Speech Transcript - Sound Patterns in Language

0:00:16	Good morning everyone.
0:00:19	Thank you so much Haizhou for your
0:00:23	very kind introduction.
0:00:26	And I also want to thank the organising committee
0:00:30	for giving me this chance to comment,
0:00:34	interact with colleagues from so many different fields, from so many different areas.
0:00:42	Speech
0:00:45	is a fundamental defining characteristic of our species.
0:00:51	Sometimes
0:00:53	our species is called
0:00:55	Homo sapiens,
0:00:58	but I'm not sure we deserve the word sapiens.
0:01:03	More likely,
0:01:05	some people said, we should be called
0:01:07	Homo loquens (talking man),
0:01:10	because
0:01:11	we're talking all the time, we're using speech all the time.
0:01:16	Because it's so fundamental, it's also
0:01:20	extremely complex.
0:01:23	I'm reminded of the cartoon that
0:01:27	Ray Liu showed yesterday
0:01:30	of an ocean of knowledge and
0:01:33	a few islands popping out.
0:01:37	Our job
0:01:38	is to build bridges
0:01:40	among those islands.
0:01:43	An island of electronic engineering,
0:01:46	island of linguistics,
0:01:49	computer science,
0:01:50	psychology,
0:01:52	neuroscience.
0:01:54	Each one
0:01:56	gives us an important window on speech.
0:02:00	And we need to connect them together.
0:02:05	The theme of this
0:02:07	conference
0:02:08	is the Diversity
0:02:10	of Spoken Language.
0:02:14	So this is the outlining
0:02:18	of my remarks today.
0:02:21	I will begin by asking, How is speech possible?
0:02:27	and I'll say something about
0:02:29	African origins,
0:02:32	the diversity that we see today in genes
0:02:37	and in words,
0:02:40	then I will proceed to
0:02:43	present three
0:02:45	case studies
0:02:47	of sound patterns in motion.
0:02:51	Sounds
0:02:52	hardly ever
0:02:54	behave as individuals.
0:02:57	They always
0:02:58	configure in patterns and move in patterns.
0:03:03	And I will
0:03:06	discuss consonants.
0:03:08	How they moved
0:03:10	with respect to Grimm's Law.
0:03:14	I will talk about vowels and
0:03:17	the Great Vowel Shift
0:03:20	in the history of English.
0:03:23	And I will talk about something that's not
0:03:25	found in all languages. Consonants and vowels are found in
0:03:29	all languages.
0:03:32	But in several thousand languages of the World you find tones.
0:03:36	And I will say something about how tones move.
0:03:41	And if there's time, I'd like to say something about speech and music.
0:03:45	Two cultural universals.
0:03:48	Two
0:03:49	cultural aspects that's present in every group
0:03:53	of human beings.
0:03:55	And I will close off.
0:04:04	I saw this cartoon few months ago and I thought maybe it's relevant.
0:04:10	Here you have one dolphin talking to another.
0:04:15	Although humans make sounds with their mouths,
0:04:19	occasionally look at each other,
0:04:22	there is no solid evidence that they actually communicate among themselves.
0:04:30	If somebody came up to me and looked at me and spoken Tamil.
0:04:37	I wouldn't understand him.
0:04:40	Or
0:04:42	Swahili,
0:04:44	or Caucasian,
0:04:46	or Abkhaz.
0:04:51	So in that sense our species is unique.
0:04:56	There is no other species
0:04:58	that have such great systematic diversity
0:05:02	among system of communication.
0:05:05	Diversity of language
0:05:07	is something unique to our species.
0:05:12	How did this
0:05:13	diversity come about?
0:05:17	Well, here's one story.
0:05:19	Lori Lamel
0:05:20	yesterday also
0:05:22	showed this beautiful painting by
0:05:24	Bruegel.
0:05:28	It is based on the story of The Book of Genesis.
0:05:34	God was annoyed
0:05:37	that people
0:05:39	had the hubris and
0:05:41	the arrogance
0:05:43	of building
0:05:45	a tower so high to reach the sky.
0:05:48	So God said, Go to, let us go down,
0:05:52	and there confound their language,
0:05:56	so that they may not understand
0:05:58	one another's speech.
0:06:03	Actually people all over the world
0:06:07	ethnographies told us,
0:06:10	have a very similar accounts
0:06:12	of creations and origins and so on.
