The objective of this study was to investigate inter-observer and test-retest reliability of different behavioural observations to be used in an on-farm, animal welfare monitoring system for veal calves. Twenty-three veal calf farms, varying in size, housing system, feeding regime and age of the calves were visited twice with two observers, simultaneously. Behavioural tests were conducted in eight pens per farm, measuring the response of calves to: a human entering the barn; a novel object; a passive, unfamiliar person; disturbance in the pen and an active approach by an unfamiliar and a familiar person. Furthermore, behaviour was recorded 20 min before and 20 min after feeding in eight other pens per farm. For all behavioural tests, inter-observer reliability was very high. Farm effects and test-retest reliabilities were high and significant for all behavioural tests, except for the test measuring response to disturbance in the pen. Although the active approach test with the familiar person was reliable, it was not feasible in practice due to the availability of the farmer. Since the active approach test with the unfamiliar person gave similar results, this test was recommended for an on-farm animal welfare monitoring system. For most behavioural elements recorded around feeding, farms differed significantly and interobserver and test-retest reliabilities were high as well as being significant. The behavioural tests with entering the barn, novel object and unfamiliar person, and the behavioural observations before and after feeding were feasible and distinctive and reliable enough to be performed on-farm. These methods are promising tools to use as a monitor of animal welfare in veal calves.