2. Classical definitions of tool use

Although ethologists had for decades documented instances of tool use—like extracting, pounding, and pulling, mostly among primates and birds—it was not until the late 1960s and early 1970s that explicit tool use definitions were offered. We will begin our survey of tool use definitions with the classical definitions of Hall (Reference Hall1963), van Lawick-Goodall (Reference Van Lawick-Goodall, Daniel, Robert and Shaw1971), and Alcock (Reference Alcock1972), starting with Hall:

This served as a basis for the definitions of van Lawick-Goodall (Reference Van Lawick-Goodall, Daniel, Robert and Shaw1971) and Alcock (Reference Alcock1972):

These definitions share core similarities. Hall, Goodall, and Alcock all conceive of tool use as an action that establishes a relationship among three items:

1. (a) A subject, which is the organism that acts (also called the agent or actor).

2. (b) A tool, which is the mediating object (also called the agent of change).

3. (c) A target object, which is the thing to which the tool is applied (also called the object of change).

Furthermore, these items exhibit a specific causal connection: (a) uses (b) to affect (c). This causal structure allows us to distinguish proper tool use from other instances of object manipulation. If tool use is an object-mediated instrumental action, then it involves a triadic relational process, not a dyadic one. Thus, a capuchin monkey using its tail to hang from a pole, or a climber using the protrusions of a boulder to help it climb, are not instances of tool use because they do not involve a triadic relation. In both scenarios, the subject may be skillfully dealing with an object, but they are not using a mediating object to alter a target object.Footnote ¹ Nest building also falls in this category.

To best understand the accounts of Hall, Goodall, and Alcock, we will focus on how they answer these two questions: How must an object be used for it to qualify as a tool? What counts as a target object? As we will see, despite some divergences among them, their conceptions overlap in most cases.

3. Contemporary tool use definitions

For an example of a contemporary definition of tool use, we will focus on Shumaker et al. (Reference Shumaker, Walkup and Beck2011), which is an updated version of Beck’s (Reference Beck1980) definition. We highlight their definition because it is widely considered the main point of reference in the contemporary tool use literature. Shumaker et al. (Reference Shumaker, Walkup and Beck2011) criticize the classical approach to defining tool use along several fronts. For example, they point out that the classical account can be overly permissive and will include as tool use cases like a rat that learns through operant conditioning to press a lever to obtain a reward. Since only instances of manipulation that involve dexterity and control over an object are generally considered tool use, this is a problematic inclusion. Classical accounts can also be overly restrictive. If the sponge a dolphin is using is alive, then at least some classical definitions of tool use (such as Alcock’s) will exclude this as a case of tool use. In response to these and other limitations, Shumaker et al. define tool use as:

This definition maintains the same kind of tripartite dynamic captured by Hall, Goodall, and Alcock, but offers a more precise characterization of the tool, its use, and the target object.

Under Shumaker et al.’s definition, any manipulable external object can be a tool, even living or deceased organisms. For instance, macaques sometimes grasp and hold infants when charged by aggressive conspecifics. This behavior, called agonistic buffering (Deag and Crook Reference Deag and Crook1971), prevents or lessens the severity of the attack. In such a case, the baby is used as a tool (a shield). Additionally, tools can be unattached or partially detached from their location and can be externally or internally manufactured (allowing feces or exuviae to be used as tools).

The definition also takes a permissive approach to target object qualification. Any object whose condition, position, or form can be altered via an appropriate mediating object counts as a target object. Targets can include the organism itself, other animals, or environmental objects. The target could be a hermit crab that changes its condition by inhabiting a shell, a digger wasp burrow whose form has been altered after having been hammered with pebbles, or a fly that falls into the water after having been shot by an archerfish.

