False-belief (FB) tasks have been widely used to study the ability of individuals to represent the content of their conspecifics' mental states (theory of mind). However, the cognitive processes involved are still poorly understood, and it remains particularly debated whether language and inner speech are necessary for the attribution of beliefs to other agents. We present a completely nonverbal paradigm consisting of silent animated cartoons in five closely related conditions, systematically teasing apart different aspects of scene analysis and allowing the assessment of the attribution of beliefs, goals, and physical causation. In order to test the role of language in belief attribution, we used verbal shadowing as a dual task to inhibit inner speech. Data on 58 healthy adults indicate that verbal interference decreases overall performance, but has no specific effect on belief attribution. Participants remained able to attribute beliefs despite heavy concurrent demands on their verbal abilities. Our results are most consistent with the hypothesis that belief attribution is independent from inner speech.