This paper describes an experimental study that aims at assessing the actual contribution of voice system messages to visual search efficiency and comfort. Messages which include spatial information on the target location are meant to support search for familiar targets in collections of photographs (30 per display). 24 participants carried out 240 visual search tasks in two conditions differing from each other in initial target presentation only. The isolated target was presented either simultaneously with an oral message (multimodal presentation, MP), or without any message (visual presentation, VP). Averaged target selection times were thrice longer and errors almost twice more frequent in the VP condition than in the MP condition. In addition, the contribution of spatial messages to visual search rapidity and accuracy was influenced by display layout and task difficulty. Most results are statistically significant. Besides, subjective judgments indicate that oral messages were well accepted.