While Shumaker et al. are permissive about what can be a tool or target object, their definition excludes some cases that they identify as problematically included by classical definitions, such as lever pressing. Despite such advantages over classical accounts, Shumaker et al.’s definition is not immune to critique. Critics have mostly focused on the exclusion of some behaviors that they hold should be considered tool use, such as using objects to mediate information flow (Crain et al. Reference Crain, Giray and Abramson2013; Cenni and Leca Reference Cenni, Leca, Vonk and Shackelford2020). However, Shumaker et al. do include social signal modification as a possible function of tool use, as well as tools more generally used for directing the flow of information.Footnote ³ Examples of information-focused tool use include gorillas checking stream depth with a stick and orangutans concealing themselves using a branch (Breuer et al. Reference Breuer, Ndoundou-Hockemba and Fishlock2005; St Amant and Horton Reference St Amant and Horton2008). When a gorilla checks the depth, there is no external object they aim to alter (as is necessary for tool use under Schumaker et al.’s definition), but the tool is simply used to gain information about the water’s depth.Footnote ⁴

There is also the symbolic function of tool use. For example, consider the case of an orangutan who repeatedly handed a coconut to humans. When the humans kept giving it back, the orangutan held a stick and was claimed to have mimicked humans opening the coconut with a knife (Russon Reference Russon, Simon and Kevin2003). Although they do not present a compelling explanation as to why it should count as tool use, Shumaker et al. (Reference Shumaker, Walkup and Beck2011) interpret this event as a case of a tool used to symbolize and communicate. The coconut is not (or need not be) altered for the signal to be effective. But perhaps it is the observing human and not the coconut that is the object. In this case, tool use would be effective if it altered the human, leading them to open the coconut.

Shumaker et al.’s definition classifies some actions in ways that seem hard to justify. While our focus here is non-human animal tool use, flaws in their framework can be highlighted through cases of human tool behavior. Consider the distinction between spear hunting, bow hunting, and rifle hunting. The first two are tool use under their definition, while the latter is classified as construction.Footnote ⁵ As such, using a gun is classified with nest building, not with bow or spear hunting. Shumaker and colleagues explain this distinction by noting that the shooter is not entirely responsible for all the connections between the rifle, bullet, and animal, and does not directly manipulate the object in its entirety (i.e., it does not manipulate the bullet). Certainly, there are evident mechanical differences between these performances, but we believe that hunting with a rifle is closer to spear hunting—hence tool use—than nest building.

One can quibble about other questionable classifications of Shumaker et al.’s definition and one can offer further refinements to their definition that allow for better classifications of tool use behavior. But instead of considering such refinements, we will now turn to an account of tool behavior that calls not just Shumaker et al.’s definition into question, but challenges the entire history of tool use definitions and the possibility of obtaining a scientifically useful definition.

4. Limitations of tool use as a scientific category

A rising tide of literature argues that tool use definitions are rather limited when it comes to generalizing, explaining, or predicting ecological, psychological, or evolutionary phenomena, especially in comparative studies (Mangalam and Fragaszy Reference Mangalam and Fragaszy2016; Cenni and Leca Reference Cenni, Leca, Vonk and Shackelford2020; Colbourne et al. Reference Colbourne, Auersperg, Lambert, Huber and Völter2021; Heersmink Reference Heersmink2022). In the face of these critiques, some have called for an approach to tool use that is more deeply connected to the ecology, psychology, and neurology of the behavior (Jacobs and Osvath Reference Jacobs, Osvath, Weekes-Shackelford, Todd and Viviana2016; Fragaszy and Mangalam Reference Fragaszy, Mangalam and Naguib2018; Mangalam et al. Reference Mangalam, Fragaszy, Wagman, Day, Kelty-Stephen, Bongers, Stout and Osiurak2022). Such critiques are not directed at Shumaker et al. (Reference Shumaker, Walkup and Beck2011) or any other specific tool use definition, but at the overall framework through which animal tool use has been conceived and its value as a scientific category.

