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. On Monday, we began with an introduction from Piotr Kozak. Commentaries follow Tuesday through Thursday.

* * *

Thinking in Images is a rich and elaborate book that defends a view that images play an indispensable role in thinking as measurement devices. In articulating that view, Kozak covers a lot of ground by critically engaging with a wide range of literature and carving out new paths in the theoretical landscape.

Inevitably, this short commentary cannot do justice to the wealth of detail in Kozak’s book. For instance, the reader can find there a fascinating discussion of the representational role of knot diagrams and black hole pictures. The book also presents an innovative account of recognition-based identification in terms of construction invariants. These are just a couple of examples of what the book offers but that I don’t have space to discuss here.

The book has two parts. The first part is about the issues and challenges regarding the idea of thinking with or in images. The second part presents Kozak’s positive proposal. My commentary will focus on the first part, and the third chapter in particular. In that chapter, Kozak considers and criticizes ‘Neo-Lockean’ approach and Wittgenstein-Ryle’s approach as two candidates for understanding the idea of thinking with images. According to the former, imagistic thoughts are theoretical postulates that are supposed to mediate between perceptual and discursive representations; according to the latter, the talk of imagistic thoughts just lacks sense. To further narrow my focus in this commentary, I won’t be addressing Kozak’s criticisms of Wittgenstein-Ryle’s approach, mostly because I found his critique convincing.

Regarding the neo-Lockean views, Kozak targets Jesse Prinz’s neo-empiricist theory of concepts and Chris Gauker’s theory of imagistic thought. Since the criticism of the neo-Lockean view reappears later in the book when Kozak appeals to that very same criticism in arguing against the view of mental imagery as weak perception, I take this part to be significant also for the arguments later in the book.

In describing the gist of the neo-Lockean view, Kozak moves seamlessly between talk about imagistic thoughts mediating between perceptual and discursive representations and imagistic thoughts grounding thoughts in perceptual experience or linking perception and thought. I think that the idea that images mediate between perception and thoughts can be understood differently from the idea that images ground thoughts in perceptual experience. The second is essentially the statement of the empiricist view of concepts, à la Prinz. The first can be cashed out in different ways, some of which give images a different theoretical role from the one the empiricist view like Prinz’s does. Since Kozak is not entirely clear on what that mediation is supposed to amount to, it merits further discussion.

On p. 58, he lays out what I take to be the most explicit statement of mediation:

In the most general sense, the role of an image is to mediate between perception and thinking, being both the product of perception and the basis for thoughts at the same time. (58)

Kozak’s main criticism of the view seems to appeal to there being a categorical difference between thought and perception. However, it can be argued that such putative categorical difference is exactly what imagistic representations, in their mediative role, challenge. There are reasons to think that imagistic representations can be more abstract than perceptual representations. For instance, during mnemonic replay in memory consolidation, memory processes extract more abstract information from more concrete and detailed representations of past experiences (Cowan et al. 2021). By prioritizing the features of those experiences that are more goal-relevant, the resulting memory representations tends to be more schematic and gist-like. There are thus mental operations on imagistic representations that can abstract common features from them and form more schematic representations as a result. If we accept this, then there is a path open to understanding the mediating role of imagistic representations in more generous terms. For instance, if imagistic representations encompass abstract contents that are formed through schematizing stored perceptual representations, then such schematic representations are not exactly like percepts or discursive thoughts but instead could constitute a bridge between two.

To put a bit more pressure on the categorical distinction between perception and discursive thought, I was also wondering what Kozak would say about the data that seem to support the view that perceptual states involve something like predication (a claim Kozak denies) in the sense that they track particulars and attribute properties to them (for a nice recent overview, see Quilty-Dunn & Green 2021). If that is the case, the strict boundary between perception and discursive thought seems to break down.

Let’s now look at Kozak’s response to Gauker’s view. Treating the latter together with Prinz’s as neo-Lockean is quite surprising, given that Gauker has explicitly criticized empiricism about concepts (Gauker 2011). Furthermore, by appealing to the categorical differences between perception and discursive thought, Kozak’s own view seems to be very close to Gauker in that Gauker has also stressed the radical distinction between imagistic and conceptual thought and has argued that there are insurmountable challenges of building concepts through abstraction from perceptual experiences. In addition, Gauker’s view of imagistic thought looks similar to the measurement-theoretic approach in that it involves locating an object or scenario in the objective quality space. Granted, Gauker’s theory has narrower application than Kozak’s, but it seems to present a comparable outlook.

According to Kozak, Gauker’s theory is problematic because it cannot account for the translatability thesis which states that images are expressible with propositions, and it cannot meet the epistemological challenge of accounting for the epistemic role of imagistic content. Admittedly, these do look like challenges for Gauker’s view. However, given that the translatability thesis is not explicitly addressed after p. 49 in Kozak’s book, I was left wondering what exactly Kozak’s own way of accounting for it was. As for the epistemological challenge, Kozak’s own proposal of how to understand the epistemic relevance of images is to appeal to their contribution to understanding, instead of propositional knowledge. However, I was wondering why Gauker couldn’t likewise appeal to the contribution of imagistic representations to understanding? At least there was no argument that he couldn’t. Given these considerations, it seems to me that Kozak did not disprove Gauker’s theory of imagistic thought and exaggerated the differences between their views.

