Subliminal determinants of cue-guided choice

By anticipating potential rewards, external cues can guide behavior to achieve a goal. Whether the conscious elaboration of these cues is necessary to elicit cue-guided choices is still unknown. The goal of the present study is to test whether the subliminal presentation of a visual cue previously paired with a reward is sufficient to bias responses that can lead to the same or a similar reward. To this aim, three experiments compared the subliminal and supraliminal presentation of reward-associated cues during a Pavlovian-to-Instrumental Transfer task. In line with previous evidence, results showed that the supraliminal presentation of reward-associated Pavlovian cues biased participant’s choice towards motivationally similar rewards (general transfer) as well as towards rewards sharing the precise sensory-specific properties of the cue (outcome-specific transfer). In striking contrast, subliminal cues biased choice only towards motivationally similar rewards (general transfer). Taken together, these findings suggest that cue-guided choices are modulated by the level of perceptual threshold (i.e., subliminal vs supraliminal) of reward-associated cues. Although conscious elaboration of the cue is necessary to guide choice towards a specific reward, subliminal processing is still sufficient to push towards choices sharing the motivational properties of the cue. Implications for everyday life, clinical conditions, and theoretical accounts of cue-guided choices are discussed.

. Pavlovian-to-Instrumental Transfer (PIT) task. Visual representation of the three task phases. During instrumental training, participants learned the association between three instrumental responses (R1, R2, R3) and the associated reward (juice) or no-reward (X). Response options were represented by the three middle white squares and outcomes appeared in the bottom square. During Pavlovian training, visual cues (fractals) appeared in the top square and their associated outcomes in the bottom square. During the transfer test, participants performed the instrumental choices under extinction (i.e., in the absence of rewards), while being presented with task-irrelevant Pavlovian cues. As a control condition (mask-only), no CS was presented between the two masking images. In Experiment 1 the cues were presented for either 600 ms (explicit trial) or 16 ms (subliminal trial); in Experiment 2 the cues were presented for either 600 ms (explicit trial) or 33 ms (subliminal trial); in Experiment 3 the cues were presented for 33 ms only. The grey letters (R1, R2, and R3) are for illustrative purposes of the response options and were not actually visible to the participants. www.nature.com/scientificreports/ Evidence from the PIT literature shows behavioral [13][14][15]19 and neural [16][17][18][20][21][22] dissociations between outcomespecific and general transfer. While general transfer is more dependent on the motivational properties of the Pavlovian cue, outcome-specific transfer is more related to the sensory-specific properties of the cue 23 . Thus, we expected that the lower level of elaboration consequent to the subliminal presentation of Pavlovian cues 5,11 may be sufficient to elicit general but not outcome-specific transfer. In other words, we expected the supraliminal processing condition to induce both general and outcome-specific transfer, while the subliminal processing condition to induce general transfer only.

Results
For each experiment, we first report the analyses on performance during Instrumental and Pavlovian training to confirm that participants correctly learned response-outcome contingencies (Instrumental training) and cue-outcome contingencies (Pavlovian training). Then we report the results on outcome-specific and general transfer. The models used in each analysis are the same across the 3 experiments. Therefore, we report a detailed description of the models only in Experiment 1.

Experiment 1. Instrumental training. Number of responses.
To test the learning of instrumental contingencies, the number of rewarded (R+ 1 , R+ 2 ) and unrewarded (R−) responses (mouse clicks) performed during each 10-s trial was compared 16,24 . A linear mixed-effects model was used, with the response (R+ 1 /R+ 2 /R−) as the independent variable and the number of responses performed as the dependent variable. Results showed a main effect of response (F (1.58,45.81) = 91.61; p < 0.0001; part.η 2 = 0.76; BF 10 = 2.07e+24). Post-hoc analysis showed a significantly higher number of R+ 1  Explicit learning. An analysis of explicit associations showed on average 77.01% correct response-outcome associations.
Pavlovian training. To test Pavlovian learning, two measures were considered: reaction times to the presentation of the patch preceding the outcome and the pre-post change in the liking of the conditioned cues (CSs).
Explicit learning. An analysis of explicit associations showed on average 88.5% correct cue-outcome associations.
