Animal foraging strategies are a delicate balance between food preference and food selection influenced by various factors like nutritional requirements, availability of food, predation, competition etc (Schoener, 1971). Food preference is defined as the discrimination exerted by animals on different food types resulting in the selection of one food type over the other when no constraints bear on their choice (Rozin and Vollmecke, 1986). Multiple experiments have shown that, given the chance, animals do indeed prefer certain food items more than others (Berteaux et al., 1998; Ganas et al., 2008). But such idealised situations rarely occur in the natural world and thus animals may not always be able to acquire food according to their preference. For example, goats and giant anteaters make different dietary choices when foraging in the wild versus in experimental conditions (Redford, 1985; Nefzaoui et al., 1995). Thus, it is necessary to distinguish between food preference and food selection, defined as preference modified by environmental conditions (Hodgson, 1979). The search for the best possible food with the least risk and energy investment, thereby having the highest survival advantage, gives rise to different foraging strategies. Foraging can thus be viewed as a sequence of events- searching, assessing, handling, and utilization of resources and food with the ultimate aim of gaining energy and fitness and each component of this process may be optimised according to the species, trophic level or situation.

The decision-making processes that underlie what and how an animal feeds are of import not only to understand individual foraging behaviour but are also important in the broader sense of population dynamics and community structure amongst other ecosystem processes at many levels. Optimal foraging theory is a behavioural ecological model that seeks to understand the foraging behaviour of animals in the context of the efficiency of a certain behaviour or choice in the presence of local and global constraints (Kamil et al., 1987). According to classical optimal foraging models (Pyke, 1977), an animal must make multiple decisions whilst foraging – (a) choice of what to eat (food selection), (b) choice of where to search and feed from (patch selection), (c) amount of time to be allocated to different patches, and (d) choice of movement patterns and speed between patches. As the availability and abundance of food is neither uniform nor continuous in nature, the outcomes of these decisions must be dynamic, governed by the stochastic nature of the challenges. Urbanisation, in recent years, has been instrumental in transforming ecological conditions like habitat and food sources, introducing further heterogeneity in animal foraging ecology. Animals living in urban habitats often face new challenges, apart from stochasticity in resources, in the form of novel, anthropogenic food and scarcity of their natural food. One way in which urban animals adapt to this challenge is by diet-switching, a foraging behaviour wherein the animal starts consuming an alternative food/prey when the primary food is scarce in number (Dasgupta et al., 2021; Sarkar and Bhadra, 2022).

The Canidae family includes species that show a range of foraging patterns, including several that are opportunistic and generalist predators and are increasingly being found in and around urban settlements (Kjellander and Nordström, 2003; Contesse et al., 2004). Hence the canids are a highly suitable taxon for studying foraging behaviour in urban centres. Being group-living animals, their challenges in food procurement are not only environmental, but also social. Members of a group-living species might forage alone or in groups depending on the food abundance in the area (Le Roux et al., 2009). Foraging in groups provides animals with the advantages of lower predation risk and higher efficiency (Clark and Mangel, 1986), but group-living comes with various disadvantages too. Intra-group competition is one of the costs of group living (Janson, 1985). Competition has an effect on the quality and quantity of food acquired by individuals. As a trade-off to competition, individuals may change dietary preferences or forage from lower quality patches (Schulke and Ostner, 2012) resulting in eating inferior food (Taylor, 1970) or eating less (Janson, 1985). Furthermore, individuals in a group may exert more effort and forage faster, as a manifestation of competitive interaction, in order to obtain their share (Clark and Mangel, 1986).

The free-ranging dog (Canis lupus familiaris) is an ideal model canid to study the effect of various factors and challenges of an urban, human-dominated environment on food preference and selection, if any. They have been living in human habitations for centuries, much earlier than any other canids, and have become well adapted to their scavenging lifestyle and dependence on anthropogenic food (Pal, 2001; Debroy, 2008). Despite subsisting on a carbohydrate-rich omnivorous diet, they maintain a strong preference for meat in cafeteria-type trials (Bhadra and Bhadra, 2014; Fleming et al., 2017). They find their preferred food using the Rule of Thumb: “If it smells like meat, eat it” (Bhadra and Bhadra, 2014). This preference is also exercised whilst scavenging by the dogs in a more realistic scenario of finding food from garbage bins. Individual dogs were found to preferentially find and eat meat pieces from the noisy background of garbage first, as compared to other types of food using the “Sniff & Snatch” strategy (Sarkar et al., 2019). They show qualities of a periscopic forager by optimising the order of sampling and eating from resource patches (in this case, the boxes) on simultaneously encountering them (Mitchell, 1989). This type of feeding strategy has characteristics of an optimal foraging model where the highest ranked food (in both quality and quantity) is sought out first and then less preferred food are added to the diet (Pyke, 1977). In the previous experiment, although the dogs seemed to prefer eating from the protein box, it was not immediately clear whether the dogs carefully assessed available patches before making a decision or they behaved impulsively or were able to hone on the best available resource immediately using their strong olfactory capabilities. Thus, whilst we knew their feeding preference, we did not know of the decision-making and the strategies that were instrumental for exercising this preference. Furthermore, we observed this preference during foraging by solitary dogs, as the trials in the previous experiment were performed on individuals, and not groups of free-ranging dogs.

