“Considered incorrect”
On wide-spread “mis”pronunciations in native speech recorded in LPD

1 Introduction

This short piece addresses particular cases where native speakers of the two major reference accents, RP and GA, pronounce words in ways that are not considered normative, that is standard pronunciation. To date, the Longman Pronunciation Dictionary (hence, LPD) by J C Wells is the only pronouncing dictionary that incorporates such useful information in its entries: “For a few words, LPD includes a pronunciation variant that is not considered correct. These variants are included because of the fact that they are in widespread use” (Wells 2008: xx). These transcriptions are assigned “a special warning sign: the exclamation mark enclosed in a triangle” (Wells 2014: 2) (Wells occasionally claims, as in the case of tortoise (2014: 20-21), that he gave such a warning triangle in LPD to some words which do not actually have it in LPD.) Drawing attention to cases of “incorrect variants” in “wide-spread use”, some 150 recorded in LPD, has a practical pedagogical import, for instance in phonology/phonetics or pronunciation courses at an advanced level, since it gives a better picture of native pronunciation from a variationist point of view and also focuses learners՚ attention on what native speakers do who get the pronunciation “right”. A phonologistʼs view will not bear on the “incorrectness” (or otherwise) of these cases, and this coincides with Wells՚ lexicologist position (2014: 2). From a phonological perspective, such examples are revealing because they are informative about phonotactic, morphophonological or accentuation patterns and analogies that native speakers apparently draw upon. Irrespective of whether, from a prescriptivist point of view, they could be seen as incorrect, they are definitely not so for a phonologist taking stock of what constitutes phonological knowledge to native speakers.

Not much is known about the collection process Wells used to collect his data and about just how widespread these forms really are, although he did use native informants to answer a detailed questionnaire for “many words of uncertain pronunciation” (Wells 2008: xviii). At least on one occasion, such a poll investigated a case of “incorrect” pronunciation, the alternative pronunciations of asterisk, standardly pronounced /'æstərɪsk/ both in BrE and AmE. While in BrE /'æstərɪks/ is marked “incorrect” in LPD without further comments, the poll found the following preferences in AmE: /'æstərɪks/ 9%, /'æstərɪk/ 29%, /'æstərɪsk/ 62%, but LPD does not mark the least preferred variant as “incorrect”. Also, nothing is known about whether certain “incorrect” variants occur cumulatively in certain speakers or not. I have not carried out any empirical research on how widespread the alternative pronunciation is for the lexical items in LPD, nor have I investigated if further variants exist for the same lexemes or indeed for others LPD might ignore. I have considered these items essentially as minor alternative pronunciations and I cannot (and would not) draw the line between those marked as “incorrect” and other alternative pronunciations in LPD. I have generally ignored cases like ylang-ylang (when BrE /jəˌlæŋjə'læŋ/) as well as trivial spelling pronunciations (such as Ynys /'ɪnɪs/ for /'ʌnɪs/), haplologies (like /aɪ'tɪnəri/ in BrE for itinerary /aɪ'tɪnərəri/), personal names (often used as brand names) and place names, since many of these are commented on in Wells (2014). I have limited my remarks to a selection of specific cases LPD records and have concentrated on cases offering phonological or morphophonological matter to discuss.

2 Variants based on falsely identifying morphemes

The following variants can be described as falsely identified compounds or words with pseudo-prefixes. They are examples of folk etymology. The word acupuncture when pronounced in BrE as /'ækwəpʌŋktʃə/ for /'ækjupʌŋktʃə/ seems to contain a first element aqua-, which is unetymological here. The word asphalt, pronounced sometimes in BrE as /æs'felt/ or /æʃ'felt/ for /æs'fælt/ is treated as if it contained felt or even ash as its components. The name of the shrub cotoneaster, BrE /kəˌtəʊni'æstə/ and GA /kə'toʊniæstɚ/, also shows a variant based on folk etymology: BrE /ˌkɒtən'iːstə/, AmE /'kɑ:təniːstɚ/. The word tuberose, pronounced /'tjuːbrəʊz/, “by folk etymology” (remark in LPD entry), for /'tjuːbərəʊz/, looks as if it was a compound of tube+rose, while etymologically it is tuber-ose (cf. Hungarian tubarózsa for a similar folk etymological solution for this word and French tubéreuse for an etymologically better fit). The BrE “incorrect” pronunciation of hyperbole as /'haɪpəbəʊl/ for /haɪ'pɜːbəli/ can be described as driven by treating hyper- as phonologically free, so hyper+bole. Some would probably treat some of these examples as spelling pronunciation.

