Leadership Behavior in Relation to Dominance and
Reproductive Status in Gray Wolves, Canis lupus
Methods
Leadership data were recorded during ground observations made between September and March in 1997-1999. Behavioral evidence of leadership in gray wolves has not been systematically evaluated, so we selected five plausible behavioral indicators of leadership that were observed regularly in the field: scent-marking; frequency and time in the lead while packs were traveling or hunting (frontal leadership); initiation of pack behavior; and nonfrontal leadership. Scent-marking was included as a metric of leadership because it is an integral component of travel and dominance expression (Peters and Mech 1975). Observations were made using binoculars and spotting scopes up to 75× power from an average distance of 1 km. When packs traveled out of visual range, radiotelemetry was used to locate them and confirm the presence of breeding individuals.
Data were collected during September through mid-January and late January through March. Complete field seasons included early and late winter in 1997-1998 and 1998-1999, as well as early winter 1999. Included in both seasons were 30-day periods of daily tracking and observations beginning on 15 November and 1 March. All three packs observed in this study were frequently visible from a road that provided access for teams of two observers per pack.
All three packs were formed as wolf pairs initially reproduced in Yellowstone National Park in 1995 or 1996. Individual wolves in the three study packs were identified by the presence or absence of a radio collar and by distinctive pelage and other physical characteristics. The Rose Creek pack contained 14-23 wolves (mean 19.0) and the Leopold pack and Druid Peak pack included 8-13 (mean 10.4) and 7-8 (mean 7.8) wolves, respectively. The average numbers of pups and older subordinate nonbreeding wolves were 8.6 and 8.0 for the Rose Creek pack, 4.0 and 4.6 for the Leopold pack, and 3.0 and 2.6 for the Druid Peak pack, respectively. Pack size changed little during this study, but the number of nonbreeding yearlings and older wolves was generally greater during the second year of the study. The Leopold pack had a single breeding pair, the Rose Creek pack had two breeding females (mother and daughter ≥3 years apart in age), and the Druid Peak pack had two breeding female siblings, after a third breeding female sibling dispersed in November 1997. In the Druid Peak pack, one sister became dominant even though the other initially reproduced first. All breeding individuals were either radio-collared or easily distinguished from other wolves. Dominant individuals were identified from body postures such as raised tail and direct stare with ears forward, or because they elicited submissive responses from other wolves (Schenkel 1947, 1967). Dominant wolves were never observed exhibiting submission to other wolves. Breeding wolves, whether dominant or subordinate, were identified as such because they were seen copulating in winter or with newborn pups the following spring. Once a subordinate breeding female was identified in a pack, the pack was considered to have multiple breeders as long as she was present. Breeding females often shared whelping dens, so it was not always possible to determine individual reproductive success. No subordinate breeding males were detected during the study. However, because it was impossible to observe the wolves constantly, additional subordinate breeding males and females (if they did not whelp) could have been present without our knowledge.
Behaviors classified as scent-marking included raised-leg urination, scratching, and double scent-marking (Peters and Mech 1975). A double scent mark was recorded when two wolves in quick succession marked the same location.
During this study all three packs were highly cohesive. Individual subordinate wolves did separate (N = 9) temporarily or permanently from their pack, but packs rarely split into two or more groups. Data on frequency and time individual wolves led packs were collected only when the positions of both dominant breeding wolves in a pack were clear during the observation. A leadership bout was defined as a period when any individual wolf was identified at the head of the line during travel or pursuit of prey. A new bout began with a change in leading wolf or activity. Recorded for each bout were date, total duration, breeding and social status of the leading wolf (dominant breeder, subordinate breeder, or nonbreeder), pack identity, pack size, position of breeders, activity, and snow condition. Snow was classified by depth on the front legs of wolves: low or high if below or above the midpoint on the radius, respectively.
"Activity initiation" occurred when one wolf prompted the following: arousing the pack from rest, traveling after group rallies (greeting ceremonies), chasing prey, changing direction during travel, or defending the pack from trespassing wolves. "Nonfrontal leadership" occurred when a wolf not in the lead broke ranks and led the pack in a new direction or activity.
