Under domestic conditions, most foals are artificially weaned prior to the time of natural weaning, usually between the age of 4 and 7 months [1
]. The breaking of the mare–foal bond (associated with the disruption of milk intake) is usually abrupt, by placing the foal out of sight and sound of its dam, sometimes more gradual, but in all cases, takes place when the foal is still closely bonded to its mother [1
]. In addition, the foal generally undergoes additional nutritional, social and environmental challenges. Even though the precise origin of early artificial weaning is unclear, it is important to go back to the sources of such a practice. In the second half of the 19th century, several studies highlighted two main results that probably led to the routine practice of early artificial weaning. Firstly, it has been found that the maternal milk production decreases sharply by the third month of lactation [4
] and secondly, that the nutritional requirements of 3–4 month-old foals exceed the level of nutrients available from maternal milk [5
]. From that point, early weaning may have been considered the best decision to make to optimize the physical development of domestic foals. This practice became rapidly widespread in professional breeding farms followed by non-professional breeders of one or two mares. Nowadays, there are many practical, economical and safety reasons to proceed to the early weaning of foals, such as allowing for an early marketing of foals, switching the foal’s attention from the mother to humans [7
], facilitating the management of the foal’s nutritional intake without the mare interfering, or even optimizing the subsequent reproductive efficiency of the mares by limiting the potential negative impact of a prolonged nursing. Some of the reasons are based more on habits and tradition, perhaps even on false beliefs, and clearly not on the prospects for improving the welfare of domestic foals.
While the impact of artificial weaning on mares has not yet been properly examined [9
], it is now well admitted that artificial weaning is one of the most stressful events in a foal’s life [1
]. The well-known behavioral responses to early weaning (e.g., increased long-distance whinny calls, increased active locomotion and eliminative behaviors) and the associated risks of injury, typically peak within the first two post-weaning days [2
]. However, other behavioral changes such as altered feeding and sleeping patterns, aggressiveness, suspension of play and redirected suckling towards peers stemming from frustration, may be observed for much longer periods [14
]. Weanlings also experience elevated glucocorticoid (stress hormones) levels, changes in heart rate, as well as a decline in growth rates [1
]. Most of these behavioral and physiological manifestations are more intense in the first post-weaning days and vanish within two-three weeks. Artificial weaning results also in chronic problems (which are more difficult to link to artificial weaning by foals owners). Thus, stress hormones lead to a subsequent decrease in immune response [20
]. They have also recently been reported to have a negative impact on the maturation of the gut microbiota which can lead to the potential growth of harmful microorganisms such as Escherichia coli
], while, in other species, recent studies emphasize the key role played by intestinal microbiota in the long-term health of the host [23
]. Numerous scientific studies also point out, in foals [2
], as in other domestic or captive young mammals [27
], that early weaning is a major cause of stereotypic behaviors and thus of animal welfare impairment. Thus, a 4-year prospective study performed on 225 young horses (70% were weaned between the age of 4 and 6 months) revealed that crib-biting was initiated by 10% of the foals at a median age of 20 weeks (i.e., 5 months), 30% showed excessive wood-chewing (i.e., lignophagia) by 30 weeks of age (i.e., 7.5 months) and 7% suffered from locomotor stereotypies by the age of 64 weeks (i.e., 1.5 years) [25
]. Although this has not been demonstrated to date in weanlings, it is worth noting that most studies found altered learning abilities in adult stereotypic horses as compared to healthy horses [28
]. According to Wolter [31
], early artifical weaning can also lead in the long term to other disorders, especially in future breeding mares who could later be less fertile and produce weaker, slower-growing foals with a weaker skeleton. Some proposals to limit the distress induced by weaning have been put forward [1
], such as progressive retrieval of the mares [13
], repeating separations [32
] and introducing adults in groups of weanlings [13
]. However, few studies investigated the importance of the timing of artificial weaning [1
] and, overall, none of the previous studies questioned the practice of artificial weaning.
