The main point of the final part of the case study on agent nominals is to show that a generalization in terms of phase theory explains the distribution of Icelandic palatalization in a well-motivated way. In the general case, the process causes an underlying /k/ to be pronounced [c] when immediately followed by /e/ or /i/ (adjacent segments). This phonological contrast is demonstrated by the following pair of examples; further examples of this type will be given in (257) below.
(254) a. ví[k] ‘bay’
The core data to be considered are drawn from the plural form of Icelandic -and
nominals. In VAN like leik-and-i, then morpheme whose phonological exponent is
-and in the singular is -end in the plural. The declension of leikandi is shown in
the singular and the plural for all four cases in (255) below. The plural form of the nominalizer is underlined.
(255) Realization of Root-n-nInfl in leikandi ‘actor’
sing plur
nom leik-and-i leik-end-ur
acc leik-and-a leik-end-ur
dat leik-and-a leik-end-um
gen leik-and-a leik-end-a
Rögnvaldsson (1990b) points out a puzzle that the n-initial /e/ never triggers
palatalization (256) even though /e/ and /i/ are robust palatalizers in the lan- guage. In other environments, /k/ is pronounced [c] when immediately followed by /e/.
(256) a. lei[k]-end-ur ‘actors’
b. dýr[k]-end-ur ‘worshippers’ c. þáttta[k]-end-ur ‘participants’
The analysis to be pursued here is that the phonological representation of the root is not phase-theoretically active when the phonological exponent of n, -end, enters the derivation as a potential trigger of palatalization. Phase inactivity does not mean that the deactivated representation is completely inert for the purpose of PF computation. By hypothesis, the inactive material “has a phonological represen-
be “seen or altered by non-cyclic or phrasal phonological rules, but not by cyclic phonological rules” (Embick 2014). I propose that palatalization is part of the cyclic phonology of Icelandic and thus subject to phase locality.
The diversity of the environments for palatalization can be illustrated by Indriðason’s (1994) Lexical Phonology classification of triggering environments (257).
(257) a. [c]efa ‘give’ vs. [k]af ‘gave’ (Morpheme Internal)
b. ví[c]-ing-ur ‘viking’ vs. ví[k] ‘bay’ (Level I Suffix)
c. fan[c]-elsi ‘prison’ vs. fan[k]-ar ‘prisoners’ (Level I/II Suffix)
d. bó[c]-in ‘book-the’ vs. bó[k] ‘book’ (Suffixed Definite Article)
In the above examples, /k/ alternates between [c]/[k] as determined by the palatal- ization process. One exception is that recent loanwords occasionally resist palatal-
ization. For example, the word dis[k]etta ‘diskette (an obsolete type of electronic
storage)’, did not stay in the language long enough to fully adapt to the native phonology. It is reasonable to assume that those are facts about loanwords rather than palatalization. When we have taken loanword phenomena into account, the generalization is that palatalization is triggered by any suffix with-e/-i which is not
-end. The process iswug-productive and there is neurolinguistic evidence of its pro- ductivity from a patient who suffered a stroke which damaged his ability to palatal- ize, as reported in the Icelandic medical journal (Magnúsdóttir and Sighvatsdóttir 2009).
In contrast, palatalization does not apply across major derivational boundaries such as in compounds (258) and between words (259). Note that if we assume that every root must be categorized by its own category-defining head (Embick and Marantz 2008) and if every category-defining head is a phase head, we expect the individual roots of compounds to be spelled out in distinct phase cycles. This view of compounds is supported by the findings of Shwayder (2015:148-160) who argues that each part of a compound is effectively its own “word” in his analysis of Spanish;
see also Harðarson (2016) for a thorough study of Icelandic compounds.3
(258) a. bla[k]-innviðir ‘volleyball-infrastructure’ b. strá[k]-ermar ‘boy-sleeves’ (259) a. María Mary hitti met strá[k] boy inni í inside bókabúð. bookstore
‘Mary met a boy inside a bookstore.’
b. Jón John spilar plays bla[k] volleyball eftir after vinnu. work
‘John plays volleyball after work.’
The proposed analysis is that Icelandic palatalization requires the target to be phase-active, explaining its underapplication in VAN. This hypothesis is interest- ing and useful because it accounts for the palatalization facts based on locality conditions which are independently motivated.
3There are some exceptional cases in which palatalization is used across compound boundaries, e.g., Indriðason (1994:63) on variable palatalization infis[k/c]eldi‘fish farming’ but these are rare
(260) Analysis
Palatalization requires phase locality.
The idea behind the analysis is not that all phonological processes operate in terms of locality over a representation which requires phase activity. As hypothesized above, non-cyclic and phrasal phonology may apply freely to inactive representa- tions in which morphemes are not identifiable as morphemes and there are indeed many cases in language of phonological interactions between pieces which are not phase-local to each other. For example, contracted auxiliaries in English interact with their host even if the host is clearly spelled out in a different phase cycle (see MacKenzie 2012 for a detailed up-to-date investigation of contraction). Ideally, the part of the phonology which must obey phase locality will have clearly distinguish- able properties and the cyclic phonology is a prime candidate; see Embick (2014).