The soybeans are grown mainly in North America, South America and Asia; The United States of America and Brazil are the most producing and exporting countries (^{Kumudini, 2010}). In Mexico, production is concentrated in the south of Tamaulipas with an average yield above 1.5 t ha^-1. However, ^{Maldonado et al. (2013)} mentioned that during 2011, the average yield was 0.96 t ha^-1 due to the drought that prevailed during the reproductive stage of the crop. Since according to ^{Liu et al. (2003}) ^{Liu et al. (2003)}, flowering, pod formation and grain filling are the most susceptible phases of this adverse factor.

The drought is the main limitation in food production worldwide, so resistance to this factor has been studied extensively in different crops. ^{Muñoz (1992)} stated that drought resistance is the ability of the plant to accumulate energy against the adverse factor, which can be measured by model 1, which states that the resistance is a function of the generic effects estimated by the component G (genotype) and the specific effects to drought estimated by the G*S component (genotype*drought).

Therefore, the most resistant varieties will be those with the highest proportion of the two. The model 2 was then established (^{Muñoz, 1997}), considering that resistance to drought is not constant throughout the plant cycle, since if early, intermediate and late genotypes are sown at a given date, a drought occurs when Precocious variety is flourishing, its response will not be comparable with the intermediate or the late, because the phenomenon did not affect the same ontogenetic development.

In Mexico, ^{Gill et al. (2008)} identified under greenhouse conditions soybean genotypes to derive lines with combined resistance characteristics: a water deficit and Macrophomina phaseolina. However, the selection of drought resistant genotypes under field conditions has not been explored. Therefore, the objective of this research was to characterize the resistance to drought in 25 national soybean genotypes, considering the yield, its components and other morphological and physiological characteristics.

The experiment was established in Altamira, Tamaulipas. The precipitation was 20 mm during the evaluation cycle: December 2012 To April 2013. The averages of maximum and minimum temperatures were 26.8 and 15.2 °C respectively. An experimental design was used with two replications divided (^{Sloane et al., 1990}) plots. The drought levels corresponded to the large plots and to the subplots the genotypes. The experimental unit was two rows of 5 m. The germplasm used were four varieties generated by INIFAP for the humid tropics of Mexico and 21 advanced lines of the soybean breeding program of the same Institute.

Two drought levels were compared: without drought (S0) and with drought (S1). In S1 five drip irrigations were applied at 30 days after sowing (DDS), irrigation was suspended (beginning of the reproductive period). In S0 it continued to be irrigated to 100 DDS. In each experimental unit, the volumetric water content (CVA) was measured at 60 DDS and expressed in%. Full bloom (DR2) and physiological maturity (DR7) days were also recorded; height of plant to full bloom (AR2) and beginning of physiological maturity (AR7); chlorophyll index (CL90) and stomatal conductance (CE90), both at 90 DDS; harvested plants (PLC); plants (PxM) and pods m^-1 (VxM); weight of 100 seeds (P100S) and yield (REND). It should be noted that CL90 and CE90 were measured only in the intermediate and late genotypes, since at 90 DDS the precocious had total wilt or leaf senescence; PxM and VxM were not counted in the late group.

The stress response was evaluated with model 2 drought resistance (^{Muñoz, 1997}): Y=Ro+RF+Ro*RF , where Y represents the total variation, R_O the change in resistance through the ontogenic stages, R_F the resistance variation through precocity groups and R_O*R_F is the interaction. First genotypes were grouped into early (less than 50 DR2), intermediate (50-59 DR2) and late (over 59 DR2) genotypes, then model 1 was applied (^{Muñoz, 1992}). In order to know the effect of S and G, analysis of variance and Tukey’s mean comparison test were performed. For specific drought effects, REND interactions were plotted in the three genotypes with the highest proportion of generic effects, according to the method ^{Muñoz (1974)}.

The differences humidity at 60 DDS is only presented between levels of S and were highly significant (p≤ 0.01): in S0 the CVA averaged 49.9% and S1 29.6%. In contrast, the CVA between the experimental units assigned to the genotypes was not significant.

The effect of S in early genotypes showed highly significant differences (p≤ 0.01) in four of the seven variables evaluated: S1 reduced by 41.5% the REND, the above as a result of decreased PLC, VxM and P100S. Accordingly, ^{Desclaux et al. (2000)} indicated that the number of pods per unit of vegetative dry matter was significantly affected by drought also noted that stress in early stage of grain filling reduced the number of seeds plant^-1, while stress in late stage caused a decrease in the weight of the grain.

