80} \hfill \\ \end{array} } \right] \)

\( \tau = \left[ {\begin{array}{*{20}c} {554.08} \hfill & { - 145.45} \hfill & {48.50} \hfill \\ {} \hfill & {128.49} \hfill & { - 18.52} \hfill \\ {} \hfill & {} \hfill & {42.76} \hfill \\ \end{array} } \right] \)

\( \tau = \left[ {\begin{array}{*{20}c} {938.29} \hfill & { - 367.87} \hfill & { - 2.73} \hfill \\ {} \hfill & {258.66} \hfill & {1.57} \hfill \\ {} \hfill & {} \hfill & {26.37} \hfill \\ \end{array} } \right] \)

Level-1

σ12 (beginning, English) = 19.10

σ12 (beginning, English) = 60.20

σ12 (beginning, English) = 59.21

 

σ22 (beginning, Spanish) = 72.52

σ22 (beginning, Spanish) = 1.98

σ22 (beginning, Spanish) = 36.68

 

σ32 (middle, English) = 269.68

σ32 (middle, English) = 105.55

σ32 (middle, English) = 105.69

 

σ42 (middle, Spanish) = 323.62

σ42 (middle, Spanish) = 61.20

σ42 (middle, Spanish) = 56.00

 

σ52 (end, English) = 284.12

σ52 (end, English) = 117.93

σ52 (end, English) = 84.63

 

σ62 (end, Spanish) = 609.72

σ62 (end, Spanish) = 60.63

σ62 (end, Spanish) = 69.11

*** p < 0.001

In first grade (see the first column of Table 2), students started, on average, with significantly higher Spanish pseudoword reading fluency scores than English pseudoword reading fluency scores (β10 = 13.41, t = 9.83, p < 0.001). The average estimated English pseudoword reading fluency initial score was 17.09 (β00) while the average estimated Spanish pseudoword reading fluency initial score was 30.50 (β00 + β10) at the beginning of first grade. The pseudoword fluency growth rate was also significantly higher for Spanish than for English (β30 = 13.39, t = 12.50, p < 0.001). The average estimated fluency growth rate in English was 27.91 (β20) and in Spanish 41.30 (β20 + β30). This was across the three testing time points during first grade. In other words, on average, students gained about 83 letter sounds in Spanish as measured by the IDEL FPS from the beginning of the year to the end of the year, and 56 letter sounds in English as measured by the DIBELS NWF. Thus, ELs in first grade read more pseudowords in Spanish than in English at the beginning of first grade, and also developed their pseudoword reading fluency faster in Spanish than in English.

In second grade (see the second column of Table 2), the average estimated oral reading fluency initial score in English on the ORF was 29.70 (β00). The difference in initial oral reading fluency between Spanish (FLO) and English (ORF) at the beginning of second grade was not statistically significant (β10 = 1.80, t = 1.64, p > 0.10). Growth rates in oral reading fluency, however, were significantly different between languages (β30 = −4.25, t = −7.01, p < 0.001). Unlike the pattern in first grade, student oral reading fluency growth rates in second grade were higher, on average, in English reading (β20 = 22.85) than in Spanish reading (β20 + β30 = 18.60). Second-grade ELs in this study gained approximately 46 words on oral reading fluency in English from the beginning to the end of the year, and 36 words in Spanish. The difference in growth rates is worth noting, particularly given that students started the school year at similar oral reading fluency levels in each language.

In third grade (see the third column of Table 2), the estimated English oral reading fluency score at the beginning of the year was 53.57 (β00) and the difference in initial oral reading fluency scores between Spanish (FLO) and English (ORF) was not statistically significant (β10 = −0.29, t = −0.19, p > 0.10). However, the average fluency growth rate was significantly higher in English than Spanish (β30 = −11.26, t = −17.06, p < 0.001). The average estimated growth rate in English was 23.10 (β20), and the average growth rate in Spanish was 11.84 (β20 + β30). Similar to the second grade pattern, ELs in third grade performed similarly in Spanish and English at the beginning of the year, but during the year they gained 46 words in English compared to 23 words in Spanish, a larger difference in growth rates between English and Spanish than in second grade.

