Disclosed are methods for determining the pregnancy status of a cow by detecting a pregnancy-associated glycoprotein (PAG) level that comprises proteins present in an acidic protein fraction of day 55 to day 60 bovine placental tissue. These methods may involve using novel polyclonal and monoclonal antibodies that have been generated using the acidic protein fraction as an immunogen. According to the method, the cow is pregnant when the PAG level is present at an elevated level in a sample obtained from the cow. The invention provides accurate methods of detecting pregnancy at early stages and has the benefit of allowing early post partum detection of pregnancy status with few false positive results. The antibodies for detecting the PAG level will also be useful in combination with detecting elevated progesterone levels, thus providing an even more effective pregnancy detection method. The benefit of this early pregnancy detection is that identifying cows that are not pregnant very shortly after breeding allows for timely rebreeding and minimizes the amount of time the cow is open. Also disclosed are methods for making a breeding decision for a cow. These methods allow a herd manager to make breeding decisions based on the PAG level and optionally, progesterone levels, detected in a biological sample taken from a cow suspected of being in the early stages of pregnancy.