ArticlePDF Available

Ultimate Grounding of Abstract Concepts: A Graded Account



Abstraction, one of the hallmarks of human cognition, continues to be the topic of a strong debate. The primary disagreement concerns whether or not abstract concepts can be accounted for within the scope of embodied cognition. In this paper, we introduce the embodied approach to conceptual knowledge and distinguish between embodiment and grounding, where grounding is the general term for how concepts initially acquire their meaning. Referring to numerous pieces of empirical evidence, we emphasise that, ultimately, all concepts are acquired via interaction with the world via two main pathways: embodiment and social interaction. The first pathway is direct and primarily involves action/perception, interoception and emotions. The second pathway is indirect, being mediated by language in particular. Evidence from neuroscience, psychology and cognitive linguistics shows these pathways have different properties, roles in cognition and temporal profiles. Human development also places revealing constraints on how children develop the ability to reason more abstractly as they grow up. We recognize language as a crucial cognitive faculty with several roles enabling the acquisition of abstract concepts indirectly. Three detailed case studies on body-specificity hypothesis, abstract verbs and mathematics are used to argue that a compelling case has accumulated in favour of the ultimate grounding of abstract concepts in an agent’s interaction with its world, primarily relying on the direct pathway. We consolidate the debate through multidisciplinary evidence for the idea that abstractness is a graded, rather than a binary property of concepts.
Prof. Igor Farkaš
Department of Applied
Informatics, Faculty of
Mathematics Physics and
Informatics, Comenius
University in Bratislava,
Bratislava, SK
embodiment; grounded
cognition; graded
abstractness; language;
conceptual metaphor
Reinboth, T., & Farkaš, I.
(2022). Ultimate Grounding of
Abstract Concepts: A Graded
Account. Journal of Cognition,
5(1): 21, pp. 1–26. DOI: https://
Ultimate Grounding
of Abstract Concepts:
A Graded Account
Abstraction, one of the hallmarks of human cognition, continues to be the topic of a
strong debate. The primary disagreement concerns whether or not abstract concepts
can be accounted for within the scope of embodied cognition. In this paper, we
introduce the embodied approach to conceptual knowledge and distinguish between
embodiment and grounding, where grounding is the general term for how concepts
initially acquire their meaning. Referring to numerous pieces of empirical evidence,
we emphasise that, ultimately, all concepts are acquired via interaction with the
world via two main pathways: embodiment and social interaction. The first pathway
is direct and primarily involves action/perception, interoception and emotions. The
second pathway is indirect, being mediated by language in particular. Evidence
from neuroscience, psychology and cognitive linguistics shows these pathways have
different properties, roles in cognition and temporal profiles. Human development
also places revealing constraints on how children develop the ability to reason more
abstractly as they grow up. We recognize language as a crucial cognitive faculty with
several roles enabling the acquisition of abstract concepts indirectly. Three detailed
case studies on body-specificity hypothesis, abstract verbs and mathematics are used
to argue that a compelling case has accumulated in favour of the ultimate grounding
of abstract concepts in an agent’s interaction with its world, primarily relying on the
direct pathway. We consolidate the debate through multidisciplinary evidence for the
idea that abstractness is a graded, rather than a binary property of concepts.
*Author affiliations can be found in the back matter of this article
2Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
We need concepts to organise our knowledge about the world. One aspect of these concepts
is their abstractness, meaning the extent to which concepts reflect the idiosyncratic details
of the experiences from which they are derived (Burgoon, Henderson & Markman, 2013).
This establishes our experiences as the source for abstract concepts, but it leaves open the
question of how exactly such concepts are formed. As a basis for the rest of the paper, we use
a combination of the most commonly used definitions of abstraction to define it broadly as:
a conceptual process by which general regularities and resulting concepts are derived from
the usage and classification of specific examples (Barsalou et al., 2003a). The related notion
of abstractness suggests that abstraction operates hierarchically, with some concepts being
more abstract than others. The accepted view is that “democracy” does not just appear to
be a more abstract concept than “pain”; instead, more abstract concepts are said to rely on
cognitive processes that are in some important way different from those of more concrete
concepts. The crucial open question about abstraction is what kinds of cognitive processes are
the basis for how we understand more or less abstract concepts.
While many attempts have been made to explain abstraction, cognitive science lacks a
consensus regarding the scope of these explanations (Yee, 2019). One set of examples are
the embodied approaches, which argue, in one way or another, that cognition is non-trivially
dependent upon the physical body (Wilson, 2002). Put another way, these approaches argue
for different ways in which how we think and act depends on our bodies. Fincher-Kiefer (2019)
summarises compelling empirical evidence for this idea from decades of research, centred
around three major theories – metaphoric extensions (Lakoff & Johnson, 1980), action
schemas (Glenberg & Robertson, 2000) and situated simulation (Barsalou, 2008). Besides
compelling empirical support, embodied approaches also benefit from their consistency with
the ascension of situated action in cognitive science, which is solving old problems by shifting
emphasis from how humans think to how humans act within their environments (Robbins &
Aydede, 2009). In this sense, embodied cognition reflects how cognitive science is maturing
past the historically-privileged notions of mentalism, logic and a latent Cartesian dualism.
Taken together, these reasons encourage making embodied approaches the starting point of a
unifying look at abstract concepts.
Though the embodied paradigm offers compelling accounts of many of the basic features of
cognition (Varela, Thompson & Rosch, 2017), it struggles with a full account of abstraction
and abstract concepts specifically (Borghi et al., 2017). The challenge resides in the need to
ground abstract concepts in concrete, sensorimotor interactions, or simulations (Barsalou,
1999). However, even within the embodied literature on abstraction, we are faced with a
heterogeneous set of theoretical commitments. For instance, some attempts have adopted a
representationalist view of abstract reasoning (Barsalou & Wiemer-Hastings, 2005), while others
propose a radical embodied claim that completely eliminates the mental representations from
its view of cognition (Thompson & Varela, 2001). That alone is the difference between retaining
or completely abandoning one of the central concepts of cognitive science (Shea, 2018). This
paper avoids getting bogged down in discussions of such underlying issues as much as possible.
Instead, we use the considerable evidence for some role of embodiment in abstraction as a
driver to approach the topic through the issue of symbol grounding more broadly (Harnad,
1990). That means we are addressing the question of how abstract concepts come to have
meaning. To keep the discussion focussed, a large part of our paper presents three case studies
through which we aim to breathe new life into different aspects of the theoretical debate.
One starting point from which to approach abstractions is the distinction that concrete concepts
typically refer to categories of physical referents, whereas abstract concepts refer to sets of
situations and events. This abstract–concrete distinction finds wide empirical support and
dichotomization is a useful general approach in science when trying to contrast the two sets of
entities. For example, Kousta et al. (2011) provide a number of factors by which to distinguish
abstract and concrete concepts, such as imageability, context availability, or emotional
valence. Aspects of these distinctions are also robustly supported by neuroimaging studies with
patients with language disorders (Hoffman, 2016), as well as by findings indicating qualitatively
distinct neural representational systems (Crutch & Warrington, 2005). Moreover, even studies
that aim to move beyond the dichotomy report results that reinscribe a qualitative difference
3Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
between the two types of concepts: abstract and concrete (Della Rosa et al., 2010). On the one
hand, Wiener-Hastings & Xu (2010) identified differences between 18 abstract and 18 concrete
words, when they asked participants to list features of different concepts: abstract/concrete
concepts had more subjective/intrinsic properties respectively. On the other hand, based on
ratings of 425 abstract concepts on 15 dimensions, Villani et al. (2019) concluded that the
abstract–concrete dichotomy is too simple to properly understand differences between more
concrete and more abstract concepts. In this paper, we reconcile such results by arguing that
the differences these studies identify do not have to be interpreted in terms of a dichotomy
between abstract and concrete. An alternative explanation is that there is a continuum of
abstractness, from the most concrete to the most abstract concepts.
Alongside controversy around categories/degrees of abstractness, there is a different aspect of
abstraction that cannot be agreed upon. Namely: how does the meaning of concepts, especially
of abstract concepts, emerge in the first place? Put another way: how are abstract concepts
grounded (Dove, 2011; Machery, 2009; Meteyard et al., 2012). This is the sense in which we are
dealing with the symbol grounding problem (Harnad, 1990), which refers to how words – and
symbols in general – first acquire their meaning. Abstract concepts, we suggest, can be thought
of as just such a symbol (expressed by a word or a phrase), for which we must determine
the source of its meaning. To understand the problem, imagine arriving in a new country in
which all words and gestures are unfamiliar, and there is no way to begin to understand what
someone is trying to communicate to you. The signs (words and gestures) have no meaning
for you, because you cannot meaningfully relate them to anything you know. In other words,
they are not grounded in your experiences of sensing/acting in the world, or your experiences
of communicating with people using signs (a language, gestures) you do understand. Harnad’s
(1990) point is that such experiences ground the meaning of all of our knowledge. Without
grounding, we see someone mouthing and gesturing at us in a strange country, but understand
nothing. That makes how concepts can be grounded a crucial question.
With the introduction of the concept of “grounding” it should be clarified that we use the term
embodied in the context of the body, whereas grounding has a wider meaning, covering also
other forms of interaction with the world (via language and social interactions). Grounding,
not embodiment will be the central notion going forward, making the central question of this
paper how abstract concepts are grounded. To reflect the distinction between embodiment
specifically and grounding generally, we introduce the idea of two pathways of grounding.
The first pathway captures embodiment in the sense of sensorimotor interactions with the
world, interoception, or emotions. The second pathway builds on this to take into account
how social aspects and language, in particular, add an additional route of grounding abstract
concepts. This coincides with our idea of a continuum of abstractness that reflects different
ways that concepts are grounded, specifically the extent to which they depend on either of
the two pathways outlined above. This is consistent with differences between concrete and
abstract concepts (e.g. Wiemer-Hastings & Xu, 2010) and differences between more or less
abstract concepts as such (Villani et al, 2019, Villani et al, 2021a). We emphasise a continuum
of abstractness mediated by two more/less direct pathways to make 1) gradedness and 2)
directness of grounding the focal aspects of our contribution.
This paper is organised as follows. Section 2 is intended to shore up claims we have already
made about the central role of embodiment, grounding and the role of two pathways. To do
so, we briefly summarise the strong empirical support for embodiment of knowledge and
outline the various roles language is assumed to play in cognition. This will provide grounds
for developing our ideas in Section 3, on the basis of three detailed, purposefully chosen case
studies of increasing abstractness. The case studies provide specific examples of different ways
that concepts are grounded and provide further evidence for an embodied basis of abstraction.
In Section 4, we complement the case studies with a review of neurodevelopmental literature
that illustrates the contrasts and asymmetry between concrete and abstract domains. This
provides both a closer look at the recent neuro-imaging data and a glance at how results from
developmental studies shed light on a continuum of concreteness-abstractness. In Section
5, we summarise the supporting evidence for treating the abstract–concrete dichotomy as
a continuum, and hence viewing abstractness as a graded phenomenon. Section 6 briefly
summarises the major conclusions from the paper and offers a short outlook on open questions.
4Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
Ahead of the main argument, this section outlines key aspects of an embodied approach to
conceptual knowledge. Embodied cognition provides the basis for a continuum of abstractness
by emphasising the role of interactions with our world across all domains of cognition. This claim
is supported by a vast empirical literature that contains many internal disagreements about
exactly how embodiment works. One particular challenge has been to provide an embodied
account of language, which appears to provide a separate route to ground conceptual knowledge.
Embodied approaches to abstraction argue that there are two pathways to grounding conceptual
knowledge: one tied closely to sensing and acting in the world, and the other tied more to
language and social interactions, which are themselves embodied in various ways.
The starting point of this contribution is embodied cognition. In a nutshell, embodiment is
one kind of grounding (Barsalou, 2010). Specifically, embodiment refers to the notion that
conceptual knowledge derives its meaning from the apparatus, faculties and states of the
(physical) body of the agent. For this, interactions with the environment are crucial (Barsalou,
2005), including those an agent has with itself, e.g. in interoception (Damasio, 1999).
The credited original description is that embodiment has two basic manifestations, one in
the lived, experiential structure of the body, and the other in the body as the active vessel
and immediate environment of cognition (Varela et al, 2017, Chapter 8). This reflects 1) an
inner and 2) and outer view (Reid & Mgombelo, 2015). The corresponding claims are that 1)
cognitive capacities depend on having a body with various sensorimotor capacities and 2) these
capacities are embedded in the situation of the agent, encompassing biological, psychological
and cultural factors (Varela et al. 2017, Chapter 8). This characterisation of embodiment reflects
an enactivist or radically embodied perspective on embodiment that other authors, including
ourselves, are not necessarily committed to or consistent with (Núñez, 1999).
From a neural perspective, that a certain cognitive ability is embodied is also understood as
that ability being “mapped within our sensorimotor system” (Gallese & Lakoff, 2005, p. 3). In
that sense, embodied knowledge is “structured by our constant encounter and interaction with
the world via our bodies and brains” (Gallese & Lakoff, 2005, p. 3). As an example, consider
grasping: the claim, from an embodied view, is that the information present in the sensorimotor
system at the neural level is sufficient – and necessary – to create a conceptual structure of the
concept “grasping/to grasp”. In this view, the sensorimotor system is necessary to understand
a concept, because conceptual knowledge of the action “to grasp” is based on a sensorimotor
simulation of grasping (Gallese & Lakoff, 2005). The simulation of grasping is the concept.
