Issues in Grounded Cognition and How to Solve Them – the Minimalist Account

2.1. Perceptual symbol systems

Perceptual symbol systems as a theory states that concepts are represented by multimodal re-enactments of percepts involved when experiencing those concepts (Barsalou, 2008, 2020).⁵ It argues that fragments of multimodal perceptions are isolated to become represented as perceptual symbols (Barsalou, 1999). When later representing the concept, these perceptual symbols are recombined into so-called simulators that are equivalent with concepts and form the basis of all cognition. For example, when someone mentions that “there is a cup on the table”, various contextually-relevant perceptual symbols (the shape of a table or the color of a cup) are activated in a simulator, which constitutes the mental representation. This holds even for abstract concepts which differ from concrete concepts by their grounding substrate: Concrete concepts mostly involve simulations of perceptual symbols from the modalities of the sensorimotor system, while abstract concepts involve simulations of interoception, or meta-cognitive states (Barsalou & Wiemer-Hastings, 2005; Kiefer & Barsalou, 2013). For example, in order to verify the concept of truth, the simulation of that coffee cup on a table can be mapped to the visual perception of seeing it, and the meta-cognitive appraisal of successfully matching the two becomes the abstract concept of truth. Perceptual symbol systems emphasizes the importance of interoception and meta-cognitive states, but this is not to say that other theories do not also purport grounding via these modalities.

The details of this theory, and how it is applied, differ depending on the authors and the foci of their respective research. For example, it has been differently named with different emphases. Situated conceptualization argues that simulators are rich context-dependent representations, often with a focus on emotion (Lebois et al., 2020; Wilson-Mendenhall et al., 2011). The perspective effect, where the perspective from which a scenario is described affects what attributes of an object become particularly salient, also supports this view (Borghi et al., 2004; Borghi & Barsalou, 2021). Another example includes modality-switch costs, which reflect concepts described in one modality inhibiting concepts associated with another modality (Pecher et al., 2003). This view was extended further to become the language as situated simulation hypothesis (Barsalou et al., 2008). It emphasizes more than its predecessors the role of language as a shortcut or pointer to concepts. This understanding of cognition as experiential simulation connects to the larger situated action cycle which describes how such modal simulations connect to action (Barsalou, 2020, 2021).

There is an important difference between the perceptual symbol systems theory and the mechanism of simulation, despite them often being equated. The theory brings with it the constructs and processes posited in seminal theory papers (Barsalou, 1999, 2008), including simulators which are equivalent to a concept and convergence zones, which enable abstraction (Simmons & Barsalou, 2003). Yet when other work references this theory, it typically intends to signify allegiance to the single mechanism, not the theory. The mechanism of simulation only reflects the covert re-activation of modal perceptions (Pezzulo, Candidi, et al., 2013; Reilly et al., 2023), without the added theoretical baggage, like simulators or supramodal convergence zones (although, indeed, abstraction will be an important ability for future research to explain). This distinction will become important when we propose our minimalist account below.

2.2. Conceptual metaphor theory

In contrast to perceptual symbol systems, conceptual metaphor theory states that concepts are represented by re-using representations from other, more concrete, domains to represent a target domain (Gibbs, 2005a; Kövecses, 2020a; Lakoff & Johnson, 1999). More specifically, it argues that throughout human experience primitive schemas (e.g., CONTAINER, which originates from the many instances of bottles, cups, or boxes holding other objects) repeat and become cognitively represented in the form of structured patterns (Johnson, 2018). These patterns, called image schemas, are transformed in a limited number of ways, such as by the ‘multiplex-to-mass’ or ‘following-a-trajectory’ transformations (Gibbs & Colston, 1995). These transformations turn image schemas into conceptual metaphors (Johnson, 1987; Lakoff, 1987; Langacker, 1986; Talmy, 1983) which constitute the mental representation of the concept. An example for this chain of transformations is the image schema of SOURCE-PATH-GOAL which is applied to the target domain of love, to become the conceptual metaphor LOVE-IS-A-JOURNEY (which is the origin for sayings like “look how far we’ve come” or “we have reached the end”; Gibbs, 2005a, p. 92). This then also elicits embodiment signatures in behavior (Gibbs, 2006). For example, the abstract concept of time is understood by redeploying the more concrete domain of space, such that forward movement in space leads to a change in the representation of the abstract concept time, toward the future (Boroditsky, 2000; Loeffler et al., 2017). In short, conceptual metaphor theory argues that, via image schemas, all conceptual processing is reasoning that is either image-based (Lakoff, 1990) or “visual, auditory, kinesthetic, and tactile” (Gibbs & Colston, 1995, p. 394). Primary metaphors are also relevant, although their relation to the other constructs is not fully clear (Grady, 1997, 2005) and they do not appear in all texts describing conceptual metaphor theory (e.g., B. Winter & Matlock, 2017). A more recent development of this theory argues for different levels in the hierarchy of representation, consisting of the image schema at the most abstract level, followed by the domain, the frame, and the mental space as the least abstract level (Kövecses, 2017, 2020a).

