Читать книгу Stress Variation in English - Alexander Tokar - Страница 14

Enlarging on what was stated in the previous sections, we note that with regard to phonetic realization of stress, languages are traditionally classified into pitchpitch-accent languages and stress-accentstress-accent languages. Pitch-accent languages rely on (typically a high) pitch as their primary cue, whereas in stress-accent languages other cues such as durationduration, intensityintensity, and vowel quality play an important role as well (see, e.g., van der Hulstvan der Hulst 2010b: 12). According to HymanHyman (1977: 37–38), JapaneseJapanese is a prototypical example of a pitch-accent language, whereas English exemplifies a stress-accent language. As GussenhovenGussenhoven (2006: 216) observes in this connection:

Whereas the most prominent syllable in a word in English is longer than other syllables and has a less reduced vowelreduced vowel, in addition to being marked with tonetone in focused pronunciation, […] the accented syllable of a JapaneseJapanese word stands out virtually exclusively because it is associated with pitchpitch features. (GussenhovenGussenhoven 2006: 216)

As regards stress placement, world’s languages are often classified into fixed-stress languages and variable-stress languages (van der Hulstvan der Hulst 2010b: 33). An example of the former is DyirbalDyirbal, in which primary stress falls on the first syllable in a word (van der Hulst et al. 2010: 672). The latter category (in addition to being exemplified by contemporary English) is also exemplified by Modern GreekGreek, in which stress falls upon one of the last three syllables in a word (van der Hulst et al. 2010: 803).

The category of fixed-stress languages can be further divided into languages in which stress assignment rules count from the beginning vs. the end of the word. For example, in DyirbalDyirbal, in which stress falls on the first syllable in a word, stress assignment can be said to be left-bound, counting from the beginning of the word (its left edge), whereas, e.g., MacedonianMacedonian is a right-edgeright-edge fixed-stress language, i.e., stress in Macedonian falls upon the antepenultimate syllable (van der Hulstvan der Hulst et al. 2010: 807), which is the third syllable counting from the end of the word.

As van der Hulstvan der Hulst (2010b: 33) observes, so far typological studies on stress have discovered only three stress location possibilities for both left- and right-edgeright-edge languages with fixed stressfixed stress. That is, in left-edgeleft-edge languages, stress can fall only on the first three syllables (stress in such languages can thus be said to be initial, pen-initial, or post-pen-initialpost-pen-initial). Similarly, in right-edge languages stress can fall only on the last three syllables (i.e., the final syllable or the ult, the pre-final syllable or the penult, or the pre-pre-final syllable or the antepenult). There seem to be no left-edge fixed-stress languages in which stress is placed on the fourth syllable counting from the beginning of the word. Likewise, there seem to be no right-edge fixed-stress languages in which stress is placed on the pre-antepenult (i.e., the fourth syllable from the end of the word).

Just like fixed-stress languages, languages with variable stressvariable stress can be further classified into several subcategories. On the one hand, there are languages such as the above mentioned Modern GreekGreek, where the location of stress is “completely unpredictable and thus lexically marked” (van der Hulstvan der Hulst 2010b: 42), i.e., as pointed out above, stress in a word of Greek falls upon one of the last three syllables in a word. On the other hand, there are languages with variable stress in which variability can be attributed to what has traditionally been called the weight of a syllable. With regard to this parameter, syllables are usually classified into heavyheavy and lightlight syllables, but note that an answer to the question of which syllables count as heavy and stressable and which as light and unstressable varies from language to language and, what is more, typically within one and the same language, syllables are in different positions required to satisfy different conditions in order to be eligible to receive stress. A case in point is Modern High GermanGerman, in which the placement of primary stress is said to obey the following rules: 1) final stress if the vowel in the ult is long or there are two coda consonants, 2) penultimate stress if Condition 1 is not fulfilled and if the penult contains a coda consonant, and 3) antepenultimate stress in all other cases (van der Hulst et al. 2010: 803).