Without human interference, the situation is of course dramatically different. It is often reported that weaning takes place gradually over several months through the joint initiative of the dam and the foal, and that foals are usually not weaned before the age of 9-11 months or until shortly before the birth of the next foal. In addition, “natural weaning” only implies a nutritional aspect, as the close dam-offspring bond remains afterwards for much longer time [35
]. Beyond these general statements, the weaning process (such as the gradual process of maternal rejection or distancing between the mare and its foal over time) is however still relatively unknown as few scientific studies have been done at the later stages of the lactating period (i.e., after the first 6 months) and the potential factors of variation in the weaning process remain poorly documented [37
]. After reviewing the scientific literature on the temporal dynamics of the dam-offspring bond in horses living in naturalistic conditions, we present the results of a study conducted in Iceland on mare–foal dyads (n = 16) kept outdoors in stable social groups with limited human interference (i.e., no intervention in foaling and no artificial weaning). We believe that better knowledge of the natural weaning process can provide new insights on the management of weaning under domestic conditions, and even lead us to reconsider such routine practice.
3. A Study on Spontaneous Weaning in Domestic Foals
In horses, the temporal dynamics of the mother-offspring bond during the late stage of the lactation period (i.e., when foals are older than 6 months of age) remains nowadays poorly documented because studies are few in number and/or mostly based on discontinuous observations. It should be noted that the natural process of weaning and individual variations have been little studied, while the immediate effects of weaning on the mare–foal relationship and behavior of both partners remain unexplored. To partly address these gaps, we performed a study around the weaning period on Icelandic mares and their foals, kept outdoors with minimal human interference. In some places in Iceland, horse management practice provides a unique opportunity to observe such situations under semi-natural domestic conditions, with foals allowed to stay with their dams until they are at least 1 or 2 years old. First, we focused on the age of foals at weaning and potential factors of variation including the age, parity, conception rate and body condition of mares, as well as the date of next foaling and the sex of both the current and upcoming foal (as all mares were again pregnant). Second, we examined, retrospectively, whether the mare–foal relationship, as well as the behavior of mares and foals, exhibited changes just prior to and/or just after the weaning of foals.
3.1. Subjects and Study Site
This study was conducted on 16 Icelandic mare–foal dyads living in three stable social groups, composed of other mare–foal dyads and yearlings (Table 1
). The mares, aged 7 to 23 years old, were all multiparous with parity ranging from 2 to 13 (an average of 4.5 prior foals) and again pregnant. All the mares were in the group with their foals of the year and their yearlings. The foals, 9 females and 7 males, were all born between May and July, and were issued from 11 sires. Horses belonged to the Farm at Holar University College, Iceland (study groups 1 and 2) and to a professional breeder (study group 3) at the farm Kalfstadir, close by Holar. In both sites, all the groups were maintained all the year round in semi-natural settings (large pastures with natural resources, stable social groups). All horses from the same social group have been kept together at least for a year prior to the start of the study. Animals were supplemented with hay (in winter only) and salt. The water supply came from natural streams and sometimes in winter by the ingestion of snow. Human interference was very limited outside of the distribution of forage and prophylaxis care: foaling took place outdoors without human assistance, juveniles were left with their mothers up to the age of 2 years. In both sites, young males are usually removed from the natal group around the age of 12 months, either temporarily or permanently, respectively to undergo castration or prevent them from covering females.
Daily observations were conducted from February to May, when most foals were over 6 months of age. Special attention was paid to various variables commonly used to assess the quality of the mother-offspring relationship in horses. These include mare–foal distances, choice of the dam as preferred neighbor, frequency and quality of mare–foal interactions including suckling activities [37
]. All observation sessions took place during the daytime period between 10:00am to 5:00pm, with 3 to 6 observation days per week and per study group (depending on the weather conditions, e.g., snowstorms, temperature, etc.). Observation periods changed every day following a rotation schedule. Observations were tape-recorded and transcribed later. Three observers were involved in data collection. Reliability was controlled using the kappa coefficient of Cohen [72
] that rated at k = 0.95.
