Aim of the present study was to test a new experimental paradigm to quantify imagination abilities. For this we used a classical conditioning paradigm with ambiguous stimuli and unambiguous stimulus variants from two different categories, i.e. cubes and letter/number stimuli (Real Conditions). We extended the classical conditioning paradigm by a second experimental condition, where the conditioning stimulus wasn’t observed but instead had to be imagined (Imaginary Conditions). The basic underlying idea was to use the probabilities of imaginary conditioning as objective quantitative measures of participants imagination ability by correlating them to related probabilities of real conditioning. We further compared the probabilities of imaginary conditioning with VVIQ questionnaire scores as a subjective measure of the participants’ imagination ability.
We neither found systematic priming nor systematic adaptation effects on the group level. In contrast, some individual participants showed strong priming effects and some participants showed strong adaptation effects and again others showed neither. This pattern of results was observed across the classical Real Conditions of the two experiments and the Imaginary Conditions. Interestingly, our subsequent correlation analysis revealed that most of the participants with strong real priming or adaptation showed the same direction of conditioning and similar intensities in the Imaginary Conditions. This pattern of correlation results was consistent across cube and letter/number stimuli. We did not find a systematic relation between real or imaginary conditioning and VVIQ scores. However, on a more qualitative level, we observed a small cluster of three participants with high scores on the VVIQ scale in combination with low Imaginary Condition probabilities. These three participants identified themselves as aphantasics.
Interpreting the absence of systematic adaptation and priming effects at the group level
A considerable number of experimental studies used ambiguous and unambiguous stimuli in selective conditioning paradigms, very similar to the present Real Conditions 26,38–45. Most of those studies reported group adaptation and priming effects. It was particularly found that conditioning effects come with a retinotopic pattern. For example, no conditioning was found if the conditioning stimulus was presented in one visual hemifield and the subsequent test stimulus in the other visual hemifield, or if conditioning and test stimuli differed substantially in size 43. Moreover, both the quality (priming or adaptation) and the amount of conditioning was shown to be a function of the presentation duration of both the conditioning stimulus and the temporal gap between conditioning and subsequent test stimulus. More and larger priming effects were found with short conditioning periods and long gaps, whereas more and larger adaptation effects occurred with long conditioning periods and short gaps 44–46. Possible dependent variables to assess conditioning effects included the initial percept of the ambiguous test stimulus immediately after its onset, the number of subsequent perceptual reversals across the duration of test stimulus observation and the total time of perceptual stability of the two different perceptual interpretations across the total time of test stimulus observation 45. In the present study we focused on the initial percepts as dependent variable.
One interesting question is why we did not find any systematic classical priming nor systematic adaptation effects on the group level, like the previous studies? The reason for this may be that we used a fixed single value for the conditioning time (5 s) and for the gap duration between conditioning stimulus and test stimulus (0.3 s). In their seminal study, Toppino and Long 43 tested conditioning times between 0 and 90 s and found priming effects for short conditioning times, and adaptation for long conditioning times. For conditioning times at around 1 s neither priming nor adaptation effects were found. The authors suggested that the two opposite conditioning effects neutralized each other in this case. According to their results, we should have found mainly adaptation effects with our 5-seconds conditioning time. Of course, the two studies differ in a number of experimental parameters, including the choice of stimuli. Another major difference is that in our study each Necker cube trial was followed by a letter/number trial (see Fig. 3). In Toppino and Long’s study the conditioning trials were executed in succession without interruption. Whether this or the different stimuli explains the difference in results between their and our study has to be accessed in a follow-up study. In the present case, it is well possible that priming and adaptation time constants vary to a certain amount within a population and that the conditioning time we used supports priming in some participants, adaptation in others and a neutralization of the two opposite effects in again other participants, possibly resulting in a null effect on the group level.
Perception and Imagination
The retinotopic character of the conditioning effects, as reported in the literature and described above, indicate mechanisms at lower levels of the processing hierarchy, where receptive fields are reasonably small. However, similar priming and adaptation effects have been reported for a variety of stimuli at very different levels along the perceptual hierarchy, from the perception of contrast 47, motion 48,49, line orientations 50, up to the perception of the emotional content of faces 51 and even with numerosity 52.
Mental imagination is the ability to activate perceptual representations from memory without any sensory input, inducing the experience of “seeing with the mind’s eye” 1. It has been proposed that the brain uses almost the same “hardware” for visual imagination as for visual perception. In confirmation of this proposition, neuroimaging studies indicate similar neural representations when objects are imagined and when they are perceived (e.g. 1). A number of related findings further confirm this approach: Some behavioral studies reported that imagination content can particularly affect perception (Perky, 1910; McDermott and Roediger, 1994; Pearson et al., 2008). For example, imagination has been shown to affect visual detection thresholds (Ishai and Sagi, 1997) and performance on a visual acuity task (Craver-Lemley and Reeves, 1992). Repeatedly visualizing the critical region of a visual bisection stimulus (visual spatial judgment) or a low-contrast Gabor pattern (contrast judgment) can enhance performance on subsequent perceptual tests (Tartaglia et al., 2009). Similarly, imagination of performing motor acts can improve performance on related tasks, probably by improving the connections between key areas of the motor cortex (Driskell et al., 1994; Weiss et al., 1994; Feltz and Landers, 2007). In addition, imagination can give rise to adaptation effects (Gilden et al., 1995) as much as a corresponding sensory stimulus.
