An Experimental Study of Group Selection Author(s): Michael J. Wade Source: Evolution, Vol. 31, No. 1 (Mar., 1977), pp. 134-153 Published by: Society for the Study of Evolution Stable URL: http://www.jstor.org/stable/2407552 Accessed: 02/10/2008 20:16 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=ssevol. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected]. Society for the Study of Evolution is collaborating with JSTOR to digitize, preserve and extend access to Evolution. http://www.jstor.org AN EXPERIMENTAL STUDY OF GROUP SELECTION MICHAEL J. WADE Department of Biology, University of Chicago, Chicago, Illinois 60637 Received November 17, 1975. Revised March 25, 1976. Darwin (1878) suggested in The Origin of Species (p. 230-233) that selection between families or communities had given rise to sterile castes in the social insects. He argued that sterility could not be favored directly by selection among individuals but that parents or communities might benefit, in terms of reproductive success, from the presence of sterile individuals. It is clear from his arguments that Darwin believed that selection among groups could result in a change in the types of individuals within a species. It was much later that Wright (1931) perceived intergroup selection as one of the primary forces controlling evolutionary change. He stated that the only practicable method of bringing about a rapid and non-self-terminating evolutionary advance would be through the subdivision of populations into isolated and differentiating groups, among which selection could be practiced. Haldane (1932) and Wright (1945) both considered the case of the evolution of altruistic behavior where individual selection acts against the altruist, but the altruist is favored by group selection. Group selection became a more controversial subject with the publication of Wynne-Edwards' book Animal Dispersion in Relation to Social Behavior (1962). It was Wynne-Edwards' thesis that group selection was a general phenomenon and was responsible for the evolution of mechanisms which regulated population size. Wynne-Edwards believed that population size was limited at some level below the "carrying capacity" (Errington, 1934) of the environment by means of individuals within the population withholding reproduction. To the extent that an individual withholds reproduction that indiEVOLUTION 31:134-153. March 1977 134 vidual suffers a loss of fitness, where fitness is measured by reproductive success. It was for this reason that WynneEdwards thought individual selection could not be responsible for the evolution of the mechanisms for withholding reproduction and therefore, some form of group selection was required. Territoriality, ritualized aggression, the establishment of hierarchies and other social orderings, the onset of senility, and the optimum reproductiveeffort were some of the regulatory mechanisms of populations that Wynne-Edwardsproposed had evolved by means of group selection. The regulation of population size has been and continues to be an interesting and central problem in the field of population ecology and several authors, among them Andrewartha and Birch (1954), Williams (1964), Hamilton (1966), and Amadon (1964) have cogently discussed the evolution of the above mentioned regulatory mechanisms by means of individual selection. The focus of much of the controversy has been the question of whether or not group selection is required to explain the evolution of a particular character. In general, the viewpoint of most ecologists and theorists has been that individual selection is a faster and more efficient form of natural selection than is group selection for the following reasons: 1. In iteroparous species the generation time of the individual is short relative to the turnover time of the population and therefore selection proceeds faster when the unit of selection is the individual. 2. The numbers of individuals greatly exceed the numbers of populations. If the units of selection are numerous it is less likely that chance events will significantly influence the outcome of selection. There- GROUP SELECTION fore selection is more efficient when the unit of selection is the individual. 3. Just as individual selection requires that there be differences between individuals, group selection requires that there be differences between populations; however, the origin of variation between populations and the maintenance of that variation in the face of even low levels of migration has heretofore been considered a major obstacle to the operation of group selection (Maynard-Smith, 1964). Reasoning such as the above led Williams (1966) to formulate a principle of parsimony which appears to be generally adopted at the present time. This principle states that if the evolution of a trait can adequately be explained by means of individual selection, there is no need to invoke group selection. The principle of parsimony clearly restricts the use of group selection to those cases in which it operates in a direction opposite to individual selection. Serious theoretical inquiry into the process of group selection was stimulated in part by MacArthurand Wilson's Theory of Island Biogeography, published in 1967. One of the concepts espoused by MacArthur and Wilson was that a species could be viewed as an array of populations which are formally equivalent to islands. When populations are viewed in this way, the processes of extinction, dispersion, and recolonization assume a role of central importance in the investigation of many ecological problems. During the last five years, several authors, among them Levins (1970), Boorman and Levitt (1973), E. 0. Wilson (1973), Levin and Kilmer (1974), D. S. Wilson (1975), Gadgil (1975), and Gilpin (1975), have directly addressed the genetic problems of group selection. All of these authors implicity or explicity consider the case of individual selection acting in one direction and group selection acting in the opposite direction to be the one of major importance. The general conclusion has been that 135 group selection will be a significant force in natural populations only under a very restrictive set of conditions. The purpose of this paper is not to criticize the aforementioned theoretical works but to present the results of an experimental investigation of the process of group selection. The basic questions which underlie the research to be presented here are: Can the differential extinction and/or differential proliferation of populations cause genetic change? And if so, what are the parameters which determine the rate and the extent of the changes? Groupselection is defined as that process of genetic change brought about or maintained by the differential extinction and/or proliferation of populations (Wright, 1945, 1956; Wynne-Edwards, 1962; MaynardSmith, 1964; Williams, 1966; Lewontin, 1970). (In keeping with this definition, I will not consider selection between species [Van Valen, 1975] or between ecosystems [Dunbar, 1960], both of which have been called "group"selection.) The gene frequency changes caused by group selection, as is also true for individual selection, will consist of changes in the genetic makeup of individuals within populations. Any trait which significantly increases the probability of survival or proliferation of the population and at the same time increases the relative probability of survival or reproductionof the individuals within populations will be selected for by both group and individual selection. In such a case the rate of change of gene frequency is expected to be significantly greater than the rate of change when either group selection or individual selection is acting alone. These two different processes may or may not result in the same genetic equilibrium and it is their interaction which will determine the rate of approach to, and the position of, the equilibrium. EXPERIMENT I The purpose of this experiment was to investigate the genetic effects of a process 136 MICHAEL J. WADE of differential extinction and recolonization of populations of the flour beetle, Tribolium castaneum, and the trait chosen for selection was the number of adult beetles in a population. The ecology of Tribolium has been well studied and much is known about the factors governing changes in its population size (King and Dawson, 1972; Mertz, 1972; and Sokoloff, 1974). It was believed that this large body of information would facilitate the interpretation of any changes in population size which might occur during the course of the experiment. The developmental time is an important factor in determining both the turnover time of the population and the rate of increase of population numbers. T. castaneum is a holometabolousinsect, passing through egg, larval, and