The idea of markedness goes back to the Prague school of linguistics, in particular to N.S.Trubetzkoy. It was originally conceived of as applicable to phonemes standing in a privative opposition to each other, i.e. when one phoneme is differentiated from the other by an additional relevant feature, e.g. voicing, nasalization or rounding. The phoneme which possesses this distinctive mark is called "marked", the other is called "unmarked". Trubetzkoy later supplemented this material criterion of phonemic markedness with a distributional diagnostic: in positions (defined syntagmatically) where the given phonemic contrast is neutralized it is the unmarked phoneme which actually appears. Voiceless obstruents crop up word-finally in German, Russian and Polish, for example.

The descriptive and explanatory potential of markedness was quickly noticed and exploited by other members of the Prague circle, Roman Jakobson in particular, who extended markedness to morphology and semantics (see Batistella 1996 for a thorough review). Only much later was markedness theory successfully applied to syntax. Predictably, new "content" criteria of markedness were added to the classical formal and distributional ones: marked elements are more semantically determinate and elaborate but less prototypical and iconic.

Later, with the growing realization of the heuristic and explanatory fruitfulness of the concept of markedness it was applied to just about every level and component of language, from shorthand (Lipski 1975) to pragmatics (Zuber 1973, Verhagen 1980) and discourse (Fletcher 1985, Sobkowiak 1997), as well as to just about every viewpoint on language, from applied (Eckman 1977, Marquardt et al. 1979) through diachronic (Lass 1972, Vennemann 1972, Enkvist 1979) to anthropological (Greenberg 1987). With the accumulated observations, analyses, descriptions, dependencies, etc. a theoretical edifice was erected from what was once an isolated remark.

The main aim of markedness theory, as seen by most of its proponents and practitioners, is to offer hypotheses on the relative naturalness and commonness of linguistic entities: representations, systems, processes. After appropriate verification the hypotheses serve as a basis for the construction of an evaluation metric for particular grammars on the one hand and the theory of language (universal grammar) on the other (cf. eg. Vennemann 1972 and 1983, Mayerthaler 1988). In both its functions markedness theory closely collaborates with (and, according to some, is indistinguishable from) the theory of language universals.

The process of hypothesis formulation and testing in markedness theory is governed by the widely agreed upon principles of empirical science, and is subject to the equally wide-spread tendency to error. In what follows I will look at some markedness argumentation which appeared in the relevant literature and discuss some of the errors and misconceptions which I discovered there. I will also make some general remarks and pose some unresolved questions with the aim of raising methodological awareness in this important area of linguistics.

While few linguists would now consider going back to Chomskyan absolute determinism and categoricality, the association between nomological and absolute or exceptionless in the linguistic mind is very strong. Consider Lieb (1978:164): "The occurrence of non-probabilistic laws makes the explanation 'nomological'" (my emphasis - WS). And yet, while the formulation of universally true laws (beyond the enumeratively inductive ones) disagrees with the very nature of empirical science, the nomological status of linguistics need not be compromised if probabilistic statements are allowed. Lacey (1986:161) defines 'nomological' simply as "concerning or involving laws", implicitly allowing probabilistic ones in the scope of this definition. While the standard of falsification of probabilistic claims must of necessity be lower than of those boasting full categoricality, it is of course not true that they are in principle unfalsifiable: their falsifiability is guaranteed by the application of stringent statistical procedures.

There are a number of reasons why markedness assignments are seldom, if ever, absolute. One source of the relativity of markedness is theory-internal: it is evident in cases of those oppositions which are not strictly privative in Trubetzkoyan formulation, i.e. where properties are distributed more gradually over the opposing categories: plural is marked relative to singular (both intra- and inter-linguistically), but unmarked relative to dual. A three-term opposition like this would be dubbed "equipollent" in Trubetzkoy's parlance.

Another sense in which one can speak of relativity of markedness values is expressed by Battistella (1990:45) in the following way:

Similarly Dressler (1989:118): "Several parameters of markedness must be considered simultaneously, and they may be in conflict with one another". And Wurzel (1989:229): "grammatical phenomena are not marked purely and simply but always relative to certain of their elements or properties". And Postal (1968:168): "such a theory does not claim that in every case the Unmarked phonetic element will actually appear in the position of neutralization. It claims only that this will be the case in the majority of instances".

