A common method to describe maize phenology during the vegetative phase is to determine the number of leaves. The thermal interval between the appearance of successive leaves, called the phyllochron, is critical for predicting the duration of vegetative development. In general, there is a high correlation between the final number of initiated leaves and the duration from sowing to silking. The development of a crop depends on temperature and photoperiod and genetic response to these environmental factors, thus bio-meteorological indices are needed to describe this process. The goal of this study was to obtain a better understanding of the phyllochron and its impact on the final number of leaves. A second objective was to evaluate three thermal indices for predicting the duration of sowing to silking and silking to physiological maturity for four maize hybrids grown off-season in a subtropical region of the Southern Hemisphere (São Paulo State, Brazil). The modified general thermal index (mGTI) was compared with the classical thermal time concept (TT) with two sets of cardinal temperatures, including TT(10) with 10 °C as minimum base temperature (Tb) and TT(8) with 8 °C as Tb. Three field experiments that included four hybrids with different relative maturity ratings were conducted during the autumn and winter growing seasons of 2001 and 2002 at the University of São Paulo, in Piracicaba, SP, Brazil. Phenology was recorded daily. Also, an independent evaluation of three methods for estimating phenology was conducted with data obtained from the variety trials of the ‘Instituto Agronômico’ for four regions of the State of São Paulo, Brazil. The results indicated that the value for the phyllochron was about 50 °Cd/leaf until approximately the appearance of the 12th leaf, and then declined. For estimating the planting to silking period, the TT(10) method resulted in a coefficient of variation (CV) of 2·01%, while for TT(8) the CV was 1·37% and for mGTI the CV was 1·04%. For the period from silking to physiological maturity, the CV was 2·91% for the TT(10), 2·64% for the TT(8) and 2·01% for the mGTI method. All CVs were very small, suggesting that there was only a slight difference between algorithms. It can be concluded from this study that the mGTI can be used for estimating both silking and physiological maturity dates for maize grown off-season in subtropical regions of Brazil because it resulted in the smallest CV. However, additional studies with different maize hybrids are necessary to confirm the advances in estimating the main phenological phases for maize based on the mGTI.