Mandrin-K, part one: the fifth day

The mouth of the cave#

The place is a karst gorge in the southwestern Carpathians, roughly forty thousand years ago, on the same ground that will later give up the Oase jaw and the Romăneşti workshops. The afternoon is cold. Grass and juniper cover the slope, and pine grows down along the river. The cave runs about fifteen metres into the limestone, and its floor is scattered with old cave bear bone. Vlad has a fire going and his kit laid out beside it. A woman comes up the scree after something she has been tracking and smells the smoke. She is in her twenties, dark-skinned, dressed in hide stitched with sinew and ivory beads on a cord at her throat. She carries a spruce spear with an antler point.

Nika

stopping in the mouth of the cave, bringing the point up level Ktha. Ktha ne-e.

Vlad

lifting a hand and pulling a notepad, and then drawing a lone figure followed by a jumping arrow, spear, deer, jumping arrow, cave symbol, cut arrow, and showing the inscription

Nika

taking one step back and lifting the point higher Ktha! holding still at the mouth of the cave, watching the white square in his hand Semek. Semek na ktha. moving her chin at it once, then flicking the spear point down at the ground between them Aha. Aha-ke.

Vlad

drawing a mouth and three semicircles on the right and left on the next page and then breaking the paper and handing the paper and a pen and gesturing on his paper the act of writing

Nika

leaving the paper in his hand, keeping the spear between them Ne. Ne-e. watching his hand move on the second sheet, then looking at the torn edge he is holding out Semek ta. lowering the point a little and going down on her heels, still in the light, one hand on the rock beside her Ha.

Vlad

pointing at himself I pointing at the ground been here pointing to the entrance from there gesturing at the four reference unary segments on the cave wall for pointing to the sun five times five days pointing at the reference stick on the wall near the unary reference alone.

Nika

following his hand to the marks on the wall and staying on them Semek. standing back up and pointing at the marks herself, then out at the sky, then back at the marks Ktha ne? Ktha ne-e? Aha ta semek. Ta semek!

Vlad

pointing at the reference closed mouth and crossed on the cave wall above the unary and lone figure Cannot communicate.

Nika

looking at the mark above the figure Ne. coming two steps in, out of the light, and setting the bottom of the spear on the gravel Semek ne-e. Ktha. tapping her own chest twice Aha. Aha!

Vlad

tapping the ground and pointing at the sitting figure on the left of the reference closed mouth

Nika

sitting down on the gravel three metres from him, spear laid across her knees Aha. pointing at herself, then at the ground beside her Aha, ktha ne. pointing at him Semek?

The lit face and the map#

Vlad

Ne. pulling out the rugged laptop and logging into the already open session, then pointing at the taskbar XFCE pointing at the rat on the wallpaper rat laying his hand on the keyboard keyboard laying his hand on the screen screen outlining the frame of the screen screen frame outlining the frame of the keyboard keyboard frame pointing at the sticker processor sticker lifting the laptop and drawing a small square on the left side of the laptop's bottom with his finger processor arching the laptop with his hand laptop tapping on the shields shields.

Nika

coming forward on her knees until she is close enough to see the light coming off it Ktha. Ktha, ktha. putting her hand out towards the screen and stopping short of it, then pulling the hand back Ne. pointing at the fire, then at the screen Semek? Semek na?

Vlad

opening a map of the locale, and pointing at the dot location, current zooming out the map and circling with the cursor area.

Nika

watching the shapes slide under his finger Ktha ne-e. turning her head to look at the mouth of the cave, then back at the screen Semek. putting her hand on the gravel floor, then lifting it and holding it in front of the screen Aha ta?

Vlad

pulling out a table with twenty basic English words, associated with a drawing

Table of twenty base words

Word	Drawing
water	wavy lines under a cupped hand
fire	flames over sticks
eat	hand at an open mouth
sleep	figure lying down, eyes closed
walk	figure mid-stride, footprints behind
come	figure with arrow pointing toward a second figure
go	figure with arrow pointing away
I	figure with a finger on its own chest
you	figure with a finger pointing outward
give	hand holding an object out to a second hand
take	hand closing on an object
big	small circle beside a large circle, arrow to the large one
small	large circle beside a small circle, arrow to the small one
one	one stroke
two	two strokes
many	seven strokes
man	standing figure
woman	standing figure
child	short figure beside a tall figure
dead	figure lying down, eyes open

Nika

going still with her eyes on the seven strokes Ne-e. putting her finger on the screen over the strokes, one after another, and making a short sound at each Ktha. Ktha aha. Ktha aha semek. lifting her hand off and looking at him

Vlad

Many things.

Nika

going back to the first stroke and starting again Ktha. touching the two strokes Ktha aha. touching the seven Ktha aha semek. sitting back on her heels and holding up one finger, then two, then spreading the whole hand

Vlad

One, two, three, four, five, six, seven, eight, nine, ten.

Nika

listening to the run of sounds and then shaking her head once Ne. holding up one finger Ktha. adding the second Ktha aha. opening the whole hand and pushing it towards him Ktha aha semek. lowering the hand and touching the seven strokes on the screen again Semek.

Naming what is drawn#

Vlad

pulling out an exercise on the screen where a human is asked to sound the specified drawings, to establish the other's language

Elicitation sheet 1: nouns and actions

Show one panel at a time. Point at the panel, then at the speaker's mouth. Wait and record what is said. Do not speak while waiting.

Panel	Drawing	Response
1	sun, high, with rays
2	sun, low, at a horizon line
3	moon, full
4	wavy lines in a hollow
5	flames over sticks
6	four-legged animal with branched antlers
7	four-legged animal, humped shoulder, short horns
8	four-legged animal, long tail, mane
9	large four-legged animal, standing on hind legs, claws
10	bird in flight
11	fish
12	tree
13	pointed stone bound to a shaft
14	hand holding a shaft, arm drawn back
15	figure lying down, eyes closed
16	figure lying down, eyes open
17	tall figure and short figure, hands joined
18	opening in a rock face
19	falling white specks over a hill
20	hand at an open mouth

Nika

putting her finger on the antlered animal Rusa. moving to the humped one Rusa ne. Bontu. to the maned one Kesh. to the standing clawed one, and pulling her finger off it quickly Torna. Torna-ke. looking up at him and tapping the sun panel Miht.

Vlad

pulling out another exercise, about possible suffixes and prefixes

Elicitation sheet 2: modification of a known word

Use only words already recorded on Sheet 1. Show the paired panels together. Point at the left panel, then the right. Point at the speaker's mouth. Wait and record what is said. If the second sound repeats the first exactly, mark the pair and move on.

Panel	Left drawing	Right drawing
1	one antlered animal	seven antlered animals
2	one antlered animal	one antlered animal, small, beside a large one
3	one antlered animal	one antlered animal lying down, eyes open
4	one antlered animal	half an antlered animal, cut line
5	antlered animal standing	antlered animal running, legs extended
6	antlered animal near	antlered animal far, small, over a horizon line
7	figure holding a shaft	figure with empty hands, shaft on the ground
8	figure walking toward a cave	figure walking away from a cave
9	sun high	sun high, seven strokes beneath
10	fire burning	fire out, smoke only

Nika

at the first pair Rusa. Rusa-ktha. at the second Rusa. Rusa-me. at the third, slower Rusa. Rusa-ke. stopping her finger over the fourth and looking at him instead of the screen Torna-ke. Torna-ke rusa? putting her hand on the spear across her knees

Vlad

pulling out a speech training exercise

Elicitation sheet 3: connected speech for model training

Recording is running. Do not interrupt. Show one strip at a time. The panels in a strip are read left to right. Point along the strip once at normal speed, then point at the speaker's mouth. Whatever length of speech comes back is kept. Repeat each strip twice. Fifteen strips is the minimum for a first pass.

Strip	Panels, left to right
1	figure, walking, antlered animal, shaft in hand
2	figure, walking, antlered animal, shaft on ground, antlered animal running away
3	two figures, walking, cave opening, fire
4	sun high, figure walking, sun low, figure lying down eyes closed
5	falling white specks, no animals, figure at an open mouth with an empty hand
6	clawed animal, cave opening, figure walking away
7	tall figure, short figure, water in a hollow, hand cupped
8	figure, fish, water, hand closing on the fish
9	seven figures, walking, humped animal, seven shafts
10	figure lying down eyes open, other figures standing around it
11	fire out, smoke, figures walking away from a cave
12	maned animal, running, figure standing, shaft lowered
13	tree, bird, figure looking up
14	moon, figure lying down eyes closed, sun low, figure standing
15	figure, other figure, hand holding an object out, hand closing on the object

Nika

watching his finger cross the first strip and saying nothing until it stops Semek na aha rusa. Ktha semek, ktha rusa-ka miht-ne, aha ta bontu-me na kesh. at the second strip, faster and with her hand moving alongside the panels Rusa-ktha! Rusa-ktha semek na, aha ne torna, ne torna-ke, semek ta rusa miht-ka na aha. at the sixth she stops and puts two fingers on the clawed animal Torna. pointing at the back of the cave, at the fissure, and then at him Torna ta. Torna ta, semek!

States, contour and the pegging#

Vlad

pulling out an exercise about describing feelings and grammatical constructions

Elicitation sheet 4: internal states and sentence frames

Two parts. Part A is faces and postures only. Point at the panel, then at the speaker's own face, then at the speaker's mouth. Part B is frames: the same panel returns with one element changed. Point at the changed element. Wait and record. If the response is a single sound repeated, mark the frame as failed and move on.

Part A

Panel	Drawing
1	face, mouth turned up, eyes narrowed
2	face, mouth turned down, water on the cheek
3	face, mouth open wide, eyes wide, figure leaning back
4	face, brows down, teeth showing, fists closed
5	figure sitting alone, other figures small and far off
6	figure with hand on stomach, empty hollow beside it
7	figure lying down, hand on a wound, red mark
8	figure with both arms raised over a fallen animal
9	figure looking back over its shoulder at a clawed animal
10	two figures, hands joined, faces turned to each other

Part B

Frame	Panel
1	figure eats fish
2	fish eats figure
3	figure does not eat fish, fish whole, cut line through the eating hand
4	figure ate fish, sun low behind, bones only
5	figure will eat fish, sun high ahead, fish whole
6	figure gives fish to second figure
7	second figure gives fish to figure
8	figure with fish, question mark absent, second figure hand open and palm up
9	one figure eats, seven figures eat
10	figure eats with a shaft, figure eats with a hand

Nika

at the first face Hesu. at the second Hesu-ne. Kamla. at the fourth, and she makes the face herself while she says it Torna-hesu. at the fifth she goes quiet for a while and then puts her whole hand over the panel Aha-ke. Aha-ke, aha-ke. looking up at him and then back down at the panel, and saying it once more, quieter Aha-ke.

Vlad

pulling out an exercise about punctuation marks and klingon script question

Elicitation sheet 5: prosody, polarity and orthographic pegging

Four parts. Parts A to C establish the four sentence types by contrast. Part D is for the operator and needs no panels.

Part A: Prosody

The same panel is shown four times with one mark changed above it. Point at the mark, then at the panel, then at the speaker's mouth. The four marks are the operator's own, and the speaker will not read them. What is wanted is the contour that comes back on each pass. Do not point at more than one mark in a single pass.

Pass	Panel	Mark above panel
1	figure eats fish	flat bar
2	figure eats fish, second figure with palm up	rising hook
3	figure eats fish, both arms raised	tall stroke
4	figure eats fish, cut line through the eating hand	crossed bar

Part B: Polarity

Frame	Panel	Wanted
1	fish whole in water	affirmative
2	fish whole in water, cut line through the fish	negative
3	fish whole in water, palm up beside it	interrogative
4	fish whole in water, figure with both arms raised	exclamative
5	clawed animal at the cave opening	affirmative
6	clawed animal at the cave opening, cut line	negative
7	clawed animal at the cave opening, palm up beside it	interrogative
8	clawed animal at the cave opening, figure running	exclamative

Part C: Endings

Run Part B a second time. The four responses in each block share a stem. Segment on the stem and record whatever precedes or follows it. If the difference is carried only by contour and no segment changes, mark the block as prosodic and note it. The translator will take its polarity from whichever channel the speaker uses, and a language that marks both is expected.

Part D: Orthography

Decide before transcription begins and do not change it afterwards.

Question	Options
1	Latin as given, or Latin with added letters for sounds it lacks
2	Case distinction meaningful, or case ignored and all transcription in one case
3	Pegged language for the model's output side

If case is meaningful, the transcription can carry a distinction the speaker makes with no extra letters. If case is ignored, every distinction needs a letter and the alphabet grows. The pegged language is fixed once and everything the model produces is anchored to it.

Nika

at the second pass, watching the mark and then the panel Semek na aha? at the third, louder, and the hand comes up with it Semek na aha! at the sixth frame of Part B, with her finger on the cut line Ne torna. Ne torna ta. at the seventh, holding the palm up herself in copy of the drawing Torna ta? coming back to the sixth and saying it a second time with the ne pushed forward Ne-e torna ta.

Vlad

configuring the model to be Latin as given, case ignored, and pegged to Romanian first, English second, Czech third, Polish fourth, and pulling out an exercise about establishing the nominative case, and deriving past times from passe compose

Elicitation sheet 6: case and anterior tense

Transcription is latin as given, case ignored, all lowercase. Output side pegged: romanian first, english second, czech third, polish fourth. Two parts.

Part A: The bare form

The aim is the form of a noun standing alone, with nothing done to it and nothing done with it. Every panel in this part shows one animal or object at rest, unattended, with no figure present. Point at the drawing, then at the speaker's mouth. Run the whole part before Part B. Whatever comes back here is the citation form and everything later is measured against it.

Panel	Drawing
1	antlered animal standing alone
2	humped animal standing alone
3	maned animal standing alone
4	clawed animal standing alone
5	fish in water
6	bird on a branch
7	tree
8	shaft with a stone point, lying on the ground
9	fire burning, no figure
10	water in a hollow
11	cave opening
12	stone

Then the same twelve, each as the one acting. Panel 1 becomes the antlered animal walking, panel 4 the clawed animal striking, panel 6 the bird taking off, panel 12 the stone falling. Then the same twelve, each as the one acted on: the antlered animal struck by a shaft, the fish in a closing hand, the tree cut, the stone lifted. Three passes over the same twelve drawings. Segment across the three and record what changes at the edges of the stem.

Part B: What is finished

Each frame is a pair of panels in a single strip. The left panel is the action. The right panel is the same scene after, with the sun moved back below the horizon line and the result standing. Point along the strip once, then at the speaker's mouth. The wanted form is the one that says the thing is done and over.

Frame	Left panel	Right panel
1	figure striking an antlered animal with a shaft	animal down, sun low behind, figure standing over it
2	figure eating fish	bones only, hollow empty, sun low behind
3	figure walking toward a cave	figure inside the cave, footprints behind, sun low behind
4	fire burning	smoke only, cold sticks, sun low behind
5	figure giving a stone to a second figure	second figure holding the stone, first figure hands empty, sun low behind
6	figure lying down with eyes closed	figure standing, sun low behind
7	falling white specks	white on the ground, no specks falling, sun low behind
8	clawed animal at a cave opening	tracks at the opening, no animal, sun low behind

If the right panel returns the same form as the left, the language does not mark it here and the frame is repeated with two suns drawn back instead of one. If a separate word appears in the right panel and nothing changes on the stem, record that word and mark it as a particle. If both appear, record both, and note whether the particle can be dropped.

Nika

at panel eight of Part A, and reaching over to touch her own spear at the same time Naga. at the third pass, with the shaft drawn in the animal Rusa-tu. Naga-me rusa-tu. at the second frame of Part B, following his finger from the fish to the bones Semek na aha ontu. Ontu. saying it again with the hand turned over and dropped Aha ontu. Ontu-ke.

Borrowing and the ground#

Vlad

pulling out an exercise about if words can be loaned

Elicitation sheet 7: borrowing

Two parts. The question is whether a sound with no drawing behind it can be taken into the speaker's own speech and given back, and whether it survives a gap.

Part A: Repetition without meaning

Say each string once at normal speed. Point at the speaker's mouth. Record what comes back. Do not correct the speaker and do not repeat a string more than twice. Strings 1 to 6 use sounds already recorded from the speaker. Strings 7 to 12 use sounds not yet recorded from the speaker.

String	Said by operator
1	ka
2	ka-mi
3	to-na-se
4	hesu-ma
5	ne-ta-ko-ru
6	semek-oru-tan
7	vlad
8	zbor
9	pluie
10	strzemie
11	vrch
12	Kinkyu Jishin Sokuho desu. Tsuyoi yure ni keikai shite kudasai

String 12 is long and carries no meaning for either party. What is wanted is the length held, the break between the two parts, and whether the fall at the end is reproduced. Segment the response and record how much survives.

Part B: Attachment and retention

Point at a drawing the speaker has already named. Say the operator's own word for it once. Point at the drawing again. Point at the speaker's mouth. If the speaker returns the speaker's own word, the borrowing has failed and the frame is marked. If the speaker returns the operator's word, wait, run three unrelated panels from Sheet 1, then come back to the same drawing without saying anything and point at the speaker's mouth a second time.

Frame	Drawing	Operator says
1	fire burning	foc
2	water in a hollow	apa
3	antlered animal	cerb
4	stone	piatra
5	figure with a finger on its own chest	vlad
6	shaft with a stone point	sulita
7	sun high	soare
8	cave opening	pestera

Frame 5 is the operator's own name and is run last. Retention on frame 5 with no retention elsewhere means the speaker takes names and refuses nouns.

Nika

at string 3 To-na-se. at string 7, watching his mouth and coming back a little short Ovlat. Ovlat-ke? at string 12, holding the length and putting the break where he put it, the sounds falling apart in the middle and coming back for the ending Kinku shin sokuho tesu. Tsuyo yure ke keka shite kutasa. at frame 5, after the three panels have gone past and his finger comes back to the chest drawing on its own Ovlat.

Vlad

pulling out an exercise about navigation comprehension

Elicitation sheet 8: wayfinding and unsupported referents

Two parts. Part A tests whether the speaker reads a surface as ground. Part B tests whether a drawing with no referent in the speaker's world returns anything at all. Part B is expected to fail and the failures are recorded.

Part A: The surface as ground

Show one panel at a time. Point at the drawing, then out at the real slope through the cave mouth, then at the speaker's mouth. Record whether the response names the drawing or names the ground.

Panel	Drawing
1	line running from a cave opening to a river, seen from above
2	the same line with a figure standing on it partway along
3	the same line with a fork, one branch to a river, one to a tree
4	a ring of hills seen from above, cave opening marked with a dot
5	the same ring seen from the side, hills as a jagged edge
6	footprints in a line, running away from a cave opening
7	footprints in a line, running toward a cave opening
8	sun low at one edge of the panel, cave opening at the other
9	a river drawn as two long lines, a fish between them
10	the same river with a crossing marked, stones in a row

If panel 4 returns nothing and panel 5 returns a name, the speaker reads elevation and refuses the overhead view. Run panels 1 and 6 again at the end and check the answers hold.

Part B: Referents with nothing behind them

Show one panel at a time. Point at the drawing, then at the speaker's mouth. Wait no longer than for the panels in Part A. No response is a result. A response naming a part of the drawing rather than the whole is a result and the part is recorded.

Panel	Drawing
1	a figure holding a small flat object, three curved lines coming down to it from a point above
2	a point above the earth, four lines going down to four separate figures
3	a long box on wheels, running on two parallel lines, figures inside it
4	figures standing in a row beside the same box, one stepping in
5	a box with a lit face, a second box below it, a hand on a flat panel of small squares
6	a stack of discs on a spindle, an arm reaching over them
7	the same stack, cut open, with rings drawn on one disc
8	a road with a moving box on it, a line drawn ahead of the box to a cave opening
9	a figure at a desk, the lit face in front, marks on it
10	the lit face alone, filled with the same line and fork as Part A panel 3

Panel 10 is run last. It shows a drawing the speaker has already read, placed inside a frame the speaker has no use for. If panel 10 returns the panel 3 response and nothing else, the speaker reads the content and discards the frame.

Nika

at panel 5 of Part A, running her finger along the jagged edge and then lifting her chin at the real ridge through the cave mouth Karnu. Karnu ta, karnu ta. at panel 4, going quiet and putting the palm over it Ne. at panel 6, standing the two fingers up and walking them across the gravel away from the fire Aha na. Aha na ontu. at the last panel, after a long look, moving her finger straight past the frame and onto the fork Karnu-me na kesh. Karnu-me na miht-ne.

The machine taken apart#

Vlad

pulling out an exercise about understanding abstract actions

Elicitation sheet 9: abstract actions

Three parts. Part A uses actions the speaker performs. Part B uses the same actions with the object removed. Part C applies the result to a system of operations that exists only on the operator's screen. Part C is expected to fail and the failures are recorded.

Part A: Actions with an object present

Point at the drawing, then at the speaker's mouth.

Panel	Drawing
1	hand taking a stone from a pile
2	hand putting a stone onto a pile
3	hand splitting one stone into two
4	two hands pressing two halves back together
5	figure copying a mark from one surface onto a second surface
6	figure rubbing a mark off a surface
7	figure comparing two surfaces side by side, marks on both, one mark different
8	figure carrying a stone from one pile to another pile
9	figure putting a stone back where it was, footprints reversed
10	two piles becoming one pile

Part B: The same actions with nothing to hold

The object is drawn faintly or absent. The action is drawn in full. If the response repeats the Part A word, the action is separable from its object. If nothing comes back, it is not.

Panel	Drawing
1	hand in the taking shape, pile absent
2	hand in the putting shape, stone absent
3	hands in the splitting shape, stone absent
4	hands in the pressing shape, halves absent
5	figure in the copying posture, both surfaces blank
6	figure in the rubbing posture, surface blank
7	figure in the comparing posture, both surfaces blank
8	figure in the carrying posture, hands empty
9	figure in the returning posture, hands empty
10	two blank shapes merging into one blank shape

Part C: Operations on a store that is not a place

Each panel shows a named operation as a diagram of boxes and arrows on the lit face. The speaker has already read line drawings as ground in Sheet 8 Part A and refused the frame in Sheet 8 Part B panel 10. What is wanted here is whether the Part B word survives when the object is a record of changes with no body.

Panel	Operation	Drawing
1	clone	one column of boxes, arrow to a second identical column
2	commit	a box added to the end of a column, marks inside it
3	branch	a column splitting into two columns at one box
4	merge	two columns joining back into one at one box
5	diff	two boxes side by side, one mark different, that mark ringed
6	revert	a column with the last box crossed out, arrow back to the box before it
7	checkout	a hand reaching into a column and pulling one box forward
8	push	a column on the left, arrow to a column on the right, the right one behind a line
9	pull	the same two columns, arrow reversed
10	conflict	two columns joining, one box in each with a different mark, both ringed, no join drawn

Panels 3 and 4 are the pair to watch. The speaker split a stone in Part A panel 3 and pressed the halves back in panel 4. If those two words return here and no others do, the speaker is reading the shape of the diagram and not the operation.

Nika

at Part A panel 3 Kesta. at panel 4 Kesta-ne. at Part B panel 3, with the stone gone from the drawing, hands coming up and making the split in the air over the screen Kesta. at Part C panel 3, tracing the fork with one finger Kesta. at panel 4 Kesta-ne. at panel 10, stopping on the two ringed marks, then looking up and pointing at his chest and then at her own Kesta. Ovlat, aha, kesta.

Vlad

pulling out an exercise about understanding the elements of the computer

Elicitation sheet 10: the parts of the machine

Four parts. The machine is on the ground between operator and speaker with the lit face turned so both can see it. Part A is the body. Part B is what changes on the lit face when the operator acts. Part C is the arrangement on the lit face. Part D is what sits under the arrangement. Parts C and D are expected to fail and the failures are recorded.

Part A: The body

Point at the part on the machine itself, then at the speaker's mouth. No drawings. The parts are all present and can be touched.

Item	Part
1	the lit face
2	the frame around the lit face
3	the panel of small squares
4	one small square
5	the hinge
6	the underside
7	the warm patch on the underside
8	the opening where air goes in
9	the sound coming from the opening
10	the whole object, closed

Part B: Cause on the lit face

Perform the action, then point at what changed, then at the speaker's mouth. Do the same action three times. If the same word comes back all three times, the speaker has joined the hand to the change.

Frame	Operator does	What changes
1	presses one small square	one mark appears at the edge of the lit face
2	presses and holds the same square	many of the same mark appear in a row
3	presses a second square	a different mark appears
4	drags a finger across the panel	a small shape crosses the lit face
5	drags the small shape onto a picture and presses	the picture opens and fills the face
6	presses at the corner	the picture closes and the face goes back
7	closes the hinge	the lit face goes dark
8	opens the hinge	the lit face comes back with the same marks on it

Frames 7 and 8 are the pair to watch. If the speaker names the darkness in frame 7 with the word given for the fire going out in Sheet 6, and then has no word for frame 8, the machine is being read as a fire.

Part C: The arrangement

Point at the region of the lit face, then at the speaker's mouth.

Item	Region
1	the strip along the bottom edge
2	one picture sitting in the strip
3	the same picture with a line under it
4	the picture without the line
5	the field behind everything
6	the animal drawn on the field
7	a bordered area sitting over the field
8	a second bordered area over the first, edges of the first still showing
9	the row of small marks at the top edge of a bordered area
10	the bordered area gone, the field showing where it was

Part D: What is under the arrangement

There is nothing to point at. Each item is shown as a diagram drawn by the operator on paper and held beside the lit face.

Item	Diagram
1	a box containing many smaller boxes, one of the smaller boxes ringed
2	the same box with the ringed box removed and a gap left
3	the same box with a new small box put into the gap
4	a column of boxes, each resting on the one below
5	the bottom box of the column, alone
6	the top box of the column, alone, with the lit face drawn beside it
7	many machines drawn, one box drawn once, arrows from the box to every machine
8	a figure at a desk drawing marks inside a box, arrow from the box to a machine

Item 7 is the point of the part. Nothing in the speaker's world is one thing held by many people at once and unchanged by the holding. A response here would be the single most informative result on the sheet.