Evidence for such a simulation comes from neuroscientific studies of humans either grasping or
imagining grasping an object, which showed that both behaviours activate partially overlapping
of the brain (Jeannerod, 1995; for a summary of neuroscientific data, see Jeannerod & Frak,
1999). Gallese & Lakoff (2005) go on that this same information is not only necessary, but also
sufficient to ground conceptual knowledge, because all the sensorimotor structures through
which we understand the world are present in the sensorimotor system, at least for concrete
concepts. Staunch opposition to this view persists on the basis of data interpreted to show
that even motor actions, arguably the most concrete concepts, are organised by conceptual,
not sensorimotor properties (Bedny et al., 2008). The idea comes from functional magnetic
resonance imaging (fMRI) data that show different patterns of reaction to verbs and nouns,
which is taken to suggest that the partially-overlapping neural activity is rooted in event
concepts (“I am moving a part of my body”) or grammatical information processed through
these regions. In Bedny et al’s (2008) view, conceptual properties like these are of a higher
order than “mere” sensorimotor cognition. This separates concepts from modal processing,
i.e. information received through the sensory modalities. Doing so reiterates the question of
how more abstract concepts may be grounded, if they are separate from the substrate of our
sensory and motor experiences. What happens during the upstream integration of modal
sensory processing that justifies talking about the kind of understanding in terms of conceptual
properties that the abstract–concrete dichotomy depends on. The crux of this debate is more
whether embodiment is necessary, and less whether it is sufficient to ground abstract concepts.
5Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
Whereas the following sections, and the case studies in particular, will address the controversy
about whether or not we should think of concepts as being grounded, the take-away at
this point is that embodied approaches focus on sensorimotor processing and the sensory
modalities. In doing so, they emphasise that it is the ways that an agent perceives and the
ways that it is able to act that determine what concepts and categories can be formed. Indeed,
proponents go so far as to claim that the nature of concepts referred to by the very technical,
linguistic notion of aspect – concepts denoting the structure of events and our reasoning about
them – is such that they are ideally realised in premotor schemas (Gallese & Lakoff, 2005). This
is a fine-grained argument from within cognitive linguistics that illustrates the external validity
of embodied approaches by how its proponents have found neat overlaps and consistencies in
the distinct concepts of sensorimotor and language cognition.
There exists a vast literature on such overlaps from the various theories of embodied cognition.
These map cognitive linguistics onto embodied cognition, but also decision-making (Lepora &
Pezzulo, 2015) or conceptual knowledge (Barsalou, 2008). However, this breadth incurs a variety
of embodied theories that are not all completely compatible. One approach to taxonomizing
these theories places different variants along a continuum, ranging from strongly embodied,
through intermediate, to weakly embodied theories (Meteyard et al., 2012). The distinction
between strong and weak embodiment is most closely related to the discussion whether
embodied knowledge is sufficient to ground conceptual knowledge. Strong embodied theories
argue yes, while weak theories accept that other types of knowledge are necessary as well.
This distinction is important, because it captures one of the central issues of this article: to what
extent abstract concepts are grounded via the first pathway.
Another distinction that is important is related to how different theories of embodied cognition
can also be divided into groups based on the role that they assign to representations when it
comes to concrete and abstract concepts (Scorolli et al., 2011). This distinction between theories
that eliminate or include representations is more closely related to the question of whether
grounding is necessary. Within the embodied approach, a frequent view is that embodiment
is necessary to ground all conceptual knowledge, but that abstract and concrete concepts are
represented differently in the two systems the perception and action system, and the language
domain (e.g. Borghi et al., 2017). Based on recent neuro-imaging data, Pulvermüller (2018)
argues that symbols are necessary to bind causal actions and their goals, from which they
derive their meaning. The same study also attributes linguistic symbols (e.g. words) a key role
in this binding when it comes to abstract concepts. This suggests that understanding abstract
words may rely more on the linguistic system, whereas understanding concrete words depends
more on the perception and action system. Besides reiterating our two pathway model, this
also raises a crucial question about how exactly concepts (conceptual knowledge) and words
(language) are related in the embodied approach.
In general, many accounts on the organisation of semantic knowledge assume that the human
cognitive system standing behind the acquisition of concepts consists of two main systems,
the conceptual system and the linguistic system. Especially to those to whom concepts
are inseparable from words, this is an indefeasible distinction. Nevertheless, separating the
conceptual and linguistic systems has strong evolutionary support. At the same time that
the human conceptual system appears more complex, in general, than that of other species,
part of that complexity comes from the presence and importance of language in human lives
and society (Barsalou, 2005; Šefránek, 2008; Evans, 2016). The cognitive processes within
conceptual and linguistic systems also interact during development, also bootstrapping each
other. On one hand, empirical evidence suggests that conceptual knowledge influences lexical
acquisition in infancy (Booth, Waxman, & Huang, 2005). On the other hand, there is a lot of
evidence in favour of the linguistic relativity hypothesis, according to which language shapes
the way we think (e.g. Boroditsky, 2003). The crucial addition of our graded approach is to put
these views on a continuum.
The two systems, conceptual and linguistic, are central to a variety of theoretical models of
conceptual knowledge (for a recent overview see, e.g. Borghi, 2020). Within an embodied
6Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
approach, the conceptual system is multimodal, meaning it is always tied to the senses and
our ability to act. The main differences between such theories are in the assumed details
(example theories include e.g. LASS by Barsalou et al., 2008; WAT by Borghi & Cimatti, 2009b;
or LCCM by Evans, 2016). Most embodied approaches maintain that sensory modalities are the
ultimate grounding of all conceptual knowledge, no matter how abstract. But other embodied
approaches take the idea of distinct systems so far that they inadvertently reintroduce
amodal concepts, concepts that are not tied to the body at all, in the end. Dove (2011), for
example, offers the notion of dis-embodied concepts, positing a new kind of concepts,
which “are dynamic and multimodal but, in contrast to other forms of embodied cognition,
do not inherit semantic content from this embodiment [and are instead] embodied in the
neurophysiological sense that they rely on sensorimotor simulation.” In Dove’s (2011) view,
all linguistic representations of concepts are dis-embodied in this sense. He emphasises the
value of language as an internalized, amodal symbol system itself grounded in an embodied
substrate.1 The idea is reiterated in Dove (2018) where it is also suggested that language
somehow acquires a fundamental character in our experience and thereby becomes its own
fundamental grounding for concepts. We are concerned that this ignores that language mirrors
the structure of sensorimotor interaction, rather than ever being independent of it. On the other
hand, sensorimotor processes may not always be automatically activated during linguistic
processing, since their activation may depend on the task and the context (see Tomasino &
Rumiati, 2013 and references in that issue).
The challenge about language for embodied views of conceptual knowledge lies in determining
how language plays an additional, but still embodied role for abstract concepts. For example,
Borghi and Cimatti (2009b) specify an explicitly not-independent lexical semantic code,
that allows conceptual content to be captured by relationships between different linguistic
representations. To Borghi and Cimatti (2009b), in contrast to Dove (2011), language confers
a number of new representational affordances through compositionality, recombination
and a rich expressive medium (cf. Borghi & Binkofski, 2014). In their view, linguistic symbols
cannot provide an amodal grounding for abstract concepts, as Dove (2011) suggests. Instead,
as a secondary pathway (together with social interactions), they offer a particular cognitive
scaffolding that exploits word associations to provide unique benefits to human cognition (cf.
Clark, 2008).
Despite these disputes, there seems a general agreement about the fundamental importance
of language in human cognition within embodied approaches, where a number of language
roles have been identified: (1) Language allows linguistically-mediated communication (Evans,
2016), (2) it enables inner speech that helps us in monitoring our knowledge and in searching
for meaning (Clark, 1998), (3) language is involved in controlling mental representations
(Lupyan & Bergen, 2016), (4) it provides an efficient shortcut in certain tasks involving
conceptual processing (Connell, 2019; Barsalou et al., 2008) and (5) it serves as a cognitive tool
that enhances cognition as such in various ways and enables the acquisition and processing of
abstract concepts (Mirolli & Parisi, 2011; Borghi, 2020), as discussed later. Consistently with the
above roles, language also serves as an important medium for using conceptual metaphors. In
these ways, language provides a second, indirect pathway for grounding conceptual knowledge.
We make the case that more and less abstract concepts depend on the two pathways of
grounding in different ways. Less abstract (i.e. more concrete) concepts depend more directly
on the body, while more abstract concepts are grounded in important ways by language and
social interaction. Specific examples of concepts that are different in their degree of abstractness
provide evidence for a continuum of abstractness, as more abstract concepts still depend on the
body, but less directly so. At the very abstract end of the continuum, the two pathway hypothesis
of grounding even provides an embodied basis of technical, mathematical concepts like “infinity”,
which, in their actual form, cannot have a physical referent by definition.
1 Dove’s dis-embodied view on language differs from traditional disembodied view (e.g. Mahon and
Caramazza, 2008) which does not rely on sensorimotor simulations in linguistic processing of concepts, but
rather postulates disembodied processes using amodal symbols.
7Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
Despite the widely recognized important role of sensorimotor experience in grounding,
especially in the context of concrete concepts, it has been recognized by many that an
extension beyond a simple embodied approach is needed. Borghi et al. (2018) list three other
forms of grounding, namely inner experience (interoception and metacognition), linguistic
experience and social experience. Abstract concepts are all constituted differently and do not
form a homogeneous set – numbers, emotions, evaluative concepts (e.g. aesthetic and moral
ones) and social concepts are among examples of abstract concepts. This variety of abstract
concepts is a challenge for any single explanation for how all of them are grounded.
Villani et al. (2021a) found evidence for this view in a study in which participants rated the
difficulty of more concrete and more abstract concepts while also performing a second
task. “Difficult” was not defined for the participants, but left to them to interpret. The tasks
participants performed while they rated words were designed to interfere with conceptual
processing in specific ways, in order to see if different concepts would be disrupted more by
different tasks. Participants either 1) squeezed a ball in their hand, 2) chewed gum, 3) held the
hand of a confederate or 4) held an instant warm/cold pack. The conditions were designed
to interfere with 1) the hand motor system, 2) mouth motor system (related to speaking,
therefore language), 3) social cognition and 4) interoception. The results showed that concepts
vary widely in the extent to which these conditions affect difficulty ratings, compared to a
control condition in which there was no second task to perform. More abstract concepts were
rated as more difficult while holding an instant hot-cold pack, and concrete concepts appeared
less affected by the mouth-motor system interference. This suggests abstract concepts might
be more closely tied to interoception, while concrete concepts may depend less on language
for their grounding. What the study also showed is that difficulty ratings of abstract concepts
were affected differently by different conditions, such that some abstract concepts appeared
to be interfered with much more than others by the manual versus the interoceptive task.
This could be interpreted to mean that some abstract concepts rely more on direct grounding
via the sensory-motor system, while others depend more on interoception or even mostly on
indirect grounding via language. A second study by Villani et al. (2021b) not only reiterated
that abstract concepts appear grounded in different ways from each other, but also showed
that how concepts are grounded depends on personal experiences, in this case expertise in
the legal domain. Their findings showed differences in how abstract concepts are understood
by legal experts/non-experts, in terms of the ratings they gave words on 16 dimensions. These
results showed that the more personal experience someone possessed with a concept, the
more highly they rated it on direct dimensions like “hand and mouth involvement”. In general,
this study reiterates the richness of abstract concepts and the differences between them in
how they are grounded. For a complex example, the study found that emotion has a different
significance for abstract concepts depending on expertise.
We tackle the problem that this richness of abstract concepts poses by grouping dimensions
of grounding into two pathways. The first one is characterised by bodily experience, which
includes both sensorimotor processes but also inner processes (interoception) – for a discussion
of the role of interoception for abstract concepts see Connell, Lynott, and Banks (2018) – and
affective states. The second pathway comes from interactions with others and includes both
linguistic and social processes. Of course, language is very important for multiple reasons,
as mentioned earlier, but it can only work if coupled by an earlier body-related mode of
knowledge acquisition, in order to avoid the symbol grounding problem (see more about the
developmental aspects in Section 4). The boundaries between the two pathways are not clear-
cut, because, for example, social processes are often necessarily related to emotions (van Kleef
et al., 2016). These can be analysed at individual, dyadic, group or cultural context but what
seems to be important, though, are the properties of each concept that determine to what
extent the concept is resistant to embodiment. In other words, from the perspective of the
“owner” of an emotion, the same emotion can be evoked in different contexts which serve
as its source of grounding. Experienced emotions are typically grounded through one’s own
body, but they can also be grounded via observation within a social (not necessarily a linguistic)
communication, potentially mediated by the mirror neuron system (Bekkali et al., 2021).
In the following, we provide evidence from three case studies, proceeding in an ascending
order with respect to their “distance” from the “source” of embodiment. In the first case study,
the body-specificity hypothesis points to a clear, immediate role of the body in the grounding
8Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
of abstract concepts; the second case study discusses the grounding of abstract verbs, which
we argue is often mediated through the body; as a third case study, we show how grounding
mathematics reveals a highly abstract hierarchical set of concepts that poses a formidable
challenge precisely because of how indirectly related to embodiment it appears. In fact, all
three case studies illustrate the deep dependence of abstract concepts mostly on embodiment
and partially on social interaction. In each case, the first pathway, via embodiment, provides
the grounding and consequently constitutes the fundamental source of meaning. Whereas
language, the second pathway, provides an additional, indirect means to ground especially
abstract concepts. These case studies provide an opportunity to disentangle the multifaceted
concept of abstraction one step at a time along a continuum of abstractness.
There is a range of evidence that motor experience shapes affective judgments, for example
from an expert typist preferring easily typed letter pairs to a pair more difficult to type fluently
(Van den Bergh, Vrana, & Eelen, 1990). Moreover, motor fluency has been implicated in a broad
range of non-motor judgements (Oppenheimer, 2008). This case study focuses on a compelling
series of five experiments makes the case that the association of valence (good vs. bad) with
the left or right side of agent-centred space is the result of the differential experience of
perceptuo-motor fluency with the corresponding hand (Casasanto, 2009). The five tasks, with
five unique samples were 1) choosing where to draw a good/bad animal; 2) indicating where a
corresponding animal should be drawn; 3) making an oral response about where animal should
be drawn; 4) indicating preference for fictional creatures presented on left/right; 5) choosing,
from a side-by-side presentation, which of two products to buy / which of two candidates to
hire. In all five tasks – participants consistently associated positive valence with their dominant,
and negative valence with their non-dominant hand (Casasanto, 2009). We interpret this as
evidence that abstract concepts must, in some way, be tied to sensorimotor processes.
