Eight experiments investigated the role of semantic and associative relatedness and short-term memory in speech planning. Subjects produced a conjoined noun phrase corresponding to picture pairs that were semantically related only, semantically and associatively related, and associatively related only. Interference effects, i.e., longer onset latencies were observed for neurologically intact undergraduates and elderly controls in both semantic conditions but were not apparent in the associative only condition. Interference effects only occurred under the following conditions: (1) The degree of semantic relatedness was high between picture pairs. (2) Each picture was presented several times. (3) Related and unrelated matched pairs occurred in the same block and (4) Several pairs appeared from the same category. A phonological STM patient showed interference effects for semantically related pairs within the range of elderly controls. However, a semantic STM patient showed mixed results across several experiments. His inconsistent performance pattern was attributed, in part, to long term priming effects resulting from repeated testing with the same materials. Overall, the results suggest that the lexical-semantic information for both pictures was planned prior to production, and that semantic and associative relatedness differentially affect planning in speech production.