Skip to main content

Wedding bells or bedding wells? Lexical and semantic influences on phoneme binding

Final Report Summary - SEMBIND (Wedding bells or bedding wells? Lexical and semantic influences on phoneme binding)

The SEMBIND project successfully characterised the mechanisms that support our ability to maintain speech sounds in the correct order – a fundamental capacity that underpins all aspects of language processing. We described semantic binding behaviourally, examined how this capacity breaks down as a consequence of semantic impairment in patients with stroke and dementia, characterised the neural mechanisms underpinning this aspect of cognition using neuroimaging, obtained evidence for a causal role for these brain regions using inhibitory brain stimulation in healthy participants, and detailed the underlying causes of individual differences in the aspects of cognition that support binding within and beyond language. Our findings show that our capacity to maintain language information in the correct configuration emerges from the interplay of two sets of representations: (i) The phonological system can hold on to speech sounds in sequence, given that order information is an inherent property of language. Learning within the phonological system (i.e. familiarisation with the phonological forms of new words) reduces the frequency with which phonemes from different items split apart and recombine with the elements of other items during tasks such as immediate serial recall. We have shown that electrical stimulation delivered to a brain area linked to phonological processing facilitates the acquisition of new phonological forms, and this reduces the frequency with which phonemes migrate between items the following day. (ii) In addition, we have obtained evidence across several converging methods (experimental psychology, magnetic brain stimulation studies, patient data) that interactions between semantics and phonology support phoneme sequencing further. Patients with semantic dementia, who have impaired understanding of specific words, show frequent errors in immediate serial recall and paced reading tasks: the phonological elements of words that these patients no longer fully understood are liable to split apart and recombine with the phonemes of other list items, even though these patients do not have pronounced phonological deficits. Healthy participants show similar effects of “semantic binding” in difficult language tasks: for example, in immediate serial recall, phoneme migrations are reduced when words are categorised according to their meaning at the time of presentation. These errors are also reduced for new words when participants are familiarised not only with the phonological forms of these items but also with their meaning. In addition to using simple lists of words, we have successfully explored these issues with more naturalistic language input, such as sentences/stories, in which effects of semantic binding were found to be much stronger. We have also explored parallel effects of semantic binding beyond language – for example, in understanding actions, and in scaffolding our dynamic spontaneous cognition.