All behaviors of foals and their dams were continuously recorded (“focal sampling” [73
]) for 5 minutes, 6 times a week (30mn. per individual per week), for details on frequency (occurrences per hour) and type of behaviors expressed by each foal and each mare. The following behavioral items were recorded: locomotion, exploration (e.g., sniffing the ground), grazing, feeding (e.g., hay), drinking, resting standing or lying down, solitary play (including manipulation of an object and locomotion play), self-grooming and social interactions with mother/foal and/or other social partners. Interactions recorded were as follows [40
]: (a) social investigations (e.g., sniffing through naso–nasal, nasal–body or naso–genital contact); (b) affiliative interactions (e.g., mutual grooming, other physical contacts, social play); and (c) agonistic interactions (e.g., threats to bite or kick, bite, kick).
Suckling activities and associated behaviors were also recorded each time they were observed in the group, specifying the identity of the dyad concerned (“all occurrences of behavior") [73
]. Each week, two observation sessions of 1.5 hours each (as foals suckle on average once per hour at this age) were carried out. During these sessions, we noted the number of suckling attempts (i.e., foal suckling initiatives rejected and terminated by the dam, duration < 5 sec.), the number of (successful) suckling bouts, which partner (dam or foal) terminated the suckling bout and, if it was the dam, the associated behaviors (e.g., distancing from the foal, threat to bite…) were recorded.
The behavior of each mare and each foal, the mare–foal distance, as well as the identity and distance to the nearest neighbor of both partners, were recorded, using the “instantaneous scan sampling" method [73
]. Scans were performed every 5 minutes for an hour, three times a week (36 scans per week). Distances were measured in horse body-lengths and six distance categories were used:  (i.e., contact), ]0;1], ]1;5], ]5;10], ]10;20], more than 20 “horse body lengths" [59
]. Six social categories were used: [mother or foal of the focal individual], [other mares], [other foals], [related yearling], and [other yearlings]. Scan sampling yielded three types of data: (1) time (in %) devoted to the different behavioral items (time-budget); (2) time (in %) spent at different distances to the dam (or the foal) by the foal (or the dam) (dam-foal spatial relationship); (3) time (in %) spent near different neighbors by the foal (social preferences).
Lastly, the body condition of mares was scored monthly following the method of Arnaud et al. [74
] (in addition to the follow-up carried out by the breeders). The score ranged from 0 (emaciated) to 5 (obese) with a score of 3 corresponding to an optimal body condition.
As data were not distributed normally, non-parametric statistical tests were used [75
]. The statistical tests were performed using Statistica® 13 (Statsoft, Tulsa, USA). All means are given ± SE (standard error).
Two main types of analyses were performed: (1) we first investigated the age of foals at weaning and the potential factors of variation (for all study groups); (2) we retrospectively compared the behavior of both partners during the two weeks prior to weaning to that scored in the two weeks following weaning (only for study groups 1 and 2).
Weaning age was defined as the age at which the foal was last observed performing nutritive suckling. We also determined the duration of the dry period (i.e., the time from the last suckling of the current foal until the birth of the next foal) by recording the date of birth of the next foal. We calculated the coefficients of variation (CV), which represent the ratio of the standard deviation to the mean, to examine individual variation. Several factors of variation have been examined: the age, parity, conception rate and body condition of mares, as well as the date of next foaling and the sex of both the current and upcoming foal. Mann–Whitney U-tests were used to compare two independent samples (e.g., sex differences), while Spearman tests were used for correlations.
In order to examine whether weaning induces or reflects a clear change in the mother-offspring bond, we compared the behavior of both partners during the two weeks prior to weaning to that scored in the two weeks following weaning. Changes between the two time periods in mare–foal distance, social preferences and behavior of both partners were addressed using Wilcoxon signed rank tests. Preferential spatial partners were identified within family groups and based on the spatial proximity to the nearest neighbors. Preferential spatial partners of individual A were those that were more frequently the closest to A than expected by chance (partitioned chi-square goodness-of-fit test).