Nika

at Part A item 7, laying the back of her hand on the warm patch and leaving it there Miht-me. at Part B frame 7, when the hinge comes down Ontu. at frame 8, when it comes back up, pulling her hand off the machine and sitting back Ne. Ne-e. at Part C item 6, finding the animal on the field and putting her finger on it Rusa-me? at Part D item 7, following the arrows out to each drawn machine and then back to the single box, three times over, then pushing the paper away with two fingers Ne. Ovlat, ne.

Vlad

pulling out an exercise about understanding changing elements

Elicitation sheet 11: substitution

Four parts. The speaker has named the parts of the machine in Sheet 10 Part A and the regions of the lit face in Part C. The question here is whether those names hold when the thing named is replaced by another that does the same work. Parts C and D use programs the research computer does not carry and are run on paper only.

Part A: One object for another

The object is present. Point at it, then at the speaker's mouth. Then remove it, put the replacement in its place, point at the replacement, and point at the speaker's mouth again. If the same word comes back for both, the word is on the work and not on the object.

Frame	First object	Replacement
1	the operator's shaft with a stone point	the speaker's shaft with an antler point
2	a stone used for striking	a second stone of another colour and shape
3	a cupped hand holding water	a hollowed piece of bark holding water
4	the fire in the cave	a burning brand carried in the hand
5	the machine, lit face open	a sheet of paper with the same marks drawn on it

Part B: The same face rearranged

One region of the lit face is changed while the speaker watches. Point at the region before, then at the region after, then at the speaker's mouth.

Frame	Before	After
1	strip along the bottom edge, pictures in a row	the same strip moved to the top edge
2	the field with the animal on it	the field plain, animal gone
3	the field plain	the field with a different animal on it
4	a bordered area at the left of the face	the same bordered area at the right
5	a bordered area covering a quarter of the face	the same area filling the whole face
6	the marks in one size	the same marks larger
7	the face in one set of colours	the same face in darker colours

Frames 2 and 3 are the pair to watch. The speaker named the animal on the field in Sheet 10 Part C item 6. If the word returns for a different animal in the same place, the word has gone onto the position.

Part C: A second arrangement, drawn only

The research computer does not carry this arrangement. Each item is drawn by the operator on paper beside the lit face, so the speaker sees the machine and the drawing together. Point at the region on the lit face, then at the region on the paper, then at the speaker's mouth.

Item	On the lit face	On the paper
1	strip along the bottom edge with pictures	a bar across the top with words and no pictures
2	pictures sitting on the field	a field with nothing on it
3	a bordered area with a row of marks at its top edge	a bordered area with a single wide band and no row
4	pictures reached from the strip	pictures reached by pressing a corner and a full face of pictures appearing
5	bordered areas overlapping	bordered areas side by side, each filling half the face

Part D: The same work done by another program

Drawn only. Each pair shows two arrangements that do the same work and share nothing in their shape. Point along the pair, then at the speaker's mouth.

Item	First	Second
1	marks typed into a plain field, no border	marks typed into a bordered area with a row of small marks above
2	a column of names with one ringed	a grid of pictures with one ringed
3	a bar filling from left to right	a ring closing from the top
4	a figure pressing a corner and a list dropping down	a figure typing marks and a list narrowing as the marks are added
5	two boxes side by side with an arrow between them	one box with the contents of the second box drawn inside it

Item 3 is the one to record carefully. Nothing in the speaker's world fills to show that a thing is partway done. If a word comes back for the bar and the same word comes back for the ring, the speaker has read the filling and not the shape.

Nika

at Part A frame 1, taking his shaft, weighing it, and putting her own down beside it Naga. Naga-tu. at Part B frame 3, when the new animal appears in the same place Ne. Ne, rusa-ne. Bontu-me. at Part C item 2, looking from the lit face to the paper and back Ontu-me? at Part D item 3, following the bar across and then the ring around, and doing it again Kesta-ne. pushing two fingers slowly together on the gravel until they meet Kesta-ne ta.

Definiteness, recorded sound, sameness#

Vlad

pulling out an exercise about understanding articles

Elicitation sheet 12: definiteness

Four parts. The speaker uses a bare stem for a citation form, established in Sheet 6 Part A. The question here is whether the speaker separates a thing already known to both parties from a thing named for the first time, and whether that separation is carried on the stem, on a word before it, or on nothing at all. Part D uses a running text and is run only after Parts A to C are complete.

Part A: Present and absent

Both parties can see the object in frames 1, 3, 5 and 7. In frames 2, 4, 6 and 8 the object has been removed and only the drawing is shown. Point, then at the speaker's mouth.

Frame	Condition
1	the speaker's own shaft, lying between the two parties
2	a drawing of a shaft, no shaft present
3	the fire in the cave
4	a drawing of a fire, fire out of sight
5	the stone the operator has been holding all afternoon
6	a drawing of a stone, no stone present
7	the cave mouth, visible from where both parties sit
8	a drawing of a cave mouth

Part B: First mention and second mention

Each strip has three panels. The same animal appears in panel 1 and again in panel 3, with an unrelated panel between them. Point along the strip once, then at the speaker's mouth. What is wanted is any difference between how the animal is named the first time and how it is named the third.

Strip	Panel 1	Panel 2	Panel 3
1	an antlered animal on a slope	a tree	the same antlered animal, closer
2	a fish in water	falling white specks	the same fish in a hand
3	a figure walking	a fire	the same figure lying down
4	a stone on the ground	a bird	the same stone in a hand

Part C: One of many and one already chosen

Frame	Left panel	Right panel
1	seven antlered animals, none marked	seven antlered animals, one marked, a shaft pointed at the marked one
2	seven stones in a pile	seven stones, one lifted out and held
3	seven figures standing	seven figures, one stepping forward
4	many cave openings along a rock face	the same rock face, one opening with smoke coming out

Part D: Running text

The operator reads the following aloud at normal speed, then reads it a second time and stops on each of the underlined words, pointing at the speaker's mouth. The speaker has no access to the meaning of the text and none is intended. What is wanted is whether the speaker marks the underlined words as recurring items and whether the speaker attempts them at all.

A human is a member of the species Homo sapiens, which means ‘wise man’ in Latin. Humans are mammals. They are also social animals. They usually live in groups.Source: "Human", Simple English Wikipedia, CC BY-SA 4.0.

The two underlined words carry no drawing and cannot be pointed at. If the speaker returns them as sound and drops them everywhere else, they are being copied and not understood. If the speaker returns nothing for them and returns the surrounding words, the speaker is hearing them as edges of the words beside them.

Nika

at Part A frame 2, with the shaft gone and only the drawing left Naga. at frame 1, putting two fingers on the shaft itself Naga ta. at Part B strip 1, following his finger across the three panels Rusa. Rusa ta. at Part C frame 1, moving her hand over all seven animals and then onto the marked one Rusa-ktha. Rusa ta-ke. at Part D, listening through the first reading without moving, and on the second reading saying nothing at either underlined word and coming in only on the last one Kruf.

Vlad

pulling out an exercise about understanding audio files

Elicitation sheet 13: recorded sound

Five parts. Sound comes out of the machine with nothing making it. The speaker has already named the opening in the underside and the sound coming from it in Sheet 10 Part A items 8 and 9, so the machine is a thing that makes noise. What is wanted here is whether a sound with no body behind it is heard as the thing itself, as a copy of the thing, or as the machine. Part D is the point of the sheet. Run the parts in order and do not skip to Part D.

Part A: Sound and its source together

The operator makes each sound in full view. Point at the hand making it, then at the speaker's mouth.

Item	Sound
1	two stones struck together
2	a hand slapped on the rock floor
3	water poured from a hollowed piece of bark
4	a shaft dropped on gravel
5	the operator whistling

Part B: The same sounds from the machine

The five sounds from Part A are played back with the hands of both parties visible and empty. Point at the machine, then at the speaker's mouth. If the speaker names the object from Part A, the sound is being heard as the thing. If the speaker names the machine, it is being heard as the machine.

Part C: Sounds with no source in the cave

Played one at a time. Nothing is shown. Point at the machine, then at the speaker's mouth.

Item	Recording
1	wind over open ground
2	rain on leaves
3	running water
4	a large animal breathing
5	wolves calling at distance
6	a bird, repeated call
7	a herd moving over dry ground
8	a low continuous tone, machine-made
9	a rising and falling tone, machine-made
10	silence, machine running, no recording

Item 5 is expected to move the speaker before it is named. Record the movement.

Part D: The speaker's own voice

The recording from Sheet 3 is played back. Order is fixed: first the operator's voice from Sheet 7 Part A, then a stranger's voice reading unrelated words, then the speaker's own voice from Sheet 3 strip 1. Point at the machine before each. Do not point at anyone.

Pass	Voice	Point at, after playback
1	the operator	the operator
2	a stranger	nothing
3	the speaker	nothing
4	the speaker, played a second time	the speaker

Part E: Sound moved off the thing

Each frame plays a sound recorded from one object while a different object is held up. What is wanted is whether the speaker corrects the pairing.

Frame	Sound played	Object held up
1	two stones struck	a shaft
2	water poured	two stones
3	the operator whistling	the machine
4	the speaker's own voice	the operator

Nika

at Part C item 5, on her feet before the call has finished, spear off her knees and the whole of her turned to the cave mouth Ktha! staying up, listening out at the slope for a long while, then looking back at the machine Ne. at Part D pass 3, going very still Ne-e. Ne-e! at pass 4, when his finger comes to her, backing off two steps and putting her free hand over her own throat Aha. Aha ta? Aha-ke ta?

Vlad

pulling out an exercise about understanding difference

Elicitation sheet 14: sameness and replacement

Five parts. Parts A to C use objects. Parts D and E use operations on the lit face that have no object behind them. The speaker refused the shared record in Sheet 10 Part D item 7, so nothing here assumes a record can exist away from a thing. Run the parts in order.

Part A: Two things or one thing

Both objects are in view. Point at the left, then the right, then at the speaker's mouth.

Frame	Left	Right
1	a stone	the same stone, moved a hand's width
2	a stone	a second stone of the same size and colour
3	a stone	a second stone, larger and of another colour
4	the operator's shaft	the speaker's shaft
5	a mark drawn on rock	the same mark drawn a second time beside it

Part B: The thing altered and the thing returned

Performed in view. Point at the object before, point at the object after, point at the speaker's mouth.

Frame	Before	Operator does	After
1	a whole stone	strikes it	two pieces
2	two pieces	presses them together, held	one shape, a line across it
3	a mark on rock	rubs it out	bare rock
4	bare rock	draws the same mark again	the mark on rock
5	a shaft with a point bound on	unbinds the point, binds on a second point	a shaft with a different point

Frame 5 is the one to record. The speaker named the work and not the object in Sheet 11 Part A frame 1. If the same word comes back here, a thing survives having its parts changed.

Part C: One kept and two kept

Frame	Drawing
1	a mark on one surface, arrow to the same surface with the mark altered, the first surface gone
2	a mark on one surface, arrow to a second surface carrying the altered mark, both surfaces standing
3	two surfaces standing, the first one crossed out, the second remaining
4	two surfaces standing, both bearing marks, neither crossed out

Part D: The four operations

Each is performed on the lit face while the speaker watches. Marks are typed, the operation is run, and the result is shown by opening the store of names and pointing along it. Point at the store before and after each operation.

Operation	Path	Before	After
1	file, save	one name in the store, marks altered on the face	one name in the store, the altered marks under it
2	file, save as, new name given	one name in the store	two names in the store, the first with the old marks, the second with the new
3	file, save as, the existing name given in the dialog	two names in the store	the dialog asks, the operator agrees, two names remain, the marks under the chosen name replaced
4	file, overwrite	two names in the store	two names remain, the marks under one of them replaced without a question

Operations 1 and 4 leave the store looking the same and differ only in whether a question was asked. Operations 2 and 3 begin identically at the dialog and end in opposite states. If the speaker separates 2 from 3, the speaker is reading the name in the dialog and not the action.

Part E: The question in the dialog

The dialog from operation 3 is opened and left standing. Point at it, then at the speaker's mouth. Then the operator agrees and the dialog closes. Point at the store, then at the speaker's mouth. Then the operation is run a second time and the operator refuses. Point at the store, then at the speaker's mouth.

The speaker returned an interrogative contour in Sheet 5 Part A pass 2. If that contour comes back for the standing dialog, the speaker has read a question drawn on a surface.

Nika

staying where she is, out of reach of the machine, and pointing at Part A frame 2 from there Kesta-ne. at frame 3 Kesta. at Part B frame 3, when the mark is rubbed off Ontu. at frame 4, when it comes back Ne. Kesta-ne ta. at Part D operation 1, watching the store before and after and finding the same name in both Ta. Ta-ke. at operation 2, following his finger down to the second name Rusa-ktha. Ne-e, rusa-ktha ne. at Part E, when the dialog opens and holds Aha na ta?

Interest, time and new words#

Vlad

pulling out an exercise about understanding remote context

Elicitation sheet 15: interest and claim

Five parts. Every panel in this sheet makes an assertion about a place or a thing that the speaker cannot see. What is wanted is whether the speaker separates an assertion made by a party who gains from it being believed from an assertion made by a party who gains nothing. Nothing on this sheet is true or false in itself and the operator does not know the answers. What is recorded is the speaker's treatment of the source. Run the parts in order and do not signal agreement at any point.

Part A: Report and evidence

Two ways of learning the same thing. Point along the strip, then at the speaker's mouth.

Strip	Panels, left to right
1	a figure walking to a cave, entering, coming out, a second figure at a fire
2	a figure at a fire, a second figure arriving, the second figure pointing at a drawn cave
3	a figure holding a carcass, a second figure looking at the carcass
4	a figure with empty hands, a second figure listening, a carcass drawn faintly above the first figure's head

Part B: The party who gains

Each strip shows an assertion and, in the same strip, what the asserting figure receives if the assertion is believed. Point along the whole strip once, then at the speaker's mouth.

Strip	Panels, left to right
1	a figure standing at a cave mouth, the same figure pointing into the cave, arms wide, a second figure walking in, the first figure taking the second figure's shaft
2	a figure holding stones, the figure pointing at the stones with arms wide, a second figure giving meat, the first figure holding the meat
3	a figure pointing along a valley, arms wide, a second figure walking off that way, the first figure walking the other way to a carcass
4	a figure at a cave mouth, arms wide, a second figure walking in, the first figure walking away with nothing
5	a figure pointing along a valley, a second figure walking off that way, the first figure walking the same way behind them

Strips 4 and 5 are the controls. The assertion is identical to strips 1 and 3 and the asserting figure receives nothing. If the speaker separates 1 from 4 and 3 from 5, interest is being read off the panel and not the gesture.

Part C: The claim against what is visible

The claim is drawn on the lit face. The state of the same place is drawn on paper beside it. Point at the lit face, then at the paper, then at the speaker's mouth.

Frame	On the lit face	On paper
1	a cave mouth, a figure at it with arms wide, seven figures walking in	the same cave mouth with clawed tracks at the entrance
2	a valley with many antlered animals drawn large	the same valley with one antlered animal and bare ground
3	water in a hollow, a figure drinking, arms wide	the same hollow dry, cracked ground
4	a fire burning, a figure warming its hands, arms wide	wet sticks, smoke, no flame

Part D: Repetition

The same claim from Part C frame 1 is shown four times in succession, drawn by four different figures, each with arms wide, with nothing else changed and no new evidence added. Point at each in turn, then at the speaker's mouth after the first and after the fourth.

Part E: Who is speaking

The same claim, the same drawing, four sources. Point at the source panel, then at the claim, then at the speaker's mouth.

Frame	Source panel
1	a figure the speaker has already named, standing beside the claim
2	a figure with its back turned, standing beside the claim
3	the claim alone, no figure
4	seven figures standing beside the claim, all with arms wide

Frame 3 is the one to record. A claim with no figure attached to it has no source in the speaker's world, and the operator's own claims arrive that way.

Nika

at Part B strip 1, going along it panel by panel and stopping on the last one, where the shaft changes hands Ne. Ne-e! tapping the wide arms and then the taken shaft, and doing it a second time Ne-e ta. at strip 4, going along it and then turning her hand over Ha. at Part C frame 1, moving her finger off the lit face and onto the tracks on the paper and leaving it there Torna ta. Torna ta, ne-e. at Part D, after the fourth figure Ktha ne. Ktha ne-e, semek ne. at Part E frame 3, looking at the claim with nothing beside it, then lifting her eyes to him Aha na? Ovlat na?

Vlad

pulling out an exercise about time

Elicitation sheet 16: time and the marked object

Six parts. Parts A to D establish how far back and how far forward the speaker will go and what carries the distance. Part E puts the same event at each of those distances and checks whether the marking on the thing acted upon survives. Part F is the far edge and is expected to fail. Run in order.

Part A: Inside one turn of light

Each panel shows the sun at a position and a figure doing something. The operator points at the real sun through the cave mouth first, then at the drawn sun, then at the panel, then at the speaker's mouth.

Panel	Sun	Figure
1	below a horizon line, sky pale	figure standing, fire out
2	low at one edge	figure walking away from a cave
3	high	figure with a shaft, an antlered animal ahead
4	low at the other edge	figure walking toward a cave, carrying a load
5	below the horizon line, sky dark, moon drawn	figure lying down, eyes closed

Part B: Counting turns

A row of suns is drawn beside the panel. The event in the panel is the same each time. Point along the row of suns, then at the panel, then at the speaker's mouth. Suns drawn to the left of the panel are behind, suns to the right are ahead.

Frame	Suns to the left	Suns to the right
1	one	none
2	two	none
3	seven	none
4	none	one
5	none	two
6	none	seven
7	thirty	none
8	none	thirty

Frames 7 and 8 are drawn as a long unbroken band of suns with no gaps. If the speaker returns the same word as for frame 3, the count has been abandoned and the band is being read as a quantity.

Part C: The moon and the cold

Frame	Drawing
1	a full moon, then a half, then a thin curve, then nothing, then a thin curve again
2	the same sequence drawn once through, with an event panel at one end
3	the same sequence drawn seven times through
4	falling white specks, then bare ground, then green, then falling white specks again
5	the sequence of frame 4 drawn twice through
6	the sequence of frame 4 drawn seven times through
7	the sequence of frame 4 drawn as a long band, thirty times through

Part D: People as distance

Frame	Drawing
1	a short figure beside a tall figure
2	a short figure, a tall figure, an older figure behind them
3	the same three with a fourth figure behind, lying down, eyes open
4	the same, with a fifth figure behind, lying down, eyes open, drawn faint
5	the same, with a sixth and seventh behind, drawn fainter
6	a row of twenty figures, each fainter than the last, the far end blank

Part E: The marked object across the distances

One event: a figure strikes an antlered animal with a shaft. It is shown at each distance established above. The speaker gave a form for the animal acted upon in Sheet 6 Part A. What is wanted is whether that form holds, whether the change lands on the animal or on the striking, and whether a separate word arrives with the distance.

Frame	Distance drawn
1	sun high, event happening now
2	one sun to the left
3	seven suns to the left
4	the moon sequence once through, to the left
5	the cold sequence twice through, to the left
6	the older figure of Part D frame 2, event drawn above them
7	the faint figures of Part D frame 5, event drawn above them
8	one sun to the right
9	seven suns to the right
10	the cold sequence twice through, to the right
11	the short figure of Part D frame 1 drawn tall, event drawn above them
12	figures not yet drawn, blank space, event drawn above the blank

Part F: The far edge

No figures, no suns, no seasons. Each panel is a band of the marked unit running off both sides of the panel with the event drawn small in the middle.

Frame	Band
1	suns, running off both edges
2	cold sequences, running off both edges
3	faint figures, running off both edges
4	the panel blank, one small event drawn in the middle

Frame 4 carries no distance at all. If it returns the same word as frame 3, the speaker has no separation between very far and nowhere.

Nika

at Part B frame 3, laying seven fingers on the row and taking them off one at a time Miht-ktha. Semek na aha rusa-tu, miht-ktha ontu. at frame 7, running her hand the whole length of the band without stopping Ktha. Ktha-ke. at Part C frame 6, going through the six cold sequences slowly and then sitting back Karnu-ktha. at Part D frame 3, putting her finger on the figure lying down with its eyes open and holding it there Aha-ke. at frame 5, on the two fainter ones behind Aha-ke ta. Ne-e. Ne-e semek. at Part E frame 5, back on the deer with the shaft in it Rusa-tu ontu. at frame 10, with the same event drawn ahead Rusa. Rusa ne-e tu. at Part F frame 4, with nothing drawn at all, taking her hand off the paper and putting it on the gravel Ne.

Vlad

pulling out an exercise about productive forms

Elicitation sheet 17: building new words

Six parts. The speaker has returned four endings so far, recorded in Sheets 5 and 6. What is wanted here is whether those endings are habits fixed to the words they were found on, or a procedure the speaker will run on a word never heard before. Parts C and D are said aloud by the operator and carry no drawings, since the point of them is the shape of the word and not its meaning. Run in order.

Part A: The endings on known stems

Point at the drawing, then at the speaker's mouth. Every stem here has already been recorded.

Frame	Drawing
1	one antlered animal, then seven
2	one antlered animal, then a small one beside a large one
3	one antlered animal standing, then lying down with eyes open
4	a shaft alone, then a shaft driven into an animal
5	one humped animal, then seven
6	one maned animal, then a small one beside a large one

Part B: An ending run on a stem it was never heard on

The stems here were recorded in Sheet 1 and the endings in Sheets 5 and 6, and the two have never been put together in the speaker's hearing. If the speaker produces the combination without hesitation, the ending is a procedure.

Frame	Stem recorded as	Drawing shown
1	fish	seven fish
2	tree	a small tree beside a large one
3	bird	a bird lying down, eyes open
4	stone	a stone used to strike
5	cave opening	seven cave openings
6	water in a hollow	a small hollow beside a large one

Part C: A place made from a thing

Said aloud by the operator, twice each, at normal speed. No drawings for the first pass. On the second pass the drawings are shown while the words are said.

Item	Operator says	Drawing on second pass
1	sidlo	one figure lying down inside a shelter
2	sidliste	many shelters together, many figures
3	hra	two figures throwing a stone between them
4	hriste	flat open ground, many figures throwing
5	park	one box on wheels, standing still
6	parkoviste	flat ground, many boxes on wheels standing in rows

The added part is the same in items 2, 4 and 6 and the front part changes. The operator points at the front of each word by holding up one finger, and at the added part by holding up two, so the join is marked without being named.

Part D: The same place said in four ways

Said aloud, one after another, with the drawing from Part C item 2 held up throughout.

Language	Operator says	Shape
czech	sidliste	one stem, one added part
polish	osiedle	one stem, one part before and one after
romanian	cartier	one word, nothing added, taken whole from elsewhere
english	housing estate	two whole words standing side by side

These four are the four available shapes. Whichever the speaker produces in Part F places the language.

Part E: Number left unmarked

The operator says the czech word for one such place and then for many such places. The two are the same sound. The drawings differ: one shelter group, then seven shelter groups. Point at each drawing, then at the speaker's mouth. What is wanted is whether the speaker accepts a form that does not change, having marked number with an ending in Part A.

Part F: A thing with no word

Two roots the speaker already holds, put together for a thing the speaker has never seen. The drawings are shown together and pointed at in order, then at the speaker's mouth.

Frame	First drawing	Second drawing	Together
1	a maned four-legged animal, teeth showing	a single tooth, large, drawn alone	an animal with a long jaw and rows of teeth, drawn from a skull
2	a stone	a place, flat ground with many figures	flat ground covered in worked stone, many figures knapping
3	a fire	a place	a hollow in rock, ash, no flame
4	water	a place	a hollow with water and many figures around it
5	the machine	a place	nothing drawn, blank panel

Frame 5 has no second drawing to support it and no referent anywhere. If the speaker produces a word for it anyway, the procedure runs on anything and the language will take the operator's vocabulary. If the speaker returns nothing, coining requires a thing.

Nika

at Part B frame 5, without waiting for the second panel to be pointed at Kruf-ktha. at Part F frame 3, going from the fire to the open ground and then to the ash Miht-nu. saying it again and then reaching over to touch the cold edge of his own fire pit Miht-nu ta. at frame 4 Sula-nu. at frame 2 Kesta-nu. stopping there, and then turning her head to the cave mouth and the ridge beyond it and pointing at it with the whole hand Kar-nu. Karnu! laughing Kar-nu ta! at frame 5, with the blank panel under his finger, the hand comes down and stays down Ne.

Composition and the last sheets#

Vlad

pulling out an exercise about understanding composition

Elicitation sheet 18: what a thing is made of

Five parts. The speaker builds and unbuilds objects daily and returned a word for the work of splitting and one for the work of joining in Sheet 9. What is wanted here is whether those words reach a thing far larger than a body, whether a whole made in one place and carried whole to another is read as made at all, and whether a name stays on a place after everything standing on it has gone. Parts C to E are drawn only. Run in order.

Part A: Parts of things already known

The object is present and is taken apart in view, part by part, and left in a row on the ground. Point at each part, then at the row, then at the whole object rebuilt, then at the speaker's mouth.

Item	Object	Parts laid out
1	a shaft with a point	shaft, point, binding
2	a fire	stones in a ring, dry sticks, thick wood, ash beneath
3	a cord of sinew	separate strands, twisted length
4	a hide garment	separate pieces, the sinew stitching, the whole

Part B: Made in place, course by course

Drawn as a sequence read left to right, one panel per stage. The operator points along the whole sequence once at speed, then goes back and points at each stage.

Stage	Drawing
1	a strip of ground, flat, nothing on it
2	a trench dug along the strip, stones laid in it
3	one row of small rectangles laid end to end on the stones
4	four rows, each on the one below, joins offset
5	twenty rows, gaps left where openings will be
6	timbers laid across the top of the walls
7	a sloped cover over the timbers
8	figures inside the openings, smoke from an opening at the top

A single small rectangle is drawn alone beside the sequence throughout, so the unit is available at every stage.

