It is surprising how little it has been generally realized what excellent objects mutations afford for the controlled study -- by contrast with them -- of the functions of members and features normal to the organism. The gene change provides us with a subtle inner operation which is often capable of excising, remolding or changing the properties of an organ, tissue, cellular component or substance, far more deftly than this could he done by the surgeon's or experimental physiologist's knife or other relatively crude procedure, or even by the pharmacologist's or biochemist's medication.

    It has, to be sure, been evident for some years to some geneticists with physiological and embryological leanings what rich opportunities visible and other mutations offer for the investigation of the ontogenetic, the physiological, and ultimately the biochemical processes whereby the observed characters are brought into being, and a few exemplary researches have been made along such lines, in a number of different organisms.

     But it has not been so well appreciated how valuable this material also is for investigations of the functions of the given characters, in the individual's living and perpetuation (although of course these two lines of study are closely connected and in fact not ultimately separable). For this reason my examples of conclusions that may be drawn in this way in Drosophila are derived mainly from the results of casual inspection of easily visible mutants, and seldom upon researches carefully directed to these ends. Nevertheless, even the incidental observations have already been sufficient to show the distinct trend of the data, with reference to the chief question here at issue.

    Let us first recall, by a scrutiny of our illustrations of normal male and female Drosophila (Fig. 1), how very complicated even this external anatomy is, how many details it has which possess no value obvious to us, but how nearly constant, on the whole, most of these features are from individual to individual, so that every bristle, every row of hairs, and every veinlet of the wings, has been given its name or number.

    Yet, recall also that by careful search, mutations changing or abolishing one or another of each of these characters to a greater or lesser degree, and either singly or in some combination, may be and have been found, and that there are probably still more frequent mutations which alter them so slightly as not to have been noticed. Undoubtedly the same story could be repeated in every higher organism. Hitherto, however, only the mutations with more conspicuous effects have been much worked with, on account of the greater ease with which they are recognized, and so our examples will be confined to these for the present.

   If we now turn our attention to the fly's most conspicuous appendage, the wings, examining the wing mutations from their functional aspect, we find, as we should expect, that those mutants which do not fold their wings back completely while these are at rest are apt to get them caught and bedraggled. That is, the normal position of the wings is more advantageous. So is the normal shape and size, as shown by the fact that those with wings abnormal in shape or unusually small are visibly hampered in flying. We may by the way surmise that, conversely, oversize wings also are less efficient, in relation to the weight of the fly and to the energetics of the muscular mechanism for beating them. The many-millions-of-years-old pattern of the wing veins has been by no means fortuitously fixed upon either. Thus, if even a little cross-vein is absent, or a portion of a long vein, the wing is unduly flimsy, and likely to become torn. Moreover, branched veins having blind ends to their branches are likely to leak fluid into the wing at these points, resulting in blisters or even ballooning.

    The bristles and hairs afford another set of external features in which conspicuous mutations occur. They cannot be there for warmth in this tiny object, or for their pleasing appearance. The presence of any particular one in its regularly assigned place might seem to be a matter of no importance. Yet an inspection of cultures of given mutants, in which particular bristles are absent or reduced in size, shows that such a seemingly trivial deficiency is apt to cause trouble. For such a tiny, weak creature as an insect the capillarity of water and the adhesiveness of the moist soft materials with which it must often come into contact constitute an extremely powerful force and a serious menace. The waxed hairs and bristles tend to keep this danger at a distance and flies lacking them are much oftener found stuck in the food or on any moist surface. Even where one small group of bristles is absent, as in Fig. 2b showing the so-called "scute" flies, which chiefly lack the bristles on the small scale or "scutellum" that forms the posterior end of the thorax, we may frequently see that a bit of the viscous culture medium has become plastered upon them in just that region.

   The same disasters are likely to meet flies which have bristles or hairs that are unusually small, or that are split and curled, like the so-called "forked" bristle flies of Fig. 3, or that are otherwise misshapen. Similarly, it has been shown, mutants with those chemical changes in their cuticle that go with the color change to "yellow", and to certain other types of pigmentation, are less resistant to desiccation (Kalmus), as well as more subject to getting glued up.

  An additional danger, for all flies defectively equipped with bristles or hairs, is that on them their external parasites, the mites, have a much easier job in becoming attached, as they do, to the epidermis, so that one commonly sees these flies much more heavily infested.

A last example of bristle function may be drawn from the minute pair of hooked bristles at the end of each foot. For the mutants in which these are underdeveloped (such as those which, by a second effect of the same gene, have "spectacled" eyes) are, unlike the normals, incapable of getting a good grip on vertical or inverted surfaces and walking readily upon them.

    It is widely known that the normal brick-red eye color in Drosophila is produced by the cooperation of dozens of genes and that most of their mutations result in the pigmentation being lighter. Can it be by mere chance that this complicated outfit for keeping the color dark continues to be maintained? Although even the white eye is sensitive to light, and can distinguish the general direction of its source, tests by Kalmus and others have shown that light colored eyes do not enable the flies to orient as well to moving patterns (stripes). This is presumably because in them the light is more indiscriminately diffused over the photosensitive retinal cells, rather than in the form of the sharp, erect images produced by the light that passes through the well-pigmented light-insulated optic tubes of the normal compound eye.

    Now, the same genes which give rise to this optic pigmentation produce pigmentation also in some internal organs, notably the excretory tubules and the sheath of the testis. And tests of Mackenzie and this writer have shown that flies without this screen over their gonads, thin though it may be, are considerably more sensitive than the normal to the damaging effect of ultraviolet light (of wave lengths present to some extent in outdoor sunlight) in producing mutations in their germ cells. The normal epidermal pigment likewise helps in this respect.

    The above observations, superficial though most of them have been, will suffice for our present purposes. Many other externally visible features of little understood function, such as the three small ocelli, the parts of the antennae, the individual joints of the legs, projecting mouth parts, secondary sexual characters, and characteristics of the behavior, could be studied in a similar way, and would in fact justify much more intensive studies. At the same time, the interior characteristics of the body, including especially those pertaining to the cells themselves, and to their metabolism, remain as yet a largely unread library.

    While few other types of visible mutations than those above discussed have been studied enough to show their modus operandi in affecting the life of the organism, one general fact concerning them deserves emphasis here. This is that practically all of them do result in some quantitatively demonstrable reduction in the expectation of life, even under the relatively sheltered conditions of laboratory culturing. In many of these cases, to be sure, it is evident that there must be some hidden damage, besides that stemming from the visible change itself, inasmuch as even the larva, in which the given external feature has not yet had a chance to develop, is less viable.

    This illustrates the multiple effects which most individual genes have. We have however intentionally chosen, for our above examples, cases in which the consequences of those changes that occurred in the visible characters themselves could be traced. For only thereby was it possible to show that all these characters too, non-essential though they might seem, did have their own utilitarian roles to play in the great system of processes that serves biological perpetuation.

    At the same time it should be remembered that all so-called visible mutations together, including those which have hidden harmful effects accompanying them, form only a small minority -- at least as low as a twentieth -- of all the mutations that can be demonstrated to occur in Drosophila, and that the vast majority of all these changes is somehow detrimental, just as one would expect for fortuitous alterations in processes that themselves play some serviceable role in the organic system.

    That this is true on the biochemical level is clearly evident from the work of Beadle and his co-workers on moulds, and this work also demonstrates the value of such mutations for the elucidation of the biochemical functions of the genes there studied, just as in the Drosophila cases above cited the functions of the morphological characters are brought to light.

    In passing we may note, for the benefit of possible objectors here, that there is no real contradiction between mutations having provided the building blocks of evolution, on the one hand, and the fact of such a large majority of them being detrimental, on the other hand. For the very fact of the organism's having been fashioned largely through a step-by-step natural selection of those rare mutations which did happen to fit, carried on over many millions of years, would necessarily have resulted in its being so organized now that a random alteration in almost any feature would be far more likely to be harmful than helpful to it.

    This could still leave open, however, various devices and special routes for its further improvement, especially those which fitted it to hitherto unusual conditions, and it is evident that upsets of the organic balance in any such directions would in turn be likely to open up still further possibilities. Thus only the very rare mutation, preferably too under rare conditions or in rare combinations with other mutations, can be expected to be useful in evolution, if in fact organisms have evolved through an accumulation of useful mutations. And, as we have seen, the data on the usefulness of existing characteristics quite fit in with this interpretation.

    It might however still be maintained that, although the majority of organic features do have their useful functions, evidence of the above kind indicates this only in a rough way, and that we are still at liberty to hold that alterations in somewhat less conspicuous characters, or finer grades of change than those studied in the above characters, would be of indifference so far as the organism's success in living was concerned. It is thus evident that more exact methods, or a quite new mode of attack, would be desirable for throwing light on the question of the adaptive value of character differences of a more minute order of magnitude. Fortunately, evidence of an unexpected nature has in fact turned up, in Drosophila genetics of a much more technical kind, concerning the adaptiveness of differences of an astonishing degree of refinement, even in cases where the reason for the functional importance of these differences is not known.

This evidence arose out of studies of the observable effect on the organism produced by changing the quantity of given genes. For this purpose, comparisons were made of individuals having the normal number, or "dose", of these genes, with those which were known to have inherited more or fewer than the normal number, or dose. Such differences in dose were made possible by first irradiating the germ cells of individuals having the desired genes and then, in appropriate crosses, picking out and breeding -- for extra doses -- those offspring which happened to have received a small extra fragment of chromosome, broken in just the right way to contain that gene, or -- for subnormal doses – those offspring which received all of the chromosome in question except for a corresponding fragment which in this case had been broken out of it and lost.

    An illustration of the most typical sort of situation is given in Fig. 4, which shows the results with the sex-linked mutant gene for apricot eye color. This is a rather light color produced by a mutation of the same normal gene which, by a more extreme mutation, sometimes gives rise instead to the well known white eye. Let us confine our attention at first, for simplicity's sake, to the results in females.

    The chromosome constitution of a series of these is represented on the upper row of cytological diagrams in this figure. The diagram in the second position from the left end of this row indicates the constitution of the ordinary apricot female, having two X-chromosomes each containing one apricot gene, and therefore having two doses of apricot in all.

     As the picture of an eye just above this on the top row shows, the resulting hue is rather like that of the fruit after which it is named. The female with three doses, to the right of it, having two apricot-containing X's with the addition of a small fragment that also includes one of these genes, has a distinctly darker color, though by no means as dark as the normal red. On the other hand the female with only one dose, shown at the left end, having apricot in one of her X-chromosomes but a small section which could have contained it missing from her other X-chromosome, has a color only about half as dark as the regular apricot.

    The above comparison shows, for one thing, that the amount of pigment varies with the dosage, and in a nearly proportionate manner in these cases. In this connection it may be mentioned that, as a more refined check on the conclusion that it was really the change in dosage of the gene for apricot itself which was responsible for the color differences, and not (which seemed but a remote possibility) that of some other genes which happened to be in the same small chromosome fragment, the effect of adding a slightly smaller fragment, taken from the same region but not containing the locus of apricot, was tested out. This effect was found, as expected, to be zero.

    A second conclusion to be drawn from the results is that, since the observed color varies directly rather than inversely with the dosage, the gene which we call "the gene for apricot" must be in some way engaged in furthering rather than in interfering with the production of eye pigment. However, it is not as effective in its work of aiding in pigment production as is the normal gene, since the presence of two or even of three doses of the gene for apricot fails to lead to the intense brick-red coloration found in normal individuals.

     Before such dosage tests had been made this conclusion could not yet legitimately be drawn. For it might as well have been assumed, as an alternative, that the mutant gene called apricot exerts some actual inhibiting action on the process of pigment manufacture, which normal genes simultaneously present in other positions in the chromosomes are working to carry out, and it might even have been postulated, further, that the normal gene from which apricot arose by mutation (its normal "allele", as we say) itself played little or no positive role in the pigmentation process (or even a somewhat negative one). Since however both the normal and the apricot genes alike are shown by the above seriation to play a positive role, with that of the apricot similar to but less potent than that of the normal, we have applied the term hypomorphic here to designate the relationship of the mutant to the normal gene's type of activity.

