As commonly happens in epidemiological research, none of the reported studies were totally free of methodological problems. Studies have considered the influence of social relationships on dementia, but the mechanisms underlying these associations are not perfectly understood. We look at the possible impact of selection bias. For their first memory consultation, patients may come alone or accompanied by a relative. Our objective is to better understand the impact of this factor by retrospective follow-up of geriatric memory outpatients over several years. All patients over 70 who were referred to Bretonneau Memory Clinic for the first time, between January 2006 and 2018, were included in the study. The patients who came alone formed group 1, the others, whatever type of relative accompanied them, formed group 2. We compared the Mini-Mental State Examination (MMSE) scores of patients; and for all patients who came twice for consultation with at least a 60-day interval, we compared their first MMSE with the MMSE performed at the second consultation. In total, 2,935 patients were included, aged 79.7 ± 8.4 years. Six hundred and twenty-five formed group 1 and 2,310 group 2. We found a significant difference in MMSE scores between the 2 groups of patients; and upon second consultation in group 2, but that difference was minor in group 1. Our finding of a possible confounding factor underlines the complexity of choosing comparison groups in order to minimize selection bias while maintaining clinical relevance.