SIZE-FECUNDITY RELATIONSHIPS AND THEIR EVOLUTIONARY IMPLICATIONS IN FIVE DESMOGNATHINE SALAMANDERS

STEPHEN G. TILLEY Museum of Zoology, University of Michigan, Ann Arbor

Received June 5, 1968

The plethodontid salamander genus Des­ genus has been "the gradual transformation mognathus comprises seven species accord­ of a population with a high egg production ing t.o current taxonomy, of which six and low survival into a population with low occur in, and four are endemic to the Ap­ egg production and a high survival to ma­ palachian Mountain system of eastern turity." He found no differences in num­ North America. Over most of the southern ber of clutches per year or age at maturity Blue Ridge Physiographic Province several that might compensate for differences in of these forms may exist sympatrically. An clutch size. apparent maximum of sympatry occurs in This paper attempts to further explore the Balsam Mountains of southwestern Vir­ and compare the relationships between size ginia, where five species, D. quadramacu­ and fecundity among the five sympatric latus, D. monticola, D. fuscus, D. ochro­ species of Desmognathus and to elaborate phaeus, and D. wrighti occur in abundance. on the hypotheses drawn by Organ relative Organ (1961) studied the ecological rela­ to evolutionary trends in reproductive tionships and population dynamics of these habits. In particular I have sought to species at Whitetop Mountain and Mt. determine whether fecundity differences Rogers in the Balsam Mountains. Hairston among the five species are explained simply (1949) compared the ecologies of D. by differences in their sizes, or whether size­ quadramaculatus, D. monticola, D. ochro­ fecundity relationships are more complex. phaeus, and D. wrighti in the Black Moun­ tains of North Carolina where those four MATERIALS AND METHODS forms occur sympatrically. Both authors This study is based mainly on Organ's found that, in the order given above, these series of Desmognathus from the Balsam species form a series in which decreasing Mountains. With regard to the species size is paralleled by an increasing tendency wrighti and quadramaculatus, I restricted toward terrestrial habits. Organ further this study to these series. I have included found that the trend is also paralleled by specimens of D. monticola from western increasing juvenile survivorship. He con­ North Carolina in addition to specimens cluded that heavier juvenile mortality is from the Balsam Mountains in an attempt to associated with aquatic sites, and that D. include more of the total snout-vent length quadramaculatus, D. monticola, D. fuscus, range for that species. Data on series of D. ochrophaeus, and D. wrighti represent D. ochrophaeus from the vicinity of High­ an evolutionary trend in which selection lands, North Carolina and D. fuscus from has favored progressively more terrestrial Licking County, Ohio are included where habits. comparisons with Balsam Mountain ma­ Size and egg production are positively terial of those two species are appropriate. correlated in the genus Desmognathus, as Whenever possible, only females with they are in many animals. Organ (1961) eggs greater than 1 mm in diameter were also emphasized that the larger species lay examined. Otherwise it was difficult to larger clutches than the smaller ones, and distinguish smaller eggs from those to be that the evolutionary trend within the deposited the following year, since in the

