Syntactic lexicons, which associate each lexical entry with information such as valency, are crucial for several natural language processing tasks, such as parsing. However, because they contain a rich and complex information, they are very costly to develop. In this paper, we show how syntactic lexical resources can be merged, in order to take benefit from their respective strong points, and despite the disparities in the way they represent syntactic lexical information. We illustrate our methodology with the example of French verbs. We describe four large-coverage syntactic lexicons for this language, among which the Lefff, and show how we were able, using our merging algorithm, to extend and improve the Lefff.