Background: Deafness has an adverse impact on children's ability to acquire spoken languages. Signed languages offer a more accessible input for deaf children, but because the vast majority are born to hearing parents who do not sign, their early exposure to sign language is limited. Deaf children as a whole are therefore at high risk of language delays. Aims: We compared deaf and hearing children's performance on a semantic fluency task. Optimal performance on this task requires a systematic search of the mental lexicon, the retrieval of words within a subcategory and, when that subcategory is exhausted, switching to a new subcategory. We compared retrieval patterns between groups, and also compared the responses of deaf children who used British Sign Language (BSL) with those who used spoken English. We investigated how semantic fluency performance related to children's expressive vocabulary and executive function skills, and also retested semantic fluency in the majority of the children nearly 2 years later, in order to investigate how much progress they had made in that time. Methods & Procedures: Participants were deaf children aged 6-11 years (N = 106, comprising 69 users of spoken English, 29 users of BSL and eight users of Sign Supported English--SSE) compared with hearing children (N = 120) of the same age who used spoken English. Semantic fluency was tested for the category "animals". We coded for errors, clusters (e.g., "pets", "farm animals") and switches. Participants also completed the Expressive One-Word Picture Vocabulary Test and a battery of six non-verbal executive function tasks. In addition, we collected follow-up semantic fluency data for 70 deaf and 74 hearing children, nearly 2 years after they were first tested. Outcomes & Results: Deaf children, whether using spoken or signed language, produced fewer items in the semantic fluency task than hearing children, but they showed similar patterns of responses for items most commonly produced, clustering of items into subcategories and switching between subcategories. Both vocabulary and executive function scores predicted the number of correct items produced. Follow-up data from deaf participants showed continuing delays relative to hearing children 2 years later. Conclusions & Implications: We conclude that semantic fluency can be used experimentally to investigate lexical organization in deaf children, and that it potentially has clinical utility across the heterogeneous deaf population. We present normative data to aid clinicians who wish to use this task with deaf children.