Granular calcined magnesites originating from Spain, China and Greece and magnesium hydroxide powder were evaluated in a series of balance experiments conducted with wether sheep (c. 42 kg liveweight) given a basal ration of dried grass at Glasgow University Veterinary School between 1989 and 1994. Four Spanish ‘Agma’ products and three Spanish ‘Navarras’ products contained less material (0–104 g/kg)>1000 μm diameter than four Chinese and four Greek materials 178–483 g/kg). The Spanish products had lower (18–45 g/kg) losses on ignition compared with the Chinese and Greek materials (53–112 g/kg). A fifth Greek product contained only 4 g/kg>1000 μm diameter. In three experiments comparing supplementation of the basal diet with 2·0 g Mg as either magnesium hydroxide or as one of the three separate purchases of the commercial calcined magnesite sources, the powdered magnesium hydroxide had the highest mean apparent availability coefficient (0·30) (faecal increase method). The corresponding values were 0·24 (Agma and Navarras), 0·20 (Chinese) and 0·15 (Greek) (pooled s.e. 0·014, d.f. 10). Powdered magnesium hydroxide also had the highest (0·18) mean availability coefficient (urinary increase method) compared with 0·12 (Agma), 0·13 (Navarras), 0·14 (Chinese), 0·10 (Greek) (pooled s.e. 0·005, d.f. 10). In two subsequent experiments, Agma had a superior apparent availability coefficient (0·26) (faecal method) than either the Chinese (0·14) or the Greek materials (0·19). A finer grade of Greek calcined magnesite (0·25) was equivalent to Agma. For the 12 calcined magnesites evaluated in the three main experiments, significant (P<0·05) correlations (r) were found between the magnesium availability coefficient (faecal method) and loss on ignition (−0·65); particle size proportions (<250 μm, 0·59; and >1000 μm, −0·65) and tended towards significance for rumen solubility in 24 μm nylon bags in hay-fed cows over two days (r=0·54, P=0·07) and rate of reaction with citric acid (0·57, P=0·055). No correlations were found with magnesium availability coefficients (urinary increase method). This confirms the limitations of both in vitro predictions and in vivo (faecal) estimates of magnesium availability for different calcined magnesites. The mean magnesium availability coefficient (faeces) of the 17 calcined magnesites was 0·21 (s.d. 0·05). If this were to be confirmed by other observations, it would have implications for the amounts of dietary calcined magnesite required to meet recommended dietary magnesium allowances for ruminants.