[page 19↓]

Chapter 2 
Effects of controlled seasonal breeding on reproductive performance traits of pastoral goat herds in northern Kenya

2.1 Introduction

In African pastoral systems, poor and fluctuating nutritional levels can cause reproductive inefficiencies in goat flocks. Despite the ability of many local goat breeds such as the Small East African to be non-seasonal, breeding females may exhibit prolonged periods of anovulation or anoestrus, reduced ovulation rates, as well as high embryonic and perinatal losses as a result of poor nutrition (Delgadillo & Malpaux, 1996; Walkden-Brown & Restall, 1996). Traditionally, pastoral producers in northern Kenya exert very little control over reproductive activity in their goat herds. In these systems, the impact of variations in pasture growth and forage quality on reproductive and other aspects of herd performance is minimized by optimizing migratory pathways between quality patches (Western & Finch, 1986). However, the increasing human demographic pressure on semi-arid rangelands in Kenya as well as a number of profound changes in the socioeconomic environment that have occurred over the last decades have triggered a rapid decrease in the mobility of pastoral herds and settlements (Schwartz & Schwartz, 1985; Hogg, 1986; Bennett, 1988). Also, the occurrence of sedentary pastoralism in the vicinity of small towns and trading centres is often characterised by a gradual commercialisation of pastoral production (Roth, 1990; Roth & Fratkin, 1991).

With respect to pastoral livestock production, the main problems arising from these developments are that the impact of climatic seasonality on all aspects of herd performance will tend to be aggravated. At the same time, improvements in the efficiency with which scarce rangeland resources are used are urgently needed to support an increasing human population. However, due to the low and erratic supply of nutrients from semi-arid rangelands it is difficult to envisage alternatives to herd mobility for overcoming seasonal nutritional deficiencies. One possible management intervention for achieving a better balance between nutrient requirements and supply in pastoral livestock operations is to concentrate reproductive activity in a single short breeding period. Restricting mating to a time that synchronizes the most physiologically demanding phases of the reproductive cycle with peak pasture production has been argued to be an important step to improve the productivity of pastoral goat herds in northern Kenya in particular (Field et al., 1984), as well as of small ruminant herds in tropical regions in general (Carles, 1983; Bradford & Berger, 1988; Delagadillo & Malpaux, 1996). Nevertheless, detailed information on the effects of controlled seasonal breeding on reproductive performance of goats in semi-arid regions in Africa is currently not available. Additionally, the few published results on this subject were obtained mainly from observational studies (e.g., Wilson et al., 1984; Wilson & Traoré, 1988) and there is a clear lack of experimental tests of the effects of such a management intervention on various reproductive traits, including probabilities of conception and abortion, as well as of rates of fertility, prolificacy, fecundity, and the proportion of kids weaned. Data on these parameters are indispensable for a sound assessment of how a controlled breeding regime affects biological (technical) efficiency of pastoral goat herds.

The objective of this study is to examine the effects of a restricting breeding regime on reproductive traits of Small East African (SEA) goats maintained under pastoral management on a semi-arid thornbush-savannah in northern Kenya. The effects of such a management intervention are evaluated by dividing the year into six consecutive mating seasons each of two months duration, so that the impacts of within-year variations in forage supply and production conditions are explicitly taken into account. This range of mating periods is intended to allow the identification of the most favourable period to which breeding could be restricted in this region to achieve optimal reproductive performance.

[page 20↓]

2.2 Materials and Methods

Study area

Data for this study were part of an experiment conducted at the University of Nairobi’s Ngare Ndare research station in the Isiolo District of northern Kenya. The research station is situated on a former holding ground of the Livestock Marketing Division, Ministry of Agriculture and Livestock Development, approximately 18 km southwest of Isiolo township, on the northern foothills of Mt. Kenya (centerpoint 0°22’N, 37°26’E). According to Ojany and Ogendo (1973), the climate in northeastern Kenya can be classified as tropical continental, semi-desert type. Rainfall in the study area is distributed over two distinct rainy seasons which receive 75 percent of total annual rainfall, a long rainy season from March to May, and a short rainy season from October to November. Rainfall records taken at the nearest meteorological station in Isiolo township display a long-term annual average of 615 mm (Kenya Meteorological Department, 1931-1991, n=49). The long-term median annual rainfall is 583 mm; in 50 percent of the years rainfall amounts ranged between 498 and 694 mm.

The predominant vegetation type is open Acacia tortilis-Lippia carviodora dwarf shrub bushland (Cornelius & Schultka, 1997; Schultka & Cornelius, 1997), occurring on upland sites, valley slopes and valley bottoms. On heavily grazed pastures in these areas, the herblayer is characterized by high frequencies and coverages of the grasses Sporobolus nervosus and Tragus bertorianus while the dominant dwarf shrub is Indigofera spinosa. Chrysopogon plumulosus and Hibiscus micranthus dominated communities are found in transition areas between uplands and riparian zones. Both of these communities are important forage resources for livestock (Walsh et al., 1993). Acacia mellifera-Acacia paolii shrublands are associated with deeply layered Vertisols on upland sites and valley bottoms. The dominant grasses are Setaria acromelena and Cyperus blysmoides, while Farsetia stenoptera and Blepharis linnarifolia are characteristic forb species occurring on these sites. Acacia reficiens and Acacia horrida thickets are confined to sites on eroded Vertisols in valley bottoms. Acacia xanthophloea or Acacia tortilis riparian gallery forests and woodlands occur in areas with permanent ground water access. Seasonally flooded plains are dominated by semi-terrestric Cynodon plectostachyous swards and are frequently grazed during dry seasons. According to the classification of Pratt and Gwynne (1977) the research area falls into ecozone V.

Experimental design

The study commenced in early 1983, when a herd of Small East African goats comprising approximately 60 mature and immature females was provided by the Research Division of the Ministry of Livestock Development. By culling unsuitable animals and further local purchases, a total of 90 mature does and 50 immature females were available at the end of 1983. This herd was used to start a systematic breeding programme designed to achieve year-round mating, kidding, and weaning. Experimental animals were between one to six years of age. Female replacements required at later stages of the study were obtained from immatures born during the course of the experiment. The female herd was stratified according to age and body weight, and animals assigned to treatment groups were randomly sampled from each stratum. Females weighing less than 20 kg were excluded from the sampling procedure. Each treatment group comprised approximately 18 does.

In this experiment the treatment structure consisted of a sequence of 18 treatment groups, into each of which a buck was introduced for a period of two months. At the end of each two-month period the buck was transferred to the next group, so that the experiment generated six different treatment groups per year. Hereafter, the latter will be referred to as the "breeding groups" of the experiment. The first breeding group was set up at the end of January 1984, and the breeding programme was concluded in January 1988, ending with breeding group number 18. Thus, three complete productive cycles were obtained for each two-month breeding group over the four years of the experiment. Summary information on the number of breeding females and number of kids born in the first seventeen breeding groups is displayed in Table 2.1.

For the purpose of analysis, breeding group eighteen was discarded because of incomplete records. The three replications for each breeding group were labelled with an identifier, mating season, which was used as the treatment variable in subsequent analyses. Except for mating seasons 4 and 5, all treatment levels refer to a breeding season of two months length. Due to a delay which occurred in setting up breeding group 4 in 1984, breeding in mating seasons 4 and 5 were assumed to have taken place over a 3 month period (Table 2.1). For the same reason, mating season 6 was not observed in the first year of the experiment, and only two complete reproductive cycles were available for that group. Note that the experimental design did not include a control group in which does were allowed to breed continuously throughout the year for the entire experimental period. Therefore, no "aseasonal" mating group was available as a reference for comparative purposes.

