Alexander Tokar

Stress Variation in English

Narr Francke Attempto Verlag Tübingen

Inhalt

Acknowledgements

I am indebted to Professors Irmengard Rauch (University of California at Berkeley) and Jean-Michel Fournier (l’Université François-Rabelais, Tours), who reviewed this monograph for the Language in Performance series. It is difficult to say how many mistakes, inadequacies, inconsistencies, etc., this book would contain if not for, in particular, Prof. Fournier’s thorough critique of an earlier version of this manuscript. All remaining errors are, of course, entirely my own responsibility!

My sincere gratitude goes also to the editor of the Language in Performance series, Professor Rainer Schulze (Leibniz Universität Hannover), whose comments on the structure of the book have made it more reader-friendly. To Prof. Schulze, my thanks are also due for adding this title to the Language in Performance series!

I also thank 1) Professors Ulrich Busse and Alexander Brock (Martin-Luther-Universität Halle-Wittenberg), for inviting me to present parts of this research in their Forschungskolloquium in October 2016; 2) the UC Berkeley Library (represented by Lydia Petersen), for providing access to its vast resources, without which I would have been unable to write this book; 3) Samantha Tanner, whose native-speaker competence was of great help during the earlier stages of this project back in 2014; and 4) a number of people—colleagues, family, and friends—with whom I had the pleasure of discussing this topic during the many years I spent studying English stress. Some of these people include, in alphabetical order, Denis Balagurov, Sergei Danilov, Yuri Dyomin, Tibaut Houzanme, Natalia Peters, Karen Sullivan, and Viktor Treshchev.

For their moral support and much more, I thank my parents Boris and Natalia Tokar. The book is dedicated to them!

Abbreviations

1 Introduction

This monograph is about English words with stress doublets, i.e., words in which stress is interchangeably placed upon more than one syllable without changing the meaning. For example, in the YouTube video whose unique identifier is ‑AnsohxXnQU (17.09.2016; to retrieve the video the identifier should be preceded by https://www.youtube.com/watch?v=), an American English speaker can be heard stressing applicable initially in This name change will be more functionally ˈapplicable (00:01:25.899 --> 00:01:33.979), but in we’ve made the name change to make it more applicable (00:03:01.980 --> 00:03:04.880), the very same American English speaker can be heard using the stress pattern apˈplicable. (According to Longman Dictionary of Contemporary English (henceforth LDOCE), the variation /əˈplɪkəbəl/ vs. /ˈæplɪkəbəl/ occurs in both British and American English, but observe that in YouTube videos featuring the voices of British English speakers, the author heard only the stress pattern /əˈplɪ-/.) A very similar case is adult, which in British and American English is interchangeably stressed /ˈædʌlt/ and /əˈdʌlt/ (LDOCE) (with, however, initial stress being the preferred stress pattern in British and Australian English vs. final stress being more frequently used (than initial stress) in American English); a YouTube video in which a British English speaker can be heard vacillating between /ˈædʌlt/ and /əˈdʌlt/ is S5hXrgfwK8o (10.04.2017).

1.1 Scope of the study

In addition to the obvious questions raised by variably-stressed words such as applicable and adult—1) What are the causes of stress variation exhibited by these and other English words with stress doublets? 2) Why do only some English words have stress doublets? 3) Do words with stress doublets prefer particular stress patterns and if so, why?—the present monograph will attempt to give a more precise answer to the general question of why English words (either with or without stress doublets) are stressed the way they are stressed. According to a popular view, Present-day English is a LatinLatin-like language as far as its stress system is concerned (HayesHayes 1995: 181). Thus, since “[i]n the three hundred years that intervened between the Norman ConquestNorman Conquest and Chaucer, the [English] language was inundated by Romance words” (Halle & KeyserHalle &Keyser 1971: 97), the stress rule of contemporary English is essentially the stress rule of Classical Latin: “Stress falls on the penultimate syllable if it contains a long vowel or is closed. […] Else stress is antepenultimate” (van der Hulstvan der Hulst 2010a: 459). An illustration is the word paprika, which in British English, according to LDOCE, is interchangeably stressed /ˈpæprɪkə/ and /pəˈpriːkə/. As one can notice, when stress in paprika is penultimate, the vowel in the penultimate syllable /ˈpriː/, which bears stress, is long. When, by contrast, paprika is pronounced /ˈpæprɪkə/, the vowel in the unstressed penult /prɪ/ is short. The stress patterns /ˈpæprɪkə/ and /pəˈpriːkə/ can thus both be said to be in accordance with the Latin Stress Rule.

