One of the situations that lead to polymorphism is, then, based upon the consideration that the advantageous effects of a gene become dominant and the disadvantageous ones become recessive. In a recent account of this aspect of the subject (Ford, 1964), I assumed that concept without discussion. Yet is seems important in the present context briefly to mention certain aspects of it, for the evolution of dominance and, indeed, of the effects of the major genes in general, is fundamental to the theory of polymorphism and to evolutionary genetics as a whole. For I believe that grave misconceptions are once more arising in regard to it.
These reiterate in a new form the old idea, long disproved, that genes are themselves potentially dominant or recessive, the two types differing in their structure. That fallacy is being revived by molecular biologists; indeed I recently heard it stated unhesitatingly by one of the most distinguished of them.  The reasons for this error are, perhaps, not far to seek. Those who have researched one genic structure, and obtained outstanding results thereby, are not unnaturally inclined to interpret all genetic phenomena in terms of it. Moreover, having had no cause to follow the progress of general genetics, they are probably ignorant of the facts which demonstrate their concept of dominance to be erroneous. Thus it seems necessary to draw attention to certain features of that phenomenon here. A detailed account of the evolution of dominance, and of the mechanisms responsible for it, is, however, outside the scope of this work, but it can be obtained if required from Sheppard (1958, pp. 129-145). It may incidentally be remarked that Crosby (1963) has recently called in question the whole concept of the evolution of dominance and recessiveness through selection operating upon the effects of genes. His criticisms are not new: they have often been discussed before. Indeed his definition of dominance, as a property of genes not of characters, is itself erroneous. [CybeRose note: Crosby even claimed that his computer simulation was more "realistic" than actual experiments.]
It has already been pointed out that dominance and recessiveness are properties not of genes but of characters. For a single mutant often has multiple effects some of which are recessive and others not, being dominant or else intermediate in the heterozygote. Moreover, we are not here dealing only with super-genes consisting of closely linked units having complementary effects, but with mutational sites within the same cistron. These have mutated on several occasions giving rise to the same series of characters, as with ebony in Drosophila melanogaster. Thus dominance and recessiveness cannot be conditions to which different types of genes give rise.
We are indeed only beginning to realize how frequently genes controlling completely recessive characters also affect viability when heterozygous: indicating, that is to say, that they have important physiological effects which are not recessive (see, for instance, Oshima, 1962). That situation was generally obscured in the earlier work because so few studies on differential survival had been carried out and, indeed, special techniques are required for that purpose. This is due to the fact that the achievements of Man in breeding plants and animals are far superior to those of nature.  The product of one pair of organisms is normally one breeding pair only or its equivalent; a result which, total failure apart, is greatly surpassed in artificial conditions.
Since, then, much less elimination generally takes place in the laboratory than in the wild, experimental breeding is not usually well fitted to detect the differential destruction of less viable types except when their handicap is rather a heavy one. Special methods are, however, available for doing so. Thus Dobzhansky (1951), working on Drosophila pseudoobscura and D. persimilis, has made great advances in studying the survival-value of genes, using population-cages in which the numbers can be maintained for many generations at a limit imposed by the food-supply. Somewhat similar methods are available for increasing the effects of selection in other organisms; for example, by means of semi-starvation. Perhaps the first time that the latter technique was applied purposely, with this end in view, it disclosed differential viability favouring an industrial melanic in the moth Boarmia repandata (Ford, 1940b). That situation, which leads to transient polymorphism, is remarkable since the normal form, which is pale and cryptic, proved to be at a physiological disadvantage compared with the recently established and dominant black one owing, apparently, to the superiority of the heterozygote. A slight, but non-significant, excess of the latter had become apparent in segregating families bred by ordinary means. That difference was, however, accentuated and became significant in broods that were partially starved.
R. A. Fisher (1928) was the first to suggest that positive and negative selection for the heterozygous manifestation of a gene can give rise respectively to the evolution of its dominance and recessiveness. That effect can be reproduced artificially. Thus Ford (1940b) so altered the expression of the yellow (compared with the normal white) form of the moth Abraxas grossulariata that it became approximately dominant in one line and recessive in another.  This result was reached in three generations, of which there is one a year. The normal condition, with an intermediate heterozygote, was recovered on out-crossing to ordinary wild-type specimens, so establishing that the observed effect was not due to the selection of distinct alleles but to the differing responses of the organism to the controlling gene. This was the first occasion on which the evolution both of dominance and recessiveness had been demonstrated in wild material, though Fisher (1935) had already proved that dominance-modification had occurred in poultry as a result of domestication. Fisher and Hold (1944), working with a gene reducing tail-length in mice, subsequently obtained an almost exactly comparable result to that recorded in Abraxas, and in the same number of generations.