As in perceptual symbol systems theory, reference to conceptual metaphor theory often does not signify allegiance to the full theory, but rather to its core mechanism, which in this case is metaphoric mapping (see Table 1). The full theory involves image schemas, transformations, and primary metaphors. The mechanism of metaphor, and the way it is most frequently referenced, merely describes that an abstract domain is represented by re-deploying representations from a concrete domain (Casasanto & Gijssels, 2015; Jamrozik et al., 2016). This view of metaphor relates to the action schema approach (Glenberg et al., 2008), a theory often posited to be distinct from conceptual metaphor theory (see Table 1). A prototypical example from this theory is the work of Glenberg and Kaschak (2002) who showed that linguistic descriptions of object transfer along the midsagittal plane (toward or away) influence response time when responding with movements along the same axis. Importantly, in relation to metaphoric mapping, this effect holds also for abstract descriptions like “she told you the story”. This action sentence compatibility effect (ACE) suggests that meaning comprehension is related to action planning and that comprehension of even abstract concepts relies on action schemas. The action schema approach, despite relying on the same mechanism as conceptual metaphor theory is treated as separate, and this repetition is left unaddressed.

3.1. Lack of Conceptual Clarity

Much progress in science revolves around commensurability of theories: when two theories address the same phenomenon, they can be integrated to inform each other and gradually expand to explain novel observations (Kuhn, 1962; Oberheim & Hoyningen-Huene, 2009). Incommensurability, on the other hand, describes a state in which theories cannot be combined, threatening progress. Incommensurability is often the result of a lack of conceptual clarity (Hagger, 2014) because theories that use different terms to address the same phenomena cannot be integrated. This often results from conceptual clutter (Jones et al., 2016; Morrison & Grammer, 2016) and related, the jingle-jangle fallacy (Block, 1995; Kelley, 1927), where similar terms for different constructs (jingle) and different terms for the same construct (jangle) co-exist (Dang et al., 2020; Wulff & Mata, 2023). Conceptual clutter prevents the cumulative progress of science because theories cannot converge in order to build on one another (Gerring, 1999; Locke, 2003).

Conceptual clutter is found throughout grounded cognition theories, highlighting the need for commensurate constructs. For example, conceptual metaphor theory proposes the existence of image schemas (also called “cognitive primitives” in Lakoff, 2012, or “schema structure” in Lakoff, 2014) and primary metaphors. In a coherent theoretical framework, these two components should be co-occurrent, yet many important works exist where only image schemas, and not primary metaphors, are mentioned (e.g., Gibbs, 2005b; Richardson et al., 2003). Alternatively, sometimes only primary metaphors are discussed (Lakoff, 2014; Winter & Matlock, 2017; although Lakoff uses the term “primitive metaphor”, which may be the same as cognitive primitives, another name for image schemas; Lakoff, 2012), or image schemas and primary metaphors as taking place on the same level (e.g., Kövecses, 2020b). Furthermore, beyond image schemas and primary metaphors, a wealth of other categories exists. Some authors distinguish structural, orientational, and ontological metaphors (e.g., Łącka-Badura, 2016; Lakoff & Johnson, 1980), or spatial relation- and force-type ‘cogs’ (another term for cognitive primitives, i.e., image schemas; Lakoff, 2012; also attributional vs. relational metaphors; Gentner et al., 2001; or resemblance vs. correlation-based metaphors; Grady, 2008). Even when these terms are mentioned alongside one another, their functions within theories are not consistent: Some authors state that image schemas are the basic building blocks (Kövecses, 2008), others focus on primary metaphors (Grady, 1997; Lakoff, 2014). Without this consistency in their function, these terms become useless as a reference. A similar problem arises from the absence of an accepted or widely used definition of image schemas (Pecher et al., 2011). They are at times described as kinetic analog representations of spatial relations (Gibbs & Colston, 1995), experiential (Gibbs, 2005b) or schematic (Kimmel, 2009) gestalts, preconceptual structures, etc. (cf. Hampe, 2005; Mandler & Pagán Cánovas, 2014). It is also unclear how many there are, ranging from around two dozen (Johnson, 1987; Lakoff, 1987) to hundreds (Cienki, 1997; Johnson, 2018).