I conclude with a big picture comment. Much of the book is devoted to the analysis of external/material images like diagrams and pictures and the results of that analysis are smoothly applied to the discussions of mental imagery and imagistic mental content more broadly. Perhaps from the measurement-theoretic perspective, both mental and material images can be treated as measurement devices without hesitation. However, it was difficult to get rid of the suspicion that the differences between two kinds of images are more substantial than Kozak lets it seem. For instance, he sometimes talks about thinking in images and sometimes about thinking with images. I find the first way of speaking much more naturally applicable to mental images, while the second way of speaking fits better with external images. Maybe this is only a superficial linguistic worry, but I suspect that it reflects a more serious issue of missing certain significant differences between external and mental images. Unless Kozak wants to embrace the extended mind thesis (which is entirely possible but is not given any indication of), external images presumably discharge their representational role in a different way from mental images and only count as thinking aids, not as something in which we think.

References

Cowan, E.T., Schapiro, A.C., Dunsmoor, J.E. and Murty, V.P, 2021. Memory consolidation as an adaptive process. Psychonomic Bulletin & Review, pp.1-15.

Gauker, C. 2011. Words and Images: An Essay on the Origin of Ideas. Oxford: Oxford University Press.

Quilty-Dunn, J. and Green, E.J. 2021. Perceptual attribution and perceptual reference. Philosophy and Phenomenological Research. pp. 1-26. DOI: 10.1111/phpr.12847

Reply to Uku Tooming

Authors love when their books are being read, but they love even more when their books are read insightfully. I am in such a position.

Uku Tooming, in his fair and perceptive comment, raises three points. First, he poses a clarification question on the link between thoughts and perception. In the book, I hold that one of the main aims of the Lockean and neo-Lockean theories of imagistic thinking is to explain how thoughts can be derived from perception. Taking mental images to mediate between thoughts and perception seems to be a straightforward answer to this problem. Thoughts can be derived from perception in terms of being mediated by iconic representations. Playing a mediatory role means that images share features that can be ascribed to perceptions and thoughts.

However, such an answer seems to be unsatisfactory if we accept that there is a thought-perception border. This border can be understood differently. Most commonly, it is understood in terms of differences in format and content (see, e.g. Block, 2023).

Tooming is perfectly right when he states that the thought-perception border can be challenged, which is exactly what neo-Lockean accounts want to achieve. I am not saying that it cannot be done (see, e.g., Stokes, 2021). Yet, I do believe that there are strong reasons to be sceptical about it (see, e.g., Raftopoulos, 2009; Raftopoulos and Zeimbekis, 2015). However, whether such a border exists and where it runs is a completely different topic that must be elaborated separately.

My argument is more moderate, though. I conditionally hold that if we accept the thought-perception border claim, which I do, then images cannot play a mediatory role by sharing incompatible properties of both representational kinds. However, it does not mean that there is no imagistic link between thoughts and perception. In the book, I am trying to cash this link out in terms of perceptual skills. According to the measurement-theoretic framework, mental imagery is a skill of perceiving based on procedural knowledge of construction rules. It involves the skill of identifying the parameters of perceptual space in order to localize an object or event. Although it does not involve predication (that would require possessing logical form), it does attribute properties to the represented objects. The only difference is that I consider the attributed properties to be measurement predicates.

Second, Tooming worries that I unjustly interpret Gauker’s (2011) theory of imagistic cognition, which, Tooming argues, is compatible with the measurement-theoretic framework. I completely agree with Tooming. Gauker’s theory is ingenious, and I am far from giving it all the credit in the book.

However, I have never claimed, I hope, that Gauker’s theory is wrong or incompatible with the measurement-theoretic account. In general, I consider the measurement-theoretic framework to be an extension of the structural-resemblance approach, which Gauker’s theory is a part of. Thus my argument is not that Gauker’s theory is wrong. However, I claim that it lacks explanatory power. I hold that it cannot explicate the relation between imagistic and propositional knowledge and content and explain the role of images in scientific practice. Moreover, I doubt whether it can explain the compositionality and systematicity of imagistic thoughts. However, it is not an argument against Gauker’s theory since that theory had never been aimed at explaining these phenomena.

Last but not least, Tooming points out that there is a linguistic difference between thinking in images and thinking with them. According to Tooming, it reflects the substantial difference between thinking in mental images and thinking with external images, such as diagrams and pictures. Moreover, he holds that any attempt to identify thinking in images and with them would require accepting the extended mind hypothesis.

I am afraid I have to disagree with these points. On the one hand, although it is not the place to elaborate on this remark, I take the extended mind hypothesis as deeply flawed. On the other, I hold that the linguistic difference is superficial. It involves thinking about thoughts as mental objects being bearers of some content. This is the view I reject explicitly. Drawing on Frege, I consider thoughts to be a kind of operation that must be expressed with some representations. In contrast to Frege, I hold that these representations do not have to be linguistic in nature. They can be imagistic. The whole book project can be seen as an attempt to show how it is possible.

Reference:

Block, N. (2023). The Border Between Seeing and Thinking. New York: Oxford University Press.

Gauker, C. (2011). Words and Images. An Essay on the Origin of Ideas. Oxford: Oxford University Press.

Raftopoulos, A. (2009). Cognition and Perception: How do psychology and neural science inform philosophy? Cambridge, MA: MIT Press.

Raftopoulos, A., Zeimbekis, J. (2015). The cognitive penetrability of perception—An overview. In J. Zembeiskis and A. Raftopolous (Eds.), The Cognitive Penetrability of Perception: New Philosophical Perspectives (pp. 1-56). Oxford: Oxford University Press.

Stokes, D. (2021). Thinking and Perceiving. New York: Routledge.