Thus, the reasons why typically the diagnostic criteria "will have converging results" (Moravcsik & Wirth 1986:3) but will not necessarily hit the same target have to do both with the probabilistic nature of natural language and the non-existence of a logically coherent, minimal, empirically substantiated and widely accepted set of diagnostics. Metaphorically speaking, then, not only must we use different yardsticks for different linguistic animals (because morphemes have 'content' while phonemes do not, for example), but the yardsticks themselves are not always quite reliable. All this being rather straightforward, I find it difficult to understand how one could seriously formulate a categorical claim like the following: "If a lexical item W_i is more marked than another lexical item W_j, then it must also be semantically more complex" (Kiefer 1989:123; my emphasis - WS).

There are also extra-theoretical reasons for the relativity of markedness values which come into play if we agree - as I believe we should - that both language-specific and cross-linguistic claims are legitimately included in markedness theory. An example of the former would be Chomsky and Halle's (1968) marking conventions predicated over the phonology of English in the famous chapter nine of the Sound Pattern of English. Language universals would be a substratum of the latter. Across-the-board markedness pronouncements informing the evaluative component of universal grammar are thus mitigated by intralinguistic sub- and counter-regularities. For example (cf. Bardovi-Harlig 1987), while preposition stranding (Who did John give the book to?) is a universally marked syntactic construction, its language-specific status in English is - at least with respect to frequency of occurrence - unmarked; and it is this construction which is acquired before preposition pied piping (To whom did John give the book?) by learners of English as a foreign language [1].

The double conditioning of markedness contributing to its relativity is what Batistella (1996:8-9) in his recent monograph calls a "consistency problem": "One aspect of the consistency problem concerns whether markedness relations are universals or language-particular asymmetries or both. In other words, do we determine markedness values - what is marked and what is unmarked - by looking at the facts of a single language or at the facts of many? And if markedness is both language-particular and universal, are we dealing with two different concepts? " (the answer of this writer is: no).

Finally, another reason for the relativity of markedness values: on a micro scale, local inversion of markedness assignments occurs in specific (marked) contexts (Andersen 1972:44, Enkvist 1979:8, Tiersma 1982:832, Mayerthaler 1988:36, Andrews 1990:144 ff). For example, voicing is marked in consonants generally, but unmarked in a marked subclass of sonorants. Notice that here the context is defined paradigmatically (homosegmentally) by reference to a feature of this same consonant. Thus, voicing is unmarked in sonorants, but reversed again to marked in the (syntagmatic) context of a preceding voiceless obstruent: sonorants are devoiced (and liquids spirantized) in smack, trip, clay, sneak. Similarly, the otherwise marked velarity is relatively unmarked in the syllable-final position. Markedness inversion is thus bound to contribute to the overall relativity of markedness values.

In strictly logical terms this fallacy could also be classified as vicious circle in explanation or chicken-and-egg problem (Batistella 1996:8). An apt formulation of the problem comes from Herbert (1986:32), who remarked with reference to Zipf's principle of least effort:

In other words: some sounds are frequent because they are unmarked, and they are unmarked because they are frequent. It is not true, however, as believed by Fenk-Oczlon (1991:373) that "die Feststellung hoher Übereinstimmung [...] lä8 t keinen Schlu8 zu, was von beiden bedingendes und was bedingtes Faktum ist". The vicious circle is immediately broken once we realize that the first "because" is used explanatorily, but the second heuristically, epistemically or argumentatively. In Botha's (1981:188 and 281) words, the first "because" introduces reasons for being, the second - reasons for knowing. The former is used in scientific explanation, while the latter - in scientific discovery or justification. The distinction is nicely summarized in Lieb (1978:165) with reference to the other side of the same coin, the "why":

In the present context we can say that markedness is used to explain observed sound frequency [2], but that frequency is used as a diagnostic to discover or justify markedness assignment. It is only when "because" (and likewise "therefore", "hence", "since", "why", etc.) is used in its explanatory sense in both implicational directions that misunderstanding arises. This reasoning is, of course, valid for all of the other observational variables subsumed under markedness: age of L1 acquisition, degree of syncretism, direction of neutralization, etc.

