24.914 | Spring 2019 | Undergraduate

Language Variation and Change

Assignments

Due session 5

Using the handout on Wells’ standard lexical sets (PDF) as a guide, identify the vowel phonemes in your dialect of English and transcribe their basic realizations. The second examples on the handout are often a better guide to basic realizations than the keywords.

Due session 26

Possible Topics

These are just examples—anything on language variation or language change is potentially fair game, in any language. Projects should generally include collection of some primary data (e.g. usage data, pronunciation data from the historical record or from recordings), and formal analysis of those data (e.g. analyzing the grammatical difference between two stages or variants of a language, computational modeling).

Regularization of past tense forms in English

Regular past tense formation in English involves suffixing /d/ (grab, grabbed), but a number of verbs have irregular past tense forms, e.g. run-_ran_, sleep-_slept_ etc. Over time, verbs with irregular past tenses have regularized, e.g. climb-_clomb_ > climb-_climbed_. An obvious theory of regularization is that it is the result of lack of data: language learners default to the productive  regular pattern in the absence of sufficient evidence for an irregular verb. This leads to the prediction that regularization should apply first to verbs with lower token frequencies. Lieberman et al (2007) found support for this hypothesis, but there is also evidence that  frequency alone cannot explain the patterns of regularization. For example, some irregular patterns seem to be more resistant to regularization than others, independent of frequency (e.g. ɪ-ʌ pasts like sling-_slung_ vs. aɪ-oʊ past tenses like climb-_clomb_). Investigate some past tense patterns to test for effects of factors other than frequency.

Does lenition target more frequent words?

Bybee claims that sound changes/optional processes involving lenition are more likely to apply to higher frequency word. Is that true? Does optional lenition target predictable words rather than frequent words? Data on t-d deletion in the Buckeye corpus are available and could be used to test this claim. Other data sets might be available.

The grammar of style-based variation

The dominant grammar models for analyzing variation (e.g. Stochastic OT) treat all variation as completely random, but we know that much of it is related to factors such as speech style. Explore proposals for analyzing stylistic variation in phonology/phonetics. A possible source of data is Labov (1972) ‘The isolation of contextual styles’. Another source of data would be to collect recordings of someone (a public figure?) speaking in two contexts (e.g. formal speech, informal conversations) and compare pronunciations in those two contexts.

(Variation in) the environments of TRAP-tensing

A lot of open questions remain from our investigation of TRAP-tensing in Long Island: What are the precise environments in which this change has applied? The environments in which it applies variably? The reasons why those environments favor tensing? How does the phonetic realization of the tensed TRAP vowel vary in fluent speech vs. isolated words? When does tensing apply in derived words (e.g. tensing in manning (man+ing), but not in Manning (the name))? These questions could be addressed through further investigation of Long Island English, or some other mid-Atlantic variety (we have recordings of Philadelphia English), or a comparison between dialects. It might also be interesting to look for antecedents of the change in other dialects.

Lexical diffusion

Lexical diffusion is still a big mystery. Why does it often seem to apply as a generalization of a regular sound change? Are the mechanisms of phonetic conditioning in lexical diffusion the same as those that operate in regular sound change? Does word frequency have anything to do with it? My best candidate for data at the moment is TRAP-tensing, but there should be other possibilities.

Modeling sound change

Implement and test a model of some aspect of sound change.

Syntactic change

Identify a recent syntactic change. Come up with analyses of the ‘before’ and ‘after’ grammars, and propose/test hypotheses concerning the mechanism for getting from one to the other (e.g. reinterpretation of ambiguous input). A candidate change: The recent rise in double ‘is’ constructions, as in ‘The funny thing is is that Lisa was there too.’ See the Yale Grammatical Diversity Project’s “Double is” for more information.

Due session 4

1. Transcribe the sentences in the following sound files:

2. The Texan speaker in the recording at www.dialectsarchive.com/texas-1 0:06 to 1:05 uses two allophones of the /ɪ/ phoneme (the vowel in words like ‘this’, ‘it’, etc). Transcribe these two allophones. Do you think this is free variation or could it be conditioned variation? Explain why. If you think it might be conditioned variation, suggest a hypothesis about the conditioning factors.

