Dyslexia dilemmas: Are your shortcomings specific or general?

 

“Specific learning disability” is a baleful diagnosis: something is wrong with you, but the rot has not spread throughout the system. Much is still well with you, in that, like the curate’s egg, you are good in parts. Despite your specific shortcomings, you will no doubt wish to be judged as mostly good, albeit with a blemish which is worthy of commiseration. On the other hand, a generalized shortcoming puts you in a less fashionable category: you are merely dull. Hence the popular demand for research that shows that your shortcomings are highly specific, and preferably closely allied to creative genius.

A question arises: given that abilities are normally distributed, why should the bottom 2% be any different from the brighter 98%, other than being on the wrong side of the cutoff? There is nothing specific about that. A problem arises: since intelligence is general, because there is a positive manifold between abilities, how can a specific disability be detected? By common convention the answer is: when allowance has been made for general intelligence, any disability which remains must be due from some other specific cause. So, for a shortcoming to be considered specific, it must be 2 standard errors of prediction below the levels expected on the basis of your general intelligence.

I am fond of standard errors of prediction, particularly when they are seen in the light of the standard deviation of the original variable. For example, imagine you had to guess children’s reading scores, whilst knowing nothing about them. The least error prone strategy would be assume that each child was average. Your errors would be equal to the standard deviation. Next, assume you know their intelligence scores, and that intelligence correlates with reading at about r=0.6. Using a regression equation you can now predict their reading scores with more accuracy, though still with error. You will have reduced the uncertainty, and your errors will be smaller than the standard deviation by some measureable amount, which gives you a measure of the utility of the prediction.

As far as I know, there is little high quality epidemiology on reading, apart from the total population study on the Isle of Wight, and that is as far back at 1974. Let me know of more recent findings. Current fashion is for convenience samples, all with a diagnosis of dyslexia, drawn from individual school and university settings. Here as some population results.

W.Yule, M.Rutter, M.Berger and J.Thompson. Over‑ and under‑achievement in reading: distribution in the general population. British Journal of Educational Psychology (1974) 44,1‑12.

 

https://drive.google.com/file/d/0B3c4TxciNeJZeGFGVnU1NGEzVHM/view?usp=sharing

 

If you look at these population distributions you will see that 1) they vary in skewness and kurtosis 2) there tends to be a tail of poorer readers, but not in all samples 3) your conclusions about the distribution of reading ability would differ if you were restricted to analysing one population.

This is what the authors say in their abstract: Empirical findings are presented on the distribution of over- and under-achievement in reading in five general population groups encompassing four age-groups and two parts of the country-a major city and an area of small towns. It is shown that reading achievement does not exactly parallel IQ at all levels of intelligence, confirming the inappropriateness of the achievement ratio and like statistics. It is argued that over- and under-achievement are best defined in terms of a regression equation based on IQ scores. Defined in this way, reading ability follows a generally normal distribution, overachievement and under-achievement occurring with roughly the same frequency. However, there is a significant departure from normality at the extreme lower end of the curve such that gross under-achievement in reading occurs at well above the expected frequency. This suggests that there is a meaningful group of children with specific reading retardation which is not explicable simply in terms of the bottom of a continuum.

On close inspection of the paper, the “meaningful group” are absent in 2 of the 5 samples, (could be a prediction restriction) and there is very little “over-achievement” probably because of ceiling effects on the reading test. Notice the very healthy total population sizes. Every child was tracked down, so we can be confident about the population characteristics.

Now to modern times.

The structure of intelligence in children with specific learning disabilities is different as compared to typically development children David Giofrè, Cesare Cornoldi. Intelligence 52 (2015) 36–43

https://drive.google.com/file/d/0B3c4TxciNeJZVXkteDB6QVFtSlE/view?usp=sharing

Children with specific learning disabilities (SLDs) are characterized by a poor academic achievement despite an average intelligence. They are therefore typically assessed not only with achievement tests, but also with intelligence tests, usually the Wechsler Intelligence Scale for Children (WISC). The assumption of a discrepancy between IQ and achievement in children with SLD has been questioned, however, and the implications of using different measures in batteries of intellectual subtests have not been thoroughly investigated. The present study examined these issues, taking advantage of a large database of scores obtained in the ten core subtests of the WISC-IV by a group of 910 Italian children with a clinical diagnosis of SLD, who were compared with the children considered for national standardization purposes. Our results support the doubts raised concerning the IQ-achievement discrepancy model, showing that relevant discrepancies can emerge even within the WISC profile. The four main WISC-IV indexes were found differently related to intelligence (measured by means of the g-factor) and the g content of many subtests differed in children with SLD vis-à-vis typically-developing children. These results have important implications both theoretical, indicating that the g-factor is weakly identified in children with SLD children, and practical, indicating that the QI obtained with the WISC-IV may not be a good measure of intellectual functioning for children with SLD, which are discussed.

First, we do not know if the sample is representative of the population of those with specific learning disabilities, and reading difficulties are mixed in with other specific difficulties. Even though the diagnostic criteria follow ICD 10 we do not have epidemiological safeguards on this sample. They may simply be backward children who have pushy parents. They used the full 10 core tests of the Wechsler, which is good.

The learning disability sample are above average on Comprehension 10.92, Picture Concepts 10.98, and somewhat on Matrix Reasoning 10.56, and below on Digit Span 8.04, Letter Number Sequencing 8.34, Coding 8.28, and Symbol Search 9.17. Mind you, these results would not have surprised researchers 40 years ago.

To my eye there is not all that much difference in g variances between the standardisation sample and this specific learning difficulty group, other than in the usual culprits, Digit Span and Letter-Number Sequencing. Picture Concepts and Vocabulary are also somewhat discrepant, though less clear why.

 

I think this is an interesting paper, but in the absence of a better understanding of the representativeness of the children with learning disabilities studied I cannot be sure that their findings have the implications they state: when assessing intellectual abilities in children with SLD, it therefore seems reasonable to prefer a less biased measure, such as the General Ability Index. Giving the full Wechsler and showing the 10 subtest results would be far better, in my view.

Children with SLD struggle with working memory and processing speed tasks and this can considerably lower their IQ estimates, even though these factors are weakly related to the g-factor compared to typical developing children. Such evidence further confirms that intelligence and other basic aspects (e.g., working memory or processing speed) may be very highly related in the normal population but not necessarily in other groups, suggesting that these constructs should be considered separable.

In my view that is silly. If you struggle with working memory and processing tasks you have lower overall ability. In the decathlon of ability you do badly on some of the main events. If you are honest you would warn potential employers about your real shortcomings. If you want those aspects to be “separable” then you should accept that some employments and achievements are also “separable”.

Our results can also have clinical implications, indicating that it may be very difficult to assess intelligence in children with atypical development, so examiners will need to use their own “intelligence” when interpreting the results of such scales.

No, it is not more difficult to assess intelligence in children diagnosed with learning difficulties. The pattern of abilities on the Wechsler is a very useful guide to explaining strengths and weaknesses. Many children without a diagnosis of learning difficulty also have strengths and weaknesses in their abilities. I myself am particularly susceptible to drafts……

Nice study, but I disagree with what the results mean. See what you think.