The Relationship between Paired Associate Learning and Phonological Skills in Normally Developing Readers

Journal of Experimental Child Psychology 80, 160–173 (2001) doi:10.1006/jecp.2000.2625, available online at http://www.idealibrary.com on

The Relationship between Paired Associate Learning and Phonological Skills in Normally Developing Readers Kirsten L. Windfuhr University of Manchester, Manchester, United Kingdom

and Margaret J. Snowling University of York, York, United Kingdom Seventy-five 6- to 11-year-old children were administered tests of phonological awareness, verbal short term memory (STM), and visual–verbal paired associate learning (PA learning) to investigate their relationship with word recognition and decoding skills. Phonological awareness was a stronger concurrent predictor of word recognition than verbal STM, and phonological awareness but not verbal STM was a predictor of learning in the PA learning task. Importantly, measures of phonological awareness and PA learning both accounted for independent variance in word reading, even when decoding skill was controlled. The results suggest that PA learning and phonological awareness tasks tap two separate mechanisms involved in learning to read. The results are discussed in relation to current theories of reading development. © 2001 Academic Press Key Words: paired associate learning; verbal short-term memory; phonological awareness; connectionist models; reading development.

It is widely accepted that phonological awareness is one of the best predictors of reading ability (e.g., Bradley & Bryant, 1983; Hoien, Lundberg, Stanovich, & Bjalid, 1995; Lundberg, Olofsson, & Wall, 1980; Muter, Hulme, Snowling, & Taylor, 1997; Wagner, Torgesen, & Rashotte, 1994; Yopp, 1988). Indeed, phonological awareness is viewed as a necessary requirement for setting up an efficient reading system. Children who acquire the skill to reflect upon and explicitly manipulate the constituent speech sounds of language are capable of “cracking the alphabetic code” involved in reading (Byrne, 1996). The corollary of this is This research was conducted as part of a doctoral dissertation by the first author at the University of York, UK. Thanks are due to the children and staff from Woodthorpe Primary School in York, England. We also thank Joe Torgesen and Anne Castles for insightful reviews of a previous version of the manuscript. Address correspondence and reprint requests to Kirsten Windfuhr, Human Communication and Deafness, School of Education, University of Manchester, Oxford Rd., Manchester, M13 9PL, UK. E-mail: [email protected]. 160 0022-0965/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved.

LEARNING AND PHONOLOGICAL SKILLS

161

that children with poor phonological skills have difficulty learning to read (Bradley & Bryant, 1983; Elbro, 1996; Swan & Goswami, 1997). A related body of work has emphasized the distinction between phonological processing abilities and phonological awareness (Hulme & Snowling, 1992; Wagner et al., 1994). Within this view, cognitive tasks such as verbal short-term memory, confrontation, and serial naming tasks (RAN) tap underlying phonological representations directly, whereas phonological awareness requires conscious metalinguistic awareness of the phonological forms of words. To the extent that learning to read requires awareness of phonemes, it depends on the child’s language processing system having reached a stage of development when phonological representations are segmental in form (Fowler, 1991). However, in recent years, the undoubted importance of phonological skills to the acquisition of the alphabetic principle has detracted attention from the fact that reading acquisition involves a substantial amount of learning. Indeed, orthographic learning can be conceptualized as a specific instance of visual–verbal paired associate learning in which the attributes of printed words (their graphemic features) must be associated with their phonological forms (Hulme, 1981; Manis et al., 1987). More fundamentally, paired associate learning is required to establish letter-sound and letter-name knowledge, both of which are strong predictors of reading attainment (Adams, 1989; Bond & Dykstra, 1967; Muter et al., 1997). Successful learning of each of these aspects of literacy development depends upon being able to activate and maintain two temporary representations, a visual and a verbal one, and to form a new association in long-term memory. Although the learning process is implicit in accounts of reading development, models of reading vary in the extent to which they address it. In the dual-route model (Coltheart, 1978; Coltheart et al., 1993) it is reasonable to suppose that learning underlies the operation of the direct route used for reading highly familiar and exception words. From the perspective of reading development, Share (1995) proposed that the phonological analysis of unfamiliar words was a prerequisite to the establishment of orthographic representations. However, both of these models, and other “localist” accounts that incorporate the notion of word-specific lexical nodes (Besner, 1999), leave the learning mechanism under-specified. In contrast, a learning algorithm renders the process of acquisition computationally explicit in current connectionist models. According to such models, reading development is not conceptualized as the accumulation of word-specific lexical entries as in localist models, but as the gradual development of more complete patterns of activation distributed across orthographic input and phonological output units (Seidenberg & McClelland, 1989). Learning gradually alters the weights of the connections between these representations according to the frequency of cooccurrence of graphemic and phonemic sequences. As a consequence, the network abstracts the statistical regularities of orthography–phonology mappings such that it learns to read regular words with consistent mappings more efficiently