A total of 145 does were utilized to set up the 17 breeding groups. From these, 93 animals were allocated to two or more different breeding groups in the course of the experiment. A total of 65 does were observed in at least two consecutive production cycles within the same mating season group. Thus, with regard to the statistical analysis of the experimental data, individual does were nested within the observed combinations of mating seasonxproduction cycle.

[page 21↓]

Table 2.1. Summary information on breeding and mating season groups

Herd management and data collection

The animals were maintained under traditional management, and all day-to-day herding decisions were left to local pastoralists. The goats were herded every day from 07:00h and brought back to the station for watering at 12:00h. Afternoon grazing started at 15:00h and ended at 18:00h, when the animals returned to night enclosure. During a grazing day the animals walked between 5 and 10km, where longer distances were travelled during the dry season.

Newborn kids were weighed and ear-tagged as soon as possible after birth. All male kids were castrated within the first week of life using the rubber ring method. Until approximately two months of age kids were kept for the whole day in the night enclosure, and thereafter until weaning at 16 weeks of age were left to roam freely in immediate vicinity of the night enclosure. Kids were allowed to suckle their dams from their return from afternoon grazing until the next morning; does were not milked during lactation except for milk yield measurements. Surplus young stock were disposed of at one year of age. All female kids were allocated to the breeding herd at one year of age, with the exception of those culled for poor conformation.

In order to minimise the effects of infectious diseases and internal parasites on animal performance, a health care programme was carried out regularly in the experimental herd. The animals were drenched twice a year with Panacur® (Fenbendazole, Hoechst AG, Frankfurt a.M.) prior to the rainy seasons and were vaccinated once a year against Contagious Caprine Pleuro-Pneumonia (CCPP), which is prevalent in the area. Whenever tick infestation was noted, the animals were sprayed with Bacdip® (Organophosphorus, Bayer, East Africa Ltd.). A salt lick, Maclick® (KFA Ltd. Nairobi) was offered in the night enclosures, and injuries were treated with antibiotic spray.

Time in the experiment was divided into periods of two weeks, each period starting and ending on a Monday. Routine measurements of weight (kg) of all animals, milk yields (ml) were carried out at the beginning of each two-week period and entered in record sheets. All events such as abortion, birth, and death were recorded continuously. Adult animals were weighed using a weighing crate and measurements were made to the nearest 0.5 kg. Kids were weighed immediately after birth using a spring balance to the nearest 0.25 kg.

Pasture condition was judged every two weeks using a subjective phenological pasture condition score for [page 22↓]the herblayer (range condition score [I]), including grasses, herbs, and small dwarf shrubs. Scores ranging from 1 to 4 (low to high) were allocated at each occasion according to forage availability and quality. The condition score for the herblayer was upgraded to a maximum score value of 5 to integrate the contribution of the bush and tree layers with regard to browse availability and the production of high quality litter such as leaves, flowers and fruits (range condition score [II]). Scores of 1 to 3 generally occurred during the dry seasons, and 4 and 5 during and immediately after the rainy seasons.

Traits studied

The definition of parameters of reproductive performance was based on the recommendations of Terrill and Foote (1987). In addition to the parameters suggested by these authors (i.e., fertility, prolificacy, fecundity, and proportion weaned), conception (=number of does pregnant per doe exposed) and abortion rate (=number of does aborting per doe pregnant) also were estimated in the present study. The higher resolution with respect to the components of reproduction was meant to facilitate investigating the influence of seasonality in forage supply on compound measures of reproductive performance such as fecundity (=number of offspring born alive or dead per doe exposed) and proportion of kids weaned (=number of kids weaned per doe exposed). Birth rate was defined as the number of does kidding (without abortions) per doe exposed and differs from the definition of fertility reported by Terrill and Foote (1987) in that abortion events do not enter into the calculation. The same applies to prolificacy rates, which in the present study were defined as the number of offspring born alive or dead (without abortions) per birth event.

The control of breeding to a single breeding period dictates once-a-year reproduction. The estimation of parturition intervals has little relevance in this setting and was therefore not considered. Moreover, the number of breeding does available in the experiment precluded the estimation of the probability of rebreeding for consecutive production cycles within the same mating season group; as mentioned above, only 65 does were allocated at least twice to the same mating season treatment.

Statistical analysis

Conception, abortion, and birth rate are dichotomous outcomes and were analysed through logit models (extensive discussions of logit models can be found in Agresti (1990) and Christensen (1990)). Poisson regression models were fitted to prolificacy, fecundity, and weaning rate. By definition, these outcomes are rates where the measurements to be analysed are the number of events of a specified type divided by a relevant baseline measure E (Agresti, 1990) - e.g., for prolificacy, the number of offspring born divided by the number of does kidding. In contrast to logit models where the baseline is given by the frequency of the second outcome of the dichotomous response, an appropriate and context specific measure E (e.g., the number of does kidding) must be specified in order to properly define the rate to be modelled. The logarithm of this baseline measure, the so-called offset, enters as a constant parameter into the models being estimated.

Logit and Poisson regression models for rates and proportions are special cases of generalized linear models (GLM), a class of models first suggested by Nelder and Wedderburn (1972). An extensive treatment of these models is given by McCullagh and Nelder (1989). GLMs can be extended to generalized linear mixed models (GLMM) which allow for simultaneously taking into account random model effects (or dispersion components) and non-normally distributed errors. In the generalized linear mixed model approach, instead of fitting a fixed-effects model to expected outcomes, a mixed model is fitted to the conditional mean of the observations, given the random model effects using quasi-likelihood or pseudo-likelihood procedures (see Wolfinger & O'Connell, 1993; Littell et al., 1996).

The SAS procedure GENMOD (SAS Release 6.12, 1996) and the SAS macro GLIMMIX (October 1995 version) were used to fit fixed and mixed effects logit and poisson regression models, respectively. REML estimation was used for models containing random effects. If appropriate, best linear unbiased predictions (BLUP) of random effects were obtained through the ESTIMATE statement provided by the SAS macro GLIMMIX. The likelihood ratio statistic or deviance served as a criterion to test for overall agreement between modelled and observed responses and to exclude from consideration models that did not fit observed data well. In analogy to analysis of variance the deviance of a sequence of models, each including one term more than the previous one, can be used to produce an analysis of deviance (McCullagh & Nelder, 1989). Here, the change in deviance between nested models is used as a measure of discrepancy and compared to a c² distribution with p degrees of freedom as a test of statistical significance, where p is the difference in degrees of freedom between two nested models.

Identification of parsimonious models for the data was based on backward selection from a model including all possible interaction terms up to the third degree, with terms sequentially removed if the reduction in deviance, adjusted for all other terms in the model, was not significant at the 15 percent level. However, the [page 23↓]mating season treatment effect was retained in all models, regardless of its significance, in order to provide smoothed estimates of all traits for the six mating seasons.

