There are now about 50 randomized controlled trials into rehabilitation packages, physiotherapy or related specific and non-specific techniques in multiple sclerosis (MS). Generally these, and related systematic reviews, report benefits. Particular problems arise, however, with the blinding of assessment, determination of what is the active or beneficial input by the therapist, the use of multiple domains of assessment of quality of life and function without, sometimes, a clear statement of a trial hypothesis or primary outcome and the short-term nature of many studies. Therapy inputs can be broadly broken down into verbal interactions with the patient, physical inputs and referral/recommendation processes. Each may be relevant to the outcome. 'Response-shift' may be an important internal mechanism of mind whereby changes in 'quality of life' may not always parallel function emphasizing the case for clearly separating quality of life from functional assessment and attempting to make the latter as objective as possible. Trials of such complex interventions will need to randomize specified components of therapy against appropriate placebos or active treatment arms rather than no therapy, which will be ethically harder to sustain. Classification of physiotherapy inputs by type and 'dosage', a primary hypothesis under test and attention to concealed allocation of treatment, assessor blinding and intention to treat analysis together with improved measurements of function will assist in the consolidation of the evidence base for physiotherapy as an important component of management for MS patients.