Articulation rate, naming speed, verbal short-term memory, and phonological awareness: Longitudinal predictors of early reading development?

Articulation Rate, Naming Speed, VerbalShort-Term Memory, and PhonologicalAwareness: Longitudinal Predictors of

Early Reading Development?

Rauno ParrilaUniversity of Alberta, Canada

John R. KirbyQueen’s University, Canada

Lynn McQuarrieUniversity of Alberta, Canada

This study examines how measures of articulation rate, verbal short-term memory(STM), naming speed, and phonological awareness tasks administered in kindergar-ten and again in Grade 1 jointly and uniquely predict word reading and passage com-prehension variance in Grades 1, 2, and 3. Results from regression and commonalityanalyses indicated that (a) when measured in Grade 1, phonological processing taskswere better, but not significantly better, predictors of later reading than when mea-sured in kindergarten; (b) articulation rate and verbal STM did not uniquely predictreading if phonological awareness and naming speed were controlled; (c) when mea-sured in kindergarten, both phonological awareness and naming speed accounted forunique variance in reading measures, and (d) when measured in Grade 1, phonologi-cal awareness was the strongest predictor of reading. Commonality analyses indi-cated that kindergarten letter recognition shares large parts of its predictive variancewith phonological awareness and naming speed measures. Finally, controlling for theautoregressive effect of Grade 1 word reading reduced the usefulness of phonological

SCIENTIFIC STUDIES OF READING, 8(1), 3–26Copyright © 2004, Lawrence Erlbaum Associates, Inc.

Requests for reprints should be sent to Rauno Parrila, Department of Educational Psychology, 6–102Education North, University of Alberta, Edmonton, AB T6G 2G5, Canada. E-mail:[email protected]

Do Not

Cop

y

awareness and naming speed as predictors of Grade 3 reading, but both still accountedfor significant unique variance.

Learning to recognize words of increasing complexity and decreasing familiarity isone of the most important academic challenges that children face during the firstyears of formal schooling. There is a growing consensus among reading research-ers that success in this task depends largely on the child’s phonological processingskills (see reviews in Adams, 1990; National Reading Panel, 1999; Snow, Burns, &Griffin, 1998), that is, their ability to use information about the sound elements oflanguage in processing written and oral language (Wagner & Torgesen, 1987).Several studies have indicated that phonological processing measures obtained ei-ther prior to reading or at the very early stages of reading development are strongpredictors of individual differences in word recognition performance up to severalyears ahead (e.g., Bradley & Bryant, 1985; Kirby & Parrila, 1999; Lundberg,Olofsson, & Wall, 1980; MacDonald & Cornwall, 1995; Maclean, Bryant, &Bradley, 1987; Mann, 1993; Wagner, Torgesen, & Rashotte, 1994; Wagner et al.,1997; for a review, see Adams, 1990).

In a similar vein, several studies have shown that children’s early ability to readwords (or nonwords) is a powerful predictor of their concurrent and future abilityto understand sentences and longer passages (e.g., Blackmore & Pratt, 1997; deJong & van der Leij, 2002; Kirby & Parrila, 1999; Muter & Snowling, 1998). Pho-nological processing tasks have also predicted passage comprehension signifi-cantly (e.g., Hansen & Bowey, 1994), sometimes even after controlling for priorpassage comprehension and word reading ability (Kirby & Parrila, 1999). In sum,existing research suggests a developmental progression from phonological pro-cessing skills to word reading, and further to passage comprehension.

Several important and partly interrelated questions about phonological process-ing have received relatively little attention. First, only a handful of longitudinalstudies have included measures of more than two facets of phonological process-ing and, as a result, researchers have limited knowledge about the independent pre-dictive relationships these skills have with reading development. The fact thatphonological processing skills are frequently highly intercorrelated makes thisquestion even more pertinent. High correlations between phonological processingskills also lead to our second question. Several authors have suggested that indi-vidual differences in various phonological processing skills derive from a singlelatent ability (see following sections). However, most studies have focused onunique effects, and we are aware of only two studies (Bowers, 1995; Manis,Seidenberg, & Doi, 1999) that have examined whether the common ability, or thecommon variance between any of the tasks, can account for the observed relation-ships between different phonological processing skills and reading. Finally, re-searchers have limited knowledge regarding the predictive power of phonological

4 PARRILA, KIRBY, MCQUARRIE

Do Not

Cop

y

processing measures obtained prior to reading instruction, compared with thoseobtained later. We provide a short review of these issues, focusing on longitudinalstudies with unselected samples of children. This is followed by an overview ofthis study.

INDEPENDENT PREDICTIVE RELATIONSHIPSBETWEEN PHONOLOGICAL PROCESSING SKILLS AND

READING DEVELOPMENT

A critical issue for those interested in the relationships between different phonolog-ical processes and reading is whether various phonological processing skills haveindependent predictive relationships with reading development. Several studieshave included measures of phonological awareness together with either phonologi-cal memory (verbal short-term memory [STM]) or rapid automatized naming(naming speed) measures, thus including two of the three facets of phonologicalprocessing elaborated on by Wagner and Torgesen (1987) in their influential arti-cle. These studies have generally established that after controlling for verbal STMor naming speed, phonological awareness makes a unique contribution to explain-ing reading variance (e.g., Hecht, Burgess, Torgesen, Wagner, & Rashotte, 2000;Muter & Snowling, 1998). As well, after phonological awareness is controlled, ver-bal STM (e.g., Bradley & Bryant, 1985; Muter & Snowling, 1998; Rohl & Pratt,1995) and naming speed (e.g., Badian, 1994; Hecht et al., 2000) can make uniquecontributions to reading development, although these latter effects are not alwaysverified (e.g., Hecht et al., 2000).

Further complicating the interpretation of independent predictive relationshipsis the fact that those few longitudinal studies that have included phonologicalawareness, verbal STM, and naming speed measures have produced somewhat in-consistent results. Two of these studies (Kirby & Parrila, 1999; Wagner et al.,1994; Wagner et al., 1997) report unique effects of phonological awareness toword reading and passage comprehension 1 or 2 years later, after controlling forthe autoregressive effect (the effect of the same skill at an earlier time point). Inboth studies, naming speed made a unique but smaller and sometimes time-limited(Wagner et al., 1997) contribution to predicting reading development. The effectof verbal STM, in contrast, was fully accounted for by the autoregressors or otherphonological processing variables.

de Jong and van der Leij’s (1999) results with Dutch-speaking children indi-cated that when phonological awareness, verbal STM, and naming speed weremeasured in kindergarten, only naming speed was a significant predictor of Grade1 and Grade 2 word reading. When phonological awareness, verbal STM, andnaming speed were measured in the beginning of Grade 1, naming speed and pho-nological awareness predicted unique variance in year-end Grade 1 word reading,

LONGITUDINAL PREDICTORS 5

Do Not

Cop

y

and all three were significant predictors of Grade 2 word reading. A follow-up ofthe initial study (de Jong & van der Leij, 2002) indicated further that when mea-sured at the end of Grade 1, both phonological awareness and naming speed madeunique and significant contributions to predicting Grade 3 word decoding speed,after controlling for Grade 1 word decoding speed and vocabulary.

