A selection experiment was established in Coopworth sheep in 1981 to breed for increased or reduced ultrasonic backfat depth (scan C). Foundation females came from four flocks recorded for scan C and live weight, with weight-adjusted scan C data within flock being used for initial screening and subsequent selection. Three groups of animals per source flock, comprising proportionally the fattest 0·12, a random sample, and the leanest 0·12, were used to establish the F, control and L lines, respectively. Ewe flock numbers from 1981 to 1992 averaged 51 per line. Foundation rams were selected in the same manner from four different farms (two sources each in 1981 and 1982) to provide F line (proportionally the fattest 0·04), control line and L line (leanest 0·04) rams for use in 1981 and 1982, with four mated per line per year. Thereafter homebred rams were selected, with 41 or 42 homebred rams being used per line until 1992. Average generation intervals were 2·13 years and annual inbreeding rates per line 0·004. Bivariate heritability estimates for log scan C and log live weight, and a univariate estimate for log scan C using restricted maximum likelihood with an animal model, were 0·28, 0·22 and 0·38, respectively (all with s.e. 0·03). There was a suggestion of lower heritabilities in the L line for log scan C after adjustment for live weight. Realized heritabilities in the F and L lines were 0·34 (s.e. 0·07) and 0·26 (s.e. 0·03), respectively. Deviations of back-transformed weight-adjusted scan C in the last 2 years ofF and L data analysed (1991 and 1992 birth years) from the control flocks were 2·08 and −0·85 mm, which corresponded to responses of +2·50 and −1·03 phenotypic standard deviations, respectively. In addition there were relatively large responses in live weight taken at scanning, with F and L lines averaging 34·0 and 40·2 kg, compared with 38·0 kg for controls in the 1991 and 1992 birth years. The genetic and phenotypic correlations between log scan C and log live weight at scanning were 0·16 (s.e. 0·07) and 0·46 (s.e. 0·01) respectively.