Previous neuroimaging evidence has consistently shown the role of the insular cortex in conflict monitoring and decision-making (Canessa et al., 2013, 2017; Markett et al., 2016). Several studies have reported the involvement of this region in a wide array of functions, placing it at the interface between interoception, emotion and cognition (Chang et al., 2013; Uddin et al., 2017; Molnar-Szakacs and Uddin, 2022). In particular, insular activity related to interoceptive representation, performance monitoring, as well as awareness of emotional states and body movement, has been interpreted in terms of bodily physiological signals guiding decision-making based on the outcome of behavioral learning processes (Craig, 2009). This hypothesis was supported by studies investigating the neural bases of decision-making under risk and/or uncertainty, showing increased activation of the right insular cortex when making riskier compared to safer choices (Paulus et al., 2003) and higher anticipatory insular response preceding riskless choices (Kuhnen and Knutson, 2005).

Based on anatomical, functional and connectivity features, the insula has been parceled in multiple subregions, and particularly in posterior and anterior sections divided by the central insular sulcus (Wysiadecki et al., 2018). The anterior section, highly connected with frontal and limbic areas, has been mainly related to emotional and cognitive functions, whereas the posterior insula (PI) plays a crucial role in interoceptive processing and behavioral regulation via connections to posterior-temporal and parietal areas, as well as sensorimotor cortices (Cloutman et al., 2012; Centanni et al., 2021). These sectors appear to underpin a posterior-to-anterior flow of neural signals, whereby the progressive integration of sensory information into mechanisms of “salience” detection underpins the modulation of cognitive control and goal-directed behavior by sensory-interoceptive and affective stimuli (Craig, 2009; Centanni et al., 2021). Accordingly, neuroimaging studies have shown the PI role in representing homeostatic states related to the experience of risk, in turn shaping upcoming decisions (Xue et al., 2010). One such experience concerns the mental anticipation and evaluation of choices entailing both positive and negative prospective outcomes, typically associated with the overweighting of potential losses compared with gains, and thus with the individual’s tendency to prefer avoiding losses to acquiring equivalent gains (i.e., loss aversion; Kahneman and Tversky, 1979). Previous neuroimaging evidence shows that individual differences in the extent of loss aversion reflect the strength of activation of multiple brain structures, including the striatal and dorsal anterior cingulate (dACC) sectors of the meso-cortico-limbic pathway (Tom et al., 2007), and the right PI—extending into the supramarginal gyrus—both when making real choices (Canessa et al., 2013) and at rest (Canessa et al., 2017).

Consistent with the role of the insula in loss aversion, clinical evidence shows that operculo-insular resection for drug-resistant epileptic seizures impairs patients’ sensitivity to expected value when facing a potential loss during decision-making under risk (Von Siebenthal et al., 2017). Moreover, insular damage has been associated to a reduction of the typical gambling-related cognitive distortions in tasks simulating real-life gambling (Clark et al., 2014): compared both to controls and patients with lesions in the amygdala or ventromedial prefrontal cortex (vmPFC), those with lesions confined to insula did not show the so-called near-miss effects (i.e., higher motivation to play a gamble after non-wins close to the jackpot) or gambler’s fallacy (i.e., the belief that non-recent outcomes are more likely to happen).

By showing a role of the insula in modulating adaptive behavioral learning via sensory-affective signals, these data suggest its possible involvement in conditions characterized by poor adaptation to changing environments. This view is supported by growing evidence of structural and functional abnormalities, in this region, in patients with major depressive disorder (Herwig et al., 2010; Surguladze et al., 2010; Takahashi et al., 2010; Stratmann et al., 2014), bipolar disorder (Ellison-Wright and Bullmore, 2010; Hulvershorn et al., 2012; Wise et al., 2017), anxiety disorders (Stein et al., 2007; Terasawa et al., 2013), schizophrenia (Wylie and Tregellas, 2010), psychopathy for abnormal socio-emotional processing (Blair, 2013), anorexia nervosa (Kaye et al., 2009), addiction (Ibrahim et al., 2019), as well as different deficits associated with neurological disorders (Löffler et al., 2016; Namkung et al., 2017). Overall, this evidence highlights the importance of assessing insular responsiveness to tasks tapping evaluation and control as a potential translational biomarkers (or “endophenotypes”; Sharp et al., 2012) of adaptive behavioral learning, that might also represent the target of innovative neuromodulation protocols with therapeutical purposes (Downar et al., 2016).

