1. Gamete Selection in Corn Breeding. The method of corn improvement commonly known as "selection in self-fertilized lines" has been remarkably effective in the development of types of corn far superior to any previously existing variety in yield and in other agronomic characters of practical value.

The general experience of corn breeders and the results of the experimental studies of breeding methods which they have made indicate that, if this job were to be done over, it would be possible to make comparable advances at a much smaller cost in time and labor. The chief results of the method experiments, as related to yield improvement, may be summarized as follows:

1. Visual selection for yield is practically ineffective. The extent to which a plant of given genotype will contribute to yield in hybrids can only be determined by yield testing of its hybrid progeny. The factor limiting the scope of breeding operations is the number of items which may be adequately tested for yield.
2. The combining value of a given genotype varies considerably in combinations with different genotypes. General combining value may be tested effectively in practice by crosses on mixed populations.
3. The inheritance of yield genotype is in general in agreement with expectation based on the hypothesis of complementary dominant favorable factors.
4. There is little or no advance in yield genotype in the course of inbreeding and selection as ordinarily practiced in the production of inbred lines. This fact, convincingly demonstrated by Jenkins, is the basis for current attempts to improve the efficiency of the breeding technic, for it shows that the method owes its success not to selection in self-fertilized lines, but to the unrecognized differences in genotype of the foundation plants.

Jenkins' results suggest the possibility that an appreciable fraction of the individual plants in open-pollinated varieties may be as high in yield genotype as the best present inbred lines. Obviously, the identification of these plants near the beginning rather than near the end of the breeding operations would make for greater efficiency, for it would concentrate the analysis upon populations with the highest content of desirable genotypes. In the few outstanding selected strains it would be feasible to use test-controlled selection in the first selfed generation, where genetic variability is at its maximum. Such selection might reasonably be expected to accomplish further improvement in yield.

This is an effective and practicable method for the further sampling of the open-pollinated varieties. It is not widely used in corn breeding at present, chiefly for these reasons:

1. The frequency of high yield genotypes among the plants of open-pollinated varieties is low enough to make their identification much less economical than that of comparable genotypes in populations of various types which may be produced by the use of the highly improved lines now at hand.
2. The exceptional genotypes identified are virtually unselected as regards characters other than yield. Some of these characters are very important in practice, often more important than a considerable increment in yield.

The critical factor determining the practical feasibility of varietal sampling is the frequency in the varieties of genotypes approximating the yield level of the present elite strains. The limiting data available (all for trials in single seasons) indicate rather high variability in yield genotype among plants of open-pollinated varieties, averaging about 9% of the mean yield after removal of the variance due to experimental error. The distribution of yield level in these populations is normal. The data unfortunately do not show where the present elite lines would fall upon these distribution curves. The general experience of corn breeding in the past 20 years is probably a better basis for estimating the frequency of plants in the foundation varieties which approximate the elite yield level. On this basis a fair estimate of this frequency is 1 or 2 per cent.

Despite its relatively low return, the further sampling of the open-pollinated varieties is essential. The greater part of the hybrid corn now grown is the product of various combinations of about a dozen inbred lines. Each of these represents a single gamete genotype, fixed as a homozygous diploid for controlled combination. These, with the additional lines of promise for further breeding, constitute a minute sample of the gamete populations of the foundation varieties. To confine further breeding to the recombinations of this small group of genotypes is to reduce its ultimate possibilities to an extent which cannot be accurately estimated from available evidence but which must be pretty drastic. Moreover, any new line produced from the recombination of the old lines is limited in its practical use, for no line gives good combinations with lines to which it is related.

Now these varietal populations, in which 1 or 2 per cent of the members reach the elite level, are populations of open-pollinated plants. Each plant represents a random combination of two gametes of the varietal gamete population. The yield potential of the plant is the result of dominant factors contributed by the two parental gametes. The frequency of genotypes of unusually high (or low) yield-potential must be much higher in the gamete population than in the population of open-pollinated plants. In a variety in which plants of yield potential equal to the elite lines occur at a rate of about 1 per cent, gametes of correspondingly high average yield potential constitute almost 10% of the gametic population. This group includes the tail of the frequency curve, and the best 1-2% may be genotypes well in advance of the elite level. Gametes constituting 1% of the population represent a level of yield potential occurring among the open-pollinated plants with a frequency of only about 1 in 10,000. Such genotypes may represent a level of efficiency in grain production which has not been closely approached by selections made from the open-pollinated plants.

The term "yield potential" (YP) as here used refers to the capacity of the genotype for contributing to yield in specific hybrid combinations. Detailed definition and illustration of the concept of yield potential must be omitted here for brevity, but it may be briefly described as follows: The yield potential of a homozygous individual, with reference to any homozygous biotype used as a tester, is (for given conditions) the excess in yield of the F₁ or test-cross over the tester biotype. The YP of the gamete genotype of this individual is one-half of this value. When the tester is a hybrid or mixed population, the YP of the tested individual is the excess of the F₁ over a hypothetical yield which would be produced by biotypes representing the gamete population of the tester. This quantity is indeterminate, but since it affects all test cross yields equally its determination is unnecessary. In practice, YP with reference to a hybrid or mixed tester may be determined as accurately as to a homozygous tester, since the number of plants of each testcross required for an adequate yield test is large enough to render negligible any variation due to individual plant variability.