pupal stages before becoming an adult. The total duration of the developmental period, from egg to mature adult, is approximately 33 days under the conditions of 29 C and 70% relative humidity (Park and Frank, 1948; Young, 1970). These life stages are generally spent in and around flour or other stored products. Although the intrinsic rate of natural increase, r, (Mertz, 1970) varies with the climatic conditions, an r of approximately .10 beetles per day can be achieved under the standard conditions mentioned above, and this represents a doubling in population size once every seven days (Leslie and Park, 1949). In an experimental study of interspecies competition over a wide range of climatic conditions, Park (1954) found that a significant correlation existed between a species proportion of adults in the early months of husbandry and the time to elimination of that species from mixed species populations. The more T. castaneum adults relative to the total present in the second and third months of competition the longer the time to extinction of the T. castaneumpopulation. In a study of single species populations of T. confusum on an unrenewed resource, McDonald and Stoner (1968) established several replicate populations from each of several genetic strains. Within genotypes, large numbers of adults in the early months implied a long time to extinction. The opposite correlationheld in a between genotypes comparison, i.e., large numbers of adults implied a short time to extinction. Nathanson (1975) has shown that the number of T. castaneum adults at 60 days is a good indicator of the time to extinction of the T. castaneum population when in competition with T. confusum on an unrenewed medium. In this case low numbers of adults at day 60 predict a long time to extinction, while high numbers at day 60 indicate a relatively early extinction. These experiments suggest that population size may be associated in a causative way with the probability of extinction and that the likelihood of extinction and the time to extinction can be evaluated at a relatively early stage in the population's history. In addition, Ziegler (1972) discovered that 80 to 90% of T. castaneum adults if allowed to emigrate would do so six or seven days after eclosion. This result implies that T. castaneum individuals are prone to dispersion at an age of approximately 3 7 days. Materials and Methods Adult numbers in a population at 3 7 days was the criterion chosen in studying selection on the basis of earlier empirical studies. In this study a population is a founding propagule of 16 adults, whose sex ratio is determined by chance, in an 8 dr shell vial containing 8 g of a flouryeast medium (95% by weight Elam's stoneground whole wheat flour and 5% by weight Fleischman'syeast type #7B). The standard techniques used in the laboratory of Dr. Thomas Park were rigorously followed in the handling and censusing of the experimental populations (Park, 1948). A stock culture of high genetic variability was created by mass mating 12 adult males and 12 adult females from GROUP SELECTION each of four "inbred" strains of T. castaneum (for more information on the "inbred" strains see Park et al., 1961; 2 sexes X 12 adults per sex X 4 strains = 96 adults total). The F3 generation of this mass mating was the source of the founding propagules for generation 1 for all treatments. The propagule size of 16 adults was chosen to minimize the effects of inbreeding (Crow and Kimura, 1970) and to produce a manageable number of adults per population. The effect of inbreeding upon numbers of adults will be discussed briefly in ExperimentII below. The experiment consisted of four treatments with 48 populations per treatment (4 treatments X 48 populations per treatment = 192 populations total). One hundred and ninety-two groups of 16 adults were chosen at random from the common stock described above and one population was founded with each group of 16. Each population was assigned a position in one of six racks in such a way that the populations were evenly distributed by treatment among the racks but randomly positioned within the racks. The racks were placed on one shelf in a dark incubator and systematically rotated, as racks and by position, once a day to avoid the effects of temperature stratification. The incubator was successfully maintained at essentially 29 C and 70% relative humidity throughout the course of the experiment. After 37 days the populations were removed from the incubator and a census of adult beetles was taken for every population. Immediately following the census, selection was imposed upon the populations accordingto treatment. The four selection treatmentsare labelled for convenience: Treatment A: Selection by differential extinction and recolonization of populations, i.e., group selection, for higher numbersof adults per population. Treatment B: Selection by differential extinction and recolonization of populations, i.e., group selection, for lower numbers of adults per population. 13 7 Treatment C: No group selection; individual selection within the populations was allowed to determine the numbers of adults. Treatment D: Selection and recolonization of populations by means of a table of random numbers,i.e., a random extinctions process. In Treatment A, that population with the largest number of adults at the 37-day census was selected and divided into as many groups of 16 adults as possible. (Remainders less than 16 were discarded along with any individuals who appeared "unhealthy," e.g., missing a limb or having a split elytra. The number of "unhealthy" individuals never exceeded three of four per population of size 160, and some of these individuals appeared in all treatments. Many of these "unhealthy" individuals are the result of unsuccessful cannibalistic attack during the pupal or teneral stage. For these reasons, the effect of this kind of selection against manifestly "unhealthy" individuals is of little or no consequence to the interpretation of the experimentalresults.) One new population was founded with each group of 16. The population with the second highest number of adults was then chosen and likewise divided into propagulesof 16 adults. Group selection for high number of adults was continued in this manner until 48 new populations had been established. In Treatment B, the procedure was identical to that described for A except that the populations with the lowest numbers of adults were selected and divided. More populations from Treatment B are required to found 48 new populations than are populations from A. (Throughout the course of the experiment, the B founding populations were approximately three to five times the number of the A founding populations.) Treatment D was similar to A and B, but in this case a table of random numberswas used to select the populations. Treatment C was designed to be a control treatment which would determine the effect of individual selection upon adult numbers at 37 days. In this treatment one MICHAEL J. WADE 138 EXPERIMENTAL DESIGN COMMONSTOCK- e B = LOWGROUP C =NO GROUP SELECTION SELECTION SELECTION SELECTION 48 POPULATIONS 48 POPULATIONS 48 POPULATIONS 48 POPULATIONS 16ADULTS/POP 16ADULTS/POP 16ADULTS/POP 16ADULTS/POP A= HIGHGROUP D=RANDOM TREATMENTS. I I 37-DAYINTERVAL DATA GATHERED NUMBER OF ADULTS IN EACH POPULATION /RANDOM | SELECTION K 16 16 C 16 16 48 POPULATIONS 16ADULTS/POP REPEATED 8 TIMES | 16 16 V 0 X 16 16 16 T 16 16 16 48 POPULATIONS 48 POPULATIONS 48 POPULATIONS 16ADULTS/POP 16ADULTS/POP 16ADULTS/POP I FIG. 1. Schematic outline of the design of Experiment I. See text for further explanation. group of 16 adults was chosen at random from each of the 48 C populations and a new population was founded with each group of 16. In this manner, each population in one generation gave rise to exactly one population in the succeeding generation for C. There could be no group selection in treatment C because there was no differential extinction or proliferation of the populations. Any changes in the mean numbers of adults at 37 days in C could be attributed to individual selection operating within the populations. A schematic diagram of the selection program is presented in Figure 1. Conclusions were drawn from the experimental data by a between-treatments statistical comparison of the distributions of adult numbers. In deriving these conclusions it is reasonable to assume that individual selection is operating in the same direction within the populations of A, B, and D, as it is within the populations of C, at least in the initial stages of the experiment. Results The censusing and selection procedures were continued as described for nine 37day intervals, hereafter called generations. The experiment was terminated at generation nine because