The three consecutive reproductive cycles were considered to be random blocked replications. Therefore, the random factor of reproductive cycle was tested for significance in all models that were developed. The treatment structure was an n-way factorial arrangement, n denoting the number of treatment and classification factors included in the analysis. The sole treatment factor considered was mating season. Classification factors evaluated in preliminary models fitted to rates of conception, abortion, prolificacy, and fecundity are listed in Table 2.2. In analysing birth rate and proportion of kids weaned, lagged median range condition scores [I] and [II] at mating were not taken into account.

Table 2.2. Classification factors used in statistical analyses.

Since the effect of mating season and that of lagged range condition scores [I] and [II] at mating were confounded, fitting separate models for assessing the effects of these variables was warranted to avoid estimation problems arising from multicollinearity in the linear predictor. In order to control the experimentwise error rate at the prespecified level of α=0.05 when making multiple comparisons of factor level means, both the bonferroni and Tukey multiple comparision procedures were used. The procedure giving the narrower confidence limits was then chosen to report significance probabilities of differences in factor level means. This choice is proper since it does not depend on the observed data (Neter et al., 1996).

[page 24↓]

2.3 Results

Range condition

This section provides a summary of environmental conditions prevalent during key phases of the productive cycle, as captured by the pasture condition assessment procedure described above. It is intended to facilitate interpretation of the impact of the breeding season treatment on performance traits. The biweekly values for range condition scores [I] and [II] recorded throughout the experimental period were used to compute median scores for each of the six mating season groups over a production cycle of one year duration (Figure 2.1). The graphs show the timing of mating, kidding and weaning events in relation to the average pasture condition experienced over three (mating seasons 1 to 5) and two (mating season 6) consecutive production cycles.

The most favourable pasture conditions during the breeding period were observed in mating season 2 (long rainy season) and 1 (end of short dry season). For mating groups 3 (beginning of long dry season) and 4 (middle of long dry season), the onset of mating was characterized, respectively, by rapidly deteriorating and poor forage quality and quantity. Fair range conditions prevailed during the short rainy season in November and December, when mating took place in group 5. Somewhat less favourable forage conditions where encoutered by breedong does in mating group 6 during the short dry season. The poorest range conditions at kidding were experienced by mating season group 2, in the middle of the long dry season. Similar range conditions, with respect to score [I], prevailed at the time of kidding in group 1 (beginning of long dry season). The peak in the adjusted range condition score [II] during the same period was caused by an increased availability of high quality litter such as leaves, fruits and flowers of Accacia species. According to Schultka and Schwartz (1987), these are the most important dry season browse for goats in this region.

Figure 2.1. Observed median range condition in the herblayer (score [I]) and browse-adjusted range condition (score [II]) by mating season group (MS) over a production cycle of one year duration. Mating starts in week 0; the x-axis label "-4" refers to the time point four weeks prior to mating.

[page 25↓]

All other treatment groups gave birth under favourable environmental conditions typical of the short rains (group 3), the short dry season (group 4), as well as the beginning of during the long rainy season (groups 5 and 6). Kids born in mating season groups 2, 4, and 5 were weaned during periods with fair to good quality and quantity of forage on offer, such as prevails during the two rainy seasons. Increased availability of the high quality litter mentioned above helped to maintain range condition score [II] at fairly high levels toward the end of the long rainy season, when kids in group 5 were weaned. Somewhat less favourable pasture conditions were observed at the same production stage in group 3. In contrast, very poor forage conditions were encountered by kids born in groups 1 and 6, which were weaned in the middle and at the end of the long dry season, respectively.

Conception rate

The average ages at breeding, for animals in the various parity classes, for which age records were available were 67.8 weeks (SD 12.6) for first breeders, 117.6 (SD 19.3), 170.8 (SD 20.6), and 202.7 (SD 19.7) weeks for parity one to three animals, respectively. Only one age record was available for does with four or more prior kiddings (243 weeks). Mean body weights at breeding by parity stage are shown in Figure 2.2. Body weight was highest in third parity does with an average of about 38 kg and tended to decline thereafter. Of the 287 does mated during the experiment, 255 or 88.5 percent conceived. Except for range condition score [I], none of the fixed effects tested (parity; weight at breeding; lagged RC score [II]) had a significant influence on conception rate.

Figure 2.2. Box and Whisker plot of doe body weight at mating against parity number. The fitted function is a second-order polynomial.

Table 2.3. Analyses of deviance for logit models fitted to conception rates.

[page 26↓]

The mating season treatment and production cycle effects were also non-significant. The effect of lagged RC score [I] at mating on conception rates was analysed in a separate model (Table 2.3). As expected from the analysis of deviance for model a) in Table 2.1, none of the pairwise comparisons of expected conception rates across mating seasons were significant. However, the highest conception rate was observed when does were joined just before the onset of the long rainy season (mating season 1, Table 2.4), and the lowest when does were mated during the rainy and at the beginning of the dry season (mating season groups 2 and 3). Relatively high conception rates were observed in mating seasons four (0.89) and five (0.93), both of which were characterized by improving nutritional conditions around the time of mating, especially during the month of October.

Estimated conception rates in relation to lagged RC score [I] in Table 2.4 reveal a curvilinear trend, with the lowest values being observed under very poor and good range conditions, and the highest at intermediate scores of 2 and 3 (0.97 and 0.93, respectively). The latter were typically measured just prior and immediately after the rains, and thus typify situations in which forage conditions were gradually changing in direction of either the lower or upper extremes of the score scale.

Table 2.4. Parameter estimates for logit models fitted to conception rates (f denotes the extra dispersion parameter).

Abortion rate

A total of 15 abortion events out of 255 pregnancies (5.9 percent) were observed during the three production cycles. The small number of events made it difficult to fit higher level models to the data, and none of the effects tested, i.e. mating season, parity, doe body weight at mating, and lagged RC scores [I] and [II] at mating significantly affected abortion rates. The corresponding analysis of deviance is displayed in Table 2.5. The deviance of the mating season effect (MS) was substantially reduced after the non-constancy of mating season effects between production cycles (PC) was accounted for (σ2 PC x MS>0 versusσ2 PC x MS=0). Note that the production cycle dispersion component, σ2 PC, was set to zero by the REML estimation procedure and therefore omitted from the final model. The PCxMS interaction was caused mainly by large differences in abortion rates among production cycles in mating season groups 4 and 5. As expected from the analysis of deviance, no significant difference in abortion rates between the six mating seasons groups was found (Table 2.6).

[page 27↓]

Table 2.5. Analyses of deviance for logit models fitted to abortion rates.

However, approximately 40 percent of all observed abortion events occurred in mating season 5, compared to 27 percent in mating season 4. Expected abortion rates in the latter two groups of approximately 15 and 9 percent, respectively, represented major productive inefficiencies. A comparison of model predictions pooled over mating seasons 4 and 5 on the one hand with those pooled over all other mating seasons on the other, revealed a significant difference at the 5 percent level (F(1, 13.8)= 5.3). In general, fetal losses increased as pasture condition scores declined around the time of mating (group 5) and during mid (group 4) and late pregnancy stages (group 3). The lowest abortion rates (<3.5percent) were observed among goats mated during the short (group 6) or at the beginning and during the long rainy season (groups 1 and 2). In these cases, favourable forage conditions prevailed throughout the gestation period.

Table 2.6. Parameter estimates for the final model fitted to abortion rates.

