Phonological processing and emergent literacy in younger and older preschool children
Anthony, Jason L
Phonological awareness, phonological memory, and phonological access to lexical storage play important roles in acquiring literacy. We examined the convergent, discriminant, and predictive validity of these phonological processing abilities (PPA) in 389 3-, 4-, and 5-year-old children. Confirmatory factor analysis supported the validity of each PPA as separate from general cognitive ability and separate from each other. Multigroup structural equation modeling (SEM) with mean structure demonstrated that older preschoolers have better developed latent PPA than younger preschoolers but that the structure of PPA is equivalent. RAN was found uniquely associated with letter knowledge and text discrimination in younger preschoolers, and PA was found uniquely associated with word reading skills in older preschoolers. Finally, general cognitive ability was only indirectly associated with emergent literacy via PPA. These results highlight the importance of PPA in the early literacy development of English-speaking preschool children.
Keywords Phonological awareness * Phonological memory * Rapid naming * Emergent literacy * Phonological processing * Children * Reading * Phonological sensitivity * Short-term memory * Preschool
Consensus has emerged from three decades of literacy research that difficulty with the mental processing of phonological information is a core deficit that accounts for many children’s difficulties in learning to read (Adams, 1990; Stanovich, 1988). Phonological processing refers to the use of the sound structure of oral language in processing written and oral information. More specifically, research with school-age children has identified three interrelated phonological processing abilities (PPA) that are important for reading and writing: phonological awareness (PA), phonological memory (PM), and efficiency of phonological access to lexical storage (a.k.a. RAN; for review see Wagner & Torgesen, 1987). How the various PPA are related to each other and what roles they play in literacy development are issues of considerable theoretical and practical importance. The answers to these questions are likely to have significant implications for the development of assessment batteries, for the improvement of our ability to identify children at risk for reading difficulties, and for pinpointing potential loci of intervention.
PA refers to one’s ability to detect or manipulate the sounds in his or her oral language (for review, see Anthony & Francis, 2005). PA encompasses phoneme awareness, the ability to manipulate individual sounds (phonemes) in words, and rudimentary phonological awareness skills, such as judging whether two words rhyme. PM refers to the coding of information in a sound-based representation system for temporary storage. PM is utilized during all cognitive tasks that involve processing sound information. Individuals’ PM capacity is often operationalized by auditory span tasks, like digit span. RAN refers to the efficiency of retrieving phonological codes from memory. Individual differences in efficiency of retrieving phonologically stored information from memory are typically operationalized by performance on rapid autonomic naming tasks in which individuals verbally identify common objects, letters, or numbers as quickly as possible.
Numerous correlational and longitudinal studies demonstrate that PA, PM, and RAN are reliable individual predictors of reading achievement. Moreover, increasingly more comprehensive and sophisticated multivariate research has been conducted in an effort to determine which PPA are the best predictors of reading achievement and which PPA are uniquely important for literacy development. Depending on what demographic, phonological, and reading-related constructs are assessed and included in the statistical models, some studies have shown that PA is uniquely predictive of reading (Bryant et al., 1989, 1990; Wagner et al., 1994), some studies have shown that PM is uniquely predictive of reading (Rapala & Brady, 1990; Rohl & Pratt, 1995; Snowling et al., 1986; Stone & Brady, 1995), and some studies have shown that RAN is uniquely predictive of reading (Felton & Brown, 1990; Griffiths, 1991; Wagner et al., 1997; Wolf & Obregon, 1992). Of course a superior approach is to examine the relative predictive contributions of PA, PM, and RAN in the context of all other PPA and all known predictors of literacy (e.g., language ability and orthographic knowledge). Employing a latent variable approach further strengthens such investigations by providing more reliable estimates of the true associations among the psychological constructs than those obtained by observed scores. To our knowledge, only three longitudinal investigations of the relations among all three latent PPA and latent reading abilities have been conducted (see below).
First, Wagner and colleagues have conducted the most comprehensive line of research in the area of PPA to date (Wagner & Torgesen, 1987; Wagner et al., 1993, 1994, 1997). Their research boasts impressive sample sizes, both cross-sectional and longitudinal designs, inclusion of general ability covariates and autoregressors, and use of structural equation modeling (SEM) to examine relations among latent PPA and latent literacy skills. In a number of studies, Wagner and colleagues compared alternative models of the factor structure of PPA in over 200 typically developing, English-speaking, school-age children. Initial findings and subsequent replications supported a measurement model of PPA in which PA, PM, and RAN were distinguishable but moderately correlated phonological abilities. Additionally, PA and RAN demonstrated unique predictive relations with word reading.
Second, de Jong and van der Leij (1999) conducted an equally impressive study that was very similar to that of Wagner et al. (1994) except that de Jong and van der Leij examined the structure of PPA and the relations of PPA with reading and math achievement in 166 Dutch children followed from kindergarten through second grade. Like Wagner and colleagues, de Jong and van der Leij found that PA, PM, and RAN were separate but moderately intercorrelated PPA that demonstrated unique patterns of relations with achievement outcomes. Specifically, PA and RAN as measured in first grade were uniquely predictive of second grade literacy, and PM was uniquely predictive of second grade arithmetic.
Third, Sprugevica and Hoien (2004) conducted a study similar to those of Wagner et al. (1994) and de Jong and van der Leij (1999) except that Sprugevica and Hoien examined the structure of PPA and the relations of PPA with reading achievement in 55 Latvian children followed from first through second grade. Similar to studies with American and Dutch children, Sprugevica and Hoien found that (a) PA, PM, and RAN were clearly distinguishable phonological abilities, (b) PA and RAN were uniquely predictive of word reading, and (c) RAN was uniquely predictive of reading comprehension after controlling for the other PPA and decoding.
In summary, research with school-age children learning to read different orthographies has consistently shown that PA, PM, and RAN are distinguishable from each other and from general cognitive ability. This research has also demonstrated that, from among the PPA, PA tends to be the strongest predictor of individual differences in word reading (e.g., Wagner et al., 1997). Furthermore, the influence of PPA on literacy development appears to be greatest during the first few years of formal schooling when children are learning to read and write. Research consistently shows that PA and RAN have unique relations with literacy. However, most studies also show that there is a large amount of shared variance among the PPA, independent from verbal ability, that is related to literacy development (e.g., Wagner etal., 1994). In fact, some researchers have proposed that there may be a single core phonological factor that underlies the development of all PPA (Kamhi & Catts, 1989; Shankweiler et al., 1995; Stone & Brady, 1995). Indeed, some researchers have proposed a core phonological processing deficit that underlies the majority of reading difficulties (Hoskyn & Swanson, 2000; Stanovich & Siegel, 1994; Stone & Brady, 1995).
Although a number of generalizations can be made concerning the nature of school-age children’s PPA and the relations PPA have with reading development, much less is known about preschool-age children’s PPA. This is an important gap in the literature to fill because a number of longitudinal studies have demonstrated that preliterate children’s PPA are reliable individual predictors of beginning reading achievement. For example, the National Early Literacy Panel (NELP, 2007) summarized peer-reviewed published articles of longitudinal studies that examined the power of preschool and kindergarten emergent literacy skills (e.g., PPA, letter knowledge, oral language) to predict school-age literacy. The NELP’s meta-analysis revealed that all of the emergent literacy skills were reliable predictors of school-age decoding and reading comprehension abilities. Such findings demonstrate that the foundations for literacy are established while children are preliterate and they underscore the current need to better understand how the various emergent literacy skills are interrelated and work together to support literacy acquisition.
