development of emergent literacy and early reading …faculty.swosu.edu/stephen.burgess/development...

Developmental Psychology2000, Vol. 36, No. 5, 596-613

Copyright 2000 by the American Psychological Association, Inc.0012-1649/00/S5.00 DOI: 10.1037//OOI2-1649.36.5.596

Development of Emergent Literacy and Early Reading Skills in PreschoolChildren: Evidence From a Latent-Variable Longitudinal Study

Christopher J. Lonigan, Stephen R. Burgess, and Jason L. AnthonyFlorida State University

Although research has identified oral language, print knowledge, and phonological sensitivity asimportant emergent literacy skills for the development of reading, few studies have examined therelations between these aspects of emergent literacy or between these skills during preschool and duringlater reading. This study examined the joint and unique predictive significance of emergent literacy skillsfor both later emergent literacy skills and reading in two samples of preschoolers. Ninety-six children(mean age = 41 months, SD = 9.41) were followed from early to late preschool, and 97 children (meanage = 60 months, SD = 5.41) were followed from late preschool to kindergarten or first grade. Structuralequation modeling revealed significant developmental continuity of these skills, particularly for letterknowledge and phonological sensitivity from late preschool to early grade school, both of which were theonly unique predictors of decoding.

Reading skills provide a crucial piece of the foundation forchildren's academic success. Children who read early and wellexperience more print exposure and consequent growth in numer-ous knowledge domains (Cunningham & Stanovich, 1997; Echols,West, Stanovich, & Zehr, 1996; Morrison, Smith, & Dow-Ehrensberger, 1995). In contrast, children who lag behind in theirreading skills receive less practice in reading than other childrendo (Allington, 1984), miss opportunities to develop reading com-prehension strategies (Brown, Palincsar, & Purcell, 1986), oftenencounter reading material that is too advanced for their skills(Allington, 1984), and may acquire negative attitudes about read-ing itself (Oka & Paris, 1986). Such processes may lead to whatStanovich (e.g., 1986) termed a Matthew effect, in which poorreading skills impede learning in other academic areas (Chall,Jacobs, & Baldwin, 1990), which increasingly depend on readingacross the school years.

Although the development of skilled reading occurs withoutsignificant problems for the majority of children, an estimated one

Christopher J. Lonigan, Stephen R. Burgess, and Jason L. Anthony,Department of Psychology, Florida State University.

Stephen R. Burgess is now at the Department of Psychology, South-western Oklahoma State University.

Preparation of this article was supported, in part, by grants from theNational Institute of Child Health and Human Development (HD36067,HD36509) and the Administration for Children and Families (90-YF-0023); the views expressed herein are the authors' and have not beencleared by the grantors.

We wish to acknowledge the contributions of the child-care centers, thedirectors, and the personnel who assisted with this project as well as thechildren and parents who made it possible. We thank Sarah Dyer, BrenleeBloomfield, Crystal Carr, Tracy Ferguson, Kimberly Ingram, DanielleKarlau, Nikki Sutton, Emily Shock, and other students at Florida StateUniversity for their assistance with data collection.

Correspondence concerning this article should be addressed to Christo-pher J. Lonigan, Department of Psychology, Florida State University,Tallahassee Florida 32306-1270. Electronic mail may be sent [email protected].

in three children experience significant difficulties in learning toread (Adams, 1990). There is strong continuity between the skillswith which children enter school and their later academic perfor-mance. Those children who do experience early difficulties inlearning to read are likely to continue to experience readingproblems throughout the school years (Baydar, Brooks-Gunn, &Furstenberg, 1993; Felton, 1998; Stevenson & Newman, 1986;Tramontana, Hooper, & Selzer, 1988) and into adulthood (Bruck,1998). For instance, Juel (1988) reported that the probability thatchildren would remain poor readers at the end of the fourth gradeif they were poor readers at the end of the first grade was .88.Children who enter school with limited reading-related skills are athigh risk of qualifying for special education services. In fact, themajority of school-age children referred for special educationevaluation are referred because of unsatisfactory progress in read-ing (Lentz, 1988).

Whereas more traditional approaches to the study of readingoften take as their starting point children's entry into the formalschool environment, an emergent literacy approach conceptualizesthe acquisition of literacy as a developmental continuum with itsorigins early in the life of a child, rather than as an all-or-nonephenomenon that begins when children start school. An emergentliteracy approach departs from other perspectives on reading ac-quisition in suggesting that there is no clear demarcation betweenreading and prereading. Emergent literacy consists of the skills,knowledge, and attitudes that are presumed to be developmentalprecursors to conventional forms of reading and writing (Sulzby &Teale, 1991; Teale & Sulzby, 1986; Whitehurst & Lonigan, 1998),and thus it suggests that significant sources of individual differ-ences in children's later reading skills are present prior to schoolentry. Previous research has identified a number of potentiallyimportant components of emergent literacy. Whitehurst and Loni-gan (1998) recently outlined different components of emergentliteracy skills and identified three factors that appear to be asso-ciated with preschool children's later word-decoding abilities: orallanguage, phonological processing abilities, and print knowledge.

596

EMERGENT LITERACY AND EARLY READING 597

Reading is a process of translating visual codes into meaningfullanguage. In the earliest stages, reading in an alphabetic systeminvolves decoding letters into corresponding sounds and linkingthose sounds to single words. A substantial body of research hasdemonstrated positive correlations and longitudinal continuity be-tween individual differences in oral language skills and laterdifferences in reading (e.g., Bishop & Adams, 1990; Butler,Marsh, Sheppard, & Sheppard, 1985; Pikulski & Tobin, 1989;Scarborough. 1989; Share, Jorm, MacLean, & Mathews, 1984).Whereas the connection between oral language and reading is clearfor reading comprehension (e.g., Snow, Barnes, Chandler, Hernp-hill, & Goodman, 1991), some studies indicate that vocabularyskills also have a significant impact on decoding skills very earlyin the process of learning to read (e.g., Wagner et al.. 1997).Additionally, oral language appears to be related to a secondemergent literacy skill, phonological sensitivity, as defined below.Studies of both preschool (e.g., Burgess & Lonigan, 1998; Chaney,1992; Lonigan, Burgess, Anthony, & Barker, 1998) and earlyelementary school children (e.g., Bowey, 1994; Wagner, Torgesen,Laughon, Simmons, & Rashotte, 1993; Wagner et al., 1997) havedemonstrated significant concurrent and longitudinal correlationsbetween children's vocabulary skills and their phonologicalsensitivity.

Phonological sensitivity refers to sensitivity to and ability tomanipulate the sound structure of oral language. Research with avariety of populations and using diverse methods has converged onthe finding that phonological sensitivity plays a critical and causalrole in the normal acquisition of reading (e.g., Adams, 1990; Byrne& Fielding-Bamsley. 1991; Slanovich, 1992; Wagner & Torgesen,1987). Children who are better at detecting and manipulatingsyllables, rhymes, or phonemes are quicker to learn to read, andthis relation is present even after variability in reading skill owingto factors such as IQ, receptive vocabulary, memory skills, andsocial class is partialed out (e.g., Bryant, MacLean, Bradley, &Crossland, 19%; Wagner & Torgesen, 1987; Wagner, Torgesen, &Rashotte, 1994). Moreover, studies of disabled and poor readersindicate that there is a core phonological deficit in nearly all poorreaders regardless of whether their reading abilities are consistentor inconsistent with their general cognitive abilities (Stanovich,1988; Stanovich & Siegel, 1994; Torgesen, 1999).

In addition to oral language and phonological sensitivity, as-pects of children's print knowledge seem to be important emergentliteracy skills. For example, knowledge of the alphabet (i.e., know-ing the names of letters and the sounds they represent) at entry intoschool is one of the strongest single predictors of short- andlong-term success in learning to read (e.g., Adams, 1990; Steven-son & Newman, 1986). Understanding the conventions of print(e.g., left-to-right and top-to-bottom orientation of print, the dif-ference between pictures and print on a page; Clay, 1979a, 1979b)and the functions of print (e.g., that the print tells a story or givesdirections; Purcell-Gates, 1996; Purcell-Gates & Dahl, 1991) alsoappears to aid in the process of learning to read. For example,Tunmer, Herriman, and Nesdale (1988) found that children'sscores on Clay's (1979a) Concepts About Print (CAP) Test at thebeginning of first grade predicted their reading comprehension anddecoding abilities at the end of second grade even after Tunmer etal. controlled for differences in vocabulary and metalinguisticawareness. Some emergent literacy advocates have also suggestedthat children's faculty with environmental print (e.g., recognizing

product names from signs and logos) reflects their early printawareness by demonstrating the ability to derive the meaning oftext within context (e.g., Goodman, 1986).

Despite some evidence for associations between emergent liter-acy and later reading, there have been relatively few studiesexamining the relations between these multidimensional aspects ofemergent literacy or between these components during the pre-school period and later reading skills. As noted above, aspects oforal language appear to be related to phonological sensitivity.Children's letter knowledge also appears to be associated withsome aspects of phonological sensitivity (Bowey, 1994; Stahl &Murray, 1994) and growth in these skills (Burgess &. Lonigan,1998; Wagner et al., 1994, 1997). Evidence from school-agechildren indicates that these three components of emergent literacyare causally related to each other and to later reading (e.g., Wagneret al., 1997); however, how they relate to each other during thepreschool period is not known. Consequently, it is not clearwhether there are interactions between these different emergentliteracy skills or whether they are relatively independent of oneanother, and thus a well-elaborated developmental model of pre-school emergent literacy and its relation to conventional literacycannot be advanced. Moreover, basic questions concerning thenature of preschool phonological sensitivity currently remain un-answered, as discussed below.

The majority of evidence linking phonological sensitivity inprereaders with the development of reading has come from studiesthat have assessed phonological skills at the point of school entrybut prior to formal reading instruction (e.g., Bradley & Bryant,1983, 1985; Share et al., 1984; Stanovich, Cunningham, & Cra-mer, 1984; Wagner etal., 1994, 1997). Compared with research onschool-age children's phonological sensitivity, there has been sig-nificantly less systematic study of preschool children's phonolog-ical sensitivity, and many of these studies have been limited bysmall sample sizes, use of only one or two measures of phonolog-ical sensitivity, and other methodological weaknesses.

In one of the more extensive studies to date, MacLean, Bryant,and Bradley (1987) administered a rhyme detection task and aknowledge-of-nursery-rhymes task to a group of 66 three-year-oldchildren. When the children were 4'/2 years old, their ability toread 12 simple high-frequency words was assessed. Comparedwith nonreaders, children who could read some of these wordsscored higher on the earlier rhyme and alliteration measures.Bryant et al. (1990) reported additional data on these children, whocompleted additional rhyme and alliteration detection tasks whenthey were about 4'/i years old, phoneme deletion and phonemetapping tasks when they averaged about 6 years of age, and readingand spelling tests when they averaged about 6V2 years of age.Bryant et al. found that the rhyme and alliteration detection teststhat had been administered when the children were AV2 werecorrelated with the later phoneme deletion and phoneme tappingtasks (average r = .48), and scores on these rhyme and alliterationtasks significantly added to the prediction of reading and spellingscores independent of mothers' educational level, child age, IQ,receptive vocabulary, and the score on either the phoneme deletionor phoneme tapping tasks.