Some possible explanations for variability in the results include orthographicregularity of the language (Dutch is much more regular than English), variation inthe outcome measures, what other variables were controlled for, and what kinds ofphonological processing tasks were used. For example, all studies included wordreading as an outcome measure, although the first two studies focused on accu-racy, and de Jong and van der Leij (1999, 2002) focused on decoding speed. Per-haps naming speed is more important for the development of word reading speedthan it is for word reading accuracy (Bowers, 1995), particularly in transparent or-thographies such as Dutch or German (Wimmer, Mayringer, & Landerl, 2000). Interms of what other factors are controlled for, a possible problem arises when gen-eral cognitive ability is operationalized as a vocabulary measure, as it is likely thatvocabulary is differentially related to various phonological processing abilities(see e.g., Baddeley, Gathercole, & Papagno, 1998; Fowler, 1991; Metsala, 1999).More generally, it seems necessary that if general cognitive abilities are controlledfor, we should also examine closely how they are related to both reading and otherpredictor variables.

Autoregressive Effect

The three longitudinal studies previously reviewed included the autoregressive ef-fect in statistical analyses. Including an autoregressor changes the question fromone of predicting growth in general into one of predicting further growth, or unex-pected growth that cannot be accounted for by the skill itself at an earlier time. Aproblem arises, however, when we are examining the early developmental levels ofa skill. For example, what is the proper kindergarten autoregressor for Grade 1word reading? Wagner et al. (1994) and Kirby and Parrila (1999) used both kinder-garten word reading and letter recognition as autoregressors for Grade 1 readingmeasures. In both studies, kindergarten reading measures showed a considerablefloor effect. de Jong and van der Leij (1999) acknowledged that Dutch kindergartenchildren are not expected to read and, therefore, they used letter knowledge as theirautoregressor. Conceptually, it is not clear that letter knowledge is an autoregressorfor word reading or passage comprehension. Moreover, Burgess and Lonigan(1998) showed that letter knowledge and phonological sensitivity were recipro-cally related prior to the onset of formal reading instruction, similar to previousfindings with older children (e.g., Perfetti, Beck, Bell, & Hughes, 1987). To avoidinterpretation problems, it seems necessary for the researchers to establish that their


Do Not

Cop

y

autoregressors are indeed representative of the early developmental levels of thetarget skill and that important reciprocal relationships are not overlooked.

Unique and Common Contributions

Interpretation of the independent predictive relationships that different phonologicalprocessing skills have with reading development is further complicated by the factthat phonological processing measures are intercorrelated. This could possibly re-flect a single latent construct (e.g., Elbro, 1996; Fowler, 1991; Hansen & Bowey,1994; McDougall, Hulme, Ellis, & Monk, 1994; Metsala, 1999; Muter & Snowling,1998). An important issue that has not been sufficiently addressed is whether this la-tent ability can account for the observed relationships of phonological processingtasks with reading development (see also Goswami, 2000). Most empirical studieshave used analytical methods aimed at examining the unique contributions of differ-ent predictors rather than their common or shared contributions. The two notable ex-ceptions are studies by Manis et al. (1999) and Bowers (1995). These studies usedphonological awareness and naming speed measures in Grades 1 and 2, respectively,to predict various reading outcomes 1 year later. In both studies, commonality analy-ses indicated that the common component did not account completely for the ob-served relationships of phonological awareness and naming speed tasks with readingmeasures. For most reading measures, the common component was a poorer predic-tor of performance than one or both of the unique components.

Kindergarten and Grade 1 Phonological Processes asPredictors of Reading

The final issue we address is the predictive power of phonological processing mea-sures administered in kindergarten when compared to phonological processing mea-sures administered in Grade 1. This question has significant ramifications for the al-location of limited assessment and intervention resources available in schools. Asindividual differences in phonological processes appear to be relatively stable priorto reading instruction (e.g., Lonigan, Burgess, & Anthony, 2000) and into the schoolyears (Lonigan et al., 2000; Wagner et al., 1997), early identification of phonologicalprocessing deficits seems a realistic possibility. If, however, phonological processesmeasured in Grade 1 provide significantly better predictive power than measures ad-ministered earlier, limited resources may be better spent later.

OVERVIEW OF THIS STUDY

This study examines longitudinally how the same phonological processing mea-sures administered in kindergarten and again in Grade 1 predict word reading and


Do Not

Cop

y

passage comprehension in Grades 1, 2, and 3. We first employ regression and com-monality analyses to examine what unique and common contributions phonologi-cal processing measures make in predicting reading variance up to 3 years ahead,and to compare the relative power of kindergarten and Grade 1 predictors. Next, weexamine what effect controlling for letter knowledge or previous reading skills hason the relationship between phonological processing and reading measures.

We use measures of phonological awareness, verbal STM, naming speed, andarticulation rate to predict both word reading and passage comprehension accu-racy. We call the predictor variables jointly, and loosely, phonological processingvariables (see Wolf & Bowers, 1999, for an alternative conceptualization). The in-clusion of articulation rate was motivated by two observations. First, articulationrate has predicted reading variance in several studies comparing different readingability groups (see e.g., Ackerman, Dykman, & Gardner, 1990; Avons & Hanna,1995; Catts, 1989; Das, Mensink, & Mishra, 1990; Das & Mishra, 1991; Das,Mok, & Mishra, 1994; Gallagher, Laxon, Armstrong, & Frith, 1996; McDougall etal., 1994). Second, Hulme (e.g., Hulme & Roodenrys, 1995) has argued that verbalSTM is related to reading via articulation rate: Poor readers have poor (subvocal)articulation rates which then leads to poor memory spans as a consequence of slowprocessing of information in the articulatory loop of the working memory system.There is, however, little information regarding the predictive power of articulationrate in unselected samples. Only one longitudinal study has included articulationrate measures. de Jong and van der Leij (2002) reported that articulation rate mea-sured at the end of Grade 1 correlated only modestly with word decoding speedand reading comprehension measured at the end of Grade 3 (r = .20 for both). Inthe following, we examine both the unique and shared effects that articulation ratemay have on reading development.

METHOD

Participants

In the 1st year of the study, 161 children in Senior Kindergarten (mean age = 66.7months) were recruited to participate (Senior Kindergarten is the 1st year of com-pulsory schooling in Ontario, although there is no formal reading instruction). Let-ters describing the study were sent to parents of all children in the targeted classes.All participating schools were in Kingston, Ontario, and served a broad range of so-cial class neighborhoods. In Grade 1, 121 of these children were tested again, andfull data were available for 117 children. In Grade 2, 105 of the original participantswere tested for the third time, with full data available for 102 participants. In Grade3, 95 of the original participants were administered a subset of the original tests, in-cluding both reading tests (see the following). Complete data across the 4 years


Do Not

Cop

y

were available for 93 children. Attrition was due to children moving out of the area,being unavailable for testing, or parents withdrawing their child’s participation.Comparisons between the children who remained in the study and those who left in-dicated no significant differences between the groups, all ts < 1.90.

Tasks

Articulation Rate. The Speech Rate (SR) task from the Das–Naglieri Cogni-tive Assessment System (Naglieri & Das, 1997) was used to assess articulation rate.The SR task requires the participant to say a series of words, 10 times, as fast as pos-sible. There are eight items, each containing three very common one- or two-sylla-ble words. The participants’ raw score was the time taken to complete all eightitems. Repetitions with errors were not included, but children were allowed to cor-rect themselves. The reported split–half reliabilities for this task are .85 for age 5and .87 for age 6 (Naglieri & Das, 1997).