EVOLUTION 22:806-816. December, 1968 806 SIZE-FECUNDITY RELATIONSHIPS IN FIVE SALAMANDERS 807 early stages of yolk deposition egg sizes Harrison's work indicates that annual egg are so variable as to obscure the differences laying may occur in D. aeneus. The evi­ between the clutches of successive years. dence for bienniality in Leurognatkus is Snout-vent length measures the distance not, in my opinion, compelling. In view of from the tip of the snout to the posterior these arguments it would seem dangerous angle of the vent. The method of covari­ to assume bienniality in D. wrigkti, D. ance analysis was utilized to compare the monticola, and D. quadramaculatus. While slopes and elevations of the regression lines more data on this point are sorely needed, for the five species, using the methods and I shall assume all five species dealt with forms of calculations given in Snedecor here to be annual reproducers. (1956). The squared deviations and cross products of deviations were calculated from Nesting Habits computer calculated raw sums, sums of All five species exhibit parental care, in squares, and sums of cross products. In which the females remain with the eggs the case of D. quadramaculatus rounding until hatching. Pope (1924) found that error prevented the use of computer calcu­ females of D. monticola and D. quadra­ lated raw sums, and hand-calculated sums maculatus attach each egg individually to were used instead. the roof of the nesting cavity so that the eggs form a flat mass one or two eggs thick, SUMMARY OF HABITS AND whereas D. fuscus, D. ochrophaeus, and D. LIFE HISTORIES wrigkti (information on the latter species The following accounts are based pri­ from Organ, 1961) generally deposit their marily on the studies of Hairston and eggs in spherical clusters with only a few Organ, supplemented in certain cases by actually attached to the roof of the nesting my own observations. cavity. Females of the five species utilize some­ Female Reproductive Cycles what different sites for egg deposition. D. Organ concluded that all five of the spe­ quadramaculatus utilizes the most aquatic cies deposit eggs biennially, but Martof sites, depositing eggs on the undersides of and Rose (1963) found evidence for an­ rocks in the beds of streams and beneath nual egg laying in D. ochrophaeus in the small waterfalls. D. monticola, D. ochro­ southern Appalachians. Tilley and Tinkle pkaeus, and D. wrigkti deposit eggs in less (1968) have presented evidence for an an­ aquatic sites than D. quadramaculatus, be­ neath stream banks and, in the case of D. nual or possibly even biannual cycle in D. ochrophaeus in the vicinity of Mt. Mitchell ochrophaeus, beneath moss on rocks and and in the Balsam Mountains. Spight logs in seepage areas. D. fuscus appears to (1967) concluded that D. fuscus deposits lay eggs in situations intermediate between eggs annually in eastern North Carolina those utilized by D. quadramaculatus and and Harrison (1967) found that a sixth those used by the other species. species, D. aeneus, also lays annually. Martof (1962) found evidence for biennial Lengtk of tke Larval Period reproduction in another desmognathine, The length of the larval period varies Leurognathus marmoratus. greatly among the five species, and reflects The same weaknesses which Tilley and the aquatic to terrestrial trend discussed Tinkle found in the arguments for biennial above. In the terrestrial D. wrigkti meta­ egg laying in D. ochrophaeus probably morphosis occurs prior to hatching (one apply as well to D. wrigkti, D. fuscus, wonders, therefore, why it deposits its eggs D. monticola, and D. quadramaculatus. in aquatic sites). Organ gave estimates of Spight's study bears this out for D. fuscus. 26-28 months for D. quadramaculatus, 12- 808 STEPHEN G. TILLEY

4.0

3.5

0 Z ...w U ... 3.0 0 u. o· 0...

2.5

2.0

3.0 3.5 4.0 4.5 LOGe SNOUT -v ENT LENGTH FIG. 1. Relationships between size and fecundity in five species of Desmognathus. Lines represent least squares regression lines of log. follicle number on log. snout-vent length.

13 months for D. monticola, 13-16 months cola prefers the banks of streams, as does for D. fuscus, and 10-11 months for D. D. fuscus. Organ found that both D. ochrophaeus. The estimate for D. ochre­ ochrophaeus and D. wrighti occur in ter­ phaeus is in error (Tilley and Tinkle, 1968); restrial situations. Males of both species the larval period of that species is from 4 are largely terrestrial during the summer to 6 months at Mt. Mitchell, North Caro­ months, moving to aquatic sites such as lina. seepage areas and spring heads during the Age at Maturity winter. His data, as well as Hairston's and Organ estimated that males of all five my own observations, indicate that of the species matured at 3.5 years of age, while two, D. wrighti is the more terrestrial. The Spight (1967) estimated 3 years for males aquatic sites preferred by D. ochrophaeus of D. fuscus. Both authors concluded that are of the seepage area-spring head variety, females require an additional year to pro­ rather than the stream or stream side habi­ duce their first clutch of eggs. tats occupied by D. quadramaculatus and D. monticola. Habitats The species are also segregated altitudi­ D. quadramaculatus occurs chiefly in nally to a certain degree. D. ochrophaeus larger and faster streams, while D. monti- occurs from the lowest to the highest avail- SIZE-FECUNDITY RELATIONSHIPS IN FIVE SALAMANDERS 809