Level-1 random effects shown at the bottom of Table 2 provide another interesting insight about the patterns in student fluency development in English and Spanish for each grade. The variances in student scores on fluency measures are presented for each time point (beginning, middle, and end) for each language (English and Spanish). Two patterns are noticeable from these random effects. First, there are larger variances in Spanish fluency scores than in English fluency scores for first grade while the reverse is true for second and third grades. Second, for all three grades, the variances in fluency scores become larger at the middle or end of the year compared to those at the beginning of the year.

In summary, Spanish-speaking ELs in our study were more fluent in Spanish pseudoword reading than in English pseudoword reading in first grade, and they made more growth on pseudoword reading in Spanish than in English. ELs in second and third grades made more growth on oral reading fluency in English than in Spanish, even though these students began the year with similar oral reading fluency levels in the two languages. Particularly striking was the difference in oral reading fluency growth in third grade where students had nearly twice the gain in oral reading fluency in English compared to Spanish.

Prediction of reading comprehension by reading fluency in Spanish and English

Next, we examined how student initial reading fluency scores and year-long reading fluency gains in Spanish and English predicted their reading comprehension scores in each language at the end of the year. For this purpose, we used sequential regression models to predict reading comprehension scores in each language by first including the fluency variables (initial scores and gains) of the other language and then adding those of the same language. This approach was adapted from the mediation analysis method suggested by Kenny et al. (Baron & Kenny, 1986; Judd & Kenny, 1981). We assumed that reading fluency in the native language (F1) would predict reading comprehension in the second language (C2) via fluency in the second language (F2). If this assumption were true, the results would confirm the mediation pattern among the variables (e.g., reading fluency in Spanish → reading fluency in English → reading comprehension in English). This can be concluded if a significant effect of F1 on C2 decreases in magnitude and becomes nonsignificant when F2 is added to the regression equation, while F2 as a mediator remains a significant predictor of C2. The results from our analysis are presented in Table 3 for Spanish reading comprehension and in Table 4 for English reading comprehension. Each table includes the results from three grades separately, in different columns.
Table 3

Prediction of Spanish reading comprehension by Spanish and English reading fluency initial scores and gain scores

Model

Variable

Grade 1 (NWF/FPS)

Grade 2 (ORF/FLO)

Grade 3 (ORF/FLO)

b

se

t

R2

b

se

t

R2

b

se

t

R2

1

Constant

538.39

3.20

168.39***

0.20

571.34

3.60

158.90***

0.15

564.42

5.34

105.70***

0.21

 

English fluency initial

0.82

0.13

6.39***

 

0.52

0.10

5.07***

 

0.51

0.09

5.74***

 

2

Constant

525.37

5.73

91.68***

0.24

562.47

5.08

110.76***

0.18

547.16

9.02

60.67***

0.24

 

English fluency initial

0.84

0.13

6.68***

 

0.45

0.11

4.23***

 

0.51

0.09

5.84***

 
 

English fluency gain

0.23

0.08

2.71**

 

0.24

0.10

2.43***

 

0.37

0.16

2.35*

 

3

Constant

534.09

3.25

165.63***

0.28

562.55

3.75

150.10***

0.27

549.07

6.06

90.56***

0.31

 

Spanish fluency initial

0.60

0.08

7.82***

 

0.77

0.10

7.36***

 

0.81

0.11

7.58***

 

4

Constant

512.25

5.98

85.61***

0.35

550.32

5.82

94.61***

0.30

536.05

7.92

67.66***

0.35

 

Spanish fluency initial

0.64

0.07

8.67***

 

0.74

0.10

7.15***

 

0.88

0.11

8.10***

 
 

Spanish fluency gain

0.25

0.06

4.25***

 

0.36

0.13

2.71**

 

0.42

0.17

2.48*

 

5

Constant

511.54

6.30

81.16***

0.35

550.11

5.83

94.38***

0.31

534.64

9.01

59.32***

0.35

 

English fluency initial

0.10

0.20

0.49

 

−0.19

0.16

−1.17

 

−0.01

0.14

−0.09

 
 

English fluency gain

0.03

0.09

0.35

 

−0.07

0.13

−0.57

 

0.05

0.17

0.31

 
 

