Monthly and fortnightly milk production records were analysed from 59 autumn-calving Angus and Angus × Holstein crossbred cows. Half the cows had been administered 10 mg thyroxine per day from day 60 to 120 of lactation and half were controls. Four published equations to characterize individual lactation curves were compared. These were: (1) log Y(n) = log –a1 + b1log n – c1n (Wood); (2) equation 1 with each log Y(n)2 weighted by Yin)2 (Wood weighted); (3) log [Y(n)/n7 = log l/a3 – k3n(Jenkins); and (4) log Y(n) = a4 – b4n‘(l + 25·5 n’) + c4n2 = d 4/ n (Morant), where Y(n) is milk yield on day n of lactation, n' is n–110 (the mid point of lactation), and the a, b, c, k and d parameters are estimated from solution of the equations. The lactation curve from the Jenkins equation projected peak milk yield to occur some 30 days later than estimates from the other equations. It underestimated production early and late in lactation and overestimated it during mid lactation. For several cows, the Morant equation projected that peak production occurred at the end of lactation. Also, analysis of variance of milk production variables was less sensitive when the traits were estimated by the Morant equation than when they were estimated by one of the others. The Wood weighted equation resulted in estimates of peak day of lactation and peak yield that were less variable and more realistic than estimates from the Wood equation. Collectively, therefore, the Wood weighted equation was deemed most suitable to characterize variability among and within these beef cows in milk production. All four equations, however, ranked the 59 cows similarly for estimated 220-day yield.