    It must be understood that this terminology does not imply that the mutant gene's action is qualitatively, or in a biochemical sense, exactly the same as that of the normal allele, only less. In fact, there are several good reasons for inferring that the mutation caused a change in the chemical constitution of the gene and of its product. What is implied is only that the mutant gene works in such a way as to produce a final effect similar to that of the normal, but a lesser effect, like that which would presumably have been brought about if the concentration or the activity of the normal gene itself (or of its products) had been reduced. It is in this sense that the mutant gene is less active.

     Results similar to those above described for apricot have been obtained with a good many other mutant genes in Drosophila, both with those in the X and in other chromosomes. We now have available for conclusions in this field not only the studies on extra doses, which were the first done expressly for this purpose, but also the much more numerous, still earlier found cases of so-called "deficiencies", that had been investigated especially by Bridges and by Mohr. These could not with confidence be used for such conclusions until -- long after their finding -- it was shown by cytological study of salivary gland chromosomes that a small bit of chromosome is in fact really absent, rather than altered in some way, in such cases, and that their use accordingly gives us a means of changing the number of genes present.

    Now, a perusal of all the available cases involving either extra or missing doses of mutant genes, or both, shows clearly that the great majority of them are, like apricot, hypomorphs. That is, they have an action like, but lesser than, that of their normal allele, and varying in amount directly with their dosage. This is true for instance of the scute gene, previously referred to, as shown by observations on bristle number in the presence of one, two and three doses, and it is true of the forked gene, as shown by the frequency and amount of bristle deformation.

    There are some mutant genes, however, like that for white eye, the action of which on the observed character turns out to be so very weak, if present at all, as to be indetectible, as judged by the ineffectiveness of change in their dosage. These latter are accordingly termed "amorphs", and may be regarded as standing at the bottom of the ladder of hypomorphism. On the other hand, there are probably a few, to be termed "hypermorphs", with even greater effectiveness than normal, though these are hard to detect for technical reasons. And there are certainly some, though rare, which we call "neomorphs", with a qualitatively different type of effect than the normal.* [Footnote: Possibly too there is a rare class of mutants, opposite in type of action to their normal alleles, that may be termed "antimorphs". However, our earlier criteria for these were inadequate. For, as C. Stern has shown, the property of competition between mutant and normal genes is separate from that of the direction of their effects. Thus the latter must be judged only from experiments in which the dosage of one allele is varied in the total absence of the other, a procedure which has been insufficiently tried in the cases in question. This competition, it may be pointed out, may be either for substrates, as Stern has postulated, or, conceivably, may involve an interference of some of the gene's products with the total gene activity of the two alleles.] Although these last are of particular importance in their bearing on evolutionary possibilities, the point of greater immediate interest for us in connection with our present theme is that the great majority of mutations turns out on these tests to be hypomorphic or amorphic; that is, they involve some sort of weakenings or losses of biochemical processes normal to the organism, processes the strength of which varies directly with gene dosage.

    These results acquire a greater interest when considered in connection with those for the normal alleles of these same hypomorphic mutants. In the great majority of cases a change either from two down to one or up to three doses of the normal gene results in no detectable change in the character, or at most in a change of trifling magnitude, seldom noticed except under special conditions. Yet we have just seen that the effect of these normal genes must be regarded as like that of a quantitatively greater amount of mutant genes.

    It must therefore be concluded that, with increase in the dose (or with an equivalent increase in the activity) of the genes at these loci, although there is at first, at the lower levels represented by the hypomorphic mutants, a proportionate rise in the effect, as shown in the left hand portion of the curve in Fig. 5, there ensues a progressively lesser rise, that is, the curve levels off in its right hand portion (that representing the grades of effect given by the normal alleles), until it becomes nearly horizontal. In this right hand portion, then, the so-called "saturation effect" is approached. It is not surprising that gene effects, like so many others, should at high doses obey this "law of diminishing returns."

   We may however ask, why is it that the normal genes should in such a majority of cases be at a level of effectiveness which puts them in this "saturation" portion of the curve? The answer to this would at the same time go far towards explaining the observed fact that such a majority of normals appear almost completely dominant to their mutant alleles. For the finding that most mutants are hypomorphs or amorphs would lead to the conclusion that the heterozygous individual, i.e. that having one mutant and one normal allele, has the equivalent of one or somewhat more than one dose of the normal gene, and this, being at or near the saturation level, would have an effect practically equal to two normals; thus the normal would appear dominant.* [Footnote: We are here assuming that the mutant gene's competitive action, mentioned in the preceding footnote, is usually a minor one, as the facts indicate it to be. In the relatively few cases in which it is more pronounced, however, it must cause the level of activity in the heterozygote to be lower, approaching more nearly, or even going appreciably below, the level of one dose of the normal gene.]

    In seeking an interpretation of this high effectiveness of the normal gene we may discount the possibility that the most advantageous level of a character can only be attained by, so to speak, pushing each gene to its utmost activity. There is no a priori reason why characters in general should represent the maximum biochemical possibilities. The body pigmentation, for example, is not the darkest chemically possible, nor is even that of the eye, since the normal types of some Drosophila species, as well as some mutant types of Drosophila melanogaster itself, have darker bodies, and some have darker (though perhaps not redder) eyes. That is, higher levels are possible by readjusting other interacting gene processes, and the present lower level might then be reattained by lowering the effectiveness of the particular gene we are primarily considering. Thus we should be able to obtain a substitute normal type, which would in the main be indistinguishable from our present one.

    We are also very skeptical of the thesis that the high potency of the normal has been attained, as Fisher has suggested, by reason of the advantageousness of having heterozygotes (individuals with one normal and one mutant gene) rendered nearly normal. For, as Wright and the author have separately pointed out, such heterozygotes are so few as hardly to give a selective advantage great enough to withstand the disruptive tendencies of mutation pressure, drift, and selection for more important correlated characters.* [Footnote: Even if Fisher's explanation were adopted, however, it would lead us to the same conclusion concerning the high adaptive value of precisely the "normal' degree of gene expression now existing, as that arrived at on our own interpretation.]

    The clue to an interpretation not subject to these difficulties is (as was pointed out independently by Plunkett and the present writer in 1932) provided by the fact that mutant genes are on the whole much more variable in their expression than normal genes. This is obviously related to the further fact that most of these mutant genes (the hypomorphs and the neomorphs as well) have a level of effectiveness, in regard to the characters studied, such as to place them on the obliquely sloping, left-hand portion of the previously discussed curve.

     For, since they occupy this position, any agency which operates, during the time of their activity in determining the character, in such a way as, in effect, somewhat to diminish or increase this activity, will result in a visible alteration of the character, for it will have an effect 1ike that of changing the dose. Agencies, both of an environmental and of a genetic kind, which exert such influences, are quite evidently very common, as we can see from the character's variability.

    Yet, if we were to judge from the relative invariability of the normal type, these influences are not at all common. The reason for this apparent discrepancy clearly is that the normal gene has such a high level of effectiveness that even when its activity (along the abscissa of our curve) is shifted backwards or forwards by the given influences as much, relatively * (Footnote: i.e. when the variations in activity are expressed as fractions of the mean activity.), as that of the mutant gene was found to be (a process which we have no reason to doubt), there results no discernible alteration of the character (the ordinate) since the curve of effect is here almost horizontal (see Fig. 5).

     But this conclusion in itself furnishes us, ready made, with the explanation of why the normal gene has been given this nearly saturation level of activity. It is because variations in the character exhibited by the normal individual are thereby minimized. All this implies as its basic proposition that such variations are disadvantageous. That is, that the precise grade ordinarily attained by the normal character must, under most circumstances, be optimal, and even small deviations from it must result in an appreciable lowering of the individual's expectation of life or reproductive rate. When we here say "appreciable" we mean, great enough to afford an effective handle for the operation of natural selection.

    We can even get some idea of the amount of character alteration which affects survival enough to be of importance in evolution, on the basis of these considerations. This may be done as follows. Let us first measure the amount of phenotypic (i.e. visible) variation of a hypomorphic character, such as apricot, taking, say, its variance under natural conditions as our index. We may then determine, from our curve of the variation of the character with gene activity or dosage (Fig. 5.), how great a change in activity, measured on the abscissa, this observed change in character, measured on the ordinate, corresponds to. Let us then divide this "activity variance" by the value for the average activity in the case of apricot itself, so as to put the variance in relative terms, constituting a "coefficient of variance of activity" for the apricot gene.

     Next, assume that a like variation in activity of the normal gene, relative to its own value, occurs as a result of the same fluctuating influences. We may then see how much absolute variation in activity this amount of relative variation would correspond to in the case of two doses of the normal gene. After plotting the resulting values for this in their appropriate positions on the abscissa, on either side of the value for the average activity of two doses of the normal gene, we may then read off on the corresponding ordinates how much alteration in the character ("phenotype") of individuals provided with two normal genes would thereby result. It will certainly be found that in the given case this amount of variation is quite imperceptible to our eye. Yet, on this interpretation, a greater amount of variance than this has been prevented by reason of the disadvantage thereby entailed. In other words, this quite imperceptible variance represents, in a sense, the limit of the naturally allowed changes.

    It might, as an alternative, be held that it is not the ordinarily observed variants which are selected against, but rather the larger ones which may appear under more extraordinary conditions (environmental and/or genetic). Just because of the rarity of such conditions we should have to make the selective detriment of these supposedly larger variations correspondingly higher, for them to be so effective in selection. But we should thereby arrive at a value similar to that reckoned above for the average disadvantage of a given amount of change in the character. In any case, then, we would conclude that extremely high stability has been selected for, allowing only a subliminal variance.

    Summing up the foregoing argument once more, we may say that the finding that normal genes are in general so near the saturation level of their effectiveness appears to require the interpretation that there is usually an appreciable disadvantage to the organism if one of its features fails to be developed to exactly that grade characteristic of the type which has been established as normal. In the cases dealt with, in fact, the required exactitude even surpasses our powers of direct visual discrimination.

The above evidence of adaptive precision is, however, surpassed by that arising from a further series of findings of an unexpected nature, concerning the dosage effects of genes in Drosophila. These surprising phenomena came to light when genes contained in the X chromosome were considered and a comparison was made between the dosage effects found in females, which, having two X chromosomes, regularly have two doses of such genes, and those found in males, which regularly have one dose. The relationships found have been denoted as "dosage compensation".

    A typical example of this is shown by the mutant gene for apricot eye, the dosage effects of which in the female have already been considered. The lower row of chromosome diagrams in Fig. 4 shows the genetic composition of the males which have the eye shades shown in corresponding positions along the topmost row. It is at once apparent:

• First, that in the males, as in the females, the color varies about proportionately with the dose, at this level of activity.

• Secondly, however, it is equally clear that the color of the ordinary male, with one dose, is very nearly the same as that of the ordinary female, with two doses. The one-dose female, then, is considerably lighter than the one-dose male, and the two-dose female, correspondingly, is lighter than the two-dose male.

These relations are also shown in the activity-effect curve in Fig. 5, where the bottom line shows the male doses that produce the effects indicated by the ordinate heights shown on the curve above them. Here the dotted lines show the approximate latitude of the error possible in the determinations of color grade. It will be seen that this error is relatively so slight as to justify us in giving the rule as: one dose in the male has very nearly the effectiveness of two in the female.

    This peculiar sex alteration in the effect of any given dose of the gene for apricot must clearly be produced by the interaction, in the pigment-producing process, of other genes than apricot that differentiate the sexes. That genes in the Y-chromosome of the male have nothing to do with this matter is not only an a priori likelihood based on previous evidence of the relatively small influence of the Y in general, but has been directly demonstrated here by observations on apricot males without any Y and on apricot females possessing a Y in addition to their two X's, inasmuch as these individuals are found to have the same color as the regular apricots.

