A romantic caress is a primary expression of affiliative behavior. It reflects the disposition of individuals to seek close contact between them, and it promotes socio-emotional relationships, pair bonding and reproduction (Dunbar, 2010; Gallace and Spence, 2010; Morrison et al., 2010). Despite the interactive character of tactile interactions among humans, they are addressed almost exclusively as a receptive experience by psychological and neuroscientific investigations. Moreover, active romantic caress essentially has a socio-emotional intention through somatosensory interaction with another individual. Hence, its anticipatory neural processes already could be uniquely modulated according to sensory-motor, emotional and social factors. Indeed, during action, predictions are made by the brain about its consequences in order to motivate initiation, anticipate effects and optimize performance (Wolpert et al., 1995; Blakemore et al., 1998; Knoblich and Flach, 2001; Schutz-Bosbach and Prinz, 2007; Haggard, 2008). Elucidating the anticipatory neural mechanisms underlying active romantic caress will add elementary information to the understanding of how social interaction is driven by brain function, motivating and coordinating behavior.

In addition to its sensory-motor component, romantic caress is inherently associated with an affective component (Hertenstein et al., 2006; Gallace and Spence, 2010; Morrison et al., 2010). Regarding the anticipation of passive tactile experiences, sensory and affective consequences are coded by brain circuits related to sensory-motor and affective processes (Porro et al., 2002; Lovero et al., 2009; Gazzola et al., 2012; Morrison et al., 2013). Behavioral evidence suggests that the prediction of outcomes of performed actions also comprises an affective component that modulates the perception of action consequences (Wilke et al., 2012). Furthermore, touch is a primary reinforcer strongly connected with motivational and affective processes that may drive social behavior (Rolls, 2010), and some studies demonstrated that rewards influenced the processing of tactile stimuli (Pleger et al., 2008, 2009). Thus, active romantic caress could be anticipated by neural processes associated with emotions, motivating social behavior and modulating effects of touch performance. However, romantic caress is a rather intimate domain where its affective evaluation is subjective and depends on many factors (Fisher et al., 1976; DiBiase and Gunnoe, 2004). A powerful factor influencing this affective component is the relationship between the interacting individuals. For example, Gazzola et al. (2012) showed that, when manipulating the perceived affective quality (pleasantness) of a passively experienced caress, primary somatosensory cortex (SI) encoded, and was modulated by, the perceived gender of the caresser. Likewise, touching a person for whom one has strong feelings of passionate love could be associated with hedonic experiences and motivated by a strong desire for physical contact (Hatfield and Sprecher, 1986; Gallace and Spence, 2010).

Whereas previous studies started to elucidate the neural basis of affective, social touch experiences, it remains poorly understood whether and how anticipatory neural processes of active romantic caress performance vary as a function of the desire for affective contact with the other. Co-variance of anticipatory neural processes with passionate feelings for the other could indicate the anticipation of hedonic experiences and prediction of reward due to tactile interactions. The present study aims at investigating how the anticipatory neural processes of an active romantic caress are modulated by the desire for affective contact as reflected by the intensity of feelings of passionate love for the other. An additional issue concerns the question whether anticipatory neural activity patterns associated with desire are independent of perceptual affective qualities, like pleasantness. Although the desire for union with another likely is related to pleasantness, it also could comprise more abstract representations of love and affection, partly associated with distinct neural substrates (Fisher et al., 2002; Cacioppo et al., 2012). Alternatively, anticipatory processes associated with the desire for romantic caress could correspond to those for anticipating the perceived affective qualities of touch.

For this purpose, functional magnetic resonance imaging (fMRI) scanning was performed in romantically involved partners using a paradigm that allowed to isolate the specific anticipatory representations of romantic caress, compared with a control caress. Main statistical analysis tested for the relationship between neural activity anticipating active romantic caress, compared to control caress, and measures of feelings of passionate love for the other. It was hypothesized that brain regions known to be associated with the affective valence of experienced romantic caress as well as reward processing and prediction also contribute to the anticipation of active romantic caress, subsequently modulating their response to actual experience.

A particular role is expected for posterior insula in relationship with feelings of desire for union with the other. Posterior insula is generally involved in processing somatosensory stimuli with affective or motivational significance, and is considered a central brain region for interoception, whereas, along with the integration of cognitive and emotional responses, anterior insula underpins the subjective awareness of bodily feeling states (Craig, 2002, 2009). Based on these functions, posterior insula has been proposed to generate predictions about bodily feeling states (Paulus and Stein, 2006) and to provide a crucial neural substrate for reward-processing (Paulus, 2007). Especially relevant in the context of pleasant skin-to-skin contact (Morrison et al., 2010), coding of pleasant romantic caress has been associated with posterior insula (Morrison et al., 2011) based on the C-tactile fiber system projecting to posterior insula (Olausson et al., 2002; Löken et al., 2009). Insular cortices also have been related to sexual desire and love with a posterior-to-anterior pattern, suggesting that desire and love are on a spectrum that evolves from integrative representations of affective visceral sensations (Cacioppo et al., 2012). Moreover, an fMRI study combined with selective serotonin reuptake inhibitors administration provided further evidence for anticipatory processing of emotional stimuli in posterior insula (Simmons et al., 2009). Thus, posterior insula could have a peculiar role in the anticipation of active tactile contact with others mediating between the desire for and the experience of performing a romantic caress.

