Extending Wells' Lexical Set to Prelateral Vowels

At some point in the past few years, I’ve analyzed pretty much every English vowel before laterals. They’re pretty cool because they’re understudied, they’re somewhat infrequent, and they’re involved in a lot of different mergers in different parts of the country. When referring to these prelateral vowels, several labels have been used in the past, but none do the job quite right. So, I think prelaterals should get a standardized set of Wells-style labels. The problem is figuring out what they should be. In this post, I explain why existing labels aren’t great and then propose a complete set of new labels for prelateral vowels.

To cut to the chase, this table—particularly the column labeled “Prelateral Keyword”—shows the labels I’ve come up with:

A standard set of prelateral keywords.

Yes, this is probably another case of competing standards, but it illustrates what I think lexical sets for allophones should look like.

For additional background how and why I came up with these labels, read on.

Wells Lexical Sets

As a small bit of background, John C. Wells. 1982. Accents of English.John C. Wells came up with standardized labels for sets of words that historically share the same vowel. The labels are also used to refer to the vowel itself. So something like fleece may refer to words like creep, speak, leave, feel, key, and people but also to the phoneme /i/. This is useful in cross-dialect comparisons when the actual phonetic quality of that vowel may systematically differ in that set of words.

The keywords themselves are not just arbitrary words within that lexical set, but rather they were carefully selected:

The keywords have been chosen in such a way that clarity is maximized: whatever accent of English they are spoken in, they can hardly be mistaken for other words. Although fleece is not the commonest of words, it cannot be mistaken for a word with some other vowel; whereas beat, say, if we had chosen it instead, would have been subject to the drawback that one man’s pronunciation of beat may sound like another’s pronunciation of bait or bit. (Wells 1982:123)

Presumably then, a good set of keywords for prelateral vowels should be deliberately chosen to follow this same principle of maximal clarity.

To be clear, what I am proposing is a set of keywords that refer to the prelateral allophones of each English vowel. That is, a label like zeal would refer to /i/ when followed by an /l/, as in the words feel, peel, and deal. As explained in my previous post on non-canonical allophonic extensions to Wells’ lexical sets, zeal would be a subset of the umbrella label fleece, which refers to all tokens of /i/ regardless of phonetic environment.

Existing Prelateral Labels

Of the research I’ve seen on prelateral vowels, if labels are given to them, they usually take the form of a minimal pair or set (or as close to one as possible). See, for example, Bowie (2000), Thomas (2004), and Tillery & Bailey (2004).Some of the more common labels include pool, pull, and pole for back vowels, or feel, fill, fail, and fell for front vowels. I’ve used them myself, actually.

I presented an abbreviated version of this table already in a previous post since it’s what popularized the b_t frame.To my knowledge, the only attempt at standardizing prelateral labels was by Thomas & Yeager-Dror (2009). In their table, reproduced below, they list the labels they use in that volume. In addition to standardizing other allophones’ labels, prelateral vowels get an entire column, and a nearly complete set of labels.

The lexical set, including prelateral and other allophonic expansions, from Thomas & Yeager-Dror (2009:6).

It appears the authors wanted to select as many keywords as possible that started with /b/. The now somewhat widespread b_t frame is established, and most prerhotic labels start with /b/ as well. The prelateral vowels had the most exceptions: there are seven keywords with b_l, but there are several consonant onsets to fill in the gaps (three with /p/ and one each with /k/ and /h/). This inconsistency in the prelaterals creates a somewhat haphazard set of labels.

The draw of using minimal sets for prelaterals is tempting, but I believe they have the same problems that the b_t frame has. In a previous post, I’ve already shared my thoughts on why I think the b_t frame isn’t very helpful, so I won’t rehash it further. The gist is this: if you hear someone say [poɫ] in isolation, are they saying pole or are do they merge /ʊl/ and /ol/ and are actually saying pull? This ambiguity goes against Wells’ original intentions of the words being “hardly be mistaken for other words.”

To put it another way, just because we call something the pull-pole merger, that doesn’t mean we should refer to the entire lexical sets as pull and pole. In fact, I think we should actively avoid referring to them as pull and pole for the very reason that they do form a minimal pair.

Why a minimal set won’t work for prelaterals

Let’s entertain the idea of using a minimal set to unite all prelateral vowels. We’ve got f_l for front vowels and p_l for some of the back vowels. Can we find one frame that works for all vowels?

No. The following table, which is grouped approximately by natural class, lists the frames that have the most words and would therefore be the best candidates for new labels:

Minimal sets of and potential keywords for prelateral vowels. Words in gray are not ideal for a variety of reasons (infrequent words, proper nouns, or inconsistent vowel assignments). Blank cells are, as far as I can tell, accidental gaps in the English lexicon.

