It is often claimed that clitics are unaccented in Czech (e.g. Palkova 1994: 280). It would be incorrect, however, to conclude that clitics are inherently prosodically deficient, as some authors do (e.g. Rivero 1991, Fried 1994, King 1996, Franks & King 2000). Rather, as this papers shows, accent for all words is equally nondistinctive in Czech. The prosodic properties of words (including clitics) is a by–product of discourse–related word order and phonological weight. If this is true, analyses based on the prosodic deficiency hypothesis are ruled out.
Jakobson observes the stress is automatic in Czech. Stress is assigned to the first syllable of the word, as in (1).
(1) a. nouns: 'sva&chachek;ina, 'chlebichek, 'pomazanka, 'dzhus
b. 'pect, 'zrat, 'mazat, 'pit
Clitics, on the other hand, may be stressed or unstressed. Stressed is the unmarked case. Either way, no contrast results (Palkova 1994, PMC 1997).
(2) a. 'Vchera jsem mu ho 'predstavila. (PMC 1997: 649)
b. 'Vchera 'jsem 'mu 'ho 'predstavila.
Shmilauer (1972) and Travnichek (1951) describe metrically–based clitic accentual rules in which stress is automatically assigned to the fifth or seventh mora in the sentence. For example,
(3) Priblizhi–li 'se mi na dostrel.(Shmilauer 1972: 310)
If he comes within range
The key observation is that—under the proper conditions—clitic can be stressed. This stress, however, is automatic (conditioned) and nondistinctive.
The sort of prosodic variation is not unique to clitics, however. Phonological words can be unstressed too. Compare the data in (1) with their prefixed or post–prepositional counterparts in (4).
(4) a. nouns: 'na svachinu, 'na chlebichek, 'do pomazanky, 'za dzhusem
b. 'upect, 'sezhrat, 'namazat, 'vypit
In each case the stress moves off the word onto the preceding syllable. More than that, this variation occurs with one and the same wordform as well, as je and to in (5a–b) reveal.
(5) a. 'Je to 'jeji 'vina. (Palkova 1994: 281)
b. 'To je 'jeji 'vina.
Neither word is a clitic (both occur stressed and sentence initially). The remarkable thing is that the very same words are unstressed in 2P! I conclude that prosodic properties of words are a reflex of discourse–related word order, and not a lexically inherent property per se. The role played by word length is investigated as well.
A proper understanding of the grammatical status of prosody in Czech is not trivial, different analyses with different predictions emerge as a result. Analyses which postulate the inherent prosodic deficiency of clitics lack descriptive inadequacy and cannot account for the correlation of stress and word order. Analyses which derive prosodic representations from word order, on the other hand (e.g. Avgustinova and Oliva 1995, Veselovska 1995) are formally motivated and arrive at the prosodic facts for free (short, pragmatically unmarked words are unstressed).