HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (2) Citing Articles via Google Scholar Google Scholar Articles by Lee, E. A. Articles by Kannenberg, L. W. Search for Related Content PubMed Articles by Lee, E. A. Articles by Kannenberg, L. W. Agricola Articles by Lee, E. A. Articles by Kannenberg, L. W. Related Collections Crop Genetics Maize Biometrics

CROP BREEDING, GENETICS & CYTOLOGY

Genetic Components of Yield Stability in Maize Breeding Populations

University of Guelph, Department of Plant Agriculture, Crop Science Building, Guelph, ON, N1G 2W1 Canada

^* Corresponding author (lizlee{at}uoguelph.ca).

	ABSTRACT

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

 
Phenotypic stability has long been recognized as an important target in plant breeding. Stability is influenced in part by the genetic structure, i.e., level of heterogeneity and heterozygosity, of the cultivar. Yet, very little is known about the genetic components underlying stability, and how population improvement strategies influence stability. We examined 12 maize (Zea mays L.) breeding populations selected via reciprocal recurrent selection (RRS), selfed progeny recurrent selection (S), or a method combining RRS and S (COM), to examine changes in the genetic structure of the phenotypic stability of three traits (grain yield, grain moisture, and broken stalks), and two associated selection indices. Partitioning of the genotype x environment sums of squares from diallel matings of the original (C₀) and advanced (C_A) cycle populations into linear trends indicated that only grain yield and the unadjusted performance index (UPI) followed a predictable linear response. Grain yield and UPI linear trends were further partitioned by Gardner and Eberhart Analysis III to examine the genetic components of stability. We found that recurrent selection (RS) improved grain yield stability, and that this trait is heritable, predictable, and mostly controlled through additive gene action. Improvement in grain yield stability was observed both in cross and per se performance and was accompanied by significant improvement in the mean performance of the populations. However, the improvement in grain yield stability did not result in substantial changes in the general combining ability (g_i) estimates of most populations. Our results indicate that grain yield stability can be improved through RS by selecting solely for mean performance across multiple environments.

Abbreviations: API, adjusted performance index • C₀, original cycle • C_A, advanced cycle • COM, combined recurrent selection • GCA, general combining ability • RS, recurrent selection • RRS, reciprocal recurrent selection • S, selfed-progeny recurrent selection • SCA, specific combining ability • UPI, unadjusted performance index • P, phenotype • G, genotype • E, environment • ANOVA, analysis of variance • CG, Canada-Guelph • CCGP, Cross Canada Gene Pool • RCBD, randomized complete block design • LR, linear regression • AMMI, additive main effects and multiplicative interaction • IPCA, interaction principal component axis

	INTRODUCTION

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

 
QUANTITATIVE GENETIC THEORY states that an individual's phenotype (P) is the product of the genetics (G) of the individual, the environment (E) that the individual is exposed to, and the interaction that occurs between the genotype of the individual and the environment (G x E). Large G x E effects tend to be viewed as problematic in breeding because the lack of a predictable response hinders progress from selection (Dudley and Moll, 1969). This idealized predictable response across multiple environments is generally referred to as stability (Cannon, 1932). When stability concepts are applied to breeding–selection for increased grain yield, the idealized genotype is one that is capable of utilizing the resources available in higher yielding environments and has a mean performance that is above average in all environments.

Yield stability is influenced in part by the genetic structure of the variety. More heterozygous varieties and more heterogeneous varieties are less affected by environmental differences (Lerner, 1954; Lewontin, 1957; Allard and Bradshaw, 1964). For example, maize double-cross hybrids have smaller G x E interactions and are more stable than maize single-cross hybrids (Sprague and Federer, 1951). More homogeneous generations, inbred and F₁, have larger G x E interactions than more heterogeneous generations, F₂ and BC₁, and a more homozygous generation, inbred, has a larger G x E interaction than a more heterozygous generation, F₁ (Valdivia-Bernal and Hallauer, 1991). To reduce the complications that the G x E interaction creates when selecting superior genotypes, many attempts have been made to (i) understand the environmental components causing the G x E interaction (Epinat-Le Signor et al., 2001; Jeutong et al., 2000), (ii) examine the G x E interaction biometrically (Lin et al., 1986; Zobel et al. 1988; Finlay and Wilkinson, 1963; Yan et al., 2001), and (iii) develop selection strategies that involve a stability parameter (Magari and Kang, 1993). However, very little effort has been focused on examining the genetic components underlying, in particular, yield stability. In maize, the limited information regarding the genetics of yield stability comes from diallel studies involving inbred lines. These results indicate that stability (both positive and negative) is heritable and controlled by additive gene action (Eberhart and Russell, 1966, 1969).