Notably, several studies indeed showed strong functional connectivity between the PI and the dACC (Taylor et al., 2009; Ghaziri et al., 2017), i.e., a key node of the “executive” network (Shenhav et al., 2016), which highlights a possible neural basis of the insula role in cognitive control via salience processing (Dosenbach et al., 2007; Seeley et al., 2007). This notion highlights the potential translational implications of investigating the role of insula and dACC, since both regions map strikingly well with a common core of brain areas involved in different psychiatric conditions (Goodkind et al., 2015; Downar et al., 2016), that might be targeted in innovative treatment protocols. Since this aim can be pursued with non-invasive neurostimulation and neuromodulation techniques, studies investigating their feasibility are required to pave the way to possible clinical applications (Downar et al., 2016). Transcranial direct current stimulation (tDCS) is a relatively inexpensive, portable and well tolerated candidate for this purpose (Brunoni et al., 2012), especially with high-definition montages (HD-tDCS), that are expected to improve the intensity and/or focality of stimulation compared with the “conventional” bipolar approach (Edwards et al., 2013; Huang et al., 2019). It is worth noting that optimization methods for HD-tDCS highlighted a non-trivial trade-off between focality and variability of the effects with different montages (Dmochowski et al., 2011; Mikkonen et al., 2020; Lee et al., 2021). The available evidence from computational models suggests that 4 × 1 ring montages (i.e., one small stimulating electrode surrounded by a ring of four small reference electrodes) grant higher focality, but at the cost of larger inter-individual variability in strength and distribution of electric fields (Mikkonen et al., 2020; Van Hoornweder et al., 2022). On the other hand, model-based HD-tDCS montages produce larger current dispersion, which is however compensated by higher stimulation intensity at the target site (Lee et al., 2021), resulting in an overall optimal trade-off between focality, individual variability, and intensity (Mikkonen et al., 2020).

By using a model-based approach, we previously reported evidence of the significant effect of dACC stimulation, via HD-tDCS, on cognitive control and loss aversion in decision-making under risk (Mattavelli et al., 2022). It is still unknown, instead, whether such modulation might be also elicited by HD-tDCS of the insula. Since this area is a potential key target for the treatment of different neuropsychiatric conditions (Downar et al., 2016), this is a critical gap that should be filled to investigate its feasibility and its effects.

On this basis, we assessed the effects of anodal and cathodal HD-tDCS over the right PI in modulating executive control on the Flanker task, and both loss and risk aversion in well-established tasks of decision-making under risk (Kahneman and Tversky, 1979). We replicated the experimental paradigms and methodological procedures used in our previous study (Mattavelli et al., 2022), in which decreased “Flanker” conflict effect and increased loss/risk aversion suggested an improvement of conflict monitoring—involving both visual-attentional skills and behavioral control—after cathodal HD-tDCS of the dACC. As mentioned above, the dACC and PI are highly connected, and the activity of both regions has been reported to track individual differences in response inhibition (Wager et al., 2005; Fedota et al., 2016) and loss aversion (Canessa et al., 2013). We therefore predicted that active HD-tDCS of the right PI, compared to sham stimulation, might confirm the modulation of behavioral performance in the Flanker and gambling tasks observed in our previous study.

We aimed to assess the effect of HD-tDCS of the right PI on different facets of cognitive control such as the conflict effect on the Flanker task, and both loss and risk aversion in a decision-making task based on mixed gambles, in healthy individuals.