At the same time, however, in the case of the stress variation /ˈɛkskwɪzɪt/ vs. /ɪkˈskwɪzɪt/ (Oxford English Dictionary, henceforth OED), the vowel in the penultimate syllable remains short irrespective of whether stress in exquisite is penultimate or antepenultimate. Similarly, it does not matter whether stalactite and stalagmite are stressed pen-pen- or antepenultimately. In both /ˈstælæktaɪt/ vs. /stəˈlæktaɪt/ (OED) and /ˈstælæɡmaɪt/ vs. /stəˈlæɡmaɪt/ (OED), there is a short vowel in the penult that is followed by a coda consonant, i.e., /læk/ and /læɡ/; stress in the trisyllables stalactite and stalagmite is thus supposed to be penultimate in accordance with the LatinLatin Stress Rule. A fairly similar case is necropsy, for which the OED gives segmentally identical British English transcriptions /ˈnɛkrɒpsi/ and /nɛˈkrɒpsi/: As one can notice, these transcriptions differ from each other only with regard to the location of the stress symbol (ˈ).

Similar examples can be found among disyllabic English words, which are not covered by the LatinLatin Stress Rule, stating that “[i]n words with 2 or fewer syllables, primary stress occurs on the initial syllable” (StressTyp2 database), i.e., in contrast to a disyllabic English word, in which stress is either initial or final (e.g., /ˈædʌlt/ vs. /əˈdʌlt/ of adult), in a disyllabic Latin word, stress can only be initial, i.e.,

According to van der Hulstvan der Hulst (2010a: 445), in the English language “[p]rimary stress falls on the final syllable in nouns if the vowel is long, in verbs if the vowel is long or there are two closing consonants.” The diachronic basis of this assertion is the view, expressed in Halle & KeyserHalle & Keyser (1971: 99–101), that the stress system of contemporary English was shaped not only by LatinLatin but also by (Old) FrenchFrench:

A convincing critique of this view can be found in FournierFournier (2007: 232), who argues that:

From a purely synchronic perspective, the view that final stress in English crucially depends upon the length of the vowel in the final syllable/the number of closing consonants when the vowel is short cannot be accepted because especially words with stress doublets provide too many counterexamples. For instance, for the verb migrate the OED gives segmentally identical British English transcriptions /mʌɪˈɡreɪt/ vs. /ˈmʌɪɡreɪt/ and the American English transcription /ˈmaɪˌɡreɪt/, i.e., migrate is always pronounced with a diphthong in the ult (which counts as a long vowel), but stress in migrate is not always final in Present-day English. A similar case is the adjective overt, for which LDOCE gives segmentally identical British English transcriptions /ˈəʊvɜːt/ vs. /əʊˈvɜːt/ and American English transcriptions /ˈoʊvɜːrt/ vs. /oʊˈvɜːrt/. The adjective under consideration is thus also always pronounced with a long vowel in the ult, even when stress in overt is non-final. The noun decade is interchangeably stressed /ˈdekeɪd/ and /deˈkeɪd/ (LDOCE), with both the stressed ult /ˈkeɪd/ of the latter and the unstressed ult /keɪd/ of the former containing a diphthong.

Of the 268 variably-stressed disyllables in LDOCE (including words such as caffeine, which, according to the dictionary, is stressed only /ˈkæfiːn/ in British English vs. only /kæˈfiːn/ in American English), 173 (~64.55 %) are disyllables such as caffeine, in which stress differences are not accompanied by segmental differences involving the quality of the vowel in the ult. Thus, if the ult bearing stress contains a long vowel or diphthong, then also the unstressed ult of an initially-stressed alternative pronunciation likewise contains a long vowel or diphthong (e.g., /kæˈfiːn/ vs. /ˈkæfiːn/ of caffeine). Similarly, if the ult bearing stress contains a short vowel, then (a qualitatively unreduced) short vowel also occurs in the ult of an initially-stressed alternative pronunciation. E.g., address is, according to LDOCE, only /əˈdres/ in British English, but it is /əˈdres/ vs. /ˈædres/ in American English. The ult /res/, which contains a short vowel (being followed by only one coda consonant), thus occurs in both the finally-stressed pronunciation /əˈdres/ and the initially-stressed alternative /ˈædres/. (Notice, however, that in YouTube videos featuring the spoken occurrences of address, initial stress was heard by the author only in environments such as IP address or street address, whereas in contexts such as State of the Union Address, Gettysburg Address, inaugural address, commencement address, etc., where address expresses the meaning “public speech,” stress in address was exclusively final. The variation /əˈdres/ vs. /ˈædres/ is thus, at least in part, a matter of semanticssemantics.) Anyway, the fact that disyllables such as caffeine and address constitute the majority of variably-stressed disyllables in English strongly suggests that the quality of the vowel in the ult plays a relatively insignificant role in the assignment of stress in disyllabic English words.

