| Abstract [eng] |
In the present thesis, we present a novel phonological universal, which we name the XYZ effect. The effect in question is foremostly related to the phonotactic well-formedness of words and pseudo-words. In any specific language variety, the XYZ effect is said to hold, if sampling any set of three words with the shape {XY, YZ, XZ} implies the well-formedness of the (pseudo-)word XYZ. For instance, in English, the set of words {XY, YZ, XZ} = {smell, mellow, sow (the verb)} can be hereby “recombined” into XYZ = °smellow. The words in the set {XY, YZ, XZ} need not be meaningful; they can also be well-formed pseudo-words. The goal of the thesis is to verify the cross-linguistic universality of the XYZ effect. To this end, we test it against various typical phenomena of local phonotactics and against the harmony systems in Navajo, Slovenian, Yaka, and Nyangbo. To serve as the only proxy for these phenomena, we use their descriptions in grammars and articles. The outcome of our verification is mostly favorable: the XYZ effect can be preliminarily considered to be universal. To corroborate this conclusion, it would serve well to verify the effect against more data sets that are encountered in the verification of other models of phonotactics. However, it must be noted that one of the ways of interpreting the Yaka data already leads to a counterexample for the XYZ effect. Additionally, we construct an algorithm for the learning of phonotactics, based on the XYZ effect and several other minimal assumptions. Some early tests of this algorithm have also been favorable, but a more exhaustive analysis of its inductive capacity would require an additional study. |
