Visual mental images can be ambiguous: insights from individual differences in spatial transformation abilities

Abstract

The debate about whether objects in mental images can be ambiguous has produced ambiguous results. In some studies, participants could not reinterpret objects in images, but even in the studies where participants could reinterpret visualized patterns, the results are not conclusive. The present study used a novel task to investigate the reinterpretation of ambiguous figures in imagery, which required the participants to mentally rotate a figure 180 degrees before attempting to “see” an alternate interpretation. In addition, the participants did not know the purpose of the study in advance, nor did they see alternate interpretations of the stimuli; moreover, we explicitly measured individual differences in key mental imagery abilities. Eight of the 44 participants discovered the alternate version while they were memorizing the figure; 16 reported it after mentally rotating an image; and 20 were not able to “see” the alternate version. The ability to rotate images, assessed with an independent task, was highly associated with reports of image reversals, whereas measures of other imagery abilities were not.

Introduction

Visual perception is driven by the nature of the external world, and percepts typically are only subtly affected by the viewer's intentions and beliefs. In contrast, mental images are created by the imaginer, and thus may not be independent of their creator's intentions and beliefs. Indeed, some researchers have claimed that objects in images are inextricably tied to the way a shape was interpreted when it was encoded into memory, and cannot subsequently be reinterpreted during imagery (e.g. Chambers & Reisberg, 1985). According to this view, images are more like descriptions than they are like pictures that can be readily reinterpreted during perception.

The question of whether images are descriptions or depictions is of interest in part because it bears on the “imagery debate” (e.g. Kosslyn, 1980, Kosslyn et al., 2001, Pylyshyn, 1973, Pylyshyn, 1981, Pylyshyn, in press). This debate revolves around the question of whether the depictive aspects of imagery that are evident to introspection reflect functional aspects of the underlying representation, or whether such pictorial characteristics are purely epiphenomenal (like the heat from a light bulb when one is reading, which plays no functional role in the reading process). If instead of describing information, images depict information in a picture-like way, they would retain some of the “raw material” of the perception – and hence could later be interpreted.

Chambers and Reisberg, 1985, Chambers and Reisberg, 1992 concluded from their experiments that people cannot reinterpret patterns in mental images, and hence that mental images are more like descriptions than pictures. However, other researchers subsequently showed that under some circumstances imagined patterns can be reinterpreted (Anderson and Helstrup, 1993, Finke et al., 1989, Peterson et al., 1992, Rouw et al., 1997). At this writing, it is fair to say that the available findings about whether imaged objects can be ambiguous remain ambiguous. Two features of the studies seem particularly pertinent. First, in many of the previous studies, the participants were told in advance, as part of the instructions, that the stimuli were reversible figures. On the one hand, such instructions could help the participants develop appropriate reversal strategies. Indeed, Rock and Mitchener (1992) found that the nature of the instructions has a clear influence on the number of reversals of ambiguous figures in perception. Only one-third of their participants ever reversed when they were not told about the possibility of reversals. However, when told that the figures were reversible, all participants experienced reversals. On the other hand, it is also possible that such instructions could instill inappropriate reversal strategies when participants are acquainted with different ambiguous figures prior to the experiment. For example, Chambers and Reisberg (1985) report that none of their 55 participants (across three experiments) ever reinterpreted an imagined duck/rabbit ambiguous figure. In sharp contrast, Peterson et al. (1992) reported that such reversals commonly occur in mental imagery. In Experiment 1 of their study, 40% of the participants reported image reversals. In fact, Peterson et al. (1992) used the same duck/rabbit ambiguous figure used by Chambers and Reisberg (1985), only changing the demonstration figure they used in the instructions prior to the task (they used the goose/hawk ambiguous figure instead of the ambiguous Mach book, see Fig. 1). Peterson et al. (1992) conclude that using a demonstration figure that was more appropriate for the duck/rabbit ambiguous figure helped the participants discover the alternate version.

However, it is possible that participants who reinterpreted objects in images might not have actually operated on the image, but instead they might simply have recalled a different interpretation that they stored when they initially memorized the stimuli. In order to rule out such an alternative account, in the present study we examine the ability to reinterpret ambiguous figures in mental images but do not alert the participants in advance that they will see ambiguous figures or try to reverse them. We use a drawing that changes its apparent identity only when it is inverted. The face appears as a young woman, and, when turned upside-down, as an old woman. Unlike other ambiguous figures (e.g. the duck/rabbit figure, or chef/dog figure), the inverted alternate version is hardly noticeable without prior experience. This type of reversal requires a new interpretation of the image components (e.g. the nose in one orientation becomes part of the hair upon inversion), and a new reference frame (as discussed by Peterson et al., 1992). In particular, the top and bottom of the figure must be reassigned to discover the alternate version. Thus, this ambiguous figure requires a rotation of the reference frame by 180 degrees (from right-side-up to upside-down), which makes an incidental discovery (by merely looking at the figure) much less likely than with other figures.

The second key aspect of previous studies is that even when some participants could interpret the imaged pattern in novel ways, not all participants could do so. People clearly differ in their mental imagery abilities (e.g. Kosslyn et al., 1984, Richardson, 1994), and people who have difficulty with specific types of imagery processing may have difficulty in challenging imagery tasks that draw on such processing. Thus, in the present study we assess four key individual imagery abilities, specifically the ability to generate (i.e. form on the basis of stored information) high-resolution images, the ability to compose images from separate parts, the ability to inspect imaged patterns, and the ability to rotate objects in images. We seek to discover whether individual differences in specific aspects of mental imagery predict performance when people visualize ambiguous figures.