3.4.1. Age at Weaning and Factors of Variation
Spontaneous weaning occurred when foals reached the average age of 9.0 ± 0.1 months. The dry period was on average of 3.2 ± 0.2 months (Table 2
). There was no difference between the three study groups either for the age at weaning nor the duration of the dry period (Kruskal–Wallis: p
> 0.1, Table 3
). Considerable variations between the different mare–foal dyads were found. The age of foals at weaning ranged from 7.7 to 10.0 months, while the duration of the dry period ranged from 1.9 to 4.6 months (Table 3
). Individual variations were higher for the duration of the dry period (Coefficient of variation: CV = 22.9%) as compared to the age at weaning (CV = 7.7%; Table 2
and Table 3
The age of the foal at weaning (Nfemales = 9, Nmales = 7; Xfemales = 8.9 ± 0.2, Xmales = 9.0 ± 0.3; Mann–Whitney test: U = 30.5, p = 0.96) and the duration of the dry period after weaning (Xfemales = 3.1 ± 0.3, Xmales = 3.3 ± 0.2; U = 27, p = 0.68) were not sex-dependant. The sex of the upcoming foal had no impact on the age of foals at weaning (weaning age: Nfemales = 11, Nmales =5; Xfemales = 9.0 ± 0.2, Xmales = 8.8 ± 0.4; Mann–Whitney test: U = 31, p = 0.73) nor the duration of the dry period (Xfemales = 3.2 ± 0.2, Xmales = 3.2 ± 0.3; U = 28, p = 1). The age of foals at weaning and the duration of the dry period were not correlated with the age (Spearman: rs = 0.31, p = 0.25 and rs = -0.07, p = 0.81, respectively), the parity (rs = 0.08, p = 0.75 and rs = -0.06, p = 0.81, respectively) or the body condition (scored in the month of weaning) of mares (rs = 0.18, p = 0.57 and rs = -0.16, p = 0.59, respectively). However, it is worth noting that mares had an optimal body condition, with an average score of 3.0 ± 0.1, that remained stable over time. Only the conception rate (i.e., mean number of foals produced per year) of mares was positively correlated with the duration of the dry period (rs = 0.55, p = 0.028), while no significant correlation was found with the age of foals at weaning (rs = -0.31, p = 0.25): in other words, the more foals the mares produced on a regular basis, the longer the dry period lasted between two successive “lactation“ periods.
3.4.2. Impact of Weaning on Time-Budget and Dam-Offspring Relationship
In the two weeks prior to weaning, mares and their foals spent the majority of time in close proximity (from 28% to 40% of the time at less than 1 horse length and from 44% to 67% of the time at less than 5 horse lengths depending on the study group; Figure 1
). The dams were the most preferred spatial partner of foals among all the other individuals available within the group: foals spent from 21% to 40% of the time with their mothers as the nearest neighbor (Figure 2
a,b). Similarly, the dams showed a clear preference for their foals over the other members of the group (the foal was the closest neighbor from 30% to 44% of the time depending on the study group; Figure 2
Interestingly, the dams rarely displayed agonistic behaviors towards their foals (i.e., on average less than one occurrence every two hours), either during or outside of suckling activities. There was no decrease in the frequency of suckling bouts, with foals still suckling on average slightly less than once per hour (0.68 suckling bouts per hour), and they terminated most of the suckling bouts (69.2%) until the very last weeks prior to weaning.
No significant change in spatial proximity between the two partners was scored prior to and after weaning: mares and foals spent most of their time in close proximity (i.e., within one horse length; Figure 1
). In addition, foals showed an equally strong preference for their dams after weaning, still choosing the mother more often as the nearest neighbor compared to other individuals in the group (Figure 2
a,b). Similar findings were scored in mares in both groups (Group 1
: Xprior to weaning
= 44.4 ± 4.5%, Xafter
= 29.5 ± 6.1%; Group 2
: Xprior to weaning
= 34.1 ± 4.8%, Xafter
= 38.2 ± 4.1%; Wilcoxon test: p
> 0.1 in all cases). Furthermore, it should be noted that no suckling attempts were observed after weaning. Finally, apart from suckling activity, no significant change in the foals’ time-budget was observed (Wilcoxon tests: 2.50 < w < 11.50, p
> 0.1 in all cases): feeding and resting remained the two predominant activities, either two weeks before or after the weaning date.
