The present study examined whether speakers avoided the production of lexically ambiguous target labels in a referential communication task (e.g., avoiding the ambiguous bare homophone mouse when pictures of both a computer mouse and a mammal mouse were present in the referential display), and whether this avoidance was increased after either direct or indirect prior experience with a different set of lexical ambiguities as a listener. Participant speakers described target pictures in ambiguous displays and unambiguous displays to an experimenter matcher partner. Ambiguous displays contained pictures representing both alternative meanings of a homophone noun (e.g., computer mouse and mammal mouse, plus two unrelated filler pictures). Unambiguous displays contained only one interpretation of the homophone noun (e.g., computer mouse, plus three unrelated filler pictures). Prior to this, participant speakers gained either direct experience (completing the same task with a different set of ambiguities, fulfilling the role of matcher) or indirect experience (watching a video of two naïve participants completing the task) or no prior experience (control), using a between-participants design. Responses were coded as bare homophones (eg., mouse) or modified homophones (e.g., computer mouse) and analysed using a mixed-model ANOVA. Avoidance of bare homophone target labels on ambiguous trials was taken as a measure of lexical ambiguity avoidance. As predicted, participants produced somewhat fewer bare homophone labels for targets in ambiguous displays (72%) than for targets in unambiguous displays (85%). This main effect of trial type (ambiguous versus control) was significant by speakers and by items, reflecting a degree of lexical ambiguity avoidance, but this effect was not mediated by prior experience as an interaction between prior experience and trial type did not reach statistical significance. It is suggested that speakers may be somewhat egocentric, failing relatively often to monitor the referential situation for ambiguity and to consider the comprehension needs of an addressee. Instead, speakers may favour the production of ambiguous forms, which are less cognitively demanding to produce. Keywords: lexical ambiguity; language production; referential communication; prior experience