Verbal STM. Word Series (WS) from the Das–Naglieri Cognitive Assess-ment System (Naglieri & Das, 1997) was used as a measure of verbal STM. WS is aword memory span test in which the participant is asked to repeat a series of wordsin the order they were presented. Words are presented orally at a rate of one per sec-ond. The first two series consist of only two words and the length of each series isgradually increased with the most difficult series consisting of nine words. Ninedifferent single syllable words are used as targets. Presentation of items was dis-continued after four consecutive failures and the participant’s WS score was thenumber of series correctly reproduced. Naglieri and Das reported split–halfreliabilities of .80 and .83 for ages 5 and 6, respectively.

Rapid Serial Naming. Color Naming (CN) required the participant to statethe names of four colors (blue, green, red, and yellow) as quickly as possible. Thecolors were presented in 4 × 4 randomly ordered arrays on two separate pages. Priorto beginning the timed naming, each participant was asked to name the colors to en-sure familiarity. The two pages were timed separately and the combined time inseconds to name all 32 targets was the raw score. No reliability data are availablefor this measure. However, Kirby and Parrila (1999) reported a kindergarten corre-lation of .68 and Grade 1 correlation of .67 between CN and similarly constructedObject Naming tasks.

Phonological Awareness. Participants’ phonological awareness skillswere assessed with two tasks from Wagner, Torgesen, Laughon, Simmons, andRashotte (1993). Sound Isolation (SI) requires the participant to identify the first,the last, or the middle sound in a word. There were 6 practice items and 15 test items


Do Not

Cop

y

consisting of three- and four-phoneme one-syllable words. In Blending Phonemes(BP) the participant is presented individual phonemes orally at the rate of two persecond and then asked to pronounce the word that results when these phonemes areblended together. The task consisted of 6 practice items and 15 test items. The firstitem consisted of two phonemes whereas the most difficult items had six pho-nemes. Both tests were discontinued after four mistakes in the last 7 items. A partic-ipant’s score was the number of correct items. Wagner et al. (1993) reported coeffi-cient alpha reliability estimates of .89 for SI and .88 for BP in kindergarten.

Letter Knowledge and Reading Tasks. Participants’ letter knowledgewas assessed in kindergarten and Grade 1 by administering the Letter Identificationtest (Clay, 1993). This test asks the participant to identify each of the uppercase andlowercase letters. Two lowercase letters, a and g, are presented in two differentfonts, so the total possible score is 54. Clay reported a split–half reliability of .97 forage 6 in this task.

Reading performance was assessed at all grade levels using tests from theWoodcock Reading Mastery Tests–Revised battery (Woodcock, 1987). Form Htests were used in Grades 1 and 3 and Form G in Grade 2. The Word Identificationtest requires the participant to read isolated words aloud. Words are graded in diffi-culty from preprimer to adult level. Woodcock reports split–half reliabilities of .98and .97 for Grades 1 and 3, respectively.

The Passage Comprehension test requires the participant to read short passages(usually two to three lines) and identify a key word missing from the passages.Woodcock reported split–half reliabilities of .94 and .92 for Grades 1 and 3, re-spectively.

Procedure

All tests were administered in kindergarten and in Grade 1 (with the exception ofthe Passage Comprehension test, which was not administered in kindergarten). TheWoodcock (1987) reading tests were also administered in Grades 2 and 3. All par-ticipants were tested individually in their respective schools during school hours bytrained experimenters (graduate students). Testing was divided into sessions last-ing roughly 20 to 30 min.

RESULTS

Descriptive Statistics

Table 1 displays the means and standard deviations for all phonological processingtasks and Letter Recognition in kindergarten and Grade 1. As would be expected atthis age, the participants’ performance level improved in every task from kinder-


Do Not

Cop

y

garten to Grade 1. A closer examination of the distributional properties of the tasksindicated several problems. Log transformations were performed on all speededmeasures (Articulation Rate and Naming Speed). Only the log-transformed scoreswere used in subsequent analyses. Contrary to most other predictor variables, Let-ter Recognition scores were negatively skewed. Both were reflected by subtractingthe actual score from X, where X was equal to the largest score + 1 (Tabachnick &Fidell, 1996). Grade 1 Letter Recognition was also kurtic and subsequent analyseswere performed with a log transformed score. As both Letter Recognition scoresused in subsequent analyses were reflected, results are corrected for direction tosimplify their interpretation. Finally, raw scores of both phonological awarenesstasks showed a floor effect (about 20% of the sample scored 0), and the sum of rawscores also showed a floor effect (with 12% of the sample scoring zero). Standard-ized SI and BP (correlations between the two were .69 and .58 in kindergarten andGrade 1, respectively) were combined to make the Phonological Awareness score.In kindergarten, the sum of standardized scores was slightly positively skewed buttransformations did not improve the situation. In Grade 1, the floor effect wassmaller (only 1 in the sum score), but 4 participants solved all items correctly. Notransformations were made. Only kindergarten and Grade 1 Phonological Aware-ness construct scores were included in the remaining analyses.

Table 2 displays the correlations between Letter Recognition, ArticulationRate, Verbal STM, Naming Speed, and Phonological Awareness in kindergarten(below the diagonal) and in Grade 1 (above the diagonal). Values at the diagonalrepresent the intravariable correlations across time. Table 2 shows that all correla-tions between different predictor variables are significant and that the phonologi-cal processing variables share up to 20% of their variance. Letter Recognition andPhonological Awareness, in turn, share up to 35% of their variance, and kindergar-ten Letter Recognition in particular was highly correlated with all the phonologicalprocessing variables. Moreover, individual differences in these variables are rather


TABLE 1Means and Standard Deviations of Kindergarten and Grade 1 Predictor Variables

Kindergarten Grade 1

M SD M SD

Letter Recognition 35.91 17.16 48.50 8.40Articulation Rate 137.88 34.54 107.96 21.69Word Series 10.50 3.88 12.67 3.30Color Naming 38.46 12.10 30.97 10.45Sound Isolation 4.03 4.36 7.72 4.65Blending Phonemes 5.91 5.24 9.75 4.24

Note. N = 117.

Do Not

Cop

y

stable across the first 2 years of schooling, as indicated by the high correlations be-tween the same tasks at two different time points displayed on the diagonal.

Table 3 shows the means and standard deviations for Grade 1, Grade 2, andGrade 3 Word Identification and Passage Comprehension scores. Both tasks showconsiderable growth, as would be expected during the first three grades. Examina-tion of the distributional properties of the scores indicated that the Grade 1 WordIdentification score was positively skewed and had several outliers. A log-trans-formed score was used in subsequent analyses. The Grade 1 Passage Comprehen-sion score showed a floor effect with 30 participants scoring zero and 27participants scoring 1. No transformation resulted in a normally distributed score,and the raw score was used in all analyses. All remaining reading scores were nor-mally distributed.

Finally, Table 4 displays the correlations between Word Identification and Pas-sage Comprehension scores, and all kindergarten and Grade 1 predictor variables.These include the scores themselves in earlier years, gender, Letter Recognition,and the four phonological processing variables.


