Where traditional dialectology usually asks only such questions as, 'What does the vowel of this word sound like at each point under investigation?', structural dialectology goes on to ask, 'How does this vowel sound fit into the phonemic system of each point?' Examples taken from the Sprachatlas der deutschen Schweiz illustrate how this second, structural question can add new insights not only to dialectology but also to general phonological theory, both synchronic and diachronic. Dialects are one of our great linguistic laboratories. In order to exploit them to the full, we must no longer regard the field worker merely as a recording machine (he does that badly, and has higher values) and his informant merely as a talking machine. Even so simple a matter as determining (with the informant's help) whether two words rhyme will both correct recording errors and yield valuable structural information.

It is not generally recognized that near-homonymy may, like homonymy, be a causative factor in language change. Relevant instances are cited from the Germanic languages where near-homonymy has resulted in homonymy contrary to laws of regular sound change, thus proving the causative influence. It then becomes possible to accept similar causation in cases where near-homonymy is followed by loss of a word, as happens frequently with homonymy. Finally, implications of the effects of near-homonymy for the concept of distinctiveness in language are considered.

In Asturian and Central-South Italian dialects, there is a category of collective nouns, pronouns, and definite articles which is usually called ‘neutro de materia’ and considered a continuation of the Latin neuter. This category is not a neuter in the traditional sense of the term, but represents a grammatical category hitherto unrecognized in Romance, the mass-noun, and is to be traced to an (also unsuspected) survival of the Latin ablative, through its use in partitive constructions, first with and then without dē ‘of’.

This study proposes a transformational analysis for Spanish adverbial clauses of time, place, and manner. Sentences such as He arrived when I left are treated as reduced versions of sentences like He arrived at the time when I left. This latter sentence, in turn, is viewed as an alternate form of the following: He arrived at the time at which I left (the Spanish equivalent of this sentence is somewhat less artificial sounding than the English). This last sentence is then derived from two simple sentences (I left at some time; he arrived at the time) by application of a relative clause transformation.

As a result of analogical change, a bundle of phonetic features may arise which occurs simultaneously only in a given morphological category. Such a segment can be accounted for by a low-level morphologically conditioned phonetic rule. A subsequent phonetic change may normalize the segment and delete the first rule from the synchronic grammar. While the diachronic grammar will include both phonetic rules, the resultant change will appear in the synchronic grammar as a change in the morphophonological structure of the underlying form. This principle is illustrated by the anomalous nasal vowel letter attested in certain Old Church Slavonic participles, and by the fate of the velar stop in adjectival and pronominal inflection in northern Russian dialects.

The importance of the morphophoneme as a structural unit has not been fully recognized. Post-Bloomfieldians explicitly rejected this unit and methods of description making use of it. This paper contends that morphophonemes are indeed required for the most efficient description of morpheme alternants. It is further argued that morphophonemes are real phonological units and not simply abstractions fictions. These points are illustrated by a problem in Tera, a Chadic language of northern Nigeria. In order to account for word-final vowel alternations, some /i/'s must be interpreted as the morphophoneme |ə|, others as |i|. Surprisingly, allophones of preceding consonants are variably determined by these underlying morphophonemes.

The following alternative analyses of the Korean vowel system are compared: (1) By the view of traditional phonemics, we recognize nine vowels /i e ø ε ə a u o/, plus autonomous /w y/. (2) By distinctive feature analysis, regarding w and y as on-glide realization of features [+round] and [+palatal], we recognize nineteen vowels: i e ε i ə a u o, wi we (= φ) wε wa wə, ve vε və va vu vo. (3) By the view of generative phonology, we recognize four vowels—high , low a, mid rounded o, and mid unrounded ə—with w and y as on-glides, and with y as a fronting off-glide. (4) If we extend the analogy of off-glide y, and recognize a backing off-glide w, then we have three vowels, ə a, and two distributionally free glides, w and y. It is argued that alternative (3) gives us the truest and simplest picture, when considered in conjunction with morphophonemic phonomena, and especially in the light of the generality that several phonological rules gain by this analysis. It is hoped that this paper illustrates the possibility of choosing the ‘descriptively adequate’ solution from among several competing, non-unique, ‘observationally adequate’ solutions.

Similarities between Aymara and contiguous Quechua dialects have been considered by linguists simply as evidence of borrowing. This study, based on the application of standard comparative procedure to 520 Quechua sets and 255 Quechua-Aymara sets, gives evidence rather for genetic relationship between the two languages. A section on isoglosses clarifies the relationship between the northern and southern areas of the Quechua-speaking territory and proposes an older and important dialectal stage in central Peru.