3 Transparent derivations

A handful of examples take a semantically related free morpheme as the base for the derived form: they are essentially transparent derivations. The word diarist illustrates this point in both reference accents when it is pronounced as if it was diary+ist in a transparent derivation, that is without truncation: /'daɪəriɪst/ for truncated /'daɪərɪst/. A pronunciation based on transparent derivation also occurs in both reference accents in the deverbal adjective preferable: /pri'fɜːrəbəl/ (for /'prefərəbəl/). This carries over to preferability as well. One only wonders if there is a semantic difference that goes unnoticed in this case? All the other examples LPD gives are exclusively from BrE, never AmE:

Notice that the variant with /dʒ/ in tautologous is considered incorrect in LPD, while the (morpho)phonologically similar analogous, which also has a variant in /dʒ/ in both accents, is considered to have a secondary variant and not an “incorrect variant”. One sometimes wonders where the boundary was drawn. When tautologous and purgative are pronounced with /dʒ/, the sound values of 〈g〉 can also play a role, of course, as is the case with margarine /ˌmɑːdʒə'riːn/. The pronunciation of the derivation from burgle is particularly amusing since the verb to burgle is etymologically a backformation from burglar and now it is interpreted, for these native speakers, as the base of their morphologically transparent pronunciation /'bɜːɡəlri/. On the other hand, it also shows that the word is treated as a complex word with a free base, much like deverbal revelry, rivalry and unlike words with a bound morpheme such as salary or vocabulary.

4 Problems when choosing the standard suffix or the allomorph of a suffix

Among -logy words, in BrE only two have general currency that have -a- rather than -o- as a combining vowel-letter (besides analogy, but it does not have the same morphological structure and -logy definitely does not mean a branch of science). The difference does show in the standard pronunciation because it is the nucleus of the stressed syllable: /æ/ versus /ɒ/. In BrE both these -logy words show levelling from /æ/ to /ɒ/ for some speakers:

These examples illustrate sharp morphological knowledge from native speakers since the word tetralogy does not seem to show a similar wide-spread “incorrect” variant *tetrology: this word does not belong semantically with the science -ology words. Moreover, it is interesting to note that the standard AmE pronunciation corresponds to the “incorrect” BrE pronunciation since it contains the lot vowel (in the sense of Wells 1982) rather than the trap vowel unlike the standard BrE variant. This latter observation is not commented on in LPD. Note that the American pronunciation is consistent in all -logy words: /te'trɑːlədʒi/, thus belonging to the lot lexical set. These -logy words can be taken to belong to the lot set because they do not resemble another, much larger, group of words where the lexical incidence of /æ/ versus /ɑː/ varies between BrE and AmE: the group of mainly Spanish and Italian words whose nativization hinges precisely on the choice of the vowel corresponding to their original /a/. BrE /æ/ corresponds to AmE /ɑː/ in the following words among others: Vivaldi, taco, pasta, macho, mafia. These can be regarded as different from the -logy words under discussion since they do not share their morphological structure or their origin. Finally, another -logy word needs to be mentioned because it shows an “incorrect” variant in BrE in the first element this time: meteorology /ˌmiːtiə'rɒlədʒi/, and its derivatives, can have /ˌmiːtə'rɒlədʒi/, apparently interpreted as having meter as its first element. This word could, however, be placed in (4b) below.

