The Role of Phonological Awareness and Rapid Automatized Naming in the Prediction of Reading Dif fi culties in Portuguese

This study investigated the contribution of phonological awareness (PA) and rapid automatized naming (RAN) to the persistence of beginning reading diffi culties in Portuguese. The study was longitudinal and involved 13 children who had shown little progress in learning to read during their 1st school year, and 13 children with typical reading development. At the beginning of the study, when participants were, on average, six years and nine months of age, their ability to read and spell words, and their PA and RAN skills were assessed. One year later, the children who had shown diffi culties in learning to read were divided into two groups, based on their progress in learning to read during the course of the study: a group with persistent reading diffi culties and a group whose diffi culties had been overcome. Results showed that both groups performed equally poorly and below controls on the PA tasks at the beginning of the study. In contrast, only the children with persistent reading diffi culties performed more poorly than controls on the RAN tasks at that time.

Developmental dyslexia is a neurodevelopmental learning disorder characterized by diffi culties in the ability to read words with accuracy and/or fl uency, despite appropriate literacy instruction and normal or above normal intelligence.It is one of the most frequent developmental learning disorders, affecting approximately 9% of the school age population (e.g., Pennington, 2009).
According to the phonological defi cit model (Pennington, 2009), developmental dyslexia results from defi cits in phonological processing, most notably in phonological awareness, that is, the ability to pay conscious attention to the sounds that comprise the words we hear and speak.These defi cits compromise the ability to recode letters or letter sequences into the sounds they typically represent in words, an ability that is fundamental for success in learning to read (Cardoso-Martins & Ehri, 2014).In support of this hypothesis, there is evidence that children with dyslexia fi nd it particularly diffi cult to read pseudowords, that is, words that do not exist in their language and therefore can only be read by phonological recoding.
Despite its prominent position in the literature, the phonological defi cit model has recently become the subject of increasing scrutiny (Snowling, 2008).In particular, it has been suggested that phonological processing defi cits are not suffi cient (e.g., Pennington, 2006;Snowling, 2008) or even necessary (e.g., Pennington, 2006;Wolf & Bowers, 1999) for diffi culties in learning to read to occur.For example, Snowling, Gallagher and Frith (2003) have noted that, notwithstanding their diffi culties in phonological awareness tasks, not all children at high familial risk of developmental dyslexia present with diffi culties in learning to read.The same is also true of children with phonological defi cit disorder, a condition frequently associated with defi cits in phonological awareness (Pennington & Bishop, 2009), suggesting that diffi culties in phonological processing can be mitigated or compensated by other processes.One moderating factor that seems particularly important, and the one that is the focus of the present study, is rapid (or automatized) naming of series of familiar stimuli such as letters, colors or pictures of objects (e.g., Peterson & Pennington, 2012).
Since the seminal work of Denckla and Rudel (1976), a large body of research has demonstrated a strong relationship between rapid automatized naming (RAN) and both concurrent and future literacy skills (see Kirby, Georgiou, Martinussen, &Parrila, 2010, andNorton &Wolf, 2012, for reviews).This relationship has been shown in practically every language in which it has been tested, including Portuguese (e.g., Albuquerque, 2012;Justi & Roazzi, 2012), and regardless of variations in other predictors of literacy ability, such as verbal and non-verbal skills, general processing speed, and phonological awareness (Kirby et al., 2010).Nonetheless, there has been much disagreement regarding the explanation underlying this relationship.According to one theoretical standpoint (e.g., Wagner & Torgesen, 1987), RAN predicts reading ability because, comparable to phonological awareness (PA), it is an index of phonological processing ability.On the other hand, Wolf and her colleagues (e.g., Bowers & Wolf, 1993;Wolf & Bowers 1999) have suggested that RAN indexes processes that are, at least in part, independent of phonology, including attention, visual discrimination, integration of visual information with stored phonological and semantic representations, and access and retrieval of phonological labels.In support of their view, they have shown that defi cits in RAN and PA often dissociate, and that defi cits in RAN can predict literacy diffi culties even in the absence of defi cits in PA.Perhaps more importantly for the purpose of the present study, there is evidence that defi cits in both RAN and PA are often associated with more severe and persistent literacy diffi culties than isolated defi cits in RAN or PA (Kirby et al., 2010).This may be particularly true of orthographies more consistent than the English orthography.Indeed, given the claim that PA plays a relatively modest role in the prediction of later literacy skills in more consistent orthographies (e.g., de Jong & van der Leij, 1999;Papadopoulos, Georgiou, & Kendeou, 2009;Wimmer, Mayringer, & Landerl, 2000), it is unlikely that isolated defi cits in PA predict severe and persistent reading diffi culties in these orthographies.
The fi ndings reviewed above suggest that, when used in conjunction with PA, RAN may be a particularly useful tool in the prognosis of children who experience diffi culties in the beginning phases of reading acquisition.In the present study we investigated this hypothesis in a group of children who were learning to read in a relatively consistent orthography -Brazilian Portuguese -and who had demonstrated diffi culties in learning to read during their fi rst school year.Children's reading and spelling ability was evaluated twice, at the end of their fi rst and second school years.Both PA and RAN were assessed at the beginning of the study, that is, at the end of children's fi rst school year.Based on the evidence reviewed above, we predicted that skills in both RAN and PA would better distinguish the children who overcame their reading diffi culties over the course of the study from those who did not than skills in PA alone.

