The House Carpenter

@favorite-phoneme-creature

There are 16 distinct phonemes, which fall into four natural classes:

Checked vowels: /ɪ o ɛ a ɑ/

Schwa: /ə/

Long vowels: /oː əː ɑː/

Diphthongs: /ij ʉw oj ɛj əw aj aw/

These are distinguished on the following basis:

Checked vowels only occur immediately before a tautomorphemic consonant; schwa only occurs in unstressed syllables; long vowels and schwa never occur before a vowel; diphthongs can appear anywhere.

Checked vowels and schwa are always short, long vowels are always long, and diphthongs vary in length depending on the environment.

Diphthongs are phonetic diphthongs at least some of the time; the other vowels are always monophthongs.

The constraint that long vowels and schwa never occur before a vowel is enforced across morpheme and word boundaries, by insertion of /r/ between any long vowel or schwa and any following vowel. For example drawing /droːrɪng/, there it is /ðəːrɪtɪz/.

checked vowels

Lexical sets:

/ɪ/ = KIT

/o/ = STRUT, FOOT

/ɛ/ = DRESS

/a/ = TRAP, BATH

/ɑ/ = LOT, CLOTH

Geoff Lindsey's description of /ɪ ɛ a/ for SSB is also accurate my accent. For the LOT vowel, however, which Lindsey transcribes as /ɔ/, I don't see any rounding in my lips when I pronounce it, nor do I hear any difference in quality between it and the /ɑː/ vowel of START, so I transcribe it as /ɑ/. Admittedly, from listening to the bossy / posso recordings in that Lindsey blogpost it seems I am unable to distinguish [ɒ] from [ɔ] by ear, so it's quite possible there is a quality difference between /ɑ/ and /ɑː/ that I simply don't perceive.

As for /o/, this is quite different from either the /ʌ̈/ of SSB STRUT or the /ɵ/ of SSB FOOT. If I listen closely, I can perceive a slight quality difference between it and /oː/: it's a little higher and more centralized. So /ʊ/ would not be an unreasonable transcription either. But since I think length is the primary distinguishing factor between /o/ and /oː/ ([bod] sounds much more like bud than board, and [bʊːd] sounds much more like board than bud), it seems more appropriate to use the simpler symbol /o/ for both vowels.

long vowels and schwa

Lexical sets:

/ə/ = commA, lettER

/oː/ = THOUGHT, NORTH, FORCE, CURE

/əː/ = NURSE, SQUARE

/ɑː/ = PALM, START

The qualities of the vowels are the same as in SSB, though note that SSB doesn't merge NURSE and SQUARE; my /əː/ is pronounced like SSB NURSE, rather than SSB SQUARE.

Some possible quality differences that may exist between /o ɑ/ and /oː ɑː/ were noted above, but they are so subtle compared to the quantity difference that it seems unreasonable to not consider this a principally quantity-based contrast, with at most a minor contribution from quality. Therefore I'd say that my accent pretty much has phonemic vowel length.

/əː/ is a slightly different case. The quality of /ə/ is pretty variable; I haven't considered it in much detail but I think it's generally low-mid word-finally, and higher elsewhere. The quality of /əː/ is more fixed: it's consistently low-mid, like word-final /ə/.

SSB has up to two additional long vowel phonemes /ɪː/ and /ɵː/ found in the NEAR and CURE lexical sets, respectively. But /ɵː/ has a strong tendency to merge with /oː/, and this merger is fully complete for me. As for /ɪː/, for me it corresponds to /ij/, realized as [iː], before /r/, and a sequence of two phonemes, /ijə/, realized as [ɪjə], elsewhere. The /ijə/ pronunciation can also be levelled into forms with a following /r/: for me nearer can be either /nijrə/ or /nijərə/, the latter based on near /nijə/. I would regard /ijə/ as occupying two syllables, but this is somewhat arbitrary (I don't believe that there is a context-independent phonetic definition of "syllable" that could decide this question).

diphthongs

Lexical sets:

/ij/ = FLEECE, happY

/ʉw/ = GOOSE

/oj/ = CHOICE

/ɛj/ = FACE

/əw/ = GOAT

/aj/ = PRICE

/aw/ = MOUTH

The basic phonetic realizations are similar to SSB. One clear difference is that my PRICE is /aj/, rather than /ɑj/, with a first element identical in quality to that of MOUTH.

Unlike the monophthongs, which are all pretty much pronounced the same in all environments, except for schwa, the diphthongs tend to vary allophonically quite a bit.