There are a couple words whose suffix in the standard pronunciation is replaced by another, similar, suffix. The word gypsophila /dʒɪp'sɒfɪlə/ has an “incorrect” variant in BrE: /ˌdʒɪpsə'fɪliə/. It is one among three biological words listed in the OED Online with -(o)phila (drosophila is a fruit-fly, and gypsophila and Nemophila are a flower and a genus of plants, respectively), as opposed to more than 20 items listed with -(o)philia, such as haemophilia, p(a)edophilia, necrophilia, Anglophilia, cinephilia, audiophilia, among others. These latter often have an even more frequent corresponding adjective such as cinephile, audiophile, etc. Thus the affiliation of -phila to -philia is not particularly surprizing, all the more so since gypsophila has the right number of syllables to accommodate the stress pattern ˌ– –'– –. No such “incorrect” variants are recorded for drosophila and Nemophila in LPD, however. A similar case is provided by urethane and polyurethane, which have -thane in standard /(ˌpɒli)'jʊərəθeɪn/. In BrE the variant /(ˌpɒli)'jʊərəθiːn/ is also recorded, which corresponds to the suffix -thene in the same semantic domain.

Finally, there are cases of allomorphy. The adjective promissory, normally pronounced as BrE /'prɒmɪsəri/, AmE /'prɑːməsɔːri/, has a variant in AmE, /'prɑːməseri/, that is analogical to necessary and commissary, a different and more frequent allomorph of the same suffix. (BrE does not show this pattern since here the penultimate vowel is reduced to [ə].) Another example is grievous, where besides standard /'griːvəs/, “incorrect” /'griːviəs/ is also recorded. Rather than treating this word as belonging to those discussed in (5) below, it is important to point out that a comparison between -vous and -vious adjectives (-veous is negligeable) shows that the second set includes very frequent and some phonologically very simlar adjectives like previous, devious and obvious, envious, etc. The -vous set interestingly includes mischivous, whose /mɪs'tʃiːviəs/ variant is considered standard! The only other frequent word in that set is nervous. These considerations weigh in favour of treating /'griːviəs/ as analogical. A similar reasoning can be proposed for heinous and intravenous. The standard pronunciation is /'heɪnəs/ (or /'hiːnəs/) and /ˌɪntrəˈviːnəs/, respectively. The “incorrect” pronunciations are /'heɪniəs/ (and /'hiːniəs/) and /ˌɪntrəˈviːniəs/, where the penultimate stress is more consistent in -neous, -nious words (spon'taneous, homo'geneous, har'monious, acri'monious, etc), since frequent -nous words have antepenultimate stress (in'digenous and family, 'luminous, 'ominous, 'poisonous, etc). The interaction of stress-patterns and allomorph selection favour analogical reassignment in /'heɪniəs/ and /ˌɪntrəˈviːniəs/.

The following two words are interpreted by some native speakers of both accents as containing the suffix -ant:

This usage extends the occurrence of the suffix -ant to words that quite likely do not contain any suffix for most native speakers of the reference accents. Selection of a different suffix is found in the case of dominant where AmE shows a variant that apparently corresponds to an adjective *dominate, making it look like one of the adjective-verb pairs in -ate such as approximate_v–approximate_adj: AmE /'dɑːmɪnət/ for /'dɑːmɪnənt/.

5 Alternation between C{ə i u} and C{jə ju}

Possibly the most interesting set of words to discuss is those where the reduced vowels [ə u] are preceded by [i j]. This is partly phonotactics, partly a question of allomorphy. This is also the group of words where it would matter to take the broader morphophonological context and patterns into account.

The first group of such words is a set of two types where Cjə alternates with Cə, and the latter is considered “incorrect”. The two types differ in whether the Cjə sequence is from /ju/ or from /iə/ and what the preceding consonant is. They are presented in (4a)–(4b) below.

In (4a), the variants have [ju] or [ə] in an unstressed syllable following a (primary- or secondary-)stressed syllable. In the case of particular this can result in contraction from ?*/pə'tɪkələ/ to /pə'tɪklə/. This happens after velars in BrE only, otherwise the only example in both reference accents also has a non-coronal /m/ before the alternation site. Words in (4b) have /jə/ in the same phonological context as those in (4a) but here (/iə/>) /jə/ in the standard variant does not derive from /ju/ and is after coronals. The distribution of these examples across the reference accents shows less coherence, however, BrE still being ahead.

