Neonatal diabetes can either remit and hence be transient or else may be permanent. These two phenotypes were considered to be genetically distinct. Abnormalities of 6q24 are the commonest cause of transient neonatal diabetes (TNDM). Mutations in KCNJ11, which encodes Kir6.2, the pore-forming subunit of the ATP-sensitive potassium channel (K_ATP), are the commonest cause of permanent neonatal diabetes (PNDM). In addition to diabetes, some KCNJ11 mutations also result in marked developmental delay and epilepsy. These mutations are more severe on functional characterization. We investigated whether mutations in KCNJ11 could also give rise to TNDM. We identified the three novel heterozygous mutations (G53S, G53R, I182V) in three of 11 probands with clinically defined TNDM, who did not have chromosome 6q24 abnormalities. The mutations co-segregated with diabetes within families and were not found in 100 controls. All probands had insulin-treated diabetes diagnosed in the first 4 months and went into remission by 7–14 months. Functional characterization of the TNDM associated mutations was performed by expressing the mutated Kir6.2 with SUR1 in Xenopus laevis oocytes. All three heterozygous mutations resulted in a reduction in the sensitivity to ATP when compared with wild-type (IC₅₀∼30 versus ∼7 µM, P-value for is all <0.01); however, this was less profoundly reduced than with the PNDM associated mutations. In conclusion, mutations in KCNJ11 are the first genetic cause for remitting as well as permanent diabetes. This suggests that a fixed ion channel abnormality can result in a fluctuating glycaemic phenotype. The multiple phenotypes associated with activating KCNJ11 mutations may reflect their severity in vitro.