Reading and phonemic awareness and RAN: The rest of the story

Is it possible that the NICHD sponsored reading disability research has produced an emperor that is only partially clothed? Have the constructs of rapid automatized naming (RAN) and phonological awareness (PA) become more powerful than they should?

While rooting around the literature, I recalled the following meta-analytic review that contributes to answering these questions. Two things to note:

• The article is published in the none-to-shabby journal the Review of Educational Research. RER is one of the highly respected flagship journals of the American Educational Research Association (AERA). AERA is a tough crowd, as anyone who has presented a paper at their conference soon learns during the Q/A session. These folks know their research and methods. Making an AERA presentation is not for the weak.

• By using meta-analysis, which is a statistical review technique that provides a quantitative summary of findings across an entire body of research, the study provides for a more quantitative/objective perspective regarding the importance of PA and RAN.

Article reference

• Swanson, H. L., Trainin, G., Nechoechea, D. M., & Hammill, D. D. (2003). Rapid naming, phonological awareness, and reading: A meta-analysis of the correlation evidence. Review of Educational Research, 73(4), 407-440.

Article Abstract

• This study provides a meta-analysis of the correlational literature on measures of phonological awareness, rapid naming, reading, and related abilities. Correlations (N = 2,257) were corrected for sample size, restriction in range, and attenuation ffom 49 independent samples. Correlations between phonological awareness (PA) and rapid naming (RAN) were low (.38) and loaded on different factors. PA and RAN were moderately correlated with real-word reading (.48 and .46, respectively). Other findings were that (a) real-word reading was correlated best (r values were .60 to.80) with spelling and pseudoword reading, but correlations with RAN, PA, vocabulary, orthography, IQ, and memory measures were in the low-to-moderate range (.37 to .43); and (b) correlations between reading and RAN/PA varied minimally across age groups but were weaker in poor readers than in skilled readers. The results suggested that the importance of RAN and PA measures in accounting for reading performance has been overstated.

My comments (aka; Dr. IQ's 2 cents)

Unless a school-based assessment professional has been living in a cave for the last 10 years, it would be rare to find an assessment professional who has not heard of the prominence the double-deficit dyslexia hypothesis (Wolf and Bowers, 1999). This hypothesis suggests that some deficits in reading may be related to the speed with which one can name aloud a series of letters, objects, and numbers (rapid automatized naming-RAN), as well as to deficits in phonological awareness (PA). This reading disability model has dominated reading disability research this past decade (particularly that sponsored via NICHD). Few assessment professionals working in the schools have not learned the lingo of RAN and PA.

The Swanson et al. (2003) review raises questions about the strength of the RAN/PA double-deficit model position, a position that continues to exert a significant influence on federal educational policy decisions. The meta-analysis based conclusions suggest that the hot RAN/PA position may be overstated. The study serves to remind us of the potential problem of model specification error in research.

Briefly, specification error occurs when potentially important variables in predictive or explanatory studies are omitted, an error that can lead to biased estimates of the effects of predictive variables. In reading research, specification error can be particularly problematic as it may result in:

• (a) abilities being excluded from the analysis as they are incorrectly judged to be unimportant, when they are vital,
  
• (b) the accurate relative contribution of multiple abilities to reading achievement remains unknown,
  
• (c) researchers have a tendency to follow the “bandwagon” and may continue to build on studies where an incomplete set of predictor/causal ability variables are included [which may lead to premature reliance and focus on certain predictors measures – a "premature hardening of the predictors"], and
  
• (d) consumers of this research may be misled by less than accurate conclusions
• [See Evans, Floyd, McGrew (yes, it be me) & Leforgee, 2002. The relations between Cattell-Horn-Carroll (CHC) cognitive abilities and reading achievement during childhood and adolescence, School Psychology Review, 31(2), 246-262 – these issues are discussed in greater detail in the context of CHC theory in this article.]

According to Swanson et al (2003), aside from PA and RAN, a significant number of other studies have identified other cognitive processes as important to reading such as those related to orthography, semantics, and working memory span. Our own research (Evans et al, 2002) has suggested that the CHC ability of associative memory (MA) may be a victim of specification error in contemporary reading disability research.

Swanson et al. (2003) concluded that:

• “When corrected for sample size and sample heterogeneity (variations in SES, ethnicity, and age), the majority correlations related to PA and RAN are in the low range (mean r= .38).”
• “Furthermore, correlations between real-word reading, PA, and RAN are also in the low-to-moderate range (.35 to .50).”
• “These findings remain stable even after partialing for variations in the samples as a function of age, SES, distribution of reading ability, gender, and ethnicity.”
• “The present synthesis appears to support the observations of Wolf and Bowers that performance on RAN measures and performance on PA measures are the results of independent processes.”
• “Thus, it appears to us that less emphasis should be placed on PA and RAN measures in attempts to classify children at risk for reading and more emphasis should be placed on spelling.”
• “The double-deficit theory of reading disability does not fit at all.”
• “Our synthesis is consistent with the current literature suggesting that isolated processes, such as phonological coding, do play a modest part in predicting real-world reading and pseudoword reading. However, our study highlights additional processes as playing equally important roles in reading. It suggests the importance of phonological awareness may have been overstated in the literature.”

Final musing/comment

Regardless of the independence of PA and RAN measures in predicting reading, the meta-analysis suggests that RAN and PA measures, although related to reading achievement and reading disabilities, are not the only measures of important constructs that should be used in reading research and reading-related applied assessments by practitioners.

From my reading of the literature, when cast within the context of CHC theory and when focusing on the early elementary grades, the potentially important set of reading “marker” variables (with CHC notations included) should include, in addition to PA (Phonetic Coding, Ga-PC) and RAN (Naming Facility, Glr-NA), associative memory (Glr-MA), working memory (Gsm-MW), processing speed (Gs), and vocabulary/semantics (lexical knowledge, Gc-VL).