This paper examines the effect of using construction heuristics on the evolutionary process for the domain of school timetabling. The paper firstly presents a genetic algorithm to solve the school timetabling problem. The paper compares the performance of a GA with randomly creating potential solutions in the initial population with that of a GA using a construction heuristic for developing the initial solutions. Two construction heuristics, namely, largest degree and saturation degree are studied. The GA's are applied to a data set of six real world problems made available for the Greek school timetabling problem. This study has revealed that the use of construction heuristics improves the performance of the GA to produce better quality timetables. Furthermore, different construction heuristics produce better results for different problems.