The redundancy of self-organization as an explanation of English spelling

Kristian Berg and Mark Aronoff (2017; henceforth B&A) use historical data on the spellings of various English suffixes to argue for the role of ‘self-organization’ as a process by which system emerges spontaneously in an unregulated linguistic domain. They argue that choices of spelling for given phonological suffixes gradually settled down in such a way that spellings became indicators of part of speech. This is an attractive idea, but B&A’s data fail to support it, because they do not adequately consider the availability of other explanations. Data cannot constitute evidence for a novel theory, if they are already convincingly accounted for independently of that theory. This, I believe, applies to most if not all spellings cited by B&A. Many are explainable in terms of phonological detail, or of the nature of education during the relevant period; some have more specialized explanations.

B&A discuss four cases: the suffixes -ous, -ic, -al, and -y, together (in each of the four cases) with suffixes that are phonologically similar but spelled differently. To keep the length of this commentary within bounds, I do not deal exhaustively with every example word cited by B&A. I begin by showing that for their first case, -ous, all of their data are convincingly accounted for by explanations independent of ‘self-organization’. Then from B&A’s three other cases I pick out examples that introduce considerations different in kind from those already discussed.

I do not claim to be able to ‘explain away’ every last one of B&A’s data. But that is to be expected, considering that I do not have (and perhaps no one has) an ideally complete knowledge of the evolution of English spelling and English pronunciation (particularly the pronunciation of vowels in unstressed syllables). However, the proportion of B&A’s examples not accounted for by any of the principles I discuss is not large enough, I believe, to leave the case for self-organization persuasive.

B&A introduce the -ous case by displaying a table (their Table 2, p. 45) that lists alternative graphic patterns for words that they regard as sharing a common phonological suffix,¹ which they transcribe as /ɨs/.

(1)

<ous>	as in	hazardous, nervous
<us>		bonus, genius
<is>		glottis, tennis
<ess>		hostess, princess
<ice>		office, service

B&A note (p. 41) that the CELEX database used for their research makes a phonetic distinction between [ǝs] as in nervous and [ɪs] as in office, but they see that as not relevant because ‘[t]his distinction is not justified phonetically in many varieties of English’; they suggest, citing Flemming and Johnson (2007), that in American English the distinction only occurs in the special case of pairs like Rosa’s, roses. However, what matters for B&A’s argument is not the phonology of any present-day English dialect, but that of the language during the period when spelling conventions were still evolving. (For -ous and all but one of the other phonological suffixes discussed by B&A, the endpoint of the period they consider is 1710.) One cannot assume that this resembled present-day American English in the relevant respect, considering that in standard British Received Pronunciation (RP) the five graphic patterns listed above correspond to at least three different phonological patterns. A. C. Gimson (1967) gives hazardous, nervous, bonus, and genius with /ǝs/; glottis, tennis, office, and service with /is/; princess with /es/ (thus not wholly unstressed); and hostess with /is/ but /es/ as a ‘less frequent’ variant. Failure to distinguish the reduced vowels of tennis and office from those of hazardous and nervous is unusual in Britain, as demonstrated by the fact that it is a striking feature of the speech of Tony Blair, prime minister from 1997 to 2007, which otherwise conforms to RP norms.²

B&A’s decision to ignore the contrast between [ǝs] and [ɪs] would be justified if American English reflected the original position in this case and RP usage were innovative. For some of the obvious phonological differences between the two dialects, that is so, but here the truth...