The Role of Working Memory in Phoneme Awareness

The primary goals of this study were to examine the association between working memory and phoneme awareness, and to explore the relationships of these phonological processes with decoding and comprehension components of reading ability. Previously, only a few studies had examined the role of working memory in phoneme awareness, and the findings were inconclusive. A further goal of the present study was to examine whether simple phoneme awareness tasks (requiring counting or segmenting of phonemes) relate differently to reading skills than do compound phoneme awareness tasks (requiring deletion or manipulation of phonemes). Second-grade children whose reading skills were normally distributed served as subjects (n~l40). In addition to examining the abilities of the entire subject pool, two subgroups of skilled and lessskilled decoders were compared to determine if children having difficulty learning to read perform differently on phoneme awareness and working memory tasks. The subjects were given three working memory measures, three simple phoneme awareness tasks, and three compound phoneme awareness measures. The results suggested that working memory ability is not strongly linked to performance on phoneme awareness tasks. Thus, these constructs represent relatively independent phonological processes. Second, the simple phoneme awareness tasks were found to factor separately from the compound phoneme awareness measures, suggesting it is important to distinguish between these two levels of metaphonological awareness. For the entire pool of subjects, the working memory, simple phoneme awareness, compound phoneme awareness, and IQ measures were differentially related to decoding and comprehension performance. Decoding, a more basic skill, related highly to expertise on compound phoneme awareness measures, with simple phoneme awareness making a emaller contribution. Similarly, the subgroups of skilled and lessskilled decoders were significantly differentiated by their performance on both simple and compound phoneme awareness measures, but not on working memory. Comprehension performance, in contrast, was best predicted by compound phoneme awareness, verbal and nonverbal IQ, and working memory. The value of phoneme awareness measures for prediction and assessment purposes was discussed, as was the issue of how to control for IQ when examining phonological processes related to learning to read. ACKNOWLEDGEMENTS I am especially grateful to Dr. Susan Brady, my major professor, for being supportive during this latter phase of my education, and for sharing her time, ideas, and expertise to benefit my professional development. I am also very grateful to Dr. Joseph Rossi for sharing his statistical knowledge and creative insights. Lastly, I thank Dr. Allan Berman, , Dr. Janet Kulberg, and Dr. Susan Trostle for their highly regarded input and the contribution of their time. I also extend a special thanks to the many school personnel, parents, and children who volunteered their services and assisted me in this project. Gratitude is extended to: Sr. Ann Marie (principal), Ma. Barrios, Ms. Hedden, and to a very good friend, Mrs. Jenny Abate of st. Philomena's school; Ms. Cathy Harnett (principal), Ms. Busabos, Ms. Kittner, and Ms. Pitou of Warren schools; and Mr. Hayes (principal), Ms. Christensen, Ms. Eisenhaure, Mr. Garzilli, and Ma. Smith of Narragansett school. I am also especially grateful to Heather MacLeod and Jennifer Thuber who gave generously of their time in assisting me in testing children, score protocols, conduct literature searches, and to complete work on the computer. Their assistance was invaluable to me. Lastly, and most importantly, I want to thank my very special husband and group of U.R.I. friends who have been an inspiration to me and whose encouragement I have always appreciated. And, thanks be to


INTRODUCTION
The intent of this study is to .examine the degree to which phoneme awareness relies on phonological codes in working memory. Thia investigation was conducted within the context of Wagner's (1988) model ■ that described the relationship of three phonological processes.
Presently, the evidence is inconclusive about how working memory and phonological awareness relate to one another. It is proposed that working memory and phoneme awareness are independent proce ■ ses, but that each latent ability makes a unique contribution to reading performance.
This conclusion was based on the well-supported evidence that measures conjectured to rely on a phonological code are highly discriminating of good and poor readers (e.g., verbal working memory and phonological awareness). Also, it has been hypothesized that phoneme awareness tasks can be classified according to the degree of cognitive complexity the task involves (Yopp, 1988). Therefore, multiple measures reflective of these latent constructs, were administered in order to assess the factorial structure of these manifest variables and the relationship of their respective constructs.
In addition to exploring the relationship of the constructs to overall reading ability, the skill requirements for decoding versus passage comprehension are contrasted. Also, the profiles of leaaakilled and skilled decoders are examined on a subset of indicator variables.