The body-specificity effect occurs in left-handed participants despite a strong cultural association
of “right” with “good” (e.g. “my right-hand man”) and left with bad (e.g. “two left feet”). In the
case of left-handers, the studies showed that the effect occurs not as a result of culturally-
mediated metaphors but in opposition to them. Moreover, the assignment of right-is-good and
left-is-bad is unsystematic, i.e. there is no idiom of “two right feet”, or one’s “left-hand woman”.
In addition, differences on left/right ratings that were related to handedness were not present on
top/bottom control conditions. A further study found positively and negatively-valenced speech
acts to be preferentially associated with the dominant and non-dominant hand respectively
(Casasanto & Jasmin, 2010). Beyond that, a subsequent study found that this association
was reversed following unilateral stroke resulting in a change in motor fluency (Casasanto &
Chrysikou, 2011). The same study observed a similar reversal for artificial handicaps induced
in a laboratory setting after only minutes of experience. These results are strong evidence that
motor experience plays a causal role in shaping abstract thought.
One way to explain this effect is via implicit motor simulations (Barsalou, 1999). This means
that though the association is established and grounded in action and motor experience,
the representation of the association may be decoupled from activity. Indeed, it is an open
question to which extent, if at all, spatio-motor representations are constitutive of valenced
concepts. Casasanto and colleagues fall back on the idea that it is impossible to demarcate
concepts (cf. Wittgenstein & Anscombe, 1953), with the aim of using Wittgenstein’s concept
of meaning-in-use to argue that the dissociable assignment of valenced labels in left- vs.
right-handers implies that these groups have different concepts. It is not that handedness
leads us to make different decisions, but to make them differently. By emphasising the act
and process of decision making, Casasanto and colleagues shift the focus to what ultimately
grounds the laterality of valence via perceptuo-motor fluency. In this view, concepts are not
part of packages of preformed knowledge, but processes of activating stored information ad
hoc based on sensorimotor experience (Casasanto & Lupyan, 2015). That way, this effect of
how one uses one’s hands on how one thinks about and attributes abstract properties also
makes a compelling case for embodiment (Casasanto, 2011b).
Complementing the behavioural data, an early question was whether the conventional
hemispheric lateralization of valence in neuroimaging studies conducted with exclusively
9Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
right-handed samples might be an artefact of body-specificity (Casasanto, 2009), especially
considering functional lateralization as a result of handedness (Casasanto, Hagoort, & Willems,
2009). There is already evidence of lateralized activity for motor imagery of hand-related action
and for reading hand-action related phrases in premotor cortex (PMC) (Willems et al., 2009).
Moreover, the application of disruptive transcranial magnetic stimulation (TMS) reiterated that
premotor activity was functional, not epiphenomenal to hand-related language processing
(Willems et al., 2011). Specifically, bursts of TMS over lateral PMC reduced reading rates. Taken
together, this points to an embodied semantics in the neural processing of one set of abstract
concepts, namely valence.
Given the role that the body plays in emotion (Damasio, 1999), and the body-specificity of
valenced, often emotionally-laden concepts, there is a case to be made that body-specific
patterns of activation mediate a functional connection between the neural substrates of motor
control and emotional mechanisms such as motivation, specifically approach-avoidance
systems (Casasanto, 2011a). Indeed, there is evidence of neurophysiological correlates of
handedness across the human frontal lobe (Brookshire & Casasanto, 2012), which can be
interpreted as evidence of such a functional link. This converges with data that support the
differential grounding of abstract concepts in emotional states in contrast to more concrete
concepts (Kousta et al., 2011). Through this link to emotion and its neural correlates, the
body-specificity hypothesis provides evidence that motor fluency, a fundamentally embodied
experience, is one of the organising factors of conceptual structure. Casasanto and colleagues
tie the grounding of valence to embodiment directly. These studies attribute a subordinate
role to the second pathway of grounding, namely social interactions and culture transmission.
The next case study moves on to consider a different way that our experience ground abstract
Different word classes appear to be processed by partially dissociable neural circuits (Wiemer-
Hastings & Xu, 2005). Verbs, in particular, can also be characterised in terms of their relative
embodiment – a measure derived from Borghi and Cimatti (2009a) that captures the bodily
sense and strength thereof with which a given verb is associated (Sidhu et al., 2014). On three
different tasks, Sidhu et al. (2014) showed a facilitative effect of relative embodiment in terms
of reduced reaction times, even when controlling for lexical factors such as imagability. From
these results, embodiment appears to be a central aspect of verb processing and by extension
of verb meaning.
Previous work suggests there are also differences within the verb class (Barsalou & Wiemer-
Hastings, 2005). For example, neuroimaging data show that, in terms of their involvement of
the cortical motor system, abstract verbs with motor stems (e.g. German begreifen, “to grasp
an idea”) are dissociable from verbs with explicit motor meanings (e.g. German greifen, “to
grasp a cup”), but not distinguishable from verbs with abstract stems (Rüschemeyer, Brass
& Friederici, 2007). At first, this suggests that having a motor stem rather than an abstract
stem does not make verbs more connected to the cortical motor system. However, this does
not necessarily weigh against an embodiment of motor-stem abstract verbs. It establishes a
similar processing network for abstract verbs, whether they possess a motor stem or not. This
could reflect the equal embodiment of abstract verbs irrespective of their morphological form.
Both abstract and concrete verbs extensively, but differentially recruit sensorimotor cortices
(Sakreida et al., 2013). Both kinds of verbs appear embodied to some extent, but rely on the
sensorimotor and conceptual linguistic subsystems to different extents. This is in line with
various attempts at synthesis (Dove, 2011; Arbib, 2017) and with our idea of a continuum of
abstractness. One interesting study showed that the processing of abstract actions recruits
motor systems independent of the presentation format (schematic visual displays versus
words) (Quandt, Lee & Chatterjee, 2017). This indicates that there is some level of processing
beyond the modality in which we usually experience an action. However, the activity appears
specific to individual abstract actions, without a clear pattern. As a result, this section will
examine an individual abstract verb to better understand how it is grounded, and thereby to
improve our understanding of the grounding of this word class as a whole.
10Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
One example is the intransitive verb “to meditate”, which means “to engage in mental
exercise (such as concentration on one’s breathing or repetition of a mantra) for the purpose
of reaching a heightened level of spiritual awareness” (Merriam-Webster, 2016). At first this
seems like a concrete verb, but a closer look at the goal-directed nature reveals this may not
be the case. This becomes more apparent in the definition by editors of the American Heritage
Dictionaries (2018) of “to meditate” as “to train, calm, or empty the mind, often by achieving
an altered state, as by focusing on a single object, especially as a form of religious practice
in Buddhism or Hinduism.” That definition is not in terms of any specific actions, so much
as in terms of the purpose of this action, which is the abstract emptying or calming of the
mind. As an abstract verb, “to meditate” is a particularly lucid example of how each concept
is grounded through both pathways, which also makes it a good example of the gradedness
of abstractness–concreteness. As a verb, it refers to an action, from which the concept draws
a certain concreteness. At the same time, the theory of meditation contributes a rich abstract
character that explicitly embeds the act of meditating in a system of values and norms. How
we understand the concept falls somewhere in between.
One way to approach abstract verbs like this is by taking into consideration the two types of
knowledge under discussion: conceptual/intellectual and embodied. This reflects a different
aspect of differences in how directly concepts are acquired, as in how immediate our
experience of their referent is. Starting from the experiential nature of some knowledge, Pagis
(2010) explored the concept of dissatisfaction in Buddhist tradition to show that individuals
who ‘possess’ an abstract concept prior to lived experience, have a rather vague notion of what
that concept represents. They can connect to it “intellectually”, but it often hardly poses any
meaning to them. Through repeated acts of lived experience that constitute the referent of the
concept, participants reported that they acquired a new meaning for the concept. This example
also reflects the fundamental role of non-sensorimotor sources of knowledge in the grounded
paradigm, such as meta-cognition and affect (Barsalou, 2016). At the same time, the example
outlines an experiential understanding that is rooted in the sensory impressions and actions
that make up that experience. However, this leaves open to what extent the representation
of the concept of dissatisfaction changes as someone becomes more familiar with it in an
experiential way.
To understand how the basic representation of “to meditate” first develops, it is worth noting
that the definitions cited above make use of embodied language, like the image of “emptying
(pouring out) a mind”, or “training the mind (like a muscle)”. These kinds of images are the
subject of the theory of conceptual metaphors, one of the first embodied accounts of abstract
thought (Lakoff & Johnson, 1980). In this view, metaphors shape our conceptual system, which
in turn, shapes our view of the world.2 This results not only in metaphorical speech, but also
in metaphorical thought, where the latter precedes the former. Within conceptual metaphor
theory, understanding a grasping movement is taken to facilitate the comprehension of
metaphorical phrases like “grasping a concept” (Gibbs, 2011). Or the act of balancing our bodies
when we move in the world, for example, can serve as the embodied metaphoric foundation
for the concept of justice, which is often imagined as a scale to be balanced (Antle, Corness
& Bevans, 2011). In this view, abstract concepts are grounded by mapping our experiences
of balancing our bodies onto the concept of justice. Through this mapping, we understand
that justice has something to do with managing excess and deficiency. At the same time,
the converse need not hold. So deficits in the ability to balance need not necessarily produce
difficulties understanding the concept of justice – though a possible relationship, especially
during childhood, offers an interesting avenue of research.
The question of the mechanism of this mapping remains. In conceptual metaphor theory, it is
our experiences of the co-occurrence of a source (balancing) and a target (justice) that creates
a neural mapping to connect the feeling of balancing with the concept of justice (Gallese &
Lakoff, 2005). The circuit that is formed in this way is the metaphor and the experience results
in the metaphor of justice-is-balance in the cognitive domain (Lakoff, 2014). In this model,
since the concrete experience of balancing involves the sensorimotor system, that system is
indirectly involved in representing abstract concepts, like justice, that are based on metaphor
2 The conceptual metaphor theory is not universally accepted. For instance, McGlone (2011) argues that it is
ambiguous, not parsimonious, it suffers from “homunculus problem” and employs circular reasoning. Mahon and
Caramazza (2008) present more general criticism, defending the symbolic nature of abstract knowledge.
11Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
of balance. In this account, the meaning of those concepts depends on being mapped back
into the concrete domain, and therefore back to sensorimotor representations. Pulvermüller
(2018) investigated the mapping that conceptual metaphor theory proposes at the neuro-
mechanistic level of distributed cortical circuits for the abstract verb “to cause”. He argues that
abstract concepts and words can be learned by and grounded through real-life interaction,
because it is only through the neural activity associated with our experiences that the meaning
of abstract concepts can be consolidated.
It is important to keep a distinction in mind between concepts and words. On a neural level,
associations between concepts are not straightforward, because concepts form dense
multimodal networks, related via mental simulators (Barsalou et al. 2008). By contrast,
associations between words are more direct since lexical units are much simpler entities.
This allows the formation of fast associative lexical networks. To some extent, this reflects
the double structure of the two grounding pathways again. In fact, our two-pathway model
recapitulates an aspect of conceptual metaphor theory when we distinguish between a
“deeper” understanding which has a motor component (within the first grounding pathway),
and “shallower” understanding that does not have it (second grounding pathway). This
distinction was originally proposed within the direct matching hypothesis (Rizzolatti & Sinigaglia,
2010), which defines deeper understanding as being mediated by mapping observed actions
to one’s own motor repertoire via mirror neurons (here, the term “direct” is unrelated to its
use in the context of pathways). Consequently, shallower understanding results from cases
where no such mapping is available (e.g. observing someone doing a somersault without
his/her own experience). By extending this idea to experiential context, it is still possible to
conceptually understand something (e.g. justice; via language and social interactions), without
an understanding that maps the concept directly to embodied experience. In this account,
an embodied, first pathway route just provides an added value, understanding “from inside”.
That is not to say that a second pathway grounding is not also embodied, only it is embodied
indirectly, via the way that language and social interactions themselves are ultimately
grounded in embodiment. Within conceptual metaphor theory, this distinction produces the
“dimensions” that constitute the representation of the concept. The motor dimension is one of
these dimensions and its absence does not eliminate understanding as such. Such a distinction
will be useful also in other contexts throughout the paper.3 The next case study takes up this
point and moves even further along the continuum of abstractness, to concepts that cannot
have a physical referent by definition. This would appear to rule out any direct mapping at all
to the sensorimotor system, suggesting they are entirely dependent on the second pathway.
On the surface, it appears that mathematical knowledge should pose a formidable challenge
to embodied cognition. Matrices, vector-spaces and imaginary numbers are all conceptually-
rich entities that do not immediately suggest a practical, action-based origin. Of course,
the foregoing discussion has opened the possibility of grounding concepts in sensorimotor
experience indirectly, via simulation or conceptual metaphors. Mathematics education reveals
that we intuitively exploit the scaffolding role of physical engagement during the acquisition of
mathematical concepts – just consider the role of counting on fingers during early numeracy
training (Hutto, Kirchhoff & Abrahamson, 2015).
This example lends itself to an interpretation according to which mathematical cognition is based
on interactive dynamics with limited to no involvement of contents or concepts – consistent
even with the radical embodied and enactive views (Chemero, 2011). Hutto, Kirchhoff & Myin
(2014) propose a view partially based on basic numerical capacities that exist long before
the formation of mathematical concepts in human development. In part, the argument also
rests on a phylogenetic perspective that emphasises the role of an environment that affords
dynamic interaction as the structuring force behind human cognitive abilities. Again, it is our
experience with them that accounts for the physical reality that abstract concepts, in this case
mathematical ones, appear to have for us; the key may be our basic experiences of size, weight
and even emotive significance (Núñez, 1997).
3 As a limitation, it should be acknowledged, direct mapping applies to (some) concepts related to motor
actions, so it does not provide an overarching principle, especially when it comes to abstract concepts.
12Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
Previous work has argued that an embodied understanding of number concepts can account
for heuristics and biases in mental arithmetic (Fischer & Shaki, 2018). Hayes and Kraemer (2017)
also discuss the grounding provided through laboratory-based and hands-on methodologies in
science and engineering, but somewhat neglect mathematics, which we rectify here. To do so,
we focus on the Kinemathics project (Abrahamson et al 2011, 2012; Abrahamson & Trninic,
2015), as summarised in Hutto et al (2015). The project was intended to design a tutoring
method to help students (United States’ grades 4–6) understand the concept of proportional
relations. The solution that was developed encapsulates the role of embodied learning in
mathematics. In the implementation of the method analysed for the Kinemathics project,
Howison et al (2011) invited students to sit at a desk in front of a large screen. Students see
that the height at which they position their hands in front of a screen moves two cursors up and
down, one for each hand. They are then given the instruction to “make the screen green” by
moving their hands in front of the screen. To make the screen green, students have to hold their
hands apart according to a fixed interval that is unknown to them. The next instruction they
receive is to keep the screen green while moving their hands. This provides an embodied basis
to understand proportionality. The task is successively modified as students learn to orient
and move their hands according to different ratios. First a grid is introduced in the background,
then the gridlines are numbered, then students start moving the cursors by inputting numbers
on a keyboard. This process gradually adds layers of abstraction on top of the newly learned
manual skill to slowly introduce mathematical concepts and notation that are grounded in
sensorimotor interaction.
Micro-ethnographic studies of human development suggest that, throughout cognition, it is
likely goal-oriented physical interaction such as this that acts as the psychological basis of
the human ability to recognize and reproduce symbolic tokens (Hutto et al, 2015; cf. Piaget,
Inhelder, & Szeminska, 1960). In a similar sense, physical operations on objects may constitute
the basis of arithmetic operations and their signs, which simply re-encode a previously implicit
understanding of the corresponding transformation (Radford, 2013). A child might play with two
groups of objects, combine them into a single group being aware of the persistence of the total,
and only later acquire the arithmetic notion of addition and the corresponding notation. This
is what is reported in the Kinemathics project. Here, the grounded cognition literature happens
upon what seems to be a basic process of knowledge acquisition during child development,
which leads it to embrace nativist constraints on how we are able to learn as children (Barsalou,
2016). The idea is that pre-symbolic knowledge identified in sensorimotor interactions captures
a precursor of mathematical knowledge.
What the Kinemathics studies suggest is that the episodes with the training method constitute
the grounding of both the simple mathematical concepts they refer to and, separately, the
grounding of highly abstract mathematical notions. To substantiate this, a subsequent study
used non-invasive eye-tracking to study patterns in the children’s eye gaze (Hutto et al, 2015).
These data indicate that immediately prior to learning a new ratio between their hands,
children’s eye-gaze was directed towards the relevant position that marked the target ratio (e,g,
half way up one side for a ratio of 1:2). Then students typically exclaimed they had understood
the task. This shows an attentional and motor understanding prior to any knowledge of having
“got it”.
One concept that is important for understanding these studies, and by extension for
understanding the embodied argument presented here, is the attentional anchor (Hutto et al,
2015). The term refers to the focus of an agent’s attention during a specific interaction with
its environment. The anchor is made explicit to the learner as her skill increases by engaging in
related tasks (Ingold, 2000). This anchor may serve as a higher-order invariant, rooted in situated
motor interaction that grounds action in order to reduce complexity and enhance control
(Kostrubiec et al., 2012). Achieving attentional anchors is positively related to performance
and learning rate, while the anchor itself reflects the convergence of different idiosyncratic
solutions (Hutto et al, 2015). In the Kinemathics studies, the physical grounding offered by the
interval between the participants’ hands attracted their attention and was brought forward
in experience, as an entity to be manipulated and monitored (Hutto et al, 2015). Individuals
adjusted their motor behaviour prior to their own awareness of such changes (Kelso, 1984),
reflecting what may be a preferred processing route for conceptual knowledge through
physical interaction, as well as a facilitating role of the sensorimotor system for understanding
13Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
abstract concepts. The interval that acts as an attentional anchor is brought into existence by
exploratory motor behaviour and becomes a stimulus for learning in its own right (Hutto et al.,
2015). Furthermore, that gaze preceded speech about a relevant location during the task, as
revealed by the eye-tracking data, suggests that it and the learned motor schemas may be
what provides the basis for discriminations made in speech later on, via implicit, sensorimotor
based reasoning. Consistent with the role this suggests that sensorimotor interactions have for
learning, Abrahamson & Trninic (2015) reported a positive learning outcome for students who
engaged in manual exercises directly or vicariously.
Evidence from adults substantiates this claim by linking gesturing with how (well) we understand
mathematical concepts. On the one hand, adults use embodied language in the domain of
mathematics. For example, mathematicians refer to “never going beyond a limit”. Like with
conceptual metaphors though, the question remains whether this aspect of language points
to anything like a functional embodiment of our understanding of the underlying concepts. The
notions employed, e.g.“approaching a limit”, are likely rooted in physical experience. Without
a distinct mathematical coding, this at least suggests that mathematics recruits embodied
notions similar to how conceptual metaphors work (Marghetis & Núñez, 2013). It also suggests
that mathematical knowledge builds on domain-general cognitive mechanisms, as would be
expected (Fauconnier & Turner, 2003).
On the other hand, gesturing has been shown to be functional in language (Pulvermüller
et al., 2005). Gesturing also facilitates the acquisition of new concepts in proportion to the
faithfulness of gestures to the mathematical notion they are related to (Goldin-Meadow,
Cook, & Mitchell, 2009). By itself, that gestures are used to communicate meaning does not
necessarily implicate embodiment in the fundamental encoding of an abstract concept being
communicated through gestures. However, co-speech gesturing also reflects implicit cognitive
processes not evident in speech (Goodwin, 2000). Marghetis and Núñez (2013) argue that such
implicit, embodied processes are why their data show that dynamic thought, which is reflected
in increased gesture dynamism, facilitated the generation of rigorous mathematical proofs.
Taken alone, this still does not disambiguate the role of gesture dynamism however, ie. whether
it structures mathematical thought or simply supports how we understand or communicate
mathematical knowledge. Nonetheless, Zdrazilova, Sidhu & Pexman (2018) reiterate a crucial
role for gestures in communicating and grounding abstract meaning.
Crucial evidence comes from the ethnographic side again, as it did with Kinemathics. One
study noted that gesturing while explaining scientific figures mirrors the physical structure of
the recording device used to collect the represented data (Roth & Lawless, 2002). That puts
embodiment forward as the basic means of understanding and/or communicating concepts.
In the context of mathematics, gesturing is even, and unusually so, recruited in association
with abstract noun phrases, like “limit” (McNeill, 1992), despite being preferentially related
to verbs in other contexts. Recalling that language may be built upon the gesturing system
on the evolutionary scale (Arbib, 2017), these studies suggest that the gesturing practice of
mathematics educators does provide a meaningful source of grounding during the pedagogical
interaction, for for themselves and, by enacted example, for their students (Roth & Lawless,
2002). Reminiscent of Barsalou et al. (2003b), this also emphasises the social interaction at the
root of learning (Cobb, 1994; Voigt, 2013) – see also Balacheff (1991).
All of this is evidence for an involvement of sensorimotor interaction in learning and doing
mathematics. However, it has been difficult to show that this embodied knowledge constitutes
the ultimate grounding of mathematical concepts. Practically acquiring a mathematical
concept remains a very different process from acquiring a manual skill like playing piano or
basketball (Hutto et al, 2015). Indeed, embodied interaction with the display used to teach
proportion in Kinemathics still requires rich mental constructions (de Freitas & Sinclair, 2012),
which may be of a different kind than those that subserve the practice of piano or basketball.
On the other hand, gesture itself has been argued to reflect the underlying perceptual and
motor simulations that constitute the sediment of embodied language and cognition generally
(Hostetter & Alibali, 2008). In this case, the functional act of gesturing would indicate the deep
embodiment of the multimodal concept in reference to which it occurs.
What all of this is distinct evidence for then, is that mathematical concepts can be brought
forth through action (Abrahamson & Trninic, 2015). It also shows that this interaction can be
14Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
conceptualised in dynamical terms (Hutto et al., 2015). It also seems clear that embodied
engagement has advantages beyond strictly abstract learning methods (Marghetis & Núñez,
2013). To that extent, it is even exhibited by adult experts (Roth & Lawless, 2002). Taken
together, this is compelling evidence from various sources that sensorimotor interaction is
one fundamental mode of cognitive operation with respect to mathematics. Against such a
backdrop, it may be up to the disembodied paradigm to substantiate the long undisputed claim
that mathematical knowledge is an encapsulated, purely abstract domain (Reid & Mgombelo,
Obviously, there exist even more abstract concepts than the examples mentioned above.
Nevertheless, we think that our explanatory approach would be the same. It seems true that
the more abstract the concept is, the more reliance on the second pathway it requires. In the
language pathway, the contextual words should provide information about the semantics
of the target word. Regarding grammatical categories, it is difficult to say what can be more
abstract, whether nouns or verbs. Even though there exist large collections of word concreteness
ratings for various languages (e.g. Brysbaert, Warriner & Kuperman, 2014, for English), their
distributions of ratings are highly overlapping and have very large variance. But, regardless of
the target grammatical category, recent interesting results by Naumann, Frassinelli and Schulte
im Walde (2018) reveal that 1) concrete target nouns, verbs and adjectives primarily co-occur
with concrete nouns, but abstract verbs and adjectives, while 2) abstract target words (nouns,
verbs, adjectives) primarily co-occur with abstract words. This means that the significant
portion of both concrete and abstract target word context refers to words that themselves are
difficult to ground, pointing to the symbolic merry-go-round (Harnad, 1990). On this semiotic
carousel ride, we keep spinning in circles so long as we try to ground abstract concepts on
other abstract concepts. The only way off the merry-go-round is to find concrete concepts in
which to ground more abstract ones. With regard to Naumann et al (2018) we predict that only
the complementary portion of contexts provided by concrete words mediates second pathway
grounding through language. In our three case studies, we focus on the contribution of the first
pathway, which serves as the primary source of grounding. However the “strength” of that direct
grounding may be, admittedly, negatively correlated with the abstractness of the concept.
A crucial aspect that will need to be examined more closely is how exactly learning might
proceed from concrete to more embodied abstract (first pathway) and finally more socially/
interactively mediated abstract concepts (second pathway). The acquisition of mathematical
knowledge, for example, is a gradual process that has to proceed in a certain, incremental
manner as we grow up. To address this, the next section outlines a neurodevelopmental
perspective that provides another fruitful source of evidence for the dynamics of abstract
cognition that can be integrated into the overall picture.
Looking at how our ability to understand concepts develops through childhood is another way
to approach degrees of abstractness. Various neurodevelopmental studies have shown that
abstract conceptual knowledge builds on understanding concrete concepts in particular and
reliable ways. This supports an embodied view of abstraction, if it can be shown that the concrete
basis continues to underpin how we understand abstract concepts, even once children develop
the ability to reason more abstractly. A neurodevelopmental view also sheds light on how the
grounding of abstract concepts shifts from being embodied (first pathway) to being more strongly
tied to language and social interaction (second pathway).
Delving into the neurodevelopmental perspective offers great and essential insight into
how the ability to think abstractly arises (Yee, 2019). As argued by Barsalou, Dutriaux and
Scheepers (2018), abstract thinking is the ability to integrate various relational properties of
elements in a given situation. Humans tend to fully acquire this ability only in late adolescence.
Evidence suggests that growth of new connections and myelination of neurons lasts until early
adulthood (Nickel & Gu, 2018). Recent studies have shown that one of the most prominent
parts of the brain responsible for relational/abstract thinking is the rostro lateral prefrontal
15Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
cortex (RLPFC) (Wendelken et al., 2011; Wright et al., 2008; Dumontheil, 2014). RLPFC has a
prolonged development in children. It has been shown that this neural structure plays a vital
role in processing relational integration of stimuli and aspects of episodic memory retrieval
(Dumontheil, 2014). During tests of processing semantic relations, RLPFC had engaged later in
children than in adults. In children it was activated after a decision was made and the older
the child was, the faster the activation and the bigger the role it played in correctly finishing the
task (Wright et al., 2008). This neuroanatomical evidence proves what is readily observable in
children, they tend to acquire concrete concepts much faster than abstract concepts. Although
our ability to think abstractly comes only later in life, it seems that it is with us from the start.
Concept acquisition starts with natural categories and as mentioned above, goes roughly from
concrete to abstract. Nevertheless, when it comes to common categories that are hierarchically
organised, the acquisition starts at the basic level (Rosch, 1973) that presumably provides
(ecologically) the most relevant conceptual information and/or right amount of variability
among instances (e.g. dog, chair). Subordinate level categories are characterised by a low
degree of generality and by clearly identifiable, detailed and specific features (e.g. Dalmatian, a
rocking chair). On the contrary, superordinate level typically implies a high degree of generality
and allows to store general information (e.g. mammal, furniture). Hence, subordinate categories
are more concrete than basic categories, which in turn are more concrete than superordinate
categories. Interestingly, however, despite the order of increasing abstractness, the acquisition
of knowledge in children does not in this case proceed bottom up (Bloom, 2002).
Actually, according to Borghi and Binkofski (2014), this type of abstraction (e.g. mammal vs. dog)
should be distinguished from abstractness that does not, unlike the former, provide concrete,
perceivable referents (e.g. an electron). They contend that abstractness, not abstraction, is the
real problem posed by abstract concepts. We view abstraction and abstractness as related but
distinct problems posed by abstract concepts. In a developmental context, we can also return
to numbers. Izard et al. (2009) identify a fundamental prerequisite for abstraction, namely
the ability to match events to one another, which they show is already present in newborns.
They conclude that some abstract numerical representations are present from birth. Arguably
therefore, this ability should also be present in utero, and raises a question of when and how
exactly it develops. Such findings raise the profile of nativist approaches and indicate that
a number of essential functions formed prior to birth. Language adds an additional layer of
complexity that most likely builds on such a fundamental function, but also requires abilities
only acquired during post-natal experience. Developmentally speaking, abstraction then
remains difficult, for example, because using some superordinate labels is itself a challenge,
among other reasons due to the sheer variability of referents (including those that very much
depart from typical examples; e.g. a whale does not really look like a mammal). Acquiring this
skill appears to take time.
The common developmental trajectory also points to some differences between more
concrete and more abstract concepts; it is not by chance that more abstract concepts can only
be acquired after many more concrete concepts have been learned and labelled (by words).