the mean number of adults in treatment B, the low group selection populations, had declined to nearly the lower limit of 16 set by the experimentaldesign. The census data indicated an obvious departure from normality and for this reason the non-parametric Kruskal-Wallis rank sum test was employed in all betweentreatmentscomparisons. The mean, standard deviation, and co- GROUP SELECTION 139 and the mean of treatment B was less than that of either C or D. However, the differences were not statistically signifi<00 cant (see Table 1 and Fig. 2). LL ~~~~~~~~~~~~~~~~~A After two generations of selection, there ot150 were highly significant differences between 0 00 treatments. In generation 3, the mean Z number of adults in the A populations exceeded the mean of the control (C) popu0 lations by more than 40 adult beetles. A 4 2 5 6 3 7 8 9 GENERATIONS test of the means using the non-parametric FIG. 2. The mean numbers of adults for the Kruskal-Wallis rank sum statistic gave a High (A) and Low (B) group selected populaP < .0001. Similarly, the mean number tions and the Control (C) populations of Experiof adults for treatment B was significantly ment I for all generations. Each point is the less than the mean for C (P < .0005). mean of 48 observations. Treatments A and B differed at this time by an average of over 70 adults per popuefficient of variability (%) for all generalation. tions of all treatments are given in Table These differences persisted and became 1. Figure 2 depicts the actual mean more extreme during the course of the population size for treatments A, B, and experiment. At generation 9, after 8 C through generation time. Treatment D generations of selection, the mean differis not presented here for treatment D was ence between the A and B treatments had statistically identical to treatment C until increased to 158 adults per population generation 7. The A populations in There were no statistical differences (see Table 1). 9 recruiting an average of were generation between treatments in generation 1. This 40 times the number of adults reover was expected because the founding propain the same the B populations cruited by gules had been chosen at random from the common stock and selection had not yet period of time! The means of the B and C treatments been imposed. In generation 2, after one generation of selection, the mean of treat- tend to decline through generation time ment A was greater than that of C or D while the mean of the A treatment fluctu300 00 250 0 - HIGHSELECTION" - NO Z LOW SELECTION l] ' SELECTION 50' " - C '. TABLE 1. The mean, standard deviation, and coefficient of variability (%) for all treatments of Experiment I. Rf Ab S.D.g C.V. x Bc S.D. C.V. x Cd S.D. C.V. x De S.D. C.V. G2 256 223 57 67 22 30 271 211 57 77 21 37 278 215 53 68 19 32 284 214 49 58 17 27 G3 G4 G5 196 178 279 58 60 66 30 34 23 122 88 108 52 42 25 43 48 23 152 137 170 46 58 61 31 42 36 137 121 168 46 52 70 34 43 42 G6 228 69 30 76 27 35 155 68 44 158 59 37 G7 219 64 29 48 29 60 106 59 56 122 53 43 G8 129 55 42 26 7 28 68 27 39 102 46 45 G9 178 57 32 a G = generation. b A - high selected lines. B = low selected lines. 20 4 18 49 41 85 69 44 64 Gla C = control lines. e D = randomly selected lines. f= mean of 48 vials. d e S.D. = standard deviation of 48 vials. h C.V. = coefficient of variation in %. As of G3 the following relationship is true at the 0.005 significance level: A > C = D > B. 140 MICHAEL J. WADE ates but does not change substantially (Fig. 2). (One should not conclude from this result that treatment A has remained genetically similar to the initial stock, for large differences between A and the maintained stock culture in fecundity, percent larval survivorship to adulthood, body weight, and female development time have been found and will be presented in a later paper.) It can be inferred from these data that individual selection (C) operated, in a way that is yet unknown, to decrease the mean number of adults per population at 37 days. Possible mechanisms responsible for this decline will be mentioned in the Discussion section, but a similar decline in productivity has been found by Dr. David McCauley (pers. comm.) for T. castaneum husbanded in much greater numbersand in discrete generations. The upsurge in mean numbersat generation 5 and the precipitous decline at generation 8 remained unexplained. None of the important climatic variables, temperature, relative humidity, or age of the medium, are unusual in any respect for the intervals of time preceding these generations. Fortunately, the fluctuations in mean population size occur in all treatments and therefore it can safely be assumed that these fluctuations do not affect the between-treatmentscomparisons within any given generation. In any generation, the deviations of the A, B, and D treatments from the control mean (C) are an explicit measure of the effect of group selection (A, B, D) relative to individual selection (C) (Fig. 3). Group selection (A) acting in the opposite direction to individual selection (C) rapidly achieved a difference in mean population size of over 100 adult beetles. Group selection (B) acting in the same direction as individual selection (C) was so able to accelerate the rate of change of population size that a mean difference in excess of 60 adults per population was produced. The apparent convergence of the mean population sizes of the B and C treatments, seen in the final generations of Figure 3, should K 140 A 0100 60 0 0-20 D 20 ir0 z o 60 Q10 w 40 LDI4O - HIGH SELECTION LOW SELECTION -- RANDOM SELECTION 2 E 3 -- 4 5 6 - E 7 E 8 , 9 GENERATIONS FIG. 3. The deviations of the mean numbers of adults of the High (A), Low (B), and Random (D) selection treatments from the mean of the Control (C) treatment for all generations of Experiment I. not be construed to imply a genetic convergence. It will be shown in a later paper that the B and C treatments differ in their respective life history parameters, and probably achieve low population numbers by different mechanisms. In generation 4, the populations of treatment D began a gradual increase in mean adult numbers relative to the control populations. This increase continued and resulted in the mean of D becoming significantly different from that of C in generations seven through nine (P < .0001). An analysis of variance of the randomly selected populations reveals the cause of this increase in the mean population size of the D treatment. The explication of this analysis is the topic of the following subsection. The Origin of Variation Between Populations.-For any generation the D treatment populations can be partitioned into "lines" on the basis of the "parent"population in the previous generation. A "line" is defined to be all the populations in a given generation that were founded by choosing propagules from the same "parent"population in the previous generation. Since each population in a given generation must be descended from some population in the previous generation, every population can be assigned to a "line." Because the entire array of 48 populations can be partitioned in this fashion, the variance in the population size of the GROUP SELECTION 141 lines variance is a measure of the variation that is available for group selection. In the D treatment, the "parent" populations for one generation are a random sample of the populations from the precedo ing generation. Therefore, it is possible to generalize the information obtained from D~~~~~~~~~~~~~~~~L the analysis of variance and make a reliable statement concerning the betweenand within-populations variation of the entire preceding generation. A similar partitioning can be made for any generation of the A or B treatments. However, in these cases, the "parent" populations are a deliberately biased FIG. 4. Analysis of variance based upon the sample of the total array of populations square-root transformation of the data for each generation of the Random (D) selection treatand one is not permitted to draw a general ment. inference from the analysis of variance. The analyses of variance for the D treatment for generations two through nine array members can also be partitioned. were performed after square root transSpecifically, the total variance for a genera- formation of the data due to the sensitivity tion can be partitioned into two com- of the F statistic to deviations from the ponents: A between lines component and normal (Fig. 4). (A logarithmic transa within lines component. These com- formation was also investigated and similar ponents have an important biological results obtained.) The between-lines variinterpretation which bears a direct rele- ance for the random treatment increases vance to the comparison of group and in- relative to the total variance through