Teacher question:
I am a literacy interventionist at an elementary school, and we use DIBELS for our progress monitoring. While I recognize the value of DIBELS as a screening tool, I have concerns about the appropriateness of the current fluency benchmarks my school has adopted. I have found some research that identifies fluency goals calibrated to reading comprehension. Studies by O'Connor (2017) and Cogo-Moreira et al. (2023) identify specific words-per-minute benchmarks to establish a cut-off point for reading speed and accuracy to obtain minimum values for comprehending texts. These wpm goals are much lower than our fluency goals. If the ultimate goal of reading is understanding the text, I wonder if these research-based targets would be more appropriate goals for many of our students.
Shanahan responds:
Several years ago, a friend of mine was developing a remedial reading program. He wanted to set fluency benchmarks.
I hadn’t thought much about that problem. I had chaired the National Reading Panel subcommittee on fluency instruction. I knew that fluency teaching improved fluency and, consequently in most studies, reading comprehension.
But how fluent did kids have to be?
There were no studies that had addressed the problem in quite that way (I thought), but there were some fluency norms that could provide a clue.
My first thought was that they should aim for the 50th percentiles. For example, the norms indicated that the average second grader ends the year able to read about 96 words correct per minute (wcpm) (Hasbrouck & Tindal, 2017). For me, that would have been the second-grade target.
My reasoning was straightforward. My thinking was that kids who reached the 50th percentile would not have a fluency problem. If they were struggling to make sense of text, it wouldn’t be because they struggled too much with the words, etc.
My friend was not satisfied. He wondered, “Why wouldn’t the 40th or 45th percentile be adequate?”
Many years ago, Keith Stanovich (1984) described reading as an “interactive-compensatory” process. What he was getting at was that reading involves a constellation of varied skills and abilities.
Think, for instance, of the “simple view of reading.” That model describes reading comprehension as the product of two sets of abilities: decoding and language comprehension. To read you must translate print to language and then you must do what you do to understand language. Decoding and language abilities must interact.
What happens if readers struggle with one of these skills? According to Stanovich and scads of research, readers try to rely on their relative strength to compensate for the limitation. When readers struggle to decode, they don’t quit in a snit, they make like AI, trying to guess the word, using what they know about the semantics and syntax of language to compensate for decoding limitations.
Fluency – like reading in general – is a bit of a mash up. It relies on both decoding skills and language knowledge (along with executive functioning, reasoning ability, and knowledge). Kids who reach the 50th percentile in fluency are not necessarily all the same. Some kids may rely more on decoding, while others may compensate with some of their other abilities. Achieving average on a fluency test won’t guarantee average decoding skills, but it seems very unlikely that such students would be particularly low in decoding. You can only compensate so much.
The research that you noted is interesting. The researchers were doing exactly what you said, they were trying to identify the degree of fluency that was necessary to enable adequate comprehension.
There is a long line of research on this topic – something those modern researchers seem to be unaware of. Back in the 1940s, Emmett Betts (1946) wasn’t trying to establish a fluency learning objective. No, his purpose was to determine an appropriate level of text to use for instruction. He theorized, without evidence, that kids could only improve their reading when working with texts they comprehended, and he decided, again without any evidence, that an adequate degree of comprehension meant kids could answer 75-89% of the questions about a text.
He concluded, based on the kinds of studies researchers are doing now, that kids were best taught with texts they could read with 95% accuracy. When someone refers to “reading levels” that’s what they mean (or some variation on those criteria).
Later studies (Dunkeld, 1970; Powell, 1968) accepted Betts’s theory but challenged his criteria. They reported a lot of variation across the grades. In other words, different degrees of fluency were needed to ensure the target comprehension levels depending on grade levels.
The more recent studies don’t just consider accuracy – that is, the percentage of words read correctly. They look at a combination of speed and accuracy: the numbers of words students can read correctly per minute (wcpm). This approach provides a more reliable estimate of fluency, especially if conducted with multiple texts or for longer reading durations.
These newer studies, like those from the 1960s and 1970s, are reporting that kids don’t need to be especially fluent to comprehend. For instance, the fluency norms report that at the end of Grade 2, the average student can read about 100wcpm. Various studies say that 43 (Alves, et al., 2020), 47 (Cogo-Moreira, et al., 2023), or 78 wcpm (O’Connor, 2017) are all that is needed to allow successful comprehension.
