This week at The Junkyard we’re hosting a symposium on Piotr Kozak’s recent book: Thinking in Images: Imagistic Cognition and Non-propositional Content. Today we begin with an introduction from Piotr Kozak. Commentaries will follow Tuesday through Thursday.

* * *

When we are asked how many windows are in our flat, we will probably form and inspect a mental image of the flat and count the windows. If architects design a house, they use drawings. You may use a map to get from Berlin to Paris. These are instantiations of what can be called imagistic thinking. The main question of my book is what such thinking can be.

Imagistic thinking is commonly contrasted with propositional thinking and thinking with words (e.g., Slezak, 2002; Zhao et al., 2020). For instance, an architect’s thinking results are not something stateable. They are not true or false. They have no logical form that would determine which step is necessarily next. That is, however, only a negative description.

And yet, one may ask: what on earth is the problem? If we have no issues with the concept of thinking in words, then it is natural that we can think in images, too. However, the very idea of thinking in images does not seem to fit well into our traditional theories of thinking.

In the most general terms, there are two philosophical traditions to think about thinking. According to propositionalism, thinking can be modelled according to truth-preserving rules of logic (Fodor, 1991; Frege, 1984). Thinking is a set of operations on true-evaluable representations and a matter of holding propositional attitudes.

Propositional theories are most often contrasted with associationist theories of thought. According to associationism, thoughts are related to each other because, in their causal history, certain facts have causally linked the mental states into pairs. For instance, the frequency with which a certain organism came into contact with events X and Y determines the frequency with which that organism will have related thoughts about X and Y.

How does one find a place for imagistic thinking within this division? There are two dominant strategies. We can hold that imagistic thinking is explainable within the associationist theories framework (e.g., Quilty-Dunn and Mandelbaum, 2020). That is a natural move if you are keen on British empiricism.

However, adopting an associationist framework to explain imagistic thinking is a dead-end, for it cannot explain the systematicity of imagistic thinking. Associationism cannot do that since it leaves no place for inferential transitions between thoughts. 

To meet these objections, we can try to interpret images in propositional terms and adjust images to the requirements set by the propositional framework. Let us suppose that it is true that images are not truth-evaluable and have no logical form. In this case, we can weaken the propositional theory and formulate hybrid positions (Fodor, 1975, Langland-Hassan, 2020). They all state that a proper part of the content of imagistic thoughts is propositional.

There are two reasons to adopt the propositional framework to explain imagistic thinking. First, it rightly assumes that imagistic thoughts are translatable to propositions. For instance, every map can be translated into a set of propositions in the form of coordinates expressing the location and relations between spatial objects. Second, we want to maintain the idea that imagistic representation may correctly or incorrectly represent the world. If I depict a thief as a blond girl and the thief turns out to be a dark-haired man, I can say that I was wrong because the proposition the thief is a dark-haired man is true.

However, propositionalism, even in its weak form, does not address the redundancy objection. Let me explain it. The proponents of the propositional accounts have to explain the contribution of the iconic content of thoughts to cognition and its functions in the broader cognitive economy. If propositional content does all the job in the machinery of thinking, then the role of images is instrumental and replaceable. Images can accompany thoughts; they can make them more accessible. Yet, they are not necessary for having thoughts. Thus, neither propositionalism, nor associationism can explain the nature of imagistic thinking.

In Thinking in Images, I argue that any attempt to formulate a non-trivial theory of imagistic thought has to be able to find a place for two seemingly contradictory theses. On the one hand, it has to take into account the so-called Irreducibility Thesis, according to which images are necessary for having some thoughts. On the other, it has to be reconciled with the Translatability Thesis, according to which some imagistic thoughts are expressible in propositions.

Moreover, any attempt to explain imagistic thinking must meet the general requirements for any theory of thought. First, it should be able to deliver a theory of knowledge. Next, it should be able to provide a theory of thought’s content. Last but not least, it should be able to explain the systematic and compositional nature of thoughts. The main advantage of propositionalism is that it meets these requirements.

One of the main reasons for the scepticism regarding the idea of imagistic thinking is based on the belief that images, taken loosely as resembling the represented objects, do not meet these conditions. Therefore, they cannot be bearers of thought (e.g., Bennett and Hacker, 2003; Pylyshyn, 2002).