Birth rate

Overall birth rate in the experiment was 80.1 percent. Of the 255 breeding does that conceived, 21 aborted or died during pregnancy. Except for lagged range condition [I] at mating, the statistical analyses revealed no significant effect on birth rate of any of the explanatory variables considered. As before, the mating season treatment effect was retained in one of the two final models fitted to fertility data in order to provide smoothed estimates of birth rate for each mating season. Likewise to the analysis of conception and abortion rates presented above, the inclusion of the PCxMSinteraction as a random effect led to a noticeable decrease in deviance for the mating season effect (Table 2.7). Although significant differences in expected birth rates between mating seasons were found when the fixed-effects model was fitted to the data, all pairwise comparisons were nonsignificant in the mixed-effects model.

[page 28↓]

Table 2.7. Analyses of deviance for logit models fitted to birth rates.

Table 2.8. Parameter estimates for final logit models fitted to birth rates.

Lagged range condition score [I] at mating was analysed separately, and was found to have a significant influence on birth rates (p <0.01). Given that birth rate is a function of conception rate, abortion rate, and doe survival until parturition, some of the patterns described previously were carried over to the analysis of birth rates (Table 2.8). The high conception and fetal survival rates observed for mating season group 1, for instance, resulted in a high birth rate of approximately 95%. In contrast, the low number of fertilisations in group 2, as well as the high fetal mortality in mating season 5 were responsible for the comparatively poor performance of does joined during these periods. Large differences occurred between replications for does bred in mating season 5, with a very low birth rate of 43 percent in the second production cycle. Intermediate [page 29↓]levels of performance with birth rates averaging about 81 percent were observed in mating season groups 3, 4 and 6.

Lagged range condition score [I] at mating had an effect on birth rates similar to that it had on conception rates. Low and high values of this index corresponded to significantly lower birth rates than the intermediate values of 2 and 3. The latter, for instance, typically occurred around the time of mating in group 1, whereas scores of 1 and 4 characterised pasture conditions in groups 2 and 5, which achieved the lowest birth rates among all groups.


The most frequent litter size was unity accounting for 59 percent of births, followed by 40.6 percent for twins and only one triplet, yielding an overall mean litter size of 1.44. Models fitted to prolificacy rates revealed a significant main effect of doe parity (p <0.01), and main effects for lagged range condition scores [I] and [II] effects at mating (Table 2.9). Two separate models were fitted to the two range condition scores because of the high correlation between the two indices. After adjusting for the effect of range condition, additional dispersion components were set to zero by the REML estimation procedure (PC and PCxMS) and were, therefore, omitted from further consideration.

A third poisson regression model fitted contained mating season and doe parity at mating as the linear predictors. Mating season neither significantly affected predicted prolificacy rates in the fixed effects models, nor in the model incorporating the random PCxMSinteraction. In contrast, differences between parity numbers remained highly significant in both of these models.

Body weight at mating exerted a significant effect (p <0.01) when no allowance was made for the effect of parity on prolificacy rates, but was nonsignificant (p>0.5) after adjusting for the latter. This result was largely due to the strong positive correlation between body weight at mating and parity (seeFigure 2.2). Conversely, the effect of parity on prolificacy was not influenced by the weight at mating term. Based on this finding, it

Table 2.9. Analyses of deviance for Poisson regression models fitted to prolificacy rates.

was concluded that the effect of weight at mating was confounded with that of parity. Therefore, only the latter term was retained in the final models presented in Table 2.9. Does bred in the first mating period, just prior to the long rains, produced the largest litter size per birth event (1.58 kids, Table 2.10). The conspicuous depressing effect of mating season 2 (1.30 kids) on prolificacy was unexpected, since, in this group, does were joined at the peak of the long rainy season when, in general, good feeding conditions prevailed. The fact that fairly large numbers of twin birth events occurred in mating groups 4 and 5 was also unexpected given that these seasons coincided with the middle and the end of the long dry season, respectively. Prolificacy [page 30↓]rates in mating seasons 3 to 6 showed only minor deviations from the predicted mean value over these four groups of 1.45. With respect to lagged range condition scores at mating, the patterns described previously for other traits persisted in the analysis of prolificacy rates.

Table 2.10. Parameter estimates for final models fitted to prolificacy rates.

Very poor (level 1) and favourable (level 4) pasture conditions prior to mating had a depressing effect on prolificacy. A significant difference was found between the largest litter sizes predicted for level 3, and the lowest ones predicted for level 4. The relation between RC score [II] and conception rates was less consistent (p>0.1) and, consequently, no significant differences among score levels could be detected. However, estimated mean litter sizes across RC score [II] levels displayed an overall pattern similar to that observed for RC score [I].

The most important influence on prolificacy rates was exerted by parity number. As can be seen from Table 2.10, predicted values of prolificacy rates by parity number remained relatively consistent across all three [page 31↓]models, indicating a very small dependence of model predictions for this term upon other explanatory factors. A considerable increase occurred in prolificacy rates with increasing number of lactations, starting from a minimum of 1.1 kids per birth for first breeders to a maximum of about 1.7 kids per birth at the fourth kidding. Litter sizes declined in older does with at least four prior kiddings. Orthogonal polynomial contrasts revealed significant linear (F(1, 52.2)=26.5) and quadratic (F(1, 54.5)=4.3) trends in prolificacy rates with increasing parity stage.

In order to ascertain that no differential effects of body weight at mating were present within each of the five parity classes, all observed combinations of body weight and parity levels were combined into a single variable with 20 levels. A model including the mating season effect and the weight-within-parity variable was then fitted to the prolificacy data. The weight-within-parity variable was found to be highly significant χ2=61.8, df=19). However, no significant differences in mean prolificacy between body weights within parity levels could be detected.


Fecundity rate is a function of birth rate and prolificacy. The results of the statistical analysis presented below thus summarize, in a way, the joint effects of these two traits. As might be expected from the foregoing, the analysis revealed no significant effect of mating season once adjustment was made for the random PCxMS interaction (Table 2.11). In contrast, the significance of parity number was not affected by the inclusion of this dispersion component. The production cycle dispersion component, σ2 PC, was set to zero by the REML estimation procedure and therefore omitted from the final model.

Since body weight at mating and parity were confounded, the former effect was not considered in the analysis. Furthermore, no significant effect of range condition score [II] was found. The analysis of deviance (Table 2.11) for the model containing RC score [I] and parity number revealed significant effects of both variables on predicted fecundity rates (p <0.05). Parameter estimates in Table 2.12 show that the predicted response patterns of fecundity rates across mating season groups resembles that described previously for birth and prolificacy rates.

Does bred in the first mating period performed particularly well, with an expected number of kids born per doe exposed of 1.46. In contrast, reproductive efficiency was severely reduced when does were mated at the peak of the long rainy season (group 2). Although not statistically significant, the difference between these

Table 2.11. Analysis of deviance for models fitted to fecundity rates.