Only two comprehensive latent variable examinations of preschool children’s PPA have been published to date. Wagner et al. (1987) administered tests of PA, PM, and RAN to 111 native English-speaking 4- and 5-year-old American children. Wagner et al. found that a two factor model of PPA best characterized these preliterate children’s performances. Specifically, this two factor model of PPA consisted of a combined PA/PM factor and a separate RAN factor, implying that latent PA and latent PM were indistinguishable from each other at this age. An unpublished study by Lonigan et al. (2007; submitted for publication) found similar results with 433 native English-speaking preschoolers. In contrast, Anthony et al. (2006) administered tests of PA, PM, and RAN to 147 native Spanish-speaking 3-, 4-, and 5-year-old American children. They found that a three-factor model best characterized these preschooler’s performances. In other words, Spanish PA, Spanish PM, and Spanish RAN were clearly distinguishable from each other and from general intelligence. Finally, a noteworthy contribution to the literature of the Anthony et al. study was that because it included measures of other emergent literacy skills, this study was able to demonstrate that each of the PPA was uniquely related to early literacy, even after controlling for general intelligence.
Conflicting findings concerning the nature of PPA in preschool children are potentially attributable to three factors: (a) how phonological awareness has been assessed, (b) what populations have been studied, and (c) which covariates have been included in statistical models. The first possible explanation for different findings is that the PA tasks used by Wagner and colleagues were more difficult and had greater memory demands than those used by Lonigan et al. (2007; submitted for publication) and Anthony et al. (2006). For example, all of the PA tests used by Wagner et al. required the unassisted holding of phonological information in auditory short-term memory while a given task was performed on that information, and some of the tasks used are now known to be quite difficult for preschoolers (e.g., elision of syllables into nonwords and counting syllables). In contrast, more than half of the PA tests used by Lonigan et al. and Anthony et al. used illustrations to assist children’s short-term memory, and many of the tests required only a pointing response to multiple choice items, which can arguably be achieved using a rudimentary phonological comparison strategy. second, Wagner et al. and Lonigan et al. examined English PPA in native English-speaking preschoolers, whereas Anthony et al. examined Spanish PPA in native Spanish-speaking preschoolers. Finally, Anthony et al. examined relations among PPA after statistically controlling for both chronological age and general cognitive ability, whereas Wagner et al. only controlled for general ability and Lonigan et al. only controlled for age.
The purpose of the present study was to extend the work of Wagner and colleagues with native English speaking children in two important ways. One, we examined the nature of PPA in younger preschool-age children, and two, we included a broad range of emergent literacy skills so that associations among younger and older preschoolers’ English PPA and their English emergent literacy skills could be identified. Less central to our aims but still important, we hoped to help elucidate the source of conflicting findings in the preschool literature by administering measures of PA that were minimally confounded with short-term memory, by comparing results from a variety of covariate models, and by comparing the structure of PPA in younger and older preschool children.
As with our prior research, we faced three challenges to meeting our objectives, and these challenges stemmed from the age of the population of interest and from the nature of our research questions about the interrelations among various cognitive abilities. First, although the preschool period lasts only a few years, these years represent a large portion of these children’s lives. Moreover, the developmental abilities of preschool age children vary greatly as a function of children’s chronological age. Thus, any investigation of relations among preschoolers’ developmentally associated skills must account for children’s ages so that the relations among such skills are not inflated by the age range of the sample. Because we were first interested in identifying the best model of PPA across the full range of preschool-age children, we partialled out the effects of chronological age from children’s scores and then modeled the residual covariances. Later, when we compared the associations among a group of younger preschoolers’ PPA to those of a group of older preschoolers, we partialled out the effects of age within each group and modeled the residual covariances.
Second, it cannot be assumed that a completely valid and reliable estimate of a preschooler’s ability is obtained by a single measure administered at a single point in time, nor can it be assumed that all measures in a given assessment battery have equivalent reliabilities and distributional properties. The problem with differential reliabilities and differential deviations from normality among measures is that they influence the covariances among scores, and such influences can lead researchers to faulty conclusions about the relations among phonological processing skills and the associations these skills have with literacy outcomes. Our approach to this commonplace problem was to administer multiple measures of each PPA and to estimate the true associations among latent PPA.
Third, children’s performances on cognitive tasks, including measures of PPA, tend to be intercorrelated at least partially as a function of children’s general cognitive ability. For this reason, the most carefully designed investigations of PPA have employed various means of controlling for individual differences in general cognitive abilities so that (a) associations among PPA were not inflated by a common cause and (b) so that PPA could be demonstrated as distinguishable from general cognitive ability. Both issues are particularly relevant to the investigation of preschool children’s PPA because it is possible that at this young age children’s general abilities may fully explain individual differences in one or more PPA. In the present study, we simultaneously modeled children’s latent general ability and children’s latent PPA so that we could directly examine whether preschoolers’ PPA were distinguishable from their general cognitive ability.
With the general aim of examining age-varying relations among PPA and other emergent literacy skills in native English-speaking preschool children, we posed four research hypotheses:
1. Each phonological processing ability is distinguishable from general cognitive ability.
2. Phonological processing abilities are distinguishable from each other.
3. Relations among phonological processing abilities are different in younger and older preschool children.
4. Phonological processing abilities demonstrate different relations with emergent literacy skills at different points in children’s literacy development.
To test our first hypothesis, we used confirmatory factor analysis (CFA) to evaluate four a priori models that collectively tested whether each PPA was distinguishable from general cognitive ability. To test our second hypothesis, we used SEM to evaluate three a priori models that reflected three current conceptualizations of the nature of PPA. These included a specific abilities model comprised of three separate but correlated PPA (i.e., RAN, PA, and PM), a model resembling Wagner et al. (1987) conceptualization of preliterate children’s PPA comprised of two correlated PPA (i.e., RAN and combined PA/PM), and a phonological core model comprised of one general PPA (i.e., combined RAN/PA/PM). To test our third hypothesis, we performed multigroup SEM with mean structure to compare the mean and covariance structures of PPA in younger and older preschool children. Finally, to test our fourth hypothesis and assess the practical merit of distinguishing among PPA in this population, we used multigroup SEM to determine if RAN, PA, and PM have different relations with emergent literacy skills at different points in children’s preliteracy development.
Materials and methods
Data for the current study were collected from children who were part of a program evaluation project directed by the first author. The program evaluation project was a simple longitudinal correlational study. There was no control group and no special interventions beyond what is considered standard early childhood education programming by Head Start. The aim of the program evaluation was to describe the amount of children’s learning of key emergent literacy skills (e.g., letter names and phonological awareness) that occurred over the course of the school year. The present study includes both end-of-year outcome data from the program evaluation (e.g., letter names and phonological awareness) and supplemental data (e.g., phonological memory and RAN) that were collected concurrently from the same sample.
The organization that contracted the program evaluation coordinates Head Start services among 24 sites in or around a large city in Texas, USA. All 24 sites provide classroombased early childhood education services. About 1/4 of the sites are located on public elementary school campuses and include coteaching by a Head Start teacher and a public school teacher. Because Head Start programs serve low-income families, virtually all participants were from economically disadvantaged backgrounds.
Two sampling strategies were employed, one for each year of the project. In year 1, we randomly selected 236 monolingual, native English-speaking children from the pool of children in classrooms where instruction was provided in English. In year 2, we first randomly selected 44 of the 74 classrooms in which instruction was provided in English. Next, we randomly selected approximately 4 monolingual, English-speaking children from each of the selected classrooms, which totaled 153 children after minor attrition.
