Breeding of cattle with docile temperament can simplify routine practices and enhance welfare of cattle in common production systems. The aim of the study was to compare and validate some aspects of assessment methods for temperament traits in cattle for their potential use in breeding programmes. Calves of different beef and dairy cattle breeds were video-recorded during a 2 min tethering (n = 160) and a 2 min crush test (n = 185) at the age of 38 ± 25 days. Subsequently, two observers blinded to the identity of records each analysed the video clips three times with the software Interact® in intervals of 1–2 days between viewings. Frequency and duration of behaviour patterns such as head movement, tail flicking or defecation were recorded. In addition, each behaviour pattern was assessed on a 10 cm visual analogue scale (VAS) and in the end of each observation an overall behaviour score on a numerical scale from 1 to 5 and on a VAS was assigned to each animal. Using variance components from mixed model analysis, results revealed that both inter- and intra-observer reliability (OR) were similar or superior for the majority of behaviour patterns when assessed with the VAS compared to the more time-consuming assessment of exact frequencies and durations (e.g. Intra-OR ‘tail flicking’: Observer I: rVAS = 0.83 and rFreq = 0.26; Observer II: rVAS = 0.91, rFreq = 0.37 and rDur = 0.87 for the tethering test; Inter-OR ‘body movement’: rVAS = 0.81 and rFreq = 0.33 for the crush test). The overall behaviour score showed a high intra-observer reliability when assessed with the VAS as well as with the numerical scale (Observer I: rVAS = 0.97 and rNum = 0.77; Observer II: rVAS = 0.82 and rNum = 0.84 for the crush test). However, not all traits, and particularly including those assessing rarely occurring behaviour patterns, proved to be appropriate for recording with a VAS. In conclusion, a VAS that is based on carefully selected features can be considered a reliable and practical method for temperament assessment in cattle.