[page 32↓]

Table 2.12. Parameter estimates for final models fitted to fecundity rates.

two treatment groups was substantial, amounting to 0.49 kids per doe exposed. Comparatively high fecundity rates were also achieved in group 4. Estimates for groups 3, 5, and 5 were of similar magnitude but somewhat lower than those for groups 1 and 4. Predictions for the former were quite unreliable, as can be seen from the standard errors of the predicted means. Parity number had an outstanding effect on fecundity rates, which, based on estimates accounting for the effect of mating season, ranged from 0.97 kids per doe exposed for first parturitents, to 1.38 kids per doe exposed for breeding females with three prior kiddings. A significant overall linear trend across parity levels was demonstrated (F(1, 54.5)=6.51), but not the presence of a quadratic curvature, suggested by the observed decline in theresponse of does with at least four prior kiddings, had to be rejected. Mean separation tests showed significant differences in the expected numbers of kids born per doe exposed belonging to levels 1 and 4 versus levels 2 and 3, respectively, of range condition score [I] at mating. The highest fecundity rates were achieved for a score of 3 at mating.

Weaning rate

The number of offspring weaned per doe exposed is a measure that combines two components of reproductive performance, fecundity and postnatal survival of kids. Survival rates of kids until weaning are discussed in Chapter 3, and a detailed account of the factors affecting kid survival, and thus weaning rate, is therefore deferred to that chapter.

In the fixed effects model, only parity exerted a significant effect on weaning rates (Table 2.13). For the same reasons as before, weight at breeding was dropped from the final model. The variation due to production cycle and its interaction with mating season caused the parity main effect to become insignificant and, consequently, pairwise comparisons of expected weaning rates across parity numbers failed to reveal any significant differences (Table 2.14). However, does with higher lactation numbers seemed to achieve better weaning rates than those in the first and second lactation. The response peaked for fourth parity, and declined [page 33↓]afterwards. The slightly lower expected response for primiparous does compared to first breeders was suspected to be related to an interaction between mating season and parity number. Although the interaction term appeared to be an unimportant source of variation (p=0.303), a test of simple effects showed a substantial difference in weaning rates between the first two parity levels within mating season four (parity0=0.928; parity1=0.436). This might account for the observed attenuation of the response for primiparous females.

Table 2.13. Analysis of deviance for Poisson regression models fitted to weaning rates.

When compared to fecundity rates in Table 2.12, a considerable change in relative performance of does bred in the six mating seasons is apparent. High mortality rates in kids obviously depressed the reproductive efficiency of animals joined in the first and sixth period, with expected weaning rates of 0.77 and 0.82 kids weaned per doe exposed, respectively. Weaning rates in mating season groups 2 and 3 were also affected by elevated kid mortality, although to a lesser extent. The best performance was achieved by does mated in the middle of the long dry season in mating season group 4, with a predicted weaning rate of 1.12 kids weaned per doe exposed.

Table 2.14. Parameter estimates for final Poisson regression models fitted to weaning rates.

[page 34↓]

2.4 Discussion

Measurement of reproductive performance

A high rate of reproductive efficiency is often thought to be the most important prerequisite for the production of meat. milk, skins, and breeding stock (Terrill & Foote, 1987; Steinbach, 1988; Wilson, 1989). The term reproductive efficiency, however, is often poorly defined in the literature (Baptist, 1992b), and a confusing multitude of different measures have been employed in livestock production studies, some of which have recently been reviewed by Bosman et al. (1997). Steinbach (1988) for instance, in assessing reproductive performance equates the term “biological productivity” with weaned weight per metabolic doe weight per year. Similar indices have also been used by Knipsheer et al. (1984), Peacock (1987), and Wilson (1984). Strictly speaking, these types of indices do not qualify as true measures of technical (biological) efficiency of reproduction in livestock herds, since they do not relate outputs yielded to inputs required in the production process (Baptist, 1992b). More precisely, they reflect yield levels but fail to relate this to a given level of production inputs. Measuring reproductive performance in terms of a compound measure such as weaned weight per breeding female (or unit metabolic weight thereof) provides little insight into the relative importance and impact of component traits on the calculated value (Bosman et al., 1997b) and, additionally, implies a value judgement as to the objectives of the decision maker. For instance, while maximisation of output in terms of weaned liveweight can be a relevant criterion in commercial meat production systems, this is not necessarily true in dairy or dual-purpose operations, and even less so in subsistence oriented production systems (Behnke, 1985).

In the present study the effect of seasonal breeding on reproduction was evaluated in relation to specific responses that are linked to physiological processes affecting reproductive performance, such as ovulation rate, fertilization rate, embryonic and/or fetal survival, and proportion of dams and offspring surviving to parturition. Defining reproductive performance in its most specific components bears several advantages in this context. Firstly, the level of resolution provided by this approach facilitates the identification of factors influencing the biological potential of a herd to produce offspring, and allows the assessing of how these factors are affected by experimental treatments. Also, for analytical purposes, it would seem appropriate to isolate reproductive responses from kid survival and growth, because the component of environmental variance attached to the latter two traits is often high. Secondly, whereas the estimation of parameters such as probability of conception and abortion, birth rate, prolificacy, and fecundity is straightforward, special problems arise in defining measures such as kidding rate (offspring born per breeding female per time unit) and kidding interval (average time interval between parturitions) which, in many cases, are required for computing compound indices of reproductive performance. For instance, Baptist (1988) and Upton (1989) demonstrated that failure of taking into account fractions of temporarily unreproductive as well as sterile females in a herd are sources of inconsistencies and errors in estimating kidding rate and kidding interval.

General impact of environmental conditions on reproduction

Reproductive performance in goats is a composite of several processes which are influenced by environmental, developmental, genetic, and managerial factors (Terril & Foote, 1987). The present work focused on investigating the influence of management (mating season), environment (range condition) and developmental history (parity) on these traits.The mating season treatment was hypothesized to act on reproductive performance by altering the nutritional status of dams around breeding and throughout the pregnancy stage, while range condition was assumed to be an indicator of nutritional conditions prevailing just prior to mating only. However, mating season did not exert a significant influence on any of the traits analysed. Except for conception rates, all models fitted revealed a substantial variation in the mating season effect across production cycles, thus suggesting that inter-year variability in environmental conditions was very large and hence masked a clear expression of differential effects on reproductive function of controlled breeding. In contrast, range condition, and thus forage quantity and quality on offer, during the pre-mating period significantly affected some of the reproductive responses analysed (conception rate, birth rate, prolificacy, and fecundity). This is not surprising, since it is a well established fact that in the absence of photoperiodic cuing, nutrient supply is the main environmental regulator of all aspects of reproduction (Walkden-Brown & Restall, 1996). According to Bronson (1989), a reproductive strategy that is not regulated by seasonal predictors, photoperiod or otherwise, can be classified as opportunistic. In its extreme form, opportunism dictates that males remain sexually ready at all times of the year, and that the breeding behaviour of females is responsive to short-term changes in energetic and nutritional conditions.

[page 35↓]Conception rate

The conception rates in excess of 82 percent across all mating season groups are a clear evidence of non-seasonality of reproduction in SEA goats. Moreover, since conception rate was defined as the fraction of does kidding and aborting relative to the total number of does present at mating, without taking into account embryonic and fetal losses that may have occurred, it is likely that the conception rates predicted for each mating season group underestimated the true rates of fertilization achieved during the experiment. Unfortunately, there are only few reports available to which the present results could be compared. This is partly due to the fact that most of the relevant reports do not provide figures on conception rates achieved in the flocks under study, but only birth or kidding rates, or merely production indices (e.g., Mellado et al., 1991, 1996; Ndlovu & Simela, 1996; Sachdeva et al., 1973; Wilson, 1989; Wilson et al., 1984; Wilson & Light, 1986). On the other hand, experimental studies dealing with aspects of reproductive physiology in tropical, non-photoresponsive goat breeds are scarce (Walkden-Brown & Restall, 1996; Wilson, 1989).