Participants
The sample was initially comprised of two groups of 6-to 7-year old children (M = 7 years and 1 month; SD = 3.65 months) enrolled in the former "pré-primário" (kindergarten) of private schools in a large Brazilian city: 13 children with beginning reading diffi culties and 13 children without reading diffi culties, individually matched for age, gender and kindergarten class.They all had normal or above normal intelligence as assessed by the Brazilian version of the Wechsler Intelligence Scales for Children--III (WISC-III; see below).To be assigned to one of the two groups, children had to meet two criteria.First, they had to be nominated by their teachers as having or having not manifested diffi culties in learning to read throughout the school year.In all cases, teachers' indication occurred towards the end of the school year, approximately 5 to 6 months after the beginning of reading instruction.In addition, the children had to have scored in the middle-inferior to inferior range, in the case of the children with reading diffi culties, or in the middle-superior to superior range, in the case of the children without reading diffi culties, on the reading subtest of a Brazilian test of academic performance -TDE (see below) -that was administered at the end of their kindergarten year.Among the children indicated by the teachers, only one did not meet both criteria and could not be included in the study.As illustrated in Table 1, the children assigned to the group with beginning reading diffi culties performed signifi cantly worse than the children assigned to the group of children without reading diffi culties (henceforth the control group) on that subtest.They also performed signifi cantly below the control group on the spelling subtest of the same test and on an experimental pseudoword reading task that were also administered at the end of kindergarten.One year later, at the end of 1 st grade, the children were re-tested on the reading and spelling subtests of the TDE.Results showed that the children in the group with beginning reading diffi culties varied largely with regard to their ability to read words.In view of this, they were divided into two groups: the group with persistent reading diffi culties, that is, the 6 children whose scores on the reading test were at least one SD below the mean score of the control group, and the group of children with transient reading diffi culties (N = 7), whose scores on the reading test were above that cut-off point.As illustrated in Table 2, the children with persistent reading diffi culties performed signifi cantly worse than the children with transient reading diffi culties and controls on both the reading and the spelling subtests of TDE at the end of 1 st grade.No differences were found between the latter two groups at this occasion.

Procedure
Participants were evaluated at two different times with an interval of approximately 12 months: Time 1 occurred towards the end of kindergarten, and Time 2 at the end of 1 st grade.At both Time 1 and Time 2, children were administered the reading and spelling subtests of the TDE (Stein, 1994).At Time 1, they were also administered the WISC-III (Wechsler, 2002), and tests of PA, RAN and pseudoword reading ability.Except for the WISC--III, which was administered at a private clinic or at the child's home, all tests were administered in a quiet place at the children's schools.The study was approved by the Ethics Committee of the Universidade Federal de Minas Gerais and, in all cases, the children's parents signed a written consent form allowing their children to participate in the study.

Instruments
Word Reading and Spelling.The ability to read and spell words was assessed with the corresponding subtests of TDE (Stein, 1994).The reading subtest comprises 70 words printed in lower case letters in a card, in order of increasing diffi culty.Children are asked to read aloud the words at their own pace.Thirty-fi ve words are included in the spelling subtest.The words are dictated in order of increasing diffi culty, and the test is discontinued after 10 incorrect responses in a row (or after completion of the subtest).Reliability (alpha) is reported to be .86and .90, for the reading and spelling subtests, respectively.At the end of kindergarten, children were also asked to read a set of 20 pseudowords (e.g.bude, calvilho, etc.) printed in capital letters at the center of individual cards.
Phonemic Awareness.Two experimental tasks were used.In the phoneme detection task, children were asked to identify among three different words, the two that started with the same sound.The task is divided into two parts, each one of which is comprised of two or three training trials, followed by 10 test trials.In the fi rst part, all words begin with a CV or CVC syllable (e.g., sorvete 'ice-cream'-machado 'axe' -cigarro 'cigarette'); in the second part, two of the words in a trial begin with a consonant cluster (e.g., praça 'square'-coroa 'crown'criança 'child').For each trial, the examiner showed a card with drawings of the words' referents, and pronounced the words out loud.The number of correct responses was scored.
In the phoneme deletion task, the child is asked to repeat a word enunciated by the examiner after deleting its fi rst sound (e.g., gato 'cat': ato; fl anela 'fl annel': lanela).The task has 10 training trials and 20 tests trials.The number of correct responses to the test trials was scored.Performance on the two phonemic awareness tasks was signifi cantly correlated (r = .54,p < .01).A PA composite score consisting of the corresponding summed z-scores divided by two was thus derived and used in the analyses comparing the three groups of participants.
Rapid Serial Naming was measured with a version of the RAN task (Denckla & Rudel, 1976).The participant was asked to name as fast as possible series of repeating familiar stimuli printed on a chart, in the form of a matrix consisting of fi ve rows and 10 columns.There were four different charts, each comprised of a different type of stimulus (letters, digits, objects, and colors).In each chart, fi ve different stimuli were presented 10 times each in a random order.Responses were timed using a stopwatch.The time taken to name all the stimuli was scored.Students' scores on the RAN tasks were signifi cantly correlated (rs ranging from .51 to .77).
In view of this, a composite score consisting of the mean of the summed corresponding z-scores was used in the analyses described below.
Intelligence.Intelligence was measured with the Brazilian version of the WISC-III (Wechsler, 2002).Reliability (alpha) for full IQ is reported to be .93.