Evidence suggests that there is a developmental hierarchy ofchildren's sensitivity to linguistic units at different levels of com-plexity. Children achieve syllabic sensitivity earlier than theyachieve sensitivity to phonemes, and children's sensitivity to in-

598 LONIGAN, BURGESS, AND ANTHONY

trasyllabic units (i.e., onset-rime) also precedes sensitivity to pho-nemes (Fox & Routh, 1975; I. Liberman, Shankweiler, Fischer, &Carter, 1974; Lonigan et al., 1998; Treiman, 1992). However,there is controversy concerning whether sensitivity to lower levelsof linguistic complexity (i.e., syllables, onset-rime) represents pro-cesses important for reading. Sensitivity to phonemes is oftenassumed to have special status in the relation between phonolog-ical sensitivity and reading both because it is at this level thatgraphemes correspond to speech sounds in reading and becauseindividual phonemes do not have separable physical reality (e.g.,A. Liberman, Cooper, Shankweiler, & Studdert-Kennedy, 1967;Morais, 1991; Muter, Hulme, Snowling, & Taylor, 1997; Nation &Hulme, 1997; Tunmer & Rohl, 1991). Other authors have sug-gested that children's abilities to detect rhyme facilitate readingthrough a mechanism different than sensitivity to phonemes (e.g.,Goswami & Bryant, 1990, 1992).

Both of these views, however, assume that there is more thanone type of phonological sensitivity and that the different typesmay be more or less related to reading. Most extant studiescompared the ability of different phonological sensitivity measures(e.g., measures of rhyme vs. phonemic sensitivity) to predictreading. In these studies, when one measure of phonologicalsensitivity predicts reading better—typically defined as a signifi-cant semi-partial correlation obtained while controlling for theother measure of phonological sensitivity—the results are taken assupporting the crucial importance of the skill supposedly measuredby that task (e.g., phonemic sensitivity). For example, in support ofthe importance of phonemic sensitivity, both Muter et al. (1997)and Nation and Hulme (1997) reported that children's abilities toperform a phoneme segmentation task were more strongly relatedto reading and spelling than were their abilities to detect andproduce rhyme. Goswami and Bryant (1992) also reported dataconsistent with separate domains of phonological sensitivity. Intheir study, when they controlled for phonemic sensitivity, rhym-ing abilities facilitated children's abilities to make use of analogyin reading unfamiliar words. There are several problems with thispredictive approach. First, it assumes a priori that there are differ-ent types of phonological sensitivity. Second, it ignores the factthat the most predictive element in such studies is the overlapbetween measures (i.e., the degree of shared predictive variance issubstantially larger than the degree of unique predictive variancefor a given test). Finally, these types of analyses fail to take intoaccount the effects of differential reliability of the measures. Thatis, the more reliable of two variables measuring equally predictiveconstructs (or the same construct) generally will capture the uniquepredictive variance.

Prior to examining differential predictive validity, therefore, it isimportant that we determine whether there is more than one typeof phonological sensitivity. Extant evidence suggests that sensitiv-ity to onset-rimes, syllables, and phonemes represents the sameunderlying ability. For example, Stahl and Murray (1994) admin-istered to 113 kindergarten and first-grade children four differenttasks varying in linguistic complexity. Separate factor analyses ofthe four tasks across linguistic complexity and of the four levels oflinguistic complexity across tasks each yielded a single-factorsolution that explained a majority of the variance in children'sperformance on the measures. Using confirmatory factor analysis,Anthony et al. (2000) found that a single factor provided the bestfit to preschool children's scores on measures of rhyme, syllable,

and phoneme sensitivity (see also Anthony & Lonigan, 2000). Incontrast to these findings, however, Hoien, Lundberg, Stanovich,and Bjaalid (1995) reported evidence for the distinction betweensensitivity to phonemes and sensitivity to rhyme and syllables.They found separate factors for phonemic sensitivity, syllabicsensitivity, and rhyme sensitivity in 6- and 8-year-old Norwegianchildren, and scores on all three factors independently predictedreading abilities for the older group of children. However, it isdifficult to interpret the results of Hoein et al. because only onetask defined the Rhyme Sensitivity and Syllabic Sensitivityfactors.

Several predictive studies also support a unitary view of pho-nological sensitivity. For example, Lonigan et al. (1998) demon-strated that preschool children's performance on tasks measur-ing phonological sensitivity at the syllable, onset-rime, and pho-neme levels was associated with measures of letter knowledge anddecoding. Similarly, studies of school-age children by Wagner andcolleagues (Wagner et al., 1993, 1994, 1997) found that a latentvariable defined by rhyme, syllable-level, and phoneme-level taskswas associated strongly, concurrently, and longitudinally withchildren's reading skills. These findings indicate that the variancecommon to phonological sensitivity tasks measuring different lev-els of linguistic complexity represents the predictive aspect of thephonological sensitivity construct. However, the majority of pre-dictive studies have involved either school-age children or rela-tively small samples. Consequently, little is known about thenature and predictive importance of preschool children's develop-ing phonological sensitivity.

Questions concerning the nature of preschool phonological sen-sitivity (i.e., whether it is a unitary or a multidimensional con-struct), the independence of phonological sensitivity, oral lan-guage, and print knowledge, as well as the significance of thesethree components of emergent literacy for later reading are impor-tant because studies demonstrate that there are highly stable indi-vidual differences in these abilities from kindergarten forward(Wagner et al., 1994, 1997). Such findings suggest that the pre-school period is an important source of development in skillsassociated with later reading. Our goals in this study were toexamine the nature of preschool emergent literacy as well as thejoint and unique predictive significance of preschool emergentliteracy skills for later reading. We examined the development ofemergent literacy and early reading longitudinally in two samplesof preschool children who overlapped in age at different assess-ment points and in the measures they completed. We used struc-tural equation modeling to address questions about the nature ofpreschool phonological sensitivity, the independence of differentemergent literacy skills, and the developmental significance ofthese skills across time from the early preschool period to kinder-garten and first grade.

Method

Participants

Data from two groups of preschool-age children, recruited through 13different preschools and child-care centers serving middle- to upper-


income families, were used in this study.1 One group of children consistedof 96 younger preschoolers who completed follow-up testing (i.e., Time 2)approximately 18 months after their initial (i.e., Time 1) assessments.These children ranged in age from 25 to 61 months (M = 41.02 months,SD = 9.41) at Time 1. The majority of the younger group of children wereWhite (89.6%), and 56.3% were girls. The second group of childrenconsisted of 97 older preschoolers who completed follow-up tests (i.e.,Time 2) approximately 12 months after their initial (i.e., Time 1) testing.This group of children ranged in age from 48 to 64 months (M = 60.04months, SD = 5.41) at Time 1. Most of the older group of children wereWhite (96.9%), and 52.6% were girls. Only children who had completed allmeasures at both Time 1 and Time 2 were included in these samples. Anadditional 58 children (18 from the older sample) who had incomplete dataon measures at Time 1 or Time 2 either because they refused continuedparticipation or because they could not be located for the Time 2 testingwere excluded from the results reported in this study. With the exceptionthat excluded children in the younger sample scored lower on the rhymeoddity task (p = .04), excluded children did not differ from includedchildren on any measure.

These two samples of children originally were recruited for separate butrelated research projects, and there was substantial overlap in the primarymeasures administered in each project. The grouping used in this studymaintained the original division between samples because of the differentfollow-up periods associated with each and the commonality of measuresadministered at each assessment within a sample. Information concerningchildren's family and home literacy experiences obtained on a majority ofparticipants (85%) revealed that the samples were similar to each other onthese variables. English was the primary language spoken in the home forall children, and fewer than three mothers or three fathers reported thatEnglish was not their native language. Mothers and fathers of children inboth samples had completed on average 16 years of education (in bothsamples over 70% of mothers and over 63% of fathers were collegegraduates). Parents reported a significant number of children's books in thehome for both the younger (M = 89.47, SD = 67.5) and older (M =137.75, SD = 92.9) samples. Children in both samples were reported to beread to frequently at home (younger sample, M = 6.69 times per week,SD = 3.22; older sample, M = 5.71, SD = 2.35), and shared reading hadstarted early for children in both the younger (M = 6.37 months,SD = 6.18) and older (M = 7.33 months, SD = 4.48) samples.

Preschool and Child-Care Centers

Although we did not conduct formal observations of the centers, ourmultiple opportunities for observation allowed us to identify salient fea-tures of their educational environments. There was significant variabilitybetween centers in terms of materials available to the children and activitystructure. Generally, the curriculum in the centers was designed to fostersocial and interpersonal skill growth and to introduce the children to avariety of educationally relevant concepts such as letters, numbers, andstorybooks. We never observed any explicit attempts to teach the childrento read in any center. Some of the centers had at least some letterknowledge instruction, but most of this was informal. A number of direc-tors commented that they discouraged explicit teaching. The centers hadsimilar daily activity schedules, including free play, storytime, and small-group arts and crafts projects. Each of the centers incorporated someteacher-directed classroom activities (typically arts and crafts); however,the majority of children's time was spent in self-directed activities in andout of the classroom.

Procedures and Measures

After parents provided informed consent for their children to participate,trained research assistants tested children individually in their centers. Testadministration for individual children was conducted over two to four

sessions within a 2-3-week period to ensure optimal performance on alltasks. Children in the younger sample completed four standardized tests oforal language, four tests of phonological sensitivity, and two tests ofnonverbal cognitive ability during Time 1 testing, and they completed fourtests of phonological sensitivity, two tests of letter knowledge, an envi-ronmental print task, and a print concepts task during Time 2 testing.Children in the older sample completed one test of oral language, four testsof phonological sensitivity, two tests of letter knowledge, an environmentalprint task, and a print concepts task during Time 1 testing, and theycompleted four tests of phonological sensitivity, two tests of letter knowl-edge, a print concepts task, and two text decoding tasks during Time 2testing.

Phonological sensitivity measures. Each of the four phonological sen-sitivity tasks was preceded by practice trials to teach children the task (e.g.,blending or deleting word sounds). For all tasks, corrective feedback wasgiven during the practice trials, but no feedback was given during the testtrials. Many items on the phonological sensitivity tasks used pictures toreduce memory demands on the children. Except as noted below, tasksadministered at Time 1 and Time 2 and to younger and older childrenincluded the same items. Previous analyses of these four tasks indicatedthat they had moderate to high internal consistencies for 4-year-olds (as =.47 to .96) and 5-year-olds (as = .69 to .94) but lower internal consisten-cies for 2- and 3-year-olds (see Lonigan et al., 1998).

A rhyme oddity detection task and an alliteration oddity detection task,patterned after the tasks developed by MacLean et al. (1987) and usingtheir word lists, required children to demonstrate awareness of rhyme orawareness of singleton word onsets. In both tasks, children were presentedwith three pictured words (e.g., boat, sail, nail; car, cat, sun), which werenamed by the examiner, and were asked to select the one that did not rhymewith (or that sounded different from) or did not sound the same at thebeginning of the word as (or that sounded different at the beginning of theword from) the other two words. Two practice trials and 11 test trials werepresented to all children.

A blending task required children to combine word elements to form aword. Three practice items and the first eight test trials were presented bothverbally and with pictures; the remaining test trials were presented verballyonly. In both picture and nonpicture trials, the first five items requiredblending single-syllable words to form compound words, and the remain-ing items required blending syllables or phonemes. For picture itemsinvolving compound words, the examiner showed the child two pictures,named them, and then asked the child what word would be produced if heor she said them together (e.g., "What do you get when you say cow . . .boy together?"). All practice items required the blending of compoundwords, and during the practice the examiner emphasized the nature of thetask by putting the pictures together. For the Time 1 assessment of the oldersample and both Time 1 and Time 2 assessments of the younger sample,there were 18 test trials, consisting of 10 word-blending items, 4 syllable-blending items, and 4 phoneme-blending items. At the Time 2 assessmentof the older children, there were 37 test trials, consisting of all of theTime 1 items followed by 3 additional syllable-blending items and 16additional phoneme-blending items. These additional syllable and pho-neme items were included in the older children's Time 2 assessment toreduce the chances of children's scoring at ceiling levels. During bothassessments, testing was discontinued after a child missed 5 consecutivetrials.