TABLE 2Correlations Between the Predictor Variables in Kindergarten and Grade 1

LetterRecognition

ArticulationRate

VerbalSTM

NamingSpeed

PhonologicalAwareness

Letter Recognition .70*** –.22* .32*** –.32*** .56***Articulation Rate –.34*** .69*** –.46*** .31** –.26**Verbal STM .38*** –.46*** .77*** –.31** .24**Naming Speed –.48*** .40*** –.27** .74*** –.42***Phonological Awareness .59*** –.39*** .35*** –.36*** .68***

Note. Kindergarten correlations are below the diagonal; Grade 1 correlations are above thediagonal; N = 117. STM = short-term memory.

*p < .05. **p < .01. ***p < .001.

TABLE 3Means and Standard Deviations for the Word Identification and Passage Comprehension

Scores from Grade 1 to Grade 3

Grade 1a Grade 2b Grade 3c

M SD M SD M SD

Word Identification 12.98 18.72 40.62 18.38 56.62 15.74Passage Comprehension 6.62 9.79 19.94 9.98 29.12 8.14

aN = 117. bN = 102. cN = 95.

Do Not

Cop

y

Table 4 indicates that gender did not correlate with any of the reading outcomevariables, and consequently it was dropped from further analyses. All other vari-ables correlated significantly with at least some of the reading outcome variables.Table 4 indicates further that reading scores are very stable from year to year, sug-gesting that if the same skill measured a year earlier is included as an autoregressor,very little ‘unexpected’ growth will be left for other variables to explain.

Relative Power and Unique Contributions of Kindergartenand Grade 1 Phonological Processing Variables

Table 5 presents results from regression analyses with reading measures (Grades 1to 3) as dependent variables and kindergarten and Grade 1 phonological processingscores as independent variables. Only R2, standardized beta coefficients, and sig-nificance levels are reported. All variables were entered simultaneously, and betasreflect unique contributions after the effects of all other variables are controlled.


TABLE 4Correlations Between Kindergarten and Grade 1 Predictor Variables and Grade 1, 2, and 3

Word Identification and Passage Comprehension Scores


WI PC WI PC WI PC

KindergartenGender .07 –.06 .03 .06 .02 .08Letter Recognition .62*** .53*** .58*** .59*** .50*** .48***Articulation Rate –.24* –.29** –.22* –.27** –.23* –.29**Verbal STM .25*** .24** .27** .32** .28** .27**Naming Speed –.51*** –.55*** –.46*** –.50*** –.52*** –.51***Phonological

Awareness .71*** .70*** .58*** .54*** .47*** .40***Grade 1

Letter Recognition .60*** .45*** .62*** .62*** .53*** .48***Articulation Rate –.18 –.18 –.16 –.24* –.16 –.23*Verbal STM .24** .26** .22* .26** .23* .26*Naming Speed –.48*** –.52*** –.40*** –.43*** –.45*** –.47***Phonological

Awareness.68*** .59*** .70*** .69*** .65*** .60***

Word Identification .86*** .76*** .73*** .68*** .59***Passage

Comprehension .71*** .69*** .61*** .55***

Note. WI = Word Identification; PC = Passage Comprehension; STM = short-term memory.aN =117. bN = 102. cN = 95.*p < .05. **p < .01. ***p < .001.

Do Not

Cop

y

Table 5 shows that with the exception of Grade 1 analyses, R2 values for Grade 1predictors are higher than those for kindergarten predictors. To assess whether oneset of predictors was significantly better than the other, we calculated z values for thedifferences following procedures outlined in Tabachnick and Fidell (1996). Theseanalyses indicated that kindergarten phonological processing variables were signif-icantlybetterpredictorsofGrade1PassageComprehension thanGrade1phonolog-ical processing variables, z = 2.69. All other z values were within the critical ±1.96range for a two-tailed test (range = –1.54–1.41).

In terms of the different phonological processing measures, Verbal STM did notaccount for significant unique variance in any of the regression analyses reported inTable 5, despite being significantly correlated with all outcome measures (see Table4). Kindergarten Articulation Rate, in turn, was a significant predictor only of Grade1 Word Identification after controlling for Verbal STM, Naming Speed, and Phono-logical Awareness. However, the direction of effect is positive, indicating thatslower articulation rate was associated with better performance (the same was truefor all reading outcome measures). These results suggest that Articulation Rate mayhaveactedasasuppressorvariable (Cohen&Cohen,1983).Acloser lookat the rela-tionship between Kindergarten Articulation Rate and Grade 1 Word Identification


TABLE 5Unique Contributions of Kindergarten and Grade 1 Phonological Processing Variables in

Accounting for Word Identification and Passage Comprehension Variancein Grades 1, 2, and 3


WI PC WI PC WI PC

KindergartenArticulation Rate .146* .084 .131 .084 .135 .010Verbal STM .006 –.033 .088 .133 .136 .109Naming Speed –.333*** –.373*** –.333*** –.364*** –.425*** –.408***Phonological

Awareness .653*** .611*** .497*** .411*** .348*** .244*R2 .598*** .599*** .438*** .421*** .389*** .331***

Grade 1Articulation Rate .084 .084 .057 –.032 .063 –.015Verbal STM .065 .083 .028 .030 .054 .061Naming Speed –.241** –.329*** –.145 –.170* –.229** –.262**Phonological

Awareness .586*** .454*** .650*** .605*** .567*** .475***R2 .517*** .446*** .511*** .506*** .475*** .425***

Note. WI = Word Identification; PC = Passage Comprehension; STM = short-term memory.aN =117. bN = 102. cN = 95.*p < .05. **p < .01. ***p < .001.

Do Not

Cop

y

indicated that suppression did not occur when Articulation Rate was entered withany one of the other variables but did occur with all the combinations of two othervariables.Whenkindergarten regressionswithoutArticulationRatewerecomparedto those with it, including Articulation Rate seemed to enhance the relationship be-tween all other predictor variables and word reading. When measured in Grade 1,ArticulationRatefailed topredictuniquevariance inanyof thereadingvariables.

Kindergarten Naming Speed and Phonological Awareness both accounted forlarge unique variance in all reading measures. The Grade 1 predictors present adifferent picture with Phonological Awareness clearly being the stronger predictorof reading variance. It is possible that Grade 1 Phonological Awareness is alreadyinfluenced by earlier reading acquisition (e.g., the correlation between kindergar-ten Letter Recognition and Grade 1 Phonological Awareness was .60; see alsoPerfetti et al., 1987).

Commonality Analyses

Coefficients shown in Table 5 reflect the unique contributions of each independentvariable. What is missing is an account of proportions of variance that can be attrib-uted to various combinations of independent variables, although these commonal-ity elements (Pedhazur, 1982) are reflected in the amount of variance explained.Given relatively high correlations between predictor variables (Table 2), we can as-sume that a large proportion of reading variance is accounted for by the commonal-ity elements, possibly including the element capturing the commonality of Articu-lation Rate and Verbal STM. Commonality analysis is a method of variancepartitioning that allows us to identify proportions of variance in the dependent vari-able attributed uniquely to each of the predictor variables, and the proportions ofvariance attributed to various combinations of independent variables (seePedhazur, 1982, for details).

Table 6 displays the results from commonality analyses involving kindergartenphonological processing variables and Grade 1 to Grade 3 reading outcome vari-ables. Articulation Rate and Naming Speed were first multiplied by –1 to reducethe likelihood of negative commonalities and to simplify their interpretation(Pedhazur, 1982) should they exist. Note that the sum value at the bottom of the ta-ble is equal to R2 values in Table 5 and that the individual values add up to it(within rounding error).

