This paper reports on the annotation and maximum-entropy modeling of the semantics of two German prepositions, mit (‘with’) and auf (‘on’). 500 occurrences of each preposition were sampled from a treebank and annotated with syntacto-semantic classes by two annotators. The classification is guided by a perspective of information extraction, relies on linguistic tests and aims at the separation of semantically transparent and opaque meanings (that is of collocational constructions). Apart from descriptive statistical material, we present results of experiments using monolingual and multilingual evidence (the latter from informative English and Spanish translations) in order to predict the semantic classes.