An elision task required children to say a word minus a specific sound.Two practice items and the first eight test trials were presented both

1 These children represent a subset of the children from middle-incomefamilies included in Lonigan et al. (1998). The results reported previouslyconcerned age- and SES-related performance differences in phonologicalsensitivity tasks from the children's initial assessment (i.e., Time 1 in thepresent study).


verbally and with pictures; the remaining test trials were presented verballyonly. In both picture and nonpicture trials, the first four items requireddeleting a single-syllable word from a compound word to form a new word.Subsequent items in both picture and nonpicture trials required deletion ofa syllable or a phoneme from a word to form a new word. For picture itemsinvolving compound words, the examiner showed the child two pictures,named them (e.g., "This is a bat, and this is a man."), asked the child to saythe compound (i.e., "batman"), and then asked the child to delete part of it.During both practice trials, which used compound words, the examineremphasized the nature of the task by removing the picture of the word tobe deleted. For the Time 1 assessment of the older children and bothTime 1 and Time 2 assessments of the younger children, there were 17 testtrials, consisting of 10 word-level items, 4 syllable-level items, and 3phoneme-level items. At the Time 2 assessment of the older children, therewere 34 test trials, consisting of all of the Time 1 items followed by anadditional 17 phoneme-level items. These additional phoneme items wereincluded in the older children's Time 2 assessment to reduce the chancesof children's scoring at ceiling levels. During both assessments, testing wasdiscontinued after a child missed 5 consecutive trials.

Oral language and cognitive ability measures. At Time 1, children inthe younger sample completed four standardized tests of oral language.Receptive vocabulary was assessed with the Peabody Picture VocabularyTests—Revised (PPVT-R; Dunn & Dunn, 1981). Expressive vocabularywas assessed with the Expressive One-Word Picture Vocabulary Test—Revised (EOWPVT-R; Gardner, 1990). The Verbal Expression subtest ofthe Illinois Test of Psycholinguistic Abilities (ITPA-VE; Kirk, McCarthy,& Kirk, 1968) was used to assess children's descriptive use of language,and the Grammatical Closure subtest of the Illinois Test of Psycholinguis-tic Abilities (ITPA-GC; Kirk et al., 1968) was used to assess children'sexpressive grammar. At Time 1 for the older children, oral language wasassessed with the ITPA-GC. In addition to these oral language measures,children in the younger sample completed the Picture Completion andObject Assembly subtests from the Wechsler Preschool and Primary Scalesof Intelligence—Revised (Wechsler, 1989) at Time 1.

Letter knowledge measures. For the Time 2 assessment of the youngersample and both Time 1 and Time 2 assessments of the older sample, twotasks assessed different aspects of letter knowledge. A letter-name knowl-edge task required children to name all 26 uppercase letters that werepresented individually in random order on individual 3 X 5 in. index cards.A letter-sound knowledge task required children to name the sound madeby each letter when it appeared in a word. All 26 uppercase letters werepresented individually in random order on individual 3 X 5 in. index cards.If children responded with the letter name or a word that started with theletter (e.g., "dog" for D), they were prompted to provide the letter sound;however, credit for a correct response was given if children provided thelong vowel sound for vowels.

Environmental print measures. The older sample at Time 1 and theyounger sample at Time 2 completed an environmental print task. On thistask, children were shown 11 pictures of print in environmental context(e.g., a stop sign, a Coke machine, a McDonald's sign) and were askedwhat each said. Children were also shown the same print as printed text outof context and were asked what it said.

Print concepts measure. At Time 2 for the younger sample and at bothTime 1 and Time 2 for the older sample, portions of Clay's (1979a) CAPtest (the"Sand" task) were used to assess the children's print knowledge.Items in this test require children to demonstrate understanding of theleft-to-right and top-to-bottom direction of print in a book, the sequenceand direction in which print progresses from front to back across pages, thedifference between the covers and the pages of a book, the differencebetween pictures and print on a page, and the meaning of elements ofpunctuation, including spaces between words and periods at the ends ofsentences.

Word decoding measures. At Time 2, children in the older samplecompleted the Word Identification subtest of the Woodcock Reading

Mastery Test—Revised (Woodcock, 1987) and a task requiring them todecode 25 frequent words printed individually on 3 X 5 in. index cards.

Results

Descriptive Statistics and Preliminary Analyses

Separate scores for word, syllable, and phoneme items on theblending and elision tasks were computed. Descriptive statisticsfor raw scores on all variables for the younger sample at bothTime 1 and Time 2 are shown in Table 1. Descriptive statistics forraw scores on all variables for the older sample at both Time 1 andTime 2 are shown in Table 2. The tables also list the internalconsistency reliabilities for the eight phonological sensitivityscores at both assessments; with a few exceptions, these reliabili-ties were at least moderate. Analyses of variance (ANOVAs)revealed that the older sample scored substantially higher than theyounger sample on the phonological sensitivity measures atTime 1 (all ps < .001). For all tasks that were the same betweenTime 1 and Time 2 assessments (i.e., all phonological sensitivitytasks for the younger sample and the letter knowledge and word-level phonological sensitivity tasks for the older sample), within-subject ANOVAs revealed that there was significant growth fromTime 1 to Time 2 (all ps < .001).

Standardized scores for both samples were computed by re-gressing chronological age onto the raw score for each variablewithin sample and time of assessment (i.e., Time 1 and Time 2) toremove statistically the reliable variance that was due to children'schronological age from the scores on the observed variables.Inspection of the distribution of scores for each variable revealedsome moderate departures from normality (i.e., skew) but noobvious outliers. Further inspection revealed that positive skew forthe younger children's Time 1 scores was due to a moderatenumber of children scoring at low levels on the blending andelision tasks, whereas negative skew for the older children's scoreswas due to a moderate number of children scoring at high levels onthe word blending, word elision, and letter knowledge tasks. Al-though these distributions accurately reflect task difficulty for theage ranges included in the samples, the use of nonnormal data mayattenuate relations among variables and compromise model fits;consequently, we conducted confirmatory factor analysis (CFA)using robust maximum-likelihood estimation, the Satorra-Bentlerscaled chi-square (S-B;^2), and adjustments to the standard errorsto account for nonnormality in model fit statistics and significancetesting (Bentler & Dudgeon, 1996).

Evaluation of Measurement Models

We conducted separate CFAs using EQS (Bentler, 1995) toevaluate measurement models for both samples at Time 1 andTime 2. All CFAs were conducted on covariance matrices. Prior toevaluating the adequacy of measurement models that included allemergent literacy tasks, we evaluated the adequacy of a one-factormodel to explain scores on the phonological sensitivity tasks ateach measurement period for both samples. In all models, taskvariance for the different phonological sensitivity tasks was mod-eled by allowing correlated residuals between similar tasks (i.e.,parameter estimates for covariances between error terms for thetwo oddity tasks, three blending tasks, and three elision tasks were


Table 1Descriptive Statistics for Younger Sample of Children at Time 1 and Time 2 Assessments

Variable

Age (in months)Rhyme oddityAlliteration oddityBlending wordsBlending syllablesBlending phonemesElision wordsElision syllablesElision phonemesPPVT-R (MA)EOWPVT-R (MA)ITPA-VE (MA)ITPA-GC (MA)WPSSI Object AssemblyWPSSI Picture CompletionLetter namesLetter soundsConcepts About Print TestEnvironmental print: picturesEnvironmental print: text

M

41.054.543.452.570.680.251.770.430.22

42.5442.7148.5445.4812.4010.48

—————

Time 1

SD

9.362.001.814.471.350.622.770.950.70

11.6012.9013.0914.875.316.77—————

a

.30

.18

.97

.90

.52

.91

.79

.86

M

57.566.935.557.301.511.315.731.951.12——————

14.516.847.265.220.97

Time 2

SD

10.092.522.665.881.341.483.831.591.20

—————

10.068.323.322.692.13

a

.90

.85

.98

.89

.87

.96

.88

.85

Note. N = 96. All means are for raw scores unless otherwise noted. Internal consistency reliabilities (alphas)are provided only for phonological sensitivity measures. Dashes indicate tasks not administered at an assessmentperiod. PPVT-R = Peabody Picture Vocabulary Test—Revised; MA = mental age score; EOWPVT-R =Expressive One-Word Picture Vocabulary Test—Revised; ITPA-VE = Verbal Expression subscale of theIllinois Test of Psycholinguistic Abilities; ITPA-GC = Grammatical Closure subtest of the Illinois Test ofPsycholinguistic Abilities; WPPSI = Wechsler Preschool and Primary Scales of Intelligence.

specified in the models).2 For the younger sample at Time 1,S-B^CB, N = 96) = 13.15, p > .25, RCF1 (robust comparativefit index) = 1.00, and Time 2, S-Bx^B.iV = 96) = 5.23,p > .25,RCFI = 1.00, and for the older sample at Time 1, S-Bx^B, N =97) = 17.89, p > .10, RCFI = .98, and Time 2, S - B ^ D , N =97) = 10.78, p > .25, RCFI = 1.00, a one-factor model providedan excellent fit to the data. Following these analyses, differentone-, two-, and three-factor measurement models that included allemergent literacy tasks were compared in the younger and oldersamples at the Time 1 and Time 2 assessments.

Younger sample. Fit indices for the different measurementmodels for the younger sample of children are shown in Table 3.For the Time 1 assessment (upper half of Table 3), the fits ofmodels that included different combinations of phonological sen-sitivity, oral language, and nonverbal IQ measures were compared.A three-factor model with separate Phonological Sensitivity, OralLanguage, and Nonverbal IQ factors provided a significantly betterfit than all of the alternative models (all ps < .01 for chi-squaredifference tests) except the model with the phonological sensitivityand oral language measures represented by one factor. The differ-ence (diff) between the three-factor model and this two-factormodel was only marginally significant, ^ iff<2, N = 96) = 4.09,p = .11; however, examination of the other fit indices (Bentler &Bonett, 1980; see Table 3) and factor loadings, which indicatedthat the majority of phonological sensitivity tasks did not loadsignificantly on the factor, supported the superiority of the three-factor model.

For the younger sample's Time 2 assessment (see lower half ofTable 3), the fits of models that included different combinations of

phonological sensitivity, letter knowledge, and environmentalprint measures were compared. These models also included theCAP test as a separate measured variable. A three-factor modelthat included different Phonological Sensitivity, Letter Knowl-edge, and Environmental Print factors provided a significantlybetter fit than the one-factor model, XdiffC5- N - 96) = 49.34, p <.001, a two-factor model with phonological sensitivity and letterknowledge measures represented by a single factor, Xdiff(3, N =96) = 42.07, p < .001, and a two-factor model with phonologicalsensitivity and environmental print measures represented by asingle factor, *jjiff(3, N = 96) = 39.12, p < .001. The two-factormodel with letter knowledge and environmental print measuresrepresented by a single factor was not significantly different fromthe three-factor model (p > .10); however, when the CAP measurewas excluded from the model, the three-factor model provided abetter fit to the data, xjiiff(2, N = 96) = 6.40, p < .05, supportingthe use of three separate factors to represent phonological sensi-tivity, letter knowledge, and environmental print.

