It was a balmy day in Taiwan in November 2019, and I used to be rummaging by means of the Family Mart adjoined to the Qishan Bus Station. It was my final probability for 9V batteries and spicy tuna rice balls earlier than taking a taxi into the mountains, the place lots of the remaining Indigenous languages of the island are spoken, the remaining having been changed by Chinese—the language of settlers from the Asian mainland who slowly took over the arable plainsland over the previous couple of hundred years, in addition to of the present ROC regime.

The 16 Indigenous languages nonetheless spoken in Taiwan in the present day—the Formosan group—are tragically endangered, with three Formosan languages right down to a single-digit variety of audio system and a fourth quickly encroaching. The languages are very nicely documented in some areas of their grammar and really poorly in others. The obtainable documentation is the results of efforts by neighborhood members who create assets for his or her language’s revitalization motion and from native and overseas students.

The aim of my PhD dissertation venture is to analyze probably the most poorly documented facets of language. And I’m going to make use of a secret weapon, which I purchased at B&H. To document, I take advantage of a Sony PCM-M10 recorder and a Røde Videomic, which I purchased in a $379 bundle marketed to aspiring YouTubers, which I’m not. Thankfully, it’s a directional (or ‘shotgun’) microphone, which data no matter you level it at louder than sound coming from different instructions. This has allowed me to document analyzable elusive information in a sawmill, throughout a navy drill, and whereas surrounded by canine. (Not on the identical time, fortunately!)

The gaping gap in documentary linguistics which requires such tools is one thing known as prosody, which is simple to really feel however arduous to listen to. To illustrate, I’m going to make use of a easy instance from English.

How many sounds does English have?

You is perhaps tempted to say English has 26 sounds, one for every letter of the alphabet. But that’s not fairly proper: Some letters like ‘c’ and ‘k’ could make the identical sound. Some sounds, like ‘sh’ and ‘ng,’ aren’t represented by single letters within the alphabet. And how might we neglect ‘ch’? Or, after all, ‘th’? How concerning the rising tone on the finish of a query?

In faculty, we typically study two kinds of speech sounds: consonants and vowels. But I promise, there’s extra! One layer of further construction in our speech is stress. As Mike Myers demonstrates in View From the Top (2003)—“You put the wrong em-pha-sis on the wrong sy-lla-ble!”—in English, one particular syllable in multisyllabic phrases is extra outstanding than the others. Stress is one a part of prosody, which is a big umbrella of speech phenomena that happen in bigger domains like syllables and phrases, as an alternative of smaller items like consonants and vowels.

But the actual enjoyable (should you’re me) begins whenever you ask how we all know a syllable is harassed within the first place. The greatest clue is how the phrase interacts with intonation, the a part of prosody that investigates how languages use tonal melodies.

For instance, let’s say you’re at work, and somebody walks into the break room and utters one of many following:

1. “There’s coffee.”

2. “There’s coffee?”

Same consonants and vowels. Same context. The first is a press release informing that there’s espresso. The second is a query, probably the place somebody is shocked to listen to that there’s espresso. Aside from durations and query marks, purely the area of writing, what precisely is the distinction between the 2?

The commonest method to modeling intonation is through the use of the constructing blocks H (excessive tone) and L (low tone). An increase may be described as LH, and a fall as HL. These and longer melodies are used for certainly one of two functions: 1) a ‘pitch accent’ which marks a harassed syllable; or 2) a ‘boundary tone’ which marks the sting of a phrase (like a comma may do in writing).

These notations can get very nuanced. The gold commonplace mannequin of English intonation, Janet Pierrehumbert’s dissertation, counts seven distinct pitch accent melodies: our good buddy L+H*, in addition to H*, L*, L*+H, H*+L, H+L*, and H+!H. The asterisk * right here notes which tone within the melody is aligned with the harassed syllable. Pierrehumbert additionally counts 4 boundary tones: H- and L-, which mark minor phrase boundaries (like a comma), and H% and L%, which mark main phrase boundaries (like a interval). While there have been efforts to tease apart how all of these are used, it’s not a simple activity. Was that L*+H presupposed to be sarcasm, or disbelief? Sass? Are they mad at me?!