Spanish fluency initial

0.59

0.13

4.53***

 

0.93

0.19

5.02***

 

0.88

0.20

4.44***

 
 

Spanish fluency gain

0.24

0.07

3.34**

 

0.45

0.18

2.46*

 

0.39

0.20

1.98*

 

p < 0.05; ** p < 0.01; *** p < 0.001

Outcome variable = Spanish reading comprehension. English fluency variables are italicized

Table 4

Prediction of English reading comprehension by Spanish and English reading fluency initial scores and gain scores

Model

Variable

Grade 1 (NWF/FPS)

Grade 2 (ORF/FLO)

Grade 3 (ORF/FLO)

b

se

t

R2

b

se

t

R2

b

se

t

R2

1

Constant

492.70

3.43

143.76***

0.31

545.21

4.22

129.35***

0.25

564.42

5.34

105.70***

0.21

 

Spanish fluency initial

0.72

0.08

8.65***

 

0.78

0.12

6.73***

 

0.51

0.09

5.74***

 

2

Constant

474.90

6.56

72.43***

0.35

529.19

6.61

80.11***

0.30

547.16

9.02

60.67***

0.24

 

Spanish fluency initial

0.74

0.08

9.14***

 

0.76

0.11

6.67***

 

0.51

0.09

5.84***

 
 

Spanish fluency gain

0.21

0.07

3.15**

 

0.47

0.15

3.09**

 

0.37

0.16

2.35*

 

3

Constant

493.51

3.12

158.30***

0.36

546.15

3.67

149.03***

0.30

549.07

6.06

90.56***

0.31

 

English fluency initial

1.22

0.13

9.70***

 

0.82

0.11

7.76***

 

0.81

0.11

7.58***

 

4

Constant

473.91

5.41

87.53***

0.43

525.66

4.65

113.08***

0.46

536.05

7.92

67.66***

0.35

 

English fluency initial

1.24

0.12

10.38***

 

0.66

0.10

6.85***

 

0.88

0.11

8.10***

 
 

English fluency gain

0.35

0.08

4.33***

 

0.55

0.09

6.17***

 

0.42

0.17

2.48*

 

5

Constant

468.71

6.41

73.09***

0.44

528.49

5.84

90.45***

0.46

534.64

9.01

59.32***

0.35

 

Spanish fluency initial

0.22

0.13

1.67

 

−0.06

0.18

−0.35

 

−0.01

0.14

−0.09

 
 

Spanish fluency gain

0.09

0.07

1.26

 

−0.14

0.18

−0.78

 

0.05

0.17

0.31

 
 

English fluency initial

0.96

0.21

4.61***

 

0.70

0.16

4.32***

 

0.88

0.20

4.44***

 
 

English fluency gain

0.27

0.09

2.86**

 

0.61

0.12

5.01***

 

0.39

0.20

1.98*

 

p < 0.05; ** p < 0.01; *** p < 0.001

Outcome variable = English reading comprehension. Spanish fluency variables are italicized

Results of sequential regression in Table 3 indicate that English reading fluency variables (initial status and gain) were significant predictors of Spanish reading comprehension when Spanish fluency variables (initial status and gain) were not included in the models (see models 1 & 2). Spanish reading fluency variables (initial status and gain) were also significant predictors of Spanish reading comprehension when English fluency variables were not included in the model (see models 3 & 4). However, once we included all Spanish and English fluency variables in the regression model, only Spanish fluency variables were significant predictors of Spanish reading comprehension (see model 5). This pattern was consistent across all grades.

From Table 3, we can observe the significant contribution of fluency gains over initial reading fluency levels to later reading comprehension. While student initial reading fluency scores in Spanish significantly explained about 27–31% of the variance in Spanish reading comprehension scores (model 3), their year-long reading fluency gains in Spanish also explained an additional significant 3–7% of the variance in Spanish reading comprehension (model 4), depending on their grade (p < 0.01). Adding English fluency variables to the model did not significantly improve the variance explained as shown by comparing R squares in models 4 and 5 (for all grades, p > 0.05).