It is not difficult to see that word-finally, Modern High GermanGerman only treats syllables whose rhymes exhibit the structures V: or VCC, where C stands for a consonant, V: for a long vowel, and V for a short vowel, as stressable. Thus, “a word-final syllable must have more consonants to be counted as heavyheavy, since word-final consonants are often extrametricalextrametrical” (HayesHayes 1995: 59) i.e., invisible to metrical construction rules so that the rhymerhyme structure VC is at the end of a word equivalent to the rhyme structure V and counts therefore as lightlight rather than heavy. (By contrast, the rhyme structure VCC is word-finally equivalent to the rhyme structure VC and counts therefore as heavy.) In the penultimate position, by contrast, only syllables with a coda count in German as stressable, i.e., the length of the penultimate vowel does not play a role (with regard to which Modern High German is different from Present-day English, where a codaless penult also counts as heavy when it contains a long vowel). Finally, in the antepenultimate position, syllables in German are (just as in English) not required to exhibit a particular structure in order to be able to attract stress. Just like Present-day English, Modern High German is therefore also often referred to as a language with LatinLatin-like stress (Hayes 1995: 181; cf. Domahs et al. 2014: 80, who conclude that “German, DutchDutch, and English must be considered quantity-sensitivequantity-sensitive languages, with the three languages showing very similar patterns overall,” i.e., stress is penultimate when the penult is heavy and antepenultimate when the penult is light.)

In some quantity-sensitivequantity-sensitive languages, the weight of the stressed syllable depends on factors other than the length of the vocalic nucleusnucleus and the presence/number of consonants in the coda position, which is typical of Modern High GermanGerman. For example, in KeraKera stress is normally placed on a syllable that contains a long vowel, but if no such syllables occur, stress falls on the first syllable that is pronounced with a high tonetone (van der Hulstvan der Hulst et al. 2010: 840). Similarly, in Serbo-CroatianSerbo-Croatian stress is normally placed on the first syllable that is pronounced with a tone (van der Hulst et al. 2010: 814). In Shipibo-ConiboShipibo-Conibo stress is placed on a second syllable if it ends in a consonant or a nasal vowel (van der Hulst et al. 2010: 782). Similarly, in ApurinãApurinã stress is placed on the final syllable when the vowel occurring in it is characterized by a nasal manner-of-articulationmanner-of-articulation (van der Hulst et al. 2010: 775). Observe also that in MamMam the stress placement rules normally assign stress to the last long vowel in a word, but when no such vowels occur, stress is placed on the last syllable that ends in a glottal stop (van der Hulst et al. 2010: 772).

Controversies surround the issue of whether consonants occurring in the onsetonset position have any effect on syllable weightsyllable weight. A popular, often repeated view is that this is not the case. According to HayesHayes (1995: 51; author’s boldface), “prevocalic segments in the syllables (i.e. onset segments) are prosodically inert: […] VC is prosodically equivalent to CVC and CCVC, V: [is equivalent] to CV: and CCV:, and so on.” As Hayes (1995: 51) adds, “[w]hile this claim is not fully valid at the observational level […], it is so well supported across languages that it serves as the central observation for formal theories of syllable weight.”

The present monograph suggests, however, that the question of whether onsetonset segments play a role in stress assignment should be split into two separate questions: 1) Is the presence of an onset consonant a necessary condition for a syllable to receive stress and 2) does the presence of an onset consonant increase the probability of a syllable receiving stress?

As for Question 1, the answer to it varies from language to language. For instance, in ThaypanThaypan and MbabaramMbabaram, stress is placed on the first syllable with an onsetonset, whereas Lamu-lamuLamu-lamu, Western ArandaWestern Aranda, and AlyawarrAlyawarr place stress on the first syllable if it contains an onset, otherwise stress in these languages is pen-initial (van der Hulstvan der Hulst et al. 2010: 685, 680, 677, 669). Similarly, “[i]n MadimadiMadimadi, an Australian language of New South Wales […], stress on words of three or more syllables is sensitive to the presence of onsets that have coronal consonants” (DavisDavis 1988: 4).

By contrast, in the case of, e.g., the English language, the answer to Question 1 is a clear no, which is strongly supported by the following fact. In the entries of Merriam-Webster Online in which boundaries between syllables are marked by means of (-) (e.g., for animal, the dictionary gives the syllabified transcription /ˈa-nə-məl/), there are 51,090 onsetless syllables, i.e., those that begin with a vowel. Since 22,179 of these syllables (~43.41 %) bear either primary or secondary stress (e.g., the first syllable in animal), we can say that a syllable beginning with a vowel is in the English language almost as frequently stressed as it is unstressed; the presence of an onsetonset consonant is thus clearly not a conditio sine qua non of stress in the English language.

