Sometimes referred to in sports as “visualization,” motor imagery (MI) involves the imagining and vivid mental rehearsal of movements without overt motor output (Moran et al., 2012). It is commonly classified according to predominant type as kinesthetic (“feeling”) or visual (“seeing”), with visual taking either the first- or third-person perspective. In practice, multiple modalities may be used concurrently. Psychological evidence for the use and effectiveness of visualization in sports is well established (Weinberg, 2008). Early sport psychology research focused on methods for optimizing MI practices to maximize performance, leading to the development of the well-known kinesthetic method referred to by the acronym PETTLEP (Holmes and Collins, 2001). Conceptualized as a neuroscientific approach to MI in sport, the PETTLEP model proposed that seven key features—physical, environment, task, timing, learning, emotion, and perspective—be incorporated into an athlete’s imagery practices to be most effective. Grounded in the theory of functional equivalence, which presumes that overt and covert movement share similar patterns of activation in the brain (Lang, 1979; Jeannerod, 1995; Kraeutner et al., 2014), parallel to visual imagery and perception (Farah, 1989; Pearson et al., 2015), PETTLEP has been widely adopted as a tool in sport psychology, although questions persist regarding underlying neurophysiological mechanisms and the concept of neuroanatomical functional equivalence (Wakefield et al., 2013).

For “open” sports, defined by unpredictable environments and a need for athletes to react rapidly to complex behaviors, visualization of potential scenarios requires a multimodal approach to MI that integrates manifold aspects of human perception, experience, and action. Ice hockey is a fast, open team sport in which players make rapid decisions in response to an infinite number of different situations, decisions based on cognitive—conscious or subconscious—associations between multiple types of sensory stimuli and motor schema. One might intuitively expect that MI would be most useful in “closed” sports (e.g., diving, figure skating, and golf), and experimental studies of MI have commonly investigated simple movements (e.g., sequential finger tapping and object grasping) in a laboratory setting or, in sport-specific contexts, have focused on closed sport actions (e.g., golf swing) (Decety et al., 1994; Ross et al., 2003; Guillot et al., 2008; Kraeutner et al., 2014).

We were particularly interested in the more complex form of MI used by open sport athletes, where PETTLEP-based “scripts” of multifaceted scenarios are used as the foundation for an athlete’s imagery practice. At a fundamental level, complex MI provides a model of how covert thought—imagination—can affect subsequent behavior, a mind–brain phenomenon that presents an opportunity to explore the relationship between consciousness and action (Moran et al., 2012). The specific goal of this exploratory study was to determine if the brain region(s) most involved in sustaining this form of dynamic imagination could be revealed by magnetoencephalography (MEG).

The script used in this study of eight elite female ice hockey players incorporated all PETTLEP features with as much detail as possible (see Appendix: Visualization Script). Using an experimental block design, the athletes visualized a specific 2-on-1 ice hockey play (a goaltender used a different script) while their brain activity was monitored using MEG. In brief, 3-min baseline recording (resting state, no task) was followed by 10 trials of 30 s of imagery, alternating with 30 s of a mental counting task. MEG beamforming was then used to contrast the data from the imagery condition with the two different control (resting state or mental counting) conditions.

All participants demonstrated good imagery ability on the MIQ-RS (Supplementary material) and reported no difficulty with visualization inside the magnetically shielded room.

The MEG beamforming results obtained at the group level, comparing relative change in power at the FOI for the different conditions (imagery versus resting state or mental counting), revealed prominent desynchronization—that is, decreased oscillatory beta power, reflecting neuronal activation (Miller et al., 2014)—during the imagery condition in the left posterior hemisphere. This effect was maximal around the intraparietal sulcus and extended to the inferior aspect of the superior parietal lobule and the superior aspect of the inferior parietal lobule.

Figure 1 shows the contrast between imagery and the resting state baseline control.

Figure 2 shows the contrast between imagery and the mental counting control.

The statistical analysis confirmed the observed visualization-related desynchronization to be significant in amplitude and spatial extent in both contrast settings. Specifically, the cluster-based permutation test identified a significant cluster in the posterior left hemisphere parietal region (p < 0.01, corrected for multiple comparisons).

As suggested in the group-level figures, along with the maximal left parietal activation, significant desynchronization was also present during the imagery condition in the left > right pre- and post-central gyri. This activation in primary sensorimotor cortex was most evident when the imagery condition was compared to the mental counting control, as opposed to the resting state, implying that the resting state contained greater baseline sensorimotor neural activity than the more focused non-motor mental counting state.

Remarkably, the same pattern of maximal activation in left posterior hemisphere cortex was evident at the individual level in 100% of subjects (8 out of 8 participants). This consistent finding included the goaltender, who used a different visualization script, indicating that the left parietal locus of activation was not dependent on the exact content details of an individual’s MI but rather reflected the dynamic cognitive process of the visualization.

At the individual level, lesser significant activation in the left and/or right sensorimotor cortex was observed in six out of eight participants (three unilateral left, two bilateral, and one unilateral right). Some of the athletes also showed lesser significant activation in adjacent left occipital cortex (two out of eight participants) and/or in contralateral right parietal cortex (two out of eight participants, most apparent in the goaltender, Figure 3).