TABLE 1. Summary of regression data. n =sample size, r =correlation coefficient. Equations and correlation coefficients are for regressions of log. foUicle number on log. snout-vent length. v = Balsam Mountains (Virginia), 0 =Ohio.

Significance Regression Species " level equation D. wrighti 23 0.458 0.050 Y = 2.4130X - 5.7235 D.ochrophaeus 25 0.598 0.010 Y = 2.1773X - 5.1407 D.fuscus (v) 26 0.837 0.001 Y = 2.2845X - 5.2727 D. [uscus (0) 18 0.687 0.010 Y = 1.7484X - 3.3944 D. monticola 35 -0.163 < 0.100 D. quadramac. 26 0.390 0.050 Y = 1.0700X - 0.8825 able habitats in the Appalachians. In the sion lines relatively homogeneous, but not Balsam Mountains D. fuscus has a similar in removing all of the curvilinearity. D. distribution, but does not extend as high wrighti appears to have somewhat higher as does D. ochrophaeus, while D. quadra­ loge follicle number values than might have maculatus and D. monticola occur from the been expected from extrapolation from the lowest elevations in the Balsam Mountains regression for D. ochrophaeus. D. quadra­ to just above 5000 feet. Organ found that maculatus, on the other hand, appears to D. fuscus and D. monticola have mutually have slightly lower values. Table 1 gives exclusive distributions, with D. fuscus oc­ tile correlation coefficients and regression cupying tile banks of the headwaters of equations for the five species, Table 2 the streams at high elevations and D. monti­ results of the covariance analyses in which cola occupying the banks of streams at the regression for D. ochrophaeus was com­ lower elevations. D. wright; is essentially pared with those for the other four species. a high elevation species, whose populations Whereas there is a suggestion of curvi­ reach their maximum densities in the linearity when all the species are graphed spruce-fir forests of the southern Appa­ together, the regressions for D. wrighti and lachians. However, it extends to as low as D. quadramaculatus do not differ signifi­ 3000 feet near Mt. Mitchell (personal ob­ cantly from that for D. ochrophaeus at the servation and Hairston, 1949) and Organ five percent level, neither with respect to found it as low as 3500 feet in the Balsam the slopes nor to the elevations of the re­ Mountains. gression lines (Tables 2 and 3). The analysis suggests, rather, that the relation­ RESULTS ship between snout-vent length and fe­ Figure 1 shows the relationship between cundity is identical in the species wrighti, snout-vent length and follicle number in ochrophaeus, and quadramaculatus. tile five species studied. Plots of the un­ Since the correlation coefficient for D. transformed data were curvilinear, presum­ monticola is insignificant (in fact, negative) ably because follicle number is a function at the five percent level, that species was of body volume rather than body length. excluded from the covariance analysis. The Furthermore, the variance in follicle num­ hypothesis that D. monticola lies on the ber was positively correlated with snout­ regression line common to D. wrighti, D. vent length. For these reasons both vari­ ochrophaeus, and D. quadramaculatus was ables were converted to their natural tested as follows: A single regression line logarithms, which are plotted in Figure 1. was calculated for the three species, having All of the statistics discussed below are the equation Y = l.5027X - 2.7020. From based on the natural logarithms of the data. this equation the mean loge follicle number The loge-loge transformation succeeded corresponding to the mean loge snout-vent in making the variances along tile regres- length' was calculated (3.3931). This value 810 STEPHEN G. TILLEY

TABLE 2. Summary of covariance analyses. F tests for similarity of slopes and elevations constructed as in Snedecor, 1956. Subscripts as in Table 1. Based on loge-log. transform.