As you can see, no one frame works for all vowels. The main culprits are /ʌl/ and /ɑl/ and maybe also /ʊl/ and /æl/. Let’s look at each frame individually:

• b_l—It’s closest to the established b_t frame, but other than the diphthongs (which are probably least studied), none of the By natural classes, I’m referring to front vowels, back vowels, low vowels, and diphthongs—environments within which mergers are most commonly reported.natural classes gets a full minimal set. In fact, from what I’ve been able to find, there are no monosyllabic English words with /#bil/, so you can’t even add extra coda material to make it work.

• p_l—If the b_l frame is the goal, p_l makes for a good backup since it’s articulatorily closest. For the back vowels, it’s probably the least bad option. (If so, I prefer pole over poll.) For the front vowels, it works fine except the word pell isn’t exactly that common.

• f_l—This one is best choice for front vowels with its four common words. Conveniently, it works for the diphthongs too. I don’t speak hypothetically here: I’ve included foal in wordlists and people sometimes do mistake it for foul. See also Wade (2017:7) and Arnold (2015:3) who explicitly mentioned this problem too.It’s a little less ideal for the back vowels because foal isn’t a particularly common word, and some people confuse it with foul.

• d_l—This one works well for a lot of vowels, but only if you use a couple proper names like Dale and Doyle. In a lot varieties of American English dual works fine here, but not for non-American varieties. Furthermore, dole isn’t a very well-known word I don’t think. The benefit of looking at d_l though is that it is the only frame that contrasts /ɑl/ and /ʌl/.

• h_l—It works for the front vowels, but I don’t really want to plaster hell all across my papers. Besides, who’ll isn’t ideal either because if we’re okay with contractions, we might as well include bee’ll, boo’ll, or any number of words to fit the frame. I will say though that the benefit to using h_l is that is has a nice minimal pair contrasting /ɔl/ and /ʌl/, which some varieties merge.

• k_l—Most of the benefits of h_l apply to k_l but none of the broad natural classes are complete.

• t_l—It’s comparable to f_l, except there’s no /ʊl/.

• g_l—The benefits are similar to k_l, except there’s no /ɡil/, and I had to fudge /ɔɪ/ with Goyle (though Harry Potter fans may enjoy it).

Several frames work pretty well, but they all have their problems. More importantly, no one frame works well for all front vowels (if you include /æl/), no one frame works for all back vowels (if you include /ʌl/), and no one frame works for the low vowels (thanks, /ɑl/). To get a complete set of labels, you’re going to have to pick three different frames—at a minimum. Such a patchwork of frames is inconsistent and confusing.

So, should we ditch the minimal sets and go for unambiguous, truly Wells-style lexical sets? Yes.

Unambiguous labels for prelateral vowels

My goal in all this was to select prelateral vowel keywords that are, in Wells’ words, “unmistakable no matter what accent one says them in.” So, rather than finding the optimal minimal set, I decided to choose words that were were very different from one another.

To identify potential words, I would need a list of all monosyllabic English words with a prelateral vowel. Fortunately, I’ve got that (and I’ve had it for some time)!

List of all ~401 monosyllabic English words with a coda lateral, organized by vowel and syllable structure? Check. My favorite? Squelched.

— Joey Stanley (@joey_stan) May 7, 2017

This spreadsheet is based on the CMU dictionary. I searched for all words that contained just one vowel and where that vowel was followed by an L. I then just organized them by their consonants in a spreadsheet. It’s pretty handy.

With this spreadsheet, I can easily identify candidate representative words. So, loosely based on Wells’ original keywords, here were the criteria that I established for selecting a word, approximately in order of importance:I’ve outlined more general criteria for selecting a new Wells-inspired label for an allophone in a previous post.

1. They must be monosyllabic.

2. They must be monomorphemic.

3. Words should not form another English word when the vowel is replaced with another (i.e. it should have no minimal pairs).

4. Onsets should ideally have just one consonant. Glides are avoided and liquids are acceptable as a last resort.

5. Codas consist of a single voiceless obstruent following the /l/. I have relaxed the coda requirement a little bit from what Wells says (1982:123):
   
   “As far as possible the keywords have been chosen so as to end in a voiceless alveolar or dental consonant: a voiceless consonant minimizes the likelihood of diphthongal glides obscuring a basic vowel quality, while coronality (alveolar or dental place) minimizes the possible allophonic effect of the place of a following consonant. An exception here is trap for the /æ/ correspondence, where no items in /-t, -s, -θ/ are altogether suitable; another one is palm.”
   
   I figure the presence of the lateral is enough of a buffer between the vowel and any following consonant (though phoneticians may prove me wrong).

6. Nouns are preferred.

With this organized spreadsheet, applying the six criteria for each vowel class was relatively straightforward. The below table shows the list I came up with:

The prelateral keywords I propose in this post.

Allow me to share a few comments about each one, starting with the front vowels.

• zeal—The only other word with /il/ and no minimal pairs was squeal, and I opted for the one with a less complex onset.

• guilt—If I wanted to relax the coda requirement, I really like chill as an alternative option.Lots of other words fit the critera, like filth, lilt, milk, and tilt, so I arbitrarily chose guilt as the representative word. Also, milk wouldn’t work because a lot of people have /ɛ/ in that word.