My guess is that this new army of environmentalists, although certainly welcome to the cause, will never produce a person who inspires quite the passion as Aldo Leopold. I make that guess because, as a hunter myself, I dare think I understand Leopold’s passion.
Statistical methods
The data were classified to year, pack, activity (travel or pursuit of prey), season (early or late winter), and identity of the wolf first in line (dominant male, dominant female, subordinate breeder, or nonbreeder). Observations of wolves pursuing prey were too sparse for analysis. The observations were organized into a three-dimensional contingency table, the dimensions being season, pack, and social status/sex categories.
The scent-marking data were relatively sparse, so for each season we conducted a χ² test of homogeneity (Gibbons 1997) across packs to see if data pooling across packs was feasible. We then conducted χ² tests of equal cell proportions (Gibbons 1997) to determine if scent-marking events were equally distributed across social status/sex categories (Martin and Bateson 1993). If significant differences were found we conducted one-tailed sign tests (Gibbons 1997) to evaluate differences in scent-marking rates between dominant males and dominant females. The null hypothesis was P = 0.50 and the alternative hypothesis was P > 0.50, where P denotes the proportion of scent-marking events performed by the dominant male. The postulated alternative hypothesis was based on higher expectations of leadership derived from earlier studies, primarily done on captive animals (i.e., male > female and dominant > subordinate) (Fox 1980; Zimen 1981).
For analyses of frequency and time at the head of the line, data were used only from bouts during which the positions of both dominant breeding wolves were known. For the frequency data we fit a three-dimensional log-linear model, the effects being pack, season, and social class/sex. The pack and season main effects were not of interest, as they arose from differences in numbers of observations across packs and seasons. The social status/sex effect was of primary interest, as well as the two-way interactions pack × social class, pack × season, and season × social class. If these interactions were significant, pooling of data across packs and (or) seasons was precluded, and a significant interaction of social class with pack and (or) season indicated that behaviors varied across packs and (or) seasons. We fit the loglinear model containing all main effects and all two-way interactions and conducted tests of significance. In some cases there were no subordinate breeding females and this created structural zeroes (Agresti 1990) in the frequency table, which we incorporated into the analysis.
The log-linear model analysis tests for equal distribution of leadership across social classes, but there were typically several nonbreeders and in at least one case more than one subordinate breeding female. For each pack and season we tested for equal distribution of leadership across social classes while accounting for the number of wolves in each category. We computed the expected leadership frequency based upon the null hypothesis of equal leadership distribution across wolves. The expected frequencies (E) were computed as follows: E = (number of wolves in category/total number of wolves in categories being compared) × total number of bouts for the categories being compared. The observed and expected frequencies were then compared using a χ² test. Note that the nature of the scent-marking data did not require this adjustment.
For each pack and season we first tested for equal distribution of leadership across wolves in all social classes. For nonbreeding wolves we included yearlings and adults but excluded pups (in this study pups led 3% of travel bouts).
Including pups could have biased the comparison between breeders and nonbreeders, owing to the age discrepancy. If we found a significant difference across classes, we proceeded to compare the dominant breeding male and dominant breeding female. There is only one dominant breeding male and one dominant breeding female in each pack, so their leadership rates were compared using a one-tailed sign test. In those tests we used only data from those two animals. We also compared frequency of leadership between the dominant breeding female and subordinate female breeders, accounting for the number of subordinate breeders. Before testing for differences between breeders and nonbreeders we compared subordinate breeding females with nonbreeders. In addition to frequency of leadership, proportion of time spent at the head of the line was calculated, to compare frequency of leadership between social classes and sexes of wolves.
For each set of hypothesis tests we controlled the overall level of significance at 0.05 using the Bonferroni procedure. Data for activity initiation and nonfrontal leadership were too sparse across packs and seasons for hypothesis testing, so we report only frequency of occurrence.
