Are Bordered Plots Necessary for Corn Performance Tests?

Search PMN

Kraig L. Roozeboom, Assistant Professor, Department of Agronomy, Kansas State University, Manhattan 66506; and Mark M. Claassen, Professor, Harvey County Experiment Field, Kansas State University, Hesston 67062

Abstract

The range of heights and maturities present in a corn hybrid test has the potential to influence the relative importance of border rows, yet many breeding and performance testing programs use unbordered, two-row plots. A series of experiments was conducted at three sites in two years to evaluate the need for border rows in small-plot corn hybrid tests containing hybrids that differ in maturity and height. The six environments provided average grain yields of 77 to 227 bu/acre. Four corn hybrids provided contrasting combinations of relative maturity (RM) and height: 102 RM, medium-short; 110 RM, medium-short; 110 RM, medium-tall; and 116 RM, tall. Hybrids were planted in plots with four rows, 30 ft long, in a randomized complete block experimental design. The outer two rows of each plot functioned both as the border rows for the 4-row, bordered plot treatment and as the plot rows for the 2-row, unbordered plot treatment in a type of split-plot arrangement. Hybrid yield ranks did not change significantly in response to presence or absence of border rows in any of the six environments, supporting the conclusion that bordered plots are not necessary for accurate estimation of yield ranks in corn hybrid performance tests.

Introduction

The necessity of border rows for corn (Zea mays L.) performance test plots is still open to debate. The range of heights and maturities in a particular hybrid test has the potential to influence the relative importance of border rows (8). Several studies have documented competition between genotypes in the same trial, generally benefiting the taller, later genotypes and hindering the shorter, earlier genotypes. In a variety test with favorable early-season growing conditions followed by hot, dry conditions after August 1, Kiesselback (7) harvested each row of three-row plots separately to examine competition effects. In those conditions, taller varieties yielded more in rows beside shorter varieties than they did in rows beside taller varieties. Likewise, shorter varieties yielded less in rows beside tall varieties than they did in rows beside shorter varieties. Pendleton and Seif (10) used normal and dwarf versions of the same corn genotype to demonstrate that rows of normal height bordered by dwarf rows received a slight yield benefit, but dwarf rows bordered by normal rows demonstrated a significant yield decrease. They concluded that plant height and light competition were of primary importance. David and others (2) came to a similar conclusion more recently in France and recommended four-row plots as the best way to avoid such competition. Genter (4), comparing well adapted early and late maturing corn hybrids, concluded that for both maturities, hybrids had greater yields if bordered by the early hybrid on both sides and lesser yields if bordered by the late hybrid on both sides. Yields were intermediate if the hybrid was bordered by the early hybrid on one side and the late hybrid on the other. Hayes (5), reviewing a number of early studies dealing with controlling experimental error and field plot design in a number of crops, found inconsistent results but concluded that borders were necessary to avoid possible competition effects as well as other factors such as lodging. Esgar and Bullock (3), assuming the necessity of border rows, concluded further that border rows should be thinned as well as the center two "harvest" rows to minimize experimental error. Although demonstrating the value of border rows for removing competition effects and increasing experimental precision, some of these studies examined unbordered single-row plots (4,10). In other situations (3,4) the large number of replications facilitated the detection of differences much smaller than those typically found to be significant in hybrid performance tests (8).

The issue is complicated further by several studies that demonstrate little or no benefit from border rows. Olson (9) documented a negligible influence of border rows on grain yields in adjoining plots, even with two genotypes that differed drastically in stature and yield. Ross (11) concluded that sorghum plots with two rows do not need to be bordered if all entries were of a similar type. However, he stated that borders are required when entries possess large differences in height or growth habit. Bowman (1) demonstrated a greater gain in efficiency from increasing the number of replications than from adding borders to plots. Keng and Hall (6), examining corn plot design in western Canada, found an increase in precision using bordered plots in some situations, but with uncertain implications for the practical interpretation of test results. They recommended adequate randomization and north-south row orientation to minimize border effects for single-row plots in that environment.