The elements common to Articulation Rate and Verbal STM in Table 6 are eithervery close to zero or negative, possibly reflecting suppression. When consideringthis, and nonsignificant unique contributions (Table 5), it seems reasonable to con-clude that neither of these tasks is a useful (Pedhazur, 1982) predictor of readingvariance in our sample when Naming Speed and Phonological Awareness are in-cluded (note that all commonality elements including Naming Speed or Phonologi-


Do Not

Cop

y

cal Awareness reflect variation in reading variables that would be accounted for bythese variables alone).

The elements common to all tasks accounted for 6 to 8% of the sum and were al-ways smaller than the unique contributions of Naming Speed and PhonologicalAwareness. Similarly, the elements common to Naming Speed and PhonologicalAwarenessweresmaller than theuniquecontributionsof these taskswithoneexcep-tion(Grade1WordIdentification).Thiswouldseemtosuggest thatwhat isunique tothese tasks is more important in terms of prediction of reading variance than whatthey share. An interesting detail about the commonality between Naming Speed andPhonological Awareness is that it seems to decline over the years; in other words,when measured in kindergarten the two tasks predict later reading performance inmore distinct ways than they predict early reading performance.

Table 7 presents the results from commonality analyses involving Grade 1 pho-nological processing variables and Grade 1 to 3 reading outcome variables. Aswith the kindergarten predictors, Articulation Rate and Verbal STM do not seem


TABLE 6Unique and Common Contributions of Kindergarten Phonological Processing Variables in



WI PC WI PC WI PC

Unique contributions1. Articulation Rate .015 .005 .012 .005 .012 .0002. Verbal STM .000 .001 .006 .014 .015 .0103. Naming Speed .088 .110 .090 .107 .144 .1334. Phonological

Awareness.329 .287 .198 .136 .098 .048

Common contributionsCommon to 1 & 2 .002 .002 –.004 –.004 –.007 .001Common to 1 & 3 –.012 –.004 –.011 –.005 –.012 .010Common to 1 & 4 –.015 –.003 –.012 –.005 –.011 .002Common to 2 & 3 .000 –.001 .005 .008 .007 .005Common to 2 & 4 .011 .005 .015 .016 .010 .006Common to 3 & 4 .100 .103 .062 .055 .056 .038Common to 1, 2, & 3 –.003 –.002 .003 .008 .009 .013Common to 1, 2, & 4 –.001 .002 .006 .010 .005 .006Common to 1, 3, & 4 .032 .041 .027 .028 .026 .028Common to 2, 3, & 4 .013 .011 .011 .012 .008 .005Common to all .038 .042 .029 .034 .028 .026

Sum .598 .599 .438 .421 .389 .331

Note. WI = Word Identification; PC = Passage Comprehension; STM = short-term memory.aN =117. bN = 102. cN = 95. Do

Not C

opy

to be useful predictors of reading variance. Both their unique contributions andtheir commonality elements are either very small or negative. Also similar to Ta-ble 6, the elements common to all tasks do not account for a large proportion of thereading variance. What is different in these analyses is that Phonological Aware-ness is now clearly the most useful predictor of reading variance and that the ele-ments common to Phonological Awareness and Naming Speed are considerablylarger than the elements unique to Naming Speed. This would seem to suggest thatwhen measured in Grade 1, what Naming Speed shares with Phonological Aware-ness is more important as a predictor of reading variance than what is unique to it.

To summarize, our results indicate that (a) Articulation Rate and Verbal STMare not useful predictors of reading variance if phonological awareness or namingspeed measures are included and (b) how Phonological Awareness and NamingSpeed predict reading variance in unique and shared ways depends on when theyare measured.


TABLE 7Unique and Common Contributions of Grade 1 Phonological Processing Variables in



WI PC WI PC WI PC

Unique contributions1. Articulation Rate .005 .005 .002 .001 .003 .0002. Verbal STM .003 .005 .001 .001 .002 .0033. Naming Speed .044 .082 .016 .022 .041 .0544. Phonological

Awareness.248 .130 .288 .248 .200 .153

Common contributionsCommon to 1 & 2 –.002 –.009 –.021 –.024 –.026 –.012Common to 1 & 3 –.004 –.004 –.002 .002 –.002 .002Common to 1 & 4 –.001 .001 .000 .000 .000 .000Common to 2 & 3 .005 .008 .002 .002 .004 .005Common to 2 & 4 .009 .009 .006 .006 .007 .007Common to 3 & 4 .161 .164 .155 .149 .175 .133Common to 1, 2, & 3 .000 .009 .020 .028 .026 .019Common to 1, 2, & 4 .022 .034 .078 .086 .090 .054Common to 1, 3, & 4 .005 .002 .003 .017 .003 .013Common to 2, 3, & 4 .016 .016 .036 .019 .019 .017Common to all .005 –.007 –.058 –.052 –.067 –.023

Sum .517 .446 .511 .506 .475 .425

Note. WI = Word identification; PC = Passage comprehension; STM = short-term memoryaN =117. bN = 102. cN = 95. Do

Not C

opy

Controlling for Letter Knowledge and Prior Reading Skills

Letter knowledge. To examine the effect of including measures of letterknowledge and prior reading skills on the usefulness of phonological processingvariables as predictors of reading variance, we first report results from commonal-ity analyses that include kindergarten Letter Recognition, Naming Speed, and Pho-nological Awareness as predictors. This approach was chosen instead of fixed-or-der regression analyses, because conceptually it is not clear that letter knowledge isan autoregressor for word reading or passage comprehension. The use of letterknowledge as an autoregressor would ignore the bidirectional relationships that al-ready exist between letter knowledge and other preliteracy skills prior to the onsetof formal reading instruction. As Articulation Rate and Verbal STM did not con-tribute uniquely in the previous analyses, they were dropped from these analyses tokeep the number of elements in Table 8 within a manageable range. As in the previ-ous analyses, Naming Speed was multiplied by –1 to reduce the likelihood of nega-tive commonalities.

Table 8 presents the results from commonality analyses involving kindergartenLetter Recognition, Naming Speed, and Phonological Awareness, and Grade 1 toGrade 3 reading outcome variables. Replacing Articulation Rate and Verbal STMwith Letter Recognition has little or no positive effect on the predictive power ofthe models. Phonological Awareness is clearly the most useful predictor of Grade


TABLE 8Unique and Common Contributions of Kindergarten Letter Recognition, Naming Speed,

and Phonological Awareness in Accounting for Word Identification and PassageComprehension Variance in Grades 1, 2, and 3


WI PC WI PC WI PC

Unique contributions1. Letter Recognition .027** .001 .045** .060** .023 .032*2. Naming Speed .038** .084*** .042** .058** .094*** .097***3. Phonological

Awareness.163*** .200*** .080*** .046** .041* .016

Common contributionsCommon to 1 & 2 .035 .019 .044 .060 .054 .064Common to 1 & 3 .162 .091 .128 .111 .062 .046Common to 2 & 3 .018 .030 .011 .010 .011 .007Common to 1, 2, & 3 .164 .168 .118 .119 .106 .090

Sum .608 .592 .469 .465 .392 .353

Note. WI = Word Identification; PC = Passage Comprehension.aN =117. bN = 102. cN = 95.*p < .05. **p < .01. ***p < .001.