162

WINDFUHR AND SNOWLING

than words containing inconsistent mappings, such as exception words. Furthermore, Harm and Seidenberg (1999) showed that manipulating the learning parameter to produce a nonoptimal learning rate, had a large effect on exception word reading and a smaller but nonetheless significant effect on the development of nonword reading. In light of these observations, it is noteworthy that several studies have reported that dyslexic readers have difficulties associating verbal labels with visual stimuli (Castles & Holmes, 1996; Manis, 1985; Vellutino, Scanlon, & Spearing, 1995; Vellutino, Steger, Harding, & Phillips, 1975). Moreover, there is some evidence that this paired associate learning deficit could produce effects on learning to read that are independent of those that follow from problems of phonological awareness. Wimmer, Mayringer, and Landerl (1998) showed that dyslexic readers of German (a transparent orthography) who do not show the poor performance on tests of phonological awareness that is typical of dyslexic readers of English (an opaque orthography) had more difficulty learning to associate pictures of animals with nonsense names than normal readers (see also Mayringer & Wimmer, 2000). They interpreted these findings as suggesting that paired associate learning ability taps the ability to establish wordspecific orthographic identities for words (cf. Ehri, 1992). In turn, this learning difficulty might prevent the development of automaticity in reading, even in children who can decode accurately (cf. Manis et al., 1999; Wolf & Bowers, 1999). An alternative possibility is that effective visual–verbal paired associate learning, just like reading, depends upon the ability to segment the speech stream into phonemic units. De Jong, Seveke, and van Veen (2000) found that the performance of Dutch kindergarten children on a test of phoneme awareness was a strong predictor of their ability to learn new names. In a second experiment, training in phonological awareness improved novel word learning ability, perhaps because it directed the children’s attention to the phonological structure of the words to be learned. These findings suggest that the relationship between paired associate learning and reading might be explained by shared phonological processes. If this is correct, then paired associate learning and phonological awareness should influence the development of reading in similar ways. The main aim of the present study was to clarify the relationship between phonological skills, paired associate learning, and reading ability in normally developing readers. A battery of tasks was administered comprising tests of verbal short-term memory (implicit phonological processing), phonological awareness (explicit phonological processing), word reading, and nonword decoding skill. It was hypothesized that, if paired associate learning taps the creation of word-specific associations, it should be a better predictor of individual differences in word reading than of decoding skill. Second, it should account for independent variance in word reading, after decoding skill is controlled. If, on the other hand, the relationship between paired associate learning and reading is due to shared variance with phonological awareness, then its contribution to