dividual selection. The within lines com- generation time. In generation two, the ponent of the variance is a measure of the between-lines component of the variance amount of genetic variability affecting represents less than 2% of the total varipopulation size which exists within a ance but by generation seven it accounts population in the previous generation. This for over 70% of the total variance. This is not an exact measure of the genetic increase in the between-populations comvariance but external environmental fluc- ponent of the variance is highly favorable tuations have been stringently controlled to the operationof group selection. by the experimental design and, as is inIt was noted above that the random dicated below, a large part of this variation extinctions treatment (D) began a slow is heritable. It is Fisher's Fundamental but steady increase in mean population Theorem of Natural Selection (1930) that size relative to the control treatment (C) the intensity of individual selection is in generation four (Fig. 3). The initiation proportional to the genetic variance within of this increase in mean numbers coincides a population. The between lines comwith the large increment in the betweenponent of the variance is a measure of the populations variance in generation four variation between populations in the pre(Fig. 4). The random extinctions treatvious generation and, as was stated in the introduction, group selection requires that ment becomes group selection for large there be differences between populations. population size once the between-populaThus, the within lines variance is a mea- tions variation reaches 35% of the total sure of the variation that is available for variance. This group selection is the reindividual selection. while the between sult of a differential proliferation of the ANALYSIS OF VARIANCE RANDOM LINES T 7 WL 4 W 2 5 4 6 GENERATIONS VARIANCET TOTAL * WITHINLINES . WL BETWEEN LINES . BL 3 7 8 9 MICHAEL J. WADE 142 TABLE 2. Analysis of variance for D = randomly selected lines (square-root transformation of data). Generation Total sum of squares Within lines sum of squares 186 158 243 216 279 291 237 311 184 127 159 129 138 89 91 162 G2 G3 G4 G5 G6 G7 G8 G9 Between lines sum of squares randomly selected populations. That is, large parent populations will found more new populations than will small parent populations and this tendency becomes heritable when the between-populations variance reaches a certain, unspecified, level. The total and component sums of squares, the F statistic, and the level of significance for each generation of treatment D are compiledin Table 2. It is of interest to determine whether or not an increase in the between-populations variance also occurredin the control treatment where there was no extinction and recolonization.In the case of neutral alleles, population genetics theory predicts that dividing a large population into several smaller isolated populations will result in an increase in the between-populations variance and a decrease in the within-populations variance. It is not known whether such a phenomenon would be observed if the trait in question were undergoing strong directional selection as was the case for the adult population size in the control (C) treatment. It was not possible to perform an analysis of variance of the C populations during the experiment because the one-to-one mapping of one 2 31 84 87 141 202 146 149 F Statistic Probability level 0.25 3.57 5.68 4.60 10.98 23.83 11.00 3.67 > < < < < < < < .05 .025 .005 .005 .005 .005 .005 .005 Degrees of freedom 2,45 3,44 4,43 6,41 4,43 4,42 6,41 9,36 generation onto the next prohibited measuring the within-populations variance. However, at the time Experiment I was terminated, eight C populations were chosen by using a table of randomnumbers and each population was divided into as many propagules of 16 adults as possible. One new population was then founded with each group of 16. This procedure, which was identical to the handling of the D populations, permitted an anaylsis of the within- and between-populations variance in adult numbers. The results of this analysis of variance are illustrated in Table 3. There was no significant betweenpopulations variance in adult numbers for the control populations. This result implies that, when there is strong unidirectional selection within populations, the isolation of populations in and of itself will not result in an increase in the between-populationsvariance for the character undergoing selection. However, a process of random extinctions with recolonization will cause an increase in the between-populations variance of a character despite directional selection against that character within the populations. It is likely that the rate of extinction and 3. Analysis of variance for the control treatment of Experiment I (square-root transformation of the data). TABLE Total Variance Between Lines Sum of Squares Within Lines Sum of Squares F Statistic Degrees of Freedom Probability 21.1 9.01 12.09 1.81 7,17 > 0.25 GROUP SELECTION recolonization will determine the rate of increase in the between-populationsvariation. That is, a high rate of random extinction would be expected to produce a more rapid partitioning of the variance than would a low rate of extinction. The intensity of individual selection within populations must also be consideredin this hypothesis. The rate of extinctions in the D treatment of this experimentvaried from 93.7% to 79.2% with a mean of 88.0%, certainly a high extinction rate. (The rate of extinctions in the A treatment was similar to that in the D treatment whereas the rate of extinctions in the B treatment averaged 62.2%.) However, the individual selection within populations was of sufficient intensity to change the mean population size from 215 adults at generation two to 49 adults at generation nine in treatment C. Thus it can reasonably be assumed that a similar phenomenon could occur when both the rate of extinctions and the intensity of the individual selection were operating at the lower levels hypothesized in natural populations. It can be inferred from the analysis of variance of the randomly selected populations (D) that a process of random extinctions and recolonization will establish the ideal and favorable conditions for group selection to occur. In this way group selection need only be a sporadic event in nature and still accomplish large genetic change. 143 about that expectation, which is important for the operation of group selection. On the other hand, one would expect that any organism whose persistence depended heavily upon dispersal and recolonization would experience significant levels of migration between populations. In order to approximate more closely the conditions expected in natural populations, a second experiment was conducted to examine empirically the effects of migration upon the process of group selection. By "migration" in this experiment, I specifically mean that new populations are founded by propagules composed of adults from more thanaone parent population. Materials and Methods In Experiment II, three levels of migration were examined for each of the treatments A, B, C, and D of the previous experiment. 1) 0% migration; propagules from selected populations exchanged no migrants (identical to Experiment I). 2) 25% migration; pairs of populations were selected according to treatment and one member of a pair contributed 12 adults and the other 4 adults to a founding propagule of 16 adults. For example, in treatment A, the two populations with the highest numbers of adults were selected. The largest population then contributed 12 adults to as many propagules as possible and the second largest population added 4 more adults to each group of 12, for a total of 16 adults per II EXPERIMENT propagule. When the largest population Wright (1931) and, more recently, was exhausted, the remainderof the second Maruyama (1970) have shown, on theo- largest population was divided into retical grounds, that populations which propagules of 12 adults and the third exchange on the average one migrant every largest population contributed groups of 4 two generationswill be genetically identical adults to these propagules, and so on. An at equilibrium. For this reason, the identical procedure was followed in treatcomplete isolation of populations has pre- ments B and D except that in B, the parent viously been considered necessary for populations were selected for low numbers group selection to occur (Williams, 1966). of adults and in D, the parent populations However, it is not the expected value of were chosen at random. Treatment C was designed to be a conthe mean gene frequency but the variance 144 MICHAEL J. WADE trol which would determine the effect of individual selection upon the numbers of adults at 37 days in populations exchanging migrants in a manner comparable to the other treatments. Thus, in treatment C pairs of populations were chosen at random each generation and 12 adults were taken from the first population of a pair and 4 from the second population and combined to form a new propagule of 16 adults. From this same pair of populations, another propagule was formed by drawing 4 adults from the first population and 12 from the second. One new population was founded with each of the two propagules. For each generation every population in treatment C was a member of exactly one pair of populations. For this reason, there was no differential extinction or proliferation of populations and there could be no group selection. Any change in the distribution of adult numbers in treatment C could be attributed to individual selection operating within the populations. 