DIAGNOSTIC PROCESS: MORE DETAILS

There are several key issues that work together to make LD diagnosis complicated at least. These are:

1. different operationalizations of LD
2. misinterpretations of them, lack of understanding / qualifications, funding in the public school system, lack of funding in adult education, lack of cooperation with other agencies.
3. bias in diagnosis: regression toward the mean

Different operationalizations

In spite of the varying definitions of what constitutes a learning disability, only one measurement process (operationalization) is recognized: the discrepancy between achievement and potential (IQ). A key question is: how big should that discrepancy be? Different agencies in Texas use different cut-off points. For example, TEA lets the public schools use one standard deviation (interpreted to be 15 points) as the minimum allowable difference for LD to be diagnosed. On the other hand, Texas Rehabilitation Commission, which serves adults, considers two standard deviations (30 points) the minimum. This effectively means that someone can be diagnosed LD in one system and not in another. The differences in standards are, at least in part, due to (intentionally) ambiguities in the DSM-IV, the diagnostic manual for psychologists. In order not to pigeonhole LD too much as a specific disorder, language was intentionally kept 'soft'. The 'two-standard-deviation' criterion also stems from the DSM-IV. However, a two-SD criterion is merely suggested, not prescribed. Traditionally, for any diagnosis to be considered 'severe', a two-SD deviation from the norm has been used as a cut-off. Considering the seriousness of such a diagnosis, it would seem appropriate to use equal standards within a state, and of course, preferably nationwide. However, this is not the case.

Misinterpretations of these operationalizations

It is commonly believed that one SD equals 15 points and 2 SD's equals 30 points. This may be true for standardized test results like achievement and IQ, but not for the difference between the two. When subtracting one score from another, a third score, with its own distribution characteristics, is created. In the case of LD, it has a mean of zero (on average, people will perform up to their potential, so the average difference of IQ and achievement is expected to be zero), and a SD not equal to 15. The exact size of the SD depends on the correlation between the IQ test and the achievement test scores. In general, the SD of the difference score will be around 7 or 8. But in practice, 15 points or 30 points is the norm - leading to a more stringent cutoff score in order to be considered LD.

Regression toward the mean

Regression toward the mean is a statistical artifact that often goes unrecognized, as it does in the case of LD diagnosis in Texas. The basic implication of this phenomenon is that people who score above average have likely been helped, rather than hurt, by measurement error. Those who score below average have likely been hurt, rather than helped. To put a number on this: if someone scores 130 on an intelligence test, it is quite likely that their 'true' intelligence is closer to 125 on most tests. If they then take the next test, in this case achievement, their score will drop almost automatically. In fact, this person 'needs' to score 115 to be considered LD. However, in reality, their 'true' discrepancy is only 10. This leads to an over identification of LD among more-than-average intelligent people. Conversely, below-average people will be under identified. Those who score 70 on their IQ, likely have a 'true' score of 75. In order to be considered possible LD, they need to score 15 below their score, which is 55. However, it is 20 below their true score, so the discrepancy needs to be much larger than intended for it to show up as a learning disability.

What this means is that the way in which LD is diagnosed in Texas discriminates on the basis of ability (intelligence). The more intelligent the people are, the smaller the discrepancy is that needs to exist in order to (possibly) be considered LD.

This can easily be countered, by adopting a so-called regression model. Such a model compensates for this regression effect. In fact, Dr Cecil Reynolds, School Psychology professor at Texas A&M University, chaired a federal task force in the early '80's, tackling the problem of diagnosing LD. The committee recommended adopting a regression-model, and 16 States changed their laws to adopt such a model. Texas was not one of them.