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


Here, we demonstrate that visual hallucinations, much like any other type of visual experience, engage the entire visual system. We additionally report that, unlike other types of visual experiences, visual hallucinations are related to a spontaneous build-up of neural activity, which is most pronounced in early visual cortex, and decays along the visual hierarchy. Taken together, our findings propose a network view of visual hallucinations in CBS, under which a slow build-up of spontaneous neural activity in early visual regions may bring about a perception of vision by triggering a cascade of activity throughout the visual system.

Our findings suggest a plausible mechanism underlying the emergence of visual hallucinations in CBS. Specifically, deafferentation of the visual system may lead to over-excitability in visual areas (Desai et al., 1999; Burke, 2002; Reichert et al., 2013; Painter et al., 2018), and to enhanced susceptibility to noise fluctuations. Given the normally-developed hierarchical structure of the visual system in (previously sighted) individuals with CBS, when an activation threshold in early visual regions is crossed by a spontaneous fluctuation, a cascade of activations across the visual hierarchy may lead to the emergence of visual hallucinations (Moutard et al., 2015). Further imaging studies are needed for clarifying the unique characteristics of spontaneous brain activity in individuals with CBS compared to non-hallucinating blind individuals.

Unlike our findings, studies of sighted patients with dementia-related visual hallucinations reported a reduction in visual activation (Meppelink et al., 2009; Goetz et al., 2014), which may indicate that hallucinations are a product of impaired bottom-up processing of external visual stimuli. However, we propose that hallucinations in CBS are related to intact bottom-up visual processing, but are a product of internal, rather than external, visual inputs (as individuals with CBS are visually deprived). Thus, our findings do not offer a unified account for visual hallucinations across many disorders, but instead illuminate the possible role of spontaneous brain activity in evoking visual hallucinations consequent to visual deprivation.

In this interpretation, we assume that the activity seen in the visual system of blind individuals is spontaneous (Echlin et al., 1952; Loeser and Ward, 1967; Segal and Furshpan, 1990). Yet, it is also possible that the visual system of these individuals is activated by inputs from other brain areas processing non-visual information (cross-modal plasticity; Edelman, 1993; Amedi et al., 2003; Collignon et al., 2011; Renier et al., 2014). However, the extent of cross-modal plasticity in the late blind is still debated (Collignon et al., 2013; Voss, 2013). Additionally, hallucinations in our sample of CBS participants appeared in the absence of any external sensory trigger, suggesting that sensory inputs did not elicit the hallucinations.

It could also be argued that the slow build-up of activity observed in the hallucination condition may simply reflect a delay between hallucination onset and its report, due to the elusive nature of the hallucination. However, all CBS participants included in this condition indicated that they could report their hallucinations as promptly as they could report visual stimuli before their vision deteriorated. Moreover, if this hypothetical confound actually held, hallucinations should evoke an increase in BOLD signal, similarly to veridical vision (Figure 1B). A consequence of this is that there should be rapid propagation of activation across the visual system (Dijkstra et al., 2020) leading to an 'artificial build-up' of activity across the entire visual hierarchy. However, this possibility is contradicted by our findings of a decaying activity build-up across the visual hierarchy (Figures 4 and 5). It is therefore more likely that the anticipatory signals reflect accumulation of spontaneous activity prior to hallucination onset, rather than hallucination-evoked activity which precedes the onset of report.

While our findings indicate an endogenous trigger for visual hallucinations in CBS, a recent EEG study proposed that hyperexcitability in the early visual cortex of elderly, partially blind CBS individuals, is related to external visual stimulation (Painter et al., 2018). Another case study of an elderly individual with CBS and mild sensorineural deafness ascribed hallucinations to external auditory stimulations of the early visual cortex (Vacchiano et al., 2019). Given the difficulty in recruiting large samples of CBS participants (Teunisse et al., 1996; Plummer et al., 2007; Cox, 2014), and given the demographic and clinical differences in the recruited samples in the three studied, we cannot exclude the possibility of a heterogeneous mechanism for hallucinations, combining both internal and external stimulation of the hyperexcitable early visual cortex. However, in our sample of participants, the observation of a slow, anticipatory build-up of activity in the visual cortex prior to hallucination onset cannot be accounted for by external stimulation. Our study clearly demonstrates that hallucinations in CBS can be evoked in the absence of external visual information.

As such, our findings provide unique insight into the possibility that spontaneous brain activity can evoke conscious percepts by triggering existing neural cascades. This observation is compatible with previous hypotheses regarding the role of the slow spontaneous activity in initiating internally-generated behaviour (Schurger et al., 2013; Moutard et al., 2015), which could be directly tested here due to the unique characteristics of CBS. Taken together, these findings raise the intriguing possibility that spontaneous network dynamics may underlie many types of internally-generated behaviours. For example, it is tempting to conjecture that if accumulation of spontaneous activity in the visual system of blind individuals can ignite visual hallucinations, a similar mechanism may also be implicated in more typical forms of unprompted and deprivation-related percepts, such as in dreaming.


Since CBS is rare, our sample size was small and heterogeneous in both the phenomenology of hallucinations, and in neural responses to visual hallucinations. Some of the small reported effects may therefore relate to a reduction in statistical power. It is worth noting, however, that despite the heterogeneity, all three participants evinced the same activation profile with the slow build-up of activation in early visual cortex. Additionally, we acknowledge that it is impossible to simulate accurately the content/timing of hallucinations because (i) our only access to the visual experiences of the CBS participants was the participants' subjective reports; (ii) the realistic 3D attributes of the hallucinations could not be simulated when presented on a computer screen; and (iii) our simulated stimuli were much smaller and more confined to the fovea compared to the real hallucinatory percepts, as we needed to provide enough movement space for stimuli on the display screen. Nevertheless, we made every effort to capture the gist and the reported temporal dynamics of the hallucinations.