Milking cows with a 1-min manual stimulation (treatment 1) and without any udder preparation (treatment 2) was compared by application of an improved, highly sensitive radioimmunoassay for oxytocin and recordings of milk-flow curves. Both treatments caused the release of oxytocin, but treatment 2 generally seemed to be less efficient. Milking characteristics supported the advantage of manual stimulation; milk yield and milk flow were significantly higher, while 'machine-on' time was shorter. This clearly indicates the importance of the right timing of release of oxytocin before commencement of milking. Substitution of stimulation by an i.v. injection of 0·5 i.u. oxytocin (treatment 3) resulted in milking parameters very similar to those of treatment 2. This implies that manual stimulation has other effects besides the secretion of oxytocin which are also responsible for optimal milk removal.