LEARNING AND PHONOLOGICAL SKILLS

163

individual differences in reading should be subsumed by measures of phonological awareness. A related aim of the study was to investigate the predictors of paired associate learning. Following de Jong et al. (2000), we hypothesized that phonological awareness and paired associate learning would be significantly related, consistent with the notion that awareness of the phonological structure of spoken word facilitates the creation of new phonological representations. We also investigated the extent to which paired associate learning depends upon the ability to store verbal items in a temporary form, as indexed by short-term memory measures. STUDY 1 Method Participants The study involved 75 children from five classes that spanned the school grades of two through six in the city of York, England. The children ranged in age from 7 years, 1 month to 11 years, 11 months and in reading age from 6 years, 3 months to 15 years, as measured by the Wechsler Objective Reading Dimensions Basic Reading subtest, a test of single word recognition (WORD; Wechsler, 1993) (see Table 1). Design and Materials Reading skills. There were two tests of reading, one comprising words, the other nonwords. The WORD Basic Reading is an untimed test of single word recognition (Wechsler, 1993). The Graded Nonword Reading Test is a standardized test comprising 10 one-syllable and 10 two-syllable nonwords to be decoded without imposing a time limit (Snowling, Stothard, & McLean, 1996). Phonological awareness. There were two tests of phonological awareness. In the phoneme deletion task, the children were required to delete a sound from the initial, medial, or final position of a set of 24 nonwords, for example, barp, kelm (after McDougall, Hulme, Ellis, & Monk, 1994). The sounds to be deleted were either single phonemes or one phoneme of a consonant cluster (e.g., [b] from bice, or [l] from clart). In each case, the response was a real word (e.g., ice). In the rhyme oddity task the children heard 26 blocks of four one-syllable words (e.g., pad, had, bat, mad). One word in each block of four words did not rhyme with the others (in this case bat), its position being varied randomly in positions 1 to 3 across trials (after Snowling, Hulme, Smith, & Thomas, 1994). The child’s task was to say which word was the odd one out. Verbal short-term memory. The children’s verbal memory was assessed using standard span tasks. A pool of items for the word and nonword span tasks consisted of 14 monosyllabic items of high frequency (e.g., bed, test). Nonwords were constructed from each word by consonant substitution of either the initial or medial phoneme (e.g., voice, roice), and the items for each trial were selected at random from this pool. Presentation was computer generated. The children first

164

TABLE 1 Means and Standard Deviations for Test Battery Administered to 7- to 11-Year-Olds Measures Group

RAa

88.93 3.49

88.07 7.79

9.47 2.29

99.67 2.32

104.80 29.28

113.07 3.36

WISC WISC Block Vocabularyb Designb

WORD Reading1

Nonword Reading2

Rhyme Oddity3

Phoneme Deletion4

PA Learning5

Word Span6

Nonword Span6

7.53 3.56

24.73 5.13

10.66 5.14

14.53 4.32

10.20 3.09

25.46 11.92

4.73 1.03

2.13 .99

12.80 2.57

8.47 3.07

31.16 8.41

12.58 4.95

17.06 4.62

13.82 4.25

35.02 16.80

4.38 1.88

2.61 1.14

124.00 23.14

9.67 3.18

7.33 2.85

37.66 5.84

17.46 3.83

20.07 5.23

17.15 3.19

45.85 22.81

5.31 1.60

3.23 1.32

122.27 2.89

117.33 19.10

9.27 3.39

7.53 2.77

36.93 5.11

17.47 3.14

19.66 3.84

18.33 2.41

37.73 17.35

5.33 1.40

2.87 .91

137.00 3.62

139.33 18.88

9.20 2.48

10.00 2.88

42.53 3.11

18.40 1.76

21.66 2.69

18.06 3.97

51.53 12.98

5.93 1.16

3.26 .70

Note. 1maximum ⫽ 55; 2maximum ⫽ 20; 3maximum ⫽ 26; 4maximum ⫽ 24; 5maximum ⫽ 80; 6maximum ⫽ 6. a Age in months. b Scaled scores.