Across grounded cognition we find jingle and jangle in constructs, constituting many slightly-different versions of the same theory, thereby preventing continuity between and within theories. The various names given to grounded cognition theories are examples of jangle (for an overview see Table 1), being very similar but different-sounding. In conceptual metaphor theory some theoretical enclaves (new versions of theories which differ slightly from the original but nonetheless rely on many of the same central posits) are Lakoff’s (1987) work, Boroditsky’s metaphoric structuring account (Boroditsky, 2000), the neural theory of language (Gallese & Lakoff, 2005), a novel extended conceptual metaphor theory (Kövecses, 2020a), or just metaphor theory (Merritt et al., 2010). Similar jangle exists in perceptual symbol systems theory which has been called: perceptual symbol systems (Barsalou, 1999) or perceptual symbol systems theory (Horchak et al., 2014), situated conceptualization (Barsalou, 2009), situated simulation (Barsalou, 2003b), language as situated simulation (Barsalou et al., 2008), or often just simulation (Glenberg et al., 2008). These are, of course, not to be confused with the ‘embodied simulation theory’ (Gallese, 2005; Gallese & Sinigaglia, 2011) or the many other different names for ‘simulation’ theories (Pezzulo, Candidi, et al., 2013), which presents an instance of jingle.

These examples of conceptual clutter and jingle-jangle are symptoms of premature advancement, and generate obstacles to meaningful theoretical progress. Firstly, theories that attempt to be elaborate and detailed, without adequately understanding their most fundamental components, result in many enclaves and disputes about details of constructs that are in themselves speculative. Second, prematurely specific theorizing closes the field off to valuable descriptive and exploratory (as opposed to ever more theory-laden) research, which only exacerbates the development of theoretical enclaves. This triggers a cycle in which development is impeded because the theories are not constrained by basic phenomena, which cannot be discovered as a result of little descriptive research, which is in turn the result of conceptual clutter, which prevents discovering basic phenomena to constrain theories, and so on. Together, these trends all prevent pruning of unsupported constructs. The resulting impediment of theoretical progress is evident. If even within conceptual metaphor theory there are many different versions of a theory, which each only permits tests within itself, how should this form a coherent whole to contest against perceptual symbol systems theory?

Philosophers of science have proposed different solutions for a lack of conceptual clarity (cf. Podsakoff et al., 2016; Wulff & Mata, 2023), but we argue that grounded cognition is just not yet ready for grand theories which can justifiably posit such specific constructs. Instead, the field should ‘go back to basics’ (Bringmann et al., 2022). We argue that going back to basics consists of reverting to a silhouetted and non-detailed point of departure, and progressing with bottom-up, inductive research (cf. Aguinis & Vandenberg, 2014). This approach ensures that there are relatively few assumptions.

We choose to focus on two basic mechanisms in the following minimalist account, without assuming any further details. This simplification is in contrast to conceptual metaphor theory’s postulating a limited number of image schema transformations, which already takes for granted the existence of image schemas. Its further claim about a limited number of transformations in turn assumes that something innately limits the amount of transformations, and so on. Without making such unfounded assumptions, the minimalist account should be attractive, no matter which theory’s prior assumptions one disagrees with. Moreover, when using the minimalist account, any departure from, or extension to, the theory should be well-evidenced, which should in turn limit the creation of theoretical enclaves. A common overarching theoretical framework such as this prevents conceptual clutter by generating commensurability between camps of grounded cognition, and across cognitive science, therefore encouraging a coherent research program.

3.2. Unsystematic Empirical Work

How does empirical work position itself when multiple incommensurable enclaves of theories exist? Some authors have argued it generates “a potpourri of disconnected empirical findings” (Muthukrishna & Henrich, 2019, p. 2), or in other words, unsystematic empirical work. Empirical work in grounded cognition is characterized by two strategies: authors either do not align with one specific theory, preferring to avoid the murky waters of conceptual clutter and theoretical baggage; or they already agree on the theory’s basic assumptions (e.g., the existence of simulators). Both strategies fail to serve the goal of falsification. Perhaps most intuitively, this problem is reflected in the fact that empirical work prefers to reduce overly elaborate theories to their basic mechanisms during operationalization, suggesting that experimenters intuit that this is the boundary of what is currently known.

