Dyskinesia affects approximately 30 to 40% of patients with Parkinson's disease but treatment options for the prevention of dyskinesia induction and for the suppression of established dyskinesia are limited. This situation is made more difficult by a poor understanding of the pathphysiology of the processes underlying both the priming for dyskinesia and the manifestations of involuntary movements. Loss of tonic stimulation of striatal dopamine receptors in PD and its replacement by pulsatile dopaminergic stimulation using short acting drugs has been proposed as leading to the abnormalities that cause dyskinesia induction. As a consequence, the concept of continuous dopaminergic stimulation (CDS) was introduced to explain why longer acting dopamine agonists do not produce the same intensity of dyskinesia. Key to these ideas has been the use of both 6-OHDA lesioned rodent models of PD and, in particular MPTP-treated primates. Comparison of the ability to induce dyskinesia of the same dopamine agonists given by pulsatile or continuous administration or more constant administration of Levodopa (L-dopa) has shown that constant drug delivery (CDD) dramatically reduces dyskinesia induction. Similar conclusions have been reached from clinical investigations in PD. Recent studies in MPTP-treated primates have also suggested that switching from pulsatile drug delivery to CDD can be utilized to inhibit dyskinesia expression. However, CDS does explain some important features of dyskinesia induction in PD but it may not apply to early PD when remaining dopaminergic neurones buffer against pulsatile stimulation. In addition, CDS may not apply when comparing between drug classes and it appears that it is CDD which is more important in regulating therapeutic efficacy. Recently, studies in MPTP-treated primates have suggested that a range of nondopaminerigic drugs might be useful in suppressing dyskinesia. These have included 5-HT-1A agonists and alpha-2 adrenergic antagonists and a variety of other molecular entities. Unfortunately, these findings are not always reproducible in the same models and do not translate into clinically useful effects. Preclinical studies have suggested a number of directions that might be utilized to prevent dyskinesia in PD. However, much of what is proposed is empirically-based and we still do not have a good understanding of why dyskinesia appears, why it persists or how to bring the movements under control. Certainly, the use of CDD can reduce dyskinesia intensity but other factors also influence its appearance and it is these that we need to study at the preclinical level if effective therapies are to be developed.