Do Not

Cop

y

1 reading performance, as well as of Grade 2 word reading, whereas NamingSpeed seems to be a more useful predictor of passage comprehension and Grade 3reading performance. In most analyses, Letter Recognition makes a smallerunique contribution than the other two variables. However, contrary to results inTable 6, common elements seem to account for a large proportion of variance inthese analyses. In general, common elements account for approximately two thirdsof the explained variance, and the elements common to all three tasks account formost variance in all but one of the analyses. In sum, Letter Recognition shares alarge part of its predictive variance with Phonological Awareness and NamingSpeed. The fact that both Phonological Awareness and Naming Speed frequentlymake larger unique contributions than Letter Recognition further challenges thepractice of automatically controlling for the effect of letter knowledge.

Prior reading skills. Table 9 reports results from fixed-order regressionanalyses with Grade 1 predictor variables. These analyses allowed us to examinethe usefulness of phonological processing variables as predictors of reading vari-ance after controlling for the effect of prior reading skills. In other words, Grade 1phonological processing variables are now used to predict further growth in read-ing that cannot be accounted for by Grade 1 reading performance. To simplify thepresentation, we focus only on Grade 3 reading outcomes. The values represent the


TABLE 9R2 Change Associated With Adding Phonological Processing Predictors Into the

Regression Equation After the Autoregressive Effect Was Controlleda

Grade 3

Step Grade 1 Predictor Variables WI PC

1. Word Identification .455*** .353***2. Articulation Rate .004 .0203. Verbal STM .003 .0053. Naming Speed .023* .039*3. Phonological Awareness .075*** .063**2. Verbal STM .006 .0173. Naming Speed .022 .041*3. Phonological Awareness .074*** .067**2. Naming Speed .026* .051**3. Phonological Awareness .068*** .059**2. Phonological Awareness .078*** .074***3. Naming Speed .015 .036*

Note. WI = Word Identification; PC = Passage Comprehension; STM = short-term memory.aN = 95.*p < .05. **p < .01. ***p < .001.

Do Not

Cop

y

R2 change resulting from adding the phonological processing variables into theequation. The autoregressive effect of Word Identification was entered on the firststep, followed by phonological processing variables in varying order, as indicatedin Table 9. Grade 1 Word Identification was used as an autoregressor with bothreading measures, as it had considerably better distributional properties and its cor-relation with Grade 3 Passage Comprehension was as high as that of Grade 1 Pas-sage Comprehension (see Table 4).

After the autoregressive effect of Word Identification was controlled, Articula-tion Rate and Verbal STM did not predict unique reading variance. Subsequently,they were not included in the second set of analyses in which the effect of otherphonological processing variables was controlled. Naming Speed remained a sig-nificant predictor of word reading after controlling for the autoregressive effectand of Passage Comprehension in all of the analyses. It was not, however, a signif-icant predictor of word reading after controlling the effect of both prior word read-ing and Phonological Awareness or Verbal STM. Finally, PhonologicalAwareness significantly predicted reading variance in all analyses. In sum, con-trolling for the autoregressive effect reduced the usefulness of Grade 1 NamingSpeed as a predictor of reading variance 2 years later. In contrast, PhonologicalAwareness remained a useful predictor in all analyses.

DISCUSSION

The central question examined in this article is whether different phonological pro-cessing skills have independent predictive relationships with reading development.We examined this question first without including the autoregressive effect andemploying both regression and commonality analyses, and then with theautoregressive effect included in fixed-order regression analyses.

Results from all analyses converged in indicating that Articulation Rate was nota useful predictor of reading development in our sample. Zero-order correlationsbetween Articulation Rate and reading variables were relatively small, and Articu-lation Rate shared its predictive variance with other phonological processing tasks.These results are in agreement with earlier findings by de Jong and van der Leij(2002). Together, these studies indicate that when measured in kindergarten orGrade 1, articulation rate is not a useful predictor of reading development in an un-selected sample of children.

Verbal STM correlated significantly with all reading outcome measures but didnot account for significant unique variance in any of the regression analyses re-ported in Table 5 after the effect of other phonological processing variables wascontrolled. Together with commonality analyses showing no independent contri-bution of Verbal STM, this suggests that Verbal STM shares its predictive vari-


Do Not

Cop

y

ance with other phonological processing tasks, as has been previously found by deJong and van der Leij (1999) and Wagner et al. (1994; Wagner et al., 1997).

In contrast, Naming Speed, particularly when measured in kindergarten, madea unique and lasting contribution to predicting reading variance. This was evidentin the analyses without the autoregressor, in which the effect of all other phonolog-ical processing variables was controlled, and the analyses with the autoregressiveeffect included. This result is in agreement with de Jong and van der Leij (1999),but only partly with Wagner et al. (1997), who reported time-limited effects ofnaming speed. There are some clear differences between Wagner et al. (1997) andthis study that may explain differences in results. First, Wagner et al. (1997) usedkindergarten naming speed only to predict Grade 2 word decoding. At the sametime, Grade 1 naming speed was used to predict Grade 3 word decoding and Grade2 naming speed was used to predict Grade 4 word decoding. Their conclusion oftime-limited effects was based on the fact that the effect of naming speed de-creased with each analysis with kindergarten naming speed being the best predic-tor. Our results show a similar trend. Grade 1 Naming Speed was not an equallystrong predictor as kindergarten Naming Speed, perhaps due to sharing more of itspredictive variance with Phonological Awareness. Wagner et al. (1997) also useda different autoregressor in each analysis as they consistently controlled for worddecoding ability 2 years earlier. Stability of the scores across these intervals in-creased considerably, thus decreasing the amount of unexpected growth to accountfor in their latter two analyses. In our analyses, Grade 1 Naming Speed was a sig-nificant predictor of Grade 3 reading even after controlling for the autoregressiveeffect. Wagner et al. (1997), however, included vocabulary as a second controlvariable, which could explain the differences.

A second possible reason for divergent results is the naming speed task used.Wagner et al.’s (1997) naming tasks involved naming alphanumerical stimuli, ei-ther letters or digits, whereas our task involved naming colors. As discussed byWagner et al. (1997), it is possible that including alphanumerical stimuli makesthese tasks “mere proxies for individual differences in early literacy and print ex-posure” (p. 476; see also Neuhaus & Swank, 2002). Although their results do notentirely support this conclusion, it is possible that naming colors is less dependenton early literacy skills than naming letters and digits, and as a consequence, itwould also be less affected by the inclusion of an autoregressive effect. This inter-pretation is partly supported by de Jong and van der Leij (2002), who reportedlower correlations between object naming and decoding than between letter ordigit naming and decoding (note, however, that the same is not true for readingcomprehension).

Both kindergarten and Grade 1 Phonological Awareness accounted for uniquevariance in all reading measures after the effect of other phonological processingvariables was controlled. When measured in Grade 1, Phonological Awarenesswas clearly the strongest predictor of reading across the 3 years. This was evident


Do Not

Cop

y

also in analyses that did include the autoregressive effect. It is also possible that theincreased significance, particularly at the level of single phonemes (as opposed tosyllables or onset and rime) as measured in this study, reflects the bidirectional re-lationship between emerging reading abilities and phonological awareness (e.g.,Perfetti et al., 1987; Wimmer, Landerl, Linortner, & Hummer, 1991).

