Phonological implications

This section will highlight how the findings of the two experiments have contributed to the understanding of linguistic universals in L2 acquisition.

Phenomena related to the notion of markedness were discussed in Section 2.2, where it was mentioned that markedness may account for certain features of interlanguage not found in either the L1 or the L2/FL (cf. Eckman 2008). Examples of such phenomena have been found by Altenberg & Vago (1983), who examined two Hungarian ESL learners of English and found that the learners exhibited final obstruent devoicing (FOD) in their English speech. This observation has been widely cited in the literature to illustrate the phenomenon of “The Emergence of the Unmarked” (TETU, cf. Section 2.2), because neither English nor Hungarian has final devoicing in their phonological systems.

However, there are at least two reasons why this might not be the best example to support TETU. Firstly, as pointed out in Section 2.2 (in footnote 24), some instances of devoicing observed in the subjects’ pronunciation may be a result of successful acquisition of a target language feature, because utterance-final obstruents are indeed voiceless in English (cf. Section 3.4). Secondly, Mády & Bárkányi’s (2015) experiment has shown that FOD can be attested in Hungarian too, although FOD in Hungarian is a phonetic detail and not a neutralising process, so the contrast between words like mész and méz is not lost (just like in the case of English).

For these reasons, although at the phonetic level both English and Hungarian may feature FOD to some extent, it is not the type of final devoicing that characterises languages such as German, in which word-final devoiced lenis obstruents become equivalent to fortis ones, and as a result the contrast between, for example, [t] and [d] is neutralised in word pairs like bunt ‘colourful’

and Bund ‘federation’. Altenberg & Vago’s (1983) observation of FOD might therefore not necessarily be a perfect example of TETU.

Nevertheless, one particular finding in each of the experiments presented in Chapter 5 of this thesis can be considered an instance of TETU. The first one of these is the observation that certain Hungarian learners acquiring a non-rhotic accent of English systematically drop the [r]

before consonants (e.g., in words like market), but variably drop it in word-final position (e.g., in words like car). This indicates that apparently the suppression of the [r] is “easier” pre-consonantally than word-finally, which leads to an intermediate stage in the acquisition of R-dropping. Intriguingly, it is well-known that universally, processes that delete consonants or

vocalise them (replace them with vowels) apply word-finally only if they also apply before consonants – that is, such processes are more marked in the former position than in the latter.

It can be illustrated by a number of examples taken from various languages that a word-final coda position is stronger than a preconsonantal (word-medial) one (cf. Balogné Bérces &

Honeybone 2012, Ségéral & Scheer 2008), which can explain why word-final R’s are less prone to lenition (deletion in this particular case) than preconsonantal ones.

For example, as demonstrated in, for example, Scheer (2004: 629, cf. Table 6.1), the diachronic process of Old French L-vocalisation, leniting /l/’s into vowels, left unaffected non-coda L’s (e.g., luna > lune ‘moon’; flore > fleur ‘flower’; vela > voile ‘sail, veil’) as well as those in final codas (e.g., sal > sel ‘salt’, caball > cheval ‘horse’), whereas it systematically weakened preconsonantal L’s (e.g., alba > aube ‘dawn’, talpa > taupe ‘mole’).

Table 6.1: Old French l-vocalisation, Scheer (2004: 629)

A similar example is Harris & Kaisse’s (1999: 158) account of s-debuccalisation (referred to by the authors as “aspiration”) in two varieties of Argentinian Spanish (the porteño dialect of Buenos Aires and in the dialect of Rio Negro, abbreviated as PO and RN, respectively, cf. Table 6.2). In PO, /s/ is aspirated preconsonantally, irrespective of whether the sC string falls within a morpheme (ca[h]pa), within a word across a morpheme boundary (de[h]-cargar), or across a word boundary (ve[h] do[s]). However, this phenomenon is not attested prevocalically (e.g., ca[s]a, de[s]-armar, ve[s] uno) or before a pause (e.g., ve[s]...do[s], even if the pause is a word-medial one (ca[s]...pa). (Pauses are indicated by “…” in the table).

Table 6.2: Harris & Kaisse (1999: 158)

Such universal tendencies in consonant lenition may also apply to produce the semi-rhoticity patterns found in Jamaican-type varieties of English as well as Hunglish, and preconsonantal-only deletion is an instance of TETU.

The other instance of TETU (the one found among the observations of Experiment 2) is Hungarian learners’ tendency to pronounce otherwise unstressed (C)VV and (C)VC syllables as stressed, of which the former was more frequent. Let us see why stress assignment to (C)VV syllables is less marked than to (C)VC.

In languages with weight-sensitive stress systems, stress assignment is dependent on syllable weight in that heavy syllables attract stress. In some languages, heavy syllables are determined qualitatively (i.e., it is the quality of the segments in the syllable that account for syllable weight), while in others syllable weight is quantitative, which means that in order for a syllable to be heavy (and thus stressable), it needs to contain two moras (cf. Gordon 2004b).

The fact the bimoraic syllables attract stress is so frequent a characteristic of weight-sensitive stress systems that the principle expressing this regularity in metrical phonology (the so-called Weight-to-Stress Principle)⁷⁵ has also been formalised as a markedness constraint in Optimality Theory (cf. Gordon 2004a, Prince & Smolensky 2004). In this respect, stress assignment to bimoraic syllables can be considered an unmarked property of weight-sensitive systems.

In such systems, (C)V syllables are light (which means they are unable to attract stress);

however, what counts as a heavy syllable varies greatly (cf., e.g., Gordon 2004b): while there are languages in which only (C)VV syllables are heavy (i.e. syllables containing a long vowel, including diphthongs), in others both (C)VV and (C)VC syllables are heavy (cf. Gordon 2006).

In other words, if a stress system treats (C)VC syllables as heavy, it also treats (C)VV syllables

75 Despite its misleading name, the WSP expresses that if a syllable is heavy, then it is stressed – cf. Chomsky &

Halle (1968), Prince (1990), etc.

as heavy, but not vice versa. (Similar implicational relations were discussed in Section 2.2.) This means that in a quantity-sensitive stress system, it is less marked for the WSP to assign primary stress to (C)VV than to (C)VC. Since Hungarian is not weight-sensitive, the participants’ tendency to stress (C)VV syllables to a greater extent than (C)VC can be regarded as another instance of TETU.

What the above examples illustrate is that learners possess an innate knowledge of what is universally less marked (even if the feature in question is not part of their L1), and what makes this case special is that these instances were found in purely EFL settings (i.e., with learners’ exposure to the target language limited to the classroom) in the pronunciation of learners who are well beyond the critical or sensitive period in FL language learning. It can thus be concluded that these universal strategies are available to learners totally irrespective of such crucial differences as learning setting and age.