Results
The aim of this study was to investigate the role played by PA and RAN in predicting persistent, as opposed to transient, reading diffi culties in Brazilian Portuguese.Table 2 lists the mean composite scores for the PA and RAN tasks, along with the mean scores for the other tests that were administered at Time 1, separately for the three groups of children (Given the small number of participants, whenever possible, non-parametric tests were used to test the signifi cance of the differences found between the groups.Specifi cally, the Kruskal-Wallis test was used to evaluate the signifi cance of the overall group difference, and the Mann-Whitney test was used to evaluate the signifi cance of any pairwise differences).
At Time 1, both groups of children with beginning reading diffi culties performed signifi cantly worse than controls on the pseudoword reading task and on both the reading and the spelling subtest of the TDE (Persistent vs. Controls: pseudoword reading: p < .001,Cohen's d = 3.34; TDE reading: p < .001,d = 5.18; TDE spelling: p < .001,d = 2.62; Transient vs. Controls: pseudoword reading: p < .001,d = 3.03; TDE reading: p < .001,d = 3.78; TDE spelling: p < .01,d = 1.59).As illustrated in Table 2, the group with persistent reading diffi culties also performed worse than the group with transient reading diffi culties on both the reading and spelling subtests of the TDE at Time 1.The difference between these two groups on the spelling subtest was statistically signifi cant (p < .05,d = 1.74).These results suggest that the children with persistent reading diffi culties were already poorer readers than the children with transient reading diffi culties at Time 1, and this might be the reason why their reading diffi culties were relatively more persistent.However, it is important to note that, except for two children, the scores of the children with transient diffi culties on the reading subtest of the TDE at Time 1 fell well within the range of the scores of the group of children with persistent reading diffi culties.In addition, as illustrated in Table 2, no difference was found between these two groups with regard to the hallmark of alphabetic literacy acquisition, namely, the ability to read by recoding letters or groups of letters into the sounds they represent in words.
The comparisons involving the PA and RAN measures showed a signifi cant effect of group (p < .001,for PA and p < .01,for RAN).Pairwise comparisons revealed that both the children with transient reading diffi culties and the children with persistent reading diffi culties performed equally poorly and signifi cantly worse than controls on the PA tasks administered at the beginning of the study (Persistent vs. Controls: p < .001,d = 2.04; Transient vs. Controls: p < .01,d = 2.03).In marked contrast, only the children with persistent reading diffi culties performed signifi cantly worse than controls on the RAN measure at that time (p < .001,d = 1.29).As illustrated in Table 2, their scores were also signifi cantly poorer than those of the group with transient reading diffi culties (p < .01,d = 1.62).Similar results were obtained when the two children in the group of transient reading diffi culties who had performed relatively well on the word reading subtest of the TDE at Time 1 (see above) were excluded from these analyses.
Although our three groups of participants did not in general differ signifi cantly with regard to IQ, there was a trend for both groups with beginning reading diffi culties to score below controls on the WISC-III.In view of this, two ANCOVAs controlling for individual differences in full IQ were conducted, one for PA and the other for RAN.Similar to the results reported above, the group factor was signifi cant for both dependent variables, F (2, 22) = 6.57, p < .01,= .37,for PA and F (2, 22) = 4.70, p < .05,= .30,for RAN.Furthermore, contrasts with the control group as the reference category revealed that both the children with persistent diffi culties and the children with transient diffi culties signifi cantly underperformed controls on the PA measure, both ps < .01.On the other hand, similar to what we had found before, while the group with persistent reading diffi culties performed signifi cantly below controls on the RAN measure (p < .05), the children with transient reading diffi culties performed as well as controls.

