The behaviour of young cattle was studied in a series of learning experiments and in two observational studies. In a first experiment, it was shown that six yearling heifers could discriminate between familiar herd members as stimuli in a discrimination learning task in a Y-maze. As part of the experimental protocol, the heifers stood in a start area overlooking the two Y-maze arms for about one minute before they were allowed to enter a maze arm. Their behaviour during this time period was studied on the basis of video records. The orientation of their heads predicted the direction of the maze arm that they chose when they made a correct choice. When they made an incorrect choice, both body and head orientation tended to predict the maze arm that would have been correct. In a second experiment, the emotional reactions of heifers during acquisition of an operant task were studied. Six heifers learned to press a panel, while alone in a start area visually isolated from the rest of the herd. Pressing the panel made a gate open, which allowed them to exit the start area and gave access to a race, at the end of which the heifers received a food reward. Six control heifers were yoked to the experimentals, so they received the same treatment, but contingent on their matched partner's - not on their own - behaviour. The heart rate of heifers, and their behaviour when moving down the race were measured. When experimental heifers made marked learning improvements, they were more likely than on other occasions to have higher heart rates just before the gate opened, and to move more vigorously down the race in comparison with their controls.

No description available.

No description available.

A modelling framework is developed which transfers the uncertainty inherent in estimating the size of parr populations from electrofishing data into sub-models which predict the wealth of salmon biomass and age structure of such populations. The focus of much of this thesis is on data collected on the River Conon and we find here that such a model is able to account for 36.5% (posterior median, 95% credible interval (2.8%, 58.4%)) of the observed variation in the habitat “score” – a proxy measure for the estimated size of a population. Having looked at production, the next stage involves examining migrations of smolts within freshwater. Specifically, a methodology for using RJMCMC methods for model choice when considering a range of models for dealing with capture-recapture data with subsequent captures at a smolt trap, is developed. The smolt trap allows for estimates of a migration propensity for salmon parr by recasting migrations as recoveries of dead animals within the traditional capture-recapture-recovery framework. As a by-product, estimates for over-winter, over-spring and in-migration survival probabilities based on model-averaged posterior distributions can be achieved. This methodology is then extended to study the feasibility of incorporating highly variable time-dependent covariate histories into estimates of survival and migration. Despite the presence of a low capture probability, the data appears to hold sufficient information as to be able to identify a strong relation between migration propensity and fish length, which also appears to be positively correlated with in-migration survival.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.