WINDFUHR AND SNOWLING

7 years M SD 8 years M SD 9 years M SD 10 years M SD 11 years M SD

CAa

LEARNING AND PHONOLOGICAL SKILLS

165

heard a list of two stimuli spoken aloud, presented at the rate of one per second. They then repeated these back to the experimenter, in the order in which they had been presented, for two trials at each list length. List length was gradually increased by one item each time until the child reached his or her limit and made errors on both trials at a given list length. An individual’s span was taken as the longest list length where both trials were correct, plus 0.5 for any further trial in which the items were recalled in the correct order. Paired associate learning (PAL). The paired associate learning task was constructed by pairing 2-D abstract visual shapes with spoken nonwords. Four shapes were selected from a group of 12-point shapes rated as moderate in complexity and of low associability (Vanderplas & Garvin, 1959). Children were taught to associate these shapes with four nonwords, two of one-syllable, stosp and taith, and two of three-syllable, meferal and balio. Following practice, there were 20 trials per block of 4 test items, and the child received corrective feedback throughout. As the focus of the study was on individual differences, the same shape was associated with the same nonword in each case, rather than rotating shapes with nonwords across subjects as is usually the case in paired associate experiments. Procedure Participants were assessed in a quiet area within their school. All participants received two testing sessions. The first session lasted approximately 30 min in which they were given the standardized tests in the following fixed order, and subsequently the tests of phonological awareness: WORD Basic Reading, WISCIII Vocabulary and Block Design, the Graded Nonword Reading Test. For the rhyme oddity task, each child was instructed to listen to four words spoken aloud by the experimenter and asked to choose the “odd one out.” Each child was first given three practice trials with corrective feedback before proceeding to the test items. For the phoneme deletion task, each child was asked to first listen to a nonsense word spoken aloud by the experimenter. The experimenter then deleted a phoneme from the target item and the child then verbalized the resulting word. Eight practice trials were given, and corrective feedback was provided before proceeding to the test items. The second testing session lasted approximately 20 min. It comprised the paired associate learning task and the memory tests. In the paired associate task, each child was shown four abstract shapes and taught the nonsense name of each shape. Participants were asked to repeat the nonword while carefully looking at each abstract shape. The whole procedure was repeated a second time. Test trials did not begin until each child was able to pronounce each of the nonwords. After the abstract shape/nonword pairs were shown to the children twice, the children were instructed to try to recall the name of each of the shapes. The four shapes were presented in random order over 20 blocks of four trials. Following each trial, the child heard the name of the shape irrespective of

166

WINDFUHR AND SNOWLING

whether they had named it correctly themselves or made an error. In addition, if the participant correctly remembered the match, they were praised to maintain interest and motivation. The memory span task was administered last, first span for words and then nonwords. Each child was instructed to listen to the items in each list and repeat them back in the order the experimenter had spoken them. Results The means and standard deviations describing the performance of the children across the tasks, by group, are shown in Table 1. The distributions of all the measures were examined to assess normality. Only nonword reading showed a slightly negative skew, indicating that the decoding skills of the older children were approaching ceiling. Partial correlations (controlling for chronological age) among the variables are shown in Table 2. As expected, word recognition ability was highly correlated with nonword reading and both measures of phonological awareness; rhyme oddity (p ⬍ .001) and phoneme deletion (p ⬍ .001). However, the correlations with memory span measures were lower, only reaching significance for word span (p ⬍ .05). Rhyme oddity correlated significantly with phoneme deletion (p ⬍ .001) and with nonword reading (p ⬍ .001). The correlation between phoneme deletion and nonword reading was also significant (p ⬍ .001). Importantly, there was a significant correlation between paired associate learning and reading (p ⬍ .001) as well as paired associate learning and nonword reading (p ⬍ .001). Paired associate learning was correlated with phoneme deletion (p ⬍ .001) but the correlations with rhyme oddity and short-term memory were not significant. Predictors of Paired Associate Learning Two regression analyses were conducted to assess the contribution of verbal, nonverbal, and phonological skills to paired associate learning. Composite variTABLE 2 Partial Correlations (Controlling for Chronological Age) between Measures of Reading Phonological Processing and Paired Associate Learning Measures (1) Word reading (2) Nonword reading (3) Rhyme oddity (4) Phoneme deletion (5) PA learning (6) Memory for words (7) Memory for nonwords *p ⬍ .05. **p ⬍ .01. ***p ⬍ .001.

