During a survey carried out in Madagascar the use of photographs to support interviews (photo-elicitation interview) generated a number of « misunderstandings » that revealed themselves to be eventually « productive » for the understanding of the object of study. This shows the heuristic fruitfulness of what could be rapidly described as « methodological errors ». From this perspective of a reflexive analysis the author discusses the methodology proposed by John Collier (1967) and Dona Shwartz (1989) for this survey technique.