Nevertheless, Peacock (1984) investigated the reproductive performance of Maasai SEA goat flocks which were exposed to a buck during the months of July and August. Her results can thus be compared to those obtained for mating season group 4 in the present study. The author analysed the reproductive performance achieved in two different groups of flocks. One had been moved to pastures on which Acacia tortilis was one of the dominant plant species in order to allow the goats to feed on the high quality litter, predominantly Acacia tortilis pods shed by these trees at the peak of the long dry season. The other group was herded on pastures on which animals had no access to Acacia tortilis litter. The analyses revealed that in two flocks of pod-fed animals about 77 and 83 percent of the does conceived, compared to only 20 percent in the flocks not pod-fed. As the major difference between the two treatments was in the animals actually mated, Peacock (1984) concludes that the consumption of pods mainly affected the occurrence of oestrus. By comparison, the conception rates achieved in the pod-fed flocks are somewhat lower than those predicted for mating season group 4 in the present study (89 percent). The comparatively good performance of this group was surprising, since one would have expected that the poor nutritional conditions prevailing at the peak of the long dry season, and their associated effect on body condition of does, should have had a deleterious impact upon oestrus activity, ovulation, and fertilization rate, such as was reported, for example, by Sachdeva et al. (1973)

Figure 2.3. Frequency distribution by mating season of the duration between onset of mating and parturition in fertile females.

[page 36↓]

and Mani et al. (1992). Peacock’s study provides at least a partial explanation for this phenomenon, in spite of the fact that the browse adjusted range condition score [II], which was supposed to reflect changes in the availability of high quality litter, did not exert a significant influence on conception rates.

Although differences in conception rate among treatments were not statistically significant, it is notable that conception was minimal when does were joined under the most favourable forage conditions. Only 83 percent of the does mated at the peak of the long rainy season in April (mating season 2) became pregnant, and a similar value (82 percent) was predicted for a median pre-mating range condition score [I] of 4. Conversely, maximum conception rates occurred in groups mated during the short (group 1) and towards the end of the long dry season (group 5), or for median pre-mating range condition scores [I] of 2 or 3. In mating season group 5, however, nutritional anoestrus may initially have occurred in some of the does exposed as suggested by the distribution of the duration between first exposure to the buck and parturition, depicted in Figure 2.3. Given that the length of the breeding period was two months, and assuming a gestation period of roughly 22 weeks, does giving birth after more than 28 weeks must have been mated towards the end of the breeding period, which, in case of group 5, coincided with the peak of the short rainy season in November. Resumption of ovarian and oestrus activity thus may have occurred in response to improving nutritional conditions. It is also interesting to note that in most fertile females joined during the short (group 1), at the middle of the long dry season (group 4), as well as towards the end of the short rainy season (group 6), fertilization must have occurred within a few weeks from the onset of mating. In contrast, the proportion of non-cycling does appears to have been larger throughout the long rainy season (groups 2 and 3).

These results suggest, firstly, that whenever feeding conditions start to improve oestrus and ovulation are positively affected. The sexual activity of SEA goats thus seems to be sensitive to short-term changes in nutrient supply. This observation fits well into the theoretical framework advanced by Bronson (1989), which postulates that in the absence of photoperiod as a predictive cue, external climatic and dietary factors may act as predictors of conditions promoting reproductive success in mammals. Such an opportunistic reproductive strategy may also rely on short-term predictors obtained from plants. Based on Bronson (1989), the use of plant predictors can be expected to be particularly advantageous for strict herbivores of short life span that live in highly unpredictable environments. Also, previous studies on goats reviewed by Walkden-Brown and Restall (1996) provide some evidence that the sudden availability of good nutrition may induce oestrus and ovulation in non-photoresponsive breeds. According to Landau et al. (1996), these short-term effects are achieved through provision of nutrients that modify the hormonal environment, with no alteration of body condition. In a similar vein, Peacock (1984) conjectured that the improved reproductive performance of goats fed on Acacia tortilis pods prior to mating could have been due to the high protein content of the pods and/or to their content in compounds with possible oestrogenic properties. It is worth noting that nutritional supplementation may also improve oestrus activity and ovulation rate in females indirectly through the so-called male effect. Walkden-Brown et al. (1993) reported higher oestrus responses and conception rates in Australian Cashmere goats exposed to bucks fed a high quality diet compared to bucks fed a low quality diet.

Secondly, abundant availability of good quality forage over prolonged periods (>4 weeks) before mating seems to have a detrimental effect on oestrus activity and ovulation. This longer term nutritional effect on reproductive function is likely to be mediated through improved body condition (Landau et al., 1996; Walkden-Brown et al., 1996). The contention that high levels of body condition may be responsible for reduced oestrus activity and fertilization rate in goats is difficult to corroborate from the literature. Landau et al. (1996) state that no negative effect of over-condition on reproductive performance appears to have been documented for goats, as has been done for sheep. Rhind et al. (1984), for example, found that excessively high levels of body condition of ewes at mating can have detrimental effects on both oestrus activity and ova or embryo survival. Adverse effects of high body condition on conception rates in sheep have also been reported by Gunn et al. (1991). In contrast, Henniawati and Fletcher (1986) showed that a supramaintenance level of feeding throughout the pre-mating and mating period had no effect on the incidence of oestrus in Indonesian sheep and goats, but significantly increased mean ovulation rate in both species, a fact reflected in subsequent lamb and kid production. In view of these inconsistent results it is difficult to rationalise the observed reduction in conception rates in does exposed to favourable forage conditions over a prolonged period prior to mating. More research is needed to strictly characterise the relationship between seasonal forage supply, body condition and conception rates in goats maintained on semi-arid tropical pastures.


Generally, the appearance and continuation of oestrus activity in goats has been reported to be less dependent on nutrition than does ovulation rate (Landau et al., 1996). Although the physiological mechanisms by which nutrition affects ovulation rate are not yet fully understood (Dunn & Moss, 1992; Mani et al., 1992), improved nutrition has been found to increase ovulation rate and the incidence of multiple births in goats. Besides the study of Henniawati and Fletcher (1986) noted above, Sachdeva et al. (1973) also hold the view [page 37↓]that a high plane of nutrition, particularly with respect to the energy content of the diet, increases the number of twin birth events. In the present study, the models fitted to prolificacy rates confirmed these findings in that litter size increased with increasing pasture quantity and quality during the pre-mating period, as expressed by the parameter estimates for levels 1 through 3 (score I) and levels 1 through 4 (score II) of the two range condition scores. However, the decline in litter size associated with levels 4 and 5 of range condition scores I and II, respectively, that were presumed to reflect maximum availability of forage of good quality, are inconsistent with the reports noted above. This tendency was also expressed in the estimated prolificacy rate of mating season group 2 (1.3 kids per birth), as opposed to that achieved in all other groups (>1.4 kids per birth). Likewise to conception rates, corroborative evidence for a depressing effect of excessive body condition or high planes of nutrition on ovulation rate, and/or embryonic and fetal survival, and hence the probability of multiple birth events in goats seems to be unavailable at this time.