Traditional approaches to defining tool use aim for a concept free from anthropocentric bias and broad enough to encompass all instances of what they consider to be tool use. In crafting such definitions, scholars have increasingly aimed to decouple the classification of behavior as tool use from the associated skills, ecology, and psychology of the tool users. As exemplified by Shumaker et al., all that is required for tool use is a kind of object-mediated interaction. So long as this requirement is fulfilled, any object can be a tool or a target. While the traditional approach to tool use preserves inclusivity and allows for a rather efficient individuation of what qualifies as tool use, this inclusivity appears to have a downside. As Fragaszy and Mangalam (Reference Fragaszy, Mangalam and Naguib2018) argue, by divorcing tool use from specific mechanical, anatomical, neurological, or ecological parameters, the ability to use the category of tool use for explanatory and predictive power is undermined:

What does this mean more concretely? Tool use in animals has long been considered to be a proxy for intelligence and cognitive abilities, such as problem solving, causal understanding, and goal directedness (Thorpe Reference Thorpe1963; Parker and Gibson Reference Parker and Gibson1977; Peacocke Reference Peacocke, McCormack, Hoerl and Butterfill2011; Woodward Reference Woodward, McCormack, Hoerl and Butterfill2011). However, its value for these tasks has been questioned, especially in light of empirical research indicating that sophisticated cognition is not needed for many forms of tool use (Shettleworth Reference Shettleworth2010; Emery Reference Emery, Crickette, Call and Boesch2013; Cenni and Leca Reference Cenni, Leca, Vonk and Shackelford2020).

One example of evidence for this comes from an experiment by Teschke and colleagues (Reference Teschke, Cartmill, Stankewitz and Tebbich2011), in which the problem solving abilities of the tool-using member of the woodpecker finch, Cactospiza pallida, were tested against non-tool-using conspecifics (tool use in this species varies across populations, habitat, and food abundance) as well as Camarynchus parvulus, a related tree finch species that does not engage in tool use. The groups were challenged to solve four problems designed to test different aspects of cognition, including problem solving and general learning abilities. The hypothesis was that the tool user group would learn to solve these novel problems more quickly than the other groups, given that the tasks resembled their natural tool use scenarios. Contrary to the experimenters’ predictions, the group of tool users did not outperform the non-tool users in the physical cognition tasks and excelled in only one general learning task.

Their study suggests that tool users are not necessarily more intelligent than non-tool users, and that higher cognitive skills are not required for all forms of tool use. However, other studies highlight cases of tool use linked to higher cognitive abilities (Mendes et al. Reference Mendes, Hanus and Call2007). Thus, the relationship between tool use and cognition is less straightforward than has been imagined in the past (Emery and Clayton Reference Emery and Clayton2009). Because of this, it may be difficult to make generalizations about the cognitive characters of tool users from the mere fact that they use tools.

Reports of the lack of explanatory and predictive value of tool use are not confined to the cognitive sciences. Colbourne et al. (Reference Colbourne, Auersperg, Lambert, Huber and Völter2021) claim that the neutrality (with respect to the neurological and psychological underpinnings) of traditional approaches of tool use undermines its value for interspecific comparative analyses. They also claim that traditional approaches to tool use fail to offer insights into early human evolution based on the study of animal tool use. The neutrality of the traditional definitions—often considered an advantage and a way of avoiding anthropomorphism—thus appears to drain the category of tool use of scientific value.

In light of this apparent lack of scientific value, some scholars have begun to work on novel ways of conceptualizing animal tool behavior (Mangalam et al. Reference Mangalam, Fragaszy, Wagman, Day, Kelty-Stephen, Bongers, Stout and Osiurak2022). In the following section, we will focus on tooling (Fragaszy and Mangalam Reference Fragaszy, Mangalam and Naguib2018), an approach gaining currency in the empirical literature on nonhuman animals (Cenni and Leca Reference Cenni, Leca, Vonk and Shackelford2020; Colbourne et al. Reference Colbourne, Auersperg, Lambert, Huber and Völter2021; Bastos et al. Reference Bastos, Horváth, Webb, Wood and Taylor2021; Jacobs and Osvath Reference Jacobs and Osvath2023).

