A Companion to Chomsky. Группа авторов
of its evolution,” then not only the highly specific constraints mentioned above, for particular morphemes in particular languages, but pretty much every putative expression of ill‐formedness or markedness, needs to be banished from theorizing about UG.
Our own interpretation of the legacy of SPE includes the following components:
1 We assume that phonological UG contains innate features, and that this assumption is in line with what was referred to in SPE as “universal phonetics,” including a set of substantive features that are “independent of any particular language” and that determine “a certain infinite class of possible phonetic representations from which the phonetic forms of sentences of any human language are drawn” (SPE, p. 8). Despite the disrepute into which the universal phonetics view has fallen, we believe that the arguments given in SPE and Chomsky and Halle (1965), remain convincing.
2 Possibly some other representational primitives or structure‐building algorithms to account for syllables, feet, stress, and other phenomena that we assume are also part of phonological UG.
3 Some basic logical operations that define the mappings among the representations. This part of the model can be called a theory of structural changes for the rules in a derivation. In our own recent work (e.g., Leduc et al. Forthcoming; Bale and Reiss 2018), we have deconstructed the traditional SPE arrow and replaced it with set subtraction and unification for a wide range of processes.
4 A theory of environments that determines the conditions under which one part of a representation can trigger a nonvacuous mapping to another representation in a derivation. For example, there appear to be rules in which a consonantal change in a syllable onset is triggered by a consonant in the immediately preceding onset. Could a rule be triggered by any preceding onset at any distance? Many scholars have worked on such questions about locality in phonology. Our own prejudice is that actual segmental adjacency, which is the most typical condition in rules is just a special case of long‐distance interaction.
5 Closely linked to the innateness of features as components of a universal representational code, is the notion, accepted in SPE, of a universal phonetic interpretation of features (and other representational primitives). In addition to participating in phonological representation and computation, features also serve as the basis for phonetic interpretation at the interface between linguistic competence and the sensory‐motor system: “the distinctive features provide a representation of an utterance which can be interpreted as a set of instructions to the physical articulatory system, or as a refined level of perceptual representation” (p. 65).3 This no longer falls under the purview of phonology, and so SPE provides only cursory remarks about the interpretation of features in speech production or perception, suggesting that it is governed by “universal interpretative conventions” (p. 403). In other words, the output of the grammar, the surface or “phonetic” representation, is transduced to speech output in an invariant manner, not specific to any particular language. This position is either ignored or presented as patently false by much recent work, but a full defense is not possible here (see Volenec and Reiss (2018) for details).The issue arises when putatively identical featural representations (using pretty much any current model of phonological primes) are said to have different pronunciations in different languages. To make this concrete, recall our discussion of the vowels [i] and [I] in English and Québec French. In fact, the [i] of petit and the [i] of beet are somewhat different. The universal phonetics perspective of SPE requires that the two vowels in fact be phonologically different, encoded differently in long‐term memory. The rejected alternative (adopted by most other work) is that the two can be (type)‐identical but realized differently by post‐phonological language‐specific rules of phonetic implementation.
6.5 Naturalism in Phonology
The rightful legacy of SPE includes the naturalism, internalism and nativism found also in Chomsky's syntactic work.4 By naturalism, we refer to the idea that language, including phonology, can fruitfully be studied as a natural object (see especially Chomsky 2000). Chomsky has labeled the resistance to naturalistic inquiry of language among certain philosophers “methodological dualism.” As James McGilvray sums up the idea in the introduction to Chomsky (2009b), certain philosophers “might have anything from a reasonably clear to a very good […] idea of what naturalistic scientific methodology is, but they clearly refuse to hold that language and concepts could be investigated using this methodology” (p. 22) and “when it comes to crucial features of the mind, the empiricists abandon not just internalism and nativism, but the methods of the natural sciences” (p. 21).
We detect a related methodological problem within contemporary phonology which also seems tied to neglect and/or ignorance of foundational arguments concerning internalism and nativism. Phonology has seen a number of widely adopted innovations in recent decades, such as the use of more complex graph‐theoretic representations for both segment‐feature relations (e.g., Sagey 1986) and higher‐order units like syllables (e.g., Zec 2007), that are consistent with a trend towards increased rigor and explicitness expected in science. However, phonology has a long way to go before it can be considered a mature science in practice. The rejection of radical internalism has lead to the biggest problem that characterizes much of phonological reasoning since SPE, the notion of markedness, which results in forms of teleological reasoning that, we hold, have no place in modern scientific inquiry. Markedness‐based empiricist phonological theories are trying to be theories about too many things at the same time—not only phonological competence, but also typological trends concerning the distribution of patterns in the languages of the world; verbal behavior factors related to speech rate; inter‐ and intra‐speaker variation related to dialect and register differences; and even the intelligibility of the speech of young children.5 In trying to do too much, such work fails to adopt the normal strategies of natural sciences like isolation, idealization and simplification.
Probably a majority of working phonologists believe that “Many if not most phonological phenomena [are] characterizable in terms of output restrictions” (Tesar 2014), or else that phonological grammars manifest “conspiracies” for “reoccurrence of a common output factor” as Kager's (1999, p. 56) OT discussion summarizes Kisseberth's (1970) early post‐SPE work. McCarthy's (2011, p. 2) discussion of Kisseberth asserts that a conspiracy involving two rules ensures that “[f]inal vowel deletion cannot create bad syllables in surface forms, and epenthesis exists to eliminate” bad syllables. In some work, the purpose is expressed in terms of pathologies: phonological computation has to “cure” a “condition” (Yip 1988). Phonology is said to contain “principles of well‐formedness (the laws) that drive it” (Prince and Smolensky 1993, p. 216), taking input representations and making them somehow better, more harmonic or optimal. Such passages, demonstrate that phonology has not, in general, achieved the rigor of a natural science—naturalism is not explicitly rejected, but such teleological rhetoric makes it clear that it hasn't been universally embraced either.
Markedness and other expressions of “surface‐orientation” in phonology are reminiscent of Appelbaum's (1996) critique of the protean reference to “gesture” in the motor theory of speech perception (e.g. Liberman and Mattingly 1985) and Articulatory Phonology (Browman and Goldstein 1989): “By leaving the referent of ‘phonetic gestures’ ambiguous between an articulatory interpretation and a neural one, proponents of the motor theory try to exploit the theoretical benefit of each interpretation, without incurring the theoretical burden of either.” In a similar vein, surface‐oriented, output‐driven phonology oscillates between actual phonetics (sound and articulation) and the output of the grammar (surface phonetic representations). For example, McCarthy and Prince (1995, p. 88) refer to an OT constraint as the “phonologization of Boyle's law.”6 There is no way that Boyle's law, or the Bernoulli effect or details about tongue and lip position of actual articulation can motivate the computation between input and output