The present study investigated the production of British English (BrE) vowels by Yemeni-Arabic learners of English. Specifically, the most problematic BrE vowels for those learners were explored in relation to Lados’ (1957) contrastive analysis hypothesis (CAH) and Flege’s (1995) Speech Learning Model (SLM). Sixty-seven Yemeni EFL learners at different proficiency levels completed a questionnaire, which mainly explored how difficult the BrE vowels are for them, and a vowel production test, which measured the learners’ production accuracy of the BrE vowels. Overall, the results revealed that Yemeni EFL learners encountered some difficulties when pronouncing BrE vowels, but their production accuracy rates increased as their levels grow. It was found that the same vowels, /e/, /ɒ/, /eə/, /ɔ:/, /ʊ/ & /u:/, were found to be the most inaccurately produced sounds by learners at all levels. The subtle differences between vowels might have caused the learners’ production problems. Findings also showed that learners’ L1 vowels that are different from the BrE vowels were easier to produce, and this is congruent with the SLM.