I reviewed the maize breeding literature with a focus on the United States, delineating when key changes occurred and by whom. The yield potential of OP cultivars in the United States from the mid-1800s to the early 21st century was then addressed. Reports of multiple cultivar and multi-environment trials, particularly where improved management was more likely to have been practiced, were used to generate yield means under a variety of conditions. Averages for better-adapted cultivars over many environments were sought to best represent the potential of this type of cultivar.

Rates of gain for methods of maize improvement were generated using published summaries [10,47]. Data from the experiments of Weyrich et al. with the BS11 maize population were also incorporated [48]. Very high rates of gain (>10%/cycle) observed with some tropical and other populations were ignored for the sake of generating conservative estimates more relevant to potential outcomes for high yielding populations in the higher latitudes.

Conservative estimates of rates of gain from selection, 0.6% annually for gridded mass selection and 1.3% annually for modified ear-to-row selection (based on an average of 83 kg/ha applied to a hypothetical population yielding 6270 kg/ha), were used to predict the yield of OP cultivars had they been bred for 80 years (since 1930), 60 years (since 1950) or 40 years (since 1970) because these two methods do not require controlled pollinations, directly provide seed for planting, and are possibly the most adaptable to on-farm breeding [35,48]. Estimated yields of OP cultivars under modern management, from the review of OP cultivar yields, were multiplied by these generated rates of gain. The predicted yield gain quantity from that first cycle was then added to the base yield for successive years to predict accumulated gains over time: steady gains can continue for many years [29]. I also sought out literature concerning synthetic and composite populations, including tropical maize breeding reports. Published results were used to generate predictions of the potential of this technique to form better OP cultivars.

Indigenous farmers in the Americas developed maize and methods for seed production; farmers around the world continued this same process of mass selection wherein seeds from good ears or plants were saved each year [14-17,20,21,49]. Selection produced widely used cultivars, such as Improved King Philip Flint, Leaming, and Silver King [18,50,51]. Composite breeding, a technique developed by Native and Immigrant farmers, formed the entire race of Corn Belt Dent and cultivars like Reid's Yellow Dent, Krug's Yellow Dent and Falconer by crossing two racial types [16,26,48-50].

In the early 20th century, maize breeders and extension educators promoted a wide variety of selection techniques that were considered to be an improvement over traditional techniques [13,23,50,52-54]. Most focused on mass selection for ear and seed qualities thought to be related to yield in “pure strains,” although some more intensive systems of selection were also being promoted. Universities and associations of corn growers also established corn shows at fairs to exhibit “perfect ears” that were well matured and matched for uniformity. Rist wrote that, “If we are to estimate the value to be derived from corn shows in increased yields alone it probably would not amount to much, as we know that ‘show ears’ do not as a rule increase the yield of corn. On the other hand farmers of this state have paid thousands of dollars in tax money for scientific investigation along the line of corn improvement, yet such investigations have resulted in practically no increase in the yield of corn. Therefore, farmers could object to further investigations fully as well as could those who are opposed to corn shows object, as their results have been about the same so far as increased yields are concerned” [55]. In 1921, W.L. Burlison at the University of Illinois wrote that, “from the experiments conducted at this station it appears that while selection has had some effect in increasing the yield of corn, it has not given the results that were anticipated” [23].

Early breeding experiments with maize populations did not often yield impressive results for many breeders at many locations, and breeding based on ear type, though wildly popular and heavily promoted by universities, was even associated with yield reductions over time [13,52,53,56-59]. Datta visited maize breeders across the country and reviewed their techniques [52]. What he found was a mindset of progress through science and the land grant mission, along with a general abhorrence of inbreeding. However, though selection had improved the uniformity and local adaptation of many populations [49,50,60], “scientific” selection techniques promoted before 1920 appear not to have been fully evaluated before their widespread promotion [16].

During the 1920s, the selection program of the University of Illinois Extension Service was apparently successful in increasing yields using mass selection. J.C. Hackleman promoted selection against male plants that appeared barren, against plants with disease, and for high rates of germination. He reported yield gains of 300–600 kg/ha (5–15%) with this method [23]. Given the yields of the day, these were reasonable increases, especially since this was indirect selection for yield.

Reported yields of OP cultivars in university and other scientific trials were usually more than 2000 kg/ha in most parts of the United States when averaged across years, locations, and often times cultivar as well (Table 1), however, it is possible that very poor yields were under-reported [35]. It is clear that the yield potential of many OP cultivars with good management was often over 3000 kg/ha before modern management practices. In recent trials, yields of better-performing OP cultivars were often over 4400 kg/ha, a yield substantially lower than most hybrid checks [9,29,33,61-63].