It is unfortunate that tests of this kind have not been carried out more frequently, though it is important to notice that they seem to have been successful when attempted. The fact is that relatively few genetic studies have aimed at adjusting the effects of major genes by means of selection until the recent and outstanding work of Clarke and Sheppard on mimetic butterflies.
A considerable number of experiments in addition to that on Abraxas, have now demonstrated that dominance and recessiveness are the result of evolution in wild populations. Three examples of these may be referred to here.
(i) Ford (1955a) studied a dimorphism in the moth Triphaena comes involving the typical and the dark curtisii forms. It was possible to show that the latter had become nearly dominant due to selection operating on the gene-complex. That is to say, the heterozygote proved intermediate when the gene concerned was placed in a genetic setting derived equally from the very isolated localities (obtained by crossing individuals from Barra, in the Outer Hebrides, and from Orkney): a result which indicates that different modifiers having the same total effect had been utilized by this species in building up the dominance of curtisii in different parts of its range.
(ii) The Selidosemid moth Biston betularia, a European (and British) species, is normally whitish speckled with black. Its carbonaria form (p. 63)  is uniformly black except for a white dot at the base of each fore-wing and two others where the antennae join the head. This is a complete dominant to the ordinary betularia. By crossing the British species with the allied B. cognataria from the north-east of the U.S.A. and the adjoining parts of Canada, Kettlewell (1965) was able to examine the effect of the carbonaria gene in a genetic setting to which it had not been adjusted, and in these circumstances its dominance completely broke down.
(iii) Clarke and Sheppard have carried out remarkably thorough genetic experiments on the various races of the butterfly Papilio dardanus which is widespread in Africa south of the Sahara. In the course of their work they obtained living material from many parts of the Ethiopian region, including even Abyssinia and Madagascar, and bred the insect in Britain for a number of generations. Among the important results they obtained, some throw light upon the evolution of dominance (Clarke and Sheppard, 1960a).
Papilio dardanus is a large insect (9 cm to 11 cm across the expanded wings). The males are sulphur-coloured with some marginal and sub-marginal black markings except that the hind-wings, which are always tailed, are clouded with brownish shades on the underside in such a way as to make the insect cryptic when at rest. The females are in most areas utterly unlike them, being polymorphic and generally mimetic. Their various phases are determined by autosomal super-genes (the H locus), sex-controlled in expression, responsible for the colour-pattern and giving the simulacrum of multiple alleles. They copy tail-less models belonging to the families Danaidae and Acraeidae and, accordingly, they are themselves tail-less (except in Abyssinia) due to the operation of a pair of alleles (TNTN), also sex-controlled by assorting independently with H. The heterozygous females (TTTN) are intermediate between the tailed (TTTT) and the tail-less conditions.
In Madagascar, however, the females are monomorphic and much resemble the males. The hind-wings of both sexes bear tails which have a mean length of 14.4 mm in the males and 14.8 mm in the females.  When crossed with specimens from South Africa, the females are equipped with short tails of variable length, the mean being 4.4 mm. They are further reduced in back-crosses to the South African race which, therefore, must carry modifiers that shorten the tails of the heterozygotes.
The situation is different in Abyssinia where, unlike that in South Africa, 80 per cent of the females are male-like while 20 per cent are entirely distinct from them, being polymorphic and mimetic. In this region, alone on the African mainland, the females always have tails which, in the non-mimetic forms, are shorter than those of the males (the mean lengths being, 14.2 mm for the males and 12.5 mm for the females) and shorter still in those that are mimetic (9 mm). When crossed with the South African race the female tails are more reduced than in the Madagascan hybrids, having a mean length of only 1.9 mm, while about half the specimens are tail-less. In these, therefore, the gene for tail-production has actually become fully recessive in effect.
Another aspect of this work may be mentioned here. Dominance in the various (female) polymorphic forms of P. dardanus in South and Central Africa takes an order down to hippocoon, which constitutes the bottom recessive, and is generally complete when sympatric crosses are made. It fails to be so in 7 out of 21 heterozygous types tested. Yet these exceptions if anything strengthen the argument for the evolution of dominance in such circumstances. For, as Clarke and Sheppard point out, either such forms are very rare and resemble only imperfectly some mimetic phase, or else selection for mimicry is relaxed on account of the scarcity of the model in the region in question. In the crosses between phases belonging to the main African races and males form Abyssinia and Madagascar, dominance broke down, though sympatrically complete, in 12 out of the 14 forms tested.