Older sample. Fit indices for the different measurement mod-els for the older sample of children are shown in Table 4. For theTime 1 assessment (see upper half of Table 4), the fits of models

2 The structure of all measurement and longitudinal models was identicalwhether or not these correlated residuals were included in the models;however, model fits were improved when correlated residuals were in-cluded because they accounted for significant covariance between itemsthat was due to similar task methods (e.g., blending vs. deleting wordsounds) or other sources of systematic variance.


Table 2Descriptive Statistics for Older Sample of Children at Time 1 and Time 2 Assessments

Variable

Age (in months)Rhyme oddityAlliteration oddityBlending wordsBlending syllablesBlending phonemesElision wordsElision syllablesElision phonemesLetter namesLetter soundsConcepts About Print TestEnvironmental print: picturesEnvironmental print: textITPA-GC (MA)Decoding frequent wordsWRM Word ID

M

60.046.495.467.732.701.785.592.131.14

20.029.097.635.730.97

68.64——

Time 1

SD

5.412.752.642.951.391.402.371.431.067.378.913.322.171.91

16.22——

a

.71

.68

.93

.69

.67

.80

.70

.57

M

72.888.898.739.44

10.166.967.562.326.42

24.7220.4511.41——

11.9814.32

Time 2

SD

5.712.132.421.162.703.650.690.973.873.686.681.70—

8.4612.12

a

.71

.80

.75

.61

.90

.50

.44

.88

Note. N = 97. All means are for raw scores unless otherwise noted. Internal consistency reliabilities (alphas)are provided only for phonological sensitivity measures. Dashes indicate tasks not administered at an assessmentperiod. ITPA-GC = Grammatical Closure subtest of the Illinois Test of Psycholinguistic Abilities; MA = mentalage score; WRM Word ID = Word Identification subtest of the Woodcock Reading Mastery Test—Revised.

that included different combinations of phonological sensitiv-ity, letter knowledge, and environmental print measures werecompared. These models also included the CAP test as a sep-arate measured variable. Both chi-square difference tests andevaluation of the other fit indices indicated that a three-factormodel that included separate Phonological Sensitivity, LetterKnowledge, and Environmental Print factors provided a signif-icantly better fit than the one-factor model, ^ i f f ( 5 , N =97) = 32.49, p < .001, a two-factor model with phonological

sensitivity and letter knowledge measures represented by asingle factor, )&i{f(3, N = 97) = 10.89, p < .05, a two-factormodel with phonological sensitivity and environmental printmeasures represented by a single factor, Xdiff(3, N =97) = 24.39, p < .001, and a two-factor model with letterknowledge and environmental print measures represented by asingle factor, Xdiff(3, N = 97) = 18.91, p < .001, supporting theuse of three separate factors to represent phonological sensitiv-ity, letter knowledge, and environmental print.

Table 3Fit Indices for Measurement Models for Younger Sample at Time 1 and Time 2 Assessments

Model (and factors)

1-factor (PS + OL + IQ)2-factor (PS + OL, IQ)2-factor (PS + IQ, OL)2-factor (PS, OL + IQ)3-factor (PS, OL, IQ)

1-factor (PS + LK + EP)2-factor (PS + LK, EP)2-factor (PS + EP, LK)2-factor (PS, LK + EP)3-factor (PS, LK, EP)

S-B*2

102.70**89.82*

130.77***99.43**85.73

123.05***115.78***112.83***80.45*73.71*

df

Time 1

7069696967

Time 2

5856565653

CFI

Assessment

.89

.91

.86

.93

.94

Assessment

.88

.89

.90

.96

.96

RCFI

.91

.94

.82

.91

.95

.88

.89

.90

.96

.96

TLI

.85

.88

.82

.90

.92

.84

.84

.85

.94

.94

RMSEA

.10

.09

.11

.08

.07

.11

.11

.11

.07

.07

AIC

-9.57-18.20

3.74-29.98-36.14

7.445.252.51

-32.85-32.26

Note. All models include correlated residuals between like phonological sensitivity tasks. All models of Time 2assessment include scores on the Concepts About Print Test as a measured variable. N = 96. PS = PhonologicalSensitivity; OL = Oral Language; IQ = Nonverbal IQ; LK = Letter Knowledge; EP = Environmental Print;S-B^2 = Satorra-Bentler chi-square; CFI = comparative fit index; RCFI = robust comparative fit index; TLI =Tucker-Lewis index; RMSEA = root mean square error of approximation; AIC = Akaike information criterion.*p<.05. **p<.0l. * * * / > < . 0 0 1 .


Table 4Fit Indices for Measurement Models for Older Sample at Time 1 and Time 2 Assessments

Model (and factors)

1-factor2-factor (PS + LK, EP)2-factor (PS + EP, LK)2-factor (PS, LK + EP)3-factor (PS, LK, EP)

1-factor (PS + LK + RD)2-factor (PS + LK, RD)2-factor (PS + RD, LK)2-factor (PS, LK + RD)3-factor (PS, LK, RD)

S-B*2

117.15***95.55***

109.05***103.57***84.66**

103.00***69.0879.18*77.09*51.50

df

Time 1

5856565653

Time 2

5856565653

CFI

Assessment

.87

.91

.88

.89

.93

Assessment

.90

.96

.95

.951.00

RCFI

.86

.91

.87

.89

.92

.90

.97

.95

.951.00

TLI

.82

.88

.84

.85

.90

.86

.95

.93

.921.00

RMSEA

.10

.09

.10

.09

.08

.10

.06

.07

.07

.02

AIC

-0.34-18.16

-4.68-10.24-22.76

-6.59-36.89-30.50-28.19-51.38

Note. All models include scores on the Concepts About Print Test as a measured variable and correlatedresiduals between like phonological sensitivity tasks. N = 97. PS = Phonological Sensitivity; LK = LetterKnowledge; EP = Environmental Print; RD = Word Reading (decoding); S-B^2 = Satorra-Bentler chi-square;CFI = comparative fit index; RCFI = robust comparative fit index; TLI = Tucker-Lewis index; RMSEA = rootmean square error of approximation; AIC = Akaike information criterion.* p < . 0 5 . **p<. 01. ***/><.001.

For the older sample's Time 2 assessment (see lower half ofTable 4), the fits of models that included different combinations ofphonological sensitivity, letter knowledge, and text decoding werecompared. These models also included the CAP test as a separatemeasured variable. Both chi-square difference tests and evaluationof the other fit indices indicated that a three-factor model thatincluded different Phonological Sensitivity, Letter Knowledge,and Reading (Decoding) factors provided a significantly better fitthan the one-factor model, ^ i f f (5 , N = 97) = 51.50, p < .001, atwo-factor model with phonological sensitivity and letter knowl-edge measures represented by a single factor, Xaiff(3, N =97) = 17.58, p < .001, a two-factor model with phonologicalsensitivity and decoding measures represented by a single factor,x3iff(3, N = 97) = 27.68, p < .001, and a two-factor model withletter knowledge and decoding measures represented by a singlefactor, ^ i f f (3 , N = 97) = 25.59, p < .001, supporting the use ofthree separate factors to represent phonological sensitivity, letterknowledge, and text decoding.

Sample comparisons. To facilitate comparisons betweenyounger and older samples and to allow preliminary hypothesesconcerning the development of reading-related skills across theage range covered by both samples (i.e., continuity between theyounger sample's Time 1 assessment and the older sample'sTime 2 assessment), we compared raw scores and measurementmodels for the emergent literacy measures from the youngerchildren at the Time 2 assessment with those from the olderchildren at the Time 1 assessment. ANOVA revealed that childrenin the younger sample at Time 2 were somewhat younger thanchildren in the older sample at Time 1, F(l, 191) = 4.54, p = .03.ANOVAs on children's raw scores also revealed that children inthe younger sample at Time 2 scored lower on letter knowledge,F(l, 191) = 18.86, p < .001, syllable blending, F(l, 191) = 36.87,p < .001, and phoneme blending, F(l, 191) = 5.18, p = .02, thandid children in the older sample at Time 1 (see Table 1 and Table 2for descriptive statistics). Differences on phoneme blending were

rendered nonsignificant in an analysis of covariance controlling forchronological age (p = .16); however, the differences for letterknowledge and syllable blending remained significant (ps <.001).3

Multisample CFA was carried out on the data from the youngerchildren's Time 2 data and the older children's Time 1 data toexamine structural invariance of the three-factor measurementmodel across samples (see Table 5). A multisample model withseparate Phonological Sensitivity, Letter Knowledge, and Envi-ronmental Print factors, with the CAP test as a separate measuredvariable and with none of the parameters across groups constrainedto equality, served as a basis for testing whether adding constraintsto the model across groups would yield a significantly worse fit.

A significant change in the chi-square when factor loadingswere constrained across groups suggested there was a statisticallysignificant lack of invariance. However, fit indices that are morerobust to sample size supported the invariance of factor loadings,factor correlations (including correlations with the CAP measure),and correlated residuals. The comparative fit index (CFI), Tucker-Lewis Index (TLI), root mean square error of approximation(RMSEA), and Akaike information criterion (AIC) remained es-sentially unchanged when these invariance constraints were im-posed, and the imposition of all of these constraints did not resultin a significant reduction in the overall model chi-square from theunconstrained model, ^ i f f(25, N = 193) = 32.79, p > .10.Consequently, the majority of fit indices indicated that the slightlack of invariance noted for the factor loadings was of little

3 These significant differences may have been the result of the age rangeof the younger group. That is, the youngest child in the older samplewas 60 months old, whereas the youngest child in the younger samplewas 38 months old. Alternatively, these differences may have been afunction of the different preschool environments of the younger and oldersamples.


Table 5Fit Indices for Multisample Analysis of Three-Factor Measurement Model for Younger Sampleat Time 2 and Older Sample at Time 1

Model constraints

None (unconstrained)Factor loadingsFactor loadings and factor

correlationsFactor loadings, factor

correlations, and correlatedresiduals

Factor loadings, factorcorrelations, correlatedresiduals, and residualvariances

x2

156.98***176.89***

185.34***

189.77***

221.87***

df

106118

124

131

143

CFI

.95

.94

.94

.94

.92

TLI

.92

.92

.92

.93

.91

RMSEA

.05

.05

.05

.05

.05

AIC

-55.02-59.11

-62.66

-72.23

-64.13

Xdiff

19.91*

8.45

4.43

32.10**

df

10

6

7

12

Note. All models include scores on the Concepts About Print Test as a measured variable and correlatedresiduals between like phonological sensitivity tasks. Chi-square difference tests reflect comparison of a modelwith the previous model and thus reflect the change associated with the addition of the specified constraint. N =193. CFI = comparative fit index; TLI = Tucker-Lewis index; RMSEA = root mean square error ofapproximation; AIC = Akaike information criterion.* p < . 0 5 . ** p < . 01. ***/?<.001.

practical importance and was outweighed by the large gain inparsimony. In contrast, all fit indices decreased substantially whenitem residuals were constrained. Thus, the results indicated that themeasurement model explained children's emergent literacy skillswell across both the younger and older samples of children (i.e.,the factor structure was equivalent) but that there may have beensystematic sample differences in measurement errors that were oflittle substantive interest to the present study.

Longitudinal Prediction Models

Structural equation modeling in EQS was used to examine thelongitudinal relations between emergent literacy and either lateremergent literacy skills (younger sample) or both later emergentliteracy skills and text decoding (older sample). The measurementmodels identified in the previous analyses served as the basis forthe longitudinal models. We first calculated the cross-time zero-order correlations between latent constructs. Because our interestwas in identifying significant sources of influence on children'sdevelopment, we began by examining models that included au-toregressive paths (i.e., paths between the same factors at differenttime points). Inclusion of other paths was guided by results fromanalyses of zero-order correlations as well as theoretical consid-erations. Modifications to these base models were made by exam-ining the results of both (a) Lagrange multiplier (LM) tests, todetermine the value of adding parameters to the models that wouldsignificantly increase the model fit at the p < .05 level, and (b)Wald tests, to determine the statistical necessity of parameterswhose elimination would not significantly decrease the model fit atthe/? > .10 level.

