This thesis is concerned with the problems that novices have in learning to program: in particular it is concerned with the difficulties experienced by novices learning at a distance, using instructional materials which have been designed especially for novices. One of the major problems for novices is how to link the new information which they encounter with their existing knowledge. Du Boulay, O'Shea and Monk (1981) suggest helping novices to bridge the gap between their existing knowledge and new information by teaching via a conceptual model, which serves to explain the new information in familiar terms. In this thesis the difficulties which novices have when learning to program with the help of a conceptual model were investigated. The curricula and conceptual models of four different programming languages are examined, all of which were designed to teach novices. Du Boulay, O'Shea and Monk (1981) have suggested criteria for analysing conceptual models. It is argued that these criteria, however, do not address the presentation of the conceptual model, and so are insufficient to evaluate them. An additional form of analysis was proposed and used, in addition to the criteria offered by Du Boulay et al. This is a way of describing the conceptual model which distinguishes three views of the conceptual model: state, procedure and function, and which highlights the different aspects which are important for the novice learner by identifying the different kinds of knowledge which are necessary to understand the conceptual model. This analysis of the conceptual models showed that the environments are not as exemplary as the du Boulay et al's criteria suggest, and indicated that three of the environments, SOLO, PT501 and DESMOND, lack a functional representation, and that the fourth, Open Logo, has other different problems. An empirical study was carried out to study the transfer effects of learning two of the languages, a high level and a low level language, sequentially. There was no evidence for such transfer effects. The difficulties novices have in learning the four different languages were also investigated. These studies show that even though the novices were studying environments designed for novices learning at a distance, they did not develop good levels of competence, and the problems they had fall into two main categories: programming and pedagogical. Although the different languages had different aims and curricula, novices had some problems which were common to all or most of the languages. These included understanding flow of control, developing and using programming plans, developing accurate mental models, and in the high level languages, understanding recursion. It is argued that some of these problems are related to the conceptual models. In particular, the difficulties novices had in developing and using plan knowledge, which is one of their main problems, can be explained by the lack of an appropriate functional description in the languages. The subjects' pedagogical problems arose from the relationship between the style and structure of the curriculum, its content, and the subjects themselves. In all the four texts the teaching material is very carefully structured and it is suggested that this may encourage the learner to adopt an over-dependent attitude towards the text, and in some cases, to work at an inappropriate syntactic level. The relationship between the distance learning situation and the novice programmer is discussed, and recommendations are made for improving the curricula used for teaching novices programming.