During the last two decades the traditional subjective appraisal of the body fat stores in farm animals, made by eye and touch, has been rationalized by the introduction of numerical systems of rating specific anatomical points.

Palpation of the lumbar vertebrae, the pin and hook bones (tail head) (Lowman et al. 1973), and, occasionally, width behind the shoulders (Treacher et al. 1986) provides an assessment of the fatness of the animal. This is calibrated from standard photographic charts, in the use of which the operator is trained. Body condition scoring (CS) by this method has been widely developed for dairy cows (Earle, 1976; Mulvany, 1977; MacCarthy, 1978; Scott & Smeaton, 1980; Wildman et al. 1982; Aalseth et al. 1983; Anon. 1986; Garcia Paloma, 1990).

The actual numerical scales have varied among systems. Thus over the range from very thin to very fat cows the scale is 1, 2…7, 8 (Earle, 1976) and 0½, 1,…4½, 5 (Mulvany, 1977), for example. CS is thus a discrete variate with a limited number of readings. Several authors have sought to decrease the division size, for example 1, 1 1/3; 1 2/3;…4 2/3, 5 (Ducker et al. 1985a, see also Dewhurst et al. 1996) and by using the means of the separate values obtained by two operators judging CS independently (W. H. Broster, V. J. Broster, A. J. Clements, J. W. Siviter and T. Smith, unpublished results). In this review CS will be stated in the units given by Mulvany (1977), to which other ranges have been scaled. Inspection of CS data has usually been conducted by analysis of variance, but also by the Mann–Whitney signed rank test (Moorby et al. 1996).

Evans (1978) and Nicoll (1981) studied the variation in recording CS. Both investigators concluded that a system of observation by two independent assessors per recording occasion is advantageous and that revision training is necessary to maintain operator standardization. Of total variance in CS, some 60–70% was found attributable to ‘between-animal’ differences, <5% to assessor variation and <10% to animal×assessor variation. Croxton & Stollard (1976) found good repeatability of CS measurements ‘between’ and ‘within’ operators. It is regrettable that few reports of experimentation give full details of method of and number of operators engaged in body score recording.