Another important number is that in Oxford Dictionaries (henceforth OD), there are 48,428 solidly-spelled polysyllables whose only stress pattern (given in the dictionary) is non-initial stress. E.g., inhibit is stressed only /ɪnˈhɪbɪt/ (OD), with stress in the word not falling upon its first syllable /ɪn/. Of the 48,428 polysyllabic words in the OD such as inhibit, 28,944 (~59.77 %) have righthand strings such as, e.g., /‑ˈhɪbɪt/ of /ɪnˈhɪbɪt/, which occur in at least one other English word. E.g., apart from occurring in the transcription /ɪnˈhɪbɪt/ of inhibit, the righthand string /‑ˈhɪbɪt/, which begins with the primary stress symbol (ˈ), also occurs in the transcription /prə(ʊ)ˈhɪbɪt/ of prohibit (OD). (The percentage of such cases rises to ~66.55 % if, apart from counting only exact matches such as /‑ˈhɪbɪt/ of /ɪnˈhɪbɪt/ and /prə(ʊ)ˈhɪbɪt/, we ignore the onsetonset of the primary-stressed syllable. E.g., /kəˈlɒtəmi/ of colotomy is the only transcription in the OD that contains the righthand string /‑ˈlɒtəmi/. At the same time, however, apart from /kəˈlɒtəmi/ of colotomy, the dictionary also has the transcription /ləˈbɒtəmi/ of lobotomy, which shares with /kəˈlɒtəmi/ the righthand string /‑ˈ-ɒtəmi/.)

Notice further that of the 48,428 polysyllabic words such as inhibit, 35,525 (~73.36 %) have transcriptions such as /ɪnˈhɪbɪt/, in which the righthand string that begins with the primary stress symbol (ˈ) is segmentally longer (i.e., contains more phonetic segments, with the durationduration symbol (:) not counting as one of them) than the remaining lefthand string preceding it. Thus, for instance, the righthand string /‑hɪbɪt/ of /ɪnˈhɪbɪt/ consists of four phonetic segments while the immediately preceding lefthand string /ɪn/ has only two. (If also the duration symbol (:) is taken into consideration (e.g., in /ʃɪˈkɑːɡəʊ/ of Chicago (OD), the righthand string /‑kɑːɡəʊ/ consists of six rather than five phonetic segments), 36,017 (~74.37 %) non-initially-stressed words in the OD dictionary can be said to have segmentally longer righthand strings such as /‑kɑːɡəʊ/ of /ʃɪˈkɑːɡəʊ/.)

The connection between these numbers is the Old EnglishOld English Stress Rule: “Primary stress falls on the first syllable (of the root/stem)” (van der Hulstvan der Hulst 2010a: 446). Assuming that this rule is still alive in contemporary English, we can argue that inhibit is stressed /ɪnˈhɪbɪt/ because the righthand string -hibit-hibit counts morphologically as the root of the word. Thus, because apart from occurring in inhibit this righthand string also occurs in, e.g., prohibit, both inhibit and prohibit count for an English speaker as morphologically complex words consisting of the components in-in-/pro-pro- and -hibit-hibit. These components do not have meanings of their own, but as many authors suggest, components of which morphologically complex English words are made up are not necessarily traditional morphemesmorphemes (i.e., meaning-bearing units). As, e.g., AronoffAronoff (1976: 12) aptly observes in this connection: “What even vague sense does prefer share with confer and transfer? or commit with remit and submit? or receive with conceive and perceive? or consume with presume and assume? or reduce with induce and deduce? None.”

The (purely formal) segmentation in-in-/pro-pro- + -hibit-hibit raises, however, the question of which of these components should count morphologically as the root of inhibit and prohibit, for the root is the only obligatory element in a word (e.g., Mel’čukMel’čuk 2001: 69), i.e., a word may be affixless, but it may not be rootless: Any word contains at least one root. As Mel’čuk (2001: 69–79) explains, roots are different from affixes both semantically and formally. With regard to the former, roots are associated with more concrete meanings than those usually expressed by affixes. With regard to the latter, however, roots usually contain more segments than affixes (Mel’čuk 2001: 69). Needless to say, since the components in-/pro- and -hibit-hibit are not associated with discernible meanings of their own, it is only the formal length-criterion that can be relied upon to answer the question of which of these two components counts in inhibit and prohibit as the root. Since the righthand component -hibit-hibit contains more segments than the lefthand components in-/pro-, the former is clearly more like a root whereas the latter are more like prefixes. Inhibit and prohibit are, therefore, both stressed /‑ˈhɪbɪt/, with the location of stress in these words being the root–prefix boundaryroot–prefix boundary location (and precisely because of this fact, it does not matter that the penult /ˈhɪ/, which receives stress, ends in a short vowel and is thus not supposed to be stressed in accordance with the LatinLatin Stress Rule). Likewise, given the numbers presented above, it can be conjectured that similar analyses apply to the majority of non-initially-stressed English words. E.g., colotomy is stressed /kəˈlɒtəmi/ not (or at least, not only) because the penult /tə/ ends in a short vowel—stress in /kəˈlɒtəmi/ should thus be antepenultimate in accordance with the Latin Stress Rule—but (also) because the segmentally longer righthand string -otomy-otomy, which colotomy shares with lobotomy, counts morphologically as the root.

