Most current psycholinguistic models assume that the majority of inflected forms are processed as discrete morphemic constituents (Levelt et al., 1999; Pinker, 1999). Recent word recognition studies, however, challenge this assumption by suggesting that the information about individual inflected forms, and paradigmatic relations is available in the long-term memory (Milin et al., 2009; Lõo et al., 2017). To shed light on this issue in production, we conducted a word naming study with Estonian case-inflected nouns. Estonian is a Finno-Ugric language with a large number of inflected forms (800,000 in a 15-million token corpus). Interestingly, these forms do not exist in isolation, but they are organized into extensive inflectional paradigms (up to 50 inflected words sharing a lemma), and morphological families (up to 1,000 derived and compound words sharing a constituent). This organization seems to play a crucial role in Estonian lexical processing. The current study investigated how the whole-word frequency, the size of inflectional paradigm, and morphological family affect word naming, specifically, speech onset time and articulation duration. In Experiment 1, twenty-six native speakers of Estonian (18 female; 21-67 years) read aloud 200 isolated case-inflected words. We found that words with higher frequency, more paradigm and morphological family members were produced faster and in shorter time. In Experiment 2 with 2,800 items and 33 participants (20 female; 22-69 years), similar effects were found. However, they were stronger in general. In summary, the current study is in line with previous studies on paradigmatic effects in lexical processing. It extends our knowledge in three important ways. First, we established that the whole-word frequency effect for inflected forms is present for all words. Second, we replicated the well-established effect of morphological family size in an another domain — production of case-inflected nouns. Third, we documented that a novel paradigmatic measure, inflectional paradigm size, is not specific to word recognition, only. These results are in line not only with linguistic theories like Word-and-Paradigm morphology (Blevins, 2006), but also by usage-based and discriminative models (Bybee, 2006; Baayen, et al., 2011), and indicate that the amount of information about word use in the mental lexicon might be substantially larger than previously assumed.