We have been concerned with people’s performance in our tasks. We can gain further insights into working memory by considering the biological mechanisms that make this performance possible. Dous it literally involve covert speech and thus moevements of the tongue? Data indicate, that people who can’t produce speech, show sond-alike confusions in their data, just as ordinary participants do. This observation suggests that actual muscle movements are not needed for subvocalization. ‘inner speech’ probably relies on the brain areas responsible for planning the muscle movements of speech. Recent developments in brain-imaging technology tell us that when a participant is engaged in working-memory rehearsal, considerable activity is observed in brain areas that we know are crucially involved in the production of spoken language, as well as in areas that play an important role in the perception and understanding of spoken language. We can also gain insights by comparing diverse populations: Deaf people rely on a different assistant for working memory: they use an inner hand rather than a inner voice. It is important that we have built our argument with multiple lines of evidence. It’s only when we take the results as a packeage that we can make headway, and if we have done our work well, there will be just one theoretical account that fits with the entire data pattern.