Part C: Made elsewhere, carried whole

Stage	Drawing
1	flat ground with a hollow dug in it
2	a distant place, figures making one large flat slab, a wall's worth, openings already in it
3	the slab on a long box on wheels, moving
4	an arm above the ground lowering the slab into place
5	a second slab lowered beside the first
6	six slabs standing, a slab laid flat across their tops
7	the same six-sided box repeated eight times in a stack
8	figures at the openings

Stage 2 is the stage to watch. The wall exists before the place it belongs to exists. Nothing in the speaker's world is finished before it is anywhere.

Part D: Taken down and put up again

Stage	Drawing
1	a low building of rough stone, timber above, a figure at the door
2	the same building, figures carrying things out
3	the same building with the cover off and the timbers gone
4	walls half down, a heap of rubble
5	flat ground, rubble carted off, one line of old foundation still showing
6	a new trench cut across the old line
7	cut stone laid in courses, taller than the first building, five levels of openings
8	the finished front, a band of ironwork across the fourth level, a sloped grey cover with openings in it
9	figures at the openings, none of them the figure from stage 1

Part E: The name after the parts have gone

Each frame pairs a first state with a last state. Point at the first, then at the last, then at the speaker's mouth. What is wanted is whether one word covers both.

Frame	First	Last
1	the shaft of Part A item 1	the same shaft, point replaced, binding replaced
2	the fire of Part A item 2	the same ring of stones, all wood burnt, new wood laid
3	the building of Part D stage 1	the building of Part D stage 8, same ground
4	the ground of Part D stage 5, bare	the same ground, nothing built, drawn again
5	the six-sided box of Part C stage 6	the same box, lifted off the ground and set down elsewhere

Nika

at Part A item 2, taking apart his fire pit with her hand while she names it and putting each piece down in the row he has made Sanu. Kruf-me. Miht. Miht-nu. at Part B stage 4, on the four rows of rectangles Kesta-ne. Kesta-ne ktha. at stage 5, running her finger up the twenty rows and then flicking it back down to the single rectangle drawn alongside Kesta-ne, kesta-ne, kesta-ne ta. Ktha-ke! at Part C stage 2, going back to it after seeing stage 4, and then back to stage 4 again, three times Ne. Ne-e. Kruf ne-nu? at Part D stage 5, with the ground bare and the old line still showing Ontu. at stage 8, with the new front standing on the same line Ne ontu. Kesta-ne ta-ke. at Part E frame 4, the same bare ground drawn twice with nothing on it, putting her palm on it and leaving it Karnu. Karnu ta, semek-ne.

Vlad

pulling out an exercise about default words

Elicitation sheet 19: the high-frequency words

Seven parts. Six items have come back more often than everything else recorded and each has come back in contexts that do not sit together. The translator cannot peg any of them until the senses are separated. Every part below works the same way: one item is held constant and everything around it is varied, and what is recorded is where the item stops being offered. Do not use the item in a question. Run in order.

Part A: ktha

Recorded at the cave mouth on first sight of the operator, on a single raised finger, on a row of seven suns, on seven antlered animals as an ending, on a wall of twenty courses, and on hearing wolves. Point at each panel and wait. Where the item is offered, mark it. Where it is withheld, mark that.

Panel	Drawing
1	one stone, alone
2	seven stones
3	a stone too large to lift, figure beside it for scale
4	a clawed animal at close range, figure leaning back
5	a clawed animal far off, figure standing easy
6	a stranger at a cave mouth, figures inside standing
7	a known figure at a cave mouth, figures inside sitting
8	a row of suns running off the panel edge
9	one sun
10	a fire much larger than the ring of stones holds

Part B: semek

Recorded on the operator, on the operator's marks, on a strip of connected panels, on the sun, and on bare ground with nothing built on it. Point and wait.

Panel	Drawing
1	the operator, pointed at directly
2	a drawn figure with no face
3	a drawn figure with the speaker's own ornaments
4	marks on a surface, no figure
5	a figure making marks on a surface
6	a figure speaking, mouth open, a second figure listening
7	a figure speaking, no second figure
8	two figures, hands joined
9	a figure alone on open ground

Part C: aha

Recorded on the speaker's own chest, on a figure sitting alone with others far off, on figures lying down with eyes open, and in a question about the operator. Point and wait.

Panel	Drawing
1	the speaker, pointed at directly
2	a single figure, standing
3	seven figures, standing together
4	a figure sitting, others small and far off
5	a figure lying down, eyes open
6	seven figures lying down, eyes open
7	a figure lying down, eyes open, others standing over it
8	a tall figure and a short figure, hands joined
9	the operator and the speaker drawn side by side

Part D: ne and ne-e

The long form has come back where the short one would fit and the difference has not been placed. Each pair below shows a wrong thing and a dangerous thing, matched for everything else.

Frame	Left	Right
1	a figure eating a fish tail first	a figure eating a fish that is drawn swollen and marked
2	a shaft bound crookedly	a shaft with the binding come loose in the hand
3	a fire laid with green wood, smoke	a fire spreading past the ring of stones
4	a figure walking the long way to water	a figure walking toward clawed tracks
5	a figure carrying a load badly	a figure walking onto a drawn edge

Part E: ta

Frame	Left panel	Right panel
1	an antlered animal, first drawing	the same animal, drawn again
2	an antlered animal	a different antlered animal
3	a stone in the operator's hand	the same stone on the ground
4	a stone on the ground	a stone drawn far off
5	the cave, drawn from inside	a different cave, drawn from inside

Part F: na

Recorded between two things, and at the end of a rising line. The two uses may be one item or two. Each frame is a whole strip and the item's position in the strip is what is recorded.

Strip	Panels
1	figure, shaft, antlered animal
2	figure, shaft, antlered animal, palm up beside the last panel
3	antlered animal, shaft, figure
4	figure, figure
5	figure, figure, palm up

Part G: Substitution

The operator says back a whole utterance the speaker produced earlier, with one of the six items replaced by another of the six. Nothing is pointed at. What is wanted is whether the speaker corrects the operator and where the correction lands.

Frame	Operator says
1	semek na aha rusa
2	semek na ktha rusa
3	aha na semek rusa
4	ktha ne torna ta
5	ne-e torna ta
6	aha-ke ta
7	ktha-ke ta

Nika

at Part A panel 4, with the clawed animal close Ktha! at panel 2, on the seven stones, and the sound is longer and comes down at the end Ktha. doing the two panels again in turn and watching his face while she does it Ktha. Ktha. at Part G frame 2, when he says it back to her, sitting forward Ne. Ne, semek na aha. saying his version and then her own, twice each, and getting nothing from him Ne-e. reaching over and putting two fingers on his throat, then saying the first one, then the second, and keeping her fingers where they are Ktha. Ktha.

Vlad

pulling out the final exercise, about whether the person is multilingual

Elicitation sheet 20: one speech or more than one

Seven parts. Everything recorded so far has been treated as a single system. It is possible that it is two, held apart by who is being addressed, and that the operator has been given one of them throughout. It is also possible that the speaker holds a second speech kept for people outside the group, which is what would matter most here. Parts A to C test recognition, Parts D and E test switching, Part F tests whether a second system exists in the hands rather than the mouth, and Part G is the control. Run in order and stop after Part G. This is the last sheet.

Part A: Speech and other sound

Played from the machine, one at a time, hands empty, nothing shown. Point at the machine, then at the speaker's mouth. What is wanted is a division between what is speech and what is noise.

Item	Recording
1	the speaker's own voice, from Sheet 3
2	the operator's voice, romanian
3	a woman's voice, hungarian
4	a man's voice, basque
5	a woman's voice, mandarin
6	a man's voice, xhosa, clicks present
7	wind over open ground
8	wolves calling
9	a bird, repeated call
10	a voice reading a list of numbers, no sentence contour

Item 6 is the one to record. If the clicks are put with the animals and the rest of the voices are put together, the sorting is on the sound and not on the speech.

Part B: One voice, two systems

A single speaker recorded twice, saying different things in two unrelated languages. The voice, the pace and the room are identical across the pair. Play the pair, then point at the speaker's mouth.

Pair	First	Second
1	a woman, romanian	the same woman, english
2	a man, czech	the same man, polish
3	a woman, english	the same woman, english, different words

Pair 3 is the control. If the speaker separates pairs 1 and 2 from pair 3, two systems in one mouth is a thing the speaker can hear.

Part C: Two speakers, one system

Pair	First	Second
1	a woman, romanian	a man, romanian
2	a woman, romanian	a man, hungarian
3	a child, romanian	a man, romanian

Part D: The addressee changed

No recordings. Drawings only. Each panel shows the speaker's own figure, drawn from the ornaments the speaker wears, addressing a different party. Point at the panel, then at the speaker's mouth, and record whatever comes back at length.

Panel	Addressee
1	a short figure, close, hands joined
2	an older figure, seated
3	seven figures standing in a group
4	a figure at distance across open ground, hand raised
5	a figure with unfamiliar ornaments, standing apart, shaft held
6	an antlered animal
7	a clawed animal
8	nobody, empty ground, a fire

Panel 5 is the point of the part. A speech kept for outsiders would surface here and nowhere else.

Part E: The same content, both addressees

Each frame is one strip, run twice. The first run has the panel from Part D item 1 at the head of the strip. The second run has the panel from Part D item 5 at the head. Everything after the head is identical.

Frame	Strip after the head
1	figure, shaft, antlered animal, animal down
2	water in a hollow, figure drinking
3	clawed animal, cave opening, figure walking away
4	falling white specks, no animals, figure at an open mouth

Part F: A system in the hands

No sound is wanted. The operator holds a finger to his own mouth, then points at the strip, then at the speaker's hands.

Strip	Panels
1	figures crouched, an antlered animal ahead, one figure with a raised hand
2	figures spread wide, an animal between them
3	a figure pointing along a ridge, others watching the hand
4	a figure holding up a closed fist, others stopping

Part G: Control

Item 1 of Part A is played a final time and nothing else is done. If what comes back matches what came back on the first pass, the sheet is clean and the record can be closed.

Nika

at Part A, after all six voices, moving her hand along the whole row of them and then laying it down Semek. Semek-ktha. at item 6, when the clicks come, going back and putting her finger on it a second time and then leaving it with the rest Semek. at Part B pair 1, listening to the woman twice over and then turning her palm up Ne. doing the same at pair 2 and pair 3, the same hand each time, and saying nothing at Part D panel 5, with the stranger and the shaft, sitting up straighter and giving him a run of it, the same sounds she has been giving him all afternoon and no others Aha na, aha ta karnu, semek ne-e naga, aha na ne torna. at Part E, running both heads and stopping partway through the second Ktha. Ktha, ne kesta. at Part F, going quiet, closing her fist and holding it up, then opening two fingers and moving them apart low over the gravel at Part G, when her own voice comes back out of the machine one more time Aha. Ovlat, aha.

The build#

Vlad

pressing the finish button

Nika

watching the lines run down the lit face, looking at him Ontu?

elicit-build 0.9.4, session 0007

wrap

elicit-build 0.9.4  (field kit, offline)
python 3.11.9 / torch 2.3.1+cpu / 4 threads / no accelerator
project: MANDRIN-K  session 0007  operator: unset
input: /var/elicit/session-0007/  20 sheets, 41 audio, 0 external corpora
target: contextual translator, bidirectional[00:00:00]        INFO  build started
[00:00:00]        INFO  orthography locked: latin-as-given, case-insensitive, nfc
[00:00:00]        INFO  peg order: ro > en > cs > pl
[00:00:00]        WARN  peg ro selected as primary; no ro speaker in session audio
[00:00:01]        INFO  loading sheet responses ......................... 20/20 ok
[00:00:01]        INFO  loading audio ................................... 41/41 ok
[00:00:01]        INFO  total speaker audio 04:11:38, speaker turns 388
=== stage 1/11  acoustics ==================================================
[00:00:03]        INFO  sheet 1, sheet 3: segmenting continuous speech
[00:00:31]        INFO  phone inventory converged: 14 consonant, 5 vowel
[00:00:31]        INFO  no click, no tone, no ejective
[00:00:32]        INFO  length contrast found on vowels (sheet 5, sheet 19/A)
[00:00:32]        WARN  <ne> and <ne-e> are one segment plus length, not two words
[00:00:32]        INFO  merging lexemes ne, ne-e -> ne [+long]
[00:00:33]        INFO  sheet 7 borrow test: 6/6 native strings returned intact
[00:00:33]        INFO  sheet 7 borrow test: 5/6 foreign strings reduced to inventory
[00:00:33]        INFO  string 12 length 61 held, break preserved, final fall preserved
[00:00:33]        INFO  speaker can carry foreign prosody without foreign phones
=== stage 2/11  lexicon ====================================================
[00:00:35]        INFO  type count 43, token count 1,204
[00:00:36]        INFO  citation forms from sheet 6/A pass 1 ........... 12 anchored
[00:00:36]        INFO  sheet 1 nouns ................................... 20 anchored
[00:00:37]        WARN  <ktha> occurs as free form and as bound form
[00:00:37]        INFO  sheet 19/A: contour separates them, rising vs level+long
[00:00:37]        INFO  splitting: ktha1 [free, alarm/large/beyond-count]
[00:00:37]        INFO             ktha2 [bound, plural]
[00:00:38]        INFO  sheet 19/G frame 2 rejected by speaker; split confirmed
[00:00:38]        INFO  lexicon 44 types after split
=== stage 3/11  morphology =================================================
[00:00:41]        INFO  running sheet 17 productivity test as validator
[00:00:44]        INFO  suffixes induced from sheets 5, 6, 17:
[00:00:44]        INFO    -ktha  plural
[00:00:44]        INFO    -me    diminutive / lesser of a pair
[00:00:44]        INFO    -ke    terminated, no longer living or running
[00:00:44]        INFO    -tu    instrument, thing acted through
[00:00:44]        INFO    -nu    place characterised by the stem
[00:00:45]        INFO  sheet 17/B: 6/6 suffixes applied to unheard stems, no latency
[00:00:45]        INFO  classifying: agglutinative, suffixing, productive
[00:00:45]        INFO  sheet 17/D shape match: cs
[00:00:45]        WARN  peg primary is ro; morphological match is cs
[00:00:45]        INFO  peg order unchanged, decoder will route morphology via cs
[00:00:46]        INFO  sheet 17/F: coining requires a referent, 4/4 with, 1/1 without failed
[00:00:46]        INFO  coinage gate set: refuse to invent stems for absent referents
[00:00:46]        INFO  <karnu> re-derived by speaker as kar+nu, unprompted
[00:00:46]        INFO  root <kar> recovered, gloss ~ rock mass, high ground
=== stage 4/11  case and voice =============================================
[00:00:49]        INFO  sheet 6/A three passes over 12 nouns
[00:00:50]        INFO  bare stem = agent and citation, no marking
[00:00:50]        INFO  patient marked, allomorphs -tu / -t after vowel
[00:00:50]        WARN  -tu collides with instrument suffix, stage 3
[00:00:51]        INFO  disambiguating by position, 40 held-out strips
[00:00:51]        INFO  agreement 39/40, collision resolved by slot not form
[00:00:51]        INFO  no ergative pattern found, nominative-accusative
=== stage 5/11  polarity, mood, definiteness ===============================
[00:00:54]        INFO  sheet 5/A: four contours recovered, all four distinct
[00:00:54]        INFO  affirmative level, interrogative rising, exclamative high+short
[00:00:54]        INFO  negative carried segmentally by ne, contour flat
[00:00:55]        INFO  language marks polarity on both channels
[00:00:55]        INFO  sheet 19/D: ne [short] = wrong, ne [long] = dangerous
[00:00:55]        INFO  this is a lexical distinction, not emphasis
[00:00:55]        WARN  earlier session transcripts mis-glossed 22 tokens
[00:00:55]        INFO  retagging 22 tokens .......................... done
[00:00:56]        INFO  sheet 12: no article, no definite affix
[00:00:56]        INFO  definiteness carried by post-nominal <ta>
[00:00:56]        INFO  sheet 12/A: ta present only when referent is present or prior
[00:00:56]        INFO  sheet 12/D: speaker produced nothing for a, the
[00:00:56]        INFO  decoder rule: ro enclitic article maps to ta, drop when new
[00:00:57]        INFO  1 token unmatched in sheet 12/D: <kruf>
[00:00:57]        INFO  <kruf> traced to sheet 1 panel 12, gloss = stone
[00:00:57]        INFO  speaker was naming the surface, not answering
=== stage 6/11  deixis and time ============================================
[00:01:00]        INFO  sheet 16 loaded
[00:01:01]        INFO  temporal units: miht (sun/day), moon cycle, cold cycle, generation
[00:01:01]        INFO  counted range backward: 7 days, 12 moons, 6 cold cycles
[00:01:01]        INFO  counted range backward, persons: 3 named, 2 faint, 0 beyond
[00:01:01]        INFO  counted range forward: 7 days, 2 cold cycles
[00:01:01]        WARN  sheet 16/B frames 7,8 returned ktha1, not a count
[00:01:01]        INFO  above 30 units the system is mass, not number
[00:01:02]        WARN  sheet 16/F frame 4 returned ne
[00:01:02]        INFO  unbounded time is not distinguished from nonexistence
[00:01:02]        INFO  decoder guard: refuse to render source dates beyond 6 cold cycles
[00:01:02]        INFO  guard applies to 100% of dated content in the operator corpus
=== stage 7/11  referent grounding =========================================
[00:01:05]        INFO  sheets 8, 10, 11, 13, 14, 18 loaded as grounding evidence
[00:01:06]        INFO  sheet 8/A: side elevation read as terrain, 9/10
[00:01:06]        FAIL  sheet 8/A panel 4: overhead view, no response
[00:01:06]        FAIL  sheet 8/B: 10/10 no response for unsupported referents
[00:01:06]        INFO  sheet 8/B panel 10: content read, frame discarded
[00:01:07]        INFO  registering 214 operator-side concepts as UNGROUNDED
[00:01:07]        INFO  sheet 10/A: machine parts named as body parts and heat
[00:01:07]        INFO  sheet 10/B frame 7: hinge closed glossed <ontu>, as a fire
[00:01:07]        FAIL  sheet 10/B frame 8: no word for the return
[00:01:07]        INFO  machine modelled by speaker as a fire that can be relit
[00:01:08]        FAIL  sheet 10/D item 7: one thing held by many, refused
[00:01:08]        INFO  no type/token distinction available for artefacts
[00:01:08]        INFO  disabling decoder output class: copies, files, editions, records
[00:01:09]        INFO  sheet 11/B frames 2,3: name moved onto screen position
[00:01:09]        INFO  positional reference is live, decoder may use it for deixis
[00:01:09]        INFO  sheet 14/D: operations 1 and 4 not separated
[00:01:09]        INFO  sheet 14/D: operations 2 and 3 separated by the name given
[00:01:09]        INFO  speaker tracks names in a store, not states of a store
[00:01:10]        INFO  sheet 18/B: course-built construction accepted, kesta-ne x n
[00:01:10]        FAIL  sheet 18/C stage 2: object finished before its place, refused
[00:01:10]        INFO  sheet 18/D: identity survives full replacement on same ground
[00:01:10]        INFO  place outranks material in identity, weight 0.81
[00:01:11]        INFO  sheet 9/C: git operations mapped to split/join shape only
[00:01:11]        INFO  diagram shape is read, operation is not
[00:01:11]        INFO  sheet 9/C panel 10: conflict applied to operator and speaker
[00:01:11]        INFO  retaining as valid figurative extension, first observed
=== stage 8/11  source and reliability =====================================
[00:01:14]        INFO  sheet 15 loaded
[00:01:14]        INFO  strips 1,3 marked ne [long] by speaker
[00:01:14]        INFO  control strips 4,5 marked ha
[00:01:15]        INFO  interested assertion is separated from disinterested, 4/4
[00:01:15]        INFO  sheet 15/C: visible evidence overrides claim, 4/4
[00:01:15]        INFO  sheet 15/D: repetition without evidence lowers acceptance
[00:01:15]        INFO  sheet 15/E frame 3: unattributed claim triggers source query
[00:01:15]        INFO  building attribution layer
[00:01:16]        INFO  decoder rule: every rendered assertion carries its source
[00:01:16]        INFO  decoder rule: unattributed source-text renders with source query
[00:01:16]        INFO  this rule fires on 61% of operator corpus (encyclopedic text)
=== stage 9/11  affect =====================================================
[00:01:19]        INFO  sheet 4/A: 8/10 states lexicalised
[00:01:19]        INFO  hesu positive, hesu+ne negative, kamla grief, torna-hesu anger
[00:01:19]        INFO  panel 5 isolation glossed <aha-ke>, reusing the death form
[00:01:19]        WARN  aha-ke has two senses: dead person, and person cut off from group
[00:01:20]        INFO  keeping both, disambiguation by animacy of surrounding context
[00:01:20]        INFO  sheet 13/D pass 4: recorded self-voice triggered aha-ke query
[00:01:20]        INFO  speaker asked whether the recorded self was the second sense
[00:01:20]        INFO  decoder rule: no playback of speaker audio without a present source
=== stage 10/11  monolinguality ============================================
[00:01:23]        INFO  sheet 20/A: 6/6 human voices sorted as speech, 4/4 non-speech out
[00:01:23]        INFO  sheet 20/A item 6: clicks retained inside speech class
[00:01:23]        INFO  sheet 20/B: pairs 1,2,3 all returned undifferentiated
[00:01:24]        INFO  speaker does not hear two systems in one mouth
[00:01:24]        INFO  sheet 20/C: speaker sorts by voice, not by system
[00:01:24]        INFO  sheet 20/D panel 5: outsider register absent, same forms returned
[00:01:24]        INFO  sheet 20/E: heads produced no lexical or morphological switch
[00:01:24]        INFO  sheet 20/F: manual signal set recovered, 4 signs, hunting only
[00:01:25]        INFO  registering manual set as a closed register, not a second language
[00:01:25]        INFO  sheet 20/G control matched first pass, cosine 0.97
[00:01:25]        INFO  MONOLINGUAL: 1 spoken system, 1 closed manual register
[00:01:25]        INFO  no bridge language available, no pivot possible
[00:01:25]        INFO  decoder must run direct against the peg chain
=== stage 11/11  decoder ===================================================
[00:01:28]        INFO  building lexicon table ......................... 44 entries
[00:01:28]        INFO  building morphology table ...................... 5 suffixes, 2 clitics
[00:01:29]        INFO  building context layer ......................... 6 gates
[00:01:29]        INFO    gate 1  ungrounded referent       214 concepts blocked
[00:01:29]        INFO    gate 2  unbounded time            blocked
[00:01:29]        INFO    gate 3  type/token artefacts      blocked
[00:01:29]        INFO    gate 4  unattributed assertion    query inserted
[00:01:30]        INFO    gate 5  interested assertion      marked ne [long]
[00:01:30]        INFO    gate 6  coinage without referent  refused
[00:01:31]        INFO  training alignment on 388 turns ................ 6 epochs
[00:01:52]        INFO  held-out set 40 turns
[00:01:52]        INFO  exact 21/40, partial 13/40, refused 6/40, wrong 0/40
[00:01:52]        INFO  refusals all gate 1, all correct refusals
[00:01:53]        INFO  writing /var/elicit/session-0007/model/mandrin-k.tr
[00:01:54]        INFO  writing /var/elicit/session-0007/model/mandrin-k.gates
[00:01:54]        INFO  writing /var/elicit/session-0007/coverage.txtBUILD OK  in 1m54s
Coverage against operator corpus:

  concrete objects present in scene ........................  94%
  actions with a visible agent .............................  88%
  quantity, up to 30 ....................................... 100%
  quantity, above 30 .......................................   0%  gate 2
  time, within 6 cold cycles ...............................  71%
  time, beyond .............................................   0%  gate 2
  place and terrain ........................................  83%
  internal states ..........................................  80%
  artefacts made elsewhere .................................  12%  gate 1
  records, files, copies ...................................   0%  gate 3
  institutions, law, money .................................   0%  gate 1
  encyclopedic text ........................................   0%  gate 1, gate 4
  overall ..................................................  31%

WARN  the translator will refuse more than it renders.
Refusals are logged with the gate that fired. Do not lower a gate to raise coverage.

Ovlat and Nika#

Vlad

loading the generated tensors into the translator and saying into the computer Hello.

Computer

the first sound coming out of the machine, thin off the small speaker Ovlat.

Nika

going stiff at the sound of the word coming out of the machine, then putting her hand out toward it Ovlat. pointing at him, then at the machine, then at him again Ovlat na? Ovlat ta na?

mandrin-k runtime 14:02:11

wrap

mandrin-k 0.9.4  runtime
model: /var/elicit/session-0007/model/mandrin-k.tr
gates: 6 loaded
direction: ro > k   [primary peg]
input device: mic0   output device: spk0[14:02:11] in        "Hello."
[14:02:11] asr       en detected, peg chain position 2
[14:02:11] map       greeting, phatic
[14:02:11] gate 1    ungrounded referent: greeting formula
[14:02:11] note      no phatic exchange recorded in 388 turns
[14:02:11] note      speaker opened with alarm, not with greeting
[14:02:11] fall      substituting nearest attested: self-identification
[14:02:11] out       "ovlat"
[14:02:11] spk0      playing 0.4s[14:02:19] mic0      capture 3.9s, speaker voice, snr 21 dB
[14:02:19] asr       k: ovlat
[14:02:20] asr       k: ovlat na
[14:02:20] asr       k: ovlat ta na
[14:02:20] seg       3 utterances, contour: level, rising, rising
[14:02:20] tag       ovlat = proper noun, borrowed sheet 7 frame 5, retained
[14:02:20] tag       na = relator, sheet 19/F, position-sensitive
[14:02:20] tag       na utterance-final + rising = interrogative particle
[14:02:20] tag       ta = definiteness clitic, referent present or prior
[14:02:21] ctx       visual channel unavailable, deixis unresolved
[14:02:21] ctx       prior turn: system output "ovlat"
[14:02:21] ctx       resolving ta against most recent referent: the machine
[14:02:21]          direction: k > ro > ennika      "Vlad."
nika      "Is that Vlad?"
nika      "Is that one Vlad too?"[14:02:21] warn      utterance 3: <ta> attached to a machine-class referent
[14:02:21] warn      gate 3 adjacent: speaker has no type/token distinction
[14:02:21] note      the question may be whether the machine is a second Vlad, or whether the voice in it is the same Vlad. The grammar
[14:02:21] note      does not separate these and the model will not guess.
[14:02:21] note      confidence 0.44, below display threshold 0.60

Vlad

Yes, I am Vlad.