Slopes Elevations Conclusions Significance Significance Test F level F level Slopes Elevations Overall (less monticola) 1.100 <0.05 27.235 0.01 equal different ochrophaeus vs. wrighti 0.040 <0.05 1.754 < 0.05 equal equal ochrophaeus vs. [uscus (v) 0.024 <0.05 10.273 0.01 equal different ochrophaeus vs. [uscus (0) 0.282 <0.05 4.495 0.05 equal different ochrophaeus vs. quadramac. 1.821 <0.05 0.349 <0.05 equal equal juscus (v) vs. fuscus (0) 0.786 <0.05 5.373 0.05 equal different

was compared with the actual value for D. is not significant at the five percent level monticola (3.3001) by means of at-test. (t = 1.0057, .4> P > .3). Thus D. juscus, The resultant t = 2.1577, df = 33 is sig­ for its size, has a higher fecundity than any nificant at the five percent level. Thus D. of the other species. monticola appears to lay somewhat fewer The Balsam Mountain region is one of than the expected number of eggs. the few areas in the Blue Ridge Physio­ In order to further test the hypothesis graphic Province where D. juscus occurs, that the relationship between clutch size and is perhaps the only part of its exten­ and body size is the same in both D. wrigkti sive range where it is sympatric with as and D. ochropkaeus, the clutch sizes of fe­ many other species of Desmognatkus. It male D. wrigkti were compared with those seemed possible, therefore, that the repro­ of very small female D. ockropkaeus from ductive properties of the species in the the southwestern portion of the latter's Balsams might differ from those in other range near Highlands, North Carolina. parts of its range. In order to test for such These data were not analyzed statistically differences, the regression in the Balsam but, as shown in Figure 2, female D. ochro­ Mountain material was compared with that pkaeus of the same size range as D. wrigkti in a series of 19 females from Licking lay approximately the same number of County, Ohio (UMMZ 125516-17). One eggs. of the latter was discarded from the analy­ The overall covariance analysis (exclud­ sis because she contained abnormally few ing D. monticola) shown in Tables 2 and 3 follicles. The regressions are superimposed indicates that all five species lie on lines in Figure 3, and the results of the covari­ with the same slope, but that the hypothesis ance analysis comparing them are given in that all lie on a single line is rejected at the Tables 2 and 3. The difference between the one percent level. Comparison of D. ochro­ slopes is not significant at the five percent pkaeus with the other species has shown level but that between the elevations of the that it lies on the same regression line as regression lines is, the elevation for the Bal­ D. wrigkti and D. quadramaculatus but on sam Mountain series being higher. Fur­ a different line from D. juscus. At any thermore, the Balsam Mountain specimens given size then, D. juscus lays more eggs are larger. Both of these factors would ap­ than D. ochrophaeus whereas the increase pear to indicate that D. juscus in that in fecundity per unit increase in size is the region are considerably more fecund than same for all the species. The mean num­ at least those from the Ohio population. ber of eggs carried by females of D. juscus The covariance analysis also indicates examined (31.4), in fact, exceeded the that Ohio specimens differ significantly mean number contained in females of D. from D. ockropkaeus, although less than monticola (28.1), although the difference Balsam Mountain D. juscus do. SIZE-FECUNDITY RELATIONSHIPS IN FIVE SALAMANDERS 811

0 ceo 0 10 2.5 0/0 00 Ell 0 cj 0 z o EIlEl;) + ...III 3.5 U Ill: ... + + + + oC±> 0 &AI if ...

