The notion of markedness and its role in pronunciation acquisition

The notion of markedness has been defined in a variety of different ways. The three most widely cited definitions of markedness (i.e., the criteria serving as the basis for evaluating a category as more or less marked than another one) are as follows:

- The definition based on implicational relations: In any linguistic system, the presence of a marked²² category (a sound segment or a structure) implies the presence of less marked categories but not vice versa. (In other words, focussing on how to define

21 The reason why a separate section is devoted to markedness is that it is not only highly relevant in the field of SLA in general, but it will have a fundamental role in some of the discussions in Chapters 5 and 6.

22 For the sake of simplicity, the terms “marked” and “unmarked” will sometimes be used in the thesis as if they were discrete categories. In reality, markedness is a scalar property, therefore whenever a category is labelled

“marked” throughout the dissertation, it is to be interpreted as “more marked (than another category)”, and

“unmarked” as “less marked”.

markedness: if the presence of a category implies the presence of another one, it is more marked.)

- The aspect of frequency of occurrence: Marked categories are less common in languages than unmarked ones are (or if a category exists in numerous languages, while another one is rare, the latter is called more marked than the former).

- The aspect of language acquisition and language impairment: Unmarked categories are acquired earlier in the course of L1 acquisition than marked ones (or if a category takes longer to acquire than another one, it is said to be more marked). Similarly, unmarked categories are retained the longest in the case of language impairment (such as aphasia).

The most widely cited examples illustrating the above definitions (which can therefore be regarded as the “classic” examples) are as follows: As for individual sound segments, there are languages which only have three vowels (viz. /i/, /u/ and /a/). More complex vowel systems also include these three; in fact, these three vowels can be found in the vowel inventory of nearly all languages, thus they are the most frequent and least marked vowels of languages. The presence of certain more marked vowels in a system implies the presence of certain less marked ones, but not vice versa – a detailed account is not relevant to us here and is therefore beyond the scope of the discussion, but for example, a language that has /y/ also has /u/ and /i/; one that has /ø/ also has /e/, etc. (though the former example is less interesting because /i/ and /u/ are one of the three vowels mentioned above that can be found in all vowel systems). It inevitably follows from this that the acquisition of a more marked vowel in a language is more likely to cause problems for non-native speakers of the language in question, simply because the vowel inventory of the learner’s L1 may be less complex than that of the L2/FL and may not contain the vowel that is to be acquired.

Still with regard to sound segments, there exist languages whose segment inventory does not contain voiced obstruents, only voiceless ones, while no language has voiced obstruents only and no voiceless ones – voiced obstruents are thus more marked than voiceless ones. In a number of languages such as German or Polish, voiced obstruents do exist, but they do not occur in word-final position: a phonological rule systematically changes voiced obstruents into voiceless ones word-finally, this is dubbed final obstruent devoicing or simply final devoicing – the unmarked form displaces the marked one, which does not happen vice versa. The same conclusions can be drawn based on the third definition of markedness: in the course of language acquisition (in both L1 and L2 acquisition), voiced obstruents are acquired earlier/easier word-initially and word-medially than word-finally (it is easier to pronounce a voiceless consonant word-finally than a voiced one).

Another often-cited example concerns syllable structure. The most unmarked syllable type is CV: there exist languages (such as Hawaiian) in which CV is the only syllable type. In languages allowing more marked syllable structures (CVC, V and VC), the presence of a more marked type implies the presence of the less marked one (though it is again not interesting if the less marked structure implied by a more marked one is CV, since CV is the least marked syllable structure and exists in every language).

Introducing such markedness phenomena into the field of SLA is credited to Eckman’s (1977) famous Markedness Differential Hypothesis (MDH), which served as a partial solution to the criticism voiced against CA (cf. the previous section), as one of the arguments against the CA approach was that it is unable to account for the varying degrees of difficulty in terms of the acquisition of different pronunciation features. The MDH filled this gap as its main observation was that unmarked categories are acquired easier (and therefore earlier) than marked ones. Of course, this is far from being this simple – difficulty does not always go hand in hand with markedness, because beside markedness, similarity also plays a role in shaping how difficult a target language category will be. The details of how exactly the four possible permutations of markedness and similarity can predict difficulty is beyond the scope of the present discussion as it is too complex and not even relevant for our purposes.

Apart from accounting for differences in difficulty, markedness grew to be a central issue in at least one more area of SLA: it is able to explain some learner errors that are not part of either the L1 or the L2/FL in the contact situation. Cases where markedness (or rather:

unmarkedness) accounts for an error unattributable to L1 transfer has been called “The Emergence of the Unmarked” (TETU, cf. McCarthy & Prince 1994). Such instances have been discussed in the field of Hunglish: Altenberg & Vago (1983: 433) observed final obstruent devoicing and stressed vowel lengthening²³ in their Hungarian participants’ accent of English, and grouped these examples under the category “unmarked rule application”. Relevant examples of the former include ban[t] increases, en[t] of in one of their subjects’ accent, and en[t] in sentence-final position²⁴ pronounced by the other subject. No specific examples of the

23 There is a third example they discuss in their paper, but it has been deliberately omitted from the list. This third example is two words (sun’s and difference) pronounced with a [ts] at the end, described as a “natural assimilation process” (1983: 434). This is not supposed to be unmarked rule application, but successful acquisition of an L2 feature (the homorganic stop insertion rule).

24 This last one is not even a perfect example, as the plosive at the end of end is voiceless in English. Although it is possible that the devoicing of [d] to [t] happened due to universal unmarkedness, it may as well have been an instance of successful acquisition of an L2 feature. These are the examples that illustrate what was discussed at

latter phenomenon are mentioned in the paper, but stressed vowel lengthening concerns words like decision pronounced with a long vowel in the stressed syllable, which is interpreted as an instance of TETU because stress and vowel length often go hand in hand in languages (but not in English or Hungarian).

Here we have only overviewed the most widely cited, therefore “classic” examples of markedness, and those aspects of TETU that have been discussed in the literature on Hunglish – further issues will be examined later when interpreting the data gained in the case studies presented in Chapter 5 and drawing the conclusions in Chapter 6.