Ask LLOG: "friends" vs. "flense"

Query from reader RR:

Just trying to get unpaid labor from a phonetician here…

I've written a puzzle which involves swapping out one phoneme for another in various words. A couple of testsolvers have objected that "flense" doesn't become "friends" if you change the second phoneme; they insist they pronounce the D in "friends" (or don't have a D in the transition from N to Z in "flense", if you prefer).

Try as I might, I can't pronounce those two words such that they don't rhyme exactly, at least without sounding like an idiot. And like all people, I of course believe my self-judgment of phonetics is better than average. :-)

To start with, there's an issue with flense.

The OED gives three pronunciations, rhyming with bench, finch, and dense — and none of those pronunciations are an appropriate basis for RR's puzzle, since to pair with friends he needs something that ends in [nz], not [nʃ] or  [ns] — that is, something that rhymes with dens, not bench or  dense. Merriam-Webster gives only the rhymes-with-dense pronunciation (which is what I assumed, though not with much confidence), and so does Wiktionary.

Apparently RR has /flɛnz/ rather than /flɛns/, as indicated by his focus on the issue of whether people "pronounce the D in 'friends' or don't have a D in the transition from N to Z in 'flense'". That's an interesting and difficult question, if we're applying it to tends vs. tens, or  tents vs. tense, etc.

In case of English monosyllabic wordforms ending in /ndz or /nts/, we're in the domain of what sociolinguists call "t/d deletion". This is an unfortunate and misleading choice of terminology, referencing a range of phenomena from gradient reduction to full lexical substitution.

From an articulatory point of view, you can think of such words as a trio for velum, tongue, and larynx. After the release of the syllable onset (if any), the velum opens through the vowel, and remains open for the nasal murmur if there is one, but needs to be shut again for the  [d] or [t] (if any) and the [z] or [s]; the tongue tip closes for the [n] and [d] or [t] (if any) and then arranges a narrow opening to generate the fricative noise; and the larynx maintains voicing at least through the vowel and the [n] and [d] (if any), but usually arranges for voicing to weaken or stop for the fricative, even if it's nominally a /z/.

The coordination among these articulators spans an rich space of rate, amplitude, and phase relations, whose acoustic consequences are widely and interestingly variable. My opinion, FWIW, is that there is no such thing a rule of "t/d deletion" or a rule of "t/d epenthesis". Rather, that space of rate, amplitude and phase relations creates overlapping distributions of phonetic realization for words ending any of /ndz/, /nz/, /nts/, /ns/. It's easy for phonologists to interpret this phonetic variation in terms of deletion or epenthesis of segments, even when the speaker's phonological intent is what the standard dictionary says it should be. And sometimes speakers do the same thing, and internalize alternative pronunciations with fewer or more consonants than the standard dictionary prescribes.

In any case, there's  no question that in (most varieties of?) English, normal fluent productions of these phonological categories are (often) ambiguous.

To illustrate this point, here's a little quiz on lend v. lens. There are 12 audio clips, taken from passages in LibriSpeech, a small (1000 hour) sample of the LibriVox collection of public-domain audiobooks. Each one is either the word lend or the word lens — see if you can guess the answers:

1 Your browser does not support the audio element. 2 Your browser does not support the audio element. 3 Your browser does not support the audio element. 4 Your browser does not support the audio element. 5 Your browser does not support the audio element. 6 Your browser does not support the audio element. 7 Your browser does not support the audio element. 8 Your browser does not support the audio element. 9 Your browser does not support the audio element. 10 Your browser does not support the audio element. 11 Your browser does not support the audio element. 12 Your browser does not support the audio element.

The answers are here, along with links to the phrases that the election came from.

This is a bit unfair, since there are lot of reasons that it's hard to process words taken out of phrasal context, but it's a start.

My belief is that a larger and more careful experiment would show that this is an example of a "near merger", where the realizations of two historically distinct phonological classes comes to overlap (at least in some contexts) to the point that listeners can no longer distinguish them reliably, although speakers continue to produce statistically different distributions.