Younger sample. Zero-order correlations between the latentvariables at Time 1 and the latent variables at Time 2 for theyounger sample of children are shown in the upper half of Table 6.For the younger sample of children, the base longitudinal predic-tion model included paths from the Phonological Sensitivity factorat Time 1 to the Phonological Sensitivity and Letter Knowledgefactors at Time 2. Paths from the Time 1 Oral Language factor to

the Time 2 Phonological Sensitivity factor, the Time 2 LetterKnowledge factor, the Time 2 Environmental Print factor, and theTime 2 CAP variable also were included in the model. Finally, apath from the Nonverbal IQ factor to the Time 2 CAP variable wasincluded.

On the basis of Wald tests, we dropped the paths from theTime 1 Phonological Sensitivity factor to the Time 2 PhonologicalSensitivity factor, from the Time 1 Oral Language factor to theTime 2 Environmental Print factor, and from the Time 1 NoverbalIQ factor to the Time 2 CAP variable. On the basis of LM tests, weadded paths from the Time 2 Letter Knowledge factor to theTime 2 Phonological Sensitivity factor, the Time 2 EnvironmentalPrint factor, and the Time 2 CAP variable. The resultant model forthe younger sample is shown in Figure 1, S-B^2(305, N = 96) =383.05, p < .01, RCFI = .93, RMSEA = .06. Time 2 Phonolog-ical Sensitivity was significantly predicted by Oral Language atTime 1 and Letter Knowledge at Time 2 (R2 = .25). LetterKnowledge was significantly predicted by both Time 1 Phonolog-ical Sensitivity and Time 1 Oral Language (R2 = .20). Environ-mental Print was significantly predicted by Time 2 Letter Knowl-edge only (R2 = .45). Finally, scores on the CAP measure atTime 2 were significantly predicted by both Time 1 Oral Languageand Time 2 Letter Knowledge (R2 = .23).

Because the absence of significant cross-time stability in thePhonological Sensitivity factor suggested a problem with the mea-surement of phonological sensitivity at Time 1 for the youngersample, we examined a model with a Time 1 Phonological Sen-sitivity factor that included only those tasks with significant cross-time stability (syllable blending and all three elision measures).Evaluation of the measurement model supported three separatefactors to represent Phonological Sensitivity, Oral Language, andNonverbal Cognitive Abilities, S-B^2 (30, N = 96) = 28.83, p =.53, RCFI = 1.00. There was a significant cross-time correlationbetween this reduced Phonological Sensitivity factor and theTime 2 Phonological Sensitivity factor (r = .35, p < .01). Weexamined the full longitudinal model starting with the same base


Table 6Zero-Order Correlations Between Time I Emergent Literacy Skills and Time 2 EmergentLiteracy and Reading Skills for Younger and Older Samples

Time 1 variables

Phonological sensitivityOral languageNonverbal cognitive

Phonological sensitivityEnvironmental printLetter knowledgeConcepts About Print Test

Phonologicalsensitivity

.14

.36***

.16

1.00***.59***.64***.60***

Time

Letterknowledge

Younger sample

.33**

.39***

.15

Older sample

.48***

.42***

.80***

.35***

2 variables

Environmentalprint

.23

.33**

.19

Reading.60***.51**.51***.40***

Concepts AboutPrint

.1437***

.32*

44***.18.37**.62***

Note. Correlations are between latent variables for each construct, except for the Concepts About Print Test,which is a measured variable.* p < . 0 5 . ** />< .01 . ***/?<.001.

model described previously. The resultant final model is shown inFigure 2, S-B^(212, N = 96) = 265.74, p < .01, RCFI = .94,RMSEA = .06. In this modified model, the reduced Time 1Phonological Sensitivity factor was significantly related to OralLanguage. Time 2 Phonological Sensitivity was significantly pre-dicted by both Phonological Sensitivity and Oral Language atTime 1 (R2 = .17). Letter Knowledge was significantly predictedby both Time 2 Phonological Sensitivity and Time 1 Oral Lan-guage (R2 = .26). Environmental Print was significantly predictedby Time 2 Letter Knowledge only (R2 = .49). Finally, scores onthe CAP measure at Time 2 were significantly predicted byboth Time 1 Oral Language and Time 2 Phonological Sensitivity(R2 = .20).

Older sample. Zero-order correlations between the latent vari-ables at Time 1 and the latent variables at Time 2 for the oldersample of children are shown in the lower half of Table 6. For theolder sample of children, the base longitudinal prediction modelincluded paths from the Time 2 Phonological Sensitivity andLetter Knowledge factors to the Reading factor, from the Time 1Phonological Sensitivity factor to the Time 2 Phonological Sensi-tivity and the Time 2 Letter Knowledge factors, from the Time 1Letter Knowledge factor to the Time 2 Letter Knowledge andPhonological Sensitivity factors, and from the Time 1 CAP mea-sure to the Time 2 CAP measure.4

The paths between the Time 1 Phonological Sensitivity factorand the Time 2 Letter Knowledge factor and between the Time 1Letter Knowledge factor and the Time 2 Phonological Sensitivityfactor were dropped on the basis of Wald tests. On the basis of LMtests, a path between the Time 1 Phonological Sensitivity factorand the Time 2 CAP variable was added. The resultant model forthe older sample is shown in Figure 3, S-B^2(276, N = 91) =428.18, p < .001, RCFT = .87, RMSEA = .08. Time 2 Phono-logical Sensitivity was perfectly predicted by Phonological Sensi-tivity at Time 1 (R2 = 1.00). Time 2 Letter Knowledge waspredicted by Time 1 Letter Knowledge only (R2 = .72). Scores onthe CAP measure at Time 2 were predicted by scores on the

Time 1 CAP measure and Time 1 Phonological Sensitivity (R2 =.44). Finally, Time 2 Phonological Sensitivity and Time 2 LetterKnowledge were the only significant predictors of reading (R2 =.54).56 As would be expected from the results of the Wald and LMtests, when paths between the Time 1 Environmental Print factor,the Time 2 CAP measure, and the Reading factor were included inthis model, these paths were not significant, indicating that neitherthe Environmental Print factor nor the CAP measure added uniquevariance to the Reading factor once the Phonological Sensitivityand Letter Knowledge factors were in the model.

To confirm these findings and to ensure that our model devel-opment strategy was not biased against finding significant effectson reading for the environmental print and print concepts mea-sures, we conducted model testing starting with a base model thatincluded just the autoregressive paths and paths from both theTime 1 Environmental Print factor and the Time 2 CAP variable tothe Time 2 Reading factor. The resultant final model followingWald and LM tests was the model shown in Figure 3, with theexception that the path between Time 2 Letter Knowledge and

4 The identical final model was obtained when the base model includedpaths for all of the Time 1 variables with significant zero-order associationswith Time 2 variables.

5 As would be expected given the high correlations between the Time 1and the Time 2 Phonological Sensitivity and Letter Knowledge factors,when the Time 1 Phonological Sensitivity and Letter Knowledge factorswere used to predict the Reading Factor at Time 2, they also accounted for54% of the variance in decoding.

6 We also tested the unique variance associated with each of the blendingand elision measures by adding (in separate sequential models) a path fromeach measure's residual to the Reading factor. In none of these models wasthere an improvement in model fit or increment in the R2 for the Readingfactor. These results suggest that, in these data, it was the variance commonto the eight phonological sensitivity measures that was predictive ofdecoding rather than the variance unique to the manipulation of phonemes,syllables, or words.


-.61.48

Figure 1. Structural equation model of longitudinal relations between emergent literacy abilities for youngersample of children. Circles represent latent variables, and rectangles represent observed variables. Variables onthe left of the figure are from the Time 1 (Tl) assessment (mean age = 41.1 months, SD = 9.4); variables onthe right of the figure represent Time 2 (T2; mean age = 57.6 months, SD = 10.1), reflecting development overan 18-month period. All paths shown as solid lines are significant at p < .05. Wrd = word-level items, Syl =syllable-level items, Phon = phoneme-level items, Ltr = letter; Env = environmental; Pics = pictures; CAP =Concepts About Print Test; PPVT-R = Peabody Picture Vocabulary Tests—Revised; EOWPVT-R = Expres-sive One-Word Picture Vocabulary Test—Revised; ITPA-VE = Verbal Expression subtest of the Illinois Testof Psycholinguistic Abilities; ITPA-GC = Grammatical Closure subtest of the ITPA; PicComplet and ObjAs-sem = Picture Completion and Object Assembly subtests of the Wechsler Preschool and Primary Scales ofIntelligence—Revised.

Reading was not included. An additional model examined theinfluence of ITPA-GC scores on the Time 2 factors. In this model,ITPA-GC scores were not a significant predictor of any Time 2factor and did not alter the significance of the paths shown inFigure 3. Finally, we also examined the independence of phono-logical sensitivity from oral language by regressing ITPA-GCscores from both reading measures. In this analysis, both thePhonological Sensitivity and Letter Knowledge factors continuedto be significant and substantial predictors of the Reading factor(R2 = .39).

Discussion

The results of this study demonstrate that the developmentalorigins of a large component of children's reading skills in kin-

dergarten and first grade can be found in the preschool period. Anumber of the emergent literacy skills present during the preschoolperiod (i.e., phonological sensitivity, letter knowledge) reflecthighly stable individual differences and have substantial uniquepredictive relations with later reading abilities. Together, phono-logical sensitivity and letter knowledge accounted for 54% of thevariance in kindergarten and first-grade children's decoding abil-ities. In contrast, other emergent literacy skills, such as environ-mental print and print concepts, although present during the pre-school period and relatively stable, do not appear to be uniquelyimportant for children's later reading. Taken together, these resultshighlight the developmental continuity between emergent literacyand later reading from the early preschool period to the earlyelementary school period. Additionally, these results provide im-


Blend Syl-.40

jElision W r d i * ^

-.16

-.39 " - ^Elision Syl

\ -.25'

Vision Phog

PicComplet« -,g

ObjAssem ..98.

CAP

2. Structural equation model of longitudinal relations between emergent literacy abilities for youngersample of children including modified Phonological Sensitivity factor for Time 1 assessment. Time 1 measuresincluded in the modified Phonological Sensitivity factor were those with significant cross-time stability. Circlesrepresent latent variables, and rectangles represent observed variables. Variables on the left of the figure are fromTime 1 (Tl) assessment (mean age = 41.1 months, SD = 9.4); variables on the right of the figure representTime 2 (T2; mean age = 57.6 months, SD = 10.1), reflecting development over an 18-month period. All pathsshown as solid lines are significant at p < .06. Wrd = word-level items, Syl = syllable-level items, Phon =phoneme-level items, Ltr = letter; Env = environmental; Pics = pictures; CAP = Concepts About Print Test;PPVT-R = Peabody Picture Vocabulary Tests—Revised; EOWPVT-R = Expressive One-Word PictureVocabulary Test—Revised; ITPA-VE = Verbal Expression subtest of the Illinois Test of PsycholinguisticAbilities; ITPA-GC = Grammatical Closure subtest of the ITPA; PicComplet and ObjAssem = PictureCompletion and Object Assembly subtests of the Wechsler Preschool and Primary Scales of Intelligence—Revised.

portant information concerning issues of the development and themeasurement of several key emergent literacy skills.