5. Tooling

Tooling is not a mere iteration in the attempts to define tool use. Instead, it breaks with tradition and represents a rejection and replacement of tool use as a scientific category. In doing so, tooling offers a novel analytical approach to object-mediated instrumental action. Fragaszy and Mangalam (Reference Fragaszy, Mangalam and Naguib2018) define tooling as:

The domain of tooling so defined only partially overlaps with tool use as traditionally defined. Like tool use, tooling is an object-mediated instrumental behavior. In particular, it is an instance of problem solving that involves a relationship among a subject, a mediating object, and a target object. However, influenced by movement sciences and ecological psychology, tooling is centered on the analysis of specific kinematic, perceptuomotor, and spatiotemporal features of a particular form of object-mediated instrumental action. Thanks to this focus, tooling offers a novel conceptualization and categorization of both the relationship between the subject and the mediating object and between these and the target object. Proponents of tooling argue that because of its grounding in ecological psychology and the kinematic sciences, tooling has better explanatory and predictive power than the traditional tool use category.

According to Fragaszy and Mangalam (Reference Fragaszy, Mangalam and Naguib2018), when a user grasps an object to reach a target, the tool becomes a part of the subject’s body, transforming the dyadic user–object relationship into one system: the body-plus-object system. This embodiment of the tool has a double effect on the user: First, it influences what the subject perceives as affordances. Second, it alters how it coordinates and interacts with its environment. These changes are most clearly observed if one focuses on the mechanical and psychological consequences of the embodiment (Mangalam and Fragaszy Reference Mangalam and Fragaszy2016).

While traditional approaches to tool use suggest that tools simply extend the reach, strength, or range of an animal’s movement, Mangalam and Fragaszy (Reference Mangalam and Fragaszy2016) propose that we conceive the relationship between the body and the tool in terms of degrees of freedom (DoFs). In particular, they stress that during tooling, the animal’s body acquires at least one novel DoF. The concept of degree of freedom is derived from a kinematic approach to locomotion (Bernstein Reference Bernstein1967) and it is used to map out how a rigid object moves in a three-dimensional space. An object can change its position by moving along three axes: forward–backward (surge), up–down (sway), and left–right (heave). In addition, it can rotate on each of these axes. Thus, we have a total of six DoFs. As an example, if we assume that hand, elbow, arm, and forearm form one block, then raising an arm involves only one DoF, namely the rotation of the arm on the “left–right” axis, which is made possible by the shoulder.

In the interaction with a tool, a subject acquires at least one novel DoF (and thus the possibility of a novel movement) since there is one more rigid item that composes the body (now body-plus-object system). This acquisition triggers a redistribution of the overall DoFs, since using a tool changes the way the subject coordinates and manages their own body and, consequently, how it interacts with the environment (Fragaszy and Mangalam Reference Fragaszy, Mangalam and Naguib2018). An example will help to elucidate this point. Grasping a screwdriver enriches one’s movement repertoire by adding at least one novel DoF since one can now rotate a screw to drive it in or remove it. A series of conditions need to obtain for this action to be successful: the fingers need to be firmly wrapped around the handle of the screwdriver, which needs to be more or less parallel to the screw, and the arm has to be extended. These conditions correspond to constraints that the new action imposes on the body. It is in this sense that tooling involves the acquisition of novel movements and the reorganization of the extant DoFs.