Similar results were found for English reading comprehension as presented in Table 4. Spanish reading fluency variables (initial status and gain) were significant predictors of English reading comprehension when English reading fluency variables were not included in the model (models 1 & 2). When English reading fluency variables (initial status and gain) were added (model 5), Spanish reading fluency variables became nonsignificant for all grades. English initial fluency significantly explained about 30–36% of the variance in English reading comprehension (model 3). English fluency gain significantly explained an additional significant 4–7% of the variance in English reading comprehension across grades (p < 0.05). Adding Spanish fluency variables did not significantly improve the prediction of English reading comprehension (for all grades, p > 0.05).

In sum, results of the sequential regression analysis indicate that growth on pseudoword reading and oral reading fluency significantly explained variance in reading comprehension controlling for initial status on fluency in Spanish and in English. However, when reading fluency variables in the same language as reading comprehension were added to the model, the role of the fluency variables in the other language faded away. Interpretations of these findings will be addressed in the “Discussion” section.

Relations between reading fluency and reading comprehension within and across languages

Results of path modeling confirmed that the direct relation between reading fluency and reading comprehension was significant only within the same language but not across different languages (see Figs. 1, 2, 3). In first grade (Fig. 1), for example, a 10-point higher initial score on English pseudoword reading fluency (NWF) significantly predicted a 9.4-point higher English reading comprehension score (p < 0.001), and a 10-point higher gain score on pseudoword reading fluency (NWF) predicted a 2.6-point higher English reading comprehension score (p < 0.001). However, neither initial fluency nor fluency gain on English pseudoword reading had a significant direct prediction of Spanish reading comprehension (p > 0.05). Similarly, a 10-point higher initial score on Spanish pseudoword reading (FPS) significantly predicted a 5.4-point higher Spanish reading comprehension score (p < 0.001) while a 10-point higher gain score on FPS significantly predicted a 2.3-point higher Spanish reading comprehension score (p < 0.001). Neither initial fluency nor fluency gain in Spanish pseudoword reading had a significant direct prediction of English reading comprehension (p > 0.05).

In second grade (Fig. 2), a 10-point higher initial score on English oral reading fluency (ORF) significantly predicted a 7.3-point higher English reading comprehension score (p < 0.001) and a 10-point higher gain score on ORF predicted a 6.1-point higher English reading comprehension score (p < 0.001). Neither initial performance on oral reading fluency nor oral reading fluency gain in English directly predicted Spanish reading comprehension (p > 0.05). Similarly, a 10-point higher initial score on Spanish oral reading fluency (FLO) significantly predicted a 9.0-point higher Spanish reading comprehension score (p < 0.001) while a 10-point higher gain score on FLO significantly predicted a 4.8-point higher Spanish reading comprehension score (p < 0.01). Just like before, neither initial performance on oral reading fluency nor oral reading fluency gain in Spanish had a significant direct prediction of English reading comprehension (p > 0.05).

In third grade (Fig. 3), a 10-point higher initial score on English oral reading fluency (ORF) predicted a 1.9-point higher English reading comprehension score (p < 0.001) and a 10-point higher fluency gain on ORF predicted a 1.3-point higher English reading comprehension score (p < 0.05). Similarly, a 10-point higher Spanish oral reading fluency initial score significantly predicted an 8.8-point higher Spanish reading comprehension score while a 10-point higher Spanish oral reading fluency gain score predicted a 4.0-point higher Spanish reading comprehension score (p < 0.001).

Although neither initial scores nor gain scores showed a significantly direct prediction of reading comprehension across different languages, path analysis showed high correlations between Spanish fluency variables and English fluency variables. For all grades, the correlations between Spanish and English fluency initial scores were very high (0.82 to 0.83) and the correlations between Spanish and English fluency gain scores were moderate (0.43 to 0.65). This may imply that initial scores and gain scores in Spanish reading fluency (F1) have an indirect effect on English reading comprehension (C2) through English reading fluency (F2), and initial status and gain scores in English reading fluency (F2) have an indirect effect on Spanish comprehension (C1) through Spanish reading fluency (F1). However, it is difficult to determine whether high correlations imply a mediation or a confounding effect. This should be examined further through experimental research (MacKinnon, 2008).