In the absence of direct evidence, it is necessary to make certain assumptions regarding the inheritance of YP. The validity of these assumptions for the present purpose does not require that they be precisely correct in specific instances but rather that they represent correctly the general or average interaction of the factors involved. All assumptions regarding inheritance of YP in this discussion are derivable from two postulates which are in harmony with the evidence now available but which still require direct experimental verification. These postulates are as follows:

1. The YP of an individual is the sum of the YP's of its parental gametes.
2. The mean of the YP's of the gametes produced by an individual is equal to the mean of the YP's of its parental gametes.

In the initial stage of an isolated corn breeding program, the gamete cannot be made the unit of selection, since there is no homogeneous gamete population with which the varying gametic series may be combined for comparative testing. It is therefore necessary to select among the plants produced by the random combination of gametes of all levels. After an initial series of inbreds distinctly superior to the varietal means has been established, it is possible to use these inbreds in further sampling of the varieties, and in this procedure the gamete may be the unit of selection.

Gamete selection in practice would ordinarily involve two steps:

1. The selection, on the basis of outcross yield tests, of individual plants of a variety/inbred population, and
2. A similar test-controlled selection in the first generation self-progeny of the outstanding individuals identified in the first step. This would ordinarily be followed by continued selfing, with visual selection, to fix a line homozygous for the desired agronomic characters as well as yield genotype.

For some purposes continued selfing would be unnecessary; notably for the extraction of plants of value in complex crossing. Complex crossing for the extraction of improved lines has been little used in corn breeding, chiefly because of the limited number of good lines available. But homozygosis is not essential in the strains used in complex crossing, and the heterozygous strains identified in the plant selection and gamete selection tests may be used without sacrifice of the established inbreds.

The technic may be illustrated by an experiment now in progress. The variety used is Midland, which has given exceptionally good yields among open-pollinated varieties in central and southern Missouri and in other localities in the southern Corn Belt. The inbred used is WF9, which is outstanding in performance among lines now available in the Corn Belt, though it is a little too early to make full use of the growing season in Missouri. It is one of the parents of U. S. 13(WF9/38-11 × L317/Hy) the hybrid now most widely grown in Missouri.

Each Midland/WF9 plant is selfed and is outcrossed on a tester stock, in this case L 317/Hy. Each outcross tests the yield potential of one Midland gamete added to that contributed by the uniform gametes of WF9. Similar outcross tests on L317/Hy are made for comparison from the line WF9, and from F₁'s of WF9 with various inbreds of outstanding performance in this region.

Any Midland/WF9 plant which excels the performance of WF9 in outcross yield tests under varying and representative conditions represents a Midland gamete superior in yield potential to WF9, in a combination in which WF9 is very effective. The selfed progeny of such a plant provides a population in which further improvement by test-controlled selection should be possible. This selfed progeny is comparable to the F₂ of a cross of WF9 with an unrelated elite line. As compared to such F₂'s it has, in addition to its possible advantage in yield genotype, the merit of avoiding interbreeding of the tested lines. A derivative of WF9 × L317 cannot be used effectively with either WF9 or L317; a derivative of WF9 × Midland can be used with any other line except WF9.

In comparison with selfs of plants selected from the pure variety, the variety/inbred selfs have certain distinct advantages and disadvantages. For brevity the former will be referred to as the plant-selection series and the latter as the gamete-selection series.

The chief advantage of the gamete-selection series is the expected superiority in yield potential of the best individuals in the population, or in the limited sample of the population which may be effectively tested for yield-genotype. It has in addition the following noteworthy advantages:

(1) A probably greater range of segregation for yield potential in the selfed progeny of the selected individual. This segregation is the basis for any further improvement in yield which may be made by a second application of test-controlled selection in the selfed progeny of the selected plant. The extent of this segregation is dependent upon the difference in the specific yield-controlling genes contributed by the parental gametes. The yield potential of the selected plant would benefit as much, on the average, from five such genes, each contributed by both parents, as from ten, each contributed by only one of the parents. But the possibility of further improvement in yield potential would come only from the latter.

It would be expected that a self of an outstanding Midland plant, representing a combination of one superior Midland gamete with another, would be heterozygous for fewer yield factors than a self of a Midland/WF9 plant of equal yield potential, representing a combination of a superior Midland gamete with a superior gamete type of unrelated origin. The evidence available is very limited, but indicates that this difference is an important one.