In addition to these mechanical consequences, the transformation of the body into the body-plus-object system carries psychological significance. Fragaszy and Mangalam (Reference Fragaszy, Mangalam and Naguib2018) suggest that tooling entails a change in how tool users perceive the boundaries of their own bodies. In particular, the interaction with a tool extends the perceptual range of the subject, since the subject’s “end effector” (the part of the system that interacts with the environment) is now the end of the tool, and not the end of the body. This phenomenon is called distalization of the end effector (Arbib et al. Reference Arbib, Bonaiuto, Jacobs and Frey2009) and has been assessed as a crucial factor for a correct representation of the user–tool relationship (Miller et al. Reference Miller, Luca Montroni, Salemme, Hayward and Farnè2018; Osiurak and Federico Reference Osiurak and Federico2021; Mangalam et al. Reference Mangalam, Fragaszy, Wagman, Day, Kelty-Stephen, Bongers, Stout and Osiurak2022). It consists of the transfer of the locus of perception from the boundaries of the body to the boundaries of the tools via the conduction of mechanical feedback from the tool to the body. In other words, since the user seems to sense with the tool itself, the user perceives the tool to be part of its body.

Finally, Fragaszy and Mangalam’s approach introduces a new strategy for analyzing and categorizing the interaction between the body-plus-object system and the target object. This interaction is conceived as a spatial relationship: the body-plus-object system interacts with at least one target object, thus creating a spatial connection between the two. The assessment of this relationship is based on different spatial features including the egocentric or allocentric framework of reference, the number of spatial relations, the temporal dynamic, and the specificity or precision in the action required to succeed in tooling.

The determination of these precise kinematic and perceptuomotor parameters, left unspecified in the literature on tool use, radically transforms the individuation and classification of object-mediated instrumental actions. By providing finer, measurable criteria across multiple dimensions, this approach yields more precise and interpretable data, leading to analyses that are more robust and replicable. Improved replicability is a critical advancement for cognitive and behavioral research, particularly in a field plagued by a replication crisis and historically reliant on anecdotal and sparse observations of potential tool use instances.

These parameters also enable more consistent grouping of behavioral instances, but also lead to the exclusion of many actions traditionally categorized as tool use, such as projectile throwing. Projectile throwing involves perceptuomotor features, spatial relationships, and organismal and environmental constraints that fall outside of tooling. For example, the distalization of the end effector—a hallmark of the body-plus-object system in tooling—is absent in projectile throwing, as is the specific perception of environmental constraints central to tooling behaviors.

In summary, the tooling framework introduces a methodology grounded in specific kinematic and psychological markers. This allows for more precise measurements of tool behavior, more coherent comparisons across species, and better interpretability of findings, ultimately offering promising advancements for the study of object-mediated instrumental actions and their associated cognitive and neurological dimensions. The apparent advantages of the tooling approach of Fragaszy and Mangalam (Reference Fragaszy, Mangalam and Naguib2018), and the limitations of the traditional tool use framework, lead us to ask whether we should simply replace tool use with tooling.

6. Should tooling replace tool use?

Although tooling is offered as an alternative to tool use, it has limitations that make it problematic as a complete replacement.

First, since tooling is restricted to direct mechanical interactions between a body-plus-object system and a target object, much of what has been traditionally understood as tool use is excluded. Behaviors like projectile throwing, dropping, and baiting require that the tool be released or detached from the subject during its use, which tooling forbids. Some of the excluded behaviors represent the very foundation of the interest in animal tool behavior. For Hall (Reference Hall1963), primates breaking off and throwing branches to scare off intruders was taken to be a standard of what tool use means. Along these lines, van Lawick-Goodall refers to instances of birds “picking up stones to drop or throw on hard shelled food” (1971, 198) as paradigmatic cases of tool use. Recent tool use discoveries include tool behavior that does not require the grasping of the tool, such as crocodilians positioning small sticks on top of their snout to lure birds (Dinets et al. Reference Dinets, Brueggen and Brueggen2015).

Being restrictive is especially problematic for cases of behaviors that are excluded from tooling yet are the same—under some descriptions—as ones included in tooling. For instance, consider applying one object to another. This can be manifested by different forms of interaction between the subject, median object, and target object. Some cases meet the requirements for tooling, while others do not. One that meets tooling criteria is hyacinth macaws’ application of wood wedges or leaves around nuts placed in their beaks to facilitate nut cracking (Borsari and Ottoni Reference Borsari and Ottoni2005). By contrast, a squirrel applying a snakeskin to their fur to repel predators is not considered tooling (Clucas et al. Reference Clucas, Owings and Rowe2008). Despite the latter not being tooling, both are examples of problem solving achieved by applying an object on a target (namely, the nut and the squirrel itself). The sameness of these behaviors is not reflected by tooling and it is therefore possible that tooling is masking useful similarities.