Transfer effect. Outcome-specific transfer. A linear mixed-effects model was used with cue duration (16 ms/600 ms) and kind of response (congruent/incongruent) as the independent variables, and the number of responses as the dependent variable. Results showed statistically significant duration by response interaction (F (1,29) = 5.00; p = 0.03; part. η 2 = 0.15; BF 10 = 6,117,174). All other effects were not statistically significant (p > 0.21). Post-hoc analysis showed a significantly higher number of congruent (mean = 3.43; SD = 1.47) than incongruent (mean = 2.33; SD = 1.34) responses when the CS was presented for 600 ms [β = 1.1; p < 0.001; BF 10 = 10.16], but not when it was presented for 16 ms (p = 1; BF 10 = 0.369) [β = 0.33; p = 1; BF 10 = 0.36]. Although confidence intervals slightly overlap (Fig. 4), participants showed a significant increase of congruent over incongruent responses in the 600 ms condition only. Thus, outcome-specific transfer was observed only when the reward-associated Pavlovian cue was presented supraliminally (600 ms), but not when it was presented subliminally (16 ms Thus, general transfer was observed both when the reward-associated Pavlovian cue was presented supraliminally (600 ms) and when it was subliminally presented (16 ms). Mask-only. As a further control condition, the number of responses performed during the transfer phase was compared between mask-only trials (in which no Pavlovian cue was presented between the two masking images) and those in which a Pavlovian cue was presented for a subliminal (16 ms)    Explicit learning. An analysis of explicit associations showed on average 77.33% correct response-outcome associations. www.nature.com/scientificreports/ Pavlovian training. Reaction times. Results showed a significant main effect of CS (X 2 (2) = 11.81; p = 0.002; BF 10 = 1.13; e = 0.01%). Post-hoc comparisons revealed that reaction times were significantly faster for CS+ 1 (mean = 558.21; SD = 631.54) compared to CS− (mean = 668.62; SD = 735.22) [p = 0.004; BF 10 = 2.99], and for CS+ 2 (mean = 562.01; SD = 635.56) compared to CS− [p = 0.002; BF 10 = 2.26]. Crucially, there was no difference between CS + 1 and CS + 2 [p = 0.82; BF 10 = 0.06]. Although confidence intervals slightly overlapped, overall participants showed a tendency to be faster when a reward was anticipated (CS + 1, CS + 2) as compared to when no reward was expected (CS−), thus indicating discrimination between the cues and, thus, Pavlovian training.
Explicit learning. An analysis of explicit associations showed on average 84.52% correct cue-outcome associations.
Interim discussion. Overall, the results of experiments 1 and 2 show that supraliminal (600 ms) rewardassociated Pavlovian cues biased choice towards motivationally similar outcomes (i.e., general transfer) as well as outcomes with the same sensory-specific properties (i.e., outcome-specific transfer), in line with previous evidence 8,1415,24 . The novel result is that the subliminal (16 ms and 33 ms) presentation of reward-associated Pavlovian cues biased choice towards motivationally similar outcomes (i.e., general transfer). Additionally, while Experiment 1 showed no outcome-specific transfer for the 16 ms condition (subliminal processing-imperceptible cues), Experiment 2 reported such effect for cues lasting 33 ms (subliminal processing-perceptible cues). Thus, the extent to which subliminally presented reward-associated Pavlovian cues influence choice towards outcomes with the same sensory-specific properties remains to be clarified. These results might be explained by the individual variability of the subjective visibility with a subliminal presentation of 33 ms. Indeed, studies comparing behavioral responses and brain activation indicated that a proportion of cues presented for 33 ms could still be categorized as clearly visible 11 . In this case, a precise differentiation between seen and unseen trials could help clarifying the results from Experiment 2. For this reason, in a third experiment, the PIT paradigm was implemented using only the subliminal presentation of 33 ms, and the transfer effect was compared between "seen" and "unseen" trials, as assessed by participants' judgment of their subjective experience (see "Methods"). , with clearly separate 95% confidence intervals (Fig. 3), Crucially, there was no difference between CS + 1 and CS + 2 [p = 0.61; BF 10 = 0.27]. These results indicated that, after Pavlovian training, participants significantly increased their liking for reward-paired cues (CS+ 1 and CS+ 2 ), as compared to the unrewarded cue (CS−), thus confirming successful Pavlovian training (Fig. 3).
Explicit learning. Analysis of explicit associations showed on average 68% correct cue-outcome associations.