Consider also the stress variation /ˈɒksᵻdʒᵻneɪt/ vs. /ɒkˈsɪdʒəneɪt/ of oxygenate (OED; only British English). To begin with, observe that in the view of many authors, suffixes in English have intrinsic properties with regard to stress; e.g., KettemannKettemann (1988: 290) speaks of a diacritic featurediacritic feature that is contained in the lexicon entry dedicated to a particular English suffix. This diacritic feature is what triggers, in a particular suffixed derivative, a particular stress-related behavior. Thus, for instance,

A challenge to views similar to this is, however, posed by “mixed suffixesmixed suffixes,” which “have two or more distinct modes of operation” (FudgeFudge 1984: 45). E.g., the suffix -ate-ate of /ˈɒksᵻdʒᵻneɪt/ is a stress-neutralstress-neutral suffix: Stress in the derived verb oxygenate falls upon the same syllable as in the base noun oxygen: /ˈɒksɪdʒ(ə)n/ (OD). Similarly, vaccinate is stressed /ˈvaksɪneɪt/ (OD) because vaccine is stressed /ˈvaksiːn/ (OD) whereas adsorbate is /adˈsɔːbeɪt/ or /adˈzɔːbeɪt/ (OD) because adsorb is /ədˈzɔːb/ or /ədˈsɔːb/ (OD); the suffix -ate-ate thus again acts as a stress-neutral suffix. In /ɒkˈsɪdʒəneɪt/, by contrast, the very same suffix -ate-ate acts as a stress-shiftingstress-shifting suffix, causing the derived verb oxygenate to have a different stress pattern than the base noun oxygen. What is interesting about the case of oxygenate is that the preferred stress pattern of this verb is not the preferred stress pattern of a similar variably-stressed -ate-ate-derivative hydrogenate, which is /ˈhaɪdrəʊdʒəneɪt/ vs. /haɪˈdrɒdʒəneɪt/ according to the OED. Of 69 native English speakers who were found to have pronounced oxygenate in YouTube videos, everybody (100 %) used initial stress in this verb, i.e., /ˈɒk-/, but of 14 native English speakers who were found to have pronounced the verb hydrogenate, 10 (~71.43 %) used the antepenultimately-stressed version /haɪˈdrɒ-/. In agreement with these findings, the OD gives for oxygenate only the initially-stressed transcription /ˈɒksɪdʒəneɪt/, whereas in the case of hydrogenate the transcription /haɪˈdrɒdʒəneɪt/ is placed before the transcription /ˈhʌɪdrədʒəneɪt/. This stress difference can only be accounted for if we assume that (for the majority of contemporary English speakers) hydrogenate is not a suffixed but prefixed derivative, with the segmentally longer righthand string -rogenate-rogenate, which hydrogenate shares with nitrogenate (vacillating, according to the OD, between the stress patterns /ˈnʌɪtrədʒᵻneɪt/ and /nʌɪˈtrɒdʒᵻneɪt/), counting morphologically as the root and hence receiving stress on its first syllable. Oxygenate is, by contrast, the only -ygenate-word in the OD dictionary; English speakers have therefore no reasons to morphologically segment oxygenate into the prefix ox-ox- and the root -ygenate (or into the prefix o-o- and the root -xygenate). The morphological structure of oxygenate is thus oxygen + -ate-ate, i.e., the verb is a genuine -ate-ate-derivative and therefore, like other genuine -ate-ate-derivatives, preserves the stress of its base oxygen.

Another English suffix that has more than one distinct mode of operation with respect to stress is -al-al. Thus, for instance, while both the trisyllabic derivative personal and the disyllabic base person are stressed initially, /ˈpəːs(ə)n(ə)l/ and /ˈpəːs(ə)n/ (OD), in the derivative–base pair parental vs. parent only the latter is stressed on the first syllable: /pəˈrɛnt(ə)l/ vs. /ˈpɛːr(ə)nt/ (OD). It is tempting to say that the stress patterns of personal and parental are both in accordance with the LatinLatin Stress Rule: In /ˈpəːs(ə)n(ə)l/, stress is antepenultimate because the penult /sə/ ends in a short vowel, whereas in /pəˈrɛnt(ə)l/, stress is penultimate because the penult /ˈrɛn/ is closed. Note, however, that also in the trisyllabic derivatives parentage and parenthood, the penultimate syllable is closed, but they are nonetheless stressed /ˈpɛːr(ə)ntɪdʒ/ and /ˈpɛːr(ə)nthʊd/ (OD), preserving the initial stress of the disyllabic base parent.