Here is a transcript to help you locate relevant words:

First part of recording:

When the sunlight strikes raindrops in the air, they act as a prism and form a rainbow. The rainbow is a division of white light into many beautiful colors. These take the shape of a long round arch, with its path high above, and its two ends apparently beyond the horizon. There is, according to legend, a boiling pot of gold at one end. People look, but no one ever finds it. When a man looks for something beyond his reach, his friends say he is looking for the pot of gold at the end of the rainbow.

(public-domain text from International Dialects of English Archive)

Second part of recording:

(Uh) This is (oh)—it’s just a jail delivery that we (uh)—when we went through the (uh), the Heritage Association’s tour, this last year, (uh) we obtained copy of this. And (uh) this—I remember when my grandparents (uh), who had a farm…

(transcription © International Dialects of English Archive. This content is excluded from our Creative Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.)

Due session 8

The core of the assignment is to describe some of the phonetics and phonology of the TRAP vowel (/æ/) in the accent of a speaker from Long Island. The context is the lack of clarity in many descriptions of the phonological process often referred to as /æ/-tensing —it is unclear exactly what the difference between so-called tense and lax variants of /æ/ is, and it is often unclear exactly where these allophones occur. This is unfortunate because /æ/-tensing plays an important role in the literature on the mechanisms of sound change.

We have an opportunity to try to get a better understanding of this process by analyzing data from a speaker from Long Island who has tense /æ/. The data consists of recordings of relevant words, in a sound file on the course website [not accessible to OCW users]. Download the file and use Praat to analyze the patterns and write up your results in the form of a short paper. Illustrate it with relevant spectrograms. Details of the contents of the recording are below.

Points to address:

  • Introduction: The questions you are addressing and their context. (The discussion in Labov (1981) may be useful here).
  • The transcription of the lax and tense versions of the TRAP vowel (and any other allophones that you identify)
    • Use your ears and spectrograms.
    • Use comparisons of the formant frequencies in tense /æ/ to formant frequencies in other vowels (e.g. lax [æ], KIT, DRESS, etc) to justify your transcriptions.
    • Duration measurements may be useful also.
  • Are there only two main allophones of the TRAP vowel, or are there more?
  • The phonological contexts in which the allophones of the TRAP vowel appear.

Suggestions

Identifying the contexts for allophones of TRAP:

According to Labov (1981), tense /æ/ appears in monosyllabic words when followed by the boxed consonants in Philadelphia English (dashed box) and New York City (solid box):

A grid of consonants, with a solid line around p, t, j caron, g, m, n, theta, s, and esh, representing the consonants causing tensing of previous /ae/ in New York City, and a dotted line surrounding m, n, theta, s, and esh, representing the consonants causing tensing in Philadelphia dialect.

© Linguistic Society of America. All rights reserved. This content is excluded from our Creative Commons license. For more information, see https://ocw.mit.edu/help/faq-fair-use/.

  • Start with the monosyllables—the main conditioning factor is expected to be the identity of the following consonant.
  • Once you’ve worked out which consonants condition tensing (or other allophones of TRAP), see if the same consonants condition tensing (or related allophones) when /æ/ is in an open syllable (i.e. followed by a single consonant, which is in turn followed by a vowel, e.g. ‘cabin’).
  • How about when non-final /æ/ is followed by a consonant cluster (e.g. ‘amber’)? 

Are there just two versions of the TRAP vowel?

Discussion of the TRAP vowel in these dialects focuses on a distinction between tense and lax variants of this vowel, but it seems likely that the situation may be more complicated in many accents. Some places to look for more variants:

  • There is supposed to be a tense-lax distinction in non-final syllables. If there is, it seems to be somewhat different from the distinction found in monosyllabic words.
  • There are mixed reports on whether velar nasal [ŋ] conditions tensing. This may be because this nasal can condition a third allophone of the TRAP vowel. 

Again, use spectrograms and acoustic measurements to characterize the differences between any allophones that you identify. Try to transcribe them, and specify the contexts in which they appear.

The recording

The recording [not accessible to OCW users] includes the following words illustrating TRAP and DRESS vowels. (Most of the words with DRESS vowels are probably not relevant at the moment.)

