This preliminary study was aimed at testing the theoretical assumption that a virtual experience elicits emotional responses comparable to those produced by real exposure. In addition, we also assumed that the sense of presence induced by the VR immersion makes the virtual experience more realistic, and consequently more effective than static pictures, in producing emotional responses in humans. In accordance with the first hypothesis, our data show that virtual food is as effective as real food, and more effective than photographs of food, in producing psychological and physiological responses in patients with ED, suggesting a possible advantage of using virtual stimuli instead of static pictures as an alternative to real stimuli to induce emotional reactions in subjects. This finding appears to be not specifically related to the diagnosis (AN or BN), as suggested by the fact that there were no significant differences in the emotional response recorded between the two groups of patients. Not even the severity of illness seems to influence the patients' reactions, as subjects with a mild, moderate or severe eating disorder did not significantly differ in their emotional responses to real or virtual food. However, we did not find any significant variation in the controls' emotional reactions in any of the experimental conditions. This is not surprising because, as happens in real life, food does not represent a stressful stimulus for healthy people.
Regarding our second hypothesis, we found an effect of subjects' degree of presence experienced in the VR condition on their level of perceived anxiety (STAI-S and VAS-A): the higher the sense of presence, the higher the level of anxiety. The sense of presence in virtual reality is defined as 'the participant's sense of "being there" in the virtual environment' [4
] and it is obtained through two factors: immersion and interaction. Immersion is provided by the use of technological devices such as HMDs that permit a 3 D experience, while interaction is the possibility given to the users to interact in real time with the virtual environment. The higher the sense of presence, the more realistic the virtual experience, and more intense the emotional involvement. Immersion and interaction are the key distinctive factors that make the difference between the VR and the PH conditions. In the latter, subjects can only passively observe static pictures, while in the VR condition they can actively explore the environment, approach the food and virtually touch it, as they would do in real-life situations. We argue that the effectiveness of virtual and real stimulations is the reason why both psychological (STAI-S, and VAS-A) and physiological (HR and SCR) responses appear to be consistently higher in the RF and VR than in the PH condition. Thus, this result showing a similar pattern of psychological and physiological responses is rather new considering that, to date, there have been many studies that separately investigated psychological or physiological responses during VR exposure, but only few assessing the effects of stressors presented in a virtual environment on the subjective and objective response of anxiety [28
]. Regarding the general lack of significant variations on respiration, we hypothesize that it may be due to the fact that only respiration rate was assessed and not tidal volume, and anxiety mainly affects tidal volume rather than rate [30
To date, despite the large amount of data demonstrating the efficacy of VR-based approaches for the treatment of different psychological disorders [2
], none of the previous work had directly investigated if the exposure to virtual stimuli is able to elicit emotional reactions similar to those elicited by real-life exposure, which is the added value of using VR instead of simple static pictures. Even though it was accomplished on only two small samples of ED patients, these preliminary data encourage the use of VR in clinical (exposure therapy) and even non-clinical (task learning) settings in which a highly customizable and controllable stimulation is preferred to a real-life one. Additionally, our data emphasize the role of presence in the emotional processes, proving that, even if definitively more expensive, VR is preferable to static images for generating affective responses in humans. So, in accord with the previous studies [17
], the present research adds some evidence that virtual stimuli can be used instead of the real ones to elicit patients' emotions.
Despite the clearness of the present findings, this study has some important limitations. First, the small number of subjects per group makes us cautious about the generalization of the results. A future randomized controlled study including a larger sample will address this issue. Second, in the VR condition subjects were exposed to virtual food in a virtual restaurant, while in the other two conditions they were exposed to food only. A restaurant is a broader stimulus than food because it elicits a complex context possibly inducing a greater level of anxiety than food alone, and also other fears, not strictly or necessarily related to food (for example, agoraphobia). In order to control this aspect in future studies, virtual food could be presented in neutral virtual environments not specifically related to eating contexts. Thus, even if considered a limitation in the present study, the possibility to measure subjects' reactions in a complex virtual environment is a great advantage offered by virtual reality, with poor feasibility for testing the subjects' responses in a real complex environment such as a restaurant.