Please imagine the following scene, as vividly as you can:
A person walks into a room, and knocks a ball off a table.
Once you've imagined it, read on.
Consider the scene that you imagined. Did you imagine the color of the ball? What about the ball's size? Did you imagine the color of the person's hair, or the pattern of their clothes? Can you trace with your finger the trajectory that the ball took?
If you're like most people, you imagined some of these properties, but not others.
Perhaps you can confidently say the ball was red, and moved from left to right, but did not imagine the person's hair color, or clothes, or gender. You could fill in such details if needed, but they were not there to begin with.
In this paper, we study non-commitment in mental imagery. That is, we investigate the possibility that people do not represent basic properties of an imagined scene.
Non-commitment in the imagination has been discussed in philosophy of mind, and has played a theoretical role in debates about mental imagery, but there has been little empirical study of this topic.
Mental images have been a topic of intense study and debate in philosophy, neuroscience, cognitive science, and beyond. Researchers have debated their format, cognitive development, neural underpinnings, and phenomenology. Mental images and the imagination also play a practical role, and recent research in marketing has studied their use in product design (Dahl, Chattopadhyay, & Gorn, 1999; DeRosia & Elder, 2019), and their affect on consumer choice (Jiang, Adaval, Steinhart, & Wyer Jr, 2014; Lee & Qiu, 2009).
In philosophy, Descartes suggested that mental images are similar to rough engravings (Cottingham, Stoothoff, Murdoch, Kenny, et al., 1985), in that they directly capture some aspects of an actual scene, but leave out crucial details. More recently, philosophers have discussed non-commitment, using a zoo of thought experiments that examined non-commitment to different properties in mental images, from speckled hens (Ayer, 1940) to bald men (Shorter, 1952), striped tigers (Dennett, 1986), and purple cows (Dennett, 1993). Such examples have also been used in discussion of the inexactness of perception, but that is not our focus here.
The notion of non-commitment in mental imagery has also played a small role in the ongoing debate over the format of mental images (the ‘Imagery Debate’), although the phenomenon of non-commitment was not itself empirically established and studied.
Researchers on one side of the imagery debate have argued that mental images are propositional in nature, and that the subjective phenomenon of seeing images in the mind is epiphenomenal (Pylyshyn, 1973, Pylyshyn, 1978, Pylyshyn, 2002, Pylyshyn, 2003). On the other side, researchers have argued that mental images have a depictive, picture-like format (Kosslyn, Pinker, Smith, & Shwartz, 1979; Kosslyn, Thompson, & Ganis, 2006; Pearson, 2019; Pearson & Kosslyn, 2015). Non-commitment has been used to argue against the pictorial-depictive view (Pylyshyn, 1978, Pylyshyn, 2002). Defenders of the pictorial view, however, have countered that the possibility of non-commitment need not rule out a depictive format for mental images. For example, Kosslyn et al., 2006 has claimed that supposed non-commitment is due to inattention or other cognitive limitations when reporting on mental images (which are themselves highly detailed), just as perceptual and cognitive limitations may prevent you from accurately counting the numbers of dots in a Styrofoam ceiling. Block, 1983 has argued that mental images are more like sketches than photographs, in which various details can be left unspecified. Outside of the mental imagery debate, it has recently been suggested that the default values of properties in imagined scenes explain why people fall prey to ‘stumpers’ (Bar-Hillel, Noah, & Shane, 2018), riddles whose solution falls outside the dominant construal of the scene set up by the riddle.
Regardless of the format of mental images, people vary in the self-reported vividness of their imagery, and various measures have been developed to study these individual differences (e.g. Andrade, May, Deeprose, Baugh, & Ganis, 2014; Hall, Pongrac, & Buckholz, 1985; McKelvie, 1995; Reisberg, Pearson, & Kosslyn, 2003; Sheehan, 1967). Here, we are concerned with non-commitment to the properties of a mental scene that should be readily apparent in a perceptual image depicting that scene, although we also examine the relationship between non-commitment and the self-reported vividness of a mental image.
But beyond reports of very low vividness, some people report that they lack a subjective experience of visual imagery entirely. This condition is known as ‘aphantasia’ (Dance, Ipser, & Simner, 2022; Galton, 1880; Keogh & Pearson, 2018; Zeman et al., 2020; Zeman, Dewar, & Della Sala, 2015). Non-commitment in mental imagery is different to aphantasia: in principle, a person may experience a visual mental scene of a ball rolling off a table without being committed to the color of the ball, while another person may know the ball is red, without the subjective experience of a mental image.
