Calves born on most dairy farms are separated from their dams and reared artificially (1). Disease is a major concern in these calves, with studies suggesting that 9–48% of artificially reared calves have bovine respiratory disease and 19–46% have diarrhea in the first 9 weeks of life (2, 3). Recently, targets have been set for reductions in both calf mortality and antibiotic use in the first 6 months of life (4). The key to a reduction in the mortality, welfare, financial and production costs of these diseases is early diagnosis. Thus, early diagnosis of disease is an area of considerable research. The use of behavior- based disease detection tools for calves has been increasingly researched (5). It is becoming more common to keep artificially-reared calves in group housing, with the provision of milk through automatic milk feeders. This provides us with the opportunity to record individual calf feeding behaviors. Similarly, the use of animal-mounted sensors is increasing globally. These technologies can aid with disease detection, for example tri-axial accelerometers can be utilized to record activity variables such as lying and standing time (6). Daily lying time and mean lying bout length have been shown to increase in calves with bovine respiratory disease (7), and a reduction in milk consumption has also been seen (8). Calves with neonatal calf diarrhea have been shown to increase lying bout length and decrease lying bout frequency in the days prior to disease development (9) in addition to drinking less milk, and drinking more slowly than their healthy counterparts (10). Recent work has utilized calf behavior to predict both neonatal calf diarrhea (11) and bovine respiratory disease and these models may include calf factors such as live-weight (12).

While there is a considerable body of literature analyzing the effects of group housing, nutrition, and disease on calf behavior in the pre-weaned period, there is very little in the published literature regarding the effects of calf factors such as age, live-weight, and breed on behavior in the pre-weaned period. Many previous studies have been designed to control factors such as age (13), sex (7, 8) and breed (6), frequently by only using one breed or sex. Thus, there is very little information to support our assumptions that these intrinsic calf factors require controlling for or as to which factors should be used in predictive models. For instance, age of the calf has been shown to influence behavior in the limited amount of literature available. Calves had more lying bouts and less time at the feeder at 40 days of age than at 26 days of age (14). Standing time has been shown to increase by 0.52 min per day of age (15). Age has also been shown to influence behavior during transport of calves < 10 day old (16).

While there is literature available on the effect of breed and sex on, for example, flight speed, behavior in the handling chute (17), avoidance behaviors (18) and curiosity (19), there is no research into the effects of calf breed on lying and feeding behaviors in the pre-weaned period. One previous study has analyzed differences in activity variables between sexes with female calves spending more time walking than their male counterparts but no differences in lying or standing time (20). Another factor that affects calf behavior is the social environment. Calves that were group-housed from 3 days of age had more social interactions and performed more play behaviors than those grouped from 7 or 14 days of age (21). However, there was no information given on activity and feeding behaviors.

A variety of environmental factors may also influence calves, including heat stress, with calves in a non-shaded environment changing posture more frequently than those in the shade (22). However, to the authors' knowledge, there is no published information on whether the behavior of pre-weaned calves is affected by season in the UK climate. One study did show that calves were more likely to seek refuge at higher induced wind speeds (23), but there was no work looking at transitions between lying and standing or time spent lying or standing.

Maternal factors may also influence the calf. The pre-natal environment may influence factors such as birthweight, dystocia, immune function, and calf survival (24). Calves born to dams suffering from metabolic stress in late lactation have lower birthweights and an increased basal inflammatory response compared to calves born to dams experiencing a lower level of metabolic pressure (25). Another study found increased inflammation and metabolic stress in calves born to cows with a low body condition score 24 days prior to calving (26). Heifers born to primiparous mothers were found to be at higher risk of both mortality and culling prior to first calving than those born to multiparous mothers (27). Calf lying behavior may be influenced by maternal management during the dry period with calves born to cows kept at pasture during the dry period having longer lying periods than calves born to cows kept indoors but exercised daily (28). Calves born to dams that suffered heat stress in late lactation have altered standing variables when compared to calves born to cows that did not suffer from heat stress in late lactation (29). Furthermore, these effects may be long lasting, as heat stress in late gestation has been found to effect the milk production of both the heifer born from that gestation and the production of the heifer's daughter (30). Disease during the lactation where the heifer was conceived is associated with an increase in age at first calving and a reduction in second lactation fat yield (31).

As detailed above, the evidence suggests that several calf, dam, and environmental characteristics can affect behavior. As changes in behavior are being used in studies that aim to detect disease, we need to determine what factors are associated with key behaviors and what the size of these associations are, so that they may be accounted for in future study designs. To identify calf factors that may need to be controlled for in comparison studies or included in predictive models, an understanding of how these factors are associated with feeding and activity variables in a healthy pre-weaned calf is needed. This study aims to address this gap by analyzing the association of intrinsic calf factors with behavior in healthy pre-weaned artificially reared calves.

