Behavioural tests are often used to describe dogs’ responses to a variety of stimuli. However, the expressed behaviours do not always give an accurate indication of the dogs’ mental state. Cardiac measures can be recorded in parallel to the behavioural test to provide additional information about arousal level. Heart rate (HR) reflects the combination of sympathetic and parasympathetic stimulation, whereas the beat-to-beat variation in heart rate (heart rate variability: HRV) provides more details about the balance between sympathetic and parasympathetic activation. In studies validating cardiac measures typically the dog's movement is controlled or the response to only one stimulus at a time is tested. In this study, the validity and reliability of cardiac measures were tested during a behavioural test at home (N = 18 shepherd dogs). HR, HRV and behaviour were recorded in six randomised subtests, in which stimuli of different emotional valences were presented (neutral, positive, negative). Positive and negative subtests differed significantly from the neutral subtests in terms of HR and behaviour (P = 0.000). Compared to the neutral subtests (77.85 ± 21.00 bpm), HR was increased (positive: 91.67 ± 21.64 bpm; negative: 100.29 ± 18.54 bpm), and the dogs showed more interactions with the presented stimuli (contact/avoid, approach, look), more oral behaviours and were more active (walk, stand, less lying). However, HR only significantly correlated with activity during the neutral subtest (r = 0.511, P = 0.030). Hence, activity only appeared to influence HR in the absence of significant emotional stimuli, whereas the increased HR in both the positive and negative subtests reflected arousal. This is supported by the fact that stress‐related behaviours and HR were similar in dogs tested a second time (test-retest reliability; N = 15 dogs), despite significant differences in activity between both test moments (Z = −2.510; P = 0.009). For the HRV variables, on the other hand, no significant differences were found between subtests of different emotional valences. This was possibly caused by errors in recording, due to poor electrode conduction caused by movement of the dogs. In conclusion, this study showed that HR can be a useful measure of arousal during behavioural tests, though it did not allow for a distinction between positive and negative emotional valences. HRV appeared more difficult to record (with the Polar® RS800CX with Wearlink) during constant movement of the dog.