In response to questions about corn performance test plot design in Kansas and given the inconclusive nature of the literature, a series of experiments was conducted at three sites in two years (six environments) to evaluate the need for border rows in small-plot corn hybrid tests containing hybrids that differ in maturity and height.

Field Studies to Assess Importance of Border Rows

Four corn hybrids were selected to provide contrasting combinations of maturity and height based on company classifications. Average days to silk, grain moisture, and plant height for each hybrid ranked as expected in the current study and provided ranges of three days to silk, 2% grain moisture, and 10 inches of plant height (Table 1).

Table 1. Characteristics of hybrids used to compare bordered with unbordered plots.

Hybrid	Relative maturity	Height classification	Mean days to silk	Mean grain moisture (%)	Mean plant height (inch)
1	102	Medium-short	73	12.8	94
2	110	Medium-short	74	13.0	94
3	110	Medium-tall	74	13.4	99
4	116	Tall	76	14.8	104

Four replications of plots with four rows, 30 ft in length, were planted with each hybrid in a randomized complete block experimental design. The outer two rows of each plot functioned both as the border rows for the bordered plot treatment and as the plot rows for the unbordered plot treatment in a type of split-plot arrangement (1). Planting dates, fertilization, and weed control practices were as close to recommended best management practices for each location as weather conditions allowed. Uniform plant populations were achieved by overplanting and thinning to standard at one location and by overplanting and removing extra, closely spaced plants in the other locations.

Hybrid Rank Changes

Interaction graphs comparing hybrid yield ranks in 2-row and 4-row plots for each environment are presented in Figure 1. Yields differed between hybrids in each environment, and hybrid ranks differed between environments. In less favorable environments (Manhattan Year 2 and Hesston Year 2), the earliest maturing hybrid had the greatest yield. In the remaining, relatively favorable environments, the full-season hybrid had the greatest yield. In the four environments with yields greater than 100 bu/acre, hybrid ranks did not change when yields were estimated using 4-row plots rather than 2-row plots.

Fig. 1. Graphs showing hybrid-by-plot-type interactions for corn yield in six environments.

One rank change occurred when yields were estimated in 4-row plots rather than 2-row plots in each of the two environments with yields less than 100 bu/acre. Changes in the yield estimates for each hybrid involved in the rank changes were small relative to typical least significant difference (LSD) values from corn performance tests at the same locations (8). In Manhattan Year 2, the change in yield estimated from 2-row vs 4-row plots was -0.5 bu/acre for the 110 RM-Tall hybrid and +4.8 bu/acre for the 110 RM-Short hybrid. In Hesston Year 2, yield estimates for the 110 RM-Short hybrid changed by +7.1 bu/acre and by +4.5 bu/acre for the 116 RM-Taller hybrid in 4-row vs 2-row plots. Differences in yield between these hybrids also were small when estimated in either plot type. Yield estimates of hybrids that changed rank were within 1.2 bu/acre of each other in Hesston Year 2 and within 4.1 bu/acre in Manhattan Year 2 when estimated using 4-row plots. When yields were estimated in 2-row plots, these differences were 1.5 and 1.0 bu/acre respectively. These differences also were within typical error levels for corn hybrid tests, implying a low probability of erroneous ranking with 2-row plots compared to the rankings obtained in 4-row plots.

Analysis of variance using a mixed model with hybrid and plot type as fixed effects supplied additional evidence for the conclusion that hybrid yield ranks did not change significantly in response to presence or absence of border rows. The hybrid by plot-type interaction effect was not significant (α = 0.05) in all six environments, supporting the conclusion that rank changes which did occur were not significant. Contrasts testing 2-row vs 4-row plot yield estimates for each hybrid were not significant (α = 0.05) in all environments, indicating both plot types supplied a similar yield estimate for each hybrid.

