This study adopts Nelson’s (2014) methodological framework to investigate core and peripheral lexical bundles (i.e. recurrent multi-word sequences) in conversation, using data from the British, Canadian, Singapore and Hong Kong components of the International Corpus of English (ICE). The overlap and non-overlap comparisons reveal (dis)similarities in the use of bundles across the four World Englishes (WEs). Findings suggest that in terms of discourse building blocks, the more advanced a variety is according to Schneider’s (2007) Dynamic Model of New Englishes, the more lexical bundles it shares with the common core in conversation. Canadian English (CanE) shares the most common ground with British English (BrE). As a nascent variety, Hong Kong English (HKE) differs most from BrE, while Singapore English falls between CanE and HKE. Though the results do not correlate with Schneider’s Dynamic Model at the level of recurring chunks, they allow to test predictions of WEs models. Quantitative and qualitative analyses enable the identification of bundles with significantly high frequency in each regional variety, thus enriching comparative research of WEs.