An alternative explanation is that the morphological structure of parental is not the “correct” (i.e., semanticssemantics-based) parent + -al-al but pa-pa- + rental, i.e., the disyllabic righthand string rental, which occurs in English as a separate word, counts morphologically as the root and therefore attracts stress on to its first syllable. (Cf. /satɪsˈfakt(ə)ri/ of satisfactory and /ɒˈkʌlt/ of occult, whose righthand strings /‑ˈfakt(ə)ri/ and /‑ˈkʌlt/ also occur in the OD as the transcription of the initially-stressed trisyllable factory/the monosyllable cult. Cases such as satisfactory–factory or occult–cult constitute ~29.82 % of the 48,428 non-initially-stressed polysyllabic words in the OD dictionary, i.e., in 14,440 non-initially-stressed English words, the righthand string that begins with the primary-stressed syllable also occurs in the OD as the transcription of either an initially-stressed polysyllable or a monosyllable.) Likewise, simply because there is the word oral, mayoral is in American English stressed not only /ˈmeɪər(ə)l/ (OED), preserving the stress of mayor, but also /meɪˈɔr(ə)l/ (OED); pastoral is (also in American English) stressed not only /ˈpæstər(ə)l/, preserving the stress of pastor, but also /pæsˈtɔr(ə)l/ (OED); electoral is, according to Merriam-Webster Online, not only eˈlectoral, preserving the stress of elect(or), but also elecˈtoral. In the YouTube video 1oTFB9wdGl4 (14.06.2016), the stress pattern cliˈtoral is used by an American English speaker whereas in InJCUD0K2co (14.06.2016) a British English speaker can be heard saying cliˈtoral; a YouTube video in which clitoral is interchangeably stressed ˈclitoral and cliˈtoral by one and the same English speaker is _U_wKwVj8i8 (14.06.2016). These stress patterns strongly suggest that some Present-day English speakers analyze the (t)oral of these words as the root, i.e., these -al-al-derivatives are for them not suffixed but prefixed derivatives. Another similar case is the variation /ˈkant(ə)n(ə)l/ vs. /kanˈtɒn(ə)l/ of cantonal (OD). In addition to regarding it as the inherited variation /ˈkantɒn/ vs. /kanˈtɒn/ of canton (OD), the stress pattern /kanˈtɒn(ə)l/ can also be seen as the preserved initial stress of the disyllabic tonal, which occurs in English as a separate word and may therefore (from a purely formal point of view) be perceived as the root of the trisyllabic cantonal.

To reiterate, words with stress doublets challenge the popular assumption that stress in a polysyllabic English word should fall upon a particular syllable when it exhibits a particular segmental makeup (e.g., a long vowel in the penult). Additionally, cases of stress variation among suffixed derivatives raise the question of why in the case of some English suffixes, there is apparently more than one diacritic featurediacritic feature triggering a particular stress-related behavior. It is thus the author’s hope that a systematic account of the phenomenon of stress variation, which will be presented in the following chapters of this monograph, will contribute to a better understanding of the general stress assignment principles by which contemporary English speakers abide when deciding where to place stress in a polysyllabic English word.

1.2 Previous studies

With the exception of BergBerg’s (1999) study of stress differences between British and American English, none of the previous publications (on English stress) can be referred to as a systematic analysis of English words with stress doublets; the phenomenon of stress variation is only briefly touched upon in these publications as a side aspect of the more general issues pertaining to English stress.

For example, Chomsky & HalleChomsky & Halle (1968: 73) observe that in the word umbilicus, which is interchangeably pronounced /ʌmbɪˈlaɪkəs/ and /ʌmˈbɪlɪkəs/ (OED), “we have penultimate stress if the penultimate vowel is taken to be tensetense in the underlying representation, and antepenultimate stress if the penultimate vowel is taken to be laxlax.” (Note that Chomsky & Halle (1968: 73) do not speak of long vs. short vowels because “[i]n few cases in American English as a whole is time length, or durationduration, of vowels significant—that is, used to distinguish from each other words otherwise alike” (Kenyon & KnottKenyon & Knott 1953[1944]: xxvi). E.g., the phonetic contrast between the “long” /iː/ of meal and the “short” /ɪ/ of mill is in American English by and large a matter of vowel quality rather than of duration.) A problem with this explanation is, however, the above mentioned fact that apart from cases such as /ʌmbɪˈlaɪkəs/ vs. /ʌmˈbɪlɪkəs/, with stress being penultimate when the penult /ˈlaɪ/ is heavyheavy vs. antepenultimate when the penult /lɪ/ is lightlight (the LatinLatin Stress Rule is thus in both cases abided by), there are also cases such as /ˈɛkskwɪzɪt/ vs. /ɪkˈskwɪzɪt/, where stress differences do not correlate with segmental differences, i.e., stress in exquisite can be penultimate even when the vowel in the penult is phonetically realized as the short /ɪ/.

Another highly problematic explanation for the phenomenon of stress variation in English is CruttendenCruttenden’s (2008: 245) notion of rhythmic pressurerhythmic pressure. (Cruttenden (2008) is, however, the seventh edition of GimsonGimson’s Pronunciation of English. An earlier edition of the same book (Gimson 1970: 232) also mentions rhythmic pressures as one of the causes of stress variation in contemporary English.) What is meant by this is that “[i]n some words containing more than two syllables there appears to be a tendency to avoid a succession of weak syllablesweak syllables, especially if these have /ə/ or /ɪ/” (Cruttenden 2008: 245). For example, deficit is /ˈdɛfɪsɪt/ vs. /diˈfɪsɪt/ (Cruttenden 2008: 246), with the latter pronunciation representing in the view of the author a more rhythmic alternative to the former, i.e., the initially-stressed /ˈdɛfɪsɪt/ contains a sequence of two unstressed syllables both of which have the qualitatively reduced vowelreduced vowel /ɪ/. Notice, however, that the OD dictionary has 3,466 initially-stressed words in which the first syllable, bearing stress, is followed by two unstressed syllables that have /ə/ or /ɪ/ in the nucleusnucleus position. E.g., in both /ˈaldʒɪbrə/ (OD) and /ˈanɪm(ə)l/ (OD), the stressed antepenults /ˈal/ and /ˈa/ are followed by the unstressed penult–ult sequences /dʒɪbrə/ and /nɪm(ə)l/, which contain the vowels /ɪ/ and /ə/. Of the 3,466 antepenultimately-stressed trisyllables such as algebra and animal, only 58 (~1.67 %) are, according to the OD, also pronounced by contemporary English speakers with penultimate stress. E.g., in addition to the above mentioned /ˈpaprɪkə/ vs. /pəˈpriːkə/ of paprika (OD), also the adjective integral is interchangeably stressed /ˈɪntɪɡr(ə)l/ and /ɪnˈtɛɡr(ə)l/ (OD).