For the thirty-one calves included in the analysis, the mean age was 27.7 days (range 11–39 days) and the mean live-weight 56.5 kg (range 41.9–79.9 kg). For further descriptive statistics of the intrinsic factors, please see, Supplementary Table 3. Descriptive statistics for each feeding and activity variable are available in Supplementary Table 3.

Univariable analysis identified thirteen out of fourteen behaviors as suitable to move forward to the model building stage (Table 1). The factors tested were sire breed type, season of birth, sex, birthweight, age at inclusion into the group pen, age, live-weight, and dam lactation number. All factors tested were taken forward to the model building stage for at least two of the thirteen behaviors that underwent multivariable analysis. The multinomial analysis of total time at milk feeder yielded no variables with a p < 0.2 at the screening stage and therefore this analysis was not taken forward to the multivariable stage. The forward model building for daily standing time yielded only a univariable result at p = 0.066 for season of birth (Wald = 6.06, F = 3.03. ndf = 2, ddf = 26) as the multivariable model did not converge, therefore this model was disregarded.

This study found that live-weight, age, sex, and season of birth influenced activity and feeding behavior variables in pre-weaned artificially weaned calves. Birthweight, sire breed type and age at inclusion into the group pen did have significant associations but with fewer behaviors. Dam lactation had no association with behavior in pre-weaned calves.

The association of live-weight with healthy calf behavior has not previously been explored. However, some studies have used live-weight to pair diseased and healthy calves e.g., Duthie et al. (7) and included in disease prediction models e.g., Chisholm et al. (46). The current study found that heavier calves visited the milk machine more frequently (Figure 1A), for a shorter visit length (Figure 1B), drank faster (Figure 1C) and were slightly more likely to drink their milk allowance (Table 3) than their lighter counterparts. This concurs with a recent study that suggested a fast-drinking personality type is associated with increased live-weight gain (47). Another possible explanation is that due to the feed allowance being the same for all calves the larger calves were receiving a lower % of their bodyweight in milk and this was influencing their behavior. A previous study found time drinking was reduced in calves on a lower level of milk replacer (48). Calves fed less milk have been shown to visit the milk feeder more often (49).

Unlike in previous work, no increase in either the number of lying bouts (14) or daily standing time (15) with age was seen in this study. This study did, however, find that older calves had a lower total daily MI (Figure 2C), mean MI per standing bout (Figure 2B) and standing bout length (Figure 2A). This may reflect calves playing less as they get older. Calves have previously been found to have a reduction in play between 3 and 6 weeks of age (50). Another possible explanation is that stabilization of the group composition means that less movement is required. It may also reflect reduced exploration as calves become more familiar with their environment, for example, weaned calves visit different areas of the pasture more often when first introduced into a pasture but this behavior reduces over time (51).

Only one previous study has examined the association of birthweight with milk drinking behaviors and activity variables. It found that calves born at a normal (42.7 ± 2.6 kg) birthweight spent more time drinking milk and had a greater daily standing time than their lighter (34.9 ± 2.4 kg) counterparts (52). In contrast the current study found that calves that were born heavier had fewer lying (Figure 3B) and standing (Supplementary Figure 1) bouts each day, a longer mean standing bout length (Figure 3C) and drank less milk per feed (Figure 5B). However, in the current study the range of birthweights was greater (32–52 kg). The reduction in the number of lying and standing bouts was an interesting and unexpected finding, especially as birthweight is positively associated with play behavior in piglets (53). All the calves included in this analysis had an unassisted birth, so this association is unlikely to be due to dystocia. It is worth noting that birthweight in this study was influenced by both sex (p = 0.050, Supplementary Table 5) and dam lactation (first lactation-second lactation p = 0.013, first lactation- third lactation and over p = 0.002, Supplementary Table 5). In turn, birthweight had a significant effect on live-weight in a mixed model (p < 0.001, Supplementary Table 5). Interestingly a recent study found that as well as starting at a smaller size, calves with a smaller birthweight also grow more slowly (52). A larger study size may allow examination of interactions between the factors examined in this study, but this was not possible with only thirty-one calves.