Monosyllables Polysyllables - Open syllable Polysyllables - Closed syllable
Voiced Stops Voiced Stops Voiced Stops
sad said radish meadow admiral Edna
pad Ted ladder pedal abdomen hebdomadal
bad bed cabin rebel Agnes segment
mad   cabbage debit magnet  
glad   maggot beggar    
slab   baggage   Voiceless Fricatives
bag beg     after  
stag   Voiceless Fricatives Alaska  
    affable effort asterisk ester
Voiceless Stops baffle      
mat met catheter method Nasals
bat   Catherine   candidate Kendall
slap slept castle lesson anchovy  
tap kept passive   canyon Kenyon
back beck fashion   amber ember
stack   passion special lambda  
        amnesty  
Voiceless Fricatives Voiced Fricatives anger  
laugh   cavern Kevin    
half deaf avenue      
bath Beth hazard   Reference Vowels
pass   dazzle desert heed  
mass mess     hid  
mash mesh Nasals big  
    banish   head  
Voiced fricatives planet   hog  
halve   manner menace    
jazz   famish feminine    
    camel chemical    
Nasals hangar      
man men        
ban Ben        
slam hem        
hand          
dance dense        
bang          
hang          
           
Liquids        
pal pell        

The words were recorded in the following random order:

hangar, bath, meadow, laugh, magnet, sad, ladder, beck, mash, abdomen, Beth, after, bang, heed, canyon, affable, ember, met, dense, pal, cabin, pad, bad, bag, said, Catherine, big, tap, amnesty, passion, Ted, hid, castle, stag, camel, glad, jazz, Kevin, admiral, half, mad, amber, man, Ben, ester, mesh, slap, hebdomadal, mess, pedal, slam, men, pass, Kenyon, hog, famish, chemical, dance, anchovy, bed, hazard, asterisk, effort, head, banish, deaf, manner, rebel, planet, stack, passive, avenue, catheter, desert, anger, hang, cavern, baggage, candidate, ban, beg, hand, fashion, maggot, Kendall, Alaska, slab, mat, slept, lesson, mass, bat, beggar, Agnes, segment, dazzle, hem, feminine, baffle, bad, back, menace, halve, pell, cabbage, radish, kept, lambda, Edna

Due session 12

Investigate a possible case of lexical diffusion or similar irregular sound change. Determine whether the sound change is in fact regular. If so, see if it is consistent with selected hypotheses about the properties of lexical diffusion. Write up your results as a short paper (about 1000 words).

Possible Topics

1. TRAP-BATH split

A series of sound changes that have lead to some US English TRAP vowels corresponding to PALM vowels in Southern British accents like RP, e.g. dance US [dæns] / RP [dɑːns], while other words have TRAP vowels in both accents, e.g. mass US/RP [mæs]. This looks like an irregular correspondence because it doesn’t seem to be predictable from phonological context whether a US TRAP vowel will correspond to TRAP or PALM in RP English:

  US RP
mass mæs mæs
pass pʰæs pʰɑːs
ant ænt ænt
grant ɡɹænt ɡɹɑːnt
asp æsp æsp
clasp kʰlæsp kʰlɑsp

Wells (1982) argues that these irregular correspondences are the result of a vowel lengthening change that was diffusing through the lexicon, but stopped before it affected all the words containing vowels in the relevant contexts. There is more data on this Wikipedia page.

2. LOT-CLOTH split

In US English, some Middle English short ‘o’ vowels have developed into the LOT/PALM vowel (e.g. [ɑ]), as in lot, while others have merged with the THOUGHT vowel [ɔ] (in dialects with a LOT-THOUGHT contrast), as in cloth, but it does not seem to be entirely predictable which outcome you get in a given word. So the short ‘o’ phoneme has split, suggesting an irregular sound change. (Actually there is probably more than one change involved: short ‘o’ lengthened before voiceless fricatives in the 17th Century, and subsequently merged with long [ɔ] in US English). The unlengthened short ‘o’ vowels unrounded to [ɑ] in most of the US.

Earlier form  lɒt klɒθ
o lengthening - klɒːθ
THOUGHT merger - klɔθ
LOT unrounding lɑt -
Later form lɑt klɔθ

The ɒ > ɔ change seems to have applied more or less regularly before word-final fricatives, but seems to be irregular before medial fricatives (e.g. foster, but not roster), and has been extended in some dialects to apply before some voiced velar stops and nasals (dog, long), and a few alveolar nasals (usually just on, gone). So this change looks like a possible example of lexical diffusion.

