The purpose of this article is to analyse item bias in a measure of self-reported functional ability among 75-year-old people in three Nordic localities. The present item bias analysis examines whether the construction of a functional ability index from several variables results in bias in relation to geographical locality and gender. Information about self-reported functional ability was gathered from surveys on 75-year-old men and women in Glostrup (Denmark), Göteborg (Sweden) and Jyväskylä (Finland). The data were collected by structured home interviews about mobility and Physical activities of daily living (PADL) in relation to tiredness, reduced speed and dependency and combined into three tiredness-scales, three reduced speed-scales and two dependency-scales. The analysis revealed item bias regarding geographical locality in seven out of eight of the functional ability scales, but nearly no bias in relation to gender in the combined data. The conclusion is that only one tiredness-scale (Lower Limb-T), one reduced speed-scale (Lower Limb-S) and the two dependency-scales (PADL-H and Mob-H) can be proposed for use in comparisons between the three localities (with removal of one or more items).