Decoding and reading comprehension: A meta-analysis to identify which reader and assessment characteristics influence the strength of the relationship in English

Garcia, J. R., & Cain, K. (2014). Decoding and reading comprehension: A meta-analysis to identify which reader and assessment characteristics influence the strength of the relationship in English. Review of Educational Research, 84(1), 74-111.


Two realizations struck me as I read this article: 1) Reading is a complex process, with many interlocking components, all of which influence how (or how well) meaning is constructed, and 2) How the components of the reading process are measured reflects what the researcher believes reading is, with the choice of measure in turn affecting what kinds of relationships will be perceived among the components measured. A third thought also ran through my mind, and that was this question: Could it be that the relationships among reading components might vary across readers? In other words, could the reading process, and the developmental paths learners take as they become readers, be different for different readers? Yes, there may be some basic relationships that tend to be seen more often, but is it possible that we will never have perfect correlations because of natural human variations in how literacy develops?

Those of us who have worked closely with young readers know that different readers can show us different patterns of literacy skills. It seems intuitive that if one is skilled at decoding words, one would then be better able to comprehend what was read—but it doesn’t always work out that way. Everyone who has assessed more than a few readers has observed readers who could read every word accurately, and sometimes even fluently and expressively, but do not comprehend what they read. Conversely, most of us have also observed readers whose oral reading made us think they could never get meaning from the text, and yet, somehow, they did, and sometimes even did it quite well, and occasionally even better than the readers we considered better decoders. For some readers, the decoding-comprehension relationship is clear, but it is not so clear for others.

Interestingly, the authors of this meta-analysis find support for the idea that the older readers get, the smaller the relationship between decoding skills and comprehension. They pinpoint the age at which this relationship begins to decrease at about 10 years old. Maybe this is because basic word skills are often already learned by age 10, and other aspects of the text begin to be more important for meaning construction. Could it also be because the human motivation to make sense of things is so strong that readers who struggle with decoding will sometimes develop compensatory strategies that enable them to construct meaning in ways that work for them? Could it be that if they are able to develop these strategies, they will do so by the late elementary grades? I don’t know the answers to these questions, but the beauty and mystery of the reading process continues to fascinate me, and I suspect will continue to keep researchers busy for the foreseeable future, at least until someone figures out how to measure the complex phenomenon that is reading more directly.

Until we do find ways to measure complex cognitive processes directly, we must rely on indirect measures that require inferential thinking, and that’s where we run into the problem of assessment. How we define good word decoding, and how we define good reading comprehension, can vary widely, and will affect our choices of assessment tools. The authors here clearly show us how those choices can lead to different research findings. If a researcher believes that comprehension is being able to read short passages and answer literal-level questions about them, that will lead to different findings than those that would be gotten by a researcher who believes comprehension is being able to discuss themes and concepts in an extended text and generate questions about that text. If a researcher believes that decoding skills can be measured by reading lists of words, that will lead to different findings than those that would be gotten by a researcher who believes that word decoding can only be accurately measured by looking at how readers read words in contexts. As long as there has been literacy research, there has been debate about what reading is, what its components are, what the relationships among those components are, how best to measure them, and of course, what we can conclude about how best to help readers develop. These debates are likely to continue; thanks to Garcia and Cain for their contribution to the discussion, and for helping at least one literacy educator refresh her own thinking on the nature of the reading process.