In the target article, Quilty-Dunn et al. not only revive the language-of-thought hypothesis (LoTH) in the original domain, but also extend it to new domains such as perception, where the hypothesis did not cover in the past (Fodor, Reference Fodor1975, Reference Fodor2008; Schneider, Reference Schneider2011), for obvious reason: Even if LoTH can be made plausible enough for propositional thoughts, it is unlikely to make it work for imagistic contents, such as those in perceptions (e.g., Block, Reference Block2007; Campbell, Reference Campbell1997; Fodor, Reference Fodor, McLaughlin and Cohen2007). The authors, however, argue that “[i]f cognition is largely LoT-like, and perception feeds information to cognition, then we should expect at least some elements of perception to be LoT-like” (target article, sect. 4, para. 2). More specifically, they invoke six criteria to make the case that some elements in perception such as object files and structured relations are LoT-like. Although their case here is indeed strong, their examples are exclusively visual, and therefore limited in this regard. In what follows I will extend their proposal by providing examples from touch that also exemplify some LoT-like properties, though with the proviso that there are some grey areas to be concerned about.
The relevant examples are the so-called “tactile field” cases in the recent empirical and philosophical literatures (Green, Reference Green2022; Skrzypulec, Reference Skrzypulec2021, Reference Skrzypulec2022). In those cases, multiple tactile stimuli constitute spatial patterns that facilitate varieties of tactile judgements (Cheng, Reference Cheng, Cheng, Deroy and Spence2019, Reference Cheng, Ferretti and Glenney2020, Reference Cheng2022; Fardo, Beck, Cheng, & Haggard, Reference Fardo, Beck, Cheng and Haggard2018; Haggard & Giovagnoli, Reference Haggard and Giovagnoli2011). Although in those cases researchers often emphasise the holistic characters of tactile pattern perceptions, the tactile fields also have some LoT-like structures identified by the authors, and this might strengthen their case that LoTH can extend to perceptions, including nonvisual ones. Let's look into some relevant details.
The tactile field cases typically involve multiple tactile stimuli, each of them exists independent of one another. This exemplifies discrete constituents (property 1). Those tactile stimuli can exhibit different properties at different times; for example, some of them can vibrate while the other ones remain still. This exemplifies predicate–argument structure (property 3). Moreover, the multiple tactile stimuli can jointly vibrate to generate geometrical representations such as lines, triangles, and squares; as long as the stimuli in question generate neutral touch (as opposed to thermally salient or nociceptive feels), those shapes can be equally represented. This exemplifies abstract conceptual content (property 6). Thus, tactile fields at least exemplify three core properties of LoTs.
One might argue that there are six core properties identified by Quilty-Dunn et al., but tactile fields might exemplify only three of them. Is this enough? Well, in their discussion of object files in section 4.1, they also primarily point out that properties 1, 3, and 6 are exemplified by them. Moreover, later they write that “perceptual representations of individual objects contain discrete constituents that are organized in a predicate-argument structure and predicate abstract conceptual contents” (target article, sect. 4.2, para. 1), and these are again properties 1, 3, and 6. It seems that amongst the six core properties, somehow these three are more important, or even almost definitive.
Now, the good news is that the tactile field cases are at least as good as the object file cases, so if the latter fits LoTH, so does the former. However, this generates a potential worry about the status of the six core properties. It should be quite clear that Quilty-Dunn et al. are not engaging the traditional project of offering necessary and sufficient conditions for LoT: They write that “[m]any, perhaps all, of these properties are not necessary for a representational scheme to count as an LoT…. We regard these properties as (somewhat) independent axes on which a format can be assessed for how LoT-like it is” (target article, sect. 2, para. 3). But isn't this too weak? If none of them is necessary, and presumably none of them is by itself sufficient, how do we assess whether a given format fits LoTH? Relatedly, to say “LoT-like” might make the situation worse, as similarities are too vague to be useful without further explications. The positive story offered by Quilty-Dunn et al. invokes the notion of “cluster,” and they write that “LoTH predicts that these sorts of evidence should tend to cooccur” (target article, sect. 2, para. 13). For them, such “clustering-based approach” provides “an abductive, empirical argument for LoTH” (target article, sect. 2, para. 13). There is nothing wrong with this approach as such, but one is justified in asking for more concrete criteria: Do these six core properties constitute a weighting system? Are properties 1, 3, and 6 indeed more important than the others? It is even more worrying that Quilty-Dunn et al. use the term “core” to name these six properties, as that signifies that even if they are not necessary, they might be near enough.
As mentioned above, Quilty-Dunn et al. write that “[i]f cognition is largely LoT-like, and perception feeds information to cognition, then we should expect at least some elements of perception to be LoT-like, because the two systems need to interface.” This general point is well taken, but again, it might be too vague to be truly useful. Consider a similar thought in the conceptualism debate: If cognition is largely conceptual, and perception feeds information to cognition, then we should expect at least some elements of perception to be conceptual, because the two systems need to interface. Here we face two worries: First, “largely” is unclear, and second, if only some elements are conceptual, what about those nonconceptual elements? How do they feed information to cognition (McDowell, Reference McDowell1996)? If these are legitimate challenges to partial conceptualism (Peacocke, Reference Peacocke1992), then the same doubt can be cast on the general point made by Quilty-Dunn et al.
Let's head back to the good news. In the target article, Quilty-Dunn et al. provide many rationales for the thesis that LoTH is the best game in town. They might well be right about that. The tactile field cases, and potential cases from other sensory modalities, should strengthen their hypothesis. That said, it will be very helpful if the status of the six core properties can be further clarified.