By excluding many instances of tool use, the tooling approach risks losing sight of the broader goal of the traditional approaches to tool use research. Tool use is not merely about the mechanical action of manipulating an object to act on a target. It is also about understanding the possibilities and consequences created by manipulating an object in a specific way and context (Baber Reference Baber2006; Baber et al. Reference Baber, Parekh and Cengiz2014). This ability is expressed in diverse forms across various contexts. While these expressions differ in their biomechanical and psychological details, they all share a connection to the overarching ability that tool use represents. Traditional approaches to tool use, despite their shortcomings in providing detailed parameters for biomechanical and psychological analysis, succeed in capturing this generality.

If we accept that tool use represents a general ability, it becomes essential to differentiate it from tooling. Tooling, while offering valuable insights, cannot be equated with the broader concept of tool use. Rather, it represents a specific instance—a particular pattern of interaction that exemplifies one way in which the general ability is expressed. This distinction highlights a key point: tool use and tooling target different explanatory domains. Tool use focuses on the overarching ability to manage objects to achieve goals, whereas tooling emphasizes the specific kinematic and perceptuomotor dynamics involved in one form of tool use.

Rather than viewing tool use and tooling as competing frameworks, they should be seen as complementary. Tool use captures the breadth of the ability to recognize and use objects to achieve goals, while tooling offers the analytical precision needed to investigate the dynamics and features of a specific kind of tool use behavior. Instead of choosing one or the other, we will argue below that we can retain both, allowing us to illuminate the diversity of tool use behaviors while at the same time articulating the specific cognitive and neurological mechanisms underlying its various forms, paving the way for more unified and insightful research.

In sum, while the tooling framework of Fragaszy and Mangalam (Reference Fragaszy, Mangalam and Naguib2018) carries many predictive and explanatory strengths, viewed as a replacement for the category of tool use, it has shortcomings. One of the biggest limitations is the fact that the domain of tooling is far smaller than the category of tool use. Faced with the benefits and drawbacks of both tooling and tool use, the question is whether we need to choose one or the other, or if there is a way of retaining the best of both. In the following section, we propose a way of building on the tooling framework in order to capture more of the domain of tool use, but in a way that retains the precision of tooling.

7. Synthesis

Animals use objects in a variety of manners and for a diversity of ends. Sometimes they maintain a grasp on the object and use it to physically manipulate a target object (in line with tooling), but other times the object is used for informational purposes or is released from grasp (St Amant and Horton Reference St Amant and Horton2008). These latter uses are excluded from tooling but fall within the broader category of tool use. We see the appeal of the tool use category but recognize the validity of the criticisms arising from the tooling camp. In light of this, we seek a synthesis.

To synthesize the accounts, we propose to distinguish two modes of interaction between the subject and the tool (retention and releasing) and two modes of interaction between the tool and the target (mechanical and informational). Retention refers to when the object is held throughout the duration of its use. Releasing, by contrast, occurs when an object is discharged from the direct physical control of the subject, though its movement in space and/or its orientation may be determined by the user. The mechanical-informational distinction is between object-mediated instrumental actions that generate their effect on the target object through direct contact and actions that generate their effect via changing the flow of information (between the subject and the target or between the target and its environment). These two distinctions can be ordered into a two-dimensional array, resulting in four categories of tool use (see table 1).