Free-ranging dogs typically live in groups, and whilst foraging, show different levels of social organisation, from solitary, to pairs, and groups of three or more (Sen Majumder et al., 2014). We carried out a multiple-choice task with free-ranging dog groups to understand whether they preferentially feed on meat using the Rule of Thumb, “If it smells like meat, eat it” (Bhadra et al., 2016), in the presence of competitors. The objectives of this paper are to examine the feeding patterns and strategies of free-ranging dogs in groups and to compare these against those of solitary dogs in the context of the predictions of optimal foraging models. We could only compare the foraging behaviour of the free-ranging dogs against the first three decision-making points of the optimal foraging theory-what to eat, where to eat from and how much time and effort to expend in a patch. Our experiment design did not require the dog to travel any amount of distance to access the different resources and thus decision making in relation to movement patterns and speed was not studied. We hypothesized that presence of other group members will have an effect on the foraging strategies of dogs as compared to when they are solitary; they would try to minimise competition by relaxing resource selectivity, rather than entering into active competition for the most preferred resource.

The results have been broadly categorised under various components of the foraging process-(a) Sampling (this includes the search, exploration and evaluation components), (b) Handling and Allocation (this includes the time and effort expended on a “patch”), (c) Feeding (this includes the utilization of resources and preference components), and (d) Risk Management (this looks at the various behavioural tools and solutions to manage competitors, predators, and other challenges).

Resources are often distributed in a heterogenous and stochastic fashion across time and space in patches. The acquiring of such resources is dependent on the ability to make profitable foraging decisions by foragers. Such decisions can only be made by gathering information about patch heterogeneity. Patch quality information can be gathered whilst exploiting it. This is known as sampling for information (Stephens and Krebs, 1986). Foragers harvest information from the environment by sampling and investigating through visual, olfactory, and chemical cues and using such information to modify their behaviour accordingly (Lima, 1984; Sandlin, 2000). In uncertain environments, animals spend a considerable amount of their time investigating, with a significant proportion of that spent in resource-rich patches (Lima, 1984; Naef-Daenzer, 2000). As the resources in a patch deplete, foragers may have to frequently reassess patch choice and make decisions whether to stay, move on or re-visit a previous patch (Smith and Sweatman, 1974; Krebs et al., 1977).

In our experiment, the boxes simulating dustbins acted as heterogeneous resource patches about which the dogs had no prior knowledge. Thus, the only way to collect information and assess the quality of the patches was through sampling them. A major decision node around which we split the dogs’ sampling efforts was the first eating event. Previous research had shown that solitary dogs harvested the best resource patch first but showed no such preference during sniffing (Sarkar et al., 2019). Our current experiment showed that sampling played an important role during foraging in free-ranging dogs. Dogs in groups sampled multiple “patches” (boxes), leaving behind sub-optimal food as evidenced by the fact that sampling likelihood increased if dogs came across the carbohydrate box first, until they encountered the best available resource and then ate from it. Solitary dogs, on the other hand, were more likely to keep sampling even if they encountered the best “patch” first, the protein box. The fact that dogs who sampled multiple boxes before eating did not keep sampling afterwards elucidated the role of sampling as an information gaining process with specific purpose of finding the best available resource “patch” and not a random, undirected behaviour. Furthermore, the post-eating sampling tendency increased with an increase in group size. Thus, dogs in groups were more likely to “eat first, sample afterwards” as compared to solitary dogs. This behaviour hinted at an implicit effect of competition, in the presence of conspecifics, at play. Overall, these results highlight several salient features of the dogs’ sampling strategy. Whilst sampling individually, it might be hypothesized that free-ranging dogs were always on the lookout for better opportunities. Thus, even after encountering protein box, they often chose to leave it without eating from it to sample other boxes. In groups, dogs were aware of the risk of opportunity and food loss posed by the presence of other group members and thus paused their sampling as soon as they found their preferred food, regardless of quantity. This strategy of quickly initiating eating from a near optimal (mixed) or an optimal (protein) “patch” they encountered and then collecting information from other “patches” assured them of some amount of energy gain whilst keeping their options open for moving onto better “patches,” if they became available. This demonstrates that dogs were able to assess and evaluate the quality and quantity of resources available in a patch. They were also aware and adaptable to the presence of other members and took all of these into account before making a decision of eating. Additionally, in both cases, solitary, and group, no particular box was sniffed first more than the others, thus demonstrating that the dogs did not have an instantaneous clue of the location of preferred food and thorough sampling was the only way for them to gather information about the “patches.” This is further supported by the fact that once profitable patches had been located through sampling, dogs in groups tended to eat more and eat first from the protein and mixed boxes as compared to the carbohydrate box and solitary dogs ate more and ate first from protein box only (Sarkar et al., 2019).

