Most cognitive tests are standardiased to detect cognitive decline, e.g., in establishing a neuropsychological diagnosis. They are designed to be sensitive to deficit, with their norms quantifying sub-normal performance. But how well can the same tests measure superior performance? e.g., what would be an equivalent of IQ=140 in the Stroop, the Wisconsin Card Sort or Flanker test? would it be expressed in respnse velocities, error rates or both? how would cogntive performance norms look like for non-clinical applications? e.g., to measure educational gains from a training program? can they be baselined and calibrated on a broadly representative sample? are there any such norms in existence?
Thank you.
Dear Eugene, I don't see a categorical difference in the way how norms for IQ or other cognitive measures are/could be developed (i.e., Digit-span, Verbal fluency, or Stroop test). However, the problem may be in part due to the great variety in task parameters (i.e., non-uniform administration), which influences the outcome variable. These aspects are usually decided ad hoc, i.e., fitting the current experimental procedure as a whole. But if data from a representative/stratified sample, including all important aspects (e.g., sex, age), were collected, such data (i.e., response times, errors, memory span, fluency scores) could be transformed into a standardized form (such as IQ 100+/-15, T-scores, or Stens). In non-clinical settings, I would say this might not be as important (correlation, comparison, and training effect / gain score can be usually assessed using absolute/raw scores). On the other hand, some form of standardized scores might facilitate the outcomes of comparative studies, especially if we consider how arbitrary comparison groups can be (e.g., a convenient sample median split used for comparing "high"- vs "low"-achieving individuals). Looking forwards for other opinions, best. M.
Thank you, Martin. I agree that for a typical research question you can get by with raw timings and error rate values, wihtout even bothering to fit percentile rank distributions under a Gaussian curve to derive standard scores. But practical applications - such as occupational selection - would require mature individual differences metrics that are meaningfully baselined and calibrated. The Vienna Test System seems like an attempt at developing such calibrated metrics but I am not sure how it would handle non-clinical references pointing to high-performing individuals. E
There's no reason one could not develop norms for any cognitive function that span the range from marked deficit to extraordinarily high performance. Of course doing so requires one to recruit substantial samples of more-or-less ordinary people (i.e., not just patients) and so can be costly, but that's purely a practical problem, not a conceptual one. (I think at least one published version of the Stroop does yield standard scores, by the way.)
The test designer's main challenge here is to devise tasks that even people with deficits can perform to some degree, and that even people who are not just intact but gifted will find challenging. If it's too difficult, you get floor effects (lots of people getting the lowest possible score). If it's too easy, you get ceiling effects (lots of people maxing out). The standard Wisconsin Card Sort Test, for example, is really pretty easy - most neurologically intact people of average-range intelligence will master the maximum number of categories.
Thank you, Stephen. You have nailed my question: do we have a test that would produce neither ceiling nor floor effect?
Quite a few, I should think. Any supraspan learning test (such as RAVLT or SVLT), most working memory tests (such as Digit Span or Spatial Span/Corsi Blocks), anything that involves speed (such as Stroop, Coding, etc.) - for starters.
- Badges
- Science topic
- Similar topics
- Psychology