Including an autoregressor considerably affected the usefulness of bothNaming Speed and Phonological Awareness as predictors of reading. Including anautoregressor changes the question from one of predicting growth in general intoone of predicting further growth, or unexpected growth that can not be accountedfor by the skill itself at an earlier time. As de Jong and van der Leij (2002) alsopointed out, absence of such an additional direct effect does not by itself allow theconclusion that the predictors are no longer of importance. According to de Jongand van der Leij (2002), additional effects are observed only when two conditionsare met. The first condition is that individual differences in the target skill are notentirely stable over time, allowing for unexpected growth. In this study, for exam-ple, Grade 1 Word Identification correlated .68 with Grade 3 Word Identificationand .59 with Grade 3 Passage Comprehension. This level of stability leaves littlereliable variance for other predictors to account for. The second condition is thatany predictor variables should be relatively more strongly related to the target vari-able at the later point in time. In our case, this would mean that the unexpectedgrowth should be due to the earlier individual differences in Phonological Aware-ness or Naming Speed. This “strengthening of relationships” may be true for kin-dergarten Naming Speed in this study, but likely not for Phonological Awareness.

Commonality analyses distinguishing between unique and shared predictivevariance produced some interesting results. With kindergarten predictors, the ele-ments common to all tasks were always smaller than the unique contributions ofNaming Speed and Phonological Awareness. Similarly, the elements common toNaming Speed and Phonological Awareness were always smaller than the uniquecontributions of these tasks. This would seem to support Wolf and Bowers (1999)in that what is unique to Naming Speed and Phonological Awareness seems to bemore important in terms of the prediction of reading variance than what they share.Our results are somewhat stronger than those reported by Bowers (1995) andManis et al. (1999), who found that the component shared by Phonological Aware-ness and Naming Speed tasks was frequently a poorer predictor of reading perfor-mance than the better of the two unique components. In these two studies,however, the common component was almost always larger than the poorer of thetwo unique components.

An interesting detail about the shared predictive variance between kindergartenNaming Speed and Phonological Awareness is that the commonality elements ap-pear to decline over the years; in other words, the two tasks predict later readingperformance in more distinct ways than they predict early reading performance.These results are in line with suggestions that orthographic strategies emerge to


Do Not

Cop

y

supplement alphabetic strategies during the early years of schooling (e.g., Ehri &McCormick, 1998).

Results from Grade 1 commonality analyses, however, suggest caution withthese interpretations. As in kindergarten, the elements common to all tasks do notaccount for a large proportion of the reading variance. In contrast to kindergarten,Phonological Awareness is now clearly the most useful predictor of reading vari-ance and the elements common to Phonological Awareness and Naming Speed areas large as, or larger than, the elements unique to Naming Speed. This would seemto suggest that when measured in Grade 1, what Naming Speed shares with Phono-logical Awareness is equally or more important as a predictor of reading variancethan what is unique to it. Why Grade 1 Naming Speed shares more of its predictivevariance with Phonological Awareness than does kindergarten Naming Speed isnot clear to us. Correlations between the two variables do not change significantly(from –.36 in kindergarten to –.42 in Grade 1; see Table 2) and stability of bothmeasures is relatively high at .74 and .68, casting doubt on any explanation ofqualitative change in what the tasks measure. The unexpected growth of phonolog-ical awareness could be related to early reading development, but the same expla-nation does not seem to hold for naming speed, as Grade 1 Naming Speed does notcorrelate any better with concurrent or later reading skills than kindergartenNaming Speed.

Finally, we examined whether the predictive power of phonological processingmeasures administered in kindergarten is equal to that of the same measures ad-ministered in Grade 1. As in the previous studies, phonological processing mea-sures were highly stable from kindergarten to Grade 1. Our results, with theexception of the Grade 1 analyses, indicated that Grade 1 predictors accounted formore, but not significantly more, variance in reading measures than did kindergar-ten predictors. Why the opposite is true for Grade 1 is not clear. It is possible thatsmall changes in distributional properties of the key tasks could account for someof the differences, although we attempted to negate such a possibility in advanceby transforming the problematic scores as much as possible. However, kindergar-ten Phonological Awareness (but not Grade 1 Phonological Awareness), Grade 1Word Identification and Grade 1 Passage Comprehension all exhibited a smallfloor effect that may have affected the results.

Some limitations of this study are worth mentioning here. First, we did not in-clude any measures of general cognitive ability, as has been done in many previousstudies. We previously argued that controlling for general ability can be ill advisedunless the relationship of the chosen general ability measure to other included pre-dictors is well examined and understood. Our argument would naturally be stron-ger if we could demonstrate the issue further with this sample. We have alsoavoided references to causality, which would require the examination of the effectof general ability in more detail. Second, we used observed variables as opposed tolatent variables. Contributions of different predictor variables on reading measures


Do Not

Cop

y

thus reflect both specific method effects as well as effects of the constructs thatthese measures are representing. In principle, it is possible that the unique contri-bution of each of these variables reflects more the former than the latter. This is notlikely, however, as our findings mostly replicate and expand on findings of exist-ing studies that have used both observed and latent variables.

Finally, our measure of reading comprehension was highly correlated withword reading. More complex and complete measures of comprehension wouldhave likely lead to smaller R2 values and perhaps to smaller correlations betweenword reading or phonological processes and reading comprehension.

ACKNOWLEDGMENTS

An earlier version of this article was presented at the annual meeting of the Societyfor the Scientific Study of Reading in Stockholm, July 2000. This research was sup-ported by a Social Sciences and Humanities Research Council of Canada grant to J.R. Kirby and J. P. Das, and a postdoctoral fellowship to R. Parrila.

We thank anonymous reviewers and the editor for their very helpful sugges-tions, and Richard Wagner and J. P. Das for making some of the tests available.

REFERENCES

Ackerman, P. T., Dykman, R. A, & Gardner, M. Y. (1990). Counting rate, naming rate, phonologicalsensitivity, and memory span: Major factors in dyslexia. Journal of Learning Disabilities, 23,325–327, 319.

Adams, M. J. (1990). Beginning to read. Cambridge, MA: The MIT Press.Avons, S. E., & Hanna, C. (1995). The memory-span deficit in children with specific reading disability:

Is speech rate responsible? British Journal of Developmental Psychology, 13, 303–311.Baddeley, A., Gathercole, S. E., & Papagno, C. (1998). The phonological loop as a language learning de-

vice. Psychological Review, 105, 158–173.Badian, N. (1994). Preschool prediction: Orthographic and phonological skills, and reading. Annals of

Dyslexia, 44, 3–25.Blackmore, A. M., & Pratt, C. (1997). Grammatical awareness and reading in Grade 1 children.

Merrill-Palmer Quarterly, 43, 567–590.Bowers, P. G. (1995). Tracing symbol naming speed’s unique contributions to reading disabilities over

time. Reading and Writing, 7, 189–216.Bradley, L., & Bryant, P. (1985). Rhyme and reason in reading and spelling. Ann Arbor: University of

Michigan Press.Burgess, S. R., & Lonigan, C. J. (1998). Bidirectional relations of phonological sensitivity and

prereading abilities: Evidence from a preschool sample. Journal of Experimental Child Psychology,70, 117–141.

Catts, H. W. (1989). Speech production deficits in developmental dyslexia. Journal of Speech and Hear-ing Disorders, 54, 422–428.

Clay, M. (1993). An observation survey of early literacy achievement. Portsmouth, NH: Heinemann.Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the behavioral sci-

ences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.


Do Not

Cop

y

Das, J. P., Mensink, D., & Mishra, R. K. (1990). Cognitive processes separating good and poor readerswhen IQ is covaried. Learning and Individual Differences, 2, 423–436.