Discussion
In the present study, we examined the role played by PA and RAN in the prediction of literacy skills of children experiencing diffi culties in the beginning of reading acquisition in a relatively consistent orthography -Brazilian Portuguese.We predicted that defi cits in both PA and RAN, rather than deficits in PA alone, would be associated with persistent reading diffi culties in our sample.As mentioned previously, Pennington (2006) has cogently argued that isolated defi cits in PA (or in any other cognitive skill for that matter) may not be suffi cient to explain reading diffi culties.In addition, it has been suggested that PA plays a relatively modest role in the prediction of literacy skills in more consistent orthographies (e.g., de Jong & van der Leij, 1999;Papadopoulos et al., 2009;Wimmer et al., 2000; but see Caravolas, Volín, & Hulme, 2005, for a different view).In line with our prediction, although both groups of children with beginning reading diffi culties had defi cits in PA at the beginning of the study, not all of them showed word reading and spelling diffi culties one year later.Our results suggest that, in addition to defi cits in PA, children with persistent reading diffi culties in Brazilian Portuguese are likely to present defi cits in rapid naming at the beginning of literacy acquisition.As described previously, only the children with persistent reading diffi culties performed significantly worse than controls on the RAN tasks administered at the end of kindergarten.Indeed, while none of the children with transient reading diffi culties scored above the cut-off of one standard deviation above the mean reaction time of the control children on the RAN tasks, as many as four of the six children with persistent reading diffi culties did.
It is unlikely that the differences observed between the children with persistent reading diffi culties and those whose reading diffi culties were transient resulted from differences in their intelligence.In fact, the two groups did not differ signifi cantly on either Verbal or Performance IQ, as measured by the WISC-III (Wechsler, 2002).Nor did they seem to differ with regard to learning ability more generally or to the presence of other developmental disorders (see, e.g., Michalick, 2005, for a detailed description of the present sample).
As noted previously, some researchers have argued that RAN is best viewed as a measure of phonological processing skill (Wagner & Torgesen, 1987).In this case, the presence of RAN defi cits in children with persistent reading diffi culties could simply be a sign of more severe phonological diffi culties which, arguably, would place a greater toll on the ability to learn to read by processing and remembering letter-sound relations in words, a foundational skill in alphabetic reading acquisition (e.g., Cardoso-Martins & Ehri, 2014).In line with this, the children with persistent reading diffi culties performed signifi cantly worse than those with transient diffi culties on the digit span subtest of the WISC-III (Mean raw score = 6.17 and 8.86, respectively, p < .05), a task that assesses verbal short-term memory.However, the two groups with beginning reading diffi culties performed equally poorly and signifi cantly below controls on measures that clearly rely on phonological processing skills, namely, the PA and the pseudoword reading tasks.Furthermore, similar to the results of other studies (Wolf & Bowers, 1999), the correlation between RAN and PA (r = .41,p < .05)or between RAN and the digit span subtest of the WISC-III (r = -.43,p < .05)for the whole group was very modest, suggesting that RAN and PA should not be considered as two measures of the same underlying construct (Wolf & Bowers, 1999).Last but not least, there is evidence that children with defi cits in both PA and RAN show more severe and persistent reading diffi culties than children with defi cits only in PA, even when the two groups of children are matched for PA (e.g., Kirby, Parrila, & Pfeiffer, 2003;Papadopoulos et al., 2009).This was certainly the case in the present study.
Taken together, our results suggest that defi cits in RAN constitute an important risk factor in the development of persistent reading diffi culties in Portuguese, at least when associated with defi cits in PA.Our results should however be interpreted with caution.This study comprised a small sample of children, all of them coming from high socioeconomic status family.In addition, the children were followed for only one year.Clearly, more studies are necessary to evaluate the generality of the present fi ndings to a more representative sample of children at risk of developing reading disabilities.In case our results are confi rmed, it will be important to investigate the cognitive factors behind RAN.As mentioned in the introduction, there is still no consensus regarding the nature of these factors and how they contribute to the development of reading.
Note.WISC-III = Wechsler Intelligence Scales for Children-III; TDE = Test of Academic Performance; PA = Phoneme awareness; RAN = Rapid automatized naming; * composite z scores.a cell differs signifi cantly from cells with superscripts b or c; b cell differs signifi cantly from cells with superscripts a or c; c cell differs signifi cantly from cells with superscripts a or b.

Table 1
Mean Age and Mean Scores (and Standard Deviations) on the Literacy Tests, as a Function of Group

Table 2
Mean Age and Mean Scores (and Standard Deviations) on the Literacy and Intelligence Tests, as a Function of Group and Time