By conducting a production experiment, this study investigates whether speech melody in English is categorical or gradient, whereby the melody of speech indicates a categorical function. Test stimuli consisted of two data sets. In the first set, neutral conditions were directly compared to uncertainty conditions. In the other set, speakers produced both neutral-focus and contrastive-focus conditions to test one of the communicative functions. The results of the production experiment revealed that focused areas are realized with increased duration, fundamental frequency, and intensity compared to non-focused areas. Also, significant differences can be found between neutral and uncertainty conditions. In contrast to the neutral condition, the latter exhibits a sharp rising F0 curve following the release [m] in millionaire. On the other hand, the neutral condition is defined by a continuous declining F0 contour after [m]. While this distinction seems to support the categorical differences between the two data sets, the conditions are certainly attributed to different communicative functions. More importantly, the distributions of the uncertainty conditions turn out to be broad, which implies that each speaker delivers a different tonal pattern.