It is well known that stress assignment in English noun–noun compounds is non-uniform (compare e.g. left-prominent ópera glasses and right-prominent steel brídge), and recent corpus-based studies (e.g. Plag et al. 2007, 2008) have shown that categorical, rule-based approaches that make use of argument structure (e.g. Giegerich 2004) or semantics (e.g. Fudge 1984) are not able to account satisfactorily for the existing variability. Using data from the corpus studies by Plag and collegues, I argue in this paper that an exemplar-based approach is better-suited to accounting for stress assignment in English noun–noun compounds than a traditional, rule-based paradigm. Specifically, it is shown that two current implementations of exemplar-based algorithms, TiMBL (Daelemans et al. 2007) and AM::Parallel (Skousen & Stanford 2007), clearly outperform comparable rule models in terms of how well they predict stress assignment in the corpora. Furthermore, systematic testing reveals that the reasons for the differences between exemplar and rule models mainly lie in their ability to incorporate detailed, non-abstract information (specifically, constituent family information). The present study therefore adds to the growing evidence in favour of the importance of constituent family information in compounding (e.g. Gagné 2001, Krott, Schreuder & Baayen 2002).

Unlike most other grammatical domains, that of personal pronouns is clearly under-researched in works on grammaticalization. One reason can be seen in the fact that personal pronouns differ in their diachronic behavior from most other grammatical categories to the extent that they present a challenge to grammaticalization theory. In the present paper it is argued that in order to account for this behavior, an extended understanding of grammaticalization is needed.

This paper argues that neutralizing phonological alternations are sensitive to how much homophony they create among distinct lexical items: neutralizing rules create fewer homophones than expected. Building on a case study of Korean by Silverman (2010), I compare the neutralizing rules of Korean to a large number of hypothetical alternatives generated by Monte Carlo simulations. The simulations reveal that the actual rules of Korean frequently create far fewer homophones than similar (but unattested) rules, even when the rules that are compared are controlled for the number of phonemic contrasts they eliminate. These results suggest that phonological patterns are sensitive not only to high-level contrasts among phonemes but also to contrasts among individual lexical items. The effect is most pronounced when homophones are not weighted by frequency, a result that adds to evidence in the literature that the relevant measure of lexical frequency for many lexicon-sensitive phonological patterns is type frequency, not token frequency.

Blocking in inflection occurs when a morphological exponent prevents the application of another exponent expressing the same feature value, thus barring the occurrence of multiple exponents of a single morphosyntactic feature value. In instances of extended exponence, more than one exponent in the same word realizes the same feature value. We provide a unified account of blocking and extended exponence that combines a realizational approach to inflection with Optimality Theory (Realization Optimality Theory), encoding morphological realization rules as ranked violable constraints. The markedness constraint *Feature Split bars the realization of any morphosyntactic feature value by more than one exponent. If *Feature Split ranks lower than two or more realization constraints expressing the same feature value, then we observe extended exponence. Otherwise, we find blocking of lower-ranked exponents. We show that Realization Optimality Theory is superior to various alternative approaches to blocking and extended morphological exponence.