    The influence must therefore be exerted by a gene or genes lying in the X chromosome, at some other locus or loci than that of apricot itself. This conclusion follows directly from a comparison of the genetic compositions of any male and female that have the same dosage of the apricot gene. For example, the genetic difference shown in Fig. 4 between the regular apricot female, with its two doses, and the darker colored two-dose apricot male, that has one apricot gene present in an X of normal size and another in a small extra fragment of an X, lies (if we disregard the Y) purely in the fact that the female contains not merely this fragment of a second X but, in addition, all of the remainder of a normally constituted X chromosome. In this remainder of the X, then, must lie a gene or genes whose presence in double dose, when the rest of the genetic composition is held constant, causes the apricot color to be lighter than that developed when they are present only in single dose -- in fact, just enough lighter so that the two-dose apricot female is brought down to approximately the level of the one-dose apricot male.

      This interacting gene or genes elsewhere in the X thus have the effect of compensating almost exactly for that difference in dosage of the gene in question (apricot) which ordinarily exists between the sexes. It is for this reason that such modifying genes are called "dosage compensators". Their action is such as to keep the two sexes alike in their grade of expression of the given sex-linked character, despite the difference in dosage of the primary gene under investigation.

    Of the many different mutations that have occurred at the locus of apricot, giving different shades of eye color, a considerable majority, including both those of grades darker and of grades lighter than apricot, show the same phenomenon of compensation as apricot itself. A few however (notably eosin and ivory) show no compensation, the color being virtually identical in males and females having the same dosage of the gene, even though in these cases as in the others the color increases with rise in dose within the same sex. We may suppose either that, through some qualitative chemical change, the process of pigment formation mediated by them has become insensitive to the action of the normal compensators, or that they represent more nearly the normal process and that it is the others which, through a chemical alteration, have come to have effects sensitive to the compensators. For reasons to be given below, however, the former interpretation, according to which compensation is a normal phenomenon, appears the only reasonable one.*[Footnote: Direct proof of this has been obtained since the above was written (see footnote on p. 192)]

    That the phenomenon is, in Drosophila, a general one, and not to be explained by the accidents of a particular type of mutation, is shown by a scrutiny of the data derived from other sex-linked mutations. So for example the hypomorph scute, which has been investigated by bristle counts of males and females with different doses of this gene, brought about by the addition and subtraction of small fragments produced by irradiation for this purpose, just as in the case of apricot, gives exactly the same kind of results as apricot. That is, the two-dose female is much the same in degree of bristle development as the one-dose male and it has far fewer bristles than the two-dose male.

     Various other hypomorphic sex-linked mutants have been studied similarly, by comparing the effects of one dose in the male with those of two doses and of one dose in the female (the single dose being brought about by having a small piece removed by irradiation, from the appropriate region of one of the X-chromosomes). Included in this study are the loci of yellow, achaete, facet, singed, garnet, and forked. Of these cases only one, facet, fails to show an accurate compensation for the difference in dosage between the sexes. In facet, the one-dose male is less normal than the two-dose female, as expected for incomplete compensation, although it is not as abnormal as the one-dose female, so that some compensation does exist. In all the rest, however, the male with its one dose looks very much like the female with two doses, and much more nearly normal than the female with only one dose of the gene.

    As a final example we may take the case of Bar eye, shown in Fig. 6, for which quantitative data based on counts of the number of optic elements, or ommatidia, present were gathered by Offermann working in collaboration with the present author. This particular allele of Bar, known as Stone's Bar B^s, had been obtained by Stone as a result of an irradiation whereby the X-chromosome, already containing Bar, was broken into two pieces, in such a way that the gene for Bar became included in the right hand fragment. This was the smaller fragment, and would in fact appear very much smaller, relatively to the other piece, than in the diagram, if only the genetically active portions of the chromosomes were shown.

       By constructing, through appropriate crosses, flies with one or more of these small fragments, then, the effects, on the size of the eye, of varying the doses of B^s in both the male and female could be observed. Since in this case the mutant is not a hypomorph but a neomorph we find that, unlike what occurs with apricot, scute, etc., the more doses of Bar there are the more abnormal is the effect, as manifested in reduction of eye width, while on the other hand changes in the number of doses of the normal locus have no effect on the eye width.

      Nevertheless, the size of the eye of the female with two doses of Bar is not very different from that of the male with one dose of Bar, while much smaller than that of the female with one dose, and much larger than that of the female with three doses or of the male with two doses. In other words, there is a strong tendency to dosage compensation in this case, even though the developmental reaction here, a neomorphic one, must be taking a somewhat different course, qualitatively, from that in the normal organism, and one not summative with the latter. The only other sex-linked mutation yet known to be neomorphic, however, that known as Hairy-wing, fails to show dosage compensation.

    Despite the rare exceptions to, and the comparatively minor imperfections in dosage compensation, the phenomenon is so general, especially in the case of hypomorpliic mutants, and the resemblance of one-dose males is so very much closer, on the whole, to two-dose females than to one-dose females, that it is impossible to regard it as accidental. We are therefore confronted with the questions:

• (1) What has caused the organism to develop this genetic mechanism?

• (2) What is its nature in more detail?

Turning our attention to the question of the cause of the phenomenon, we must, to begin with, reckon with the at first sight surprising fact, that the above cited manifestations of dosage compensation are all exhibited by mutant genes (mainly hypomorphs). It is evident that there could be only the most minimal advantage to the species in having these mutant genes attain one degree of expression rather than another, in view of the fact that they occur so rarely in the population anyway, and that they are especially rare in that homozygous condition in females in which the compensators exert their characteristic effect.

     Moreover, it is found in the case of the locus of apricot, and of some others in which several grades of hypomorphic alleles are available, that in general the compensation works as well for one grade of mutant allele as for another, making the female about equal to the male. Now there can hardly be an advantage in keeping the apricot female's color as low as it is, if at the same time there is an advantage, in the case of the darker allele coral, in having the grade of the female (though here too kept down to that of its own corresponding male) kept as high as it is in this latter case. The effect on the manifestation of the mutant character must therefore be an automatic result of the operation of processes the reason for existence of which lies in their effect in some other connection.

    What then is this "other connection"? There can be but one answer here.

    The compensators must be genes, established in the normal type, which have the effect of equalizing the expressions of the normal alleles of the mutant genes which we have been studying, so as to make the female's two doses of the normal gene produce the same degree of development of the given character as does the male's one dose of the same normal gene. And the fact that the phenomenon is so general in its occurrence, indicates that this result must entail some advantage to the individual and to the species.

      This advantage cannot lie in the fact of the resemblance itself, between male and female, that is thereby brought about in the given characters. Rather it must mean that, as regards most features and characters, whatever is the best grade of expression for an individual of one sex is also, and for the same reasons, the best grade for one of the other sex as well. For after all the males and females of the same species usually have much more in common, in their needs, way of life, physiology, ecology, etc., than they diverge in; and, while secondary sexual differences certainly should and do occur, these are, from the standpoint of the total organismic make-up, a comparatively minor part of the whole.

     But the genes in the X-chromosome are numbered in the hundreds or even thousands, and they concern all sorts of bodily processes and structures. Therefore, in the case of the great majority of them, that grade of development of their effects which is best for one sex is pretty much the same as the best grade for the other sex. Since, however, the female has two doses of each such gene and the male but one, the attainment of that same grade by both would require some special genetic mechanism, one only to be achieved by what we have termed "dosage compensators".

    A converse proof of the existence of dosage compensation as an adaptation to the end of equalizing the expression of genes of which the two sexes have unequal doses is provided by an examination of the dosage effects of non-sex-linked genes. Here too we find, mainly from studies involving chromosomes with small sections missing, that most mutant genes are hypomorphs, giving a more normal grade of character as a result of two doses than of one. But here the two doses in the female give an effect like that of two doses in the male, in accordance with the fact that here both male and female do regularly have the same dosage. Thus, the sexes are again usually equal, but, compensators being unnecessary to achieve this result, genes having the property of compensators do not occur.

    A still more crucial test of the same point is furnished by the case of the hypomorphic mutant called "bobbed," which reduces bristle size, a case worked out by Stern (1929) before dosage compensation was known. This gene lies in the X-chromosome, but in that small part of it which has a homologue in the Y-chromosome, so that, unlike the vast majority of sex-linked genes, the male has the locus represented in both its sex chromosomes, there being ordinarily one so-called normal allele in the X and, in effect, at least one* in the Y. [Footnote: It is probably present in full or nearly full strength in just one arm of the Y (the short arm), and in the form of a weaker allele in the other arm, as Neuhaus first showed. Thus the male would have it represented thrice, but its total dose is probably not much, if any, more than double in this sex, so that the male's dose is not very different from the female's.]

     Now the dosage studies showed that in this case, as with non-sex-linked genes, there was no dosage compensation, inasmuch as a male with one dose of the gene for bobbed, and lacking a Y, looked much like a female with one dose of it, while a female with two doses had much better developed bristles than either. A remarkable check on this case, being in a way the converse of the converse, is found in the genetic situation with respect to bobbed in the related Drosophila species simulans, as disclosed by data of Sturtevant's (1929). Here the Y does not contain an allele of bobbed like the normal allele present in the X, but contains in some strains an amorph (i.e. fails to influence the bobbed character), and in others an allele which, like an antimorph, acts to intensify the abnormality when the mutant gene for bobbed is present in the X.

     Thus, dosage compensation of bobbed would certainly be expected in D. simulans, if it really exists to serve the end of character equalization. And this is in fact found to be the case. For, unlike what occurs in D. melanogaster, the simulans bobbed male without a Y, having its one dose of bobbed in the X, possesses as well-developed bristles as the female with two doses of bobbed. All these cases, then, combine to show that it is not the fact of a gene's being in the X-chromosome, per se, which somehow leads to its being subject to dosage compensation, for this phenomenon appears only in the case of those genes whose normal alleles are regularly in different doses in the two sexes.

    Further evidence that the chromosome configuration in itself has nothing to do with the matter, is seen in the cases in which a piece of the X-chromosome has become broken off and attached to another chromosome and/or, conversely, in which a part of another chromosome has become translocated onto the X. Whether the pieces are large or small, or derived from one or another chromosome region, the result is the same: the genes, both those originally of the X and those of other chromosomes, still have the same dosage effects as they did in their old positions. Compensation is a chemical mechanism, or rather, system of mechanisms, stably established in the distant past, with reference only to those particular genes which regularly existed in different doses in the two sexes, and so it continues to operate now even when we change the very conditions that must once have called it forth.

    If our thesis is correct, dosage compensation is a mechanism normally at work to equalize the expressions of the different doses of normal genes present in the X-chromosomes of the two sexes, and it must have become established as a result of the advantage conferred by this regulation of the expression of the normal genes. Why then have we drawn our above cases from the effects observed in the case of mutant genes?

    Quite obviously this is because of the mutant genes having a level of activity so low that differences in their dosages or activity give an easily observable result, whereas for normal genes, lying as they do near their saturation level, the difference in expression resulting from two as compared with one or three doses are usually imperceptible, as has previously been pointed out. Dosage tests, like those above described for mutants, have in fact been carried out for the normal alleles of all these mutants. But since the individuals, whether male or female, with one or two or three doses looked sensibly alike in the great majority of cases, no conclusion as to whether dosage compensation was or was not occurring could be directly derived from these observations. The mutants, in other words, were required by us to serve as indicators, or sensitizers, for processes and relations which we would not otherwise have been aware of, but which we cannot avoid concluding must really be present in the normal organism as well.