Outcome-specific transfer. To confirm the presence of the outcome-specific transfer, the transfer index was tested against zero, where zero represented an equal number of congruent and incongruent responses, i.e. absence of transfer. A higher transfer index (indicating a higher number of congruent than incongruent responses) was reported during seen trials (t (14) (Fig. 4). In a second analysis, a linear mixed-effects model was used with CS (seen/unseen) as the independent variable and the outcome-specific transfer index as the dependent variable. Results showed a statistically significant difference (F (1,14) = 7.13; β = 0.84; p = 0.03; part.η 2 = 0.47; BF 10 = 2.18) with a higher transfer index during seen (mean = 0.63; SD = 0.54) than unseen (mean = − 0.03; SD = 0.89) trials (Fig. 4). Thus, outcome-specific transfer was observed only when the rewardassociated Pavlovian cue was explicitly perceived (seen), but not when it was not explicitly perceived (unseen). Overall, these results demonstrate that outcome-specific transfer depends on the conscious elaboration of the reward-associated cue.

Discussion
The present work investigated the extent to which conscious elaboration represents a boundary condition within which environmental cues can influence individual choice. To this end, we compared the effect of subliminal and supraliminal presentation of reward-associated Pavlovian cues during the transfer phase of a Pavlovian-to-Instrumental Transfer (PIT) task. In line with previous evidence 2, 8,14,15,24,25 , the supraliminal presentation (600 ms) of reward-associated cues biased choice towards rewards sharing the precise sensory-specific (outcome-specific transfer) and motivationally similar (general transfer) properties of the cues. On the other hand, the subliminal presentation of reward-associated cues was always able to bias choice towards motivationally similar rewards (general transfer), but not sufficient to bias choices associated with rewards sharing the precise sensory-specific properties of the cue (outcome-specific transfer). More precisely, perceptible subliminal Pavlovian cues (33 ms) triggered outcome-specific transfer only when the cues were explicitly perceived (seen trials), but not when they were only subliminally processed (unseen trials). Crucially, imperceptible subliminal Pavlovian cues (16 ms) were never associated with outcome-specific transfer.
Scientific RepoRtS | (2020) 10:11926 | https://doi.org/10.1038/s41598-020-68926-y www.nature.com/scientificreports/ Taken together, the present findings suggest that cue-guided choices are modulated by the level of perceptual threshold (i.e., subliminal vs supraliminal) of reward-associated cues 14,15,19 . While a conscious elaboration of the reward-associated cue is necessary to bias choice towards the exact same reward, subliminal processing is only sufficient to bias choice towards rewards sharing the same motivational properties.
The present results provide crucial evidence for the theoretical accounts of cue-guided choice. This study supports previous evidence highlighting behavioral dissociations at the core of outcome-specific and general transfer 14,15,19 . For example, high-level cognitive abilities, such as working memory, appear to be crucial for the expression of a sensory specific biasing effect, but not of a general biasing effect 14 . In line with this, human neuroimaging studies reported the activation of separate brain regions for outcome-specific (dorsal striatum and ventral amygdala) and general (ventral striatum and dorsal amygdala) transfer [16][17][18]20,22 . Such differences could be attributed either to separate mechanisms, possibly mediated by different neural networks, or to a continuum of the same system ending on a different path as a function of quantity and quality of the information carried by the cue. While a lower level of processing provided by the subliminal presentation of the cue 5,11 might be sufficient to convey the necessary motivational properties responsible for general transfer, a higher level of processing would be required to activate the sensory-specific representation of the reward characterizing outcome-specific transfer. At least in humans, outcome-specific and general transfer could thus be supported by different mechanisms, one more explicit and another more automatic 14,15,19,23 . Future studies might explore whether these differences can be related to the activation of different brain networks. For example, the fronto-striatal loops involved in flexible behavioral adaptation 21,26-30 could be differentiated in a cortical component related to supraliminal processing, and a subcortical component involved in subliminal processing 16,23,25,31,32 . Another critical issue is whether individual differences in the prevalence of the automatic component over the explicit one can indicate a higher predisposition to develop a maladaptive behavior 13,33 .