One approach to examining stability is to further partition the G x E interaction from a traditional analysis of variance (ANOVA) into linear trends and a departure from linear (residual) (Finlay and Wilkinson, 1963; Eberhart and Russell, 1966). Linear trends can then be further partitioned by Analysis III of Gardner and Eberhart (1966) to arrive at estimates of genetic effects. It is then possible to examine the genetic structure underlying yield stability in terms of heterosis, general combining ability (GCA), and specific combining ability (SCA). The linear portion of the G x E interaction presumably has an additive and a nonadditive genetic component. Understanding the relative importance and genetic nature of these components and how selection influences them will greatly enhance our knowledge of yield stability and will help determine if selection strategies designed to enhance stability are important in breeding programs. In this paper, we present the results from two diallels involving, respectively, the originating gene pool (C₀) and the most advanced cycle (C_A) of 12 breeding populations. Specifically, we were interested in examining (i) if genetic progress for yield stability has been achieved indirectly through RS, (ii) if RS produced changes in the genetic structure of the populations for yield stability, and (iii) the underlying genetic components that influence yield stability.

	MATERIALS AND METHODS

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

 
Genetic Materials
The 11 C₀ and 12 C_A (C₂–C₆) early (2550–2900 Ontario Crop Heat Units [OCHU] [Brown and Bootsma, 1993]) University of Guelph breeding populations in this study included CG-SynA (S), CG-SynB (S), CG-Wigor (RRS) and CG-SynA (RRS), CG-HOPE Elite A (RRS) and CG-HOPE Elite B (RRS), CG-Stiff Stalk (COM) and CG-Lancaster (COM), CG-CBI (RRS) and CG-CBII (RRS), and CG-Cross Canada Gene Pool A (RRS) and CG-CCGP B (RRS). The two CCGP populations were selected from a common C₀, resulting in 11 C₀ populations and 12 C_A populations. All other populations were unrelated except that CG-SynA (RRS) and CG-SynA (S) originated from different cycles of selection in the CG SynA gene pool. Four to six cycles of RS had been completed in each population except for the CCGP populations, which had undergone two cycles of RS. For each cycle, families within a population were selected visually during development and subsequently in replicated trials in two environments, i.e., locations. Further information regarding the populations can be found elsewhere (Lee, 2002). Unadjusted (UPI) and adjusted (API) performance indices were used as the selection criteria for advancement until the populations had attained a level of maturity comparable to adapted germplasm for 2600 OCHU. The UPI was the ratio of grain yield to the percentage of grain moisture at harvest, while the API was the same ratio, but based on standing plants only, i.e., whole-plot grain yield x the percentage of plants neither lodged nor with broken stalks. Progenies were ranked by each index and selected based on the sum of ranks to identify those progenies for intercrossing to form the next cycle. When satisfactory grain moisture at harvest was reached, selection was based only on the sum of ranks for grain yield and grain yield of standing plants. In general, 20 selected families were intercrossed to form a new cycle.

