Good, evil, their balance and what they tell us about language


Philippe Ségéral (University Paris 7) and Tobias Scheer (University of Nice)


It seems to us that one point of convergence between Natural Phonology and the kind of Government Phonology that we are doing is a very broad one: the existence of two basic forces that counterbalance each other. That is, existing patterns in language are viewed as an equilibrium state that has arisen through the interplay of these forces, and diachronic change is the consequence of a change in prominence among them. In Natural Phonology, this description corresponds to the interplay of higher principles such as the law of the least effort vs. the need to produce contrast/ to be expressive. These principles then are implemented into linguistics and produce cross-linguistic Preferences that are scalar and violable such as "CV is the best/ most natural syllable, anything that departs from it will be less frequent/ more marked, diachronic evolution and acquisition strive toward CV." The area of Government Phonology that bears a similar conflict resolving system is consonantal lenition/ fortition. We have proposed in the Coda Mirror (Ségéral & Scheer 2001) that a consonant potentially experiences two distinct and antagonistc forces, i.e. Government (the evil one) and Licensing (the good one): Government inhibits the melodic expression of its target, while Licensing backs it up. This system defines four and only four logical combinations, i.e. 1) [+gov, +lic], 2) [+gov, -lic], 3) [-gov, +lic], 4) [-gov, -lic]. 1) = intervocalic consonant, 2) = impossible, 3) = the Strong Position = word-initially and after a Coda, 4) = in Coda position. The effects of the resulting positional plight or ease are various and gradual: cross-linguistic observation allows to assign gradual strength/ weakness the four positions mentioned: consonants are the strongest in 3) (the Coda Mirror) and the weakest in 4) (the Coda), 1) (the intervocalic position) being intermediate, but more often on the weak side of 4). Note that this scale is not absolute but relative: we do not claim that there can never be any lenition in 3) (the Coda Mirror), nor that there can never be any fortition in 4) (the Coda). These are events that may occur, but we do not expect them to be frequent. What we do firmly rule out is that any higher ranked position on the strength scale does undergo some lenition, while lower ranked positions do not. Or, that any lower ranked position experiences strengthening, but the higher ranked ones do not.


Now let us look at the potential differences between the Natural Phonology approach and our own scenario. 1. we do make hard predictions and define the conditions under which our system is falsified, cf. above. For example, we hold that a language where consonants spirantize in Codas and in the Strong position, but not intervocalically, is not human. We do not see how any logically possible pattern can ever be ruled out by the kind of higher principles of Natural Phonology that have been mentioned. That is, the balance may go way towards the expected natural centre of gravitation (strict CV languages), but a hypothetical language where words must begin with at least 5 consonants will not be declared impossible. Rather, it will be viewed as improbable. 2. this leads us to another difference: there does not seem to be any such thing as the notion of "possible language" in Natural Phonology. Linguistic structures are not possible/ well-formed or impossible/ ill-formed, but just more or less probable: this is the very essence of what a Preference is. This property is shared by OT ("I assure you, this constraint exists in the brain of every human being, but unfortunately you can never see it in any language because it is so low ranked"). In other words, nothing is ever discrete, linguistic structures are only gradual. By contrast, we have shown how the two antagonistic forces Government and Licensing build a network which is 1) predictive, 2) discrete and 3) splits the logically possible area into "human" and "non-human". 3. we believe that the non-discreteness, the non-predictiveness and the absence of conditions of falsification in Natural Phonology stem from its non-linguistic foundations. Dziubalska-Kolaczyk (2001:73) writes: "in Natural Linguistics linguistic preferences are explanatory since they are derived from non-linguistic levels", and p.74 "Preferences of Natural Linguistics are explicitly based on extralinguistic fundaments." And indeed, the higher principles invoked which govern the linguistic implementation of Preferences make no reference to linguistics at all: being natural means being "cognitively simple, easily accessible (especially to children), elementary and therefore universally preferred, i.e. derivable from human nature, or unmarked/ less marked" (Dressler 1999:135). We hold that any scientific theory must be predictive and falsifiable. Moreover, we believe that the human brain is not equipped for handling graduality: all objects that are manipulated by our cognitive system are discrete. One aspect of the Chomskian research programme is to explain how "infinite states can arise from finite means". 4. the foregoing paragraph actually provides good illustration why Natural Phonology falls outside of the generative paradigm: in the philosophical debate between empiricists and mentalists, it stands on the empiricist side. That is, "nothing in language is specifically linguistic". Rather, linguistic behaviour is but one facet of more general human behaviour, i.e. the strive toward naturality. Therefore, Natural Phonology has to deny the existence of any specifically linguistic endowment that is transmitted by the human genetic code: infants do not build on any linguistic knowledge in order to acquire natural language. Rather, the capacities that they use in order to acquire language are of more general, non-linguistic kind (induction, categorization etc.). They are the same as those that are used for the acquisition of bipedal movement or the ability to drive a car. Thereby, Natural Phonology takes the behaviouristic stance, in both acquisition and adult practice of natural language. 5. but Natural Phonology does not only fall outside of the generative paradigm. It appears that it is also incompatible with the Saussurian insight that has grounded modern linguistics. Saussure holds that there is a level of analysis (Langue) that obeys exclusively its own rule and remains entirely unaffected by any external influence that may arise through social pressure etc. (Parole). Not only does such a level that is purely lingusitic and obeys only the linguistic rule exist, but the study of language, says Saussure, cannot possibly succeed if this level, i.e. the Langue, is not studied independently and before all other things. The Chomskian equivalent, of course, is competence/ I-language vs. performance/ E-language. Now the higher principles that control the content of Preferences in Natural Phonology clearly fall into the Parole, if anything: they are non-linguistic. It thus appears that Natural Phonology denies the existence of the Saussurian opposition Langue vs. Parole since nothing in language is Langue, and hence no such thing is genetically encoded, which means that infants do not use any such information in order to acquire natural language.


We realize that we may have misconstrued some of the aspects of Natural Phonology as stated above, and we would be happy to learn through discussion in Poznan in which way we are mistaken or imprecise.



Dressler, Wolfgang U. 1999. What is Natural in Natural Morphology (NM)? Prague Linguistic Circle Papers 3, 135-144.


Dziubalska-Kolaczyk, Katarzyna. 2001. Phonotactic constraints and preferences. Constraints and Preferences, edited by Katarzyna Dziubalska-Kol''aczyk, 69-100. Berlin, New York: Mouton de Gruyter.


Ségéral, Philippe & Tobias Scheer. 2001. La Coda-Miroir. Bulletin de la Société de Linguistique de Paris 96, 107-152. Older English version available at