The fact that fairly large litter sizes were observed for mating season groups 4 and 5 was also surprising given that mating in these groups coincided with the middle and end of the long dry season, respectively. A possible explanation for this pattern is that the consumption of high quality litter from Acacia species could have had a stimulating effect on ovulation rates. Also, the superior quality of different plant parts of Acacia species that are consumed by goats during the long dry season tends to upgrade and balance the deteriorating quality of other forages in the diet during this time (Schultka & Schwartz, 1987), and may have prevented excessive losses in body condition before mating. However, the point to be made here is that no clear picture has emerged yet as to the mechanisms by which changes in plane of nutrition, body condition score, and body weight affect ovulation rate in livestock (Dunn & Moss, 1992). Although body condition was not assessed in this study, the responses for mating season group 4 in particular seem to suggest that short-term changes in feeding conditions could exert a more direct influence on ovulation rate than body condition status at mating. Supportive evidence for a rise in ovarian activity during the dry season was reported by Hambolu and Ojo (1985, p. 281), who attributed the observed pattern to the higher temperature and the “ability of goats to convert any available leaves and dry forage into useful nutrients during the dry season”. On the other hand, Peacock (1984) did not observe higher twinning rates in goats fed on high quality Acacia litter prior to mating during July and August, which conflicts with the statement made above that there exists a short term nutritional effect on ovulation rate.

Apart from the above anomalies in responses, the present work confirmed the finding of Wilson et al. (1984) that litter sizes in SEA goats were greatest for parturitions taking place in both dry season (groups 1 and 4). With 1.44 kids per birth, however, average litter size was markedly larger than the 1.23 kids per birth recorded in the latter study. In contrast to other African breeds, such as Red Sokoto and West African Dwarf goats, which have been reported to produce 1.56 (Adu et al., 1979) and between 1.53 and 1.89 kids per birth (Bosman, 1995), respectively, triplet birth events seem to occur very rarely in SEA goats. Only one triplet litter out of 234 kiddings was observed in the present study. The overall probability of twin birth events of about 41 percent is comparable to the twinning rate of 30 percent estimated by Sacker and Trail (1966) for Mubende SEA goats in Uganda, whereas Ndlovu and Simela (1996) reported a twinning rate of only 3 percent for Mashona goats in Zimbabwe. Hence, the genetic potential for prolificacy of the SEA breed type appears to be inferior to that of other African breeds, particularly to that of many West African breeds. In any case, according to Bradford and Berger (1988) prolificacy in excess of twins is not desirable in goats dependent on grazing in arid lands, and the primary management goal should be the achievement of a mean prolificacy of 1.5-1.8 kids born per parturition. On average, the performance of SEA goats observed herein is close to the lower end of this range, while at third and later kiddings does can be expected to produce up to 1.7 kids per parturition.

The observed quadratic trend in litter size with increasing number of lactations is well established, and has been reported, among many others, by Wilson et al. (1984) for SEA goats, by Wilson (1984) and Wilson and Traoré (1988) for Sudan desert and Sahel goats, and by Adu et al. (1979) for Red Sokoto goats. Also, Osuagwuh (1991) found that the incidence of multiple births increased linearly with doe age. Given that the probability of a twin birth event increased from 14 percent at first to 69 percent at fourth kidding, doe parity appears to be one of the most important variables determining litter size in SEA goat herds. This emphasizes the importance of taking into account the relative abundance of does in different parity stages when making productivity assessments in goat herds. In comparative trials, performance criteria computed by averaging over the entire breeding female herd are likely to produce biased results and, therefore, invalid inferences, if relative parity stage abundances in the flocks considered are different, or, alternatively, if they are not in a stable state.

Other studies have reported an increase in litter size with liveweight of dams at mating, presumed to be indicative of body condition (e.g., Adu et al., 1979; Amoah et al., 1996). Similarly, changes in liveweight during the pre-mating period have been assumed to reflect changes in body condition. However, as has been pointed out by Devendra and Burns (1983), liveweight and age, and liveweight and parity number tend to be [page 38↓]confounded. Since the liveweight of breeding females is generally positively correlated with the number of parturitions or age, it is unclear to what extent an increase in multiple birth events can be attributed to physiological maturity, body weight or body condition at the time of mating. The statistical analysis performed in the present study did not reveal a significant effect of liveweight within parity class on prolificacy rates. Hence, liveweight at mating appears to be a poor predictor of litter size, after adjusting for the effect of parity. It is also doubtful whether this variable can be used as an indicator for body condition. Body weight is mainly determined by frame size, conformation, and relative fatness. It follows that a small body weight can be due to either a poor body condition or a small frame with adequate condition (Gunn & Moss, 1992). Moreover, the assumption that a change in body weight can be treated as an indicator for a concomitant change in body condition appears to be problematic, since, as has been demonstrated in the present study, body weight changes in does at all stages of the reproductive cycle are closely related to their parity stage (see Chapter 4). As such, it is suggested that parturition number is an important source of variation in reproductive responses in goats that should be controlled for experimentally.

Abortion and birth rate

Since embryonic and fetal mortality rates were not determined, the estimates obtained for birth rate were a function of (apparent) conception rate, abortion rate, and survival of pregnant does until parturition. The overall mean abortion rate of 5.9 percent was within the range reported in the literature on African goat breeds. Osuagwuh (1991) observed an incidence rate of 4 and 2.3 percent in 1.5 and 4 year old West African Dwarf goats, whereas that recorded by Traoré and Wilson (1988) in a Mali traditional system was much higher at 12.6 percent. Abortion rates were generally low (<3.5 percent) when moderate to good feeding conditions prevailed at mating and throughout the gestation period (mating season groups 1, 2 and 6), while a noticeable increase in incidences occurred with further displacement of mating into the long dry season (mating season groups 3, 4 and 5). At 8.7 and 15 percent of all pregnancies, the number of incidences in mating season groups 4 and 5 was particularly high. The majority of the abortion events occurred during late pregnancy which, in these two groups, coincided with the onset of the short and long rainy season, respectively. Hence, it is unlikely that abortion was linked to poor nutrition and starvation. At the beginning of the rainy seasons in semi-arid climates, animals change from a dry and limited diet to a green and abundant one, which is usually rich in nitrogen but poor in fibre. According to Lebbie et al. (1996), such a rapid modification of the ration can lead to an outbreak of enterotoxemia, caused by a sudden increase in the Clostridium perfrigens population in the intestine where the fast growing bacteria produce toxins which diffuse into the blood. This disease as well as other metabolic disturbances linked to nutritional disorders may have precipitated the incidence of abortions in groups 4 and 5.

Estimated doe mortality during the gestation period was of the order of 4 to 5 percent (mating season groups 2 and 3, respectively) when mid- and late pregnancy stages occurred during the long dry season, and less than 2 percent otherwise. The effects of high abortion rates and/or high doe mortality until parturition on weaning rates were most pronounced when does were joined towards the end of the long dry season (mating season group 5), in which case on average only 83 percent of the observed pregnancies were carried to term, while animals mated just prior to the long rains were least affected, with a birth rate of 98 percent in fertile does. In the other four groups, about 90 (group 2), 94 (groups 3 and 6), and 92 percent (group 4) of the pregnancies were estimated as having been carried to term.