In keeping with the unpredictable environment that they live in and the fluctuating food resources available therein, free-ranging dogs were found to investigate and engage with all the available resource “patches” (boxes), though they concentrated their maximal efforts on the energy-rich boxes (protein and mixed), choosing to handle them more number of times than the energy-poor boxes (carbohydrate). Furthermore, instead of allocating all their time to the best “patch” once located, the dogs allocated different proportions of their available time to different “patches” depending on their quality with better “patches” allocated more time but suboptimal patches allocated a small proportion of time too. Although this seems to be a deviation from optimal behaviour, it is hypothesized that the dogs are actually prioritising a long-term adaptation to a fluctuating environment, akin to titmice foraging (Smith and Sweatman, 1974). In their natural habitat, food resource clumps such as dustbins and garbage dumps vary in food abundance across time such that high abundance food sources can become low abundance ones at a later time and vice versa. In such a scenario, it is advantageous to spend some time handling all “patches,” especially if the distance between the “patches” are minimal, as is the case here, in order to continually track the status of the environment and change the allocation as and when necessary. Additionally, dogs spent a lower proportion of time in each handling attempt for the carbohydrate box as compared to the mixed and protein boxes. This can be because of two reasons. One, apart from handling carbohydrate box fewer times, the smaller time allocation for each attempt shows that dogs expend less effort for a suboptimal resource. The other reason seems to be that between feeding from protein and mixed boxes, dogs would sample from carbohydrate box too and would quickly assess the value to be lower than that of the other two boxes and move back to them. Unfortunately, we were unable to observe the amount of food eaten by individual dogs. Thus, we were unable to find out the point of resource depletion at which the dogs shifted to another patch.

The decision-making displayed in the sampling and allocation components of free-ranging dogs’ foraging behaviour implies that the dogs employ a random search optimization process between patches initially when information about the patches is low, and then switch to systematic foraging when information is obtained (Baum, 1987; Bartumeus and Catalan, 2009). By heterogenous sampling of the patches, the dogs attempt to increase their chances of encountering their preferred food quickly. Once located, more time and effort are spent in some patches than others. Further research is required to provide a definitive understanding of the optimization process, for example, whether dogs revisit previous patches, and to check the generalizability of the process observed.

Free-ranging dogs are known to forage singly as well as in groups (Sen Majumder et al., 2014). Group living animals have to balance between access to quality nutrition and conspecific competition. Decreased resource selectivity in the presence of conspecifics reflects an adaptive response to competition. Such decreased selectivity can also be observed in the free-ranging dogs in the current experiment. Whilst solitary dogs sought to maximise both quality and quantity of the available food by preferentially feeding from the protein box (Sarkar et al., 2019), dogs in groups show a relaxation in their preference by feeding from both protein and mixed boxes equally. Whilst it might be hypothesized that the dogs moved to the mixed box only after they believed that the protein box had no valuable resource to offer them, we do not think that is the case here. The dogs ate first from both the boxes equally showing that instead of the protein box being the top “patch” in preference hierarchy, the dogs have relaxed their choice to include both as equally preferable. As dogs ate chicken pieces in overwhelmingly high amount than bread (80.5% of the total food eaten; Supplementary Figure 2), it implies that even whilst eating from the mixed box, they selectively, but not completely, preferred eating the chicken pieces over bread. Thus, whilst there is increased acceptance in terms of feeding from lower valued “patches” (lower quantity of preferred food), there is little change in acceptance of lower valued food as compared to solitary dogs. Solitary dogs follow the SnS strategy exclusively for the protein box whereas the dogs in groups apply the strategy to both mixed and protein boxes. Taken together, these results show that even in groups, dogs do follow the Rule of Thumb, “If it smells like meat, eat it” (Bhadra et al., 2016). These results lend credence to optimal foraging prediction, with a slight deviation, that states that if foods of higher value are available, low value items should be rejected, regardless of their abundance (Pyke, 1977). Similarly, free-ranging dogs sequester the preferred food, meat, first, followed by bread. But, instead of complete rejection, there is preference priority in that whilst they prefer eating chicken, they are not averse to eating bread if they do come across it in the mixed box. This seems to be an adaptation of the scavenging lifestyle that these dogs lead wherein they try to maximise nutrients in any form they can whilst maintaining their preference.