Two of those components have seen a good quantity of consideration in pop science, particularly from non-expert authors who prefer to police millennial ladies’s speech. ‘Uptalk’ is simply the recurring use of H-, and ‘vocal fry’ is what occurs when one’s L% is low sufficient that the larynx produces creaky voice as an alternative of modal voice. These two intonational components have routinely been maligned as undesirable and even bodily dangerous: Naomi Wolf as soon as known as vocal fry a “destructive speech pattern.” In actuality, components like H- and L% are neither injurious nor unusual in intonational methods. If the usage of these components is as dangerous for the English language because it’s made out to be, then I’ve dangerous information about a number of thousand different languages.

How can we analyze intonation?

These days, analyzing a recording is simple sufficient. The commonest software program utilized in linguistics is named Praat, which is open supply, thank goodness. Plunk in a .wav and also you’ve acquired all of phonetics at your fingertips. If you may hearken to your individual voice on a recording with out freaking out (I can’t), you must strive it out your self. Here’s a screenshot of “There’s coffee.” open in Praat:

Here, the waveform is proven on prime, with the spectrogram in grayscale beneath. This exhibits all of the frequencies sounding concurrently at every time limit, with the totally different shades of grey exhibiting the depth of every frequency. Overlaid are the general depth (or ‘loudness’) proven with the yellow curve, the pitch in blue, and the formants (that are what make vowels sound totally different) in crimson.

On the underside are two tiers of transcription, one with the consonants and vowels in IPA, a language-neutral manner of transcribing speech sounds: [ðɛɹzkɔfi]. (I’ve written the ‘open o’ [ɔ] right here, however I really use [ɑ] as an alternative as a result of I’m not a real New Yorker. Shh!) The backside transcription exhibits a label for L+H*, the pitch accent aligned with the harassed first syllable of espresso. It feels like an increase in pitch, which reaches its crest in the direction of the top of the syllable.

There’s lots to take a look at right here, however what we’re curious about is the pitch monitor. Praat really has a extra complicated pitch-tracking system than what’s proven within the blue squiggles above, and you may manually filter out different detected frequencies. This is helpful whenever you’re an extremely awkward particular person like myself, who usually by accident talks over their interviewees. If the pitch tracker picks up your embarrassing interruptions, you may simply click on them away on a display just like the one proven beneath. Here, the pink dots are frequencies included within the ultimate pitch monitor, whereas the remaining have been filtered out.

With your ultimate, non-awkward pitch monitor, you should utilize Praat’s ‘smoothing’ software with the default 10Hz buffer to clean out the bumpiness. You don’t desire a bumpy pitch monitor, like how embarrassing would that be? Once the pitch monitor is publication-ready, you may generate an illustration within the Praat Picture window, as you may see beneath.

“There’s coffee.”

It’s smoothed. It’s annotated. Our pitch monitor is *chef’s kiss* and now we now have a significantly better view of what’s happening in our intonation. The rising tone of the L+H* pitch monitor is aligned with the harassed first syllable [kɔ] of espresso, and the utterance ends on a low tone proven by the boundary tones L-L% (as each main phrase boundary can be a minor phrase boundary).

Now examine this with the ‘question’ intonation.

“There’s coffee?”

Instead of an increase on the primary syllable of espresso, there’s a low tone, so the pitch accent is L* slightly than L+H*. And on the finish of the utterance there’s a pointy rise, so the boundary tones are H-H% slightly than L-L%.

Why don’t we see extra intonation in descriptive linguistics?

Many of the world’s 7,000-ish languages are each endangered and poorly documented by linguists. And of the languages that do see devoted examine, prosody and intonation are sometimes an afterthought. In ‘grammars,’ a sort of ebook that serves as an in-depth description of all facets of a language’s phonology and syntax, usually primarily based on years of subject examine, it’s not unusual for the one mentions of prosody to be 1) which syllable within the phrase is harassed, and a pair of) an impressionistic description of the intonation on questions. (Spoiler alert: there’s in all probability a ultimate rise.) That shouldn’t be sufficient.

