In their thought-provoking article, Gilead and colleagues provide a much needed unified conceptualization of abstraction and show how modality-specific, categorical representations up to scripts and more complex mental structures form a hierarchy of increasingly abstract mental representations. Recently, we have proposed an embodied/grounded theory of “abstract concepts,” the Words As social Tools (WAT), that similarly highlights the role of linguistic and social interaction (Borghi & Binkofski Reference Borghi and Binkofski2014; Borghi et al. Reference Borghi, Binkofski, Castelfranchi, Cimatti, Scorolli and Tummolini2017; Reference Borghi, Barca, Binkofski and Tummolini2018a; Reference Borghi, Barca, Binkofski and Tummolini2018b; Reference Borghi, Barca, Binkofski, Castelfranchi, Pezzulo and Tummolini2019a; Reference Borghi, Barca, Binkofski, Castelfranchi, Pezzulo and Tummolini2019b).

Here, we focus on one of the highest rungs in their abstraction ladder – categorical representations – and address three related issues: the distinction between “tangible” and “intangible” abstracta, the role that goal-derived categories might play in the ontogeny of intangible abstracta, and the role of language and sociality for their representation.

First, we argue that, even if the level of conceptual abstractness is extremely variable and context dependent, intangible abstracta differ from other tangible categorical representations, like superordinate concepts, more than the authors acknowledge. Intangible abstracta are considered among the hallmarks of human cognition, indeed many authors argued that concrete (tangible) and abstract (intangible) concepts rely on at least partially separate systems (Shallice & Cooper Reference Shallice and Cooper2013). It is to highlight the difference between intangible abstracta and other categorical representations that we have distinguished between abstraction (the process leading to categorizing and representing concepts hierarchically, e.g., “collie”–“dog”–“animal”) and abstractness (the process leading to the formation of abstract concepts like “freedom” whose referents are more detached from sensory modalities and are not bounded single entities) (Borghi et al. Reference Borghi, Barca, Binkofski, Castelfranchi, Pezzulo and Tummolini2019a). In our view, sensorimotor, interoceptive, linguistic, and social dimensions are relevant for both intangible and tangible abstracta, but to a different extent. Linguistic and social experience is certainly relevant also for tangible abstracta: we learn from others that bats are mammals. Still, “mammals” refer to a collection of perceivable exemplars that ultimately ground their meaning; the case is different for concepts like “freedom”: Linguistic and social inputs become essential to cohere otherwise disparate perceptual experiences. Thus, the suggestion that, compared to multimodal abstracta, “categories” engage areas linked to linguistic processing as the left inferior frontal gyrus (LIFG) fits well with our view, but holds more for intangible abstracta than for superordinate concepts (Borghi et al. Reference Borghi, Barca, Binkofski, Castelfranchi, Pezzulo and Tummolini2019a; see also Dove Reference Dove2018).

We believe, however, that this evidence does not support the claim that their representation should be considered as amodal. Much evidence shows that the processing of intangible abstracta also activates sensorimotor areas (Sakreida et al. Reference Sakreida, Scorolli, Menz, Heim, Borghi and Binkofski2013). Relying uniquely on meta-analyses that focus on areas activated to a greater extent by abstract than by concrete concepts risks to downplay the importance of areas engaged by both; furthermore, meta-analyses typically collapse across different kinds of intangible abstracta, and this can exclude important sensorimotor information. Finally, neuroimaging studies generally include highly concrete items and collapse intermediate and highly abstract concepts, thereby increasing the variability for abstract concepts (Pollock Reference Pollock2018). Evidence shows that specific abstract concepts are grounded in event-based, interoceptive, introspective, sensorimotor areas: for example, temporal concepts rely on perisylvian locations generally identified in time perception studies (Lai & Desai Reference Lai and Desai2016), spatial processing areas in the posterior parietal cortex are activated for numerical concepts whereas emotion concepts engage regions of amygdala and orbitofrontal cortex involved in emotional experience (Desai et al. Reference Desai, Reilly and van Dam2018). Finally, claiming that their representation is amodal is especially problematic for tangible abstracta like superordinate concepts. Behavioural studies have indeed shown that superordinates activate multiple exemplars and their sensorimotor features through an instantiation principle (Borghi et al. Reference Borghi, Caramelli and Setti2005; Heit & Barsalou Reference Heit and Barsalou1996; Murphy & Wisniewski Reference Murphy and Wisniewski1989).

Second, we believe that ad hoc and goal-derived categories (Barsalou Reference Barsalou1983; Reference Barsalou1985) should play a more important role within the framework outlined by the authors. We propose that the capability to form and use goal-derived categories constitutes one of the bases enabling the formation of intangible abstracta. Learning intangible abstracta implies the ability to form and acquire categories that do not have single objects as referent, whose members are not perceptually similar. Such flexibility is present in goal-derived categories, which generally cross the boundaries of standard taxonomic categories, as in the case of “birthday presents” that may include exemplars like flowers, animals, and artefacts. Learning goal-derived categories can provide a bootstrapping mechanism useful for further acquisition of intangible abstracta.

Third, several studies conducted in our lab and in other labs have shown that abstract concepts activate linguistic experience through the involvement of the mouth motor system. Participants rate abstract concepts, particularly mental state ones, as more associated with mouth than hand actions (Ghio et al. Reference Ghio, Vaghi and Tettamanti2013; Granito et al. Reference Granito, Scorolli and Borghi2015); consistently, fMRI has shown that the mouth motor system is engaged during processing of these concepts (Dreyer & Pulvermüller Reference Dreyer and Pulvermüller2018). When participants process abstract concepts, they are facilitated in responding using the mouth rather than the hand (e.g., Borghi & Zarcone Reference Borghi and Zarcone2016; Mazzuca et al. Reference Mazzuca, Lugli, Benassi, Nicoletti and Borghi2018). Furthermore, blocking the mouth selectively influences abstract concepts acquisition, as demonstrated by studies with children who used the pacifier until late (Barca et al. Reference Barca, Mazzuca and Borghi2017; in press). We have proposed that mouth involvement might be linked to a mechanism of social metacognition (Borghi et al. Reference Borghi, Barca, Binkofski and Tummolini2018b; Reference Borghi, Barca, Binkofski, Castelfranchi, Pezzulo and Tummolini2019a; Fini & Borghi Reference Fini and Borghi2019; Villani et al. Reference Villani, Lugli, Liuzza and Borghi2019): Because abstract concepts are more complex, we feel less competent (Shea Reference Shea2018) and need help from authoritative others (Prinz Reference Prinz2012b), preparing ourselves to ask for information. Gilead et al. propose that abstractness has an important evolutionary function because groups that define their beliefs in terms of intangible ideas might become more cohesive. We argue that our social metacognition mechanism can help to increase group cohesion. Differently from multimodal abstracta, the intrinsic complexity of intangible abstracta induces individuals to assess their own competence and to rely more on others. This mechanism can be powerful in creating social bonds, because it helps individuals to recognize the role of others as dispensers of knowledge, and induces competent others to share their knowledge. It can also contribute to explain social hierarchies, based on the different competences people have in matters relevant for their group.