Semantic distances of WAIS Similarities word pairs in non-demented adults: An item-level index of semantic memory granularity

Abstract

Innovative approaches to test scoring can help neuropsychologists detect subtle semantic memory alterations. We focussed on the Wechsler Adult Intelligence Scale Similarities (SIM) test and calculated an item-level score based on the ‘Leacock & Chodorow ‘(LCH) semantic distance expressed by each SIM item. We hypothesised that LCH would predict 1) performance on standard semantic memory tests; 2) an Alzheimer aetiology; 3) transentorhinal grey-matter integrity. Six hundred sixty-nine non-demented participants completed a neuropsychological battery inclusive of SIM and consolidated tests of semantic memory and executive functioning. Hierarchical linear regressions were designed to test the association between LCH and semantic memory performance after controlling for major confounders. A hierarchical logistic regression was then designed to test the association between LCH and underlying aetiology (Alzheimer/cerebrovascular) in a mild cognitive impairment sub-cohort. Finally, we tested the association between LCH and both whole-brain grey-matter density and transentorhinal thickness using voxel-based-morphometry and region-of-interest models. LCH predicted semantic memory performance but not on a test significantly supported by executive resources. LCH also predicted clinical aetiology and grey-matter density in the transentorhinal cortex and in other regions involved in linguistic-semantic processing. No significant association was found with regional thickness. Post-hoc LCH scoring in 89 people with dementia revealed the presence of a gradient of diagnostic severity, i.e., healthy adults < mild cognitive impairment < dementia. Item-level scores of SIM performance are associated with neurocognitive constituents of semantic memory. LCH is a valuable construct that could help clinicians detect semantic memory decline in ageing adults with suspected neurodegeneration.