The lack of rigorous theory-testing is harmful, and yet pervasive in the field, as demonstrated by both experimentally-driven and theoretical work. First, experimental work that is clearly located within grounded cognition, and that also draws on related theoretical and empirical work (as indicated by citations), often does not refer explicitly to either “embodied cognition” or “grounded cognition” at all, let alone to specific theories (e.g., Dils & Boroditsky, 2010; Kaschak et al., 2005). On the other hand, the authors could declare allegiance to a specific theory (e.g., Richardson et al., 2001), but this incurs a full allegiance to all detailed postulates, which is rare. Additionally, because of the strong assumptions which accompany these theories, the outcomes cannot subsequently be compared against another theory. For this reason, common practice is less stringent: In most cases where allegiance is proclaimed, it smudges over the theories within the field. There is reference to so-called “embodied effects” that are theoretically backed by citing vague entities like “theories of embodied cognition” (Jostmann et al., 2009, p. 1169) or the “embodied cognition approach” (Fischer & Brugger, 2011, p. 5; Tolentino-Castro & Raab, 2021, p. 1; Werner et al., 2019, p. 171), which are similarly-vaguely defined by e.g., “thoughts comprise simulations of bodily experience” (Casasanto, 2009, p. 351) or “a bidirectional interplay between cognitive processes and the human body” (Werner et al., 2019, p. 171), followed by a listing of citations including popular works from both perceptual symbol systems and conceptual metaphor representatives of the grounded cognition field. Without allegiance to a specific theory, the empirical work evidently cannot provide strong evidence in support or against it.

Second, theoretical work often involves authors cherry-picking experimental results collected under the name of a different theory. To illustrate, Anderson (2008) underlines three findings that have been deployed to support multiple different, ostensibly contradictory, grounded cognition theories. One of these findings is the seminal ACE (Glenberg & Kaschak, 2002), which has been cited in support of the authors’ own theoretical framework of action schemas, as well as in support of perceptual symbol systems and also its opponent, conceptual metaphor theory (see below for the minimalist account’s parsimonious solution to this). Even work by critics of grounded cognition is chosen as support of grounded cognition when deemed suitable (Anderson, 2008; Ostarek & Bottini, 2021). In other words, the ‘theory-testing’ practices in grounded cognition research often involve theories not being tested, but rather experimental results being selected post-hoc, resulting in confirmation bias.

In contrast to these practices, successful research programs engage in small-step theory-building, followed by systematic testing of predictions deduced from these theories (Kitcher, 1995; Popper, 1959). The field of grounded cognition is currently performing discovery-oriented research that is focused on confirming ostensibly surprising ‘embodied cognition effects’, but should move to theory-testing research (Oberauer & Lewandowsky, 2019). It has found enough headline results (discoveries) to be taken seriously as standard cognitive science’s antagonist, and should now begin testing its theories by falsification (cf. also Barsalou, 2010; Fischer, 2024; Popper, 1959). It could even go a step further and perform strong-inference testing, pitting two ‘critical’ hypotheses derived from opposing theories against one another (Ostarek & Bottini, 2021; Platt, 1964). To foster engagement in falsification or strong-inference, a grounded cognition account without unnecessary assumptions must be put forward to which experimental work aligns itself comfortably without buying into a host of theoretical addenda. Therefore, this problem too, is counteracted by going back to basics with a framework of reduced detail and agreed-upon mechanisms.

Beyond not serving falsification, prematurely advanced theories are also closed off from descriptive research, which constitutes the foundation of the topic under investigation (Bringmann et al., 2022; Rozin, 2001). In contrast to discovery-oriented research, which has a surprising hypothesis that is based on a theory which it aims to confirm, descriptive work merely involves understanding a phenomenon better, often with few top-down expectations that it aims to confirm. Prematurely advanced theories force exploratively-minded researchers to buy into significant amounts of theoretical detail or into smudging over theoretical allegiance. The problem with examples from empirical research listed above does not rest with their authors. Researchers often perform basic descriptive or explorative work, which is a valuable and necessary endeavor. This holds especially in a field such as grounded cognition which has limited background knowledge of its context and basic foundation, as explicated above. Yet, researchers are rightly hesitant to align with overly elaborate theories for risk of incurring unwanted theoretical baggage (e.g., assuming image schemas buys in transformations and primary metaphors). This leads researchers to instead smudge over the specific theoretical allegiance and instead reference a wide set of theories. Yet, experimental research, whether explorative or not, is required to be situated within the bigger theoretical picture. If not situated in a bigger theoretical picture, as is the case for much research on grounded cognition, empirical work cannot build on one another and each experiment must restart at zero. This is why the premature advancement has stifled progress: For a field like grounded cognition, which may not yet have such expansive foundational knowledge, descriptive and explorative research is critical; prematurely advanced theories are closed off from, and indeed obstruct, such work.