Free-ranging dogs are known to be bolder in groups (Bhattacharjee et al., 2020). The reduced time demonstrated by groups in responding to the experimental set-up and in eating as compared to solitary dogs imply the presence of such traits during foraging. Indeed, it was found that solitary dogs are more likely to indulge in vigilance behaviour as compared to those in groups, perhaps due to the likelihood of facing competition from other dogs and scavengers whilst foraging. Since this was not the focus of this experiment, we did not attempt to look into this behaviour in great detail. Dogs in groups are also known to spend more time engaging with the boxes in each foraging attempt. Foraging in groups confer multiple benefits to animals, some of which such as reduced vigilance (Beauchamp, 2003) and more time spent in foraging (Gotmark et al., 1986) seem to be behind the dogs’ action here. However, we must specifically investigate other factors that might elicit similar behaviour in future studies. Intragroup competition and social rank appropriate behaviour might be two such factors (Elgar, 1989).

To summarise, dogs in groups initiated their sampling randomly, choosing to keep sampling multiple boxes if they first encountered a suboptimal “patch” (carbohydrate box). On encountering an optimal (protein box) or near optimal (mixed box) “patch,” they followed a “eat first, sample afterwards” strategy. The likelihood of this strategy being followed increased with increasing group size. These dogs allocate varying proportions of time to different “patches” based on their quality with higher quality “patches” (both protein and mixed) allocated a higher portion of the time. Dogs, in general, expended more effort (quantified through attempts) in handling these higher quality “patches” with more time spent in these “patches” with each attempt. A similar preference of “patches” is observed in the feeding choice of groups too. Overall, dogs in groups spent a higher proportion of time in each of their handling attempt, initiated sampling and eating quicker, and were less likely to display vigilant behaviour as compared to solitary dogs.

Solitary dogs initiated their sampling randomly and were likely to sample all three boxes at least once. Unlike dogs in groups, solitary dogs were likely to keep sampling boxes even if the first box they encountered was the protein box. They were least likely to sample once they started eating. They preferred feeding from protein box over others and allocated more time to handling protein box over carbohydrate box.

In this study, we investigated the foraging behaviour of free-ranging dogs in groups and compared it against that of solitary dogs. This study is the first of its kind to explore in detail the foraging strategies of these dogs under different social conditions. We found that the foraging tactics of free-ranging dogs fulfil multiple characteristics of a stochastic optimal foraging strategy and that these animals are highly plastic in their behaviour and adaptable in their strategies, as befits an organism living in an unpredictable environment. Further research may focus on finding out the energy threshold that make a “patch” valuable and how animals assess that. Studies may seek to explore the mechanism behind the inverse relationship between group size and vigilance and disentangling the various factors that come into play when foraging in group such as competition, hierarchy and food density. Examining the effect of these confounding factors in tandem and alone will shed greater insights into the foraging behaviour of these animals and the causal mechanism behind such behaviours as reduced latency. Another thing worth noting would be the differing motivations of individuals, like pregnant females, across different time of the year (pre-mating and post-mating). Such studies can help to understand the strategies and behavioural patterns that enable urban adaptation in animals that adopt a scavenging strategy.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at: https://osf.io/ma6uq/.

Ethical review and approval was not required for the animal study because the experiment did not involve any invasive procedure, and the food provided to the dogs were fit for human consumption. Dog feeding on streets is permitted by Prevention of Cruelty to Animals Act 1960 of the Parliament, and this experimental protocol did not need any additional clearance from the Institute ethics committee, as it did not violate the law.

RSa and ABha: conceptualization, project administration, and writing—reviewing and editing. RSa: data curation, formal analysis, methodology, visualization, and writing original draft. ABha: funding acquisition, resources, supervision, and validation. RSa, ABho, DD, RB, PC, AN, RSr, AR, RSo, and AM: investigation. All authors contributed to the article and approved the submitted version.

RSa was supported by IISER Kolkata Institute fellowship. AN was supported by JNCASR Summer Research Fellowship Programme, 2019. RSo was supported by the INSPIRE fellowship of the Department of Science and Technology, India. This work was supported by the Science Engineering and Research Board (grant: EMR/2016/000595), Department of Science and Technology, Government of India.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.