1Contribution from the Zoological Laboratory of the University of Illinois No. 50.
2Tice, S. A. 1914. A new sex‑linked character in Drosophila. Biological Bulletin, vol. 26, pp. iii to 230.

The selection experiment described in the present paper was made with a view to testing the germinal uniformity as regards the distinguishing characteristic in a recent mutant, the 'bar eye' race of Drosophila. In this race the ommatidia are reduced in number and the facets are restricted to a vertical band or 'bar' as shown in figure 1. The characteristic appeared in a single male during 1913 (Ticc, '14)2 and the whole 'bar eye' stock is descended from this individual. The race has undergone no apparent change during the two years of its existence. Our material was obtained in January 1914 through the kindness of Prof. T. H. Morgan.

There is a pronounced sexual dimorphism in the number of facets, the males averaging 98.03 and the females 65.06. In every case in the present paper the female number is transformed to the male basis by multiplying it by 98.03/65.06 = 1.51.

There is often a slight difference between the number of facets in the right and that in the left eye. For one hundred individuals the average difference was 0.245 per cent in favor of the left eye. This difference is obviously not significant. The right eye only is given in the present records.

The number of facets seems not to vary with the length of the period of development. In five broods counts were made of the facets of the earlier emerging individuals and compared with those of the later emerging ones. The averages for the two are approximately equal.

A sample of five hundred individuals, 250 males and 25€ females, taken from the general population of the mutant showed a range of variation in the number of facets from 45 to 182 with a mean of 98.03 0.73 and a standard deviation of 24.30 The variation curve is shown in figure 2. The average for ten individuals of the normal wild race was 701.1 facets.

Unselected stock was carried as a control through the period of the experiment. There was practically no change in the number of facets, the average of 250 individuals at the beginning of the experiment being 98.04 and of the same number at the end of the experiment 98.03. There is thus no change in facet number in the absence of selection.

In making the first selection, food containing eggs and larvae was removed from the culture of the 'bar eye' stock and placed in glass vials. Every twelve hours the individuals which had emerged from the pupal cases were slightly etherized and an estimate was made of the number of facets under the low power of the microscope. Males and females with high or low numbers were selected out, high being mated with high and low with low. Each pair was placed in a small bottle with sufficient food to last until all the offspring of the first brood had reached the adult stage. When larvae began to appear in the bottle the parents were killed by etherization and an exact count was made of the facets. As the offspring emerged in the adult form estimates of facet number were again made and high selections were made in the 'high' lines and low selections in the 'low' lines. The final exact facet counts were in all cases made in killed individuals.

The experiment has proceeded far enough so that data are complete for three successive selections in each of three 'high' lines, called A, B, and C, and in each of three 'low' lines, called D, E and F. Fifty individuals, 25 males and 25 females from a single pair, were measured for each generation in each of the lines, with the exception of the third generation in line B where only forty‑six were available.

The data for the individual lines are given in tables 1 to 6 and a summary of the six lines is given in table 7. Figures 3, 4 and 5 give in graphic form the course of the selections, each figure combining a 'high' with a 'low' selection line. The mean values, the extreme variates and the mid‑parental values are here represented in diagrammatic form for each of the selections.

In all eases there is a significant shifting of the mean as a result of selection. The general population mean of 98.0 is changed in plus line A to 108.7 by the first selection, to 127.5 by the second and to 135.5 by the third. In plus line B the corresponding figures for the three selections are 110.1, 128.6 and 141.9, and in plus line C, 116.9, 133.5 and 141.0. In minus line D the general population mean of 98.0 is changed to 88.3 by the first selection, to 85.5 by the second and to 81.7 by the third. In minus line E the corresponding figures for the three selections are 93.9, 89.6 and 84.8 and in minus line F, 94.6, 89.8 and 84.7.

As a result of the three minus and the three plus selections the mean value of the individuals in the plus lines (139.5) is greater than the highest extreme individual of the three minus lines (137.0) and the mean of the three minus lines (83.7) is lower than the lowest extreme individual of the three plus lines (89.0).

The coefficients of variability as given in table 7 show on the average a slight decrease in the selected stock as compared with the general unselected population. This difference however can­ not be considered as significant because of its irregularity and slight amount.

The progression of the mean is much more rapid in the plus than in the minus lines. The averages for the three plus selec­tions give respectively increases of 13.9, 18.0 and 9.6.The corresponding decreases in the minus lines are only 5.7, 4.0 and 4.6.