The estimated birth rate for mating season 4 (77 percent) is in agreement with observations made by Peacock (1984) on two flocks of SEA goats mated during the long dry season and fed on high quality Acacia litter prior to mating, which achieved birth rates of 75 and 83.3 percent. It is noteworthy that the birth rate in goats mated at the same time, but which had no access to Acacia litter was as low as 13.3 percent. In both treatments, prenatal reproductive wastage in fertile does was limited to the occurrence of abortions, which were recorded to be 2.1 and 0 percent in the flocks fed on Acacia litter, and 6.7 percent in the other group. Likewise to conception rates, other studies, including the review of Wilson (1989) on reproductive performance of African indigenous small ruminants, conducted previously on African goat breeds kept under similar environmental conditions do not provide information on birth rates (e.g., Wilson et al., 1984; Wilson & Light, 1986; Wilson & Traoré, 1988; Ndlovu & Simela, 1996). However, overall birth rate in SEA goats appears to be high and is comparable to the 83.8 percent achieved after artificial insemination with follow-up mating by 20 months old Cashmere goats in a temperate environment in Australia (Ritar et al., 1994). Under arid conditions in northeast Mexico, a similar performance (82 percent) was recorded by Mellado et al. (1994) in crossbred goats in good body condition at mating, whereas poor body condition at mating resulted in a birth rate of only 46 percent. As reported by Mellado et al. (1996), the length of the breeding period appears to be a decisive factor in determining reproductive performance in goats kept under extensive range conditions. Birth rate was severely depressed when the mating period was less than 21 days. Indeed, the distribution of the duration between the onset of mating and parturition in fertile females observed in the [page 39↓]present study (Figure 2.3) underlines the importance of extending the breeding period beyond the average length of one oestrus cycle under controlled breeding. For instance, allowing non-cycling does to return to service was certainly an important factor in maintaining satisfactory conception rates in does mated at the end of the long dry season (mating season group 5).


In spite of the fact that differences in fecundity rates between mating season groups failed to reach statistical significance, it is nevertheless clear that mating at the start of the long rainy season by far produced the largest number of kids per doe exposed (1.46, mating season group 1). Controlled breeding under semi-arid conditions in northern Kenya clearly has a detrimental effect on fecundity when goats are joined in the middle of the long rainy season. The results obtained in this study indicate that restricting breeding to this period can be expected to lead to fecundity rates of less than unity. Estimates obtained for does joined between the months of June to December (groups 3 to 6) exhibited only small differences, with the exception of the depressed performance caused by an elevated incidence of abortions during pregnancies that were established at the end of the long dry season in October (group 5). Also, the patterns reflected by fecundity estimates across lagged median range condition scores at the time of mating have to be interpreted in relation to the change in range condition scores throughout the gestation period. Favourable forage conditions at mating, as observed in group 2, produced the lowest fecundity rates. Contrary to what might be expected, this apparently was due only in part to the fact that in this case late pregnancy occurred during a period of low quality and quantity of forage on offer. This gave rise to somewhat higher death rates in pregnant does. Although not yet documented in goats, the appearance and continuation of oestrus activity and/or ovulation and fertilization rates in this case may have been negatively affected by the development of excessive body condition before mating. Likewise, poor pre-mating range condition, as observed in group 5, was probably not responsible for the predicted depression in performance. Rather it was due to a drastic increase in prenatal wastage during late pregnancy, which seemed to be related to rapid changes in forage quality at the onset of the long rains in February-March.

Although at a substantially lower level, fecundity rates showed a trend with increasing parity similar to that reported previously by Adu et al. (1979) for Red Sokoto goats. The authors provided estimates up to the third kidding at which stage goats produced on average 2.0 kids per parturition; about 60 percent more than in the present study. Results of the present analysis suggest that this trend in fecundity rates can, to a large part, be attributed to the effect of physiological maturity on litter size. There is a paucity of performance data concerning fecundity rates in African goat breeds. Nonetheless, it can be concluded that reproduction does not seem to constitute an acute problem in the SEA goat type studied in this work. Moreover, the statement made by Peacock (1984) that the attempt of Maasai herders to restrict breeding to the months of July and August is an “artificial obstacle to reproduction” (p.361) does not seem to be unequivocally true. In the present experiment, the overall average fecundity rate, adjusted for parity, was 1.18 kids per does exposed. This is the average reproductive performance which one could expect to observe upon implementing an aseasonal breeding regime consisting of joining an equal number of does in each of the six consecutive mating seasons. By comparison, there was some evidence that restricting breeding to a period at the onset of the long rains is likely to result in a larger kid crop per doe exposed (1.46), and with 1.07 to 1.23 kids born per doe exposed even mating during the long dry season failed to exhibit the deleterious impact on reproductive performance advocated by Peacock (1984).

Weaning rate

The ranking of weaning rate for the six mating season groups differed markedly from that of fecundity rate due to large differences in postnatal kid survival. It should be stressed, however, that the reported estimates of weaning rates can be biased downwards, because no adjustment for censored observations was made in calculating this criterion. In principle, weaning rate is a function of fecundity and kid survival to weaning, the latter estimates being affected by kids lost to follow-up with no information about their survival status at the time of weaning. Although the number of censored observations was small, their effect on estimates of weaning rate are difficult to quantify and, hence, these results should be interpreted cautiously. With 1.11 kids per doe exposed, the largest litter size at weaning tended to be produced by goats mated at the peak of the long dry season (mating season group 4). By comparison, the aseasonal breeding regime described above could be expected to produce, on average, 0.91 kids at weaning per doe exposed, thus pointing to the fact that the restricted breeding management practised by Maasai herders in southern Kenya might indeed be beneficial.

Clearly, mating during the short dry season should be avoided (mating season group 1), since here large pre-weaning losses are likely to severely depress the number of kids weaned per doe exposed. In the latter group, [page 40↓]about half of the kids born died before weaning. This leads to the rather obvious conclusion that real improvements in goat flock performance cannot be achieved by concentrating efforts solely on increasing reproductive performance. Such improvements can only be obtained upon integrating the whole production process, for which kid survival is of paramount importance (Delgadillo & Malpaux, 1996). Kid survival not only affects the number of surplus animals available for offtake, but also the number of female youngstock available as replacements. Reduced availability of replacements imposes restrictions on selective culling of less productive animals, which is instrumental in achieving and maintaining the most productive demographic structure in the breeding herd.

2.5 Conclusions

The results of this study indicated that seasonality in the availability and quantity of forage on offer as it prevails throughout northern Kenya can have significant impacts on individual reproductive traits in SEA goats. The most pronounced influence, as captured by the subjective phenological pasture condition scores, was detected in relation to prolificacy and fecundity rates. The observance of the latter effect was a direct consequence of the fact that litter size is one of the determinants of fecundity as defined in this work. The mating season treatment effect, however, could not be shown to have a statistically significant effect on reproductive traits. Though the range of predicted values for traits such as conception and prolificacy rates among mating season groups were quite large, the inter-year variability in seasonal environmental conditions were too high to permit a clear expression of statistical differences between individual groups with respect to the traits considered. Nevertheless, it appears that joining does during the short dry season, just before the onset of the long rains, could produce the largest number of offspring per doe exposed. But the results with respect to weaning rate indicate that this superiority is entirely lost due to high mortality in young kids during the following long dry season. To summarize, controlled breeding is most likely to be an inefficient management tool to improve reproductive performance of pastoral goat herds, as long as no remedy is found for reducing kid mortality until weaning.

© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 3.0Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML generated: