Objective: Development and validation of a short Observation List of possible early signs of Dementia (OLD) for use in general practice. Design: Stepwise development using reviews of publications and expert consensus. Field study for evaluation of reliability. Validation study (interviews, family forms) using existing valid and reliable measures. Use of data reduction techniques to construct a short version. Setting of field study: Twenty-two GPs in 19 Dutch practices. Participants: The first two patients seen on 15 working days (n=470) were observed. Inclusion: age > 75, without a known diagnosis of dementia. Exclusion: psychiatric treatment, severe depression, acute illness with confusion. Division of patients into three groups with no, intermediate, and the most signs (total of interviewed patients, n=60; family forms, n=39). Outcome measures: Reliability (Cronbach's alpha and factor-analysis). Convergent validity using the Cognitive Screening Test (CST), the Word Learning Test (WLT, total and retention), the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE), the Groningen Activities Restriction Scale (GARS), and an IADL scale. Discriminant validity using the geriatric depression scale (GDS). Construct validity using a Principal Component Analysis (PRINCALS). Incremental validity using the intuitive opinion of the GP (McNemar test). Results: Reliability in the total group 0.88, first factor explained variance 42.5%. Convergent validity (two-way ANOVA) results: CST (p=0.00), WLT-total (p=0.001), WLT retention (p=0.00), IQCODE (p=0.09). No statistically significant differences for GARS and IADL. GDS (p=0.30) not different. PRINCALS first factor explained 48% of variance. The OLD added to the GP opinion (McNemar p=0.00). Reliability short version 0.89 (interviewed group), 0.86 (total group). Conclusion: The OLD is a valid and reliable method to detect early signs of dementia in general practice that can indicate when it may be useful to employ existing screening instruments.