The present study aimed at investigating the neural correlates of the desire for romantic caress by means of fMRI scanning. An fMRI-compatible experimental paradigm was applied in order to isolate the specific anticipatory representations of romantic caress (intending to caress another human being), compared with control caress (intending to caress an object). Specifically, we investigated the relationship between anticipatory neural activity and measures of feelings of passionate love for the other, that is, the intensity of the desire for union with the other. It was hypothesized that BOLD-response in anticipation of active romantic caress, compared with active control caress, in brain regions associated with the processing of affective and reward information of touch co-varied with the intensity of the desire for union with the other. Functional connectivity analysis was subsequently applied in order to gain more insight in the functional interactions in terms of long-range communication of brain regions that could be associated with the desire for romantic caress.

Summarizing the main results, whole-brain, voxel-wise analysis showed that BOLD-response in right posterior insula in anticipation of romantically caressing one’s partner, compared with an object, significantly co-varied with PLS scores, that is, the desire for union with the other. Furthermore, although pleasantness may be considered part of passionate love, this correlation between anticipatory activity in posterior insula and PLS score was independent from pleasantness ratings. Although posterior insula activity at the average group level did not significantly differ between the anticipation of romantic caress performance and the anticipation of control caress, additional ROI analysis indicated that posterior insula activity significantly differentiated between the actual performance of a romantic and a control caress, being increased for the former, compared with the latter. As evidenced by covariance analysis, the difference in posterior insula activity between the anticipation of romantic caress performance and control caress performance depended on how the participant felt for the other rather than on what was going to be touched per se. Connectivity analysis revealed that right posterior insula was embedded in a functional network including regions associated with reward processing and anticipation (e.g., Elliott et al., 2000; Schultz, 2000) and, consistent with previous functional connectivity studies, sensory-motor circuits (Taylor et al., 2009; Kurth et al., 2010; Deen et al., 2011; Ebisch et al., 2011; Chang et al., 2013). Moreover, functional connectivity between right posterior insula and right parahippocampal cortex positively co-varied with PLS scores.

The insula is considered a central brain structure for sensorimotor, visceral, interoceptive, homeostatic/allostatic functions, interacting with limbic, somatosensory and motor regions (Augustine, 1996; Critchley, 2005; Craig, 2009). Thalamo-cortical pathways provide a direct representation of homeostatic afferent information to posterior insula that engenders distinct bodily or interoceptive feelings by projections onto the anterior insula for subjective and conscious emotional evaluation (see also Saper, 2002). More specifically regarding the cutaneous senses, posterior insula could constitute the primary cortical locus of an interoceptive system regulating affective feelings states from the skin. Specific afferent pathways have been identified projecting threatening (Craig, 2002) or comforting (Olausson et al., 2002; Löken et al., 2009) cutaneous information via thalamic nuclei to posterior insula.

Beyond these merely perceptual functions, the insular cortex contributes to active behavior, too. For example, intracortical electrical microstimulation of the macaque insular cortex evoked a variety of emotional as well as non-emotional behaviors (Jezzini et al., 2012). In particular, stimulation of mid-posterior insula evoked affiliative and communicative behavior (Caruana et al., 2011). Based on processing of homeostatic emotions reflecting the physiological needs of the body (Craig, 2002), posterior insula also could accommodate affective/motivational functions driving behavior (Craig, 2003). For example, it has been argued that the insular cortex, integrating interoceptive information about the internal state of the body, also generates predictions about bodily feeling states (Paulus and Stein, 2006; Lovero et al., 2009) and provides a crucial neural substrate for reward-processing (Paulus, 2007). Regarding romantic relationships, insular cortex has been associated with sexual desire and love with a posterior-to-anterior pattern, implying that desire for pleasant physical contact with others could be grounded in representations of affective visceral sensations related to pleasant sensorimotor experiences and their anticipation (Cacioppo et al., 2012).

