The present study deals with the phenomenon of extremely restricted collocability (monocollocability) in Czech and Italian. On the basis of a list of words with the highest degree of monocollocability extracted from corpora with the aid of the so called Herfindahl-Hirschman Index (HHI), we try to analyse this little-explored phenomenon in both languages. Monocollocable words (MW, often referred to as cranberry words) and the fixed combinations in which they occur are investigated in terms of syntactic and collocation structures, frequency and diacronic development. In the light of corpus evidence, the phenomenon of extremely restricted collocability appears more complex than originally believed. The paper shows that MW in both languages form a very heterogeneous category and that many differences between Czech and Italian monocollocable structures can be explained by typological (e.g. the degree of nominal inflection) or historical factors.