The Value of a Number
A while ago, I completed an assessment for a student who had transferred from another school district. The student’s family had moved while the evaluation was in process. The sending school had finished the standardized testing, and I completed language sampling, an observation, an interview, teacher rating scales, a hearing screening and the report. The district administrator and clinician from the other district attended the eligibility meeting. I shared results of the evaluation. The student had received a standard score of 79 on a comprehensive language test. The staff from the other district announced that if the student had remained in their district, he would not have qualified for services, as they used a cutoff score of 78. I stated that I weighed multiple factors in the decision-making process and was not required to decide the presence or absence of a disability using one sole numeric criterion.
At the federal level, standardized scores are not specified. The Individuals with Disabilities Education Act (IDEA) states that a communication disorder “adversely affects a child’s educational performance.” It does not list a test score. At the state level, each state receives funds from the federal government for establishing their own rules, regulations and policies to support children with disabilities. At the local level, each local educational agency (LEA), like a school district, is required to have their own “policies, procedures and programs that are consistent with the State policies.”
A few additional notes:
· Significantly discrepant: Some districts use cutoff or guideline scores that they consider to be significantly discrepant from the normative sample (typically -1, -1.5 or -2 standard deviations).
· Performance: Standardized tests measure performance at a given moment in time. We cannot measure underlying knowledge of linguistic concepts. We can only measure an individual’s physical behaviors (speaking, pointing, etc.) in response to sets of decontextualized stimulus items.
· Standard measure of error: There is no true test score. Standard measure of error is an estimate of an individual’s performance on the same measure with multiple administrations (without the effects of learning the test items). It shows the likely range of that person’s performance.
· Optimal cut-score: Researchers determine the best score to prevent over or under-identification of disorders. Sensitivity shows the probability that people who have a disorder will test positive for the disorder. Specificity shows the probability that people who do not have a disorder will test negative for a disorder. Researchers find the best balance of sensitivity and specificity to recommend an optimal cut-score. The Clinical Evaluation of Language Fundamentals-Fifth Edition has an optimal cut-score of 80 (-1.33 standard deviations).
· Missed items: Two different individuals could receive the exact same standard score and miss completely different stimulus items. The score alone does not provide any information about areas of strengths and challenges.
We have a cultural belief that standardized testing is an accurate measurement of knowledge and skills throughout the educational system, with benchmarks from preschool through college. Perhaps it’s time to question the origin of cutoff scores and what one number actually means. We can think about how we are interpreting numerical data and the significance that we give it.