High OP yields were familiar to many before 1930. Yields over 6200 kg/ha were recorded during the mid-1800s [20,21]. Montgomery reported that the four largest yields of maize on record in the United States at that time were over 12542 kg/ha for ears at husking [99], while Bowman reported the largest yield of grain belonged to South Carolina's Z.J. Drake in 1889 (15990 kg/ha) [50]. Hartley wrote that “good farmers frequently raise from 75 to 100 bushels of corn per acre [4700–6270 kg/ha]” and yields this high were regularly observed in Iowa and Wisconsin [51,54]. Cornell University cooperative trials showed average dry grain yields over 3140 kg/ha and at times over 4390 kg/ha, similar to yields reported by notable New York farmers in the 1840s [21,57,73,100,101].

There were two major problems for maize yields prior to 1930. Many farmers were not producing high quality seed (not well dried, freeze damage, not well selected, inbred) and were therefore obtaining poor stands and yields lower than the genetic potential of their cultivars [51,53,54,59,74,102,103]. The other problem was that soil and crop management improvements were not widely adopted, leading to unsustainable production in some cases [34,36]. Kent wrote, “By intense cultivation and proper rotation, most of the farms of Iowa would produce from 75–80 bushels of corn per acre [4700–5020 kg/ha], and under favorable climatic conditions still more” [70]. Yields in Missouri of 4870 vs. 740 kg/ha and in Illinois of 4000 kg/ha vs. 1690 kg/ha were reported for rotated vs. continuous cultivation [78,104]. Still, many farmers did not rotate or fertilize enough because: (1) fertilizers were not always economical [76,77], (2) the value of continuous maize was higher than the value of some alternative crops [105], or (3) for some other reasons. Smith wrote that “the land was corn sick” [104].

One example of the importance of management for yields comes from Kansas where maize yields had fallen steadily, leading to an unsustainable situation before 1920 [79]. At that time one of the highest yielding OP cultivars was Pride of Saline, with an average yield over 1880 kg/ha [78]. In the 1940s, Pride of Saline was yielding 3140–3760 kg/ha and after 1960 it often yielded above 6270 kg/ha (Figure 3). This is more than three times the Pride of Saline yield from university trials before 1920 with no reported breeding effort.

A complete consideration of the effects of plant breeding during the 20th century includes all of the factors that changed [22,36,114]. No more than 60–70% of the yield gain since 1930 should be attributed to breeding and using hybrid cultivars [29,115]. Hybrid yields in Kansas tests were on average 8–22% higher than Pride of Saline prior to 1960, and this difference increased after 1960 (Figure 2). However, while hybrids often outperformed their OP checks in the 1930s and 1940s [22,30,31,85-88], Reid's Yellow Dent has yielded about as well as many commercial hybrids from the 1940s and early 1950s, but substantially less than post-1960 hybrids in recent tests in Iowa [9,29]. Also, improvements in standability in hybrids have been more dramatic still, making for greater gains for machine harvested yields and for more flexible grain harvest schedules [9,28-30,97,98]. Could new, more competitive OP cultivars that stand well be bred now?

Farmers had been making crosses among cultivars for thousands of years, and then reselecting new and more vigorous offspring from among the following generations with some success [49,50]. In Michigan, Beal suggested the use of F1 varietal hybrids in the 1870s as did Carrier in Virginia [4,26,116,117]. These new hybrid seeds did find some commercial use in the early 20th century, although like any hybrid seeds they required annual regeneration and further improvement in these hybrids would require ongoing improvement of each parent population.

Around 1910, some breeders began to think in new ways, to see vigorous cultivars, both uniform and genetically diverse, that had first passed through a period of inbreeding [11,17,44,52,80,114,118]. Early data with inbred hybrids showed great potential in the method, although it would be expensive and removed from the hands of farmers [23,49,94]. Richey said evidence suggested “pure line methods as the only sound basis for real improvement of corn” [119].

Richey reported that F1 varietal hybrids on average did not yield more than the parent cultivars and also noted small or no gains on average from ear-to-row breeding [119]. However, he then expounded on the best yields from Jones' hybridization work rather than showing that 11 of the 25 Leaming hybrids yielded less than the OP parent [80,119]. Concerning hybridization experiments in Nebraska, Richey wrote, “It is unlikely that all of the crosses yielded equally, and some, therefore, probably exceeded this 0.8 bushel [50 kg/ha] increase” [119]. Had Richey applied the same focus on the best gains from the other breeding methods he would have reported that 2.5% of F1 varietal hybrids outyielded the better parent by 26% or more and would have noted good yield gains in some ear-to-row breeding experiments [119]. Therefore, Richey's conclusion about pure line methods was only partly correct considering the data presented. On average, hybrids did not yield better than their OP parent cultivars, but some hybrids yielded much better, as is still true today.

