A major difficulty in studies of the brain, from the molecular to large-scale network level, is ensuring the accuracy and reliability of results, since repeatability has been a problem in studies utilizing functional magnetic resonance imaging (f-MRI) near-infrared spectroscopy (NIRS), and positron-emission tomography (PET). More generally, an effort to replicate psychological studies has shown that the original results were unambiguously reobtained only 39% of the time. It has been suggested that researchers must undertake studies to identify factors that reduce reliability and conduct more carefully controlled studies to improve reliability. In our previous work, we examined whether changes in hand/arm posture can have a confounding effect on task-related brain activity. Here we show a solution to enhance reproducibility in a NIRS study in a hearing task. The results showed that crossed posture can lead to different results than parallel posture with respect to asymmetric functional connectivity, especially during non-resting state. Even when the only task is listening to speech stimuli, participants should be asked to place their hands on a surface and feet on the floor and keep the same stable posture to increase reproducibility of results. To achieve accurate reliability and reproductively of results, stable hand posture through the experiment is important.