Assuming a constant egg volume-body '--_...... '_'--_....L..-_...J...._....J1 volume relationship in the five species, the 3.5 ".0 regression of clutch size on body length LOG.SVL would appear to indicate that a reasonably FIG. 3. Relationships between size and fecun­ constant relationship is maintained be­ dity in D. [uscus from the Balsam Mountains tween clutch volume and body volume, at (hollow circles) and Ohio (solid circles). Regres­ least among the species wrighti, ochre­ sion lines and transformations as in Fig. 1. phaeus, and quadramaeulatus. From the comparison of regressions one might expect that the ratio of clutch volume to body vol­ compared among the five species. Those ume is higher in D. fuseus and lower in D. species between which the ratios did not montieola. Ratios were calculated for the differ significantly at the five percent level specimens examined, and their means were all assigned the ratios' mean value for among the five species are compared in those species. Trunk-length to body-length Figure 4. The clutch volumes were based ratios were also compared and found not on the diameters of the largest eggs, found to differ significantly at the five percent in each of the species. These were: D. level among the five species. Thus snout­ wrighti, 2.5 mm; D. ochropkaeus, 3.0 mm; vent length could be used to compute body D. fuseus, 3.0 mm; D. montieola, 4.0 mm; volumes. Body volumes were then calcu­ and D. quadramaeulatus, 4.0 mm. Body lated as the following functions of snout­ volumes were calculated as perfect cylin­ vent lengths: D. wrighti, D. ochrophaeus, ders using snout-vent length as length and and D. fuseus, V = .0142(SVL3); D. monti­ width at the axilla as diameter. The ratios cola, V = .0191(SVL3); and D. quadra­ of axillary width to snout-vent length were maeulatus, V = .0172(SVL3). 812 STEPHEN G. TILLEY

TABLE 3. Details of individual covariance analyses, constructed after Snedecor, 1956. f =degrees of freedom; x = deviations irom log. mean snout-vent lengths, y =deviations from log. mean foUicle numbers, b =slopes, SS =sum of squared deviations from regression line, MS =mean squared deuia- tions from regression line (residual variances).

Deviations from means Deviations from reg. Source 1:",' 1:",y 1:y' b 55 M5 wrighti 22 0.0414 0.0999 0.9482 2.4130 21 0.7071 0.0336 ochrophaeus 24 0.1618 0.3523 2.1465 2.1773 23 1.3795 0.0599 fuscus 25 0.2390 0.5460 2.0302 2.2845 24 0.7829 0.0326 quadramac. 25 0.1128 0.1207 0.8542 1.0700 24 0.7251 0.0302 Pooled 92 3.5946 0.0390 Reg. coef. 3 0.1288 0.0429 Common 0.5550 1.1189 5.9791 2.0160 95 3.7234 0.0391 Adj. means 3 3.1948 1.0649 Total 11.1105 17.3461 33.9995 98 6.9182 ochrophaeus 24 0.1618 0.3523 2.1465 2.1773 23 1.3795 0.0599 wrighti 22 0.0414 0.0999 0.9482 2.4130 21 0.7071 0.0336 Pooled 44 2.0866 0.0474 Reg. coef. 1 0.0019 0.0019 Common 0.2032 0.4522 3.0947 2.2253 45 2.0885 0.0464 Adj. means 1 0.0814 0.0814 Total 1.3800 2.2639 5.8838 46 2.1699 ochrophaeus 24 0.1618 0.3523 2.1465 2.1773 23 1.3795 0.0599 fuscus (v) 25 0.2390 0.5460 2.0302 2.2845 24 0.7829 0.0326 Pooled 47 2.1624 0.0460 Reg. coef. 1 0.0011 0.0011 Common 0.4008 0.8983 4.1767 2.2412 48 2.1635 0.0450 Adj. means 1 0.4623 0.4623 Total 0.7731 2.3092 9.5231 49 2.6258 ochrophaeus 24 0.1618 0.3523 2.1465 2.1773 23 1.3795 0.0599 quadramac. 25 0.1128 0.1207 0.8542 1.0700 24 0.7251 0.0302 Pooled 47 2.1046 0.0447 Reg. coef. 1 0.0814 0.0814 Common 0.2746 0.4730 3.0007 1.7225 48 2.1860 0.0455 Adj. means 1 0.0159 0.0159 Total 4.7530 7.0763 12.7371 49 2.2019

