Real-time fMRI (rtfMRI) neurofeedback (NF) provides online contingent information to individuals about their own brain hemodynamic activity, enabling volitional control of BOLD signal in local or extended brain regions, potentially leading to behavioural changes [1,2]. The learning process involved into achieving self-regulation of brain hemodynamics has been associated to several factors such as type of feedback, reward, the use of mental imagery, session duration, among others. [3] A common practice in rtfMRI-NF experiments is to explicitly instruct subjects to use mental imagery, under the assumption that those instructions will enhance the learning process. On the other hand, monetary reward according to performance has also been used as reinforcement to improve brain self-regulation. However, it is yet to be defined which is the optimal strategy to improve self-regulation capacity, a critical aspect for designing efficient protocols in rtfMRI-NF studies. Furthermore, the underlying neural mechanisms ruling BOLD signal control are still an unresolved issue.Our aim is to compare the efficiency in BOLD signal self-regulation of two experimental strategies used In NF training: giving explicit instructions (i.e. motor imagery) and monetary reward according to performance. As target area for hemodynamic signal control the Supplementary Motor Area (SMA) was selected, area involved in motor imagery and used in previous fMRI-NF experiments [4,5]. Additionally, we explored variations of brain patterns generated during BOLD self-regulation training using the selected strategies.Four groups of 5 healthy subjects each (22.75 ± 1.6 y.o., males, right handed) were trained in an rtfMRI-NF protocol with contingent visual neurofeedback. Groups: NF only, NF + explicit instructions (motor imagery, IM), NF + monetary reward (R), and NF + IM + R (figure 1). Each subject was trained during 2 days, 4 training runs per day (4 minutes each) alternating baseline and up-regulation blocks. After completing NF training a “transfer run” was included in which subjects were instructed to try to self-regulate SMA, but without receiving NF. The rtfMRI system (figure 2) was implemented using a 1.5T MR Scanner (Philips Achieva, The Netherlands) running a real-time reconstructor package (DRIN) with a functional image acquisition protocol: FFE-EPI sequence with TR/TE=1500/45 ms, voxel size=3.2x3.3x4 mm³ and 150 measurements (10 dummy scans). We used a standard PC running Turbo Brain Voyager rtfMRI software (Brain Innovations, The Netherlands) and custom MATLAB scripts to generate feedback information. A second PC with Presentation software (NBS, USA) was used to present visual feedback to the subjects (display of a graphical "thermometer") in an MR-compatible Visual System (NNL AS, Norway). Anatomical T1-weighted images were acquired both days. FMRI volumes were preprocessed and analyzed using SPM and additional in-house MATLAB scripts. To evaluate self-regulation success we used r_SMA : percent change of BOLD signal during up-regulation compared to baseline blocks for each run in a selected ROI of SMA (x: -8, 8; y: -8, 8; z: 52, 68 in MNI coordinates). Group pattern analysis was performed for each study group using a 2nd level analysis in SPM. We compared SMA self-regulation levels across groups. An ANOVA analysis was performed on the values of r_SMA for each group (figure 3). A significant group effect was found (F_3-156=4.643; p<0.01) and a post-hoc analysis revealed that BOLD self-regulation levels were significantly higher in the NF+R group than in NF and NF+IM groups (p<0.01 & p<0.05; respectively). The changes in r_SMA mean values between day 1 and 2 reported a significant increment only for NF group (Wilcoxon signed-rank test; p<0.05). The transfer session reported that all groups were able to maintain the self-regulation capability without receiving contingent NF (figure 4) after the training. Brain patterns for each individual group are depicted in Figure 5. A Two-way ANOVA was used to determine regions activated exclusively for each of the tested strategies. Only a small cluster at right precentral cortex (x=40;y=-17;z=60; k_e=19,p_FWE-corr<0.05) was significantly activated for NF+R group effect. No significant differences were found at a pattern level for the other groups.Results indicate that the use of explicit instructions (IM) in rtfMRI-NF has not necessarily a significant impact in learning efficiency, at least in SMA. Additionally, when monetary reward was included a tendency to stronger amplitude in BOLD self-regulation is observed. These results are in line with the importance of operant conditioning in learning to self-regulate brain hemodynamics in human NF studies. Regarding brain patterns, highly similar brain activated areas were found across different strategies. This similarity can indicate the predominance of contingent NF on brain activations in comparison to additional factors during NF training.