Forty-eight horses subjected to elective surgery were randomly assigned to three groups of 16 horses. After premedication with 0-1 mg/kg acepromazine intramuscularly and 0.6 mg/kg xylazine intravenously, anaesthesia was induced either with 2 g thiopentone in 500 ml of a 10 per cent guaifenesin solution, given intravenously at a dose of 1 ml/kg (group TG), or with 100 mg/kg guaifenesin and 2.2 mg/kg ketamine given intravenously (group KG), or with 0.06 mg/kg midazolam, and 2.2 mg/kg ketamine given intravenously (group KM). Anaesthesia was maintained with isoflurane. The mean (sd) end tidal isoflurane concentration (per cent) needed to maintain a light surgical anaesthesia (stage III, plane 2) was significantly lower in group KM (0.91 [0.03]) than in groups TG (1.11 [0.03]) and KG (1.14 [0.03]). The mean (sd) arterial pressure (mmHg) was significantly lower in group KG (67.4 [2.07]) than in groups TG (75.6 [2.23]) and KM (81.0 [2.16]). There were no significant differences in the logarithm of the heart rate, recovery time or quality of recovery between the three induction groups. However, pronounced ataxia was observed in the horses of group KM, especially after periods of anaesthesia lasting less than 75 minutes.