Until relatively recently, general practitioners (GPs) have been allowed to work independently, with no requirement to consider the resource implications of their referral and prescribing decisions. In order to align the interests of GPs with the overall objectives of health systems a number of countries have introduced primary care based capitation, funds pooling and budget holding either as experiments or as an overall policy. Are these experiments and policies likely to work? This paper presents evidence from the UK total purchasing experiment, which was the first major quasi-market development in the NHS to be independently evaluated from the outset. Total purchasing gave volunteer groups of practices freedom to purchase all hospital and community health services for their patients. The evidence suggests that whilst GPs have great potential as purchasers, they also have considerable limitations. The expectation that they will be able to improve the quality of patient experience of care, or to alter the use of resources, may not be generally realised. GP-based purchasing may be more appropriate where the task is to alter the balance or location of care between hospital and extramural settings. However, budgetary incentives are not 'magic potions' which have similar effects on behaviour wherever they are introduced. Holding budgets and having independent contracts, while important pre-requisites for being taken seriously in a quasi-market, were not sufficient for effective total purchasing. The paper concludes that health systems should not only value innovation and experimentation and encourage learning from evaluative research; they should also recognise the importance of supportive circumstances for any innovation to effect real and sustained change.