3) 50% migration; pairs of populations were selected according to treatment and each member of a pair contributed 8 adults to the founding propagule. The exact procedure in this case was similar in every respect to that outlined in (2) above. There are several other ways in which migration among populations could have been added to the process of differential extinction and recolonization but this was considered to be the simplest. The case examined here is analogous to an ecological situation in which only those populations with a particular characteristic persist and send out migrants but migrants from more than one of the surviving populations can recolonize vacant habitats. The case in which even the populations destined for extinction contribute some migrants is currently under study in this laboratory. There were 18 populations in each of the four treatments (A, B, C, and D) for each of the three levels, of migration (18 populations X 4 selection treatments X 3 levels of migration = 216 populations total). The number of populations per treatment in Experiment II was thus, approximately one-third the number of populations per treatment in Experiment I (18 vs. 48). Because fewer parent populations were required to found 18 populations than were required to found 48 populations, the variance between populations in any treatment of Experiment II would be less than that in Experiment I. For this reason, the intensity of group selection in Experiment II was not as great as in ExperimentI. It is likely that the random sex ratio of the propagules exerted a proportionately greater influence upon the progress of group selection in Experiment II relative to Experiment I as a result of the reduced number of treatment populations. A smaller proportion of the total adult population is required in the A and D treatments to form 18 propagules than is required to form 48 propagules. Hence, both random effects and the intensity of individual selection are increased in ExperimentII relative to ExperimentI. With regard to Experiment II, 18 propagules of 16 adults comprise a greater proportion of a population than 18 propagules of 4, 8, or 12 adults. Thus, in the A treatments of Experiment II (and to a lesser extent in the D treatments) the intensity of individual selection will increase with the level of migration as a consequenceof the experimentaldesign. In order to ask whether or not the differences obtained in Experiment I would persist despite migration and because of the considerable effort involved in the monthly censusing of 408 populations (192 + 216 = 408 total populations), it was decided that Experiment II should be made ancillary to Experiment I. It was believed that some information concerning the problem of migration vis a vis Experiment I could thus be obtained while postponing a thorough study of migration to a later experiment. The starting stocks for the migration GROUP SELECTION 140 120 0MGATIN0 80f 60- - .200 _ 0 '- 120 40, 6080+ loof I20 A B-B + 40 z 1202 X8 O0- ~60? 40~ 25%MIGRATION A D- D 140 -.200 '_ _ - __ ___ L '- r-~~ .__ ?_ _ _ _ _ _ 60 o 204008 10- 0140 LL_ 140i 0 640 Z G20T 100H-8060 > 1201 40~ LLJ.20- 40- ' - - 50%MIGRATION A 14 5 initial generation of Experiment I. Similarly, C, "control" and D, "random", stocks were set up from six C and seven D populations, respectively. The D stock was composed of descendents from two original populations and the C stock from six. Each of the four stocks contained approximately 1,000 adult beetles. The A, B, C, and D stocks, respectively, were the source of the founding propagules for the A, B, C, and D treatments in generation 1 of the migration experiment. Thus, the treatment means in the initial generation of the migration experiment were not identical across the board, and, for this reason, all comparisonsbetween treatments in Experiment II must be made with referenceto generation 1. D- ~ B, I 2 ~ -- ~~ 3 ~~4 5 GENERATIONS 6 7 FIG. 5. The deviations of the mean numbers of adults of the High (A), Low (B), and Random (D) selection treatments from the mean of the Control (C) treatment for all generations of Experiment II. The data for the three levels of migration are presented in the following order: 5a 0% migration, 5b 25% migration, and 5c 50% migration. experiment were composed of adult beetles collected from excess populations of the fourth generation of Experiment I. A "high-selected" stock was created by combining all the adult beetles of four excess populations from treatment A. These four populations were descended from two separate populations in generation 1. The adults were allowed to mingle in a stock jar for approximately40 minutes at which time groups of 16 adults were chosen at random and used as founding propagules for the A treatments of all three levels of migration in generation 1 of Experiment II. The B or "low-selected" stock was established by combining the adults of 12 B populations in one stock jar; these 12 populations were descended from three separate populations of the Results migration experiment the of The results are summarized in Figures 5a, b, and c and Figures 6a, b, c, and d. (Figure 5 is analogous to Figure 3 of Experiment I above.) This form of presentation (Fig. 6) was chosen to facilitate the examination and interpretation of the effect of migration upon the process of group selection. In Generation-1 for the case of 0% migration the means of the A, C, and D treatments were statistically identical (.10 < P < .25) although the B treatment mean was significantly less than the mean for the C or control treatment (P < .005). After one generation of selection the mean of the A treatment exceeded that of the control by more than 90 adult beetles per population (P < .005) and after three generations of selection the difference was in excess of 130 adults per population. The mean of the D or random treatment became statistically greater than the mean of the control at Generation-5 (D - C = 41 adult beetles, P < .005) and remained so for the duration of the experiment. There was a difference between the means of B and C treatments in Generation-1 of 95 adults and this difference increased to 122 adults by Generation-3. After that time there was a convergencein mean pop- MICHAEL J. WADE 146 s 300 300 HIGH A " 250. - 200' 200- 10-- D " 150 - ' 150 Cl) LOW 250- , 0O%MIGRATION -25%MIGRATION ---50%MIGRATION \ / / 50 50 (A) 6712(B)67 m~~~~~~~~~~~~~~ LL CD Z CONTROL C 2501 LUI200 _ 250t 1001 100t 50 50 (2 \ 200- '- ) (C) 5 6 7 ~GENERATIONS RANDOM 0--' N 2 3(D)4 5 6 7 FIG. 6. The mean numbers of adults for the High (A), Low (B), and Random (D) populations and the Control (C) populations for all generations of Experiment II. Each graph contains the observations for three levels of migration, 0%0,25%,Xand 50%0for one selection treatment. Each point is the mean of 18 observations. ulation size between the B and C treatments similar to that previously observed in Experiment I, although the B treatment mean remained significantly lower than the control mean (P < .005) for all generations. The magnitude of the deviations from the controls for all treatments is essentially the same in both Figures 5a and 3. Thus the smaller number of populations per treatment in Experiment II relative to Experiment I did not significantly affect the process of group selection. Although both the A and D means for the case of 25% migration (Fig. 5b) initially exceeded the control mean this difference diminished and by Generation-3 the A, D, and C means were statistically identical. With continued selection the A mean exceeded that of the C treatment by an average of 45 adults per population during generations four through seven (P < .005 for Generations-4, 5 and 7; P < .01 for Generation-6). The D treatment mean was greater than that of the C treatment by 14 or 15 adults per population in