Because of the nature of the sampling, all participants were native speakers of English. However, the sample was ethnically diverse, with 51% Hispanic American, 47% African American, 1 % European American, 1 % Asian American, and 1 % other. Fifty-three percent of the sample was male. Participants ranged in age from 43 to 67 months (M=55, SD=7). Children demonstrated average motor abilities (M=101, SD=15) and low average to average language abilities (M=92, SD=13) on the Developmental Indicators of the Assessment of Learning-Third Edition (DIAL-3; Mardell-Czudnowski & Goldenberg, 1998) at the beginning of the school year.
A nonstandardized, prepublication version of the Preschool Comprehensive Test of Phonological and Print Processing (PCTOPPP; Lonigan et al., 2002) was used to assess children’s phonological awareness, phonological access, phonological memory, letterknowledge, and print awareness. The subtests of the PCTOPPP are also described in detail below under their associated conceptual headings. All measures employed age appropriate tasks, colorful and interesting pictures whenever possible, brief instructions, and corrective feedback on practice items. The first author made a number of modifications to the PCTOPPP. Specifically, the directions were simplified to make them more understandable for young children. In addition, discontinuation rules were added as noted below to reduce the likelihood of children becoming frustrated. Finally, high-quality color pictures that more clearly illustrated the stimuli and response words were substituted for all items.
Three measures were used to assess phonological awareness: initial sound matching, elision, and blending. All three measures required the children to identify one of three or four pictures in response to a phonological stimulus, and each task was preceded by two practice items that were followed by correction, explanation, and readministration if the child gave an incorrect answer, or confirmation and explanation if the child gave the correct answer.
For Initial Sound Matching, a stimulus picture was shown and identified (e.g., “This is a sock.”), after which three target pictures were displayed and labeled (e.g., rake, car, sun). The children were asked to identify which of the three pictures began with the same sound as the stimulus. The task consisted of 14 test items, but testing was discontinued if a child made three incorrect responses in four consecutive items (i.e., a ceiling criterion of three out of four).
The Elision task assessed, in two parts, the children’s ability to identify or produce a target word resulting from the deletion of a part of a stimulus word. For both parts, testing was preceded by two practice items, and confirmation and/or feedback was provided. The first half of the task consisted of ten multiple-choice items. For each multiple-choice item, an examiner named and displayed four target pictures (e.g., pen, ball, car, cup). Next, the examiner stated a stimulus word and asked the child to repeat the stimulus word (e.g., pencil). Finally, the child was told to point to the picture that illustrated the stimulus word without a particular sound (e.g., “Point to pencil without /cil/”). Testing was discontinued if the child answered incorrectly on three out of four consecutive items. The second half of the elision test consisted of nine items in a free-response format. The stimulus word was presented orally as before, but the children were asked io delete a unit of sound from the stimulus word to produce the target word without the help of pictures. Responses were coded as either correct or incorrect, and a ceiling criterion of four consecutive incorrect responses was used. Items in both parts of the elision test were grouped into three levels of linguistic complexity. Initial items required the children to delete a one- or two-syllable word from a compound word (e.g., seesaw without see). Middle items required deletion of a syllable from a word (e.g., candy without /di/), and final items required removal of a phoneme from a word (e.g., lamp without /p/).
The Blending test assessed the children’s ability to identify or produce a target word that resulted from the combination of words or parts of words. The task followed the same format and ceiling criteria as the elision task. Both parts were preceded by two practice items, and confirmation and/or feedback was provided. The first half followed a multiplechoice format, the second half required an unaided response, and both parts were divided into three levels of linguistic complexity: word (e.g., hot-dog), syllable (e.g., cir-cle), and sub-syllable (e.g., f-i-sh).
Phonological memory was assessed with two measures from the PCTOPPP (i.e., memory for words and memory for nonwords) and one measure from the Woodcock JohnsonRevised (i.e., memory for sentences). All three phonological memory tasks were preceded by at least one practice item, and for the first two measures, feedback was provided on practice items. Examiners were careful not to readminister any items from these memory tests unless a major distraction had occurred, in which case at least three other new items were administered before readministration.
The Memory for words task measured children’s ability to reproduce a list of one-syllable words. The measure included 21 items, divided into seven groups of three items, according to the number of words to be recalled on a given item. The first group of three items required repetition of one word, the second group of items required repetition of two words, and so on. Items were scored as either correct or incorrect. Failure to correctly recall all three items within one group constituted the ceiling criterion.
To assess children’s memory for phonemes, the Memory for Nonwords task required children to recall up to 20 “pretend words.” Stimuli were 1 to 4 syllables in length and 3 to 12 phonemes in length. Responses were scored as either correct or incorrect, and administration of the test was stopped after four consecutive incorrect responses.
The Memory for sentences task required children to repeat orally presented sentences. This test consisted of 32 items, beginning with one- and two-syllable words, followed by phrases and sentences of increasing length and complexity. Responses were coded as correct, one error, or more than one error, and testing was discontinued after four consecutive responses in which more than one error was made.
Two measures assessed efficiency of phonological access to lexical storage. The Rapid Object Naming stimulus card consisted of seven rows of four pictures. The pictures illustrated single syllable words common in young children’s vocabularies. Each row presented the same four pictures in random order, and the first row served as practice for the remainder of the task. During practice, the examiner pointed to each object and named it until the child could correctly identify all four pictures without assistance. Once this was accomplished, the child was told that he or she should say the names of each picture on the page as fast as possible without making a mistake or skipping any pictures. During the test, correct identification of each picture was required before continuing to the next picture. In the case of an incorrect identification, the examiner would first request another attempt by pointing to or tapping on the misidentified picture. If another incorrect response was given, the examiner would prompt the child with “This is a ________. Say ________. Now go ahead.” Two trials were conducted, and the minutes and seconds required for the child to complete each trial were recorded. The score assigned on the Rapid Object Naming test was the average time required to complete both trials.
The Rapid Size Naming test was parallel in structure and had the same administration procedures as the Rapid Object Naming test. The stimulus card included pictures of a big circle, little circle, big square, and little square. Children were required to identify whether the object was “big” or “little”. Two trials were administered, and the average time required to complete both trials was recorded as the score on the test.
Children’s letter knowledge was assessed by the Print Awareness subtest of the PCTOPPP. This subtest included ten free response letter name identification items and four free response letter sound identification items. A child was shown an uppercase letter or lowercase letter and asked to produce either the name or sound of the letter; only the name or sound was accepted as correct depending on which was asked of the child. All 14 items were administered to each child. The children’s scores on this Letter-Knowledge measure reflect the total number of correct letter names and letter sounds produced.
The Text Discrimination test consisted of eight multiple-choice items from the Print Awareness subtest of the PCTOPPP. The Text Discrimination items assessed children’s ability to distinguish printed letters and printed words from nonalphabetic characters and illustrations. More specifically, for each item, examiners presented children with an array of four choices and then asked, “Which is a letter/word?” or “Point to a letter/word.” Correct response choices were single letters, strings of letters, or individual words (e.g., C, SMK, CAR). Incorrect response choices were strings of numbers, symbols, or illustrations (e.g., 589, Ω, picture of a car). All eight items were administered to each child.
