Alzheimer's disease (AD) is thought to be caused by accumulation of amyloid-β protein (Aβ), which is a cleavage product of amyloid precursor protein (APP). Transgenic mice overexpressing APP have been used to recapitulate amyloid-β pathology. Among them, APP23 and APPswe/PS1deltaE9 (deltaE9) mice are extensively studied. APP23 mice express APP with Swedish mutation and develop amyloid plaques late in their life, while cognitive deficits are observed in young age. In contrast, deltaE9 mice with mutant APP and mutant presenilin-1 develop amyloid plaques early but show typical cognitive deficits in old age. To unveil the reasons for different progressions of cognitive decline in these commonly used mouse models, we analyzed the number and turnover of dendritic spines as important structural correlates for learning and memory. Chronic in vivo two-photon imaging in apical tufts of layer V pyramidal neurons revealed a decreased spine density in 4-5-month-old APP23 mice. In age-matched deltaE9 mice, in contrast, spine loss was only observed on cortical dendrites that were in close proximity to amyloid plaques. In both cases, the reduced spine density was caused by decreased spine formation. Interestingly, the patterns of alterations in spine morphology differed between these two transgenic mouse models. Moreover, in APP23 mice, APP was found to accumulate intracellularly and its content was inversely correlated with the absolute spine density and the relative number of mushroom spines. Collectively, our results suggest that different pathological mechanisms, namely an intracellular accumulation of APP or extracellular amyloid plaques, may lead to spine abnormalities in young adult APP23 and deltaE9 mice, respectively. These distinct features, which may represent very different mechanisms of synaptic failure in AD, have to be taken into consideration when translating results from animal studies to the human disease.