Let us consider each of these four forms of tool use, beginning with retention-mechanical. This category corresponds with tooling as spelled out by Fragaszy and Mangalam (Reference Fragaszy, Mangalam and Naguib2018). Thus, all the forms of tool use that fall into this category display the defining features of tooling. Examples include New Caledonian crows using twigs and other objects to extract prey from crevices (Hunt and Gray Reference Hunt and Gray2004) and capuchin monkeys using stones to dig up roots (Mannu and Ottoni Reference Mannu and Ottoni2009).

Next consider retention-informational tool use, which involves object-mediated instrumental actions performed with the aim of changing the flow of information (between the tool user, the target object, and/or other subjects). Gorillas using sticks to learn about the depth of water falls within this category (Breuer et al. Reference Breuer, Ndoundou-Hockemba and Fishlock2005). Interestingly, this behavior involves the formation of a body-plus-object system, some degree of distalization of the end effector, and a direct interaction of the body-plus-object system with a target surface. However, this does not count as tooling since the action is not performed to modify the stream or pond bed, but to serve as an extension of the gorilla’s arm, thereby letting information about the water flow through the tool to the gorilla. Thus, it is a mechanical interaction that satisfies an informational goal.

Next, we have releasing-mechanical tool use. While this category falls outside of tooling, the same kind of kinematic, perceptuomotor approach used in tooling can be applied to this form of tool use. We expect that commonalities among different instances of projectile throwing or dropping will be found by investigating this kind of tool use. This category includes many of the classic cases of projectile throwing such as Egyptian vultures launching rocks onto eggs (van Lawick-Goodall Reference Van Lawick-Goodall, Daniel, Robert and Shaw1971) and archerfish spitting water at insects to make them fall into water (Schuster et al. Reference Schuster, Wöhl, Griebsch and Klostermeier2006).

Finally, we have releasing-informational tool use. This is farthest from tooling, since it involves neither the continual grasp of the tool, nor the tool’s use for mechanical interactions with the target. It is nevertheless an important category of tool use. Unaimed throwing frequently occurs during displays by male chimpanzees. They toss branches, sticks, stones, sand and grass without specific targets during agonistic displays. This behavior has been interpreted as communicating the individual’s readiness to respond to intra- and interspecific threats (Goodall Reference Goodall1986; Lombardo and Deaner Reference Lombardo and Deaner2018). Another example is that of California ground squirrels (Spermophilus beecheyi) throwing material to keep snakes from discovering their burrows (Coss and Biardi Reference Coss and Biardi1997).

Our synthesis acknowledges the diversity of tool use, understood as the ability to identify objects as means for action and to manipulate the objects to act on a target. At the same time, we recognize the need to categorize different forms of tool behaviors based on specific biomechanical features. The approach of tooling has the advantage of tying tool behavior to specific mechanisms, but excludes most tool use behavior. Our synthesis extends the methodological advancements proposed by the approach of Fragaszy and Mangalam (Reference Fragaszy, Mangalam and Naguib2018) to encompass forms of object-mediated instrumental action not including within tooling.

This framework takes tooling to be a kind of tool use together with three additional categories. The sum of these four categories thus retains the advantages of tooling but no longer excludes most of what is traditionally considered tool use. Note that we say most of tool use. While we believe that the vast majority of documented tool use cases fall within our four categories, we do not claim that every instance of tool use does. It is possible that some tool behavior falls outside of our framework. If so, additional categories could be introduced without undermining the relevance of our framework.

We are optimistic that generalizations within each of the four tool use categories will be more robust than generalizations across tool use as a whole. Thus, this synthesis is a way of keeping the general category of tool use but dividing it into kinds to better serve scientific needs. In particular, tooling’s focus on specific kinematic, perceptuomotor, and spatiotemporal features provides a solid foundation for generating reliable predictions, comparisons, and generalizations. Our synthesis broadens the scope of these features to include instances of tool use that the tooling framework overlooks, thereby expanding the epistemic benefits it offers. Furthermore, by grounding tool use on a firmer conceptual foundation, we not only enhance our understanding of tool behaviors but also pave the way for more effective studies of the cognitive abilities associated with tool use, such as future planning.