Children were administered a 25-item test of word reading. Stimuli included 2-, 3-, and 4-letter words, all of which were one syllable in length. Items were arranged in order of difficulty based on Item Response Theory scaling, and a discontinuation criterion of five consecutive errors was imposed.
General cognitive ability
The Developmental Indicators of the Assessment of Learning-Third Edition (DIAL-3; Mardell-Czudnowski & Goldenberg, 1998) is a standardized developmental screener used to identify young children who are at risk for school failure. It includes a variety of ageappropriate manipulatives and tasks. The 21 subtests require naming picture vocabulary items; solving verbal problems; providing personal information; identifying shapes, colors, and body parts; understanding relative positions and measurement concepts; counting; building with blocks; copying line drawings; cutting; etc. Exploratory and confirmatory factor analysis of the subtest scores of over 2,000 children indicate that there are two dominant factors and one less influential factor that drive children’s performances on the DIAL-3 (Anthony et al., 2007). Specifically, the Verbal Ability factor and the Nonverbal Ability factor reflect children’s fluid abilities concerning verbal comprehension and integration of visual, spatial, and motor information, respectively. The Achievement factor reflects children’s crystallized knowledge for curriculum material, e.g., letter names, letter sounds, shapes, colors, and counting.
Design and procedures
This study was carried out over the course of two school years. The children were tested with the DIAL-3 at the beginning of a school year, and then the same children were tested with the phonological processing and emergent literacy battery at the end of the same school year. Preliminary analysis conducted after the first year of the project revealed unusually low scores on the Initial Sound Matching test. In fact, the sample’s average score (M=4.5) was equal to chance (i.e., 4.6). Therefore, we decided to not administer Initial Sound Matching in year 2. Additionally, budgetary constraints forced us to drop two supplemental measures in year 2. To retain the ability to address the present study’s research questions with both cohorts, we dropped one RAN measure and one memory measure, i.e., RAN Size and Memory for Words. Thus, the missing data by design pattern are such that children from the first year have complete data and children from the second cohort are missing Initial Sound Matching, RAN Size, and Memory for Words.
Children were tested individually by carefully trained examiners. Examiners were trained in standardized DIAL-3 testing procedures by the publisher of the DIAL-3. Testing took place at children’s preschools in a quiet location that was dedicated to testing. Testing with the DIAL-3 required approximately 30 min, and testing with the phonological processing battery required approximately 45 min. Examiners were allowed to divide the testing into multiple sessions as need was established on a case by case basis. At the end of the assessments, children were rewarded with colorful, cartoon stickers. Testing of individual children with the phonological processing battery was completed within a 2-week period, and testing of a given cohort with the phonological processing battery took place over a 4-week period.
Exploratory factor analysis of DIAL-3
In an effort to determine if the DIAL-3 data could be appropriately used to index children’s general cognitive abilities, we performed exploratory factor analysis to examine the extent to which the factor structure of the DIAL-3 conformed to that of commonplace intelligence tests. This secondary analysis of data from Anthony et al. (2007) was possible because the DIAL-3 had been administered to all 2,144 English-speaking children enrolled in the Head Start program that provided services to the current sample. Four of the twenty-one DIAL-3 subtests were excluded from the current factor analysis because they measured early literacy or phonological skills included in the present study, and we did not want to artificially inflate associations of general cognitive ability with some of the study’s measures. The four excluded subtests were Letter Knowledge, Writing, Rhyming, and Rapid Color Naming.
Exploratory factor analysis of the remaining 17 DIAL-3 subtests using generalized least squares extraction yielded two components with eigenvalues greater than 1.0. Specifically, the Verbal Ability factor made substantial unique contributions to the variance of all subtests involving verbal comprehension (see pattern matrix in Table 1). In contrast, the Nonverbal Ability factor made substantial unique contributions to the variance of all subtests involving visual spatial abilities and visual motor integration (see pattern matrix in Table 1). These two components closely resembled the verbal and performance factors found on most intelligence tests, although the DIAL-3 is not an intelligence test per se. Following oblique rotation, the two components remained naturally correlated at .60, and as such, most subtests were moderately correlated with both components (see structure matrix in Table 1).
Factor scores, based on the standard regression method, were output for the 389 participants in the present study. These factor scores were assumed to index individuals’ verbal ability and nonverbal ability based on the pattern of factor loadings reported in Table 1. Furthermore, the two factor scores were considered indices of a General Cognitive Ability factor (g factor) in subsequent CFA and SEM. This specification was based on Carroll’s (1993) theory of intelligence as a hierarchical arrangement with a general intelligence factor at the apex and various more specialized abilities arrayed below, which is currently the most widely accepted conceptualization of intelligence (Neisser et al., 1996).
Preanalysis data inspection and transformation
Preanalysis data inspection included examination of patterns of missing data, potential outliers, and departures from normality. By design, approximately 40% of children were missing data on Initial Sound Matching, RAN Size, and Memory for Words. In addition, 21 children did not have verbal and nonverbal ability factor scores because they were not tested with the DIAL-3 at the beginning of the year for various reasons. Apparently, random missingness was evident for less than a handful of cases on Memory for Sentences, RAN objects, Word Reading, Text Discrimination, and age (see Table 2). To maximize power and to avoid biased estimation of parameter estimates from listwise deletion of cases with missing values, we employed full-information maximum likelihood estimation in Mplus 4.2, which makes use of all data present. This approach is acceptable under conditions of missing data by design and missing data by random. Nonetheless, a check on the equivalence of cohorts was performed by using cohort as a grouping variable in a multigroup analysis that demonstrated equivalent mean and covariance structures among all observed scores present in the two cohorts.
Some moderate outliers were detected, but examination of raw data provided no justification for eliminating these observations. Descriptive statistics for raw scores are reported in Table 2. The mean for Initial Sound Matching in this sample was unusually low, falling below the chance level of performance, so it was eliminated from further analyses. To control for the age range of the participants, we regressed raw scores from each phonological processing variable, each emergent literacy variable, and each factor score from the DIAL-3 onto chronological age and modeled relations among the residuals in all subsequent latent variable analyses. Mild to moderate departures from normality were found for several variables. Consequently, maximum likelihood robust methods were employed using Mplus 4.2, the Satorra-Bentler scaled chi-squares (S-B χ^sup 2^), and adjustments to the standard errors to the extent of the nonnormality in the raw data.
Data analysis plan
Data analysis progressed in four stages, in accord with our four research hypotheses. Within each stage of data analysis, most models were on a continuum of conceptual and empirical parsimony. On one end of the continuum was a least parsimonious base model with freely intercorrelated factors and freely estimated path coefficients. Progressing down the continuum were increasingly more parsimonious models with fewer estimated parameters (i.e., fewer estimated factor intercorrelations or path coefficients). Reduction of parameter estimates and corresponding increase in parsimony were the result of adding constraints on less parsimonious models. For example, constraining the freely estimated correlation of PA with PM in the separate but correlated abilities model in Stage 2 to 1.0 yielded Wagner et al. (1987) model of preliterate children’s PPA. In other words, many of the models of interest in a given stage of data analysis were nested in that they could be derived by imposing constraints on other models. This relation between models is important because the utility of one model to explain data can be statistically compared to the utility of another model in which it is nested by using chi-square difference tests. The base model always yields the best fit because it includes the most estimated parameters and is therefore the least restrictive model. However, the most restrictive nested model that fails to yield a significantly worse fit is accepted as superior in the interest of parsimony.