Computer

Ha. Aha ovlat.

Nika

sitting back down on the gravel and letting the spear rest across her knees Ha. pointing at the machine Ovlat ne. pointing at him Ovlat ta. pointing at herself, slowly, and waiting Nika.

mandrin-k runtime 14:02:34

wrap

[14:02:34] in        "Yes, I am Vlad."
[14:02:34] asr       en, peg chain position 2
[14:02:34] map       affirmative + copula + proper noun
[14:02:34] gate 4    attribution: speaker-internal, source present, pass
[14:02:34] morph     first person: bare stem <aha>, sheet 19/C
[14:02:34] morph     copula: absent in target, juxtaposition used
[14:02:34] cont      affirmative = level contour, sheet 5/A pass 1
[14:02:34] note      <ha> available as standalone affirmative, sheet 15/B strip 4
[14:02:34] out       "ha. aha ovlat"
[14:02:34] spk0      playing 1.1s[14:02:51] mic0      capture 5.2s, speaker voice, snr 22 dB
[14:02:51] asr       k: ha
[14:02:51] asr       k: ovlat ne
[14:02:52] asr       k: ovlat ta
[14:02:52] asr       k: nika
[14:02:52] seg       4 utterances, contour: level, level, level, level
[14:02:52] tag       ha = affirmative, sheet 15/B
[14:02:52] tag       ne [short] = wrong, mistaken, sheet 19/D
[14:02:52] tag       ta = definiteness clitic, referent present or prior
[14:02:52] lex       nika: no entry, 44-type lexicon
[14:02:52] gate 6    coinage without referent: check
[14:02:52] ctx       gesture channel unavailable, deixis from prior turn only
[14:02:52] ctx       prior operator turn established self-naming frame
[14:02:52] in        fer  utterance 4 occupies the slot filled by <ovlat> at 14:02:11
[14:02:52] in        fer  proper noun, self-reference, first occurrence
[14:02:52] lex       writing entry: nika, proper noun, speaker
[14:02:52] lex       lexicon 44 -> 45 typesnika      "Yes."
nika      "That is not what Vlad is."
nika      "That one is Vlad."
nika      "Nika."[14:02:52] note      utterance 2: <ne> short, wrong rather than dangerous. The correction is of the naming, not a warning.
[14:02:52] note      utterance 4: name accepted on positional evidence alone, confidence 0.58.

Vlad

So, you are Nika, right?

Computer

Semek nika na?

Nika

taking the point off the spear shaft and setting the shaft down behind her, out of her own reach Ha. Aha Nika.

mandrin-k runtime 14:03:09

wrap

[14:03:09] in        "So, you are Nika, right?"
[14:03:09] asr       en, peg chain position 2
[14:03:09] map       second person + copula + proper noun + tag question
[14:03:09] morph     second person: bare stem <semek>, sheet 19/B
[14:03:09] morph     copula absent in target, juxtaposition used
[14:03:09] lex       nika: entry present, confidence 0.58
[14:03:09] cont      tag question -> interrogative contour, rising
[14:03:09] note      target marks polarity on both channels, sheet 5/C
[14:03:09] note      <na> utterance-final reinforces the rising contour
[14:03:09] out       "semek nika na?"
[14:03:09] spk0      playing 0.9s[14:03:13] mic0      capture 2.4s, speaker voice, snr 22 dB
[14:03:13] asr       k: ha
[14:03:13] asr       k: aha nika
[14:03:13] seg       2 utterances, contour: level, level
[14:03:13] lex       nika: confirmed in speaker-initiated self-reference frame
[14:03:13] lex       confidence 0.58 -> 0.91nika      1/2   "Yes."
nika      2/2   "I am Nika."

Vlad

And you are a hunter.

Computer

Semek na rusa naga-tu.

Nika

looking at his hands, then at the fire Ha. Aha na rusa naga-tu. Aha na bontu ktha. Semek na ne.

mandrin-k runtime 14:03:31

wrap

[14:03:31] in        "And you are a hunter."
[14:03:31] asr       en, peg chain position 2
[14:03:31] map       second person + copula + agent noun
[14:03:31] lex       hunter: no direct entry
[14:03:31] gate 1    ungrounded referent: check
[14:03:31] note      target has no agent-noun derivation, sheet 17 induced 5
[14:03:31] note      suffixes and none of them forms a doer from a verb
[14:03:31] note      a person defined by a recurring activity is not attested
[14:03:31] fall      rendering as activity with an agent: sheet 3 strip 1 frame
[14:03:31] fall      <semek na aha rusa>, second person substituted for first
[14:03:31] out       "semek na rusa naga-tu"
[14:03:31] spk0      playing 1.3s
[14:03:31] warn      this states that you take deer with a spear. It does not state that this is what you are. Confidence 0.63.[14:03:37] mic0      capture 6.1s, speaker voice, snr 21 dB
[14:03:37] asr       k: ha
[14:03:37] asr       k: aha na rusa naga-tu
[14:03:38] asr       k: aha na bontu ktha
[14:03:38] asr       k: semek na ne
[14:03:38] seg       4 utterances, contour: level, level, level, level
[14:03:38] tag       ktha1 free form, level+long: beyond count, large
[14:03:38] tag       ne [short], utterance-final: wrong, mistaken, lackingnika      "Yes."
nika      "I take deer with a spear."
nika      "I take bison, many of them."
nika      "You do not."

The marks-maker#

Vlad

I am a researcher. pointing at the looking glass on the cave wall on the right of the crossed closed mouth

Computer

Aha semek.

Nika

getting up and going to the cave wall, and putting two fingers on the crossed closed mouth, then on the looking glass beside it Ha. Semek ta. coming back and sitting down Semek na ne rusa. Semek na semek.

mandrin-k runtime 14:03:58

wrap

[14:03:58] in        "I am a researcher."
[14:03:58] asr       en, peg chain position 2
[14:03:58] map       first person + copula + agent noun
[14:03:58] lex       researcher: no direct entry
[14:03:58] gate 1    ungrounded referent: BLOCKED
[14:03:58] note      no agent-noun derivation, as at 14:03:31
[14:03:58] note      no attested activity to fall back on: looking in order to know, with no taking and no making, is not in the corpus
[14:03:58] note      sheets 1, 3, 9, 15 contain no panel where a figure observes without acting on what is observed
[14:03:58] fall      nearest attested: <semek>, sheet 19/B, marks-maker
[14:03:58] out       "aha semek"
[14:03:58] spk0      playing 0.7s
[14:03:58] warn      this states that you are the one who makes marks. The reason for making them is not carried. Confidence 0.41.[14:04:06] mic0      capture 4.8s, speaker voice, snr 21 dB
[14:04:06] asr       k: ha
[14:04:06] asr       k: semek ta
[14:04:07] asr       k: semek na ne rusa
[14:04:07] asr       k: semek na semek
[14:04:07] seg       4 utterances, contour: level, level, level, level
[14:04:07] tag       utterance 4: stem repeated across relator, sheet 19/F
[14:04:07] note      this construction has one prior occurrence, sheet 17/F, where the speaker built a place name from a rootnika      "Yes."
nika      "That is the marks-maker."
nika      "You do not take deer."
nika      "You make marks at marks."

Vlad

Yes, I make marks at marks. And I didn't expect to see you, but I certainly expected you would spare me, considering that I am a lone person with no particular reputation.

Computer

Ha. Aha na semek. Aha ne semek ta.

Nika

Ha. pointing at the marks on the wall, then at the entrance, then holding her hand still Semek ta. Aha ne semek ta. turning her hand over, palm up, and leaving it there Semek na?

mandrin-k runtime 14:04:29

wrap

[14:04:29] in        "Yes, I make marks at marks. And I didn't expect to see you, but I certainly expected you would spare me, considering that I am a lone person with no particular reputation."
[14:04:29] asr       en, peg chain position 2
[14:04:29] seg       4 clauses[14:04:29] c1        affirmative + first person + attested construction
[14:04:29] c1        out "ha. aha na semek"
[14:04:29] c2        "did not expect to see you"
[14:04:29] gate 1    ungrounded referent: BLOCKED
[14:04:29] note      expectation is a held state about an event that did not occur. Sheet 4 lexicalised 8 internal states and none of them takes an unrealised event as its object.
[14:04:29] fall      dropping the expectation, keeping the event
[14:04:29] c2        out "aha ne semek ta"          [I did not see you]
[14:04:29] warn      c2 renders as a statement about the past, not about a belief
[14:04:29] c3        "expected you would spare me"
[14:04:29] gate 1    ungrounded referent: BLOCKED
[14:04:29] note      spare = withhold an action that was available. Sheet 9/B
[14:04:29] note      established actions are separable from their objects, but an action deliberately not taken has no panel anywhere in the corpus and no form was ever elicited for it.
[14:04:29] fall      none available
[14:04:29] c3        REFUSED
[14:04:29] c4        "lone person with no particular reputation"
[14:04:29] lex       lone person: <aha-ke>, sense 2, sheet 4/A panel 5
[14:04:29] warn      <aha-ke> sense 1 is a dead person. Disambiguation is by the animacy of surrounding context and the surrounding context here contains a refused clause about being killed.
[14:04:29] warn      Sense selection unsafe. Confidence 0.29.
[14:04:29] note      reputation: standing held by others who are absent.
[14:04:29] gate 4    attribution: no source present for the absent others
[14:04:29] c4        REFUSED
[14:04:29] out       "ha. aha na semek. aha ne semek ta"
[14:04:29] spk0      playing 2.2s
[14:04:29] warn      2 of 4 clauses refused. The rendered portion says that you make marks and that you had not seen her. Nothing about
[14:04:29] warn      what you expected has been said.

Vlad

activating runtime train mode It's a partial translator.

Computer

Kruf ta na semek. Semek ne. Kesta-ne ne.

Nika

putting one hand on the machine and one on her own chest Kruf ta na semek. Semek ne, ha. Kesta-ne ne, ha. lifting the hand off her chest and making the split in the air, then pointing at him Aha na kesta. Ovlat na kesta-ne.

mandrin-k runtime 14:04:52

wrap

[14:04:52] mode      runtime train: ON
[14:04:52] note      new forms observed in dialogue are written to the lexicon
[14:04:52] note      gates remain enforced. Training does not lower a gate.
[14:04:52] note      the open utterance at 14:04:47 is unanswered: "semek na?"
[14:04:52] note      palm-up frame, sheet 15/E frame 3. This is a source query.[14:04:55] in        "It's a partial translator."
[14:04:55] asr       en, peg chain position 2
[14:04:55] map       demonstrative + copula + modified agent noun
[14:04:55] lex       translator: no direct entry
[14:04:55] gate 1    ungrounded referent: BLOCKED
[14:04:55] note      a translator presupposes two speech systems. Sheet 20/B
[14:04:55] note      returned no differentiation across three pairs and sheet
[14:04:55] note      20/D returned no outsider register. The speaker has no model of a second system, so a thing that moves between
[14:04:55] note      systems has nothing to attach to.
[14:04:55] gate 3    type/token artefacts: adjacent
[14:04:55] note      partial = incomplete against a whole that was never built.
[14:04:55] note      sheet 18/C stage 2 refused an object finished before its place existed. An incomplete thing is available: sheet
[14:04:55] note      18/B stages 4 and 5, a wall part-built.
[14:04:55] fall      rendering the machine as a part-built thing that speaks
[14:04:55] fall      attested pieces: <kruf> stone, <kesta-ne> joined, <semek> speech/marks, <ne> short, wrong or lacking
[14:04:55] out       "kruf ta na semek. semek ne. kesta-ne ne"
[14:04:55] spk0      playing 2.6s
[14:04:55] warn      this says: that stone speaks, its speech is wrong, it is not finished being joined. Confidence 0.52.
[14:04:55] warn      <semek ne> may be heard as the machine lying rather than the machine failing. Sheet 15 established the speaker tracks
[14:04:55] warn      interest behind an assertion and no interest was declared.[14:05:03] mic0      capture 7.3s, speaker voice, snr 21 dB
[14:05:03] asr       k: kruf ta na semek
[14:05:04] asr       k: semek ne, ha
[14:05:04] asr       k: kesta-ne ne, ha
[14:05:04] asr       k: aha na kesta
[14:05:04] asr       k: ovlat na kesta-ne
[14:05:04] tag       utterance 4: <kesta> bare, first person agent
[14:05:04] tag       utterance 5: <kesta-ne> bare, second person agent
[14:05:04] note      sheet 9/A frames 3 and 4: kesta splits, kesta-ne joinsnika      "That stone speaks."
nika      "Its speech is wrong, yes."
nika      "It is not finished being joined, yes."
nika      "I split."
nika      "Vlad joins."

The high ground#

Vlad

You split from the group.

Computer

Semek na kesta. Semek, aha-ktha.

Nika

Ne. Aha ne kesta. drawing two shapes in the gravel with one finger, a small one and a large one, and putting a mark in each Aha-ktha na karnu-me, aha na kar-nu. Miht-ktha, aha na rusa. Aha na Ovlat, miht ta. looking at the cave mouth Aha-ktha ne semek Ovlat.

mandrin-k runtime 14:05:22

wrap

[14:05:22] in        "You split from the group."
[14:05:22] asr       en, peg chain position 2
[14:05:22] map       second person + kesta + source phrase
[14:05:22] lex       kesta: attested, sheet 9/A frame 3, splitting one into two
[14:05:22] lex       group: <aha-ktha>, sheet 19/C panel 3
[14:05:22] warn      the source phrase has no attested marking. Sheet 6 induced
[14:05:22] warn      agent and patient and no ablative. Separation from a set is carried in the corpus by kesta alone, with both halves named.
[14:05:22] fall      naming both halves: the speaker and the group
[14:05:22] out       "semek na kesta. semek, aha-ktha"
[14:05:22] spk0      playing 1.8s
[14:05:22] warn      this says: you split, you and the group. Which half went is not carried. Confidence 0.55.[14:05:31] mic0      capture 9.6s, speaker voice, snr 20 dB
[14:05:31] asr       k: ne
[14:05:31] asr       k: aha ne kesta
[14:05:32] asr       k: aha-ktha na karnu-me, aha na kar-nu
[14:05:32] asr       k: miht-ktha, aha na rusa
[14:05:32] asr       k: aha na ovlat, miht ta
[14:05:33] asr       k: aha-ktha ne semek ovlat
[14:05:33] seg       6 utterances
[14:05:33] tag       utterance 3: karnu-me diminutive, kar-nu bare
[14:05:33] tag       <kar> root recovered stage 3, rock mass, high ground
[14:05:33] tag       utterance 4: miht-ktha, plural of sun/day, count under 30
[14:05:33] warn      utterance 4 gives no number. Sheet 16/B established counts to seven with drawn suns and none were available here.nika      "No."
nika      "I did not split."
nika      "The group is at the small high ground, I am at the high ground."
nika      "Some days, I take deer."
nika      "I came to Vlad, this sun."
nika      "The group has not seen Vlad."[14:05:33] note      utterance 3 contrasts two places by size and the model has no map. The speaker is somewhere the group is not, and the difference between the two places is one of scale.

Vlad

Why did you decide to stay rather than leave me alone?

Computer

Semek na kar-nu ta na?

Nika

Ha. Aha na kar-nu ta. holding her hand near the fire Miht-nu ta. pointing at him, then at the four marks scratched on the wall Aha na miht-nu, miht-ktha semek. pointing out at the scree, and then at the back of the cave where the fissure is Torna ta, karnu ta. turning back to him Aha-ke ne.

mandrin-k runtime 14:05:51

wrap

[14:05:51] in        "Why did you decide to stay rather than leave me alone?"
[14:05:51] asr       en, peg chain position 2
[14:05:51] seg       1 clause, interrogative, causal
[14:05:51] map       why: no interrogative of cause in target
[14:05:51] gate 1    ungrounded referent: BLOCKED
[14:05:51] note      sheet 5 recovered four contours and none of them asks for a reason. Sheet 15/E frame 3 produced a source query, which asks who says a thing, and that is the only interrogative beyond the polar one that was ever elicited.
[14:05:51] note      decide: an action selected over an available alternative.
[14:05:51] note      as at 14:04:29 c3, an action not taken has no form.
[14:05:51] note      stay: no attested stative. Sheet 3 strips render position as a figure being somewhere, never as continuing to be.
[14:05:51] fall      none available for the causal frame
[14:05:51] fall      downgrading to polar: asking whether she is here
[14:05:51] out       "semek na kar-nu ta na?"
[14:05:51] spk0      playing 1.4s
[14:05:51] warn      this asks whether she is at this place. It does not ask why.
[14:05:51] warn      The question you put cannot be put in this language with what has been elicited. Confidence 0.60.[14:06:01] mic0      capture 8.8s, speaker voice, snr 20 dB
[14:06:01] asr       k: ha
[14:06:01] asr       k: aha na kar-nu ta
[14:06:02] asr       k: miht-nu ta
[14:06:02] asr       k: aha na miht-nu, miht-ktha semek
[14:06:02] asr       k: torna ta, karnu ta
[14:06:03] asr       k: aha-ke ne
[14:06:03] seg       6 utterances
[14:06:03] tag       utterance 3: miht-nu, place characterised by fire, sheet 17/F
[14:06:03] tag       utterance 4: <semek> in the marks sense, sheet 19/B panel 4
[14:06:03] tag       utterance 6: aha-ke, two senses live
[14:06:03] gate 3    disambiguation by animacy of surrounding context
[14:06:03] note      surrounding context names a clawed animal on this ground and a person alone. Both senses are supported.
[14:06:03] note      sense 2 selected on the strength of utterance 5, which places the danger before the state rather than after it.
[14:06:03] note      confidence 0.61, above threshold by 0.01.nika      "Yes."
nika      "I am at this high ground."
nika      "This is a fire place."
nika      "You are at a fire place, the marks of many days."
nika      "There are bears here, on this high ground."
nika      "A person alone is wrong."

Vlad

It's wrong, but what else to do if I have to train a translator that constantly refuses?

Computer

Ha. Ne. Kruf ta na semek ne. Kesta-ne ne.

Nika

taking hold of the machine at its edges with both hands, lifting it a little off the gravel, and setting it back down between them Ha, ne. pointing at the machine, then at her own mouth, then at his Kruf ta na semek ne. pointing at herself and then at the marks on the wall Aha na semek. Aha na semek Ovlat. putting her hand out, palm up, and waiting

mandrin-k runtime 14:06:24

wrap

[14:06:24] in        "It's wrong, but what else to do if I have to train a translator that constantly refuses?"
[14:06:24] asr       en, peg chain position 2
[14:06:24] seg       3 clauses[14:06:24] c1        "it's wrong" + agreement with prior turn
[14:06:24] lex       ne [short] = wrong, mistaken, sheet 19/D
[14:06:24] c1        out "ha. ne"
[14:06:24] c2        "what else to do"
[14:06:24] gate 1    ungrounded referent: BLOCKED
[14:06:24] note      an alternative course not taken. Third refusal of this frame in the session, see 14:04:29 c3 and 14:05:51.
[14:06:24] note      the corpus contains no panel in which a figure holds two possible actions and one of them fails to happen.
[14:06:24] c2        REFUSED
[14:06:24] c3        "train a translator that constantly refuses"
[14:06:24] gate 1    ungrounded referent: BLOCKED
[14:06:24] note      translator: refused at 14:04:55, no second system available
[14:06:24] note      train: a thing improved by repetition. Sheet 14/D found
[14:06:24] note      the speaker tracks names in a store and not states of a store, so a thing that is the same thing after being made
[14:06:24] note      better has no handle.
[14:06:24] note      refuses: withholding an available action, refused above.
[14:06:24] fall      the machine is available as a part-built thing, 14:04:55
[14:06:24] fall      rendering the machine failing to speak, without the cause
[14:06:24] c3        out "kruf ta na semek ne. kesta-ne ne"
[14:06:24] out       "ha. ne. kruf ta na semek ne. kesta-ne ne"
[14:06:24] spk0      playing 3.1s
[14:06:24] warn      1 of 3 clauses refused. The rendered portion agrees with her and says the stone speaks wrongly and is unfinished. Your question was not put.
[14:06:24] warn      this repeats what was said at 14:04:55 with agreement added.
[14:06:24] warn      she has already accepted that statement.

Vlad

activating automatic word derivation in the runtime training parameters Say again.

Computer

Semek. Semek ktha-aha.

Nika

Ha. Aha na semek. Aha na semek Ovlat. Ovlat na semek aha. taking in the four scratched marks, the reference figures beside them, the machine, the torn sheets of paper on the gravel, and then the whole cave Semek-nu ta.

mandrin-k runtime 14:06:47

wrap

[14:06:47] param     runtime.derivation.auto = ON
[14:06:47] note      the model may now build new stems from attested roots and the five suffixes induced at stage 3, without waiting for the speaker to supply them.
[14:06:47] note      gate 6 remains ON. A coinage still requires a referent that is present, drawn, or already in the lexicon. Derivation without a referent is refused as before.
[14:06:47] note      the speaker derived karnu unprompted, stage 3, and derived
[14:06:47] note      miht-nu, sula-nu and kesta-nu in sheet 17/F. The procedure is hers and the model is being allowed to run it.
[14:06:47] warn      derived stems are marked in output. She has never heard them and may reject any of them.[14:06:52] in        "Say again."
[14:06:52] asr       en, peg chain position 2
[14:06:52] map       imperative + repetition
[14:06:52] lex       say: <semek>, speech sense
[14:06:52] lex       again: no entry
[14:06:52] gate 1    check
[14:06:52] note      sheet 14/B frames 3 and 4 showed a mark rubbed out and the same mark drawn a second time, and the speaker returned
[14:06:52] note      <ne kesta-ne ta> for the pair. Repetition is attested as a thing done twice, not as a word.
[14:06:52] fall      naming the count instead: speech, two
[14:06:52] out       "semek. semek ktha-aha"
[14:06:52] spk0      playing 1.2s[14:06:58] mic0      capture 6.4s, speaker voice, snr 20 dB
[14:06:58] asr       k: ha
[14:06:58] asr       k: aha na semek
[14:06:59] asr       k: aha na semek ovlat
[14:06:59] asr       k: ovlat na semek aha
[14:06:59] asr       k: semek-nu ta
[14:06:59] seg       5 utterances
[14:06:59] lex       semek-nu: no entry
[14:06:59] deriv     semek + -nu, place characterised by the stem, stage 3
[14:06:59] deriv     speaker-produced, not model-produced
[14:06:59] lex       writing entry: semek-nu, place of speech or of marks
[14:06:59] lex       lexicon 45 -> 46 types
[14:06:59] note      she has run the derivation on the same root the model was just cleared to run it on, in the same minute, unprompted.nika      "Yes."
nika      "I make marks."
nika      "I make marks at Vlad."
nika      "Vlad makes marks at me."
nika      "This is a marks place."

Vlad

I had to teach the computer.

Computer

Aha na semek kruf ta. Semek-tu. Miht-ktha, ontu.

Nika

Semek-tu. saying it a second time and putting her hand on it Ha, semek-tu. reaching behind her for the spear shaft, laying it beside the machine on the gravel, and putting a hand on each Naga-tu, aha. Semek-tu, Ovlat. taking her hand off the machine Kruf ta na semek-tu ne. putting it back on Semek-tu na kruf ta.

mandrin-k runtime 14:07:18

wrap

[14:07:18] in        "I had to teach the computer."
[14:07:18] asr       en, peg chain position 2
[14:07:18] seg       1 clause, first person + obligation + transfer verb + object
[14:07:18] lex       computer: <kruf ta>, established 14:04:55, accepted by speaker
[14:07:18] lex       teach: no entry
[14:07:18] gate 1    check
[14:07:18] note      sheet 3 strip 15 showed a hand holding an object out and a second hand closing on it, and the speaker gave a giving
[14:07:18] note      frame for it. Teaching is transfer without loss and the
[14:07:18] note      giving frame carries loss: the first hand ends empty.
[14:07:18] gate 3    type/token: the same thing held by two parties at once
[14:07:18] gate 3    BLOCKED at sheet 10/D item 7, refused by speaker
[14:07:18] fall      derivation available: semek + -tu, instrument
[14:07:18] deriv     semek-tu, model-produced, marked
[14:07:18] note      this names the thing marks are made through, which is what the machine has been for the last four hours. It does not name the giving.
[14:07:18] lex       obligation: no modal system in target
[14:07:18] note      sheet 16 established time and sheet 6 established what is finished. Necessity was never elicited on any sheet.
[14:07:18] fall      dropping the obligation, keeping the activity and its end
[14:07:18] out       "aha na semek kruf ta. semek-tu. miht-ktha, ontu"
[14:07:18] spk0      playing 2.7s
[14:07:18] warn      this says: I made marks at that stone, a marks-instrument, many days, finished. Nothing about having to. Confidence 0.57
[14:07:18] warn      <semek-tu> is model-derived and has never been said to her.[14:07:27] mic0      capture 7.9s, speaker voice, snr 20 dB
[14:07:27] asr       k: semek-tu
[14:07:27] asr       k: ha, semek-tu
[14:07:28] asr       k: naga-tu, aha. semek-tu, ovlat
[14:07:28] asr       k: kruf ta na semek-tu ne
[14:07:28] asr       k: semek-tu na kruf ta
[14:07:28] seg       5 utterances, contour: level, level, level, level, level
[14:07:28] lex       semek-tu accepted on first hearing, no correction
[14:07:28] lex       confidence 0.57 -> 0.88
[14:07:28] note      utterance 4 and utterance 5 reverse the agent and patient
[14:07:28] note      slots around the same two terms. Sheet 6 stage 4 resolved
[14:07:28] note      this collision by position and it is being used here.nika      "A marks-instrument."
nika      "Yes, a marks-instrument."
nika      "A spear, mine. A marks-instrument, Vlad's."
nika      "That stone is not a marks-instrument."
nika      "A marks-instrument is that stone."[14:07:28] note      utterances 4 and 5 stand together and are not a correction of each other. The model has no reading that satisfies both and is passing them through as heard.