Or for grade 4: the norms say 133 wcpm, while the amounts of fluency needed to ensure comprehension in these studies is 71 (Alves et al., 2021), 79 (Cogo-Moreira, et al., 2023) and 70 wcpm (O’Connor, 2017).
The researchers who have published these findings are appropriately cautious; they recognize that with different sets of texts or larger and more diverse samples of kids, the results are likely to vary quite a bit. This is because the standard deviations are large for this ability both in their studies and in the norms.
Those DIBELS targets are not just a seat-of-the-pants estimate like my notion of aiming for the 50th percentile. Nor are they an attempt to predict reading comprehension. Their targets are based on the connection of their oral reading fluency scores and performances on state tests – a more distant and generalized measure of reading ability than used in these studies. Basically, their benchmarks are more linked to learning progress than to comprehension (University of Oregon, 2020). Accordingly, their targets are much closer to those averages that I had recommended. For second graders, the norms say 100 wcpm, and DIBELS aims for 94; for grade 3, it’s 112 and 114; for Grade 4, 133 and 125, and so on.
What is it that DIBELS (and other test makers) are claiming with their target criteria? They are not claiming that the accomplishment of those levels of fluency will guarantee high performance on your state tests. Reading is too complex for that.
No, they are saying that if your kids are that fluent, you can check that off as a reason for low reading comprehension performance. Another way of saying this is, that given those levels of fluency, if your kids are also adequate in all their other abilities (like vocabulary, for instance), then they should have good enough reading comprehension.
Personally, if my school was using DIBELS or one of these other testing regimes, I would use their targets. Without those kinds of tools, aiming for the 50%ile may be a bit high, but only a bit. It is a reasonable target. In any event, since there is more than one way to comprehend a text, preparing kids to only be fluent enough to comprehend a given text is just too low a standard if we want our kids to able to read a wide range of texts well enough.
References
Alves, L. M., Santos, L. F. D., Miranda, I. C. C., Carvalho, I. M., Ribeiro, G. L., Freire, L. S. C., Martins-Reis, V. O., & Celeste, L. C. (2021). Reading speed in elementary school and junior high. Evolução da velocidade de leitura no Ensino Fundamental I e II. CoDAS, 33(5), e20200168. https://doi.org/10.1590/2317-1782/20202020168
Betts, E. (1946). Foundations of reading instruction. New York; American Book Co.
Cogo-Moreira, H., Molinari, G. L., Carvalho, C. A. F., Kida, A. S. B., Lúcio, P. S., & Avila, C. R. B. (2023). Cut-off point, sensitivity and specificity for screening the reading fluency in children. Pontos de corte, sensibilidade e especificidade para rastreamento da fluência leitora em crianças. CoDAS, 35(3), e20210263. https://doi.org/10.1590/2317-1782/20232021263pt
Dunkeld, C. G. (1970). The validity of the informal reading inventory for the designation of instructional reading levels: A study of the relationships between children’s gains in reading achievement and the difficulty of instructional materials. Unpublished doctoral dissertation, University of Illinois at Urbana-Champaign.
Hasbrouck, J. & Tindal, G. (2017). An update to compiled ORF norms (Technical Report No. 1702). Eugene, OR. Behavioral Research and Teaching, University of Oregon.
O’Connor, R. E. (2017). Reading fluency and students with reading disabilities: How fast is fast enough to promote reading comprehension? Journal of Learning Disabilities, 51(2), 124-136.
https://doi.org/10.1177/0022219417691835
Powell, W. R. (1968). Reappraising the criteria for interpreting informal inventories. In D. L. DeBoer (Ed.), Reading diagnosis and evaluation (pp. 100-109). Newark, DE: International Reading Association.
Stanovich, K. E. (1984). The interactive-compensatory model of reading: A confluence of developmental, experimental, and educational psychology. RASE: Remedial & Special Education, 5(3), 11–19. https://doi.org/10.1177/074193258400500306
University of Oregon. (2020). Dynamic Indicators of Basic Early Literacy Skills (DIBELS, 8th ed.). Eugene, OR: University of Oregon. https://dibels.uoregon.edu
Comments
See what others have to say about this topic.