In all CFA and SEM analyses, the RAN factor was indexed by Rapid Size-Naming and Rapid Object-Naming. The PM factor was indexed by Memory for Sentences, Memory for Words, and Memory for Nonwords. The PA factor was indexed by Blending and Elision but not Initial Sound Matching like originally planned because the sample’s average performance on Initial Sound Matching was below the measure’s chance level of performance. The General Cognitive Ability factor (g) was indexed by the DIAL-3 Verbal Ability factor score and the DIAL-3 Nonverbal Ability factor score. Factor variances or factor disturbances were fixed to 1.0 to identify and standardize the RAN, PM, PA, and g factors.
Stage 1: Distinguishableness of phonological processing abilities from general cognitive ability
To evaluate the discriminant validity of PPA relative to general cognitive ability in English speaking preschoolers in general, we tested four a priori models using the full sample of 389 cases. The base model included freely estimated correlations among first-order RAN, PA, PM and g factors. Because this model, Model 1, included the most estimated parameters among the models examined in Stage 1, it provided the best characterization of the data and served as a basis for comparing alternative models that were more parsimonious. Using the commonly accepted rules of thumb that CFIs greater than 0.95, TLIs greater than 0.95, and RMSEAs less than 0.06 reflect excellent fitting models (Hu & Bentler, 1999), one can see from Table 3 that Model 1 explained the data very well (e.g., CFI=0.98). As expected, all observed variables loaded substantially and reliably onto their respective factors (λs=.42 to .76, ps
We compared the base model above to three three-factor models in order to test the distinguishableness of each PPA from general cognitive ability. Model 2 included a separate PA factor, a separate RAN factor, and a factor in which indices of g and PM loaded on a single factor. Model 3 included a separate PA factor, a separate PM factor, and a factor in which indices of g and RAN loaded on a single factor. Model 4 included a separate RAN factor, a separate PM factor, and a factor in which indices of g and PA loaded on a single factor. None of these three-factor models (i.e., Model 2, Model 3, and Model 4) characterized the data as well as the base four-factor model that specified all PPA as distinguishable from general cognitive ability (χ^sup 2^ differences [3, N=389]=35.5-55.8, ps
Stage 2: Distinguishableness of phonological processing abilities from each other
We first compared the utility of competing conceptualizations of PPA using the entire sample of 389 preschool children. To evaluate the commonplace conceptualization of PPA as specific abilities independent of their associations with general cognitive ability (de Jong & van der Leij, 1999; Sprugevica & Hoien, 2004; Wagner et al., 1993, 1994), we tested an a priori model that included freely estimated correlations among RAN, PA, PM, after covarying the effects of g on PPA. In other words, Model 5a included freely estimated correlations among the residuals of the latent RAN variable, latent PM variable, and latent PA variable. This model served as the base model for which to compare alternative, more parsimonious models of the structure of PPA in Stage 2 that are described below. Table 5 reports that Model 5a, representing the specific abilities conceptualization of PPA, explained the data extremely well (e.g., CFI=.98). It is important to note that the correlations among PPA factors were small, and only two of them were significant (i.e., PA with PM, r=.22, f=4.34,/> .40; see Model 5b in Table 5). Parameter estimates for this model are reported in Fig. 2 and Table 4.
Model 6a tested Wagner et al. (1987) conceptualization of preliterate children’s PPA as a separate RAN ability that is correlated with a latent phonological processing ability that drives children’s performances on both PA tests and PM tests. In other words, this model specified a separate RAN factor and a combined PA/PM factor. Model 6a did not fit well (e.g., TLI=.832, RMSEA=.12; see Table 5), and this model characterized the data significantly worse than the specific abilities model (x2 difference [3, /V=389]=114.10,p .; see Model 6b in Table 5). However, this more parsimonious version of Wagner et al. (1987) model still fit significantly worse than the two versions of the specific abilities model.
Finally, Model 7 tested one possible operationalization of a phonological core conceptualization of PPA. Specifically, Model 7 included a single latent PPA variable, separate from general cognitive ability, which drives children’s performances on RAN tests, PA tests, and PM tests. Model 7 poorly described these data (e.g., TLI=.79), and it described these data significantly worse than all models in which it was nested (see Table 5). In summary, results from Stage 2 analyses indicated that the specific abilities conceptualization of PPA best characterized children’s performances on the various phonological processing measures across the preschool-age period.
To explore the effects of different covariate models, we examined the magnitude of associations among latent RAN, PA, and PM in the specific abilities model under four conditions. Table 6 reports moderate correlations among PPA (rs=.46 to .68) when neither chronological age nor general ability was partialled. Statistically controlling for the effects of age had a small impact, reducing the range of correlations among PPA to .30-.59. In contrast, controlling for general ability had a big impact in reducing the size of correlations among PPA (rs=.00-. 16). Similarly, controlling for both age and cognitive ability had a big impact, reducing the range of correlations to .06-.26. Noteworthy findings were (a) RAN was no longer significantly correlated with PM when general cognitive ability was partialled, (b) the impact of partialing general cognitive ability was larger than the impact of partialing age, and (c) all three PPA were clearly distinguishable regardless of what covariates were included in the models.
Stage 3: Age-related differences in the mean and covariance structure of phonological processing abilities
Rather than presume the factor structure of PPA was consistent across the preschool years when children’s cognitive skills develop so rapidly, we examined the extent to which relations among phonological processing abilities may differ in younger and older preschool children. Examination of age-related differences was also examined as a potential explanation for researchers’ conflicting findings about the structure of PPA. We created groups of younger and older preschool children based on a median split of the sample on children’s chronological ages on the day of testing with the phonological processing battery. Model 5b (i.e., the separate ability model with a zero correlation between RAN and PM) served as the foundation to test a hierarchy of invariance constraints across the two age groups.
Model 8 is the multigroup, means structure, SEM model that corresponds to Model 5b. Model 8 characterized the data only moderately well (CFI = .93, TLI=.88) because of the stringent default that groups have equal means or intercepts on all observed scores. Allowing the older group of preschoolers to have higher means on the nonverbal ability factor score and on the memory for sentences score that were not accounted for by higher factor means on IQ and PM resulted in a much improved model fit (CFI=.96, TLI=.93; see Model 9 in Table 7). Table 7 reports the similar fits of models that imposed equality constraints across groups on factor loadings (i.e., Model 10); factor loadings and factor correlations (i.e., Model 11); factor loadings, factor correlations, and path coefficients from IQ to each PPA (i.e., Model 12); and even factor loadings, factor correlations, path coefficients from IQ to PPA, and all regressions of observed scores on children’s chronological age (i.e., Model 13). Amazingly, the highly constrained Model 13 that evidenced a strong degree of measurement equivalence in the two groups still characterized the data quite well (CFI=.95, TLI=.94). Constraining the residual variances, which are of little substantive interest, resulted in a poorly fitting model (i.e., Model 14).
Given the relatively strong measurement equivalence of the factors in the two age groups, it is appropriate to compare the groups’ latent abilities. The cleanest comparisons to be made between groups concern latent RAN abilities and latent PA abilities because complete measurement equivalence was demonstrated for each of these factors. The older group of preschoolers had a reliably higher factor mean on RAN (z=3.04, /> A0), but the older group had a higher residual mean score on the Memory for Sentences measure.