In 1999, separate half-diallel matings, not including reciprocals, were made with the C₀ populations (n[n - 1]/2 = 55 crosses) and C_A (n[n - 1]/2 = 66 crosses). For each cross, six ears were pollinated with bulked pollen from six plants from the other population. Reciprocal crosses were then bulked, resulting in a total of 12 ears per cross. Individual plants were used as either males or females in the crosses, but not both, for a total of 24 S₀ plants, representing 24 gametes from each population. This approach samples predominant alleles and linkage blocks; rare alleles and linkage blocks are likely not represented in the population crosses and is similar to other studies utilizing population crosses (Moll et al. 1977; de Leon and Lonnquist 1978; Douglas et al. 1961; Gouesnard et al. 1991). Seed was increased for each of the 11 C₀ and 12 C_A parental populations via chain sibbing and at least 70 individual ears were harvested from each population. Equal amounts of seed from each ear of the populations per se were bulked for use in the experimental trials.

Experimental Procedure and Data Collection
The experiment included 144 entries, 66 crosses from the C_A diallel mating, 55 crosses from the C₀ diallel mating, 12 advanced populations per se, and 11 C₀ populations per se. Yield trials were grown in both 2000 and 2001 at three southwestern Ontario locations (Alma [2500 OCHUs], Elora [2600 OCHUs], and Woodstock [2850 OCHUs]) in a 12 x 12 partially balanced lattice design with three replications at each location. The soil type at all locations is Guelph loam (Typic Hapludalf). Fertilizer (N, P₂O_5, and K₂O, respectively) was applied on the basis of soil tests at the rate of Alma (2000) 155, 58, and 74 kg ha^-1; Alma (2001) 122, 75, and 57 kg ha^-1 supplemented with 33 690 L ha^-1 liquid swine manure; Elora (2000) 140, 50, and 50 kg ha^-1 and 135, 32, and 76 kg ha^-1; Elora (2001); Woodstock (2000) 150, 66, and 108 kg ha^-1; and Woodstock (2001) 167, 59, and 75 kg ha^-1. Weeds were controlled by conventional herbicides. Experimental units were two-row plots, 5.78 m long, with a spacing of 0.76 m between rows. Trials were overplanted and thinned at the six-leaf stage to uniform stands of 68 000 plants ha^-1 (60 plants plot^-1). All trials were machine planted (New Idea four-row planter New Idea Company, Cold Water, OH) in May with Almaco planting cones and machine harvested (Almaco SPC40-2 two row combine [ALMACO, Allan Machine Company, Nevada, IA)] in October or November. Three traits and two indices were measured, respectively: machine-harvested grain yield (kg ha^-1) adjusted to 155 g kg^-1 grain moisture grain moisture at harvest, percentage of broken stalks (plants broken below the ear or inclined more than 45° from the vertical), UPI, and API. All five parameters were analyzed and will be referred to as traits.

Statistical Analysis
Data were combined across locations and analyzed as a partially balanced lattice by means of PROC MIXED and a randomized complete block design (RCBD) by means of PROC GLM (SAS, 1996). The efficiency of the lattice design was tested by determining the ratio of the lattice MS_error x 100/RCBD MS_error for each analyzed trait. Analysis of variance indicated only slight gains in efficiency when a lattice was used rather than a RCBD, and thus the RCBD analysis and means were used. Data across locations were pooled and analyzed as a randomized complete block design (RCBD) by PROC GLM for the plot-based linear regression (LR) and PROC GLM and PROC PRINCOMP for the mean based AMMI models using SAS (1996). The LR on environmental mean index model is