Das, J. P., & Mishra, R. K. (1991). Relation between memory span, naming time, speech rate, and read-ing competence. Journal of Experimental Education, 59, 129–139.

Das, J. P., Mok, M., & Mishra, R. K. (1994). The role of speech processes and memory in reading dis-ability. Journal of General Psychology, 121, 131–146.

de Jong, P. F., & van der Leij, A. (1999). Specific contributions of phonological abilities to early readingacquisition: Results from a Dutch latent variable longitudinal study. Journal of Educational Psy-chology, 91, 450– 476.

de Jong, P. F., & van der Leij, A. (2002). Effects of phonological abilities and linguistic comprehensionon the development of reading. Scientific Studies of Reading, 6, 51–78.

Ehri, L. C., & McCormick, S. (1998). Phases of word learning: Implications for instruction with delayedand disabled readers. Reading and Writing Quarterly: Overcoming Learning Difficulties, 14,135–164.

Elbro, C. (1996). Early linguistic abilities and reading development: A review and a hypothesis. Read-ing and Writing, 8, 453–485.

Fowler, A. (1991). How early phonological development might set the stage for phoneme awareness. InS. Brady & D. Shankweiler (Eds.), Phonological processes in literacy: A tribute to Isabelle Y.Liberman (pp. 97–117). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

Gallagher, A. M., Laxon, V., Armstrong, E., & Firth, U. (1996). Phonological difficulties in high-func-tioning dyslexics. Reading and Writing, 8, 499–509.

Goswami, U. (2000). Phonological and lexical processes. In M. L. Kamil, P. B. Mosenthal, & P. D.Pearson (Eds.), Handbook of reading research (pp. 251–267). Mahwah, NJ: Lawrence Erlbaum As-sociates, Inc.

Hansen, J., & Bowey, J. A. (1994). Phonological analysis skills, verbal working memory, and readingability in second-grade children. Child Development, 65, 938–950.

Hecht, S. A., Burgess, S. R., Torgesen, J. K., Wagner, R. K., & Rashotte, C. A. (2000). Explaining socialclass differences in growth of reading skills from beginning kindergarten through fourth-grade: Therole of phonological awareness, rate of access, and print knowledge. Reading and Writing, 12,99–127.

Hulme, C., & Roodenrys, S. (1995). Verbal working memory development and its disorders. Journal ofChild Psychology and Psychiatry, 36, 373–398.

Kirby, J. R., & Parrila, R. K. (1999). Theory-based prediction of early reading. Alberta Journal of Edu-cational Research, 45, 428–447.

Lonigan, C. J., Burgess, S. R., & Anthony, J. L. (2000). Development of emergent literacy and earlyreading skills in preschool children: Evidence from a latent-variable longitudinal study. Develop-mental Psychology, 36, 596–613.

Lundberg, I., Olofsson, A., & Wall, S. (1980). Reading and spelling skills in the first school years pre-dicted from phonemic awareness skills in kindergarten. Scandinavian Journal of Psychology, 21,159–173.

MacDonald, G. W., & Cornwall, A. (1995). The relationship between phonological awareness and read-ing and spelling achievement eleven years later. Journal of Learning Disabilities, 28, 523–527.

Maclean, M., Bryant, P., & Bradley, L. (1987). Rhymes, nursery rhymes, and reading in early child-hood. Merrill- Palmer Quarterly, 33, 255–281.

Manis, F. R., Seidenberg, M. S., & Doi, L. M. (1999). See Dick RAN: Rapid naming and the longitudinalprediction of reading subskills in first and second graders. Scientific Studies of Reading, 3, 129–157.

Mann, V.A. (1993). Phoneme awareness and future reading ability. Journal of Learning Disabilities,26, 259–269.

McDougall, S., Hulme, C., Ellis, A., & Monk, A. (1994). Learning to read: The role of short-term mem-ory and phonological skills. Journal of Experimental Child Psychology, 58, 112–133.


Do Not

Cop

y

Metsala, J. (1999). Young children’s phonological awareness and nonword repetition as a function ofvocabulary development. Journal of Educational Psychology, 91, 3–19.

Muter, V., & Snowling, M. (1998). Concurrent and longitudinal predictors of reading: The role ofmetalinguistic and short-term memory skills. Reading Research Quarterly, 33, 320–337.

Naglieri, J. A., & Das, J. P. (1997). Das-Naglieri Cognitive Assessment System. Chicago, IL: RiversidePublishing.

National Reading Panel. (1999). Teaching children to read: An evidence-based assessment of the scien-tific research literature on reading and its implications for reading instruction. Baltimore, MD: Na-tional Institute of Child Health and Human Development.

Neuhaus, G. F., & Swank, P. R. (2002). Understanding the relations between RAN letter subtest compo-nents and word reading in first-grade students. Journal of Learning Disabilities, 35, 158–174.

Pedhazur, E. J. (1982). Multiple regression in behavioural research: Explanation and prediction (2nded.). New York: CBS College Publishing.

Perfetti, C. A., Beck, I., Bell, L. C., & Hughes, C. (1987). Phonemic knowledge and learning to read arereciprocal: A longitudinal study of first grade children. Merrill-Palmer Quarterly, 33, 283–319.

Rohl, M., & Pratt, C. (1995). Phonological awareness, verbal working memory and the acquisition of lit-eracy. Reading and Writing, 7, 327–360.

Snow, C. E., Burns, S. M., & Griffin, P. (Eds.). (1998). Preventing reading difficulties in young children.Washington, DC: National Academy Press.

Tabachnick, B. G., & Fidell, L. S. (1996). Using multivariate statistics (3rd ed.). New York:HarperCollins.

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

Wagner, R. K., Torgesen, J. K., Laughon, P., Simmons, K., & Rashotte, C. A. (1993). Development ofyoung readers’ phonological processing abilities. Journal of Educational Psychology, 85, 83–103.

Wagner, R. K., Torgesen, J. K., & Rashotte, C. A. (1994). Development of reading related phonologicalprocessing abilities: New evidence of bi-directional causality from a latent variable longitudinalstudy. Developmental Psychology, 30, 73–87.

Wagner, R. K., Torgesen, J. K., Rashotte, C. A., Hecht, S.A., Barker, T. A., Burgess, S. P., et al. (1997).Changing relations between phonological processing abilities and word-level reading as childrendevelop from beginning to skilled readers: A 5-year longitudinal study. Developmental Psychology,33, 468–479.

Wimmer, H., Landerl, K., Linortner, R., & Hummer, P. (1991). The relationship of phonemic awarenessto reading acquisition: More consequence than precondition but still important. Cognition, 40,219–249.

Wimmer, H., Mayringer, H., & Landerl, K., (2000). The double-deficit hypothesis and difficulties inlearning to read a regular orthography. Journal of Educational Psychology, 92, 668–680.

Wolf, M., & Bowers, P. G. (1999). The double-deficit hypothesis for the developmental dyslexias. Jour-nal of Educational Psychology, 91, 415–438.

Woodcock, R. W. (1987). Woodcock Reading Mastery Test–Revised. Circle Pines, MN: AmericanGuidance Service.

Manuscript received May 4, 2001Accepted April 30, 2003


Do Not

Cop

y

Articulation rate, naming speed, verbal short-term memory, and phonological awareness: Longitudinal predictors of early reading development?

Documents