Notice, for example, how "since" is used by Kiefer (1989:121): "The word dog is semantically unmarked (or neutral), since it can be applied to either males or females". The logical structure of this argument is: "X is unmarked because [X meets one markedness criterion]". The "because" here is quite obviously heuristic in the above sense, but with no explicit remark to this effect, the intended logical force may be unclear and confusion may arise. And it is practically guaranteed in those cases where the wording is ambiguous, as in the abstract of a paper by Allan & Bauer (1991:151): "more frequent, and therefore less marked, segments should have less complex representations, and vice versa" [3]. Or as in the body of this same paper by Allan & Bauer (1991:152-3): "Segment types, processes, and systems which are relatively more frequent [...] are assigned a low markedness value" (epistemic relation here, presumably), but "Those segments, systems or rules which have a low 'cost' are those which are less marked [...] and therefore more likely to recur" (causal relation here, presumably).

Researchers who are particularly prone to collapse the two meanings of "because" are those working in the "autonomous linguistics" tradition. This is because when they reach the stage of argumentation where markedness is an explanandum, there appears to be no extralinguistic law or regularity that could be used as an explanans: after all, linguistics is supposed to be autonomous and must not take recourse to extralinguistic explanation. Consider, for example this quote from Schane (1970:292): "There is no inherent logical reason for the masculine plural form to be used for 'mixed' gender. (I exclude from consideration such extralinguistic explanations as the patriarchic nature of Indoeuropean society.)" [4]. In this situation there is just one way out - to fall back on strictly linguistic regularities as explanatory of markedness [5]. These would have been used as markedness criteria in the first place; hence vicious circle.

In the non-autonomous approach, markedness patterns are ultimately explained by extralinguistic causes (cf. Lieb 1978:199, Vennemann 1983:13): contingencies of human anatomy and physiology (mostly in phonetics and phonology, e.g. Mayerthaler 1982:220, Ohala 1983, but also in semantics), psychology (on all levels; Mayerthaler 1987:27 - "more or less easy for the human brain"), semiosis (from phonology upwards; e.g. Dressler 1989), cognition and culture (from morphology upwards; e.g. Lenneberg 1967, Rosch 1973, 1975, 1977), finally what Cooper and Ross (1975) picturesquely, but quite adequately, refer to as "world order", the overall structuring of our earthly habitat (cf. also Tiersma 1982, Mayerthaler 1987 and 1988). The possibility that "prevalent patterns [...] are not the result of innate limits or pressures to select desirable traits independently applying to many separate languages, but are the result of preservation of traits, possibly quite accidental ones, of a parent language which is ancestral to many or even all of the surviving languages" (Maddieson 1991: 351) is also to be taken seriously in the non-autonomous approach.

Some authors, apparently aware of the danger of vicious circle, choose to refer to the relation between markedness and its criteria in non-directional terms, for example: "The markedness of grammatical phenomena correlates with the complexity of their structure", or "The markedness of individual segments (phonemes and sounds) corresponds to their phonological complexity" (Wurzel 1989:229). While this is admittedly a cleverly non-commital trick, it also smacks of (methodological) avoidance strategy, and may muddle things up even more.

The drive to reductionism is responsible both for some of the greatest achievements and the gravest fallacies of the human mind. Here we will deal with the latter in the context of markedness theory.

It is only natural that with the development of the theory and the explosive growth in the number and range of markedness criteria there should have been attempts to reduce the apparent variety to the more 'basic' or logically primary set.

Lyons (1977:306 ff) chose semantic marking as logically and causally superordinate to both formal and distributional marking: the lexeme more specific in sense (semantically marked) would also be formally (morphologically) and distributionally marked, e.g. lioness, unhappy, disrespectful. Lehmann (1989:176-7) opts for syncretism: "The chief operational criterion in establishing a marking relationship between two elements is an implicational one [...]. Speaking loosely, the unmarked term exhibits at least as much diversification in terms of subcategories as the marked one". Yet, meaning as the main criterion of markedness has obvious limitations, inapplicability to sound structures among them; and the syncretism criterion is conceptually and operationally difficult to apply and tends to generate conflicting results. This is why Greenberg's 1966 choice of frequency seemed - prima facie - a much better solution.