As for the effect of G, the differences were highly significant (p≤ 0.01) for AR2, AR7 and PLC, and significant (p≤ 0.05) for VxM and P100S, although VxM Tukey test did not separate groups. The H02-1991 genotype reached the highest REND which was 1145.2 kg ha^-1, but it was not statistically different from the others. The Figure 1 shows the interaction S*G, reflecting the specific effects to drought of the early genotypes selected by greater proportion of generic effects in REND. According to model 1 to evaluate drought resistance (^{Muñoz, 1992}), H02-2295 was the genotype that decreased less its REND when passing from S0 to S1, so it is inferred that it has a greater proportion of genetic effects specific to drought than allows you greater stability with or without moisture deficit.

Figure 1 Effect of the levels interaction of drought for genotypes on the performance of early, selected based on generic effects. S0= no drought; S1= with drought.  

Between levels of S in the intermediate genotypes, differences in CL90, CE90, VxM, P100S and REND were highly significant (p≤ 0.01). The characters that showed the most noticeable reductions due to S1 were VxM, CL90, P100S and CE90, thus reducing the REND by 48.7%. Studies ^{Makbul et al. (2011)} determined that chlorophyll content and stomatal conductance were reduced by 42% in drought-stressed soybean. Meanwhile, ^{Sadeghipour and Abbasi (2012)} confirmed that drought decreases, between 24 and 50%, the yield of soybeans both in the field and in the greenhouse. Considering the effect of G, it was highly significant (p≤ 0.01) in most of the variables, except for CE90 and VxM. The genotypes with the highest REND were H02-1656, H98-1240, Huasteca 400 and H06-0560. Of these, H98-1240 was the one of greater stability when passing from S0 to S1, whereas H02-1656 presented specific effects for favorable conditions (Figure 2); that is, for environments where most of the years are good in terms of rainfall.

Figure 2 Effect of the levels interaction of drought for genotypes on the performance of intermediate selected based on generic effects. S0= no drought; S1= with drought.  

For late, the effect of S was highly significant (p≤ 0.01) in CL90, CE90 and REND, and significant (p≤ 0.05) in P100S. As with the intermediates, S1 reduced the level of CL90 and CE90, in this case by 21% and 51.1%, respectively. It also affected the P100S by 34.6% and together they decreased the REND by 60.9%. According ^{Ku et al. (2013)}, the stages most sensitive to soybean drought vary according to latitudes and cultivars; so ^{Eck et al. (1987)} found in the Douglas cultivar that drought reduced REND by 45% and 88% when stress was applied at R5 (start of seed formation) and R7 (start of ripening) respectively. The effect of G in the late group was significant (p≤ 0.05) only for CE90 and AR7. The REND biggest was for H10-0556 with 1267.3 kg ha^-1, although the analysis of variance and Tukey test indicated that was no different from other genotypes.

Unlike this study ^{Chinchilla et al. (1990)} concluded that soybean varieties differed in the ability to maintain REND under limited moisture conditions, and could be selected for their tolerance to drought. However, the separation of effects proposed by ^{Muñoz (1997)} was not done. In the case of stomatal conductance, they have been widely reported genotypic differences to keep stomata open despite drought stress (^{Vignes et al., 1986}). In this case the REND interaction in the total of the genotypes was plotted, because the group was composed of few members.

No specific effects to drought were observed, which can be seen by the parallelism of the lines of H10-0556 with that of the remaining genotypes (Figure 3). This suggests, according to ^{Muñoz (1974)}, that this behavior is due to the additive action of the effect of S and the effect of G and not by the multiplicative action of the two factors. So in this case, it is better to select by generic effects. The genotype H10-0556 was superior in both S0 and S1 (Figure 3), being able to be selected as resistant considering the generic effects; While in the other genotypes, H10-0242 decreased less the REND from S0 to S1 (Figure 3), due to a higher efficiency of CL90, CE90 and P100S.

Figure 3 Effect of the levels interaction of drought for genotypes on the performance of late yield. S0= no drought; S1= with drought.  

When comparing the effect of S between precocious groups (Table 1), P100S and REND decreased more drastically in the three: the second decreased less in the early and more in the late ones, considering the application of model 2 (^{Muñoz, 1997}).

Table 1 Effect of drought in percentage [(S0 - S1)/S0)*100] on each of the variables evaluated genotypes early, intermediate and late soybeans.  

CL90= clorofila a los 90 DDS; CE90= conductancia estomática a los 90 DDS; AR2= altura a R2; AR7= altura a R7; PxM= plantas m^-1; VxM= vainas m^-1; PLC= plantas cosechadas; P100S= peso de 100 semillas; REND= rendimiento.

In the case of CL90 and CE90 they showed important reductions in the intermediate and late, indicating that they are very sensitive characteristics to the water deficit in soybean: CL90 decreased almost twice in the late ones than in the intermediates, whereas the reduction of CE90 was similar in both groups, showing that the effect on perspiration is not determined by differences in phenological development. The decrease in P100S was more or less stable in three groups, probably because the grain density depended rather on the effect of G and S and not on the phenotype cycle of the genotypes. It should be noted that in some variables, S1 caused a positive effect (negative values), but the increase is not appreciable.