Finally, some previous studies associated posterior insula with anticipatory functions attenuating subsequent neural responses. Blakemore et al. (1999, 2000) showed decreased neural activity in S2/posterior insula for self-produced tactile stimuli on the palm of the hand, compared with externally produced tactile stimuli, suggesting attenuation of tactile perception by sensory prediction mechanisms (Blakemore et al., 1998, 1999; Carlsson et al., 2000; Hughes et al., 2013). Furthermore, Simmons et al. (2009) showed stronger deactivations in posterior insula in anticipation of either positive or negative emotional stimuli after administrating selective serotonin reuptake inhibitors to participants. These findings indicate a role of posterior insula in a bottom-up suppression of physiological arousal induced by emotional stimuli.

In accordance with this information, the present results confirm the hypothesis that posterior insula could accommodate anticipatory functions underlying affiliative behavior, and active romantic caress in particular. These findings add to the existing literature by providing further insight on how subjective factors, like the psychological relationship between individuals, mediate between predictive brain function and social interaction, that is, how anticipatory processes may shape social behavior and the experience of social interaction.

However, what would such anticipatory activation patterns in posterior insula functionally entail? It is already known that one of the consequences of such anticipatory processing is the modulation of neural responses to sensory input, either suppressing or amplifying experience (Hughes et al., 2013; van Ede et al., 2013). A closer look at the fMRI group statistical maps of the present study reveals that posterior insula was characterized by negative BOLD-responses in anticipation of touch performance, compared with baseline, whereas it was positively modulated during touch performance. Similar patterns could be observed in somatosensory cortices, especially left PostCS and right SI. Whereas a stronger deactivation (negative BOLD regulation) was detected in somatosensory cortices during the anticipation of active romantic caress, compared to control caress, during performance neural activity was stronger for romantic caress, compared to control caress in these regions. In addition, the results from co-variance analysis suggest that higher PLS scores of participants were accompanied by a weaker deactivation in posterior insula in anticipation of romantic caress, compared with the anticipation of control caress. Hence, based on our, and previous (Blakemore et al., 2000; Carlsson et al., 2000; Simmons et al., 2009; Schafer et al., 2012) results, we propose that suppression of physiological arousal in posterior insula in anticipation of active romantic caress could be modulated as a function of the desire to experience that touch; the stronger the desire, the weaker the suppression (and, in turn, the stronger the affective intensity of touch experience).

Some previous studies demonstrated a direct link between reward and somatosensation suggesting that reward anticipation modulates tactile processing (Pleger et al., 2008, 2009). One may hypothesize that also posterior insula contributes to such effects in the context of an affective, romantic caress. Interestingly, although left PostCS and right SI, like posterior insula, differentiated between a romantic and a control caress during performance, unlike posterior insula, left PostCS and right SI activity did not co-vary with PLS scores. This suggests that SI/PostCS activity patterns may be related to distinct aspects of romantic caress processing, possibly based on interoceptive and exteroceptive/proprioceptive functions generally attributed to posterior insula (Craig, 2002) and postcentral somatosensory cortices (Keysers et al., 2010), respectively. Further studies would be necessary to confirm this hypothesis more directly, and to investigate more in detail the relationship between anticipatory activation patterns and subsequent stimulus processing as well as the neurophysiological or psychological arousal induced by the stimulus, and possibly distinguishable roles of postcentral somatosensory cortices and posterior insula.

Finally, functional connectivity analysis, measured as temporally-correlated patterns brain activity across brain regions reflecting an index of brain long-range communication (Fox and Raichle, 2007; Van Dijk et al., 2010), suggests that other brain regions involved in reward anticipation (Elliott et al., 2000; Schultz, 2000) may support the role of posterior insula in anticipation of romantic caress, i.e., globus pallidus (GP), and its modulation by the desire for that touch, i.e., parahippocampal areas. Indeed, GP was functionally linked to feelings of love for another (Acevedo et al., 2012) and is anatomically connected with posterior insula via the ventral striatum (Chikama et al., 1997) also involved in associating tactile stimuli with reward (Pleger et al., 2008, 2009). Moreover, previous studies showed that parahippocampal repetition suppression is sensitive to reward-predicting properties of stimuli (Zweynert et al., 2011). Reward-predicting properties of stimuli have been also shown to enhance activity in the hippocampus, in addition to recollection of the same stimuli (Wittmann et al., 2005). Thus, it has been proposed that reward prediction might enhance hippocampus-dependent memory formation (Wittmann et al., 2005). Although it needs to be recognized that the present results do not provide direct evidence for involvement of reward circuitry, based on this knowledge it is possible to speculate that the modulation of posterior insula activity in anticipation of romantic caress by the desire for that touch might rely on reward-dependent memory of romantic caress experience.