Perhaps the most striking finding from the present study con-cerned the high level of stability in children's phonological sensi-tivity. The latent variable representing the phonological sensitivityof 5-year-old children attending preschool perfectly predicted thelatent variable representing phonological sensitivity of 6-year-oldchildren attending kindergarten and first grade. These results in-dicate that there was no change in the ordering or spacing ofchildren's performance from preschool to kindergarten and firstgrade despite the fact that there was significant growth in theseskills (see Table 2). These findings are similar to those found with

older children by Wagner and colleagues (Wagner et al., 1994,1997). For example, Wagner et al. (1997) reported that year-to-year stability coefficients for their latent phonological sensitivityvariable ranged from .83 (kindergarten to first grade) to .95 (sec-ond grade to third grade and third grade to fourth grade). Ourresults indicate that this high degree of stability is present earlier indevelopment and is not the result of formal reading instruction.

In contrast to the extraordinary stability of phonological sensi-tivity from late preschool to early elementary school, phonologicalsensitivity was less stable from early preschool to late preschool.In fact, very early phonological sensitivity, represented by all eightmeasures of the construct we administered, was not a strong or


Rhyme26 \^—^Alliteration L .57

Figure 3. Structural equation model of longitudinal relations between emergent literacy abilities and readingfor older sample of children. Circles represent latent variables, and rectangles represent observed variables.Variables on the left of the figure are from Time 1 (Tl) assessment (mean age = 60.4 months, SD = 5.4);variables on the right of the figure represent Time 2 (T2; mean age = 72.9 months, SD = 5.7), reflectingdevelopment over a 13-month period. All paths shown as solid lines are significant atp < .05. Wrd = word-levelitems, Syl = syllable-level items, Phon = phoneme-level items, Ltr = letter; Env = environmental; Pic =pictures; CAP = Concepts About Print Test; ID = identification; Freq = frequent.

unique predictor of phonological sensitivity in the late preschoolperiod. There was some developmental continuity between thisearly phonological sensitivity construct and later phonologicalsensitivity; however, this continuity appeared to be mediated bylater letter knowledge, which was a significant concurrent predic-tor of phonological sensitivity. These results indicate that therewere problems with the measures of phonological sensitivity forthe early preschool group. That is, whatever variance was sharedacross all eight measures in the early preschool period was notphonological sensitivity. Based on the longitudinal empirical rela-tions of this factor, it is possible that the shared variance repre-sented letter knowledge or a proxy measure of print exposure.When we examined the longitudinal relations of a reduced Pho-nological Sensitivity factor that included only Time 1 measureswith significant cross-time stability, there was evidence for devel-opmental continuity of the Phonological Sensitivity factor from

early to late preschool. Interestingly, the variables that defined thisreduced factor were mainly those with weak relations to the factordefined by all eight measures, indicating that the variance sharedbetween these four variables, and hence the construct represented,was distinct from that included in the original factor.

Taken together, the results from these two models are similar tothe results from other studies of young children that have found apredictive relation between phonological sensitivity and later letterknowledge (Burgess & Lonigan, 1998; Wagner et al., 1994) andbetween letter knowledge and both current and subsequent pho-nological sensitivity (Bowey, 1994; Burgess & Lonigan, 1998;Johnston, Anderson, & Holligan, 1996; Stahl & Murray, 1994;Wagner et al., 1994, 1997). The mechanisms by which phonolog-ical sensitivity influences the development of letter knowledge andletter knowledge influences the development of phonological sen-sitivity are not clear. It is possible that the development of these


skills simply indexes exposure to literacy-related activities. Alter-natively, it is possible that children with greater sensitivity to thephonological structure of words and more letter knowledge maybenefit more from the formal and informal exposure to print thatmany preschoolers receive (e.g., Lonigan, 1994; Whitehurst &Lonigan, 1998). Perhaps the ability to discriminate word andsyllable boundaries makes the significance of letters more trans-parent. Similarly, understanding the significance of letters mayfacilitate the segmentation of language.

In addition to the effects of letter knowledge on phonologicalsensitivity, oral language had direct and indirect effects (dependingon the model) on phonological sensitivity in the late preschoolperiod. This finding is consistent with results from a number ofother studies of both preschool (e.g., Burgess & Lonigan, 1998;Chaney, 1992; Lonigan et al., 1998) and early elementary schoolchildren (e.g., Bowey, 1994; Wagner et al., 1993, 1997) that havedemonstrated significant concurrent and longitudinal correlationsbetween children's vocabulary skills and their phonological pro-cessing skills. These results suggest that oral language develop-ment has an influence on the acquisition of this key emergentliteracy skill. Past studies of preschool children have suggestedthat productive phonology (i.e., speech intelligibility) is related toperformance on phonological sensitivity tasks (e.g., Webster &Plante, 1995). As discussed by Metsala and Walley (1998; see alsoFowler, 1991), this evidence suggests that lexical representationsbecome more segmental in early childhood as a result of vocabu-lary growth. The emergence of phonological sensitivity may belimited by these speech representations.

Despite direct and indirect effects of early oral language andphonological sensitivity skills, all measured factors accounted foronly 17% to 25% of the variance in phonological sensitivitymeasured in the late preschool period. Although these resultsindicate that children's phonological sensitivity in the late pre-school period is partially a function of early phonological sensi-tivity, oral language skills, and letter knowledge, they highlight thefact that the origins of the majority of children's reading-relatedphonological sensitivity are unknown. Like the results of ourearlier cross-sectional study (Lonigan et al., 1998), these findingsindicate that significant growth in phonological sensitivity occursbetween 3 and 4 years of age. Consequently, efforts to identify theorigins of phonological sensitivity are likely to be most productiveduring this period. Our results also suggest, however, that screen-ing of children for phonological sensitivity deficits is unlikely tobe productive prior to the late preschool period, at least with thepresent measures because of their limited predictive power forlater phonological sensitivity.

The results of this study are also informative concerning thenature of preschool phonological sensitivity. As noted previously,phonemic sensitivity is often given special status in relation toreading, with a number of authors arguing that phonemic sensitiv-ity is the critical influence on reading skills (e.g., Morais, 1991;Muter et al., 1997; Nation & Hulme, 1997; Tunmer & Rohl, 1991).In contrast, we have argued elsewhere (Anthony & Lonigan, 2000;Anthony et al., 2000; Lonigan et al., 1998) that it is children'sgeneral sensitivity to the sound structure of language that is im-portant for learning to read an alphabetic system. Our finding thatchildren's phonological sensitivity, broadly defined (i.e., sensitiv-ity to words, syllables, onset-rime, and phonemes), was best char-acterized as a unitary construct at each of the four assessments of

children across different ages provides strong support for thisposition. Even in the reduced factor for the younger children'sTime 1 assessment, Phonological Sensitivity was represented bysensitivity to words, syllables, and phonemes. Across analysesfrom the late preschool and early grade school periods, only oneindex of phonological sensitivity did not have a significant asso-ciation with the phonological sensitivity construct. The word-blending measure did not contribute to the latent variable at theTime 2 assessment for the older group. This effect was likely dueto the fact that scores on the word-blending measure for the olderchildren were at near ceiling levels. Regardless, this same analysisdemonstrated that word-level and syllable-level blending wereassociated with manipulation of phonemes (i.e., alliteration, pho-neme blending, phoneme elision), which supports the broadlydefined phonological sensitivity construct.

Two additional aspects of our results support the importance ofthe broader construct of phonological sensitivity. First, whereasthe measures of phonological sensitivity for the younger group'sTime 1 and Time 2 assessments and for the older group's Time 1assessment were weighted heavily in favor of lower levels oflinguistic complexity (i.e., words, syllables, onset-rime), the mea-sures of phonological sensitivity for the older children's Time 2assessment were weighted heavily in favor of higher levels oflinguistic complexity (i.e., phonemes). The fact that the earlierPhonological Sensitivity factor perfectly predicted the later Pho-nological Sensitivity factor for the older group of children indi-cates that sensitivity to lower and higher levels of linguistic com-plexity represents a continuum rather than distinct abilities. Thesefindings are consistent with the results obtained by Stahl andMurray (1994), who found that a single-factor solution explaineda majority of the variance in kindergarten children's performanceon four tasks that varied by linguistic complexity.

Finally, the global construct of phonological sensitivity, definedby variance common to sensitivity to words, syllables, onset-rime,and phonemes, was a significant and strong predictor of children'sdecoding skills. This finding demonstrates that this global phono-logical sensitivity, rather than just phonemic sensitivity, is influ-ential in the development of children's decoding skills. Moreover,like other studies (e.g., Bryant et al., 1990; Lonigan et al., 1998;MacLean et al., 1987; Wagner et al., 1994, 1997), our analysesdemonstrated that this relation was not the result of varianceshared between the global construct of phonological sensitivityand oral language. That is, the predictive relation between theglobal construct of phonological sensitivity and reading is not theresult of children with more developed oral language skills, such asvocabulary, or general cognitive abilities simply having greaterfaculty with tasks assessing broad levels of phonological sensitiv-ity and also having better decoding skills. It is important to note,however, that our assessment of oral language skills in the oldersample was limited to a single measure. It is possible that otheroral language measures may have shared more predictive variancewith both decoding and phonological sensitivity. However, giventhe independence of these constructs demonstrated in the youngersample and the significant loading of the ITPA-GC on the broaderOral Language factor, it seems unlikely that additional oral lan-guage measures would have substantially weakened the strongrelation between phonological sensitivity and decoding.

Whereas a number of previous studies have interpreted findingsthat one measure of phonological sensitivity (e.g., phoneme seg-


mentation) predicts reading better than another (e.g., onset-rimesensitivity) to indicate that one type of phonological sensitivity ismore important to reading than another (Goswami & Bryant, 1990,1992; Muter et al., 1997; Nation & Hulme, 1997), these analysesmake the explicit or implicit assumption that there are differenttypes of phonological sensitivity. Our results, as well as the resultsof other large studies (e.g., Wagner et al., 1993, 1994, 1997),demonstrate that such assumptions are incorrect, at least as expla-nations of the normal development of reading. That is, our anal-yses of different tasks that varied in linguistic complexity, whichindicated that a single-factor solution provided an excellent fit tothe data, established that these tasks tap the same underlyingability, phonological sensitivity. Moreover, this single factor pre-dicted a majority of the variance in later decoding skills. Theseresults are consistent with those of Wagner and colleagues andStahl and Murray (1994) in demonstrating that phonological sen-sitivity is a unitary construct represented by sensitivity to onset-rimes, syllables, and phonemes and in showing that the variancecommon to children's abilities to perform tasks requiring sensitiv-ity to onset-rime, syllables, and phonemes is a substantial predictorof decoding skills.

Our results indicated that, like phonological sensitivity from latepreschool to early grade school, letter knowledge was a very stableindividual difference, and at every assessment, letter knowledgerepresented an emergent literacy skill that was independent ofphonological sensitivity, environmental print, and decoding. Letterknowledge in the late preschool period, indexed by knowledge ofboth letter names and letter sounds, predicted 72% of the variancein kindergarten and first-grade children's letter knowledge. More-over, this level of stability was likely attenuated because of thenear-ceiling performance of the older children on both the measureof letter name knowledge and the measure of letter sound knowl-edge at the Time 2 assessment (see Table 2).

