Involuntary mental rotation and visuospatial imagery from external control

Highlights

• The Reflexive Imagery Task (RIT) triggers high-level, involuntary imagery.

• The effect involves verbal imagery, which is an easily elicited form of imagery.

• Can the RIT effect arise for tasks requiring high-level, symbol manipulation?

• Effects arose from an RIT involving mental rotation and visuospatial imagery.

• The effects require high-level processes associated with frontal cortex.

Abstract

The Reflexive Imagery Task (RIT) was developed to investigate the entry into consciousness of involuntary imagery. Subjects are presented with objects and instructed to not think of the names of the objects. Involuntary subvocalizations arise on many trials. RIT effects reveal the capacities of involuntary processing. These cognitions do not require symbol manipulation. Can mental rotation and visuospatial imagery, too, arise in this involuntary manner? In the mental rotation task, subjects were first taught to mentally rotate two-dimensional objects. Subjects were then instructed to not mentally rotate objects. In the chess task, subjects were taught how to move in their minds objects in specified ways, much as one could imagine how chess pieces move on a chessboard. Subjects were then instructed to not have such visuospatial imagery. For both tasks, involuntary imagery occurred on a substantial proportion of trials, revealing that symbol manipulation can be influenced involuntarily through external control.

Introduction

Early in the morning, the eyes open and one immediately experiences percepts and urges—the sight of sunlight entering the room, the smell of breakfast, and the urge to have a drink of water. To the observer, these conscious contents simply “just happen” (Morsella, Godwin, Jantz, Krieger, & Gazzaley, 2016a). This everyday scenario illustrates that most of the contents composing the conscious field¹ arise effortlessly, passively, and involuntarily (Morsella et al., 2016a). Perception research (e.g., Allen et al., 2016, Firestone and Scholl, 2016) reveals that entry into consciousness of this nature (“involuntary entry,” for short) is influenced by many variables.² As illustrated in our example, urges, too, can enter consciousness in this way (Loewenstein, 1996, Morsella, Gray et al., 2009, Morsella, Wilson et al., 2009).³

Other forms of involuntary entry stem from a combination of sets⁴ and external stimuli. Take, for example, the situation noted long ago by Ach (1905/1951): If one has activated the set to add before hearing “one and one,” then one will experience the conscious content “two.” If, however, the set that had been activated was to subtract instead of add, then one would have experienced “zero” instead of “two.” This form of involuntary entry has been referred to as set-based entry (Bhangal, Merrick, Cho, & Morsella, 2018), which has been distinguished from the involuntary entry mentioned above concerning percepts (e.g., a loud sound) and visceral urges (e.g., hunger).

The Reflexive Imagery Task (RIT; Allen, Wilkins, Gazzaley, & Morsella, 2013; see review in Bhangal, Cho, Geisler, & Morsella, 2016) was developed to investigate the nature of involuntary entry from a combination of external stimuli and the activation of sets. The task stems from “subjective” variants of the Eriksen flanker task (e.g., Morsella, Gray et al., 2009, Morsella, Wilson et al., 2009; see discussion in Desender, van Opstal, & van den Bussche, 2014, and in Questienne, Atas, Burle, & Gevers, 2018), in which distractors activate involuntary urges and other conscious contents.⁵ Other aspects of the RIT are based on theories (Morsella et al., 2016a; see Discussion) and on experiments by Ach, 1905, Stroop, 1935, Uznadze, 1966, Wegner, 1989, and Gollwitzer (1999).

In the RIT, subjects are instructed to not perform a mental operation (e.g., to count or name an object) on to-be-presented stimuli. For example, before being presented with three circles, subjects might be instructed to not count the number of objects presented on the screen (Bhangal et al., 2018). On most trials, despite the intentions of the subject, the undesired mental operations still arise, yielding “three.” The RIT uses negative instructions only to diminish artifacts stemming from demand characteristics and strategic processing on the part of the subject. However, without such negative instructions, RIT effects still arise at comparable rates.⁶

RIT effects of a complex nature have been observed. For example, in Merrick, Farnia, Jantz, Gazzaley, and Morsella (2015), subjects were presented with drawings of objects and instructed to (a) not think of the name of the object, and (b) not count the number of letters composing the object name. RIT effects arose for both mental operations on ~30% of the trials. In another study (Cho, Zarolia, Gazzaley, & Morsella, 2016), subjects first learned to transform words according to a rule resembling that of the childhood game of Pig Latin. After training, subjects were presented with words and instructed to not transform the words according to the newly-learned rule. Involuntary transformations arose on ~40% of the trials. It is worth noting that this involuntary effect requires, not only memory retrieval, but symbol manipulation, a process associated with frontal cortex (Miller & Cummings, 2007).

Might tasks that are more complex than those in Merrick et al. (2015) and in Cho et al. (2016) reveal the boundary conditions of the RIT effect? Such boundary conditions would shed light on the limits of involuntary processes and thereby illuminate the contributions of conscious processing.

Demand characteristics and inaccurate memories of ephemeral conscious contents (Block, 2007) could yield inaccurate self-reports in an RIT (see discussion in Morsella, Wilson et al., 2009). However, evidence suggests that subjects’ self-reports are accurate. First, in an RIT (Cushing, Gazzaley, & Morsella, 2017) in which subjects reported the occurrence of the basic RIT effect and also had to press a button if the involuntary subvocalization they experienced rhymed with a word held in mind, performance (>80% mean accuracy across trials) corroborated that subjects did experience involuntary subvocalizations. This is because detecting a rhyme requires retrieval of the phonological form of a word.

