• 20111207-kiy.zip

Notes from Thurs 8 December 2011 (RH)

Morning session: We started today by recording a wordlist, organised by tone contour, that was written by Helja Clouse, who has done a lot of work on Kirikiri. This gave us approximately 2-300 new words. The tone contours given did not seem to match those we heard, and the words in one group were often clearly heterogeneous. The translator avoided words in the list that he felt would be difficult to translate, such as 'plant sp.'

Between two recording sessions we learnt a bit more about the speaker, which I feel helps us to better understand his linguistic background. We discovered that his linguistic community is in constant contact with many other linguistic communities, which may partly explain why the tones we elicited are so different sometimes from those of Helja Clouse.

We took notes on tokens that seemed different or interesting, such as the word for 'ironwood', which seemed to have very high tone in comparison with most other words. We also took turns at watching the wordlist during the elicitation to better understand the principles of transcription.

Afternoon Session: We split into two groups to find frames for the words we had already recorded. One group found Noun Phrases, using adjectives (black, large) to describe animals. The second group looked for verbs to put with the nouns we had. We found that in-context words sometimes have different tone contours, indicating sandhi.

We also found that the L-M-H system of description was sometimes insufficient. Despite both groups eliciting separately, both transcribers resorted to using arrows in addition to the L-M-H markings to better indicate tone contours. The possibility of more levels seems increasingly likely.

Additionally, we had the speaker sit on the same side of the room as the board. As well as allowing the group to more easily look between the speaker and the board, I found it allowed the transcriber to better engage with the speaker, better facilitating transcription.

Tomorrow's Plan: Continue with frames some more - we were getting some great results with the frame 'N is making a noise', and would like to further explore the consequences of different frames.

Audio from the 2nd day is available below.

• 20111208-kiy.zip

Notes from Fri 9 December 2011 (RH)

Morning Session: We started this morning with a discussion on what was ‘burning holes in our minds’ – what stuck out to us, anything we wanted to follow up, what we wanted to experiment with, and how we thought we were going.

We also discussed glides and their distribution – it is unclear where glides are different from long vowels, if at all, but we all agree that they might be important. Additionally, we made the point that we don’t want to get too caught up on transcription of individual words, and focus rather on working out categories of tone contour.

We decided that we needed to trial using a system with more than 3 tone heights, so we started using a four-tone system. Although initially it was difficult to swap, since we were used to using L-M-H, it became easier as we went on. So far the four-tone height system seems to be working well.

We started by getting the numbers 1, 2, and 3, and the word for many (3 is the highest number). We then used 1 as a frame: 1 bandicoot, 1 dog, 1pig etc. We also got some minimal pairs in this frame.

Some categories started appearing in the disyllabic words, so we shifted focus to those. We made columns for the 3 categories we had found, which were 3.3, 2.3 and 4.2 (with 1 as low, and 4 as high). We found a fourth category – 23.32, and so far those categories have been sufficient, with the following distribution:

        3.3 – 7 words

        2.3 – 6 words

        4.2 – 2 words

        23.32 – 2 words

This took us through to lunch, so we did our recording. In the recording process, in addition to making one recording of all the new words and phrases we’d elicited, we made a second recording of our categorised disyllabic words. We discovered one definite error, and found that there were two words in the second category that, although definitely fitting there in the frame, seemed to have a different contour in isolation.

Afternoon Session: We went over the morning’s data, column by column, and evicted some words from their columns. We rearranged the words and ended up with:

        3.3 – 6 words

        2.3 – 3 words

        4.2 – 3 words

        23.32 – 5 words

We then looked for more disyllabic words and put them into our columns. We reclassified our columns to group words by their forms both within the frame (one N) and in isolation. We soon found a fifth pattern. We have also had to be careful to make sure we’re allowing for vowel height, which changes tone height, giving us allotones.

The five categories we have now are (arranged isol/context – number of words):

        Category 1: 3.1/3.3 – 7

        Category 2: 2.3/2.4 – 7

        Category 3: 4.2/4.2 – 5

        Category 4: 12.31/2.4 – 13

        Category 5: 4.4/2.4 – 3

The spreadsheet with the categories is attached below (as 20111209-3-kiy-ap-numphraseToneset.xls).

