Recently, reservoir computing techniques have been employed to elucidate the functions of the large-scale network of structural brain connections known as the connectome. Researchers investigated the role of brain connectivity patterns in reservoir computing by comparing task performance when the reservoir layer nodes of the echo state network were connected according to the connectome versus when they were connected based on rewired null networks. While the performance of the original connectome and its nulls has been compared in a standard memory capacity task, a similar comparative analysis has not yet been made in more realistic neuroscience tasks. In this study, we perform comparative analyses with null networks while applying connectome-based reservoir computing to four neuroscience tasks in NeuroGym using the conn2res toolbox. We found that the discrepancy in performance between the original connectome and its nulls varied depending on the selected neuroscience tasks. Nonetheless, for three of the four neuroscience tasks we examined, the decline in task performance as the spectral radius of the reservoir weight matrix increased was less pronounced for the original connectome than for the null networks. The observed robustness of performance to increasing spectral radius was attributable to network-level characteristics inherent in the connectome rather than the presence of high-degree hub nodes. This study offers insight into how brain connectivity patterns within the connectome influence computational ability when solving realistic neuroscience tasks.