«THESIS Date This is to certify that the thesis entitled POPULATION CHARACTERISTICS OF AGOUTI PACA ( RODENT IA) IN COLOMB IA presented by Stephen Frederick Collett has been accepted towards fulfillment of the requirements for Ph. D degree in LQQLQL Rollin H. Baker @fim Major professor 23 February 1981 0-7639 v; m- 4:31:57: ctrakv - Lil/“rm: R Y Michism State Univcrnty . "I w: 25¢ per day per item 1 \\ ‘ . I'm“ * RETURNING LIBRARY MATERIALS: ‘6' ‘ ’h "1’”: Place in book return to remove 1‘2“!!! charge from circulation records POPULATION CHARACTERISTICS OF AGOUTI PACA (RODENTIA) IN COLUMBIA BY Stephen Frederick Collett A DISSERTATION Submitted to Michigan State University in partial fUlfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Zoology 1981 Gflée’éié ABSTRACT POPULATION CHARACTERISTICS OF AGOUTI PACA (RODENTIA) IN COLOMBIA by Stephen Frederick Collett The paca (Agouti paca) is a large (> 7 kg) nocturnal rodent that is intensively hunted by peOples of tropical America. Characteristics of reproduction and age structure were evaluated fOr a sample of 205 pacas collected from narrow gallery forests in the Llanos Orientales of Colombia. Fecundity was estimated from gross examination of female reproductive tracts and from thin sections of ovaries. Individual age was determined from counts (If seasonal annulations seen in thin sections of the cementum tissue of the first upper molar. Observations on foods, behavior, use of habitat, and burrows, were made in the course of hunting. Crude estimates of abundance were also obtained. Females first gave birth at approximately one year of age, with only one of 43 females less than one year old having given birth. Forty-two of 69 females of reproductive age were pregnant. There was no indication of seasonal clustering of births, or of significant age- specific variation in fecundity. .A post-partum estrus was indicated by twelve females that were simultaneously parous and pregnant. The mean interval between births was calculated at 191 days. All Stephen Frederick Collett pregnancies were of single fetuses. The estimate of fecundity for all adult females was 0.95 QQ/Q/yr. Among subsamples defined by collecting methods or collecting sites, both sex ratios and age-structure regressions proved to be heterogeneous. Significantly more males than females were obtained when hunting from canoes. For the combined sample, adult age struc- ture was best described by log-linear regression of numbers on age, both for males and for females, implying a pattern of constant annual survivorship if the population is assumed to be stationary. Annual survivorship was estimated to be 0.800 for females and 0.867 for males. The age of the oldest individuals was estimated to be between 12 and 13 years. Examination of stomach contents and direct observations of feed- ing indicated that pacas ate fruits almost exclusively. Twenty-three specific food items were identified. Availablity of suitable fruits did not appear to be limited at any time of the year. All observa- tions of paca activity occurred in gallery forests at night. Of 108 observations of pacas in the wild, 86 were of solitary individuals. Mbst instances of intraspecific contacts involved females with small offspring. Pacas in the Llanos spend daylight hours in extensive, individual burrows, elevencfliwhich were excavated. Abundance esti- mates ranged from 38 to 56 adult pacas per km2 of forest. Relationships among fUndamental demographic parameters for the paca are described by an algebraic model in five terms, using simpli- fying assumptions of constant adult survivorship, constant fecundity, and stationary population. The model estimates survivorship of juve- niles at 0.23 from birth to the age of 1 1/2 years. Effects of Stephen Frederick Collett increasing or decreasing population size on estimated juvenile sur- vivorship are also considered. Exploitation by man appears to be a major source of mortality for the population, accounting for as much as 40% of adult deaths. The impact of local hunters is probably ameliorated as a result of hunting efforts being concentrated in relatively few, accessible forest galleries. Intraspecific interactions such as terri— torial behavior may function in determining population size and recruit- ment of young adults into heavily-harvested populations. The demographic pattern of high adult survivorship and low fecundity is more comparable to game ungulates than to most rodents. ACKNOWLEDGMENTS I wish to thank the members of my guidance committee, Rollin H. Baker (Chairman), Richard Hill, Peter MUrphy, and Howard Hagerman. In the initial stages of the paca project, I was greatly aided by John Beaman, walt Conley, Thomas Yuill, and Eustorgio Mendez. The assistance of Dr. John Hunter of the Latin American Studies Center, IMSU, was instrumental in accomplishing a preliminary trip to prospec- tive field sites in Panama and Colombia during March of 1975. Initial funding for the field work was provided by Grant 71-3432 to R. H. Baker from Midwest Universities Consortium for International Activities (MUCIA). The collaboration of Instituto del Desarollo de los Recursos Naturales Renovables (INDERENA), and of Instituto Colombiano Agropecuario (ICA), is gratefully acknowledged. I thank Carlos Uribe, Ceferino Hernandez, Joaquin Castiblanco, Raimundo Evaristo, Teofilo Villamizar, and Alejandro Jerez of INDERENA for their contributions to the field work at Territorio Faunistico El Tuparro, and Jorge Hernandez Camacho and Victor Vésquez for admini— strative support in Bogota. Invaluable assistance was provided by Cesar Lobo, Guillermo Restrepo, and Fernando Lozano of ICA. Completion of the study would not have been possible without the generous logistical, financial, and interpersonal help of the people of the Tulane University International Center for Medical Research (CIDEIM) of Cali, Colombia. In particular, I thank Paul Beaver, ii iii Thomas Orihel, Margarita de Izquierdo, Stephen Ayala, Wieslaw Kozek, Pablo Barreto, and Kirby Kloter. I am especially indebted to Antonio D'Alessandro, director of CIDEIM-Cali, for constant and unselfish sup- port and encouragement. Many members of the Centro Internacional de Agricultura Tropical (CIAT) organization provided invaluable assistance, both in Cali and at the Carimagua field station. James and Joyce Spain cheerfully pro- vided logistical and moral support, often under trying circumstances. Administrative support was provided by Fernando Bernal, Guillermo IMatéus, and Orlando Martinez. Ablelardo Cardenas, Ramon Gualdrén, Jorge Quintero, and Juan Reyes also helped in various ways. During the last two years of my stay in Colombia, I was supported by Peace Corps/Colombia. Ernesto Barriga was a great help in all phases of the study. I also thank José Sorzano and Grace de Murillo for essential non-technical help. VOlunteers Charles MCClung, Raleigh Blouch, Bill Lamar, Paul Gertler, Wally MCKay, and Marc weitzel all contributed substantially to my research. Tom Defler and Paul Vincelli provided information essential to identifying many of the plants involved in the study. I am particularly indebted to Jon Yates, Mauricio Barreto, and Sharon Halstead de Porre of the CIDEIM field station at El Porvenir for unfailing cooperation. The entire people of El Porvenir earned my eternal appreciation for their interest in furthering the research efforts of the CIDEIM program. Simon Pena of Hato Caviona collabo- rated.most willingly with all of us. The mathematical treatment presented here would not have been possible without fruitful discussions with Donald Hall, John Hart, and iv Robert Boling. Patient tutoring by Ira Longhini and Ronald Hamelink was equally essential. Donald Straney and Charles Cress provided advice regarding portions of the statistical treatments. any people in Colombia provided support in various non—profes- sional ways. I especially appreciate the help of Paul and Sue Barry, Tony and Eve Salmon, Bill and Nancy Lamar, Sharon Fish, Richard Cross, Brian Hempstead, Flavio Barney, and Cutto aniglia. The greatest contributions to my knowledge of the elusive paca came from experienced local hunters, who still know much more than I about the species. To these men, Dario Angel, Fernando Camico, Genaro Useche, Raimundo Evaristo, Jorge Ramirez, Cipriano Rodriquez, Fernando Batista, and Jose Gutierrez, I express my most humble gratitude. TABLE OF CONTENTS List of Tables ..... List of Figures ....... List of Symbols. INTRODUCTION . DESCRIPTION OF THE STUDY AREAS . . The Llanos Region. The El Tuparro Study Area. . . . I . . . . The E1 Porvenir Study Area . . METHODS. . . . Collecting Methods . . . Examination of Specimens . . . Abundance. RESULTS. Female Reproduction. . . ...... Male Reproduction. . Annual Fecundity . .Age Structure. Homogeneity of Sex Ratios. Hemogeneity of Age Structure . . . Alternative Age Structure Regressions. Abundance. . . . . . . . . . . . ADDITIONAL OBSERVATIONS. Foods. . . . . . . . . ...... Behavior . . . . . . . . . ...... Activity. Response to Hunters. . . . . . . . . . . Page vii . viii ..... ....... 0000000000 ll 11 21 26 31 31 36 38 vi Feeding Behavior . . Communication. Social Behavior. Use of Habitat . . Burrows ..... DISCUSSION . Demographic Parameters . A General Model. . Parameter Values for the Sample of Pacas . Validity of the Assumptions ..... . . Factors Affecting Demographic Parameters . . Sensitivity of the General Model . . Hunting. . . . . . . . . . . . . Use of Space . . Food. . . . Comparisons with Other Mammals . Conclusion . . . . . . . APPENDIX.A. iMETHODS USED TO ESTIMATE AGE. APPENDIX B. METHODS USED TO ESTIMATE TIME OF BIRTH. . APPENDIX C. SENSITIVITY ANALYSIS. LIST OF REFERENCES . 102 102 105 106 108 118 119 120 123 127 128 129 132 144 149 158 Table Table Table Table Table Table Table LIST OF TABLES Numbers of pacas obtained from each collecting site. . ..... . . ................ Sex ratios of subsamples and samples of pacas . . Tests of homogeneity of regression. . . Tests of three models of regression for the combined sample of paca age structure. . . . . . ...... Estimates of abundance. Food items utilized by pacas. ...... . ..... Characteristics of eleven paca burrows. vii Page 70 76 80 97 Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure 10. 11. 12. 13. 14. 15. 16. LIST OF FIGURES Characteristics of Agouti paca . . Geographical distribution of the genus Agouti. . Locations of the study areas . Gallery forests. . . . Nbp of isohyets of total annual rainfall and number of months per year with less than ten cm rainfall . . . LANDSAT images of part of the Llanos Orientales. Seasonal distribution of rainfall near the study areas ....... Map of collecting sites for the El Tuparro study area . . . . . . . . . . . . . Map of collecting sites for the El Porvenir study area . . . . . . Uterus and ovaries of a pregnant parous paca . Thin sections of paca ovaries. . Seasonal distribution of female reproductive conditions and of estimated births ........ Distribution by age class of female reproductive conditions . . Variation in estimated mean testis volume and mean testis length .............. Age pyramids for the samples of pacas from the El Tuparro and El Porvenir study areas . Age pyramids for subsamples of pacas from El Tuparro and El Porvenir. . viii 12 14 17 19 20 22 28 48 50 52 54 57 60 Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure 18. 19. 20. 21. 22. 23. A1. A2. A3. B1. C1. C2. C3. ix Age pyramids for paca subsamples grouped by collecting method. . Age pyramid for the pacas from El Tuparro and El Pervenir combined ............. Curves of the log-linear and log-quadratic models for males. . . . . ...... Curves of the log-linear and log-quadratic models for females. . . . . . . Sketches of the paca burrows examined. . Prediction and confidence limits for juvenile survivorship . . . ........ Curves for a = 0.95 for different values of A. Micrographs of thin sections of upper first molars of pacas. . variation by weeks of the year in the amount of light-staining cementum in the outermost portion of the root of M1 for 45 adult pacas . . . Distribution of cementum tissue, and crown patterns of molariform teeth . ....... Growth curve for four pacas raised in captivity. Sensitivity of A to changes in Za and p when w = m. 0 . Curves showing the relationship between juvenile and adult survivorship using different values for maximum longevity in solving Equation (14), assuming o = 1.5, m = 0.95, and A = l. . . . . . . . . . . . . . . . . . . Demographic parameters for the population of pacas in relation to values for equal sensitivity of A to la and p for the finite and infinite longevity situations . 68 73 74 98 109 115 135 136 139 146 152 153 157 Zn um LIST OF SYMBOLS Age. The midpoint of each year class. Age at which females first give birth. Infinity. Longevity. The midpoint of the oldest age class. Finite rate of increase. Survivorship. The probability at birth of an individual sur- viving to age x. Juvenile survivorship. The probability of surviving from birth to age x = a. Annual adult survivorship. The probability of surviving from age x to age (x+l) for all x;?o. Fecundity. Expected number of female offspring per year per female of age x. IMean annual fecundity for females age XE’o. Age class. The proportion of the population that consists of individuals of age x. Product function. Summation function. The base of the natural logarithms (= 2.3026). Natural logarithm. iMicrometers (= 1.0 x 10'6 meters). INTRODUCTION The paca, Agouti paca, is a large, nocturnal, frugivorous rodent found in a variety of forest habitats from southern Mexico to northern Argentina. It belongs to the infraorder Caviomorpha (sen§u_Wood 1955), a group which first appears in the fossil record in the early Oligocene of South America. Isolated from other rodent groups until later invasions from North America when the land bridge was formed in late Pliocene or early Pleistocene (Hershkovitz 1969), the Caviomorpha have had approximately 60 million years during which to diversify. Recent species occupy a wide variety of ecological niches and have become a major component of the Neotropical mammal fauna. The larger species are especially remarkable for the convergences they demonstrate with various small ungulates of tropical Africa and Asia (Dubost 1968, Eisenberg and MCKay 1974). Striking similarities can be seen in the morphology, behavior, dietary specializations, and habitat preferences of New WOrld and Old WOrld "ecological equiva- lents,” with the paca most resembling the water chevrotain (Hyemoschus aquaticus) of central Africa. Of the several larger caviomorph species which are hunted for their flesh -- and indeed are major sources of ”bushmeat" in the Neotropical Region -- the paca is regarded as having the most flavor- ful meat. Beginning with the earliest accounts of New WOrld mammals (i333, Marcgrave 1648), naturalists have repeatedly made mention of the paca's edibility and of the methods used by local hunters to Obtain it. The genus Agouti Lacépéde 1799 is included in the family Dasy- proctidae, along with the genera Dasyprocta and.Nypprocta. The syno- nym.Cuniculus Brisson 1762 has been frequently but incorrectly used for the pacas (see Hopwood 1947). TWO species of Agguti_are currently recognized: :5' (Stictomys) taczanowskii Stolzmann 1889, occurring above approximately 2000 m elevation in the Andes of venezuela, Colombia, Ecuador, and Peru; and A, paga_(Linnaeus 1766), living in virtually all forest habitats of the New WOrld tropics below 2000 m. Eight subspecies of A, paga_have been described (Ellerman 1940, Cabrera 1961). The paca has a moderately robust and stout body. Adults typically attain a total length of 550 to 700 mm and a weight of 6.5 to 12 kg. The color of the upper parts varies from reddish brown to dark chocolate or smokey gray, with a variable number (two to seven) of irregular rows of white spots on the flanks (Figure 1(a)). The underparts are white. Pacas are characterized by the presence of unique pockets formed by a ventrolateral expansion of the maxillary bones in association with large lateral plates of the jugal portion of the zygomatic arch (Figure l(b)). An external, fur-lined cheek pouch is associated with the maxillary pocket. The zygomatic plate is larger and more rugous in males than in females (Figure 1(c)). The montane A. taczanowskii, which will not be considered in the present study, is distinguishable from A. paca by its thick, wool-like fur, by splayed digits with long phalanges and weaker, characteristically curved claws, and by a somewhat reduced zygomatic plate that is less Figure 1. Characteristics of Agouti paca. (a) Profile view of year- ling male. (b) Ventral view of head of male. The external zygomatic pouches are visible as dark apertures on either side of the lower jaw. (c) Skulls of adult male (upper) and female (lower) pacas. 4s msu 293349 rugous and distinctively compressed dorsoventrally as compared to A. paga. The approximate distributions of the two species of'Agguti are shown in Figure 2. Although A, paca appears to be restricted to for- est, it does occupy a wide variety of forest habitats, including man- grove swamps, thin riparian growth, and dense upland scrub (Dalquest 1953, Leopold 1959, Mbndolfi 1972). It occasionally becomes a pest of cultivated crops (Alvarez del Toro 1977). Pacas are terrestrial and nocturnal. They spend the day in burrows, the entrances to which they conceal with wads of leaf litter (Smythe 1970a). Their diet consists of fruits of a wide variety of sizes and textures (Smythe et_al. in press). It is somewhat difficult for the North.American biologist to appreciate the commercial importance of medium- and large-sized Neo- tropical rodents. Game biologists generally think of sporting and edible nmmmmls chiefly in terms of ungulates, some lagomorphs, and perhaps only tree squirrels among the rodents. Several surveys, how- ever, have demonstrated the importance of the meat of agoutis (Dasyf procta spp.), the capybara (Hydrochoerus hydrochaeris), and especially the paca, in rural markets. In Manaus, Brazil, for example, the meat of the paca was found to be the most popular mammal dish on local restaurant menus, and ranked behind only tapirs and peccaries in total amount consumed (wetterberg et_a1, 1976). Similar preferences were f0und among peoples of eastern Peru (Pierret and Dourejeanni 1966, 1967; Mbro 1976). Rural peoples of tropical America depend upon caviomorphs such as pacas, agoutis, capybaras, and spiny rats (family Echimyidae) as much as they do ungulates, fish, and birds for bushmeat protein. The special preference for paca meat is almost universal ‘ r .I ' . AGOUTI PACA @ AGOUTI TACZANOWSKII 20 - ‘O‘Prufi \- ..V " l .' ' ‘~ '- z‘r‘. ' I. . ‘i - . )...."