Regression toward the mean of the unselected population decreases with successive selections. (Table 8.) The average regression in the three plus lines is 0.67 for the first selection, 0.60 for the second and 0.57 for the third. For the minus line the corresponding figures are 0.77, 0.64 and 0.60. On the other hand regression toward the mean of the parental generation increases with successive selections. (Table 9.) The average regression in this respect in the plus lines is 0.67 for the first selection, 0.72 for the second and 0.85 for the third. In the minus lines the corresponding figures are 0.77, 0.81 and 0.83. The decrease with respect to the mean of the general unselected population indicates that there is real progress during the successive selections. The increase with respect to the parental generation indicates that the effectiveness of the selection decreases with successive selections and that there is probably a limit to the number of effective selections. It seems probable that continued selection would not be able to change the number of facets in the mutant stock to that of the original stock from which it was derived.

The data presented thus show that selection in the 'bar eye' race of Drosophila is effective both in increasing and in decreasing the number of eye facets. As a result there can be no doubt of the existence of differences in the germinal constitution as regards this characteristic among the individuals of a generation. This fact is of special interest because the origin of the race by sudden appearance in a single individual is known and is of recent occurrence. Furthermore the behavior in crosses with the normal wild race shows that the mutant differs from the normal wild race in but a single Mendelian factor. If this were the only germinal factor involved in facet number we would be compelled to conclude that we have a case of variability in a unit factor. As far as the present selection data go such a provisional hypothesis would not be contrary to the facts. It seems more probable however that facet number in the normal wild race is represented in the germinal constitution by more than one factor and that the modification occurring in the production of the 'bar eye' race involved only one of these factors. That this factor is a most important one is of course indicated by reduction from an average of 701.1 facets in the original stock to 98.0 in the 'bar eye' stock. Selection, then, may have an effect because of variability or because of lack of homogeneity in the race as regards these other factors without regard to the 'barring' factor itself.

Three possibilities are thus open as regards the explanation of the effect of selection in this case. First, the 'barring' unit factor may be variable or may have varied since its first appearance in 1913. Second, the 'bar eye' race and by inference the original normal eyed stock from which it was derived may contain additional germinal factors affecting facet number and these additional factors may be variable. Third, the 'bar eye' race and by inference the original normal eyed stock from which it was derived may not be homogeneous with regard to these additional factors. Different factorial combinations may be present in different individuals. Selection in this case would segregate the 'high' combinations of factors on the one hand and the 'low' combinations on the other, yielding finally two homogeneous races in which further selection would have no effect. The 'highest' possible combination of factors as well as the 'lowest' possible may not exist in the original sample of the general population, but by Mendelian recombination it would finally appear.

While the data so far obtained do not enable us to decide which one of these three possibilities or which combination of them is to be considered as active in this case there is some evidence to support the view that the third is at least partly responsible. The increase in regression of the mean toward the mean of the parental population with each successive selection indicates an approaching limit to the effectiveness of selection. This is what we would expect in a population that is heterogeneous as regards factorial composition. If the unit factors themselves do not vary, selection must soon cease to have further effect.


  1. Three successive selections for high number of facets in the 'bar eye' race of Drosophila increased the mean number of facets from 98.0 to 139.5.
  2. Three similar selections for low number decreased the mean from 98.0 to 83.7.
  3. The lowest individual, 89.0, in the 'high' lines after three selections is higher than the mean of the 'low' strains, 83.7, and the highest individual of the 'low' lines, 137.0, is lower than the mean of the 'high' strains, 139.5.
  4. Significant progress was noted in each of the three selections in both 'high' and 'low' lines.
  5. There are some differences in variability in the different generations but no significant change is proven.
  6. Regression toward the mean of the general unselected population decreases with successive selections.
  7. Regression toward the mean of the parental populations increases with successive selections. This increase makes it improbable that 'bar eye' stock can be raised to the original level by continued selection.
  8. It is apparent from these data that individuals in any generation differ as regards germinal constitution.
  9. If this difference in germinal constitution is solely in the unit factor concerned in 'barring,' variability in this unit factor must be assumed.
  10. It is however more probable that there are other factors concerned in facet number. In that case the selection effect may be due either to variability of single unit factors or to presence of original differences in factorial composition. That it is due in part at least to the latter is indicated by the increase in regression toward the mean of the parental generation with successive selections.