We then recorded our session in order of the columns, and were very happy with the results. We plan on randomising the words at some point and re-recording them to ensure that the speaker isn’t being primed by saying the words in their groups.

Finally we looked at the advantages and disadvantages of moving on to trisyllables or monosyllables for next week’s work.

• 20111209-kiy.zip

Notes from Mon 12 December 2011 (RH)

Today we swapped groups, 2 members at a time (usually). From my perspective, it was interesting to see how other groups are going. Both the Iha and Alekano groups have different group dynamics to Kirikiri, with the former a bit more 'student'-led, and the Alekano group a bit more 'leader'-led. It was interesting to note that members of each group commented on how nice it felt to go back to their 'home' group after a time in other groups. I think that for the visitors to make a really good contribution to their temporary groups, more time would need to be spent to really get a handle on how the other languages work. I think also that we've tended to all get quite invested in the languages we're studying. It will be interesting to see how it works when we swap leaders on Wednesday - whether perhaps the students take over more, or whether the leaders will become just as invested in the new language.

On today's home-group work, there are details given in More Notes below. I've attached 2 versions of today's spreadsheet of extended tone categories, which (as you will learn below) now includes monosyllables and trisyllables. One version is a pretty version, which allows you to view the different words in their categorical columns. The other version is more useful for analysis, and has all the words running down the spreadsheet.

The sound file that goes with the spreadsheets is below.

• 20111212-1-kiy-ap-toneset.wav

More notes from Mon 12 December 2011 (acw)

Morning Session: On the basis of Friday's work, everyone was asked to try a tonological analysis to capture the five tone categories in their isolation and their 'one ___' contexts. Such an analysis might attempt to account for five isolation forms and the merger of Categories 2,4,&5 (see above) in the 'one___' context, using (ideally) fewer rather than more tonological elements like L, H, and M. We didn't compare analyses but my hope was that just HAVING an analysis would give everyone a further stake in today's developments.

We decided we wanted more contexts to test the disyllables; and some monosyllables. We decided to pursue contexts first.

Isolde tested several postnominal possessive pronouns, settling on /nara/ (Cat3) 'your, pl'. We tried it with the first three disyllables of each category. It appeared to preserve 5-way distinctiveness; nevertheless it introduced several minor sandhi changes.

We then began collecting monosyllabic examples, starting with minimal sets in our material and in material from the Clouses provided by Isolde. We quickly ended up with four distinct categories straightforwardly associating themselves with all categories but Cat4. Daunting 6 or 7 way sets from the Clouses were handled either by noting differences in the vowels or, in two cases, homonym pairs.

Afternoon sessions. We continued with monosyllables briefly but soon became intrigued by the possibility that longer, 3-syllable words might 'spread out' the more complex melody of Cat4. We collected one Cat4 trisyllable right off (it stretched the initial rise over two syllables and then proceeded as usual). Subsequently, all our examples were associated with Cat1 or Cat3 in that their final two syllables had the tones we expected for those categories.

In the last half hour, with Mark Donohue's help as translator, we made recordings of the monosyllables and trisyllables in isolation, with 'one', and with 'your pl.'; of the first three of each set of disyllables with just 'your pl.'. We want to make more recordings of the disyllables in context, as well as of some new disyllables that we elicited by the boards during the day.

Notes from Tues 13 December 2011 (RH)

Morning Session: We started with a discussion on what we think is going on. Peter explained his ideas, and he and Tony began to work out, step-by-step, what was going on with the underlying tone system. So here is our current analysis:

They decided that it looked like there was a high tone candidate in Cat.2, because it rises consistently. Similarly, there looked like a high first tone in Cat.3, and all of Cat.5 looks very high. We also suspected that Cat.3 was experiencing downstep.

This gave us:

Category	1	2	3	4	5
Elicitation Isolation Form	3.1	2.3	4.2	12.31	4.4
Elicitation Context Form	3.3	2.4	4.2	2.4	2.4
Underlying Isolation Form	X-L	X-H	H-H	H-L	H-X
Underlying Context Form		X-H		X-H	X-H

So why does it do this?