Another significant finding of our study was that measures ofvariables that have been the focus of traditional emergent literacyapproaches (i.e., print concepts, environmental print) had nounique predictive relation to later reading skills or other lateremergent literacy skills. Some emergent literacy advocates haveargued that children's faculty with environmental print demon-strates their ability to derive the meaning of text within context(e.g., Goodman, 1986); however, other research has not generallysupported a direct causal link between the ability to read environ-mental print and later decoding skills (Gough, 1993; Masonhei-mer, Drum, & Ehri, 1984). Although these variables were associ-ated with later reading and later emergent literacy when consideredin isolation (see Table 6), they were not significant unique predic-tors in the context of letter knowledge and phonological sensitiv-ity. Concepts of print and environmental print might reflect veryearly knowledge of literacy; however, our analyses demonstratedthat measures of environmental print reflected a construct that wasdistinct from letter knowledge and phonological sensitivity. Thefact that both the environmental print variable and the CAP vari-able were predicted by phonological sensitivity and letter knowl-edge suggests that they may best be conceptualized as proxymeasures for these other emergent literacy skills, reflect moreexposure to print and other literacy-related activities (e.g., seePurcell-Gates, 1996), or both. Two limitations to the conclusionsthat can be made concerning print concepts in this study are thatwe had only a single indicator of the construct and that we did not

administer the measure to the younger children at Time 1. Conse-quently, we were unable to represent it as a latent variable, and wecould not estimate its influence on the development of otheremergent literacy skills from the early to the late preschool period.Although future studies should address these limitations, our find-ings indicate that what is measured by print concepts that isindependent of letter knowledge and phonological sensitivity isunrelated to early decoding abilities.

Our analyses revealed that both the measurement models andthe scores obtained by both groups of children during the latepreschool period were nearly identical. Consequently, these find-ings provide a preliminary means of examining the developmentalcontinuity of emergent literacy and early reading skills from earlypreschool to early grade school. This cross-sample analysis high-lights the significance of individual differences in both oral lan-guage and phonological sensitivity. That is, individual differencesin oral language skills, such as vocabulary, appear to be an im-portant influence on later emergent literacy skills that are crucialcomponents for children's development of decoding skills (i.e.,phonological sensitivity and letter knowledge). Individual differ-ences in phonological sensitivity measured at an early age alsoappear to have a significant influence on these key emergentliteracy skills that is independent of oral language abilities.

Despite these significant findings, a number of caveats concern-ing this study are required. Although the samples used in this studywere larger than those used in most prior studies of preschoolemergent literacy (e.g., Bryant et al., 1990; Chaney, 1992; Fox &Routh, 1975; Maclean et al., 1987), they were marginally adequatefor structural equation modeling. The broad age range in theyounger sample of children may have obscured potentially impor-tant relations between some emergent literacy variables. For in-stance, it may be that greater stability in phonological sensitivityemerges at an earlier age but was not apparent because of the agerange of our younger sample. Additionally, our reliance on differ-ent samples of children to explore the developmental continuity ofemergent literacy from early preschool to kindergarten and firstgrade was not optimal. Although results of multisample analysesindicated that scores on the measures and the measurement modelswere nearly identical across both groups during the late preschoolperiod, indicating that interpretations across samples were justi-fied, conclusions derived from the same sample would be stronger.Importantly, it is unlikely that our main findings concerning thesignificant relations between emergent literacy skills within andbetween assessment phases were the result of the age range ofchildren within the samples because all analyses were conductedusing scores from which the reliable variance associated withchildren's chronological age was statistically removed. However,these results provide a preliminary examination of how thesedifferent emergent literacy skills relate to each other from the earlypreschool period to kindergarten and first grade.

Not all domains of emergent literacy were measured in thisstudy (Whitehurst & Lonigan, 1998). For instance, some writershave suggested that the constructs of emergent reading or emer-gent writing reflect children's developing conceptualizations ofliteracy (e.g., Pappas & Brown, 1988; Purcell-Gates, 1988; Sulzby,1985, 1986, 1988). Although we believe that these skills are likelyto be related to concepts of print and understanding of narratives,and therefore either reflect dimensions similar to letter knowledgeand phonological sensitivity or relate more strongly to reading


comprehension rather than to decoding (Whitehurst & Lonigan,1998), future studies should address the relative independence andspecific influences of these emergent literacy skills. In addition tophonological sensitivity, components of phonological processing,such as phonological memory and phonological naming, have beenidentified in older children as significant correlates of readingskills (e.g., Bowers & Wolfe, 1993; Wagner et al., 1994, 1997;Wolfe, 1991). A complete account of emergent literacy will re-quire an understanding of the development of these skills and theirsignificance, if any, during the preschool years.

Although the results of this study highlight the developmentalcontinuities and discontinuities in emergent literacy and the sig-nificant linkage between emergent literacy skills and later decod-ing, they do not address the question of the origins of these skills.Given the significant linkages found in this study, future studiesshould address questions concerning the developmental origins ofkey skills such as phonological sensitivity and letter knowledge.Such information will expand our knowledge of emergent literacyand provide clues for the development of interventions designed tohelp children at risk for developing later reading difficulties.Finally, our results concern the development of emergent literacyand decoding in children from English-speaking and middle-classfamilies. Consequently, our results are most relevant to childrenlearning to read an alphabetic language, and the degree to whichthese findings translate to children who may be at risk for readingdifficulties because of conditions associated with poverty or be-cause their native language is not English is unknown.

In summary, the results of this study have extended previouswork on the development of emergent literacy skills and earlyreading in several ways. First, our results highlight the develop-mental continuity between early preschool emergent literacy skills,later preschool emergent literacy skills, and early reading abilitiesof children. Second, these results clarify the nature of reading-related phonological sensitivity. Contrary to the dominant viewthat it is phonemic sensitivity that is critical for decoding, ourresults clearly establish that it is children's global sensitivity tophonological features of language that relates to decoding. Third,this study partially explains the status of other emergent literacyskills in explanatory accounts of the development of reading. Skillssuch as print concepts or the ability to "read" environmental printdo not appear to have independent predictive associations withlater reading; rather, their predictive relations with later readingappear to reflect the development of other emergent literacy skillssuch as letter knowledge and phonological sensitivity. Finally, theresults of this study highlight the significance of the study of theorigins of preschool emergent literacy skills. The high level ofstability of emergent literacy skills from late preschool to earlygrade school, coupled with the lower degree of stability of emer-gent literacy skills from early preschool to late preschool, suggeststhat efforts to identify significant sources of variability betweenchildren in these skills should be directed toward the preschoolyears.

References

Adams, M. J. (1990). Beginning to read: Thinking and learning aboutprint. Cambridge, MA: MIT Press.

Allington, R. L. (1984). Content, coverage, and contextual reading inreading groups. Journal of Reading Behavior, 16, 85—96.

Anthony, J. L., & Lonigan, C. J. (2000). The nature of phonologicalsensitivity: Converging evidence from four studies of preschool andearly-grade school children. Manuscript submitted for publication.

Anthony, J. L., Lonigan, C. J., Burgess, S. R., Driscoll Bacon, K., Phillips,B. M., & Bloomfield, B. G. (2000). Structure of preschool phonologicalsensitivity: Overlapping sensitivity to rhyme, words, syllables, and pho-nemes. Manuscript submitted for publication.

Baydar, N., Brooks-Gunn, J., & Furstenberg, F. F. (1993). Early warningsigns of functional illiteracy: Predictors in childhood and adolescence.Child Development, 64, 815-829.

Bentler, P. M. (1995). EQS structural equations program manual. Encino,CA: Multivariate Software.

Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness offit in the analysis of covariance structures. Psychological Bulletin, 88,588-606.

Bentler, P. M., & Dudgeon, P. (1996). Covariance structure analysis:Statistical practice, theory, and directions. Annual Review of Psychology,47, 563-592.

Bishop, D. V. M., & Adams, C. (1990). A prospective study of therelationship between specific language impairment, phonological disor-ders and reading retardation. Journal of Child Psychology and Psychi-atry and Allied Disciplines, 31, 1027-1050.

Bowers, P. G., & Wolf, M. (1993). Theoretical links among naming speed,precise timing mechanisms and orthographic skill in dyslexia. Reading& Writing, 5, 69-85.

Bowey, J. A. (1994). Phonological sensitivity in novice readers and non-readers. Journal of Experimental Child Psychology, 58, 134-159.

Bradley, L., & Bryant, P. E. (1983). Categorizing sounds and learning toread—A causal connection. Nature, 301, 419-421.

Bradley, L., & Bryant, P. (1985). Rhyme and reason in reading andspelling. Ann Arbor: University of Michigan Press.

Brown, A. L., Palincsar, A. S., & Purcell, L. (1986). Poor readers: Teach,don't label. In U. Neisser (Ed.), The school achievement of minoritychildren: New perspectives (pp. 105-143). Hillsdale, NJ: Erlbaum.

Bruck, M. (1998). Outcomes of adults with childhood histories of dyslexia.In C. Hulme, R. Joshi, & J. Malatesha (Eds.), Reading and spelling:Development and disorders (pp. 179-200). Mahwah, NJ: Erlbaum.

Bryant, P. E., MacLean, M., Bradley, L. L., & Crossland, J. (1990). Rhymeand alliteration, phoneme detection, and learning to read. DevelopmentalPsychology, 26, 429-438.

Burgess, S, R., & Lonigan, C. J. (1998). Bidirectional relations of phono-logical sensitivity and prereading abilities: Evidence from a preschoolsample. Journal of Experimental Child Psychology, 70, 117-141.

Butler, S. R., Marsh, H. W., Sheppard, M. J., & Sheppard, J. L. (1985).Seven-year longitudinal study of the early prediction of reading achieve-ment. Journal of Educational Psychology, 77, 349-361.

Byrne, B., & Fielding-Bamsley, R. F. (1991). Evaluation of a program toteach phonemic awareness to young children. Journal of EducationalPsychology, 82, 805-812.

Chall, J. S., Jacobs, V., & Baldwin, L. (1990). The reading crisis: Whypoor children fall behind. Cambridge, MA: Harvard University Press.

Chaney, C. (1992). Language development, metalinguistic skills, and printawareness in 3-year-old children. Applied Psycholinguistics, 13, 485-514.

Clay, M. M. (1979a). The early detection of reading difficulties (3rd ed.).Portsmouth, NH: Heinemann.

Clay, M. M. (1979b). Reading: The patterning of complex behavior.Auckland, New Zealand: Heinemann.

Cunningham, A. E., & Stanovich, K. E. (1997). Early reading acquisitionand its relation to reading experience and ability 10 years later. Devel-opmental Psychology, 33, 934-945.

Dunn, L. M., & Dunn, L. M. (1981). Peabody Picture Vocabulary Test-Revised. Circle Pines, NM: American Guidance Service.

Echols, L. D., West, R. F., Stanovich, K. E., & Zehr, K. S. (1996). Using


children's literacy activities to predict growth in verbal cognitive skills:A longitudinal investigation. Journal of Educational Psychology, 88,296-304.

Felton, R. H. (1998). The development of reading skills in poor readers:Educational implications. In C. Hulme, R. Joshi, & J. Malatesha (Eds.),Reading and spelling: Development and disorders (pp. 219-233). Mah-wah, NJ: Erlbaum.

Fowler, A. E. (1991). How early phonological development might set thestage for phoneme awareness. In S. A. Brady & D. P. Shankweiler(Eds.), Phonological processes in literacy (pp. 97-117). Hillsdale, NJ:Erlbaum.

Fox, B., & Routh, D. K. (1975). Analyzing spoken language into words,syllables, and phonemes: A developmental study. Journal of Psycholin-guistic Research, 4, 331-342.

Gardner, M. F. (1990). Expressive One-Word Picture Vocabulary Test-Revised. Novato, CA: Academic Therapy.

