History and Development of Above-Level Testing of the Gifted

none - comprehensive literature review covering above-level testing from 1919 to 2011

High-ability children given tests designed and normed for older students; the empirical base is overwhelmingly US Talent Search programmes (SMPY and successors)

A narrative literature review, so grade D by the rubric regardless of how carefully it is done - and it is done carefully, including a section that grades each rationale for the practice against the evidence rather than asserting it. Its most useful property for this archive is that the author is not a Talent Search insider and says so: he reproduces Matthews' (2008) criticism that the entire Talent Search evidence base has "a total lack of random assignment or sampling, an operational definition that equates giftedness with a high test score, and a lack of economic or cultural diversity."

Above-level testing means giving a child a test written and normed for older students. Five rationales are given for it in the literature and their evidential support is very uneven. (1) RAISING THE CEILING is the best-supported: the author reports being unable to find any study in which a group of gifted children failed to score higher on a test two or more levels above their age group than the maximum scale score of the grade-level test. Children at or above the 95th percentile on a grade-level test typically score around the average for students four or more years older on an above-level test. (2) INCREASED VARIABILITY AND DISCRIMINATION is well supported - above-level scores approximate a normal distribution where grade-level scores pile up at the ceiling, and they discriminate into the top 0.01%. (3) IMPROVED RELIABILITY is theoretically sound but essentially unevidenced: the only reliability coefficients from an above-level ACHIEVEMENT test given to a gifted sample that the author could locate are from Stanley (1951). (4) COMPARABILITY of interpretation rests on a single measurement-invariance study (Minor & Benbow 1996) which the author judges flawed. (5) REDUCED REGRESSION TOWARD THE MEAN "has not been empirically tested." The operational rule the literature converges on: do not test a child four or five levels above grade until they can already score at the 95th percentile or higher on a regular grade-level test.

Not applicable - a review of measurement practice. It touches the archive's premise only in passing, by treating high ability as something to be measured accurately rather than caused.

A review rather than a finding. The ceiling-raising claim is supported by many independent studies; the reliability and regression claims are supported by essentially none, which is the review's own verdict and the reason this source is worth holding.