0:06:16	My purpose this morning is to look at
0:06:20	diversity
0:06:22	more from
0:06:23	evolutionary perspective.
0:06:28	The key concept
0:06:31	behind diversity
0:06:33	is innovation.
0:06:37	In biological evolution
0:06:41	every generation of giraffes
0:06:45	has a certain degree of variation
0:06:48	in how long their neck is.
0:06:52	Every generation of wolves
0:06:54	have a certain variation
0:06:57	in how long their canine teeth are for fighting.
0:07:02	So nature selects,
0:07:04	there's variation.
0:07:06	As you innovate,
0:07:08	every generation has its variation.
0:07:11	And nature selects.
0:07:14	Similarly language.
0:07:19	Every infant,
0:07:21	in trying to reconstruct a language based on the
0:07:24	sounds, based on the language context in which it's born,
0:07:28	innovates.
0:07:30	And these innovations
0:07:32	mostly go by the wayside.
0:07:35	Many of them
0:07:36	are selected culturally rather than biologically
0:07:40	and persevere.
0:07:43	And as these innovations build up, accumulate more and more,
0:07:48	pretty soon
0:07:50	you won't be able to talk to each other anymore.
0:07:53	So, in biological evolution the innovations build up.
0:07:57	There's no longer mutual
0:08:00	reproductive
0:08:01	fertility.
0:08:03	You get a new species
0:08:06	in speech.
0:08:07	As these innovations become numerous,
0:08:10	build up over generation after generation,
0:08:13	you lose mutual intelligibility
0:08:16	and you have a new language.
0:08:22	So, the sounds as I said earlier
0:08:26	are not individuals.
0:08:28	They form natural patterns
0:08:31	and together
0:08:33	they
0:08:34	move in these patterns.
0:08:39	As these
0:08:40	innovations take place,
0:08:42	the social values attached to them.
0:08:47	Many of you may have enjoyed this great movie My Fair Lady.
0:08:52	Henry Higgins in the movie was actually based on the real English phonetician called Henry
0:08:59	Sweet.
0:09:01	But in the movie of course they took various liberties.
0:09:04	Here's Henry Higgins saying,
0:09:07	An Englishman's way of speaking
0:09:10	absolutely classifies him.
0:09:13	The moment he talks
0:09:15	he makes some other Englishman
0:09:17	despise him.
0:09:19	One common language I'm afraid we'll never get.
0:09:25	Oh, why can't the English learn to set a good example to people whose English
0:09:31	is painful to your ears?
0:09:33	The Scotch and the Irish
0:09:35	leave you close to tears.
0:09:38	Maybe that's why Scotch want independence.
0:09:42	Anyway according to Higgins,
0:09:45	There even are places where English completely disappears.
0:09:49	In America, they happen't used it for years!
0:09:57	Each one of us has to make a whole series of personal decisions when we
0:10:02	talk.
0:10:05	When I went to Hong Kong,
0:10:10	I noticed that in the textbook
0:10:13	to say you
0:10:14	in Cantonese
0:10:16	it's nay
0:10:18	with an n.
0:10:20	You find n in most of the Chinese dialects. In Mandarin it's nee
0:10:25	So I was told it's nay.
0:10:28	But as I
0:10:29	walked around the campus,
0:10:31	I noticed the young people don't say nay they all say lay.
0:10:38	The textbook says woo is gho
0:10:41	whether ??real/rear or nasal.
0:10:43	But the young people don't say gho, they say who.
0:10:48	So should I say
0:10:50	nay?
0:10:51	Closer to my native dialect
0:10:54	or should I try to be fashionable and say lay?
0:10:59	Lay is kinda not my age.
0:11:02	So, there are all these personal decisions
0:11:05	and these decisions build up
0:11:08	for group solidarity,
0:11:11	for group distinctions, and so on, and on and on
0:11:15	over the generations
0:11:16	new languages arise.
0:11:20	But let's take a look much for the back in time.
0:11:25	When did speech get it's first shot?
0:11:29	When did speech
0:11:31	start its trajectory?
0:11:36	I think a good case can be made for
0:11:39	over three million years ago,
0:11:41	okay.
0:11:43	In the region in Ethiopia
0:11:45	a bunch of physical anthropologists
0:11:49	were very lucky
0:11:50	and discovered
0:11:52	a relatively complete fossil of a
0:11:56	young woman,
0:11:58	middle teens,
0:11:59	called Lucy.