At the time, the difficult task of finding the best hybrid combinations provided rapid yield gains in the short run but did not make for long term improvements. However, the method proved wildly successful after years of intensive development when released hybrids often outyielded OP cultivars by 9–40% [10,11,17,22,23,29,31,37,86-88,91,118,120]. From on-farm ear selection and no inbreeding with “pure” populations, breeders appear to have leapt to no population improvement and little farmer involvement. During neither period were decisions fully evaluated through experimentation.

After 1922, most university maize programs in the USA took up inbreeding studies [89]. Some maize breeders moved out of breeding and into different roles [22]. Hartley, the head of USDA's maize improvement efforts, was pushed from the USDA Bureau of Plant Industry over his conservatism in favor of on-farm seed production and breeding [23]. This parallels changes underway in plant science departments world-wide as molecular and transgenic techniques and courses of study have come into favor and funding for conventional public plant breeding research and education has dwindled even though the approach still works well [7,116,121,122].

Interest among hybrid maize breeders in recurrent or repeated selection rebounded in the 1940s and 1950s when they discovered the need for improved populations from which to select new inbreds [16,47,123], and this time the conclusions were very different. OP cultivars had been thought to lack the genetic variation for successful selection and this was disproven via experimentation [16,124]. Gardner wrote, “mass selection would appear to be as effective as hybridization in increasing yield” [96]. Webel and Lonnquist said of modified ear-to-row selection, “The realized gain suggests the method might be as effective as hybridization in increasing yield” [125]. Lonnquist and Gardner found “that no critical experiments where yield was the main criterion for selection were reported during the period covered by Richey's report” [95]. Sprague noted “it appears that the ear-to-row method of selection for yielding ability was discredited not because of genetic limitations of the method, but because of the inadequate field plot technique” [16]. Research after Richey's 1922 conclusion showed recurrent selection and varietal hybrids could be very successful at increasing maize yields after all, although they could not provide cultivars with the yields of the best hybrids [35,47,81,95,119,120,126-131].

The Hays Golden OP cultivar from Nebraska showed a yield increase of 40% from several years of recurrent selection for yield alone as compared to the 9–40% yield advantage that many hybrids had over OP cultivars with many years of development [30,47,88,123]. Average gains in yield for gridded mass selection range from 1.8–2.6% annually [10,35,47,120,132]. Modified-ear-to-row selection has demonstrated average gains per year of about 2.1–3.5% [10,35,47,48,133]. These and other recurrent breeding techniques have been very successful in breeding cultivars for farmers in the tropics [120,128,131,132]. Conservative estimates of average potential annual yield gains of 82 kg/ha for mass selection and 83 kg/ha for modified-ear-to-row selection, largely from North American data, are comparable to those experienced in an open-ended inbred hybrid breeding program from the United States, where yields have increased by about 66 kg/ha each year since 1930 [9,29,35]. Both Coors and Duvick provide lists of reasons why these results for recurrent selection might be biased comparisons and might at times favor the results for recurrent selection over the inbred-hybrid method [35,36]. Added to that, recurrent programs rarely if ever achieve the level of selection and evaluation intensity described for commercial hybrid breeding [17]. Regardless of type of cultivar, annual improvements via breeding are essentially about the same. The fastest and cheapest method which best supports sustainable agriculture outcomes will depend on yields in the target agroecosystems, grain prices, and farmer skills and interests.

Gardner's modification to mass selection techniques was to focus on yield among plants in small grids or plots across a seed selection field [96]. This reduces the environmental effects on observed plant phenotypes and improves selection for better genotypes, although it does not allow for evaluation in multiple agroecosystems. The method is easily adapted to farm and garden situations, although it is better for qualitative traits than for yield and may need the two step procedure suggested by Hartley for larger acreages [54]. Modified ear-to-row selection involves replicated evaluation of plant progenies at three locations and is a more intensive technique that would probably require special training for farmers or even a participating breeder [134]. Prior to 1910, some breeders had adopted replication, check rows, detasseling and the use of remnant seed in some ear-to-row breeding trials for better plant evaluations [13,52,58,135]. Use of the triple lattice design for progeny tests (a component of Lonnquist's method) was unlikely before demonstrations of its utility in accounting for finer spatial variation of soils in the early 1940s and the use of bulk seed rows for cross-pollination appears to have originated with Lonnquist [52,134,136,137]. The method often results in rapid gains and the availability of computers makes it and other complex recurrent selection methods a possibility for many farmers, especially those working in groups or when partnering with breeders.

