Summary: Eleven healthy horses underwent 5 repeated abdominocenteses, with either a sharp-tipped spinal needle or a blunt-tipped teat cannula to investigate possible differences in success rate, sample volume, depth at which a sample was obtained, length of procedure, complications and cytological variables. Variables were analysed with a repeated-measures ANOVA or Fisher's exact test (α = 0.05). Success rate, sample volume, length of procedure, occurrence of haemorrhage during the procedure and incidence of grossly visible blood contamination were not different between techniques or over time. Depth at which samples were obtained was greater using a cannula than a needle (P<0.02), and samples were obtained with either technique at a greater depth than abdominal wall thickness assessed via ultrasound (P<0.014). Peritoneal fluid total and differential nucleated cell counts, and total protein concentration did not differ between techniques or over time. Red blood cell count in the least blood contaminated fraction of each sample was not affected by time, but it was lower after needle abdominocentesis than after cannula abdominocentesis (P = 0.04). Swelling of abdominocentesis sites increased with both techniques over time (P<0.05) and was more severe in horses undergoing cannula abdominocentesis (P<0.05). Enterocentesis occurred with a spinal needle in one horse, but no subsequent complications were noted. Both techniques appear to be safe and reliable for abdominocentesis in healthy horses. Using a blunt-tipped cannula, as opposed to a needle, is likely to result in greater subcutaneous swelling. Both the cannula and needle must be long enough to penetrate well beyond the thickness of the abdominal wall to achieve successful peritoneal fluid collection.