Human-robot collaborations that operate in shared spaces are anticipated to become increasingly common in coming years. Decades of social psychological research have revealed that human observers positively influence people's performance in dominant and negatively in nondominant tasks. While studies indicate moderate support for social facilitation/inhibition effects with robot observers, this evidence is hotly debated. Addressing known methodological criticism, this study investigates how a copresent robot-observer affects Stroop task performance and whether perceptions of that robot's mental capacities have explanatory value. Results reveal limitations in transferring social facilitation/inhibition theory to robots. Since participants reported high task attention levels across conditions, emerging flow states may have helped them circumvent social facilitation/inhibition mechanisms. It may thus be recommended for future research to consider flow dynamics when investigating social performance effects.