We decided that Cat.5 and Cat.4 could have right shift, with low absorption into the following syllable in Cat.4, with Cat.1 going to X-X with a default mid tone following low absorption into the following syllable. Also that on monosyllables the L tone on Cat.4 gets disassociated and becomes like Cat.2. It is possible that there isn’t enough room on the monosyllabic words for the more complex tone contour. We’ve been using what Tony calls an underspecification analysis.

So what happens with a high frame?

The second person plural possessive form is high. In ‘your wasp’ the H moves over again and there’s just no room for the L, so it just sort of vanishes. But it does actually pull down the peak of the first H in ‘your pl’ – so the peak H in Cat.4+your = lower than peak H in Cat.2+your.

I’m beginning to see how you would get multiple, highly different analyses – you get on a path and follow it through and make things fit, so if you started on a different path you’d doubtless get elsewhere. Tony discussed this. He says it’s necessary and that we now have to look at how our data fits, and he points out that if you’re putting your money in the right place, you’re putting your attention in the right place. So in Wednesday’s swapping, one of the things the other ‘leaders’ will do is act as sceptics, which is what we need.

So what we have now:

Category	1	2	3	4	5
Isolation	X-L	X-H	H-H	H-L	H-X
Context	X-X	X-H	H-H	X-H	X-H

We then did some possessives with nouns, such as pig’s tongue, dog’s tongue, etc. It looks like the context form of nouns is the same, regardless of the contextual frame.

Afternoon Session: We discussed how the morning session had gone. We decided to keep going with possessive noun phrases. We noticed that the word order for alienable forms seemed to be opposite for inalienable forms (pig’s head but string person’s). We didn’t have an inalienable word for Cat.5, and even after quite a bit of searching we couldn’t come up with one.

Fortunately though, we discovered that for some reason string (inalienable) comes after the noun anyway, so we drew up a grid and made a recording. We now think that it was the word ‘person’ that is treated differently, rather than ‘string’.

I also made a new version of the noun category spreadsheet – updated pretty and analysis versions are in the bundle.

Finally, we had a quick look at some suffixes. We looked at two – the benefactive high rising /li/, and the suffix indicating the possessor of something, low /yo/. Both were remarkably stable, and their addition to nouns puts the nouns in their still-consistent context forms, rather than leaving them in their isolation forms. We did another recording, and made sure to use a range of disyllables and monosyllables, as well as words with different final vowels.

Tomorrow we’re planning on looking further at trisyllables and seeing what it’s all about, keeping an eye on possible compounding. We’ll also be thinking about trying some larger noun phrases, such as the small person’s fish arrow. We might also try some very basic sentences – it will be a sort of fishing expedition for anything interesting, with (we think) the potential to yield some excellent stuff.

The bundle is attached below:

• 20111213-kiy.zip [manifest]

Addendum to notes from Tues 13 December 2011 (acw)

Regarding the analysis above, it's worth also laying out the rules it entails:

1. Downstep. In order for /H-H/ to represent a sequence [1-3] (1 is high, 4 is low), we must assume that the second of two H tones in a word domain will lower, fairly drastically, almost to the bottom of the range.

2. L-delinking or deletion. As mentioned above, categories 1 and 4 show a final L which is absent in the context forms: this process would delink or delete that L in a non-phrase-final (or similar) context.

3. H tone spreading. In Cat 5, this rule would spread initial H across the following X.

4. H peak delay. This rule would delay the realization of H peaks. It would account for the realization of H-L (cat 4) as [12-31]. It might also account for the rise of H in cat 2 isolation forms only to [3]. In the context forms, it could be that the same peak is fully realized because it is delayed into the next word. Possibly by sleight of hand, this rule may also be said to yield the cat 5 context forms /X-H/. I say 'sleight of hand' because rule 3, H tone spreading, could reasonably be thought to spread the H in the cat 5 context form.

5. H-L simplification in monosyllables. Cat 4 is not distinct in monosyllables. Kristine has offered an explanation for the merger in terms of the accommodation of (possibly delayed) peaks and falls on monosyllables. On that scenario, /H-L/ would be indistinct from /X-H/, the Cat 2 sequence.

