It has been contemplated for a long time that lumbar segmental instability is a contributor to LBP symptomatology. The definition of this entity has been expanded within the last decade, to include pathological processes affecting not only the passive inert spinal structures but also the neural control movement system and the muscular system. However, the theoretical level of the evidence for such a relationship even nowadays seems to be stronger than the clinical. In an unknown percentage of certain categories of back pain, like in lumbar spondylolysis and spondylolisthesis, such a link may be very strong. This by no means justifies generalisation of stabilisation training in all LBP conditions. The purpose of this thesis was dual. An important target was to select and develop a series of reliable outcome measures that could be used to assess outcome in a back rehabilitation study. The second objective was to test the hypothesis that supplementing a general endurance exercise programme with exercise techniques that preferentially activate the stabilising trunk muscles is of more benefit in non-specific recurrent LBP patients than general exercise alone. A randomised-controlled two- group experimental design with an 8-week exercise intervention was used to investigate this question. Results indicated that there was no apparent difference between the two exercise approaches in a variety of physiological, functional and cognitive status measures of relevance to the patient. These results could only be considered pilot work however, as the power to detect with confidence between-population differences was not high. The thesis makes a number of recommendations regarding outcome assessment for back rehabilitation studies and the possible approaches to treatment.