Observe now that apart from initially-stressed trisyllables such as algebra and animal, the OD also has 234 initially-stressed trisyllables such as, e.g., /ˈɔːtɑːki/ (OD). The unstressed penult–ult string in trisyllables such as autarky contains any long or short vowel with the exception of /ə/ or /ɪ/. Of the 234 initially-stressed trisyllables such as autarky, five (~2.14 %) are, according to the OD, also pronounced by contemporary English speakers with penultimate stress. E.g., in addition to the above mentioned /ˈnɛkrɒpsi/ vs. /nɛˈkrɒpsi/ of necropsy (OD), also autopsy is stressed both /ˈɔːtɒpsi/ and /ɔːˈtɒpsi/ (OD), conversely is interchangeably /ˈkɒnvəːsli/ and /kənˈvəːsli/ (OD), covertly is both /ˈkəʊvəːtli/ and /kəʊˈvəːtli/ (OD), and patchouli is /ˈpatʃʊli/ and /pəˈtʃuːli/ (OD). Since the difference of 58/3,466 vs. five/234 does not count as statistically significant—χ² (1) = 0.281, p = 0.5959—we can claim that trisyllables such as algebra and animal are in contemporary English not more frequently interchangeably pronounced with antepenultimate and penultimate stress than trisyllables such as autarky.

Notice also that among the examples provided by CruttendenCruttenden (2008: 246) to substantiate his claim that “[h]esitancy and variation of accentual pattern occurring at the present time are the result of rhythmic […] pressures […]” (Cruttenden 2008: 245) is also the word acumen, which, according to LDOCE, is /ˈækjəmən/ vs. /əˈkjuːmən/ (cf. the OED, where acumen is only /ˈakjᵿmən/ as far as the British variety is concerned). As one can see, when stress in acumen is antepenultimate, the unstressed vowels in the ult and the penult may undergo qualitative reduction, yielding thereby the rhythmically unfortunate pronunciation /ˈækjəmən/: This pronunciation contains a sequence of two weak syllablesweak syllables, both of which have schwas. Note, however, that according to the OED, “[p]ronunciation with stress on the first syllable was first noted in the mid 20th cent.”; the original pronunciation of acumen was the penultimately-stressed /əˈkjuːmɪn/, which etymologically is due to the LatinLatin acūmen, in which stress is penultimate because the vowel in the penultimate syllable is long. Proceeding from Cruttenden’s (2008: 245) notion of rhythmic pressurerhythmic pressure, the pronunciation /əˈkjuːmɪn/ should be seen as a more fortunate pronunciation of acumen with respect to rhythm (compared to pronunciations in which the stress is antepenultimate), but it has nonetheless been abandoned in British English in favor of the presumably less rhythmic antepenultimately-stressed pronunciation /ˈakjᵿmən/. Likewise, of 84 General American English speakers whose voices could be heard in YouTube videos containing the spoken occurrences of acumen, only four (~4.76 %) pronounced the word with a penultimate stress. Thus, we can say that not only in British but also in American English, the stress in acumen is close to becoming exclusively antepenultimate. It is fairly obvious, then, that if the variation between penultimate and antepenultimate stress in acumen did indeed have anything to do with rhythm in the sense of Cruttenden (2008: 245), we would now be observing a different tendency: More rhythmic pronunciations with stress on the middle syllable would be supplanting less rhythmic pronunciations with stress on the first syllable.