1

2

3

4

5

6

7

.63***

.59*** .46***

.58*** .64*** .44***

.56*** .49*** .47*** .42***

.28* .20 .22 .33** .15

.18 .16 .38** .28* .18 .44***

167

LEARNING AND PHONOLOGICAL SKILLS

ables were formed for phonological awareness and for verbal memory. The composite phonological variable was derived by taking the mean of the z scores for phoneme deletion and rhyme oddity, standardized across the whole group. Similarly, the verbal memory composite was derived by taking the mean of the z scores for word span and nonword span. In the first analysis, phonological awareness and verbal short-term memory scores were entered into a hierarchical regression predicting paired associate learning after first controlling for the effects of age, nonverbal and verbal ability in the first two steps (Table 3). Together these variables accounted for 36% of the variance in paired associate learning but only phonological awareness made a unique contribution of 20% at the final step. A follow-up hierarchical regression assessed the relative contributions of phoneme deletion and rhyme oddity to this prediction. When rhyme oddity was controlled, phoneme deletion accounted for a unique 5% of variance at the last step and when the measures were entered in reverse order, rhyme oddity accounted for an independent 8% of variance. Phonological Skills and Paired Associate Learning as Predictors of Reading The contribution of phonological awareness, verbal memory, and paired associate learning as predictors of reading skill was investigated further in the next set of multiple regression analyses. To reduce the number of variables entered into these equations, the composite variables for phonological awareness and verbal memory were used. These variables were then examined, together with paired associate learning, as predictors of word and nonword reading in two parallel analyses (see Table 4). In each analysis, Age and Block Design raw score were entered at Step 1 and Vocabulary at Step 2, to control for individual differences in age and IQ. Paired associate learning (PAL), phonological awareness, and verbal memory were then

TABLE 3 Hierarchial Multiple Regression Investigating the Concurrent Predictors of Paired Associate Learning PAL Step 1 2 3 4 3 4 3 4 3 4

Variable Chronological age Block design Vocabulary Phonological awareness Memory Memory Phonological awareness Phoneme deletion Rhyme oddity Rhyme oddity Phoneme deletion

R2 change

p

.14

.01

.00 .22 .00 .03 .20 .15 .08 .18 .05

ns .00 ns .ns .00 .00 .01 .00 .02

168

WINDFUHR AND SNOWLING TABLE 4 Hierarchical Multiple Regression Analyses Investigating the Concurrent Predictors of Word and Nonword Reading

Step

Variable

1

Chronological age Block design Vocabulary Phonological awareness Memory PAL PAL Phonological awareness Memory Memory PAL Phonological awareness

2 3 4 5 3 4 5 3 4 5

Word reading R2 change p

Nonword reading R2 change p

.48 — .04

.001 — .03

.31 — .00

.001 — ns

.21 .00 .03 .16

.001 ns .01 .001

.31 .00 .02 .16

.001 ns .05 .001

.04 .00 .17 .14

.001 ns .02 .001

.16 .00 .03 .14

.001 ns .08 .001

.08

.001

.15

.001

entered in alternating order at Steps 3, 4, and 5 to assess their relative importance in the prediction of reading ability. Age and IQ accounted for 52% of the variance in word reading and 31% of the variance in nonword reading skill. After controlling for these variables, phonological awareness accounted for 21%, paired associate learning for 16%, and verbal memory for 17% of the variance in word reading. The contribution of paired associate learning to nonword reading was similar in magnitude to that in the prediction of word reading (16%) but the contribution of phonological awareness was greater (31%) and of verbal memory smaller (3%). Importantly, phonological awareness and paired associate learning both accounted for independent variance in word reading when entered at the last step (8 and 3% of the variance, respectively) but verbal memory did not. Similarly, phonological awareness accounted for a substantial 15% of the variance in nonword reading when entered at the final step, and paired associate learning accounted for a small but significant 2%. Thus, paired associate learning and phonological awareness both make an independent contribution to variance in reading skill. In a second set of analyses, verbal memory was dropped from the equations as it did not contribute unique variance to reading skill. These analyses compared the relative importance of phoneme deletion, rhyme oddity, and paired associate learning as concurrent predictors of reading. Details of the analyses are shown in Table 5. Each of the three phonological variables accounted for substantial proportions of variance in word and nonword reading (between 15 and 30%) when entered directly after age and IQ. Moreover, all three skills accounted for significant variance on the last step as predictors of word reading (phoneme deletion:

169

LEARNING AND PHONOLOGICAL SKILLS

TABLE 5 Hierarchical Multiple Regression Analyses Assessing Phoneme Deletion, Rhyme Oddity, and PAL as Concurrent Predictors of Word and Nonword Reading Step

Variable

1

Chronological age Block design Vocabulary Phoneme deletion Rhyme oddity PAL PAL Phoneme deletion Rhyme oddity Rhyme oddity PAL Phoneme deletion