* [Footnote: * More delicate than direct observation of eye color, however, is electrophotometric determination of the relative concentration of extracted pigment. R. M. Valencia, F. Verderosa and the author have, since the above was written, applied this method to the red pigment of eyes having the normal allele of apricot. The results show unequivocally that this normal allele does in fact (like apricot but unlike eosin) have dosage compensation, just as had been expected on our theory presented above. (Footnote added in proof, May 14, 1949.)]

    But although the great majority of normal genes fail to show enough visible dosage effect to make such studies possible on the normal alleles of most known mutant genes, it would be strange if the X-chromosome, containing as it does from one to several thousand separable genes in all, did not include some normal genes that were far enough from their saturation level to be open to such investigation, either singly or collectively.

    That this is true is demonstrated by the fact that the subtraction of a comparatively small section from one of the X-chromosomes of an otherwise normal female, though it only changes the dosage of relatively few of her normal genes, frequently leads to a visible morphological abnormality (aside from effects on sex and sexual characters), of a type depending on what region is subtracted. The "deficiency," as it is called, is also apt to reduce the viability in some way, and it is certain to cause death if the piece removed is a rather large one, or if it includes any one of a number of particular loci. These results prove several points. It becomes evident:

• First, that there are a few normal genes whose effect is considerably different in one dose than in two. It is only to be expected, however, that these seldom happen to be the normal alleles of the relatively few visible mutations which we have to work with.

• Second, it is evident that there are many normal genes whose effect is slightly different in one than in two doses, so that by the reduction to single dose of many at once the combined effect becomes perceptible where the individual effect would not be.

• Third, the doses that these genes normally have, and their normal relationships to one another, are the ones most advantageous for the organism.

• Fourth, these normal genes are subject to dosage compensation.

    At the risk of laboring our argument, it is important that we see clearly how the fourth conclusion above is arrived at. It depends on the fact that, when we reduce the dosage of the normal genes in a given section of the X-chromosome from 2 to 1, by removing this section from one of the two X's of a female, we establish the dosage of these genes which the male normally has. The normal male however does not exhibit the morphological or physiological abnormalities that appear in the females with these deficiencies. This means that in it, the normal genes in question, though present in only one dose, attain about as high an effectiveness as is achieved by the two doses of them in the female. This difference in effect can only be referred to the double dose which the females have of genes in other parts of the X-chromosome, that must act as compensators.

    Further evidence of the same sort of thing is furnished by studies of the effect of adding sections of the X that contain normal genes. This operation too results in definite syndromes of visible and physiological abnormalities (aside from effects on sex and sexual characters), especially if the piece added is large, although, as might be expected from the shape of our curve of effectiveness (Fig. 5), a larger addition than subtraction can be tolerated. But in such cases it is always evident that, while addition of a given piece to a male (leading to two doses of it) results in changes similar in kind to those produced by the same addition to a female (leading to three doses of it in her), the changes in the male are considerably more pronounced.

    It will be seen that this male, having thereby been given only two doses of the normal genes in question, has, so far as these genes alone are concerned, a genetic composition which the female tolerates with entire impunity. The difference in effect obviously lies in the double doses of compensators for these normal genes which the female possesses. And that the two doses in the male cause even more abnormality than three doses of the same section of chromosome in the female is also to be expected, if the degree of effect depends on the ratio of primary genes to compensators. For the two-dose male has a ratio of these (2:1) which is further from the normal ratio (2:2 or 1:1) than is that of the three-dose female (3:2).

  All the above evidence shows clearly that the normal genes in the X-chromosome are subject to dosage compensation, even though we can seldom prove this directly in the case of previously selected loci, containing the normal alleles of known mutant genes, since most normal genes are too near their saturation level to make such observation possible. But the question then arises, should not the very fact that most of these genes are so near their saturation level make dosage compensation unnecessary in their case?

     We have seen that for the normal alleles of apricot, scute, etc., practically no difference between the effects of one and two doses can be seen in either female or male. Why then would there be a perceptible advantage in going through the motions of equalizing them still further? The answer can only be that since, even for these cases, there is evidence from the mutant alleles that the compensation mechanism does apply to the given locus, we must conclude that there is an advantage in it, even when the genes are normal ones.

Let us now try to estimate how small the sex difference in the character is, which we have thus decided must affect the organism's welfare. We may use here essentially the same method as we applied on pp. 182-3 when we calculated, on the basis of the variability of the mutant, how much variation in the normal character was consistent with survival. We may assume, as a first approximation (though one justified by existing data on multiple alleles), that the amount of effect of the compensators on the mutant genes is ordinarily about the same, relatively to the activity of these genes, as the amount of effect on the normal genes, relatively to the activity of the latter. That is, since we observe that the additional dose of compensators present in the mutant female as compared with the male, reduces the effectiveness of the two doses of the given mutant gene which she possesses so as to make them equal to the one dose of the male, we infer that the compensators are likewise, in the normal female, reducing the effectiveness of her two doses of the normal allele to that of the male's one dose, and with at least as great relative accuracy. We may tentatively mark off the amount of activity thereby attained in the normal female as a point on the abscissa of our curve (e.g. in Fig. 5), placed in a region so near that of the saturation level that a point half way from the origin to this point would still give a character not perceptibly (to our unaided observation) different from it.

    Now there has necessarily been a certain latitude in the determination of the exact grade of effectiveness of the mutant allele (e.g., apricot), owing to:

• (1) Errors of perception or of measurement of the given character.

• (2) Variations in conditions that influence it.

Moreover, in the case of some mutant genes, there is also some inaccuracy in the compensation achieved, evinced by under- or over-compensation as the case may be. A maximum estimate of the error from all these causes together has been indicated, in the case of apricot, by the dotted vertical lines arising on either side of the ordinate that marks the typical effect of two doses of the mutant in the female or one in the male. (In the given case, these limits represent mainly the degree of coarseness of our powers of observation -- i.e., the discrimination thresholds.)

      But, in harmony with our previous premise, we should, to be on the side of caution, assume that, in its action on the normal character, the process of dosage compensation has an equivalent amount of latitude to that found in the mutant, in the sense that the inexactitude may be as much, relatively to the total activity of the normal genes, as it is relatively to that of the mutant genes. This enables us to plot on the abscissa the approximate plus and minus limits of the effectiveness of the genes existing in the normal type (either male or female).

      Constructing ordinates at these limits, and marking off the points at which they intersect our curve of effect, we may then find (on our vertical axis) the maximum difference in the normal character which could be produced by one dose in the normal male as compared with two doses in the normal female. It will he seen that in the given case this difference, which might be called the maximum inaccuracy of the dosage compensation, is small, even as compared with the difference caused by a change from one to two doses in the same sex.

    Stating the above in a somewhat different way, let us first assume a maximal inexactitude of compensation for the mutant gene studied (apricot). This would here be about equal to one step of visual discrimination, since (according to observations of the author but contrary to some statements in the literature which report the male to be slightly darker) no certain difference can be distinguished between the apricot male and female. Now divide this maximal inexactitude (or, more properly speaking, uncertainty) by the amount of visible difference which the compensation mechanism is here working to offset, i.e., by the amount of visible difference between the effects of one and two doses of the apricot gene in the male.

     This gives a relative value for this maximal possible inexactitude, which turns out to be about one tenth. That is, at least nine tenths of the dosage difference has been compensated for, but possibly (at most) there is a one tenth under- or over-compensation. To be on the side of caution, we may now assume that there is as great a relative inexactitude of compensation for the normal alleles also -- although, on the average, the inexactitude should actually be less for normal than for mutant alleles, on account of the mechanism having been evolved in relation to the normals. This then leads us to the conclusion that, in the case of the normal alleles of apricot, not more than one tenth of the difference between the effects of one and two doses in the male has escaped compensation, while nine tenths at least have been compensated for.

       But we have seen that the entire difference in question, between the effects of one and two doses of the normal gene, whether in male or female, is itself (on the average, at least) imperceptible to our eye. Thus we find that selection has been sensitive to differences at least an order of magnitude smaller than the smallest difference which we can ordinarily see. Hence even subliminal gradations of this fineness are actually of adaptive significance to the organism.

   Application of the same method leads to similar conclusions in the case of scute and most of the other characters observed in regard to this point. That is, the compensation mechanism involves a process, or system of processes, that works to a much greater degree of refinement than is visible to us. We have seen, then, that natural selection has in the first place, in the interests of character stability, established the effectiveness of the normal gene at a point on the curve so near to the saturation level that one dose is not visibly different in effect from two, so far as we can see. Yet even this likeness is far from sufficient, for organisms (male versus female) that may differ regularly to this seemingly slight degree, otherwise the mechanism of compensation would not have been set up, in addition to this, in the case of the great majority of sex-linked genes, so as to reduce the difference to a tenth or less of even this invisibly small amount!

    It is evident too that the required equalization cannot so advantageously be achieved by the simpler method of stepping up the normal gene's effectiveness still more, into a still flatter portion of the curve, otherwise the more elaborate compensation mechanism which exists would not have been needed. That it does exist, and therefore was needed, shows that a relatively slight increase in the effect, in fact imperceptible to us, brought about in the above way, is disadvantageous, just as an imperceptible decrease in the effect is.

      In other words, it is as important to keep down the grade of the normal character to the given level in the two-dose female as to keep it up to that level in the one-dose male, and in most cases the gene's activity cannot advantageously be brought near enough to an actual saturation level to obviate this requirement. However, for the lesser variations in effectiveness of the normal gene, caused by fluctuating environmental conditions and by variable genetic "background", the raised stability brought about by the near-saturation level of the gene must be of great value.

    There is no doubt, however, that in many cases specific regulatory mechanisms ("servomechanisms") have been evolved, in addition to the above, to achieve an even higher stability in the face of given especially frequent or especially disturbing conditions, as many phenomena both of embryology, regeneration, physiology and pathology show. Moreover, in a relatively small but absolutely huge number of cases, the adaptive reactions go far beyond mere stability, and consist instead of positive changes of an advantageous nature following given types of changes in conditions. These reactions range from the "morphogenic" to the "physiological," there being no essential distinction between the two categories. This however is a topic too large to be entered upon here, since it would launch us upon the entire subject of how the organism works. We mention it here only to set our own problem in its proper limited place in relation to biological phenomena in general.

    In order to avoid the conclusion that differences far too small to be evident to us are nevertheless of importance to the organism, some might prefer to suppose that selection has rather worked by eliminating the occasional individual in which, owing to special conditions, the character in question had been influenced to a considerably larger degree than usual by the gene difference in question. This possibility has already been discussed in its relation to the character stabilization brought about by raising the gene's effectiveness to near-saturation levels.

     We have seen that, although this position is tenable, and is doubtless valid to a certain extent, it nevertheless brings us to much the same conclusion in the end, when this is expressed in terms of the average amount of character-difference caused by a gene difference which has an appreciable adaptive value. Moreover, there is reason to infer that, in many cases, there is no actual threshold amount of difference which suddenly emerges as disadvantageous, but that the amount (or the chance) of disadvantage may be taken as roughly proportional to the amount of deviation from the norm, even in the case of minute differences.

      In such cases, then, the amount of selection involved would be the sum of that occurring for each grade of expression, and that for a given grade would be proportional to approximately the product of the number of individuals with that grade times the amount of their deviation (i.e., the size of the grade, when the normal is designated as 0). A reckoning of this kind will show that here the selection based on relatively low grades of deviation (e.g. those less than the standard deviation) will probably in most cases considerably outweigh that based on the rarer, more extreme deviants. This presupposes, however, for the cases we have dealt with above, that survival affords a considerably more sensitive criterion of the degree of development of the character than does our own eyesight.

A different objection which may be raised to these conclusions is that, in studying the above characters of eye color, bristle development, etc., we may have been dealing only with certain unimportant superficial expressions of gene changes which have other, more important, effects, hidden from our casual observation. In other words, characters are often correlated in their development, and we may have studied the less important by-products of the gene-initiated processes.