Discussion

This study had two purposes. The first was to estimate the growth rates in Spanish and English of ELs in grades 1, 2, and 3 on fluency in reading pseudowords (grade 1) and connected text (oral reading fluency, grades 2 and 3). The second purpose was to examine whether growth on pseudoword fluency and oral reading fluency in Spanish and English had a significant effect on reading comprehension within and across languages, beyond the variance accounted for by initial status. Our findings indicate that (a) growth on pseudoword reading in first grade was higher in Spanish than in English, but growth on oral reading fluency in second and third grades was higher in English than in Spanish; (b) Spanish and English reading fluency growth significantly explained variance in reading comprehension after controlling for initial status on Spanish and English reading fluency; and (c) initial status and growth on reading fluency had direct predictions of reading comprehension in the same language, but not across languages. In the next section, we interpret our results in the context of the reading instruction the students were receiving.

Reading fluency growth in Spanish and English

The findings indicate that ELs began first grade reading more pseudowords in Spanish than in English, and they also made more growth during the year in Spanish than in English. Two factors may account for this pattern. ELs in our sample were native Spanish speakers and may have been exposed to more Spanish than English, fostering higher initial performance and growth in Spanish compared to English. These students also received more Spanish reading instruction than English reading instruction in kindergarten and in first grade, a pattern of instruction that also supports the probability that ELs would have higher performance and growth in Spanish than in English.

In second and third grades, the pattern of oral reading fluency performance and growth in Spanish and English was different from the pattern on pseudoword reading in first grade. Students began the year reading approximately the same number of words in English and in Spanish, but their fluency growth in English was substantially higher than in Spanish. We offer four possible contributions to explain this pattern of results: (a) cross-linguistic transfer, (b) differences in the orthographic system, (c) more exposure to English throughout the school day, and (d) student motivation to learn English.

Cross-linguistic transfer

A plausible contribution for the different patterns in reading fluency growth in Spanish and English in grades 1, 2, and 3 may be the concept of cross-linguistic transfer, an idea closely associated with Cummins’ interdependence hypothesis (Cummins, 1981). Cross-linguistic transfer is the relationship across languages of skills that involve cognitive and language abilities (Genesee, Geva, Dressler, & Kamil, 2006). How cross-linguistic transfer occurs and what constitutes evidence of transfer appear to depend on (a) the particular reading skill being transferred, and (b) the language of instruction (Proctor, August, Carlo, & Snow, 2006).

Cross-linguistic transfer of specific reading skills

Several studies have shown there is cross-linguistic transfer of phonological awareness, decoding skills, and word recognition from Spanish to English (Cárdenas-Hagan, Carson, & Pollard-Durodola, 2007; Carlisle, Beeman, & Davis, & Spharim, 1999; Cirino et al., 2009; Cisero & Royer, 1995; Durgunoglu, Nagy, & Hancin, 1993). One study has examined the transfer of oral reading fluency across Spanish and English (Dominguez De Ramírez & Shapiro, 2007). Dominguez de Ramirez and Shapiro found that there were generally moderate to strong relations between Spanish and English oral reading fluency across grades 1 through 5 from the beginning of the year until the end of the year, providing indirect evidence of transfer. Our study corroborates this finding specifically, as well as the general finding that transfer of reading skills involving Spanish and English does occur.

Language of instruction

Language of instruction might also play an important role in transfer. For example, Manis, Lindsey, and Bailey (2004) found that Spanish print knowledge, phonological awareness, and rapid naming in kindergarten predicted English reading fluency and comprehension in grades 1 and 2 for students attending a transition program (where they learned to read in Spanish first, then in English). However, once English variables were added into the prediction model, English print knowledge, phonological awareness, and rapid naming mediated the contribution of the corresponding Spanish variables. In fact, the English variables accounted for more unique variance than the Spanish variables. In our study, where students received both Spanish reading instruction and English reading instruction simultaneously, at first glance, initial status and growth on pseudoword reading and oral reading fluency in one language seemed to have a significant effect on reading comprehension in the other language. This significant effect, however, faded away when fluency variables in the same language as the reading comprehension measure were entered into the regression models. It is plausible that similar to the finding by Manis et al. (2004), the effect of cross-linguistic transfer of reading skills might be maximized when similar skills are taught in both languages. In the absence of this instruction, skills in one language might show less transfer or no transfer to another language.

