Abstract: Handling is a crucial component of the human-horse relationship. Here, we report data from an experiment conducted to assess and compare the effect of two training methods. Two groups of six Welsh mares were trained during four sessions of 50 min, one handled with traditional exercises (halter leading, grooming/brushing, lifting feet, lunging and pseudo-saddling (using only girth and saddle pad) and the second group with natural horsemanship exercises (desensitization, yielding to body pressure, lunging and free-lunging). Emotional reactivity (ER) and the human-horse relationship (HHR) were assessed both prior to and following handling. A social isolation test, a neophobia test and a bridge test were used to assess ER. HHR was assessed through test of spontaneous approach to, and forced approach by, an unknown human. Horses' ER decreased after both types of handling as indicated by decreases in the occurrence of whinnying during stressful situations. Head movement (jerk/shake) was the most sensitive variable to handling type. In the spontaneous approach tests, horses in the traditional handling group showed higher latencies to approach a motionless person after handling than did the natural horsemanship group. Our study suggests that natural horsemanship exercises could be more efficient than traditional exercises for improving horses' HHR.

Abstract: Group-living animals have to make trade-offs to reach consensus and travel together. We investigated the recruitment processes underpinning decision-making at departure in a group of 20 female domestic geese (Anser domesticus) kept in semi-free-range conditions. Two observers continuously videotaped the behaviours of the birds. Data were analyzed using multiple regression analyses. We found that decision-making was a continuous and distributed process. Departure was preceded by an increase in the arousal state of group members and their initial orientation influenced recruitment. Patterns of group movement could be predicted from the behaviours of individuals before departure. Individuals' locations, moves and signals could act as passive or communicative cues. A higher number of vocalisations and arousal behaviours led to a larger number of individuals recruited. Some individuals were more efficient than others in recruiting followers but any geese could initiate a movement. First movers recruited a higher number of mates when they had a greater number of neighbours. Not only the first mover but also the behaviours of the second and third movers prompted further individuals to follow. There was no evidence that geese were able to intentionally recruit others, rather they synchronized and adjusted each other's motives until reaching a consensus.

Abstract: Abstract  SOCPROG is a set of programs which analyses data on animal associations. Data usually come from observations of the social behaviour of individually identifiable animals. Associations among animals, sampling periods, restrictions on the data and association indices can be defined very flexibly. SOCPROG can analyse data sets including 1,000 or more individuals. Association matrices are displayed using sociograms, principal coordinates analysis, multidimensional scaling and cluster analyses. Permutation tests, Mantel and related tests and matrix correlation methods examine hypotheses about preferred associations among individuals and classes of individual. Weighted network statistics are calculated and can be tested against null hypotheses. Temporal analyses include displays of lagged association rates (rates of reassociation following an association). Models can be fitted to lagged association rates. Multiple association measures, including measures produced by other programs such as genetic or range use data, may be analysed using Mantel tests and principal components analysis. SOCPROG also performs mark-recapture population analyses and movement analyses. SOCPROG is written in the programming language MATLAB and may be downloaded free from the World Wide Web.

Abstract: Longitudinal behavioral data generally contains a significant amount of structure. In this work, we identify the structure inherent in daily behavior with models that can accurately analyze, predict, and cluster multimodal data from individuals and communities within the social network of a population. We represent this behavioral structure by the principal components of the complete behavioral dataset, a set of characteristic vectors we have termed eigenbehaviors. In our model, an individual’s behavior over a specific day can be approximated by a weighted sum of his or her primary eigenbehaviors. When these weights are calculated halfway through a day, they can be used to predict the day’s remaining behaviors with 79% accuracy for our test subjects. Additionally, we demonstrate the potential for this dimensionality reduction technique to infer community affiliations within the subjects’ social network by clustering individuals into a behavior space spanned by a set of their aggregate eigenbehaviors. These behavior spaces make it possible to determine the behavioral similarity between both individuals and groups, enabling 96% classification accuracy of community affiliations within the population-level social network. Additionally, the distance between individuals in the behavior space can be used as an estimate for relational ties such as friendship, suggesting strong behavioral homophily amongst the subjects. This approach capitalizes on the large amount of rich data previously captured during the Reality Mining study from mobile phones continuously logging location, proximate phones, and communication of 100 subjects at MIT over the course of 9 months. As wearable sensors continue to generate these types of rich, longitudinal datasets, dimensionality reduction techniques such as eigenbehaviors will play an increasingly important role in behavioral research.

Abstract: Researchers are increasingly turning to network theory to understand the social nature of animal populations. We present a computational framework that is the first step in a series of works that will allow us to develop a quantitative methodology of social network sampling to aid ecologists in their social network data collection. To develop our methodology, we need to be able to generate networks from which to sample. Ideally, we need to perform a systematic study of sampling protocols on different known network structures, as network structure might affect the robustness of any particular sampling methodology. Thus, we present a computational tool for generating network structures that have user-defined distributions for network properties and for key measures of interest to ecologists. The user defines the values of these measures and the tool will generate appropriate network randomizations with those properties. This tool will be used as a framework for developing a sampling methodology, although we do not present a full methodology here. We describe the method used by the tool, demonstrate its effectiveness, and discuss how the tool can now be utilized. We provide a proof-of-concept example (using the assortativity measure) of how such networks can be used, along with a simulated egocentric sampling regime, to test the level of equivalence of the sampled network to the actual network.