The Sound Pattern of English, SPE
Although most of its main theses have long been abandoned by both authors, SPE remains a milestone in linguistics. It is a testament to its quality and depth, that admirers and detractors alike refer to it by the acronym of its title rather than, more standardly, by the names of the authors and the date of publication. One SPE thesis which has not been abandoned is the notion of a ‘possible word’. SPE gives the example of BNIK as a completely impossible word in English, though possible in many languages, and almost possible in Greek and Russian.
The work of three of my child friendly notables critically exploited the notion of possible words, without using anything like this term. The idea was taken much further in SPE. This was on the basis of new evidence from a large number of languages and their histories, almost all unknown to European and American scholars until the late 19th and early 20th centuries.
Like Holder before them, but using a new and much more precise terminology, Chomsky and Halle propose that the phonemes are defined by the features, rather than the features being just their taxonomic attributes. This order of definition is still crucially relevant for the study of children’s speech.
By a taxonomic approach, as espoused by the International Phonetic Association, there are eleven points at which the airflow can be constricted, the vocal chords, the lips, and nine others in between. Recognising that no language exploits more than a few of these, typically no more than three or four, SPE proposes a more abstract but more parsimonious system, differentiating the TH phoneme in thin, by a negative valuation with respect to a feature which it calls ‘strident’, contrasting with a positive valuation with respect to F in fin, S in sin, SH in shin.
By the SPE system, features are either on or off. This was, as it remains, a quite reasonable assumption, given that this is how the nervous system works. Nerves are either activated or not.
Both the parsimony of the system and the fact that is essentially binary help describe both what happens and what doesn’t happen in incompetent speech, as shown in the case of S and Z. More generally, SPE makes it possible to describe a large range of what are commonly called ‘processes‘ – like saying key as TEA by reversing the values of what it calls [Anterior] and [Coronal] – ‘fronting’, or say as TAY by reversing the value of [Continuant] – ‘stopping’.
SPE gives the first full account of English word stress, as a universal property of all English words, obvious in those with two syllables or more, and as much a part of an English word as the phonemes. Before SPE, it was often said that English stress varied randomly, as in Canada, loaned from Mohawk and Iroquois, spaghetti from Italian, and vindaloo, from various South Indian languages, previously from Portuguese, with the stress on the first, second and final syllables respectively.
Aa one aspect of what is known as the ‘metricality’ of the segments or phonemes, in contrast to their ‘segmentality’, word stress has complex effects across the whole range of speech acquisition up until the age of seven or eight.
Until SPE, the basis of the metricality remained mysterious. John Thelwall described it in terms of ‘cadences’, stretching the meaning of the musical term. Alexander Melville Bell suspected that an account was possible. SPE provided the first such account.
A lost legacy
Chomsky and Halle did not know of the work of their child friendly predecessors because of some shenanigans in the early 1900s. But thanks to SPE, there is now lively research in all of SPE’s main areas, all matters of great concern for clinical linguistics, particularly the duality of metricality and segmentality.
The great defect of SPE is that it describes too much. In many cases it would be even easier to describe things which do not happen or only happen very rarely than it is to describe things which are very common. For instance, key by ‘fronting’ as TEA is common. But tea as KEY is rare. And the lip action in pea is barely involved at all. Of six logically possible substitutions, all easily defined by reversing the values of two features, only one is commonly attested. This is not perspicuously captured by a reversal of two feature values, rather than one. As shown by Nunes (2002), in a way consistent with the hypothesis here, there is a more perspicuous account of fronting by a failure of structure building.