Judgement bias tests of animal affect and hence welfare assume that the animal’s responses to ambiguous stimuli, which may herald positive or negative outcomes, are under instrumental control and reflect ‘optimism’ or ‘pessimism’ about what will happen. However, Pavlovian control favours responses (e.g. approach or withdrawal) according to the valence associated with a stimulus, rather than the anticipated response outcomes. Typically, positive contexts promote action and approach whilst negative contexts promote inhibition or withdrawal. The prevalence of Go-for-reward (Go-pos) and NoGo-to-avoid-punishment (NoGo-neg) judgement bias tasks reflects this Pavlovian influence. A Pavlovian increase or decrease in activity or vigour has also been argued to accompany positive or negative affective states, and this may interfere with instrumental Go or NoGo decisions under ambiguity based on anticipated decision outcomes. One approach to these issues is to develop counter-balanced Go-pos/NoGo-neg and Go-neg/NoGo-pos tasks. Here we implement such tasks in Sprague Dawley rats and C57BL/6J mice using food and air-puff as decision outcomes. We find striking species/strain differences with rats achieving criterion performance on the Go-pos/NoGo-neg task but failing to learn the Go-neg/NoGo-pos task, in line with predictions, whilst mice do exactly the opposite. Pavlovian predispositions may thus differ between species, for example reflecting foraging and predation ecology and/or baseline activity rates. Learning failures are restricted to cues predicting a negative outcome; use of a more powerful air-puff stimulus may thus allow implementation of a fully counter-balanced task. Rats and mice achieve criterion faster than in comparable automated tasks and also show the expected generalisation of responses across ambiguous tones. A fully counter-balanced task thus offers a potentially rapidly implemented and automated method for assessing animal welfare, identifying welfare problems and areas for welfare improvement and 3Rs Refinement, and assessing the effectiveness of refinements.