Abstract:French neoclassical compounds X-(o)phobie display two distinct meanings : 'fear' and 'hostility'. In order to determine which meaning is involved in a given complex word, X-(o)phobie lexemes are studied according to the contexts they are used in. In fact, in a non-compositional analysis, complex words cannot be studied in isolation since they construct their meaning whenever they are employed. The distributional hypothesis offers a new way to examine the semantics of complex words. Two methods make it possible to investigate the variation of meanings in X-(o)phobie lexemes. Moreover, the use of Google data enables an analysis of the construction of neologisms meaning.