This thesis explores barriers to using formal specification for software development in industry. Empirical assessment techniques are used initially in an exploratory stage and subsequently in testing a hypothesis arising from the first stage. A second hypothesis is investigated by construction of a method and tool with subjective assessment of its effect. The first stage consists of a survey of experienced industrial formal methods users via a questionnaire-based interview. The interview explore the practicalities of using formal methods in an industrial setting. From the many findings in this stage, two hypotheses are selected for further investigation. The first hypothesis is that formal specification are no more difficult to understand than code. This is tested by formal experiment. The subject's ability to understand the functionality of a formal specification is compared with their ability to understand its implementation in program code. The second hypothesis is derived from observations, during the survey stage, that formal specifications are difficult to write. In particular, choosing appropriate abstractions is difficult. We consider what might make formal specification difficult and compare the process with that of programming. The second hypothesis is that a tool supported, graphical modelling notation would be of benefit in the process of writing a formal specification. Such a notation is devised by adapting the UML and augmenting it with a formal text notation. A tool that converts this graphical formal specification into the formal notation, B is described and examples of its use are analysed.