Variability of Yield Differences

Another concern related to plot borders involves variability of yield estimates. It is often assumed that yield estimates from bordered plots should be less variable than those from unbordered plots. On the other hand, an experiment using unbordered plots occupies less area than it would with bordered plots, perhaps resulting in less variability in the experiment site. One measure of variability is the standard error of the difference between hybrid yield estimates (SED) (12). The SED values for each of the six possible hybrid comparisons within a location were used as replicated measures of variability in yield differences for both 2-row and 4-row plots. Analysis of variance of these SED values using plot type (2-row and 4-row) as the main effect detected significant differences in three of the six environments (Table 2). In all three instances, mean SED values associated with 4-row plots were larger than those associated with 2-row plots. The remaining three locations had no significant differences in SED values between hybrid yield estimates for 2-row vs 4-row plots.

Table 2. Mean standard errors of hybrid yield differences for 2- and 4-row plots.

Environment	Mean SED^x 2-row	Mean SED 4-row	Difference
Manhattan Year 1	5.9	8.6	2.7**
Manhattan Year 2	2.8	5.7	2.9**
Hesston Year 1	4.0	5.0	1.0*
Hesston Year 2	5.0	5.0	0.0
Inman Year 1	11.8	11.7	-0.1
Inman Year 2	20.3	24.5	4.2

*,** Significantly different than 0.0 at a = 0.05 and 0.01, respectively.

^x Standard error of difference.

Conclusion

The hybrids included in these experiments differed in maturity and height, the two factors most likely to influence neighboring, unbordered plots. The six experimental environments provided a wide range of yields and resulted in different hybrid rankings depending on the magnitude of yield. Although rank changes occurred in two of the six environments, the differences in yield were small and were less than that typically needed to conclude that the yields were different from each other. These results in addition to non-significant hybrid by plot-type interactions and 2-row vs 4-row contrasts for each hybrid at each location support the conclusion that 2-row, unbordered plots estimate yield ranks in corn hybrid performance tests no differently than 4-row, bordered plots.

Acknowledgments

We would like to thank Larry Koehn, Midwest Seed Genetics, for providing the impetus for this research and for supplying the corn seed used in the experiments. Norman and Tracy Schmidt provided land and support for the irrigated site at Inman, KS.

Literature Cited

1. Bowman, D. T. 1989. Plot configuration in corn yield trials. Crop Sci. 29:1202-1206.

2. David, O., Monod, H., Lorgeou, J., and Philippeau, G. 2001. Control of interplot interference in grain maize: a multi-site comparison. Crop Sci. 41:406-414.

3. Esgar, R. W., and Bullock, D. G. 1999. Thinning border rows differentially affects hybrids in corn yield trials. Crop Sci. 39:1358-1361.

4. Genter, C. F. 1958. Plot competition between corn hybrids. Agron. J. 50:205-206.

5. Hayes, H. K. 1923. Controlling experimental error in nursery trials. J. Am. Soc. Agron. 15:177-192.

6. Keng, J. C. W., and Hall, J. W. 1987. The effects of guard rows on variety test results of maize. Agric. Sys. 23:187-195.

7. Kiesselback, T. A. 1923. Competition as a source of error in comparative corn yields. J. Am. Soc. Agron. 15:199-215.

8. Lingenfelser, J., Roozeboom, K., Jardine, D., Whitworth, J., Knapp, M., Claassen, M., Maddux, L., Kimball, J., Long, J., Gordon, W. B., Heer, W., Evans, P., Kofoid, K., Schlegel, A., and Spangler, M. 2006. Kansas performance tests with Corn Hybrids. Agric. Exp. Stn. Rep. Prog. 968. Kansas State Univ., Manhattan, KS.

9. Olson, P. J. 1927. Competition between adjacent rows of corn. J. Am. Soc. Agron. 20:83-84.

10. Pendleton, J. W., and Seif, R. D. 1962. Role of height in corn competition. Crop Sci. 2:154-156.

11. Ross, W. M. 1958. A comparison of grain sorghum varieties in plots with and without border rows. Agron. J. 50:344-345.

12. Steel, R. G. D. and J. H. Torrie. 1980. Principles and Procedures of Statistics: A Biometrical Approach, 2nd Edn. McGraw-Hill, New York, NY.