Some other issues need to be mentioned. First, previous studies already investigated the neural correlates of pleasant and romantic caress experiences. How do they relate to or differ from the present study? Regarding the neural substrates of the emotional aspects of pleasant touch experiences, Rolls et al. (2003) showed that certain areas of the human orbitofrontal cortex and cingulate cortex are involved in representing pleasant touch. Other studies demonstrated coding of pleasant romantic caress in posterior insula (Morrison et al., 2011) based on the C-tactile fiber system projecting to posterior insula (Olausson et al., 2002; Löken et al., 2009). Additional experiments showed that the representation of pleasant romantic caress also involves right posterior superior temporal sulcus and the medial prefrontal cortex/dorsal anterior cingulate cortex (Gordon et al., 2013) and that anterior insula is involved in the anticipation of the sensory-affective impact of touch, predicting BOLD-responses during subsequent passive touch experiences in posterior insula (Lovero et al., 2009). Finally, Gazzola et al. (2012) reported activation in SI being associated with the affective processing of passively experienced romantic caress as well as its prediction. This neural activation also was modulated by the perceived gender of the other individual. Whereas we found significant modulations of BOLD-response in posterior insula, these where not detected in other brain regions associated with pleasant touch experiences. However, some relevant differences between studies can be noted.

Beside the difference with these previous studies in terms of active and passive touch, another important difference is that, here, we specifically focused on the psychological relationship between individuals (i.e., intensity of feelings of passion for the other), whereas previous studies investigated the affective perceptual qualities of the touch, like its pleasantness. Although pleasantness may be considered part of passionate feelings, results suggest that desire and pleasantness are not necessarily interdependent and reflect, at least partially, distinct phenomena. First, although there was a trend towards significance, we failed to detect a significant statistical dependency between them. Second, co-variance of posterior insula activity with PLS scores appeared independent of pleasantness ratings, suggesting that anticipatory activity for active romantic caress in posterior insula specifically depends on the intensity of longing for the other, but not on the sensory-affective qualities of the touch itself. In this sense, the present results suggest a mechanism for the desire for romantic caress that is distinct from or goes beyond the anticipation of the sensory-affective qualities of the touch.

However, it should be noted that previous studies also reported coding of the affective qualities of perceived romantic caress, like its pleasantness, in posterior insula (e.g., Morrison et al., 2011). A relevant difference with these studies is that they concerned touch on the back of the hand (C-tactile innervated skin) stimulating the C-tactile fiber system projecting to posterior insula (Olausson et al., 2002; Löken et al., 2009), whereas, here, participants caressed another individual with their fingers and palm of the hand (glabrous skin). Anticipation in posterior insula/parietal operculum based on sensory prediction mechanisms of tactile stimuli on the palm of the hand also has been reported (Blakemore et al., 1999, 2000). Future experiments would be needed to address this apparent discrepancy. Likely, different neuro-functional mechanisms within posterior insula are involved in processing desire and sensory-affective information of romantic caress, based on its interoceptive (Craig, 2003) and (exteroceptive) sensory-affective (Olausson et al., 2002) functions. A relevant detail supporting this explanation is that the association between anticipatory activity and PLS scores was detected in right posterior insula, that is, ipsilaterally to the touching hand. Moreover, differential activity was detected in the same cluster during the actual performance of a romantic caress, compared to a control caress. These findings further suggest that the detected association may not be directly related to the anticipation of basic sensory-affective experiences, more likely to involve contralateral posterior insula. Indeed, insula cortex is involved in manifold functions, including the processing of sensory, affective and bodily information about its physiological state and needs, at the basis of its integrative and social properties (Craig, 2009; Singer et al., 2009).

Second, the present study included romantically involved couples. It could be argued that a second group of participants touching unknown individuals might have provided relevant additional information. Extending the current results, in that case, a stronger deactivation in posterior insula in anticipation of a romantic caress could be expected, being the most arousing condition due to its social valence, considering that there possibly is less desire to experience that arousal in unrelated individuals. However, one also may argue that it would be difficult to isolate the desire component from other confounding factors when comparing groups, while it was not when studying variable degrees of desire within romantically involved participants. Nevertheless, we propose that it would be relevant for future work to compare different groups of interacting participants in order to investigate the influence of specific relationships between individuals.

In conclusion, the present results show that neural activity in posterior insula in anticipation of the performance of a romantic caress varies as a function of the desire for union with the other. The findings suggest that posterior insula, interacting with brain regions related by previous studies to sensory-motor and reward functions, could provide a neural substrate mediating between the desire for tactile contact with others and active romantic caress. In particular, anticipatory activity patterns in posterior insula possibly modulate subsequent affective processing of skin-to-skin contact.

Sjoerd J. Ebisch, Francesca Ferri and Vittorio Gallese designed research; Sjoerd J. Ebisch and Francesca Ferri performed research; Sjoerd J. Ebisch analyzed data; Sjoerd J. Ebisch, Francesca Ferri and Vittorio Gallese wrote the paper.

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.