Generations-5, 6, and 7. This mean difference was not statistically significant in Generation-5(.10 < P5 < .25) but became so in Generations 6 and 7 (P6 < .05, P7 < GROUP SELECTION .005). This increase in the statistical significance was achieved by a decrease in the variance in adult numbers in the D treatment and not by an increase in the magnitude of the difference between the D and C treatments. For the case of 50% migration (Fig. 5c) the trend in the difference between the A and C treatment means was similar to that described above for the case of 25% migration. After an initial period of convergence (Generations-2, 3, and 4) the mean number of adults in the A populations came to exceed that of the C populations by an average of 47 for generations five through seven (P5 = P6 = P7 < .005). The D treatment with 50% migration, unlike the D treatments with 0 and 25% migration in Experiment II and the D treatment in ExperimentI, never significantly exceeded the control treatment in mean numbersof adults. The B treatment, however, maintained a mean difference from the C of over 100 adult beetles per population during generations two through four and did not converge with C in mean population size to the extent witnessed with 0 and 25% migration. Three significant differences are illustrated by the comparisonof Figures 5a, b, and c and were noted in the discussion above: 1) The mean of the D or random selection treatment did not become significantly greater than the C mean in the case of 50% migration (Figure 5c) as it did with zero and 25% migration (Figure 5a and b). 2) The mean deviation of the A or high group selection treatment from the controls was less with migration (Figures 5b and c) than it was without migration (Figure 5a). 3) The mean deviation of the B or low group selection treatment from the controls was greatest in the case of 50% migration. The trend in the D migration treatments can be attributed to a decline in the between-populationsvariance with increasing levels of migration. The higher the level of migration the more homogeneous 147 the populations will become through generation time. In addition, the experimental design itself results in fewer pairs of parent populations with each increase in the level of migration and in this way also contributes to the decline in the between populations variance. It can be concluded that the tendency toward a differential proliferation of populations in the D migration treatments was not sufficient, given the low level of between populations variance, to result in a large difference between the D and C treatmentmeans. It is not clear that the above-mentioned differences (2 and 3) observed in the A and B migration treatments would have persisted had Experiment II been continued for a longer period of time. The inference drawn from this analysis is that the effect of group selection upon population size is not significantly altered by migration of the kind examined in Experiment II. Figures 6a, b, c, and d illustrate a decline in adult numbers through generation time. Because the decline affects all treatments it can be assumed that it does not affect the between treatments comparisons within any one generation. This decline could be attributed to some form of density dependent individual selection operating within all populations. The increased intensity of individual selection and the increased likelihood of interference by random effects in the A and D treatments could result in a lower position for the equilibrium population size determined by the opposing forces of group and individual selection. However, the actual cause of the decline remains unknown. An analysis of variance similar to that performed on the data from the random treatment of ExperimentI was not possible in this experiment. The populations in a given generation could not be partitioned according to "parents" because often one pair of parent populations had founded the entire generation. While several other interesting and important questions concerning the effect of migration upon the equilibrium between MICHAEL J. WADE 148 group and individual selection remain to be explored in later experiments, the answer to the fundamental question which motivated Experiment II is clear: The differential extinction and proliferation of populations can cause or maintain genetic differences whether the populations are isolated or exchanging migrants, at least under the conditions examinedhere. The Inbreeding Effect Due to a Propagule Size of 16 Adults.-The control populations of Experiment II provide an assessment of the effect of inbreedingupon adult population size. Each of the control populations in the "no migration" treatment, C-0%, remained isolated from all other populations for the duration of the experiment. On the other hand, each control population in both the 25% migration, C-25%, and the 50% migration. C-50%, treatments exchanged some number of migrants with another randomly selected control population every generation. Furthermore, it is extremely likely that all females in a propagule are premated (Park, 1933) and it is known that, although sperm transferred in later copulations takes precedence over sperm already in the spermatheca,the two types of sperm mix to some degree (Schlager, 1960). Both of these aspects of Tribolium ecology serve to decrease the effect of inbreeding due to a founding propaguleof 16 adults. Figure 6c illustrates the trend in the mean number of adults at day 37 for the C-0%, C-25%, and C-50% treatments of Experiment II. All three treatments undergo a similar decline in the mean number of adults. And, it can be concluded that individual selection and not an inbreeding effect is responsible for the trend toward fewer adults in the control populations. Figure 6c does show that the C-0% populations generally achieved a somewhat lower number of adults at 37 days than did either the C-25% or C-50% populations. An empirical attempt was made to assess the magnitude of this relative difference which could be attributed to inbreeding. This was done in the followlin<yOrf Wq 4. Comparison of the productivity of inbred and outbred populations from the control treatments of Experiment II. TABLE C-0% C-25% C-50% x (Outbred) x (Inbred) 115.2 89.7 91.0 57.2 76.3 111.9 x = the mean number of adults at 37 days for six replicates. In Generation-6, six populations, in addition to the usual number of 18, were established for each of the treatments: C-0%, C-25%, and C-50%. In all three cases, the six extra populations were founded by choosing 1 or 2 adults at random from each of 10 populations also chosen at random from the 18 populations at Generation-6. The sex ratio of the adults was left to chance. The six new populations for C-0% were thus maximally outbred and the mean number of adults produced by these six populations could be compared with the mean number of adults produced by the inbred C-0% populations in Generation-7. Similar comparisons could be made between the inbred and outbred populations of C-25% and C-50%. The means of the inbred populations were weighted according to the proportion of adults that a given inbred population contributed to the founding of the six outbred populations (Table 4). There are no statistical differences among the three outbred means (F = .347; df = 2,15; P > .25). There are apparent differences among the inbred means although using the weighted mean prohibits statistical analysis. However, the data from Generation-7, which are available for analysis using the Kruskal-Wallis rank sum test, support the conclusion that the order of the control means is C-0% < C-25% < C-50% (P < .025). Thus the tendency for the mean of the C-50% treatment to exceed that of the C-0% treatment can be attributed to an increase in homozygosity due to inbreeding. However, the decline in mean adult numbers through generation time by ap- GROUP SELECTION 149 proximately 175 adult beetles is large rela- which sends out ten times as many propative to the difference between the means of gules but establishes fewer new populathe C-50% and C-0% treatments, which tions. This definition, therefore, defines a averaged 35 adults during Generations 4 populational adaptation to be a trait which through 6. For this reason, I have con- is the focus of or the result of a process of cluded that individual selection, rather group selection. than an increase in homozygosity due to A large area of overlap exists between inbreeding, is primarily responsible for the this definition of an adaptation of a popugeneral decline in productivity of the C lation and the definition of an individual treatmentsin ExperimentsI and II. adaptation. Any trait which increases the probability of survival and reproductionof DIscusSION the individuals within