Kline correctly points out that the definition of teaching is a contentious issue. For many anthropologists, teaching is necessarily a human endeavor, while for most animal researchers functional definitions of teaching allow them to investigate the phenomenon without being shackled by assumptions of theory of mind or language use. From a theoretical evolutionary perspective, I find myself more aligned with the animal behaviorists, whose set of three simple but relevant conditions to identify teaching (Caro & Hauser Reference Caro and Hauser1992) lend themselves neatly to mathematical treatment (Aoki et al., in preparation; Fogarty et al. Reference Fogarty, Strimling and Laland2011). In order to identify the evolutionary roots of teaching, both comparative animal studies and mathematical modeling will be necessary.
The definition proposed by Caro and Hauser (Reference Caro and Hauser1992) states that teaching occurs when a role model changes their behavior in the presence of a pupil, the role model suffers a cost for this change in behavior (or does not immediately benefit from it), and the pupil learns information more rapidly than their otherwise would, or learns information that their otherwise would not. This operational definition does face many challenges. For example, it is difficult to incorporate nuanced differences in teaching types into such a basic framework. As a result, it proves difficult to account for the fact that humans possess a uniquely well-developed capacity for teaching that is absent in other species. The delineations of types of teaching that the author suggests, along with previous categorizations of teaching behavior (Hoppitt et al. Reference Hoppitt, Brown, Kendal, Rendell, Thornton, Webster and Laland2008; Thornton & Raihani Reference Thornton and Raihani2008) allow researchers to investigate what is unique about human teaching, without claiming that nonhuman animals cannot teach.
However, new delineations of teaching types are not useful in the absence of a clear and widely accepted definition of teaching in its most basic form. In other words, we must be able to identify teaching before we can meaningfully categorize and investigate it. All suggested modes of teaching appearing in Kline's proposed framework adhere to Caro and Hauser's functional definition of teaching. As such, the same divisive problems pointed out by Byrne and Rapaport (Reference Byrne and Rapaport2011) and addressed by Thornton and McAuliffe (Reference Thornton and McAuliffe2012), for example, remain unsolved in this framework, as in others.
The definition proposed by Caro and Hauser (Reference Caro and Hauser1992) has certain problems. Indeed, Caro and Hauser themselves recommended that their definition be altered as more information about animal teaching came to light. For example, learning on the part of the pupil is difficult to show without experimental manipulation. Such manipulation was possible in the case of meerkats (Thornton & McAuliffe Reference Thornton and McAuliffe2006) and ants (Franks & Richardson Reference Franks and Richardson2006), but has proven difficult in the case of large mammals and protected species in the wild with whom it may be unethical and impractical to experiment in such a way. Therefore, a definition of teaching that is applicable to such animals (and to humans) would be of great use to the community of researchers interested in animal learning.
Here, Kline accepts that the definitions of teaching currently in use are not sufficient. I agree with the author that the definitions from each field are problematic. However, I disagree that the functional approach gains from integration with the mentalistic approach, which requires discerning the intent of the teacher, or with the culture-based approach, which in extreme cases claims that only Western teaching is true teaching. Indeed, some of the criticisms leveled against the use of functional definitions in this article are the direct result of such definitions being underused (“there is a lack of systematic empirical work documenting that teaching happens across all human societies”; target article, sect. 2.4, para. 6). The functional approach allows for a standardized, scientific study of teaching in non-verbal animals as well as in humans. Whatever definition researchers eventually agree upon, this feature must not be compromised. Therefore, it seems that rather than attempting to combine functional definitions with mentalistic definitions that are essentially unusable in the identification of nonhuman teaching, a better approach might be to alter the functional definition of teaching so that the criteria can be met by close observation alone. Thornton and Raihani (Reference Thornton and Raihani2008) proposed that it might be possible to find a correlation between the extent of exposure to teaching behavior and the skill of a pupil's performance. However, as Byrne and Rappaport (Reference Byrne and Rapaport2011) point out, in a case where more teaching effort is directed toward slower learners, such a correlation would not exist.
A looser set of criteria to show that learning has occurred might be that the pupil should exhibit a behavior after teaching that they did not exhibit before teaching, exhibit a behavior in a context in which it did not before teaching, or exhibit a behavior with considerably greater skill or accuracy than it did before teaching. This removes the specification that we must know what would happen in the absence of teaching, and replaces it with the specification that we must know the behavioral repertoire of each pupil prior to, and after, a teaching event. Such definitions could yield false positives. For example, if an individual is likely to improve their performance of a particular task over time, it might be difficult to disentangle the effects of teaching and the effects of experience and age.
An alternative approach could focus on the change in skill after bouts of learning, rather than on absolute measures of performance. It is intuitive that teaching will not be adaptive when either inadvertent social learning or asocial learning is more effective. Therefore, it may be sensible to believe that an increase in skill associated with a teaching event should be greater than an increase in skill associated with either watching a task being performed or attempting to perform it with no role model for any given individual. This difference may be identifiable through observation alone.
To investigate teaching in a meaningful way, the core aspects of the functional definition must remain intact. However, the shortcomings that exclude very probable cases of teaching in wild animals must also be carefully considered. Once this problem of a basic, workable definition by which we can identify teaching is addressed, the teaching taxonomy suggested by Kline should yield interesting insights into both human and nonhuman teaching.
