I undertake an empirical analysis of Norwegian compounds with special focus on those that correspond to single words in Russian that are not compounds. Although in many cases the Russian single words have the same meaning as the Norwegian compounds, we frequently encounter semantic shifts of two types, which I refer to as “hyponymy” and “metonymy”. I argue that hyponymy is the default option, but that metonymy is preferred for certain types of compounds, e.g. those where the head represents a part of the non-head (e.g. kirkegolv ‘church floor’) or the head denotes a quantity of the non-head (e.g. melkedråpe ‘drop of milk’). It is furthermore suggested that the choice between hyponymy and metonymy is motivated by the desire to minimize loss of information; while hyponymy normally involves a smaller loss of information, metonymy appears to minimize information loss for certain types of compounds. Finally, I relate my findings to the trade-off between informativeness and economy in language and hypothesize that this trade-off is treated differently in languages like Norwegian, where compounding is a central word-formation mechanism, and Russian, where compounding plays a more modest role in word-formation.