Abstract: For centuries, scientists have attempted to identify and document analytical laws that underlie physical phenomena in nature. Despite the prevalence of computing power, the process of finding natural laws and their corresponding equations has resisted automation. A key challenge to finding analytic relations automatically is defining algorithmically what makes a correlation in observed data important and insightful. We propose a principle for the identification of nontriviality. We demonstrated this approach by automatically searching motion-tracking data captured from various physical systems, ranging from simple harmonic oscillators to chaotic double-pendula. Without any prior knowledge about physics, kinematics, or geometry, the algorithm discovered Hamiltonians, Lagrangians, and other laws of geometric and momentum conservation. The discovery rate accelerated as laws found for simpler systems were used to bootstrap explanations for more complex systems, gradually uncovering the “alphabet” used to describe those systems.

Abstract: For free-living or feral horses weaning takes place naturally at around 8-9 months [Gill, E.L., 1988. Factors affecting body condition of New Forest Ponies. Ph.D. Thesis. Department of Biology, University of Southampton]. Some mares will continue to suckle their foal until shortly before the arrival of their next foal, gestation being approximately 342 days depending upon the breed of the horse [Ropiha, R.T., Mathews, G., Butterfield, R.M., 1969. The duration of pregnancy in Thoroughbred mares. Vet. Rec. 84, 552-555]. Under domestic conditions, weaning tends to take place earlier, typically between 4 and 6 months of age. The weaning process has been identified as associated with potential psychological, physical and nutritional stressors that are of welfare concern. Following a review of the literature it is evident that there is a need for detailed research into what should constitute best practice with respect to foal and mare welfare. In addition, there is a need to understand the potential long-term impact of weaning on, for example, trainability and later maternal behaviour, and whether the stresses associated with early weaning have detrimental effects on the performance horse. There is also a lack of clear information concerning the most frequently observed weaning practices and the reasons why certain weaning methods are chosen. Some variables should be closely managed during weaning in order to minimise stress responses. These include: early creep feeding to familiarise the young animal with the food it will be exposed to during weaning, feeding a high fibre diet and keeping the animal in extensive conditions using a gradual approach to weaning. However, we conclude that there may not be one best method for weaning, since the chosen method must take into account a number of factors including: available resources, the housing environment, the individual foal's stage of development, the strength of the mare-foal attachment, the foal's ability to cope with changes in social conditions and the ability of the horse owner to implement the chosen method. We do however suggest that the fewest stress responses appear to occur where foals are weaned gradually and allowed to have social contact either with other foals or with older horses.

Abstract: For some time now, social network analysis provides tools to describe concisely the social structure of animals. Nevertheless, the factors that shape the social network and determine the frequency of different types of affiliations are often unclear. For example, the effects of relatedness on social networks have been only studied on relatively few species. Both among social or sub-social species the effects ranged from no effect to a substantial effect. Here we test the hypothesis that relatedness has an effect on the social network of horses (Equus caballus) that live freely in semi-natural conditions, and specifically, that individuals with high values of relatedness to their neighbors in the network will have fewer links Wolf et al, 2011). This hypothesis is based on the notion that related individuals have common interests and are likely to assist each other, and may need less links to other, unrelated individuals. In addition, individuals of similar age and sex are likely to have common needs, and thus are more likely to be associated. We thus tested a second hypothesis: homophily to individuals with similar age and sex will have a role in determining the associations within the social network. The field study was conducted on 27 horses in the Blauwe Kamer reserve in the Netherlands (1.1 sq km). We videotaped horses and their groups and used the information from 22 horses, after excluding the five foals from the analysis. Relatedness was calculated from the pedigree, which was based on parentage, determined by DNA analysis. The social network was constructed based on spatial proximity data. We assessed the influence of relatedness, age-homophily and sex-homophily on the network structure with Multiple Regression Quadratic Assignment Procedure (MRQAP) (Krackhardt 1988), with the R package sna. The results show that there was no significant effect of relatedness on the network, nor an effect of age-homophily. Nevertheless, we found a significant effect of sex-homophily, the tendency of individuals to associate with individuals of the same sex. We argue that the lack of a relatedness effect is not likely to have been caused due to the inability to detect who is kin. The structured social system in horses includes strong associations between often unrelated individuals, alongside with young individuals leaving their natal harem as part of the inherent inbreeding avoidance (Linklater & Cameron 2009; Boyd et al. in press). The significant effect of sex-homophily could stem from the protection females get from associations with other females, e.g., in the reduction of harassment from males. Previous studies on feral horses showed that mares that were better connected with other females in their harem benefited from higher survival rates to their foals (Cameron et al. 2009). The associations among bachelor males could also contribute to the strength of the sex-homophily effect.
In order to generalize from our results, one needs to examine additional populations of horses, because the conditions in the Blauwe Kamer reserve may not be representative, mainly due to the limited opportunity for dispersal in a restricted area.
Keywords: Long-term affiliation; spatial proximity, kin detection
Boyd, L., Scorolli, A. L., Nowzari, H., & Bouskila, A. (2016). Chapter 2: Social organization. In J. I. Ransom, & P. Kaczensky (Eds.), Wild equids. Baltimore, Maryland (in press): The Johns Hopkins University Press.
Cameron, E. Z., Setsaas, T. H., & Linklater, W. L. (2009). Social bonds between unrelated females increase reproductive success in feral horses. Proceedings of the National Academy of Sciences, 106(33), 13850-13853
Krackhardt, D. (1988). Predicting with networks: Nonparametric multiple regression analysis of dyadic data. Social Networks, 10(4), 359-381.
Linklater, W. L., & Cameron, E. Z. (2009). Social dispersal but with philopatry reveals incest avoidance in a polygynous ungulate. Animal Behaviour, 77(5), 1085-1093.
Wolf, J. B., Traulsen, A., & James, R. (2011). Exploring the link between genetic relatedness r and social contact structure k in animal social networks. The American Naturalist, 177(1), 135-142.

