How Do the Blind 'See'? The Role of Spontaneous Brain Activity in Self-generated Perception

Avital Hahamy; Meytal Wilf; Boris Rosin; Marlene Behrmann; Rafael Malach


Brain. 2021;144(1):340-353. 

In This Article

Abstract and Introduction


Spontaneous activity of the human brain has been well documented, but little is known about the functional role of this ubiquitous neural phenomenon. It has previously been hypothesized that spontaneous brain activity underlies unprompted (internally generated) behaviour. We tested whether spontaneous brain activity might underlie internally-generated vision by studying the cortical visual system of five blind/visually-impaired individuals who experience vivid visual hallucinations (Charles Bonnet syndrome). Neural populations in the visual system of these individuals are deprived of external input, which may lead to their hyper-sensitization to spontaneous activity fluctuations. To test whether these spontaneous fluctuations can subserve visual hallucinations, the functional MRI brain activity of participants with Charles Bonnet syndrome obtained while they reported their hallucinations (spontaneous internally-generated vision) was compared to the: (i) brain activity evoked by veridical vision (externally-triggered vision) in sighted controls who were presented with a visual simulation of the hallucinatory streams; and (ii) brain activity of non-hallucinating blind controls during visual imagery (cued internally-generated vision). All conditions showed activity spanning large portions of the visual system. However, only the hallucination condition in the Charles Bonnet syndrome participants demonstrated unique temporal dynamics, characterized by a slow build-up of neural activity prior to the reported onset of hallucinations. This build-up was most pronounced in early visual cortex and then decayed along the visual hierarchy. These results suggest that, in the absence of external visual input, a build-up of spontaneous fluctuations in early visual cortex may activate the visual hierarchy, thereby triggering the experience of vision.


During rest and independent of any external stimulation, the brain evinces activation in a spontaneous manner ('resting state' activity) (Arieli et al., 1995; Nir et al., 2006; Fox and Raichle, 2007; Raichle, 2009; Harmelech and Malach, 2013; Moutard et al., 2015). It has previously been proposed that a slow build-up of spontaneous activity can initiate unprompted (spontaneous) behaviour (Schurger et al., 2012; Moutard et al., 2015). This phenomenon has been classically observed in the readiness potential during decisions to move (Kornhuber and Deecke, 1965; Libet et al., 1983; Libet, 1985; Soon et al., 2008; Fried et al., 2011), but has also been observed during other cognitive tasks (Gelbard-Sagiv et al., 2008, 2018; Norman et al., 2019; Broday-Dvir and Malach, 2020). However, in these studies, the mere instruction to generate behaviour spontaneously prevents behaviour from being purely unprompted. Furthermore, since these studies involved task performance, brain activity was likely to be affected by the task demands, rather than being entirely spontaneous.

Here, we addressed the question of whether spontaneous brain activity evokes unprompted cognitive behaviours, by examining participants with Charles Bonnet syndrome (CBS). This condition is characterized by complex visual hallucinations in individuals with a marked visual impairment, in the absence of a cognitive or mental disorder (de Morsier, 1936, 1967; Teunisse et al., 1996). As individuals with CBS typically are unable to control the occurrence or the content of these hallucinations, hallucinations are genuinely unprompted perceptual behaviour. Furthermore, as neurons deprived of external inputs demonstrate enhanced spontaneous brain activity (Echlin et al., 1952; Loeser and Ward, 1967; Segal and Furshpan, 1990), it has been theorized that spontaneous activity in the deprived visual cortex of individuals with CBS may subserve hallucinations (Cogan, 1973; Burke, 2002; Plummer et al., 2007; Reichert et al., 2013).

This hypothesis is particularly appealing in light of studies showing that, in sighted individuals, spontaneous (resting-state) activity can activate the entire visual system in aspects that are similar to those induced by naturalistic visual stimuli (Nir et al., 2006; Gilaie-Dotan et al., 2013; Wilf et al., 2017; Strappini et al., 2018). It is therefore possible that, in the absence of visual input, the existing neural mechanism that typically underlies veridical vision, and spans the entire visual cortex, may instead be activated by random spontaneous fluctuations. The triggering of the existing network-wide neural cascade by spontaneous activity might explain why the resulting hallucinations in CBS can be as vivid as they are in normal vision (Teunisse et al., 1996; Menon, 2005).

Support for this hypothesis is, however, lacking. Previous studies of CBS have associated hallucinations with isolated visual regions (Ffytche et al., 1998; Adachi et al., 2000; Santhouse et al., 2000; Kazui et al., 2009; Painter et al., 2018), for example, in the surround of the fusiform gyrus (Ffytche et al., 1998), rather than with the entire visual cortex. Yet, given the interconnected nature of visual areas (Smith et al., 2009), and their hierarchical functional organization (Grill-Spector and Malach, 2004), it seems unlikely that local activity, which is robust enough to evoke vision, would remain confined to a small region. The role of spontaneous brain activity in evoking hallucinations is also unclear, since, on the one hand, recent imaging findings ascribed hallucinations (Vacchiano et al., 2019) or abnormal visual responses (Painter et al., 2018) in individuals with CBS to external stimulation. On the other hand, the isolated activity in the fusiform gyrus, described by Ffytche et al. (1998), appeared to build up slowly over time prior to hallucination onset, which aligns with the hypothesis that slow spontaneous fluctuations underlie the CBS hallucinations. In summary, as evident from the above, neither the spatial nor dynamic neural mechanisms of CBS have been fully clarified.

Here, we compared directly the neural correlates of hallucinations in individuals with CBS with the activation profiles evoked by veridical vision in sighted controls and by visual imagery in blind controls, across the entire visual cortex. As illustrated in Figure 1, we predicted that (i) in the spatial (anatomic) domain, visual hallucinations would be associated with distributed activity across the visual system, similar to activations evoked by veridical vision in the sighted and by visual imagery in the blind (Cichy et al., 2012; Dijkstra et al., 2017, 2019); and (ii) in the temporal domain, unlike in other visual experiences, a build-up of activity in visual areas would precede the emergence of hallucinations. Such findings would suggest that temporal, rather than spatial, aspects of neural activity differentiate between unprompted visual hallucinations and other cued visual experiences.

Figure 1.

Schematic of experimental design and participants. (A) Top: Individuals with CBS were asked to verbally/manually report their visual hallucinations while in the MRI scanner. Inset: Schematic of blood oxygenation level-dependent (BOLD) signal in the visual system, vertical line represents the timing of hallucination report. The hypothesis for this condition is that activation in the visual system would ramp up prior to the perception of hallucinations. Bottom: Schematic of hallucination content of all CBS participants (Table 1). Top row depicts the hallucinatory content of the three CBS participants who were able to report their hallucinations; bottom row depicts the hallucinatory content of the two participants who were unable to report their hallucinations (see 'Materials and methods' section). (B) Verbal reports of hallucinations of the CBS participant were illustrated as movies, which were presented to sighted control participants. Inset: The experimental hypothesis of this condition is that activation in the visual system would ramp up only after visual stimuli are perceived. (C) All participants completed a visual imagery scan, in which they were asked to imagine faces, houses, objects and patterns. Inset: The hypothesis for this condition is that activation in the visual system would ramp up only after visual imagery has initiated. As depicted in this figure, visual hallucinations, like other visual experiences, were hypothesized to activate the entire visual hierarchy.