A longitudinal design was applied to differentiate between normal variations of psychomotor development and lasting handwriting deficiency (dysgraphia). Sixteen primary school children were tested with writing tasks that were recorded on a computer-monitored'XY tablet. These tasks represented different modules of the handwriting model of Van Galen (1991). Dependent variables were spatial errors, movement time, movement dysfluencies, trajectory length, stroke curvature, and the degree of neuromotor noise in the movement velocity profiles. The latter variable was measured by means of Power Spectral Density Analysis of the movement velocity signal, which revealed that movements of poor writers were substantially more noisy than those of proficient writers, with a noise peak in the region of neuromotor tremor. At the same time, the poor writers were less accurate. It was concluded that control of spatial accuracy rather than allograph retrieval or size control is the discriminating feature in dysgraphic children. Moreover, poor writers do not catch up with their peers within the 1 year time span tested.