The age of foals at weaning, but especially the length of the dry period, varied from one dyad to another. The main explanatory factor was the conception rates of mares. It should be noted, however, that the mares in this study had a number of common individual characteristics, limiting the factors of variation: they were all multiparous, pregnant again, and in the presence of both their foals and yearlings. Contrary, perhaps to expectations, the older mares did not tend to wean their foals earlier and overall, no loss of body condition was noted over the lactation period despite severe climatic conditions (negative temperatures, presence of snow cover, etc.), the absence of food supplements other than hay and a high conception rate of mares (the majority having a foal each year). Anecdotally, in the two farms where the observations took place, the breeders never had to wean a foal artificially because the mare’s body condition required it, or only on rare occasions (e.g., once in 10 years at the Hólar University College). It is important to note, however, that on both farms, mares had access to continuous 24-hour fiber provision, which is known to improve both the body condition and fertility rate of mares [76
]. It is also possible that the hardy character of the Icelandic breed may have played a role in the good maintenance of the body condition of mares. Further studies on different horse breeds are required to test for generality.
Surprisingly, the suckling frequency did not change in the two weeks prior to weaning, and mares did not express increasing rejection responses to the suckling initiatives displayed by their foals. Similarly, no suckling attempts have ever been observed in weanlings, suggesting that there was no frustration and that perhaps foals voluntarily decided to stop suckling. Finally, it is worth noting that spontaneous weaning did not induce signs of distress, as is often assumed [77
], nor any change in the foal’s time-budget (apart from the cessation of suckling activities) nor in the relationship between the mare and the foal. The impact of natural weaning in terms of animal welfare obviously contrasts with that of artificial weaning commonly practiced in professional and non-professional breeding farms.
4. General Discussion
Modern breeding practices generally impose strong constraints as compared to the conditions of development of foals in a more natural environment [3
]. One major aspect is the early artificial weaning, which is not just a stage of diet transition but also a stage of social separation. There is increasing evidence that such a practice, although carried out on a routine basis by horse breeders, leads to short- and in some cases to long-term severe negative outcomes [1
]. There is therefore a clear need to better understand the factors at stake (e.g., cessation of milk intake, immature digestive system, maternal deprivation, absence of adult models, additional changes in feeding or housing…), to improve the domestic management of weaning and animal welfare.
In this line of thought, one interesting finding of our study aimed at investigating the natural weaning process is that it induces no frustration (such as dam-directed suckling attempts or non-nutritional sucking of other foals) or distress behaviors (such as increased locomotion or aggressiveness) in either partner. These results thus suggest that the main source of stress in artificial weaning is rather the abrupt rupture of the dam–foal bond than the cessation of suckling. This would explain why gradual weaning, allowing foals to see, hear, smell, and touch their dams through a fence, but not suckle, results in fewer behavioral responses than abrupt weaning [14
]. Similarly, a recent study that investigated the effect of a two-stage weaning method which involves separating the effects of the nutritional part (suckling) and the physical part (social bond), tends to confirm this assumption. Indeed, no significant behavioral and physiological changes were noted in foals nor in mares during the first stage when foals were only prevented from suckling their dams (by the use of udder covers) [9
]. However, once mares and foals were then physically separated in a manner similar to the typical abrupt method, classical behavioral and physiological signs of distress (e.g., vocalizations, running, foal–foal aggressions, higher fecal cortisol concentrations…) were exhibited. Taken all together, these results call into question the initial assumption according to which the main aspect of the mother-foal bond is food, thus justifying the practice of artificial weaning from the age of 4 months as the mare’s milk then become insufficient for the foal’s energy requirements [4
]. Furthermore, it has been found that artificial weaning induces a similar reduction in average daily weight gain whatever the age at which it is carried out (4.5 months or 6 months of age) [34
]. To date, there is increasing evidence that other aspects relative to the dam-foal bond, such as emotional security [66
