[p2–452]: new analyses of language processing reveal that semantics dissociates from syntax in mild cognitive impairment

Semantic memory deficits are evident in MCI. While recent evidence has also shown subtle differences in more complex language production there is little evidence on the mechanisms by which semantics and syntax interact in MCI. In this inter-institutional research, we evaluated the roles of semantics and syntax in MCI, amnestic type (n=21, aged 68–88)) and Healthy Aging (n=25, aged 62–87). Subjects were presented with an Elicited Imitation test, in which they repeated 60 sentences including Relative clauses (RC), Coordinate (CS) and Adverbial subordinate (ASC) sentences. Sentences were controlled for length and lexical frequency and varied by semantic plausibility with half semantically implausible. Subjects were also administered the Addenbrook's Cognitive Estimation Test-Revised (ACE-R) which provides a total score and scores within different cognitive domains. For both groups, there was a significant difference in elicited imitation success favoring semantically plausible sentences, and no significant interaction between semantics and group. Although MCI subjects were significantly less accurate in imitating sentences relative to HA for all sentences types (RC: HA > MCI (p=0.0012, t=3.35); CS: HA > MCI (p=0.0077, t=2.71); ASC: HA > MCI (p=0.0052, t=2.86)), both groups showed the same pattern of sentence success across both semantic conditions. Regression tests of cognitive variables against linguistic performance suggested a significant predictive effect of cognition in HA, but not MCI. Analysis of ACE-R components revealed that phonemic, but not semantic fluency significantly predicted linguistic success in both groups. Findings suggest that in MCI semantic and syntactic deficits are to some degree divorced. A more complete analysis of language is necessary to strengthen neuropsychological profiling in clinical MCI assessment. Findings suggest that a distinction of MCI from HA does not involve cognitive deficit in semantic memory alone. Given that MCI subjects were less successful on this task in both semantic conditions and that general cognition and semantic fluency do not predict sentence production, we suggest that the ability to integrate semantic memory with the syntactic computation necessary for language comprehension and production may explain our results.