Paroxysmal dyskinesia can be subdivided into three clinical syndromes: paroxysmal kinesigenic dyskinesia or choreoathetosis, paroxysmal exercise-induced dyskinesia, and paroxysmal non-kinesigenic dyskinesia. Each subtype is associated with the known causative genes PRRT2, SLC2A1 and PNKD, respectively. Although separate screening studies have been carried out on each of the paroxysmal dyskinesia genes, to date there has been no large study across all genes in these disorders and little is known about the pathogenic mechanisms. We analysed all three genes (the whole coding regions of SLC2A1 and PRRT2 and exons one and two of PNKD) in a series of 145 families with paroxysmal dyskinesias as well as in a series of 53 patients with familial episodic ataxia and hemiplegic migraine to investigate the mutation frequency and type and the genetic and phenotypic spectrum. We examined the mRNA expression in brain regions to investigate how selective vulnerability could help explain the phenotypes and analysed the effect of mutations on patient-derived mRNA. Mutations in the PRRT2, SLC2A1 and PNKD genes were identified in 72 families in the entire study. In patients with paroxysmal movement disorders 68 families had mutations (47%) out of 145 patients. PRRT2 mutations were identified in 35% of patients, SLC2A1 mutations in 10%, PNKD in 2%. Two PRRT2 mutations were in familial hemiplegic migraine or episodic ataxia, one SLC2A1 family had episodic ataxia and one PNKD family had familial hemiplegic migraine alone. Several previously unreported mutations were identified. The phenotypes associated with PRRT2 mutations included a high frequency of migraine and hemiplegic migraine. SLC2A1 mutations were associated with variable phenotypes including paroxysmal kinesigenic dyskinesia, paroxysmal non-kinesigenic dyskinesia, episodic ataxia and myotonia and we identified a novel PNKD gene deletion in familial hemiplegic migraine. We found that some PRRT2 loss-of-function mutations cause nonsense mediated decay, except when in the last exon, whereas missense mutations do not affect mRNA. In the PNKD family with a novel deletion, mRNA was truncated losing the C-terminus of PNKD-L and still likely loss-of-function, leading to a reduction of the inhibition of exocytosis, and similar to PRRT2, an increase in vesicle release. This study highlights the frequency, novel mutations and clinical and molecular spectrum of PRRT2, SLC2A1 and PNKD mutations as well as the phenotype-genotype overlap among these paroxysmal movement disorders. The investigation of paroxysmal movement disorders should always include the analysis of all three genes, but around half of our paroxysmal series remain genetically undefined implying that additional genes are yet to be identified.