Picking apart the associations between age, live-weight and birthweight and behavior is challenging. Live-weight is highly correlated with both age and birthweight (Supplementary Table 2). Interestingly, in Table 1 we can see that at the univariable level behaviors were only significantly (p < 0.05) associated with one or two of these three factors, this was unexpected due to the high level of correlation. The pattern seen in this study where when both age and weight had a p < 0.1 at the univariable stage, age produced a better fitting multivariable model for activity variables and weight produced a better fitting multivariable model for milk drinking behaviors was interesting. It is not however possible to pick out which of age and live-weight is more important.

Other studies have also recognized the importance of live-weight and age. For instance, both variables were previously included in random forest models for predicting bovine respiratory disease (54). It is worth noting that currently the inclusion of live-weight in predictive models for disease will require the automatic milk feeders to have a weighing platform, this is likely to require additional investment. Live-weight data also required significant data cleaning due to calves only partially standing on the scale or a second calf standing behind them. Thus, to use live-weight from a weighing platform in automatic disease detection tools automated cleaning of raw live-weight data prior to the disease detection model would be required or feeder design would need to be altered to ensure only a single whole calf could enter at any one time. Existing auto-weighers on the market already incorporate this data processing. An alternative would be to use a computer vision based system such as that used by Cominotte et al. (55) to estimate body weight. Using birthweight may be simpler and require less investment than live-weight. Thus, birthweight warrants further investigation as a possible alternative to live-weight for use in disease detection tools, possibly in conjunction with age.

In this study calves born in summer had a longer daily lying time (Figure 3A), more lying and standing bouts/day (Figure 3B and Supplementary Figure 1), and a shorter mean standing bout length (Figure 3C) than calves born in autumn or winter. It is highly likely that these associations are environmental in nature. The changes seen in the number of lying and standing bouts is consistent with the previous findings of differences in the number of posture changes in different levels of shade (22). The same study found no effect of shade on lying time however which does not concur with the finding that summer born calves had a longer daily lying time (Table 4) in the current study. The temperature and humidity index and the use of a ventilation system has been shown to influence the behaviors of 15 month old bulls (56), although the current study was conducted at far lower ambient temperatures (−5.5–28.2°C in current study compared to 20–30°C in bull study). Calves have been previously found to be aversive to wind (23) which may be reflected in the change in behavior in the autumn and winter. However, the association seen here is the opposite to what the authors would have expected. The associations with season in this study may not be completely due the conditions experienced by the calf. Calves born to cows that have been subject to heat stress in late lactation have been shown to have shorter standing bouts in the first week of life and shorter more frequent standing bouts in weeks seven and eight of life when compared to calves born to cows that had been cooled in late lactation (29). This is consistent with the findings of this experiment. This study has identified that season/environmental factors should be considered as it may alter behaviors. However, this analysis is limited by the fact that no calves included in this study were born between February and July and this should be considered for further studies. It may also be that future disease detection tools will also need to take into account factors such as substrate, which has been previously shown to affect locomotor play in an arena test (57).

The finding of a tendency for increased activity (Total Daily MI) in female calves in this study concurs with a previous finding that female 6-week-old calves spend more time walking than their male counterparts (20) and may reflect the increased reaction to human approach previously seen in female calves (18). Interestingly, previous work has shown that male calves required fewer training sessions to use an automatic milk feeder than female calves (58). This contrasts with the finding in this study that male calves tended to drink slower (Figure 4B) and were much less likely to consume their daily milk allowance (Table 3). However due to the small sample size the interaction between sex and age was not tested so while this may be an effect of learning to use the feeder this cannot be confirmed.

Previous studies investigating effects of calf breed have not used either the same breeds or the same behaviors as the current study. The association of sire breed type with mean lying bout length (Figure 4A) and drinking speed (Figure 4B) seen in this study may reflect an effect of the calf's personality on their behavior. Previous authors have studied breed differences in some personality traits. For example, a study has previously shown that Holstein calves had a more fearful response in the escape test when compared to Holstein beef cross calves (17). Previous work using Holstein calves looked at three personality types, interactive, exploratory-active, and vocal inactive and found that more vocal calves have a greater number of rewarded visits to the feeder pre-weaning (59). It was not possible in this study to explore the interaction between breed and sex. This is due both to the small sample size and the skewed distribution between sex and breed. This occurred in part as an artifact of the data selection process and in part due to the breeding policies of the two sub-herds.

