The heritability and stability over a 19 year period of long (23-item) and short (12-item) versions of Eysenck's Neuroticism scale were compared in a large Australian twin-family sample. Stability over 19 years of the 23-item Neuroticism scale was 0.62 and for the 12-item scale 0.59. Correlations between scores obtained by mailed questionnaire and telephone interview a few weeks apart were 0.87 for the long scale and 0.85 for the short scale; scores obtained by mail were slightly higher, particularly for females. The 12-item scale had slightly reduced power to discriminate both high and low scoring individuals on the full 23-item scale. Mean Neuroticism score for the 12-item scale was atypically low when compared to the distribution of the complete set of scores for all possible combinations (>1 million) of 12-items drawn from the full 23-item EPQ-R. Mean heritabilities for the lowest and highest 300,000 of these combinations were 43.2% and 42.7%, respectively, somewhat higher than the 41.0% for the actual EPQ-R-S 12-item scale. Heritability for the 23-item scale was 46.5%. We conclude that there is little loss of either stability or heritability in using the short EPQ-R scale, but the choice of which 12-items could have been better. © 2005 Elsevier Ltd. All rights reserved.