and the additive main effects and multiplicative interaction model (AMMI) is

where Y_gre is the yield of genotype g in replication r at environment e; µ is the grand mean; _g and ß_e are the genotype and environment deviations from the grand mean; _r(ß_e) is the replicate effect nested within an environment; _g is the yield sensitivity (regression coefficient) of genotype g to the change in the environmental mean; _n is the singular value for interaction principal component axis (IPCA) n; _gn and _en are the genotype and environment eigenvectors for axis n, respectively; _ge is the deviation from linear trends (LR model) or the residual not accounted for by retained IPCAs (AMMI model) for the G x E interaction, respectively; and _rge (or _ge) is the random experimental error. Both genotype (_n^0.5_gn) and environment (_n^0.5_en) IPCA scores are scaled so that yield is transformed to the square root of yield. When IPCAs are graphed together, they have the units of the square root of yield and the product of the scaled genotype and environment IPCAs, thus giving the value directly in terms of the model (Gauch, 1992). The _gß_e component (linear trends) in the LR model was further partitioned to the linear response due to GCA and SCA combining ability effects (Eberhart and Russell, 1966, 1969), according to Analysis III (Gardner and Eberhart, 1966). Genotypes in the LR model were considered stable when _g = 1. All effects partitioned out of the G x E interaction in the LR model were tested for significance by _gre mean squares (Finlay and Wilkinson, 1963), not pooled deviations mean squares (Eberhart and Russell, 1966). Thus, standard errors of the regression parameter estimates are from the _gre mean squares. In addition, an F test of the ratio of GCA to SCA mean squares was used to test the significance of additive to nonadditive variation or error, if the SCA is assumed randomly and normally distributed with constant variance (Lin and Binns, 1991). The _ge in the AMMI model was taken from the plot-based LR model divided by the average number of replications as suggested by Cornelius (1993) to test the significance of IPCAs by means of the F_GH2 test.

Least significant differences (LSDs) from zero and between estimates were calculated according to Griffing (1956) and Zhang and Kang (1997) for Analysis III. The calculation of LSDs for g_i estimates from zero, and between estimates were, respectively

where p was the number of parents in the diallel cross, n was the error degrees freedom, e was the number of environments, r was the number of replications, and MS_error was the respective source x environment mean square. Rank correlations were calculated between C₀ and C_A g_i estimates. All hypotheses were tested with a Type I error () of 0.05.

	RESULTS AND DISCUSSION

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

 
Analysis III
Environments, genotypes, and G x E were significant sources of variation for all traits (Doerksen et al., 2003). Genotype x environment was further partitioned into linear trends and deviations from linear. The linear trend was not significant for grain moisture, broken stalks, or API. Grain yield and UPI were the only traits where both linear trends and deviations from linear trends across environments were significant (Table 1) . However, as a percentage of the G x E sums of squares, linear trends for grain yield and UPI accounted for only 25 and 26% of the variance, respectively. Linear trends were further partitioned into C_A vs. C₀, E x C₀–linear, and E x C_A–linear. The C_A vs. C₀ contrast was significant for both grain yield and UPI, indicating significant differences in phenotypic stability between C₀ and C_A. Furthermore, there were significant differences among regression coefficients within the E x C₀–linear and within the E x C_A–linear, indicating that differences in phenotypic stability exist between populations.

Stability Analysis
Yield stability was examined initially by AMMI analysis on the C₀ and C_A separately and combined. The first two principal components accounted for 41 and 20% of the variation, respectively. However, large environmental scores and small genotype scores were obtained, resulting in a tight clustering of the genotypes at the center of the plot (data not shown). Because of the lack of informativeness of the AMMI results, the LR approaches of Finlay and Wilkinson (1963) and Eberhart and Russell (1966) were used to examine the mean response of a population both in crosses and per se. Linear regression, a Type 2 stability statistic (Lin et al., 1986), was used because we were only interested in comparisons among the set of genotypes, and the LR approach allowed us to exploit the genetic structure that exists among the entries.

Because of the additive nature (significant GCA and lack of heterosis) of grain yield stability observed with Analysis III, we used, as a measure of a population's stability, the mean b_i values for a population averaged over all crosses involving that population (Table 2) . As implied by Analysis III, all C_A populations in crosses had significantly higher b_i values than their corresponding C₀ populations, showing that RS had improved the responsiveness of the populations to higher yielding environments. Accompanying this increase in responsiveness was also a significant increase in mean yield in crosses for all populations. Taken together, it appears that the responsiveness did not come at the expense of average performance. This trend is consistent across all populations and can be easily observed graphically (Fig. 1–11) . The deviations associated with the average b_i values were not very informative in this set of material (Table 2). Also, when per se yield stability was examined, no pattern of improvement was apparent. For some populations, no significant change in b_i values was observed, while other populations did exhibit an improvement in responsiveness. Somewhat surprisingly, the SynB(S) population actually showed a loss of responsiveness from C₀ to C_A (1.23 to 0.82). Even though the responsiveness of the populations per se was not always improved, all populations exhibited significant improvement in grain yield on a per se basis. In this set of 12 RS populations, RS has improved cross yield stability while simultaneously improving mean performance of the populations.