As it happens, text frequency tends to correlate highly with both content and form criteria of markedness [6], and it is rather tempting to reduce them all as far as possible to this single predictor, which is in addition among the ones which are the most objectively measurable. Yet, as noticed (with or without approval, as the case may be) by various researchers, Greenberg

I tend to agree with Greenberg, who ultimately rejects frequency as the criterion of markedness. Notice, however, that the fact that frequency itself is an explanandum rather than an explanans (see above in section 2) in markedness theory is not in principle an obstacle to its functioning as the main markedness criterion. Frequency does require explanation, just like the other variables, but it might still be reasonable to claim that it is more economical to explain one rather than many such variables, however related they should turn out to be. In other words, the autonomy fallacy is independent from the reductionism fallacy. Dressler (1989:111) seems to conflate the two in his well-aimed criticism where he says that one of the "most common misinterpretations of naturalness" is its:

In one of the most deliberate attempts ever made in the linguistic literature to reduce markedness to frequency, Fenk-Oczlon (1991) steered clear of the autonomy fallacy in that she explicitly sought extralinguistic explanantia, such as natural salience and cultural importance, for linguistic frequency. Equally explicitly, however, did she opt for frequency as a substitute for markedness - reductionism fallacy, in my view. In this approach, frequency does not simply correlate (more or less highly, as the case may be) with other (markedness) variables, but takes the place of markedness as the factor causing their co-variation: "Die Häufigkeit stiftet den Zusammenhang zwische den 'Markiertheits-Indikatoren'" (page 384; my emphasis - WS). The passage from correlation to causation in Fenk-Oczlon is swift and unreflective. Notice, for example, the difference between the non-directional claim from Bybee (1985:119), quoted by Fenk-Oczlon (page 382): "The correlation of irregularity with high frequency can be documented in almost any language" and Fenk-Oczlon's own causally directional 'Hypothese': "In häufigen Kategorien ist eher mit Irregularität zu rechnen" (ibidem). Where Fenk-Oczlon's certainty about this unidirectionality comes from is not clear, especially in view of the lack of pertinent literature. A quote from the introduction to a standard collection of papers on markedness (Moravcsik & Wirth 1986:8):

The reasons why I reject the reduction of all markedness criteria to frequency have to do with the theoretical status of markedness as a cover term for a number of related properties of some linguistic entities (see section 4). As aptly observed by Andersen (1989:30), if one defines markedness in terms of frequency only, "one merely empties the concept of all meaning other than "relative text frequency", and the term might as well be dispensed with" (similarly Brasington 1982:86 and Batistella 1996:15 and 51). It is only when markedness extends over logically independent, but empirically related, variables that it is at all explanatorily and heuristically fruitful. Compare Cairns & Feinstein (1982:195):

Notice that at the theory-construction stage, where bridges are built between linguisitc explananda and extralinguistic explanantia, frequency - an observational variable - cannot oust markedness as a theoretical concept, Fenk-Oczlon's claims notwithstanding. And vice-versa - although Fenk-Oczlon only mentions this possibility parenthetically, it is a methodological error to believe that "Diese Rolle einer unabhängige Variable könnte das theoretische Konstrukt der Markiertheit auch schwerlich übernehmen" (Fenk-Oczlon 1991:387), especially if (and here I agree) "'Markiertheit' ist zwar ein Name für eine gelungene Kategorisierung oder Abstraktion" (a few lines down on the same page).

An approach somewhat similar to Cairns & Feinstein is taken by Winter (1989:106-107) when he claims that in his view: "there is no need for believing that "unmarked" in the sense of "simpler in form" should be coextensive with "more normal/more natural", thus rejecting the reductionist temptation. Similarly - my own, slightly oversimplified, example - a label like obstruent can only be useful in phonology if its meaning is not identical to, say, 'prone to word-final devoicing' (even if all segments prone to word-final devoicing should turn out to be obstruents).

There are other reasons why frequency appears to be a rather dangerous candidate for the sole predictor of markedness. As noticed by Schwartz (1980:319), for example, the reduction of the contextual neutralization criterion of markedness to text frequency, as considered by Greenberg, is open to the type-token fallacy because: "a marked member of a phonological opposition can accidentally appear in morphemes which will be of extremely high frequency for non-phonological [...] reasons", for example the voiced interdental fricative in English. A similar observation was made by Battistella (1990:26): "wider distribution does not simply mean having a greater frequency in the language than the opposed category, since the frequency of a token depends on such factors as lexical, grammatical, and discourse function".

The type-token fallacy is itself a reflection of a theoretically more interesting flaw of the frequency-reductionism approach. In the words of Tomić (1989:192): "one has to distinguish between grammar (or code) and text (or message); while the frequency phenomena belong to the latter, all the other manifestations of markedness obtain in the former". If we now choose text frequency as the markedness criterion we are (implicitly) claiming that an aggregate of grammatical (in the sense of code-related) properties of linguisitc entities are directly contingent on (one of) their textual properties. While we may not wish to exclude this effect from our linguistic theory - and indeed there is ample evidence that text-to-code causation exists in language - it is doubtful that it is so overwhelming as to grant it such unrestrained power. Interestingly, in Fenk-Oczlon 1991, one of the most forceful arguments for frequency as the main determinant of markedness, this line of criticism is not even acknowledged, let alone analyzed and refuted.