There are more data on this [Wikipedia page](https://en.wikipedia.org/wiki/Phonological_history_of_English_low_back_vowels#LOT%E2%8 0%93CLOTH_split). (Don’t take it on trust—it’s Wikipedia—but it’s a good starting place).

It will be important to check etymologies of words—the sound change of interest involves Middle English short ‘o’ > [ɔ], but thought [ɔ] has many other sources. Spelling provides some indication of etymology, but it’s not a perfect guide.

3. FOOT-STRUT split

In the early 17th Century, Middle English short [ʊ] unrounded to [ʌ] (STRUT) in words like cut, but not in words like put (foot words). (The change presumably involved intermediate steps, e.g. ʊ > ɤ > ʌ). Preceding labials tended to block unrounding (as in put, full), but incidence of the change is apparently not predictable (e.g. putt, fun are STRUT words in spite of beginning with labials). So the ʊ > ʌ change is a candidate for a change that was diffusing through the lexicon, but was never completed.

The main focus should be on STRUT words, since they mostly come from short [ʊ] by unrounding, whereas FOOT words can also come from later shortening of long [u], as in good, making it hard to tell which FOOT vowels are the result of blocked unrounding. (This u-shortening is also an irregular change—see below). Look for exceptions to the generalization that the STRUT vowel should not follow labials, and see if they follow any more specific generalizations, or if the change was truly irregular.

4. Later /uː/ shortening

After the FOOT-STRUT split, long [uː] shortened to [ʊ] (foot) in certain contexts. This shortening was quite consistent in some contexts, but appears to be irregular in others, and is still variable now (e.g. room [ɹuːm/ɹʊm], roof [ɹuːf/ɹʊf]).

5. trap tensing before sT in Long Island

In Long Island English, TRAP [æ] tenses before word-medial [st, sk, sp] clusters in some words, but not all, e.g. master [mɛːstə˞], but asterisk [æstəɹɪsk]. What determines which words of this type undergo tensing? I have some data and can collect more from the speaker you studied in paper 1.

6. ‘Milk-to-melk’

…or other word-specific changes represented in your speech (or your friends’ speech). In class we have heard about some apparently irregular changes that show up in your speech, e.g. KIT [ɪ] > DRESS [ɛ] in words like milk, pillow, vanilla. Investigating a change like this is a bit different because the change may only affect a few words, so it may be harder to nail down what is going on, but it should still be possible to address interesting questions, and generate hypotheses for further investigation.

Things to address:

Check whether the sound change you are investigating is actually irregular.

  • Examine the data. Although they are all reported to be irregular, perhaps people have missed some phonological factors that differentiate the words that undergo the change from those that do not. Or there could be morphological factors that condition the change.
  • If there are still irregularities, try to check whether the exceptions could be later loanwords borrowed after a regular change had been completed and was no longer phonologically active. Dates of first known uses of words are available in the Oxford English Dictionary (see resources below).
  • What are the contexts in which the change tends to occur? Are there contexts where it applies regularly?
  • If your change is not irregular, you’re done! Write up the argument that the change is actually regular.

Assuming your sound change is irregular, try to test some hypotheses concerning the nature of lexical diffusion:

  • Lexical diffusion involves substitution of one pre-existing phoneme for another in the underlying representations of individual words (Labov 1981). Is there evidence that the sounds involved in the change were contrastive phonemes before the change began? If there isn’t good evidence that the sounds are distinct phonemes, are they at least involved in surface contrasts? E.g. like the [bænə˞] ‘flag’ vs. [beænə˞] ‘one who bans’ contrasts observed in trap-tensing dialects.
  • Lexically diffusing sound changes target higher frequency words first (Phillips 2006, Pierrehumbert 2001). (See ‘Resources’ for a source of word frequency data).
  • The examples of lexical diffusion that we have seen seem to involve extension of a regular sound change to new environments. Is that the case for your change? I.e. is there a core set of environments where the change has applied regularly, then a set of (related) environments in which the change is irregular?
  • Phoneme substitution changes tend to apply in contexts where the phonemes sound similar (e.g. due to assimilation to context, coarticulatory effects etc). Investigating this hypothesis is likely to involve making and analyzing recordings.
  • Phoneme substitution changes are more likely to apply to a word if the change does not make it identical to a pre-existing word (Bertics, p.c.). E.g. milk can change to [mɛlk] because that was not already a word. Ilk is less likely to change to [ɛlk] because that is already a word.