View larger version (14K): [in this window] [in a new window]   Fig. 1. Mean grain yield of all population crosses involving either EliteA C0 () or CA (•), and per se grain yield of either EliteA C0 () or CA () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA population over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

View larger version (17K): [in this window] [in a new window]   Fig. 11. Mean grain yield of all population crosses involving either CCGP A/B C0 (), A CA (•), or B CA (), and per se grain yield of either CCGP A or B C0 (), A CA (), or B () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA populations over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

View larger version (14K): [in this window] [in a new window]   Fig. 2. Mean grain yield of all population crosses involving either EliteB C0 () or CA (•), and per se grain yield of either EliteB C0 () or CA () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA population over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

View larger version (14K): [in this window] [in a new window]   Fig. 3. Mean grain yield of all population crosses involving either Stiff Stalk C0 () or CA (•), and per se grain yield of either Stiff Stalk C0 () or CA () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA population over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

View larger version (14K): [in this window] [in a new window]   Fig. 4. Mean grain yield of all population crosses involving either Lancaster C0 () or CA (•), and per se grain yield of either Lancaster C0 () or CA () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA population over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

View larger version (14K): [in this window] [in a new window]   Fig. 5. Mean grain yield of all population crosses involving either Wigor C0 () or CA (•), and per se grain yield of either Wigor C0 () or CA () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA population over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

View larger version (15K): [in this window] [in a new window]   Fig. 6. Mean grain yield of all population crosses involving either SynA(R) C0 () or CA (•), and per se grain yield of either SynA(R) C0 () or CA () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA population over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

View larger version (14K): [in this window] [in a new window]   Fig. 7. Mean grain yield of all population crosses involving either CBI C0 () or CA (•), and per se grain yield of either CBI C0 () or CA () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA population over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

View larger version (14K): [in this window] [in a new window]   Fig. 8. Mean grain yield of all population crosses involving either CBII C0 () or CA (•), and per se grain yield of either CBII C0 () or CA () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA population over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

View larger version (14K): [in this window] [in a new window]   Fig. 9. Mean grain yield of all population crosses involving either SynB(S) C0 () or CA (•), and per se grain yield of either SynB(S) C0 () or CA () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA population over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

View larger version (14K): [in this window] [in a new window]   Fig. 10. Mean grain yield of all population crosses involving either SynA(S) C0 () or CA (•), and per se grain yield of either SynA(S) C0 () or CA () populations over six environments. Dashed lines represent the response of the C0 population and solid lines represent the response of the CA population over a range of environments. Environmental index is the grand mean grain yield of the experiment in that environment.

Consequences of Recurrent Selection in the 12 Breeding Populations
We have now examined two important aspects of the 12 RS populations. First, we were interested in what changes in genetic structure had occurred that influenced grain yield (Doerksen et al., 2003). We found that genetic improvement occurred in both the per se and cross performance of most populations. Accompanying the favorable changes in population performance were less favorable shifts from predominantly additive genetic effects in C₀ to greater nonadditive genetic effects in C_A. This shift did not substantially change g_i estimates of most populations. However, in the case of grain yield, the underlying components of g_i effects were altered in their relative importance. General combining ability effects in C₀ were caused primarily by population per se effects (v_i), while in C_A, the GCA effects were caused predominately by parental heterotic effects (h_i) (Doerksen et al., 2003).

