Want to listen to this post?
Highlight text, click the speaker, and it will be read aloud to you,
thanks to GSpeech technology!
I have seen the Great Predictor of Reading Disabilities pop up #inthenews multiple, trusted sources, so I felt the need to comment. I have to admit that my thoughts on this aren’t 100% thought out yet, but I’m trying to keep up with the rate of the internet, so I’m publishing this while it’s still a work in progress in order to encourage answers to my questions and critical debate. I have a few scientific and a few ethical/pedagogical questions and issues about the study.
As context, WNYC reported on a study from Dr. Nina Kraus’s lab (at Northwestern University), which says that with a 30-minute mostly-passive EEG task (that requires pre-literate children’s brains to process signals in noise), we can confidently predict whether or not they will have a reading disorder, given a . Dr. Kraus reported that she would love to use this technology to test children at birth. The original research article can be found here.
Where to begin?
My gut reaction was that there would be too many false negatives and false positives to make the technology “worth it,” although it is minimally-to-moderately invasive, and not superbly expensive compared to fMRI (I’m saying this because I used to work in an EEG lab).
The false negatives would come from the heterogeneity of dyslexia. The underlying difficulty in dyslexia (or its official title “Specific Learning Disorder with impairment in reading and/or writing”) has generally found to be phonological (more on this later), fluency and comprehension are also elements of dyslexia, and I don’t think either are highly correlated with auditory processing (again, this is just a postulation and a work in progress!).
Fluency is (also) predicted by rapid naming (although the present study did demonstrate a relationship between their auditory task & rapid naming, a rapid naming task would be a greater predictor, and it can also be done while children are pre-literate, although granted, not at birth). Comprehension is better(?) predicted by language development, and that can be assessed at a young age, and regularly throughout the child’s preschool years. A large Finnish study from Dr. Torppa et al., revealed: “The strongest predictive links emerged from receptive and expressive language to reading via measures of letter naming, rapid naming, morphology, and phonological awareness.” Similarly: “The measures that most consistently predict future reading difficulty in English are phonological processing/ awareness, letter-name knowledge, and RAN (Pennington & Lefly 2001, Scarborough 1998, Schatschneider et al. 2004)” (as cited in Norton & Wolf, 2012)
I know, I know! Science is all about updating what you know, but those studies are robust and replicable, so I want to take this new “news” with a grain of salt.
To further describe the heterogeneity: What is described as “reading” above, however, refers to behavioral elements only. At a neuroscientific level, there are three important brain regions involved in reading (and therefore dyslexia) – the inferior frontal gyrus (“Broca’s area”), which is involved in spoken language production and word analysis, the (left) temporoparietal region has to do with word analysis and decoding (sound-letter matching), but that region is also involved in spoken language processing (hence the language development correlation, although presumably, an auditory test would be able to assess this region best). Third, the left occipitotemporal region is involved in rapid access to words and helps create fluent, automatic readers (something an auditory test would not be able to access/assess).
What about false positives? This is less of a concern since students who have difficulty processing sounds in noise may have a difficulty with hearing or auditory processing (which can correlate to reading difficulties), but as long as the test provides parents with information about that difficulty and doesn’t explicitly say that “your child is likely to have dyslexia,” the false positives can still support early intervention – though they would require follow up assessments to find the root of the problem (see the rest of the “Other Sources of Reading Difficulties” from Reading Rockets).
Back to the study: just how “predictive” is predictive? A mini-statistics lesson for you: assuming statistical significance, a 0.9 regression (which is considered “extremely high”) accounts for 81% of variance, meaning that if I get a 1/10 on a test, there’s an 81% chance that I will get 1/10 on that same test. So, how predictive are other assessments of reading? . Even with the “predictable” features listed above, at best they’re in the 60%-ish range, and the students from this study correlated with around 40% reliability with later tests of word reading and basic reading. That’s basically chance! (Again, commenters, please let me know where I’m misreading the statistics… it’s been a while!).
Herein lies the rub: When we’re looking at predictability, we need an initial assessment (in this case, the “Wall-E” neurophysiological task) and a follow-up assessment of a different type to demonstrate one’s ability to predict the other. The difficulty is, there is no single test of dyslexia, so as all good clinicians, the scientists used multiple follow-up measure to assess their predictability, including some predictive measures mentioned above (phonological awareness & rapid automatic naming), as well as other word/pseudoword and reading tasks. There need to be multiple tests used to assess dyslexia because it is multi-dimensional and heterogeneous, and a single neurological test neither gives you conclusive information about the “presence” or dyslexia, nor does it provide you with specific goals for intervention. Only multiple, dynamic, frequent assessments from an experienced clinician can do that!
A few final notes: the “trouble” with science is you can often prove something, but its opposite may also be (statistically significantly) true as well. There are other recent studies (for example the aptly titled, “Phonological but not auditory discrimination is impaired in dyslexia,” which had a much larger sample size but also older students) that suggest that dyslexia is not auditory in nature (and this meshes with most other theories of dyslexia I have read about). However, at this point, that is almost less relevant and less important to me than how we interpret these results as a society.
If it’s a crystal ball you would like, know that researchers at Yale are looking into the genetics of language impairments and reading difficulties (like dyslexia), and while I couldn’t figure out their prediction correlates (can anyone? This research was way out of my wheelhouse!), this line of research can also be promising for early detection
Lastly, I would be remiss if I didn’t mention that while the main positive of early detection is providing evidence-based therapies to reduce frustration and increase self-esteem, I want to be clear that dyslexia is a neurological difference, therefore even with early detection, individuals with dyslexia will always have dyslexia, even if they no longer have as much trouble reading as they would have without early detection. BUT there’s a bright side to this! That means, that even with early detection and early intervention, we will still have access to the innovations, art, writing, and charm of people with dyslexia. Read more about the strengths of individuals with Learning Disabilities like dyslexia here.
To sum up, my thoughts/concerns are:
- false negatives could provide an unrealistic sense of confidence (since dyslexia is so heterogeneous, we are likely to miss many students)
- false positives may guide parents and educators down the wrong path (e.g. if they have a hearing difficulty or auditory processing difficulty)
- there are other (arguably better?) predictors of reading disabilities (rapid automatic naming, phonological awareness, and speech and language development, as well as family history, and possibly genetic testing)
- overall, though, we’re bad at predicting from one assessment, alone, so multiple, frequent, dynamic assessments are most effective in assessing both students’ progress and current level of functioning
- early detection can lead to early intervention (which has demonstrably high success rates), but thankfully, that cannot minimize the positive side and strengths of individuals with dyslexia
- changing how reading is taught for ALL learners is what is most needed now: explicitly teaching sound-to-grapheme and grapheme-to-sound correspondence in a systematic, hierarchical, multi-sensory way is essential for students with dyslexia, but helps all learners. Not enough teacher training programs provide teachers with this training, therefore teachers are ill prepared to teach this way. So regardless of early intervention or not, we need more professionals who are reading research and using evidence-based instruction!