(2) A better opportunity for extracting a line satisfactory in characters other than yield. In a series of Midland selfs, the only selection for such characters previous to yield testing would be that made among the individual foundation plants. It may be expected that the plants of highest yield potential might in many cases be unsatisfactory in other respects. The series of Midland/WF9 selfs is also virtually unselected, but since each plant is heterozygous for the favorable agronomic characters of WF9 it should be possible, in the extraction of homozygous lines from the selfed progeny, to avoid undesirable characters which are not common to the Midland selection and to WF9. This advantage will vary with the line used, but in major characters such as strength of stalk, for example, any elite line selected for use in this type of experiment would provide some insurance against the weaknesses likely to be met within unselected genotype of the open-pollinated varieties.

The chief disadvantages of the gamete-selection series are the following:

(1) In gamete selection it is impossible to fix the genotype selected from the variety; it can be used only to extract a combination of this genotype with some other genotype chosen in advance, (such as the WF9 genotype in the present example). The line ultimately derived from this combination is restricted to use in crosses not involving WF9. In plant selection a new line is derived which may be combined with other lines without restriction, and which may be crossed for further improvement with lines chosen after the properties of the selected Midland line are known.

(2) In yield testing to compare the value of the Midland gametes, the gametic genotypes compared represent only half of the genotype of the plants which are tested; in plant selection the genotypes compared are the total genotypes of the plants tested. A more accurate yield test is therefore required to detect significant differences in the gamete-selection series. The accuracy of yield tests is limited, and this imposes a minimum limit to the difference in yield potential which may be used in breeding. Furthermore, increased accuracy is expensive, and reduction of the standard error to one-half requires yield tests about 4 times as extensive. If differences only half as large are to be detected, only about one fourth as many items could be tested with equivalent outlay.

The gamete-selection series would involve smaller differences than the plant-selection series, but the differences to be expected are considerably more than half as large. The net variability of the outcross test yields, after removal of the superimposed variability due to experimental error, is the measure of the yield potential of the plants tested. The yield potentials of a series of open-pollinated Midland plants are the sum of the yield potentials of the male and female gametes combined. These may be represented as follows:

In wholly unselected series, A and B are equal and the yield potential of the open-pollinated plants is 2A � �2 σ_A

The yield potentials of the F₁ plants of WF9 × Midland would be as follows:

The number of tests of adequate precision that could be made with a given outlay would be about half as great for the gamete-selection series as for the plant-selection series. In view of the increased frequency of exceptional genotypes in the gamete selection series, the smaller sample would have a much higher probability of including exceptional Midland genotypes than the larger.

During the past season direct evidence on some of these points was secured in a yield test, conducted in collaboration with D. C. Anderson, at Malta Bend, Mo. The items tested included outcross tests (on L317/Hy) of the following:

1. 41 Midland plants
2. 37 Midland/WF9 plants
3. the line WF9, (entered for increased precision as 4 items)
4. 6 other elite lines (38-11, R136, 940, C.I.7, Kys, and K4)
5. 10 F₁'s of elite lines, included to check the additive inheritance of YP.

Groups (1) and (2) each included 27 plants representing a wholly unselected sample, with additional plants from visual selection which proved unrelated to yield. These two groups thus represent respectively the zygote and the gamete population of the Midland stock used. The test was planted as a 10 × 10 triple lattice, with 12 replications.

Calculation of the data is not yet completed but the results in general are evident from direct calculation as a randomized block experiment. On this basis the least significant difference is 4.5 bu. per acre. The test-cross yields of the Midland plants varied from 60.3 to 77.8. Those of the 7 elite lines ranged from 61.8 to 77.0, that of WF9 being 64.1 bu. per acre. The test-cross yields of the F₁'s and parent inbred lines were in general in good agreement with expectation on the additive basis, though the differences between the lines crossed are not large enough to make this a very significant test of YP inheritance. The test-cross yields of the Midland/WF9 plants indicated yield levels for homozygotes of the Midland gamete genotypes ranging from 46.8 to 83.8 bu. per acre.

Seed was produced in 1944 for a further trial of plant and gamete selection in the varieties, Kansas Sunflower, Clarage, and Midland, with certain modifications of method. It may be desirable in practice to apply gamete selection not to the unselected gamete population but to a selected population secured from the exceptional plants identified by a preliminary test-controlled plant selection. To test the feasibility of this modification, the unselected plants in the varieties mentioned are selfed and test-crossed as before and are also crossed on the inbred line selected for use in gamete selection. The gamete selection series from unselected plants may be made up from these crosses, and that from selected plants or mixtures may be made up from them after the plant selection tests have been made. Each variety thus yields three distribution curves, representing the unselected plant population, the unselected gamete population and the selected gamete population. Among the inbred lines included for comparison are K4, a line of excellent performance which was extracted from Kansas Sunflower; K201C, an excellent line extracted from Midland; and 3 Ohio lines which represent the best extractions previously made from Clarage. The position of these lines on the plant and gamete distribution curves of their parent varieties should provide a more definite basis for estimating the possibilities of plant and gamete selection as compared with the methods used in producing our present inbreds.

L. J. Stadler