Processing differences between Serbian prepositions and nouns were investigated in two lexical decision experiments. The outcome of the experiments indicate that processing of nouns presented in their citation form is affected primarily by the amount of information (bits) derived from noun's probability. Processing of prepositions appeared to be more complex because, in addition to probability of preposition, processing latency is also affected by number of syntactic functions/meanings carried by a given preposition and number of congruent noun cases. None of these factors per se is correlated with processing latencies. However, when combined into a single unit (expressed in bits) significant proportion of processing latency variability has been accounted for. It was shown that prepositions are processed faster than nouns of equivalent frequency and equivalent length. We conclude this on the basis of the observed differences in slope from linear regression, but also on the basis of differences in averaged RTs for the two word types.