What would have happened if these techniques were applied in the past had recurrent selection not been dropped prematurely? Some OP cultivars in the United States have yields of 5020–6270 kg/ha which are not currently economically advantageous [28]. Using conservative estimates of rates of gain, calculations of the predicted gains in yield suggest large gains were and are possible with enough breeding effort (Table 2). Gridded mass selection could have resulted in OP cultivars yielding 9280 kg/ha or more by now had selection begun by 1930. Lonnquist's modified ear-to-row method could have resulted in much faster improvements and in fairly competitive OP cultivars even if it had been employed only since its development [134]. Applying recurrent selection now, it would take many years to develop an economically competitive OP cultivar. Competitive commercial OP cultivars for the USA and some other regions were not developed and breeders (and farmers) would have had to overcome the stigma of being old-fashioned and the preference of governments and many farmers for hybrids [23]. Had higher yielding OP cultivars been released prior to 1960, would there have been an advantage? Would many farmers use high yielding OP cultivars today? Would there be an advantage in time, profitability, adaptation, etc.?

There are other biological problems to consider. Single plant evaluation in mass selection can increase yield under some circumstances but often results in delayed maturity and no improvement in stalk strength and other critical agronomic traits. Hyrkas and Carena and Bletsos and Goulas were unable to increase yield of improved populations using mass selection and suggested using more intensive recurrent selection techniques that could be more difficult for farmers although predicted rates of gain can be higher [35,48,133,138]. Gardner hit a yield plateau with Hays Golden after 12 cycles of mass selection; modified ear-to-row selection reached the same plateau in only six cycles [47]. Yield plateaus can be overcome by outcrossing, and then proceeding anew with the resulting composite population [131,139]. Eberhart et al. promoted the idea of cooperative work with many populations [128]. Such an approach could make available the improved populations and lines needed for outcrossing when breeding plateaus are discovered. At any time inbreds and inbred hybrids could be developed, making for a very comprehensive approach to maize breeding much akin to the successful public program at Iowa State University where theoretical studies of maize breeding technique included selection for agronomic traits as well as yield in both populations and inbreds. There need not be an either or choice with OP and hybrid cultivars.

Hybrid vigor has been harnessed by maize breeders for centuries [14,16,49]. Composite populations (intermated cultivars) and synthetic populations (intermated inbreds), like any maize population with a random mating structure, can be thought of and used as OP cultivars [140] and may offer a rapid approach to increasing OP cultivar performance. Shull and Duvick suggested that the yield of an OP cultivar is a result of all the possible hybrid combinations among the parents [36,44], and Wright demonstrated that the number and yield of parents in a synthetic population determined how much of the hybrid vigor could be retained in the F2 and later generations [141]. The equation based on his work, which was validated for both composites and synthetics by Mochizuki, predicts that as the number of parents increases the amount of retained hybrid vigor increases [142].

While some saw the potential for synthetic maize cultivars in the USA early on [80,143], little work was done and most results were not promising [16,144,145]. However, many breeders have had synthetic yields at least 15% above that of common OP cultivars and up to 90% of hybrid yields [83,93,97,118,146-149]. It appears that these high yielding synthetic populations never reached farmers in the USA or many other regions for their consideration.

The application of Wright's equation to historical data on maize inbreds and hybrids suggests the possibility of synthetics that yield more than 90% of commercial hybrids if the research were carried out (Table 3), although the work would be intensive with many questions to be answered [36,141,150]. Might such diverse populations open up new ways of dealing with diseases and pests as has worked for multilines of rice [151]? Could synthetics reduce seed costs, provide competitive yields and be further improved via selection on farms? Would many farmers be interested?