Stage 4: Age-related differences in the predictive validity of phonological processing abilities
To further investigate the discriminant validity and practical utility of RAN, PA, and PM, we examined if PPA had unique relations with nonphonological literacy skills. Model 13 served as the foundation model to which a new structural component was added. Specifically, three literacy observed variables (i.e., letter-knowledge, text discrimination, and word reading) were simultaneously regressed on the general cognitive ability factor (i.e., g) and the three phonological processing factors (i.e., RAN, PA, and PM). This structural component of the model, which is the focus of Stage 4, tests whether any of the literacy variables were uniquely associated with RAN, PA, PM, or IQ.
For the younger group of preschoolers, RAN was a significant unique predictor of letterknowledge (J3=.22, z=2.37, p
This study comprehensively investigated preschool children’s latent phonological processing abilities and the relations of these phonological abilities with emergent literacy. The study yielded five important findings. First, each of the phonological processing abilities (i.e., PA, PM, and RAN) was found to be distinguishable from general cognitive ability. second, the conceptualization of PPA as separate but correlated abilities best explained the nature of phonological processing abilities in this population. Third, relations among PPA are identical in younger and older preschool children, although these groups have different skill levels. Fourth and perhaps most important, phonological processing abilities were found uniquely related to these children’s emergent literacy skills, even after controlling for general cognitive ability. Fifth, the influence of general cognitive ability on children’s emergent literacy was indirect via their phonological processing. All of these findings are novel within the population of native English-speaking preschool children, but they are consistent with research conducted with school-age populations. Each finding and its implications will be discussed in turn.
Like many other researchers, we found that children’s phonological processing abilities were related to their general cognitive ability. Nonetheless, each of the three phonological processing abilities was distinguishable from general cognitive ability. Similar conclusions have been drawn concerning English-speaking school-age children (Wagner et al., 1993, 1994), Dutch-speaking school-age children (de Jong & van der Leij, 1999), Spanishspeaking preschool-age children (Anthony et al. 2006), and English-speaking preschool-age and kindergarten-age children (Wagner et al., 1987).
The present study went beyond examining the distinguishableness of each phonological processing ability from general cognitive ability to examining the extent that general cognitive ability explained the covariation among phonological processing abilities. This was an important extension of prior research given that prior studies have obtained moderate to high zero-order correlations among phonological processing abilities, and most authors have theoretically explained the interrelatedness among phonological processing abilities as a shared reliance on quality of phonological representations, efficiency of phonological code retrieval, or both. Our findings cast some doubt on these assertions because after covarying general cognitive ability, RAN was unrelated to PM. We obtained identical findings in a study of preschool-age speakers of Spanish (Anthony et al., 2006). Similarly, controlling for general cognitive ability reduced the correlation of PA with PM and the correlation of RAN with PA. This was also the case in Anthony et al. (2006). Of most interest, PA and PM only shared 5% of their residual variance, which is much less than commonly reported in studies that have not controlled for general cognitive ability [e.g., 100% in Lonigan et al. (2007; submitted for publication); 58% in Wagner et al, 1993; and 66% in de Jong & van der Leij, 1999]. In other words, our findings indicate that a substantial portion of the covariation among phonological processing abilities is attributable to general cognitive ability and that little covariation remains potentially attributable to quality of phonological representations.
Although a substantial portion of the covariation among phonological processing abilities appears attributable to general cognitive ability, PA and PM remained associated to a small degree even after general cognitive ability was partialled. This finding is not surprising when one considers the cognitive demands of most phonological awareness tasks. Phonological awareness tasks involve attending to the sounds in one or more words and then performing some type of cognitive operation, like comparing the sounds, blending the sounds, deleting the sounds, or moving the sounds around. That is, all phonological awareness tasks require holding acoustic information in phonological short-term memory while the central executive performs some type of working memory operation on the phonological information. Thus, it is fitting that children’s phonological awareness competencies are associated with their memory capacities, independent of general cognitive ability.
The second major finding from this study was that preschool children’s performances on phonological processing tests were best explained by the specific abilities conceptualization of phonological processing. This conclusion is identical to that drawn by other researchers who have studied school-age children who speak English (Wagner et al., 1993, 1994), Dutch (de Jong & van der Leij, 1999), and Latvian (Sprugevica & Hoien, 2004). The superiority of the specific ability conceptualization was also demonstrated with preschoolage speakers of Spanish (Anthony et al., 2006). Interestingly, the specific ability conceptualization characterized older and younger preschool children’s phonological skills equally well, although they differed in absolute skill levels. However as highlighted in the introduction, our favoring of the specific abilities conceptualization does not coincide with the conclusion reached by Wagner et al. (1987) or Lonigan et al. (2007; submitted for publication). Specifically, these other studies found that the performances of native English speaking preschoolers in the US were best described by a combined PA/PM factor and a separate but moderately correlated RAN factor.
Although similar in design and methods, the present study and that of Anthony et al. (2006) differ in a couple of important ways from that of Wagner et al. (1987). For example, our two studies statistically controlled for the effects of chronological age on children’s phonological processing scores; the Wagner et al. (1987) study did not covary the effects of age. However, this methodological difference is unlikely to explain the different findings given that controlling for age had such a small impact on the correlations among PPA in the present study. Instead, we believe the primary reason for the seemingly contradictory conclusions of our preschool studies and that of Wagner et al. (1987) is the way phonological awareness was assessed in these studies. In general, the PA tests used by Wagner et al. (1987) were more difficult and involved a much greater load on phonological short-term memory than the tests used in the present study and Anthony et al. (2006). For example, three of the five PA tests used by Wagner et al. (1987) required a verbal response from children, like producing a blended word, an elided word, or an elided nonword. Another PA test used by Wagner et al. (1987) involved tapping syllables in words. Only one of the five PA tests used by Wagner et al. (1987) could be successfully performed by using a rudimentary strategy of comparing phonological information. Most importantly though, all of the PA tests used by Wagner et al. (1987) required the unassisted holding of phonological information in memory while a given task was performed. Consistent with the nature of the PA tests administered and with the results they obtained, Wagner and colleagues concluded that “We consider performance on the phonological awareness tasks that we examined to be determined largely, or at least in part, by efficient coding in working memory” (Wagner et al., 1987, pg. 369).
In contrast, it could be argued that half of the phonological awareness items used in the present study and Anthony et al. (2006) could be successfully answered using a rudimentary phonological comparison strategy. Moreover, 50% of the PA items in the present study involved a multiple-choice format, and the response choices were illustrated to reduce the memory load while children performed the PA task. The rationale behind this testing format was that many traditional PA tests place too large of demands on phonological memory for preschool children, and therefore preschool children can perform poorly on such tests either because of limited phonological awareness development or because of limited capacity to hold the phonological information in short-term memory. By using pictures to aid memory and by using a combination of recognition tasks and production tasks, we have been able to construct more sensitive measures that tap a broader range of phonological awareness ability (Anthony et al., 2003; Lonigan et al., 1998). While tasking memory to a lesser degree and measuring lower levels of phonological awareness ability, these type of tests have been demonstrated to measure the same latent phonological ability as more traditional measures of phonological awareness (Anthony & Lonigan, 2004; Anthony et al., 2002). In summary, our view is that the different findings between our research and that of Wagner et al. (1987) are no cause for concern but are instead a product of improved sensitivity of preschool measures of phonological processing over the last two decades.