Abstract: Forty Danish warmblood colts in two replicates were used to investigate the effect of housing and handling in the rearing period on the reactions to humans. The horses entered the experiment after weaning and were housed either individually (n=16) or in groups of three (n=24). Half of the horses from each housing group were handled three times per week for a period of 10 min. Approach tests were performed in the home environment when the horses were 6, 9, 12, 18, 21, and 24 months old, and an Arena and Human Encounter test was performed in a novel environment when the horses were 12 and 24 months old, respectively. In the home environment, single-housed horses approached sooner and were more easily approached by a human than group-housed horses where no effect of handling was observed. Horses approached sooner and were more easily approached with increasing age. In the Arena and Human Encounter test, single-housed horses expressed less restless behaviour, more explorative behaviour, and less vocalisation than group-housed horses. Handled horses showed lower increase in heart rate during the test than non-handled horses. There was no difference between the number of times single or group-housed horses touched an unfamiliar person in the Arena and Human Encounter test but handled horses approached sooner than non-handled horses. It is concluded that the social environment affected the way horses reacted to humans when tested in the home environment but not in a novel environment. In contrast, handling affected the reactions to humans when tested in the novel environment but not in the home environment. However, handled horses also reacted less to the novel environment in general, thus indicating that handling is a mean of avoiding potential dangerous situations.

Abstract: Forty-four horses were used to determine if a learned stimulus could be used to increase acquisition of a response unrelated to the original learned task. Horses were paired by age, breed and sex. One randomly chosen horse from each pair served as the control while the remaining horse was trained to respond to a visual stimulus by pushing a lever to obtain a food reward. In Experiment I (n=28), the food reward was delivered in a feed box located equidistant between two levers located 2.7 m apart on one wall of the test stall. Trained horses were given 30 training trials daily until they achieved at least 85% correct responses in three nonconsecutive days. After each trained horse reached criterion, both horses in the pair were tested in 30 trials daily for five consecutive days in a modified Y maze. The stimulus that indicated the correct lever in the initial training task was used to signify the correct arm of the maze for both the control and trained horses. The correct arm of the maze was changed randomly during each daily session, and correct choices resulted in a food reward. In Experiment II (n=16), levers were located on opposite walls (11 m apart) of the test stall, and the stimulus and modified Y maze were simplified. Trained horses received 15 training trials daily until they reached at least 85% correct responses on three consecutive days. As each trained horse achieved this criterion, both horses in the pair were tested in a maze for 15 trials daily for 7 days. Again, the stimulus from initial lever-pressing task was used to signify the rewarded arm of the maze. Correct choices of trained and control horses in both experiments were compared using a paired t-test. In Experiment I, no differences (P>0.34) in mean correct responses were detected between trained and control horses on the first day in the maze (15.3 and 16.8, respectively) or after 5 days in the maze (84.0 and 82.1, respectively). Similarly, in Experiment II, trained and control horses did not differ (P>0.50) in mean correct responses on the first day (6.9 and 7.4, respectively) or after 7 days in the maze (63.6 and 61.6, respectively). These results indicate that the stimuli used to solve the lever-pressing task were not utilized by horses in the maze task, and that horses may have difficulty transferring learned visual stimuli to new tasks.