Replicated versus un-replicated factorial experiments for preliminary investigation of seed germination and dormancy: alternative approaches using fewer seeds

A 2(3) full factorial, using a total of 3,200 seeds per species, investigated the effects of GA(3), smoke-water, heat-shock and/or light on the germination of Actinobole uliginostan (Asteraceae) and Goodenia fascicularis (Goodeniaceae), following dry after-ripening (DAR) at 15 or 40 degrees C for two or four weeks. The 2(3) factorial experiment was nested within DAR environments. The power analysis (the probability of detecting a true effect at a specified significance), was performed for each environment by using the error estimated from the fully replicated factorial experiments with five replicates. The power was estimated for each main effect and any interactions for two, three, four and five replicates as well as for single replicates. For A. uliginosum data, the power was sufficient with three replicates for main effects and interactions when compared to five replicates, and would have saved 1,280 seeds. However, both main effects and interactions for G. fascicularis, as determined from five-replicate data, were lost when three replicates were analysed. Analysis of single replicates (1/5 of the total number of seeds per species) gave consistent conclusions regarding GA(3) and light main effects upon A. uliginostan. The effect of light upon G. fascicularis was consistent; however the effect of GA(3) and smoke-water were lost. In conclusion, an un-replicated full factorial can give robust conclusions regarding strong main effects upon seed germination, particularly for experiments containing little variation in the percentage of germination within treatments, in order to reduce the number of seeds sacrificed in preliminary screening experiments.