This multidisciplinary project examines the mechanisms in the mind and brain that underpin phoneme binding. We explore how knowledge of the sounds and meanings of words prevents their phonemes from recombining (so that errors such as ‘bedding wells’ for ‘wedding bells’ are avoided). Phoneme binding plays a fundamental role in understanding and producing running speech, but the underlying processes are not widely studied. We take advantage of recent methodological and theoretical advances, many of which arise from our own studies, in a programme of research designed to uncover the interactive semantic and phonological processes that underpin this process. We use three complementary techniques – neuropsychology, experimental studies of healthy participants and transcranial magnetic stimulation (TMS) – which yield complementary information. (i) Patients with semantic dementia reveal the way in which selective impairment of conceptual knowledge affects the ability to maintain sequences of speech sounds that make up words. For the first time, we compare patients on several different tasks, tapping verbal short-term memory, rapid reading and verbal memory in the absence of overt speech, to establish whether the same underlying processes are at work. (ii) We undertake a similar comparison of different tasks in healthy participants, using mixed lists containing both words and nonwords, as these stimuli have been found to elicit more phoneme recombinations for words in our past research. (iii) We adopt the highly novel approach of teaching people either the meanings or just the sounds of new words in order to examine the separate effects on phoneme binding. (iv) TMS studies of healthy volunteers allow us to examine the impact of temporarily disrupting processing within brain areas that underpin semantic and phonological aspects of language. This allows us to draw strong conclusions about the cognitive and neural processes that are essential for phoneme binding.