In the UK since the early 1990s, there has been widespread concern and extensive reporting about the difficulties encountered by engineering students with the mathematical elements of their university courses. Students’ lack of previously expected mathematical skills is of particular concern and has prompted the provision of mathematics support in many UK institutions. A related problem is students’ lack of self-confidence (or self-efficacy) in their mathematical capability, and this paper seeks to explore how this has arisen and how it affects students’ learning, and proposes suggestions for improvement.Interviews were conducted with final year engineering students at Harper Adams University College in 2009. These explored students’ experiences of and self-confidence in learning and using mathematics before and during university and what they anticipate in the future. The seven students interviewed exhibited a range of self-confidence and achievement and their responses about self-confidence and mathematics support were analysed. Despite their wide ranging backgrounds, all of the students achieved well in their first year university engineering mathematics modules, which naturally increased their self-confidence. Several students described how using the mathematics support provision had helped them with mathematics and improved their confidence.In addition to analysing the interview scripts thematically, Bandura’s model of self-efficacy (Bandura, 1997) was used as a conceptual framework with which the students’ accounts were cross-matched. Bandura’s model proposes four sources of self-efficacy (past achievement; comparison with others; what others tell you; feelings or physical states) and four mediating processes (cognitive; motivational; affective; selective processes). Additional sources of self-confidence outside of Bandura’s model were also described by the students, in particular working with peers, appropriate speed of teaching and small group sizes.The most important source of self-efficacy was found to be students’ past experience of success or failure, and all four of Bandura’s mediating processes were referred to by the students. There was no mention, however, of verbal persuasion, and it is argued that lecturers and support tutors might do more to develop students’ confidence through this means. Most importantly, students’ opportunities for success should be maximised, including careful provision of challenging tasks at the right level, in order to build students’ self-confidence in mathematics.