Apart from the intralinguistic frequency (both token and type, both text and code), there is extralinguistic frequency. In their influential paper of 1983, Witkowski & Brown claim that "over-all salience of referents determines marking values of their labels by affecting label frequency" (570). Notice the staunch determinism oblivious to such complicating pragmatic factors as taboo, for example. Sex, death and defecation are socio-culturally and doubtless also psychologically rather salient, but not very frequent as conversation topics. More generally perhaps, Zwicky (1978:141) emphasizes that "There is a compound here [...] of at least two factors, both quite independent of language: the relative frequencies of certain beings, objects, events [...], and the relative degrees to which certain things, objects, and so on are significant or salient to human interests". The relationship between the different types of extralinguistic frequency, as well as between them and the (different types of) linguistic frequency may be very complex, too complex to be captured by simple deterministic claims.

In cases where there is strong correlation between world- and language-frequency (like the Maya deer and sheep discussed by Witkowski & Brown) it is not clear with reference to which frequency we should try to explain the observed congruence of other markedness criteria. In a non-autonomous approach this would ultimately be world-frequency, but Witkowski & Brown opt for the other: "Terms of highly salient referents are used more frequently; and this in turn affects other features of marking, e.g. brevity and acquisition" (my emphasis - WS). Fenk-Oczlon does not even seem to notice the difference.

Further, while the frequency criterion is applicable to most of the asymmetrical arrangements of linguistic elements, rules or constraints which are the explananda for markedness theory, there are areas where it is not, and yet markedness assignment can be made on the basis of other heuristics. In her 1989 paper on binomial freezes Fenk-Oczlon admitted (page 525) that: "the rule we posit for binomials [more frequent first - WS] is neither applicable to nor testable against [non-coordinate freezes]: 'criss' or 'splish' do not exist on their own and therefore cannot be more or less frequent than 'cross' or 'splash'" [7]. In my paper on freezes (Sobkowiak 1993) I was able to show that "unmarked-before-marked" is a better supported principle of freeze organization, also for coordinate freezes. In another paper on the markedness of communicative silence (Sobkowiak 1997) I demonstrated that most markedness criteria intercorrelate quite highly to render silence heavily marked in communication, but I would have been hard pressed to reduce this co-variation causally to one variable - frequency.

Incidentaly, frequency may have been erroneously interpreted in the linguistic literature as theoretically and practically unproblematic: what can be easier than calculating averages and standard deviations? And yet, the true picture is much more subtle than that, as has been alluded to above: there are world frequencies, cognitive (mental, psychological, etc.) frequencies and linguistic frequencies; there are type and token frequencies; there are objective and subjective frequencies, there are written and speech frequencies (Batistella 1996:51), there are "different frequencies in different subdomains, genres, or corpora" (ibidem). And then, there are those linguistic and communicative entities, like the components of non-coordinate freezes and conversational silence(s), which cannot easily be counted at all. The whole issue deserves a serious linguistic and statistical treatment.

Finally, argumentum ex praxis: while there have been a number of attempts to reduce markedness to frequency, it seems that none of them quite succeded. Had they succeded, one could indeed expect the gradual phasing out of markedness as a concept and a complete overhaul of the whole markedness theory. The admitted failure of Greenberg was mentioned above. Schwartz, despite a rather apt attempt, is forced to conclude (1980:332): "Further examination is required to determine whether the remaining predicates of 'neutralization', 'dominance' and 'par excellence expression' are independent and independently sufficient to characterize markedness relations in syntax". And Mayerthaler (1988:101) notes with relation to diachrony: "If markedness values and frequency criteria conflict [...], then frequency fades into the background as a parameter that controls analogy, i.e. analogical change runs counter to frequency criteria".