At the same time, however, with regard to Question 2, it must be observed that of the 311,074 English syllables that (in the entries of the same dictionary that contain syllabified phonetic transcriptions) have at least one consonant in the onsetonset position, 169,910 (~54.62 %) are syllables that bear either primary or secondary stress (e.g., /ˈka-pə-təl/ of capital). Syllables with onsets are thus more frequently (~54.62 %) stressed in the English language than onsetless ones (~43.41 %), with this difference counting as statistically (hugely) significant: χ² (1) = 2,214, p < 0.000001.

Similar results are reported by KellyKelly (2004) with regard to disyllabic English words. His finding is that percentages of initially-stressed disyllables “increase monotonically with the number of consonants beginning a word” (Kelly 2004: 233). Thus, of the disyllables lacking onsetonset consonants in the penult, only 35 % are stressed initially (Kelly 2004: 233). By contrast, in the case of one, two, and three onset consonants (beginning the penult of a disyllabic English word), the corresponding percentages are 69 %, 83 %, and 98 % respectively (Kelly 2004: 233). Notice in this connection that in the English language, “prefixes are relatively common on words beginning with few or no consonants” (Kelly 2004: 234). E.g., inflame. However, as Kelly (2004: 235) reports, “the relationship between word onset and stress is not eliminated when prefixation is factored out” (author’s italics). The fact that, e.g., the verb excel is stressed /ɪkˈsɛl/ (OD) whereas the verb cancel is stressed /ˈkans(ə)l/ (OD) cannot then be attributed to the prefix status of ex-ex- in formations such as, for instance, ex-president or external (vs. internal). What accounts for the stress difference between these two disyllabic verbs is that while the first syllable in excel lacks an onset consonant, the first syllable in cancel contains it; stress is therefore final in the former, while it is initial in the latter. As Kelly (2004: 237) concludes,

statistical analyses of a large sample of disyllabic English words document a significant relationship between word stress and word onsetonset characteristics. This relationship is both large in magnitude and broad in scope, as it permeates the English lexicon and cannot be localized to a small set of atypical words. As such, it conflicts with many theories of English phonology which do not consider onset structure as relevant to stress assignment. (KellyKelly 2004: 237)

KellyKelly (2004: 237–239) also studied the assignment of stress in pseudowords (also commonly referred to as “nonsense words” or “nonwords”), i.e., potentially possible but thus far non-existent words in the English language. Thus 20 participants expressed their judgments as to which syllable stress should fall upon in pairs of pseudowords such as bontoon–brontoon differing only with regard to the number of consonants occurring in the initial syllable’s onsetonset position. The results obtained also “supported the hypothesis that onset patterns affect English stress assignments. The mean proportion of [initial-stress] assignments was .78 for CC pseudowords but only .61 for C pseudowords” (Kelly 2004: 238). That is, pseudowords such as brontoon are more likely to receive initial stress than pseudowords such as bontoon.

Consider also the placement of stress in RussianRussian, which is typologically a language with unpredictable stress: “Stress falls on the first syllable lexically marked for [stress], else on the first” (van der Hulstvan der Hulst et al. 2010: 813). Notice, however, that of the total number of the 157,877 words (of any syllabic length) for which stress markings are provided in the online version of the Russian Orthographic Dictionary (http://tinyurl.com/q6x5jj6, 11.08.2015), 1,028 are initially-stressed words beginning with a vowel, 18,189 are initially-stressed words beginning with a consonant, and 370 are initially-stressed words beginning with a /j/‑glideglide, which in the Russian orthography is associated with the letters е, ё, й, ю, and я. The /j/‑glide is often referred to as a semi-vowelsemi-vowel because phonetically it is indeed more like a vowel in that its articulation does not involve a major obstruction to the airflow; phonologically, however, the /j/‑glide is more like a consonant in that it does not function as the nucleusnucleus of a syllable (see, e.g., Stadnik-HolzerStadnik-Holzer 2009: 26). What is important here is that the 1,028 initially-stressed words beginning with a vowel constitute only ~3.82 % of the total number of the 26,907 Russian words beginning with a vowel. By contrast, the 18,189 initially-stressed words beginning with a consonant constitute ~14.04 % of the total number of the 129,582 Russian words beginning with a consonant. And the 370 initially-stressed words beginning with a /j/‑glide constitute ~26.64 % of the total number of the 1,389 Russian words beginning with a /j/‑glide. These differences all count as statistically significant: For a vowel vs. a consonant, χ² (1) = 2,159, p < 0.000001; for a vowel vs. a semi-vowel, χ² (1) = 1,464, p < 0.000001; for a consonant vs. a semi-vowel, χ² (1) = 179, p < 0.000001.