_ .. . . .J‘: ,. _'," . .Qz(‘ . I. . . .¢ -. Figure 2. Geographical distribution of the genus Agouti. throughout the species' range and is amply documented in the litera- ture (g,g,, Hingston 1932, Leopold 1959, Mendez 1970, Mbndolfi 1972, Pine 1973). The importance of pacas and agoutis in Neotropical forest com- ‘munities is profound but difficult to quantify. From a survey of'mamr mals in forest habitats at Parque Nacional Guatopo in venezuela, Eisenberg §t_§1, (1980) found that pacasauuiagoutis represented 16% and 13%, respectively, of the biomass of the thirty-five species com- prising the community of non-volant mammals. Similar representations have been demonstrated for these two species in other Venezuelan habitats (ibid.) and in Panama (Eisenberg and Thorington 1973). If these figures are representative of mammalian communities elsewhere in the Neotropics, these two caviomorphs should command attention as heavily-exploited sources of food for man. Only recently has attention been focused on the need to evaluate and to manage rationally the exploitation of edible caviomorphs. To date only the massive capybara has been studied in the context of a harvested resource (Ojasti and Padilla 1972, Ojasti 1973). Smythe (1970a, 1978) has investigated the behavior and ecology of the hare- sized Central American agouti (Dasyprocta.punctata), a diurnal species in.many ways similar to the larger paca. That author has repeatedly emphasized the developing problem of overexploitation of pacas and agoutis in Central America. Even so, long-term studies using live- trapping techniques in the relatively undisturbed situation of Barro Colorado Island, Panama (Smythe et_al. in press) have thus far pro- duced only very preliminary findings on the population density and reproduction of agoutis and pacas. Much of the present knowledge of several aspects of the biology of caviomorph rodents is summarized in the reviews assembled by Rowlands and.Weir (1974). These rodents show great diversity in reproductive patterns (Weir 1974a), behavior and social organization (Kleiman 1974, Eisenberg 1974), and ecology (weir 1974b, Rowlands 1974). Nonetheless, with the notable exception of the domestic guinea pig, only a few of the more than 40 genera of caviomorphs have been studied in any detail. Other than an excellent review by Mbndolfi (1972) and detailed observations by Smythe (1970a) and Lander (1974), the literature on pacas is scattered and consists primarily of numer- ous small notes or comments found in larger treatments of systematics or natural history. Despite its economic importance, the population biology of the paca, and the effects on paca populations of exploita- tion by man, have not been studied. The principal objective of this study was to determine and describe the demographic characteristics of the paca, based on samples collected from a pOpulation in eastern Colombia. Demographic charac- teristics are patterns of the processes of birth, death, immigration, and emigration which in combination determine the numbers of individuals in a population and thus are the ultimate means by which population size changes over time. The traditional approach used in demographic studies is to determine the rate of increase of the popu- lation by actuarial methods such as life tables or projection matrices. Hewever, these methods require large volumes of data obtained under rigorously defined conditions (Caughley 1966) and are acutely sensi- tive to sampling errors (Tait and Bunnell 1980), to inaccuracies of methods used to determine individual age (Rice 1980), and to residual effects of past fluctuations in reproduction and/or mortality (e333, Laws 1969). An alternative approach is to census population numbers repeatedly over time, which does little to explain the demographic changes underlying changes in numbers. It is perhaps more useful to attempt to understand how fundamental demographic parameters combine to form an overall life-history strategy (see review by Stearns [1976]). By life-history strategy, I mean the combination of age- specific fecundity and mortality functions by which a population main- tains a favorable balance between births and deaths. Mathematical models of hypothetical life—history strategies have long been used as heuristic tools for understanding the variety of strategies seen in nature (Cole 1954, Lewontin 1965, Emlen 1966, Charnov and Shaffer 1973). A body of theory has been developed which in large measure predicts the behavior of demographic parameters under different sets of biologically-meaningful simplifying assumptions. Mere recently, however, mathematically elegant descriptions of demo- graphic relationships have been applied to natural populations of birds (Goodman 1974) and mammals (Shaffer and Tamarin 1973, Siler 1979). The following analysis uses simple mathematical models to assess the degree to which the population of pacas is likely to exhibit tem- poral variations in the values of specific parameters such as litter size, frequency of breeding activity, age of first reproduction, length of reproductive life, and age-specific patterns of mortality. The goal of the model is to rank relevant parameters in terms of the sensitivity of the overall life-history strategy to changes in each of them. This elementary analysis serves two purposes: First, as a 10 basis for more detailed studies of the relationships of natural factors, and of exploitation by man, on the overall population pro— cess of the species; and second, as a way of comparing the demographic patterns of the paca with those of other exploited species about which more is known concerning population regulation and management tech- niques. With a view toward understanding the proximate biological mecha- nisms by which ultimate deomgraphic processes operate, observations on aspects of the behavior, feeding habits, use of habitat, and burrows, of the pacas I studied were recorded as a secondary objective in the course of collecting the sample of demographic analysis. DESCRIPTION OF THE STUDY AREAS The Llanos Region Field studies were based at E1 Tuparro, Comisaria del Vichada, from November 1975 until October 1977, and at E1 Porvenir, Departamento del Meta, from April 1978 until October 1979. Both study areas are located in the Llanos Orientales (”Eastern Plains") of Colombia, 3 large lowland zone of seasonally dry savannas dissected by narrow gallery forest dendrites bordering the various streams (known locally as EEEQE) of the watershed of the Orinoco River. The Llanos covers the entire northeastern lowlands of Colombia, comprising more than 260,000 ka, or 22% of the surface area of the country. The locations of the two study areas within the region are shown in Figure 3. Brumnschweiler (1972) provides a review of the climate, soils, geomorphology, and economic demography of the region. The savanna vegetation is discussed in Beard (1953) and Blydenstein (1967), and detailed soil studies are presented in FAQ (1965). A profile view of a typical forest gallery along the lower Tomo River approximately 10 km upstream from the Orinoco is presented in Figure 4(a). The canopy, approximately 15 to 20 m above the forest floor, is in many respects similar to most of the collecting locali- ties in the two study areas. Palms are conspicuous in many areas and are more abundant in swampy galleries than in the better-drained for- est shown in Figure 4(a). This gallery is approximately 75 to 100 m 11 12 km 300 v‘ ////' / ' ' /////' ’,// //«/,/ ./ // "/7 .717, é/%7C ////// , // /? =r////////z 7/7; / , / z \ ‘ LLANOS ORIENTALES 7/? s , / 7/. 7/77 ~ r , ’/ , K / V7 // AMAZON FOREST 7 "\ '1 AN D ES M OUNTAIN s 3' / f / C/ J L 75° 70° Figure 3. Locations of the study areas: T = E1 Tuparro; rectangle shows the location of Figure 8: P = El Porvenir; rec- tangle shows the location of Figure 9. 13 Figure 4. Gallery forests: (a) Profile view of forest along the lower Tomo River near Centro Administrativo, E1 Tuparro; (b) Aerial view Of the gallery forests of the upper Tomo west of E1 TapOn, E1 Tuparro. 14 w .41.. .1, v .7" 15 wide on either side of the river, as is typical of the galleries of smaller streams. Most forests along the main rivers are considerably wider, extending for 100 to 150 m on each side. A great variety of broadleaf tree species is present, with abundant lianas and epiphytes. Relatively fewer palm species are found, but these comprise a dis- proportionate part of the flora. In large-stream galleries, the palms Attalea regia, Acrocomia spp., and Socratea exorriza virtually domi- nate the well-drained galleries, accounting for approximately 20% of the trunks present. In swampy forests, Oenocarpus mine and other Oenocarpus spp., Mauritia flexosa, and Euterpe precatoria are the dominate palms. In most areas, understory palms such as Bactris spp. and Syagurus inajai are major elements of the flora. Species of Phytolaccaceae, Marantaceae, Melastomaceae, and Heliconia abound in disturbed areas such as treefalls and at the forest edge. The topography of the Llanos is almost flat, with elevation declining from only 600 m above sea level at Villavicencio, near the base of the Andes, to approximately 250 m at the Orinoco some 800 km east of the Andes. A noteworthy feature of virtually all Llanos gal- leries is the morichal, a nearly pure stand of Mauritia flexosa palm found at the headwaters of each gallery stream. Morichales are restricted to the uppermost several hundred meters of E1 Tuparro gal- leries, but may extend for several kilometers in flatter areas of the Llanos, as is the case just north of the Meta River near the El Porvenir study area. Mauritia flexosa also occurs in occasional patches in other parts of gallery forests, especially in areas which are flooded seasonally. 16 Within the Llanos, rainfall patterns are reflected in an overall gradient in relative forest cover from.north to south, and to a lesser degree from east to west. In the driest areas, where rainfall totals less than 150 cm per year, streams are widely separated by broad savannas and their gallery forests are narrow. An aerial view of the gallery forests near El Tapén, El Tuparro, shown in Figure 4(b), gives an impression of the relationship between gallery forest and savanna in moderately dry mid-Llanos areas. Farther south, the relative amount of gallery forest gradually increases with increasing total rainfall, finally merging into closed jungle where annual rainfall totals more than 200 cm and the dry season is less than four months long. The boundary between continuous forest versus Llanos vegetation types largely coincides with the course of the Guaviare River. Approximate total rainfall isohyets and curves showing the number of months per year with less than 10 cm of rain are presented in Figure 5, adapted from Brunnschweiler (1972:11). The resultant pattern in relative amount of forest cover is strikingly shown in the LANDSAT images presented in Figure 6. Several hypotheses concerning the ori- gin and development of the Llanos vegetation are discussed in Sarmiento and Monasterio (1975). The two study areas described below differ somewhat in total annual precipitation and in the seasonal distribution of rainfall. Histograms of the seasonal patterns of rainfall for the two reporting stations closest to the study areas (Puerto Ayacucho, venezuela, and Carimagua, Meta) are shown in Figure 7. Although total rainfall is greater at E1 Tuparro, it appears to be concentrated into a slightly shorter rainy season than is the case at El Porvenir. Indeed the 17 RAINFALL REGIME IN THE LLANOS -—200 Annual Rainfall (cm) --— 3 Months with __ . <10 cm Rain .-?33'-£3.;5-5-’" ------ N. Limit of 4‘ ... a . 1:1: ;3§§:v:?;;'\ ... Figure 5. Map of isohyets of total annual rainfall (solid lines) and number of months per year with less than ten cm rainfall (broken lines). Rectangles are the study areas: T = E1 Tuparro, P = E1 Porvenir. Solid circles are population centers (C = Carimagua, O = Orocué, PA = Puerto Ayacucho, T = Centro Administrativo, V = Villavicencio) . Figure 6. 18 LANDSAT images of part of the Llanos Orientales. Forest cover appears as irregular black lines or solid black. Savanna is light gray. The left portion is Image 1088- 14320-5, taken 19 October 1972. The right portion is Image 1501-14235-5, taken 6 December 1973. A portion of the margin of Figure 8 is indicated by the open rectangle at right. The central rectangle corresponds to the mar- gins of Figure 9. 19 .mo .05 .K .Nh .mh MONTHLY RAINFALL (cm) 20 PUERTO AYACUCHO CARIMAGUA 50 - rd /. / 4O _‘ H / 7/ "‘ //, I, /5?2 / /4 Z/ 7’ / ’,:// / 30 ~ I ' 5:? '/ // ,/ . l// / 9/76 / ’ 4/ ’ ’ / /,‘ z); ’ / / . 47”.; / // //, // /// 20 _ '/ )4?" 7/ // 4 7; . / ;/ //,/,, / ’/Z : ,l ’ <’)// / / / ,6; / //I 7/, 7/7 '// , /’// // /6/// /’ //; /// 27 / 1O — // /// /. /// -// /’// / /,/-/ ’/ /// i2 , // //// r x1; /// / /////, / // 7 / / / 7/ /, ,/ // / 3/7, / // 5 I I // ,/ ’; / A/ ‘/’/,~4//1 4/77' / / / y z / / ,4. [a 7/ / ”/ M I I I T I I I I I I I l l I I I l 1 l l .I F Rd A IA J .J A 55 0 l4 0 .J F IA A AA J .l A 55 0 II D Figure 7. Seasonal distribution of rainfall near the study areas. Histograms of average monthly rainfall for: Puerto Ayacucho (left), 56 km N of Centro Administrativo, for the years 1961-1970 (data from Instituto Geografico Agustin Codazzi 1973); Carimagua (right), 30 km S of E1 Porvenir, for the years 1973-1978 (data courtesy of Dr. J. M. Spain of CIAT). 21 severity and length of the dry season is effectively equal for both study areas. The El Tuparro wet season begins in late April and irregular strong rains continue through June. After a brief drier period in late JUne and early July, rains are more frequent but less intense, with light rains occurring almost daily in August and September. In contrast, at El Porvenir the rainy season is distin- guished by two periods of heavy rains separated by a pronounced dry spell or veranillo of two to three weeks' duration in late July and early August. The veranillo, not apparent in the histogram of Figure 7, is a totally dry hiatus in the middle of the wet season. The E1 Tuparro Study Area As an administrative unit ("faunistic territory”), Territorio Faunistico El Tuparro is part of the system of national parks and natural areas administered by Instituto del Desarollo de los Recursos Naturales Renovables (INDERENA), a division of the Mfinistry of Agriculture of the Government of Colombia. Comprising some 250,000 ha, Bl Tuparro is located in eastern Vichada between the Tomo River to the north and the Tuparro River to the south, and extends from the Orinoco on the east to a site called El Tap6n, on the Tomo some 180 km to the west of the Orinoco. The majority of the pacas collected in this study area were taken from just outside the Park itself, princi- pally from small tributaries south of the Tuparro River. INDERENA facilities for the administration of the Park are located at three sites (Figure 8): (1) Centro Administrativo (5°18‘N, 67°52'W), the park headquarters; (2) El Tap6n, at the western end of the Park; and (3) on the south bank of the Tuparro River at its mouth 22 Rio Orinoco ® ~ Centro Administrativo CD-‘El Tapdn _____ Road Scale Figure 8. Map of collecting sites for the E1 Tuparro study area. Numerals correspond to collecting sites discussed in the text. 23 on the Orinoco. The remainder of the Park is uninhabited except for several small bands of nomadic Indians. Ecologically, E1 Tuparro occupies the extreme southeast corner of the Llanos Orientales near the point where savanna merges into con— tinuous forest. Thus the gradient of increasing amounts of forest is pronounced along the eastern end of the Park. Less than 10% of the watershed of the lower Tomo River is forest whereas more than 30% of the Tuparro River watershed is forest. 