To empirically examine non-commitment, we conducted a series of studies similar in structure to the opening example of this paper. That is, people imagined a described scene, and were then asked whether specified properties were part of their mental image.
Throughout, we rely on people's subjective reports, as has much of previous research on imagination and mental imagery (see, e.g. Kosslyn, Pinker, Smith and Shwartz, 1979, Kosslyn, Thompson and Ganis, 2006; Markman, Klein, & Suhr, 2012; Pylyshyn, 2002; Shepard & Metzler, 1971), but we are also excited by the potential of methods that do not rely on self-reports (e.g. Morales & Firestone, 2023).
Our first study establishes the basic phenomenon of non-commitment, using a single scene and set of properties. People report that their imagined scenes lack basic details and properties, ones that would be readily available in real images. Our second study replicates and extends the first study with four additional scenes. Our third study examines the possibility that people's apparent non-commitment may instead reflect forgetfulness, uncertainty, or some other factor. Our fourth study examines the relationship between individual variation in non-commitment and individual differences in vividness. Our final study finds that when they are not given the explicit option of reporting non-commitment, people confabulate details of their imagined scenes.
Overall procedures and methods
In each study, participants were asked to imagine a specified scene, and were then queried, in various ways, about whether particular properties were part of their mental image. The procedures for the five studies are summarized in Fig. 1. All studies, including experimental procedures, number of participants, exclusion criteria, and analyses, were pre-registered.2All data is available at the following OSF
Study 1: do people report not imagining basic scene properties?
Our first study used a simple scene to demonstrate the phenomenon of interest: after being asked to vividly imagine a specified scene, many people report not having imagined properties of the scene that would be easily distinguishable in a real image.
Study 2: replication with additional scenes
To confirm our hypothesis that people do not imagine basic properties of their mental images, we replicated Study 1, using four additional scenes.
Study 3: probing for uncertainty and forgetfulness
In Studies 1 and 2, participants reported that they did not imagine basic properties of different scenes. However, it may be that when given a binary choice, participants opted for the “No, it was not part of my mental image” response, when they actually meant something else. Feelings of uncertainty, vagueness, and so on may lead participants to favor the ‘No’ response. In our third study, we gave participants multiple response options, to examine whether they selected the ‘No’ response due to
Study 4: non-commitment and existing mental imagery metrics
We next study the relationship between how vividly people report imagining mental scenes, and the extent to which people do not commit to properties of their mental scenes. In particular, we examine whether even people who report most vividly imagining mental scenes also report non-commitment. We used both established measures of individual differences in overall vividness for mental imagery (Study 4a), and a direct measure of self-assessed scene clarity for the specific scene people imagined
Study 5: open-ended responses
In all the studies so far, many participants reported non-commitment when explicitly given the option to do so. This is in contrast to the intuition that people can often report rich details of their mental scenes. To probe this contrast, we asked people to describe aspects of an imagined scene in an open-ended way. In principle, people may use this open-ended format to report non-commitment. But, the pragmatics of being asked to describe a given property may lead people to confabulate many of
After being asked to imagine scenes such as a person knocking a ball off a table, or a person putting fruit into a bag, participants were asked whether they had imagined various basic properties of the scenes, such as the color of the table, or the shopper's clothes. The questions were about simple scenes, and properties that would be easily perceivable in real images. Many participants reported that, in fact, such properties were not part of their mental images. We suggest that this
The authors wish to thank M. Bar-Hillel, L.A. Paul, D. Spurrett, and S. Bhatia. for helpful discussion, and the reviewers for helpful comments and suggestions. TDU is supported byNSF Science Technology CenterAwardCCF-1231216, theDARPA Machine Common Sense program, and theJacobs Foundation. All studies were pre-registered, and the data is publicly available on OSF. The authors declare no conflicts of interest.
CRediT authorship contribution statement
Eric J. Bigelow:Methodology, Data curation, Formal analysis, Investigation, Writing – original draft, Visualization.John P. McCoy:Conceptualization, Methodology, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Supervision.Tomer D. Ullman:Conceptualization, Methodology, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Supervision.
© 2023 Elsevier B.V. All rights reserved.