You have
Access
In the target article, Quilty-Dunn et al. not only revive the language-of-thought hypothesis (LoTH) in the original domain, but also extend it to new domains such as perception, where the hypothesis did not cover in the past (Fodor, Reference Fodor1975, Reference Fodor2008; Schneider, Reference Schneider2011), for obvious reason: Even if LoTH can be made plausible enough for propositional thoughts, it is unlikely to make it work for imagistic contents, such as those in perceptions (e.g., Block, Reference Block2007; Campbell, Reference Campbell1997; Fodor, Reference Fodor, McLaughlin and Cohen2007). The authors, however, argue that “[i]f cognition is largely LoT-like, and perception feeds information to cognition, then we should expect at least some elements of perception to be LoT-like” (target article, sect. 4, para. 2). More specifically, they invoke six criteria to make the case that some elements in perception such as object files and structured relations are LoT-like. Although their case here is indeed strong, their examples are exclusively visual, and therefore limited in this regard. In what follows I will extend their proposal by providing examples from touch that also exemplify some LoT-like properties, though with the proviso that there are some grey areas to be concerned about.
The relevant examples are the so-called “tactile field” cases in the recent empirical and philosophical literatures (Green, Reference Green2022; Skrzypulec, Reference Skrzypulec2021, Reference Skrzypulec2022). In those cases, multiple tactile stimuli constitute spatial patterns that facilitate varieties of tactile judgements (Cheng, Reference Cheng, Cheng, Deroy and Spence2019, Reference Cheng, Ferretti and Glenney2020, Reference Cheng2022; Fardo, Beck, Cheng, & Haggard, Reference Fardo, Beck, Cheng and Haggard2018; Haggard & Giovagnoli, Reference Haggard and Giovagnoli2011). Although in those cases researchers often emphasise the holistic characters of tactile pattern perceptions, the tactile fields also have some LoT-like structures identified by the authors, and this might strengthen their case that LoTH can extend to perceptions, including nonvisual ones. Let's look into some relevant details.
The tactile field cases typically involve multiple tactile stimuli, each of them exists independent of one another. This exemplifies discrete constituents (property 1). Those tactile stimuli can exhibit different properties at different times; for example, some of them can vibrate while the other ones remain still. This exemplifies predicate–argument structure (property 3). Moreover, the multiple tactile stimuli can jointly vibrate to generate geometrical representations such as lines, triangles, and squares; as long as the stimuli in question generate neutral touch (as opposed to thermally salient or nociceptive feels), those shapes can be equally represented. This exemplifies abstract conceptual content (property 6). Thus, tactile fields at least exemplify three core properties of LoTs.
One might argue that there are six core properties identified by Quilty-Dunn et al., but tactile fields might exemplify only three of them. Is this enough? Well, in their discussion of object files in section 4.1, they also primarily point out that properties 1, 3, and 6 are exemplified by them. Moreover, later they write that “perceptual representations of individual objects contain discrete constituents that are organized in a predicate-argument structure and predicate abstract conceptual contents” (target article, sect. 4.2, para. 1), and these are again properties 1, 3, and 6. It seems that amongst the six core properties, somehow these three are more important, or even almost definitive.
Now, the good news is that the tactile field cases are at least as good as the object file cases, so if the latter fits LoTH, so does the former. However, this generates a potential worry about the status of the six core properties. It should be quite clear that Quilty-Dunn et al. are not engaging the traditional project of offering necessary and sufficient conditions for LoT: They write that “[m]any, perhaps all, of these properties are not necessary for a representational scheme to count as an LoT…. We regard these properties as (somewhat) independent axes on which a format can be assessed for how LoT-like it is” (target article, sect. 2, para. 3). But isn't this too weak? If none of them is necessary, and presumably none of them is by itself sufficient, how do we assess whether a given format fits LoTH? Relatedly, to say “LoT-like” might make the situation worse, as similarities are too vague to be useful without further explications. The positive story offered by Quilty-Dunn et al. invokes the notion of “cluster,” and they write that “LoTH predicts that these sorts of evidence should tend to cooccur” (target article, sect. 2, para. 13). For them, such “clustering-based approach” provides “an abductive, empirical argument for LoTH” (target article, sect. 2, para. 13). There is nothing wrong with this approach as such, but one is justified in asking for more concrete criteria: Do these six core properties constitute a weighting system? Are properties 1, 3, and 6 indeed more important than the others? It is even more worrying that Quilty-Dunn et al. use the term “core” to name these six properties, as that signifies that even if they are not necessary, they might be near enough.
As mentioned above, Quilty-Dunn et al. write that “[i]f cognition is largely LoT-like, and perception feeds information to cognition, then we should expect at least some elements of perception to be LoT-like, because the two systems need to interface.” This general point is well taken, but again, it might be too vague to be truly useful. Consider a similar thought in the conceptualism debate: If cognition is largely conceptual, and perception feeds information to cognition, then we should expect at least some elements of perception to be conceptual, because the two systems need to interface. Here we face two worries: First, “largely” is unclear, and second, if only some elements are conceptual, what about those nonconceptual elements? How do they feed information to cognition (McDowell, Reference McDowell1996)? If these are legitimate challenges to partial conceptualism (Peacocke, Reference Peacocke1992), then the same doubt can be cast on the general point made by Quilty-Dunn et al.
Let's head back to the good news. In the target article, Quilty-Dunn et al. provide many rationales for the thesis that LoTH is the best game in town. They might well be right about that. The tactile field cases, and potential cases from other sensory modalities, should strengthen their hypothesis. That said, it will be very helpful if the status of the six core properties can be further clarified.
Acknowledgments
I would like to thank all of my past teachers and colleagues in New York and London for helping me think through these issues in the past few years.
Financial support
This study was supported by the Ministry of Science and Technology, Taiwan (MOST 109-2410-H-004-006-MY3).
Competing interest
None.