Journal of Slavic Linguistics

Professor Gussmann presents the phonology of Polish here in the framework of Government Phonology. The range of phenomena covered by his GP phonology is somewhat limited because most of what happens in Polish, such as "[t]he massive alternations involving palatals and palatalized segments" (32), he relegates to morphophonology, a grammar component that was swallowed up by generative phonology 40 years ago and which he seeks to restore to its previous status. I shouldn't say "happens" because in GP, which is monostratal not derivational-generative, nothing is supposed to happen; words are merely found to be in compliance with a number of constraints. To those of us who think of phonology as something that happens to morphemes, who regard allomorphy as the primary data, Gussmann says no: "Allomorphs are units of morphology, and while they may be a useful tool in the search for and identification of phonological generalizations, they do not define the data for analysis" (61).

Two phenomena that do fall under phonology are the nonoccurrence in Polish of [kɨ] and [gɨ] and of word-initial [ɨ].² Understanding how Gussmann accounts for these facts requires some familiarity with GP. Without attempting a systematic exposition of the postulates of GP as they were originally proposed in Kaye et al. 1985 and applied to Polish in Cyran and Gussmann 1999, I will attempt to briefly set forth a few of them that are relevant. First, the syllabic structure of words is not directly related to their phonetic realization. For example, we hear ich 'them' as the VC monosyllable [ix]. But words in GP are structured as a sequence of O(nsets) and N(uclei), so in ich [i] is preceded by a consonant onset (which happens to be empty) and [x] is followed by [End Page 131] an (empty) vowel nucleus to which it relates as onset. Next, the segments of a word are a combination of elements such as I (frontness in vowels, palatality in consonants), A (openness in vowels, coronality in consonants), and U (rounding in vowels, labiality in consonants). A segment may have one of the elements as its head, or it may be headless. Polish has no [kɨ] or [gɨ] because velars are headless (they lack A, i.e., are not coronal, and U, i.e., are not labial), [ɨ] contains the I element but is not headed by it; it is headless (by contrast, [i] contains I and is headed by it), [ɨ] cannot follow [k] or [g] because of the Empty Heads constraint: "An empty-headed nucleus cannot license an empty-headed onset" (52; "license" here means "follow"). To remedy this, the I of [ɨ] is promoted to its head, turning it into [i]. In addition, because "[a] nucleus shares I-head with the onset it licenses" (52), I is associated with the velar and turns it into a palatal, as in ptaki 'birds' and rogi 'corners'. Next, the reason no word can begin with [ɨ] is that the empty consonant that precedes a vowel-initial word qualifies as headless, so a word-initial [ɨ] violates Empty Heads. Hence the I of [ɨ] is promoted to head, turning it into [i].

Also a matter of phonology is the fact that the masc.-neut. inst. sg. ending -em causes fronting in stem-final velar stops, e.g., in ptakiem and rogiem, but has no effect on stem-final labials (chlebem 'bread') or dentals (bratem 'brother'). The vowel [ε] contains both A and I elements, but in inst. sg. [εm] [ε] is headless. Since velar stops are headless, the Empty Heads constraint applies and I gets associated with them as above. But labials and dentals are U-headed and A-headed respectively, and this keeps them from being palatalized by the I element.

However, my suggesting that 'bird' is underlyingly /ptak/ and that /k/ shifts to [k']³ before the ending -em may be a distortion of Gussmann's monostratal account. He writes: "[T]he notion of any underlying representation distinct from the surface is incoherent within our model. There are no phonological structures from which phonetic shapes are derived...