These results clearly indicate that dominance had been evolved. It is, in general, fully developed when the controlling gene is in its normal genetic setting but not when, in allopatric crosses, it is placed in a gene-complex to which it is not adjusted.
This can operate upon all but the more lethal mutants, for these latter do not contribute materially to posterity. However, its action is far more rapid in polymorphism than in other situations, since even the rarest phase is then not uncommon and is maintained more or less permanently owing both to the balance of advantage and disadvantage involved and to the very process now under consideration; that by which even unifactorial characters can be increasingly well adjusted to the needs of the organism. Accordingly a few instances of that aspect of evolution must be mentioned at this point. Butterfly mimicry, especially the well analysed examples found in Papilio dardanus, may be further used to provide the first of them.
Polymorphic Batesian mimicry posed a difficult problem to the early students of evolutionary genetics. In the first place, the various adaptations of the mimics are often numerous, complex and accurate; moreover, they may involve extremely diverse features: such as colour, pattern, wing-shape and habits. On the other hand, the distinct mimetic and non-mimetic phases are controlled by a pair of alleles, the distinction between which must have arisen suddenly by mutation (and this is equally true of each unit within a super-gene). Was it credible that these delicate adjustments had to await the chance occurrence of a mutant possessing all the qualities needed to produce them? Such an event would seem hardly possible even on a single occasion while, in fact, the phenomenon is widespread.
Punnett (1915) recognizing this inherent difficulty attempted to solve it by suggesting that the same gene on each occasion mutated in model and mimic, giving rise to the same set of characters in them both. This was an impossible suggestion, for a universal feature of Batesian mimicry is the complete superficiality. The mimics resemble their models visibly but in no other way: their similar colours are generally produced by chemically different pigments and corresponding effects of pattern are gained by entirely distinct means;  thus resemblance to the abdomen of a wasp may be achieved by the appearance of the wings when folded over the body. Moreover, only those parts exposed to the view of a predator are affected: the original non-mimetic colouring of a mimic is often preserved on the costa of the hind-wings where this is hidden by the overlap of the front pair.
The difficulties inherent in polymorphic mimicry when complex adaptations are determined by a single switch-control were resolved by a suggestion due to R. A. Fisher (1927). He pointed out that if a single mutant chanced to give an unprotected species some slight resemblance to a protected and warningly coloured one, then the similarities could be improved by selection operating on genetic variability in the expression of the gene responsible for it. Thus although the latter arose suddenly by mutation, we are not to suppose that its effects, in all their perfection, did so too; on the contrary, these could be evolved gradually.
This concept has been tested and fully verified by Clarke and Sheppard (see especially 1960b); work in which they principally used Papilio dardanus. They had in the first place established the genetics of its polymorphic forms. As already mentioned, these are restricted to the females, in which they segregate sharply. They may differ from one another as widely as very distinct species which, in fact, they were at first thought to be. Five of them (hippocoon, cenea, trophonius, niobe, planemoides), controlled on a unifactorial basis, are beautiful mimics, adjusted even to resemble the local races of their models if these undergo geographical variation in different parts of Africa, as does Amauris niavius (Danaidae), copied by hippocoon (with its geographical adjustment hippocoonides), on the western and eastern sides of the continent.
Clarke and Sheppard (1960b) crossed a number of well adapted mimics with males from regions where the phases in question are absent. An example of their results is provided by the cenea form. This is a simple dominant to hippocoon. It copies Amauris echeria and is common in South Africa, comprising 85 per cent of the P. dardanus females, but is absent (owing to the absence of its model) on the west coast, from  Angola northwards. When specimens from these two regions were interbred, the accurate adaptations of cenea broke down in the F1 hybrids, showing that they had been perfected by selection operating on the gene-complex and were not an essential property of the controlling gene itself. Moreover, a less complete disintegration of the cenea pattern occurred when the gene responsible for it was brought into a cross with specimens from Tanganyika, where that mimic is rare but not completely absent.
A remarkable contrast was obtained with trophonius, which is dominant both to cenea and hippocoon and copies Danaus chrysippus. South African females of this form were crossed with males from the two races used in the work just mentioned, those from the west coast and from Tanganyika. No breakdown in the mimicry occurred; the F1 trophonius were perfectly normal for, unlike cenea, that phase occurs in all three areas. Consequently, the trophonius gene when segregating in the hybrids was not placed in a genetic setting half of which was derived from a race unaccustomed to it.
