The post-1994 SAT while correlated with intellectual ability to some extent has an attenuated relationship with g relative to the pre-1994 version. De-emphasizing on analogies, raw vocabulary, the inclusion of a calculator and its removal from being an admission tool to Mensa all underscore this.

The claim breaks down into several components: 1) The post-1994 SAT is still correlated with intellectual ability but less strongly with general intelligence (g) than the pre-1994 SAT. 2) The removal or de-emphasis of analogies and raw vocabulary questions contributes to this attenuation. 3) The inclusion of a calculator and the SAT's removal as an admission tool for Mensa further underscore this shift.

Evidence shows that analogy questions were removed in 2005, which aligns with the claim about de-emphasizing analogies.en.wikipedia.orgen.wikipedia.orgSAT Vocabulary testing on the SAT has evolved over time, moving away from raw vocabulary and analogies toward context-based questions, which may affect its measurement of g. The inclusion of a calculator starting in 1994 changed the math section dynamics, potentially influencing cognitive demands. Mensa no longer accepts SAT scores for admission, reflecting a recognition that the test may not fully capture high-level general intelligence as measured by IQ tests.

However, research indicates that the SAT remains substantially correlated with measures of general cognitive ability and can serve as a proxy for intelligence, though some psychometricians note changes in test design have nuanced this relationship.pmc.ncbi.nlm.nih.govpmc.ncbi.nlm.nih.govWhat We Know, Are Still Getting Wrong, and Have Yet to Learn about the... Thus, while the relationship between the SAT and g may be attenuated relative to earlier versions, it is not eliminated.

The reasoning behind the verdict is that the claim is largely supported by historical changes to the SAT and psychometric research but lacks precise quantification of how much attenuation has occurred. The claim is mostly true but would benefit from more detailed empirical data on correlation coefficients before and after 1994.