Specific neuropsychological test batteries may be more sensitive to cognitive decline in early Alzheimer's Disease (AD) than general cognition measures like ADAS-cog or CAMCOG. Establishing clinically meaningful change using the most sensitive measures of cognition across the spectrum of dementia severity will aid the evaluation of new interventions. This study examined the ability of the ADAS-cog (total), the CAMCOG (total score and learning memory), Hopkins Verbal Learning Test- delayed recall (HVLT-DR) and the CANTAB computerized cognition battery (Paired Associates Learning[PAL], choice reaction time[CRT] and spatial working memory [SWM]) to discriminate among healthy elderly (HE, n = 31), amnestic mild cognitive impairment (aMCI, n = 20) and mild to moderate Alzheimer's Disease (AD, n = 19) and to detect change in cognitive function over 1 year. In addition to the cognitive tests, functional status (ADCS-ADL) and self-reported memory impairment (EMQ) were assessed at baseline, 6 months and 1 year. CAMCOG total, CANTAB PAL, and HVLT-DR discriminated among HE, aMCI, and AD. CAMCOG learning memory did not discriminate between HE and MCI and CANTAB SWM did not discriminate between MCI and AD. ADCS-ADL scores showed a non-significant trend toward functional decline among AD and aMCI subjects but not among HE. Both Camcog total score and ADAS-cog indicated significant cognitive decline over 1 year among AD subjects but not among MCI subjects. HE and aMCI showed significant learning effects on the PAL total errors at 1 year. Neither HVLT-DR nor SWM showed decline at 1 year for HE or aMCI. AD subjects showed no change in PAL, HVLT DR or SWM at 1 year. Floor effects were observed for HVLT DR and PAL among AD subjects. Measures of episodic memory, like PAL and HVLT-DR may be better suited to screening for cognitive impairment than evaluating treatment outcomes in aMCI over a 1 year period. Longer follow-up may be required to detect progression of cognitive decline in aMCI.