Flapping

The terms flap and tap are often used synonymously, although some authors make a distinction between them. When the distinction is made, a flap involves a rapid backward and forward movement of the tongue tip, while a tap involves an upward and downward movement. Linguists disagree on whether the sound produced in the present process is a flap or a tap, and by extension on whether the process is better called flapping or tapping, while flapping has traditionally been more widely used. Derrick & Gick (2011) identify four types of sounds produced in the process: alveolar tap, down-flap, up-flap, and postalveolar tap (found in autumn, Berta, otter, and murder, respectively).

In Cockney, another voiced variant of /t/ that has been reported to occur to coexist with the alveolar tap (and other allophones, such as the very common glottal stop) is a simple voiced alveolar stop [d], which occurs especially in the words little [ˈlɪdʊ], hospital [ˈɒspɪdʊ] and whatever [wɒˈdɛvə]. That too results in a (variable) merger with /d/, whereas the tap does not.

In Cardiff English, the alveolar tap is less rapid than the corresponding sound in traditional RP, being more similar to /d/. It also involves a larger part of the tongue. Thus, the typical Cardiff pronunciation of hospital as [ˈɑspɪɾl̩] or [ˈɑspɪɾʊ] is quite similar to Cockney [ˈɒspɪdʊ], though it does not involve a neutralization of the flap with [d].

Flapping of /t/ and /d/ is a prominent feature of North American English. Some linguists consider it obligatory for most American dialects to flap /t/ between a stressed and an unstressed vowel. Flapping of /t/ also occurs in Australian, New Zealand and (especially Northern) Irish English, and more infrequently or variably in South African English, Cockney, and Received Pronunciation.

The exact conditions for flapping in North American English are unknown, although it is widely understood that it occurs in an alveolar stop, /t/ or /d/, when placed between two vowels, provided the second vowel is unstressed (as in butter, writing, wedding, loader). Across word boundaries, however, it can occur between any two vowels, provided the second vowel begins a word (as in get over [ɡɛɾˈoʊvɚ]). This extends to morphological boundaries within compound words (as in whatever [ˌwʌɾˈɛvɚ]). In addition to vowels, segments that may precede the flap include /r/ (as in party) and occasionally /l/ (as in faulty). Flapping after /l/ is more common in Canadian English than in American English. Syllabic /l/ may also follow the flap (as in bottle). Flapping of /t/ before /ən/ (as in button) is observed in Australian English. In North American English, [t] and [ʔ] (t-glottalization) were the only realizations of /t/ before /ən/, but studies in the 2020s have found [ɾ] in younger US speakers.

Morpheme-internally, the vowel following the flap must not only be unstressed but also be a reduced one (namely /ə/, morpheme-final or prevocalic /i, oʊ/, or /ɪ/ preceding /ŋ/, /k/, etc.), so words like botox, retail, and latex are not flapped in spite of the primary stress on the first syllables, while pity, motto, and Keating can be. The second syllables in the former set of words can thus be considered as having secondary stress.

Word-medial flapping is also prohibited in foot-initial positions. This prevents words such as militaristic, spirantization, and Mediterranean from flapping, despite capitalistic and alphabetization, for example, being flapped. This is known as the Withgott effect.

In North American English, the cluster /nt/ (but not /nd/) in the same environment as flapped /t/ may be realized as a nasal flap [ɾ̃]. Intervocalic /n/ is also often realized as a nasal flap, so words like winter and winner can become homophonous. According to Wells (1982), in the United States, Southerners tend to pronounce winter and winner identically, while Northerners, especially those from the east coast, tend to retain the distinction, pronouncing winter with [ɾ̃] or [nt] and winner with [n].

Given these intricacies, it is difficult to formulate a phonological rule that accurately predicts flapping. Nevertheless, Vaux (2000) postulates that it applies to alveolar stops:

• after a sonorant other than l, m, or ŋ, but with restrictions on n;
• before an unstressed vowel within words, or before any vowel across a word boundary;
• when not in foot-initial position.

Exceptions include the preposition/particle to and words derived from it, such as today, tonight, tomorrow, and together, wherein /t/ may be flapped when intervocalic (as in go to sleep [ˌɡoʊɾəˈslip]). In Australian English, numerals thirteen, fourteen, and eighteen are often flapped despite the second vowel being stressed. In a handful of words such as seventy, ninety, and carpenter, /nt/ is frequently pronounced as [nd], retaining /n/ and voicing /t/, although it may still become [ɾ̃] in rapid speech.