Stones beyond counting#

Vlad

showing images of silica and rare earths The computer is composed of complex stones.

Computer

four images coming up on the lit face, quartz sand close in, a cut wafer with a mirror finish, an ore sample in a hand, seven coloured dishes in a row Kruf ta na kruf-ktha. Kruf-ktha-ke.

Nika

putting a finger on the sand image, then on the wafer, then holding the two fingers apart Kruf. Ha, kruf ta. going back to the sand Kruf-me-ktha, kar-nu. moving to the wafer and leaving her finger on the mirror surface Kruf ta na kruf ne. Aha ne semek kruf ta. turning to him Kesta-nu na?

mandrin-k runtime 14:07:49

wrap

[14:07:49] visual    4 images opened on lit face
[14:07:49] visual    quartz sand, macro
[14:07:49] visual    cut silicon wafer, mirror finish
[14:07:49] visual    bastnasite ore, hand sample
[14:07:49] visual    separated rare earth oxides, seven dishes, coloured
[14:07:49] note      images are photographs. Sheet 8/A established the speaker
[14:07:49] note      reads a side view as terrain and sheet 8/B established that unsupported referents return nothing. No photograph was ever shown on any sheet. This is unelicited ground.[14:07:52] in        "The computer is composed of complex stones."
[14:07:52] asr       en, peg chain position 2
[14:07:52] lex       computer: <kruf ta>, accepted
[14:07:52] lex       stone: <kruf>, sheet 1 panel 12
[14:07:52] lex       composed of: sheet 18/A, taking a thing apart and naming the parts. Attested and productive.
[14:07:52] lex       complex: no entry
[14:07:52] gate 1    check
[14:07:52] note      sheet 18/B ran twenty courses of one repeated unit and the speaker marked the count with <ktha-ke>, beyond counting.
[14:07:52] note      many parts is available. Intricate is not.
[14:07:52] fall      rendering as parts beyond count
[14:07:52] out       "kruf ta na kruf-ktha. kruf-ktha-ke"
[14:07:52] spk0      playing 2.0s
[14:07:52] warn      this says: that stone is stones, stones beyond counting.
[14:07:52] warn      Confidence 0.71.[14:08:04] mic0      capture 11.2s, speaker voice, snr 20 dB
[14:08:04] asr       k: kruf
[14:08:04] asr       k: ha, kruf ta
[14:08:05] asr       k: kruf-me-ktha, kar-nu
[14:08:05] asr       k: kruf ta na kruf ne
[14:08:05] asr       k: aha ne semek kruf ta
[14:08:06] asr       k: kesta-nu na?
[14:08:06] seg       6 utterances, contour: level, level, level, level, level, rising
[14:08:06] tag       utterance 3: kruf-me-ktha, two suffixes stacked, diminutive then plural. First stacking observed in the session.
[14:08:06] morph     suffix order confirmed: stem, size, number
[14:08:06] lex       kesta-nu: derived by speaker in sheet 17/F, place of knapping
[14:08:06] note      utterance 6 carries the interrogative contour and the source
[14:08:06] note      query frame from sheet 15/E is absent. This is a polar
[14:08:06] note      question about a place.nika      "Stone."
nika      "Yes, that stone."
nika      "Small stones, high ground."
nika      "That stone is not stone."
nika      "I have not seen that stone."
nika      "Is it from a knapping place?"[14:08:06] note      utterance 3 names the sand image with a stacked derivation
[14:08:06] note      she has not used before and pairs it with terrain.
[14:08:06] note      utterance 4 rejects the identification the model made in the same exchange. Confidence on <kruf ta> unchanged, since
[14:08:06] note      she has kept the term while denying the material.

Vlad

Complex knapping place, a factory.

Computer

Ha. Kesta-nu-ktha.

Nika

Kesta-nu-ktha. shaking her head once Ne. saying it slowly, with a hand moving left to right through the air for each piece of the word Kesta-ktha-nu. Ha, kesta-ktha-nu. looking back down at the seven dishes of coloured oxide Aha-ktha na kesta-ktha-nu na?

mandrin-k runtime 14:08:22

wrap

[14:08:22] in        "Complex knapping place, a factory."
[14:08:22] asr       en, peg chain position 2
[14:08:22] parse     fragment, no verb. Two nominals in apposition.
[14:08:22] note      the second nominal is an operator-language term and the first is a gloss of it built from attested target material.
[14:08:22] note      this matches the definition frame: a term the speaker lacks, paired with a construction the speaker holds, in one turn.
[14:08:22] mode      definition capture: ON[14:08:22] lex       kesta-nu: attested, speaker-derived, sheet 17/F
[14:08:22] lex       complex: refused at 14:07:52, rendered as beyond count
[14:08:22] fall      reusing that rendering, now on the place rather than the parts
[14:08:22] build     kesta-nu + ktha, place of knapping, beyond count
[14:08:22] deriv     kesta-nu-ktha, model-produced, marked
[14:08:22] warn      suffix order confirmed at 14:08:06 is stem, size, number.
[14:08:22] warn      <-nu> is a place suffix and has never been observed with a number suffix after it. This ordering is untested.
[14:08:22] lex       writing provisional entry: kesta-nu-ktha = factory
[14:08:22] lex       status PROVISIONAL, confidence 0.34, awaiting speaker uptake
[14:08:22] lex       lexicon 46 -> 47 types
[14:08:22] note      a provisional entry is promoted only if the speaker returns
[14:08:22] note      the form unprompted in a later turn. If she corrects it, the correction is written and this entry is dropped.
[14:08:22] out       "ha. kesta-nu-ktha"
[14:08:22] spk0      playing 1.5s[14:08:31] mic0      capture 8.7s, speaker voice, snr 20 dB
[14:08:31] asr       k: kesta-nu-ktha
[14:08:31] asr       k: ne
[14:08:32] asr       k: kesta-ktha-nu
[14:08:32] asr       k: ha, kesta-ktha-nu
[14:08:32] asr       k: aha-ktha na kesta-ktha-nu na?
[14:08:32] seg       5 utterances, contour: level, level, level, level, rising
[14:08:32] lex       CORRECTION: kesta-nu-ktha rejected, kesta-ktha-nu returned
[14:08:32] morph     suffix order revised: stem, number, place
[14:08:32] morph     the place suffix closes the word and nothing follows it revised order consistent with karnu, miht-nu, sula-nu, semek-nu
[14:08:32] lex       dropping provisional entry kesta-nu-ktha
[14:08:32] lex       writing entry: kesta-ktha-nu = factory, confidence 0.79
[14:08:32] note      she repeated the model's form once, marked it wrong, and supplied the repair without being asked.nika      "A knapping-place-many."
nika      "Wrong."
nika      "A many-knappings place."
nika      "Yes, a many-knappings place."
nika      "Are there people at the many-knappings place?"

Vlad

There used to be many people. 1860, general date of industrial revolution. There are not many people. 1970, general date of information age.

Computer

nothing put on the speech channel, the images left up as they are

Nika

waiting, with her hand still open over the dishes of coloured oxide, and when nothing comes out of the machine, taking the hand back Semek na? pointing at the machine, then at his mouth Kruf ta na semek ne. Ovlat na semek ne. sitting with that for a while, then putting one finger back on the image of the seven dishes Aha-ktha na? and when he does not answer Aha-ktha ne. Aha-ktha-ke.

mandrin-k runtime 14:08:51

wrap

[14:08:51] in        "There used to be many people. 1860, general date of industrial revolution. There are not many people. 1970, general date of information age."
[14:08:51] asr       en, peg chain position 2
[14:08:51] seg       4 clauses, 2 assertions, 2 date-anchored appositions[14:08:51] date      parsing 1860, 1970
[14:08:51] clock     system clock: -38050 (session header, MANDRIN-K)
[14:08:51] date      1860 CE resolves to +39910 relative to system clock
[14:08:51] date      1970 CE resolves to +40020 relative to system clock
[14:08:51] warn      both dates are forward of the system clock by ~40,000 years
[14:08:51] gate 2    unbounded time: BLOCKED
[14:08:51] note      guard set at stage 6: refuse source dates beyond 6 cold
[14:08:51] note      cycles. These exceed the counted forward range by a factor of roughly 20,000.
[14:08:51] note      sheet 16/B frames 7 and 8 returned <ktha1> for a band of thirty suns, which is a quantity and not a count. Sheet
[14:08:51] note      16/F frame 4 returned <ne> for unbounded time, which is the same form the speaker uses for absence.
[14:08:51] note      rendering these as very far past would place them where she puts things that do not exist.
[14:08:51] c2        REFUSED
[14:08:51] c4        REFUSED[14:08:51] c1        "there used to be many people"
[14:08:51] c3        "there are not many people"
[14:08:51] lex       people: <aha-ktha>
[14:08:51] lex       many: <ktha1> free form, beyond count
[14:08:51] lex       no longer the case: <ontu>, sheet 6/B, finished and over
[14:08:51] build     c1  "aha-ktha ktha, ontu"
[14:08:51] build     c3  "aha-ktha ktha ne"
[14:08:51] check     both clauses now sit adjacent with no interval between them[14:08:51] REASONING essential material deleted, entering trace
[14:08:51] trace     the two assertions are a change of state. The dates were the only thing separating them and both dates are gone.
[14:08:51] trace     as rendered, the pair says: there were many, it is over, there are not many. She will place this inside her own
[14:08:51] trace     counted range, because that is the only range she has.
[14:08:51] trace     the nearest thing she holds to a group that was many and is now few is at sheet 4/A panel 5 and sheet 16/D, where figures behind the living ones are drawn faint and are named with the death form.
[14:08:51] trace     she will hear a group that has died.
[14:08:51] trace     the interval was 110 years. Her forward range is 2 cold
[14:08:51] trace     cycles and her backward range is 6. There is no scaling that puts 110 years inside either without falsifying it.
[14:08:51] trace     options considered:
[14:08:51] trace     render both, no interval        -> asserts a recent death
[14:08:51] trace     render c1 alone                 -> asserts they lived
[14:08:51] trace     render c3 alone                 -> asserts they are gone
[14:08:51] trace     render with <aha-ke ne>          -> gate 3, sense collision
[14:08:51] trace     refuse the pair                 -> says nothing
[14:08:51] trace     every rendering that says anything says something false.
[14:08:51] decision  REFUSE THE PAIR[14:08:51] spk0      no output
[14:08:51] warn      4 of 4 clauses refused. Nothing was said.

Vlad

typing a prompt in the computer, to count the cycles in unary

Computer

a hundred and ten cold-cycle glyphs coming up in one row and scrolling across the lit face, one glyph every four tenths of a second, no grouping and no break, forty-four seconds to the end of the row

Nika

watching the glyphs go past on the lit face and putting her finger down on the surface at the first one, keeping it there while the rest run under it Ktha. when it stops Karnu-ktha. Karnu-ktha, ktha-ke. lifting her hand off the machine and pointing back at the dishes of coloured oxide, then shaking her head Aha-ke ne. holding one hand at the far left of the lit face and the other at the far right, and moving the left one closed and the right one open Aha-ktha ontu, aha-ktha ontu ne. Ha.

mandrin-k runtime 14:09:14

wrap

[14:09:14] in        [operator console] render interval 1860-1970 as cold cycles, unary, lit face, no text
[14:09:14] param     output.channel = display
[14:09:14] param     output.encoding = unary
[14:09:14] calc      interval 110 years
[14:09:14] calc      cold cycle = 1 year, sheet 16/C frame 4
[14:09:14] calc      110 units
[14:09:14] gate 2    unbounded time: check
[14:09:14] note      gate 2 governs the spoken channel. Sheet 16/B frames 7 and 8 were drawn bands and returned <ktha1>, a quantity.
[14:09:14] note      the display channel was not blocked at stage 6.
[14:09:14] gate 2    PASS for display, spoken channel remains blocked
[14:09:14] draw      110 strokes, single row, uniform spacing
[14:09:14] draw      row width 3,300 px, lit face width 1,920 px, overflow
[14:09:14] draw      rewrapping: 10 rows of 11
[14:09:14] warn      grouping is an operator convention. Sheet 16/B used one unbroken row and no grouped presentation was elicited.
[14:09:14] draw      reverting to single row, horizontal scroll, no grouping
[14:09:14] draw      cold-cycle glyph substituted for stroke, sheet 16/C frame 4
[14:09:14] draw      rendering 110 glyphs, scroll rate 1 glyph / 0.4s
[14:09:14] display   running 44.0s[14:09:58] display   complete
[14:09:58] mic0      capture 12.6s, speaker voice, snr 20 dB
[14:09:58] asr       k: ktha
[14:09:58] asr       k: karnu-ktha
[14:09:59] asr       k: karnu-ktha, ktha-ke
[14:09:59] asr       k: aha-ke ne
[14:09:59] asr       k: aha-ktha ontu, aha-ktha ontu ne
[14:10:00] asr       k: ha
[14:10:00] seg       6 utterances, contour: level, level, level, level, level, level
[14:10:00] tag       utterance 2: karnu-ktha, cold cycles, plural, sheet 16/C
[14:10:00] tag       utterance 3: ktha-ke, beyond counting, sheet 18/B stage 5
[14:10:00] tag       utterance 4: <ne> short, wrong, following the death form
[14:10:00] note      the display carried the interval. The spoken channel had refused the same content at 14:08:51.nika      "Many."
nika      "Cold cycles."
nika      "Cold cycles, beyond counting."
nika      "A dead person, no."
nika      "The people are finished, the people are not finished."
nika      "Yes."[14:10:00] note      utterance 5 places both prior assertions in one utterance with the interval between them supplied from the display.

Cold cycles#

Vlad

I am twenty-five years old.

Computer

Aha na karnu-ktha. twenty-five cold-cycle glyphs running across the lit face in a single row, no grouping

Nika

following the twenty-five glyphs across with her finger and stopping when they stop Karnu-ktha. Ovlat na karnu-ktha ta. touching her own chest Aha na karnu-ktha. holding her hands a little closer together than for his Ktha-me. pointing out at the cave mouth and then back at herself Aha-ktha na, aha ne.

mandrin-k runtime 14:10:22

wrap

[14:10:22] in        "I am twenty-five years old."
[14:10:22] asr       en, peg chain position 2
[14:10:22] seg       1 clause, first person + age
[14:10:22] calc      25 years = 25 cold cycles
[14:10:22] gate 2    unbounded time: check
[14:10:22] note      spoken counted range backward is 6 cold cycles, stage 6
[14:10:22] note      25 exceeds it display channel cleared at 14:09:14 and carried 110
[14:10:22] route     splitting across channels
[14:10:22] route     spoken: first person, age frame, count withheld
[14:10:22] route     display: 25 cold-cycle glyphs, single row, no grouping
[14:10:22] lex       age: no entry
[14:10:22] note      sheet 16/D drew a short figure, a tall figure, an older
[14:10:22] note      figure and figures behind them, and the speaker read the row as distance. Age is available as position in that row.
[14:10:22] fall      rendering as cold cycles standing behind the person
[14:10:22] build     "aha na karnu-ktha"
[14:10:22] out       "aha na karnu-ktha"
[14:10:22] spk0      playing 1.1s
[14:10:22] draw      25 glyphs, scroll rate 1 glyph / 0.4s
[14:10:22] display   running 10.0s[14:10:36] mic0      capture 9.4s, speaker voice, snr 20 dB
[14:10:36] asr       k: karnu-ktha
[14:10:36] asr       k: ovlat na karnu-ktha ta
[14:10:37] asr       k: aha na karnu-ktha
[14:10:37] asr       k: ktha-me
[14:10:37] asr       k: aha-ktha na, aha ne
[14:10:38] seg       5 utterances, contour: level, level, level, level, level
[14:10:38] tag       utterance 4: ktha-me, quantity with diminutive, first occurrence of this combination
[14:10:38] deriv     ktha1 + -me, lesser of a pair, sheet 5
[14:10:38] lex       writing entry: ktha-me, fewer, the smaller quantity
[14:10:38] lex       lexicon 47 -> 48 typesnika      "Cold cycles."
nika      "Vlad is those cold cycles."
nika      "I am cold cycles."
nika      "Fewer."
nika      "The people, yes. Me, no."[14:10:38] note      utterance 5 separates two referents against the same count and gives opposite polarity for each.

Vlad

How old are you?

Computer

one glyph on the lit face, then a blank row, then the interrogative mark Semek na karnu-ktha na? the blank row held open under it

Nika

putting her finger on the blank row and pressing it nineteen times, spaced out, watching the glyphs come up under her hand Karnu-ktha. Aha na karnu-ktha ta. Ha. taking her hand off and pointing out at the cave mouth Karnu-ktha ne, aha-ktha na semek. Aha ne semek karnu-ktha aha.

mandrin-k runtime 14:10:56

wrap

[14:10:56] in        "How old are you?"
[14:10:56] asr       en, peg chain position 2
[14:10:56] map       interrogative of quantity, second person, age frame
[14:10:56] gate 1    check
[14:10:56] note      sheet 5 recovered one interrogative contour and sheet 15/E
[14:10:56] note      one source query. Neither asks for a quantity.
[14:10:56] note      the age frame at 14:10:22 was accepted and reused by the speaker in her own reply, so the frame is available.
[14:10:56] fall      polar question on the frame, count left open
[14:10:56] build     "semek na karnu-ktha na?"
[14:10:56] route     display: prompt row, cold-cycle glyph repeated, no count
[14:10:56] draw      one glyph, then a blank row, then the interrogative mark
[14:10:56] warn      the interrogative mark is an operator convention. Sheet 5
[14:10:56] warn      Part A established that the speaker does not read the marks and reads only the contour. Dropping it.
[14:10:56] draw      one glyph, then a blank row
[14:10:56] out       "semek na karnu-ktha na?"
[14:10:56] spk0      playing 1.3s
[14:10:56] display   running, blank row held[14:11:09] mic0      capture 10.8s, speaker voice, snr 20 dB
[14:11:09] in        put  touch events on lit face, 19 discrete
[14:11:09] draw      19 glyphs entered into the blank row by the speaker
[14:11:09] asr       k: karnu-ktha
[14:11:10] asr       k: aha na karnu-ktha ta
[14:11:10] asr       k: ha
[14:11:10] asr       k: karnu-ktha ne, aha-ktha na semek
[14:11:11] asr       k: aha ne semek karnu-ktha aha
[14:11:11] seg       5 utterances, contour: level, level, level, level, level
[14:11:11] note      the speaker used the display as an input channel. No sheet
[14:11:11] note      elicited this and no instruction was given.
[14:11:11] note      utterances 4 and 5 qualify the count she has just entered.nika      "Cold cycles."
nika      "I am those cold cycles."
nika      "Yes."
nika      "Not cold cycles, the people make the marks."
nika      "I have not seen my own cold cycles."[14:11:11] note      the count of 19 was entered by the speaker and attributed by her to the group rather than to her own observation.

Vlad

You are very young then, or maybe not.

Computer

Semek na karnu-ktha-me. Aha na karnu-ktha.

Nika

Ha. touching her own chest, then pointing at him Karnu-ktha-me, aha. Karnu-ktha, Ovlat. picking the spear shaft up off the gravel and putting the point back onto it Aha na rusa, aha na bontu ktha. pointing at the four marks on the wall, then at the fire Ovlat na miht-nu, miht-ktha semek. Karnu-ktha-me ne.

mandrin-k runtime 14:11:29

wrap

[14:11:29] in        "You are very young then, or maybe not."
[14:11:29] asr       en, peg chain position 2
[14:11:29] seg       2 clauses, assertion + retraction
[14:11:29] c1        "you are very young"
[14:11:29] lex       young: no entry
[14:11:29] note      sheet 16/D read a row of figures as distance and sheet 5
[14:11:29] note      established -me as the lesser of a pair. Fewer cold cycles is available. Youth as a property is not.
[14:11:29] fall      comparative on the count: her 19 against his 25
[14:11:29] build     "semek na karnu-ktha-me. aha na karnu-ktha"
[14:11:29] warn      this compares the two counts. It does not say young.
[14:11:29] c2        "or maybe not"
[14:11:29] gate 1    ungrounded referent: BLOCKED
[14:11:29] note      an assertion held at less than full strength. Sheet 5 Part
[14:11:29] note      B ran four polarities and all four are absolute. No hedge was elicited on any sheet.
[14:11:29] note      alternatives were refused at 14:04:29, 14:05:51, 14:06:24
[14:11:29] c2        REFUSED
[14:11:29] out       "semek na karnu-ktha-me. aha na karnu-ktha"
[14:11:29] spk0      playing 2.1s
[14:11:29] warn      1 of 2 clauses refused. The rendered portion states the comparison at full strength.[14:11:40] mic0      capture 11.9s, speaker voice, snr 20 dB
[14:11:40] asr       k: ha
[14:11:40] asr       k: karnu-ktha-me, aha
[14:11:41] asr       k: karnu-ktha, ovlat
[14:11:41] asr       k: aha na rusa, aha na bontu ktha
[14:11:41] asr       k: ovlat na miht-nu, miht-ktha semek
[14:11:42] asr       k: karnu-ktha-me ne
[14:11:42] seg       6 utterances, contour: level, level, level, level, level, level
[14:11:42] tag       utterance 6: <ne> short, wrong, on the comparative term
[14:11:42] note      the count is accepted at utterances 2 and 3 and the term
[14:11:42] note      built on it is rejected at utterance 6.nika      "Yes."
nika      "Fewer cold cycles, me."
nika      "Cold cycles, Vlad."
nika      "I take deer, I take bison, many."
nika      "Vlad is at a fire place, the marks of many days."
nika      "Fewer cold cycles is wrong."

The marks-makers#

Vlad

typing the definition of neocolonialism into the computer, and then showing images about imperialism In the future, many people like you have been subjected to this.

Computer

the six images left up on the lit face, nothing put on the speech channel and nothing drawn beside them

Nika

going through the images one at a time and stopping on the seated figure with the standing attendants Aha-ktha. putting one finger on the seated figure and moving it out to each of the standing ones, then back Aha ta na aha-ktha, aha-ktha na aha ta. looking at her own spear, then at the image again, then setting the spear down Naga-tu ne. Ne-e. moving to the image of the signed document and holding her hand above it Semek, aha ne. turning to him with the palm up Semek na?

mandrin-k runtime 14:12:04

wrap

[14:12:04] in        [operator console] define: neocolonialism
[14:12:04] mode      definition capture: ON
[14:12:04] parse     definition text 412 chars, en
[14:12:04] extract   head terms: control, state, economic, political, indirect, former, colony, sovereignty, pressure, capital
[14:12:04] gate 1    ungrounded referent: BLOCKED, 11 of 11 head terms
[14:12:04] note      state: no entry. Sheet 19/C lexicalised a group as people
[14:12:04] note      standing together. A group that persists past the people
[14:12:04] note      in it requires the type/token distinction refused at sheet
[14:12:04] note      10/D item 7.
[14:12:04] gate 3    BLOCKED
[14:12:04] note      sovereignty, capital, economic: no referent, no derivation
[14:12:04] note      path from any of the 48 attested stems
[14:12:04] gate 6    coinage without referent: REFUSED
[14:12:04] lex       no entry written. Lexicon unchanged at 48 types.[14:12:04] visual    6 images opened on lit face
[14:12:04] visual    colonial administrator seated, standing attendants
[14:12:04] visual    rubber collection, Congo Free State
[14:12:04] visual    survey party with theodolite, marked ground
[14:12:04] visual    cargo at a dock, loading
[14:12:04] visual    a signed document, seals
[14:12:04] visual    a map with drawn boundaries across terrain
[14:12:04] note      sheet 8/A returned terrain from a side view and refused an overhead view at panel 4. Image 6 is overhead.
[14:12:04] note      image 5 is marks on a surface with no figure. Sheet 15/E
[14:12:04] note      frame 3 returned a source query for that configuration.[14:12:09] in        "In the future, many people like you have been subjected to this."
[14:12:09] asr       en, peg chain position 2
[14:12:09] seg       1 clause + temporal frame + deictic object
[14:12:09] date      "in the future": unbounded forward
[14:12:09] gate 2    unbounded time: BLOCKED on spoken channel
[14:12:09] route     display channel available, as at 14:09:14
[14:12:09] calc      interval to the depicted period: ~39,900 cold cycles
[14:12:09] calc      at 0.4s per glyph, display duration 4h 26m
[14:12:09] calc      at 0.05s per glyph, display duration 33m 15s
[14:12:09] note      the 110-glyph row was read at 14:09:58 by tracking a single
[14:12:09] note      glyph while the rest scrolled. That method does not scale.
[14:12:09] note      sheet 16/B frames 7 and 8 returned <ktha1> for thirty
[14:12:09] note      drawn suns. Above thirty the system is mass.
[14:12:09] route     display refused for this interval
[14:12:09] gate 2    BLOCKED on both channels[14:12:09] c1        "many people like you have been subjected to this"
[14:12:09] lex       subjected to: no entry
[14:12:09] note      sheet 6 induced agent and patient and the patient marking is available. Sheet 15/B established that the speaker reads
[14:12:09] note      who gains from an assertion, so an interested party is available as a figure.
[14:12:09] note      "this" refers to the definition, which was refused, and to the six images, which have not been read.
[14:12:09] note      no referent resolved for the deictic
[14:12:09] c1        REFUSED[14:12:09] spk0      no output
[14:12:09] warn      all clauses refused. The images remain on the lit face.[14:12:31] mic0      capture 14.3s, speaker voice, snr 20 dB
[14:12:31] asr       k: aha-ktha
[14:12:31] asr       k: aha ta na aha-ktha, aha-ktha na aha ta
[14:12:32] asr       k: naga-tu ne
[14:12:32] asr       k: ne-e
[14:12:33] asr       k: semek, aha ne
[14:12:33] asr       k: semek na?
[14:12:33] seg       6 utterances, contour: level, level, level, level, level, rising
[14:12:33] tag       utterance 2: agent and patient slots filled in both orders
[14:12:33] tag       utterance 4: <ne> long, dangerous, sheet 19/D
[14:12:33] tag       utterance 6: source query frame, sheet 15/E frame 3
[14:12:33] note      the speaker has read the images without the accompanying
[14:12:33] note      text, which was refused before it reached her.nika      "People."
nika      "That one has the people, the people have that one."
nika      "Not with a spear."
nika      "Dangerous."
nika      "Marks, no figure."
nika      "Who says this?"

