|Title:||Patterns and Categories in English Suffixation and Stress Placement: A Theoretical and Quantitative Study|
|Comment:||Doctoral dissertation submitted to the University of Tsukuba, Japan, 2012|
|Length:||248 + x pp.|
|Abstract:||The goals of this thesis are twofold. One is to describe as extensively as possible the behavior of English suffixes in affixation; in particular, the properties of the bases to which they attach, and the stress pattern they exhibit. The second goal is to provide a theory which properly accounts for the behavior of the suffixes, and predict the proportions of each stress/word-formation pattern in the English lexicon.
An extensive investigation is first carried out utilizing a CD-ROM version of the Shorter Oxford English Dictionary. This provides numerical data regarding word properties resulting from suffixation. The survey focuses on (i) whether the base of the suffix is a bound root or not ('root attachability'); (ii) whether base stress is preserved ('stress preservation'), and (iii) the stress pattern. These data are considered to be of some importance, as no quantitative investigation of this scale with respect to these properties has been carried out before.
The rates of the first two properties are used to determine a suffix's 'classhood'. Although each of the suffixes had been previously assumed to fall into one of two classes, our theory predicts the existence of four, given two independent variables: root attachability and stress preservation. Groups of suffixes belonging to each class are in fact observed in the present investigation: that is, each suffix falls into one of the four 'classes'. Moreover, the diversity of stress patterns suggests that there are distinctions even among suffixes which belong to the same class. In short, the results of the investigation show that a traditional dichotomic analysis is not adequate to describe the varying behavior of English suffixes.
The variation among suffixes can best be analyzed in the framework of Partial Ordering Theory, first proposed by Anttila (1997). Within the framework of Optimality Theory, this theory assumes that the grammar of a language consists of rankings of constraints. These are only partially determined, however. The remaining undetermined parts of constraint ranking can be specified for an individual suffix or a group of suffixes. Thus, each of the lexical classes -- four of which are considered, rather than the traditional two -- can be analyzed as being assigned different rankings. At the same time, potential variation in stress patterns can properly be limited to those actually observed; i.e. they are the only patterns that will arise in a factorial typology of available constraints. It will be shown that some of the constraints do not interact to produce any phonological patterns at all.
The investigation (iii) further showed that stress patterns are not always categorical for a suffix as is widely assumed in the literature. In other words, some suffixes show alternative stress patterns, while others do have categorical ones. The difference in the degree of categoricity can be attributed to the specificity of partial rankings: suffixes belonging to the first group are assigned to less specific ones -- hence have more options in their stress patterns -- while members of the second group are assigned to more specific rankings, thus showing a more consistent pattern. The specificity of individual suffixes will be shown in 'grammar lattices', where available partial rankings are listed in a tree diagram.
The investigation indicates that lexical classes and stress patterns are distributed unevenly: some classes and stress patterns are more likely to occur than others. Quantitative predictions for these kinds of facts can also be made within the present theory by calculating the proportion of partial rankings, i.e. those which produce each lexical class and stress pattern. As some of the interactions of available constraints do not produce a meaningful morphological/phonological pattern, each class and stress pattern can result from different numbers of constraint rankings. Assuming that each of the rankings has an equal chance of being picked by a suffix, our theory predicts that the proportion of rankings for a particular class/pattern will match that of the suffixes belonging to the relevant class/pattern itself. Strikingly, the observed proportions quite closely match our predictions in most cases. This is seen as a positive result, since no previous studies have provided predictions with the same degree of accuracy.
|Article:||This article has been withdrawn.|