Interestingly, in stark contrast to CruttendenCruttenden (2008: 245), FriederichFriederich (1967: 25) notes that “[i]n dreisilbigen Wörtern die Mittelsilbe zu betonen, ist ein Rhythmus, der dem Engländer nicht sonderlich liegt,” i.e., placing stress upon the middle syllable in a trisyllabic word is a rhythm that an English speaker does not particularly like. To substantiate this claim, Friederich (1967: 25) refers to stress shifts such as /kəmˈpɛnseɪt/ → /ˈkɒmpənseɪt/ (OED), which were undergone by many -ate-ate-trisyllables. E.g., apart from compensate, also concentrate, confiscate, contemplate, demonstrate, illustrate, infiltrate, inundate, etc. were originally pronounced with penultimate stress but have over the course of time abandoned this stress pattern in favor of antepenultimate stress (OED). At the same time, however, some -ate-ate-trisyllables still prefer penultimate to antepenultimate stress. E.g., for demarcate, elongate, and impregnate, LDOCE gives only penultimately-stressed American English transcriptions /dɪˈmɑːrkeɪt/, /ɪˈlɒːŋɡeɪt/, and /ɪmˈpreɡneɪt/ (whereas in British English, these verbs are stressed /ˈdiː-/, /ˈiː-/, and /ˈɪm-/). More generally, Present-day English would have relatively few penultimately-stressed trisyllables if this stress pattern were indeed dispreferred by English speakers (in trisyllabic words). LDOCE has, however, (no less than) 2,479 penultimately-stressed trisyllables (vs. 4,979 antepenultimately-stressed ones). Penultimate stress in a trisyllabic English word is thus without a doubt not a marginal stress pattern.

Another aspect pertaining to rhythm is the English Rhythm RuleRhythm Rule. Because “adjacent stressed syllables make speech sound jerky” (KingdonKingdon 1949: 149) and because some English words vacillate between final and non-final stress, English is believed to have “a rule that shifts a stress leftward when a stronger stress follows,” resulting in alternations such as thirˈteen ~ ˈthirteen men (HayesHayes 1995: 18). It is acknowledged, though, that “the Rhythm Rule is optional, at least in certain contexts” (Hayes 1995: 18). For example, LangendoenLangendoen (1975: 207), who is a native speaker of English, reports that for him, final stress in Detroit in the combination Detroit Lions is as acceptable as initial stress; similarly, in the combination Marlene Deetz, final stress in the modifiermodifier Marlene is interchangeable with initial stress.

Halle & VergnaudHalle & Vergnaud (1987: 271, f. 29) also mention that

Sometimes it is also added that the English Rhythm RuleRhythm Rule “tends to apply in frequent words, e.g., ántìque bóok, and not in rare ones, e.g., àrcáne sort” (Kraska-Szlenk & ŻygisKraska-Szlenk & Żygis 2012: 327). Thus, of the finally-stressed modifiers antique and arcane, the former, which is more frequent than the latter, is likely to be initially-stressed as /ˈæntɪk/ (OED) rather than finally-stressed as /ænˈtiːk/ (OED) when the immediately following head word is either a monosyllable or an initially-stressed polysyllable; by contrast, in the less frequent arcane, stress is likely to stay final in a similar environment.

Unfortunately, statements such as these, which describe when the English Rhythm RuleRhythm Rule should apply and when it should not, are as a rule based not upon extensive corpus or experimental studies but upon authors’ own introspective judgments of how particular English words in which stress is normally final are supposed to be stressed when in an immediately following head word stress is initial. A fortunate exception is MompéanMompéan (2014), who has recently studied the stress variation exhibited by the English cardinal numeralscardinal numerals ending in -teen-teen and the corresponding ordinal numerals ending in -teenth-teenth. For this purpose the author has drawn a corpus of 1,263 newscasts from the BBC World Service Web site. These newscasts were recorded between 1999 and 2009 and feature mainly RP accents, i.e., British English speakers whose accent is Received Pronunciation. As the author reports:

Note, however, that these results do not automatically corroborate the reality of the rhythm rule in English. Thus, MompéanMompéan (2014: 156) also reports that in his corpus “head nouns often designate fractions of 100 such as per cent (14 cases) or multiples of ten such as hundred, thousand, million and billion (31 potential cases), with stress shift applying in all potential cases except six.” If the so-called English Rhythm RuleRhythm Rule is essentially nothing more than a stress clash avoidance strategy, why is stress also retracted in combinations such as thirteen per cent, in which a -teen-teen-numeral modifies a non-initially-stressed head? That is, since the pronunciation thirˈteen per ˈcent would not involve a stress clash, why is thirteen per cent nonetheless as a rule pronounced ˈthirteen per ˈcent? (A possible answer to this question is that also a one-syllable distance between two syllables bearing stress is not entirely unproblematic with regard to rhythm. Thus, “adjacent stresses are strongly avoided; stresses that are close but not adjacent are less strictly avoided; and at a certain distance (perhaps four syllables) the spacing becomes fully acceptable” (HayesHayes 1995: 372). E.g., retraction of stress in thirteen per cent occurs because in thirˈteen per ˈcent only one unstressed syllable would separate two stressed ones. A similar example, discussed by CruttendenCruttenden (2008: 296), is the alternation Westˈminster ~ ˈWestminster Abbey, where there is also no more than one unstressed syllable separating two syllables bearing stress. An interesting case is the finally-stressed modifiermodifier Barack, /bəˈrɑːk/ (Dictionary.com), which in the combination Barack Obama is often initially-stressed as /ˈbæræk/ (LDOCE) even though the stressed syllables in /bəˈrɑk oʊˈbɑmə/ (Dictionary.com) are, again, separated from each other by one unstressed one. Retraction of stress in the finally-stressed modifier /bəˈrɑːk/ occurs even though the penult /bə/ contains a schwaschwa and can thus be said to lack stress altogether, i.e., “a syllable of English is completely stressless if its vowel is schwa” (Hayes 1995: 12). The point here is that the Rhythm Rule should, according to Hayes (1995: 19), be “unable to retract stress onto a completely stressless syllable,” i.e., the finally-stressed /bəˈrɑːk/ is supposed to be the only stress pattern of Barack, even when it modifies Obama. The alternation Baˈrack ~ ˈBarack Obama thus challenges Hayes’ (1995: 19) assertion.)