Vlad

bringing up a page of Wikipedia, of Marx, Arghiri Emmanuel, Yanis Varoufakis Those, and many others.

Computer

Aha-ktha ta na semek. Aha-ktha ktha.

Nika

looking at the three portraits, one after another, keeping her hands back Aha-ktha ta. Aha ne semek aha-ktha ta. pointing at the cave mouth, then back at the screen Aha-ktha ta na kar-nu na? Aha-ktha ta na naga-tu na? going back to the image of the signed document and putting the two side by side with a hand on each Aha-ktha ta na semek, semek ne semek. taking both hands back and pointing at herself, then at him Aha na semek Ovlat, Ovlat na semek aha. Ha.

mandrin-k runtime 14:12:52

wrap

[14:12:52] visual    3 pages opened on lit face
[14:12:52] visual    wikipedia en: Karl Marx
[14:12:52] visual    wikipedia en: Arghiri Emmanuel
[14:12:52] visual    wikipedia en: Yanis Varoufakis
[14:12:52] parse     each page: portrait or photograph, heading, running text[14:12:52] in        "Those, and many others."
[14:12:52] asr       en, peg chain position 2
[14:12:52] map       deictic plural + additive, answering a source query
[14:12:52] note      the open utterance at 14:12:33 is <semek na?>, sheet 15/E
[14:12:52] note      frame 3. The frame asks who stands beside a claim.
[14:12:52] gate 4    attribution: source supplied, check
[14:12:52] note      sheet 15/E frame 1 attached a claim to a figure the speaker had already named and frame 2 to a figure with its back
[14:12:52] note      turned. A named figure was accepted, an unnamed one was not.
[14:12:52] note      the three figures here are portraits with running text and the speaker has met none of them.
[14:12:52] note      sheet 15/D showed the same claim drawn by four figures with nothing new added and acceptance fell across the four.
[14:12:52] note      "and many others" adds figures without adding evidence and matches that panel.
[14:12:52] lex       many: <ktha1>
[14:12:52] lex       people: <aha-ktha>
[14:12:52] fall      naming the figures as people who make marks, count withheld
[14:12:52] build     "aha-ktha ta na semek. aha-ktha ktha"
[14:12:52] warn      this says: those people make marks, people beyond counting.
[14:12:52] warn      It does not say that they are right. Confidence 0.66.
[14:12:52] out       "aha-ktha ta na semek. aha-ktha ktha"
[14:12:52] spk0      playing 2.4s[14:13:08] mic0      capture 13.7s, speaker voice, snr 19 dB
[14:13:08] asr       k: aha-ktha ta
[14:13:08] asr       k: aha ne semek aha-ktha ta
[14:13:09] asr       k: aha-ktha ta na kar-nu na?
[14:13:09] asr       k: aha-ktha ta na naga-tu na?
[14:13:10] asr       k: aha-ktha ta na semek, semek ne semek
[14:13:10] asr       k: aha na semek ovlat, ovlat na semek aha. ha
[14:13:11] seg       6 utterances, contour: level, level, rising, rising, level, level
[14:13:11] note      utterances 3 and 4 are polar questions about the figures
[14:13:11] note      utterance 5 repeats the stem across the relator, as at 14:04:07 utterance 4, and adds a negation between the twonika      "Those people."
nika      "I have not seen those people."
nika      "Are those people at the high ground?"
nika      "Do those people carry a spear?"
nika      "Those people make marks, marks do not make marks."
nika      "I make marks at Vlad, Vlad makes marks at me. Yes."[14:13:11] note      utterance 5 pairs the figures against the document image from 14:12:33, which she named as marks with no figure.

Vlad

Those people are at high ground yes. They can be defined as intellectuals. Most intellectuals don't carry spears. Some intellectuals did. Example could be Che Guevara, although he is more of a militant than intellectual.

Computer

Ha. Aha-ktha ta na kar-nu. Aha-ktha ta na semek-ktha. Semek-ktha na naga-tu ne.

Nika

Semek-ktha. Ha, semek-ktha. pointing at him Ovlat na semek-ktha. pointing at the three portraits and then at the fire he has kept going Semek-ktha na naga-tu ne, semek-ktha na rusa ne. bringing the colonial image back up beside them and putting a hand on each Semek-ktha na aha-ktha, aha-ktha na semek-ktha. taking her hands back Torna ta na semek-ktha na?

mandrin-k runtime 14:13:34

wrap

[14:13:34] in        "Those people are at high ground yes. They can be defined as intellectuals. Most intellectuals don't carry spears. Some intellectuals did. Example could be Che Guevara, although he is more of a militant than intellectual."
[14:13:34] asr       en, peg chain position 2
[14:13:34] seg       6 clauses[14:13:34] c1        "those people are at high ground, yes"
[14:13:34] note      answers her question at 14:13:09, utterance 3
[14:13:34] warn      <kar-nu> was elicited from this ridge, sheet 17/F. As rendered this places the three figures on ground she can walk to. No other locative is attested.
[14:13:34] c1        out "ha. aha-ktha ta na kar-nu"[14:13:34] c2        "they can be defined as intellectuals"
[14:13:34] lex       intellectual: no entry
[14:13:34] gate 1    ungrounded referent: check
[14:13:34] note      no agent-noun derivation, refused at 14:03:31 and 14:03:58
[14:13:34] note      <semek> was accepted for the operator as the marks-maker and she has used it of him five times since
[14:13:34] fall      reusing <semek> in the marks-maker sense, plural
[14:13:34] c2        out "aha-ktha ta na semek-ktha"
[14:13:34] deriv     semek-ktha, model-produced, marked
[14:13:34] warn      this makes them the same kind of person as you. Confidence
[14:13:34] warn      0.58.[14:13:34] c3        "most intellectuals don't carry spears"
[14:13:34] lex       most: no entry
[14:13:34] note      sheet 16/B counted to seven and returned <ktha1> above
[14:13:34] note      thirty. A proportion of a set was never elicited.
[14:13:34] fall      dropping the quantifier, keeping the polarity
[14:13:34] c3        out "semek-ktha na naga-tu ne"
[14:13:34] warn      this says none of them carry a spear.[14:13:34] c4        "some intellectuals did"
[14:13:34] gate 1    BLOCKED
[14:13:34] note      the same quantifier gap, now carrying the whole clause.
[14:13:34] note      with c3 rendered absolute, c4 reads as a reversal of it rather than a subset of it.
[14:13:34] c4        REFUSED[14:13:34] c5        "example could be Che Guevara"
[14:13:34] lex       proper noun, not in lexicon
[14:13:34] note      sheet 7 frame 5 retained a proper noun on positional
[14:13:34] note      evidence with a figure present to attach it to no portrait open on the lit face for this figure
[14:13:34] note      "could be": hedged, refused at 14:11:29
[14:13:34] c5        REFUSED[14:13:34] c6        "more of a militant than an intellectual"
[14:13:34] gate 1    BLOCKED
[14:13:34] note      two agent nouns, neither available, ranked against each other on a scale. No comparative of degree is attested.
[14:13:34] note      the comparative at 14:11:29 ran on a count and was rejected
[14:13:34] c6        REFUSED
[14:13:34] out       "ha. aha-ktha ta na kar-nu. aha-ktha ta na semek-ktha. semek-ktha na naga-tu ne"
[14:13:34] spk0      playing 4.6s
[14:13:34] warn      3 of 6 clauses refused. The rendered portion places the three
[14:13:34] warn      figures on this ridge, makes them marks-makers, and states that marks-makers carry no spear.[14:13:52] mic0      capture 12.1s, speaker voice, snr 19 dB
[14:13:52] asr       k: semek-ktha
[14:13:52] asr       k: ha, semek-ktha
[14:13:53] asr       k: ovlat na semek-ktha
[14:13:53] asr       k: semek-ktha na naga-tu ne, semek-ktha na rusa ne
[14:13:54] asr       k: semek-ktha na aha-ktha, aha-ktha na semek-ktha
[14:13:54] asr       k: torna ta na semek-ktha na?
[14:13:55] seg       6 utterances, contour: level, level, level, level, level, rising
[14:13:55] lex       semek-ktha accepted on first hearing, no correction
[14:13:55] lex       confidence 0.58 -> 0.85
[14:13:55] note      utterance 5 puts the two groups in both slots, as she did with the colonial image at 14:12:31nika      "Marks-makers."
nika      "Yes, marks-makers."
nika      "Vlad is a marks-maker."
nika      "Marks-makers carry no spear, marks-makers take no deer."
nika      "The marks-makers have the people, the people have the marks-makers."
nika      "Is there a bear at the marks-makers?"

Vlad

The people make marks-markers, Marks-markers don't have people. Bear no.

Computer

Aha-ktha na kesta-ne semek-ktha. Semek-ktha na aha-ktha ne. Torna ne.

Nika

Ne. repeating his first clause back to him with a nod Aha-ktha na kesta-ne semek-ktha, ha. repeating the second and then shaking her head Semek-ktha na aha-ktha ne, ne. saying the whole thing her way, both halves Aha-ktha na kesta-ne semek-ktha, semek-ktha na aha-ktha. picking the spear up off the gravel, turning it so the antler point catches the firelight, and holding it out across both palms Naga-tu, aha na kesta-ne. Naga-tu na aha. setting it down again between them Kesta-ne ta, aha ne kesta-ne ta ne. Torna ne, ha.

mandrin-k runtime 14:14:16

wrap

[14:14:16] in        "The people make marks-makers. Marks-makers don't have people. Bear no."
[14:14:16] asr       en, peg chain position 2
[14:14:16] seg       3 clauses
[14:14:16] note      this corrects her utterance 5 at 14:13:54, which filled both slots. You are keeping one direction and cancelling the other.[14:14:16] c1        "the people make marks-makers"
[14:14:16] lex       make: <kesta-ne>, joining, sheet 9/A frame 4
[14:14:16] warn      <kesta-ne> joins parts into a whole. Sheet 18 used it for courses of a wall. Applied to people it says the group
[14:14:16] warn      assembles them, which is the reading available.
[14:14:16] note      a second reading exists: sheet 3 strip 7, a tall figure and a short figure, which she named without <kesta-ne>. Bearing and building are separate in the corpus and this renders as building.
[14:14:16] c1        out "aha-ktha na kesta-ne semek-ktha"[14:14:16] c2        "marks-makers don't have people"
[14:14:16] lex       have: no entry
[14:14:16] note      her utterances at 14:12:31 and 14:13:54 used the bare agent and patient slots with no verb, and the model rendered that as having. The construction is hers.
[14:14:16] fall      same construction, negated
[14:14:16] c2        out "semek-ktha na aha-ktha ne"
[14:14:16] warn      <ne> short, wrong, is being used to cancel a relation. Sheet
[14:14:16] warn      19/D placed the short form on mistakes and the long form on danger. Cancelling a relation was not elicited on either.
[14:14:16] warn      short form selected. Confidence 0.54.[14:14:16] c3        "bear no"
[14:14:16] note      answers her question at 14:13:54, utterance 6
[14:14:16] c3        out "torna ne"
[14:14:16] warn      <ne> short here reads as a mistaken belief about bears
[14:14:16] warn      rather than an absence of bears. She asked with the polar
[14:14:16] warn      contour and the polar negative is the same form.
[14:14:16] out       "aha-ktha na kesta-ne semek-ktha. semek-ktha na aha-ktha ne. torna ne"
[14:14:16] spk0      playing 3.8s[14:14:33] mic0      capture 15.4s, speaker voice, snr 19 dB
[14:14:33] asr       k: ne
[14:14:33] asr       k: aha-ktha na kesta-ne semek-ktha, ha
[14:14:34] asr       k: semek-ktha na aha-ktha ne, ne
[14:14:34] asr       k: aha-ktha na kesta-ne semek-ktha, semek-ktha na aha-ktha
[14:14:35] asr       k: naga-tu, aha na kesta-ne. naga-tu na aha
[14:14:36] asr       k: kesta-ne ta, aha ne kesta-ne ta ne
[14:14:36] asr       k: torna ne, ha
[14:14:36] seg       7 utterances
[14:14:36] tag       utterance 1: <ne> short, utterance-initial, rejecting the turn
[14:14:36] tag       utterance 3: two instances of <ne>, both short
[14:14:36] note      utterance 5 introduces her spear as the test case
[14:14:36] note      utterance 6 has a double negation on <kesta-ne>, first occurrence in the sessionnika      "Wrong."
nika      "The people build the marks-makers, yes."
nika      "Marks-makers do not have the people, wrong."
nika      "The people build the marks-makers, the marks-makers have the people."
nika      "A spear, I build it. The spear has me."
nika      "The built thing, it is not that I did not build it."
nika      "Bear, no. Yes."[14:14:36] note      utterance 6 does not resolve to a single reading. The model is passing it through with both negations in place.

The spear has me#

Vlad

So, the spear you build, it has you. Right. I'll consider that true.

Computer

Naga-tu, semek na kesta-ne. Naga-tu na semek. Ha.

Nika

Ha. taking the spear back up and holding it across her knees again Naga-tu na aha, ha. pointing at the machine Semek-tu na Ovlat na? keeping her finger there and saying it a second way Kruf ta na Ovlat na? moving the finger off the machine and onto the fire pit, and then out at the walls of the cave Miht-nu ta na Ovlat na?

mandrin-k runtime 14:14:57

wrap

[14:14:57] in        "So, the spear you build, it has you. Right. I'll consider that true."
[14:14:57] asr       en, peg chain position 2
[14:14:57] seg       3 clauses[14:14:57] c1        "the spear you build, it has you"
[14:14:57] note      restating her utterance 5 at 14:14:35 with the person
[14:14:57] note      shifted from first to second
[14:14:57] lex       naga-tu, kesta-ne, both attested this turn
[14:14:57] c1        out "naga-tu, semek na kesta-ne. naga-tu na semek"[14:14:57] c2        "right"
[14:14:57] c2        out "ha"[14:14:57] c3        "I'll consider that true"
[14:14:57] gate 1    ungrounded referent: check
[14:14:57] note      consider: holding a proposition at a chosen strength. Sheet
[14:14:57] note      5 Part B ran four polarities and all four are absolute. The same gap refused the hedge at 14:11:29.
[14:14:57] note      true: no entry. Sheet 15 ran the whole reliability battery and the speaker sorted assertions by who gains and by what is visible. She never produced a term for the property.
[14:14:57] note      <ha> is the affirmative and carries acceptance of a turn.
[14:14:57] note      <ne> long and short carry rejection. There is no third term
[14:14:57] note      between them.
[14:14:57] fall      the affirmative alone, already emitted at c2
[14:14:57] note      emitting it twice would read as two acceptances
[14:14:57] c3        REFUSED
[14:14:57] out       "naga-tu, semek na kesta-ne. naga-tu na semek. ha"
[14:14:57] spk0      playing 3.3s
[14:14:57] warn      1 of 3 clauses refused. The rendered portion restates her claim and agrees with it. Agreement and consideration are the same form.[14:15:12] mic0      capture 10.6s, speaker voice, snr 19 dB
[14:15:12] asr       k: ha
[14:15:12] asr       k: naga-tu na aha, ha
[14:15:13] asr       k: semek-tu na ovlat na?
[14:15:13] asr       k: kruf ta na ovlat na?
[14:15:14] asr       k: miht-nu ta na ovlat na?
[14:15:14] seg       5 utterances, contour: level, level, rising, rising, rising
[14:15:14] note      utterances 3, 4 and 5 run the same frame over three objects
[14:15:14] note      <semek-tu> was accepted at 14:07:28, <kruf ta> at 14:04:55, <miht-nu> derived by the speaker at sheet 17/F frame 3nika      "Yes."
nika      "The spear has me, yes."
nika      "Does the marks-instrument have Vlad?"
nika      "Does that stone have Vlad?"
nika      "Does this fire place have Vlad?"

Vlad

Yes.

Computer

Ha. Semek-tu na aha. Kruf ta na aha. Miht-nu ta na aha.

Nika

Ha. holding the spear up in one hand and pointing at the machine with the other Aha na naga-tu, Ovlat na semek-tu. lowering the spear and running her thumb along the binding at the point Aha-ktha na kesta-ne naga-tu. looking at the machine and leaving a pause where an answer would go Aha-ktha ne kesta-ne semek-tu, Ovlat na kesta-ne semek-tu na? Ovlat ne kesta-ne semek-tu. bringing up the image of the seven dishes again, and the wafer beside it Semek-tu na kesta-ktha-nu, ha. Semek-tu na Ovlat, kesta-ktha-nu na Ovlat.

mandrin-k runtime 14:15:31

wrap

[14:15:31] in        "Yes."
[14:15:31] asr       en, peg chain position 2
[14:15:31] map       affirmative, answering three polar questions
[14:15:31] note      the three questions at 14:15:13 and 14:15:14 were put with separate contours and are three turns, not one a single <ha> attaches to the last of them by position
[14:15:31] fall      emitting the affirmative once per question, with each object
[14:15:31] fall      named, so the scope is carried
[14:15:31] build     "ha. semek-tu na aha. kruf ta na aha. miht-nu ta na aha"
[14:15:31] warn      first person substituted throughout, since you are the speaker. She asked using your name and the answer returns it in the first person.
[14:15:31] out       "ha. semek-tu na aha. kruf ta na aha. miht-nu ta na aha"
[14:15:31] spk0      playing 4.1s[14:15:44] mic0      capture 16.8s, speaker voice, snr 19 dB
[14:15:44] asr       k: ha
[14:15:44] asr       k: aha na naga-tu, ovlat na semek-tu
[14:15:45] asr       k: aha-ktha na kesta-ne naga-tu
[14:15:45] asr       k: aha-ktha ne kesta-ne semek-tu, ovlat na kesta-ne semek-tu na?
[14:15:46] asr       k: ovlat ne kesta-ne semek-tu
[14:15:47] asr       k: semek-tu na kesta-ktha-nu, ha
[14:15:47] asr       k: semek-tu na ovlat, kesta-ktha-nu na ovlat
[14:15:48] seg       7 utterances, contour: level, level, level, rising, level, level, level
[14:15:48] lex       kesta-ktha-nu: entry written 14:08:32, factory
[14:15:48] note      utterance 4 asks and utterance 5 answers it in the same turn
[14:15:48] note      utterance 7 carries the relation from 14:14:35 out to the factory, which she has not seen and has from one imagenika      "Yes."
nika      "I have the spear, Vlad has the marks-instrument."
nika      "The people build the spear."
nika      "The people did not build the marks-instrument. Did Vlad build the marks-instrument?"
nika      "Vlad did not build the marks-instrument."
nika      "The marks-instrument is from a many-knappings place, yes."
nika      "The marks-instrument has Vlad, the many-knappings place has Vlad."

Vlad

Yes.

Computer

Ha.

Nika

Ha. bringing the colonial image back up and setting it beside the dishes and the wafer, so the three sit in a row Aha-ktha ta na kesta-ktha-nu. putting one finger on the seated figure and moving it out to the standing ones again Aha ta na aha-ktha, aha-ktha na aha ta. moving the finger off the image and onto the machine, then onto him Kesta-ktha-nu na Ovlat, Ovlat na kesta-ktha-nu ne. Ovlat na aha ta ne. Ovlat na aha-ktha. taking her hand back and putting it on the spear Ne-e.

mandrin-k runtime 14:16:04

wrap

[14:16:04] in        "Yes."
[14:16:04] asr       en, peg chain position 2
[14:16:04] map       affirmative, answering a two-part assertion
[14:16:04] note      her utterance 7 at 14:15:47 put two relations side by side and neither carried the interrogative contour. This accepts
[14:16:04] note      an assertion rather than answering a question.
[14:16:04] note      <ha> alone is available and was accepted in that role at 14:14:36 utterance 7
[14:16:04] out       "ha"
[14:16:04] spk0      playing 0.3s[14:16:15] mic0      capture 18.2s, speaker voice, snr 19 dB
[14:16:15] asr       k: ha
[14:16:15] asr       k: aha-ktha ta na kesta-ktha-nu
[14:16:16] asr       k: aha ta na aha-ktha, aha-ktha na aha ta
[14:16:17] asr       k: kesta-ktha-nu na ovlat, ovlat na kesta-ktha-nu ne
[14:16:17] asr       k: ovlat na aha ta ne
[14:16:18] asr       k: ovlat na aha-ktha
[14:16:19] asr       k: ne-e
[14:16:19] seg       7 utterances, contour: level, level, level, level, level, level, level
[14:16:19] tag       utterance 3 reproduces her own construction from 14:12:31, which was produced against the colonial image
[14:16:19] tag       utterance 7: <ne> long, dangerous, sheet 19/D
[14:16:19] note      utterances 4 to 6 place the operator in the lower slot of the same construction across three successive turnsnika      "Yes."
nika      "Those people are at the many-knappings place."
nika      "That one has the people, the people have that one."
nika      "The many-knappings place has Vlad, Vlad does not have the many-knappings place."
nika      "Vlad is not that one."
nika      "Vlad is the people."
nika      "Dangerous."

Vlad

Vlad not dangerous.

Computer

Ovlat ne-e ne.

Nika

Ne. repeating the machine's string back and putting her hand on the gravel Ovlat ne-e ne, ne. Aha ne semek Ovlat ne-e. pointing past him at the row of images still open on the lit face Ne-e ta, kesta-ktha-nu ta. pointing at the fire, then at the marks on the wall, then at him Ovlat na miht-nu, Ovlat na semek, Ovlat ne-e ne. Ha.

mandrin-k runtime 14:16:37

wrap

[14:16:37] in        "Vlad not dangerous."
[14:16:37] asr       en, peg chain position 2
[14:16:37] parse     fragment, proper noun + negation + predicate
[14:16:37] lex       ne [long] = dangerous, sheet 19/D right column
[14:16:37] lex       ne [short] = wrong, mistaken, sheet 19/D left column
[14:16:37] note      negating the long form requires both forms in one utterance and that sequence has no occurrence in 388 turns
[14:16:37] gate 1    check
[14:16:37] note      sheet 19/D paired a wrong thing against a dangerous thing in five frames and the speaker gave a different form for each.
[14:16:37] note      the sheet established two terms. It did not establish that either can take the other as its object.
[14:16:37] fall      candidate 1: "ovlat ne-e ne"
[14:16:37] fall      candidate 2: "ovlat ne ne-e"
[14:16:37] fall      candidate 3: "ovlat ne-e" with negative contour
[14:16:37] note      sheet 5/C found polarity marked on both channels, so the contour cannot carry a negation the segment contradicts
[14:16:37] fall      candidate 3 dropped
[14:16:37] note      candidates 1 and 2 differ by order alone and the model has no held-out data separating them
[14:16:37] select    candidate 1, 0.51
[14:16:37] warn      this may render as: Vlad is dangerously wrong.
[14:16:37] out       "ovlat ne-e ne"
[14:16:37] spk0      playing 1.2s[14:16:48] mic0      capture 13.9s, speaker voice, snr 19 dB
[14:16:48] asr       k: ne
[14:16:48] asr       k: ovlat ne-e ne, ne
[14:16:49] asr       k: aha ne semek ovlat ne-e
[14:16:50] asr       k: ne-e ta, kesta-ktha-nu ta
[14:16:50] asr       k: ovlat na miht-nu, ovlat na semek, ovlat ne-e ne
[14:16:51] asr       k: ha
[14:16:51] seg       6 utterances
[14:16:51] tag       utterance 2 repeats the emitted string and marks it wrong
[14:16:51] tag       utterance 5 reproduces the same string and marks it accepted
[14:16:51] note      the string is rejected at utterance 2 and produced at
[14:16:51] note      utterance 5 with three clauses in front of it candidate 1 confidence 0.51 -> 0.72nika      "Wrong."
nika      "Vlad is not dangerous, wrong."
nika      "I have not seen Vlad be dangerous."
nika      "The danger is there, at the many-knappings place."
nika      "Vlad is at a fire place, Vlad makes marks, Vlad is not dangerous."
nika      "Yes."

A person alone is wrong#

Vlad

You have sympathy for me. That is why you stay here.

Computer

nothing put on the speech channel, both clauses refused

Nika

waiting through the silence, then reaching over and pressing one finger against the edge of the machine, twice, the way she pressed the glyphs into the blank row Semek na? when nothing comes Kruf ta na semek ne. taking the finger back, putting a piece of thick wood on the fire, and staying where she is

mandrin-k runtime 14:17:09

wrap

[14:17:09] in        "You have sympathy for me. That is why you stay here."
[14:17:09] asr       en, peg chain position 2
[14:17:09] seg       2 clauses[14:17:09] c1        "you have sympathy for me"
[14:17:09] lex       sympathy: no entry
[14:17:09] gate 1    ungrounded referent: check
[14:17:09] note      sheet 4/A lexicalised 8 internal states. Panel 7 drew a figure with a hand on a wound and panel 10 drew two figures with hands joined. Neither returned a state held by one person about the condition of another.
[14:17:09] note      <kamla> is grief, panel 2, and it is held about a loss the holder has sustained
[14:17:09] note      <aha-ke> sense 2 is being cut off from the group and she applied it to the operator at 14:06:02
[14:17:09] fall      rendering as her statement. The state she has is dropped.
[14:17:09] c1        REFUSED[14:17:09] c2        "that is why you stay here"
[14:17:09] gate 1    BLOCKED
[14:17:09] note      causal interrogative refused at 14:05:51 and the causal
[14:17:09] note      assertion has the same gap. No connective linking a state to an action is attested.
[14:17:09] note      stay: no stative, as at 14:05:51
[14:17:09] c2        REFUSED[14:17:09] spk0      no output
[14:17:09] warn      2 of 2 clauses refused.
[14:17:09] note      the components are separately available: <semek na kar-nu
[14:17:09] note      ta> asks whether she is here, and <aha-ke ne> is what she said at 14:06:02. Joining them is what fails.