Referential uncertainty of abstract words may be the reason why they are acquired later than
concrete words (Bergelson & Swingley, 2013). In more detail, one can expect that the age of
acquisition of a concept linked to a word will be correlated with its degree of abstractness. This
scaffolding effect is consistent with a view on development that argues that infants exploit
what is offered by their environment and their own bodies to construct an understanding of
their world. Such a view introduces a distinction between objects that can be manipulated and
ones that cannot. Indeed, Scorolli et al. (2011) argue that differences between more abstract
and more concrete concepts may be traceable to the way they are acquired. They suggest that
abstract concepts are acquired more linguistically than perceptually, and that this mode of
acquisition produces the way they remain grounded, also in neural terms. On the other hand,
despite the crucial role of language for acquiring abstract concepts, it seems that from neural
perspective, it is not lexical-grammatical categories that drive the organisation of abstract
concepts in the brain but their semantic properties (Moseley & Pulvermüller, 2014).
D’Angiulli, Griffiths, and Marmolejo-Ramos (2015) develop a broad framework of the grounding
of conceptual knowledge of preschool children, using a comprehensive Event-Related Potential
(ERP) investigation of a picture matching task following aural presentation of more abstract or
more concrete words. Their approach draws on results consistent with the dynamic interaction
16Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
vision–language approach (DIVLA) (Mishra & Marmolejo-Ramos, 2010). While this approach
is an embodied one, it differs from the mainstream by emphasising the visual aspects of
sensorimotor interactions, and suggests these have primacy over motor components. While
other modalities also contribute to linguistic processing (e.g. audition, especially considering
the social role of prosody), evidence of a central role of visual processing suggests this is an
especially important system, with a special role in the phylo- and ontogenic development of
language (Givón, 2002).
The results of Mishra and Marmolejo-Ramos (2010) can be interpreted as evidence of a weak,
but pervasive embodiment. The ERP-evidence revealed a dominant presence of activity patterns
associated with visual perceptual processing. This was the case both for more concrete and more
abstract concepts. Concepts on different ends of the continuum tended to differ on the amount
of time during which they engaged visual processing, suggesting a persistent involvement of
the vision-language interface for abstract words. ERPs also offer independent indications of the
involvement of linguistic processes. The overall ERP dynamics suggests a fast visualisation for
more concrete concepts, in contrast with a gradual embodiment of more abstract concepts,
followed by visual processing and a broad recruitment of the vision-language interface for the
purpose of word comprehension. This can be seen as a consequence of the mode of acquisition,
work on which also suggests a developmental trajectory of conceptual knowledge in which more
abstract concepts require more and more complex visualisations (Borghi & Binkofski, 2014).
The particular role of social components of more abstract concepts is a prime example of the
additional complexity that takes a bigger role for more abstract concepts. In the context of DIVLA,
it should be made clear that the perceptual aspects of sensorimotor interactions contribute at
least equally, and possibly more than motor components (Mishra & Marmolejo-Ramos, 2010).
These results are discussed here because they offer an account of cognitive processes necessary
in the developmental context. Importantly, visual and other modal processes take such an
essential role regardless of whether a concept is more abstract or more concrete. Reflecting
on the major developmental contribution of visual contexts of word learning, Iossifova and
Marmolejo-Ramos (2013) found evidence of shift towards body-based processing of abstract
concepts for blind children, compared to sighted and visuomotorily impaired children. This also
reiterates the central role of sensorimotor aspects of grounding. Further underwriting this for
the developing brain, it appears that sensory input during action perception only activates the
motor system after experience of own interactions with objects (James & Swain, 2011). In the
context of the developing sensorimotor system and sensorimotor representations, this places
emphasis on one’s own actions (motor and sensory domain), rather than perceived actions
(sensory domain). Own actions are motor and sensory since there is online sensory perception of
own actions, also via proprioception. That being the case, an early function of motor association
areas may be the association of previous experiences and current sensory input.
Importantly, recalling the indirect pathway of grounding referred to by the present work, there
is also a social dimension to DIVLA; this highlights the role of the shared construction of social
situations through aspects including gestures and prosody, which play a central role not just in
mature language, but also in the process of language acquisition and cognitive development
in general, which is in itself fundamentally based on sensorimotor interactions (Smith, 2005;
Vygotsky, 1964). In this context, words can be understood as tools (Borghi, 2020), and it does
appear that linguistic labels, in addition to and built on sensorimotor experience, are crucial to
the developmental acquisition both of language and of conceptual knowledge. For example,
the higher age at which children tend to acquire abstract concepts might be accounted for by
the referential uncertainty that arises because the physical referent of an abstract is word is
absent more often during mother-infant speech, than is the case for concrete words (Bergelson
& Swingley, 2013). Crucially, this points to infants’ social-cognitive abilities as important aspects
of concept acquisition.
Concerning social interactions, these are believed to facilitate the acquisition of more abstract
concepts, for instance, mediated by valence. Given the profound relevance of emotion to
human life, it is not unsurprising that both positive and negative more abstract words are
acquired earlier than neutral ones (Ponari, Norbury & Vigliocco, 2016). Nonetheless, this
points to another important driver of the development of conceptual knowledge. Meanwhile,
valence is only one aspect of a dimensional approach to emotions and similar effects may
exist for other affective factors, e.g. arousal. The age of acquisition rating method used in this
17Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
study is broadly validated (Scorolli et al., 2011). This method has revealed a central role of the
abstractness–concreteness of the noun during language acquisition, which may facilitate the
acquisition of more abstract verbs over that of similarly abstract nouns. Consistent with this
linguistic contextual effect, Della Rosa et al. (2010) offer the mode of acquisition (Wauters et
al., 2003) as an independent predictor of abstractness–concreteness, in line with the different
contribution of perceptual, motor and linguistic processes during development. To these, the
affective processes should also be added.
Recalling the mapping of abstractness–concreteness and directness of grounding presented at
the outset of the section, an intuitive hypothesis is that the more abstract the word is, the more
difficult it is to acquire (manifested by its later age of acquisition). However, this does not hold
along the entire continuum, as we have seen in case of hierarchically organised object categories.
We have stressed three points about abstractness. First, conceptual knowledge is fundamentally
grounded, and dependent upon embodied experience in interaction with the world. Second, the
production of abstract concepts is not qualitatively different from that of concrete concepts; as
such, they differ only quantitatively in the extent to which they depend on certain information
and not, in principle, in the kinds of information (sources of grounding) a given concept may
depend upon. Third, given the nature of conceptual knowledge and the absence of a qualitative
distinction of concrete and abstract concepts, the abstractness–concreteness dimension should
be thought of as a continuum, as argued also by others.
This paper has outlined several different processes whereby grounded knowledge is produced,
some referring more, some less directly to the sensory periphery. This being the case, we
suggest that these two scales – abstractness–concreteness and directness of grounding – can
be mapped onto one another. Then, the longer and more complex the route back through
metaphors, schemata and simulations, the more abstract a concept will appear. But, there is a
risk that this mapping only reflects the difficulty of accounting for different abstract concepts in
embodied terms; for whatever reason, the grounding of a particular concept may be especially
elusive and therefore tempt us to conclude this must be a very abstract concept. The challenge
is to find a principled solution.
The situation is that the degree of abstractness seems correlated with the “distance” to – and
maybe the complexity of – the sensorimotor processes. This relates to the work on the BOI
effect, which measures how “easy to interact with” words’ referents are perceived. Past research
revealed facilitated lexical and semantic processing for words rated high in BOI than for words
rated low in BOI (Hargreaves et al., 2012). Unsurprisingly then, there are close similarities
between objects high in BOI and the referents of concrete words, and likewise for low BOI and
those of abstract verbs. This close relation of difficulty (of imagining interacting with something)
and concreteness-abstractness reinforces the idea that more abstract concepts are further
from their sensorimotor grounding. For abstract concepts, this “distance” may be much larger
because their meaning is mediated by related words (hence involving the linguistic system), that
in turn might have a shorter distance to sensorimotor grounding. However, in our view, this does
not justify positing the concept of hybrid grounding (e.g. Dove, 2011; Louwerse, 2011). Words as
contextual symbols cannot often provide grounding themselves since they are primarily abstract
(Naumann, Frassinelli & Schulte im Walde, 2018). Still, the second pathway also provides concrete
context consisting of concrete words that provide room for direct grounding. In the paper, we
emphasise the singularly fundamental role of embodiment, which persists as the primary
grounding pathway throughout complex linguistic interactions that are necessarily built upon it.
And as mentioned at the end of Section 2, language has multiple important roles in cognition.
Finally, we argue that the degree of abstractness of a concept is determined by the variability
of its referents, which probably correlates with the aforementioned distance. As an example,
consider the visual domain; the lowest variability occurs for physical objects as a whole (i.e.
the concept refers to the whole object, not its parts or properties). The most concrete concepts
only have one reference (e.g. the moon). Then, there exists the basic level of categorization that
provides a vast number of referents with relatively low variability (e.g. a dog). Next, going toward
a superordinate level, the variability starts to increase (e.g. a mammal). The same applies to
object parts, for instance, body parts that apply to morphologically distinct organisms (e.g.
18Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
the head). This line of thinking is similar to Pulvermüller (2018) who also points to a higher
variability of referents for abstract concepts (such as beauty) compared to concrete concepts
(eye). He also attributes a crucial role of words (i.e. the linguistic system) in establishing the
abstract concepts (by binding operation), which applies, in our view also to the superordinate
level of categorization of concrete objects. The variability of referents of abstract concepts,
typically providing relevant situational contexts, is in general very high, and most probably even
within this domain one could demarcate a certain hierarchy.
The view on graded abstractness has emerged in other recent works, drawing on evidence from
psychology, neuroscience and computational linguistics. One empirical study revealed systematic
quantitative differences between concrete and abstract words, represented in latent semantic
space (Troche, Crutch & Reilly, 2014), obtained from subjective assessments. The authors
investigated word meaning as originally distributed in multidimensional space using hierarchical
cluster analysis. Participants rated 400 English nouns across cognitive dimensions (e.g., polarity,
ease of teaching, emotional valence) and these vectors were then projected onto three latent
factors, corresponding roughly to perceptual salience, affective association, and magnitude. The
visualisation of words in this latent space showed that abstract and concrete words overlapped in
their topography but also differentiated themselves in semantic space. In addition, the method
allowed the authors to represent the degree of concreteness of each word (on Likert scale). This
obviated the need for an artificial dichotomy by treating all psycholinguistic variables as continuous.
Hill, Reichart and Korhonen (2014) also point to a gradual contrast in patterns of organisation
along a continuum from concrete to abstract concepts. The terms association and similarity
refer to the ways the concept pairs (onion, knife) and (onion, carrot) are related: Onion is said
to be (semantically) similar to carrot and associated with (but not similar to) knife. Repeated
analyses using existing word corpora and human judgments have shown the following (see
Hill et al. and references therein): (1) abstract concepts are mainly organised according to
association, whereas concrete concepts are organised according to (semantic) similarity (2)
abstract words have more, but weaker, associations with other words, (3) abstract words have
more symmetric associations than concrete words. This analysis supports a gradual contrast in
patterns of organisation along the continuum.
Chatterjee (2010) takes a neuroscientific approach to graded abstractness. He refers to cases
of patients with motor deficits whose conceptual understanding remained intact, suggesting a
disembodied level of understanding. In our view, this is reminiscent of the distinction between
deep, sensorimotor understanding and shallow, conceptual understanding, as articulated by
Rizzolatti and Sinigaglia (2010) in the context of the mirror neuron theory. Even in the case of
healthy individuals, such a distinction appears reasonable, when considering motor actions
being in one’s own motor repertoire (e.g. nail hammering), compared to those beyond it (e.g.
eagle flying). Drawing on his own work on spatial cognition and language, Chatterjee (2010)
also proposes three functional-anatomic axes in the brain that could be related to graded
abstractness. These are lateral axis that could reveal laterality differences in the processing of
sensory and motor attributes as they relate to concepts, ventral–dorsal axis that appears to
correlate with a shift from rich conceptual information to more schematic relationships, and
centripetal axis that goes from sensory and motor cortices to perisylvian language cortices.
Neuropsychological literature has converged to three hypotheses how concrete and abstract
words differ quantitatively, as reviewed by Hoffman (2016): (1) They differ in their representational
substrates, with concrete words depending particularly on sensory experiences and abstract
words on linguistic, emotional, and magnitude-based information; (2) Abstract words place
greater demands on executive regulation processes because they have variable meanings that
change with context; (3) The relationships between concrete words are governed primarily by
conceptual similarity, while those of abstract words depend more on associations.
A global picture of knowledge organisation in the brain has been proposed by Taylor et al. (2015)
who performed a large-scale meta-analysis of fMRI data. The hypothesis was tested, using
formal methods based on a new cortical graph metrics (network depth) that regions deeper
in the brain (i.e. remote from the sensory inputs) represent more abstract functions. Data-
driven analyses defined a hierarchically ordered connectome, revealing a related continuum
of cognitive functions. The data were collected from participants performing cognitive tasks,
such as sorting words or phrases according to the estimated degree of abstractness. Taylor
19Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
et al. conclude that progressive functional abstraction over the network depth may be a
fundamental feature of the human brain.
Analogical approaches in computational linguistics take advantage of huge linguistic corpora.
The fact that covariation in the world is reflected in the structure of language (Glenberg &
Mehta, 2009) has been repeatedly confirmed in comparing semantic similarities between
words with human judgements. We can also assume that semantically similar words tend to
have similar degrees of abstractness. Hence, in the linguistic representational space there could
exist a nonlinear concrete–abstract dimension (or dimensions) along which the words could
be organised. Alternatively, rather that going deep into semantic space via word embeddings
(which is used in methods such as Latent Semantic Analysis, Global Vectors, Word2vec and
Bidirectional Encoder Representations from Transformers, or BERT), one could remain on the
surface level (see the arguments in Louwerse, 2011) and look at word covariation when searching
for word abstractness. The abstractness of a word could be estimated from characteristics of
words associated with it (by co-occurrence), where we could expect that word abstractness
would be correlated with a higher percentage of abstract words associated with it.