The other aspect of these populations, addressed in this paper, is phenotypic stability for grain yield. Yield stability is heritable and predictable. It has also been improved through recurrent selection. In both gene pools, i.e., C₀ and C_A, there appears to be sufficient heterozygosity and heterogeneity that heterosis (i.e., population crosses) does not confer any additional stability. Mostly additive genetic affects are controlling grain yield stability, with no apparent shift toward more nonadditive effects in C_A. Recurrent selection for only grain yield across multiple environments has resulted in improved cross yield stability while simultaneously improving mean performance of the populations. It is important to note that the number of test environments during RS was only two per cycle, i.e., four cycles = eight performance tests, six cycles = 12 performance tests. Apparently, the two environments per cycle, plus the different years of testing each cycle, even though a relatively small number (eight to 12 for most populations), has been sufficient to select (indirectly) for stability of grain yield. Or, interpreted another way, the reduction of deleterious gene frequencies for grain yield, which would be especially effective in the early cycles of RS, develops a more refined heterogeneous, heterozygous population, and therefore a more stable one. In effect, a good yield gene is also a good stability gene.

We also can draw some preliminary conclusions regarding the effectiveness of the various methods of improvement, S, RRS, and COM. One of the most surprising observations from the diallels is that in many cases in C_A, the crosses involving the RRS pairs were not superior to crosses involving other populations (Table 3) . And in some instances, Wigor-SynA(R) and Stiff Stalk-Lancaster, most of the crosses involving other populations were superior to the RRS pair cross. This is contrary to RS theory. Given this set of data, it appears that RRS will not result in population pairs that are consistently superior to populations improved independently from one another.

	NOTES

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

Received for publication May 27, 2003.

	REFERENCES

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

 