These results have implications in both daily life choices and clinical contexts. In everyday life, this result translates to the fact that although an explicitly perceived logo can persuade to buy a specific product (e.g., viewing the McDonald's logo along the street can lead to driving the car towards a McDonald fast-food restaurant), a much lower level of processing of the same information can still bias choice towards the purchase of similar products (e.g., the presence of McDonald's logo on the street, even if not consciously perceived, can lead to driving the car towards any fast-food restaurant). From a clinical perspective, cue-guided choices are related to several conditions, ranging from compulsive conducts, to neuropsychiatric disorders (like depression), and addiction (for its strong implications with relapse) 7,9,[34][35][36][37][38][39][40][41][42][43] . The prevalence of automatic responding is indeed generally associated with addictive, impulsive, and obsessive traits 3439, [44][45][46][47] . In this sense, the lack of conscious processing that characterizes general transfer could potentially be a marker of vulnerability to maladaptive forms of cue-guided choices 3 . In the case of addiction, for example, although no direct correlation between drug or food forms of addiction and outcome-specific transfer has been reported in literature 48,49 , it cannot be excluded that this is instead related to forms of cue-driven choice requiring a lower level of processing, like in the case of general transfer 50 .
Finally, some limitations of the present study shall be discussed. Differently from previous studies, the participants did not receive an actual reward for themselves (but rather played to quench the thirst of a camel crossing the desert) and were exposed to Pavlovian cues for a short time [13][14][15] . Although these manipulations may have weakened their motivation to perform the task and could partially explain the presence of small effect sizes, both outcome-specific and general transfer were still observed. Moreover, the operationalization of general transfer in the present study differs from the one adopted in animal literature 25 , thus limiting the translational potential of the results linked to general transfer. However, the same operationalization has been used in the human literature [13][14][15]17,18 , thus enabling a comparison of our results with previous studies.

conclusions
The present findings highlight how environmental cues can influence human choice depending on their level of processing. If explicitly perceived, reward-associated cues can guide behavior towards choices aimed at the same or similar rewards. Crucially, subliminal processing of such cues is sufficient to bias choice towards motivationally similar, yet not identical, rewards. The current understanding of cue-guided choices, and of their implications in daily life and clinical conditions, warrants further studies to explore the extent to which reward-associated cues can exert a maladaptive control over choice. We suggest that the dissociation outlined in this as well as in previous studies [14][15][16][17][18][19][20]23,25 between outcome-specific and general transfer is crucial to shed new light on the mechanisms at the core of cue-guided choices.

participants.
The three experiments were performed on independent samples: Experiment 1 involved 15 females and 15 men, mean ± sd 25.28 ± 2.69; Experiment 2 involved 15 females and 15 men, mean ± sd 24.73 ± 2.58; Experiment 3 involved 8 females and 7 men, mean ± sd 25.73 ± 2.73. All participants gave their written informed consent to take part in the experiment. The study was conducted following institutional guidelines and the 1964 Declaration of Helsinki and was approved by the Ethics Committee of the Department of Psychology of the University of Campania. Participants were naive about the hypotheses of the experiment. All volunteers had no history of neurological or psychiatric diseases were recruited from the student population of the University of Campania "L. Vanvitelli". The number of participants was determined using power analysis for F test family (power = 0.9, alpha = 0.05, effect size f = 0.35), estimating the effect size based on previous literature using a similar PIT task [13][14][15] . This analysis was computed using the software G*Power version 3.1.9.4. pavlovian-to-instrumental transfer task. The task mirrored the structure used in previous studies 13-18 : (1) Instrumental training, in which participants learned a response-contingent reward; (2) Pavlovian training, Scientific RepoRtS | (2020) 10:11926 | https://doi.org/10.1038/s41598-020-68926-y www.nature.com/scientificreports/ in which participants learned a cue-contingent reward; (3) Transfer test, during which the influence of taskirrelevant Pavlovian cues on instrumental responding was tested. The same visual settings were used in all phases of the PIT task (see Fig. 1). A computer running OpenSesame software 51 controlled cue presentation. Five white (2 mm thickness) squared frames were displayed on a 17-inch color monitor with the image of a desert as background. One frame was positioned in the upper portion of the screen, highlighting the area where the Pavlovian conditioned cues would be displayed during the Pavlovian training and transfer test. Three squares were positioned horizontally next to each other in the middle portion of the screen, highlighting the regions of the screen to be clicked with the mouse to make a response during instrumental training and transfer test. One square was positioned in the bottom portion of the screen, highlighting the area where the outcome would be displayed during instrumental training, Pavlovian training and transfer test.