As mentioned above, all diphthongs vary in length depending on the environment: each diphthong has a short allophone and a long allophone. Some diphthongs also vary in quality; the conditioning environments for lengthening, and the ways in which length manifests, also vary; on this basis, we can distinguish three overlapping subgroups:

the high diphthongs, /ij/ and /ʉw/

the centralized diphthongs, /ʉw/ and /əw/

the low diphthongs, /aj/ and /aw/

The distinctive feature of high diphthongs is that they aren't always diphthongs. They are clearly diphthongs word-finally, or before a vowel, where their first element is distinctly near-high in contrast to the fully high second element: key [kɪj], if you replace the [j] with a [d], would sound the same as kid. Elsewhere, however, I would describe them as phonetically monophthongs, [i ʉ] when short, [iː ʉː] when long. In Lindsey's blog post, the recording of big leaves from Watch With Mother doesn't sound that different from how I'd say it (though the word usually does sound different, because its [uː] isn't centralized, and so do the late Queen's least [liːst] and few [fjuː], since for me these would be [list] and [fjʉ̞w]). The transcriptions /ij/ and /ʉw/ are meant as a compromise between the diphthongal and monophthongal pronunciations. Note that regardless of the phonetics, I think it's clear that the phonemes /ij/ and /ʉw/ are phonologically diphthongs, because they behave like all other diphthongs in exhibiting allophonic length variation and being able to occur immediately before another vowel without /r/ being inserted in between.

The centralized diphthongs are centralized in most environments except before dark L, where they preserve their older fully back pronunciations: pool [puːɫ], pole [pɔwɫ]. The starting point of backed /əw/ is audibly somewhere between /o/ and /ɑ/. When the diphthongs are centralized, I think /ʉw/ centralizes a bit more than /əw/; the starting point of /ʉw/ is a bit more front than back, so precisely something like [ÿ] or [ʏ̈], while the starting point of /əw/ is a bit more back than front, so precisely something like [ʌ̈]. Also despite the transcriptions I just gave, I'm not sure there is any difference in rounding between the two diphthongs; I think for both the rounding is weak to nonexistent at the start, and only gets stronger towards the end.

The low diphthongs are distinguished by two features: first, they are long word-finally, whereas all other diphthongs are short word-finally; secondly, when low diphthongs are long, it's clearly their first element that's lengthened, while for other diphthongs, the lengthening is more spread out across both elements, and if anything is more on the second element. So pie is [paːj], pine is [paːjn], how is [haːw], loud is [laːwd]; but fade is [fɛjːd], noise is [nojːz], etc.

There is one odd exception: before dark L, /aw/ behaves like a non-low diphthong, having the second element lengthened instead of the first: howl is [hawːɫ]. As for /aj/, it doesn't occur before dark L at all: /ə/ has been inserted wherever a diphthong ending in /j/ historically preceded dark L, as in reel [rɪjəɫ].

So what determines the length of a diphthong? As noted already, in word-final position, low diphthong are long and the rest are short. Before a vowel, all diphthongs are short. Before a single voiceless consonant all diphthongs are short, and before a single voiced consonant all diphthongs are long. Before consonant clusters, it seems the voicedness of the last consonant is what determines the length. So bind is [baːjnd] but pint is [pajnt], despite both words having a voiced consonant immediately after the /aj/.

For the record, here is a recording of me saying words with /r/ in a variety of positions (rat, arrow, law and order, prank, trick, thrush, creak, string, dictionary):

https://vocaroo.com/12xuKNF5E0lE

I used to think my /r/ was a labialized postalveolar approximant [ɹ̠ʷ], which is a pretty normal pronunciation of /r/ for English speakers. However, I've recently come to the conclusion that it's actually something much closer to a labiodental approximant [ʋ], a less normal but well-known pronunciation that is quite common among younger people in England.

One reason I haven't noticed this before is that I thought [ʋ] sounded like something in between /w/ and /v/, and at any rate quite dissimilar to /r/. Indeed, the recording on the Wikipedia page (https://en.wikipedia.org/wiki/Voiced_labiodental_approximant) sounds like that to me, and I get a similar non-rhotic sound when I try to produce a [ʋ] by taking the description "labiodental approximant" literally by vocalizing while raising my lower lip so that it approaches but doesn't touch my teeth.

Moreover, I had read that the [ʋ] realization of /r/ is commonly heard as /w/, hence why people call Jonathan Ross (a British TV presenter who has this realization) "Wossy", based on the way he pronounces his surname. So I thought, since my /r/ doesn't sound at all like /w/ to me, it surely can't be [ʋ]. I've also never had anyone comment that my /r/s sound like /w/s to them, although perhaps I don't talk to people often enough for this to be evidence of much weight.

Curiously, however, despite having watched many TV programmes involving Jonathan Ross, I had never actually noticed this feature of his speech myself---only read about it. In fact, if I listen to clips on YouTube of Ross speaking I don't hear anything odd about his pronunciation of /r/. Even in this clip where he's actually being mocked for his pronunciation, I don't hear any difference between his pronunciation and that of the guy mocking him!

