Kresimir Durdevic
Affiliation: University of Stirling
Category: Psychology
Keywords: LoT, Maps, Language of Thought, Mental representation, Comparative cognition, Animals, Iconic representation
Date: Thursday 4th of September
Time: 14:30
Location: Gen. Henryk Dąbrowski Hall (006)
View the full session: Spatial Cognition
In order for us to understand what is it like for an animal to have mental representations about the world, we must specify the nature of her representational vehicle which allows her to know it. Traditionally, it was thought that language was the only viable system for modelling mental representation, leading to proposals such as the language of thought (LoT, Fodor, 1975; 2008). Perhaps due to LoT's tight parallels with (natural) language, many comparative psychologists are reluctant to generalise LoT to nonhuman animals (henceforth, animals) and are left without a suitable explanatory framework. In order to fill this gap and avoid a general scepticism towards mental representation in animals, one must first show that an alternative, non-propositional format can be a viable representational vehicle, and second, that it can be characterised in a conceptually and empirically tractable way.
While my focus here is on the second, descriptive goal, it is worthwhile briefly addressing what criteria a viable representational vehicle must satisfy. The systematicity argument (e.g., Fodor & Pylyshyn, 1988) argue that any representational vehicle must be composed of discrete parts and combined according to systematic rules. Language is an obvious candidate here, and so it motivated thinking of mental representation as instantiated in a propositional architecture like LoT. Camp (2007), however, argues that (nonpropositional) topological maps are also systematic, which opens up space for a representational format for minds that are unlikely to have language-like architectures.
Towards the descriptive goal, Camp (2007) begins to characterise the cartographic format as guided by an overarching principle of isomorphism, i.e., a resemblance to the physical world which it purports to represent. Isomorphism implies a more direct relation between the mental representation and its referent (in the world), which is in contrast to a propositional format that exhibits an arbitrary relation between its representations and referents -- e.g., the word 'dog' need not resemble the animal, whereas a picture of a dog, somehow, should. Over-reliance on the isomorphism principle, however, risks leaving the cartographic format underspecified: any representational format must itself specify how the world is presented to the mind, and cannot presume the world as having that character specified for it. A description of the format must specify the computational function which transforms the input from the world into output for the mind.
A good working description of a representational format is found in Quilty-Dunn et al. (2022) where they divide the LoT architecture into 6 properties: it must be organised so that its 1) constituents are discrete and 2) their roles are independent of the content that fills them, so that parts of phrases can be swapped without collapsing the phrase. This makes up a 3) predicate-argument structure, which cannot be reduced to simple concatenations. Content can be transformed using 4) logical operators like OR, NOT, etc., which are 5) built into the architecture and do not require conscious effort or external scaffolding. Finally, we need to allow for LoT to represent 6) abstract content, i.e., categories without specific details particular to a modality (sensory or otherwise). Whatever a characterisation of an alternative format ends up being, one can use the above model as reference.
Leveraging Camp's initial descriptions, other proposals of animal-specific cognition, as well as empirical distinctions from comparative and developmental cognition, I advance a sketch of the cartographic representational character. With reference to Quilty-Dunn et al. (2022) description, I identify 4 main types of constraints which one can populate with relevant properties of a given representational format: the combinatory principle, inferential mode, content specification, and referential structure. Populated with cartographic properties, the constraints are: a topological combinatory principle whose discrete constituents are related to their roles, an inductive inferential mode that relies on exact quantity estimation, concrete content specification with modality-specific information, and extensional (as opposed to intensional) referential structures. Should this be a meaningful description, it should deliminate clearly from propositional formats like LoT, and it should obtain empirically. With this in mind, I contrast some conceptual and operational distinctions implied by LoT and the cartographic systems, and, where available, provide empirical evidence.
References
Camp, E. (2007). Thinking With Maps. Philosophical Perspectives, 21(1), 145–182. https://doi.org/10.1111/j.1520-8583.2007.00124.x
Fodor, J. A. (1975). The Language of Thought. Harvard University Press.
Fodor, J. A. (2008). Lot 2: The Language of Thought Revisited. Oxford University Press.
Fodor, J. A., & Pylyshyn, Z. W. (1988). Connectionism and cognitive architecture: A critical analysis. Cognition, 28(1), 3–71. https://doi.org/10.1016/0010-0277(88)90031-5
Quilty-Dunn, J., Porot, N., & Mandelbaum, E. (2022). The Best Game in Town: The Re-Emergence of the Language of Thought Hypothesis Across the Cognitive Sciences. Behavioral and Brain Sciences, 1–55. https://doi.org/10.1017/S0140525X22002849