The other group illustrates the inverse of the previous group in that the “incorrect” variants are associated with the presence of /j i/ where the standard has none. This insertion occurs in an unstressed syllable following a (primary- or secondary-)stressed syllable, just like in (4) above. But the preceding phonological context is not identical at all to that in (4). Words in (5a) include those cases where a nasal or oral stop or an affricate precedes an inserted /ju/. However, /t/ only occurs here when followed by /ɪ/, and /tʃ/ only when it is the result of palatalization across a morpheme boundary.

The examples /ˌkɒnju'teɪʃn/, /'pestjʊlənt(s)/ and /'pɜːkjuleɪt/ are relevant because they reveal that in fact it is not simply a palatal glide that gets inserted but /ju/. When the “incorrect” variant has [jə] and no /ju/ variant is recorded, the former can therefore be described as a reduced variant of /ju/ in all the examples in (5a). This generalization also holds for cases where, following the palatal affricate, /u/ appears before the schwa. Under this approach, nuclear /'njuːkjələ/ and nucleus /'njuːkjələs/ are described as cases where native speakers using these forms have a hetero-syllabic cluster /k.l/ rather than /.kl/ as the result of a putative suppression of /kə.l/, which they hypercorrect to /kju.l/. Of course, there is ample support for /ju/ in the graphic arrangement of these particular words and this can also be assumed to play a rôle. An analogical explanation for these cases is proposed for at least some of the words above in Wells (2014: 2) where spectacular, circular, molecular are evoked as analogical models. Such an explanation, however, makes reference to relatively long strings of multiple syllables in, sometimes, rare words to serve as a model for a presumed analogy: what would trigger /ɪk'stræpjəleɪt/? It would be a very interesting task to tease out the rôle of the various morphological influences: they are much more complex here than in those cases where analogy was evoked in this paper.

In (5b), the variation site is preceded by a liquid or /t d/, in other words, all non-fricative alveolars. (Fricatives do not participate in these variants.) In this sense, (5a) and (5b) are complementary and the inserted material is different accordingly: /ju/ in (5a) but /i/ in (5b). This /i/ is inserted before a vowel and there is no trace of /u/ following it. The inserted /i/ triggers palatalization when following /t/.

6 Variants based on phonotactic restrictions

These words illustrate various cases related to phonotactic restrictions, affecting single sounds or parts of syllables. Some speakers do not pronounce the preconsontal /k/ in the following words:

The inverse pattern is also found: they do sometimes pronounce a /k/ in homiletic, /ˌhɒmɪ'lektɪk/ in BrE, for /ˌhɒmɪ'letɪk/. It seems that in this case it is the specific environment /-le_tɪk/ that triggers an unetymological /k/. A search in OED Online shows that although -etic actually has more item frequency than -ectic, there are more -lectic words in the semantic field of homiletic.

Cases where consonants in consecutive syllables metathesize are instructive because they reveal phonotactic preferences and morphological or semantic parallels. Notice, crucially, that the “incorrect” variants include syllables that are sometimes existing words:

Sometimes there is a semantic feature that seems to trigger metathesis. When in BrE prodigy is pronounced /'prɒdʒədi/ (as opposed to standard /'prɒdədʒi/), the word becomes nearly homophonous with progeny, and often there is semantic overlap between the two in this case (cf. child prodigy). The word remuneration, /riˌmju:nə'reɪʃn/, is also pronounced as if it was renumeration: /riˌnju:mə'reɪʃn/ (also in remunerative, but not in the verb remunerate in LPD). The OED Online entry notes the persistence of this “mis”pronunciation, sometimes given in spelling too, since the mid-16^th century in the meaning “to remunerate”.

In this group of words showing metathesis of consonants, the word enmity, normally /'enməti/, is particularly interesting. The “incorrect” variant in BrE is /'emnəti/. There are no other *#emn- words in English so the initial cluster in the “incorrect” variant is surprizing at first sight as a phonotactically better fit. However, it turns out that -en.m- is not very frequent in general, occurring in enlighenment, Denmark, and enmity itself among a couple of other low-frequency items. But -em.n- is more frequent, occurring in words often with a relatively rich morphological network such as indemnity–indemnify, solemnity, condemnation or remnant and this seems to be enough to give some currency to /'emnəti/ as a better phonotactic fit.

