The performance of a cylindrical enzyme electrode with a soluble mediator was analysed theoretically. The normalized current response was calculated at various values of both the electrode radius (a) to enzyme layer thickness (l) ratio, a/l, and the relative catalytic activities, sigma(S) and sigma(M) {= (k(cat)[E]l2/K(MS,M)D(S,M))1/2}. The calculated results demonstrated that at high mediator concentration, larger a/l ratio and increased relative catalytic activity for the substrate, sigma(S), a wider linear range could be obtained for a calculated glucose calibration graph. However, the linearity was found to be less dependent on a/l than sigma(S). Alternatively, at high glucose concentration, the mediator concentration - response curve calculated at low a/l values and a high catalytic activity for the mediator, sigma(M), indicated a wide linear range. Immobilized glucose oxidase (GOD) cylindrical microelectrodes were fabricated and their characteristics were evaluated by using 1,4-benzoquinone as electron mediator. GOD was immobilized in a photo-cross-linkable polymer on two types of cylindrical microelectrodes of 2 and 25-mu-m diameter. The linear ranges of the observed calibration graphs were wider than that obtained using a disc electrode of 1 mm diameter. This was probably due to the larger sigma(S) values obtained with the glucose sensors of 2 and 25-mu-m diameter. Moreover, the response of the 2-mu-m glucose sensor based on hydrogen peroxide detection was compared with that using mediators. This result showed that the wider measurable range was obtained using mediators.