Much of the confusion affecting markedness arguments appears to be due to what I call here the realism fallacy whereby markedness is hypostatized into an objectively existing property of linguistic entities of all sorts. Consider the following quotation from Bardovi-Harlig (1987:387): "In a case where the more common form is also the marked form, will markedness or salience (in this case, the availability of data) determine acquisition order?". Notice that markedness and frequency are equated here in terms of their theoretical status, as if they were equivalent properties of (in this case) syntactic constructions. This, I believe, seriously misrepresents the highly abstract status of markedness, which is an aggregate derived from a number of empirically observable linguistic variables, such as frequency, for example. Markedness is a theoretical construct with no ontological force, similar to obstruent (to use the example introduced above). In yet other words: "Markedness values do not belong to the language of observation but to the language of theory" [8] (Mayerthaler 1988:7).

The realist interpretation of markedness is partly strengthened by the convenient parlance of markedness argumentation: "this or that linguistic state of affairs obtains because it is the less marked", or "Considering that most of the effects of markedness ... " (Tiersma 1982:847; my emphasis - WS). Seldom is it remembered that this is not much more than a shorthand way of referring to a number of intricately connected parameters (properties, variables, criteria). Result: treating salience (frequency) as "a second factor" which "also plays a role in determining acquisition order" (Bardovi-Harlig 1987:353), on a par with markedness - a misrepresentation of an obviously hyponymic relation.

Or: Fenk-Oczlon's (1991:363, 373 ff) juxtaposition of markedness/naturalness with frequency and her bold question: "Was ist Henne, was Ei?", a question which only makes sense if the two have comparable ontological status, which they do not. I am tempted to say that if frequency (and other exponents of markedness) should be hen or egg, then markedness will count as Gallus domesticus.

Or: Andrews' (1990:129) "counterexamples to Greenberg's eight criteria for markedness", one example of which is that "the imperfective (unmarked - WS) in Russian occurs 46.9 percent, the perfective 53.1 percent in usage [...] thus, here is one example of the marked category occurring more frequently than the unmarked". Notice that Andrews does not bother to say "morphosyntactically marked", thus creating an erroneous impression that (a) markedness (as such) is an autonomous dimension rather than an aggregate of values, (b) misalignment of one of these values (here - frequency) categorically negates whatever markedness assignment has been obtained for a given opposition.

Or: Kiefer's (1989:122) observation that "the range of applicability should be considered to be a consequence of marking rather than one of its defining properties", where in its first occurrence "marking" is used realistically, while in its second ("its") - nominalistically, and the (spurious) conflict is resolved in favour ("rather than") of the former.

Of course, if marked(ness) is no more than "a cover term for a vague categorial assymetry in which one element dominates its opposite" (Batistella 1996:15), it makes no sense to use it as an explanans of, say, frequency, otherwise than - again - as a convenient façon de parler to shorten the unwieldy "frequency of this representation (process) is low because (first): this representation (process) is at the same time: (1) neutralized out of existence in most contexts, (2) non-prototypical, (3) structurally complex, (4) non-iconic, (5) universally disfavoured, etc.", and (second): "all these properties are jointly explained by the extralinguistic property X" (this was briefly mentioned in footnote 4). Similarly, "because it is an obstruent" would at best furnish a first (analytic rather than synthetic) approximation to an explanatory statement in answering a question like: "Why does /z/ devoice word-finally in Polish?".

Granted that markedness in itself "provides no explanations at all" (Smith 1981:53), it does not follow that "it seems to make no empirical predictions which are not false or untestable" or that "markedness, too, is superfluous". It is not superfluous if understood as an integrating conceptual aggregate of empirical observations, as explained above. And it is in this sense that it can make interesting predictions. After all, to use my own example one more time, once we know that a sound has many features of an obstruent, we can predict some of its other features or processes it is likely to undergo. Natural classes in phonology may be ontologically void, but are definitely heuristically fruitful.

To take the final example from the other end of the linguistic spectrum: in a bout of rather wild hypothesizing closing my paper on the markedness of communicative silence, I formulated the following tentative predictions where markedness functions as a projectans/explanans with respect to a number of projectanda/explananda (Sobkowiak 1997):

This unsurprisingly hypothetico-deductive approach to markedness, I believe, is one which is (more often than not, implicitly) adopted by most practicing linguists and markedness theorists. Among the latter, Battistella takes it in his 'concluding remarks on markedness patterns' (1990:149):

