In a field experiment with 178 first-grade pupils, the effects of an experimental beginning reading programme were investigated. Both an experimental and a control group worked with the most frequently used Dutch beginning reading programme, "Learning to Read Safely." The instructional approach implemented in the experimental group was guided co-construction (GCC); the instructional approach implemented in the control group was direct instruction (DI). The results of an overall analysis of the development of word recognition (WR) over time (i.e., throughout the 1st grade) showed the pupils in the experimental group to outperform those in the control group. However, the better performance by the experimental group attenuated over time with better performance by the control group on the last measurement occasion. Majority pupils benefitted more from GCC but minority pupils more from DI. Minority pupils in the control group showed greatest progress. (Contains 8 tables and 3 figures.)