Complex PETTLEP-guided visualization in this group of elite female ice hockey players was primarily associated with functional activation of left hemisphere parietal lobe areas involved in somatosensation, kinetic visuospatial processing, and multimodal sensorimotor integration (Farah, 1989; Sirigu et al., 1996). A similar activation pattern was evident when the MI task was contrasted against resting state or active counting conditions, indicating its specificity to the cognitive operations involved in this type of visualization. Despite, or perhaps because of, the complexity of the MI, this region was reliably activated in all participants and could be identified by MEG even at the individual level. Moreover, the same left parietal region was maximally activated in the goaltender, who used a different visualization script, demonstrating that it was not the precise imagery content (beyond being a dynamic multisensorial hockey-specific moving image) but rather the conscious performance of complex MI—the imagining process itself—that was detected by MEG beamforming.

Previous studies of MI using positron emission tomography (PET) or functional MRI (fMRI) have identified multiple additional cortical and subcortical regions of brain activation, including frontal premotor and supplementary motor areas, the cingulate gyrus, the caudate nucleus, and the cerebellum (Hétu et al., 2013; Jiang et al., 2015; Filgueiras et al., 2018), all areas that showed no significant activation identifiable by MEG beamforming. The differences in localization across studies and imaging methods could conceivably reflect differences between complex and simple MI, between open and closed sports, between good imagers and less adept individuals (Guillot et al., 2008), or perhaps especially between PET/fMRI and MEG beamforming.

Our use of MEG beamforming provided a direct measure of neural activity with millisecond temporal resolution, different from PET or fMRI, which could explain the more discrete localization found with MEG, as non-parietal areas identified by PET/fMRI could represent downstream network connections of the principal locus of image generation. The fast temporal resolution of MEG, combined with the short sliding window of the beamforming technique, may have been ideal for the identification of consistent, prolonged activations during visualization. In this context, the beamformer, in essence, acts as a low-pass filter to de-emphasize transient or spatially inconsistent activations that might be more likely to appear as metabolic changes in PET or fMRI (see Supplementary material for more discussion of this issue).

It is conceivable that complex MI in elite hockey players would not significantly activate anterior motor planning areas if the athletes subconsciously “filter out” or automatize the basic motor skills they are already expertly familiar with during visualization. Experienced athletes expend less energy and activate fewer brain areas than novice athletes during MI and sport-related decision-making (Ross et al., 2003; Milton et al., 2007; Filgueiras et al., 2018). At the elite level, where these differences would be expected to be most pronounced, complex MI may be limited largely to the cognitive aspects of imagination and localized primarily to posterior left hemisphere parietal lobe cortex. Dynamic sensorimotor integration in this left parietal region during visualization could facilitate novel connections and optimize the cognitive hub of existing performance-related cortical–subcortical neural networks, providing a possible mechanism whereby covert PETTLEP-guided thought can improve an athlete’s subsequent performance.

The MEG identification of a left parietal cortical hub for complex MI provides both neurophysiological and neuroanatomical support for certain theoretical models derived from cognitive neuroscience; specifically, hierarchical internal forward (Tian and Poeppel, 2010) and predictive-processing (Ridderinkhof and Brass, 2015) models, which propose that “efference copies” or “emulations” of imagined actions are generated and dynamically updated in posterior parietal cortex, upstream from frontal motor planning and subcortical motor coordination areas. Viewed in terms of network connectivity, the continuously active parietal integration hub may transiently recruit other areas not needing sustained activation during imagery, but these areas would either be unidentified by MEG beamforming or show only mild or individual-level activation (e.g., primary sensorimotor, left occipital, or right parietal cortices).

Our findings, while preliminary, are hypothesis-generating, in particular about the brain mechanisms underlying image generation, dreaming, and phenomenal consciousness. Historically, neuropsychological lesion studies using componential analyses to identify focal deficits in image generation found a significant association with lesions in the left posterior hemisphere (Farah, 1984; Stangalino et al., 1995). Interestingly, some of the affected patients also reported a loss of dreaming (Farah, 1984), and a left posterior hemisphere lateralization of dreaming was later supported in a neurological review of these and other cases (Greenberg and Farah, 1986).

Dreaming is quite obviously a natural sleep analog to complex MI, and our demonstration of a similar neuroanatomical localization for complex MI supports and refines the classical imagery localization hypotheses. Localization of the neural correlates of dreaming has been a subject of interest in modern consciousness studies, with the identification of a posterior hemisphere “hot zone” for dreaming used to support the postulate of a similar neural locus for all phenomenal consciousness (Tononi et al., 2016; Siclari et al., 2017). Complex MI is a form of phenomenal consciousness that applies covert thought as a tool to alter subsequent motor behavior. The means by which thought can later influence human behavior, whether in the case of PETTLEP-guided MI for sports performance or as long-reported anecdotally in the case of dreaming, is an unresolved mind–brain problem. Nevertheless, the physical substrate of the process appears to predominantly involve the posterior cortex of the left hemisphere.