Instrumental training. This training aimed to learn the association between a specific instrumental response and its corresponding outcome. In each trial, participants made a response, i.e. they clicked on one of the three frames placed in the middle portion of the screen. Two responses (R+ 1 and R+ 2 ) were each paired with one of two different reward outcomes (i.e. the picture of an apple and orange juice) on 80% of trials, and with the no-reward outcome (i.e. picture of an 'X') in the remaining 20% of trials. The third response (R−) was always paired with the no-reward outcome. The response-outcome association was counterbalanced across participants. In each trial, only two out of the three responses were available, and this was indicated to participants by a white patch appearing within the square. After each response, a corresponding no-reward or reward outcome appeared for 1 s in the bottom square. The task consisted of a total of 90 randomized trials (30 trials for each pair of responses, see Table 1), and lasted about 5 min. Three example trials (one for each pair of responses) were performed before beginning.
Pavlovian training. This training aimed to learn the association between a specific conditioned cue and its corresponding outcome. In each trial, one of three conditioned cues (CS, i.e. fractal images balanced for luminance, complexity, and color saturation) appeared within the upper frame for 2.5 s. Then, a white patch appeared within the bottom square. Participants were instructed to press the left-Ctrl button on the computer keyboard as quickly as possible to remove the patch and discover the outcome hidden behind it. The outcome was presented for 1 s. Two CSs (CS+ 1 and CS+ 2 ) were associated with a reward (apple or orange juice) on 80% of trials and with the no-reward outcome (i.e. picture of an 'X') in the remaining 20% of trials. The third CS (CS−) was associated with the no-reward outcome on all trials. The cue-outcome association was counterbalanced across participants. The task consisted of 60 randomized trials (20 per CS, see Table 1), for a total duration of about 6 min. Three example trials (one for each CS) were performed before beginning. This Pavlovian speeded reaction-time response has been successfully used to obtain a behavioral measure of Pavlovian training 13,14,16 . To avoid a possible instrumental component, participants were explicitly told and demonstrated that the reward delivery was not dependent upon their response. Indeed, even without a button press, the patch would disappear after 3 s and reveal the hidden outcome. The main reason for using a speeded response was to mirror PIT studies on animals, in which Pavlovian training is measured by a behavior performed to gain the reward 31,52,53 . The rationale is that the higher arousal produced by reward expectancy shall produce faster reaction times when a reward rather than no reward is predicted (i.e. CS+ vs. CS−) 13,14,16,31,52,53 . As a further measure of learning, subjective liking for each CSs was rated by participants on a 9-items Likert scale, before and after Pavlovian training. The rationale was to observe an increase in liking scores after Pavlovian training for CS+ 1 and CS+ 2 , as compared to the CS−. At the end of the two training sessions, explicit learning was assessed. Participants saw each response option and CS and were required to associate them with their corresponding outcomes.
Transfer test. This phase aimed to test the influence of the Pavlovian CSs to drive instrumental responses. The task mirrored instrumental training, with two differences. First, while performing responses, task-irrelevant Pavlovian CSs were presented within the upper square, at the beginning of the trial. Crucially, the duration (supraliminal/subliminal) of the CSs varied across the three experiments: in Experiment 1, the cues lasted for either 600 ms (supraliminal processing) or 16 ms (subliminal processing-imperceptible cues); in Experiment 2, the cues lasted for either 600 ms (supraliminal processing) or 33 ms (subliminal processing-perceptible cues); in Experiment 3, all cues lasted for 33 ms. In a control condition (mask-only), no CS was presented between the two masking images. Only in Experiment 3, after each response participants were presented with all CSs and a "no-image" option and invited to indicate whether one of the images appeared in the previous trial. Trials of the PIT phase were consequently categorized as "seen" if the CS was correctly recognized, or "unseen" if the CS was either not correctly recognized or the no-image option was selected. Second, the whole task was performed under extinction, so that no reward was ever delivered. Extinction is a standard procedure for assessing transfer effect, both in human and non-human animal research, since it allows to test the influence of Pavlovian cues on instrumental responding without the confounding effects of the reward 16,18,54 . Subliminal presentation of visual cues. Experimentally, the conscious perception of a target visual cue occurs only if the interval between two masking images preceding and following the target exceeds a certain duration 5,11,55 . The threshold of such duration has been widely investigated and is commonly held to range between 16 and 33ms 5,11,55 . At a duration of 16 ms, visual cues are not accessible to consciousness and are imperceptible; at a duration of 33 ms they are still not accessible to consciousness but perceptible 5,11 . A duration of 600 ms allows supraliminal processing 5,11,55 . In this experiment, a backward masking paradigm with two different masks (random black and white dots) presented immediately before and after the presentation of the cues was used 11,55 following a pilot study aimed to test the implementation and validity of the chosen duration of the Scientific RepoRtS | (2020) 10:11926 | https://doi.org/10.1038/s41598-020-68926-y www.nature.com/scientificreports/ cues (16 ms/33 ms/600 ms). The duration of the masks was equally divided between the two (pre/post cue) and varied according to the duration of the CS so that the total duration of the three cues was always equal to 1 s. For instance, when the target cue lasted 600 ms, the two masks lasted 200 ms each.