In the previous, it made sense to omit prosody and intonation from subject research, because the recording and evaluation tools was cumbersome and costly. I do know I’m not lugging my phonograph and wax cylinders to the sector! Worse but, fieldwork usually occurs in noisy environments, and background noise can intervene with the evaluation.

Røde’s directional mic, coupled with the pitch monitoring in Praat, has allowed me to satisfy and work with audio system the place they actually converse, as an alternative of needing to carry them to a lab. While any language can be utilized to explain something, languages don’t exist in a vacuum, and the communities and cultures related to a language are vital context for linguistic examine. This is particularly so when eliciting intonation: Often, one of the best ways to get a recording of a particular intonational contour is to be in a scenario the place it will naturally be used. If you need to get an English speaker to say “no, there are two dogs,” it’s going to be tougher to conduct your interview in an empty recording sales space than out in a canine park, as an illustration.

Unfortunately, the exclusion of prosody and intonation from descriptive linguistics has persevered into the present period, regardless of the rising availability and utility of kit. While there’s rising curiosity in prosody/intonation, it’s usually within the type of standalone works. This has the disadvantage of being less-integrated with work on different facets of phonology and syntax, even after they naturally interface with many facets of prosody. We can solely hope to see extra H’s and L’s in grammars and different documentation work going ahead.

What is intonation like in Taiwan?

The journey to Family Mart was a part of my dissertation work, which sought to explain intonation in Formosan languages by way of pitch accents and boundary tones, like Pierrehumbert’s mannequin of English. I labored on as many languages as I used to be capable of finding audio system of, throughout 4 journeys to the sector in 2017-19, and wound up with unique information on 10 languages/dialects. I managed about 20% of what I wished to do, and wrote 800 pages about it.

Elicitation classes concerned all the things from asking a local speaker to translate a thesaurus to having them act out a dialogue or a real-world state of affairs which may evoke distinctive intonation. My favourite query to ask is “do you know any really long words?” which, as dumb because it sounds, will all the time both elicit a novel piece of information or on the very least break the ice. The longest phrases I discovered have been a tie between kinamakasusususuan, the phrase for “family” in Piuma Paiwan, and maisasavusavuanʉ, the Saaroa phrase for “doctor”; each 9 syllables.

The examine resulted in a wealth of descriptive details about intonation in these languages. Some Formosan languages like Seediq and Saaroa had a pitch accent L+H* identical to English, whereas others like Kanakanavu had a extra complicated pitch accent L+H*L, or simply H*L as in Mantauran Rukai. Two languages, Amis and Kavalan, had glottal stops (like when British folks say ‘butter’) that might present up on the finish of statements however not questions. Some languages had distinctive intonation to indicate sarcasm or incredulity or to mark objects in an inventory. And extra importantly, what I discovered was merely the tip of a large prosody iceberg, one which sadly is melting by the day.

How is covid-19 affecting language endangerment?

Endangered languages are such as a result of the language shouldn’t be being transmitted to youthful generations, in favor of a dominant language like English or Chinese. This implies that in lots of communities with an endangered language, it’s the elders who converse the language. Given that age is a predictor for the severity of covid-19 infections, these audio system are particularly in danger. Worse but, many communities with an endangered language have used in-person lessons as a significant element of their language revitalization motion. These are troublesome to conduct with out placing these elder audio system, who usually function the teacher, at elevated danger of an infection.

Taiwan’s prudent covid-19 response might have spared audio system of Formosan languages from a few of what different communities dealing with language endangerment are going by means of with regard to the pandemic, nonetheless, language endangerment has been a problem in Taiwan nicely earlier than covid-19. Of course, the issue and danger of worldwide journey brought on by the pandemic has additionally prevented linguists from engaged on languages exterior their very own nation. Remote fieldwork might be an possibility given the increase in recording quality seen in newer smartphones, however this received’t work with out pre-existing contacts, or if the expertise isn’t obtainable.