     On this view, natural selection may in truth have resulted in the establishment of dosage compensation for the normal gene, because of the advantage of attaining a given level of development of its most important effect. This effect itself, though hidden, might be subject to a very considerable dosage influence. Yet, studying only the side effect, we may have been able to obtain evidence of only a subliminal action of the compensators. Granting this viewpoint, we might avoid the conclusion that selection has had to deal with extremely fine differences.

    It is true that genes do have multiple effects, thus causing correlations between characters, and that in no individual case with which we are dealing can we know that the advantage of the dosage compensation mechanism lies in its influence over that particular effect of the given gene which is obvious to us. However, it would be strange if we had missed the major effect of the gene in such a large majority of the cases, especially since some at least of the genes concerned (the normal alleles of yellow, achaete and white) have been proved physiologically dispensable, i.e., the individual can live in the total absence of these genes. But, if we have caught the major effect in any of the cases, then that effect does show the type of compensation at issue -- a type so refined as to operate on subliminal grades of the character. Furthermore, we have seen that with but very rare exceptions (relative to the total number of genes) the effects of individual genes, whether in the X or other chromosomes, are so near their saturation level as to make direct discrimination between one and two doses impossible. Since a very wide sampling has consistently given this result, it would be highly unlikely that this relation did not hold for the major effects of genes in general as well as for their minor effects.

     In fact, if our interpretation is correct it is particularly for the major effects that this aid to stability has been established. But if this is true of the major effects, and if the dosage compensation found is, as the present criticism started out by assuming, also a phenomenon developed primarily for its influence on the major effect, then the main conclusion still remains valid, for the compensation results in much finer grades of difference even than those, already subliminal, which are due to doubling or halving the dose. Thus, in any case, the compensation mechanism must be concerned with the equalization of exceedingly minute differences.

    A third point to be considered in connection with the question here at issue is the following: The great majority of characters which a gene affects are not immediate products of it but are end-results of complex webs of biochemical processes, in which a given thread or chain of reactions can, theoretically, be picked out that has its root in the gene in question, and leads from the latter to the character through a considerable series of events (see Fig. 7).

      Now when the given gene has multiple effects, it is evident that the gene either serves as the root of two or more entirely separate chains of reactions which respectively lead to the two or more characters affected, or else that, although but one basal chain of processes has its start at the given gene, this chain somewhere along the line becomes branched, so as finally to diverge to the two or more correlated characters. Owing to the complexity of the whole web of ontogenetic and physiological processes, the branching will, in the case of different genes, occur at different points. In some cases it will happen to occur early, near to or at the gene itself, while in others it will occur late, near the character which the observer examines. This diversity in the position of the branching has been amply demonstrated by many studies in physiological genetics, as well as in other fields of biology and medicine.

     Now influences derived from other genes (as well as, ultimately at least, from environmental sources) impinge upon these chains of reaction at practically every link, thus playing their part in the determination of the final characters. Unless we make very unusual assumptions the compensating genes work in this way too. That is, the compensations may be achieved by influences impinging on the reaction chain at any point.

          And as it is a matter of indifference to the organism at which point this anastomosis of effects takes place, so long as the final advantageous result, compensation, is attained, it will sometimes be above and sometimes below the branching that leads to a given character (Fig. 7), if this character alone is important enough to have compensation established for it.

      Moreover, in such a case, if the compensation is the accumulated result of a number of influences, derived from different, cooperating compensator-genes, some of these influences, ordinarily, will be brought to bear above, and others below, the point of branching. Only those influences which act below the branching, however, would then affect the secondary characters as well as the important one. For this reason, we should not expect to find dosage compensation to be nearly so widespread, or so accurate, as it is if we had been studying only secondary characters, such as were of too little importance for the mechanism to have been developed in relation to them themselves.

     All in all, then, we may reject the objection based on correlation of characters, and conclude that dosage compensation has in fact become established because of its advantage in regulating more precisely the grade of characters whose variation in grade, even without it, would be exceedingly minute.

   We may here digress to point out that the considerations above set forth concerning the way in which the reactions initiated by genes bifurcate to produce multiple effects (Fig. 7) also throw light on the related question of the extent to which selection for a given important character may be expected automatically to carry, in its train, a developmentally correlated but not advantageous change in another character.

     It should rarely happen that two apparently separate characters, especially if they be characters governing only degrees of development or proportions of already present parts or substances, will be so closely connected, with the branching occurring so very near to the distal end of the reaction chain of either of them, that they cannot both be affected separately, by influences impinging beyond the point of branching.

      Therefore, if the so-called secondary effect is a disadvantageous one, it will usually be corrected, without disturbance of the primary effect, by means of a selection acting upon genes which affect the secondary branch but not the main stem. Moreover, it is to be expected -- especially in view of our present results -- that most visible characters, even those of an apparently superficial nature or of very minor degree, will nevertheless have enough importance in themselves for their variations to be felt in the selective process. For this reason we should be very wary of any interpretations which would explain the existence of a given grade of development of a character as an automatic by-product of selection for a more primary character.

    So, for example, the finding that the mutation from the normal grayish-tan to yellow body color in Drosophila results in a cuticle which is not so resistant to environmental changes in dampness, in no way explains why the flies do not normally have a yellow body color. For if that color were in itself advantageous it could doubtless be arrived at by selection of mutations in other genes which, though not having the undesirable chemical action, nevertheless gave effectively the same color, or by selection of "modifying" mutations which, with the gene for yellow in question, counteracted the undesirable chemical effect, while leaving the color yellow. In fact, there are other species which normally look rather like the yellow mutant of D. melanogaster and it is probable that in them some such arrangement has in fact been attained by selection (without giving them the above mentioned handicap), as a result of ecological conditions which put a premium on such a color.

    The organism, then, is genetically highly plastic, at least so far as its proportions and the relative degrees of development of its different features are concerned. Thus we may assume it, in general, to have attained optima in all these respects, largely regardless of developmental correlations. In fact, the correlations normally occurring, such as the observed principles of heterogonic growth, Gloger's rule, etc., have themselves come down through the long mill of selection and hence are themselves in large measure advantageous adaptive reactions. Moreover, they can themselves be changed again by further mutation and selection, as observations of specialists in such fields have shown. Nevertheless, the organism does of course have its grave limitations, causing change in some directions, and to some degrees, to be more difficult than in others, or even impossible.

In the above discussions we have implied that dosage compensation has, for each sex-linked gene that is subject to it, become established through a natural selection of all available and innocuous mutations in the X-chromosome that chanced to result in a compensating action upon the expression of the sex-linked gene in question. (Included here of course would also be such mutations in the given gene itself as furthered its sensitivity to such compensation.) This interpretation would put the method of evolution of this phenomenon in line with that which we have good reason to conclude has been followed in the case of most other functional systems of organisms.

     It might however be postulated that there are one or more peculiar genes in the X-chromosome which are especially suited to serve as compensators to all others in it, and that the only mutations then needed to achieve compensation are those, perhaps of a simple specific type, in the genes to be compensated, which allow the developmental processes caused by them to become responsive to the influence of these particular compensators. As a plausible guess, it might further be postulated that these compensators are none other than the gene or genes determining sex itself, and that their compensating influence is exerted through processes which form a part of those that differentiate the sexes.

    Even though it is now known that, in Drosophila, the turning of the scales which decides which sex is to develop depends upon the cumulative action of numerous sex-differentiating genes of individually small effect, scattered through the X-chromosomes, this would not vitiate the possibility here in question. For the effects of these numerous genes obviously converge to set going one fundamental sex-differentiating process, or "focal reaction", on which in turn most or all of the secondary features of sex differentiation depend (see Muller, 1932a).

     This is shown, among other things, by the essential similarity of the whole system of results, so far as sex is concerned, whenever any limited group of these sex-differentiating genes comes into play, in experiments in which the dosage of different parts of the X-chromosome is varied. Thus it might well be supposed that either this main sex-differentiating process itself, or one derived from it, served as the root mechanism for dosage compensation, which would thereby take on the aspect of a secondary sexual phenomenon of fairly familiar type.

    This possibility was never thought to be very probable, in view of certain results derived from a study of X-chromosome fragments, to be mentioned presently. It has, however, been pretty definitely disproved recently by some experiments carried out in 1946 by the present writer and Miss Margaret Lieb in collaboration, but not yet published (see Lieb, '46). These made use of a mutant gene called "transformer" found by Sturtevant, a recessive lying not in the X but in the third chromosome, which has the effect, when homozygous, of causing flies with two X-chromosomes, which otherwise would have become normal females, to develop into males.

     Although sterile (even when provided with a Y), these males are well developed in virtually all other respects, so that it can be inferred that the main sex-differentiating process has been initiated in them in very much the same form as it has in regular males. If now the dosage compensation of the Drosophila female is brought about through effects of the female sex-determining process, or system of processes, as such, then these males with two X-chromosomes should not have the dosage compensation characteristic of the female operative in them. If, for instance, they are provided with two doses of apricot, one in each X-chromosome, they should have the rather dark eye color characteristic of the male with two doses (as in the rightmost eye of Fig. 4), not the compensated color of the female with two doses, which looks like the one-dose male.

     The test clearly showed, however, that they had the fully compensated color, like that of the female with two doses, even though they were males with two doses. Obviously then, the compensation was caused, as in the ordinary female with two doses of apricot, by the presence of a double dose of compensators, lying in other parts of the X than at the locus of apricot itself, and these worked quite independently of the sex-differentiating genes, since all of the latter, together, although present, were here quite incompetent to cause any visible approach towards the female sex. Tests of several other compensated sex-linked characters than apricot -- scute, forked and Bar -- with the transformer gene, gave this same result quite unequivocally for all.

    These findings had been expected because tests performed much earlier, involving the addition of different fragments of the X-chromosome, first by the writer and then by Offermann working in collaboration with the writer, had shown that, in the case of different sex-linked genes, there is a different distribution of sex-compensating potencies throughout the length of the X-chromosome. That is, not only were the fragments taken from different regions unequally effective in the compensation of a given gene (when due allowance was made for their sizes), but their effectiveness relative to one another was not the same when tested with one sex-linked gene as when tested with another one. The different genes therefore had, to a considerable extent at least, different compensators, not one common set of compensators as they would have had if the compensation had worked through sex-differentiating effects. Furthermore, the results for none of the sex-linked genes agreed well with the relative values of the different regions of the X-chromosome in influencing sex-determination itself, as worked out, after some pioneer work of the present author, chiefly by Dobzhansky and Schultz, and by Patterson, Stone, Bedichek and their associates.

    A further and more elaborate investigation of the behavior of the different regions of the X-chromosome in compensating for particular genes has been carried out more recently (1946), by the writer and Lieb working in collaboration, and the results from this confirm and extend the earlier conclusions. It would takes us too far afield here to describe in detail the technique or the results of these various experiments on the location of the compensators. The method was, essentially, first to produce by irradiation fragments of the X-chromosome taken from different regions, and then to determine to what extent the addition of these fragments to a regular chromosome complement (usually of a female) affected the expression of a given sex-linked character, the locus of which lay outside of the fragment.

     In most cases the compensation proved to be a complex phenomenon, which could not be pinned down, in toto, to a particular chromosome region, nor even regarded as a simple summation of the effects of the different regions. Some regions, however, had distinctly more influence than others in the compensation of a given gene, and there were even cases where a certain region exerted what may be called a "negative compensation", working in exactly the wrong direction, an effect which was however counteracted by means of overcompensation elsewhere. That is, the surprisingly accurate compensation exerted by the chromosome as a whole was a final resultant, brought about only through an integration of individual effects which in detail appeared chaotic. And a similar chaos was evident on comparison of the manner of distribution of the effects among different regions, as seen in comparisons of the cases of different compensated genes.

