Statistical analysis
Statistical analyses were conducted using SAS statistical software (SAS Institute 9.0 Inc., 2002). All measured and calculated traits were used for analysis (Table 2). To minimize variance heterogeneity, all data were log-transformed, except for petiole angles and branch angles (square root-transformed), before statistical analysis. For plant total mass, we applied two-way ANOVA to analyze effects of germination timing, sampling time and their interactions, and one-way ANOVA to analyze the effects of germination timing or sampling time within each or across all of the other treatments. Plant size (e. g. total mass) can have very significant effects on other traits, which may bias the effects of germination time. Therefore, for all the other traits, we applied two-way ANCOVA to evaluate overall effects of germination timing, sampling time and their interactions, and one-way ANCOVAs for effects of germination timing or sampling time within each or across all of the other treatments, with total mass as a covariate. For a given trait, significant contribution of total biomass (plant size) to its variation in response to germination timing indicates an occurrence of apparent plasticity (McConnaughay & Coleman, 1999). When effects of total mass were removed, the variation due to germination timing in trait expression was an indication of true plasticity (Weiner, 2004). Multiple comparisons used the Least Significant Difference (LSD) method in the General Linear Model (GLM) program, which produced adjusted mean values and standard errors in one-way ANCOVA.