Heintz & Scott-Phillips (H&S-P) stressed that humans possess cognitive means of adaptive reaction targeted at the behavior of conspecifics: Humans cannot un-understand what others say. The neuropsychological literature on tool use in semantic dementia corroborates and extends this perspective in indicating that humans cannot un-understand what others “expect.” This literature provides insights into the neurocognitive bases of the ability to infer others' communicative intentions, which can contribute to make human evolutionary sciences less “neurocognition” blind.

Tool-use disorders – also called apraxia of tool use (Osiurak & Rossetti, Reference Osiurak and Rossetti2017) – are traditionally assessed by asking patients to show how to use familiar tools presented in isolation (i.e., single tool-use/pantomime tasks; Table 1). These tasks are considered to investigate semantic knowledge about tool function (Goldenberg & Hagmann, Reference Goldenberg and Hagmann1998). Others suggest that these tasks engage sensorimotor knowledge about tool manipulation (e.g., Rothi, Ochipa, & Heilman, Reference Rothi, Ochipa and Heilman1991), but we limit our discussion to the semantic dimension. The rationale is that the mere presence of a tool should be enough to activate the corresponding semantic knowledge and, thus, help the patient use the tool appropriately. Although these tasks appear simple at first glance, individuals, including healthy ones, frequently hesitate about the actions to perform and ask the clinician for more information about what is expected. Indeed, the absence of associated objects makes the task ambiguous and obliges the individual to infer the clinician's expectations or communicative intentions in H&S-P's terms. After all, if you frequently use newspapers to light fire, what would you do if a clinician asked you to show how to use them? Would you pretend to light fire with them or read them? To find the correct answer, you must infer that the clinician may expect that the demonstration involves the most frequent usage of newspapers. So, you would have to pretend to read them, otherwise your demonstration could be considered as pathological. These tasks are much more than simple tool knowledge tasks and illustrate how semantic knowledge and cognitive pragmatics are intertwined. The study of semantic dementia is particularly enlightening on this point.

Table 1. Neuropsychological tool-use tasks

^a The term tool will be hereafter used for the implement that performs an action (e.g., hammer) and the term object for the recipient of the action (e.g., nail).

Semantic dementia is a variant of frontotemporal dementia, which is characterized by an inaugural atrophy of the polar temporal lobes that are known to play a critical role in semantic cognition (Lambon Ralph, Jefferies, Patterson, & Rogers, Reference Lambon Ralph, Jefferies, Patterson and Rogers2017). At the onset of the disease, patients lose knowledge about unfamiliar things. Thus, they can name a giraffe as a dog with a long neck. Then, even knowledge about familiar things is lost. Patients can name a dog or any four-legged animal as “Ruben,” their dog. Their interpretation of their environment becomes progressively guided by their episodic memory, which generates communicative disorders. They become egocentric (Snowden et al., Reference Snowden, Bathgate, Varma, Blackshaw, Gibbons and Neary2001; see also Moreaud et al., Reference Moreaud, Belliard, Snowden, Auriacombe, Basaglia-Pappas, Bernard and Virat-Brassaud2008) and this egocentrism becomes also visible in single tool-use/pantomime tasks. For example, if a patient with semantic dementia is asked to demonstrate how to use a colander, the patient may be unable to name it or give details about its function, but can say it is useful for changing the oil of a car. This response can be explained by the observation that the patient has been occupied with car repair for the past several weeks. This idiosyncratic and egocentric type of reference is very common in semantic dementia.

One may consider that the difficulties reported in semantic dementia in single tool-use/pantomime tasks reflect nothing else than the selective loss of knowledge about tool function, which prevents the patients from using familiar tools appropriately. If so, then semantic dementia should be characterized by a general tool-use disorder that also impacts the real use of tools presented with objects (i.e., real or novel tool use; Table 1) as in some patients with left brain damage (for a review, see Osiurak & Reynaud, Reference Osiurak and Reynaud2020). However, evidence indicates that semantic memory is neither necessary nor sufficient for real tool use (Baumard et al., Reference Baumard, Lesourd, Remigereau, Merck, Jarry, Etcharry-Bouyx and Le Gall2019; Buxbaum, Schwartz, & Carew, Reference Buxbaum, Schwartz and Carew1997). This is also true for patients with semantic dementia, who obtain quasi-normal performance in real tool-use tasks (Baumard et al., Reference Baumard, Lesourd, Jarry, Merck, Etcharry-Bouyx, Chauviré and Le Gall2016) or when they use familiar tools in their everyday life (Bozeat, Lambon Ralph, Patterson, & Hodges, Reference Bozeat, Lambon Ralph, Patterson and Hodges2002a; Lauro-Grotto, Piccini, & Shallice, Reference Lauro-Grotto, Piccini and Shallice1997). These patients have also “normal” performance on novel tool-use tasks (Hodges, Bozeat, Lambon Ralph, Patterson, & Spatt, Reference Hodges, Bozeat, Lambon Ralph, Patterson and Spatt2000; Lesourd et al., Reference Lesourd, Baumard, Jarry, Etcharry-Bouyx, Belliard, Moreaud and Osiurak2016; Table 1), which suggests that their technical-reasoning skills are spared, allowing them to reason appropriately about mechanical actions that involve tools and objects. In fact, their good technical-reasoning skills explain why they can use familiar tools with their corresponding objects (i.e., real tool-use tasks) even if they cannot name or give details about them. The presence of both tools and objects makes the task “more mechanical-oriented,” leading them to reason about the mechanical actions that they can generate from their presence, as in novel tool-use tasks. In this case, the clinician's expectations have not to be included in the equation. By contrast, patients with semantic dementia meet difficulties in single tool-use/pantomime tasks that are strongly linked to the severity of semantic deficits (Bozeat, Lambon Ralph, Patterson, & Hodges, Reference Bozeat, Lambon Ralph, Patterson and Hodges2002b; for a review, see Osiurak et al., Reference Osiurak, Reynaud, Baumard, Rossetti, Bartolo and Lesourd2021). In sum, these patients have no tool-use disorders strictly speaking. Instead, it is the artificial nature of the single tool-use/pantomime tasks that obliges the individual to include the clinician's expectations in the equation and that makes the task difficult for a patient with a loss of semantic knowledge.

These findings, which challenge the classical semantics versus pragmatics distinction, are in line with H&S-P's perspective by suggesting that there is no reason to see cognitive pragmatics as peripheral to linguistic communication. Semantic dementia tells us that semantic memory supports the ability to infer others' communicative intentions, which complements their perspective by building a bridge with the neurocognition domain. Semantic dementia is characterized by lesions to the polar temporal lobes, which are known – as also illustrated here – as playing a key role in semantic cognition. Evidence also indicates that these brain regions are part of the mentalizing network (Gallagher & Frith, Reference Gallagher and Frith2003), thereby drawing an interesting link between semantic cognition and the ability to understand others' intentions.