    Now this is exactly the sort of result we should anticipate if each gene has had its own separate system of compensators selected for it, and if the compensation for each has, in most cases, come about by a succession of steps. Some genes could of course have gained a foothold (if we suppose them not to have been priorly present) that even worked the wrong way in the given respect, provided they possessed an advantage lying in some other direction, but even these would ultimately be corrected, by the selection of genes that counteracted such an effect, as we have pointed out in our discussion of correlated characters, and so the final resultant effect would be superior to that of any single interacting gene. The attainment of the accuracy finally arrived at must usually have required many small mutations, which, taken together, increasingly whittled the character into shape, as it were. And while in some cases these mutations may have been successive changes in the same gene, any mutation in the X-chromosome that worked in the advantageous direction would be acceptable, so that in most cases it is to be expected that a fairly complex system became at last established. Since, however, the advantage of the compensation for any single locus is only what might be called a "second-order" one, dealing as it does with such minute, to us subliminal, grades of effect, the establishment of these precisely working systems must have taken far more time than needed for visible "first order" effects. Moreover, there may more often have been setbacks which required repair, and which thus led to still greater complexity.

    If we may infer that most of the genes in the X-chromosome have compensation systems like those studied above (an inference to be accepted with some caution since we were dealing only with genes capable of giving conspicuous visible mutations), we would be led to the peculiar conclusion that, since there is usually more than one compensator for each gene of the X-chromosome, and since the compensators are themselves in this chromosome, most sex-linked genes not only have an original or primary function but also serve as compensators for one or more other sex-linked genes. Each individual gene then would be complex in its functioning and also in its potentialities of mutating so as to affect its different functions to different degrees, relatively to one another. Finally, as there are over a thousand sex-linked genes, it will be seen how amazingly complex, in all probability, is the tangle formed by the sex-linked system as a whole.

    In conceiving the mode of action of a compensator, or collective group of compensators, for a given sex-linked gene, which we will call the "primary gene," it is easy to make the mistake of thinking that each dose of the compensator or compensators effects a given total amount of reduction of the activity of the primary gene (or rather, gene-product). If this were true, however, then the female with her two doses of both primary and compensators would show an effect equivalent to twice that produced by the one dose of primary and one of compensators in the male. In other words, there would be no equalization of the sexes. Moreover, in that case compensators acting strongly with hypomorphs would not work properly for genes at higher levels. We must therefore infer that the compensators, when present in any given dose, work in such a way as to effect the same proportionate amount of reduction in primary gene activity, regardless (within wide limits) of what the dose or activity of the primary gene is. This would ordinarily be the case if the inhibiting action of the compensator were itself little influenced, in return, by the amount of primary gene-product it had to affect. An example of this would be a situation in which the compensator's "product", determined indirectly by its gene, consisted in some such pervasive condition as a relatively high pH, which the primary gene's product, no matter how concentrated, had little effect on. In such cases, then, the compensators, at a given dose, would tend to reduce the primary action by a given proportion, rather than by a given absolute amount.

    Moreover, the relations must be so fixed that the compensators, when themselves in double dose, reduce the effect of a given dose of the primary gene to half of that which would obtain in the presence of a single dose of compensators. For only thus can the effect of the female's double dose be reduced to that of the male's single dose. Such a result would be brought about most simply in a case in which the compensators, when themselves in single dose, reduced the primary effect to half what it otherwise would be, and in which, when their dose was raised, their own effectiveness rose in the usual geometrical manner. For, in such a case, one dose of the primary with one of compensators, as in the male, would have an effect which may be designated as "one half" (i.e. one half that of one uncompensated dose of the primary gene). And two doses of the primary with two of compensators, as in the female, would have an effect equal to the product of 2 doses (for the primary) times 1/2 (for the first dose of compensators) times another 1/2 (for the second dose of compensators), a product which of course is itself equal to one half, and therefore the same as the effect in the male, i.e., properly compensated. And no matter what the degree of activity of the primary gene might be in such a case (e.g., whether it were the normal allele or, say, a hypomorph having an activity equal only to one quarter of one dose of the normal), the compensation would still work to equalize the effects of the one and two doses of the primary in the two sexes. Whether this simple scheme is usually true can probably be determined definitely through quantitative studies involving several different doses of compensators.

It has been mentioned that, according to the conception above outlined, dosage compensation must be a "second-order" evolutionary phenomenon, since the selective force that establishes it must depend upon such minute advantages. It might well be so slow, therefore, as not to have caught up with the sex-linked genes in species in which they had been established as sex-linked for a period of the order of only a few million years. There are several known Drosophila species which probably meet this condition, and which might therefore serve as a partial test of our thesis. Of these, the most worked with has been Drosophila pseudo-obscura, and the results with it have therefore been reexamined from this point of view by Dr. Rowena Lamy and the present writer, in collaboration.

    A considerable body of earlier work had joined to show that what serves as the X-chromosome in D. pseudo-obscura consists of two parts or arms, one of which is, for the most part at least, descended from the original rod-shaped X-chromosome, that has remained as the X also in our common species melanogaster, while the other arm is, as a part of an X, a relatively new acquisition, since it corresponds to what forms the left arm of the third chromosome of melanogaster, and also lies apart from the X system in most other Drosophilas, as well as, presumably, in the common ancestor of both pseudo-obscura and melanogaster. The connection of it with the original rod-shaped X, to form a two-armed X, must have occurred by what we call a whole-arm translocation, and this must have been in a period geologically a good deal more recent than that at which most of the material of the original rod X was established as X-chromosome material. In view of this, it might well be expected, on our conception, that while the genes of the original X arm would show good dosage compensation, as they do in melanogaster, those in the other arm might still have this mechanism much less perfectly evolved.

    For testing this question, we have not resorted to the difficult and rigorous method of studying dosage effects as revealed by the addition or subtraction of fragments. We have, however, made use of the knowledge, derived from melanogaster, that many or most mutant genes are hypomorplic and, unless compensated, exert a more nearly normal effect in double than in single dose. This being the case, most of the mutants in a well compensated part of an X-chromosome would look alike in male and female, whereas many of those in a poorly compensated part would look more extreme in the male. There would, to be sure, be exceptions caused by the fact that some mutants are partially "sex-limited" or "sex-influenced". This phenomenon, which is most readily demonstrated in studies of non-sex-linked genes, but which must apply to some sex-linked genes as well, is quite apart from dosage compensation, and probably represents in some cases an adaptation to differing sex needs. However, the deviations from equality of sex-expression which were produced in this way would be just as likely to cause the male to be less extreme than the female, as to cause him to be more extreme, and they would not have a tendency to affect genes in one chromosome arm rather than in another.

    When the results of sex comparisons were tabulated for all those sex-linked genes concerning which data were available to us, it was found that in the "left" arm of the X of pseudo-obscura the great majority of the mutants looked alike in the two sexes, with however an occasional exception, that in one case had the male more extreme, but in the other case, the female. Now this is the arm that is descended from the original X-chromosome. In the more recently acquired arm of the X, on the other hand, the "right" arm, there was a much higher proportion of genes which showed a sex difference in their grade of expression. Moreover, in nearly all these cases the direction of the difference was the expected one, namely, that in which the female stands nearer to the normal type -- an indication of absence or incompleteness of compensation.

    As yet, the numbers of mutants examined in the above way are hardly great enough to make our conclusions as secure statistically as we would wish, yet the agreement with the theory is fully as good as we had expected. This evidence can therefore be counted as contributory to our general thesis, especially since it comes from such a different direction of attack than the previous work. It would be of great interest to have further evidence, involving exact dosage studies, obtained along this line. This would be especially valuable in such a species as Drosophila miranda, which is closely related to D. pseudo-obscura, but has been found by Dobzhansky and Tan to have, in addition, a third portion of the X-chromosome system, still more recently acquired by the latter. Drosophila americana and virilis would likewise provide an important comparison for this purpose.

Although the facts concerning dosage compensation appear to give us the most searching evidence of the amazing degree of precision of genetic adaptation, or, to put the matter conversely, of the disadvantageousness of even very subliminal (to us) departures from the already established type, nevertheless we have not to look far to find other lines of evidence, of somewhat lesser exactitude, for the same sort of thing. Among these, one of the most convincing, from a genetic standpoint, is the very slow rate at which most characters of most species are, at any given moment, undergoing evolutionary change, as shown by the very high resemblance of the organisms in most respects to their ancestors of, say, a million years before.

     True, there have been other periods, of comparatively rapid evolution, for each line of descent, as for example in the human stock during the past few million years, but these are relatively rare. They illustrate the general principle, applicable in biological evolution as elsewhere, that "one good turn deserves another". On the whole, during the past of most species, the periods of high stability have occupied much the greater portion of the duration of their lineage. Thus many insects preserved in the Baltic amber were, in the pettiest details, sensibly the same species at the remote time when they became thus entangled as they are today.

    That this stability is due to active selection in favor of the normal type, that is, to the fact that the latter is more advantageous than even the slight deviates from it, is shown by a consideration of the findings of genetics concerning the frequency of mutations. In those few but very different multicellular organisms which have served as samples for the investigation of this subject, the frequency of detectable mutations in individual genes has proved to have a range of values centering around the order of magnitude of one such mutation in the given gene in about fifty or a hundred thousand, or, at most, a few hundred thousand, germ cells. Or, if we were to follow a given gene down through a single line of descendants, this would mean that it would on the average undergo such a mutation in fifty to a few hundred thousand generations. In most higher organisms, this would occupy less than a million years. If slighter mutations than those easily detectable could have been taken into consideration, the period between one mutation and another of any given gene which we were following might in fact have been considerably less than this. Now if the mutant genes did not handicap the individuals in their living, but had an equal chance of survival with the normal, it would in consequence come to pass that after a million years had elapsed the great majority of genes, and with them practically all characters, would have become appreciably different. And in just a few million years the organisms would have become practically unrecognizable in all respects, merely as a result of "mutation pressure".

    We may, to be sure, grant that many mutations are so pronounced or so manifold in their effects as to be lethal or detrimental, and that a change in all genes would therefore be too pronounced a result to expect in the above period. However, despite this stricture, it remains to be taken into account that practically all characters depend on many genes, and that of these a large proportion produce only minor effects on them [the characters] and often mutate in such a way as not seriously to affect other characters at the same time. Therefore, if we considered the mutation frequency for a given character instead of for a given gene, we would find it to be very much higher than the above, even when only those mutations were included in the reckoning which caused but slight changes in the character and which produced little in the way of detrimental effects of other kinds. As a consequence, if these slight changes in the character were not in themselves disadvantageous, a few million years could not pass without practically every character having undergone a whole series of stepwise changes, more or less random in their directions, some steps having been smaller, some larger, and all combined having the result of making the character very different from what it had been. The only preventive to this unsettling effect of the prolonged mutation pressure would be selection, -- i.e., the dying out of the mutant individuals along the way, -- by reason of the disadvantage conferred by the slight, as well as by the great, variations in the character. Thus, the mere presence of a given character, or of a given grade of development of a character, in a species, over a period as long as a million generations, in itself demonstrates that this character, or even this particular grade of development of it, is of material value to the organism in its struggle for life and perpetuation.

    It is evident that this line of argument becomes even stronger in those numerous cases in which a given character, or a given grade of its development, is found in two or more related species, that probably diverged from a mutual ancestral type some million or more generations previously. In such cases no records of this ancestor itself are needed, provided there is good ground for it to be at least as remote as this. But there is sometimes a further kind of argument in these cases. This is illustrated by Sturtevant's finding of many years ago, that when the related species Drosophila melanogaster and D. simulans are crossed, although both these species have an identical pattern of bristles, one which must have been very long established and present in their common ancestor, nevertheless the hybrids between them show considerable disturbance in this pattern, and especially a tendency to have various bristles missing.