Goodman, K. S. (1986). What's whole in whole language? Portsmouth,NH: Heinemann.

Goswami, U., & Bryant, P. E. (1990). Phonological skills and learning toread. Hillsdale, NJ: Erlbaum.

Goswami, U., & Bryant, P. E. (1992). Rhyme, analogy, and children'sreading. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.), Readingacquisition (pp. 49-62). Hillsdale, NJ: Erlbaum.

Gough, P. B. (1993). The beginning of decoding. Reading and Writing: AnInterdisciplinary Journal, 5, 181-192.

Hoien, T., Lundberg, I., Stanovich, K. E., & Bjaalid, I. (1995). Componentsof phonological awareness. Reading and Writing: An InterdisciplinaryJournal, 7, 171-188.

Johnston, R. S., Anderson, M, & Holligan, C. (1996). Knowledge of thealphabet and explicit awareness of phonemes in prereaders: The natureof the relationship. Reading and Writing: An Interdisciplinary Journal,8, 217-234.

Juel, C. (1988). Learning to read and write: A longitudinal study of 54children, from first through fourth grades. Journal of Educational Psy-chology, 80, 437-447.

Kirk, S. A., McCarthy, J. J., & Kirk, W. D. (1968). Illinois Test ofPsycholinguistic Abilities. Urbana: University of Illinois Press.

Lentz, F. E. (1988). Effective reading interventions in the regular class-room. In J. L. Graden, J. E. Zins, & M. J. Curtis (Eds.), Alternatingeducational delivery systems: Enhancing instructional options for allstudents (pp. 351-373). Washington, DC: National Association ofSchool Psychologists.

Liberman, A. M., Cooper, F. S., Shankweiler, D., & Studdert-Kennedy, M.(1967). Perception of the speech code. Psychological Review, 74, 431-461.

Liberman, I. Y., Shankweiler, D., Fischer, F. W., & Carter, B. (1974).Explicit syllable and phoneme segmentation in young children. Journalof Experimental Child Psychology, 18, 201-212.

Lonigan, C. J. (1994). Reading to preschoolers exposed: Is the emperorreally naked? Developmental Review, 14, 303-323.

Lonigan, C. J., Burgess, S. R., Anthony, J. L., & Barker, T. A. (1998).Development of phonological sensitivity in two- to five-year-old chil-dren. Journal of Educational Psychology, 90, 294-311.

MacLean, M., Bryant, P., & Bradley, L. (1987). Rhymes, nursery rhymes,and reading in early childhood. Merrill-Palmer Quarterly, 33, 255-282.

Masonheimer, P. E., Drum, P. A., & Ehri, L. C. (1984). Does environ-mental print identification lead children into word reading? Journal ofReading Behavior, 16, 257-27'1.

Metsala, J. L., & Walley, A. C. (1998). Spoken vocabulary growth and thesegmental restructuring of lexical representations: Precursors to phone-mic awareness and early reading ability. In J. L Metsala & L. C. Ehri(Eds.), Word recognition in beginning literacy (pp. 89-120). Mahwah,NJ: Erlbaum.

Morais, J. (1991). Constraints on the development of phonological aware-

ness. In S. A. Brady & D. P. Shankweiler (Eds.), Phonological processesin literacy (pp. 5-27). Hillsdale, NJ: Erlbaum.

Morrison, F. J., Smith, L., & Dow-Ehrensberger, M. (1995). Education andcognitive development: A natural experiment. Developmental Psychol-ogy, 31, 789-799.

Muter, V., Hulme, C, Snowling, M., & Taylor, S. (1997). Segmentation,not rhyming, predicts early progress in learning to read. Journal ofExperimental Child Psychology, 65, 370-398.

Nation, K., & Hulme, C. (1997). Phonemic segmentation, not onset-rimesegmentation, predicts early reading and spelling skills. Reading Re-search Quarterly, 32, 154-167.

Oka, E., & Paris, S. (1986). Patterns of motivation and reading skills inunderachieving children. In S. Ceci (Ed.), Handbook of cognitive, social,and neuropsychological aspects of learning disabilities (Vol. 2). Hills-dale, NJ: Erlbaum.

Pappas, C. C, & Brown, E. (1988). The development of children's senseof the written story language register: An analysis of the texture of"pretend reading. " Linguistics & Education, I, 45-79.

Pikulski, J. J., & Tobin, A. W. (1989). Factors associated with long-termreading achievement of early readers. In S. McCormick, J. Zutell, P.Scharer, & P. O'Keefe (Eds.), Cognitive and social perspectives forliteracy research and instruction. Chicago: National Reading Confer-ence.

Purcell-Gates, V. (1988). Lexical and syntactic knowledge of writtennarrative held by well-read-to kindergartners and second graders. Re-search in the Teaching of English, 22, 128-160.

Purcell-Gates, V. (1996). Stories, coupons, and the TV Guide: Relation-ships between home literacy experiences and emergent literacy knowl-edge. Reading Research Quarterly, 31, 406-428.

Purcell-Gates, V., & Dahl, K. L. (1991). Low-SES children's success andfailure at early literacy learning in skills-based classrooms. Journal ofReading Behavior, 23, 1-34.

Scarborough, H. (1989). Prediction of reading dysfunction from familialand individual differences. Journal of Educational Psychology. 81,101-108.

Share, D. L., Jorm, A. F., MacLean, R., & Mathews, R. (1984). Sources ofindividual differences in reading acquisition. Journal of EducationalPsychology, 76, 1309-1324.

Snow, C. E., Barnes, W. S., Chandler, J., Hemphill, L., & Goodman, I. F.(1991). Unfulfilled expectations: Home and school influences on liter-acy. Cambridge, MA: Harvard University Press.

Stahl, S. A., & Murray, B. A. (1994). Defining phonological awareness andits relationship to early reading. Journal of Educational Psychology, 86,221-234.

Stanovich, K. E. (1986). Matthew effects in reading: Some consequencesof individual differences in the acquisition of literacy. Reading ResearchQuarterly, 21, 360-407.

Stanovich, K. E. (1988). Explaining the differences between the dyslexicand the garden-variety poor reader: The phonological-core variable-difference model. Journal of Learning Disabilities, 21, 590-612.

Stanovich, K. E. (1992). Speculations on the causes and consequences ofindividual differences in early reading acquisition. In P. B. Gough, L. C.Ehri, & R. Treiman (Eds.), Reading acquisition (pp. 307-342). Hillsdale,NJ: Erlbaum.

Stanovich, K. E., Cunningham, A. E., & Cramer, B. B. (1984). Assessingphonological awareness in kindergarten children: Issues of task compa-rability. Journal of Experimental Child Psychology, 38, 175-190.

Stanovich, K. E., & Siegel, L. S. (1994). Phenotypic performance profile ofchildren with reading disabilities: A regression-based test of thephonological-core variable-difference model. Journal of EducationalPsychology, 86, 24-53.

Stevenson, H. W., & Newman, R. S. (1986). Long-term prediction ofachievement and attitudes in mathematics and reading. Child Develop-ment, 57, 646-659.


Sulzby, E. (1985). Children's emergent reading of favorite storybooks: Adevelopmental study. Reading Research Quarterly, 20, 458-481.

Sulzby, E. (1986). Writing and reading: Signs of oral and written languageorganization in the young child. In W. H. Teale & E. Sulzby (Eds.),Emergent literacy: Reading and writing (pp. 50-87). Norwood, NJ:Ablex.

Sulzby, E. (1988). A study of children's early reading development. InA. D. Pelligrini (Ed.), Psychological bases for early education (pp.39-75). Chichester, England: Wiley.

Sulzby, E., & Teale, W. (1991). Emergent literacy. In R. Barr, M. Kamil,P. Mosenthal, & P. D. Pearson (Eds.), Handbook of reading research(Vol. 2, pp. 727-758). New York: Longman.

Teale, W. H., & Sulzby, E. (Eds.). (1986). Emergent literacy: Writing andreading. Norwood, NJ: Ablex.

Torgesen, J. K. (1999). Phonologically based reading disabilities: Towarda coherent theory of one kind of learning disability. In R. J. Sternberg &L. Spear-Swerling (Eds.), Perspectives on learning disabilities (pp.231-262). Boulder, CO: Westview Press.

Tramontana, M. G., Hooper, S., & Selzer, S. C. (1988). Research onpreschool prediction of later academic achievement: A review. Devel-opmental Review, 8, 89-146.

Treiman, R. (1992). The role of intrasyllabic units in learning to read andspell. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.), Readingacquisition (pp. 107-143). Hillsdale, NJ: Erlbaum.

Tunmer, W. E., Herriman, M. L., & Nesdale, A. R. (1988). Metalinguisticabilities and beginning reading. Reading Research Quarterly, 23, 134-158.

Tunmer, W. E., & Rohl, M. (1991). Phonological awareness and readingacquisition. In D. Sawyer & B. Fox (Eds.), Phonological awareness inreading: The evolution of current perspectives (pp. 1-29). New York:Springer-Verlag.

Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological

processing and its causal role in the acquisition of reading skills. Psy-chological Bulletin, 101, 192-212.

Wagner, R. K., Torgesen, J. K., Laughon, P., Simmons, K., & Rashotte,C. A. (1993). Development of young readers' phonological processingabilities. Journal of Educational Psychology, 85, 83-103.

Wagner, R. K., Torgesen, J. K., & Rashotte, C. A. (1994). The develop-ment of reading-related phonological processing abilities: New evidenceof bi-directional 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., Donahue, J., & Garon, T. (1997). Changingrelations between phonological processing abilities and word-level read-ing as children develop from beginning to skilled readers: A 5-yearlongitudinal study. Developmental Psychology, 33, 468-479.

Webster, P. E., & Plante, A. S. (1995). Productive phonology and phono-logical awareness in preschool children. Applied Psycholinguistics, 16,43-57.

Wechsler, D. (1989). Wechsler Preschool and Primary Scales of Intelli-gence—Revised. New York: The Psychological Corporation.

Whitehurst, G. J., & Lonigan, C. J. (1998). Child development and emer-gent literacy. Child Development, 69, 848-872.

Wolfe, M. (1991). Naming speed and reading: The contribution of thecognitive neurosciences. Reading Research Quarterly, 26, 123-141.

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

Received December 22, 1998Revision received March 29, 2000

Accepted April 7, 2000

Call for Papers: Violent Children

Developmental Psychology will publish a special issue in 2002 on the topic of ViolentChildren: Bridging Development, Prevention, and Policy. The problem of chronic vio-lence in children has reached a new level of public awareness. Developmental sciencehas much to contribute to the understanding of this problem and how it can be ad-dressed in the public domain. This special issue will highlight original empirical re-search that contributes to this understanding.

Three kinds of articles will be considered: those that contribute to knowledge of howchronic conduct problems or violent behaviors develop in children or adolescents,those that evaluate rigorous experiments in the prevention or treatment of chronicaggression, and those that evaluate public policies relevant to child conduct problems.Together, these articles will provide a bridge between developmental science and pub-lic policy.

Inquiries should be directed to the section coeditors, Gregory S. Pettit([email protected]) and Kenneth A. Dodge ([email protected]). Manuscriptsshould be standard length (fewer than 40 pages) and should be submitted for standardpeer review prior to October 1,2000, to the editorial office of Developmental Psychol-ogy, University of Wisconsin, Waisman Center, 1500 Highland Avenue, Madison, WI53705-2280. Please include the phrase "Special Issue on Violent Children" in the coverletter that accompanies the submission.

development of emergent literacy and early reading …faculty.swosu.edu/stephen.burgess/development...

Documents