When languages lose their final native speaker, any details about the language that didn’t make it into the obtainable descriptions is misplaced to historical past. Of course, it’s not simply linguists all over the world who’re curious about language information: Many communities select to revive their ancestral language after dropping the final native audio system, primarily based on archival supplies. There has even been a shift, following certain Indigenous communities in North America, to pondering of languages as ‘dormant’ slightly than ‘dead’ after they lose their final speaker, each to focus on their persisting cultural significance and to depart open the chance that the language is reawakened by the neighborhood. When these communities do reawaken their language, many won’t understand how earlier native audio system would distinguish statements from questions, or earnestness from disbelief, given the dearth of intonation in descriptive works.

Can expertise assist?

While scripting this piece, I reached out to a colleague of mine, Joe Pentangelo, a fellow linguist and a postdoctoral fellow on the Macaulay Honors College, to ask how covid-19 has affected his fieldwork. Joe’s analysis issues each endangered language documentation and the usage of expertise within the subject. His PhD dissertation was the primary use of 360º video for documentary linguistics, wherein he used a Nikon Keymission 360 digital camera and Zoom H2N audio recorder to document interviews and natural conversations with audio system of Kanien’kéha (often known as Mohawk), as spoken in Akwesasne, a Kanien’kehá:ka neighborhood on the St. Lawrence river, which straddles the border between New York State, Ontario, and Quebec. The ensuing movies may be considered in any variety of VR headsets and present the interviews and conversations of their unique context, conserving intact all the details about how audio system are interacting with one another that could be misplaced in laboratory work or audio-only recordings.

“The last recording trip I made up there was in December 2019, right before Christmas,” Pentangelo informed me. “By the end of that trip, I had nearly 11 hours of immersive video, and the corpus was essentially complete. The plan was to return a few months later to screen all of the videos I’d recorded, to get final approval from all of the participants to release these videos publicly, and to work with local experts to transcribe and translate the content. Unfortunately, with the outbreak of covid, it hasn’t been safe to return, so the videos are not yet publicly released.”

One of the targets of Joe’s examine was to make his corpus obtainable publicly, permitting it to be a useful resource for the Kanin’kéhá:ka neighborhood slightly than one thing primarily of curiosity to teachers, a aim additionally mirrored in the usage of spontaneous conversations and recordings taken in situ. In Joe’s case, it’s not solely troublesome to proceed documenting the language, however even the forms concerned in releasing the info publicly is at a standstill.

“I have been able to work remotely with Dorothy Lazore and Carole Ross, two educators from Akwesasne, to transcribe and translate content from some of the videos, but covid has greatly slowed the pace of this work, too,” he continued. “Still, I had enough of the project done to complete my dissertation… and I’m grateful that I’ll be able to continue this work—once it’s safe.”

There are some facets of Joe’s venture which have spared it lots of the difficulties confronted by different language documentation initiatives through the pandemic: the connection between Joe and the Kanien’kéha audio system he labored with already concerned a good quantity of expertise, and he didn’t have to journey internationally to satisfy with speakers. Yet, the venture has practically floor to a halt simply from the issue of staple items like transferring round and assembly with folks.

Despite the setbacks, extra expertise is perhaps a technique to mitigate the results covid-19 has had on our capacity to proceed our efforts in language documentation. It could also be some time but earlier than we will get on a aircraft and go interview folks in an enclosed area with the boldness we had in 2019, however the regular march of language endangerment has not slowed, and documentation stays as vital as ever. Hopefully some mixture of tech like directional microphones and the normalization of digital conferences will permit us to handle how little we learn about areas like prosody on the planet’s languages, regardless of all the logistical setbacks the pandemic has introduced.

Ben Macaulay is a current PhD in linguistics from The Graduate Center, CUNY, now primarily based in Malmö, Sweden. His research focuses on prosody, intonation, and endangered language documentation.

Soundcloud pictures: Getty.