Huge linguistic corpora provide rich extractable information, but despite the wide use of
distributional methods in the natural language processing field, there remains the problem
of symbol grounding (Harnad, 1990). By focusing merely on the form one cannot in principle
retrieve meaning, without additional reference to the world, not even for BERT models
(Devlin et al., 2019) whose performance is otherwise truly impressive (Bender & Koller, 2020).
Similarly, although distributional models display a number of interesting properties (Boleda,
2020), their limitation as an account of cognition seems to be their accuracy. For instance, an
extensive study (Binder et al., 2016) questions the ability of corpus-based word vectors (LSA) to
accurately reflect the similarities between concepts, when they were compared to brain-based
65-dimensional vectors of attributes (comprising sensory, motor, spatial, temporal, affective,
social, and cognitive experiences). Using both methods, the word vectors were constructed for
434 nouns, 62 verbs, and 39 adjectives, by averaging the responses of 1743 participants, and
their study revealed better within- versus between-category separation than representations
derived from distributional (LSA-based) text analysis.
The above references provide rich evidence that abstractness has its correlates, be it psychological
features (measured by subjective assessments of various words), neural features (based on brain
imaging studies), or linguistic features (revealed by corpus analyses).4 In each case, the relation
between these referents and one (or both) of the two pathways of grounding discussed herein.
In this paper we argued that, in principle, all cognition is grounded in experience via two different
pathways: primarily directly through the body (which includes sensorimotor interactions with
the environment, interoception, and emotions), and, as a secondary pathway, indirectly by
means of communication (linguistic and social processes). Departing from a dichotomous
view, we suggest abstractness can be seen as a graded phenomenon, so that abstractness–
concreteness is reconceptualised as a continuum that can be related to various psychological,
neural and linguistic features. As a novelty, we illustrated the grounding of abstract concepts by
three case studies covering the spectrum of abstractness levels (body-specificity, abstract verbs,
maths). We also reviewed neurodevelopmental insights that support differences between these
two pathways of grounding. In this way, we extend an earlier review by Borghi et al. (2017) that
had begun to summarise the crystallisation of the continuum view on abstractness. Crucial
open questions we have raised include a more detailed neuroscientific account of how abstract
conceptual knowledge depends on embodiment (direct grounding) and the sensorimotor brain
systems. At the same time, we have also challenged opponents of an embodied view to offer
alternative explanations to the results of the case studies and related literature we reported
here. One strength of our view in this respect is that it integrates different perspectives on
language as one type of grounding that is itself grounded and embodied. Major challenges
are also faced by the computational approaches to abstraction, where the examples so far
have been limited to toy cases with small vocabularies (Cangelosi & Stramandinoli, 2018).
4 Interestingly, graded abstractness has also been recently identified as one of these four dimensions that
define the notion of another high-level concept: consciousness (Jonkisz, Wierzchoń, & Binder, 2017).
20Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
One promising approach is offered by large-scale distributional methods in natural language
processing, which remain an exception in their field and will probably significantly contribute to
acquisition of abstract concepts in computer models using the second pathway, enriching the
primary sources of concrete and abstract knowledge resulting from interactions with the world.
No ethical approval was required.
The second author was partially supported by TAILOR, a project funded by EU Horizon 2020
research and innovation programme under GA No 952215, and by Cultural and Educational
Grant Agency (KEGA), under GA No K-19-006-00.
We thank both reviewers and the editor for the insightful comments towards improving this
The authors declare that the present research was conducted in absence of any commercial or
financial relationships that could have constituted a potential conflict of interest.
Tim Reinboth
Faculty of Philosophy and Education, University of Vienna, Wien, AT
Igor Farkaš
Faculty of Mathematics Physics and Informatics, Comenius University in Bratislava, Bratislava, SK
Abrahamson, D., Gutiérrez, J. F., Charoenying, T., Negrete, A. G., & Bumbacher, E. (2012). Fostering
hooks and shifts: tutorial tactics for guided mathematical discovery. Technology, Knowledge and
Learning, 17(1–2), 61–86. DOI:
Abrahamson, D., & Trninic, D. (2015). Bringing forth mathematical concepts: signifying sensorimotor
enactment in fields of promoted action. ZDM Mathematics Education, 47(2), 295–306. DOI: https://
Abrahamson, D., Trninic, D., Gutiérrez, J. F., Huth, J., & Lee, R. G. (2011). Hooks and shifts: a dialectical
study of mediated discovery. Technology, Knowledge and Learning, 16(1), 55–85. DOI: https://doi.
Antle, A. N., Corness, G., & Bevans, A. (2011). Springboard: Designing image schema based embodied
interaction for an abstract domain. In D. England (Ed.), Whole Body Interaction (pp. 7–18), London:
Springer. DOI:
Arbib, M. A. (2017). Toward the language-ready brain: Biological evolution and primate comparisons.
Psychonomic Bulletin & Review, 24(1), 142–150. DOI:
Balacheff, N. (1991). The benefits and limits of social interaction: The case of mathematical proof. In
A. Bishop, S. Mellin-Olsen, & J. van Dormolen (Eds.), Mathematical Knowledge: Its Growth Through
Teaching (pp. 173–192). Springer. DOI:
Barsalou, L. W. (1999). Perceptual symbol systems. Behavioural and Brain Sciences, 22(4), 577–609. DOI:
Barsalou, L. W. (2010). Grounded cognition: Past, present, and future. Topics in Cognitive Science, 2(4),
716–724. DOI:
Barsalou, L. W. (2005). Continuity of the conceptual system across species. Trends in Cognitive Sciences,
9(7), 309–311. DOI:
Barsalou, L. W. (2016). On staying grounded and avoiding quixotic dead ends. Psychonomic Bulletin &
Review, 23(4), 1122–1142. DOI:
Barsalou, L. W., Dutriaux, L., & Scheepers, C. (2018). Moving beyond the distinction between concrete
and abstract concepts. Philosophical Transactions of the Royal Society B: Biological Sciences,
373(1752), 20170144. DOI:
Barsalou, L. W., Niedenthal, P., Barbey, A., & Rupert, J. (2003b). Social embodiment. Psychology of
Learning and Motivation, 43, 43–92. DOI:
21Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
Barsalou, L. W., & Wiemer-Hastings, K. (2005). Situating abstract concepts. In D. Pecher & R. A. Zwaan
(Eds.), Grounding Cognition: The Role of Perception and Action in Memory, Language, and Thought (pp.
129–163). Cambridge University Press. DOI:
Barsalou, L. W., Simmons, W. K., Barbey, A. K., & Wilson, C. D. (2003a). Grounding conceptual knowledge
in modality-specific systems. Trends in Cognitive Sciences, 7(2), 84–91. DOI:
Barsalou, L. W., Santos, A., Simmons, W. K., & Wilson, C. D. (2008). Language and simulation in
conceptual processing. In M. de Vega, A. M. Glenberg, & A. C. Graesser (Eds.), Symbols, Embodiment,
and Meaning (pp. 245–283). DOI:
Bedny, M., Caramazza, A., Grossman, E., Pascual-Leone, A., & Saxe, R. (2008). Concepts are more than
percepts: the case of action verbs. The Journal of Neuroscience, 28(44), 11347–11353. DOI: https://
Bekkali, S., Youssef, G. J., Donaldson, P. H., Albein-Urios, N., Hyde, C., & Enticott, P. G. (2021). Is the
putative mirror neuron system associated with empathy? A systematic review and meta-analysis.
Neuropsychology Review, 31, 14–57. DOI:
Bender, E. M., & Koller, A. (2020). Climbing towards NLU: On meaning, form, and understanding in
the age of data. In Proceedings of the 58th Annual Meeting of the Association for Computational
Linguistics, pp. 5185–5198. DOI:
Bergelson, E., & Swingley, D. (2013). The acquisition of abstract words by young infants. Cognition,
127(3), 391–397. DOI:
Binder, J. R., Conant, L., Humphries, C., Fernandino, L., Simons, S., Aguilar, M., & Desai, R. (2016). Toward
a brain-based componential semantic representation. Cognitive Neuropsychology, 1–45. DOI: https://
Boleda, G. (2020). Distributional semantics and linguistic theory. Annual Review of Linguistics, 6, 213–34.
Bloom, P. (2002). How Children Learn the Meanings of Words. MIT press.
Booth, A., Waxman, S., & Huang, Y. (2005). Conceptual information permeates word learning in infancy.
Developmental Psychology, 41(3), 491–505. DOI:
Borghi, A. (2020). A future of words: Language and the challenge of abstract concepts. Journal of
Cognition, 3(1), 42. DOI:
Borghi, A., Barca, L., Binkofski, F., & Tummolini, L. (2018). Varieties of abstract concepts: development,
use and representation in the brain. Philosophical Transactions of the Royal Society of London B:
Biological Sciences, 373(1752). DOI:
Borghi, A., & Cimatti, F. (2009a). Embodied cognition and beyond: acting and sensing the body.
Neuropsychologia, 48, 763–73. DOI:
Borghi, A., & Binkofski, F. (2014). Words as social tools: An embodied view on abstract concepts. Springer
Science & Business Media. DOI:
Borghi, A., Binkofski, F., Castelfranchi, C., Cimatti, F., Scorolli, C., & Tummolini, L. (2017). The challenge
of abstract concepts. Psychological Bulletin, 143, 263–292. DOI:
Borghi, A., & Cimatti, F. (2009b). Words as tools and the problem of abstract word meanings. In N. A.
Taatgen & H. van Rijn (Eds.), In Proceedings of the 31st Annual Conference of the Cognitive Science
Society (pp. 31–2304).
Boroditsky, L. (2003). Linguistic relativity. In L. Nadel (Ed.), Encyclopedia of Cognitive Science (pp. 917–
922). London: Macmillan.
Brookshire, G., & Casasanto, D. (2012). Motivation and motor control: Hemispheric specialization for
approach motivation reverses with handedness. PLOS ONE, 7(4), 1–5. DOI:
Brysbaert, M., Warriner, A. B., & Kuperman, V. (2014). Concreteness ratings for 40 thousand generally
known English Word Lemmas. Behavior Research Methods, 64, 904–911. DOI:
Burgoon, E. M., Henderson, M. D., & Markman, A. B. (2013). There are many ways to see the forest for the
trees: A tour guide for abstraction. Perspectives on Psychological Science, 8(5), 501–520. DOI: https://
Cangelosi, A., & Stramandinoli, F. (2018). A review of abstract concept learning in embodied agents and
robots. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 373(1752).
Casasanto, D. (2009). Embodiment of abstract concepts: Good and bad in right- and left-handers. Journal
of Experimental Psychology: General, 138(3), 351–367. DOI:
Casasanto, D. (2011a). Bodily relativity: The body-specificity of language and thought. In Proceedings of
the 33rd Annual Meeting of the Cognitive Science Society (pp. 1258–1259).
Casasanto, D. (2011b). Different bodies, different minds. Current Directions in Psychological Science, 20(6),
378–383. DOI:
Casasanto, D., & Chrysikou, E. G. (2011). When left is “right”. Psychological Science, 22(4), 419–422. DOI:
22Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
Casasanto, D., Hagoort, P., & Willems, R. (2009). Body-specific representations of action verbs: Evidence
from fMRI in right-and left-handers. In Proceedings of the 31st Annual Meeting of the Cognitive
Science Society (pp. 875–880).
Casasanto, D., & Jasmin, K. (2010). Good and bad in the hands of politicians: Spontaneous gestures
during positive and negative speech. PLOS ONE, 5(7), 1–5. DOI:
Casasanto, D., & Lupyan, G. (2015). All concepts are ad hoc concepts. In E. Margolis & S. Laurence (Eds.),
The Conceptual Mind: New Directions in the Study of Concepts (pp. 543–566). MIT Press.
Chatterjee, A. (2010). Disembodying cognition. Language and Cognition, 2, 79–116. DOI: https://doi.
Chemero, A. (2011). Radical Embodied Cognitive Science. MIT Press.
Clark, A. (1998). Magic Words: How Language Augments Human Computation. In P. Carruthers & J.
Boucher (Eds.), Language and thought: Interdisciplinary themes (pp. 162–183). Cambridge University
Press. DOI:
Clark, A. (2008). Supersizing the Mind: Embodiment, Action, and Cognitive Extension. USA: Oxford
University Press. DOI:
Cobb, P. (1994). Where is the mind? Constructivist and Sociocultural Perspectives on
Mathematical Development. Educational Researcher, 23(7), 13–20. DOI: https://doi.
Connell, L. (2019). What have labels ever done for us? The linguistic shortcut in conceptual processing.
Language, Cognition and Neuroscience, 34, 1308–1318. DOI:
Connell, L. M., Lynott, D. J., & Banks, B. (2018). Interoception: the forgotten modality in perceptual
grounding of abstract and concrete concepts. Philosophical Transactions of the Royal Society of
London B: Biological Sciences, 5, 373. DOI:
Crutch, S. J., & Warrington, E. K. (2005). Abstract and concrete concepts have structurally different
representational frameworks. Brain, 128, 615–627. DOI:
Damasio, A. (1999). The Feeling of What Happens: Body and Emotion in the Making of Consciousness.
Harcourt Brace.
D’Angiulli, A., Griffiths, G., & Marmolejo-Ramos, F. (2015). Neural correlates of visualizations of concrete
and abstract words in preschool children: a developmental embodied approach. Frontiers in
Psychology, 6. DOI:
De Freitas, E., & Sinclair, N. (2012). Diagram, gesture, agency: theorizing embodiment in the
mathematics classroom. Educational Studies in Mathematics, 80(1), 133–152. DOI: https://doi.
Della Rosa, P. A., Catricalà, E., Vigliocco, G., & Cappa, S. F. (2010). Beyond the abstract–concrete
dichotomy: Mode of acquisition, concreteness, imageability, familiarity, age of acquisition, context
availability, and abstractness norms for a set of 417 Italian words. Behavior Research Methods, 42(4),
1042–1048. DOI:
Devlin, J., Chang, M.-W., Lee, L., & Toutanova, K. (2019). BERT: Pre-training of Deep Bidirectional
Transformers for Language Understanding. In Annual Conference of the North American Chapter of
the Association for Computational Linguistics (NAACL-HLT), pp. 4171–4186.