In an attempt to reconcile our findings in the present study and in Anthony et al. (2006) with those of Lonigan et al. (2007; submitted for publication), who like Wagner et al. (1987) found PA and PM so highly correlated that they appeared to index the same phonological ability, we again turn to the methodological similarities and differences for a potential answer. All three studies included similar measures of phonological awareness that were designed to minimize the load on phonological memory. Furthermore, all three studies controlled for the effects of age. The most notable difference between our studies and that of Lonigan et al. was that our studies controlled for general cognitive ability, whereas that of Lonigan et al. did not. This methodological difference seems a likely candidate for explaining the seemingly conflicting findings in light of the present study’s demonstration that controlling for general cognitive ability has a substantial impact of decreasing the associations among PPA. However, regardless of whether general cognitive ability was statistically controlled, the present study always supported the specific abilities conceptualization of PPA.
When conducting a study on the nature of PPA, researchers must make some critical decisions about which potential confounds to statistically partial. This is a complex issue because on the one hand, individual differences in PPA may to some extent truly covary with chronological age and general cognitive ability, but on the other hand if one is interested in the magnitude of how closely PPA are related to each other and to other literacy constructs then one must take care not to inflate these relations by using a heterogeneous sample. Our examination of the impact of different covariate models found that the decision researchers make is nontrivial and that some decisions are more likely than others to have a profound impact on the conclusions of a study.
The present findings support a special role of PA in preschool-age children’s early literacy development. These children’s phonological awareness was clearly distinguishable from their general cognitive abilities, RAN abilities, and PM abilities. More importantly, relative to all of the PPA and general cognitive ability, PA was the best predictor of older preschool children’s decoding skills. Such findings are consistent with research that has repeatedly shown phonological awareness of young children is involved in acquiring literacy (e.g., Brady et al., 1994; Bryant et al., 1989; Lonigan et al., 2000; Lundberg et al., 1988). Of course, there is also evidence that suggests phonological awareness may be involved in reciprocal relations with letter knowledge (Burgess, 2006; Burgess & Lonigan, 1998) and word reading (Anthony & Francis, 2005; Perfetti et al., 1987; Wagner et al., 1997; Ziegler & Goswami, 2005).
The present study provided strong evidence of the validity and importance of RAN in the literacy development of preschool-age children. Although correlated with other phonological processing abilities via general cognitive ability, RAN was clearly distinguishable from other phonological processing abilities and general cognitive ability. Moreover, preschool children’s RAN abilities were uniquely related to important emergent literacy skills, like letter knowledge and text discrimination. These findings are consistent with other research that shows RAN is predictive of preschoolers’ letter knowledge (Anthony et al., 2006) and school-age children’s reading abilities (Felton & Brown, 1990; Griffiths, 1991; Wagner et al., 1997; Wolf & Obregon, 1992). In accord with Wagner and Torgesen’s conceptualization of RAN, we interpret the present findings as indicating that children who more efficiently access phonological codes for lexical items also more readily learn the names of letters and the sounds associated with letters than do children who are less efficient at phonological access. Furthermore, we believe this relation is not fully explained by children’s general cognitive ability.
It should be noted that the associations we found between RAN and early literacy skills were not a function of shared method variance, as could be argued is the case in many studies. That is, the tasks we used to measure RAN involved rapid automatic naming of common objects or simple concepts (e.g., relative size), not rapid naming of letter names or letter sounds. In addition, the possibility that RAN emerged as a specific ability because of the uniquely speeded nature of the RAN tasks in the assessment battery was weakened, although not entirely disconfirmed, by findings that showed how poorly these data were explained by a two factor model in which speeded RAN tasks indexed one factor and all nonspeeded cognitive tasks (i.e., power tests) indexed a second factor. However, additional research is certainly needed to fully establish RAN as distinguishable from efficiency of different kinds of mental processing measured by speed tests. This call for research is equally relevant for school-age populations, given the field’s limited understanding of this multicomponential construct beyond the fact that it is predictive of literacy (Vukovic & Siegel, 2006; however, see Georgiou et al., 2006). Additional research is certainly needed to investigate the developmental relations of early RAN abilities with other phonological processing and emergent literacy skills.
The last major finding from the present study, which was not hypothesized, was that children’s general cognitive ability did not directly predict their emergent literacy skills. Instead, the effects of general cognitive ability were indirect via phonological processing. Similar results were obtained by Anthony et al. (2006) with preschool children who were speakers of Spanish. One implication of this finding is that children’s early literacy skills will be better predicted by assessments of their phonological processing abilities than by assessments of their general cognitive ability. This implication highlights the importance of including measures of phonological processing in early literacy screening and progress monitoring batteries. second, that general cognitive ability only had indirect effects on emergent literacy permits a variety of phonological processing abilities to influence the reading acquisition process, and each of these phonological processing abilities may serve as a potential locus of instruction or early intervention for children at risk of reading failure. This is an exciting implication that brings optimism and hope, given that early childhood educators are more successful at teaching phonological awareness, for example, than raising IQ.
Of course the conclusions we made above need to be qualified according to the limitations of the present study. Most noteworthy is that the present study fell short of providing a developmental examination of early phonological processing abilities and literacy acquisition because it essentially employed a cross-sectional, single-shot case study design. Accordingly, issues of directionality and causality of the associations among phonological processing abilities and emergent literacy could not be addressed. However, we hope to address some of these critical issues with a longitudinal follow-up of the present sample. Like all studies, the present study was also limited in its generalizability. The sample we studied was comprised of 3-, 4-, and 5-year-old minority children who were at risk for reading difficulties because of conditions associated with poverty. Given this limitation, future research should investigate the present research questions with samples of typically developing children and samples of children who speak languages other than English. Regarding second language acquisition, future research is needed to investigate the development of phonological processing within and across languages. Exciting extensions of this work will hopefully identify the extent of transfer of phonological processing abilities between languages, how phonological processing abilities in both languages relate to literacy acquisition in both languages, and how these interdependencies are impacted by instruction and different instructional models, like immersion, transition, and dual-language models of instruction.
Acknowledgement We thank Seth Allen, Deborah Corbitt-Shindler, and Yingchu Velasquez for their outstanding management and leadership, which played critical roles in the success of this project.
Adams, M. J. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA: MIT Press.
Anthony, J. L., Assel, M. A., & Williams, J. M. (2007). Exploratory and confirmatory factor analysis of the DIAL-3: What does this “screener” really measure? Journal of School Psychology; 45, 423^138.
Anthony, J. L., & Francis, D. J. (2005). Development of phonological awareness. Current Directions in Psychological Science, 14, 255-259.
Anthony, J. L., & Lonigan, C. J. (2004). The nature of phonological sensitivity: Converging evidence from four studies of preschool and early grade-school children. Journal of Educational Psychology, 96, 43-55.
Anthony, J. L., Lonigan, C. J., Burgess, S. R., Driscoll Bacon, K., Phillips, B. M., & Cantor, B. G. (2002). Structure of preschool phonological sensitivity: Overlapping sensitivity to rhyme, words, syllables, and phonemes. Journal of Experimental Child Psychology, 82, 65-92.
Anthony, J. L., Lonigan, C. J., Driscoll, K.., Phillips, B. M., & Burgess, S. R. (2003). Phonological sensitivity: A quasi-parallel progression of word structure units and cognitive operations. Reading Research Quarterly, 38, 470-487.
Anthony, J. L., Williams, J. M., McDonald, R., Corbitt-Shindler, D” Carlson, C. D., & Francis, D. J. (2006). Phonological processing and emergent literacy in Spanish speaking preschool children. Annals of Dyslexia. 56, 239-270.
Brady, S., Fowler, A., Stone, B., & Winbury, N. (1994). Training phonological awareness: A study with inner-city kindergarten children. Annals of Dyslexia, 44, 26-59.