'Vincelli (in press) has described the plant communities of Territorio Faunistico E1 TUparro, and other aspects of the vegetation are discussed by Defler (1980). Flat topography and friable soil conditions often combine to result in serpentine rivers that change course frequently, leaving "oxbow" lakes and associated swamps along old riverbeds. This is especially true of the collecting localities along the Tuparro River. Two additional factors contribute to habitat diversity near the eastern end of the El Tuparro study area. First, the Orinoco itself rises and falls approximately 15 to 18 m over the annual rainfall cycle, and since this cycle is due to rainfall patterns some 300 to 500 km to the south, the river rises well before the start of the local rainy season and falls to its lowest level just shortly after the local rains end in October. Forests on the lower 30 km of the Tuparro River appear to reflect this factor more than the local rain- fall cycle. The lower Tomo River, occupying higher, slightly better drained terrain, is much less markedly influenced by the level of the Orinoco. The second locally important factor is the presence of a row of small granitic outcrops (tepui§_in the sense of Haffer 1974) located near the Orinoco. As minor extensions of the adjacent Guiana 24 Shield of Venezuela, these hills rise approximately 200 m above the surrounding plain and support a distinct xeric forest as a result of the low water holding capacity of their thin soils. Specific collecting sites for the E1 Tuparro study area are shown in Figure 8. Although widely separated, the nine sites fall into five groups on the basis of habitat characteristics and general location. Collecting sites 1, 2, and 3 (Cafio.Alto, Cafio Macazaba, and Cafio Tuparrito), along the upper Tuparro River, are the wettest sites visited. This portion of the river is characterized by an extremely serpentine course with numerous oxbow lakes and other swamps. Swamps of these oxbows and of the numerous small tributaries of the river tend to coalesce into broad floodplains near the river itself. Although far enough upstream to be isolated from the effects of the yearly rise and fall in the level of the Orinoco, this area is very flat and so large parts of it remain flooded for three to four months during each rainy season. Specimens were collected from five small navigable streams, including two unnamed streams outside the El Tuparro Park opposite Cafio Macazaba. The forests of these streams tend to have relatively fewer palms present than is the case for most gallery forests in the Llanos. There appears to be a corresponding increase in the diversity of broadleaf tree species. This distinction reflects the same climatic factors which result in continuous wet for— est beginning only 30 km to the south at the Vichada River. Sites 4 and 5 are on the middle portion of the Tuparro River below Cafio Tuparrito. Cafio Lapa (site 4) is one of the most acces- sible and suitable sites in the region for the hunting of pacas. The lower 4 km of this stream consist of low-lying swamps which are 25 flooded for more than six months of the year. A high, well-drained, small stream is separated from the swamps by a small rapids. The stream.maintains relatively constant water level in a rich but narrow (less than 200 m) gallery forest, from which all Cafio Lapa specimens were obtained. Oenocarpus spp., Socratea, and Bactris palms are particularly abundant there, along with a variety of broadleaf trees. The forest understory is quite open. Signs of other hunting parties were seen during all five of my visits to the area. The five specimens taken from well-drained, uplifted banks along portions of the lower Tuparro River (site 5) below Cafio Tuparrito represent a sample from habitats very similar to upper Cafio Lapa. These forests are not subjected to seasonal flooding as are nearby oxbows, and are in fact somewhat narrow (generally less than 200 m) for main-river galleries. Site 6 is Cafio Rana, a tiny seasonal stream less than 3 km long. It drains seasonally-flooded swamp forests and morichal just west of the tepui§_uplands on the south bank of the Tuparro near its mouth. The stream disappears completely for several weeks in February and March, and the forest which it drains is completely innundated by the Tuparro, influenced by the Orinoco just 4 km to the east, from late June to late August. Puerto Alegre (site 7) is a large patch of forest drained by three separate small tributaries of the Tomo. These streams are not quite flooded in mid-rainy season, but in the dry season are so dry as to be inaccessible by canoe. Palms are not common, presumably as a result of the total lack of saturated soil conditions, although a belt of Mauritia flexosa pabm is found where seasonal flooding occurs at 26 the margins of the surrounding savanna. The lower portions of Cafio Peinillas (site 8) are in most respects similar to Puerto Alegre except that the stream is larger and does not dry out in the dry season. Site 9, Cafio Hormiga, is a fairly large stream located near the INDERENA guard station at E1 Tap6n. The topography there is diverse, with bluffs of old alluvium rising slightly above narrow (less than 200 m) gallery forests. During the dry season, the stream is well below the gallery floor. The stream rises dramatically in the wet season. Palms are not particularly abundant except for Oenocarpus, Socratea, and Bactris. The El Porvenir Study Area In March and April of 1977, I participated in an effort by investigators from Tulane University's Centro Internacional de Investigaciones Médicas (CIDEIM) to collect a series of pacas from the vicinity of the Centro Internacional de Agricultura Tropical (CIAT) agricultural experiment station at Carimagua, Meta. Twenty-four pacas were obtained as part of a continuing study of cestode parasites of the genus Echinococcus. Most specimens collected at that time were examined by me. After continued logistical problems at El TUparro, and in view of the success achieved at Carimagua, it was decided to move the paca study to the village of El Porvenir (4°45'N, 71°25'S), located on the south bank of the Meta River 30 km north of Carimagua. In April of 1978, a field station, with.basic laboratory facilities, was set up by CIDEIM, and field work was based there until termination of the paca study in October of 1979. 27 Although still very sparsely populated, the area around E1 Porvenir has been settled for centuries. The only significant indus- try is cattle ranching. Virtually all of the savanna land in the area is dedicated to ranching. Collecting sites involved eight large ranches and also small portions of the Carimagua property. The typical ranching operation consists of 6,000 to 30,000 ha. A limited amount of agriculture is practiced on the rich, broad floodplains of the Meta River, with corn, rice, beans, and cassava being the prin- cipal crops. The town of Orocué (pop. ca. 3,000), Intendencia del Casanare, lies 17 km downstream from El Porvenir. Specific collecting localities in the vicinity of E1 Porvenir are somewhat more scattered than was the case for E1 Tuparro. This results primarily from the cooperation of several local hunters who brought in specimens from sites of their own choosing. In all cases except for one ranch in Casanare upstream from El Porvenir, such sites were subsequently visited by me. All of the localities fall quite naturally into four categories, treated below in the order in which they are numbered in Figure 9. Sites 1 through 4 are parts of Cafio Caviona, a small tributary of the Meta River which originates in part in Carimagua and also in the ranches known as E1 Tranquero, E1 Pifial, and Hato Caviona. The localities used are all readily accessible by road from El Porvenir. Site 1 (Puente Gloria) is most significant since it is the locality which was most intensively hunted by me and is thus where many of my observations of pacas occurred. This is a single branch of forest approximately 2% km long extending to the north of Cafio Caviona from a point 4 km NE of the bridge where the road from Carimagua crosses the 28 io Vichada (I. ,. 1 . i C - [j 9P -- El Porvenir ------ Road ,.--,'~~‘ 0C " Carimagua .r \ /' .’ \“‘ : 0 km so 7 ,- ' . Figure 9. Map of collecting sites for the E1 Porvenir study area. NUmerals correspond to collecting sites discussed in the text . 29 main stream. Site 2 (La Arepa) is a much larger branch off of the south side of Cafio Caviona slightly downstream from site 1 and extend- ing 6 to 7 km toward Carimagua. Site 3 (Rincén Masato) is a segment of Cafio Caviona west of Puente Gloria and downstream.from.site l, and site 4 (Upper Cafio Caviona) includes several points west of Puente Gloria and their associated minor galleries. West of Puente Gloria, the forests of this stream are low—lying and drain poorly. There are numerous small morichales and streamside stands of Mauritia palm. The gallery forest along Cafio Caviona itself is narrow (generally less than 100 m) yet the forest is swampy and has a dense understory with abundant Bactris and Socratea palms and var- ious shrubs. The surrounding savanna is high and well-drained, but the course of the stream is so flat as to result in temporary flooding of much of the gallery forest after most heavy rains. Sites 1 and 2 are associated with slightly higher savanna and thus are drained rapidly after rains. Their forests are more open and have abundant Attalea and Socratea palm. Site 5 (Paso Ancho) is a part of the floodplain of the Meta River located approximately 7 km west of E1 Porvenir. The eleven pacas col- lected there were all taken from the vicinity of a grove of mango trees located in a small farm where corn and rice were grown. Sites 6 and 7 refer to all tributaries of the Mpco River sampled in the course of the study. Many of the specimens from site 6 were taken from four small galleries within the Carimagua station property during March and April of 1977. During 1978-79 hunting on Carimagua land was discouraged, but some pacas were collected from minor tribu- taries of the Mbco near the mouth of Cafio Carimagua. Site 7 is Cafio 3O Limici, the small tributary of Cafio Carimagua just upstream from its mouth on the Muco. During the last five months of the study, Sr. Genaro Useche Osorio, a native of the region and a former assistant at the CIDEIM field station, established a homestead on Cafio Limici. The 42 pacas from that site were all collected by Sr. Useche, whom I had previously trained to take appropriate data and to preserve skulls, genital tracts, and parasite materials. The forests of the Muco in this area are in many respects similar to Cafio Caviona, except that the larger main stream fluctuates less dramatically in water level and supports much wider forest galleries, typically 300 m wide. Numerous oxbows and other swamps are inundated by the Muco during the wet season, but for much of the dry season, no fOrested areas are flooded. Palms associated with seasonal flooding are somewhat more abundant than in Cafio Caviona, especially species of Oenocarpus and Mauritiella. Site 8 (Casanare) refers to several localities in the Intendencia del Casanare, the political division north of the Meta River. Twelve specimens were collected by a local fisherman from four ranches near the Meta River (Hato El Boral, Carocaro, and two smaller ranches on Cafio Gurripe) and examined by me. The gallery forests are low and seasonally flooded. Scheelea and.Attalea palms predominate. Savannas are very flat and borad, and separated by narrow (less than 150 m) gallery forests. Rainy-season flooding is extensive. NEHHODS Collecting.Methods It was planned originally that two parallel population assess- ments would be carried out; a live-trapping regime for estimating density, disappearance rates, and fecundity parameters; and a series of four quarterly collecting trips employing local hunters to obtain samples of 100 individuals each three months, for a sample of the standing age distribution and the occurrence of reproductive activity by season and by age. The live—trapping program, using twenty-five 22.5x22.5x80 cm double-door wire mesh traps over a period of five months, was abandoned after approximately 800 trap-nights yielded no pacas. Plans for quarterly samples were changed as it became apparent that sufficient numbers of pacas would be more difficult (and much more expensive) to obtain than had been anticipated. For these rea- sons, the principal approach used in the study was to hunt pacas when- ever logistical and financial conditions permitted. Local residents were hired as hunters, and during the course of the study, I learned much from them about hunting techniques, eventually collecting 49 specimens myself. The traditional hunting methods used by residents of the study areas were used to obtain all specimens examined. Mbst were taken by local hunters who worked in collaboration with.me and were paid on a per-specimen basis as funds allowed, and were also given the meat from 31 32 the pacas that they shot. Sites for hunting trips were chosen in consultation with these hunters. This practice had several advantages. I was able to learn which factors of habitat, season, and transporta- tion were of importance to local hunting methods, and usually discus- sions with the hunters provided details about the recent history of the use by hunters of the site. Often my travel resources enabled us to hunt at sites that had not been visited by local hunters for several years. Hunting practices were dictated by the habits of the animal being hunted. In the Llanos, there are no government regulations or sport- ing traditions constraining hunters, since hunting is generally under— taken to meet a major dietary need and not conducted for sport. Wea- pons used are shotguns and small—caliber rifles of all descriptions, including homemade muzzle-loaders. Finding a paca to shoot involves the use of either a six-volt battery-powered headlamp at night, or a tracking dog in the daytime. One can readily learn to recognize pacas at night by the strik— ingly bright orange-red reflection of the hunter's lamp in the animal's eyes. A uniform, large (ca. 10 mm) disk is seen, and it can easily be distinguished from other often-encountered forest animals by color and luminosity: The eyes of the paca do not ”sparkle” as do reptile eyes, but instead reflect a soft, uniformly bright light. Often only the eye can be detected in dense cover, yet this is usually sufficient for the hunter to identify his quarry. Usually pacas were actively pursued in the course of a night's hunting. In advance, either a portion of stream (several km) was cleared of obstructions, then traversed repeatedly by canoe (the 33 principal method used at E1 Tuparro), or a narrow trail (usually about one km long, often longer) was cut along or near the stream bank, then partially cleared of leaf litter and traversed on foot (at E1 Porvenir). Some hunters preferred either to clear a short trail (less than 50 m) or to construct a simple platform low in a tree, near a fruitfall or clear sign of paca activity, and wait quietly for a paca to approach. At best, only one paca per night was obtained by these passive methods. The primary considerations in selecting gal- leries for cutting hunting trails were the ease of transportation to the site and the acquiring of maximum visibility into the surrounding forest. This effectively restricted hunting on foot to the drier, more open, mature galleries -- much less than half of the forest in the Llanos. On the other hand, relatively few streams are suffi- ciently navigable for hunting by canoe, so that except in flat, wet regions such as E1 Tuparro, even less gallery forest is traversible by this method. Sign of paca activity and the presence of concentrated fruitfalls were always considered in selecting galleries for hunting, but were thought by the hunters to be of relatively minor importance. For hunting by canoe, the first consideration was usually the water level relative to the bank of the stream. In a flooded forest, the pacas would be remote from the navigable channel of the stream, while at low water, the hunter would not be able to see the forest floor above the stream bank. Experienced hunters in the E1 Tuparro area were very astute at visiting particular streams during those few weeks of the year when the water level was especially suitable. 34 Dogs trained to recognize paca scent can be effective, but truly skilled dogs are rare (and considered quite valuable), and so my experience with them was, unfortunately, limited. Five different dogs were used with some degree of success in this study. The procedure involved two to four men walking in the forest in the morning with one or more dogs which tracked paca scent until they located an occupied burrow. Small dogs could enter the burrow and thereby (hopefully) induce the paca to leave in search of other shelter. Three of the dogs I used were too large for this, but harrassment in the form of dogs barking or sticks poked down burrow entrances usually was suffi- cient to convince the paca to move into the open. It would then some- times seek other cover, but more generally headed for the nearest stream and tried to lose its pursuers by swimming underwater. Usually the animal could subsequently be located by carefully probing under overhanging streamside vegetation. Only four of the 42 pacas obtained using dogs had to be literally dug out of their burrows. The remainder were collected while the animals were swimming or when driven from above-ground cover. The use of dogs has several advantages over other methods. There is no need for the hunter to be especially quiet (noise being a signi- ficant factor when hunting at night) since this activity is carried out in the daytime when the animals are in their burrows. Once located, burrows can be excavated to obtain information on their structure and locations in the habitat. Eleven burrows, from eight of which the occupant was collected, were partially or fully excavated in the course of the study. The greatest advantage to the use of dogs, however, is that the hunter has a much better opportunity to 35 capture the paca alive, either in the process of burrow excavation or by grabbing the paca as it attempts to escape by swimming. Six young pacas were captured.while swimming. Seven adults were also captured but none became well adjusted to captivity. All but one, a young adult male kept for fOur months at El Porvenir, had to be sacrificed because of injuries suffered during capture or in captivity. The use of dogs is also inefficient, especially if the dog is not particularly skilled at recognizing paca scent. The most successful day of hunting using dogs yielded two adult females, each with small offspring. On only three other occasions were two pacas taken with dogs in a single day. In contrast, one skilled hunter at E1 Tuparro shot as many as seven pacas in an exceptionally good night of hunting by canoe. It was not unusual for one hunter to collect two or three pacas in one night while hunting on foot along trails. I invested 18 days in trying out dogs that were unable to locate pacas at all. At E1 Porvenir, a standing offer of 300 pesos (U.S. $10 approxi- mately) was made to local people for each.paca brought in fresh for examination. Except for nearby skilled hunters with readily-available transportation, this was not sufficient incentive, and only 24 speci- mens (many incomplete) were obtained in this manner. IMUch greater success (46 specimens) resulted from the cooperation of Sr. Genaro Useche, a local hunter trained by me to examine specimens, record data, and preserve materials from.pacas collected in my absence, from his homestead on Cafio Limici from May to October of 1979. Mbst of the pacas collected with the aid of tracking dogs were obtained by Sr. Useche. 