• Allan,W.S. & L.Bauer. 1991. "Markedness, markedness inversion, and dependency phonology". Australian Journal of Linguistics 11.151-74.
• Andersen,H. 1972. "Diphthongization". Language 48.11-50.
• Andersen,H. 1989. "Markedness theory - the first 150 years". In O.M.Tomić (ed.). 1989. Markedness in synchrony and diachrony. Berlin: Mouton de Gruyter. Pp. 11-46.
• Andrews,E. 1990. Markedness theory: the union of asymmetry and semiosis in language. Durham: Duke University Press.
• Bardovi-Harlig,K. 1987. "Markedness and salience in second-language acquisition". Language Learning 37.3.385-407.
• Battistella,E.L. 1990. Markedness. The evaluative superstructure of language. New York: SUNY Press.
• Batistella,E.L. 1996. The logic of markedness. Oxford: Oxford University Press.
• Botha, R.P. 1981. The conduct of linguistic inquiry. The Hague: Mouton.
• Brasington,R.W.P. 1982. "Markedness, strength and position". In D.Crystal (ed.). 1982. Linguistic controversies. London: Edward Arnold. Pp. 81-94.
• Bybee,J.L. 1985. Morphology. A study of the relation between meaning and form. Amsterdam: John Benjamins.
• Cairns,C.E. 1986. "Word structure, markedness, and applied linguistics". F.R.Eckman, E.Moravcsik & J.Wirth (eds). 1986. Markedness. New York: Plenum Press. Pp. 13-38.
• Cairns,C.E. & M.H.Feinstein. 1982. "Markedness and the theory of syllable structure". Linguistic Inquiry 13.2.193-225.
• Chomsky,N. & M.Halle. 1968. The sound pattern of English. New York: Harper & Row.
• Cooper,W.E. & J.R.Ross. 1975. "World order". In R.E.Grossman, L.J.San & T.J.Vance (eds). 1975. Papers from the parasession on functionalism. Chicago: Chicago Linguistic Society. Pp. 63-111.
• Dressler,W.U. 1989. "Markedness and naturalness in phonology; the case of natural phonology". In O.M.Tomić (ed.). 1989. Markedness in synchrony and diachrony. Berlin: Mouton de Gruyter. Pp. 111-20.
• Eckman,F.R. 1977. "Markedness and the contrastive analysis hypothesis". Language Learning 27.2. 315-30.
• Enkvist,N.E. 1979. "'What' and 'why': on causal explanation in linguistics". Folia Linguistica 13.1-21.
• Fenk-Oczlon,G. 1989. "Word frequency and word order in freezes". Linguistics 27.517-56.
• Fenk-Oczlon,G. 1991. "Frequenz und Kognition. Frequenz und Markiertheit". Folia Linguistica 25.361-94.
• Fletcher,Ch.R. 1985. "The functional role of markedness in topic identification". Text 5.23-37.
• Greenberg,J.H. 1966. Language universals. The Hague: Mouton.
• Greenberg,J.H. 1987. "The present status of markedness theory: a reply to Scheffler". Journal of Anthropological Research 43.4.367-74.
• Herbert, R.K.1986. Language universals, markedness theory, and natural phonetic processes. Berlin: Mouton de Gruyter.
• Itkonen,E. 1977. "Causal models in linguistics". In Papers from the Conference of General Linguistics, Lammi 22-23.9.1977. Turku: Publications of the Linguistic Association of Finland.
• Kiefer,F. 1989. "Towards a theory of semantic markedness". In O.M.Tomić (ed.). 1989. Markedness in synchrony and diachrony. Berlin: Mouton de Gruyter. Pp. 121-38.
• Lacey,A.R. 1986. A dictionary of philosophy. London: Routledge.
• Lass,R. 1972. "How Intrinsic is content? Markedness, sound change, and 'family universals'". Revue Roumaine de Linguistique 17.5.415-40.
• Lehmann,Ch. 1989. "Markedness and grammaticalization". In O.M.Tomić (ed.). 1989. Markedness in synchrony and diachrony. Berlin: Mouton de Gruyter. Pp. 175-90.
• Lenneberg,E. 1967. Biological foundations of language. New York: John Wiley and Sons.
• Lieb,H.-H. 1978. "Universals and linguistic explanation". In J.H.Greenberg (ed.). 1978. Universals of human language. Vol 1. Stanford: Stanford University Press. Pp. 157-202.
• Lipski,J.M. 1975. "Features, markedness and shorthand". IRAL 13.3.229-38.
