domains compared to HC (immediate recall F(3, 70) ¼ 6.00, p < 0.01 and delayed recall F(3,70) ¼ 3.51, p < 0.05) whilst GRN scores are comparable to HC. GRN patients performed on the other hand significantly lower on visuoconstructive ability than either HC or MAPT and C9orf72 (F(3, 64) ¼ 5.06, p < 0.01). Conclusions: These results indicate that MAPT, GRN and C9orf72 mutation carriers differ to the same amount from HC at the typical FTD domains, but distinction between the mutations can be made when comparing less typical FTD domains, namely memory and visuoconstruction. This could enable us to examine NPA as biomarker for upcoming clinical trials. P3-319 COGNITIVE PRESERVATION AND CHANGE ACROSS THREE WAVES IN MILDALZHEIMER’S DISEASE AND NON-DEMENTED CONTROLS Amber Watts 1,2 , David Kevin Johnson 2 , Eric D. Vidoni 2 , Jonathan D. Mahnken 2 , Jeffrey M. Burns 2 , 1 University of Kansas, Lawrence, KS, USA; 2 University of Kansas Alzheimer’s Disease Center, Fairway, KS, USA. Contact e-mail: amberwatts@ku.edu Background: Distinguishing cognitive impairment from normal aging, requires identification of age-related cognitive changes that differ from dementia-related cognitive changes. We aimed to establish a neuropsychological test battery that clearly differ- entiates AD from age-related cognitive change. We evaluated the pattern of cognitive changes over two years in N ¼ 515 older adults with mild AD (N ¼ 230) and without (n¼285). Methods: We used factor-analytically derived cognitive scores for attention (digits forward, digits backward, letter number sequencing), ver- bal memory (logical memory, Boston naming, selective remind- ing), and visuospatial speed (trails A, block design, digit symbol substitution). Growth curve models tested declines in each factor over three waves, each one year apart. Results: Our three factor model fit the data well (CFI ¼ 0.982, RMSEA ¼ 0.061) and was invariant across the three waves. A linear growth curve model fit the data well (CFI > 0.90). In cognitively normal participants, attention (B ¼ .095, p <.001), verbal memory (b ¼ .801, p ¼.001), and visuospatial speed (B ¼ .105, p <.001) improved over three waves. In the AD group, attention (b ¼ -.576, p ¼ .017), verbal memory (b ¼ -.434, p ¼ .001), and visuospatial speed (b ¼ -.649, p <.001) declined over three waves. Higher education was associated with higher baseline performance for visuospatial speed (b¼.178, p ¼.031 Normal; b¼.172, p ¼.030 AD), attention for controls (b¼.215, p ¼ .002 Normal; b¼.104, p ¼ .197 AD) and verbal memory (b¼.183, p ¼.008 Normal; b¼.151, p ¼ .055 AD). Conclusions: Our model detected different trajectories between cognitively normal participants and those with AD. While par- ticipants with AD declined reliably over time in all domains, cognitively normal participants reliably improved over time. It is important to recognize that cognitive declines are not inevi- table in these domains over a period of 2 years. Some scholars distinguish “successful cognitive aging” (maintenance of cogni- tive ability), from “normal cognitive aging” (age-normative cognitive decline) to emphasize this point. Improvements may be due to learning, practice effects, or attrition of the poorest performers. As in previous studies, educational attainment was associated with initial cognitive performance but not with decline. P3-320 FALSE ALARMS, RECOLLECTION, FAMILIARITY, AND CONFIDENCE IN RECOGNITION MEMORY IN HEALTHY ELDERLY INDIVIDUALS: ASSOCIATIONS WITH HIPPOCAMPALVOLUMES Dorothee Schoemaker 1 , Serge Gauthier 2 , Jens C. Pruessner 3 , 1 McGill University / McGill Centre for Studies in Aging, Montreal, QC, Canada; 2 McGill University Research Centre for Studies in Aging, Verdun, QC, Canada; 3 Centre for Studies in Aging, McGill University, Montreal, QC, Canada. Contact e-mail: dorothee.schoemaker@mail.mcgill.ca Background: Recognition performance in normal or neuropatholog- ical aging is characterized by an increase rate in false recognition. Dual-process theories posit that two separate and independent pro- cesses are involved in recognition: recollection and familiarity. In a recent literature review (Schoemaker et al., 2014), we suggested that an increased reliance on familiarity could be responsible for the increase in false recognition observed over the course of aging. However, it has also been argued that recognition confidence differ- entiates between familiarity and recollection, and that false recog- nitions are associated with lower recognition confidence. In the current study, we studied associations between recollection, famil- iarity, confidence and false alarms in a sample of healthy aging in- dividuals. We further investigated possible associations between these variables and hippocampus volumes. Methods: 86 individuals aged 55-80 (mean: 64.9, SD: 6.5) participated in the study (M/F:41/ 45). Participants completed a computerized and adapted version of the process dissociation procedure (Jacoby, 1991). From this task, we derived hit, recollection, familiarity, false alarm and high confi- dence rates. In 84 participants, we acquired a T1 MPRAGE (Siemens, 3T, voxel size: 1 x 1 x 1.2mm). Bilateral hippocampal volumes were automatically segmented using a patch-based seg- mentation algorithm (Coupe et al., 2011), and manually corrected according to previously established guidelines. Results: Recollec- tion was significantly associated with high confidence rate (r¼0.27, p<0.01), but not with false recognition (r¼0.04). Famil- iarity was significantly and associated with false alarm rate (r¼0.26, p¼0.02), but not with recognition confidence (r¼0.01). Finally, false alarm rate was not significantly associated with recog- nition confidence (r¼-0.06). Significant correlations were found between bilateral hippocampus volumes and recollection rate (r¼0.39, p<0.001). This correlation remained significant after con- trolling for age, education and gender (p¼0.006). No significant as- sociations were found between hippocampus volumes and any other variables. Conclusions: These results indicate that false alarms are associated with an increase in familiarity-based recognition rather than a decrease in recognition confidence. We further show that, while there was no significant correlation with hit rate, recol- lection rate was significantly associated with hippocampal volume. This suggest that recollection performance might be a more sensi- tive cognitive marker of neurodegenerative conditions affecting hippocampus integrity, such as Alzheimer Disease, than typical yes/no recognition paradigms. P3-321 INFORMATION-THEORETIC MODEL AVERAGING TO CHARACTERIZE EFFECTS OF AGE, GENDER, AND APOE Ɛ4 ON MID-LIFE COGNITIVE TRAJECTORIES: RESULTS FROM THE WISCONSIN REGISTRY FOR ALZHEIMER’S PREVENTION Rebecca L. Koscik 1 , Derek L. Norton 1,2 , Lindsay R. Clark 2 , Kimberly D. Mueller 1,3 , Bruce P. Hermann 1,4 , Corinne D. Engelman 1,2,5 , Poster Presentations: Tuesday, July 26, 2016 P966