] or social preferences [40
], are at stake. This is probably one of the reasons why, outside of rare exceptions, weaning (cessation of suckling) prior to the age of 7 months almost never happens under (semi-)natural conditions [37
]. In other mammal species, including humans, it is well known that young animals or infants, once the attachment bond with the mother has been established, exhibit a strong preference to stay near their mothers, even in the absence of food supply [78
Therefore, when deciding when to artificially wean the foal, it seems important not to focus only on the age of the foal and the preparation of the feeding transition (which is most often recommended) [1
], but also to pay more attention to individual variations in the strength of the social bond between mother and foal [3
]. Some foals, even relatively old ones, may be less socially independent and closer to their mothers than others of the same age [65
] and, therefore, may be more likely to respond more intensely to maternal separation [9
]. For instance, a recent study by Nicol et al. [26
] found that foals that are more prone to develop post-weaning abnormal behaviors, spent more time suckling, prior to weaning, than other foals. Taking into account the individual characteristics of the mare (e.g., if pregnant or not; if pregnant, her stage of pregnancy, her conception rate) also appears to be of high significance to choose the best time to perform artificial weaning based on the scientific knowledge on the natural weaning process [37
]. Conversely, while weaning foals in a progressive way is often assumed to more closely imitate the natural weaning process, it is far from being the case. This weaning method consists in gradually increasing over a period of several weeks the time of maternal separation (from a few minutes to several hours) and, therefore, decreasing the possibility of suckling in foals [33
]. Progressive weaning does not mimic, therefore, the natural temporal dynamics of the dam-foal bond: in line with previous scientific work [36
], our study shows that suckling frequency does not decrease in the weeks prior to weaning, with still on average one suckling bout every two hours, and that mare–foal proximity remains stable until (and even after) weaning. The same applies to gradual weaning [14
] or the weaning method consisting in habituating foals to maternal deprivation by separating them from their dams repeatedly before final weaning [32
]. This may explain, at least in part, the discrependencies in terms of benefits obtained by these various approaches of artificial weaning [14
Finally, on the basis of the scientific knowledge relative to both the deleterious short- and long-term effects of artificial weaning and the benefits of natural weaning, the question of the relevance of performing systematically early artificial weaning in domestic situations can be raised. If in professional breeding farms, natural weaning seems relatively more difficult to implement, switching to natural weaning is feasible in a number of breeding farms that own a smaller number of breeding mares (more than 80% of breeders have only 1 or 2 breeding mares). There are easy changes to make in management practices to do so, such as installing selective feeders to make sure foals have access to sufficient food or providing forage ad libitum to make sure that mares are not losing weight, especially in cases of concurrent lactation and pregnancy [60
]. Maintaining foals with their dams over a longer period of time also offers the benefit to use maternal influences to facilitate the education of foals [41
], to decrease the occurrence of stereotypic behaviors in stereotypic breeding mares [83
], and to lower aggression within groups, horses being less agressive when in groups with foals [84
]. In addition, it appears that preserving a normal ontogeny with a prolonged and secure bonding with dams is a source of protection against adult-experienced adversity: for instance, in intensive macaque breeding facilities, it has been shown that wild-born animals expressed less altered welfare (e.g., stereotypies) than captive-born animals that had been weaned early [85
]. Similar questions are raised in other domestic animals, including companion (e.g., kittens and puppies) and production (e.g., calves and piglets) animals (e.g., [86
Re-thinking weaning practices in the domestic situation is crucial for obvious welfare reasons in both the short and long term. Further studies are needed to better identify what should constitute the best weaning practices with respect to the welfare of foals and mares (e.g., age of artificial weaning, adaptation to each mare–foal dyad, or to the body condition or reproductive status of mares). However, it also seems essential to conduct longitudinal studies to compare artificially weaned foals with foals that have been naturally weaned in order to assess all the potential consequences of artificial weaning.