• flail—This was an easy choice because it’s the only one with no minimal pair. I’m not a fan of the complex onset, especially with the liquid, but it’s the only option. That’s okay: since Wells used fleece, I think flail is fine.

• shelf—Again, there were lots of other options: belch, help, health, pelt, self, wealth, and yelp. Of these, shelf just felt like the best choice—though belch would have been a great alternative!

For the back vowels, it’s especially important that minimal pairs do not exist since there are a variety of mergers that exist among prelateral back vowels, including four-way mergers.

• spool—There were no words with /ul/ that didn’t form a minimal pair with other prelateral vowels. And in fact, none of the monomorphemic words had any other coda material besides /l/. Given these limitations, I chose spool as the representative word. It forms minimal pairs with spiel, spill, spell, and spoil, but I’m not aware of any variety of English that would merge spool with any of those words. Other candidates were I’m pretty sure I’ve seen school before but I can’t find the source.school and cruel, but they form minimal pairs with skull and crawl and I wanted to avoid any words with other back vowels in the same frame.

• wolf—There really are very few words with /ʊl/, especially monosyllabic ones. As far as I know, there are five: full, pull, bull, wool, and wolf. Is it coincidental that they all start with labial sounds? Not sure. Probably. Of these, full, pull, and bull form lots of minimal pairs with other back vowels. Wool has no obstruent in the coda. That leaves wolf. I’m pretty sure my idiolect assigns /o/ to this word, so it goes against my intuition, but Wells himself assigns /ʊ/ to wolf and it’s the only real option we’ve got.

• jolt—There are lots of words with /ol/, but only two lack minimal pairs: stroll and jolt. Only jolt fits the other criteria.

• mulch—A handful of words fit the criteria: gulp, hulk, mulch, I’ve personally used pulp in the past, but I wanted to avoid anything like the established p_l frame in this new set.pulp, pulse. And unless you count the name of the department store Belk as a word, then bulk also fits. Any of these work well, but I chose mulch to add a bit of variety in coda consonants to the lexical set.

With the low vowels, the number of options was limited, and I had to relax some of the rules.

• talc—Another option is valve, and if I wanted to relax the voiceless obstruent requirement, I’d opt for that.The only other word that fit all the criteria was scalp, so I chose the one with the simpler onest.

• golf—Okay, this one was tricky. The CMU and searchable phonetic dictionaries I have access to didn’t have too many words with /ɑl/. Plus I merge /ɑ/ with /ɔ/ before coda laterals, so I don’t have intuition to back me up. In fact, the best list I know of is Aaron Dinkin’s 2016 paper, “Phonological Transfer as a Forerunner of Merger in Upstate New York.” The word golf was the most common word, and though it forms a minimal pair with gulf, it’s probably the best option. The next best might be psalm, but that only works for speakers that pronounce the [l] and who don’t have palm there. Another option was doll, but it’s part of a large minimal set.

• fault—The only two words that formed no minimal pairs, sprawl and false, aren’t ideal because they didn’t seem nouny enough. Plus, sprawl has has a complex onset, and seeing false in capital letters just seems like an error in computer code. The word fault forms a minimal pair with felt but fits all the other criteria, so I think it’ll do.

For diphthongs and pre-/ɹ/, the choices were also limited. You’ll notice that only one has a consonant in the coda, and its voiced, violating my rule 5. The reason is because none of the monomorphemic words with these vowels had anything else in the codas, other than child, wild, and mild.

• child—There was only one word with no minimal pairs, whilst, which didn’t seem like a good fit. Though child has a voiced coda, it only forms a minimal pair with chilled so it seems pretty good otherwise.

• prowl—It’s not the most common word, but it doesn’t have any minimal pairs.

• broil—There were really no good options for /ɔɪl/. They all had minimal pairs with other words, so I chose a word based on the minimal pairs it did have. Spoil forms a set with spool, spell, spill, and spiel; soil forms a set with sole/soul/Seoul, Saul, Sal, sell/cell, sale/sail, sill, and boil, toil, coil, Doyle, foil, and oil have even larger minimal sets (see the table above). In comparison, broil doesn’t seem to bad since it forms a set with just brawl and Braille.

• twirl—I’m not a big fan of the /w/, but it’s the only one that doesn’t form a minimal pair.

I believe the resulting list forms a decent set of labels to refer to prelateral vowels.

Using this new lexical set, we can now refer to things like the zeal-guilt merger, the flail-shelf merger, or the dress-talc merger among the front vowels. Among the back vowels, we might find the spool-wolf merger, the wolf-jolt merger, the jolt-mulch merger, or possibly, the wolf-mulch or the mulch-fault mergers. I personally have the golf-fault merger, though only if the /l/ is in the coda (collar and caller are still distinct).

Conclusion

I have proposed a new set of labels akin to the Wells Lexical set for referring to prelateral vowels. The ad hoc labels in existing literature are inconsistent, and the attempts at using lexical sets are unsuccessful. This new set should make labels more consistent in sociolinguistic research going forward.