Differences in the orthographic system

Another plausible explanation why students made more growth on reading fluency in English than in Spanish in second and third grades can be found in the fact that Spanish, in general, has a greater percentage of multisyllabic words than English, and thus students may have read fewer words in Spanish than in English for a 1-min timing even though their actual fluency rates may be comparable. For example, the translation of the English word “fun” (one-syllable) is “divertido” in Spanish (four syllables). Because of word length, it simply takes more time to read “divertido” than to read “fun”. In a random selection of three passages in English and Spanish used in this study, we found that 100 words in English represented an average of 121 syllables while 100 words in Spanish represented an average of 194 syllables. In other words, if students were to read 100 words in Spanish they would actually be reading 73 syllables more in Spanish than in English, a substantial number that might have had an effect on the larger fluency growth (i.e., based on the number of words students read) in second and third grades in English fluency compared to Spanish fluency.

Exposure to more English than Spanish

All four schools in our study provided content instruction in English. The only class where students were taught in Spanish was reading. Thus, it is plausible that greater exposure to English during the day enhances English reading growth, perhaps by helping students become familiar with the vocabulary and academic language (e.g., articles, prepositions, adverbs) that they are likely to encounter in reading connected text. According to Perfetti’s (1999) blueprint of beginning reading, for students to become fluent readers they need to form a mental representation of the words they encounter. The more opportunities students have to develop mental representations of words, the more automatically they will read them. In our study, in third grade for example, ELs were receiving the same amount of English and Spanish reading instruction, which means reading exposure was exactly the same. However, ELs were exposed to more English than Spanish during the school day, which is a plausible explanation for the greater reading growth in English than Spanish in third grade.

Motivation to learn English

The evidence is strong that students who are motivated to read develop their reading skills faster than students who are not (Guthrie, Wigfield, Metsala, & Cox, 1999). Perhaps to fit in with their peers and to assimilate into the school environment, ELs in this study valued learning English more than Spanish as they got older (e.g., in grades 2 and 3 compared to grade 1). Moreover, as students progress through the upper grades, they realize that English is more valued at school than Spanish for many reasons. For example, strong English skills are necessary to be able to understand what the teacher is saying in the classroom and to receive better grades. These academic and social values play a crucial role in student motivation to learn English (Eccles & Wigfield, 2002; Urdan & Schoenfelder, 2006), and might explain, in part, the higher gains on oral reading fluency in English than in Spanish in second and third grades.

Relations of reading fluency to reading comprehension within and across languages

Our findings indicated that initial status and growth on oral reading fluency were significant contributors to later reading comprehension in English and Spanish. Although other studies have demonstrated that pseudoword reading and oral reading fluency in English significantly contribute to English reading comprehension (Baker et al., 2008; Fien et al., 2008), our study extended this finding to Spanish by showing that initial status and growth on oral reading fluency in Spanish also contributes significantly to later reading comprehension in Spanish, a language with a transparent orthography. This finding is important because it suggests that the reading trajectory in a transparent orthography like Spanish, is very similar to the reading trajectory in a more complex orthography like English (Ehri, 1995; Jiménez & O’Shanahan Juan, 2008; Seymour, 2005).

Another noticeable point is that the percentage of variance explained in English reading comprehension by English oral reading fluency was lower in our study than in the Baker et al.’s (2008) study. In our study, oral reading fluency initial status and gains in English explained about 46% of the variance in English reading comprehension for second grade (95% Confidence Interval: 33–57%), and 35% of the variance in English comprehension for third grade (95% CI: 22–47%). In the Baker et al.’s (2008) study with a larger sample but the same measures in English, oral reading fluency initial status and gains explained about 70% of the variance in reading comprehension for second grade (95% CI: 69–71%), and 52% for third grade (95% CI: 50–54%). This difference in the amount of variance explained might be an indicator that, for ELs, oral reading fluency is a significant contributor to reading comprehension in English, but other skills (e.g., vocabulary, academic language) might also play a prominent role when learning to read in the second language.