In general, see if you can work out what factors affect which words undergo the change. Not all of these hypotheses will be equally relevant to all sound changes—focus on those that your data bear on.

General resources

For each sound change, I have some short readings and lists of relevant words, mostly from Wells (1982). Contact me, and I’ll send you what I have.

Dictionaries

The OED records UK and US pronunciations. The UK pronunciations are mostly RP English (e.g. they show the trap-bath split). The US pronunciations include those exemplifying the LOT-CLOTH split. The American Heritage Dictionary reports US pronunciation.

Both also contain information about etymology in case you need to find out whether a word really descends from a Middle English word with short ‘o’ (for example). The OED etymological information is more detailed, but the detail can be hard to work through sometimes.

The OED also contains dates of first recorded uses of words.

Word frequency information

Use the spoken part of the American National Corpus.

Recordings and acoustic analysis—talk to me if you’d like help making a recording in phonetics lab.

Due session 17

Pierrehumbert’s (2001) paper sketching an exemplar-based model of sound change has been very influential, and many researchers have proposed extensions and variations on her general approach. Read one of the papers listed below and write a short paper (about 1000 words) evaluating the model proposed in the paper. Make sure to summarize the original goals of the model.

The evaluation need not address all aspects of the proposed model—you can focus on one topic. All models of complex phenomena like sound change necessarily simplify and abstract away from many aspects of the phenomenon in order to keep the model tractable and interpretable, so it is not interesting to critique a model simply for ignoring some aspect of sound change, but you can address whether the simplifications and abstractions are justified, whether and how the model night be extended to account for additional generalizations or additional types of sound change, predictions of the model etc.

Our list of properties of sound change may be useful in evaluating models. Also many models are presented in relation to a motivating example of sound change—it can be helpful to consider how the model would work with variants of that motivating example, e.g. similar sound changes involving different phonetic properties, or different conditioning factors. You are welcome to include your own simulations if you are so inclined—a couple of the papers include links to code for their models (Python, R).

Suggested papers

Ettlinger, Marc (2007) Shifting Categories: An exemplar-based computational model of chain shifts (PDF). UC Berkeley Phonology Lab Annual Report, pp. 177–182_._

Johnson, Keith, and Garrett, Andrew (2013) Phonetic bias in sound change (PDF - 2.9 MB). Alan Yu (ed.) Origins of Sound Change: Approaches to Phonologization. Oxford University Press, pp. 9–61.

  • Focus on section 3.6 ‘A model of actuation’. This section proposes models of the actuation of sound change, an issue that is set aside by Pierrehumbert and many other computational models of change.

Sóskuthy, Márton (2015) Understanding change through stability: A computational study of sound change actuation. Lingua 163, 40–60.

  • I plan to discuss this paper in class, so you’ll need to coordinate with me if you’d like to discuss it, but it includes R code for the model presented. Proposes an account of actuation.

Todd, Simon, Pierrehumbert, Janet B., & Hay, Jennifer (2019) Word frequency effects in sound change as a consequence of perceptual asymmetries: An exemplar-based model. Cognition 185, 1–20.

  • This paper may not be the best option because it presents a pretty complicated model designed to account for a relatively narrow set of phenomena (frequency effects on sound change), but it builds directly on Pierrehumbert (2001) and comes with Python code.

Wedel, Andrew (2006) Exemplar models, evolution and language change (PDF). The Linguistic Review 23, 247–274.

  • Exemplar model inspired by evolutionary theory.

You could also discuss Pierrehumbert’s paper itself, but if you do, you should go significantly beyond the discussion in class, e.g. developing your own simulations to probe the model further.

Course Info

Departments
As Taught In
Spring 2019
Level
Undergraduate
Topics
Learning Resource Types
Lecture Notes
Written Assignments
Download Course
MIT Open Learning
Over 2,500 courses & materials
Freely sharing knowledge with learners and educators around the world. Learn more
© 2001–2024 Massachusetts Institute of Technology
Accessibility
Creative Commons License
Terms and Conditions
Proud member of:
© 2001–2024 Massachusetts Institute of Technology