2 3 4 5 3 4 5 3 4 5

Word reading R2 change p

Nonword reading R2 change p

.48

.001

.31

.001

.04 .15 .06 .03 .16 .06 .03 .15 .06 .03

.03 .00 .00 .00 .001 .001 .01 .00 .00 .00

.00 .30 .03 .02 .16 .17 .01 .15 .07 .13

ns .00 .02 .04 .001 .001 ns .00 .00 .00

p ⬍ .001; rhyme oddity: p ⬍ .01; paired associate learning: p ⬍ .001). In contrast, only phoneme deletion and paired associate learning were significant as unique predictors of nonword reading at the last step (phoneme deletion: p ⬍ .001; paired associate learning: p ⬍ .05). To assess whether paired associate learning accounted for knowledge of wordspecific aspects of reading, independent of decoding skill, a final set of hierarchical regressions was conducted. In these analyses, decoding ability was controlled at the first step by entering nonword reading score (Table 6). Nonword reading skill accounted for 59% of the variance in word recognition. Once decoding skill was controlled, paired associate learning accounted for 5% of the variance and phonological awareness a further unique 8% of the variance in reading. Paired associate learning also accounted for an independent 8% of variance when entered on the last step after phonological awareness was controlled. Discussion This study examined the relationship between paired associate learning, phonological processing abilities, and reading in 7- to 11-year-old normal readers. Paired associate learning, phonological awareness, and verbal memory skills were all concurrent predictors of reading but only paired associate learning and phonological awareness accounted for unique proportions of the variance. Consistent with the findings of McDougall et al. (1994), the contribution of verbal short-term memory was subsumed by that of phonological awareness. While awareness of phonemic segments and larger rime units were predictors of word reading, only phonemic awareness predicted nonword reading. A second finding of the study was that phonological awareness was a strong predictor of paired associate learning, contributing unique variance to this skill

170

WINDFUHR AND SNOWLING

TABLE 6 Hierarchical Regression Analyses Investigating the Concurrent Predictors of Word Recognition after Individual Differences in Decoding Skill Are Controlled Step 1 2 3 2 3

Variable Nonword reading PAL Phonological awareness Phonological awareness PAL

Word recognition R2 change p .59 .05 .08 .11 .08

.000 .002 .000 .000 .041

when age and IQ were controlled. Thus, children with good phonological skills can learn to associate abstract shapes with novel names efficiently, although it is not possible to conclude which component of paired associate learning depends on phonological sensitivity. The paired-associate learning task required learning four abstract visual shapes and four nonword responses, as well as establishing a link between their memorial representations. The paradigm we used minimized the demands on visual memory since the shapes were always presented in the recall phase. Response learning, on the other hand, required the acquisition of new phonological forms and it is reasonable to suppose that this would have been easier for children with better phonological skill, as indexed by performance on phonological awareness tasks (Aguiar & Brady, 1991; de Jong et al., 2000). The critical finding of the present study was that paired associate learning accounted for unique variance in reading, even when the powerful influence of phonological awareness was controlled. It can be inferred that mechanisms which account for the ability to link or “hook-up” stimulus–response pairs in the paired associate learning task may impose an independent constraint on learning to read. Furthermore, the contribution of paired associate learning to word recognition remained significant when performance in nonword reading was controlled, suggesting it has a distinct role to play in establishing lexical connections between words and pronunciations (or distributed patterns of activation representing these). Using a similar argument, Wimmer et al. (1998) proposed that paired associate learning might be used in the development of orthographic representations (Ehri, 1992). The present findings have implications for our understanding of reading development. Within the dual-route framework, word reading can proceed on the basis of lexical associations between letter strings and whole word pronunciations, whereas nonword reading depends upon the operation of a separate system containing grapheme–phoneme correspondences. The important contribution of paired associate learning to word reading, independent of decoding skill, rests easily with this view. However, the finding that phonological awareness and paired associate learning both make unique contributions to word and nonword

