Images, thought, and content. Some comments on Thinking with images

In his book Thinking with images, Kozak starts from the assumption that any theory of thinking must face three challenges: the epistemological, the semantical, and the metaphysical. So, the book can be seen as an intent to develop a theory of imaginistic thinking that meets these challenges. To do that, Kozak develops a measurement-theoretic account of imaginistic thinking together with a 2-D model of iconic reference. According to the measurement theory, images are not copies of the world but measurement devices, similar to rules and balances, used to identify measured objects and localize them in some space. According to the 2-D model of iconic reference, images denote the target and exemplify the content that identifies the referent. Let’s see.
Thus, while I agree that descriptions naturally belong to the realm of words, and it is arguable that images sustain predication, I do not think we can easily refuse the classical distinction between singular and attributive content. But is this distinction compatible with the measurement account of imaginistic thinking? This query brings me to some concerns that reminds me what Marcin Milkowski has expressed in his recent post:  Is it possible to extend the measurement theoretic account to other forms of thinking, i. e. linguistic thinking, that is, can we extend it to develop a general theory of thinking? Is this extension possible without denying the specificities of imagistic thinking? Finally, these questions lead us to consider further desiderata: Would such a general extension of the measurement account be desirable? Moreover, do we need a general theory of thought, including imaginistic and linguistic thinking? Or shall we preserve nonunitary explanations?
Nonetheless, we could push the analogy further: in as much as we can use a ruler to measure the length of several objects, we can use the same resources – drawing, color, lines – to measure different objects and identify different referents. And, in as much as the same value, for instance, 20 meters, can be applied to different objects, the same imagistic resources can also be employed with different objects and properties. Thus, in a sense, rulers and images, as measurement devices, possess a great grade of generality in that they can be applied to “measure” different entities. Besides, in both cases, the application of the measurement devices indicates the presence of properties or relations that can be used to identify the referent. In other words, images represent objects under certain aspects by isolating properties of the constructions that serve to identify the referent. In doing so, images depict some of the properties shared by the construction and the object, which is nothing but a way of characterizing the referent with certain attributives. In doing so, we can say that images articulate singular and attributive content and provide the basis for descriptions. And none of this implies that images sustain predication.
So far, Kozak has provided us with a fruitful way to satisfy these challenges. However, I would like to focus on Kozak’s notion of iconic content. Notably, Kozak defends a 2-D Model of reference that states that images denote the target (to continue with the previous example, the triangle) and exemplify the content (the construction rules). Whereas denotation is a dyadic relation between the representation and something it stands for, exemplification goes from the selected properties of the object to its representation. With these specifications, Kozak reflects his Fregean commitments. However, in the meantime, he wants to deny that images include singular and attributive content.
The epistemological challenge states that imaginistic theories cannot provide a theory of knowledge since images are not truth-evaluable and lack logical form. However, according to the measurement-theoretic account, images are tools for thinking. Thus, they provide procedural knowledge by exemplifying the construction’s rules to identify the referent. So, images can afford non-propositional knowledge based on the procedures required to construct a representation of the referred object.
The semantical challenge states that imaginistic theories cannot provide a theory of content since images lack predicative functions and cannot individuate objects. But again, for the 2-D Model, images represent the way of localizing objects. This means that depicted properties do not belong to the object but to the searching procedure to arrive at it. Thus, the picture of a triangle represents the way a triangle can be constructed. Hence, we should not confuse the properties of construction (the length of the line segments and angle measure) with the properties of the constructed object (the triangle).
On the one hand, attributives are meant to characterize or modify the referent of a representation. Thus, we can say that the content of images (Sellars, 1981) and perception (Burge, 2010) include both singular and attributive content since both represent a referent in a certain way. So, for instance, a red picture of an apple represents an apple colored in a certain way, and my perception of a coin represents it as circular. To do so, imaginistic and perceptual attributions do not need to be predicative. Thus, we can distinguish attribution from predication. And this distinction is important if we want to argue for the systematicity and compositionality of images while maintaining a difference between images and language.
Kozak, P. (2023). Thinking in Images. Imagistic Cognition and Non-propositional Content. Bloomsbury Publishing. London.
Lastly, according to the metaphysical challenge, while propositions have logical form and are systematic and canonically decomposable, images are not systematic and lack canonical decompositions; hence they cannot be bearers of thoughts. However, Kozak argues that images are systematic and decomposable by virtue of construction parameters and construction invariants which, in turn, are relative and context-sensitive. Thus, understanding an image depends on understanding its construction parameters and invariants, which then allows the comprehension of different images. In other words, images are systematic and decomposable in nonarbitrary ways and in their own right.
The distinction between singular and attributive content is a classical distinction. And it is tempting to abandon it if we aim to explain the content of images with independence of linguistic theories. But I am not sure that this strategy will work if we want to maintain a Fregean spirit and give an account of the potential links between images and descriptions, as Kozak wants. His arguments depend, firstly, on his thesis that images lack predicative function and, secondly, on the thesis that iconic content cannot be descriptive. However, whereas it is natural to think of descriptions as belonging to the realm of words, I am not convinced about his argument based on the lack of predicative function. Particularly, we can consider further alternatives:
On the other hand, Kozak rejects that the content of images can be divided into singular and attributive content because he claims that the content of images represents the ways of identifying objects by exemplifying the construction rules. So, the picture of a triangle represents how triangles can be identified by exemplifying the rules to construct triangles. So, images [of triangles] enable recognition-based identification [of triangles] by representing the construction properties [for triangles]. Moreover, Kozak suggests thinking of images as measurement devices, that is, as rulers that represent the way we isolate a spatial magnitude, i. e. a meter. So, we can use a ruler to indicate an object’s length, relate it to other values, and describe its target. Similarly, an image can represent how we isolate some construction’s properties to identify the referent – its color, shape, and angles. And although we need words to describe a target, images can be the basis of descriptions.
Sellars, W. (1981). “Mental Events.” Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition, 30(4), 325–345.
Burge, T. (2010). “Steps toward Origins of Propositional Thought.” Disputatio IV (29): 39-67.