As pointed out by Wells (2014: 2), the words Messerschmidt and schnitzel include phonotactically rare clusters in English involving sequences of sibilants and their clusters, and the “mis”pronunciations resolve them. Phonotactically syllable-initial /sn sm/ is frequent in English, while these words of German origin violate this having /ʃn ʃm/ instead. In schnitzel, the status of medial /ts/ is also rare as a heterosyllabic cluster /t.s/ (and it cannot directly correspond to affricate /ts/ in German), /tʃ/ resolves this anomaly too:

Finally, while the standard pronunciation of arthritis is /ɑː'θraɪtɪs/ (with rhoticity in the first syllable in AmE), the non-standard BrE pronunciation inserts a [ə] between [θ] and [r]: /ˌɑːθə'raɪtɪs/. The phonologically closest parallel in -thritis is urethritis and it is pronounced as a sequence of a weak foot and a strong foot: /ˌjʊərə'θraɪtɪs/. Otherwise, Cl- and Cr- clusters before -itis are very rare according to OED Online: cyclitis, *-c(.)ritis, *-p(.)litis, *-p(.)ritis, *-t.litis, ga.stritis, me.tritis. The crucial evidence for the syllabification of the words just listed should come from patterns of aspiration and the quality of /r/ following /t/: hopefully this has been done for these words to justify their syllabification in the OED (Wells 2008 only has ga.stritis from the list, hopefully checked for syllabification). The word arthritis is problematic in BrE since in the standard pronunciation the first syllable only contains a long vowel before the primary stress as opposed to the other four words that have a light CV first syllable. Also, there are no phonetic cues to decide whether [θɹ] goes in the onset or is hetero-syllabic. Furthermore, arthritis is also different from its phonological parallel, urethritis, in not having two syllables before the primary stress; that is, unless one pronounces it in the “incorrect” way: /ˌɑːθə'raɪtɪs/.

7 Lexical incidence

There are also some cases of “mis”pronunciations due a difference in the lexical incidence of particular vowels or consonants. In these cases, the vowel or consonant occurs in a phonotactically grammatical position, but this sound is different from the one in the standard variant. To begin with the vowels, the verbs catch and tack, belonging to Wells՚s trap lexical set (cf. Wells 1982), have dress for some speakers so they pronounce them /ketʃ/ and /tek/, respectively. Similarly, league belongs to the kit set for some AmE speakers: /lɪɡ/, as opposed to /liːɡ/, in the fleece set. The BrE pronunciation of the verb form ate, /et/, is considered “incorrect” in AmE. As far as consonants are concerned, a nice example is the AmE “incorrect” pronunciation of drought, for standard /draʊt/: /draʊθ/, with /θ/ in word-final position. Even the LPD entry explains that this is originally the pronunciation of a doublet, drouth, which is now archaic or obsolete. It may be archaic and obsolete in spelling, but the “incorrect” pronunciation confirms that it is not so archaic in the speech of some native speakers of AmE. Also, in a few words in BrE, instead of initial single-onset /p/ in an unstressed syllable, one finds /b/: pajama /bə'dʒɑːmə/ for standard /pə'dʒɑːmə/, and potato /bə'teɪtəʊ/ for standard /pə'teɪtəʊ/.

Concluding remarks

I have tried to show in this paper how the alternative pronunciations, considered “incorrect” by some speakers, can be described in ways that make reference to morphological or phonological information concerning the site of variation. The main idea was precisely to give a description that does not treat each word on its own but in groups, highlighting broader patterns that seem to be at work. In some case, such as the loss of preconsonantal /k/ in certain words, there is not much point in going beyond taking stock of the examples: for some speakers that is the lexical representation of those words. Other cases, however, offer much more room for discussion, especially the allomorphy cases and those presented in (4) and (5).

Ádám, I hope you donʼt agree with many aspects of what I have presented here so it can give you stuff to think about! Happy birthday!

References