How do employers use different types of signals – that is, information believed to be associated with productivity – when hiring apprentices? We use a survey experiment in Switzerland – a country with a well-developed dual training system – to study employers’ hiring behaviour. The results reveal that whenever possible, employers resort to direct signals, i.e., information such as educational credentials and aptitude test scores, when making hiring decisions. However, in contexts of high uncertainty, notably, when candidates’ profiles convey ambiguous signals, they are more likely to use probabilistic signals, such as socio-economic status indicators, to complement their assessments. Consequently, in contexts of high uncertainty, candidates with lower socio-economic status are more likely to be discriminated against in the apprenticeship market