6. Default realization of X. In general, X is realized as a mid or low mid pitch.

Discussion: All things being equal, a set of rules like this would be better justified if it applied ACROSS word boundaries, rather than only WITHIN the word domain. However, our afternoon work indicated that this is not the case. Nor is it even the case that our X slots can be filled by spreading from a prior word.  It therefore must remain a scheme for word-INTERNAL tonemics whose result is presentation of internal tonemics in the constrained, six-way space provided by: /{H,X}-{H,L,X}/

Notes from Wed 14 December 2011 (RH)

Morning Session with Tony: We had a discussion of what we wanted to do. Following on from yesterday, we knew that we wanted to look at basic sentences – Peter suggested we look at transitive and intransitive sentences rather than only intransitive. Rebecca suggested that we should try to find out whether the isolation form occurs in any contexts at all, or only in isolation. Sutri suggested we look at paradigm structure as an amendment to the basic sentences idea.

We started by eliciting some words and trying to transcribe them into out their forms. On the board we wrote:

Category #	Iso	Context
1	X-L	X-X
2	X-H	X-H
3	H-H	H-H
4	H-L	X-H
5	H-X	X-H

We began eliciting ditransitives with the verb ‘gave’, such as ‘the bat gave a net to the grasshopper’. The sentence order is Subject Object Beneficiary Verb, and we discovered that this cannot be moved around if the result is ambiguous, suggesting that we don’t have super obvious PoS markers. (So you could have ‘the bat gave a net to the grasshopper’, or ‘the bat gave a grasshopper to the net’, and both mean the first. But you can’t do this with 3 animate things, because it comes to mean something else.)

We noticed that having all three non-verb positions taken up by cat.1 means that there is a down step, or down drift, from one to the next with all three. We also noticed that the opposite happened when they were all cat.2 – we had updrift between the first 2 cat.2s, and then the third (the beneficiary) dropped down quite a bit before rising.

With Larry: We stayed with verbs, but changed to the transitive verbs ‘see’ and ‘carry’. It was all very predictable, with the nouns in their context forms regardless of whether they were subject or object. So we started on locatives – in, to, from, on, under (the house). In and from were the same. It was all predictable except for on (on top of), which stayed level where we expected a rise. It looked like cat.3, which should have been a high fall, was a high level tone, that cat.1, which should have been mid level, became low level, and cats.2 and 5, which should have been mid rise, became mid level.

Afternoon Session with Keith: We started focussing on monosyllables, finding out what’s going on with ‘on top of’ stuff. Cat.3 now seemed to have a slight fall, but cats.1, 2 and 5 all are level at times.

Keith very keen on getting EXACTLY the right pitches/tones – describing accurately what’s going on. 

In summary – we’re trying to find frames that make things do exciting, unpredictable things. Then we can work out why, rather than having our system seemingly so well-fitted.

Tomorrow we’re going to work more on monosyllables. Also – Isolde thinks he’s using ‘go to the house’ instead of ‘to the house’. So we need to try some other locations.

• 20111214-kiy.zip [manifest]

Notes from Thurs 15 December 2011 (RH with PA and LB)

Morning Session: We started with a discussion on what we want to do today (the last day as a group!). We’re all agreed that cracking monosyllables would be great, but first we need to get some more! Then we plan on working out exactly how they work. If we can we want to get a few trisyllables too, since we seem to be able to get a better hearing of tone – there’s more space for it to spread out.

We also decided how we’re going to present the data. We’ll have a comprehensive database, as well as a collaborative ‘cut-and-dried’ report documenting how the categories work, and giving the analysis and rules that we’ve worked out. In addition, we’ll have a personal paragraph from everyone to give it a bit more interest. We’re also keen to have a version that is very easy to access for people who just want a quick taste of Kirikiri without getting bogged down in data.

In addition, Tony is keen for us to write up some more articles that focus on the analyses of parts of Kirikiri in more detail, that would probably end up as individually authored.

We then started on some monosyllabic possessive NPs, such as wallaby’s thigh. We found enough sentences to put monosyllables from each category in each position to work out what is happening. We ended up with some funny stuff – Cat.5 dropped lower than expected phrase-finally, and Cat.3 that were also lower than we expected. Combined with yesterday’s work, it was looking like our analysis needed revision.

Tony pointed out that if both Keith and Tony had had problems with our Cat.3 downstepping rule, it probably needed reconsidering. So Peter came up with our…

New analysis!

