Date of Award


Degree Type


Degree Name

Doctor of Philosophy


School of Computer and Security Science


Computing, Health and Science

First Advisor

Dr Mike Johnstone

Second Advisor

Professor Geoffrey Roy


This research was motivated by an interest in novices learning to program

and a desire to understand the factors that affect their learning. The

traditional approach to performing such an investigation has been to select

factors which may be important and then perform statistical tests on a few

potential relationships. A new research model is proposed and tested to

ensure that a thorough and systematic investigation of the data is performed.

This thesis describes the data, defines the model and explains the

application and validation of the model.

The research process is managed by a control algorithm that is the heart of

the model. This algorithm is seeded by a hypothesis that connects two

variables of interest and dictates the testing of a series of hypotheses; as it

does this, it also delves deeper into the data to identify additional


In this research the model was applied to investigate the relationships

between: learning style and achievement; programming behaviour and

achievement; and learning style and programming behaviour. Learning style

was assessed using Kolb’s Learning Style Inventory, achievement was

based on exam score and programming behaviour was extracted from a log

of student activities using a programming tool. The largest number of

significant relationships was found between aspects of behaviour and


The model was validated by classifying the significant hypotheses based on

the research model’s tree structure, the section of the programming tool in

use and the literature. These three classification schemes provided a

structure to explore their similarities and differences. The model was thus

demonstrated to be robust and repeatable by comparing the results with

those from both using a programming tool, and expert opinion.

This research has revealed several attributes of the learning behaviour that

affected the students’ results within this group, including aspects of

timeliness and overall volume of activity. These are suitable targets for

future investigations.

The research model could be applied to other data sets where an in-depth

investigation into pairwise data is required.