36 Cursory notes were kept of observations made while hunting: HOw many pacas seen, what they were doing when encountered, what foods they were eating, the habitat in which they were seen, times of encounters, other pacas or other mammals seen together with them, the animal's reactions to the hunter and flight behaviors, vocalizations, or any unusual circumstances or behaviors. I personally saw approxi— mately twice as many pacas as I was able to collect, and hunters work- ing with me usually contributed observations on and descriptions of their own encounters with pacas. Some comments were also noted on the types and compositions of forests entered, hydrology of the area, stream levels, densities of ground cover, trees producing fruit, and tracks, runways and related signs of pacas and other mammals. Examination of Specimens A detailed necropsy form was prepared for recording external and reproductive characteristics, tooth-wear patterns, fat deposits, tis- sues preserved, and any abnormalities. These data were later trans- ferred to key-sort cards, indexed by weeks one through fifty-two of the calendar year, which were then used in collating and summarizing all results. Specimens were examined in detail within sixteen hours of the time of death. Each.paca was weighed with a spring scale (ZebcoR De-Liar) to 0.1 kg. Standard mammalogical measurements were taken in mm using a steel tape. Whenever possible, photographs of the pelage were taken in profile view. Any lesions or abnormalities were noted. The relative extent of fat deposits was recorded and internal organs were briefly examined for pathologies or parasites. Stomach contents 37 were classified as either fruit pulp, foliage, animal remains, or other matter. In some cases, the species of fruit present in the stomach was determined on the basis of color, texture, and/or characteristic fruit parts. Length and width of the testes and seminal vesicles were recorded in mm, and selected specimens were preserved in ten percent formalin for sectioning. vaginae were inspected for copulatory plugs. The position and size of each conceptus was noted. Each fetus was weighed to the nearest gram and the total length measured in mm. Small fetuses (less than 20 g) were subsequently weighed to 0.1 g. For very small conceptuses, the diameter of the entire sphere of fetal mem- branes was measured from the outside of the uterus and recorded in mm. The method of Huggett and Widdas (1951, and see Appendix B) was used to estimate fetal age. The degree of distortion and rugosity of the uterus (placental scarring) was noted and the greatest width of the uterus was recorded in mm. The uteri, together with ovaries and fetuses, were preserved in formalin. Females were considered to be lactating if milk could be expressed by squeezing the nipples with fOrceps. The breast tissue was also incised to check for milk. The time at which birth had occurred was estimated for all young pacas (less than 5 kg), using growth data based on four pacas raised in captivity (see Appendix B). .All ovaries and selected testes, seminal vesicles, and uteri were sectioned for detailed study of reproductive condition. Representa- tive sections of gonads, uteri, and pathological tissures from El Tuparro pacas were prepared by Dr. F. Lozano of the veterinary Medical Research Laboratory (LIMV) of Instituto Colombiano Agropecuario in 38 Bogota. Ovaries of El Porvenir specimens were serially sectioned at 0.1 mm intervals, and representative sections of placental scars and new conceptuses were prepared at CIDEIM-Cali. Right upper first molars were removed for use in determining individual age on the basis of annulations in the cementum tissue. The teeth were decalcified, sectioned at 12 um, and stained with haema- toxylin and esosin, by Matson's Commercial Microtechnique (Box 308, Nfilltown, NH‘59851). Development of the techniques used to determine age is detailed in Appendix A. Skulls from 118 specimens (MSU 29363 - 29485), three skins, and all fetuses and tissue sections, are housed in the research collec- tions at The Museum, Michigan State university. Eighty-seven skulls, three fluidrpreserved specimens, and seven skins were donated to the INDERENA collections in Bogota. Four skulls were lost. Abundance Discrete forest galleries at three sites (Cafio Rana, Puente Gloria, and Cafio Limici) were sampled for pacas with such intensity that I used the number of specimens collected at each as a crude esti- mate of the number of pacas originally present in those areas. The surface area of the forest sampled at Cafio Rana was estimated from measurements taken on the ground. Areas from Puente Gloria and Cafio Limici forest were measured from LANDSAT images 1088—14320-5 and 1501- 14235-5 (EROS Data Center, Sioux Falls, SD 57198), presented above in Figure 6, page 19. Density was estimated as the number of pacas collected divided by the area sampled. 39 Estimates of the amount of forest present (as percent of surface area) were obtained by different methods fer the two study areas. An 11" by 14" photograph of an aerial photo array of the El Tuparro study area was obtained from INDERENA. .A transparent grid of 625, l-mm2 squares was laid over representative portions of the photograph. Percent forest was calculated as the number of squares which contained at least 0.5 mm2 of forest, divided by 625, times 100%. For the E1 Porvenir study area, the LANDSAT images were evaluated using the Spatial Data Systems Datacolor 704 Color Enhancer-Densitometer video system housed at the Department of Geography, Michigan State University. The densitometer distinguishes between forest and savanna portions of the image and computes the proportion of each in any selected portion of the image. To estimate the total number of pacas in the vicinity of each study area, the total surface area of forest was estimated by the above methods and multiplied by the density estimates obtained for the three intensively hunted collecting sites. Suitable LANDSAT images are not available for most of the El Tuparro study area, and aerial photos of the El Porvenir area were not avail- able to me through INDERENA. Statistical tests and mathematical models are described where they are used in the following sections. The primary statistical reference used was Steele and Torrie (1960). Calculations were per- fOrmed on the wang 600-14 programmable calculator housed at The MUSeum, Michigan State University, using pre-recorded statistical programs developed by Wang and also my own programs. RESULTS In total, 213 pacas were examined in the course of the study. Numbers of specimens from each of the collecting sites, together with the primary collecting method(s) used at each site, are listed in Table l. Seventy-five pacas were obtained by hunting from canoes, 96 by hunting on foot, and 42 by tracking with dogs. Eight specimens (five males and three females) are excluded from.much of the following analysis because I was unable to determine their ages due to loss of appropriate materials. Female Reproduction In order to obtain a reasonably complete representation of the annual cycle, all 105 females (from both study areas and for all four years of field studies) were combined for the analysis of reproduction and the estimation of fecundity. The two study areas are approxi- mately 350 km apart (see Figure, page 12). Nevertheless, since rain- fall patterns and habitats are fundamentally similar for the two areas, I must assume that there was no significant difference in reproductive pattern for pacas between them. Fifty-five of the 57 females that were one year old or older were reproductively active (either pregnant or recently parous) when col- lected, irrespective of the time of the year. There were 69 females as heavy or heavier than the lightest (6.5 kg) pregnant female. Of 40 41 TABLE 1. Numbers of Pacas Obtained from.Each Collecting Site. STUDY AREA Collecting Name" Of SPeCimens Collecting Site .MethodsC Nbles Females Both EL TUPARROa 49 32 81 l) Cafio Alto C 11 6 17 2) Cano Macazaba C 2 2 4 3) Cafio Tuparrito C 5 1 6 4) Lower Rio Tuparro C,D 3 2 5 5) Cafio Lapa C 7 9 16 6) Cafio Rana C 6 4 10 7) Puerto Alegre C 9 3 12 8) Cafio Peinillas C 1 0 l 9) Cafio Hormiga D,F 5 5 10 BL PORVENIRb 59 73 132 1) Puente Gloria F,D 5 5 10 2) La Arepa F 0 5 S 3) RincOn Masato F 5 4 9 4) Upper Cano Caviona F,D 12 10 22 5) Paso Ancho F 6 5 ll 6) Carimagua, Rio Muco F,C 10 11 21 7) Cano Limici D,F 17 25 42 8) Ca anare C,D 4 8 12 TOTALS 108 105 213 aSite numbers for El Tuparro correspond to those in Figure 8. bSite numbers for El Porvenir correspond to those in Figure 9. CC = in canoe; F = on foot; D = with dog 42 these, 42 were pregnant. Twelve pregnant and 27 non-pregnant individuals showed placental scarring of the uterus. Three of the 12 females that were both pregnant and recently parous were also lactat— ing. One of 5 females each of which was collected together with her offspring was pregnant. This female had a conceptus measuring 17 mm in diameter and was lactating. Her young (a male) weighed 1.8 kg, which indicated that it had been born three to four weeks previously (see Appendix B). The parous horn of the uterus was distinctly wrinkled and had a maximum diameter of 23 mm, which suggests that the less-distended but distinctly wrinkled uteri observed in other speci— mens enable detection of a parous condition for a minimum of five to six weeks post-partum. The conspicuous conceptus in this same female indicated that she had conceived approximately two weeks after giving birth. Thus a post-partum estrus is possible, and lactation does not necessarily inhibit subsequent conception. From observations on 31 births by captive pacas, Matamoros (pers. comm. 1980) concluded that estrus occurs shortly after birth and again toward the end of the period of lactation. Only 21 of the 39 females with placental scarring were lactating, including 3 that were also pregnant. Six of those that were not lac- tating had large (>40 mm) wrinkled uteri indicative of recent partu- rition, suggesting that early mortality of neonates had occurred in those six cases. No lactating female was collected that did not still show uterine evidence of parturition, implying that lactation does not exceed the period of regression of the parous horn of the uterus, which apparently may take five to six weeks. No evidence of intra- uterine mortality was found. 43 Examination of sections of ovaries provided additional indica- tions of post-partum estrus. In addition to a large corpus luteum of pregnancy (CLP) in the ovary of the pregnant horn of the uterus, numerous small accessory corpora lutea (ACLs) were found in both ovaries of all pregnant females. The CLPs of females in early preg- nancy were composed entirely of large glandular cells while the.ACLs were beginning to develop from unovulated follicles by a thickening of the theca interns. The ACLs of late pregnancy were irregular and had thin thecal layers, and the CLPs were large. Maturing follicles were seen in the ovary of the non-pregnant horn. Large vesicular follicles such as these were not seen in the ovary of the pregnant side, even for near-term pregnancies. The CLP of the parous horn was persistent for females that were lactating, while the ACLs were degenerating. For these females, the cells of the CLP were large and vacuolar with small nuclei and could be distinguished readily from the cells of the CLP during pregnancy. For non—lactating parous females, the CLP was irregular and degenerating, with a loss of the thecal layers. Mbssman (1966) described similar differences between luteal cells for lactat- ing parous squirrels Tamiasciurus husonicus. In none of the 36 females with placental scarring did the condition of the ovaries indicate the presence of a new pregnancy in the parous horn. Study of the ovaries and uteri revealed no pregnancies that had not been found by gross examination alone. The genital tract and ovaries of a female that was both pregnant and parous are presented in Figure 10, where (a) is a ventral view of the uterus showing the presence of a new conceptus in the right horn (left side of the figure) of the uterus and marked wrinkling of the 44 Figure 10. Uterus and ovaries of a pregnant parous paca. (a) Ventral view of uterus. C = conceptus, P = placental scarring. (b) Thin section of left ovary. (c) Thin section of right ovary. Both sections 7 am, x10. Symbols explained in the text. 45 46 left horn. Sections of the ovaries of the parous and of the pregnant side are shown in (b) and (c), respectively. The parous ovary con- tains numerous irregular, small ACLs and the large CLP of the previous pregnancy, all of which have lost their thecal layers. Four luteiniz- ing follicles (LF) are also seen, with the dark follicle cell layer regressing from the thickening thecal layer indicating that these follicles are in the process of becoming the ACLs of the new pregnancy. These features are also seen in the ovary of the pregnant side horn in (c), where the large CLP is seen to be completely filled with luteal cells. The ovaries in Figure 11 (a) and (b) are from a female in the fourteenth (near-term) week of pregnancy. The large CLP is seen in the ovary of the pregnant side (a) along with numerous ACLs. The ovary of the non-pregnant side (b) contains ACLs and also maturing follicles (MF) which were not seen in the ovary of the pregnant side. Apparently, after parturition ova can be shed by the ovary of the non- pregnant horn but not by the ovary of the newly-parous side. A functional asymmetry of ovaries that is suggestive of this situation has been described for the porcupine Erethizon dorsatus (Messman and Judas 1949), 21 species closely allied to the caviomorphs. The seasonal distribution of reproductive activity for all females 6.5 kg or heavier is shown in Figure 12 for pregnant females (a) and for females with placental scarring (b). Although sample sizes are small and varied, as indicated at each point in (b), both pregnant and parous females were collected in ten of the thirteen fOureweek intervals of the year. The three exceptions are for samples of only two, three, and five animals per interval, and so should not Figure 11. 47 Thin sections of paca ovaries. (a) Ovary of pregnant side for a female in the fourteenth week of gestation. CLP = Corpus luteum of pregnancy. .A = accessory corpus luteum. (b) Ovary of non-pregnant side for the same Specimen. F = maturing follicle. (c) Left ovary of adult paca that evinced no reproductive activity. F = atretic follicle. All sections 7 um, XlO. 48 Figure 12. 49 Seasonal distribution of female reproductive conditions and of estimated births. (a) Percent of adult G26.5 kg) females pregnant. (b) Percent of adult females with placental scarring of the uterus. Numerals indicate the number of adult females for each four-week interval. (c) Numbers of pregnant females (diagonals) and young less than 5 kg (stipple) collected in each four-week interval. (d) Estimated times of birth for the specimens presented in (c). of 99 % of 9? NUMBER OF INDIVIDUALS 1OO--I 75- 50— 25- Pregnant v 100- 75- 50- 2-5 10— 10— 1 . . m 7 Placental SCClI' I I 1 I I C. Week CoIIectecI .\ \ \\‘\\2 .\ \ \ \ . ~\ » *4 \ \‘ \\ \‘- \\‘\ \.\\\\‘\:‘ \\\. i _\ \ I' ~ 2 \ ,2 I? \\\\\\ ‘1 \. c \ ~ \\1\:. x ‘\ '- .\ ,\ x \‘e .l . \ x \ \. ».\\\ \\\‘\ ":‘7"‘; . \ . \ . ‘ -\ \ ‘- . \\ :......... \\ \\\I\\ I .\~ I‘ I \II I‘L". 'a 1 \ \\\\ x » , \\ ~.\\ ,g..‘.o.“ . \ \ .\\\\ » x \ . \kx "4' u..-‘ \ N‘ ‘\ ‘ \ "-‘ :::::: '-.-::-::~ . \ I .- v o ‘ ’;\\ \ \\ :I\ "‘2‘". :.::O:Oe F I. . """" 24 ‘0 "MI: “'33" “ {ind-t. ’32:}: . ' - - 121‘ '.‘ PF.'.v.t."e. a - O:‘:u .L‘l:'. 'j-I. Os 'z. ‘21.... '11....‘..‘ \ ‘ 51,, 6 . i7 , ,' 4‘ //)/ /. / , / L hhhhhhh 1 )l‘ ''''''' ,“n-JFI" 51"!) .e'tvia a'l'i‘. ..' . l <‘:':’ / " ‘1’. / / I.‘ '0'!‘ '- s:,.i.:.; b7. 9" ’o ,. an”; .1; ........ ................. ‘‘‘‘‘‘‘‘ I," '..II 9. 'I '''' J ‘ +1.. a! ‘y' .:.;‘,;1’ [/55 "5‘4 / //,/// " Fk‘ /" //I :;"a'}. , / //./////, O‘. I e '1 77777 I 51 be taken to indicate temporal peaks of one condition or the other. Only 3 of the 69 females large enough to be considered potentially reproductive showed no sign of reproductive activity. These are indi- cated by the asterisks in Figure 12(a). One of the two collected in August was less than one year old, and the specimen from October was a yearling. Both of these had thin (<6 mm) uteri indicative of a nulliparous condition. Only one fully—adult female, a five-year-old collected at Cafio Caviona in August of 1978, was neither pregnant nor recently parous at the time of collection. A section of the left ovary of this individual is shown in Figure 11(c). Only a few fol- licles in various stages of atresia are seen. The absence of corpora lutea suggests that this individual had not conceived for some time. The right ovary is virtually identical in these respects. The gestation period of the paca is on the order of 116 days (Eisenberg pers. comm. 1977, Lander 1974, Kleiman §t_al. 1980). Since gestation is so long, the presence of fetuses at all times of the year does not rule out the possibility of a seasonal clustering of births. To test for such a possibility, the times when pregnant females and young pacas (<5 kg) were collected are graphed in Figure 12(c), and the estimated time of birth (see Appendix B) in Figure 12(d). No seasonal pattern is evident. Indeed, at least three birth estimates fall into each of the thirteen four-week intervals. The occurrence by age of the various female reproductive condi- tions is presented in Figure 13. Again, no particular pattern is evident other than that virtually all adults are reproductively active. A.higher proportion of simple pregnancies seems to occur in the younger age classes, but inference is limited by the small sample 52 Pregnant + Plac. scar \ \ Plac. scar + Lactating .3: Pregnant Placental scar t n a n g e r P + g .m t a t C a I— + r a C S .m t n e C m P ii riu\\\\\\\\\\§\ \\\\\\\\\\\\\\\\\\\\\\\\\ ¢%%%%%%%%%%%%%w. lmwmmmmwwu. \\\\\\\\\\\\.m /////// mm” 0 :0 I O O O O O O O O O ..... %é\\ mm m - \\ .n..u_nmaam O O \\\\\\\\i "m b H30 /////// <\\\\\\ . 75 5 ezmomma 12 10 years AGE he abscissa represent the number of females in each age class. The column at right is class of female reproductive condi- for ages 1 through 12 combined. tions. Numbers above t Figure 13. Distribution by age 53 sizes. Tests of the z-transformed normal approximations of the pro- portion pregnant in each age class (irreSpective of placental scar- ring) against the overall proportion of 42 pregnant and 27 non- pregnant adults showed no adult age class to be different from.the overall proportion (P > 0.2). It is noteworthy, however, that only 1 of the 9 reproductively active females less than one year of age is parous, versus 11 parous females out of 14 in the one—to-two-year age class. This suggests that in general, females first give birth at approximately one year of age. Male Reproduction For all males one year old or older, mean testis length and mean estimated testis volume (calculated as (4r/3)x(length/2)X(width/2)Z) were determined for each four-week interval of the year, as presented in Figure 14(a). Testis size appears to be slightly smaller during much of the dry season (December through April) than during the wet season, but pairwise t-tests of the differences between means showed no significant differences between months (P > 0.2). Although sample sizes are again small, no seasonal variation in testis size is indi- cated. Figure 14(b) indicates means for each age class of testis length and testis volume, ignoring obviously immature specimens (less than 5 kg). Note that testis size is essentially constant after one year of age. The length of the seminal vesicle was highly correlated with length of the testis, and so provided no additional information. Apparently, male reproductive activity begins at approximately one year of age, which is slightly older than the age of the youngest pregnant females seen, and remains constant thereafter. 