RESULTS & DISCUSSION
Results
The two-dimensional discriminant analysis had an explained variance of 81.82% (Can1) and 18.18% (Can2) (Figure 6). The largest positive contributions to discriminant factor 1 include “indifferent”, “relaxed”, and “happy” behaviours and eye-area temperature, while the largest negative contributions were “distressed” behaviours and reactivity to the novel object. Discriminant factor 2 had most positive contributions from “uneasy” behaviours and negative contributions from “relaxed”, “indifferent”, “active”, and “distressed” behaviours.
​
From visual inspection of the graph, there is clear separation between the positive treatment group from both the neutral and negative treatment groups. This may indicate that after being in the positive pen, the cattle exhibited more signs of being more indifferent, relaxed, and happy, and less distressed. The positive animals also had higher eye area temperatures. Future analysis will confirm which of these variables were significantly different between treatment groups.
​
Following the discriminate analysis with a MANOVA confirmed there were differences between treatment means (P=0.01). This suggests that qualitative behaviour assessments, eye area temperature, and reactivity scores during novel object tests may be sensitive indicators of positive affect.
Figure 6. Discriminant analysis results. Vectors comprised of qualitative behaviour assessment data, eye area temperature data, reactivity data, and vocalization data. Data points represent individual animals, colour coded by treatment group.
Reactivity_novel object
erature
Legend
o Negative Treatment - Cow o Neutral Treatment - Cow o Positive Treatment - Cow
After running a MANOVA on the multivariate model with Wilks' Lambda and using the Tukey adjustment to maintain the experiment wise alpha level of 0.05, the univariate ANOVA tables for each variable were obtained using the 'summary.aov()' function. This analysis revealed it was Eye Area Temperature (P=0.003) and Indifferent behaviours (P=0.02) that were primary drivers of the separation between treatment groups.
Further (univariate) analysis of the Eye Area Temperature revealed that when the animals were in the positive treatment, they had a higher Eye Area Temperature (P<0.001) than when they were in the neutral treatment based on a non-inferiority test with 90% confidence. In contrast, animals from the negative treatment, did not have a higher Eye Area Temperature (P<0.001) based on a non-superiority test with 90% confidence (Figure 7). Unfortunately, these differences were only detected with a delta of 1 degree or less but when effect size effects were calculated, it was positive treatment animals that were always highest (P=0.001) with a confidence of 90%. Based on these results, Eye Area Temperature could be worth monitoring within a herd to assess the happiness of the cows. If regular monitoring occurred, each animal would have a baseline and their individual departures from their baseline could be measured; only if the technology available would be sensitive/accurate enough to detect changes of 1 degree overtime and considering the affordability of the technology, this may only be feasible for research, and not yet on-farm.
Figure 7. Comparison of eye area temperature between treatments. Dashed line through neutral treatment is reference value for subsequent effect size calculations.
Additionally, when characterizing behaviours expressed by the animal, the negative (P=0.04) and positive treatment animals (P=0.001) were more indifferent than the animals from the neutral treatment based on a non-inferiority test with 90% confidence and a delta of 5 points (Figure 8). These results do not help with on farm monitoring of animal welfare since both the positive and negative animals were more indifferent than the neutral animals. If observing animal behaviour and seeing indifferent reactions to novel stimuli (objects or situations), this may be more indicative of arousal (high/low), and not valence (positive/negative). However, single animal behaviours are rarely an area of interest. It is common to use broader groupings of behaviours to characterize animal behaviours.
Figure 8. Comparison of displays of indifferent behaviours being observed. Dashed line through neutral treatment is reference value for subsequent effect size calculations.
Therefore, I have used the CAN1 results from the LDA (figure 6) to create a combined "happiness" value that represents the positive behaviour descriptors (indifferent, happy, calm, content, relaxed) which revealed that the positive treatment displayed more positive behaviours than the neutral treatment (P=0.02) while the negative treatment did not (P=0.51) based on an equivalence test with a confidence level of 90% and a delta of 0 (figure 9).
These results indicate monitoring dairy herds for calm, content, relaxed, indifferent, and happy behaviours of animals would be useful assessing the animal's welfare. This monitoring could be done by herdspersons daily, herd veterinarians when on site, or by video cameras eventually (as research in the remote monitoring systems progresses).
Figure 9. Comparison of displays of indifferent behaviours being observed. Dashed line through neutral treatment is reference value for subsequent effect size calculations.