The creation of nicknames in Czech, a productive category of word-based morphology, reveals properties of prosodic organization which are otherwise not overtly expressed in Czech. Although Czech is characterized by complex consonant clusters and a free distribution of vowel length, in some hypocoristics we find restrictions on syllable structure and on the distribution of quantity. Thus, while /Vladimir/ provides a straightforward base for /Vla:d'a/, /Antoni:n/ does not have the nickname *Ant'a but instead has one with a different base: /To:nša/. Both formations show vowel lengthening in the first syllable, and the lengthened vowel is transparent in that the contrast of length is /o/ with /o:/ and not so much the historical quantity alternation of /o/ and /u:/. Some names have variants: /Toma:sh/ has both /Tu:ma/ and /To:ma/. Significantly, there is no vowel lengthening in the first syllable of the following a-suffixed nicknames: Gusta, Berta, Vojta, Jindra, and others.
I argue that the formation of Czech nicknames is output-oriented and governed by constraints on preferred prosodic profiles with respect to both syllable structure and prosodic feet. The constraints are motivated by phonetic principles of articulation and perception as well as by phonological markedness restrictions. This paper thus argues against a traditional derivational rule-based grammar and against a templatic view of hypocoristics (McCarthy and Prince 1986; Scullen 1993, and others) in favor of a constraint-based grammar such as Optimality Theory (Prince and Smolensky 1993; McCarthy and Prince 1993, Kager 1998). These formations are of a different type than the hypocoristics discussed in Benua (1997) under Output-Output constraints and are therefore of interest to theory as well.