• Allard, R.W., and A.D. Bradshaw. 1964. Implications of genotype-environmental interactions in applied plant breeding. Crop Sci. 4:503–508.[Free Full Text]
• Brown, D.M., and A. Bootsma. 1993. Crop heat units for corn and other warm season crops in Southern Ontario. Ont. Minist. Agric. Food Factsheet, Agdex 111/31.Ontario Ministry of Agriculture and Food, Toronto, ON.
• Cannon, W.B. 1932. The wisdom of the body. Norton, New York.
• Cornelius, P.L. 1993. Statistical tests and retention of terms in the additive main effects and multiplicative model for cultivar trials. Crop Sci. 33:1186–1193.[Abstract/Free Full Text]
• de Leon, J.L., and J.H. Lonnquist. 1978. Heterosis in full-sibs within and between half-sib families in open-pollinated varieties of maize. Crop Sci. 18:26–28.[Abstract/Free Full Text]
• Doerksen, T.K., L.W. Kannenberg, and E.A. Lee. 2003. Effect of recurrent selection on combining ability in maize breeding populations. Crop. Sci. 43:1652–1658.[Abstract/Free Full Text]
• Douglas, A.G., J.W. Collier, M.F. El-Ebrashy, and J.S. Rogers. 1961. An evaluation of three cycles of reciprocal recurrent selection in a corn improvement program. Crop Sci. 3:157–161.
• Dudley, J.W., and R.H. Moll. 1969. Interpretation and use of estimates of heritability and genetic variances in plant breeding. Crop Sci. 9:257–262.[Abstract/Free Full Text]
• Eberhart, S.A., and W.A. Russell. 1966. Stability parameters for comparing varieties. Crop Sci. 6:36–40.[Abstract/Free Full Text]
• Eberhart, S.A., and W.A. Russell. 1969. Yield and stability for a 10-line diallel of single-cross and double-cross maize hybrids. Crop Sci. 9:357–361.[Abstract/Free Full Text]
• Epinat-Le Signor, C., S. Dousse, J. Lorgeou, J.-B. Denis, R. Bonhomme, P. Carolo, and A. Charcosset. 2001. Interpretation of genotype x environment interactions for early maize hybrids over 12 years. Crop Sci. 41:663–669.[Abstract/Free Full Text]
• Finlay, K.W., and G.N. Wilkinson. 1963. The analysis of adaptation in a plant breeding programme. Aust. J. Agric. Res. 14:742–754.[ISI]
• Gardner, C.O., and S.A. Eberhart. 1966. Analysis and interpretation of the variety cross diallel and related populations. Biometrics 22:439–452.[ISI][Medline]
• Gauch, H.G. 1992. Statistical analysis of regional yield trials: AMMI analysis of factorial designs. Elsevier, Amsterdam.
• Gouesnard, B., A. Gallais, M. Lefort-Buson, and J.P. Sampoux. 1991. Comparison of direct and correlated expected responses to intrapopluation and interpopulation selections for two forage maize populations. Maydica 36:49–56.
• Griffing, B. 1956. Concept of general and specific combining ability in relation to diallel crossing systems. Aust. J. Biol. Sci. 9:463–493.
• Jeutong, F., K.M. Eskridge, W.J. Waltman, and O.S. Smith. 2000. Comparison of bioclimatic indices for prediction of maize yields. Crop Sci. 40:1612–1617.[Abstract/Free Full Text]
• Lee, E.A. 2002. Corn breeding and genetics germplasm page. Available at http://www.plant.uoguelph.ca/research/corn_breeding/Populations2.htm (verified 4 June 2003).
• Lerner, I.M. 1954. Genetic homeostasis. Oliver and Boyd, London.
• Lewontin, R.C. 1957. The adaptation of populations to varying environments. Cold Spring Harb. Symp. Quant. Biol. 22:395–408.[Medline]
• Lin, C.S., and M.R. Binns. 1991. Genetic properties of four types of stability parameters. Theor. Appl. Genet. 82:505–509.[ISI]
• Lin, C.S., M.R. Binns, and L.P. Lefkovitch. 1986. Stability analysis: Where do we stand? Crop Sci. 26:894–900.[Abstract/Free Full Text]
• Magari, R., and M.S. Kang. 1993. Genotype selection via a new yield-stability statistic in maize yield trials. Euphytica 70:105–111.
• Moll, R.H., A. Bari, and C.W. Stuber. 1977. Frequency distribution of maize yield before and after reciprocal recurrent selection. Crop Sci. 17:794–796.[Abstract/Free Full Text]
• SAS Institute. 1996. The SAS system for windows. Version 6.12. SAS Inst., Cary, NC.
• Sprague, G.F., and W.T. Federer. 1951. A comparison of variance components in corn yield trials. II. Error, year x variety, location x variety components. Agron. J. 43:535–541.[Free Full Text]
• Valdivia-Bernal, R., and A.R. Hallauer. 1991. Estimates of genetic homeostasis in maize. Bras. J. Genet. 14:483–499.
• Yan, W., P.L. Cornelius, J. Crossa, and L.A. Hunt. 2001. Two types of GGE biplots for analyzing multi-environment trial data. Crop Sci. 41:656–663.[Abstract/Free Full Text]
• Zhang, Y., and M.S. Kang. 1997. DIALLEL-SAS: A SAS program for Griffing's diallel analyses. Agron. J. 89:176–182.[Abstract/Free Full Text]
• Zobel, R.W., J.J. Wright, and H.G. Gauch. 1988. Statistical analysis of yield trial. Agron. J. 80:388–393.[Abstract/Free Full Text]

This article has been cited by other articles:

				F. J. Kutka and M. E. Smith How Many Parents Give the Highest Yield in Predicted Synthetic and Composite Populations of Maize? Crop Sci., September 1, 2007; 47(5): 1905 - 1913. [Abstract] [Full Text] [PDF]

				E.A. Lee, R. Chakravarty, B. Good, M.J. Ash, and L.W. Kannenberg Registration of 38 Maize (Zea mays L.) Breeding Populations Adapted to Short-Season Environments Crop Sci., November 21, 2006; 46(6): 2728 - 2733. [Full Text] [PDF]

				Y. Zhang, M. S. Kang, and K. R. Lamkey DIALLEL-SAS05: A Comprehensive Program for Griffing's and Gardner-Eberhart Analyses Agron. J., June 17, 2005; 97(4): 1097 - 1106. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (2) Citing Articles via Google Scholar Google Scholar Articles by Lee, E. A. Articles by Kannenberg, L. W. Search for Related Content PubMed Articles by Lee, E. A. Articles by Kannenberg, L. W. Agricola Articles by Lee, E. A. Articles by Kannenberg, L. W. Related Collections Crop Genetics Maize Biometrics