bioRxiv preprint doi: https://doi.org/10.1101/2020.02.06.937151; this version posted August 28, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license. 1 Visual mental imagery engages the left fusiform gyrus, but not the early visual cortex: a 2 meta-analysis of neuroimaging evidence 3 Alfredo Spagna1,2, Dounia Hajhajate2, Jianghao Liu2,3, Paolo Bartolomeo2,* 4 1 Department of Psychology, Columbia University in the City of New York, NY, USA, 10027 5 2 Sorbonne Université, Inserm U 1127, CNRS UMR 7225, Paris Brain Institute, ICM, Hôpital de 6 la Pitié-Salpêtrière, F-75013 Paris, France 7 3 Dassault Systèmes, Vélizy-Villacoublay, France 8 9 * Corresponding author 10 Paolo Bartolomeo, M.D., Ph. D. 11 Inserm U 1127, CNRS UMR 7225, Sorbonne Université, Institut du Cerveau et de la Moelle 12 épinière, ICM, Hôpital de la Pitié-Salpêtrière, 75013 Paris, France 13 [email protected] 14 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.06.937151; this version posted August 28, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license. 2 1 Abstract 2 The dominant neural model of visual mental imagery (VMI) stipulates that memories from the 3 medial temporal lobe acquire sensory features in early visual areas. However, neurological 4 patients with damage restricted to the occipital cortex typically show perfectly vivid VMI, while 5 more anterior damages extending into the temporal lobe, especially in the left hemisphere, often 6 cause VMI impairments. Here we present two major results reconciling neuroimaging findings in 7 neurotypical subjects with the performance of brain-damaged patients: (1) a large-scale meta- 8 analysis of 46 fMRI studies, which revealed that VMI engages fronto-parietal networks and a 9 well-delimited region in the left fusiform gyrus. (2) A Bayesian analysis showing no evidence 10 for imagery-related activity in early visual cortices. We propose a revised neural model of VMI 11 that draws inspiration from recent cytoarchitectonic and lesion studies, whereby fronto-parietal 12 networks initiate, modulate, and maintain activity in a core temporal network centered on the 13 fusiform imagery node, a high-level visual region in the left fusiform gyrus. 14 15 Keywords: fronto-parietal networks; attention; working memory; fusiform gyrus; temporal lobe. 16 bioRxiv preprint doi: https://doi.org/10.1101/2020.02.06.937151; this version posted August 28, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license. 3 1 Introduction 2 Close your eyes and think of Leonardo da Vinci’s Monna Lisa. Is she looking at you or not? Is 3 her hair curled or straight? Visual Mental Imagery (VMI) is the set of abilities whereby we can 4 “see” things that are elsewhere (or nowhere: now imagine Monna Lisa frowning at you). These 5 capacities are important for predicting the outcome of everyday tasks (Moulton & Kosslyn, 6 2009), for example to decide whether our car fits in a narrow parking spot. The subjective 7 vividness of visual mental images varies substantially across individuals (Cui, Jeter, Yang, 8 Montague, & Eagleman, 2007; Dijkstra, Bosch, & van Gerven, 2017; Galton, 1880; J. Pearson, 9 Naselaris, Holmes, & Kosslyn, 2015), with some individuals experiencing mental images as 10 “quasi-visual” in nature, while others having less vivid images, down to the total absence of VMI 11 experience in otherwise normal individuals, a condition dubbed as “aphantasia” (de Vito & 12 Bartolomeo, 2016; Fulford et al., 2018; Jacobs, Schwarzkopf, & Silvanto, 2018; Zeman, Dewar, 13 & Della Sala, 2015). The neural bases of this remarkable set of cognitive functions are the object 14 of intense research efforts (Dentico et al., 2014; Dijkstra, Mostert, Lange, Bosch, & van Gerven, 15 2018; Dijkstra, Zeidman, Ondobaka, van Gerven, & Friston, 2017; Ishai, Ungerleider, & Haxby, 16 2000; D. G. Pearson, Deeprose, Wallace-Hadrill, Burnett Heyes, & Holmes, 2013; J. Pearson et 17 al., 2015; Winlove et al., 2018). Identifying the brain circuits supporting VMI and motor imagery 18 is also essential for clinical reasons, because detecting their activity in neuroimaging can reveal 19 consciousness in non-communicating patients in apparent vegetative state (Owen et al., 2006); in 20 addition, uncontrolled VMI activity could contribute to the vivid recollections of traumatic 21 memories resulting in post-traumatic stress disorder (Mary et al., 2020). 22 The dominant model of VMI (Dijkstra, Bosch, & van Gerven, 2019; J. Pearson et al., 23 2015) stipulates the existence of common neural substrates underlying VMI and visual 24 perception, spanning across the ventral cortical visual stream, with a crucial implication of early 25 visual areas, providing the sensory and spatial representational content of VMI (Kosslyn, bioRxiv preprint doi: https://doi.org/10.1101/2020.02.06.937151; this version posted August 28, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license. 