This problem is evident in the relative lack of basic knowledge, and comparatively much discussion, about details of concept grounding. For example, while the complicated process of how image schemas transform into conceptual metaphors has been exhaustively discussed, the basic description of the frequency of metaphor use for various abstract concepts has received comparatively little (none, to our knowledge) attention. Similarly, while there exists dispute on the intricacies of simulators in convergence zones, there have been few explorative assessments of what types of representations can form grounding substrates (see below). This practice carries real effects later in the research cycle.

One example of the detrimental effects of a lack of descriptive knowledge comes from the non-replication of the aforementioned ACE by Morey et al. (2022; see also Papesh, 2015). These authors, among them the original study’s creators, performed a stringent replication and failed to find the effect. Yet, it has since been argued that this non-replication originates from subtle differences in experiment designs, and that a lack of general understanding of grounding is to blame (Teskey et al., 2024). For example, a third variable could be responsible, as suggested by the fact that there is strong inter-individual variability in the size and direction of the effect (A. Winter et al., 2022). Alternatively, minute differences in stimuli might be to blame, because replication is successful when the focus of the sentence is on the action (Jin et al., 2024). This opaqueness is further reflected by authors of the non-replication, who themselves state that their null result does not speak against sensorimotor grounding, but rather that their experiment is not effective in uncovering the effect (Morey et al., 2022). This is a clear signal for a lack of foundational knowledge. In fact, this replication, cites various contrary approaches, but not the approach which ACE is commonly brought in connection with: the action schemas approach (see e.g., Borghi et al., 2017; Dove, 2011; Glenberg et al., 2008). The original studies on the ACE effect cited multiple frameworks, including the indexical hypothesis (Glenberg & Robertson, 1999), which is mostly absent from later work discussing the ACE effect (which frames it as support for the action schemas approach). Research on the ACE effect had no overarching theoretical framework, and therefore studies which struggled to elicit the effect could not be meaningfully connected, producing a potpourri of contextual effects. With more descriptive knowledge about the basic mechanisms, integrated into an overarching theoretical framework, empirical research on the effect could be better informed and reasoned about. This descriptive knowledge could, for example, be the time course of ACE, namely that simulations only reach a detectable threshold after repeating the inductive stimulus for longer than a certain time.⁸

In short, the prematurely advanced theories contribute to preventing a valuable description and characterization of phenomena, thereby leading to the aforementioned potpourri of disconnected empirical findings (Haig, 2013; Muthukrishna & Henrich, 2019), and finally a lack of cumulative progress. For this reason, the field requires an overarching framework which allows descriptive and explorative works to find a home and build on one another.

4.1. Exposition of the Minimalist Account

With the field of grounded cognition being defined as understanding how mental representations of concepts gain meaning by their embedding in the modalities (see above), we build on this description by proposing two central mechanisms through which modal activation can represent concepts. In short, the minimalist account consists of taking only the core mechanisms from both of the above main theories of grounded cognition, and discarding the assumption-laden theoretical baggage.⁹ First, our account of grounded cognition simply states that concepts are simulated directly (cf. Barsalou, 2003b; Barsalou et al., 2008). The mechanism of simulation posits that a concept is represented by re-activating the mental or sensorimotor states which were active during the experience of the concept (Fischer & Zwaan, 2008; Körner et al., 2023). Second, it states that these simulations can represent other, more abstract concepts. In other words, they can be metaphorically re-used (from conceptual metaphor theory; Jamrozik et al., 2016). The metaphoric mapping mechanism involves applying the representation from a concrete source domain to understand an abstract target domain. In other words, what is represented is not the abstract concept itself, but a concrete experience, similar enough to form a sufficiently useful representation of the abstract concept (Casasanto & Gijssels, 2015). This means that, under the minimalist account, concept representations consist of simulations of the experiences of the respective concept, and that abstract concepts are especially likely to involve re-deploying simulations that were previously acquired for concrete concepts. This suggests also a nesting of these concepts, with simulation being universal and metaphoric mapping being a specific type of simulation. Beyond these basic mechanisms, any further constructs will require strong empirical support to justify their inclusion into grounded cognition theorizing.

