Geminate consonants by total assimilation in English

Question

Please can someone provide an example of a geminate consonant formed by total assimilation in English?

The closest I can find is from this article written by presumably Professor Ian MacKenzie:

In [the case of stops], then, a geminate must be envisaged as two discrete articulations of the consonant in question, with the first being unreleased, i.e. having no audible release burst.

To envisage what is meant by an unreleased stop, consider English words like like act, capped or dogged, in which one stop immediately follows another. Thus in act, /t/ follows /k/; in capped, /t/ follows /p/; and in dogged, /d/ follows /g/. In careful speech, both consonants in each sequence could in principle be fully pronounced, each with its in own release burst: [akt], [kapt] and [dɒgd]. However, in relaxed speech the first stop in each pair would typically be unreleased, with the speaker moving their tongue or lips into the correct articulatory position for that consonant but delaying release until the following consonant is articulated, a state of affairs which can be indicated using the ‘combining left angle above’ diacritic, as in [ak̚t], [kap̚t] and [dɒg̚d].

I am not convinced that [k̚t] is a geminate consonant, because the place of articulation of the unreleased stop is different to the place of articulation of the stop that is then released afterwards. I (believe I) can hear the difference between [ap̚t] and [ak̚t], and some languages indeed can minimally contrast by word-final unreleased stops (e.g. Thai), and I can also hear the differences between these

So does English actually form any geminate consonants apart from at word boundaries (calm man /kɑːmman/, car man /kɑːman/)? (Does /kɑːm+man/ → kɑːmman count as total assimilation?)

Answer 1

First off, there is a distinction between geminates and fake geminates. Conventionally, we tend to write [tt, ss, nn] for all longish consonants, but the term "geminate" is usually reserved for "long single consonants" and not "sequence of two distinct but identical consonant segments". Sometimes to make the distinction clear people will write [t:] vs [tt]. The phonetic difference is that [t:] is a single segmental gesture with greater duration, and [tt] is two separate segmental gestures – a real consonant. In evaluating supposed geminates, one has to check how affricates are produced, to get the right phonetic transcription. In North Saami, the word [ba:ttsi] 'obelisk (acc.sg.)' which is spelled <bácci> has a long stop closure and only one fricative release, thus not *[ba:tstsi].

English does not generally have any geminates, thus "that toy" is a fake geminate (autonomous combination of identical consonants). It is, however, sometimes thought that there are phonological assimilations, a canonical example being "phone book" said to be pronounced like "foam book". There is some evidence that the two utterances may sound the same but they are not produced the same. This is an account promulgated in the theory of Gestural Phonology, that the tongue raising gesture of "phone book" is not actually deleted, instead the lip-closing gesture overlaps the tongue-raising gesture so that there is no acoustic sign of /n/, but there is an articulatory trace. Then in an example like "ten mice", we might have just a labial gesture, or we might have a lingual gesture covered-up by a labial gesture – a fake geminate.

MacKenzie isn't claiming that [ak̚t] has a geminate, he is explaining what an "unreleased consonant" is. This obliquely refers to the two kinds of geminate. Given that there is such a sing as [ak̚t], and there is [ask], then there could also be [ak̚k], a sequence of two identical consonants the first of which is unreleased – a fake geminate. The concern would be, how can you tell whether Finnish etc. has "long k" versus "two k's in a row", given that they are potentially indistinguishable phonetically. The literature on gemination from the 80's "solved" the problem by fiat, see all the OCP literature. One part of the "solution" was the claim that assimilations always create single long segments, thus if /pt/ becomes something like [tt], it is always a single long consonant segment.

Restricting the discussion to just English, and following conventional terminology, the question is whether "mm" etc. is a single long consonant, or is it two identical short consonants the first of which is unreleased? While I am sympathetic to the reasoning that says "tɛm mais" (ten mice) presents us with a single labial gesture with extra duration, you cannot ignore the fact that [tɛ] is not a possible syllable of English. It is well known that intervocalic consonants in English get re-adjusted in some manner, so that "hammer" either has an ambisyllabic consonant (coda of σ1, onset of σ2), or it has "coda capture" thus [hæm.ɹ̩]. The assimilated form or "ten mice" could then be [tɛm.ais]. The precise phonetic details of this supposed assimilation (to use the phonological term, as opposed to the phonetic term "coarticulation") are not clear enough that we can make a definitive declaration as to what the output of "10 mice" is, and how that relates to the phonological concept of gemination.

Answer 2

"ten mice" might be an example as I found here https://www.thoughtco.com/what-is-assimilation-phonetics-1689141 (originally David Crystal, "Dictionary of Linguistics and Phonetics, 6th ed." Blackwell, 2008)

The assimilation is total in ten mice /tem mais/, where the /n/ sound is now identical with the /m/ which influenced it.