The association of age at inclusion into the group pen with mean MI per standing bout (Figure 5C) may reflect the reduction in social interaction seen in calves grouped at an older age (21). It is unlikely however that the change in behavior purely reflected the initial post grouping period however as calves were grouped at a maximum of 13 days of age (Mean 8.7 days, Supplementary Table 3) while the mean age of the data included in the final data set was 27.7 days. It was an interesting finding that age at inclusion into the group pen was not significant for any milk feeding behaviors in this study. Calves introduced to the automatic milk feeder at less than a day of age have previously been found to drink less milk in the first week of life than those fed three times daily by bottle until 5 days of age, however this effect was not seen after 8 days of age (58). This may explain why no association with milk feeding behaviors were seen in this study, as the days immediately after introduction to the group pen were not included. However, another study found that the frequency of rewarded visits was significantly lower in the first 12 days in the group pen in calves that were introduced at 6 days of age compared to those introduced at fourteen days of age, thus there may be an effect of age at introduction for a longer time period after introduction (60). Interestingly, in a study with a slightly lower age at inclusion into the group pen (mean 7.8 days, SD 1.9 days compared to mean 8.7 days, SD 2 days in this study) a higher age at inclusion was associated with an increase in the age at first calving (61).

Dam lactation was tested in this study as a proxy for maternal management as heifers will be managed differently to other cows. As the calves are removed as soon as possible after birth and are not fed colostrum from their own mother, differences in maternal behavior between lactations are unlikely to influence the calf. Lactation number was not associated with any of the behaviors in this study (Tables 2–4). This may be due to an insufficient difference in management between the lactation groups, especially in late pregnancy.

Several studies have examined changes in the number of unrewarded visits to the automatic milk feeder with disease. The number of unrewarded visits has been shown to decline in the 2 days prior to neonatal calf diarrhea diagnosis (62) and the 3 days prior to bovine respiratory disease diagnosis (63). Unfortunately, the automatic milk feeder used in this study did not record unrewarded visits to the automatic feeder. A potential area of further research is to repeat this experiment with an automatic milk feeder that records both rewarded and unrewarded visits to the feeder.

A further limiting factor of this study is the size of dataset. Thirty-one calves out of a possible 114 calves were included with ten calf days each. Gathering a sufficient data set of healthy calves is challenging due to the high incidence of disease in young calves. Across eleven UK dairy farms, on farm incidence of bovine respiratory disease and neonatal calf diarrhea in the first 9 weeks of life were 20.4–77.8% and 24.1–74.4%, respectively (2). While more relaxed inclusion criteria could have been used, the authors deemed that it was vital to remove any confounding caused by disease due to its strong effect on behavior. The small dataset is likely to particularly impact the investigation of the effect of lactation as studies into maternal factors would frequently analyze far larger datasets, e.g., 60 calves (29), 551 calves (64) or 15,992 calves (31). However, some studies have used a smaller number of calves, Ling et al. (25) used twelve calves in their study. Larger datasets are likely to be beneficial especially when attempting to identify subtle changes in behavior.

Interestingly, a recent study has shown that individual repeatability differs between milk feeding behaviors even after adjustment for age, cohort and day number (65). This could mean that further studies should focus on those behaviors with a high level of intra-individual repeatability. Intriguingly, feeding rate and the number of meals were highly correlated in a previous study (47) but were not significantly correlated in this study when correlations of the entire dataset were calculated (Supplementary Table 6), yet the approach of calculating correlations for individual calves and then an overall correlation may warrant further investigation.

In recent years the use of sensors and automated systems to detect disease has been increasingly researched (5). However, the performance of the reported models for bovine respiratory disease in pre-weaned calves is frequently poor, accuracies reported include 59.1% (66), 75% (54) and 80% (11). While no data is currently available on calf diseases, we know from previous work done on automatic tools for lameness detection in Canadian dairy herds that accuracy is a key concern for farmers (67). Farmers required a system that both detected a high proportion of lame cows and did not give many false alerts. Thus, it is vital that we have a thorough understanding of factors other than disease that are associated with changes in sensor measured behaviors in pre-weaned calves. This study aimed to address this gap in the literature and identified several factors that are associated with sensor measured behavioral variables. However, this study was conducted in a single shed; thus, it has its own environmental and management variables. To verify the results of this study the work needs to be repeated on several units.

This study found several intrinsic calf factors that are associated with healthy calf behavior. Most notable of these were live-weight, season, sex, and age. These are important considerations for comparing behavior between sick and healthy calves and predicting disease. However, there are likely to be farm level factors that also apply, and this needs further research as all calves in this study were reared in the same shed.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by Animal Experiment Committee of the Scottish Rural College.

MH, C-AD, CM, JB, and DB designed the data collection of this work. BR developed the idea for the analysis, cleaned the data, performed the analysis, and wrote the first draft of the paper. MH, C-AD, AC, and CM provided support with the data cleaning and analysis. All authors assisted with drafting of the manuscript and approved the final manuscript.