In the book, I present an argument from the best explanation of the following kind. I hold that if images were to meet these challenges, then we should think of them as measurement devices, such as rulers. According to the measurement-theoretic account (and the so-called 2-dimensional model of iconic reference), images refer to the represented objects in two ways. On the one hand, they denote their objects. On the other, they exemplify construction rules that identify objects in some space. In a nutshell, images determine the properties of some logical or physical space in order to localize objects and properties. Consequently, imagistic thinking is a rule-governed operation of manipulation of construction parameters in order to find how something can be seen in a measurement set-up. For instance, it is thinking about how something could look if parameters like height were different.

To unpack this definition, I introduce two technical terms. The concept of construction refers to the ways of localizing the target object or property by determining the properties of some space. Construction rules are the functions that establish correct ways of determining these parameters. Next, drawing on Evans (1982), I introduce the concept of recognition-based identification, which refers to identifying objects as belonging to the same kind based on identifying the invariant properties of some constructions.

To illustrate this rather crude description, let us take a triangle diagram example. The diagram denotes an abstract mathematical object. However, the properties of the diagram, such as three lines and angles, are not copies of some abstract mathematical structure, as if there were some Platonic objects just waiting to be depicted. Instead, the properties of the diagram determine the construction parameters of Euclidean space in order to localize this structure. In short, construction rules exemplified by the diagram order the way points and lines have to be connected. Recognizing the triangle is a matter of finding the construction invariants, such as the number of angles. By manipulating the construction parameters, such as the orientation of the figure, you can discover other properties, e.g., congruency of mathematical figures, etc.

The measurement-theoretic framework allows us to acknowledge the irreducible role of images in thinking. Images are indispensable since they localize the objects of our beliefs and theories within some logical or physical space. For instance, the triangle diagram localizes the triangle concept within Euclidean space, allowing us to identify triangles. However, images are meaningless without theories and beliefs. We need propositional structures to describe what is being depicted by the triangle diagram.

Moreover, the measurement-theoretic account is able to address the general requirements for a theory of thought. Images are a source of understanding since they localize objects of our beliefs. They possess iconic content since they exemplify measures by which we identify these objects. Finally, images are systematic and canonically decomposable since they exemplify construction rules. In short, the measurement-theoretic framework offers at least a conceptual possibility to think about imagistic thinking.

To end with some closing remarks, suppose it is true that understanding the nature of language can bring us closer to understanding the nature of thoughts. Then, plausibly, the same can be said about understanding images. The only difference is that although there is a well-established tradition of studying thought processes through studying the nature of language, similar analyses in the case of images are relatively underdeveloped. In other words, we know much more about language than about images and, consequently, about thinking with language than about thinking with images. Thinking in Images takes a step towards rectifying this position.

References

Bennett, M., Hacker, P. (2003). Philosophical Foundations of Neuroscience. Oxford: Blackwell.

Evans, G. (1982). The Varieties of Reference. Oxford: Oxford University Press.

Fodor, J. (1975). The Language of Thought. Cambridge, MA: Harvard University Press.

Fodor, J. (1991). Replies. In B. Loewer and G. Rey (Eds.), Meaning in Mind: Fodor and his Critics. Oxford: Basil Blackwell.

Frege, G. (1984). Thoughts, translated by P. Geach and R. Stoothoff. In B. McGuinness (Eds.), Collected Papers on Mathematics, Logic, and Philosophy (pp. 351–372). Oxford: Basil Blackwell.

Langland-Hassan, P. (2020). Explaining Imagination. Oxford: Oxford University Press.

Pylyshyn, Z. W. (2002). Mental Imagery: In search of a theory. Behavioral and Brain Sciences, 25, 157-182.

Quilty-Dunn, J., Mandelbaum, E. (2020). Non-Inferential Transitions: Imagery and Association. In T. Chan and A. Nes (Eds.), Inference and Consciousness (pp. 151-171). New York: Routledge.

Slezak, P. (2002). Thinking about thinking: language, thought and introspection. Language & Communication, 22, 353-373.

Zhao, F., Schnotz, W., Wagner, I., Gaschler, R. (2020). Texts and pictures serve different functions in conjoint mental model construction and adaptation. Memory and Cognition, 48 (1), 69-82.