Russian nouns are inflected for six cases (nom, gen, dat, acc, ins, prep=loc) and two numbers. We wonder which cases are more difficult to process — both in Russian and elsewhere? Previous studies on different languages addressed this question by looking at isolated case forms and found RT differences primarily related to case frequency. But what about case forms in a sentence? In the current study, we ran three self-paced reading experiments comparing Russian case forms after prepositions requiring a particular case. For example, the preposition *k* ‘to’ requires dative — what happens if we use genitive, accusative, instrumental, locative?.. For example, will more frequent incorrect forms be processed faster than other incorrect forms, as it happens in isolation? The pattern we found was completely different. Correct forms are processed significantly faster than incorrect ones, as expected. Otherwise, we found mild grammaticality illusions due to the syncretism of inflections across paradigms of different inflectional classes. Here are two examples with the nouns *sestra* ‘sister’ (1st declension, or D1) and *brat* ‘brother’ (D2): (1) *k sestre*.DAT << *k *sestru*.ACC << other incorrect forms — because *-u* is used in acc in D1 and in dat in D2; (2) *k bratu*.DAT << *k *brate*.LOC << other incorrect forms — because *-e* is used in loc in D2 and in dat (as well as loc) in D1. **Conclusions.** This is a strong argument for morphological decomposition and a very independent status of inflectional affixes in the mental lexicon. Apparently, every syncretic affix activates not only the features it is responsible for in a given word form, but also other feature sets it is associated with in other paradigms. It has been shown before how syncretic affixes can trigger grammaticality illusions, but only for the syncretism inside one paradigm. Similar effects across paradigms have not been previously reported.