The proposition of involvement of the visual-perceptual system during visual imagination together with its retinotopic organization fits well with the retinotopic pattern of priming and adaptation, as found in the literature, and also with our correlation analysis results. Let us assume that the intersection between neural structures that are active during perception of an ambiguous figure and during its imagination contains those structures that are responsible for priming and adaptation. Then it is reasonable that the imagination of an unambiguous stimulus variant has a highly similar effect on the subsequent ambiguous test stimulus as the real percept of the conditioning stimulus, with more or less the same time constants for priming and adaptation within individual observers. This is exactly what we found. Most of our participants who showed strong and those who showed weak priming or adaptation effects in the Real Conditions showed the same pattern in the Imaginary Conditions.
The present study is only on the level of a “proof of the principle” and the functional details need to be further clarified in future experiments, including measurements of the underlying neural mechanisms and comparing cerebral sources. Moreover, a considerable number of our participants showed priming probability values close to the chance level (p = 0.5). Given that we only used one conditioning time, we cannot interpret these results unambiguously, as will be discussed below in more detail.
Are the real and Imaginary Condition effects good predictors for visual imagination abilities?
The strong correlation between real and Imaginary Condition indicates that vivid visual imagination activates and therewith primes or adapts the perceptual system in the same way as visual perception does. Particularly, if our choice of conditioning time induced either a strong priming or a strong adaptation effect in the Real Condition, it induced the same type of conditioning (priming or adaptation) and with about the same intensity in the Imaginary Condition. Moreover, those participants who showed weak or no real conditioning effects produced comparably weak imaginary conditioning. At first sight these results indicate that real and imaginary conditioning effects may be good objective measures of imagination abilities. However, we also applied the VVIQ questionnaire to measure the subjective vividness of imagination and correlated the resulting vividness scores with the behavioral conditioning variables. No significant relation between the vividness of imagination and the amount of imaginary conditioning was found.
One problem of the present study is that our participants with priming probability values close to chance level (i.e., 0.5) in both conditions, may be a mixture of different phenotypes, who cannot be separated based on the present data. Some participants may have high VVIQ scores but have conditioning probabilities close to chance level. They would have produced strong real and imaginary priming or adaptation effects with another choice of conditioning times. Others may a priori be unable to produce strong conditioning effects or any conditioning effects at all, independent of the conditioning time. Among those may be aphantasics, i.e. people who are unable to have a visual imagination. Keogh et al. 33 particularly focused on the phenomenon of aphantasia. They applied a similar conditioning paradigm as the present one to a group of predefined aphantasics and compared their results with a control group of non-aphantasics. In contrast to our choice of classical ambiguous figure as stimuli, they used binocular rivalry stimuli to induce conditioning effect e.g. 53. Binocular rivalry occurs, when the two eyes see different stimuli at the same time e.g. 54. Similarly, to the classical ambiguous stimuli, perception alternates between the two eyes’ input. Keogh et al. found a correlation between the VVIQ scores and imaginary conditioning. However, their correlation is mainly based on the fact that their samples of aphantasics and controls form two clusters on the VVIQ scale. Our finding of no correlation between VVIQ scores and conditioning may be explained by the invisibility of potential conditioning effects, due to suboptimal conditioning times. The observation from our Fig. 6, however, indicating also a clustering of our three aphantasic participants based on high VVIQ scores together with low conditioning probabilities is a qualitative confirmation of Keogh et al.’s results.
In a follow-up study we need to access optimal priming and conditioning times for each participant individually and execute the present experiment with individually optimized conditioning times. This may disentangle the potential mixture of subgroups with low conditioning results. We further need a separate group of aphantasics, to see whether our qualitative observations can be confirmed on a quantitative level.
Outlook
It is well-known that what we consciously perceive at a present moment not only depends on what enters our eyes (bottom-up information), but to varying degrees also on what we have perceived in the past (top-down information) 50,55,56. This is part of the constructive nature of perception. Our results, together with the findings from Keogh et al. 33, indicate that not only our previous percept (across different memory time scales) but also our recent visual imagination can have an influence on how we perceive the world around us at a present moment. The imagination literature teaches us that both real perception and imagination need the “perceptual brain” as the underlying hardware. Our correlation findings provide an easy and convincing way to behaviorally confirm this finding. Moreover, it may be possible to increase effect sizes and explanatory power, by using individually optimized conditioning time constants in the present experimental paradigm.
The latest results from aphantasia research, however, questions the generality of the relevance of the “perceptual brain” for imagination. Aphantasics may not be part of a continuum of visual imagination abilities, but instead represent a separate population with separate imaginary mechanisms. The present experimental paradigm is based on a similar idea as the binocular rivalry paradigm introduced by Keogh & Pearson 33. With the Real Condition, it contains an important extension that may allow more precise statements about the potential mechanisms underlying “normal” visual imagination. Our paradigm may thus become a promising candidate for an objective measure of imagination abilities and to identify subgroups. Importantly, the choice of stimuli (ambiguous figures) and the simplicity of the task allow to access imagination abilities in online studies and has thus the potential to acquire large amounts of data sets.
In 1995 Ishai and Sagi defined visual imagination as the activation of a perceptual representation in the absence of sensory input (Ishai and Sagi, 1995). Hallucination is another type of perception without sensory input, which can occur in any sensory modality. People sense touch, without being touched, they hear voices in silence, or they see objects, forms or shapes, their environment cannot see or hear. Such percepts without sensory input are typically involuntary and occur often but not always in combination with diagnoses of psychiatric disorders. In such cases, the intrusive nature of the phenomenon can substantially decrease the quality of life.
There are also other cases of exceptional perceptual experiences without sensory input. They also come sometimes with a severe psychological burden but without a clear clinical diagnostic and an unclear ontological status (Landolt et al. 2014; Atmanspacher and Fach 2019).
One highly relevant question for basic research is now, whether all these cases of perception without sensory input, irrespective of the differences in real-life shaping, share some basic neuronal sources and/or functions. The present paradigm may be or become a good starting point for future research in this direction.