Notes on Bybee (2000) and on Phillips (2001) Steve Fondow: On Phillips 2001 -- The author revises a previous hypothesis regarding the lexical diffusion of sound change to come up with the "Frequency Implementation Hypothesis," which is the major topic of this paper. This hypothesis indicates that sound change which requires analysis affects less frequent words first while sound change that requires no extra analysis will occur in the most frequent forms first. I assume that changes with no extra analysis are "regular" sound changes (i.e. no syntactic, morphological or non-systematic? phonological analysis, such as "straightforward assimilations or reductions"). Examples include the alveolar and palatal sibilant alternation of speakers from Georgia (e.g. street ~ strait') and Southern American English alternations of voiceless palatal affricates and alveolar stops (e.g. Tuesday ~ tumor). The primary explanation for the variants is that the higher frequency forms are less parsable or less easily analyzed segmentally and are typically treated as having "less segmental integrity." Additionally, word class is more important to sound change than word frequency. Three types of sound change are identified on the basis of word frequency and word class. The first change affects high frequency words and function words first and requires the least amount of analysis (i.e. most lacking in segmental integrity). The second affects function words first and high frequency words last, thereby necessitating greater lexical analysis. The last type of change is not well defined but affects function words last and should therefore affect high frequency words first. Word class is shown to be crucial to the diffusion of sound change if the sample is large enough to exemplify differences in frequency effects across word classes. Also, it can be shown that word frequency effects are also identifiable within word classes. Phillips attributes this to the possibility that speakers reference word class before they reference the phonology of a form. Another important factor in sound change is "neighborhood density." Words that are tightly packed into a lexical neighborhood are more likely to undergo lexical analysis. In this way, subsets of word classes may be more resistant to change since the y are analyzed in a different way than other forms within the same class. Questions 1. What defines word classes? How generic or specific can they be? 2. How independent are function words from high frequency words? Aren't function words typically high in frequency? 3. I made an assumption in my summary that there was an error in the discussion of word classes and frequency effects (131-132). The first section talks about the salience of frequency effects within word classes and shows examples across classes and the following discussion also mentiones the effects within word classes and presents corresponding examples. Did I misread this? Bybee 2000 This analysis discusses language use and its effects on sound changes and lexical alternations through environmental effects. Data of phonological variation are used to discuss the way that language is stored in memory. The first part of the discussion involves the diffusion of sound change in correlation to word frequency (i.e. high frequency forms are the first ones affects by sound change). The second section discusses the ways in which different phonological environments are involved in sound change (i.e. if a sound change is expected or not and why it may or may not occur in certain forms). The theoretical basis for discuss ion is an exemplar model theory that adjusts the representations of forms in the lexicon through the direct effects of every individual input/use of said form. This is related to the notion of entrenchment whereby high frequency forms have stronger representations, are more easily accessible, are more resistant to change and are more capable of serving as basic for a derived form. Forms are stored in the lexicon as words or units larger than words and their frequency of use determines their possible reduction or compression ("automation effect"). Higher frequency forms have a higher rate of change and factors such as speech style and introduction of the referent (i.e. whether it is the first time a form is mentioned in discourse or it has been repeated) also have a direct link to the automation effect. Alternations as discussed here are forms which occur in complementary "phonological, grammatical or lexical" environments. Examples of such alternation include Spanish syllable final s aspiration/deletion. This sound change is shown to have various stages represented by different dialects in which this phenomenon occ urs. Also, it demonstrates the extension of the conditioning environment (through a shift in the stored representations through use) and the effects of word final position as opposed to word internal position (indicating a 'word level phonology'). The resolution of this alternation is gradual but shows directionality in favor of the more frequent variant. Nonetheless, it is possible for an alternation to resist resolution if it is phrasal, that is, if the form is stored lexically as part of a highly frequent group of words or if there are phonetic environments that trigger and do not trigger a sound change in the same morpheme ("The Timberlake Effect", 259). In this case, frequency of the alternants (or their cateogories?) will predict the direction of possible resolution. The network model discussed in the following section demonstrates that morpholog y emerges in large part from the phonological and semantic neighborhoods by whic h forms are associated. The high frequency of certain morphemes will strengthen their representation and result in high productivity (a sort of generalization through association). Stable examples of variation (of either words or phrases) , as in phonologically similar but semantically unrelated forms, occur when a sound change is conditioned by one form but not the other. The problems with a Lexical Phonology model are treated and show that it may make poor predictions about sound change and that it cannot account for frequency effects since it is a structural model. Questions 1. Does the overall view of the article indicate that sound change is simply reduction or compression of forms and that anything else is analogy or some other process? 2. Why is it that all alternations need be resolved? And why should it be in favor of one alternant or the other? 3. What prevents the minimal unit of lexical storage from being a morpheme? 4. What constitutes a phrase that is "common enough" to permit lexical storage? (258) 5. How does a phonemic split occur? 6. Does my investigation on velar insert alternations in Old Spanish relate to the conclusion that "old underlying forms never resurface, even when the 'phonological rule' becomes unproductive"? (267) David Alexander: On Phillips (2001) -- The spread of a sound change is as much a product of the lexical class of words it affects than sociolinguistic or merely phonological factors. In this light, more frequent words appear to undergoe change more readily than less frequent words (ex. the stress shift for verbs of type dictate). However, several counter cases complicate matters in which a change can be seen to spread from the less frequent words to the most frequent ex. stress shift in nouns versus verbs- co'nvict vs convi'ct). Phillips argues that such discrepancies can be explained while still preserving the Frequency-Actuation Hypothesis in that across the board changes without regard to lexical class proceed from more frequent to less frequent items in the lexicon whereas changes affecting specific lexical classes usually begin in less frequent tokens and spread from there. Lexical items of greater frequency tend to undergo automation and rote processing, often becoming single unanalyzable units as in God be with you > goodbye. In contrast, less frequent words such as the noun convict undergo lexical analysis. In this case the noun convict undergoes a stress shift differentiating it from the verb form. Such changes proceed along lexical lines and therefore cannot first take hold in items which are more frequent and therefore less integrated into the lexical neighborhood. The propogation of certain sound changes from one word to another directly depends on other members of the same lexical class, as well as the level of morphological and lexical analysis an item undergoes in the mind of the speaker. How is one to define analysis? Is it that the speaker thinks about a network of words in syntactic, morphological, phonological, or lexical terms? Thereby providing a kind of model and analogy through which a change can be spread? Does analysis also have to do with processing? The less processing required the less analysis and a lesser likelihood for a change to spread lexically? However, in that less processing is related to frequency, a change that first affects the most frequent items then totally lexically blind?