4.2. Background of the Minimalist Account

To arrive at the minimalist account, we inferred from prior work those parts that are 1) frequently used, within grounded cognition and beyond, and 2) natural kinds, i.e., constructs which exist independently of specific theoretical commitments (for full explanation, see below; Fried, 2017). This second constraint, of natural kinds, is a critical constraint to enable the kind of integration which we propose (see section 4.4 Qualities of the Minimalist Account). It is possible that there are other constructs that would meet our criteria, but we focused on these mechanisms because they are prominent in grounded cognition and allow integrating its productive research history. It is also necessary for the chosen mechanisms to originate from theory because this allows research applying these mechanisms to explain phenomena, as opposed to just describe them (Brick et al., 2022; van Rooij, 2019). We now justify each criterion in turn.

Firstly, demonstrating the frequency of these terms, a search in Google Scholar shows “embodied or grounded cognition” to be mentioned alongside “simulation” 230 000 times and alongside “metaphor” 544 000 times.¹⁰ This co-occurrence pattern demonstrates that the importance of these two mechanisms has been implicitly intuited by other researchers. This is similarly reflected in the way reviews of grounded cognition work treat theories, which often involves reducing these theories to their prominent mechanisms (cf. Table 1; examples are provided in section 3.2 on unsystematic empirical work). Therefore, we feel confident in assuming these to be two frequent and important aspects in grounded cognition. We used frequency as criterion because terms that are often mentioned in empirical grounded cognition work likely capture an important space of investigation (cf. Shapiro, 2019, who uses a similar approach to evince his “three themes of embodiment”). Further, those two concepts, which are extracted by many authors who characterize descriptive experimental work as important for their own experimental reasoning, are those which generate hypotheses, and which seem to raise important questions in the eyes of researchers. Therefore, choosing those mechanisms that empirical work has extracted is likely useful. Next, by focusing on mechanisms that were also used outside of grounded cognition in other research domains within the cognitive sciences, resulting theories are more likely to generate novel insights by ensuring commensurability, thereby generating integration. Finally, as demonstrated above, much independent work has arrived at the same two scientific concepts. This suggests that they capture something about the underlying nature of cognition (unless their frequent use has an alternative explanation).

Second, to ensure a solid epistemological foundation which is not speculative, the constructs we posit should be natural kinds, as opposed to nominal kinds. A natural kind is one which corresponds to the true structure of nature (e.g., H₂O, because anything that does not have two hydrogen and one oxygen atom is no longer H₂O), while a nominal kind is a category which captures superficial relations (e.g., learning styles, which ignores underlying factors; Fried, 2017; Schwartz, 1980). It is sometimes said that natural kinds “carve nature at its joints” because they are true categories that would exist whether humans would be alive to name them or not (Bird & Tobin, 2024). Simulation is a natural kind because its definition is based only on natural boundaries. If something does not consist of this description, in other words, if something is not a mental state re-activated in absence of the referent which usually activates it, then it is no longer simulation (Reilly et al., 2023). The explication of metaphoric mapping is similarly natural because it too is defined by what it does: simulating something more concrete in order to represent a (comparatively) more abstract concept. It is important that the features we choose here are tangible and real features as opposed to other, theoretical, constructs. By choosing such strict definition rules, we ensure that this account is epistemologically secure, as no matter what other theories turn out to be true or untrue, this account can remain. Furthermore, a second advantage of having natural kinds, is that it is more likely to align with other theories that also have natural kinds. While different theoretical frameworks can sometimes be incommensurable because they draw different boundaries into the causes of phenomena, this cannot be the case for natural kinds. Therefore, by these constraints, the minimalist account is able to gather from the significant prior work those components which can form a common and solid basis, thereby reducing conceptual clutter, and encouraging systematic empirical work.