Bryant, P. E., Bradley, L” MacLean, M., & Crossland, J. (1989). Nursery rhymes, phonological skills and reading. Journal of Child Language, 16, 407-428.
Bryant, P. E., MacLean, M” Bradley, L. L., & Crossland, J. (1990). Rhyme and alliteration, phoneme detection, and learning to read. Developmental Psychology, 26, 429^138.
Burgess, S. R. (2006). Development of phonological sensitivity. In D. K. Dickinsen, & S. B. Neuman (Eds.), Handbook of Early Literacy Research: Second Edition (pp.90-100). New York: Guilford Press.
Burgess, S. R., & Lonigan, C. J. (1998). Bidirectional relations of phonological sensitivity and prereading abilities: Evidence from a preschool sample. Journal of Experimental Child Psychology, 70, 117-141.
Carroll, J. B. (1993). Human cognitive abilities. Cambridge: University of Cambridge Press,
de Jong, P., & Van der Leij, A. (1999). Specific contributions of phonological abilities to early reading acquisition: Results form a Dutch latent variable longitudinal study. Journal of Educational Psvchoiogy, 91, 450-476.
Felton, R. H., & Brown, I. S. (1990). Phonological processes as predictors of specific reading skills in children at risk for reading failure. Reading and Writing: An Interdisciplinary Journal, 2, 39-59.
Georgiou, G. K., Parrila, R., & Kirby, J. (2006). Rapid naming speed components and early reading acquisition. Scientific Studies of Reading, 10, 199-220.
Griffiths, P. (1991). Word-finding ability and design fluency in developmental dyslexia. British Journal of Clinical Psychology, 30, 47-60.
Hoskyn, M., & Swanson, L. (2000). Cognitive processing of low achievers and children with reading disabilities: A selective meta-analytic review of the published literature. School Psychology Review, 29, 102-119.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.
Kamhi, A., & Catts, H. (1989). Reading disabilities: A developmental language perspective. Boston: College-Hill.
Lonigan, C. J., Anthony, J. L., Phillips, B. M., Purpura, D. J., McQueen, J., & Wilson, S. B. (2007). The nature of preschool phonological processing abilities and their relations to vocabulary, general cognitive abilities, and print knowledge. In preparation, Florida State University.
Lonigan, C. J., Burgess, S. R., & Anthony, J. L. (2000). Development of emergent literacy and early reading skills in preschool children: Evidence from a latent variable longitudinal study. Developmental Psychology, 56(5), 596-613.
Lonigan, C. J., Burgess, S. R., Anthony, J. L., & Barker, T. A. (1998). Development of phonological sensitivity in two- to five-year-old children. Journal of Educational Psychology, 90, 294-311.
Lonigan, C, Wagner, J., & Rashotte, C. (2002). The preschool comprehensive test of phonological and print processing. Tallahassee, FL: Florida State University.
Lundberg, I., Frost, J., & Petersen, O. P. (1988). Effects of an extensive program for stimulation phonological awareness in preschool children. Reading Research Quarterly, 23, 263-284.
Mardell-Czudnowski, C, & Goldenberg, D. S. (1998). Developmental indicators for the assessment of learning (3rd ed.). Circle Pines, MN: American Guidance Service.
National Early Literacy Panel (March 2007). Findings from the National Early Literacy Panel: Providing a Focus for Early Language and Literacy Development. 16th Annual National Conference on Family Literacy. Orlando, FL.
Neisser, U., Boodoo, G., Bouchard, T J., Boykin, A. W., Brody, N., Ceci, S. J., et al. (1996). Knowns and unknowns. American Psychologist, 51, 77-101.
Perfetti, C. A., Beck, I., Bell, L. C, & Hughes, C. (1987). Phonemic knowledge and learning to read are reciprocal. Merrill-Palmer-Quarterly, 33, 283-319.
Rapala, M. M., & Brady, S. (1990). Reading ability and short-term memory: The role of phonological processing. Reading and Writing: An Interdisciplinary Journal, 2, 1-25.
Rohl, M., & Pratt, C. (1995). Phonological awareness, verbal working memory and the acquisition of literacy. Reading and Writing: An Interdisciplinary Journal, 7, 327-360.
Shankweiler, D., Crain, S., Katz, L., Fowler, A. E, Liberman, A. M., Brady, S. A., et al. (1995). Cognitive profiles of reading-disabled children: Comparison of language skills in phonology, morphology, and syntax. Psychological Science, 6, 149-156.
Snowling, M., Goulandris, N., Bowlby, M., & Howell, P. (1986). Segmentation and speech perception in relation to reading skill: A developmental analysis. Journal of Experimental Child Psychology, 41, 489-507.
Sprugevica, I., & Hoien, T. (2004). Relations between enabling skills and reading comprehension: A follow-up study of Latvian students from first to second grade. Scandinavian Journal of Psychology, 45, 115-122.
Stanovich, K. E. (1988). Explaining the differences between the dyslexic and the garden-variety poor reader: The phonological-core variable-difference model. Journal of Learning Disabilities. 21, 590-612.
Stanovich, K. E., & Siegel, L. S. (1994). Phenotypic performance profile of children with reading disabilities: A regression-based test of the phonological-core variable-difference model. Journal of Educational Psychology, 86, 24-53.
Stone, B., & Brady, S. (1995). Evidence for deficits in basic phonological processes in less-skilled readers. Annals of Dyslexia, 45, 51-78.
Vukovic, R. K., & Siegel, L. S. (2006). The double-deficit hypothesis: A comprehensive analysis of the evidence. Journal of Learning Disabilities, 39, 25-47.
Wagner, R. K., Balthazor, M., Hurley, S” Morgan, S., Rashotte, C. A., Shaner, R” et al. (1987). The nature of prereders’ phonological-processing abilities. Cognitive Development, 2, 355-373.
Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological processing and its causal role in the acquisition of reading skills. Psychological Bulletin, 101, 192-212.
Wagner, R. K., Torgesen, J. K., Laughon, P., Simmons, K., & Rashotte, C. A. (1993). Development of young readers’ phonological processing abilities. Journal of Educational Psychology, 85, 83-103.
Wagner, R. K., Torgesen, J. K., & Rashotte, C. A. (1994). Development of reading-related phonological processing abilities: New evidence of bidirectional causality from a latent variable longitudinal study. Developmental Psychology, 30, 73-87.
Wagner, R. K., Torgesen, J. K., Rashotte, C. A., Hecht, S. A., Barker, T. A., Burgess, S. R., et al. (1997). Changing relations between phonological processing abilities and word-level reading as children develop from beginning to skilled readers: A 5-year longitudinal study. Developmental Psychology, 33, 468-479.
Wolf, M., & Obregon, M. (1992). Early naming deficits, developmental dyslexia, and a specific deficit hypothesis. Brain and Language, 42, 219-247.
Ziegler, J. C, & Goswami, U. C. (2005). Reading acquisition, developmental dyslexia and skilled reading across languages. Psychological Bulletin, 131, 3-29.
J. L. Anthony (*)
Division of Developmental Pediatrics, University of Texas Health Science Center at Houston,
7000 Fannin Street, Suite 2377, Houston, TX 77030, USA
J. M. Williams * D. J. Francis
Department of Psychology, University of Houston, Houston, TX, USA
Department of Psychology, Southern Methodist University, University Park, TX, USA
Copyright International Dyslexia Association Dec 2007
Provided by ProQuest Information and Learning Company. All rights Reserved