a population and It has been shown in Experiments I and at the same time increases the probability II that both the group and the individual of survival and proliferation of the populaselection experiments have resulted in tion will be selected for by both individual changes in the mean number of adult and group selection. Such a trait is conbeetles at 37 days. It has been pointed sidered to be both an individual and a out that these changes occurred rapidly, populational adaptation. The definition of often within two or three generations, and a populational adaptation suggested here that the changes were large in magnitude, is thus very different from that suggested at times exceeding 100 adult beetles per by Williams (1966). A populational or population. The mechanism or mechanisms "biotic" adaptation in Williams' terms responsible for the changes in population "is a mechanism designed to promote the size remain to be elucidated. Experiments success of a biota, as measured by the measuring sex ratio, fecundity, fertility, lapse of time to extinction" (Williams, larval mortality, developmental time, the 1966, p. 97). If the design of such a cannibalism of eggs and pupae by adults, mechanism also promotes the success of the cannibalism of eggs by larvae, and individuals within the biota or population, body weight of adults have been completed then, by the principle of parsimony, the and will be presented in a later paper now mechanism must be judged an individual or "genic" adaptation (cf. Williams, 1966, in preparation. An adaptation at the populational level pp. 92-124). The major arguments against the operais any mechanism which enhances the tendency of a population to send out tion of group selection, which were responpropagules and establish additional popu- sible in large part for the widespread lations in new or vacant habitats. Under adoption of Williams' principle of parsithis definition the tendency for a popula- mony, were set forth in the introduction. tion to persist in a habitat for a long It is of general interest at this time to period of time is not consideredan adapta- reexamine each of these arguments in the tion unless the number of new populations light of the experimental results discussed founded by the population in question is in the previous sections. It has been correctly argued that the positively correlated with its persistence. Similarly, an increase in the rate of generation time of the individual is short production of propagules is not considered relative to the turnover time of the popuin iteroparous species. However, an adaptation unless this increase results lation this difference in and of itself does not in an increase in the rate of establishment warrant the conclusion that individual of new populations. A population which selection is a faster and more efficient sends out fewer but more successful propa- form of natural selection than is group gules is considered to be better adapted, in selection. The rate of change in the freterms of this definition. than a DoDulation quency of a given allele depends as much 150 MICHAEL J. WADE upon the magnitude and direction of the selection coefficients as it does upon the generation time. If the selection coefficients in natural populations are low, say on the order of .01 to .001 (Lewontin, 1974), then individual selection would require 50 to 100 generations to accomplish a significant change in the frequency of the allele in question. It is reasonable to expect at least one extinction of the population during such a span ok time even if the probability of extinction of any given population is very small. If the amount of between-populations variance is large a single episode of differential extinction and recolonization can result in a large genetic change. Therefore, it is important in evaluating the relative efficiencies of group and individual selection to consider not only the individual generation and population turnover times but also the magnitude of the individual selection coefficients and the betweenpopulations variance. It has been repeatedly emphasized that a significant between-populations variance is necessary for the operation of group selection. It is the origin and maintenance of this between-populations variance that has been consideredanother major obstacle to group selection. Most of the previous theoretical work on group selection has been based upon a number of assumptions which do not provide for the development of a significant amount of between-populations variance (Wilson, 1975, is a notable exception). In the absence of the required variance, it is not surprising that the conclusion reached by these papers has been that group selection can operate only under a restrictive set of parameters which are not likely to be realized in natural populations. The analysis of variance for the D treatment populations (Experiment I) revealed that a pattern of random extinctions with recolonization will create the necessary and favorable conditions for group selection to occur. This is due to the conversion of an increasing proportion of t'hi t(ltsl vu.ri.nc.e into t'he hi-twi-n-nonuila- tions component of the variance. The random treatments (D) of Experiment II offer some insight into the way in which migration between populations may affect this process. The homogenizing influence of 50% migration was sufficient to prevent an overall increase in mean adult numbers relative to the control populations (C), whereas 255%migration was not. It can be hypothesized that as the level of migration between populations increases, the rate of conversion of the variance will decrease for any fixed level of extinction. The third major argument that is offered against the process of group selection states that random events, such as temporarily unequal sex ratios or changing age distributions, will introduce a large margin of inefficiency into the group selection process unless the number of populations involved is large. This objection can be met in part by a reexamination of the design of Experiments I and II. Eighteen populations were involved in each treatment of Experiment II and despite the uncontrolled sex ratio of the founding propagules each generation, the results of Experiment II were qualitatively similar to those of Experiment I, which had approximately three times as many populations per treatment. It is reasonable, however, to expect that the degree of inefficiency that can be tolerated will be related to the intensity of group selection. In addition to modifying some of the arguments that have been raised against group selection, the experimental results may prove useful in the understanding of the process of speciation. Speciation can be viewed as a process which converts the variation within populations to variation between populations (Lewontin, 1974). Again referring to the data from the D treatment populations of Experiment I, it has been shown empirically that a process of random extinctions with recolonization will convert an increasing proportion of the total variance into the between-populations component of the variance. Thus, in those generations during which group GROUP SELECTION 151 selection is not operating, an entirely alter the progress of group selection randomprocess of extinctions will be creatrelative to the case with no such ing potentially favorable conditions for migration. group selection to occur at some later ACKNOWLEDGMENTS time. If the population structure of certain It gives me great pleasure to acknowlorganisms is such that local extinctions edge the helpful advice and criticism of occur relatively frequently, one would ex- Dr. Thomas Park, Professor Emeritus at pect a large between-populations variance the University of Chicago; I have also to accumulate rapidly. These types of benefited from the discussion of my ideas organisms might therefore be expected to with Montgomery Slatkin, David Mertz, have a greater rate of speciation than Michael Nathanson, Joel Sohn, R. C. organisms for which local extinctions are Lewontin, Janice Spofford, R. D. Alexa relatively rare occurrence. Group selec- ander, Don Willard, Richard Horwitz, and tion might very well be one of the greatest Patricia McElroy. The comments of C. creative forces for evolutionary change as Istock and an anonymous reviewer much Wright (1931) suggested over 40 years improved the original manuscript. Ora ago. Lee Watts and Timothy Wade provided SUMMARY technical assistance in the laboratory at In summary, the empirical data gathered various times during the course of the in this experimental study of group selec- experimental work. This work was supported in part by a USPS Training Grant tion permit the following conclusions: awarded to the Department of Theoretical 1. When group and individual selection Biology at the University of Chicago, operate upon a trait in opposite Grant Number GM-2037. directions, group selection can accomplish a genetic change which LITERATURE CITED occurs rapidly and is in large magni- ALLEE, W. C. 1940. Concerning the origin of tude. sociality in animals. Scientia 1940:154-160. 