[uscus (v) 25 0.2390 0.5460 2.0302 2.2845 24 0.7829 0.0326 [uscus (0) 17 0.1276 0.2231 0.8261 1.7484 16 0.4360 0.0256 Pooled 40 1.2189 0.0304 Reg. coef. 1 0.0239 0.0239 Common 0.3666 0.7691 2.8563 2.0979 41 1.2428 0.0303 Adj. means 1 0.1628 0.1628 Total 0.4247 0.9956 3.7394 42 1.4056 ochrophaeus 24 0.1618 0.3523 2.1465 2.1773 23 1.3795 0.0599 [uscus (0) 17 0.1276 0.2231 0.8261 1.7484 16 0.4360 0.0256 Pooled 39 1.8155 0.0465 Reg. coef. 1 0.0131 0.0131 Common 0.2894 0.5754 2.9726 1.9882 40 1.8286 0.0457 Adj. means 1 0.2162 0.2162 Total 0.3878 0.9404 4.3253 41 2.0448 SIZE-FECUNDITY RELATIONSHIPS IN FIVE SALAMANDERS 813

The ratios for the five species are graphed .5 in Figure 4. Non-overlap of the black o M boxes, representing twice the standard er­ F ror on either side of the means, indicates mean differences significant at approxi­ mately the five percent level. As expected from the covariance analysis, D. fuscus is .4 characterized by the highest ratios of the ... 0 W Q five species, and differs significantly from > all but D. ochrophaeus. The mean ratio >-c 0 for D. ochrophaeus is significantly higher la than those for D. wrighti and D. quadra­ maculatus. No significant differences, occur \ .3 among the latter three species. Thus, while ... 0 D. wrighti, D. ochrophaeus, and D. quadra­ > maculatus lie on the same regression line % U when follicle number is plotted against ...... ~ snout-vent length, this regression does not u maintain a constant clutch volume-body .2 volume ratio among the three species using the criteria for calculating clutch and body volumes given above.

DISCUSSION Fecundity is only one of many variables .1 subject to adjustment by natural selection in the maximization of individual fitness. FIG. 4. Comparison of clutch volume-body volume ratios among the five species of Desmog­ In the genus Desmognathus fecundity is nathus. Solid boxes represent twice the standard intimately related to body size, both within error, hollow boxes one standard deviation on and between species. At the interspecific either side of each mean. level, however, this relationship is not per­ fect. Considering the variety of habitats exploited by the species dealt with here, between size and fecundity and, more perhaps it is closer than one would in­ seriously, leaves unanswered the question tuitively expect, particularly among the of the selective advantage of high fecundity series wrighti, ochrophaeus, and quadra­ in aquatic habitats. Organ's (1961) data maculatus. Any theory concerned with the show mortality to be heavier in such situa­ evolutionary modification of demographic tions, but to hypothesize that this results parameters in the genus Desmognathus in selection for increased fecundity violates must somehow explain both the overall the arguments of Lack (1954), who pointed closeness of this relationship and its imper­ out that high mortality rates result from fections. high fecundities. One means of reconciling the similarity Assuming, as most authors have, that of the relationship in these species with the aquatic situations are the ancestral habitat diversity of their habitats would be to con­ of the genus Desmognathus (Dunn, 1926; clude that selection is able to increase fe­ Wake, 1966), and that body size differ­ cundity only by increasing body size, and ences were established in advance of ter­ that high fecundities are selectively advan­ restrial tendencies in the smaller forms, tageous in aquatic habitats. This would selection operating whenever the various ignore the imperfections of the relationship species came into contact might have re- 814 STEPHEN G. TILLEY

100

~ '.::: 50 ""- "-, ""- -, \ \ \ \ \ \ 10 \ \ \ II'l... \ ~ ~ 5 \ c \ > \ c \ z \ ... '\ 0 0 z .1