4.3. Qualities of the Minimalist Account

The minimalist account addresses the issues described above because it is general in its postulates. With this generality, it gains certainty in its ability to capture the true underlying mechanisms. Further, it also does not make strong assumptions, because it only postulates those constructs and relations which already have very good evidence. This should in turn prevent theoretical enclaves, which will allow falsification instead of conceptual clutter and jingle-jangle, because all positions have been aligned. In short, the minimalist account trades potentially speculative sophistication for a humble and limited approach. It is an intentionally silhouetted view of grounded cognition, lacking in detail. In return for this lack of detail we gain something less measurable, but ever more important: A substantiated and common point of departure, which encourages future work to use this as a tool to meaningfully inform its falsificationist research and to progress incrementally.

The minimalist account, being an overarching theoretical framework, also improves the state of theorizing in grounded cognition more generally (Muthukrishna & Henrich, 2019), as a future research program should begin with the foundation of mechanisms which are well-established and well-understood (cf. Kitcher, 1995). Furthermore, it is in line with recent calls for theory reformation (Bringmann et al., 2022; Eronen & Bringmann, 2021), serving to remedy non-replication of results (Klein, 2014). Some foundational studies in grounded cognition have also been subject to non-replication (e.g., Colling et al., 2020; Morey et al., 2022), thus underlining the need for an overarching theoretical framework; in response to another non-replication it was recently proposed that ensuring successful replications “requires abandoning general theories […] and instead focusing on specific mechanisms” (Buckwalter & Friedman, 2024, p. 1). Finally, our approach does not just look to the past to solve problems, but also forward to solicit progress. It should encourage creative research thanks to its invitation for exploration (cf. Wegner, 1992), while nonetheless providing a unified framework which coordinates and connects research. One such prospective integration is the recent proposal regarding modal and amodal cognition by Kaup et al. (2024), which is similar in its minimalist spirit, preferring few, but substantiated and natural kinds in place of elaborate theories. The minimalist account and the mechanisms of simulation and metaphoric mapping fall fully within modal cognition. This example demonstrates the increased ease of integration when reducing theoretical elaboration.¹¹ To this end, the minimalist account provides a simplified overarching theoretical framework which is necessarily limited in its detail but rather confident in its postulates, improving the current and prospective state of grounded cognition theorizing.

Beyond these strengths, the minimalist account also has limitations. It should be underlined that this framework cannot, by itself, solve all problems in the research field. Firstly, it is a theoretical framework, and not a theory in itself. This means that it cannot serve to explain many details. Instead, it serves to describe (explicitly) the components of knowledge currently available, on which most researchers are likely to agree. Further, we cannot purport that it solves problems in cognitive science beyond those which we have identified as most important above. For example, issues in non-replication can stem from the lack of a theoretical framework (Muthukrishna & Henrich, 2019), which we address; but it also comes from bad research practices, the complexity of the subject matter, and naïve use of statistical measures (Gernigon et al., 2024; Nosek et al., 2022; Open Science Collaboration, 2015). These practices causing ‘bad science’ relate also to structural issues in the field of science generally, like the preference for theory-guided over descriptive research in the reviewing of both manuscripts and funding applications, pressure to publish, and incentivizing the overselling of one’s research findings -- issues which are also prevalent in the field of grounded cognition (Zwaan, 2021). Lastly, our description of the proposed account does not provide specific recommendations for experimental practice. There are surely many ways, beyond theoretical reformation, to improve the state of empirical research in grounded cognition (see e.g., Ostarek & Huettig, 2019). In order to form a viable research program, all sides of the issue must be addressed, and the minimalist account addresses only a limited number.

4.4. Research under the Minimalist Account

The current synthesis remains intentionally agnostic about specific research practices. Of course, future grounded cognition research should adhere to the best-practice standards in science, applying principles of open science, such as pre-registration, rational sampling, and FAIR principles of data management (Appelbaum et al., 2018; Mayo-Wilson et al., 2021; Nosek et al., 2015), conducting adversarial collaborations (Logie, 2023) and performing replicable demonstrations across modalities, experimental designs, and populations (Fischer, 2024). Yet, beyond that we do not provide specific recommendations, aiming to ensure that the entirety of the proposal can be accepted unanimously by all grounded cognition researchers. Furthermore, specific recommendations would likely contain (potentially speculative) assumptions, and therefore have no place here. We emphasize that the minimalist account is a tool to improve the value of past and future research. It should be applied to enhance the ability of different studies to build on each other. We therefore make no suggestions on how to conduct research, merely to use the minimalist account as theoretical embedding (cf. Muthukrishna & Henrich, 2019). We similarly do not intend to formulate a novel research program, which is up to individual researchers to identify, based on their expertise. Nonetheless, we feel it necessary to describe in broad strokes the tone that future research can take with the minimalist account, and to flesh out our proposal with examples.