It might possibly be thought that the adaptations of trophonius were not, therefore, the result of selection acting within the ambit of its switch-control. Further work by Clarke and Sheppard (1960b), however, clearly demonstrated the contrary. They were able to bring several of the genes for mimetic forms of P. dardanus into hybrids obtained by crossing the South African race, in which these occur, with the non-mimetic one from Madagascar. The accurate adaptations of these mimics, among which was trophonius, were destroyed in the F1 individuals, producing a confused pattern unlike their normal appearance which matches that of their respective models.
The selective adjustment of the effects of a single gene seems first to have been demonstrated by Castle and Pincus (1928) in 'hooded' rats; a piebald variety, white with black dorsal markings, which is unifactorial and recessive. It is somewhat variable in expression and by breeding from the more extreme individuals, Castle and Pincus found it possible to produce strains that were entirely black, entirely white or intermediate in various degrees.  Even so early in the analysis of Mendelian phenomena, they saw that their results could be explained in either of two ways and designed an adequate experiment to decide between them. On the one hand, their choice of individuals might have favoured different multiple alleles of the controlling gene. Alternatively, this might itself have remained unaltered while the response of the organism to it changed owing to the recombination and selection of other pattern-modifiers within the gene-complex.
This latter possibility proved correct, for Castle and Pincus crossed wild rats with individuals from their selected lines and extracted the recessives in F2. These showed a considerable return to normality, which they would not have done had the gene responsible for the hooded type itself changed through selection of a different multiple allele. From this time forward, therefore, the selective modification of the effects of a major gene became a possibility which had to be reckoned with in evolutionary studies.
The same principle was demonstrated by the later and celebrated work of Morgan (1929) on the 'eyeless' condition in Drosophila melanogaster, which is also unifactorial and recessive. He found that natural selection operating in the laboratory cultures favoured those flies in which the highly disadvantageous effects of that mutant were expressed to the least extreme degree as a result of genetic variation: for its expression ranges from individuals with a somewhat reduced number of ommatidia, and therefore defective vision, down to those that are totally blind. Thus, after some generations, the homozygous stocks acquired almost fully developed eyes. The abnormality was, however, recovered on crossing such potentially eyeless, but apparently normal, flies with wild-type specimens and extracting the recessives; showing, once again, that the response of the organism to the controlling gene had changed, but not the gene itself.
Similar effects to those obtained experimentally with hooded rats and the eyeless mutant of Drosophila melanogaster have been detected in natural conditions. Several situations allow evolution to take place fast enough to be studied experimentally (Ford, 1964, p. 8).  These include polymorphism in which, necessarily, all the phases are relatively common; it therefore provides opportunities to observe adjustments in the effects of individual genes even in wild populations. Instances of this kind are provided by the industrial melanism of the Lepidoptera. This has affected over 80 species in Britain, and others in Continental Europe and the U.S.A. A general account of that occurrence may be obtained from Kettlewell (1961), who has analysed it with great success. It constitutes the most striking evolutionary change ever witnessed, though the spread of D.D.T. resistance (Brown, 1958), and that described on pp. 67-8 may be as remarkable.
The species principally used by Kettlewell in his work is the moth Biston betularia (Selidosemidae). It is a white insect marked with fine black dots and lines in such a way as to resemble a patch of lichen with great exactitude. An effective protection is normally obtained by this means, as the imago rests all day with outspread wings fully exposed on tree trunks.
However, beginning in manufacturing districts in the middle of last century a black form, carbonaria, is heavily eliminated by bird predation compared with the typical lichen-like specimens, while in industrial areas (the Birmingham neighborhood) where the trees are polluted and lichens are absent, the reverse is true.
Carbonaria first appeared, in Manchester, in 1848 and, though unifactorial, it has evolved (including the acquisition of heterozygous advantage, p. 63). The earliest specimens were marked with a scatter of white scales and often with white lines. Those found today, however, are completely black, except for a white dot at the base of the wings and another pair where the antennae meet the head, and even these are now disappearing in the Sheffield area and some other districts.  Kettlewell (1965) has been able to recover the condition in which carbonaria is not completely black by placing its controlling gene in a gene-complex that has not been adjusted to it. This he did by crossing for several generations the wholly black specimens, as seen today, with pale typical ones from Cornwall, where carbonaria is unknown. The appearance so produced is not, however, identical with that of the original melanics found in the middle of last century, since the scatter of white scales upon them is somewhat differently disposed. This may be due to the fact that selection for another carbonaria allele, of which several have been identified, has taken place in addition to the proved adjustment of modifiers which affect its expression.