     This can only mean that the present developmental mechanism of bristle production in one species has come to have such a different genetic basis from that in the other species that a mixture or compromise between the two system involves incompatibilities which obstruct the attainment of the effects of either system alone. It is thereby proved that, while the given character remained in, or regained, its original form during the evolution of both species, nevertheless the genetic and developmental basis beneath it underwent changes.* [Footnote: Better instances are provided by cases in which the introduction of a particular gene (mutant or normal) for a given character from one species into another by backcrossing (or, as Edgar Anderson terms it, "introgression") causes a change in the expression of that character, even though the normals of the two species had looked much alike in the given respect. Such cases have been reported by Harland in cotton, by Gordon and by Kosswig in fish, by the present writer in Drosophila, etc. When the conclusions are based only on the first generation hybrid it might be supposed that incompatibilities involving other genes, not normally affecting the given character, were causing disturbances that now affected it.]

    Surely the above changes would not have been just so adjusted as to allow the development of a character which was outwardly the same as before unless there had been some controlling agency which allowed only such changes to become established as failed to alter the end-result (here, the bristle pattern), or which allowed only such changes as were effectively counteracted by others which maintained that end-result. Such control, in the interests of the end-result itself, rather than of the mechanism behind it, can only signify (if we reject the animistic interpretation of a conscious guidance) an active selection favoring individuals that exhibit this end-result.

     In other words, the advantageousness of just this result for the organism is thereby implied. The numerous interspecific crosses in which, even though the first-generation hybrids may appear normal, some of the types produced by genetic recombination in later generations nevertheless manifest abnormalities, show equally well, though slightly less directly, that here too the genetic basis had become different, while maintaining a similar facade. These cases then greatly extend the number of examples of characters that must have been forcibly kept as they had been (or, in some cases, caused to undergo parallel or convergent alteration) because of the advantageousness of their outwardly seen features, and in spite of those genetic forces of mutation pressure, drift, and selection for other, "correlated" characters, which would otherwise have succeeded in making them different from what they are. And it may be recalled that, in many of these instances, not merely a given character as such, but the exact type and grade of development of it, shows this behavior, as in the case of the very precise bristle pattern of the two Drosophila species cited.

    The attainment of very nearly the same end-results, as shown in some of their characters, even by species of extremely different descent and genetic background, when their mode of life and surrounding circumstances are similar, provides even more striking evidence of the usefulness of these characters and of the precision of selection in molding them, as many comparative anatomists, paleontologists and other Darwinian evolutionists have for a long time recognized. Thus the very similar fishlike, accurately streamlined form not only of widely different fish species themselves, but even of the reptilian fossil Ichthyosaurs, as well as of modern whales, bears eloquent testimony to this.

      Another oft quoted example is that of the carnivorous marsupial called the Tasmanian wolf, which, living a wolf-like life, has became so wolf-like in its proportions and general appearance as readily to be mistaken for one despite its much closer genetic relationship to the opossum. Such illustrations could be greatly multiplied, especially if individual organs or features, such as the independently evolved eyes of cephalopods, vertebrates and scallops, were singled out. They prove that the significance of these characters for the organism lies in the end-results themselves, -- i.e., in the advantages which the latter as such confer, -- and that the genetic and developmental mechanisms whereby the results arrived at are matters of comparative indifference, varying greatly with the "accidents" of mutation and of past history, that have worked on a biological material which is essentially very plastic, at least from the standpoint of morphology and proportions. This very plasticity would, however, soon result in its becoming buffeted completely out of shape, in the course of less than one geological period, were it not for the steadying hand of selection, that persists in actively maintaining such a remarkable precision of genetic adaptation.

Nevertheless it should be recognized that some species appear much more variable that others, and that even different characters in the same species can show considerable difference in the amount of individual variation that exists in them and in the degree to which such variation seems to affect the welfare of the organism. Thus, to a European man it probably makes very little difference in his expectation of life in Europe whether his hair happens to be light or dark, sparse or thick, unlike what we have found above to be the case in our Drosophilae. The question accordingly arises: why should there be so much greater variation of characters, and apparently greater tolerance of such variation, in some cases than in others?

    In examining this question it should, in the first place, be understood that, in a large, fairly panmictic population which has long existed as such, the amount of variation found in any given character is largely a function of two sets of factors: the frequencies of mutations of different degrees affecting that character, and the average amounts of disadvantage conferred by the mutations of different degrees.

     In fact, if we make the simplifying assumption, which is probably good enough as a first approximation, that the average amount of disadvantage is proportionate to the amount of deviation, a simple relation can be calculated. This turns out to be as follows (stating it directly in words rather than in mathematical symbols): The amount of disadvantage caused by an (arbitrarily chosen) unit grade of deviation is equal to the mean absolute deviation arising by mutation in the germ cells in any one generation, divided by the total genetic variance of the population in respect to the given character. We need not trouble the reader with the rather simple mathematical derivation of our formula, but only point out that it depends upon the facts that the total variance represents an accumulation of the mutations that occurred throughout many generations, and that the amount of accumulation of those mutations that manifest themselves will be inversely proportional to their disadvantage, that is, to the rate at which they are weeded out.

    If, then, we were to determine the mutation rate for the given character in such a population, we could find approximately what grade of disadvantage a given grade of deviation from the norm entailed. (Included in this reckoning of the disadvantage, however, would be that arising from any correlated effects of the given mutant genes.) A second way of arriving at the same result, of course, would be by direct observations of the relative viabilities and fertilities of the different types, and the results of these two methods should check. Unfortunately, quantitative studies of this kind appear not yet to have been carried out, important though they would be in the attack on problems of evolution.

    In the absence of quantitative studies of these kinds, it is nevertheless already evident that most of the mutations which are viable enough to play any considerable role in the formation of the observed variance manifest themselves with a good deal higher frequency in the population than that frequency with which they arise anew in any one generation. That is, they do accumulate to a considerable degree, and the disadvantage of each individual one is, on the average, fairly small. Nevertheless, this disadvantage does prevent their accumulation beyond the observed amount, and it usually results, at any rate, in the average grade of the character, the norm, being rather well defined and comparatively stable. Moreover, we become aware of the falsity of the assumption so often made, by both biologists and medical men, which holds that variants within the so-called "normal range" (i.e. those falling within, say, the middle 80 or 90 per cent of the area of the curve of variation) are in effect "normals," possessing no or negligible disadvantage.

    Now although differences in the mutation frequency per generation must undoubtedly exist between different species and, in all likelihood even more, between different characters of the same species, it is probable that differences in the degree of disadvantage of comparable deviations from the norm are, on the whole, a much more important cause of the differences in the observed variance of populations found in different cases. An especial reason for this is that the mutation rate is itself responsive to some extent to selection, and that it would tend to creep upwards, through accumulation of mutant genes that themselves allowed mutations to occur more easily, to the point where it was held in check by reason of the disadvantage thus entailed. Hence, an especially low variance is probably, in most cases, an indication of an especially high disadvantageousness of deviates.

    We are then justified in asking, why should there be such considerable differences as the evidence indicates to exist between different species, as well as different characters, in regard to the degree of deviation that causes a given degree of disadvantage? In answer to this, many different illustrations will no doubt present themselves to the biologist and the medical man, of physiological, biochemical, morphological and ecological relations which require different degrees of precision. However, there are certain kinds of situations which are in general conducive to lower requirements of this kind, or to a poorer meeting of these requirements on the part of the population.

     One major situation which would, temporarily at least, produce an increase in the allowed variation of a character is that in which, for any reason, a character that was previously of value has now through some change of conditions (instituted not too recently) had its value diminished or possibly even reversed. A similar effect may be produced if there has been a swift change in regard to what constitutes the optimum grade of the character.

     All this is, of course, equivalent to saying that selection has relaxed, or has set up a different standard than the previously established norm. As a result, many variants that previously would have been eliminated quickly are now allowed a longer lease, or may even be caused to multiply actively, and during the disequilibration the variance will increase until a new norm is approached more closely.

     Applications of this principle to the situation among civilized humans and domesticated animals, in respect to many characters, are obvious. There are somewhat similar applications in such a case as that of the sloth, a mammal which follows a mode of life in which a high level of activity, temperature regulation, and the use of its wits, together with many features connected with these, have largely lost their advantages. In all these situations, then, we should expect to find an increase in the amount of manifested variation, as in fact we do.

    Despite the above relationships, it seems probable that many variations which hitherto have usually been regarded as valueless, especially such physical ones as distinguish human races that live in different climates, will be found to have their functional aspects, or at least to have had them, in the days when they became prevalent. And even characters which occur only as individual variations within groups, sometimes have a usefulness that maintains them up to, and keeps them down to, their given level.

     Instances in point are the cases of dimorphic mimetic butterflies, as pointed out by Ford, the cases of inversions studied in some Drosophila species by Dobzhansky and by Dubinin, and, if I am right, such a character as near-sightedness in man. The latter, according to the interpretation arrived at independently by the present writer and by Riddell (both unpublished), was probably of considerable aid, in primitive tribes, to the relatively few specialists who fashioned arrowheads and did other fine work.

     This kind of process, then, makes for intraspecies variability, but in doing so actually increases the precision of adaptation. Multiple genetic choice in regard to camouflage characters, to hinder recognition of a species by predators, as we judge to be the case in Nabours' grasshoppers, and, conversely, a high variability in pattern to promote the recognition of individuals by others of the same species, as perhaps in killifish (Lebistes), and in mammalian scents, voices and faces, would come within the same general category of phenomena.

    Another major factor in variability is, as Wright has analyzed with such remarkable insight, the extent, degree and duration of sub-division of a species into semi-isolated populations, and the effective breeding-size of these local groups. The smaller the groups the more does accident, unrelated to value, determine survival of one gene rather than another within the group, and the less precise can be the adaptation of the mean, not to speak of the individual. At the same time, in the large, these local experiments in variation should operate to further the possibilities of evolution and of a more far-seeing adaptation of the species as a whole, through giving a wider range for selection and recombination between the differentiating groups.

    Whatever the reason may be, whether because of the accidents born of small numbers in local groups, or -- probably less often -- because a given rare character or combination of characters that is useful has happened to become tied up with a variation that would otherwise be detrimental, considerable upsets in the mean values of given characters do sometimes occur in evolution. And these may cause a loss in the precision of adaptation previously attained, even where the optimum or standard of selection for the given character may still remain largely the same.

     Evidence of the occurrence of such events is to be found, for instance, in the presence of "repeat" or "duplicated" sections in the chromosomes of the normal Drosophila, and even in its X-chromosome. These must have altered the dosage of many genes at once, and, in the case of those in the X, many of the delicate dosage-compensation mechanisms must have been much disturbed simultaneously, yet these aberrations were not too disadvantageous to survive and become established. Following such occurrences, however, the population must have settled down to a prolonged process of repair and readjustment, that perhaps allowed the aberration to be worked out to the net advantage of the species in the end. Thus, there is not only the patient "whittling" process in evolution. However, as judged by the mass of results previously reviewed, this process must be so much the more usual procedure than the other that the precision of adaptation finally attained thereby is on the whole far greater than most biologists have hitherto realized.

To sum up in a nutshell the argument arising out of the consideration of dosage compensation, the subject which has consumed the major portion of our attention in this paper, we would say that this phenomenon is an illustration of the fact that organisms, starting with similar needs but different genetic make-up -- that of the male and of the female, respectively -- nevertheless tend to develop the same characters, even to the utmost degree of precision, in spite of the fact that, to attain this end, they must evolve still more different genetic and developmental mechanisms.

    For, since the female has two doses of sex-linked genes and the male but one, special compensators must be evoked, working to different degrees in the two sexes, that make the grand end-resultant of these complex biochemical activities remarkably the same, to a degree far beyond the limits of our unaided vision. This striking convergence in the effects attained in the two sexes, despite the differences in the mechanisms of attaining them, a result which so resembles that of purpose, can be rationally explained only on the ground of selective advantage which, through elimination of the unfavorable attempts and multiplication of the favorable, produces effects similar to those of conscious profiting by experience.