54 O O 0.. .m.a.m.j.j.a. 0-6 —4 2 I ' 2 n.d. “1W HLSNB'I SIlSBl NVBW o 14'- 12 - 0 1C)‘ E o LU .. E 8 :I .J C) n- > S’.’ .— (D LLI .— 2: <1 LU 55 Figure 14. Variation in estimated mean testis volume (open circles) and mean testis length (solid circles). sampling intervals. (b) By age class. indicated above the abscissas. (a) By four-week Sample sizes are 55 The age of reproductive maturity for male pacas was also examined histologically. Sections were prepared of testes from 27 specimens, collected in ten of the twelve months of the year. Eighteen of these, representing eight different months, contained.mature sperm in the seminiferous tubules. On the basis of cementum annulations, these 18 were all classed as one year old or older. None of the remaining 9, all less than one year old, contained sperm, even though the largest of these (7.9 kg) was heavier than the smallest (7.4 kg) of the pacas with sperm. The smallest testis with sperm measured 39 mm, versus 33 mm for the largest immature testis. (These were from the same 7.9 kg and 7.4 kg specimens.) All males one year old or older had testes measuring 38 mm or longer, and so can reasonably be judged repro- ductively mature. Annual Fecundity Since reproductive effort appears to be constant for all females one year old or older, irrespective of season, an estimate of overall fecundity can be derived from values for litter size, interbirth interval, and the sex ratio at birth. Since fecundity is expressed in terms of female offspring born per breeding female per year, the zero-to-one-year age class can for practical purposes be ignored since only one of 41 females in my sample of that age class had actually given birth. For all pregnancies observed in this study, a single fetus was present. Furthermore, all hunters in the Llanos whom I queried reported having seen only single fetuses. 56 The interval between births is a fUnction of the gestation period and the proportion of females pregnant. If 42/69 or 60.9 percent of the adult females are pregnant, it follows that the average adult female is pregnant 60.9 percent of the time. The gestation period thus represents 60.9 percent of the interval between births. Using 116 days for gestation period (see Appendix B), the best estimate of the interval between births is (116)X(69/42) = 191 days. Of the 87 pacas less than one year old, 44 were males and 43 were females, suggesting that equal numbers of male and female offspring are born into the population. Mean annual fecundity thus becomes (litter size)x(number of births per female per year)X(proportion of young that are female), or (1)x(365/191)x(0.5) = 0.95 female offspring per female per year. This is general for all age classes beginning with the yearling class when females first give birth. Crude confidence limits for the fecundity estimate can be obtained from the confidence limits of the binomial proportion of 42 pregnant: 27 non—pregnant adult females. The 95% confidence interval for 42/69 is 0.609 f 0.135, or approximately 47.7 to 74.4 percent. Combining these values with the high and low observed gestation per- iods of 122 and 109 days reportedlanander (1974), a reasonable range of mean adult fecundity values would be 0.799 to 1.250 female off- spring per adult female per year. Age Structure Age and sex distribution for the El Tuparro and El Porvenir paca samples are presented in Figure 15. The El Tuparro sample contains fewer young-adult females (two to five years old) and many more males 57 12 — 39 10 _ (5C5; __ 66:9 8 Odd; 6 __ 6:9999 666:99? 4 _- 666:9 6666;99 2 __ dddd;o 6656:999 6666666666666666666666;99999999999999? I I I I I I I I 20 15 10 5 0 5 10 15 BL TUPARRO 12 — ‘9 :9 66; 10 —- 66:9 8 __ 65:99 666:??? 6 _ Odddd;o 66:99? 4 _ 666299999 6666199???? 2 _ 6666:99999999 5555555299999999??? 6666666666666666666666; I 0 I I I I 20 15 10 5 9999999999999???999999999??? I I I I I 5 10 15 20 25 EL PORVENIR Figure 15. Age pyramids for the samples of pacas from the El Tuparro and El Porvenir study areas. Each row represents an age class, with age in years represented by the numbers in the vertical columns to the left. an individual male (5) Each symbol represents or female (9). Numbers of each sex in each row are indicated by the horizontal scales. 58 than females, while the El Porvenir sample has a good representation of younger females and relatively equal numbers of both sexes. These differences raise the question of whether or not the specimens that were collected adequately represent the age structure of the popula- tion; that is, whether the age distributionscHSFigure 15 are a random sample of the original standing age structure. This question was addressed by testing homogeneities among intuitively-defined a_ posteriori subsamples for (a) sex ratios and (b) within—sex regres- sions of numbers of age. Within the two study areas, subsamples large enough to be statis- tically tractable were created on the basis of several ad hgg_criteria. For the E1 Porvenir study area, the three watersheds in which the col- lecting sites were located formed three rather natural assemblages of sizeable subsamples: (a) the Muco River (sites 6 and 7), (b) Cafio Caviona (sites 1 through 4), and (c) the floodplains and northern tributaries of the Meta River (sites 5 and 8). Such convenient geo- graphical groupings of the E1 Tuparro sample tended to result in small subsamples, and so the study area was divided into heavily-hunted versus relatively unhunted sites. The former (sites 2, 5, 6 and 7) are those streams most suitable for hunting by canoe which are located on the lower portions of the Tome and Tuparro Rivers, and thus are highly accessible to hunters living nearby along the Orinoco. The latter (sites 1, 3, 4, 8 and 9), widely separated geographically, are the smaller or more remote streams identified by local hunters as hav- ing been infrequently visited. The difference in hunting pressure between the two groupings has undoubtedly been pronounced for at least several years. 59 Finally, the combined sample was divided into three groupings on the basis of how the specimens were obtained: (a) by canoe, (b) on foot, or (c) with tracking dogs. Grouping by collecting methods necessarily combined the two study areas since all but 17 of the E1 Tuparro pacas were obtained by canoe versus only 11 so obtained from El Porvenir. Age pyramids for the three categories of subsamples are presented in Figures 16 and 17. HOmogeneity of Sex Ratios Since the overall sex ratio (104 males:101 females) was not dif- ferent from 1:1 (P = 0.463), for each subsample, the exact probability was calculated of drawing the observed numbers of males and females from an infinite population with equal numbers of each sex. Proba- bilities were also calculated using juveniles only and using all ages greater than or equal to one year. A liberal critical value of P = 0.2 was chosen for distinguishing subsamples differing from the expected 1:1 ratio, since the purpose of the exercise was to avoid combining subsamples which might be different (probability of Type 11 error), as opposed to rigorously minimizing the probability of falsely concluding that there is a difference (Type I error). Ratios signi- ficantly different from 1:1 (P < 0.2) are underlined in the results presented in Table 2. The pronounced bias in favor of males in the overall El Tuparro sample was primarily due to juveniles of the unhunted subsample, although males were more numerous (but not significantly so) than females in the hunted subsample. Females, on the other hand, were 60 _ 12 __ I? :9 . —-10 —- 66; 6:9 __ __ d: 6: 8 55: :9 __ 6 __ (5:999 6:99 5519 : __ 4 __ 666:9 66; 5529? ; __ 2 __ Oddd;o 666:9 6:99 Oddddddddddddd:Ooooooooooooo 66666666:99 I I I I I I I 10 5 0 5 10 5 0 Hunted unhunted EL TUPARRO :9 3 ; 12 12 I? 6; d: 66; I? 19 dd: 7’ 8 "' 6:999 '7 8 " : 66:9 66; d: :99 " 6 " 66:9 "' 6 '— : :99999 __ 4 __ 66: __ 4 __ 6: 6:999 99:999 6: 66:999999 666:9999 69:9 6666666666666:9999999999999 66666666:9999999999 66:99999 I t I I I I I I I I I 10 5 0 5 10 5 0 S 10 0 5 Rio Muco Cafio Caviona Rio Meta EL PORVENIR Figure 16. Age pyramids for subsamples of pacas from El Tuparro and E1 Porvenir, as defined in the text. Symbols as defined for Figure 15, page 57. 61 10 - do; __ 9:99 8 666; 6 __ :9999 CANOE 666:999 4 __ 666:9 666:99 Z _ 66:9 6666;99 666666666666666666666:9999999999999999 I I I I I I 20 15 5 0 5 10 15 12 — ‘9 :9 10 — ‘ :9 6-9 8 __ . 6:999 FOOT 66666:9 6 _ 9:999 4 __ 999:999 9999:9999 2 __ dddd:999999 dddddz9999999 dddddddddddddddd:9999999999999999999 I I I I I 15 5 0 5 10 15 12 — 39 66: 10 _' dd: 66: 8 —- 66: DOG d: 6 — C5: __ :99 4 9:99 2 __ 99:99 dd:99999 6666555299999999 I I I 5 0 5 Figure 17. Age pyramids for paca subsamples grouped by collecting method. Symbols as defined for Figure 15, page 57. 62 TABLE 2. Sex Ratios1 of Subsamples and Samples of Pacas I. 'Within Study Areas E1 Tuparro E1 Porvenir Hunted Unhunted jMuco Caviona Meta <1 year 14:13 ** 8:2 13:13 7:10 2:5 E31 year 8:5 *18:11 12:23** 15:14 7:5 All ages 22:18 **26:13 25:36* 22:24 9:10 II. By Collecting Method Canoe Foot 99g l.year 21:15 24:31 15:12 All ages *42:31 40:50* 22:20 III. Sample Totals El Tuparro El Porvenir Combined 1.year **26:16 34:42 60:58 .All ages **48:31 56:70* 104:101 lExact binomial probabilities were calculated to test the null hypothesis that the proportion in the smaller class is 0.5. Significant ratios are underlined. * Probability < 0.2 ** Probability < 0.1 Asterisks appear on the side of that sex which is the larger proportion. 63 significantly favored among adult specimens from the Rio Muco water- shed. The analysis by collecting methods only partially explains different sex ratios in different categories within the two study areas. The bias in favor of males for the sample from El Tuparro was only partially reflected in the overall ratio of specimens col- lected by canoe, most of which were obtained from El Tuparro. Adult females were significantly more numerous in the combined El Porvenir sample, where most specimens were collected by hunting on foot. The preponderance of females from the Rio Muco was not evident in the ratio for specimens collected using tracking dogs. Despite the heterogeneity of sex ratios among subsamples, biases appear to cancel out when the two study areas are combined. Homogeneity of Age Structure It is apparent from all of the age pyramids presented that the zero-to-one-year age class is much larger than all older classes, but that numbers in each age class gradually decline with increasing age beyond one year. In order to test the homogeneity of age distri- butions among subsamples, least-squares regressions of animal numbers on age were performed within each subsample, with the zero-to-one-year class excluded. The model used was the log-linear form Zn (NUMBER + l) = a - b(AGE) (1) where a and b are the intercept and slope, reSpectively. The midpoint of each age class (age = 1.5, 2.5,...12.5) was used as an approxi- mation appropriate to species with aseasonal reproductive patterns (the "birth-flow" population of Caughley 1966). One was added to the number in each age class in order to include classes with zero 64 individuals (Zn(0) being undefined), up to the oldest age class in which one or more individuals was actually present in that subsample. Adding one had the disadvantage of rotating the slope of regression (increasing b), roughly in inverse proportion to the size of the sub- sample, but for samples large enough to result in a significant regression the change in b was minor. A perfect log-linear fit would result from the simplest of survivorship fUnctions -- a uniform proba- bility of any individual alive in any adult age class surviving to enter the next older age class -- for a stable population (Caughley 1977, and see Discussion). Smaller values of slope 6 indicate more individuals in the older age classes. Differences between subsample slopes for each sex within each of the three subsample categories were tested by pairwise t-tests of significance between regressions (Steele and Torrie 1960:173). Differences between combined El Tuparro versus El Porvenir sample slopes were also tested. .A critical value of P = 0.2 was chosen, again reflecting the concern for Type 11 error probabilities. Slopes, within-subsample correlation coefficients, and significant differences between slopes are displayed in Table 3. The most notable feature of Table 3 is that in only seven of the sixteen subsample regressions are numbers significantly correlated with age (within sexes). Thus tests were performed for possible differ- ences between pairs of regressions whiCh were not in themselves mean- ingful regressions. For E1 Tuparro, with only one significant sub- sample correlation (unhunted males), no differences were detected between hunted versus unhunted subsamples. Nevertheless, the results fer both sexes indicated that there were differences between sub- samples from the three watersheds of the El Porvenir study area, with 65 TABLE 3. Tests of Hemogeneity of Regression. Subsample: El Tuparro El Porvenir Methods Hunted Unhunted Muco Caviona Meta Canoe Foot Dog Males b1 .121 .124 .020 .216 .064 .165 .222 .051 r2 .368 ._584_ .145 ._8§0 .299 .606 .723 .177 n3 8 18 12 15 7 21 24 15 t-test4 ______ Females b .043 .090 .265 .183 .064 .123 .191 .041 r .069 .301 £13 _._5_6_9_ .218 .361 _.882 .433 n 5 11 23 14 5 15 31 12 t-test lSubsample regression coefficent (slope) estimated from Equation (1). All coefficents were negative (sign omitted here). 2Subsample correlation coefficient. underlined values are signifi- cant (P < 0.05). All correlations were negative (sign omitted here. SNUmber of adults (2: 1 year) in each subsample. 4Pairwise comparisons of subsample slopes within each category. Lines connect subsamples with significantly different slopes. 66 significantly older males from the Rio Muco and younger females from the Rio Meta. The analysis by hunting methods showed that signifi- cantly older males were obtained using dogs versus either by canoe or by foot. Using dogs also yielded significantly older females than did hunting on foot, while females obtained by canoe were not different from those obtained by the two other methods. Pairwise testing of subsample slopes showed that in addition to differences in sex ratio, significant differences in the age struc- tures of subsamples may be related to factors of collecting sites and hunting techniques. This poses the problem of selecting a combination of subsamples that could with confidence be taken as a random sample of the populations. To some extent, the use of different methods in different areas resulted in biases in opposite directions. Indeed, t-tests of slopes for combined samples from El Tuparro versus combined samples from El Porvenir indicated no difference, whether for males or for females. Adl four within-study-area correlations were significant (P < 0.05). The differences demonstrated between subsample slopes may have meaning to the exploitation of the population, especially regarding age- and sex-specific differences in susceptibility to different hunt- ing techniques. However, inference is extremely limited due to the lack of significant correlations, resulting from the small sizes of many samples, between numbers and age. One is obliged to ignore for the time being the observed heterogeneity of a partitioned sample and accept with caution the regression parameters of the combined sample. In the linear domain, the coefficient of the age term of the log— linear model becomes a coefficient p of annual adult survivorship from 67 the relationship p = eb (see Discussion). For males, the estimate of p is 0.867 and the limits of the 95% confidence interval are 0.823 to 0.915. The estimate of p for females is 0.800 and the interval is 0.746 to 0.858. Tested in the logarithmic domain, the slopes for the two sexes are not different (P > 0.2). The age pyramid for the com: bined sample is presented in Figure 18. Alternative Age Structure Regressions The log—linear model of age structure used in the above analysis of homogeneity is useful because it describes the parsimonious situa- tion of constant mortality for a stationary population. To examine its validity for describing the data presented here, it was tested against two competing models, (a) linear, and (b) log-quadratic. The linear model, NUMBER = a - b(AGE) (2) was chosen as a null condition. For statistical purposes, it is identical to the log-linear model in that it has the same number of terms. However, its biological meaning is minimal since a perfect fit would.imply a complex fUnction of increasing mortality with age, again assuming a stationary population. If the log-linear model does not explain statistically the variation in the observed data better than does the linear model, the parsimonious explanation of constant mor- tality must be rejected. The log-quadratic model, Zn(NUMBER) .= a - bl(AGE) i bZ(AGE)2, (3) was chosen because it uses a higher-order term which can indicate (assuming that the population is stationary) whether mortality is an 68 12'— 66- 10 6666 8 __ 6666: 666666: 6 __ 666666 66666 4 __ 666666 66666666 2 __ 66666666 66666666666 66666— — - 66666666666666666 I I I I 40 15 10 5 Figure 18. combined. :99 19 :9 999 999 :99999 :999999 :999999 299999999 1999999999 :99999999999999 :99999999999999999-"9999 I I I I 5 10 15 40 Age pyramid for the pacas from El Tuparro and El Porvenir Symbols as defined for Figure 15, page 57. 