How does pre-nasal allophony vary across dialects of the same language—gradiently, categorically, or in a hybrid fashion? This question has implications for theories of the relationship between phonological structure and phonetic output, but answering it requires a broad empirical base that has been lacking. We address this gap by comparing overlap between /ɪ/ and /ɛ/ in F1/F2 space in pre-nasal and pre-oral contexts across 25 dialects of English from North America and the UK, using data primarily from spontaneous speech and Bayesian models of vowel distributions. Because formant frequencies reflect both nasal coupling and changes in oral articulation, our measures capture the combined effect of these sources and cannot distinguish between them. With this caveat, dialects fall roughly into two groups—those with high and low effects of nasal context—but intermediate dialects are attested. Interspeaker variation is structured differently across dialect regions: North American dialects show substantially more variation in the pre-nasal dimension than British dialects, consistent with the previously hypothesized greater role of learned nasal coarticulation in North American English. These findings support a hybrid view of allophonic variation, in which dialects are biased toward particular configurations without being restricted to them, and provide new empirical constraints that were not available from smaller-scale cross-linguistic or within-dialect studies.