The neural correlates of creativity, specifically insight, is a matter of intense interest. Insight problem-solving is not deduced logically and the solution is typically very hard to get and requires “out of the box” thinking. The nine-dot problem is one such example (Fig 1). In fact Chronicle et al [1] state that “the expected solution rate for this problem under laboratory conditions is 0%”. Using transcranial direct current stimulation (tDCS), Chi et al [2] demonstrated that increasing the excitability of the right anterior temporal lobe (rATL) enabled surprisingly large number of people to solve the nine-dot problem. They hypothesized that this mitigated cognitive biases and helped subjects arrive at the solution. Here we test this hypothesis using real-time neurofeedback based on functional magnetic resonance imaging (fMRI). We hypothesized that if subjects are enabled to up-regulate activity in their rATL using neurofeedback, it must mimic the effect of tDCS and hence enable them to solve the problem. 13 healthy subjects (age range 22-28 years, mean age 21.059 years) were scanned in a 7T Siemens MAGNETOM scanner using an EPI sequence with TR=1500 ms, TE=25 ms, voxel size= 3.0×3.0×3.0 mm3 and 300 measurements per run. The experimental paradigm consisted of 4 runs per subject. Each run was a block design with 10 TRs rest followed by 20 TRs of problem solving with neurofeedback training where in the signal from rATL from the subject’s brain was given as feedback and the subjects were asked to increase the signal while trying to solve the problem. After on-line reconstruction of the data, it was sent to a computer where in the data was subjected to standard pre-processing using BioImage Suite software [3]. The feedback signal was selected from a pre-defined ROI of rATL and displayed to the subject via an MR-compatible projection system. BioImage Suite has real-time capabilities and the time lag between actual brain activity and its display as feedback to the subject was under 2 s. fMRI data was also analyzed offline using SPM12 for finding regions activated by the task vs rest contrast. Custom MATLAB code was used to calculate % signal change during the task versus rest in each run. The subjects were grouped into two groups based on whether they were able to solve the problem or not and analysis was carried out separately in both the groups. Results and Discussion: In our experiment, 4 subjects solved the problem and 9 subjects did not (i.e. 44% were successful). This is in close agreement with Chi et al [2] who found that 40% solved the problem with tDCS. Fig.2 shows the % BOLD signal change for each group across runs. Both groups were able to learn to regulate activity from their own rATL over time, subjects who were able to up-regulate more ended up solving the problem. This supports the hypothesis we stated earlier, that increasing the activity of rATL enables insight by potentially mitigating cognitive biases. Offline activation analysis from all subjects for the task vs rest contrast showed that right middle temporal gyrus, middle frontal lobe, precentral gyrus and parietal lobe (Fig.3) were activated. The solved group showed higher activation in right superior temporal gyrus, middle temporal gyrus, frontal lobe and hippocampus (Fig.4). These regions have been previously reported to be involved in insightful thinking by Zhao et al [4]. Specifically, the hippocampus has been proposed as the key brain region for forming novel associations [5], the right superior temporal gyrus for facilitating the formation of remote associations [6,7], and the middle temporal gyrus were found to be active even before the solution [4] and hence has been hypothesized to be related to the control of memory-guided saccades [8]. The frontal regions are likely involved in top-down modulation (which impose cognitive biases while solving problems) of temporal regions, forcing individuals to see a Gestalt, i.e. a square bounding box for the nine dots, which likely prevents them from “thinking out of the box”. Therefore, up-regulation of activity in rATL will likely mitigate this bias. Furthermore, our study demonstrates that neurofeedback could potentially be used to mimic effects seen in brain stimulation techniques such as tDCS. This provides a likely framework for cross-pollination between brain stimulation and neurofeedback paradigms.