Penn et al. propose a pervasive domain-general cognitive discontinuity that defines the difference between “us and them.” In doing so, we believe Penn et al. have inadvertently argued something akin to cognitive recapitulation. In many ways, human ontogeny of the cognitive abilities they discuss appears to recapitulate phylogeny, as young human children seem to display the same lack of relational insight that the authors identify in nonhuman primates. Leaving debate about recapitulation theory aside, we wondered how we might apply their discontinuity hypothesis to development within our own species. Here, we examine whether the development of relational understanding is discontinuous within two domains mentioned by Penn et al.: spatial cognition (particularly, success on scale model tasks) and social cognition.
Children younger than age 3 years have difficulty using a scale model of a room as a source of information about the location of a hidden toy in the analogous, larger room (e.g., DeLoache Reference DeLoache1989). In contrast, 3-year-olds succeed on the task as long as the locations are unique, but they fail when they cannot use object correspondences between the model and the room (Blades & Cooke Reference Blades and Cooke1994). Only by 5 years of age can children use the spatial relationships among identical locations in the scale model to find the toy in the room, thereby achieving the criteria set by Penn et al. One interpretation of the performance of 3-year-olds is that they are matching perceptual similarities between items in the model and the room, without understanding the relation between the two (Perner Reference Perner1991). However, this seems unlikely. For example, 2-year-olds can match corresponding items in the model and room, even when they cannot find the hidden toy in the room based on the hiding event in the model (Troseth et al. Reference Troseth, Pickard and Deloache2007). Evidently, perceptual matching alone is not enough to promote success in the scale model task. Instead, accuracy in the model task, even with unique locations, might require at least some understanding of the relation between the room and the model.
Evidence from theory of mind tasks may potentially offer a similar developmental trajectory. After the 3 years of age, children start to show evidence of representing behavior in terms of mental states, and by 5 years of age they can understand another's false belief as a mental misrepresentation (Wellman & Liu Reference Wellman and Liu2004). Although traditionally children below the age of 3 have not been credited with reasoning about mental states, infants have shown success on tasks ranging from understanding goal-directed actions (Woodward Reference Woodward1998) to predicting behavior based on another's perception (Luo & Baillargeon Reference Luo and Baillargeon2007), as well as false belief (Onishi & Baillargeon Reference Onishi and Baillargeon2005). We agree with Penn et al. that infant performance may be due to rule-based, rather than mental state, reasoning; however, given the breadth and flexibility displayed by infants in these tasks, it seems likely that such rules are organized within some sort of higher-order relational framework. Infants will respond similarly to a series of disparate goal-directed actions, including grasping, pointing, reaching, and looking (Woodward et al. Reference Woodward, Sommerville, Guajardo, Malle, Moses and Baldwin2001). Additionally, such responses seem to be modified correctly based on another's current and past visual access (Luo & Baillargeon Reference Luo and Baillargeon2007; Meltzoff & Brooks Reference Meltzoff, Brooks, Flom, Lee and Muir2007), previous interactions with other individuals (Kuhlmeier et al. Reference Kuhlmeier, Wynn and Bloom2003), individual versus shared knowledge or preferences (Buresh & Woodward Reference Buresh and Woodward2007; Song et al. Reference Song, Baillargeon and Fisher2005), and updated representations of otherwise meaningless actions based on context (Gergely et al. Reference Gergely, Bekkering and Kiraly2002; Kiraly et al. Reference Kiraly, Jovanovic, Prinz, Aschersleben and Gergely2003). A rule-based account that did not allow for minimal relational reasoning would, in our view, struggle to explain such flexibility.
Yet, it appears that children under 3 years of age cannot achieve the level of relational insight put forward by Penn et al. as the hallmark of human cognition. Would we consider their abilities to be discontinuous with the abilities seen at age 5? Although achieving analogical thinking is a clear developmental change (Gentner Reference Gentner, Gentner and Goldin-Meadow2003), it does not seem to qualify as discontinuous in a strong sense. Young children seem to be able to reason about unobservable explanatory mechanisms as well as map simple relations between a representation and reality, whereas more abstract relational understanding occurs later in development.
If, in this case, we are to claim that human ontogeny is continuous, how does that claim speak to phylogeny? In comparison, nonhuman primates achieve success on the same tasks in which we think children are using basic relational understanding. Kuhlmeier and Boysen (Reference Kuhlmeier and Boysen2002), for example, found that chimpanzees succeeded at using a scale model in the same task procedures that prove difficult for 2-year-old children, even though they, like 3-year-olds, seem to rely more on object correspondences. In the domain of theory of mind, Santos and colleagues have demonstrated flexible reasoning by rhesus macaques about a competitor's perceptual state, including responding correctly to changing perception across modalities (Flombaum & Santos Reference Flombaum and Santos2005; Santos et al. Reference Santos, Nissen and Ferrugia2006). We believe this suggests that nonhuman primates lie somewhere on a continuum of relational understanding, and they only fail at the later stages of higher-order relational reasoning that older children can achieve.
We pose two theoretical accounts for the development of cognitive architecture that might explain how older children, and not nonhuman primates, might come to conceptualize higher-order relations. Previously Povinelli (Reference Povinelli, Malle, Baldwin and Moses2001) has argued, at least within the domain of theory of mind, that humans have an additional system that sits side by side with evolutionarily older systems that simply activate earlier. Applied to the relational reinterpretation hypothesis, such a system might allow for analogical reasoning that is not constrained by superficial or context-specific correspondences and might be applied either across multiple specific domains or as a more domain-general “supermodule.” In our view, the addition of such a system to the existing primate mind might explain conceptual change across development, yet it does not fully constitute a violation of Darwinian continuity, particularly if such a system engages actively with the older systems (as, for example, analogical reasoning builds upon the underlying understanding of perceptual correspondences; Gentner Reference Gentner, Gentner and Goldin-Meadow2003). To account for discontinuity at the level that Penn et al. propose, we believe the authors would need to posit that the nonhuman and human minds each begin with unique mental architecture. In this case, it is only that the behavior evident in the first stages of human development looks strikingly similar to the capacities we see in other species. It is only in this latter case that we feel there would be evidence for true cognitive discontinuity.