69 increasing or decreasing fUnction of age. Equation (3) is a parabola intfimelogarithmic domain, with its curvature relative to the straight line of the log-linear model being determined by the sign of the (AGE)2 term (Caughly 1977 and see Discussion). For a stationary pOpu- lation, the sign of the coefficient b2 implies either an increasing (62 negative) or a decreasing (62 positive) probability of death with increasing age. Since all sample age classes except for ten-to-eleven-year-old females contained specimens, the natural logarithm of the actual num- ber of pacas present in each class was used instead of adding one as before, and the missing female age class was ignored. Least-squares regressions of each of the three models and F-tests of the significance of regression were performed within sexes. Correlation coefficients were tested by approximate t-transformations. All equations obtained by least—squares fit were used to generate expected age distributions, which were then compared to the observed distributions of Chi-squared tests of goodness of fit. The significance of adding the quadratic age term was also tested by computing F for the reduction in sum of squares due to the quadratic term alone. Results are presented in Table 4. Although all values of F are significant, the log-linear model did not provide an increase in F compared to the linear model, whether for males or for females. Examination of the fitted equations against the original data revealed that this was due to differences in the distributions of errors in the linear versus the logarithmic domains. Observed numbers in each age class tended to lie above the fit of the linear model for younger and older age classes, and below 70 TABLE 4. Tests of Three.Models of Regression for the Combined Sample of Paca Age Structure. Males Females -.9409* —.9l82* 69.514 * 43.592 * (1,9) (1.10) 2.819 17.032 Log Linear -.8522* -.8960* 23.878 * 40.696 * (1,9) (1,10) 1.509 * 3.086 Log—Quadratic .9170* .9037* 21.131 * 20.044 * (2,8) (2,9) 46.386 * 3.431 (1.8) (1.9) .803 * 1.498 * Fa Test of significance of overall regression. F Test of significance of adding x2 term. X2 Chi-square sum of goodnes of fit, with 9 to 11 degrees of freedom in each case. * P < .05 71 the line for classes of intermediate age, whereas deviations were more uniformly distributed above and below the exponential curve of the log-linear fit. The Chi-squared test of fit was thus a better model fer comparing these models because it could be performed in the linear domain in every case. The log-linear model, both for males and for females, resulted in major reductions in Chi-squared relative to the linear model. For females, the major source of Chi-squared error was attri— butable to the lZ-to—l3-year-old class. If the two individuals in this class had been deleted from the analysis, the Chi-squared sums for females would have been only 5.102 and 1.862 for the linear and log—linear models, respectively. The latter value would be statisti- cally significant (P < 0.05). Thus when statistical variation was tested in the linear domain, the log-linear model resulted in a signi- ficantly better fit for both males and females than did the linear model. For males, the best fit of the log-linear model was Zn(NUMBER) = 2.555 - 0.142(AGE). (4) The addition of a quadratic term resulted in a significant improvement in F over the log-linear model, as well as a major reduction in Chi- squared. The value of F for the addition of the quadratic term in age was highly significant. The predictive equation was Zn(NUMBER) 2.044 + 0.147(AGE) - 0.025(AGE)2. (5) The negative sign of the coefficient of the quadratic term in age implies that for the males of this population, mortality was an increasing fUnction of age. The positive sign of the linear term, however, means that the axis of the parabola falls to the right of the 72 the origin, indicating that the maximum number of individuals is pre- dicted to occur in some age class greater than zero. Solving for the zero value of the second derivative of Equation (5) shows that in this population, maximum numbers would occur at AGE = 2.94 years. under the restrictive assumption of a stationary population, this would imply the impossible situation of survivorship values greater than unity for the age classes of 1.5 and 2.5 years. This is clearly unrealistic, since survivorship values must always be less than unity. Figure 19 shows the transformations of the fitted curves of the log- linear and log-quadratic models of age structure for adult males, together with a histogram of the observed age structure. .Although the log-quadratic model has greater precision in describing the male age distribution statistically, its application does not provide a meaning- ful interpretation of survivorship of a decreasing function of age if the population is assumed to be stationary. The implication of an aberrant survivorship pattern must, however, be taken into account. For females, the addition of a quadratic term in age to the log- linear model produced no significant increase in F (taking into account the loss of one degree of freedom in the numerator). The value of F for the quadratic term alone was not significant (P > 0.1). There was, however, a reduction in Chi-squared. The least-squares fit of the log-quadratic model for females is Zn(NUMBER) = 3.193 - 0.313(AGE) - 0.008(AGE)2 (6) versus the log-linear fit of Zn(NUMBBR) = 2.852 — 0.223(AGE). (7) Linear transformations of Equations (6) and (7) are superimposed on a histogram of the observed female age structure in Figure 20. 73 44 42- "...” ............... Log- Linear 12‘..00000....... ooooooooooo Log-Quadratic 0 9 NUMBER . o C . ... ‘- 2 4 6 8 10 12 A G E years Figure 19. Curves of the log-linear and log-quadratic models for males, superimposed on a. histogram for the observed age distribution to which they were fitted. 74 44 ~ 42- 14 - = - '1 12 ~ "g 1C)‘ ‘2; 0: _ . “a ............... Log_Linear LU 8 _ '3‘.“ 00000.00... Log-Quadratic m _ '. I»: z 0.. 0. 3 5 - ‘ L .0 .15 I: t ‘ 4 ‘0 .0 ... O“. 4 _ 4... 0...... '7 .30.. ”A > ..... 2 - u. .. ‘ ............ - 000,}. "0.. ooooo 4:- ...... ...qhoooo. 1h.” 2 4 6 8 A G E 12 )IGPClr'S Figure 20. Curves of the log-linear and log-quadratic models for females, superimposed on a histogram of the observed age distribution to which they were fit. 75 Although in the linear domain the quadratic term results in a better fit to the original data, its effect is seen to be modest due to the relatively small least-squares estimate (0.008) of its coefficient. for the 12.5-year age class, the difference between the values pre- dicted by the two models is only 1.2 individuals. I conclude that the analysis does suggest the possibility of slightly decreasing mortality with increasing age for females, but not to a degree sufficient to warrant adopting the log-quadratic rather than the log-linear model. Disregarding problems of sample size and collecting bias, the thrust of the foregoing analysis is that all individuals older than one year have relatively the same likelihood of surviving into the following age class. In other words, all adult pacas experience approximately the same annual probability of death. Again, the entire analysis depends on the assumption, to be dealt with in the following discussion, of a stationary population. Abundance Estimates of the number of adult (one year old or older) pacas per ka of forest and of all pacas per km2 of forest are presented in Table 5. The estimates are based on the numbers of specimens col- lected at those three sites that were exhaustively hunted. Numbers of pacas from each site include only specimens from discrete galleries that were thoroughly and repeatedly hunted over a relatively short period of time. The periods of hunting effort varied, but relative hunting effort was approximately comparable at these three sites. It must be assumed that none of the individuals removed by hunting effort was replaced through immigration during each of the collecting periods. 76 .96690 90 690 966% 6:0 69m wpazfi9om Hm om mm moH.o o 09 am 699090 opcmsm -- 99co>9om Hm mm ow moH.o w m cm mcdm ommo -- 0996639 Hm muasw< proe m993p< 96909 mzm «69MNMMW9om n woa9kw mhmwgemm meHm NEM\9onE:z mmomm mo 969552 m . . .ooqmp::n< mo m696599mm .m mqm6 96009 90 6699599 90 699505 696 669669 .6996996>6 999696o6m 696 96966 69969 699 90993 999956 699905 696 6969969m909 99 u 9 ”0996939 99 u 9N .6056066909959 n 9 ”6666m u m ”99:99 69093 n 9 ”99609969 u 099 80 ................. 9.9 9 6966969 czommm: ...... 9.9 9 969639 .96 6996099 m96m9o 996969 6562 6562 09999969om m6996996>< 699902 669< 69969 505500 96609 9992<9 .66069 99 6669999: 65699 9009 .o mqm<9 %0 9999905 096 mmfimwfl .9999996996>6 96009 90 6699599 .6996996>6 999696966 696 96966 69969 699 90993 699996 699905 696 69699696< .9996>909 99 n 9 “0996939 99 n 9% 81 .6096066909999 0 9 ”66666 n m ”99:99 69092 n 9 ”99609969 n 099 ........ 9 9 .96 65099 ....... 9 9 09696965 .96 69099 m69630 ---- 9 m 0905 66 0000 .96 69699639069 6660669999069 ........ 9.9 9 0909965 09959668 6996699 m<9m9999bw ---- 9 9 960669969 .96 5659696099000 mmo ammo oozes Aw w.N H mam: mo mmueoo: came an o.m m excess: mm mammsuomo oewo fie H.o m wqwmmemwflz_ a mMES: oemu Am A.m N maze» spa: mu «Magoo: oemo he onEom z o.¢ m mum: m «weapon ammo mm H.o m :30:x:: m oppmmdh owm oozed mm o.m m ofimz_ N mommfl<.opoosm mu may cpwcoq moocwhucm ucegsooo may Emoupm opfim weapooaaou mo Honesz seam oucwpmfim .mzoposm mowm no>oam mo moflpmwhouompmnu .n m4m) (13) x=o A=1. Figure 22 presents curves for the estimated mean and for the 195% confidence limits of m as a function of p and la when A = 1. Within these limits, the shaded area is bounded to the left and right by the :95% confidence limits of p = 0.800 (solid triangle). Combining these 108 limits predicts the value of la with 95% confidence given all of the underlying assumptions: 1) Sex ratio is 1:1 at birth; 2) Litter size is always one; 3) Gestation period is between 109 and 122 days; 4) o 1.5 years; 5) m 12.5 years; 6) Annual survivorship of adults is p for all x|1.5 I ADULT SURVIVORSHIP "p” Figure 22. Prediction and confidence limits for juvenile survivorship. Symbols explained in the text. 110 estimating a and w (see below). Assumptions 6), 7), and 8) require considerable clarification. The variance of p was estimated from the log-linear regression of numbers on age must be applied with caution. Age structure is a multi- plicative function of p operating over the lifespans of all indi— viduals, so that s = A’X n p (14) Clearly p determined by regression provides an estimate of the mean of p for the interval of time when age structure was observed (in this case, the interval was four years), but not of p as a composite of the effects of px for the lifespans of the individuals in the sample. In order to estimate p properly, one would need direct estimates of sur- vivorship. Also, in obtaining subsamples from various sites, one introduces differences in habitat and history of hunting pressure among the various sites. This underlying variation was lost because regression was performed on the combined sample. I must assume that these local effects balance out when considering the entire population. Changes in demographic characteristics have been found over small dis— tances for contiguous pOpulations of long-lived species even when con- sidered over much smaller distances than those between.my collecting sites (Dapson EE.§l: 1980). Annual adult survivorship p as estimated from the coefficient b of the log-linear model was higher for males than for females. If a stationary condition is assumed, this implied that adult males exper- ience a lower annual probability of death (1 — p) than do females. Non-rejection of the log-quadratic model of age structure implies that 111 for adult males, px decreases with increasing age. Inference regard- ing the Zx schedule is limited since the log-quadratic fit based on Sx predicts the impossible situation of ZX > 1 for some real values of X. The estimated fecundity value of 0.95 must be used for all age classes 1.5<;x<;12.5 in order for Equation (12) to apply. For popu- lations of continuously-breeding species, this is often assumed to be true in the computation of instantaeous birth rates (see Lotka 1956: 115), although approximations using Equation (8) have also been used to show that the age-specific intensity of natural selection varies with age as an increasing function of age-specific differences in fecundity (Emlen 1970). .Although I detected no differences between adult age classes in the proportion of females pregnant, there was a tendency for the older age classes to include more individuals that were both pregnant and parous (see Figure 13, page 52). This trend was not tested since it involved only twelve individuals. Analyses of real populations by Schaffer and Tamarin (1973), Ricklefs (1977), and Siler (1979) suggest that the parsimonious assumptions of constant adult survivorship and constant fecundity may be applicable to a variety of long-lived species. The relationship of Equation (13) has been applied to pelagic birds by Goodman (1974), who feund the basic assumptions to be reasonable. It is often acknow— ledged (Deevey 1947) that for many species of birds and mammals, adult survivorship is nearly independent of age. The assumption of a stationary population is onerous but not unreasonable in the case of a long-lived species in a relatively stable tropical environment. The popular concept of r and K selection 112 GHacArthur and Wilson 1967) has its roots in Dobzhansky's (1950) observation that populations in the trOpics are much less subject to variation in environmental factors that can cause fluctuations in num- bers than are temperate pOpulations. K-selection refers to a life- history strategy geared toward maximizing efficiency of utilization of resources as Opposed to maximizing the production of progeny (Pianka 1970, 1972). The paca has several characteristics of K-selected species, such as low reproductive rate and maternal care of the young. The concept of rate of increase has been a source of confusion in the literature. Reviews of the various definitions and restricted meanings are found in Davis (1973) and Slade and Balph (1974). One must bear in mind that those populations which have attracted the most attention from pOpulation biologists are of r-selected species such as pests, pOpulations being manipulated to maximize yield such as fish stocks, or populations undergoing relatively unrestricted growth, such as man. A strong bias in pOpulation thinking has resulted from over- emphasis of the logistic model (Pearl and Reed 1920) (upon which the concept of r and K selection is based), which is demonstrably unreal- istic (see Pielou 1977:35). Some authors have considered.many natural populations to be at or near a stationary condition (e.g., Goodman 1974), and in considering the models of Goodman and of Ricklefs (1977), Green (19802295) posits, without elaboration, that "it is useful to bear in.mind that many animal populations are quite stable." In reality, actual evaluation of A is not possible without large bodies of data which are difficult to collect, yet the estimate obtained may not be very accurate. Tait and Bunnell (1980) have shown by computer simulation that even large 113 samples drawn at random from an infinite population with idealized age structure may yield widely-disparate results in replicated trials. The cumulative effects of sampling errors are thus a major problem in analyzing limited sets of data by means of traditional life tables or projection matrices. The greatest stumbling block to the present study is the pos- sibility of A being consistently less than one, resulting, for example, from the population having been heavily overexploited for some years. Although a stable age distribution may be apparent from a smoothly— declining sample age distribution, SX could still depart significantly from ZX consistently for all x. As Equation (12) shows, a geometric series results from constant annual survivorship as (p/A), whether or not A is actually 1.0asassumed. While the assumption of A = 1 cannot be directly demonstrated from the data, it can readily be invalidated. For a rapidly declining population, some age classes may be larger than younger ones, a condi- tion for which A = 1 would imply the impossible situation of an increase in numbers in a birth cohort, when only mortality is operat- ing. Fluctuations in A through fluctuations in any of the parameters will produce an irregular age distribution departing significantly from the stable distribution predicted by constant adult survivorship. As has been shown previously (see Figure 20, page 74), the observed age distribution is quite regular. The effect of a particular deviation in A away from the station- ary condition can be evaluated by solving Equation (12) for the rela- tionship between juvenile survivorship and adult survivorship using a variety of values for A while a remains constant at the estimated 114 value. Figure 23 presents curves for a = 0.95 as a function of 2a and p given several values for A. The shaded area is the same set of :95% confidence limits presented earlier. .A vertical distance between the line of A = l and a line for any other value of A thus represents the error in estimating Za due to (p — p/A) assuming that p is accurate, and a horizontal distance is the error in estimating p for the value of A f 1 given that all of the other parameters are accurate. Thus confidence in the model's prediction of the value of Za has a preci- sion equivalent to assuming A = 1 when the actual value of A may be as low as 0.93, even given that the remaining assumptions are true. If indeed A = 0.93, the pOpulation would be declining by 50% every ten years. Without an independent estimate of A or of la, interpretation of the model in terms of a and p is seriously restricted. That the pOpulation is approximately stationary can be shown from the data by generating an artificial estimate of juvenile survivorship. Fifty-eight females of reproductive age were collected. If each is expected to produce 0.95 female offspring in the course of one year, an artificial birth cohort expected from one interval of x would con- sist of 55.1 females. From the fit of the log-linear model for females (see Equation (7), page 72), 12.39 of these newborn would be expected to survive to age a = 1.5 years, giving 20 = (12.39/55.1) = 0.225, as the model predicts when A is assumed to be 1.0. No assump- tion was made concerning A in generating the artificial birth cohort: Estimating la in this way shows that the observed age structure expresses a net death function equal in.magnitude to the observed net birth potential, the situation which defines a stable population. JUVENILE SURVIVORSHIP ”lac" Figure 23. 