LEARNING AND PHONOLOGICAL SKILLS

171

reading processes is not compatible with the idea that lexical and nonlexical reading systems are independent. Within the connectionist framework, a single mechanism operating upon representations that are similar in form, and with the same learning algorithm, accounts for both word and nonword reading. However, individual differences in the status of the representations and in the association learning procedures can be expected to affect word and nonword reading to differing degrees. Simulations by Plaut, McClelland, Seidenberg, and Patterson (1996) indicated that, for the knowledge that a network has accrued during reading trials to generalize to allow novel words to be read, the mappings between orthography and phonology need to be at a fine grain level. In turn, this depends upon the structure and integrity of underlying phonological representations (Brown, 1997; Harm & Seidenberg, 1999). It is usual to assume that performance on tests of phonemic awareness provide an assessment of the status, and in particular, the segmental nature of a child’s phonological representations (e.g., Swan & Goswami, 1997). The present finding that the contribution of phonological awareness to nonword reading came primarily from performance on a phoneme deletion task is consistent with this view. The role of learning in the acquisition of reading, as distinct from the status of phonological representations, is highlighted by the finding that paired associate learning was a predictor of both word and nonword reading. Importantly, word recognition in a connectionist model can be disrupted by degrading phonological representations or by altering aspects of association learning. Harm and Seidenberg (1999) found that, whereas degrading phonological representations in their model had the greater effect on nonword reading, altering parameters of the learning algorithm to produce a nonoptimal learning rate had a substantial effect on the model’s performance on exception words, with a less marked effect on the model’s ability to derive pronunciations for nonwords. Similarly, altering the model’s learning resource by removing hidden units can be expected to affect its capacity to deal with inconsistencies in the mappings between orthography and phonology that characterize English exception words. Thus, the connectionist framework provides a good account of the present findings. To conclude, a considerable body of research has focused on the role of phonological skills in learning to read, and conversely, on the impact of phonological deficits. In contrast, the contribution of learning to the process of reading acquisition and as a factor in explaining reading failure has been relatively neglected. If the arguments posed here are correct, then individual differences in children’s learning skill may contribute to differing patterns of reading behavior, or “subtypes” of dyslexia. REFERENCES Adams, M. J. (1990). Beginning to read—Thinking and learning about print. Cambridge, MA: MIT Press. Aguiar, L., & Brady, S. (1991). Vocabulary acquisition and reading ability. Reading and Writing, 3, 413–425. Besner, D. (1999). Basic processes in reading: multiple routines in localist and connectionist models.

172

WINDFUHR AND SNOWLING

In R. M. Klein & P. A. McMulen (Eds.), Converging methods for understanding reading and dyslexia (pp. 413–458). Cambridge, MA: MIT Press. Bond, G. L., & Dykstra, R. (1997). The co-operative research program in first grade reading instruction. Reading Research Quarterly, 2, 5–142. Brown, G. (1997). Connectionism, phonology, reading, and regularity in developmental dyslexia. Brain and Language, 59, 207–235. Byrne, B. (1996). The learnability of the alphabetic principle: Children’s initial hypothesis about how print represents spoken language. Applied Psycholinguistics, 17(4), 401–426. Castles, A., & Holmes, V. M. (1996). Subtypes of developmental dyslexia and lexical acquisition. Australian Journal of Psychology, 48(3), 130–135. Coltheart, M. (1978). Lexical access in simple reading tasks. In G. Underwood (Ed.), Strategies of information processing (pp. 151–216). London: Academic Press. Coltheart, M., Curtis, B., Atkins, P., & Haller, M. (1993). Models of reading aloud: Dual-route and parallel-distributed-processing approaches. Psychological Review, 100, 589–608. de Jong, P. F., Seveke, M. J., & Van Veen, M. (2000). Phonological sensitivity and the acquisition of new words in children. Journal of Experimental Child Psychology, 76, 275–301. Ehri, L. C. (1992). Reconceptualising the development of sight word reading and its relationship to recoding. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.), Reading acquisition (pp. 107–143). Hillsdale, NJ: Erlbaum. Elbro, C. (1996). Early linguistic abilities and reading development: A review and a hypothesis. Reading and Writing: An Interdisciplinary Journal, 8, 453–485. Fowler, A. E. (1991). How early phonological development might set the stage for phoneme awareness. In S. Brady & D. Shankweiler (Eds.), Phonological Processes in Literacy (pp. 97–117). Hillsdale, NJ: Erlbaum. Harm, M. W., & Seidenberg, M. S. (1999). Phonology, reading acquisition, and dyslexia: Insights from connectionist models. Psychological Review, 106(3), 491–528. Hoien, T., Lundberg, I., Stanovich, K., & Bjaalid, I. K. (1995). Component of phonological awareness. Reading and Writing, 7, 171–188. Hulme, C. (1981). Reading retardation and multi-sensory teaching. London: Routledge & Kegan Paul. Hulme, C., & Snowling, M. (1992). Deficits in output phonology: An explanation of reading failure? Cognitive Neuropsychology, 9, 47–72. Lundberg, I., Olofsson, A. & Wall, S. (1980). Reading and spelling skills in the first school years predicted from phonemic awareness skills in kindergarten. Scandinavian Journal of Psychology, 21, 159–173. Manis, F. R. (1985). Acquisition of word identification skills in normal and disabled readers. Journal of Educational Psychology, 77(1), 78–90. Manis, F., Savage, P., Morrison, F., Horn, C., Howell, M. J., Szeszulski, P. A., & Holt, L. K. (1987). Paired associate learning in reading-disabled children: Evidence for a rule-learning deficiency. Journal of Experimental Child Psychology, 43, 25–43. Manis, F., Seidenberg, M., & Doi, L. (1999). See Dick RAN: Rapid naming and the longitudinal prediction of reading sub-skills in first and second graders. Scientific Studies of Reading, 3, 129–157. Mayringer, H., & Wimmer, H. (2000). Pseudoname learning by German speaking children with dyslexia: Evidence for a phonological learning deficit. Journal of Experimental Child Psychology, 75, 116–133. McDougall, S., Hulme, C., Ellis, A., & Monk, A. (1994). Learning to read: The role of short-term memory and phonological skills. Journal of Experimental Child Psychology, 58, 112–133. Muter, V., Hulme, C., Snowling, M., & Taylor, S. (1997). Segmentation, not rhyming predicts early progress in learning to read. Journal of Experimental Child Psychology, 65, 370–396. Plaut, D., McClelland, J., Seidenberg, M., & Patterson, K. (1996). Understanding normal and impaired word reading: Computational principles in quasi-regular domains. Psychological Review, 103, 56–115.