• Lyons,J. 1977. Semantics. Cambridge: Cambridge University Press.
• Maddieson,I. 1991. "Investigating linguistic universals". In Proceedings of the 12th International Congress of Phonetic Sciences, Aix-en-Provence, vol 1. Pp. 346-54.
• Marquardt,T.P. et al. 1979. "Markedness analysis of phonemic substitution errors in apraxia of speech". Journal of Communication Disorders 12.6.481-94.
• Mayerthaler,W. 1982. "Markiertheit in der Phonologie". In Th.Vennemann (ed.). 1982. Silben, Segmente, Akzente. Tübingen: Niemeyer. Pp. 205-46.
• Mayerthaler,W. 1987. "System-independent morphological naturalness". In W.U.Dressler et al. (eds). 1987. Leitmotifs in natural morphology. Amsterdam: Benjamins. Pp. 25-58.
• Mayerthaler,W.1988[1981]. Morphological naturalness. Ann Arbor,MI: Karoma Publishers, Inc.
• Moravcsik, E. & J.Wirth.1986. "Markedness - an overview". In F.R.Eckman, E.Moravcsik & J.Wirth (eds). 1986. Markedness. New York: Plenum Press. Pp. 1-11.
• Ohala,J.J. 1983. "The origin of sound patterns in vocal tract constraints". In P.McNeilage (ed.). The production of speech. New York: Springer Verlag. Pp. 189-216.
• Postal,P. 1968. Aspects of phonological theory. New York: Harper and Row.
• Rosch,E.1973. "Natural categories". Cognitive Psychology 4.328-50.
• Rosch,E. 1975. "Cognitive representations of semantic categories". Journal of Experimental Psychology, General 7. 573-605.
• Rosch,E. 1977. "Human categorization". In N.Warren (ed.). 1977. Studies in cross-cultural psychology. London: Academic Press. Pp. 1-49.
• Schwatrz,L.J. 1980. "Syntactic markedness and frequency of occurrence". In T.A.Perry (ed.). 1980. Evidence and argumentation in linguistics. Berlin: Walter de Gruyter. Pp. 315-33.
• Shane,S.A. 1970. "Phonological and morphological markedness". In M.Bierwisch & K.E.Heidolph (eds). 1970. Progress in linguistics. The Hague: Mouton. Pp. 287-94.
• Shapiro,M. 1972. "Explorations into markedness". Language 48.343-64.
• Smith,N.V. 1981. "Consistency, markedness and language change: on the notion 'consistent language'". Journal of Linguistics 17.39-54.
• Sobkowiak,W. 1993. "Unmarked-before-marked as a freezing principle". Language and Speech 36.4.393-414.
• Sobkowiak,W. 1997. "Silence and markedness theory". In A.Jaworski (ed). Silence: interdisciplinary perspectives. Mouton de Gruyter. Pp. 39-61.
• Sparkes,A.W. 1991. Talking philosophy. London: Routledge.
• Tiersma,P.M. 1982. "Local and general markedness". Language 58.4.832-49.
• Tomić,O.M. 1989. "On the assessment of the markedness status of the exponents of a grammatical category". In O.M.Tomić (ed.). 1989. Markedness in synchrony and diachrony. Berlin: Mouton de Gruyter. Pp. 191-205.
• Vennemann,T. 1972. "Sound change and markedness theory: on the history of the German consonant system". In R.P.Stockwell & R.K.S.Macauley (eds). 1972. Linguistic change and generative theory. Bloomington: Indiana University Press. Pp. 230-74.
• Vennemann,T. 1983. "Causality in linguistic change. Theories of linguistic preferences as a basis for linguistic explanations". Folia Linguistica Historica 6.1.5-26.
• Verhagen,A. 1980. "Pragmatic markedness and syntax". Journal of Pragmatics 4.5.425-47.
• Winter,W. 1989. "Markedness and naturalness". In O.M.Tomić (ed.). 1989. Markedness in synchrony and diachrony. Berlin: Mouton de Gruyter. Pp. 103-9.
• Witkowski,S.R. & C.H.Brown. 1983. "Marking reversal and cultural importance". Language 59.569-82.
• Wurzel,W.U. 1989. "Inflectional class markedness". In O.M.Tomić (ed.). 1989. Markedness in synchrony and diachrony. Berlin: Mouton de Gruyter. Pp. 227-47.
• Zuber,R. 1973. "Remarks on markedness and presupposition". Linguistische Berichte 26.1-6.
• Zwicky,A.M. 1978. "On markedness in morphology". Die Sprache 24.129-43.