.05

0-1 ~4 5-6 AGE IN YEARS F;o. 5. Survivorship curves for males of D. [uscus (solid line) and D. monticola (broken line) from the Balsam Mountains. Data on numbers of individuals in successive year classes from Organ, 1961. See text for additional explanation. sulted in their displacement into different the species examined. It seems reasonable habitats. Terrestrial tendencies would have to assume that such correlations also held been most strongly favored in the smaller among the ancestors of the present forms, forms, since these were (and still are) sub­ so that the larger forms were also the most ject to predation by larger salamanders. fecund, just as they are today. This, in The latter, by virtue of their size, would turn, would have led to the higher juvenile have been able to remain in the ancestral mortality rates among aquatic species that habitat. Organ observed in his data. The data presented in this paper indi­ The data on clutch volume-body volume cate that positive size-fecundity correla­ ratios suggest that egg size, as well as tions are evident both' within and between clutch size has been altered in the evolution SIZE-FECUNDITY RELATIONSHIPS IN FIVE SALAMANDERS 815

of the reproductive rates of the forms since he concluded that the species had a studied. Thus D. ochrophaeus, while it lies biennial laying cycle he also assumed that on the same size-fecundity regression line he could distinguish between females that as D. wrighti and D. quadramaculatus, is had brooded during the previous year and characterized by a higher clutch volume­ those which would produce eggs during the body volume ratio than either of those two current year, and used this information to species. D. monticola, on the other hand, calculate the mean annual survival rate for lays fewer eggs than expected from the brooding females. Since Martof and Rose wrighti- ochrophaeus - quadramaculatus re­ (1963) and Tilley and Tinkle (1968) have gression, but evidently amasses the same presented reasons for doubting the pres­ relative volume of yolk per clutch. ence of a biennial laying cycle, the groups Clutch volume-body volume ratios are of animals that Organ compared probably probably more meaningful indications of did not represent groups of females that the actual amount of effort expended at were depositing eggs in successive years. each reproduction than are clutch sizes Thus, with our present inability to age fe­ alone, and do not appear to correlate at all male salamanders, it is extremely doubtful closely with the aquatic to terrestrial trend. whether survivorship curves can be con­ D. juscus is particularly discrepant in this structed for that sex. regard (Fig. 4), as well as in its average If the shorter adult life expectancy of D. clutch size (Fig. 1). Of the five species juscus is in fact attributable to predation considered here, D. juscus is the only one by large aquatic Desmognathus, one might in which the adults have not evolved some expect the reproductive effort per season means of minimizing predation by larger of individuals free of that source of mor­ salamanders in aquatic habitats. D. quadra­ tality to be lower. Figure 3 indicates that maculatus and D. monticola share the ad­ this may in fact be the case. Female D. vantage of large size. Mature males of D. juscus from the Balsam Mountains do lay ochrophaeus and both sexes of D. wrighti more eggs than those from the Licking are largely terrestrial during most of the County, Ohio population. This is attribut­ year. The fact that brooding females of able both to their larger size and to a dif­ both D. ochrophaeus and D. juscus are ferent size-fecundity relationship, whereby probably subjected to heavier mortality for any given size, females of the Balsam than the other species should cause selec­ Mountain population lay more eggs than tion to favor greater reproductive efforts females from Ohio. Perhaps their larger per season (Williams, 1966). Figure 5 size is also an adaptation directed toward compares the survivorship curves for males increased fecundity. of D. juscus and D. monticola, based on The fact that D. juscus combines rela­ Organ's data for numbers of males