LEARNING AND PHONOLOGICAL SKILLS

173

Seidenberg, M. S., & McClelland, J. L. (1989). A distributed, developmental model of word recognition and naming. Psychological Review, 96, 523–568. Share, D. L. (1995). Phonological recoding and self-teaching: Sine qua non of reading acquisition. Cognition, 55, 151–218. Snowling, M., Hulme, C., Smith, A., & Thomas, J. (1994). The effects of phonetic similarity and list length on children’s sound categorisation performance. Journal of Experimental Child Psychology, 58, 160–180. Snowling, M. J., Stothard, S. E., & McLean, J. (1996). Graded Nonword Reading Test. Bury St. Edmunds: Thames Valley Test Co. Swan, D., & Goswami, U. (1997). Picture naming deficits in developmental dyslexia: The phonological representations hypothesis. Brain and Language, 56, 334–353. Vanderplas, J., & Garvin, I. (1959) The associative values of random shapes. Journal of Experimental Psychology, 57, 147–154. Vellutino, F. R., Scanlon, D. M., & Spearing, D. (1995). Semantic and phonological coding in poor and normal readers. Journal of Experimental Child Psychology, 59, 76–123. Vellutino, F. R., Steger, J. A., Harding, C. J., & Phillips, F. (1975). Verbal vs. non-verbal paired associate learning in poor and normal readers. Neuropsychologia, 13, 75–82. Wagner, R., Torgesen, J., & Rashotte, C. (1994). Development of reading-related phonological processing abilities: New evidence of bi-directional causality from a latent variable longitudinal study. Developmental Psychology, 30, 73–87. Wechsler, D. (1992). Wechsler Intelligence Scale for Children—III. San Antonio, TX: Psychological Corporation. Wechsler, D. (1993). Wechsler objective reading dimensions. London: Psychological Corporation. Wimmer, H., Mayringer, H., & Landerl, K. (1998). Poor reading: A deficit in skill-automatization or a phonological deficit? Scientific Studies of Reading, 2, 321–340. Wolf, M., & Bowers, P. G. (1999). The double deficit hypothesis for the developmental dyslexias. Journal of Educational Psychology, 91(3), 415–438. Yopp, H. K. (1988). The validity and reliability of phonemic awareness tests. Reading Research Quarterly, 23, 159–177. Received October 13, 1999; revised November 21, 2000; published online May 29, 2001