4 1 Thompson, & Ganis, 2006; J. Pearson, 2019, 2020). Neuroimaging studies in healthy volunteers 2 supported a strong version of the model, by demonstrating the engagement of early, occipital 3 visual areas in VMI (Ibáñez-Marcelo, Campioni, Phinyomark, Petri, & Santarcangelo, 2019; 4 Kosslyn, Ganis, & Thompson, 2001). Further, TMS interference on V1 was shown to impact 5 VMI (Kosslyn et al., 1999). The model also provided a principled account of inter-individual 6 differences in VMI, because the level of activation in low-level visual areas correlates with the 7 subjective experience of VMI “vividness” (Dijkstra, Bosch, et al., 2017; but see Fulford et al., 8 2018; Lee, Kravitz, & Baker, 2012). However, this neuroimaging evidence is correlational, and 9 does not directly speak to the causal role of these structures in VMI. In fact, causal evidence 10 from neurological patients is sharply discordant with an implication of early visual areas in VMI 11 (Bartolomeo, 2002, 2008; Bartolomeo, Bourgeois, Bourlon, & Migliaccio, 2013; Bartolomeo, 12 Hajhajate, Liu, & Spagna, 2020). Whereas the model would predict a systematic co-occurrence 13 of perceptual and imaginal deficits after brain damage (Farah, Levine, & Calvanio, 1988), 14 patients with brain damage restricted to the occipital cortex often have spared VMI abilities 15 (Aglioti, Bricolo, Cantagallo, & Berlucchi, 1999; Behrmann, Winocur, & Moscovitch, 1992; 16 Chatterjee & Southwood, 1995; Policardi et al., 1996), with preserved subjective VMI vividness 17 despite damaged early visual cortex (Goldenberg, Mullbacher, & Nowak, 1995), even in the case 18 of bilateral cortical blindness (Chatterjee & Southwood, 1995; de Gelder, Tamietto, Pegna, & 19 Van den Stock, 2015; Zago et al., 2010). Instead, deficits of VMI for object form, object color, 20 faces or orthographic material typically arise as a consequence of more anterior damage (Basso, 21 Bisiach, & Luzzatti, 1980; Beschin, Cocchini, Della Sala, & Logie, 1997; Bisiach & Luzzatti, 22 1978; Bisiach, Luzzatti, & Perani, 1979; Farah et al., 1988; Manning, 2000; Riddoch, 1990; 23 Sirigu & Duhamel, 2001), often extensively including the temporal lobes, especially in the left 24 hemisphere (Bartolomeo, 2002, 2008; Moro, Berlucchi, Lerch, Tomaiuolo, & Aglioti, 2008). 25 Such a strong evidence of dissociation is at odds with models proposing a crucial implication of bioRxiv preprint doi: https://doi.org/10.1101/2020.02.06.937151; this version posted August 28, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license. 5 1 early visual areas in VMI and suggests an engagement in VMI of higher-order associative areas, 2 especially in the left temporal lobe, rather than early visual cortex. Also, growing evidence 3 shows that the temporal dynamics of VMI and perception are different (Daselaar, Porat, 4 Huijbers, & Pennartz, 2010; Dijkstra et al., 2018; Ganis, Thompson, & Kosslyn, 2004; 5 Yomogida et al., 2004). VMI and perception build on distinct brain connectivity patterns 6 (Dijkstra, Zeidman, et al., 2017), characterized by a reversed cortical information flow (Dentico 7 et al., 2014), and supported by anatomical connectivity between striate, extrastriate, and parietal 8 areas (Whittingstall, Bernier, Houde, Fortin, & Descoteaux, 2014). In addition, studies on VMI 9 mainly focused on the ventral cortical visual stream, but evidence also indicates VMI-related 10 increases in BOLD response in fronto-parietal networks (Mazard, Laou, Joliot, & Mellet, 2005; 11 Mechelli, Price, Friston, & Ishai, 2004; Yomogida et al., 2004). Thus, what VMI shares with 12 visual perception may not be the passive, low-level perceptual aspect, but the active, exploratory 13 aspects of vision (Bartolomeo, 2002; Bartolomeo et al., 2013; Thomas, 1999), such as those 14 sustained by visual attention (Bartolomeo & Seidel Malkinson, 2019). Yet, the precise identity of 15 these mechanisms, and of the underlying brain networks, remains unknown. 16 To address these issues, we conducted a meta-analysis of functional magnetic resonance 17 imaging studies that examined the neural correlates associated with visual and motor mental 18 imagery. Our specific aims were to assess the role of low- and high-level visual cortex, as well as 19 the role of the fronto-parietal networks in mental imagery, by identifying the brain regions with 20 higher activation for mental imagery in healthy volunteers.