Another interesting fact regarding the stress of the -teen-teen-cardinals is that “the pattern with primary stress on the first syllable is more common when counting” (MompéanMompéan 2014: 152). That is, when an English speaker counts from 13 to 19, she usually says ˈthirteen, ˈfourteen, ˈfifteen, ˈsixteen, ˈseventeen, ˈeighteen, ˈnineteen rather than ?thirˈteen, fourˈteen, fifˈteen, sixˈteen, sevenˈteen, eighˈteen, nineˈteen. Since both the former and the latter pronunciations have a one-syllable distance between the syllables bearing primary stress, we cannot say that the initially-stressed version ˈthirteen, ˈfourteen, ˈfifteen, ˈsixteen, ˈseventeen, ˈeighteen, ˈnineteen is from the point of view of rhythm better than the finally-stressed thirˈteen, fourˈteen, fifˈteen, sixˈteen, sevenˈteen, eighˈteen, nineˈteen.

Consider also compound dates involving the -teen-teen-cardinals. When these occur in the beginning of a compound date (for instance, 1821 or 1882), stress in them is almost exclusively penultimate/initial. For instance, in the YouTube videos containing the spoken occurrences of eighteen, this numeral was found to occur in the beginning of a compound date such as 1821 and 1882 87 times. Stress in eighteen was initial in each of these spoken occurrences. By contrast, in compound dates where the -teen-teen-cardinals occur in the end, there is considerable variation. For example, compound dates that end in eighteen (e.g., 1618) were found to occur in YouTube videos 27 times. With final stress eighteen was found to have been pronounced in such compound dates 19 times (~70.37 %), the initially-stressed ˈeighteen was by contrast attested only eight times: ~29.63 %. However, if we specifically consider the compound dates 1913, 1914, 1915, 1916, 1917, 1918, and 1919, the results are different. These dates were found to occur in YouTube videos 367 times. With final stress, the components 13, 14, 15, 16, 17, 18, and 19 occurring at the end of these dates were found to have been pronounced only 66 times (~17.98 %), whereas initially-stressed pronunciations could be heard by the author 301 times: ~82.02 %. Needless to repeat, the English Rhythm RuleRhythm Rule, for which alternations such as thirˈteen ~ ˈthirteen men are nothing more than a stress clash resolution strategy, cannot adequately account for the differences in the stress behavior that the numerals ending in -teen-teen exhibit in the beginning vs. the end of compound dates such as 1821 and 1918.

A less problematic explanation for why some English words have more than one stress pattern is emphasis. Emphatic stress shifts usually occur in prefixed words. Prefixes in English are generally stress-neutralstress-neutral (e.g., CruttendenCruttenden 2008: 241), but “[w]hen there is contrast or when the idea expressed by the prefix is given special prominence, the prefix bears the primary stress and the base a secondary stress […]” (PoldaufPoldauf 1984: 24). For example, triangle is supposed to be pronounced /traɪˈæŋɡ(ə)l/ (OED), with the location of stress in it coinciding with the location of stress in the disyllabic base angle, in which stress is initial. However, since a triangle is a geometrical figure that has three angles, in contrast to, e.g., a rectangle, which contains four right angles, stress in triangle is also frequently placed upon the semantically important prefix tri-tri-: /ˈtraɪæŋɡ(ə)l/ (OED; cf. the more synchronically oriented OD, which gives for triangle only the prefix stress /ˈtrʌɪaŋɡ(ə)l/; likewise, in YouTube videos featuring more than 3,000 spoken occurrences of this prefixed derivative, the root stress /‑ˈæŋɡ(ə)l/ was heard by the author only once. The prefix stress /ˈtrʌɪ-/ can thus be said to have over the course of time become the word’s default stress pattern). A similar case is /ˈsʌbməriːn/ vs. /ˌsʌbməˈriːn/ of submarine (LDOCE), with the latter pronunciation preserving the stress of the disyllabic base marine: /məˈriːn/ (LDOCE) vs. the former pronunciation emphasizing via stress the prefix sub-sub-, which distinguishes submarine from other English marine-formations (cf. aquamarine, supermarine, transmarine, ultramarine, etc.); note also that especially the meaning of supermarine, “[t]hat is situated, takes place, or operates above or on the surface of the sea” (OED; boldface mine), is the spatial opposite of the meaning of submarine, “[t]hat exists or occurs under the surface of the sea” (OED; boldface mine).