I suspect what's going on here is that my categorical perception of English phonemes places the [ʋ] realization of /r/ in the same category as the more normal realizations of /r/, so that the difference, though it exists, is difficult for me to discern, whereas other English speakers either put it in its own category or in the same category as /w/, and thus do notice something different about Ross's pronunciation of /r/ in the above clip. I would predict that those other English speakers would be either older British English speakers or Americans, who would have had less exposure to the [ʋ] realization as they were growing up.

But if the [ʋ] realization of /r/ sounds like /r/ to me, why does the [ʋ] in the recording on the Wikipedia page for "labiodental approximant" sound so unlike /r/? I think it must be because the [ʋ] realization of /r/ has more going on with it---while it probably is both labiodental and an approximant, there are probably further adjectives that need to be added to precisely pin down its phonetics.

As for my own /r/, if I try to pay attention to what I'm doing when I pronounce it, it's evident that the "labialization" is very strong, and it certainly should be considered a primary articulation rather than a mere secondary articulation. But it's quite clearly bilabial, not labiodental. Is there any coronal co-articulation? I think probably not, although it is a bit hard to tell since I think my resting tongue position is pretty much the same as the position I'd put it in for a non-lateral alveolar or postalveolar laminal approximant. In consonant clusters such as /tr/ or /kr/, I think my tongue just goes straight from the position of the stop to the position of the following vowel, so that would suggest the /r/ is a solely labial gesture.

But "bilabial approximant" isn't quite a full description. Similarly to the labiodental approximant, if I take "bilabial approximant" literally, I just move my lips a bit closer in a vertical direction and produce something kind of like a /w/, although it doesn't really closely resemble any English sound. When I pronounce /w/, in addition to this vertical movement, there's also lip rounding---the corners of my lips move closer to each other, and the lips bend in the middle and protrude outwards a bit to produce an "o" shape. On the other hand, when I pronounce /r/, the corners of my lips get further apart, and while they still protrude outwards, they don't bend in the middle, so the resulting shape is more like a horizontal slit. It could perhaps be described as protrusion with rounding, for /w/, vs. protrusion without rounding, for /r/. I think it's not quite the protruded vs. compressed rounding described on Wikipedia---the picture of protruded rounding on Wikipedia matches what I do for /w/, but the picture of compressed rounding there certainly doesn't match what I do for /r/. Actually, the picture on the Wikipedia page for duck face is pretty similar to what I do for /r/, except my lips would be slightly open rather than fully closed.

This "protrusion without rounding" lip gesture---let's call it spread protrusion---seems to be the main thing responsible for my perception of a labial approximant as /r/ vs. /w/. If I take the basic "bilabial approximant" and add spread protrusion I think that fully reproduces my /r/. If I try to produce a labiodental approximant with spread protrusion of the upper lip, the result also sounds like /r/ to me---so I suspect this is what differentiates the [ʋ] sound used by Jonathan Ross from the [ʋ] in the Wikipedia recording.

With spread protrusion, the bilabial and labiodental sounds are very similar so it wouldn't surprise me if there's variation among speakers with a labial /r/ as to which place of articulation they use. There is an unsourced comment on Wikipedia that "bilabial realizations have also been reported" for English /r/ but that is the only thing I could find on the Internet that suggests a bilabial realization is possible.

Conclusions:

my personal realization of /r/ seems to be a bilabial approximant with spread protrusion

this is probably an instance of the so-called "labiodental r" which is becoming common among younger English speakers in England

Recently I’ve been learning about torsors. A torsor is, in informal terms, an algebraic structure that’s like a group with the identity element “forgotten”. The precise definition (whose connection to the informal notion is not meant to be obvious!) is: a torsor is a set equipped with a group acting on such that for any two elements and of , there is a unique with . My interest in torsors comes…

You can also view this article at my Github Pages website (the formatting might be better there). Consider an idealized model of a ball bouncing on the ground, where the ball is falling along a line normal to the ground, so that it remains on this line as it bounces, allowing us to treat its position and velocity as one-dimensional quantities. The only force acting on the ball while it is in the…

Sorting functions in programming languages often take a parameter which allows the user to control how comparison is done when sorting. The most direct way to do this is to have the parameter be a comparison function, which takes two arguments and returns a value indicating how the first argument compares to the second argument. For example, in Python, an implementation of insertion sort that…

Given two integers and , where is non-negative, an ordered -partition of is a -tuple of positive integers such that For example, the ordered 3-partitions of 5 are , , , , , and . The problem I'm considering in this post is: how do you enumerate all the ordered -partititions of ? In the case where , the problem is simple. An ordered 0-partition is a 0-tuple. But there is only one 0-tuple: the…