Note also that 11,923 initially-stressed RussianRussian words whose stressed syllables have only one consonant in the onsetonset position constitute ~12.1 % of the total number of the 98,538 Russian words beginning with a single consonant. However, 5,297 initially-stressed Russian words whose stressed syllables have two consonants in the onset position constitute ~18.57 % of the total number of the 28,529 Russian words beginning with two consonants and 881 initially-stressed Russian words whose stressed syllables have three consonants in the onset position constitute ~37.75 % of the total number of the 2,334 Russian words beginning with three consonants. Finally, 88 initially-stressed Russian words that have four consonants in the onset position (such consonant clusters are impossible in English, but they are possible in Russian: e.g., vstroennyj, “inbuilt”) constitute ~48.89 % of the total number of the 180 Russian words beginning with four consonants. These differences are, again, all statistically significant. For one vs. two consonants, χ² (1) = 790, p < 0.000001; for two vs. three consonants, χ² (1) = 496, p < 0.000001; for three vs. four consonants, χ² (1) = 8.76, p = 0.003. Thus, very similar to what KellyKelly (2004) could establish with regard to initial stress in English (disyllabic words), it appears that the probability of initial stress in Russian likewise increases with an increase in the number of consonants in the (initial syllable’s) onset position.

At the end of this section, it must be also mentioned that in contrast to the languages discussed thus far, few world’s languages can be referred to as stressless or at least partially stressless languages. Partial stresslessness is characteristic of, e.g., Chinanteco, which has minimal pairs whose members differ from each other only with regard to the presence/absence of stress. As van der Hulstvan der Hulst et al. (2010: 774) report, the form fi means “handle” when it is stressed, but it means “road” when pronounced without stress. Similarly, in Lak-DargwaLak-Dargwa, “there are stressless words such as turlu ‘cloud’” (van der Hulst 2010a: 482).

As for languages that completely lack stress, it must be observed that stresslessness is usually one of the possible analyses that can be found in the literature alongside analyses that do not deny the existence of stress. A classic case is modern FrenchFrench, whose stress system is analyzed differently. Sometimes it is argued that word stress in French is final unless the ult ends in a schwaschwa, which leads to penultimate stress (e.g., CarpenterCarpenter 2010: 367). The latter, however, does not occur very often because “word-final schwa in French is only pronounced under very specific circumstances nowadays” (van der Hulstvan der Hulst 2010a: 462). Stress in Present-day French is therefore predominantly final. At the same time, however, it is also suggested that French lacks stress at the word level altogether (see, e.g., NiebuhrNiebuhr 2007: 175). Thus:

when words appear on the edge of larger prosodic units, their edge syllables may be the anchor point for intonational tones […] or boundary phenomena (tonal or segmental) which create the perceptual sensation of these syllables being prominent. The linking of intonational pitchpitch movement or other properties to edges of phrases, and thus edges of words that are peripheral in phrases, may lead to the illusion of the words having primary ‘stress’, while, synchronically speaking, there is no primary word stress at all (which is one way of analyzing FrenchFrench ‘word’ stress […] (van der Hulstvan der Hulst 2012: 1515; author’s italics)

The analysis of FrenchFrench as a stressless language is much more intuitive given that French is often said to exemplify a syllable-timedsyllable-timed language, “in which all syllables are of approximately equal length” (MarksMarks 1999: 191). Although some authors question the often taken-for-granted syllable-timed nature of the French language—according to WenkWenk (1985: 158), “those syllables occurring in final position in rhythmic groups (or “phonological words”) in French are on average about twice as long as non-final syllables”—there is little doubt that syllables constituting words in French are considerably different than those that make up words in a language such as English:

In FrenchFrench, unlike English, all the syllables are fully realized. The accented syllable is always the word’s last syllable and the differences in durationduration, energy, and pitchpitch, are small between accented and nonaccented syllables. Finally, there is a high proportion of open to closed syllables in French […]. These parameters give an impression of syllable regularity for French, that can be distinguished from the stress-timedstress-timed rhythmic pattern of English […] (Dehaene-Lambertz & HoustonDehaene-Lambertz & Houston: 1998: 23)

To conclude, because in a language such as FrenchFrench, “vowel reduction is much less evident” than it is in a language such as English (Payne et al.Payne et al. 2012: 204), the former can from the perspective of the latter be regarded as a stressless language, i.e., as, e.g., JonesJones (1917: xl) observes, “French and HindustaniHindustani […] make no use of lexically significant word-stress.”