This study focused on the use of radioimmunoassay of progesterone in milk for the diagnosis of post-partum ovarian cyclicity and accurate detection of oestrus and non-pregnancy in cows in the artificial insemination (AI) programme in Bangladesh. In Investigation 1, milk samples were collected on day 0 (day of AI), day 9-13 and day 21-24 from 444 milking cows of various breeds presented for the first postpartum insemination by 413 farmers living at 182 villages/regions in Mymensingh District from 6 AI centres and sub-centres. Each cow was then examined three times after each AI until it stopped returning to oestrus. Sixty to 90 days after the last AI, the cows were examined per rectum to confirm the pregnancy. Milk progesterone data on day 21-24 contributed to a clear diagnosis with respect to non-pregnancy in 100% cows, indicating a possible use of this progesterone assay for identifying non-pregnant cows in AI programmes. In Investigation 2, milk progesterone was monitored two times in a month with a 10-day interval in 88 cows. The samples were taken between 10 days after calving and the first detected oestrus, followed by two more samples 10 days apart. The proportion of cows accurately detected in oestrus was 30%. Another 30% were stated to be in oestrus when they were not (false positive) and 40% were not detected when they were in oestrus (false negative). The mean intervals between calving and oestrus and between calving luteal activity were 40 to 362 days (median=120, n=82) and 34 to 398 (median=111, n=64) days, respectively. The body condition scores at calving and at the initiation of luteal activity influenced the interval between calving and luteal activity (p < 0.05). Cows suckled twice daily initiated luteal activity earlier than their counterparts suckled several times daily (p < 0.05). Determination of progesterone in milk on day 21-24 is a good means for detecting non-pregnant cows.