2. When group and individual selection AMADON,D. 1964. The evolution of low reproductive rates in birds. Evolution 18:105-110. operate in the same direction upon a 1954. ANDREWARTHA, H. G., AND L. C. BIRCH. trait, the rate of change of the trait The distribution and the abundance of aniin question can be significantly mals. Univ. of Chicago Press, Chicago, Illinois. accelerated relative to the rate of BOORMAN, S. A., AND P. R. LEVITT. 1973. Group selection on the boundary of a stable change of that trait under individual population. Theoret. Pop. Biol. 4:85-128. selection alone. J. 1962. Evolved regulatory mech3. A process of random extinctions will BRERETON, anisms of population control. pp. 81-73 In establish the ideal and favorable Evolution of living organisms. G. W. Leeper conditions for group selection to oc(ed), Melbourne Univ. Press, Melbourne. cur by converting an increasing pro- BROWN, W. L. 1958. General adaptation and evolution. Syst. Zool. 7:157-168. portion of the total variance in adult soft selecnumbers into the between-populations CHRISTIANSEN, F. B. 1975. Hard and tion in a subdivided population. Amer. Natur. componentof the variance. 109:11-16. 4. A process of random extinctions can CROW, J. F., AND M. KIMURA. 1970. An introbecome group selection if there is a duction to population genetics theory. Harper and Row, Publishers, New York. differential proliferation of the surC. 1878. The origin of species. D. DARWIN, viving populations. and Co., New York. Appleton 5. Migration, defined to be founding of DUNBAR, J. S. 1960. The evolution of stability new populations by varying proporin marine environments: natural selection at tions of migrants from two parent the level of the ecosystem. Amer. Natur. 94: 129-136. populations, does not qualitatively 152 MICHAEL J. WADE P. L. 1934. Vulnerability of BobWhite populations to predation. Ecology 15: 110-127. FISHER, R. A. 1930. The genetical theory of natural selection. Clarendon Press, Oxford. GADGIL, M. 1975. Evolution of social behavior through interpopulation selection. Proc. Nat. Acad. Sci. 72(3):1199-1201. GMPIN, M. E. 1975. Group selection in predator-prey communities. Princeton Univ. Press, Princeton, N. J. HARTMANN, L. VAN. 1955. Clutch size in polygamous species. Acta XI. Cong. Int. Ornith., Basel. pp. 450-453. HALDANE, J. B. S. 1932. The causes of evolution. Longmans, Green & Co., New York. HAMILTON, W. D. 1964. The genetical evolution of social behavior I, II. J. Theoret. Biol. 7:1-52. . 1966. The moulding of senescence by natural selection. J. Theoret. Biol. 12:12-45. HORN, H., AND R. H. MAcARTHUR. 1972. Competition among fugitive species in a harlequin environment. Ecology 53:749-752. HINTON, H. E. 1942. Secondary sexual characteristics of Tribolium. Nature 149:500. KING, C. E., AND P. S. DAWSON. 1972. Population biology and the Tribolium model. Evol. Biol. 5:133-227. LACK, D. 1954. The natural regulation of animal numbers. Clarendon Press, Oxford. LESLIE, P. H., AND T. PARK. 1949. The intrinsic rate of natural increase of Tribolium castaneum Herbst. Ecology 30:469-477. LEVIN, B. R., AND W. L. KILMER. 1974. Interdemic selection and the evolution of altruism: a computer simulation study. Evolution 28: 527-545. LEVIN, S. A. 1974. Dispersion and population interactions. Amer. Natur. 108:207-228. LEVINS, R. 1970. Extinction. In Some Mathematical Questions in Biology. Amer. Math. Soc. 2:75-108. LEVINS, R., AND D. CULVER. 1971. Regional coexistence of species and competition between rare species. Proc. Nat. Acad. Sci. 68:12461248. LEWONTIN, R. C. 1958. Adaptations of populations to varying environments. Cold Spring Harbor Symp. Biol. 22:395-408. . 1962. Interdeme selection controlling a polymorphism in the house mouse. Amer. Natur. 96:65-78. 1965. Selection for colonizing aLility. In The Genetics of Colonizing Species. Edited by H. G. Baker and G. L. Stebbins. Academic Press, New York. . 1970. The units of selection. Annu. Rev. Ecol. Syst. 1:1-18. . 1974. The genetic basis of evolutionary change. Columbia Univ. Press, New York. ERRINGTON, R. H., AND E. 0. WILSON. 1967. The theory of island biogeography. Princeton Univ. Press, Princeton, New Jersey. MARUYAMA, T. 1970. Effective number of alleles in a subdivided population. Theoret. Pop. Biol. 1:273-306. J. 1964. Group selection and MAYNARD-SMITH, kin selection. Nature 201:1145-1147. D. J., AND L. J. DAVIDSON. 1965. McDONALD, Population fitness in Tribolium. I. Amer. Natur. 99:463-470. D. J., AND L. STONER. 1968. PopuMcDONALD, lation fitness in Tribolium. II. Population characteristics influencing the capacity for survival. Amer. Natur. 102:323-336. D. B. 1969. Age distribution and MERTZ, abundance in populations of flour beetles. I. Experimental studies. Ecol. Monogr. 39:1-31. * 1970. Notes on methods used in life history studies. In Readings in ecology and ecological genetics. J. H. Connell, D. B. Mertz, and W. W. Murdoch (eds.), Harper and Row, New York. . 1972. The Triboluim model and the mathematics of population growth. Annu. Rev. Ecol. Syst. 3:51-78. M. 1975. The effect of resource NATHANSON, limitation on competing populations of flour beetles, Tribolium spp. Bull. Ent. Res. 65: 1-12. 1933. Studies in population physiolPARK, T. ogy. II. Factors regulating the initial growth of T. confusum populations. J. Exp. Zool. 65:17-42. . 1934. Observations on the general biology of the flour beetles Tribolium confusurn Duval. Quart. Rev. Biol. 9:36-54. * 1948. Experimental studies of interspecies competition. I. Competition between populations of the flour beetles, Tribolium confusum Duval and Tribolium castaneum Herbst. Ecol. Monogr. 18:265-308. 1954. Experimental studies of interspecies competition. II. Temperature, humidity, and competition in two species of Tribolium. Phys. Zool. 27:177-238. 1948. The fePARK, T., AND M. B. FRANK. cundity and development of the flour beetles, Tribolium confusum and Tribolium castaneum at three constant temperatures. Ecology 29: 368-374. 1964. PARK, T., P. H. LESLIE, AND D. B. MERTZ. Genetic strains and competition in populations of Tribolium. Phys. Zool. 36:97-162. PARK, T., D. B. MERTZ, W. GRODZINSKI, AND T. PRUS. 1965. Cannibalistic predation in populations of flour beetles. Phys. Zool. 38: 289-321. RAND, A. L. 1954. Social feeding behavior of birds. Field. Zool. 36:1-71. SCHLAGER, G. 1960. Sperm precedence in the MAcARTHUR, GROUP SELECTION fertilization of eggs of Tribolium castaneum. Ann. Ent. Soc. Amer. 53:557-560. SHRODE,R. R. 1960. Evidence that mating is random in Tribolium castaneum. Tribol. Inform. Bull. 2:15-16. SIMBERLOFF, D. S., AND E. 0. WILSON. 1969. Experimental zoogeography of islands: the colonization of empty islands. Ecology 50: 2 78-296. SKUTCH, A. F. 1961. Helpers among birds. Condor 63:198-226. SLATKIN, M. W. 1974. Competition and regional coexistence. Ecology 55:128-134. SOKOLOFF, A. 1974. The biology of tribolium, Vol. 2. Clarendon Press, Oxford. STANLEY, M. S. M., AND A. W. GRUNDMANN. 1965. Observations on the morphology and sexual behavior of Tribolium confusum (Coleoptera: Tenebrionidae). J. Kan. Ent. Soc. 38:10-18. STURTEVANT, A. H. 1938. Essays on evolution. II. On the effects of selection on social insects. Quart. Rev. Biol. 13:74-76. VAN VALEN,L. 1975. Group selection, sex, and fossils. Evolution 29:87-94. WMLLIAMS, G. C. 1964. Measurement of consociation among fishes and comments on the evolution of schooling. Mich. St. Univ. Mus. Publ.. Biol. Ser. 2 :351-383. 153 . 1966. Adaptation and natural selection. Princeton Univ. Press, Princeton, New Jersey. . 1971. (Ed.). Group selection. Aldine Atherton, Inc., New York. WMSON, D. S. 1975. A general theory of group selection. Proc. Nat. Acad. Sci. 72:143-146. WILSON,E. 0. 1973. Group selection and its significance for ecology. Bioscience 23:631-638. WILSON, E. O., AND D. S. SIMBERLOFF.1964. Experimental zoogeography of islands: defaunation and monitoring techniques. Ecology 50:267-278. WRIGHT,S. 1931. Evolution in Mendelian populations. Genetics 16:93-159. . 1945. Tempo and mode in evolution: a critical review. Ecology 26:415-419. . 1956. Modes of selection. Amer. Natur. 90: 5-24. J. C. 1962. Animal disperWYNNE-EDWARDS, sion in relation to social behavior. Hafner Publishing Company, New York. YOUNG, A. M. 1970. Predation and abundance in populations of flour beetles. Ecology 51: 602-619. ZEIGLER,J. 1972. Regulation of unconfined populations of Tribolium. Ph.D. Thesis, Univ. of Chicago.
© Copyright 2025 Paperzz