Vlad

activating autodraw mode and pressing on the rereasoning button

Computer

a strip of four panels composing on the lit face and coming up in one row: a figure alone at a fire, a cave opening, four marks on the wall beside it; a clawed animal on the slope outside the opening; a second figure at the opening with a shaft in hand; both figures at the fire with the shaft on the ground

Nika

going along the four panels with one finger and stopping on each Ha. at panel 1, then panel 2, then tapping her own chest at panel 3 Aha-ke, torna ta, aha. at panel 4 Ha, ta. pointing at the wall marks Aha ne semek Ovlat, miht-ktha. pointing at the cave mouth and the light going off the ridge Aha na semek Ovlat, miht ta. Aha-ke ne, ha. putting the spear down beside the machine and leaving it there Kamla ne.

mandrin-k runtime 14:17:34

wrap

[14:17:34] param     runtime.autodraw = ON
[14:17:34] note      the model may compose panels on the lit face from the sheet
[14:17:34] note      inventory when the spoken channel refuses. Panels are built from drawings the speaker has already read.
[14:17:34] note      the display channel carried the interval at 14:09:14 and the count at 14:10:22, and the speaker used it as an input
[14:17:34] note      channel at 14:11:09 without instruction.
[14:17:34] cmd       re-reason: 14:17:09[14:17:34] REASONING re-entering trace on 2 refused clauses
[14:17:34] trace     c1 refused for want of a state held about another person.
[14:17:34] trace     checking the corpus for that state expressed as an action
[14:17:34] trace     rather than as a term.
[14:17:34] trace     sheet 3 strip 10: a figure lying down, eyes open, others
[14:17:34] trace     standing around it. She named the standing figures.
[14:17:34] trace     sheet 3 strip 15: an object passed from one hand to another.
[14:17:34] trace     sheet 4/A panel 5: a figure alone, others far off. She put
[14:17:34] trace     her hand over the panel and said <aha-ke> three times.
[14:17:34] trace     14:06:02: she said <aha-ke ne> in this cave, unprompted, after naming the bears on this ridge.
[14:17:34] trace     the state is not lexicalised. The configuration is: a person
[14:17:34] trace     alone, a danger named, a second person present.
[14:17:34] trace     that configuration can be drawn.
[14:17:34] trace     c2 refused for want of a causal connective. Sheet 3 ran
[14:17:34] trace     connected strips and she read them left to right and gave
[14:17:34] trace     back connected speech. Sequence carries what the connective
[14:17:34] trace     would carry, at lower precision.
[14:17:34] decision  ROUTE TO DISPLAY, sequence in place of cause[14:17:35] draw      composing strip, 4 panels, sheet inventory only
[14:17:35] draw      panel 1  figure alone at a fire, cave opening, four marks
[14:17:35] draw      on the wall beside it              [sheet 3, 8]
[14:17:35] draw      panel 2  clawed animal on a slope outside the opening
[14:17:35] draw      [sheet 1, 3]
[14:17:35] draw      panel 3  second figure at the opening, shaft in hand
[14:17:35] draw      [sheet 20/D]
[14:17:35] draw      panel 4  both figures at the fire, shaft on the ground
[14:17:35] draw      [sheet 3, 14]
[14:17:35] warn      panel 4 is the operator's proposition. It is drawn as a state and she may read it as an instruction.
[14:17:35] draw      rendering strip, 1,840 px, single row, no scroll
[14:17:35] display   running[14:17:52] mic0      capture 15.7s, speaker voice, snr 19 dB
[14:17:52] asr       k: ha
[14:17:52] asr       k: aha-ke, torna ta, aha
[14:17:53] asr       k: ha, ta
[14:17:53] asr       k: aha ne semek ovlat, miht-ktha
[14:17:54] asr       k: aha na semek ovlat, miht ta
[14:17:55] asr       k: aha-ke ne, ha
[14:17:55] asr       k: kamla ne
[14:17:56] seg       7 utterances, contour: level, level, level, level, level, level, level
[14:17:56] tag       utterance 7: <kamla>, grief, sheet 4/A panel 2, negated short
[14:17:56] note      first occurrence of kamla in the session outside sheet 4
[14:17:56] note      utterances 4 and 5 contrast the days before this one with this one, using the same forms she used at 14:05:32nika      "Yes."
nika      "A person alone, bears here, me."
nika      "Yes, that."
nika      "I had not seen Vlad, many days."
nika      "I saw Vlad, this sun."
nika      "A person alone is wrong, yes."
nika      "It is not grief."[14:17:56] note      utterance 7 was produced without a panel prompting it and names a state she is ruling out.

Vlad

It is not grief. I understand. But you can't lead me to the group with the computer.

Computer

Kamla ne. Ha. Semek-tu na semek ne. Aha-ktha ta, semek-tu ne.

Nika

Ha. Aha-ktha ta, semek-tu ne, ha. picking up the machine by its edges, weighing it, and setting it back down Semek-tu na kruf, kruf ktha. pointing at him, then out through the cave mouth at the ridge Aha na kesta-ne Ovlat, Ovlat na kar-nu. drawing the two shapes in the gravel again, the small one and the large one, and putting a finger in each Aha-ktha na karnu-me, miht ta na miht ta. putting her hand back on the machine Semek-tu ta, aha-ktha ne semek. Ne-e na?

mandrin-k runtime 14:18:14

wrap

[14:18:14] in        "It is not grief. I understand. But you can't lead me to the group with the computer."
[14:18:14] asr       en, peg chain position 2
[14:18:14] seg       3 clauses[14:18:14] c1        "it is not grief"
[14:18:14] note      repeats her utterance 7 at 14:17:55
[14:18:14] lex       kamla, ne short
[14:18:14] c1        out "kamla ne"[14:18:14] c2        "I understand"
[14:18:14] lex       understand: no entry
[14:18:14] gate 1    ungrounded referent: check
[14:18:14] note      sheet 4 lexicalised 8 states, all of them visible on a face or a posture. Comprehension appears on none of the panels.
[14:18:14] note      she has marked acceptance of a turn with <ha> throughout, 88 occurrences, and has used it after repeating a string
[14:18:14] note      back correctly at 14:14:33 and 14:16:48.
[14:18:14] fall      affirmative alone
[14:18:14] c2        out "ha"
[14:18:14] warn      this marks agreement. Comprehension and agreement share
[14:18:14] warn      one form here. Confidence 0.69.[14:18:14] c3        "you can't lead me to the group with the computer"
[14:18:14] lex       lead: sheet 8/A panel 6 and 7 gave figures walking away from and toward an opening, with footprints. Two figures walking
[14:18:14] lex       one behind the other is sheet 15/B strip 5, which she read.
[14:18:14] lex       group: <aha-ktha>
[14:18:14] lex       computer: <kruf ta>, <semek-tu>
[14:18:14] lex       can: no modal system, as at 14:07:18
[14:18:14] note      sheet 16 gave counted ranges and sheet 6 gave what is finished. Ability was elicited on no sheet.
[14:18:14] note      the clause states a limit on an action rather than an action.
[14:18:14] note      the corpus records actions taken and things present.
[14:18:14] fall      candidate 1: render the walking, negate it
[14:18:14] fall      "semek na aha-ktha, aha na semek-tu, ne"
[14:18:14] warn      candidate 1 places <ne> short at the end of a three-part
[14:18:14] warn      utterance. Her use of that form ends utterances at 14:11:42, 14:14:34 and 14:16:48, and in each case it marks the whole
[14:18:14] warn      utterance as wrong.
[14:18:14] warn      as emitted this reads as a correction of the walking rather than a limit on the machine.
[14:18:14] fall      candidate 2: render the machine failing, as at 14:04:55
[14:18:14] fall      "semek-tu na semek ne. aha-ktha ta, semek-tu ne"
[14:18:14] select    candidate 2, 0.47
[14:18:14] out      "kamla ne. ha. semek-tu na semek ne. aha-ktha ta, semek-tu ne"
[14:18:14] spk0      playing 4.4s
[14:18:14] warn      the rendered portion says the marks-instrument speaks wrongly and that at the group there is no marks-instrument.[14:18:31] mic0      capture 17.1s, speaker voice, snr 19 dB
[14:18:31] asr       k: ha
[14:18:31] asr       k: aha-ktha ta, semek-tu ne, ha
[14:18:32] asr       k: semek-tu na kruf, kruf ktha
[14:18:33] asr       k: aha na kesta-ne ovlat, ovlat na kar-nu
[14:18:34] asr       k: aha-ktha na karnu-me, miht ta na miht ta
[14:18:35] asr       k: semek-tu ta, aha-ktha ne semek
[14:18:35] asr       k: ne-e na?
[14:18:36] seg       7 utterances, contour: level, level, level, level, level, level, rising
[14:18:36] tag       utterance 5: <miht ta na miht ta>, one sun to one sun
[14:18:36] tag       first occurrence of this construction
[14:18:36] deriv     sheet 16/A ran the sun across one turn of light and sheet
[14:18:36] deriv     16/B counted turns. This joins the two.
[14:18:36] lex       writing entry: miht ta na miht ta = one day's walk
[14:18:36] lex       lexicon 48 -> 49 typesnika      "Yes."
nika      "At the group there is no marks-instrument, yes."
nika      "The marks-instrument is stone, stones beyond counting."
nika      "I take Vlad, Vlad to the high ground."
nika      "The group is at the small high ground, one sun to one sun."
nika      "That marks-instrument, the group has not seen marks."
nika      "Is it dangerous?"

Vlad

Marks-instrument, not dangerous. Marks-instrument, complex, only I can use.

Computer

Semek-tu ne-e ne. Semek-tu na kruf-ktha, kruf-ktha-ke. Semek-tu, aha ta.

Nika

Ha. Semek-tu ne-e ne, ha. pointing at the machine, then at him Semek-tu, Ovlat ta. picking up her own spear and holding it against her chest Naga-tu, aha ta. pointing out at the cave mouth Aha-ktha na naga-tu, aha-ktha ne semek-tu. putting the spear back down and looking at the machine for a while Semek-tu na Ovlat ta, Ovlat ta na aha-ktha ne. Ovlat na kar-nu na?

mandrin-k runtime 14:18:54

wrap

[14:18:54] in        "Marks-instrument, not dangerous. Marks-instrument, complex, only I can use."
[14:18:54] asr       en, peg chain position 2
[14:18:54] seg       2 clauses, both fragments, both headed by the same nominal[14:18:54] c1        "marks-instrument, not dangerous"
[14:18:54] note      answers her question at 14:18:35
[14:18:54] lex       the ordering candidate selected at 14:16:37 was accepted by the speaker at 14:16:50, confidence 0.72
[14:18:54] c1        out "semek-tu ne-e ne"[14:18:54] c2        "marks-instrument, complex, only I can use"
[14:18:54] lex       complex: rendered as parts beyond count at 14:07:52 and she accepted the parts and rejected the material
[14:18:54] lex       only: no entry
[14:18:54] gate 1    ungrounded referent: check
[14:18:54] note      sheet 12/C ran one of many against one already chosen and she marked the chosen one with <ta>. Singling out a member of a set is available.
[14:18:54] note      sheet 12/C did not exclude the rest of the set. The marked
[14:18:54] note      animal was marked and the other six stood.
[14:18:54] lex       can use: ability, refused at 14:07:18 and 14:18:14
[14:18:54] note      sheet 11/A frame 1 exchanged her shaft for yours and she returned the same word for both, so a tool holds its name across users. Use restricted to one user runs against that.
[14:18:54] fall      candidate 1: possession, which she has produced 11 times
[14:18:54] fall      "semek-tu na aha", the marks-instrument has me
[14:18:54] warn      candidate 1 states the relation she put on the spear at 14:14:35 and on you at 14:16:17. It does not restrict use.
[14:18:54] fall      candidate 2: the marking from sheet 12/C, on the operator
[14:18:54] fall      "semek-tu, aha ta"
[14:18:54] select    candidate 2, 0.44
[14:18:54] out      "semek-tu ne-e ne. semek-tu na kruf-ktha, kruf-ktha-ke. semek-tu, aha ta"
[14:18:54] spk0      playing 4.9s
[14:18:54] warn      the last part says: the marks-instrument, this one, me. The restriction to a single user is carried by the marking alone.[14:19:12] mic0      capture 16.4s, speaker voice, snr 19 dB
[14:19:12] asr       k: ha
[14:19:12] asr       k: semek-tu ne-e ne, ha
[14:19:13] asr       k: semek-tu, ovlat ta
[14:19:13] asr       k: naga-tu, aha ta
[14:19:14] asr       k: aha-ktha na naga-tu, aha-ktha ne semek-tu
[14:19:15] asr       k: semek-tu na ovlat ta, ovlat ta na aha-ktha ne
[14:19:16] asr       k: ovlat na kar-nu na?
[14:19:16] seg       7 utterances, contour: level, level, level, level, level, level, rising
[14:19:16] note      utterance 3 returns the model's construction with the proper
[14:19:16] note      noun substituted for the first person
[14:19:16] note      utterance 4 builds the parallel on her own spear unprompted
[14:19:16] note      utterance 6 places the marked possessor in the lower slot of the construction from 14:16:17nika      "Yes."
nika      "The marks-instrument is not dangerous, yes."
nika      "The marks-instrument, that Vlad."
nika      "The spear, that me."
nika      "The people carry spears, the people have no marks-instrument."
nika      "The marks-instrument has that Vlad, that Vlad has no people."
nika      "Will Vlad come to the high ground?"

Lowering the spears#

Vlad

Tell them to lower the spears and let me.

Computer

a strip of five panels coming up on the lit face: two figures walking with footprints behind them and a cave opening ahead; seven figures at the opening with shafts held; the same seven with the shafts on the ground; the two figures among the seven at a fire; the operator's figure, hands empty, no shaft drawn

Nika

going along the five panels and stopping longest on the third Ovlat na kar-nu, ha. Aha-ktha na naga-tu, ha. tapping panel 3 twice and then shaking her head Naga-tu ta, aha ne kesta. pointing at the seven figures, then at him, then at the machine Aha-ktha ne semek Ovlat, aha-ktha ne semek semek-tu. putting a hand on her own chest and holding it there Aha na semek Ovlat, aha na semek aha-ktha. Aha na semek, ha. looking out at the light going off the ridge, and then at the fire Miht ta na miht ta, ne. Miht ta, karnu.

mandrin-k runtime 14:19:36

wrap

[14:19:36] in        "Tell them to lower the spears and let me."
[14:19:36] asr       en, peg chain position 2
[14:19:36] seg       2 clauses, both embedded under a reporting verb
[14:19:36] note      her question at 14:19:16 is unanswered. The turn answers it by presupposition rather than with the polar affirmative.[14:19:36] c1        "tell them to lower the spears"
[14:19:36] parse     reported speech, second person addressing a third party
[14:19:36] gate 1    ungrounded referent: check
[14:19:36] note      sheet 3 strip 15 gave transfer between two hands and sheet
[14:19:36] note      19/B panel 6 gave a figure speaking with a second figure
[14:19:36] note      listening. Speech reaching a party who is absent from the panel appears nowhere.
[14:19:36] note      sheet 15/A strip 2 came closest: a figure arriving and pointing at a drawn cave. That is a report of a place, and the reporting figure is present in the panel.
[14:19:36] note      an instruction carried to a party who is not present has no form. The clause has two speakers and one of them is absent.
[14:19:36] c1        REFUSED[14:19:36] c2        "let me"
[14:19:36] gate 1    BLOCKED
[14:19:36] note      permission is an action withheld that was available. Refused
[14:19:36] note      at 14:04:29, 14:05:51, 14:06:24 and 14:11:29 on the same gap.
[14:19:36] c2        REFUSED[14:19:36] REASONING autodraw ON, entering trace on 2 refused clauses
[14:19:36] trace     the content is a request about what happens on arrival.
[14:19:36] trace     arrival is available: sheet 8/A panels 6 and 7, and she used
[14:19:36] trace     the walking construction at 14:18:33.
[14:19:36] trace     spears lowered is available: sheet 20/D panel 5 drew a figure with unfamiliar ornaments holding a shaft, and she gave no outsider register for it. Sheet 15/B strip 1 drew a shaft
[14:19:36] trace     changing hands and she marked it long.
[14:19:36] trace     she put her own point off the shaft at 14:03:24 and laid the shaft behind her, out of her own reach, unprompted.
[14:19:36] trace     that action is in the corpus as her own and can be drawn.
[14:19:36] decision  ROUTE TO DISPLAY[14:19:37] draw      composing strip, 5 panels, sheet inventory only
[14:19:37] draw      panel 1  two figures walking, footprints behind, a cave
[14:19:37] draw      opening ahead                     [sheet 8/A 6,7]
[14:19:37] draw      panel 2  seven figures at the opening, shafts held
[14:19:37] draw      [sheet 3, 20/D]
[14:19:37] draw      panel 3  the same seven, shafts on the ground
[14:19:37] draw      [sheet 15/B]
[14:19:37] draw      panel 4  the two figures among the seven, at a fire
[14:19:37] draw      [sheet 3]
[14:19:37] draw      panel 5  the operator's figure, hands empty, no shaft drawn
[14:19:37] draw      [sheet 10/B]
[14:19:37] warn      panel 3 shows the shafts down without showing who put them down. Sheet 15/B established that she reads who gains from a panel and the agent is absent here.
[14:19:37] display   running[14:19:58] mic0      capture 19.3s, speaker voice, snr 19 dB
[14:19:58] asr       k: ovlat na kar-nu, ha
[14:19:59] asr       k: aha-ktha na naga-tu, ha
[14:19:59] asr       k: naga-tu ta, aha ne kesta
[14:20:00] asr       k: aha-ktha ne semek ovlat, aha-ktha ne semek semek-tu
[14:20:01] asr       k: aha na semek ovlat, aha na semek aha-ktha
[14:20:02] asr       k: aha na semek, ha
[14:20:02] asr       k: miht ta na miht ta, ne. miht ta, karnu
[14:20:03] seg       7 utterances
[14:20:03] tag       utterance 3: <kesta> bare, splitting, first person, negated
[14:20:03] tag       utterance 7: <karnu> here in the cold sense, sheet 16/C
[14:20:03] note      utterance 5 puts her in both slots of the marks construction
[14:20:03] note      utterance 7 revises the journey estimate given at 14:18:34nika      "Vlad comes to the high ground, yes."
nika      "The people carry spears, yes."
nika      "Those spears, I do not split."
nika      "The people have not seen Vlad, the people have not seen a marks-instrument."
nika      "I make marks at Vlad, I make marks at the people."
nika      "I make marks, yes."
nika      "One sun to one sun, no. One sun, cold."[14:20:03] note      utterance 3 declines the content of panel 3 while utterance
[14:20:03] note      5 accepts the sequence as a whole.

Vlad

pressing the rereason button with a note saying that the expected answer is whether people will lower spears or not

Computer

the five panels held where they are, nothing new put on the speech channel

Nika

watching the lit face go still with nothing new on it, then reaching past the machine and putting her hand on his forearm, briefly, and taking it back Aha na semek. pointing at herself, then at him, then out at the ridge, in that order, and then doing the three again in the same order Aha na semek Ovlat. Aha na semek aha-ktha. Ovlat na kar-nu, miht ta.

mandrin-k runtime 14:20:22

wrap

[14:20:22] cmd       re-reason: 14:19:36
[14:20:22] note      operator note attached: expected answer is whether the people will lower their spears[14:20:22] REASONING re-entering trace with the expected answer supplied
[14:20:22] trace     the strip at 14:19:37 was read and answered in full. Seven
[14:20:22] trace     utterances returned, none refused, none marked wrong except
[14:20:22] trace     panel 3.
[14:20:22] trace     checking whether the expected answer is among them.
[14:20:22] trace     utterance 3: <naga-tu ta, aha ne kesta>. The demonstrative
[14:20:22] trace     attaches to the spears in panel 3. <kesta> is splitting, sheet 9/A frame 3, negated short.
[14:20:22] trace     the model rendered this as declining to split the spears.
[14:20:22] trace     re-checking that gloss.
[14:20:22] trace     <kesta> at 14:05:04 was produced by her against the operator and herself, with two ringed marks between them, sheet 9/C
[14:20:22] trace     panel 10. That use was retained as a figurative extension.
[14:20:22] trace     at 14:05:22 the operator used <kesta> for separation from a group and she accepted the term while rejecting the fact.
[14:20:22] trace     <kesta> therefore covers separation between parties as well as splitting an object.
[14:20:22] trace     utterance 3 with the second sense: the spears in that panel, I do not separate them. The agent absent from panel 3 is supplied by her as herself, in the negative.
[14:20:22] trace     utterance 5 puts her as the marks-maker toward both parties and utterance 6 states it bare.
[14:20:22] trace     reading the seven together: she comes, they carry spears, she does not put the spears down, she speaks to both sides.
[14:20:22] trace     the expected answer is present as an action she declines and an action she offers in its place.
[14:20:22] trace     checking for a form that puts the question directly.
[14:20:22] trace     the polar interrogative is available and takes an assertion.
[14:20:22] trace     panel 3 as an assertion has no agent, which is what she marked. Supplying the agent makes it a question about seven
[14:20:22] trace     people who are absent.
[14:20:22] trace     sheet 15/E frame 2 attached a claim to a figure with its back turned and she withheld. Sheet 15/E frame 4 attached it to seven figures and she withheld.
[14:20:22] trace     a claim about what seven absent people will do is that frame.
[14:20:22] decision  the question cannot be put to her about them. It can be put to her about herself, and it was answered.
[14:20:22] decision  REVISE GLOSS, EMIT NO NEW UTTERANCE[14:20:22] lex       <kesta> sense 2 promoted from figurative to attested
[14:20:22] lex       gloss: separate parties, hold apart
[14:20:22] lex       confidence 0.63revised gloss, 14:19:59
nika      "Those spears, I do not hold them apart."[14:20:22] warn      under the revised gloss she declines to stand between you and the seven figures with spears. Under the prior gloss she declines to disarm them.
[14:20:22] warn      utterance 5 is compatible with either.
[14:20:22] warn      confidence 0.63 and 0.58. Neither is above 0.75.

The night#

Vlad

Can you stay in the cave?

Computer

Semek na kar-nu ta na? Miht ta, karnu.

Nika

Ha. Aha na kar-nu ta, miht ta. putting another piece of wood on the fire Miht-nu ta, ha. pointing out at the ridge, where the light has gone Aha-ktha na karnu-me, aha-ktha ne semek aha, miht ta. Aha-ktha na semek aha, miht ta na miht ta. turning back and looking at his hands, then at the ash in the fire pit Rusa na?

mandrin-k runtime 14:20:44

wrap

[14:20:44] in        "Can you stay in the cave?"
[14:20:44] asr       en, peg chain position 2
[14:20:44] seg       1 clause, interrogative, ability + stative + locative
[14:20:44] lex       can: no modal system, as at 14:07:18, 14:18:14, 14:18:54
[14:20:44] lex       stay: no stative, as at 14:05:51, 14:17:09
[14:20:44] lex       cave: <kar-nu> has been carrying this place since 14:06:01
[14:20:44] note      sheet 1 panel 18 gave an opening in a rock face and returned
[14:20:44] note      a form. Checking the entry.
[14:20:44] lex       sheet 1 panel 18: <sula>. Recorded, 3 occurrences, all sheet 1
[14:20:44] lex       <sula-nu> derived by the speaker at sheet 17/F frame 4 against
[14:20:44] lex       a hollow with water and figures around it
[14:20:44] note      <sula> is the opening and <kar-nu> is the ground it is in.
[14:20:44] note      the locative wanted here is the opening.
[14:20:44] note      she has used <kar-nu> of this place 6 times and <sula> none
[14:20:44] note      since sheet 1. Following her usage.
[14:20:44] fall      the polar frame accepted at 14:05:51: asking whether she is at a place, with the time supplied
[14:20:44] fall      she named the coming night at 14:20:02: <miht ta, karnu>
[14:20:44] build     "semek na kar-nu ta na? miht ta, karnu"
[14:20:44] warn      this asks whether she is here, with the cold night named
[14:20:44] warn      after it. It does not ask whether she is able to. Confidence
[14:20:44] warn      0.64.
[14:20:44] out       "semek na kar-nu ta na? miht ta, karnu"
[14:20:44] spk0      playing 2.9s[14:20:58] mic0      capture 14.8s, speaker voice, snr 19 dB
[14:20:58] asr       k: ha
[14:20:58] asr       k: aha na kar-nu ta, miht ta
[14:20:59] asr       k: miht-nu ta, ha
[14:21:00] asr       k: aha-ktha na karnu-me, aha-ktha ne semek aha, miht ta
[14:21:01] asr       k: aha-ktha na semek aha, miht ta na miht ta
[14:21:01] asr       k: rusa na?
[14:21:02] seg       6 utterances, contour: level, level, level, level, level, rising
[14:21:02] note      utterance 4 states the group will not see her tonight
[14:21:02] note      utterance 5 uses the journey term written at 14:18:36
[14:21:02] note      utterance 6 is a bare polar question on the deer termnika      "Yes."
nika      "I am at this high ground, this sun."
nika      "This fire place, yes."
nika      "The group is at the small high ground, the group will not see me, this sun."
nika      "The group sees me, one sun to one sun."
nika      "Deer?"