Kline correctly points out that the definition of teaching is a contentious issue. For many anthropologists, teaching is necessarily a human endeavor, while for most animal researchers functional definitions of teaching allow them to investigate the phenomenon without being shackled by assumptions of theory of mind or language use. From a theoretical evolutionary perspective, I find myself more aligned with the animal behaviorists, whose set of three simple but relevant conditions to identify teaching (Caro & Hauser Reference Caro and Hauser1992) lend themselves neatly to mathematical treatment (Aoki et al., in preparation; Fogarty et al. Reference Fogarty, Strimling and Laland2011). In order to identify the evolutionary roots of teaching, both comparative animal studies and mathematical modeling will be necessary.
The definition proposed by Caro and Hauser (Reference Caro and Hauser1992) states that teaching occurs when a role model changes their behavior in the presence of a pupil, the role model suffers a cost for this change in behavior (or does not immediately benefit from it), and the pupil learns information more rapidly than their otherwise would, or learns information that their otherwise would not. This operational definition does face many challenges. For example, it is difficult to incorporate nuanced differences in teaching types into such a basic framework. As a result, it proves difficult to account for the fact that humans possess a uniquely well-developed capacity for teaching that is absent in other species. The delineations of types of teaching that the author suggests, along with previous categorizations of teaching behavior (Hoppitt et al. Reference Hoppitt, Brown, Kendal, Rendell, Thornton, Webster and Laland2008; Thornton & Raihani Reference Thornton and Raihani2008) allow researchers to investigate what is unique about human teaching, without claiming that nonhuman animals cannot teach.
However, new delineations of teaching types are not useful in the absence of a clear and widely accepted definition of teaching in its most basic form. In other words, we must be able to identify teaching before we can meaningfully categorize and investigate it. All suggested modes of teaching appearing in Kline's proposed framework adhere to Caro and Hauser's functional definition of teaching. As such, the same divisive problems pointed out by Byrne and Rapaport (Reference Byrne and Rapaport2011) and addressed by Thornton and McAuliffe (Reference Thornton and McAuliffe2012), for example, remain unsolved in this framework, as in others.
The definition proposed by Caro and Hauser (Reference Caro and Hauser1992) has certain problems. Indeed, Caro and Hauser themselves recommended that their definition be altered as more information about animal teaching came to light. For example, learning on the part of the pupil is difficult to show without experimental manipulation. Such manipulation was possible in the case of meerkats (Thornton & McAuliffe Reference Thornton and McAuliffe2006) and ants (Franks & Richardson Reference Franks and Richardson2006), but has proven difficult in the case of large mammals and protected species in the wild with whom it may be unethical and impractical to experiment in such a way. Therefore, a definition of teaching that is applicable to such animals (and to humans) would be of great use to the community of researchers interested in animal learning.
Here, Kline accepts that the definitions of teaching currently in use are not sufficient. I agree with the author that the definitions from each field are problematic. However, I disagree that the functional approach gains from integration with the mentalistic approach, which requires discerning the intent of the teacher, or with the culture-based approach, which in extreme cases claims that only Western teaching is true teaching. Indeed, some of the criticisms leveled against the use of functional definitions in this article are the direct result of such definitions being underused (“there is a lack of systematic empirical work documenting that teaching happens across all human societies”; target article, sect. 2.4, para. 6). The functional approach allows for a standardized, scientific study of teaching in non-verbal animals as well as in humans. Whatever definition researchers eventually agree upon, this feature must not be compromised. Therefore, it seems that rather than attempting to combine functional definitions with mentalistic definitions that are essentially unusable in the identification of nonhuman teaching, a better approach might be to alter the functional definition of teaching so that the criteria can be met by close observation alone. Thornton and Raihani (Reference Thornton and Raihani2008) proposed that it might be possible to find a correlation between the extent of exposure to teaching behavior and the skill of a pupil's performance. However, as Byrne and Rappaport (Reference Byrne and Rapaport2011) point out, in a case where more teaching effort is directed toward slower learners, such a correlation would not exist.
A looser set of criteria to show that learning has occurred might be that the pupil should exhibit a behavior after teaching that they did not exhibit before teaching, exhibit a behavior in a context in which it did not before teaching, or exhibit a behavior with considerably greater skill or accuracy than it did before teaching. This removes the specification that we must know what would happen in the absence of teaching, and replaces it with the specification that we must know the behavioral repertoire of each pupil prior to, and after, a teaching event. Such definitions could yield false positives. For example, if an individual is likely to improve their performance of a particular task over time, it might be difficult to disentangle the effects of teaching and the effects of experience and age.
An alternative approach could focus on the change in skill after bouts of learning, rather than on absolute measures of performance. It is intuitive that teaching will not be adaptive when either inadvertent social learning or asocial learning is more effective. Therefore, it may be sensible to believe that an increase in skill associated with a teaching event should be greater than an increase in skill associated with either watching a task being performed or attempting to perform it with no role model for any given individual. This difference may be identifiable through observation alone.
To investigate teaching in a meaningful way, the core aspects of the functional definition must remain intact. However, the shortcomings that exclude very probable cases of teaching in wild animals must also be carefully considered. Once this problem of a basic, workable definition by which we can identify teaching is addressed, the teaching taxonomy suggested by Kline should yield interesting insights into both human and nonhuman teaching.