Mental imagery refers to the phenomenon of perceiving an event or object in the absence of the actual stimuli. Essentially, imagery represents the top down processing analogue of perception where detection in the latter stems from the stimulation of the senses and the former does not. As such, most preliminary research was dedicated to characterizing the neural basis of imagery processes – and although most studies indicate a large overlap with brain regions recruited during perception, imagery is generally marked as a modality specific experience lacking the activation of its corresponding primary sensory cortex and with more pronounced activation in associative cortex. Modalities include motor, tactile, olfactory, gustatory, auditory, and visual. The study of mental imagery is also involved in the research of popular topics such as working memory, recollection, and pain and movement perception in phantom limb patients . Moreover, the applications of mental imagery include rehabilitative procedures used on stroke patients, and the basis of neural prosthetics.
1.1 Visual Imagery
In most studies, tasks involving visual imagery generally depict a pronounced deactivation of the primary visual cortex (V1) and activation of the surrounding extrastriate cortex. In other tests the V1 is recruited during the processing of high detail mental images – which indicates a visual hierarchy that proceeds in descending acuity from the associated cortex to the V1. However, beyond that point the visual imagery network extends to include the occipital – temporal and fusiform face areas. Prefrontal and superior parietal regions are activated in the imagination of objects while early visual areas are activated in the viewing of objects – but both tasks result in the activation of ventral occipital – temporal regions. Thus, when attending images, retrieval of sensory representations within occipital – temporal regions is achieved following activation of early visual areas as a case of bottom up processing. Likewise, activity in fronto-parietal regions that stem from imagining objects also leads to retrieval of representations in a case of top down processing. These regions were believed to form part of the general imagery network, and the prefrontal region was shown to be activated during category specific recall, whereas the parietal region was activated regardless of imagined stimuli.
1.1b Decoding Mental States
Much preliminary work on mental reconstruction focused on perception. Using a mathematical model, existing neurological encoding models were used in conjunction with fMRI scanning to compensate for the slower BOLD signals to obtain a better temporal profile of underlying neural contributions. In essence, this technique was used to map out and predict visual images viewed by participants based on their activation profiles in early visual areas. In addition, multi-voxel pattern analysis techniques can be used to decode category level information of viewed objects and, recently, imagined objects. Viewing and imagining objects resulted in the same voxel activation patterns within the ventral temporal cortex, however, such is also activated albeit to a lesser extent by imagery alone – without bottom up input.
1.1c Working Memory
As prefrontal regions have been implicated in the generation and maintenance of mental images, visual imagery was also thought to be involved in working memory – sharing in the underlying mechanism for the visuospatial sketchpad. Consistent with what is known, previous experiments with image recall tasks reflect increased activation in extrastriate cortex in contrast to decreased V1 activity. Generally, tests on visual imagery involved rating the subjective experience of imagining a previously viewed image. However, a recent study analyzed the effects of tactile memory training – memorizing and subsequently tracing an images drawn by the participants themselves - on the activation profile that represents the use of the visuospatial sketchpad. In such a case, there was an activation of V1 with deactivation of extrastriate cortex – with frontal and parietal recruitment constant. Suppression of surrounding associated cortex would sever the top down propagation of imagery signals from the former toward the V1. Rather, independent activation of the V1 contradicts the visual hierarchy of processing established in previous studies.Thus, a theory of an 'amodal' sketchpad was proposed as different regions are apparently recruited in working memory depending on the nature of the initial task to purportedly generate and maintain a visual image.
Further evidence that supports the notion of amodality is provided by fMRI scans obtained in a separate study – from a patient who reportedly lost the ability to visualize images, as well as having lost all visual content in dreams. Scans show pronounced deactivation in superior temporal, occipital, and fusiform regions when compared to profiles of healthy individuals following face imagery. Despite this, the patient was able to successfully completely experimental tasks, propounding alternative methods of completing imagery tasks in absence of mental visualization – although exact compensatory mechanism for this case is not yet defined.
Findings from Likova et al, and Zeman et al appear to dissociate visual imagery from working memory. However, because it may be possible to complete standard visual imagery tasks without actually visualizing targets, future studies should explicitly measure both objective performance and subjective experience following trials as the two do not share a perfect correlation. This is reinforced by the finding that the activation of posterior visual cortices matched the subjective ratings of visual imagery but not the completion of experimental tasks.
1.2 Auditory Imagery
The network for auditory imagery consists of the STG – superior temporal gyrus and association cortices within it – whose activation is generally coupled with the deactivation of the primary auditory cortex. It is bilaterally activated in musical imagery, and recent evidence supports the correlation between self reports of vividness of auditory imagery – as reflected by STG activation levels – and accuracy of pitch imitation though not pitch discrimination. Moreover, activation of the STG is associated with the silencing of other modality networks such as the extrastriatal cortices in visual imagery, motor regions, and other primary sensory areas – and is conversely suppressed during imagery of other modalities.
1.2b Parsing Simple Imagined Words
The technique of stimulus reconstruction refers to the parsing of neural responses to various stimuli in an effort to eventually reconstruct the original stimulus based on a consensus. This can be achieved through the use of multichannel electrode recordings. Ultimately, the process of reconstructing auditory information will allow for the observation of how it is deconstructed in the brain.
1.3 Model for General Imagery
Although many studies have focused exclusively on a particular modality of imagery , a few recent studies have attempted to describe a general – or supramodal network of mental imagery. The DMN (default mode network) is comprised of posterior cingulate, lateral posterior parietal, and medial prefrontal regions and was found across several experiments to be activated during imagery tasks and deactivated during perceptual tasks. Thus it was purported as a network responsible for preliminary generation of mental representations across modalities. The mechanism of modal selection that precedes the recruitment of sensory areas - corresponding to the nature of the imagined event -remains unknown. Following DMN activation, selection was thought to be mediated by a modal independent semantic top down selection from a memory storage. However, the candidacy of the DMN was questioned in a recent meta analysis which posits that any observed activation during imagery may simply due to the lack of perceptual related processes during experiment tasks - as the DMN is characteristically active when the brain is at resting state. The same study reported consistent activation of several clusters within the left inferior frontal and bilateral dorsal parietal regions in imagery tasks across modalities – which may suggest an alternate model to a general imagery network. These correlate with previous experiments which detected content independent activation in frontal and superior parietal regions.