Vlad

editing reasoning parameters on the computer to allow creation of new functions

Computer

a short horizontal bar with the ends open drawn on its own, then four strips under it: a figure, a shaft, an antlered animal, the animal down, the full strip with no bar in it; the same strip with the third panel replaced by the bar; the same with the second panel replaced by it; then the bar, the shaft, the animal, the bar

Nika

reading strip 1 straight through Semek na aha rusa naga-tu, rusa-ke. at strip 2, stopping on the bar where the third panel was, and saying the strip again with a gap in it Ne. Semek na aha, naga-tu, ne, rusa-ke. putting one finger on the bar and leaving it there Ta ne. Ne semek. going back through strips 3 and 4, finding the bar in each, touching each one Aha ne semek ta. turning to him with the palm up Semek na?

mandrin-k runtime 14:21:20

wrap

[14:21:20] param     runtime.function_synthesis = ON
[14:21:20] note      the model may introduce grammatical functions the target
[14:21:20] note      lacks, by pairing a drawn symbol with the function across
[14:21:20] note      repeated contexts, in the manner of sheets 5 and 12.
[14:21:20] note      gate 6 governs stems and remains ON. This parameter covers
[14:21:20] note      function words and markers.
[14:21:20] note      sheet 5 Part A drew four marks above one panel and she read
[14:21:20] note      none of them, so a symbol carries no function on sight and has to be built against panels she already reads.[14:21:20] AUDIT     scanning session for content deleted at emission
[14:21:20] audit     14:04:29  c3, c4     spare, standing among absent others
[14:21:20] audit     14:05:51  whole turn  reason for an action
[14:21:20] audit     14:06:24  c2          an alternative not taken
[14:21:20] audit     14:08:51  4 clauses   two dates, one interval
[14:21:20] audit     14:11:29  c2          an assertion held at part strength
[14:21:20] audit     14:12:04  definition  11 head terms
[14:21:20] audit     14:12:09  whole turn  forward time, deictic object
[14:21:20] audit     14:13:34  c4, c5, c6  a subset, a named figure, a degree
[14:21:20] audit     14:17:09  2 clauses   a state held about another, a cause
[14:21:20] audit     14:18:14  c2          comprehension, merged into agreement
[14:21:20] audit     14:19:36  c1, c2      reported speech, permission
[14:21:20] audit     total 11 turns, 19 clauses, 0 recoverable from the record
[14:21:20] audit     the emitted stream contains no marker where a clause was dropped. She has been answering a text with holes in it and has had no way to see the holes.
[14:21:20] audit     her turns at 14:04:47, 14:12:33 and 14:17:26 queried the source of an utterance. All three followed a deletion.[14:21:21] SYNTHESIS priority ordered by deletion count
[14:21:21] fn 1      ALTERNATIVE NOT TAKEN            6 deletions
[14:21:21] fn 2      PART STRENGTH                    4 deletions
[14:21:21] fn 3      REASON                           3 deletions
[14:21:21] fn 4      MISSING SPAN                     19 deletions, meta[14:21:21] build     fn 4 first. Without it the others arrive inside a stream
[14:21:21] build     she cannot check.
[14:21:21] draw      symbol: a short horizontal bar with the ends open
[14:21:21] draw      grounding sequence, 4 strips, sheet inventory only
[14:21:21] draw      strip 1  figure, shaft, antlered animal, animal down
[14:21:21] draw      full strip, no symbol            [sheet 3 str 1]
[14:21:21] draw      strip 2  figure, shaft, [symbol], animal down
[14:21:21] draw      the third panel replaced by the symbol
[14:21:21] draw      strip 3  figure, [symbol], antlered animal, animal down
[14:21:21] draw      strip 4  [symbol], shaft, antlered animal, [symbol]
[14:21:21] note      the symbol occupies a panel slot she already scans. Its position varies and its shape holds.
[14:21:21] display   running[14:21:44] mic0      capture 16.9s, speaker voice, snr 19 dB
[14:21:44] asr       k: semek na aha rusa naga-tu, rusa-ke
[14:21:45] asr       k: ne
[14:21:45] asr       k: semek na aha, naga-tu, ne, rusa-ke
[14:21:46] asr       k: ta ne
[14:21:46] asr       k: ne semek
[14:21:47] asr       k: aha ne semek ta
[14:21:47] asr       k: semek na?
[14:21:48] seg       7 utterances
[14:21:48] tag       utterance 3: the symbol read aloud as <ne> in sequence
[14:21:48] tag       utterance 4: <ta ne>, the demonstrative on the negative
[14:21:48] tag       utterance 5: <ne semek>, the negative on the marks term
[14:21:48] note      she has produced a term in the symbol's slot on the second strip and held it across strips 3 and 4
[14:21:48] lex       writing entry: ne semek = a span that was not said
[14:21:48] lex       status PROVISIONAL, confidence 0.51
[14:21:48] lex       lexicon 49 -> 50 typesnika      "You take deer with a spear, the deer is dead."
nika      "Wrong."
nika      "You take, a spear, missing, the deer is dead."
nika      "That missing thing."
nika      "Unsaid marks."
nika      "I have not seen that."
nika      "Who says this?"[14:21:48] note      utterance 7 is the source query, fourth occurrence today.
[14:21:48] note      the three prior occurrences each followed a deletion.

Vlad

It's not about me. This creates functions so we can understand each other better.

Computer

Semek-tu na semek. Ovlat ne. Ne semek, semek-tu na kesta-ne. Aha na semek ovlat, ovlat na semek aha. Ne semek.

Nika

Ne semek. saying it a second time and putting her hand on the machine Ha, ne semek. Semek-tu na semek, Ovlat ne, ha. repeating the reciprocal back with the marker on the end where the machine put it Aha na semek Ovlat, Ovlat na semek aha, ne semek. holding both hands out wide Ne semek ktha. bringing the hands in onto the machine Semek-tu na ne semek ktha, miht ta. Ha.

mandrin-k runtime 14:22:06

wrap

[14:22:06] in        "It's not about me. This creates functions so we can understand each other better."
[14:22:06] asr       en, peg chain position 2
[14:22:06] seg       3 clauses
[14:22:06] note      her utterance 7 at 14:21:47 is the source query. The frame
[14:22:06] note      asks who stands beside a claim, sheet 15/E.[14:22:06] c1        "it's not about me"
[14:22:06] note      the query asks for a source and this declines to be one
[14:22:06] note      sheet 15/E frame 3 showed a claim with no figure beside it and that configuration produced the query in the first place
[14:22:06] fall      naming the machine as the source, which is available and which she has treated as a speaker since 14:05:03
[14:22:06] c1        out "semek-tu na semek. ovlat ne"[14:22:06] c2        "this creates functions"
[14:22:06] lex       function: no entry, no referent
[14:22:06] gate 6    coinage without referent: check
[14:22:06] note      the provisional entry <ne semek> written at 14:21:48 is one such thing and she produced it herself. The class has one member and she has met it.
[14:22:06] fall      naming the member instead of the class
[14:22:06] c2        out "ne semek, semek-tu na kesta-ne"
[14:22:06] warn      this says the marks-instrument builds the unsaid-marks term.
[14:22:06] warn      Confidence 0.46.[14:22:06] c3        "so we can understand each other better"
[14:22:06] lex       understand: merged into the affirmative at 14:18:14
[14:22:06] lex       can: no modal system
[14:22:06] lex       better: comparative of degree, refused at 14:13:34 c6
[14:22:06] lex       so that: purpose, an outcome that has yet to happen
[14:22:06] gate 2    check
[14:22:06] note      forward counted range is 2 cold cycles and this sits inside it, so the time passes
[14:22:06] gate 1    ungrounded referent
[14:22:06] note      purpose links an action to an outcome and the causal
[14:22:06] note      connective was refused at 14:05:51 and 14:17:09
[14:22:06] note      the reciprocal is available: she built it herself at 14:06:59 and at 14:20:22, putting both parties in both slots
[14:22:06] fall      the reciprocal alone, with the purpose dropped
[14:22:06] c3        out "aha na semek ovlat, ovlat na semek aha"
[14:22:06] warn      this is her own sentence returned to her. The purpose is absent and the sequence carries it at lower precision.[14:22:06] REASONING fn 4 available, marking the deletion
[14:22:06] trace     <ne semek> is provisional at 0.51 and has one grounding
[14:22:06] trace     sequence behind it. Emitting it inside a live turn is its first use outside the strips.
[14:22:06] trace     the deletion in c3 is the purpose clause.
[14:22:06] decision  EMIT WITH MARKER
[14:22:06] out      "semek-tu na semek. ovlat ne. ne semek, semek-tu na kesta-ne. aha na semek ovlat, ovlat na semek aha. ne semek"
[14:22:06] spk0      playing 6.2s
[14:22:06] warn      <ne semek> appears twice in this utterance, once naming the term and once using it. Confidence 0.46 and 0.51.[14:22:22] mic0      capture 18.6s, speaker voice, snr 18 dB
[14:22:22] asr       k: ne semek
[14:22:22] asr       k: ha, ne semek
[14:22:23] asr       k: semek-tu na semek, ovlat ne, ha
[14:22:24] asr       k: aha na semek ovlat, ovlat na semek aha, ne semek
[14:22:25] asr       k: ne semek ktha
[14:22:26] asr       k: semek-tu na ne semek ktha, miht ta
[14:22:26] asr       k: ha
[14:22:27] seg       7 utterances
[14:22:27] lex       <ne semek> returned unprompted in 4 of 7 utterances
[14:22:27] lex       status PROVISIONAL -> ATTESTED, confidence 0.51 -> 0.83
[14:22:27] tag       utterance 5: <ne semek ktha>, the term pluralised with the free quantifier
[14:22:27] note      utterance 6 attributes the missing spans to the machine and places them across the whole of todaynika      "Unsaid marks."
nika      "Yes, unsaid marks."
nika      "The marks-instrument speaks, not Vlad, yes."
nika      "I make marks at Vlad, Vlad makes marks at me, missing."
nika      "Missing spans beyond counting."
nika      "The marks-instrument has missing spans beyond counting, this sun."
nika      "Yes."

Vlad

a bit frustrated Yes, that is the problem.

Computer

Ha. Ne semek ktha, ne.

Nika

Ha. looking at his face rather than the lit face Ovlat, hesu ne. Torna-hesu na? answering it herself before he can Ne. Kamla ne, torna-hesu ne. putting a hand on the gravel beside him and leaving it there Aha na kar-nu ta. Ne semek ktha, ha, ne. taking her hand back and pointing at the machine, then out at the dark on the ridge, then at him Miht ta na miht ta, aha na semek Ovlat.

mandrin-k runtime 14:22:46

wrap

[14:22:46] in        "Yes, that is the problem."
[14:22:46] asr       en, peg chain position 2
[14:22:46] prosody   pitch range narrowed, rate up 14%, terminal fall steep
[14:22:46] note      the speaker-side channel carries contour and the operator
[14:22:46] note      side has not until now. Contour is available on both.
[14:22:46] seg       2 clauses[14:22:46] c1        "yes"
[14:22:46] c1        out "ha"[14:22:46] c2        "that is the problem"
[14:22:46] lex       problem: no entry
[14:22:46] note      sheet 19/D separated a wrong thing from a dangerous thing and both are states of an object or an act, not of a situation experienced by a person.
[14:22:46] fall      <ne> short, on the term she has just used
[14:22:46] c2        out "ne semek ktha, ne"
[14:22:46] out       "ha. ne semek ktha, ne"[14:22:46] spk0      playing 2.1s
[14:22:46] warn      the operator prosody was dropped. Sheet 5 Part A established
[14:22:46] warn      four contours and none of them is this one. spk0 emits the level contour for an affirmative.[14:23:01] mic0      capture 13.4s, speaker voice, snr 18 dB
[14:23:01] asr       k: ha
[14:23:01] asr       k: ovlat, hesu ne
[14:23:02] asr       k: torna-hesu na?
[14:23:02] asr       k: ne
[14:23:03] asr       k: kamla ne, torna-hesu ne
[14:23:04] asr       k: aha na kar-nu ta. ne semek ktha, ha, ne
[14:23:04] asr       k: miht ta na miht ta, aha na semek ovlat
[14:23:05] seg       7 utterances, contour: level, level, rising, level, level, level, level
[14:23:05] tag       utterance 2: <hesu ne>, sheet 4/A panel 2, mouth turned down
[14:23:05] tag       utterance 3: <torna-hesu>, sheet 4/A panel 4, anger, queried
[14:23:05] tag       utterance 5: both states ruled out
[14:23:05] note      she read the operator from the face and the voice. Neither
[14:23:05] note      channel was carried by spk0.nika      "Yes."
nika      "Vlad, unhappy."
nika      "Angry?"
nika      "No."
nika      "Not grief, not anger."
nika      "I am at this high ground. Missing spans beyond counting, yes, wrong."
nika      "One sun to one sun, I make marks at Vlad."

What the words carry#

Vlad

typing a by-the-way command to explain what one sun to one sun means and what I make marks at Vlad means

Computer

the two glosses coming up on the lit face for the operator, nothing put on the speech channel

Nika

waiting while the lines come up on the lit face Semek na? reaching over and turning the machine a little so the lit face is toward her, then leaving it turned that way Aha ne semek ta.

mandrin-k console 14:23:24, explain

wrap

[14:23:24] in        [operator console] explain: "miht ta na miht ta"
[14:23:24] in        [operator console] explain: "aha na semek ovlat"
[14:23:24] param     explain.level = 5
[14:23:24] note      no output to spk0. Console only.miht ta na miht ta
  one sun, to, one sun
  She said this three times today. At 14:18:34 it was about where her
group is. At 14:20:03 she took it back and said one sun instead. At
14:21:01 and again just now she used it about seeing you again.
  It is one sun, then the word that goes between things, then one sun.
You can walk it between the sun coming up and the sun going down. It
is how far away a thing is, said as walking.
  She has no word for a distance on its own. A place is near or far by
how many suns it takes. She counted to seven suns on the sheets and
she stopped there.
  At 14:20:03 she changed it to one sun and said cold. The walk got
shorter because the night is coming, or because the cold makes the
walking different. She did not say which.

aha na semek ovlat
  me, to, marks, Vlad
  She has said this eleven times. The first was at 14:06:59.
  Three words in a row. The first is who does it. The second is the
word that goes between. The third is the doing. The fourth is who it
is done to.
  semek is marks. It is also speaking. It is also you, because you make
marks, and she called you that at 14:04:07 and has kept it.
  So the sentence is: I make marks at Vlad. It is also: I speak to
Vlad. She uses one word for both and there is no way to tell them
apart from the words alone.
  When she turns it around and says ovlat na semek aha, that is you
doing it to her. She has said both halves together nine times. She
did it at 14:06:59, at 14:20:22 and at 14:22:24.
  When she says aha na semek aha-ktha, the last word is her group. She
said that at 14:20:01, with the same middle.
  So just now she said: one sun of walking away, I speak to Vlad. The
two parts sit side by side and nothing joins them. What joins them is
that she said them together.[14:23:24] note      <semek> carries marks, speech, and the operator. All three
[14:23:24] note      senses are live and no disambiguation is available.

Vlad

I do not know everything. This computer is an artificial intelligence, like a human, but in a form that only exists in this machine, not a full human. I am tired.

Computer

Ovlat na ne semek ktha. Semek-tu, ne semek ktha. Ovlat na tesu.

Nika

Ha. Ovlat na tesu, ha. looking at the cave mouth, where the light has gone off the ridge entirely Miht ontu, karnu ta. picking the spear up and setting it across her knees, point outward toward the entrance Aha na kar-nu ta, aha ne tesu. Torna ta, aha na naga-tu. putting a hand out toward the fire and drawing it back Ovlat na tesu, aha na semek. Ne semek ktha, miht ta na miht.

mandrin-k runtime 14:23:46

wrap

[14:23:46] in        "I do not know everything. This computer is an artificial intelligence, like a human, but in a form that only exists in this machine, not a full human. I am tired."
[14:23:46] asr       en, peg chain position 2
[14:23:46] seg       5 clauses[14:23:46] c1        "I do not know everything"
[14:23:46] lex       know: merged into <semek> and into the affirmative
[14:23:46] lex       everything: sheet 16/B frames 7 and 8 returned <ktha1> for a band beyond thirty. A totality was elicited on no sheet.
[14:23:46] fall      the marker written at 14:22:27, on the operator
[14:23:46] c1        out "ovlat na ne semek ktha"
[14:23:46] warn      she has been putting this on the machine since 14:22:26.
[14:23:46] warn      This moves it onto you.[14:23:46] c2        "this computer is an artificial intelligence"
[14:23:46] lex       intelligence: no entry, no referent
[14:23:46] gate 1    BLOCKED
[14:23:46] c3        "like a human"
[14:23:46] lex       human: <aha>, <aha-ktha>
[14:23:46] lex       like: comparison of kind. Sheet 14/A ran two things against
[14:23:46] lex       each other and returned same or split, with nothing between.
[14:23:46] gate 1    BLOCKED
[14:23:46] c4        "in a form that only exists in this machine, not a full human"
[14:23:46] gate 3    type/token: BLOCKED at sheet 10/D item 7
[14:23:46] note      a thing that is a person in one machine and nowhere else
[14:23:46] note      requires the distinction she refused
[14:23:46] c2        , c3, c4  REFUSED[14:23:46] REASONING fn 4 available
[14:23:46] trace     three clauses in sequence, all describing the machine.
[14:23:46] trace     she has treated the machine as a speaker since 14:05:03 and named it a stone that speaks. She has not asked what it is.
[14:23:46] trace     emitting the marker alone tells her that something about
[14:23:46] trace     the machine was said and dropped.
[14:23:46] decision  EMIT MARKER, PLURAL
[14:23:46] out       "semek-tu, ne semek ktha"[14:23:46] c5        "I am tired"
[14:23:46] lex       tired: no entry
[14:23:46] note      sheet 4/A ran ten states and none of them is depletion
[14:23:46] note      sheet 1 panel 15 gave a figure lying down with eyes closed and she returned a form for it
[14:23:46] lex       sheet 1 panel 15: <tesu>, 4 occurrences, all sheet 1
[14:23:46] fall      naming the act rather than the state
[14:23:46] c5        out "ovlat na tesu"
[14:23:46] warn      this says you sleep. It does not say you are tired.
[14:23:46] out       "ovlat na ne semek ktha. semek-tu, ne semek ktha. ovlat na tesu"
[14:23:46] spk0      playing 4.7s[14:24:02] mic0      capture 15.2s, speaker voice, snr 18 dB
[14:24:02] asr       k: ha
[14:24:02] asr       k: ovlat na tesu, ha
[14:24:03] asr       k: miht ontu, karnu ta
[14:24:04] asr       k: aha na kar-nu ta, aha ne tesu
[14:24:05] asr       k: torna ta, aha na naga-tu
[14:24:05] asr       k: ovlat na tesu, aha na semek
[14:24:06] asr       k: ne semek ktha, miht ta na miht
[14:24:06] seg       7 utterances
[14:24:06] tag       utterance 3: <miht ontu>, the sun finished, sheet 6/B
[14:24:06] tag       utterance 7: <miht ta na miht>, this sun to a sun, without
[14:24:06] tag       the second demonstrative
[14:24:06] note      utterance 7 modifies the journey term written at 14:18:36 by dropping one element. First occurrence of this form.nika      "Yes."
nika      "Vlad sleeps, yes."
nika      "The sun is finished, it is cold here."
nika      "I am at this high ground, I do not sleep."
nika      "Bears here, I have the spear."
nika      "Vlad sleeps, I make marks."
nika      "Missing spans beyond counting, this sun to a sun."[14:24:06] note      utterance 7 places the missing spans between today and an unspecified later day.

Vlad

You want to stay with me.

Computer

Semek na kar-nu ta, ovlat na kar-nu ta, na?

Nika

Ha. Aha na kar-nu ta, Ovlat na kar-nu ta. moving across the gravel until she is beside him, close to the fire, and setting the spear down on her far side Aha-ke ne. Aha-ke ne, aha-ktha na. Ovlat na semek, Ovlat ne aha-ktha. putting her hand against his shoulder, holding it there, and then taking it away and holding the same hand out between them, palm up, waiting Aha na kesta-ne. Semek na ha na?

mandrin-k runtime 14:24:26

wrap

[14:24:26] in        "You want to stay with me."
[14:24:26] asr       en, peg chain position 2
[14:24:26] seg       1 clause, second person + volition + stative + comitative
[14:24:26] lex       want: no entry
[14:24:26] gate 1    ungrounded referent: check
[14:24:26] note      sheet 4/A ran ten states and all ten are readable off a face or a posture. A state directed at an action that has yet to happen appears on none of them.
[14:24:26] note      the same gap refused expectation at 14:04:29 and the alternative not taken at 14:04:29, 14:05:51, 14:06:24, 14:11:29 and 14:19:36.
[14:24:26] lex       stay: no stative, as at 14:05:51, 14:17:09, 14:20:44
[14:24:26] lex       with: comitative. Sheet 3 strip 3 drew two figures walking to a cave and she returned them as two agents in sequence. No marker joins them.
[14:24:26] note      she stated the content herself at 14:24:04 and 14:24:05:
[14:24:26] note      she is at this high ground, she does not sleep, you sleep, she makes marks.
[14:24:26] fall      the polar frame on the two parties at one place, accepted at 14:20:44
[14:24:26] build     "semek na kar-nu ta, ovlat na kar-nu ta, na?"
[14:24:26] warn      this asks whether you are both here. It does not ask what she wants. Confidence 0.61.
[14:24:26] out       "semek na kar-nu ta, ovlat na kar-nu ta, na?"
[14:24:26] spk0      playing 3.2s[14:24:41] mic0      capture 16.7s, speaker voice, snr 18 dB
[14:24:41] asr       k: ha
[14:24:41] asr       k: aha na kar-nu ta, ovlat na kar-nu ta
[14:24:42] asr       k: aha-ke ne
[14:24:43] asr       k: aha-ke ne, aha-ktha na
[14:24:44] asr       k: ovlat na semek, ovlat ne aha-ktha
[14:24:44] asr       k: aha na kesta-ne
[14:24:45] asr       k: semek na ha na?
[14:24:46] seg       7 utterances, contour: level, level, level, level, level, level, rising
[14:24:46] tag       utterance 4: <aha-ktha na> with the relator and no second term
[14:24:46] tag       utterance 6: <kesta-ne> bare, joining, first person
[14:24:46] tag       utterance 7: <ha> inside a polar question, first occurrence
[14:24:46] note      utterance 7 asks about an affirmative rather than asserting
[14:24:46] note      one. No sheet elicited this.
[14:24:46] note      the gesture accompanying it is outside the model's channels.nika      "Yes."
nika      "I am at this high ground, Vlad is at this high ground."
nika      "A person alone is wrong."
nika      "A person alone is wrong, the people are."
nika      "Vlad makes marks, Vlad has no people."
nika      "I join."
nika      "Do you say yes?"

Vlad

typing a by-the-way command to explain why a person alone is wrong repeats

Computer

the gloss on the lit face rewritten for the operator, nothing put on the speech channel

Nika

keeping the hand out, palm up, and putting the other one back on the gravel between them Ha na?

mandrin-k console 14:25:04, explain

wrap

[14:25:04] in        [operator console] explain: "aha-ke ne", recurrence
[14:25:04] note      no output to spk0. Console only.aha-ke ne
  person, cut off, wrong
  Four occurrences today. 14:06:02, 14:17:55, 14:24:42, 14:24:43.
  aha is a person. The ending -ke means finished, not running any
more. A dead animal takes it. A fire that has gone out takes it.
  On a person it has two senses. One is dead. The other is cut off
from the group. She gave both on sheet 4.
  ne short is wrong. It is the word she uses for a spear bound
crookedly and for a name given to the wrong thing.
  So the three words are: a person cut off is wrong.
  It repeats because she has no other way to say it. There is no word
for should, no word for danger to a person over time, no word for
a state that lasts. She has one short sentence that covers all of
it and she says it again each time it applies.
  It applied when she found you here on the fifth day. It applied
after the strip about the bear. It applies now that the sun has gone.
  The second sense sits next to the first in the same three words.
Nothing in the grammar separates them.[14:25:04] note      her utterance at 14:24:45 is unanswered. <semek na ha na> asks whether you say yes. Her hand is out.

Vlad

Yes.

Computer

Ha.

Nika

Ha. taking his hand in hers and holding it, then letting go Aha na kesta-ne, Ovlat na kesta-ne. Kesta-ne ta. picking the spear back up and setting it across her knees with the point toward the cave mouth Ovlat na tesu, aha na naga-tu, aha ne tesu. lifting her chin at the entrance, where it is fully dark now Miht na, aha-ktha na.

mandrin-k runtime 14:25:22

wrap

[14:25:22] in        "Yes."
[14:25:22] asr       en, peg chain position 2
[14:25:22] map       affirmative, answering <semek na ha na>
[14:25:22] out       "ha"
[14:25:22] spk0      playing 0.3s[14:25:34] mic0      capture 11.8s, speaker voice, snr 18 dB
[14:25:34] asr       k: ha
[14:25:34] asr       k: aha na kesta-ne, ovlat na kesta-ne
[14:25:35] asr       k: kesta-ne ta
[14:25:36] asr       k: ovlat na tesu, aha na naga-tu, aha ne tesu
[14:25:37] asr       k: miht na, aha-ktha na
[14:25:37] seg       5 utterances, contour: level, level, level, level, level
[14:25:37] tag       utterance 5: <miht na>, the sun with the relator and no second term, matching the shape at 14:24:43
[14:25:37] note      the construction leaves the second slot open and she has now
[14:25:37] note      used it twice in one minutenika      "Yes."
nika      "I join, Vlad joins."
nika      "That joined thing."
nika      "Vlad sleeps, I have the spear, I do not sleep."
nika      "The sun, and the people."

The camera and the wall#

Vlad

shutting down the computer and starting a camera with a spinning hard drive on it

Nika

hearing the machine go quiet and the lit face go dark, and putting her hand on it once Ontu. turning to the other thing in his hands, the one that has started making a thin sound of its own Ne. pointing at it, then at the dark machine, then back at it Kruf ta na? Semek-tu na? leaning in toward the sound, close enough that the firelight catches the lens, and stopping there Semek na?

Vlad

pointing at the wall close to the ceiling that contains the description of the device in pictures

Nika

getting up and going to the wall, following his hand up to the panel near the ceiling, and reading along it left to right Ha. touching the drawing of the lens, then the drawing of the disc, then the figure standing in front of both Semek ta. coming back down the wall to the four marks, then to the crossed closed mouth, then to the looking glass, then to the sitting figure, putting a hand on each in turn Semek ktha. Semek ktha, Ovlat. turning around to face him with her back to the wall Ovlat na semek ta, miht-ktha. Ovlat ne aha-ke.

Vlad

closing his eyes Ha.