115 ADULT SURVIVORSHIP ”p” Curves for a = 0.95 for different values of A, assuming a = 1.5 and w = 12.5 in each case. The shaded area is the :95% confidence areas for p and a when A = 1.0. 116 There are two critical flaws in this line of reasoning. First, it is circular in that 2a is not estimated independently of'p and.m. Second, it evaluates observed data from a time-specific or "vertical" life table as if it were obtained from a time-composite or "hori- zontal" life table. The two computational ferms are equivalent only if (1) the population has attained a stable age distribution, and (2) A is constant for the time interval represented by the composite case (see Caughley 1977:119-20), which is 12.5 years for these data. The regular form of the observed age distribution suggests that these assumptions have not been violated. Thus estimated juvenile survivor- ship can be obtained indirectly whether or not the population is assumed to be stationary, although difficult assumptions are involved in either case. When A is not assumed to be 1.0, the estimate obtained is virtually 1.0. This is equivalent to obtaining a net replacement rate (R0) of 1.0. Nfichod and Anderson (1980) have shown that A and the schedule of ZX can be jointly calculated if la, the age distribution, and the fecundity schedule are known. Letting "x be the number of individuals in age class x, values for each ZX are calculated as ZX = Za (nx/na). From the artificial birth cohort, let n0 = 55 and from the observed age distribution (see Figure 20, page74 )letZfi = 14/55 = 0.225 and ZX = 0.255 (nX/l4). I used the above to solve Equation (8) for A by iteration, letting mX = 0.95 for all x 1.5, and found A = 1.001. Solutions using the limits of la predicted in Figure 22 gave a range for A from 0.88 to 1.06. Thus assuming only mx = a suggests that A is indeed in the neighborhood of 1.0. Even.when mx is calculated for each x according to the proportion pregnant in each age class, the 117 range for A is from 0.86 to 1.04. Wlth an independent estimate of la, the value of A could be calculated.more precisely. To summarize, the data cannot be used to evaluate A directly, but they do indicate that A has been relatively constant for some years at approximately 1.0. The validity of the original assumption is sup- ported in that: l) The age distribution is regular (see Figure 20, page 74); 2) the range of A (as p/A) in Equation (12)) that could have resulted in the 95% confidence interval oflp as estimated from age structure (see Figure 23, page 115) is not large (0.93<§A €1.11); 3) juvenile survivorship estimated from the net fecundity of the collected females and from the size of the a = 1.5-year age class indicates that A = 1 independently of the original assumption involved in estimating adult survivorship from age structure; and 4) if A is assumed to be constant (i.e., vertical and horizontal life tables yield equivalent solutions for A), and then is estimated jointly with age structure using the range of la as predicted by the model of Equation (13), the range of esti— mates for A is again small (0.86 100 um) layer of cementum which begins to fill in the apertures Of the root canals, but generally does not extend along the outside of the upper part of the root to the region Of the alveolus. The extent Of these early cementum layers is shown in the section of M1 from a one-to-two-year-old paca presented in Figure.A3(a). subse- quent annuli extend along the root up to the vicinity of the alveolus, where they are thin but usually distinct. In each Older specimen, the root has been extended by proliferation of'much greater cementum than is found on the lateral margins of the base of the crown, as Shown in Figure.A3(b). The area in which annular layers are seen can therefore be divided into root-crotch and alveolar regions. The root canal becomes a narrow extension of what was a large (> 2 mm) aperture in the newly—erupted tooth. Any annuli seen along these root canals are highly irregular. Thus the earliest dark annulus to be deposited appears only on the lower root, being especially conspicuous in the crotch between the two roots, and is not represented in the region of the alveolus. The upper alveolar region seen in Figure A1 (a) and (b) therefore has Figure.A3. 138 Distribution of cementum tissue, and crown patterns of ‘molariform teeth. (a) Section OfIM1 of one-year-Old female paca. (b) Section ofIMl of nine-year-Old male. Both sections 12 um, X10. (c) Occlusal view of left upper molariform teeth of female paca classed as slightly less than one year Old. Note that only the anterior half of‘M is occluded. (d) Occlusal view of male paca classed as slightly Older than one year. 1M3 is fully occludedi P: -germanent premolar, p4= deciduous pre- molar, MI1 fllM 1M first, second, and third molars, respectively. 139 140 no "first year” annulus; the dark stripe nearest the junction between dentine and cementum.is labeled (2). The view of the crotch region in Figure A1(c) shows how the first dark-staining annulus is limited to the region between and around the primitive root canals and in the crotch between them. Indeed, for all specimens for which sufficient cementum was seen in sections of M1, the number of annuli in the crotch region was generally one more than the number seen in the region of the alveolus. For eight Specimens, there were two more dark annuli in the crotch, and for three pacas, there were three more, than in the alveolar region. This suggests that late in life, very little cementum is produced near the alveolus. In order to estimate the age at which the first dark annulus appears in the cementum, and in fact to distinguish first-year (less than one-year-Old)fitmiyearling (one- to two-year-Old) individuals, a generalization had to be made on the basis of very little infor- mation concerning convenient developmental criteria. .A female paca raised by me in captivity from.the age of approximately one week was sacrificed at the age Of seven and one-half months, shortly after termination Of the work in Colombia. At that age, the third molar was not present. Observations on the development of the same animal had revealed that the second molar had begun to erupt at approximately four and one-half months of age. Since wild pacas weighing less than two kg (i.e., less than one month Of age -- see Appendix B) have only the deciduous premolar (p4) and an emerginglM1 present, it would appear that several months are required to complete the eruption and full occlusion of a molar. I assume that it would have taken an addi- tional four to five months for the captive—reared individual's M3 tO 141 form, erupt, and wear down into a fully-occluded condition. For most specimens, p4 is replaced either at or shortly after the time of attainment of a fully occluded M3. On the basis of this and.no other information, I presume that fully-occludedM3 and/or the presence Of P4 indicate that a paca is approximately one year Old or slightly older. Figure A3 (c) and (d) shows occlusal views Of the upper tooth- rows of specimens judged by these criteria to be slightly less than (c) and slightly more than (d) one year of age. Of 48 specimens classified as being less than one year Old by the above criteria, 39 had no cementum tissues in sections of M1. Only relatively thin (< 100 pm) caps of cementum were found in the remain- ing nine first-year pacas, always located around the openings of the broad primitive root canals and in the crotch between them. Twenty- five pacas classed as older than one year had teeth with rather broad (> 100 um) areas Of light-staining cementum in the crotch but no cementum around the alveolus. For eighteen of these,21dense-staining dry-season annulus was found at a variable position within the crotch cementum.. Five animals, all collected between July and October, had no annuli, and the remaining two, collected in December and March, had dark-staining cementum at the margins of the crotch area. As determined in the text, pacas give birth at all seasons Of the year. If, as Figure A2 suggests, proliferation Of cementum is largely restricted to the wet season, one can expect that the age at which time dark-staining dry-season cementum begins to appear depends as much on the season of birth as on the age of the animal. For pacas born during the wet season, cementum begins to develop before the age of one year, and the first dark-staining annulus will be deposited 142 during the dry season in the middle of the second year of life. For pacas born in the dry season, the first wet season may pass without much deposition of cementum taking place, but the first dry-season annulus will be depositedzn:just about one year of age. I conclude that the appearance of the first dark-staining annulus occurs sometime during the second year of life and each subsequent dark-staining annulus corresponds to an additional year of life. The final criteria used to estimate the ages of pacas are therefore as follows: 1) Pacas with deciduous premolars (p4) present and with M3 not fully occluded are less than one year Old; 2) Pacas with permanent premolars (P4) and/or fully occluded M3, and having no more than one dense-staining (dry season) annulus in the cementum of the crotch of the roots of M1, are between one and two years of age; and 3) For all Older specimens, the age in years is taken to be the number Of dense-staining (dry-season) annuli seen in the crotch of M1, or one less than the number of annuli in the region of the alveolus. In cases of a discrepancy between the numbers of annuli in the two regions, the value for the crotch region is used. Thus, a specimen with one (in crotch) and two (in alveolus) dry-season annuli, respectively, is placed in the 3- to 4-year age class. Of 205 pacas examined (eight others were excluded due to missing skull or damaged dentition), 106 were classed on the basis of tooth eruption (Often taking into account the amount of light—staining cementum in M1) and 99 on the basis of the number of cementum 143 annulations. In only four cases was the age estimate based on annuli in the alveolar region alone. This system is Of course subject to error, especially since no known-aged adult animals were available for study. The potential error in discriminating first-year from yearling pacas is at most several months (probably two or three). Furthermore, greater errors are likely in simply determining age on the basis Of cementum from a single tooth, Since counts of annuli may differ for teeth from Oppo- site sides Of the same individual, at least in the case of Odocoileus virginianus (Rice 1980). There is a tendency also for the boundaries of Older age classes to overlap as a result of the aseasonal breeding pattern of the paca. For example, an older paca that was born in the wet season would have dark-staining annuli representing the midpoints of chronological years Of life. If collected in the wet season, the broad outermost light-staining region might represent nine months Of life since the appearance Of the last dark-staining annulus. For such a specimen, age would be underestimated by approximately three months, and the age class to which it is assigned would be in error by one year. Such a shift in estimated age, if consistent for all Older age classes, need not have any effect on the survivorship values obtained from the Slope Of regression of the resulting age distribution. APPENDIX B METHODS USED TO ESTIMATE TIME OF BIRTH Growth Of Young Animals Age in weeks was estimated from body weight for relatively young Specimens as a means of estimating the week of their birth. Growth data were Obtained from four specimens raised in captivity; two Of unstated sex born at the U. S. National ZOO in 1972 and 1973 (J. Eisenberg pers. comm. 1977), and one male and one female purchased in August of 1979 from a fisherman living near El Porvenir, obtained at approximately three weeks and one week of age, respectively. Gross body weight was taken from the latter two individuals at variable (generally weekly) intervals. In fitting a growth curve, the ages when first Obtained Of the El Porvenir captives were estimated using the data provided by Eisenberg. All data were combined and fitted by the least-squares methods to a growth equation of the form AGEweeks = a - b(Zn(a - WEIGHTkg)) (Bl) with a and b being the intercept and slope, reSpectively, and a being an appropriate adult weight toward which the curve converges asymtotically. Trial values of 7, 7.5, and 8 kg were used for a, the best fit being obtained using a = 7.5 kg. The resulting equation is AGEweeks = 36.74 - (19.11)(Zn(7.5 - WEIGHIng. (132) 144 145 The standard error Of the estimate is 1.76 weeks. The fitted curve is plotted together with the original data points and 90% confidence limits in Figure Bl. As the figure shows, the dataeuxaconcordant at least up to a weight of approximately five kg. The curve was used to estimate the age in weeks Of specimens less than this weight, and thus to estimate the weeks in which they were born. Due to the vagaries of regression, the confidence limits suggest less precision in the first few weeks of life than is likely to Obtain. Birth weight is surely constrained by the reproductive structure of the mother to a range of 500 to 800 g and one cannot conceive cm? a newborn quadrupling in weight in the first two weeks of life. One suspects that more data, especially if replicated to permit evaluation Of resi- dual variation at each week of age, would reduce the error limits con- siderably. The limiting value of 5 kg used in estimating age, which is somewhat larger than the maximum weight of the E1 Porvenir male before he escaped, is thus somewhat arbitrary. Since the animals used for the data provided by Eisenberg were born to pacas obtained from Panama, there is some need for caution in combining these data with that obtained from the El Porvenir animals. Also, some circularity was introduced in using Eisenberg's data to estimate the ages of the E1 Porvenir animals when they were first obtained and then combining the two sets of data to derive the predic- tive curve. In any case, caution is in order when applying data obtained from captive animals to specimens collected in the field. It should be noted, however, that such data are difficult to Obtain. Indeed, three other young pacas were captured at E1 TUparrO in the 146 35 — . USNZ n=2 5. - Porvenir d4 30 — . Porvenir $3 Age = 36.74- 19.11 x ln(7.5 -Wt) 25 _ ........... I 900/0 CL 0... m d .x r 3; w ..0 .0 Q) 20 -" I; 0.. ‘ a A 3 e“ o. " A o' O O... ‘ .’ “J 15 — (D " .’ I < .0... . 1O — . A o... .0. o. I .‘o. ’0‘ ........ t 3". I ........ %-°' 5 — . .0 . A .‘fl 5‘0 5.. °. ..'.. WEIGHT k9 Figure Bl. Growth curve for four pacas raised in captivity. 147 course 1.0, a greater increase in A results from increasing juvenile survivorship, and when m < 1.0, as was the case for pacas in the Llanos, increasing adult survivorship has the greater effect. As Goodman has indicated, the apparent insensitivity of A to changes in o is an artifact Of the assumption of A = 1.0. The partial (8’1) is lost when transforming to Obtain a in ordi- derivative of pA nary form, since when (A) is assumed to be 1.0, Zn(l) = 0. For even minor deviations from.a stationary cOndition, the age of onset of reproduction can have a marked effect on population growth. In the Simplified model, however, sensitivity to changes in o is still minor. The effect is one of shifting the net maternity function (mex) along the age axis, while replacement rate (the area under the curve, ZZXmX) remains constant. Changes in rate of increase Obtain from changing the area under the curve. Thus the model is not sensitive to changes in a alone, particularly in Goodman's analysis where w = m, but rather is sensitive to changes in the net length of reproductive life, or (w - a). Figure C2 plots curves of the estimated constant fecundity value of 0.95 as a function of combinationscrfla and p which satisfy the conditions defined by Equation (14). Each curve represents solutions for a different value Of w, assuming that a = 1.5 in each case. Clearly, the model is sensitive to the length of reproductive life 152 .6 - = a .5 --~.. m=O.25 E 3.... m = 0.5 $ ,4 -+ o: a '>' 3 _ ‘: CC ' m =10 3 m LU — \ X... _l o 2 9... iii ”1::213 'nu. a D r \ o..... -w .I “ m :40 \\ 9.... m = 8.0 '9‘ 7 l l l * ADULT SURVIVORSHIP ”p” Figure Cl. Sensitivity of A to changes in la and p when 11).: co. The sensitivities are equal when r72 = 1.0 (broken line). 153 LONGEVITY ”uJN-yeors = 3.5 O Q . Q —-q .0 2 O. O 0 — 0 o .0 .0 =75 =95 =12£5 =19.5 JUVENILE SURVIVOR SHIP "la" (A) 1 I A o o 0 o o o o o o o .0 A l 1 1 I 1- .5 .6 .7 .8 .9 1.0 ADULT SURVIVORSHIP " p" Figure C2. Curves Showing the relationship between juvenile and adult survivorship using different values for maxiJme longevity in solving Equation (14), assuming a = 1.5, 172 = 0.95, and A = l. 154 expressed as variation in w, especially as p approaches 1.0, but the effect of decreasing w from infinity to the observed value Of 12.5 is equivalent to a change of less than three percent in the value of la when p is in the neighborhood Of the estimated value of 0.8. Further- more, a is subsumed in 2 so that survivorship to the first repro- a, ductive age class enters the model without regard for the length of time represented by a. Decreasing o (hastening the onset of repro- ductive life) would have the effect of increasing the difference |A - 1|, and increasing a would decrease IA - 1|. When p is 0.8 and A is l i 0.1, the relationship between la and p is affected by less than four percent when w = 13.5 and a is as large as 3.5. Thus, although it is important to estimate the value w, the model is robust even for improbably large deviations in w and/or a. To incorporate w << w into the analysis Of Equation (12), Equation (Cl) is modified by subtracting summations to account for the range of w + l to w, so that a 00 -1 1 = azdp-a ( z (p/A)X - 2 (p/Alx) - x=0 x=O ( 2 (p/AJX - z Qp/A)X) . (C5) X-O x 0 Using the definitions of Equation (C2) and rearranging superscripts, = _ _a /A a _ /A w—l 1 mZap GP 1 _ 6g7A3) , (C6a) OT 1 - (p/A) = mar“ — mzap‘0‘(p/A)w+l. (C6b) 155 Moltiplying by A“ gives 1“ - pAa—l - mZa(1 - (p/A)“’“+l) = 0. (C6c) The relevant partial derivatives are Zo(1 _ p(w-a+l)) 3A I _ '17 a T _ ’ (C73) am A 1 l - p(o-l) - fiZa(w-a+l)p(w a+l) 3_A l = 1 ' Won'wwwwa) (C7b) 5? A = 1 1 - p(o-l) - aza(w-a+1)p(w'“+l) ’ - " +1 3.1 = mu ‘ Pm a )) (C7c) Ola X _ 1 1 - p(o-l) - fiZa(w-o+1)p(w_a+l) , 8A = .5; | A = 1 0 , (C7d) 8A = 55.