Although we found that pseudoword reading and oral reading fluency had a direct prediction of reading comprehension within the same language but not across different languages, the fact that the correlations of reading fluency variables between Spanish and English were high for initial levels (above 0.80) and moderate for year-long gains (above 0.40), indicates that fluency in the same language as reading comprehension might play a mediational role between reading fluency in one language and reading comprehension in another language. To confirm the mediation hypothesis, we would need additional evidence about the causal relation between fluency in one language and fluency in another language, which our data could not provide. Another possibility is that there might be simply a confounding or spurious relation in fluency variables between different languages due to other linguistic factors (e.g., Spanish and English use the same writing system, the alphabet). From a statistical perspective, mediation and confound are similar constructs and the best way to distinguish them is to have a strong conceptual and empirical evidence that supports the mediation hypothesis (Cohen, Cohen, West, & Aiken, 2003; MacKinnon, Krull, & Lockwood, 2000). The conclusion we can make confidently from the present study is that there is no direct prediction of reading comprehension in English by reading fluency in Spanish and vice versa.

Our findings also support previous research on the predictive validity of reading fluency within and across languages. For example, Dominguez De Ramírez and Shapiro (2006) found that oral reading fluency slope was a moderate to strong predictor of oral reading fluency at the end of the year within and across languages. Our study extended these findings by showing that both initial levels and growth on reading fluency during a school year were also significant predictors of end-of-year reading comprehension scores in English and Spanish.

These findings have important implications for assessment and instruction of ELs who are receiving reading instruction in two languages. The lack of a direct relation between reading fluency and comprehension across languages suggests that to determine if ELs in the early grades (e.g., grades 1–3) are learning to read adequately, it would be advisable to assess them on reading fluency in English periodically during the year, even if they are receiving only Spanish reading instruction and improving on Spanish reading fluency. If students are not increasing their reading fluency in English perhaps as an effect of cross-linguistic transfer, this would be an indication that they would need to be provided with English reading instruction because their growth on fluency in Spanish would not indicate growth on fluency in English. Moreover, once English reading instruction is introduced (e.g., in an early or late transition bilingual model) schools should systematically monitor the progress of students on English reading fluency to ensure that their English reading skills are increasing in a manner that is commensurate to the development of their Spanish reading skills.

Limitations

Participants in this study were receiving Spanish and English reading instruction in a paired bilingual program. We don’t know if the same results would apply to students receiving Spanish reading instruction only in an early or late transition program or English only instruction in an English immersion program. However, our results do corroborate results in other studies with different bilingual programs where reading skills in English appeared to mediate the effects of Spanish reading skills on English reading comprehension (Manis et al., 2004; Proctor et al., 2006).

In addition, we could not examine the causal effects of Spanish reading skills on English reading skills and vice versa because our study was only correlational. Further research is needed using an experimental design to determine causality. Finally, our models did not include additional variables (e.g., vocabulary, oral language proficiency) other than reading fluency that are known to be crucial components of reading comprehension. For instance, Miller et al. (2006) and Crosson & Lesaux (2009) indicate that language proficiency may have an important measurable effect on reading comprehension for bilingual students.

Conclusions

This study adds to the literature on the contribution of pseudoword reading growth and oral reading fluency growth to reading comprehension in English and Spanish at the end of the year. Our findings also indicate that growth on pseudoword reading and oral reading fluency has a significant direct effect on reading comprehension only within the same language in which fluency and reading comprehension are measured. These findings are important because they suggest that ELs living in the United States should be progress monitored in English, even if they are only receiving Spanish reading instruction, to make sure that they are making adequate progress in their second language. Future research looking at the causal relations between oral reading fluency and reading comprehension in Spanish and English is necessary to increase our knowledge base on the effects of cross-linguistic transfer of reading skills from one language to another. It is particularly important to examine transfer with students of different skill levels and to examine transfer in the context of different types of instructional environments, represented, for example, by different types of bilingual education programs.

Acknowledgments

This work was supported by an Oregon Reading First subcontract from the Oregon Department of Education to the University of Oregon (8948). The original Reading First award was granted by the U.S. Department of Education to the Oregon Department of Education (S357A0020038).

Copyright information

© Springer Science+Business Media B.V. 2010