Dove, G. (2011). On the need for embodied and dis-embodied cognition. Frontiers in Psychology, 1, 242.
Dove, G. (2018). Language as a disruptive technology: abstract concepts, embodiment and the flexible
mind. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 373. DOI:
Dumontheil, I. (2014). Development of abstract thinking during childhood and adolescence: The role
of rostrolateral prefrontal cortex. Developmental Cognitive Neuroscience, 10. DOI: https://doi.
Editors of the American Heritage Dictionaries. (2018). American heritage dictionary of the English
language (5th ed.). Boston, MA: Houghton Mifflin Harcourt.
Evans, V. (2016). Design features for linguistically-mediated meaning construction: The relative roles of
the linguistic and conceptual systems in subserving the ideational function of language. Frontiers in
Psychology, 7. DOI:
Fauconnier, G., & Turner, M. (2003). Conceptual blending, form and meaning. Recherches en
Communication, 19, 57–86. DOi:
Fincher-Kiefer, R. (2019). How the Body Shapes Knowledge: Empirical Support for Embodied Cognition.
American Psychological Association. DOI:
Fischer, M., & Shaki, S. (2018). Number concepts: abstract and embodied. Philosophical Transactions
of the Royal Society of London B: Biological Sciences, 373(1752). DOI:
23Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
Gallese, V., & Lakoff, G. (2005). The brain’s concepts: the role of the sensorimotor system in
conceptual knowledge. Cognitive Neuropsychology, 22(3–4), 455–479. DOI: https://doi.
Gibbs, R. W. (2011). Evaluating conceptual metaphor theory. Discourse Processes, 48(8), 529–562. DOI:
Givón, T. (2002). The visual information-processing system as an evolutionary precursor of human. In T.
Givón & B. F. Malle (Eds.), The Evolution of Language out of pre-language (pp. 3–50). John Benjamins
Publishing. DOI:
Glenberg, A. M., & Mehta, S. (2009). Constraints on covariation: It’s not meaning. Italian Journal of
Linguistics, 20, 33–53.
Glenberg, A. M., & Robertson, D. A. (2000). Symbol grounding and meaning: A comparison of high-
dimensional and embodied theories of meaning. Journal of Memory and Language, 43(3), 379–401.
Goldin-Meadow, S., Cook, S. W., & Mitchell, Z. A. (2009). Gesturing gives children new ideas about math.
Psychological Science, 20(3), 267–272. DOI:
Goodwin, C. (2000). Action and embodiment within situated human interaction. Journal of Pragmatics,
32(10), 1489–1522. DOI:
Hargreaves, I., Leonard, G., Pexman, P., Pittman, D., Siakaluk, P., & Goodyear, B. (2012). The neural
correlates of the body–object interaction effect in semantic processing. Frontiers in Human
Neuroscience, 6(22). DOI:
Harnad, S. (1990). The symbol grounding problem. Physica D: Nonlinear Phenomena, 42(1–3), 335–346.
Hayes, J. C., & Kraemer, D. J. (2017). Grounded understanding of abstract concepts: The case of STEM
learning. Cognitive Research: Principles and Implications, 2(1). DOI:
Hill, F., Reichart, R., & Korhonen, A. (2014). Multi-modal models for concrete and abstract concept
meaning. Transactions of the Association for Computational Linguistics, 2, 285–296. DOI: https://doi.
Hoffman, P. (2016). The meaning of “life” and other abstract words: Insights from neuropsychology.
Journal of Neuropsychology, 10(2), 317–343. DOI:
Hostetter, A. B., & Alibali, M. W. (2008). Visible embodiment: Gestures as simulated action. Psychonomic
Bulletin & Review, 15(3), 495–514. DOI:
Howison, M., Trninic, D., Reinholz, D., & Abrahamson, D. (2011). The mathematical imagery trainer: from
embodied interaction to conceptual learning. In G. Fitzpatrick et al. (Eds.), In: Proceedings of the
SIGCHI Conference on Human Factors in Computing Systems (pp. 1989–1998). Vancouver: ACM Press.
Hutto, D. D., Kirchhoff, M. D., & Abrahamson, D. (2015). The enactive roots of STEM: Rethinking
educational design in mathematics. Educational Psychology Review, 27(3), 371–389. DOI: https://doi.
Hutto, D. D., Kirchhoff, M. D., & Myin, E. (2014). Extensive enactivism: why keep it all in? Frontiers in
Human Neuroscience, 8. DOI:
Ingold, T. (2000). The Perception of the Environment: Essays on Livelihood, Dwelling and Skill. Routledge.
Iossifova, R., & Marmolejo-Ramos, F. (2013). When the body is time: spatial and temporal deixis in
children with visual impairments and sighted children. Research in Developmental Disabilities, 34(7),
2173–2184. DOI:
Izard, V., Sann, C., Spelke, E. S., & Streri, A. (2009). Newborn infants perceive abstract numbers.
Proceedings of the National Academy of Sciences of the USA, 106(25), 10382–10385. DOI: https://doi.
James, K. H., & Swain, S. N. (2011). Only self-generated actions create sensori-motor systems in the
developing brain. Developmental Science, 14(4), 673–678. DOI:
Jeannerod, M. (1995). Mental imagery in the motor context. Neuropsychologia, 33(11), 1419–1432. DOI:
Jeannerod, M., & Frak, V. (1999). Mental imaging of motor activity in humans. Current Opinion in
Neurobiology, 9(6), 735–739. DOI:
Jonkisz, J., Wierzchoń, M., & Binder, M. (2017). Four-dimensional graded consciousness. Frontiers in
Psychology, 8. DOI:
Kelso, J. A. (1984). Phase transitions and critical behavior in human bimanual coordination. American
Journal of Physiology, 246(6), R1000–R1004. DOI:
Kostrubiec, V., Zanone, P.-G., Fuchs, A., & Kelso, J. A. S. (2012). Beyond the blank slate: routes to
learning new coordination patterns depend on the intrinsic dynamics of the learner – experimental
evidence and theoretical model. Frontiers in Human Neuroscience, 6. DOI:
24Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
Kousta, S.-T., Vigliocco, G., Vinson, D. P., Andrews, M., & Del Campo, E. (2011). The representation of
abstract words: Why emotion matters. Journal of Experimental Psychology: General, 140(1), 14–34.
Lakoff, G., & Johnson, M. (1980). Metaphors We Live By. Illinois: University of Chicago Press.
Lepora, N. F., & Pezzulo, G. (2015). Embodied choice: How action influences perceptual decision making.
PLOS Computational Biology, 11(4), e1004110. DOI:
Louwerse, M. (2011). Symbol interdependency in symbolic and embodied cognition. Topics in Cognitive
Science, 3(2), 273–302. DOI:
Lupyan, G., & Bergen, B. (2016). How language programs the mind. Topics in Cognitive Science, 8, 408–
424. DOI:
Machery, E. (2009). Doing without concepts. Oxford University Press. DOI:
Mahon, B. Z., & Caramazza, A. (2008). A critical look at the embodied cognition hypothesis and a new
proposal for grounding conceptual content. Journal of Physiology-Paris, 102(1), 59–70. DOI: https://
Marghetis, T., & Núñez, R. (2013). The motion behind the symbols: A vital role for dynamism in the
conceptualization of limits and continuity in expert mathematics. Topics in Cognitive Science, 5(2),
299–316. DOI:
McGlone, M. S. (2011). Hyperbole, homunculi, and hindsight bias: An alternative evaluation of conceptual
metaphor theory. Discourse Processes, 48(8), 563–574. DOI:
McNeill, D. (1992). Hand and Mind: What Gestures Reveal about Thought. University of Chicago Press.
Merriam-Webster. (2016). The Merriam-Webster Dictionary. Encyclopaedia Britannica Inc.
Meteyard, L., Cuadrado, S. R., Bahrami, B., & Vigliocco, G. (2012). Coming of age: A review of
embodiment and the neuroscience of semantics. Cortex, 48(7), 788–804. DOI: https://doi.
Mirolli, M., & Parisi, D. (2011). Towards a Vygotskyan cognitive robotics: The role of language
as a cognitive tool. New Ideas in Psychology, 29(3), 298–311. DOI:
Mishra, R. K., & Marmolejo-Ramos, F. (2010). On the mental representations originating during the
interaction between language and vision. Cognitive Processing, 11(4), 295–305. DOI: https://doi.
Moseley, R. L., & Pulvermüller, F. (2014). Nouns, verbs, objects, actions, and abstractions: Local fMRI
activity indexes semantics, not lexical categories. Brain and Language, 132, 28–42. DOI: https://doi.
Naumann, D., Frassinelli, D., & Schulte im Walde, S. (2018). Quantitative semantic variation in the
contexts of concrete and abstract words. In Proceedings of the 7th Joint Conference on Lexical and
Computational Semantics, New Orleans, USA, pp. 76–85. DOI:
Nickel, M., & Gu, C. (2018). Regulation of central nervous system myelination in higher brain functions.
Neural Plasticity, 2018, 12. DOI:
Núñez, R. (1997). Eating soup with chopsticks: dogmas, difficulties and alternatives in the study of
conscious experience. Journal of Consciousness Studies, 4(2), 143–167
Núñez, R. (1999). Could the future taste purple? Reclaiming mind, body and cognition. Journal of
Consciousness Studies, 6(11–12), 41–60.
Oppenheimer, D. M. (2008). The secret life of fluency. Trends in Cognitive Sciences, 12(6), 237–241. DOI:
Pagis, M. (2010). From abstract concepts to experiential knowledge: Embodying enlightenment in a
meditation center. Qualitative Sociology, 33(4), 469–489. DOI:
Piaget, J., Inhelder, B., & Szeminska, A. (1960). The Child’s Conception of Geometry. Harper.
Ponari, M., Norbury, C., & Vigliocco, G. (2016). Acquisition of abstract concepts is influenced by emotional
valence. Developmental Science, 2. DOI:
Pulvermüller, F. (2018). The case of CAUSE: neurobiological mechanisms for grounding an abstract
concept. Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1752). DOI: https://
Pulvermüller, F., Hauk, O., Nikulin, V. V., & Ilmoniemi, R. J. (2005). Functional links between motor and
language systems. European Journal of Neuroscience, 21(3), 793–797. DOI:
Quandt, L. C., Lee, Y. S., & Chatterjee, A. (2017). Neural bases of action abstraction. Biological Psychology,
129(Suppl. C), 314–323. DOI:
Radford, L. (2013). Three key concepts of the theory of objectification: Knowledge, knowing, and
learning. Journal of Research in Mathematics Education, 2(1), 7–44. DOI:
25Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
Reid, D., & Mgombelo, J. (2015). Survey of key concepts in enactivist theory and methodology. ZDM
Mathematics Education, 47(2), 171–183. DOI:
Rizzolatti, G., & Sinigaglia, C. (2010). The functional role of the parieto-frontal mirror circuit:
interpretations and misinterpretations. Nature Reviews Neuroscience, 11(4), 264–74. DOI: https://doi.
Robbins, P., & Aydede, M. (2009). A short primer on situated cognition. In P. Robbins & M. Aydede (Eds.),
The Cambridge Handbook of Situated Cognition (pp. 3–10), Cambridge University Press. DOI: https://
Rosch, E. H. (1973). Natural categories. Cognitive Psychology, 4(3), 328–350. DOI: https://doi.
Roth, W.-M., & Lawless, D. (2002). Scientific investigations, metaphorical gestures, and the emergence of
abstract scientific concepts. Learning and Instruction, 12(3), 285–304. DOI:
Rüschemeyer, S.-A., Brass, M., & Friederici, A. D. (2007). Comprehending prehending: Neural correlates of
processing verbs with motor stems. Journal of Cognitive Neuroscience, 19(5), 855–865. DOI: https://
Sakreida, K., Scorolli, C., Menz, M., Heim, S., Borghi, A., & Binkofski, F. (2013). Are abstract action words
embodied? An fMRI investigation at the interface between language and motor cognition. Frontiers in
Human Neuroscience, 7, 125. DOI:
Scorolli, C., Binkofski, F., Buccino, G., Nicoletti, R., Riggio, L., & Borghi, A. (2011). Abstract and concrete
sentences, embodiment, and languages. Frontiers in Psychology, 2, 227. DOI:
Šefránek, J. (2008). Knowledge representation for animal reasoning. In Proceedings of the 2nd
International and Interdisciplinary Conference on Adaptive Knowledge Representation and Reasoning
(p. 65–72). Espoo, Finland.
Shea, N. (2018). Representation in Cognitive Science. Oxford University Press. DOI:
Sidhu, D. M., Kwan, R., Pexman, P. M., & Siakaluk, P. D. (2014). Effects of relative embodiment in lexical
and semantic processing of verbs. Acta Psychologica, 149, 32–39. DOi:
Smith, L. (2005). Cognition as a dynamic system: Principles from embodiment. Developmental Review,
25(3–4), 278–298. DOI:
Taylor, P., Hobbs, J., Burroni, J., & Siegelmann, H. (2015). The global landscape of cognition: hierarchical
aggregation as an organizational principle of human cortical networks and functions. Nature, 5,
1–18. DOI:
Thompson, E., & Varela, F. (2001). Radical embodiment: neural dynamics and consciousness. Trends in
Cognitive Sciences, 5(10), 418–425. DOI:
Tomasino, B., & Rumiati, R. I. (2013). Introducing the special topic “the when and why of sensorimotor
processes in conceptual knowledge and abstract concepts”. Frontiers in Human Neuroscience. DOI:
Troche, J., Crutch, S., & Reilly, J. (2014). Clustering, hierarchical organization, and the topography of
abstract and concrete nouns. Frontiers in Psychology, 5. DOI:
Van den Bergh, O., Vrana, S., & Eelen, P. (1990). Letters from the heart: Affective categorization of letter
combinations in typists and nontypists. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 16(6), 1153–1161. DOI:
Van Kleef, G. A., Cheshin, A., Fischer, A. H., & Schneider, I. K. (2016). Editorial: The Social Nature of
Emotions. Frontiers in Psychology, 7. DOI:
Varela, F., Thompson, E., & Rosch, E. (2017). The Embodied Mind, Revised edition: Cognitive Science and
Human Experience. MIT Press. DOI:
Villani, C., D’Ascenzo, S., Borghi, A. M., Roversi, C., Benassi, M., & Lugli, L. (2021b). Is justice grounded?