| A = 1 0 . (C7e) Equation (C6c) is identical to Equation (C3c) except that the product film is now weighted by a factor of approximately (1 - pm) since a = 1.5 = 1.0 for the present purpose). Since p is necessarily a fraction, the weighting factor becomes vanishingly small as p becomes small or as w becomes large, giving partial derivatives for p and a equivalent to Equations (C4b and d). When p is large and w is small, as in the case Of the pacas, la and a have smaller effects on Equation (C6c) and so the sensitivity of A to small changes in them is increased over the infinite longevity model by a fonction of p, o, and w. The 156 partial derivatives relative to o and w are again intractable when A is assumed to be 1.0 When length of reproductive life is included in the model (i.e., w << 00), fecundity as a function of la and p is not a straight line and the critical fecundity at which A is equally sensitive to la and p is no longer a unique value. As p becomes large, the critical fecundity becomes lower, the relative sensitivities of A to la and p are defined in terms of both a and (w-o), and the slope of the curve for a given value of a does not define these sensitivities. Line (a) in Figure C3 locates the set Of combinations of la and.p for which A is equally sensitive to both, using the observed values of a = 1.5 and w = 12.5. Line (b) is the line for a = 0.95 as before, and line (c) II is where sensitivites to la and.p would be equal for w w. The open circle shows the position of the estimated values Of a 0.95 and p = 0.80, and the Shaded area defines the 195% confidence limits as before. Notice that taking length of reproductive life (w-a) into account reverses the sensitivities to la and p. If maximum longevity is 12.5 years, a greater increase in the number Of pacas will result from increasing juvenile survivorship than from increasing survivor- ship for all adults by the same amount. As before, even less is to be gained from increasing fecundity by the same amount. .As the shaded area indicates, the higher sensitivity to la holds true for virtually the entire range of confidence for the estimates of a and p. I\) 'w is in 1 1 1 1 JUVENILE SURVIVORSHIP "lac” I Figure C3. 157 - .................. :Efiifiémfi -a -C l I I I I .6 .7 .8 .9 ADULT SURVIVORSHIP "p” Demographic parameters for the population Of pacas (line (a) and open circle) in relation to values for equal sensitivity of A to la and p for the finite (w = 12.5, line (b)) and infinite (w = m, line (c)) longevity situa- tions. LIST OF REFERENCES LIST OF REFERENCES Aisbey, E. O. 1974. The grasscutter, Thryonomys swinderianus Temminck, in Ghana. Symp. 2001. Soc. Lond. 34:161—170. Advarez del Toro, M; 1977. Los mamiferos de Chiapas. Tuxtla Gutierrez, Universidad Aut6n. de Chiapas, 147 pp. Aquino, Yarihuaman Rolando, personal communication. 1979. (Ministerio de Agricultura y Alimentacion, Proyecto de Asentamiento Rural, Iquitos, Peru). Armitage, K. B., and J. F. Downhower. 1974. Demography of yellow- bellied marmot populations. Ecology 55:1233-1245. Barash, D. P. 1974. The evolution of marmot societies: A general theory. Science 185:415-420. Bates, M. 1964. The land and.wildlife of South America, by Marston Bates and the editors Of Life. New York, Time Inc., 200 pp. Beard, J. S. 1953. The savanna vegetation of northern tropical America. Ecol. Menogr. 23:149-215. Beddington, J. R., and D. B. Taylor. 1973. Optimal age-specific har- vesting Of a population. Biometrics 29:801-809. Blydenstein, J. 1967. Tropical savanna vegetation Of the Llanos of Colombia. Ecology 48:1-15. Bourliere,IL 1973. The comparative ecology of rain forest mammals in .Africa and trOpical America: Some introductory remarks. Pp. 279- 292 13: B. J. Meggers e£_al, (eds.), Tropical forest ecosystems Of.Africa and South America. ‘Washington, D. C., Smithsonian Institution Press, 438 pp. Brisson, M. J. 1762. Regnum animale in classes IX distributum, Leiden, Theodorum Haak, 296 pp. Brunnschweiler, D. 1972. The Llanos frontier of Colombia. Environ- ment and changing land use in Meta. Latin American Studies Ctr. MOnograph 7, Mich. St. Uhiv., 71 pp. 158 159 Cabrera, A. 1961. Catalogo de los mamiferos de America del Sur. Rev. Mus..Arg. Cienc. Nat. Cien. Zool. 4:1-625. Caughley, G. 1966. Mortality patterns in mammals. Ecology 47:906- 918. Caughley, G. 1977. Analysis of vertebrate populations. New York: John Wiley and Sons, 228 pp. Caughley, G., and L. C. Birch. 1971. Rate of increase. J. Wildl. IManag. 35:658-663. Charnov, E. L., and W. M. Schaffer. 1973. Life-history consequences Of natural selection: Cole's result revisited. Amer. Natur. 107: 791-793. Cole, L. C. 1954. The population consequences Of life history pheno- mena. Quart. Rev. Biol. 29:103-139. Collins, L. R., and J. F. Eisenberg. 1972. Notes on the behavior and breeding of pacaranas, Dinomys branickii, in captivity. Internat. ZOO Yb. 12:108-114. D'Alessandro,.A., R. L. Rausch, G. A, MOrales, S. Collett, and D. Angel. In press. Echinococcus infections in Colombian animals. Amer. J. Tropical Med. and Hygiene. Dalquest, W. 1953. IMammalS of the Mexican state of San Luis Potosi. Louisiana St. Univ. Studies, Biol. Ser. 1:1-224. Dapson, R. W., P. R, Ramsey, M. H. Smith, and D. F. Urbston. 1980. Demographic differences in contiguous populations of white-tailed deer. J. Wild. Manag. 43:889-898. Davis, D. E. 1973. Comments on 3. Bull. Ecol. Soc. Amer. 54(3):14- 15, 26. Deevey, E. S. 1947. Life tables for natural populations of animals. Quart. Rev. Biol. 22:283-314. Defler, T. R. 1979. On the ecology and behavior Of Cebus albifrons in eastern Colombia: I. Ecology. Primates 20:475-490. Dobzhansky, T. 1950. Evolution in the tropics. Amer. Scientist 38: 208-221. Dubost, G. 1965. Quelqes traits remarquables du comportement de Hyaemoschus aquaticus (Tragulidae, Ruminantia, Artiodactyla). Biologia Gabonica 1:282-287. Dubost, G: 1968. Les niches ecologiques des forets tropicales sud- americaines et africaines, sources de convergences remarquables entre rongeurs et artiodactyles. La Terre et la Vie 22:3-28. 160 Eisenberg, J. 1966. The social organization of mammals. Handbuch der Zoologie lO(7):l-92. Eisenberg, J. 1974. The fUnction and motivational basis of hystrico- morph vocalizations. Symp. 2001. Soc. Lond. 34:211-247. Eisenberg, John F., personal communication. 1977. (United States National 200, Front Royal, Virginia). Eisenberg, J., and G. MCKay. 1974. Comparison of ungulate adapta- tions in New WOrld and Old WOrld tropical forests with Special reference to Ceylon and the rain forests of Central America. Pp. 585-602 12: ‘V. Geist and R. walther (eds.), The behavior of ungulates and its relation to management, VOl. 2, IUCN Publ. 24. Eisenberg, J., M. O'Connell, and P. August. 1980. Density, produc- tivity, and distribution Of mammals in two venezuelan habitats. Pp. 187-207 in: J. Eisenberg (ed.), vertebrate Ecology in the Northern NeOEEOpics. washington, D. C.: Smithsonian Institution Press. Eisenberg, J., and R. W. Thorington, Jr. 1973. A preliminary analy- sis Of a Neotropical mammal fauna. Biotropica 5:150-161. Ellerman, J. R. 1940. The families and genera of living rodents. VOlume I. Rodents other than Muridae. London: British Museum, 653 pp. Emlen, J. M. 1970. .Age Specificity and ecological theory. Ecology 54:588-601. Ewer, R. F. 1968. Ethology of mammals. New York: Plenum Press, 418 pp. FAO. 1965. Reconocimiento edafolégico de los Llanos Orientales, Colombia. V01. 1. Los suelos de lOS Llanos Orientales. United Nations Food and.Agriculture Organization, 528 pp. Fleming, T. H. 1975. The role Of small mammals in tropical eco- systems. Pp. 269-298 ip: F. B. Golley, K. Petrusewicz, and L. Ryszkowski (eds.), Small mammals: Their productivity and popu- lation dynamics. International Biological Programme 5. Cambridge: University Press. French, N. R,, D. M. Stoddart, and B. Bobek. 1975. Patterns of demo- graphy in small mammal populations. Pp. 73-102 13: F. B. Golley, K. Petrusewicz, and L. Ryszkowski (eds.), Small mammals: Their productivity and population dynamics. International Biological Programme 5. Cambridge: University Press. Goodman, D. 1974. Natural selection and a cost ceiling on repro- ductive effort. Amer. Natur. 108:247-268. 161 Goodman, D. 1980. Demographic intervention for closely managed popu- lations. Pp. 171-195 in: M; Soule and B. Wilcox (eds.), Con- servation biology: An.Evelutionary-ecological perspective. Sunderland, Nbss.: Sinauer Assoc. Green, R. 1980. A note on K-selection. Amer. Natur. 116:291-296. Haffer, J. 1974. Avian speciation in tropical South America. Publ. Nutall Ornithol. Club 14:1-390. Hershkovitz, P. 1955. On the cheek pouches of the tropical American paca Agouti paca (Linnaeus, 1766). Saeugetirk. Mitteil. Muenchen 3:67-70. Hershkovitz, P. 1969. The recent mammals Of the Neotropical region: A.zoogeographic and ecological review. Quart. Rev. Biol. 44:1-70. Hingston, R. 1932. A naturalist in the Guiana forest. London: Edward Arnold and Co., 285 pp. Hopwood,.A. T. 1947. The generic names of the mandrill and baboons, with notes on some of the genera of Brisson, 1762. Proc. 2001. Soc. Lond. 117:533-536. HOward, W. E. 1960. Innate and environmental dispersal of individual vertebrates. Amer. Midland Natur. 63:151-162. Howard, W. E., and H. E. Childs. 1959. Ecology Of pocket gophers with emphasis on Thomomys bottae mewa. Hilgardia 29:227-358. Hugett, A, St. G., and W. F. Widdas. 1951. The relationship between mammalian fetal weight and conception age. J. Physiol. Lond. 114:306-319. Instituto Geografico Agustin Codazzi. 1973. Atlas de Colombia, 3rd edition. Bogota, Colombia, 269 pp. Jarman, P. J. 1974. The social organization of antelope in relation to their ecology. Behaviour 58(3):l7l-209. Kleiman, D. 1974. Patterns Of behaviour in hystricomorph rodents. Symp. 2001. Soc. Lond. 34:171—209. Kleiman, D., J. Eisenberg, and E. Maliniak. 1980. Reproductive para- meters and.productivity of caviomorph rodents. Pp. 173-183 33 J. Eisenberg (ed.), vertebrate ecology in the northern neotropics. washington: Smithsonian Institution Press. Kraus, C., M. Gihr, and G. Pilleri. 1970. Das verhalten von Cuniculus aca (Rodentia, Dasyproctidae) in Gefangenschaft. Revue Suisse He Zoologie 77:353-388. 162 Krebs, J. R. 1971. Territory and breeding density of the Great Tit, Parus major L. Ecology 52:2-22. Lacépede, G. 1799. Tableau des divisions, sous-divisions, ordres, et enres des mammiféres. 12? Buffon, G., Histoire naturelle. VOl. l4, Quadrupes.) Paris: P. Didot L'Aine et firmin Didot. Lander, A. 1974. Observaciones preliminares sobre lapas Agouti paca (Linn. 1766) (Rodentia, Agoutidae), en venezuela. Univ. Central de venezuela, Fac. Agronomia, Inst. de Zoologia Agricola, Maracay, 104 pp. Laws, R. M. 1969. Aspects of reproduction in the African elephant, Loxodonta africana. J. Reprod. Fert., Suppl. 6:193-217 Leopold, A. S. 1959. 'Wildlife of Mexico. Berkeley: Univ. Of California Press, 568 pp. Lewontin, R. 1965. Selection for colonizing ability. Pp. 77-91 12: H. Baker and G. L. Stebbins (eds.), The genetics of colonizing species. New York: .Academic Press. Linnaeus, Carolus. 1766. Systema naturae. 12th edition, revised. Stockholm: Laurentii Salvii, 824 pp. Lockard, P., and D. Owings. 1974. Seasonal variation in moonlight avoidance by bannertail kangaroo rats. J. Mamm. 55: 189-193. Lotka,.A. J. 1907a. Relationship between birth rates and death rates. Science 26:21-22. Lotka, A. J. 1907b. Studies on the mode of growth of material aggre- gates. Amer. J. Sci., Ser. 4, 24:199-216. Lotka, A, J. 1956. Elements of mathematical biology. New York: Dover Publications, 465 pp. MacArthur, R. H., and E. O. Wilson. 1967. The theory of island biography. Princton Univ. Press, 203 pp. Marcgrave, J. 1648. Historia naturalis brasiliae. Tras. J. procopio de Magalhaes. Ed, Museu Paulista de Sao Paulo, Sao Paulo 1952. thamoros, Yolanda, personal communication. 1980. (Universidad National, Heredia, Costa Rica). Mendez, E. 1970. Los principales mamiferos silvestres de Panama. Privately printed, 283 pp. Michod, R. E., and W. W. Anderson. 1980. On calculating demographic parameters from age frequency data. Ecology 61:265-269. MOndOlfi, E. 1972. La lapa O paca. Defensa Natur. venez. 2(5):4-16. 163 MOrO, M. 1976. Utilization Of wild Species in human nutrition in the tropics. Pp. 55-63 12: 'VIII Inter-American Meeting on Foot-and- MOuth Disease and Zoonoses Control. PAHO Sci. Publ. 316. MOrris, D. 1962. The behavior of the green acouchi (Myoprocta pratti) with special reference to scatter hoarding. Proc. 2001. Soc. Lond. 139:701-732. MOrris, P. 1978. The use of teeth for estimating the age of Wild mammals. Pp. 484-494 13: P. Butler and K. Joysey (eds.), Development, function and evolution of teeth. New York: Academic Press, 523 pp. MOssman, H. W. 1966. The rodent ovary. Symp. 2001. Soc. Lond. 15: 455-470. MOssman, H. W., and I. Judas. 1949. Accessory corpora lutea, lutein cell origin and the ovarian cycle in the Canadian porcupine. Amer. J. Anat. 85:1-39. Murray, B. G. 1979. Population dynamics:.Alternative models. New York: Academic Press, 212 pp. Ojasti, J. 1973. Estudio biolégico del chiguire o capibara. Caracas: Fondo Nacional de Investigaciones Agropecuarias. Ojasti, J., and G. Padilla. 1972. The management of capybara in venezuela. Trans. 37th North American Wld. Nat. Res. Conf. 37: 268-277. Pearl, R., and L. J. Reed. 1920. On the rate of growth of the popu- lation of the united States since 1790 and its mathematical representation. Proc. Natl. Acad. Sci. U.S.A. 6:275-288. Pianka, E. 1970. On :_and K_selection. Amer. Natur. 104:592-597. Pianka, E. 1972. £_and K selection or b_and.d_selection? Amer. Natur. 106:581-588. Peilou, E. C. 1977. Mathematical ecology. New York: John Wiley and Sons, 385 pp. Pierret, V), and M. Dourejeanni. 1966. La caza y la alimentacion humana en 1as riberas del Rio Pachitea, Peru. Turrialba 16: 271-277. Pierret, V;, and M. Dourejeanni. 1967. Importancia de la caza para alimentacién humana en el curso inferior del Rio Ucayali, Peru. Rev. Forest. Peru 1(2):10-21. Pine, R. 1973. Mammals (exclusive Of bats) of Belem, Para, Brazil. Acta Amazonica 3(2):47-79. 164 Pocock, R. I. 1922. On the external characteristics of some hystrico- morph rodents. Proc. 2001. Soc. Lond. 25:365-427. Pollard,.l. 1973. .Mathematical models for the growth of human popula- tions. London: Cambridge Univ. Press, 181 pp. Ralls, K. 1971. Mammalian Scent marking. Science 171:443-449. Rice, L. A. 1980. Influences Of irregular dental cementum layers on aging deer incisors. J. Wildl. Manag. 44:266-268. Ricklefs, R. 1977. On the evolution of reproductive strategies in birds: Reproductive effort. .Amer. Natur. 111:453-478. Rowlands, I. 1974. Mountain viscacha. Symp. 2001. Soc. Lond. 34: 131-141. Rowlands, I. W., and B. J. weir (eds.). 1974. The biology Of hystricomorph rodents. Zoological Society of London Symposium 34, 482 pp. Sarmiento, G., and.M;‘MOnasterio. 1975. .A critical consideration Of the environmental conditions associated with the occurrence of savanna ecosystems in tropical America. Pp. 223-250_ig: F. B. Golley and E. Medina (eds.), Tropical ecological systems. New York: Springer verlag. Schaffer, W., and R. Tamarin. 1973. Changing reproductive rates and population cycles in lemmings and voles. Evolution 27:111-124. Sharpe, F. R., and A, J. Lotka. 1911. A problem in age-distribution. Philos. Mag. 21:434-438. Siler, W; 1979. A.competing-risk model for animal mortality. Ecology 60:750-757. Slade, N. K., and D. F. Balph. 1974. Population ecology of Uinta ground squirrels. Ecology 55:989-1003. Smythe, N. 1970a. Ecology and behavior of the agouti (Dasyprocta punctata) and related species on Barro Colorado Island, Panama. P .D lssertation, university Of Maryland, 201 pp. Smythe, N. 1970b. Relationships between fruiting seasons and seed dispersal methods in a neotrOpical forest. Amer. Natur. 104: 25-35. Smythe, N. 1978. The natural history of the Central American agouti (Dasyprocta punctata). Smithsonian Contribs. ZOOl. 257:1-52. 165 Smythe, N., W} Glanz, and E. Leigh, Jr. In press. Ecology and popu- lation regulation in pacas and agoutis. IE: E. Leigh, Jr. (ed.), Seasonal rhythms and long-term changes in a tropical forest. washington, D.C.: Smithsonian Institution Press. Spinage, C. A. 1970. Incremental cementum lines in the teeth of tropicalAfricanmaJmnals. J. Zool. Lond. 178:117-131. Spinage, C. A. 1972. African ungulate life tables. Ecology 53: 645-652. Stearns, S. C. 1976. Life history tactics: A review of the ideas. Quart. Rev. Biol. 51:3-47. Steele, R., and J. Torrie. 1960. Principles and procedures of statis- tics. New York: MCGraw-Hill, 481 pp. Stolzmann, J. 1885. Description d'un nouveau rongeur du genre Coelogenys. Proc. 2001. Soc. Lond. 1885 VI:16l-l67. Taber, R. D., and R. F. Dasmann. 1958. The black-tailed deer of the Chaparral. Bull. Calif.,qunx Fish and Game 8:1-163. Tate, G. H. H. 1931. Random Observations on habits Of South American mammals. J. Mamm, 12:248-256. Tait, D., and F. L. Bunnell. 1980. Estimating rate of increase from age at death. J. Wildl. Manag. 44:296-299. Vincelli, P. In press. Estudio de la vegetacién del Territorio Faunistico E1 Tuparro. Cespedesia (Cali), 1980. Walker, E., F. Warnick, S. Hamlet, K. Lange, M; Davis, H. Uible, and P. wright. 1975. jMammalS of the world. 3rd edition, revised by J. Paradiso. Baltimore: John Hopkins Univ. Press. Weir, B. 1974a. Reproductive characteristics of hystricomorph rodents. Symp. 2001. Soc. Lond. 34:265-301. Weir, B. 1974b. The tuco-tuco and plains viscacha. Symp. 2001. Soc. Lond. 34:113-130. wetterberg, G., M. Ferreira, W. dos Santos Brito, and V. C. de Araujo. 1976. Amazon fauna preferred as food. Brazil, PRODEPEF Field Document No. 23, UNDP/FAO/IBDF Project for Forestry Development, 17 pp, (mimeograph). WOOd, A, E. 1955. A revised classification Of rodents. J; mem, 36: 165-186. fl. IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII llIIIIIIIILIIIIIIII[IIIIIIIIIIIIIIJIIILJIIIIIIII