     Essentially the same kind of phenomenon is seen in the resemblances in character between two species, like the case of bristle resemblance in Drosophila melanogaster and simulans, where again the same end is attained by means of demonstrably dissimilar genetic and developmental mechanisms, as is shown by the upset in bristle formation in the hybrid between them.* [Footnote: See qualifications expressed in footnote on pages 214-215.] Again, it is as if a goal were aimed at, regardless of the means of its attainment. And since this goal is in fact attained, it is necessary to infer here too that it has had, and in all probability still has, a value for the organism, and that to depart from it is disadvantageous.

    It is therefore justifiable, in this sense, to regard the organism teleologically. And since its adaptations are found to be so far-reaching and precise, we are entitled to look for a function or functions of some sort in practically every feature found, which has attained widespread, long-term existence in a species. Owing to the complexity and the reconditeness of many of the features of any organism's physiology, biochemistry and ecology, however, the functions in question will often be far from obvious to our limited observations and imaginations. However, we may usually assume that, until we have found them, we have not yet properly done our job of understanding the organism. In the meantime, this realization itself should often be of help to us in our effort to understand it, for we may thereby be guided to see in which directions some of the gaps in our knowledge still lie.

    Thus, the very fact that the organism can do without the tonsils or appendix should have served to stimulate further search for the long unrealized functions of these organs. And the fact that in so many common infections the temperature is raised should early have made us suspicious that, despite the inconvenience thus entailed, an adaptive mechanism was involved here, which could sometimes be taken advantage of, and guided, rather than nullified [say, by taking aspirin]. At the same time it should also be remembered, wherever there is a host-parasite or even a symbiotic relationship between two species, that a given development in one, say the host, may, like a gall, represent an adaptation achieved for the benefit of the other -- the parasite -- rather than of the host itself. And by a piling up and interlocking of the adaptations and counteradaptations in such cases, the problems involved may become exceedingly intricate.

    The above illustrations will also serve to illustrate another point not previously mentioned here. That is, that many "final" characters, like many genes themselves and many intermediate processes of development and physiology, serve more than one function, and may even be, at the same time, actually disadvantageous in other respects, but with the disadvantages outweighed and perhaps in part compensated for by special means.

     In other words, even when we seem to have found "the" function of a part or process, we may be far from finished in our search since there may be still other more or less important consequences of it for the life of the organism. In a sense, each part represents a compromise between advantage and disadvantage, if only because of the fact that simplicity is usually an advantage, for reasons of economy in construction, maintenance and repair, and for the minimizing of hazards, while on the other hand a given use may often far outweigh the drawbacks of the greater complication. This all goes to indicate how tremendously interwoven and intricate the life processes of any higher organism are, and how the weighing of the benefit of a given feature by the criterion of selection may often lead to a result quite different from what we might have guessed.

    Of course all this evidence of adaptation does not connote that natural selection has acted like an omniscient guide. It can seldom take steps that depart very far from a purely immediate advantage, even though it is helped to a limited extent by the somewhat more far-flung trials engaged in by small semi-isolated local groups, as previously mentioned. But some developments, which would require much longer, more elaborate trials before their advantage materialized, will never have a chance to be carried out. As an illustration we may cite the absence of all wheel mechanisms in organisms, occasioned by the fact that such an arrangement cannot be arrived at by a series of small steps giving intermediate stages which themselves are viable and useful.

    Conversely, other developments, which for some temporary reason confer an immediate advantage, will become incorporated despite their lower, or even negative, long-term evolutionary value. So, for example, "mistakes" are made like placing the optic nerve fibres in the path of the light, at a time when the eye is still so imperfect that this does not matter so long as the function of distributing nerves to the individual retinal cells is achieved. Once incorporated, however, such an arrangement cannot easily be revoked, since around it, as around a keystone in an arch, there have in the meantime been placed other structures, now indispensable, which would collapse if it were changed, unless very special substitute arrangements were found. This, then, is a type of developmental correlation which cannot readily be overcome.

     There is, however, no contradiction here involved to our earlier thesis that the characters representing branched effects of the same gene can be modified separately from one another. For in cases like the present one the characters in question are not caused by different branches of one primary process, but by the direct dependence of one from the other chainwise, without branching, -- i.e., in what is in effect a straight line. Thus, in our example above, the only use we can now expect to find in the given peculiar arrangement of nerve fibers is that it serves in us as a necessary ontogenetic basis for the development of other features, in themselves useful.

    Considerations similar to the above apply to many other facts of recapitulation in general, e.g., to the development of gill slits, mesonephros, etc., in human embryos. Despite this persistence of such relics, however, the course of development itself (as well as of intermediate physiology) can undergo adaptive changes, while leaving the final product relatively unchanged, so that in the end entirely new developmental (or physiological) routes to the attainment of given adult structures (or reactions) are sometimes arrived at. A striking illustration of this is seen in the substitution of "complete metamorphosis", with its inserted pupal stage, in higher insects, for the earlier "direct" processes of development. But that this substitution was itself gradual can also be seen, from the intermediate situation still persisting in some forms.

    The organism, then, is fearfully and wonderfully adapted, to a marvelous degree of nicety, but it has been constructed bit by bit, by accretion and patching, as one minute, functionally operating step was superimposed upon another. And though no unchanged part of the original fabric may any longer be present, much of the earlier, and now superseded, pattern still remains, in some way overlaid.

     Therefore it is probable that much of the present structure and working is far more complicated than would have been necessary to achieve the present results, if only they could have been planned ab initio for the most efficient attainment of the ends which are subserved today. This, however, would have required intelligence, involving long-range foresight and design of an order far beyond any which we ourselves as yet possess. The adaptation of organisms, in other words, is most meticulous and all-embracing, yet close scrutiny shows it most wanting in those respects which would have involved distant prevision, rather than short-range trial and error.

     To be sure, natural selection itself does attain longer range when choosing among much diversified and widely separated groups, and operating over longer periods, but in so far as this happens it finds itself more limited in its range of choice, since there are fewer groups to choose between as the categories in question become more general ones. Moreover, they are limited to presenting, for these higher processes of choice, or "courts of appeals", such patterns as were themselves developed through the selection processes of shorter range.

    Despite these limitations on the operations of selection, the organism is understandable in its present workings, and in its past origination, only if its ends are taken into consideration, that is, if it is regarded not only "mechanistically" but also "teleologically", in the sense previously set forth. It is the fact of its being so constituted as to work toward ends, -- i.e., toward results favorable for its own multiplication and perpetuation, -- that chiefly distinguishes an organism from an inanimate thing, even though its operations are not guided by that actual consciousness of their ends by which they so often seem to be.

     However, this also is the reason why, when what we call consciousness does at last enter the picture, it too can be utilized by organisms in the further attainment of their ends. Unlike other ends, however, these ends are at the same time beginnings, since the courses of their lives are cyclic, with the difference from other cycles [being] that these multiply as they proceed, and that they are subject to mutations which, if they serve to further these ends, themselves become reproduced.

    All the above will perhaps help to make clear the essential correctness of the aphorism of the American zoologist W. K. Brooks, to the effect that "The essence of life is not protoplasm but purpose". We must, however, understand the term "purpose" here as involving only an analogy. If we do so, we should have reopened to us many clues which the so-called mechanistic biologists wastefully disregarded, when they started us on that otherwise promising course the strict following of which has at last put us in the position of realizing that teleology, of a sort, is a part of "mechanism" after all.

The attempt to interpret the structures and reactions of living things, without reference to their adaptive significance, is a consequence of confused thinking, stemming from the opposition to the Darwinian principle of natural selection. As shown by illustrations from Drosophila, mutations provide abundant evidence of the adaptiveness of existing "normal" features, as well as material for determining how these features operate in furthering the life of the individual or the species. Moreover, a study of the effects of changing the number (dosage) of genes, and of the nature of the dominance of normal to mutant genes, leads to the conclusion that not only is each normal feature of adaptive value, in a qualitative sense, but that its exact grade of development, quantitatively considered, represents the optimum value for it under the conditions that prevailed during the period when it attained that value. For the activity of each gene is usually so near its saturation level as to result in a high stability of its effects, despite the presence of disturbing factors. It is this which ordinarily results, secondarily, in its appearing to be dominant over its more weakly acting ("hypomorphic") mutant allele.

    Beyond this, a study of the expression of mutant genes which exist in two doses in the female and one in the male (those in the X-chromosome), shows that special genetic mechanisms have been evolved for compensating for the effects of these dosage differences, so as to make the amount of effect in the two sexes much more nearly the same than it otherwise would be. The mechanisms involve the presence of other, modifying genes in the X-chromosome, that may be termed "compensators". These, in their double dose in the female, reduce the activity of her two doses of the primary gene, so as to make the effect like that of the one dose of primary gene and one of compensators in the male.

      It is shown that actually these mechanisms must have been evolved for the function of equalizing the effects of the normal genes rather than the mutants, even though we can seldom be aware of this equalization in the case of normal genes, because of their above-mentioned near-saturation level. In other words, the stabilization of expression attained through the high level of activity of the normal gene does not by itself equalize the effects in the two sexes sufficiently to satisfy the organism's needs, despite its giving effects in one versus two doses, which are not to us perceptibly different. For the compensating mechanisms which have been evolved serve to equalize them much more exactly. This shows that differences in visible characters, which are of far finer grade than can be detected by our own powers of discrimination, are nevertheless of adaptive significance in the organism's struggle for existence, -- i.e., they are of selective value.

    Experiments are cited which show that these compensating reactions are no mere consequences of the sex difference itself, that a given gene usually has a whole system of interacting compensator genes, and that the systems for different genes are different from one another. Thus each must have evolved through an accumulation of several or many steps, subliminal to our perception. The second-order magnitude of this selection pressure is illustrated by evidence of Lamy and the writer, indicating the incompleteness of the compensation for genes in parts of the X-chromosome of D. pseudo-obscura which have more recently been acquired by the X from another chromosome. The objection is considered that, instead of subliminal selection for the characters under observation, the results may be due to the developmental correlation of these characters with other, more important but unobserved characters, in regard to which major differences were caused by the 1-dose: 2-dose relation of the sexes. It is shown that this interpretation encounters serious difficulties, based on the lack of completeness of developmental correlation. Thus, we are led back to the conclusion that selection must, in fact, have dealt subliminally with the actually observed characters.

    Further evidence of the adaptive value of normal characters, and in fact of the value of the precise grade of expression which they exhibit, is found in their frequent persistence virtually unchanged over millions of years, and even in different species having similar needs. It is shown that the frequency of mutation is high enough to have prevented such stability, had selection not actively favored the given norm. The causes of deviations from high stability are considered. A formula is given (p. 217) whereby the strength of such selection can be calculated from the existing variability within a large, persistent and interbreeding population, when the amount of variation arising by mutation per generation has also been determined. Though data of the latter nature are as yet insufficient, it can in the meantime be concluded that the prevalent assumption is incorrect which would regard deviations within the so-called "normal range" (e.g., within the central 8 or 9 deciles of a population) as being of negligible detriment to individuals possessing them.

    It is concluded that long persistent morphological, physiological or biochemical features, characteristic of large populations, should in general be assumed to have, or to have had, adaptive value, under the conditions in which these groups evolved, and that the degree to which they are normally developed is also of adaptive significance. Failure to find this significance, or function, is to be taken as an indication of our ignorance, and of the need for a deeper understanding of their workings. This justifies us in applying a "teleological" viewpoint here, which is in no wise mystical. In medicine, it should, moreover, lead to caution in attempting the alteration of standard bodily reactions, or the extirpation or alteration of structures normally present, until the functions of these parts and processes, and of their normal degree of development, have been well determined. At the same time, it must also be recognized that selection itself has been subject to grave limitations. Some of these have been discussed above. The organism is not perfect in any absolute sense, otherwise there would be no use in medicine, or in any form of artificial "interference" with nature.