How expertise shapes conceptual representation of institutional concepts. Psychological Research,
1–17. DOI:
Villani, C., Lugli, L., Liuzza, M. T., & Borghi, A. M. (2019). Varieties of abstract concepts and their multiple
dimensions. Language and Cognition, 11(3), 403–430. DOI:
Villani, C., Lugli, L., Liuzza, M. T., Nicoletti, R., & Borghi, A. M. (2021a). Sensorimotor and interoceptive
dimensions in concrete and abstract concepts. Journal of Memory and Language, 116, 104173. DOi:
Voigt, J. (2013). Negotiation of mathematical meaning in classroom processes: Social interaction and
learning mathematics. In Theories of Mathematical Learning (pp. 33–62), Routledge. DOI: https://doi.
Vygotsky, L. S. (1964). Thought and language. Annals of Dyslexia, 14(1), 97–98. DOI: https://doi.
26Reinboth and Farkaš
Journal of Cognition
DOI: 10.5334/joc.214
Reinboth, T., & Farkaš, I.
(2022). Ultimate Grounding of
Abstract Concepts: A Graded
Journal of Cognition,
5(1): 21, pp. 1–26. DOI: https://
Submitted: 05 June 2021
Accepted: 27 February 2022
Published: 11 March 2022
© 2022 The Author(s). This
is an open-access article
distributed under the terms
of the Creative Commons
Attribution 4.0 International
License (CC-BY 4.0), which
permits unrestricted use,
distribution, and reproduction
in any medium, provided the
original author and source
are credited. See http://
Journal of Cognition is a peer-
reviewed open access journal
published by Ubiquity Press.
Wauters, L., Tellings, A., van Bon, W., & Wouter van Haaften, A. (2003). Mode of acquisition of word
meanings: The viability of a theoretical construct. Applied Psycholinguistics, 24(3). DOI: https://doi.
Wendelken, C., O’Hare, E. D., Whitaker, K. J., Ferrer, E., & Bunge, S. A. (2011). Increased functional
selectivity over development in rostrolateral prefrontal cortex. Journal of Neuroscience, 31. DOi:
Wiemer-Hastings, K., & Xu, X. (2005). Content differences for abstract and concrete concepts. Cognitive
Science, 29(5), 719–736. DOi:
Willems, R. M., Labruna, L., D’Esposito, M., Ivry, R., & Casasanto, D. (2011). A functional role for the
motor system in language understanding: Evidence from theta-burst transcranial magnetic
stimulation. Psychology Science, 22(7), 849–854. DOi:
Willems, R. M., Toni, I., Hagoort, P., & Casasanto, D. (2009). Body-specific motor imagery of hand
actions: Neural evidence from right- and left-handers. Frontiers in Human Neuroscience, 3, 39. DOI:
Wittgenstein, L., & Anscombe, G. (1953). Philosophical Investigations. Basil Blackwell.
Wright, B. S., Matlen, J. B., Baym, L. C., Ferrer, E., & Bunge, A. S. (2008). Neural correlates of fluid
reasoning in children and adults. Frontiers in Human Neuroscience, 1. DOI:
Yee, E. (2019). Abstraction and concepts: when, how, where, what and why? Language, Cognition and
Neuroscience, 34(10), 1257–1265. DOI:
Zdrazilova, L., Sidhu, D., & Pexman, P. (2018). Communicating abstract meaning: concepts revealed in
words and gestures. Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1752).
... Abstract concepts on the other hand do not necessarily have such referents. Therefore, we expect their encoding to mostly reflect information related to internal experience as well as more general linguistic processing, implying brain regions associated with emotional interoceptive and general linguistic processing (Reinboth & Farkaš, 2022). ...
Full-text available
Humans learn and process language in complex and dynamic contexts, involving simultaneous processing of connected speech, faces, bodies, objects and including, at times, touch, odor, and taste. How words and their associated concepts are encoded in the brain during real world processing is unclear. We investigated the representational structure of conceptual encoding in a naturalistic setting and asked to what extent brain responses dynamically depend on context. Across different contexts, we find evidence that concrete and abstract concepts encode experience-based information in roughly differential sets of brain regions. However, these differences reduce when contextual variance is considered. Specifically, the response profile of abstract concepts becomes more concrete-like when they are processed in context highly related to their meaning. Conversely, when context is unrelated to a given concrete concept, the activation pattern resembles more that of abstract conceptual processing. Our results suggest that while concepts encode habitual experiences on average, the underlying neurobiological organization is not fixed but depends dynamically on available contextual information at the time of processing.
Full-text available
Using abstract concepts is a hallmark of human cognition. While multiple kinds of abstract concepts exist, they so far have been conceived as a unitary kind in opposition to concrete ones. Here, we focus on Institutional concepts, like justice or norm, investigating their fine-grained differences with respect to other kinds of abstract and concrete concepts, and exploring whether their representation varies according to individual proficiency. Specifically, we asked experts and non-experts in the legal field to evaluate four kinds of concepts (i.e., institutional, theoretical, food, artefact) on 16 dimensions: abstractness-concreteness; imageability; contextual availability; familiarity; age of acquisition; modality of acquisition; social valence; social metacognition; arousal; valence; interoception; metacognition; perceptual modality strength; body-object interaction; mouth and hand involvement. Results showed that Institutional concepts rely more than other categories on linguistic/social and inner experiences and are primarily characterized by positive valence. In addition, a more subtle characterization of the institutional domain emerged: Pure-institutional concepts (e.g., parliament) were perceived as more similar to technical tools, while Meta-institutional concepts (e.g., validity) were characterized mainly by abstract components. Importantly, for what concerns individual proficiency, we found that the level of expertise affects conceptual representation. Only law-experts associated Institutional concepts with exteroceptive and emotional experiences, showing also a more grounded and situated representation of the two types of institutional concepts. Overall, our finding highlights the richness and flexibility of abstract concepts and suggests that they differ in the degree of embodiment and grounding. Implications of the results for current theories of conceptual representation and social institutions are discussed.
Full-text available
The paper outlines one of the most important challenges that embodied and grounded theories need to face, i.e., that to explain how abstract concepts (abstractness) are acquired, represented, and used. I illustrate the view according to which abstract concepts are grounded not only in sensorimotor experiences, like concrete concepts, but also and to a greater extent in linguistic, social, and inner experiences. Specifically, I discuss the role played by metacognition, inner speech, social metacognition, and interoception. I also present evidence showing that the weight of linguistic, social, and inner experiences varies depending on the considered sub-kind of abstract concepts (e.g., mental states and spiritual concepts, numbers, emotions, social concepts). I argue that the challenge to explain abstract concepts representation implies the recognition of: a. the role of language, intended as inner and social tool, in shaping our mind; b. the importance of differences across languages; c. the existence of different kinds of abstract concepts; d. the necessity to adopt new paradigms, able to capture the use of abstract concepts in context and interactive situations. This challenge should be addressed with an integrated approach that bridges developmental, anthropological, and neuroscientific studies. This approach extends embodied and grounded views incorporating insights from distributional statistics views of meaning, from pragmatics and semiotics.
Full-text available
Theoretical perspectives suggest that the mirror neuron system (MNS) is an important neurobiological contributor to empathy, yet empirical support is mixed. Here, we adopt a summary model for empathy, consisting of motor, emotional, and cognitive components of empathy. This review provides an overview of existing empirical studies investigating the relationship between putative MNS activity and empathy in healthy populations. 52 studies were identified that investigated the association between the MNS and at least one domain of empathy, representing data from 1044 participants. Our results suggest that emotional and cognitive empathy are moderately correlated with MNS activity, however, these domains were mixed and varied across techniques used to acquire MNS activity (TMS, EEG, and fMRI). Few studies investigated motor empathy, and of those, no significant relationships were revealed. Overall, results provide preliminary evidence for a relationship between MNS activity and empathy. However, our findings highlight methodological variability in study design as an important factor in understanding this relationship. We discuss limitations regarding these methodological variations and important implications for clinical and community translations, as well as suggestions for future research.
Full-text available
It is increasingly apparent that sensorimotor information is a constitutive part of conceptual knowledge. Yet all concepts, even highly concrete ones (e.g. dog) include information that is abstracted across individual episodes of experience, departing somewhat from direct sensory or motor input. This process of abstraction is the essence of conceptual structure. This Special Issue brings together developmental, experimental, computational and cognitive neuroscientific perspectives on abstraction. The contributions address questions like: When during development do our concepts become less directly tied to sensory or motor knowledge? How (and where) in the brain does the process of abstraction happen? And what is the role of a concept’s label in abstraction? In answering these questions, the contributions highlight that context–the developmental contexts of our first episodic experiences, and the linguistic contexts that accompany the development of conceptual knowledge in both children and adults–is at the root of conceptual knowledge.
Full-text available
The issue of how abstract concepts are represented is widely debated. However, evidence is controversial, also because different criteria were used to select abstract concepts – for example, imageability and abstractness were equated. In addition, for many years abstract concepts have been considered as a unitary whole. Our work aims to address these two limitations. We asked participants to evaluate 425 abstract concepts on 15 dimensions: abstractness, concreteness, imageability, context availability, Body-Object-Interaction, Modality of Acquisition, Age of Acquisition, Perceptual modality strength, Metacognition, Social metacognition, Interoception, Emotionality, Social valence, Hand and Mouth activation. Results showed that conceiving concepts only in terms of concreteness/abstractness is too simplified. More abstract concepts are typically acquired later and through the linguistic modality and are characterized by high scores in social metacognition (feeling that others can help us in understanding word meaning), while concrete concepts obtain high scores in Body-Object-Interaction, imageability, and context availability. A cluster analysis indicated four kinds of abstract concepts: philosophical-spiritual (e.g., value), self-sociality (e.g., politeness), emotive/inner states (e.g., anger), and physical, spatio-temporal, and quantitative concepts (e.g., reflex). Overall, results support multiple representation views indicating that sensorimotor, inner, linguistic, and social experience have different weights in characterizing different kinds of abstract concepts.
Full-text available
How can we understand causal relationships and how can we understand words such as ‘cause’? Some theorists assume that the underlying abstract concept is given to us, and that perceptual correlation provides the relevant hints towards inferring causation from perceived real-life events. A different approach emphasizes the role of actions and their typical consequences for the emergence of the concept of causation and the application of the related term. A model of causation is proposed that highlights the family resemblance between causal actions and postulates that symbols are necessary for binding together the different partially shared semantic features of subsets of causal actions and their goals. Linguistic symbols are proposed to play a key role in binding the different subsets of semantic features of the abstract concept. The model is spelt out at the neuromechanistic level of distributed cortical circuits and the cognitive functions they carry. The model is discussed in light of behavioural and neuroscience evidence, and questions for future research are highlighted. In sum, taking causation as a concrete example, I argue that abstract concepts and words can be learnt and grounded in real-life interaction, and that the neurobiological mechanisms realizing such abstract semantic grounding are within our grasp. This article is part of the theme issue ‘Varieties of abstract concepts: development, use and representation in the brain'.
Recent theories propose that abstract concepts, compared to concrete ones, might activate to a larger extent interoceptive, social and linguistic experiences. At the same time, recent research has underlined the importance of investigating how different sub-kinds of abstract concepts are represented. We report a pre-registered experiment, preceded by a pilot study, in which we asked participants to evaluate the difficulty of 3 kinds of concrete concepts (natural objects, tools, and food concepts) and abstract concepts (Philosophical and Spiritual concepts, PS, Physical Space Time and Quantity concepts, PSTQ, and Emotional, Mental State and Social concepts, EMSS). While rating the words, participants were assigned to different conditions designed to interfere with conceptual processing: they were required to squeeze a ball (hand motor system activation), to chew gum (mouth motor system activation), to self-estimate their heartbeats (interoception), and to perform a motor articulatory task (inner speech involvement). In a control condition they simply rated the difficulty of words. A possible interference should result in the increase of the difficulty ratings. Bayesian analyses reveal that, compared to concrete ones, abstract concepts are more grounded in interoceptive experience and concrete concepts less in linguistic experience (mouth motor system involvement), and that the experience on which different kinds of abstract and concrete concepts differs widely. For example, within abstract concepts interoception plays a major role for EMSS and PS concepts, while the ball squeezing condition interferes more for PSTQ concepts, confirming that PSTQ are the most concrete among abstract concepts, and tap into sensorimotor manual experience. Implications of the results for current theories of conceptual representation are discussed.
Distributional semantics provides multidimensional, graded, empirically induced word representations that successfully capture many aspects of meaning in natural languages, as shown by a large body of research in computational linguistics; yet, its impact in theoretical linguistics has so far been limited. This review provides a critical discussion of the literature on distributional semantics, with an emphasis on methods and results that are relevant for theoretical linguistics, in three areas: semantic change, polysemy and composition, and the grammar–semantics interface (specifically, the interface of semantics with syntax and with derivational morphology). The goal of this review is to foster greater cross-fertilization of theoretical and computational approaches to language as a means to advance our collective knowledge of how it works. Expected final online publication date for the Annual Review of Linguistics, Volume 6 is January 14, 2020. Please see for revised estimates.
Multi-modal models that learn semantic representations from both linguistic and perceptual input outperform language-only models on a range of evaluations, and better reflect human concept acquisition. Most perceptual input to such models corresponds to concrete noun concepts and the superiority of the multi-modal approach has only been established when evaluating on such concepts. We therefore investigate which concepts can be effectively learned by multi-modal models. We show that concreteness determines both which linguistic features are most informative and the impact of perceptual input in such models. We then introduce ridge regression as a means of propagating perceptual information from concrete nouns to more abstract concepts that is more robust than previous approaches. Finally, we present weighted gram matrix combination, a means of combining representations from distinct modalities that outperforms alternatives when both modalities are sufficiently rich.