Flapping is a specific type of lenition, specifically intervocalic weakening. It leads to the neutralization of the distinction between /t/ and /d/ in appropriate environments, a partial merger of the two phonemes, provided that both /t/ and /d/ are flapped. Some speakers, however, flap only /t/ but not /d/. Yet, for a minority of speakers, the merger can occur only if neither sound is flapped. That is the case in Cockney, where /t/ is occasionally voiced to [d], yielding a variable merger of little and Lidl. For speakers with the merger, the following utterances sound the same or almost the same:

In New Zealand, Māori loanwords may retain the original [ɾ] pronunciation of ⟨r⟩. This, in combination with flapping, leads to homophones such as Kura and coulda, both pronounced [kʊɾə].

In accents characterized by Canadian raising, such words as riding and writing may be flapped yet still distinguished by the quality of the vowel: riding [ˈɹaɪɾɪŋ], writing [ˈɹʌɪɾɪŋ]. Vowel duration may also be different, with a longer vowel before /d/ than before /t/, due to pre-fortis clipping.

In a dissertation in 1982, M. M. Withgott demonstrated that, among speakers of American English, words seem to be chunked into pronunciation units she referred to as a foot, similar to a metrical unit in poetry. Such chunking was said to block flapping in the word ‘Mediterranean’ ([[Medi[terranean] ], cf. [ [sub[terranean]]). How a word is chunked relates to its morphological derivation, as seen by contrasting morphologically similar pairs such as the following (where the vertical bar shows where Withgott argued there is boundary between neighboring feet):

The medial t in càpitalístic can be flapped as easily as in post-stress cátty [ˈkʰæɾi], in contrast to the medial t in mìlitarístic, which comes at the beginning of a foot, and so must be pronounced as [tʰ], like a t at the beginning of a word.

Long, seemingly monomorphemic words also are chunked in English for purposes of pronunciation. In such words [t]’s, as well as the other unvoiced stops, are pronounced like initial segments whenever they receive secondary stress or are at the beginning of a foot.

The T-to-R rule is a phenomenon in certain accents of English in which the flapping of /t/ is subsequently reinterpreted as the flap variant of /r/. As a result, in most such accents, this type of /t/ is then replaced by the prevailing variant of /r/ in English, the approximant [ɹ]. However, studies have found that the rule is more restrictive that usual flapping rules, with the phenomenon occurring only in a particular set of frequently used words and mainly across word boundaries (in phrases), but only very rarely does it occur word-internally (thus, it is common in shut up but much rarer in shutting).

The application of the rule means that shut in the phrasal verb shut up /ʃʌrˈʌp/ can be analysed as having a different phonemic form than the citation form of the verb shut /ʃʌt/. Other examples of phrases where [ɹ] can replace [t] in such accents include get off, let us, that again, forget all, bought it, and what is.

The T-to-R rule has been reported to occur primarily in certain 19th- or 20th-century working-class dialects, including of the English Midlands and Northern England (where it is always stigmatized) such as Geordie, older Cockney (London), certain local Dublin accents of Ireland, the Cardiff dialect of Wales (where the merged consonant can surface as either an approximant or a flap), some South African English (where only a flap is possible), and some Scottish English. In the Cardiff dialect, the rule is typically applied between any vowel (including long vowels) and /ə/ or the reduced /ɪ/ (also across word boundaries), so that starting /ˈstaːtɪŋ/ and starring /ˈstaːrɪŋ/ can be homophonous as [ˈstaːɹɪn ~ ˈstaːɾɪn]. In South African English, the merger is possible only for those speakers who use the flapped allophone of /r/ (making the starting, starring minimal pair homophonous as [ˈstɑːɾɪŋ]), otherwise the sounds are distinguished as a flap (or a voiceless stop) for /t/ ([ˈstɑːɾɪŋ ~ stɑːtɪŋ]) vs. approximant for /r/ ([ˈstɑːɹɪŋ]). There, the merger occurs word-internally between vowels in those environments where flapping is possible in North American English.