The minimalist account’s research program should depart from our simplified description of grounded cognition that is based on commonalities between theories, without the attached theoretical baggage in the form of unjustified details. Within this, and in agreement on the same basic principles, empirical research can depart in small steps, approaching a holistic description of concept grounding incrementally.

Research under the minimalist account should be less expansive and less assuming than much previous work. We do not suggest relinquishing all theoretical work. On the contrary, we argue that experimental work must align itself with a theory, but that in practice, research in grounded cognition has often exhibited poor theoretical allegiance, meaning it has not aligned with (any single) theory. Therefore, our minimalist account may, at first sight, seem less theoretical, but it is indeed more stringently theoretical than previous options. This relates to the same tension as the description above in which grounded cognition research is simultaneously framed as being too theory-laden (as opposed to descriptive) and engaging in too little theory-testing (being too discovery-oriented). Although, they sound similar, being theory-laden does not make research theory-testing. Indeed, the confirmatory, discovery-oriented research of past grounded cognition was theory-laden, but without actually attempting serious theory-testing via falsification. Being theory-laden prevents insights because it closes the research off from descriptive or exploratory research, and being discovery-oriented prevents these theories being refuted. By walking this thin line of theory-laden and discovery-oriented research, research in the field of grounded cognition has often managed to avoid paths towards more challenging but ultimately productive insights. Our solution, the minimalist account, does not assume detailed knowledge, which allows for a descriptive approach that is not closed off from discovering possible foundational principles. The issue of being closed off to new discoveries as a result of research being theory-laden is effectively reduced when giving empirical work the option of aligning with an unassuming and silhouetted theoretical framework. Beyond supporting descriptive research, future research should use the minimalist account to theory-test within the proposed framework of common assumptions. This was not possible before, where empirical work did not align itself with any theory at all, or when it did align, departed so far from another theory so as to make critical hypothesis testing impossible. Such benefits reflect our minimalist account’s focus on back-to-basics. Those explorative researchers who are less theoretically-minded can perform studies on “embodied effects”, without smudging over their theoretical allegiance, while others can pit opposing hypotheses against one another (but crucially, under the joint assumptions of the minimalist account; see also James’ distinction between knowing truth and avoiding error; James, 1897).

To give this proposal some shape, it may be valuable to provide two positive examples from past research which give an impression of what research under the minimalist account may entail. First, despite many detailed theories on metaphor, descriptive research such as the frequency of metaphor use in abstract concept representations, has been neglected. Determining such frequencies can be the basis of a strong-inference test: To what degree is metaphoric mapping involved in representing abstract concepts? Drawing on existing work, we can sketch how such an assessment could progress. The Lancaster sensorimotor norms (Lynott et al., 2020) map how strongly concepts draw on the various sensory (e.g., haptic or olfactory) or motor (e.g., foot or mouth movement) representations. The frequency of sensorimotor representations in abstract concepts can provide insight into the frequency of metaphor, with sensorimotor grounding of abstract concepts providing evidence in favor of metaphor. As an aside: The Lancaster sensorimotor norms only make such work possible because they are not theory-laden, consisting of basic, bottom-up constructs and natural kinds. The norms are therefore a product of the type of work which the minimalist account encourages, and this example demonstrates the compounding positive influences of the minimalist account.

A second demonstrative example of a minimalist research question is to determine the types of representations that can serve as grounding substrate. This question is rarely explicitly examined, despite forming an evidently important component of grounding concepts. We argue that this is the case because disputes over theoretical details and confirmatory research have taken precedence over descriptive and explorative research. It was recently pointed out that there is little a priori reason to exclude representations of physical invariants (unchanging features of the physical world like gravity or momentum) from these grounding substrates, given that they are richly represented (Friedrich et al., 2024). Indeed, recent experimental evidence building on this proposal has supported this inclusion (Friedrich et al., 2025). Yet it is only when divorced from their theoretical baggage, that the mechanisms in grounded cognition theories can account for this development. This demonstrates the importance of giving research the possibility to divorce from detailed theoretical descriptions. In order to arrive at a comprehensive understanding of concept grounding, the minimalist account should be the point of departure for future research because it does not presume to have a detailed account, and therefore allows for descriptive and explorative research that is close to nature, as opposed to theory.