Second, in Bhangal et al. (2018), subjects were presented with an array of visual objects and instructed to not count the number of objects. When the number of objects was small (2–5 objects), the involuntary counting was very accurate (~90% mean accuracy), suggesting that the counting did in fact occur. Third, in RITs involving lexical retrieval, involuntary subvocalizations are influenced by the word frequency of the name of the object: High-frequency words are more likely to yield an RIT effect than low-frequency words (Bhangal, Merrick, & Morsella, 2015). Such a frequency effect would be unlikely to stem from demand characteristics. Regarding the possibility of artifacts resulting from strategic processing, on many trials of the RIT, the effect arises too quickly to be caused by such processing (Allen et al., 2013, Cho et al., 2014).⁷ In addition, neuroimaging data from studies (not involving the RIT) in which subjects self-report about the occurrence of involuntary thoughts or about subvocalizations corroborate that subjects do not confabulate about their reported mental events (Mason et al., 2007, McVay and Kane, 2010, Mitchell et al., 2007, Pasley et al., 2012, Wyland et al., 2003).

The view that the RIT effect is involuntary stems in part from theoretical explanations of the effect. According to Wegner (1994), undesired, “ironic” effects⁸ such as the RIT effect arise from an interaction between two distinct mechanisms. One mechanism is an operating process. This process is associated with the conscious intention to maintain a particular mental state. The process tends to be effortful, capacity-limited, and consciously mediated (Wegner, 1994). The mechanism actively scans mental contents (e.g., thoughts, sensations) that can help maintain the desired mental state (e.g., to be calm). The other mechanism is an ‘ironic’ monitoring process that automatically scans activated mental contents to detect contents signaling the failure to establish the desired mental state. When such a content is detected, that content then enters the conscious field. Of import, in Wegner (1994), the monitoring process is usually unconscious and autonomous. In other accounts of the RIT effect (Ach, 1905, Bhangal et al., 2016), the effect is the consequence of sets having been activated by the verbal instructions provided to the subject. From this standpoint, merely hearing the word “add” in the instruction “Do not add the following numbers” increases the activation level of the set to add, yielding “four” in response to the stimuli “2 and 2.” This set activation account is consistent with the tenets of parallel distributed processing (Rumelhart, McClelland, & the PDP Research Group, 1986), in which stimuli can activate, in an interactive network, “units” representing sets or rules. These units can in turn then automatically influence the activation thresholds of other, related units (e.g., representing “FOUR”) in the network. The RIT effect would also be deemed to be involuntary from the standpoint of a “crossmodal” account of mental representation. From this standpoint, the mental representations of objects are inherently multimodal (Ernst and Bülthoff, 2004, Lacey and Lawson, 2013, Spence and Deroy, 2013), in such a way that the activation of one sensory attribute of an object (e.g., the image CAT) will trigger automatically the activation of the other sensory features composing the multimodal representation, including those associated with linguistic label “cat.” Of most importance, in all theorizing regarding the mechanisms underlying the RIT effect, the nature of the effect is deemed to be involuntary.

One early concern was that the RIT effect was not noteworthy because stimulus-elicited memory retrieval is often automatic (Schacter & Tulving, 1994). However, such an interpretation cannot account for the RIT effects found in Cho et al. (2016), in which the involuntary verbal imagery required symbol manipulation, which is more than just memory retrieval. However, one criticism of Cho et al. (2016), in which subjects experienced involuntary word transformations, is that the involuntary effect observed (symbol manipulation involving syntax) is not noteworthy because syntax is largely unconscious. This led to the hypothesis that such symbol manipulation should fail to arise for a modality other than language (e.g., vision), a hypothesis whose falsification has important implications for theories regarding the limitations of involuntary, unconscious processes.

Perhaps the true boundary conditions of the RIT effect lie in mental operations requiring other forms of symbol manipulation, such as visuospatial imagery and mental rotation. Addressing this possibility is informative because symbol manipulation is associated with both executive function and cognitive control (Enger, 2017). With this in mind, one could conclude that perhaps RIT effects cannot arise for operations that require symbol manipulations that are visuospatial in nature. Identifying such a boundary condition would be informative, for the boundary conditions of the RIT effect reveal some of the limits of involuntary processing and thereby shed light on the contributions of conscious processing.

In the mental rotation task, subjects were first taught to mentally rotate (30°, 60°, or 90°) two-dimensional nonsense objects. After training, subjects were instructed to not mentally rotate in these ways a different set of objects. In the chess task, subjects were taught how to move in their minds (i.e., visuospatial imagery) objects in specified ways, much as one could imagine how, in the game of chess, a given piece can navigate the chessboard. Each object was associated with a unique pattern of potential movement on a chessboard-like grid. After training, subjects were instructed to not think of where each object could move on the grid. The order of presentation of the two tasks was fully counterbalanced across subjects.

Our aim was to assess whether high-level phenomena such as symbol manipulation can be influenced involuntarily and systematically through external control. We sought to obtain substantive evidence that, under controlled laboratory conditions designed to minimize artifacts and measurement error, these effects on symbol manipulation can occur involuntarily and at a reliable rate. This would provide evidence that high-level, symbol manipulation can occur involuntarily.

We should add that the RIT is the kind of paradigm that, because it builds incrementally on robust phenomena and previous research, has been encouraged by leading researchers in the field (e.g., Fiedler, 2017, Nosek et al., 2012). The reliable, component processes of the present variant of the RIT are of interest in disparate subfields of the study of mind and brain, including consciousness, cognitive control, and imagery.