Category	Old ‘Iso’	New ‘Final’	Old ‘Context’	New ‘Non-Final’
1	X-L	X-X	X-X	X-X
2	X-H	X-H	X-H	X-H
3	H-H	H-L	H-H	H-L
4	H-L	H-X	X-H	X-H
5	H-X	H-H	X-H	X-H

So there is no more low deletion/absorption. Instead we have phrase-final Xs falling, which we would expect, and downstep in an H-L-H sequence, which is also very common.

Isolde wanted to try some adjectives with our NPs – from centipede’s breast to centipede’s big breast (or big centipede’s breast?). We ended up with both these forms, and tried other tone categories.

Afternoon Session: We talked about what our afternoon plan is. We’re going to keep going with monosyllables, but first we want to work out whether monosyllables are affected by vowel height in the same way as disyllables. We also have a lot of recordings that we want to do, so Kristine has written up a list and we’ll break them up over the afternoon (and possibly some of tomorrow, now that we’re going to have until 3.30 tomorrow afternoon).

It appears that monosyllables are indeed affected by tone height. We really need some more with high vowels to make sure, so we’re going to try and get some.

Then.. WE FOUND A CAT.4 MONOSYLLABLE!!!

This is good because it means we no longer need to find a way for our Cat.2 and 4 monosyllables to collapse or cancel. In particular, we can drop our completely arbitrary rule that said that there simply weren’t any category 4 monosyllables. Instead we can now provide a fully fleshed out analysis of monosyllables.

The rest of the afternoon we kept doing recordings, and trying to get a range of vowels in the same position, eg ‘ka, ke, ki, ku’ in the same tone category, to really contrast vowel height.

Addendum for Thurs 15 December 2011 (acw) - Our new analysis

We propose a slight revision of our representation of the tonal sequences for the five lexical tone categories of Kirikiri, as follows:

Category            Sequence           

1            X-X           

2            X-H

3            H-L

4            H-X

5            H-H

In final form, these sound as follows (1 is low, 4 is high)

1            3-1

2            2-3

3            4-21

4            12-31

5            4-4

We assume that there is a phonetic utterance-final lowering effect that accounts for the drop in the toneless category 1, and at the end of category 4.

We also assume that there is a peak-delay effect. This would delay the H target of 2 so that it reaches only the ‘3’ level. It delays the H peak in category 4 to the beginning of the second syllable when the sequence is realized on disyllables.

Linking. We assume that these tone sequences link to the last two syllables of words of two or more syllables. In monosyllables, we always hear a long vowel, suggesting perhaps a dimoraic representation. The tones then can link one per mora. The distinctiveness of category 5 supports this dimoraic hypothesis. However, there is also a merger in monosyllables of categories 2 and 4, suggesting dimoraic monosyllables neutralize X-H and H-X. The specific mechanism may be a constraint against a contrast between two single-tone syllables that depends solely on first- vs. second-mora linking.

Realization of toneless (X) syllables or moras. Aside from the cases covered above, we assume that toneless syllables or moras are realized at level 3 or 2.

In non-final forms, we initially recorded the tones as:

1            3-3

2            2-4

3            4-2

4            2-4

5            2-4

We now are not convinced that categories 2, 4, and 5 are absolutely merged in non-final form. We now prefer an account where phonetic processes apply whose extreme endpoint may be the merger thought we heard, but which also allow for degrees of distinctness among these words. The account would rely on an absence of utterance-final lowering to give the flat level-3 realization of the toneless category 1; and the lack of a fall at the end of category 4. The account would also rely on peak delay in categories 2, 4, and 5 in non-utterance final position to produce what we heard as 2-4 in all three of these cases.

• 20111215-kiy.zip [manifest]

Notes from Fri 16 December 2011

Morning session: We jumped straight into recordings – one of which was a narrative on how Kirikiri fish. We managed to get some interesting stuff, and between other recordings, spent much of the morning transcribing notes for it in Praat. Other group members were writing up bits and pieces about the research.

Afternoon session: We did the same as in the morning – recordings and writing up. Rebecca wrote up a series of possible topics for the edited volume and arranged them by theme into 4 chapters and an introduction. Rebecca also volunteered to be the ‘whip-cracker’ – the person responsible for hassling everybody to write their pieces and get them in on time.

• 20111216-kiy.zip [manifest]

Archived Materials

• http://www.archive.org/details/KirikiriProsody