Empty body (EB) composition data from 146 lactating Holstein–Friesian dairy cows were analysed, with cows selected from a large herd to represent a wide range of animal factors including parity, live weight (LW), body condition score (BCS), milk yield and stage of lactation. The objectives of the study were to examine possible relationships between EB composition and live animal characteristics and to develop prediction equations for EB weight (EBW) and EB composition. EBW was strongly related to LW with a R2 of 0·90, and addition of BCS and stage of lactation increased the R2 to 0·93 and reduced the s.e. value from 19·8 to 16·8. There was a strong relationship between LW and EB mass of crude protein (CP) or water (R2=0·89 or 0·85), and addition of BCS, milk yield and stage of lactation as supporting predictors gave little improvement in the relationship between LW and EB mass of CP or water. Similar results were also obtained for the prediction of EB ash mass, but the relationship between EB ash mass and LW or a combination of LW and parity was relatively poor (R2=0·59 or 0·63). However, including BCS and milk yield as supporting predictors in the linear relationship between LW and EB masses of lipid (0·51 v. 0·79) and dry matter (DM) (0·79 v. 0·91) and total gross energy (GE) content (0·66 v. 0·85) considerably increased the R2 values. Similar linear and multiple relationships were also developed to predict EB component proportions of lipid, CP, GE, DM and water, and the R2 values were much smaller than those for EB component masses. The equations for prediction of EB component masses were evaluated through internal validation, by developing a range of similar new equations from two-thirds of the present data and then validating these new equations with the remaining one-third of data. The validation indicated that addition of other live animal factors to support LW to predict EBW and EB masses of lipid, GE and DM considerably increased the prediction accuracy, but had little effect on the prediction accuracy for EB masses of CP, water and ash. It is concluded that in lactating dairy cows, LW is a very good predictor of EB masses of CP and water. Accurate prediction of EBW and EB masses of lipid, GE and DM can be obtained using LW together with BCS, milk yield and stage of lactation.