Definition of outcome-specific and general transfer and transfer indexes. Outcome-specific
transfer. Participants were presented with a task-irrelevant CS+ (CS+ 1 or CS+ 2 ) while required to choose between two available response options (R+ 1 and R+ 2 ) previously paired with two different reward outcomes, of which only one was also previously paired with the concurrently presented CS. So, for example, if the CS+ 1 was presented, choosing R+ 1 would constitute a congruent response, while choosing R+ 2 would constitute an incongruent response. Similarly, if the CS+ 2 was presented, choosing R+ 1 would constitute an incongruent response, while choosing R+ 2 would constitute a congruent response. Evidence for outcome-specific transfer would be seen if the presence of the CS-induced a higher rate of congruent responses, as compared to incongruent responses. Consequently, outcome-specific transfer was tested considering only those trials in which a CS+ 1 or CS+ 2 was presented while choosing between R+ 1 or R+ 2 (see Table 1 for a detailed description). The rationale of the outcome-specific transfer is to test if a CS+ (i.e., a reward-paired cue) can elicit a response independently associated with the same reinforcer. To this aim, responses were categorized as congruent (choosing R+ 1 while CS+ 1 is presented or choosing R+ 2 while CS+ 2 is presented) or incongruent (choosing R+ 1 while CS+ 2 is presented or choosing R+ 2 while CS+ 1 is presented) and compared [13][14][15] . A transfer index was calculated, for each trial, as (number of congruent responses-number of incongruent responses) / total number of responses [13][14][15] . A positive score indicated a preference for congruent over incongruent responses, while a negative score the opposite. A value of zero represented the absence of outcome-specific transfer.
General transfer. Participants were presented with a task-irrelevant CS (either reward-associated, CS+ 1 /CS+ 2, or CS−) and required to choose between two available response options none of which was compatible with the CS currently available (i.e., the responses were always associated with a different or no outcome). So, if the CS+ 1 was presented the available response options were R+ 2 and R−; if the CS+ 2 was presented the available response options were R+ 1 and R−; if the CS− was presented the available response options could be either R+ 1 and R− or R+ 2 and R−. Evidence for general transfer would be seen if participants favored the R+ during the presentation of a CS+, as compared to the CS−. The whole task consisted of 135 randomized trials (15 repetitions for each trial type, see Table 1) and lasted for about 12 min. General transfer was tested by comparing the preference for rewarded over unrewarded responses when presented with a CS+ previously paired with a different outcome, relative to the CS− (see Table 1 for a detailed description) [13][14][15] . A preference index was created by subtracting, for each trial, the number of R− from the number of R+choices and dividing for the total number of responses (reward-associated responses-unrewarded responses) /total number of responses [13][14][15] . A positive score corresponded to a preference for R+ over R− during that trial, and vice-versa. To obtain a single indicator of transfer CS− trials were then subtracted from CS+ trials. In the resulting transfer index, a positive value indicated a preference for R+ over R− during CS+ trial, while a negative score the opposite. A value of zero represented the absence of general transfer. In the whole task, trials were equally divided into three time-points. Given the known decremental effect of responses during extinction 9,13,16 , only the first is reported in the analysis.
procedure. Participants seated comfortably in a silent room and their position was centered relative to the screen, at a viewing distance of about 50 cm. At the beginning of the experimental procedure, participants were instructed to play a computer game aiming to accumulate apple and orange juice to quench the thirst of a camel crossing the desert. Then the task began. In each phase of the task, the participants were required to pay attention to the screen and follow the instructions reported at the beginning of the task.