To the best of our knowledge, this is the first study using a modeling/targeting approach to modulate insular activity with HD-tDCS, that allows to maximize stimulation intensity at the target site, within safety parameters and with good tolerability (Huang et al., 2019). To this purpose, we replicated the experimental procedure adopted in a previous study that proved successful in modulating performance in the same tasks following HD-tDCS of the dACC (Mattavelli et al., 2022). There are both scientific and translational implications of investigating the feasibility and the effects of modulating PI activity: (i) strong connectivity between insula and dACC (Medford and Critchley, 2010; Ghaziri et al., 2017); (ii) previous evidence of an asymmetric anticipatory response of activation for losses and deactivation for gains reflecting behavioral loss aversion (i.e., “neural loss aversion”; Tom et al., 2007; Canessa et al., 2013, 2017); (iii) the translational relevance of targeting the insula, which, alongside the dACC, has been highlighted as a common neural substrate for several neuropsychiatric conditions (Downar et al., 2016).

While cathodal stimulation of the dACC increased behavioral control both in the Flanker and gambling tasks (Mattavelli et al., 2022), no such effect was found, in the present study, when targeting the right PI. Bayesian statistics rather provided moderate evidence in favor of the null model for the Flanker conflict effect and for both loss and risk aversion, along with anecdotal evidence for the null model concerning the “choice consistency” index. Overall, these results are suggestive of no significant modulation of performance in these tasks by HD-tDCS of the right PI.

As mentioned in the Introduction, the modeling/targeting procedure that we applied is expected to maximize the intensity of the stimulation on the target (Huang et al., 2016) and to decrease the individual heterogeneity of tDCS effects (Mikkonen et al., 2020; Lee et al., 2021), although the electric field modeling showed some current dispersion in temporo-parietal regions (Figure 1). This represents a limitation of the model-based HD-tDCS approach, since, in absence of direct measures of the physiological effect of stimulation, we cannot rule out the interpretation that results might reflect a suboptimal modulation of the insula, or counterproductive behavioral effects of electric fields affecting the cortical regions around the target. Our previous findings on the significant effect of dACC stimulation (Mattavelli et al., 2022) however support the effectiveness of targeting regions located in the depth of sulci, and thus suggest that the null present findings might rather indicate that the right PI—despite its strong connectivity with dACC (Ghaziri et al., 2017)—is not an effective target for modulating cognitive control or decision-making. However, this claim needs further investigation to be confirmed, using neurophysiological methods or combined neurostimulation and neuroimaging techniques (e.g., Meinzer et al., 2012; Pisoni et al., 2018; Esmaeilpour et al., 2020; Vergallito et al., 2023), which would allow disentangling the hypotheses that such null findings reflect the lack of impact on these tasks despite an effective insular modulation, or rather the ineffective modulation of the cortical target.

In the case of the Flanker conflict effect, another potential target is the anterior insula, where previous neuroimaging studies identified, both in normal (Trautwein et al., 2016) and pathological (Galandra et al., 2018, 2019, 2021; Crespi et al., 2019; Canessa et al., 2021, 2022) conditions, subregions involved in attentional control and salience processing that might act as a gatekeeper in executive control (see Molnar-Szakacs and Uddin, 2022). Based also on the high connectivity between posterior and anterior insula (Craig, 2009), the latter might be targeted by further studies to investigate a modulation of performance in tasks tapping executive control, including the Flanker conflict effect.