Some answers come from the tropics where maize breeding has embraced population improvement and inbreeding and breeders develop inbred hybrids, varietal hybrids, synthetics and other OP cultivars to best fit local needs [120,128,131,152-157]. In those regions synthetic cultivars have proven useful for farmers by providing higher yields and low seed costs, although for long term improvement synthetics may require more genetic diversity than that provided by only 8–10 lines and they may require more intensive selection schemes for further improvement [133,155]. A few farmers in the United States, especially organic farmers, are looking for good cultivars of maize and other crops from which they could produce their own seed as seed prices have risen sharply in recent years and the number of commercial cultivars without transgenes has diminished [33,140,158,159]. Baltensberger et al. suggested that dryland farming and similar regions with climatic stress may be the first agroecosystems to consider for this approach as these lower yielding environments result in similar economic returns for synthetics yielding only 75% of commercial hybrids [147]. Stress environments, and those where hybrids are not economically feasible, are the targets for modern OP cultivar development in the tropics [120,152-154]. Organic farmers may be interested as well, but high organic grain prices could make lost yield opportunities economically detrimental unless cultivars without recombinant DNA were otherwise unavailable [28].

Any type of hybrid is still an option for on-farm seed production. Should sustainable farmers find it economically feasible to produce their own hybrid seed, varietal hybrids might still be a valuable option among several (single cross, top cross, double cross, varietal cross). Improved populations, whether synthetic or otherwise, could be crossed to produce F1 seed each year should those populations demonstrate hybrid vigor when crossed. Advantages of this approach could be higher yields than possible with populations per se, cheap hybrid seed given the levels of production that may be possible with populations compared to expected yields of inbred parent lines, and the possibility for ongoing improvement of the parent populations and their hybrids on-farm. Disadvantages could include separate seed production requirements, likely lower yield than single cross hybrids [129], and lower uniformity than that observed in single cross hybrids.

Cleveland suggested breeders were neither pure scientists nor only servants to social constructs, and this conclusion appears to be affirmed here [12]. Maize breeders come with their own personal interests for the work to be undertaken and certainly make subjective decisions about specific approaches and goals [8]. The widely promoted selection methods and shows before 1920 and the rejection of recurrent selection from 1922 until the late 1940s show maize breeders and their administrators sometimes got ahead of scientific validation in order to pursue exciting new options. There were some decisions involved in promoting research and development of hybrid cultivars to benefit specific seed businesses [23,37], and most professional maize breeders in the USA ignored the potential of recurrent selection in favor of the task of testing tens of thousands of lines in the 1920s and 1930s [118,119]. There were, however, many promoters of hybrid cultivars who had farmers in mind, and the science behind hybrid breeding methods eventually proved sound and useful to all forms of maize improvement [10,22,160]. Excellent and creative work was carried out that has stood the test of time and provided exceptional new cultivars of many types [10,16,17,89,123,161,162].

In the past thirty years maize breeding has changed further with patents, DNA marker assisted selection, and transgenic techniques joining the process. Public breeding has been fading in favor of breeding by major corporations around the world [116,163]. One might question whether this is once more the promotion of the novel in place of approaches that objectivity might instead focus upon [6,45,105]. Experience tells us that there can be some unanticipated problems with new technologies rushed to market [164], yet sometimes full economic and ecological evaluations are still unavailable for transgenic maize cultivars until well after commercial release [165-167]. Cox reiterated a calculation from Goodman and Carson comparing conventional hybrid breeding with the more expensive transgenic methods which are apparently no faster to produce new cultivars than conventional methods [168]. This situation is reminiscent of Gardener's favorable comparison of recurrent selection with hybrid breeding decades after the switch to hybrids [96]. At the American Seed Trade Association conference in 2003, John Dudley, then maize breeder from the University of Illinois, asked the audience, “Current corn yield goes up 1.9 Bu/A annually. What [trans]genes would improve this?” Given our experiences, an open and logical discussion of the goals, means, and costs of modern maize breeding appears in order as has always been the case.

Rather than the “March of Progress” from OP cultivars to high yielding hybrids and on to transgenic technologies, perhaps the scientific and social successes and failures of actual maize breeders (e.g., Darrah, Duvick, Hallauer, Hartley, Hayes, Gardner, Goodman, Lonnquist, Miranda Filho, Pandey, Paterniani, Richey, Sevilla, Sprague, Stringfield, Troyer, Will, etc.) in meeting the needs of farmers and consumers would be more objective and useful for students contemplating breeding for sustainability. And rather than a competition between OP and hybrid cultivars, perhaps it would be most sensible to consider this thought from Pandey et al.: “The critical question is not whether hybrids are superior to OPVs, but whether a product is superior to what the farmers grow and which new product they can afford” [152]. Breeding, like any tool with which humanity hopes to derive a better world, requires knowledge, critical thinking, hard work, humility, cooperation, objectivity, and a broad perspective in order to be successful [17,122]. If we are to make the most reasonable choices for a sustainable agriculture, it appears most fruitful to attend to all of them.