Phonological Awareness
Phonological awareness refers to a conscious awareness of the sound structure of one's language. For example, given the word "dog," one would say it consists of three sounds or phonemes, or is one syllable, and rhymes with bog. Phonological awareness tasks generally require analysis or synthesis of word units. An "analysis" task, for instance, requires a child to segment a word into its respective speech units (phonemes or syllables). In contrast, a "synthesis" task requires a child to blend speech units to form a word.
(2) All working memory tasks will load onto the working memory latent variable, with Dreyer'& Single Syllable Memory Test accounting for most of the variance.
(3) All three reading measures will load onto the reading achievement latent variables with word identification accounting for the larger portion of the variance.
(4) The two subtest& intended to reflect general intelligence will primarily load onto the IQ construct with vocabulary more atrongly related to IQ.
(5) A reciprocal relationship will exist between IQ and working memory.
(6) A unidirectional relationship will be present between working memory with the simple and compound phoneme constructs and, subsequently, reading achievement. Similarly, this directional pattern applied to the IQ construct.
(7) The latent variable for working memory will account for a large portion of the variance in the compound phoneme awareness variable and a moderate to small portion of the variance in the simple variable.
(8) Simple and compound phoneme awareness and working memory constructs will affect reading achievement independently of one another, with compound phoneme awareness accounting for more of the variance in reading achievement than the other two constructs.
The model hypothesized to fit the given data is suggested by the above predictions and represented in Figure 1. Two alternative models are proposed in Figures 2 and 3 Simple phoneme awareness measures.
(See Appendix A.) The Segmentation task used in this study was a modified version of the "Yopp-Singer" (1988).
In this task, subjects were asked to verbalize the individual phonemes heard in a real word. For example, given the test item "may," the child would say "/m/-/ay/." Twenty-seven words were presented for analysis. In order to confirm that the measures indeed represented separate con ■tructs, a factor analysis procedure was conducted (See Table 3  The results from the factor analysis also supported the hypothesis that a significant distinction exists between simple and compound phoneme awareness tasks. These four factors accounted for 671 of the variance among the variables. Table 4 provides the correlations between factors.
Factor one pertains to . the reading and compound phoneme awareness factor.
Factor two refers to the IQ construct. The construct of working memory is represented by factor three. Lastly, factor four represents simple phoneme awareness.         Table 13 summarizes the findings. Figure 7 demonstrates the relationship between these constructs.  Figure 7. Beta weights of pathways for IQ and working memory with simple and compound phoneme awareness as separate criteria.
7 Since we know that compound and simple phoneme awareness share a moderately significant relationship with one another, they were not included as a predictor variable for the other (see measurement model). The primary interest was to assess the degree working memory and IQ contributed to the individual phoneme awareness tasks.

SEGMENTATION TEST
Instructions: I'm going to say a real word and I want you to tell me what sounds you hear in the words --not letters.
For example, in the word "cat," there are three sounds --/c/, an /a/, and a /t/. Now, when you say the sounds, I want you to say them slowly and far apart from each other like I did with the word "cat." Now, let's try some together. Here's a card and ten tiles. I'm going to say some words that aren't real words and I want you to show me how many sounds you hear in the words --not letters but sounds. You'll show me how many sounds there are by placing the same number of tiles on this card. For example, with the word /ot/, there are two sounds /o/ and /t/, so I'll place two tiles on this card. OK, now let's try some together.

Practice
Practice: How many sounds do you hear in the word "ru?" You hear two sounds.
What are they? (If child says the letters, correct him/her and say: I want you to give me the sounds, not the letters. OK, now how many tiles would you put on your card? This is a memory test. I'm going to play a tape that has lists of words --words that aren't real, for example, pog. After each list is finished I want you to repeat the words back as best as you can in the order that you hear them.
Let's try practicing some together.