The lack of a significant modulation of participants’ loss and risk aversion is instead more unexpected, based both on our previous evidence of “neural loss aversion” in the right PI (Canessa et al., 2013, 2017), and on other neuroimaging (Xue et al., 2010; Purcell et al., 2021) and lesional (Clark et al., 2014) data showing its involvement in risky decision making. One possible interpretation for this null result concerns the use of a single-session stimulation, that has been questioned by meta-analytic evidence, particularly for cognitive tasks (Horvath et al., 2015; Medina and Cason, 2017). Single sessions are indeed more vulnerable to the unwanted impact of individual differences in susceptibility to stimulation and brain plasticity (Berryhill and Martin, 2018; Vergallito et al., 2022). Since multiple-sessions studies are more robust in this respect (Elmasry et al., 2015; Berryhill, 2017; see Berryhill and Martin, 2018), future studies should investigate whether repeated sessions of HD-tDCS stimulation over the insula allow overcoming individual variability in responsiveness to induce significant modulatory effects. This is not only relevant, but also more common than single-session protocols, in clinical contexts, which might benefit from usual multi-session protocols to improve symptoms in conditions associated with altered insular activity (Downar et al., 2016; Ibrahim et al., 2019). Other types of electric stimulation protocols, such as transcranial alternating current stimulation (tACS) and transcranial random noise stimulation (tRNS), have been reported as effective in modulating cortical activity, but, to the best of our knowledge, there are no studies applying these protocols to target the insula. In particular, theta-tACS seems promising in modulating executive functions (Klink et al., 2020), while there is limited evidence on the modulation of higher cognitive functions by tRNS, that has been mainly tested on motor and perceptual tasks (van der Groen et al., 2022; Brancucci et al., 2023). Thus, both the outcomes of tACS and tRNS on executive control and decision-making, and their effectiveness in targeting the insula, represent open research lines.

On the other hand, the importance of addressing the technical feasibility of insula stimulation is supported by recent multifaceted evidence of its effect on interoceptive accuracy (Sagliano et al., 2019) and compassion motivation when attending to another’s pain (Di Bello et al., 2023). Both these studies used a bipolar montage with electrodes placed over the frontotemporal region, that—as shown by computational modeling (see Figure 1)—induced larger electric fields over the frontal cortex compared to our high-density montage. This finding highlights the non-trivial trade-off between the effects of maximizing the intensity of stimulation via the targeting procedure vs. spreading electric fields in wide portions of the cortex at the cost of further decreased focality when bipolar montages are applied. Other studies targeted the insula with alternative neurostimulation techniques, such as transcranial magnetic stimulation (TMS) with H-coils, that, while generally associated with some discomfort (Levkovitz et al., 2009; Kaster et al., 2018), is considered a safe procedure allowing to reach deep brain regions (Zangen et al., 2005; Roth et al., 2007). Repetitive TMS with H-coil over the insula has been reported to reduce dopamine levels in substantia nigra and striatum (Malik et al., 2018), and to affect functional connectivity with the medial prefrontal cortex (Lee et al., 2020) and the default mode network (Moeller et al., 2022). Another study targeting the insula with H-coils in a sample of patients with anorexia nervosa reported a significant reduction of symptoms related to obsessions and compulsions, as well as depression and anxiety scores (Knyahnytska et al., 2019). In contrast, a study applying a single session in healthy volunteers reported the absence of significant modulations on blink-suppression and risk behavior (Spagnolo et al., 2019). Overall, these findings highlight inconsistent data on the effects of insula stimulation, which should therefore be interpreted with caution because of methodological variability across studies.

In conclusion, we assessed for the first time the modulatory effect of HD-tDCS applied on the right PI based on a modeling/targeting procedure. Apart from the target region, we replicated a methodological procedure that proved effective in decreasing the Flanker conflict effect and increasing both loss and risk aversion after cathodal stimulation of the dACC (Mattavelli et al., 2022). Bayesian statistical analyses provided moderate support for the absence of modulatory effects on executive control and decision-making variables following anodal or cathodal stimulation of the right PI. While confirming the technical feasibility of stimulating this region with a HD-tDCS approach grounded in targeting and modeling procedures, this negative finding suggests that further research is needed to unveil the factors shaping its actual effects and the tasks that are best suited to measure them, to increase the efficacy of translational applications.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by Ethics Committee of ICS Maugeri—Pavia Institute. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

GM and NC conceptualized and designed the study. IG and GM collected the data, performed the statistical analysis, and wrote the first draft of the manuscript. IG organized the database. NC wrote sections of the manuscript. All authors contributed to the article and approved the submitted version.

This research was partially supported by the “Ricerca Corrente” funding scheme of the Italian Ministry of Health.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.