in dif­ tively small size and aquatic habits may ferent year classes and mine on fecundities account for its absence throughout most of (Organ assumed a larger clutch size in D. the southern Blue Ridge Physiographic monticola, based on clutches of the two Province, in which its preferred habitats species that he found in the field). The are occupied by larger salamanders. As curves, similar as they are, do indicate suggested by Dunn (1926) and Organ slightly lower adult survivorship in D. (1961), competition with D. monticola, as juscus. Since in these two species both well as predation by it and other large sala­ sexes occupy similar habitats, female sur­ manders, may be important in determining vivorship curves of the two should exhibit the distribution of D. juscus. a similar relationship to one another. Or­ gan estimated the mean annual survival SUMMARY rate of brooding females of D. juscus to be Size and fecundity are positively corre­ the lowest of the five species. However, lated in four of the five species of Desmog- 816 STEPHEN G. TILLEY nathus inhabiting the Balsam Mountains LITERATURE CITED of southwestern Virginia. This correlation HAmsToN, N. G. 1949. The local distribution is manifest both inter- and intraspecifically. and ecology of the plethodontid salamanders of the southern Appalachians. Ecol. Monog. Size-fecundity relationships are identi­ 19:47-73. cal in D. wrighti, D. ochrophaeus, and D. HARRISON, J. R. 1967. Observations on the life quadramaculatus, while D. fuscus has a history, ecology and distribution of Desmog­ higher and D. monticola a lower fecundity nathus aeneus aeneus Brown and Bishop. Amer. than expected by extrapolation of the com­ Mid. Natur. 77(2) :356-370. LACK, D. 1954. The evolution of reproductive mon regression for the other three species. rates. In J. S. Huxley, A. C. Hardy, E. B. D. fuscus and D. ochrophaeus both exhibit Ford (eds.) , Evolution as a process. Allen and higher clutch volume-body volume ratios Unwin, London. than do the other species. MARTOF, B. S. 1962. Some aspects of the life history and ecology of the salamander Leurog­ It is hypothesized that selection has nathus. Amer. Mid. Natur. 67(1): 1-35. favored terrestrial tendencies in the smaller MARTOF, B. S., AND F. L. ROSE. 1963. Geo­ species because of their vulnerability to graphic variation in southern populations of predation by larger aquatic salamanders. Desmognathus ochrophaeus. Amer. Mid. Natur. 69(2) :376-425. The fact that the smaller species are neces­ ORGAN, J. A. 1961. Studies of the local distri­ sarily less fecund has resulted in their ex­ bution, life history, and population dynamics hibiting lower mortality rates. The rela­ of the salamander genus Desmognathus in Vir­ tively high reproductive efforts of D. fuscus ginia. Ecol. Monog. 31: 189-220. and D. ochrophaeus may result from their POPE, C. H. 1924. Notes on North Carolina salamanders, with especial reference to the egg­ combining relatively small size with aquatic laying habits of Leurognathus and Desmogna­ habits. thus. Amer. Mus. Nov. 306: 1-19. SNEDECOR, G. W. 1956. Statistical methods. Iowa ACKNOWLEDGMENTS State Univ. Press, Ames. SPIGHT, T. M. 1967. Population structure and I wish to thank Dr. Donald W. Tinkle, biomass production by a stream salamander. Dr. Charles F. Walker, Dr. Richard C. Amer. Mid. Natur. 78(2) :437-447. Bruce, and Kraig Adler for their criticisms TILLEY, S. G., AND D. W. TINKLE. 1968. A re­ of the manuscript, Dr. Tinkle and Dr. interpretation of the reproductive cycle and demography of the salamander Desmognathus James A. Organ for the numerous discus­ ochrophaeus, Copeia 1968(2) :299-303. sions that led to its inception, and my wife, WAKE, D. B. 1966. Comparative osteology and Mary for her assistance in its preparation. evolution of the lungless salamanders, family This research was supported by National Plethodontidae. Mem. S. California Acad. Sci. Science Foundation Grant GB-6230 to the 4: 1-111. WILLIAMS, G. C. 1966. Natural selection, the University of Michigan, for research in sys­ costs of reproduction, and a refinement of Lack's tematic and evolutionary biology. principle. Amer. Natur. 100(916) :687-690.