0:12:02	By studying Lucy
0:12:04	in great detail,
0:12:06	her hip bones
0:12:08	her ankle bones, her scull and so on,
0:12:11	they determined
0:12:14	that Lucy walked.
0:12:16	Lucy had erect posture,
0:12:18	essentially as we do.
0:12:22	So Lucy had made the transition to bipedal posture.
0:12:26	And
0:12:28	this is the reconstruction
0:12:31	of
0:12:32	Lucy.
0:12:36	This by itself, of course, is only part of the evidence.
0:12:41	Another part of the evidence which is also quite solid
0:12:45	is that they found, around the same time not very far from
0:12:48	this region called Afar,
0:12:51	footprints
0:12:53	covered by volcanic ash, so it's nicely preserved.
0:12:58	And here's the great
0:13:00	anthropologist by the name Mary Leakey
0:13:04	measuring in great detail
0:13:07	a whole string of footprints
0:13:10	and these footprints proved beyond doubt that over three million years ago
0:13:16	our ancestors were walking.
0:13:21	What are the implications of walking?
0:13:25	Well, in this excellent book written by Daniel Lieberman
0:13:28	called The Story of the Human Body
0:13:31	he contrasts
0:13:33	the
0:13:34	skeleton structure of the modern human
0:13:39	with that of the chimpanzee
0:13:41	and indeed there are many differences, many adaptations.
0:13:48	For us
0:13:49	perhaps the most important
0:13:51	is the restructuring of the entire upper body
0:13:58	and the head.
0:14:01	Rather than hanging forward
0:14:04	and help by muscle
0:14:06	the head's resting squarely
0:14:08	on the
0:14:09	cervical vertebrae,
0:14:11	or the spinal column.
0:14:16	It's
0:14:17	relatively
0:14:19	effortless for me to stand here,
0:14:22	not for a chimp.
0:14:24	I can lock my knees,
0:14:27	my skull is resting, its weight is on the spinal column.
0:14:34	There are many implications
0:14:37	of this new posture.
0:14:39	And immediate one of course, it freed our hands.
0:14:43	We became
0:14:45	tool makers in a serious sense.
0:14:50	Another very important adaptation
0:14:54	is that because of the restructuring of our heads
0:14:59	our larynx
0:15:01	has descended.
0:15:06	This is from a book
0:15:07	by Fitch, professor at Vienna.
0:15:13	And it shows the descent of the larynx
0:15:19	as well as the hyoid bone,
0:15:21	here's the thyroid cartilage and so on.
0:15:28	Even more clearly,
0:15:30	Daniel Lieberman's book,
0:15:32	makes it
0:15:34	very obvious why this is important for speech.
0:15:39	In the chimpanzee for instance whereas there's essentially just one resonance tube
0:15:46	to produce vocal sounds,
0:15:50	in our case
0:15:51	with the larynx
0:15:53	lowered
0:15:55	here's the larynx with the larynx lowered,
0:15:57	we now have two tubes.
0:16:01	A tube
0:16:03	that's the mouth
0:16:05	and another tube that's the throat.
0:16:09	And with the agile movement of the tongue
0:16:13	you can form much greater variety of phonetic distinctions.
0:16:20	And this of course is also extremely important,
0:16:27	but it's useless to have all that hardware
0:16:31	unless we have the proper
0:16:33	control mechanisms
0:16:35	and then control mechanisms
0:16:37	right in the brain.
0:16:41	Usually we see pictures of the
0:16:45	hemispheres of the brain,
0:16:47	but here's a nice picture
0:16:50	of the base of the brain.
0:16:53	If we lifted up the head
0:16:56	and looked at the brain from bottom-up,
0:17:00	we'll see a picture something like this.
0:17:04	Here's the temporal lobe, here's the frontal lobe.
0:17:09	The point that I wanna make
0:17:11	is that most of the cranial nerves,
0:17:15	twelve pairs coming out from the base brain,
0:17:20	most of them
0:17:21	are involved in speech.
0:17:24	That's how important the speech is
0:17:26	for
0:17:29	our life.
0:17:32	Henry Lenneberg,
0:17:34	some years ago,
0:17:36	published this
0:17:38	very nice diagram
0:17:40	showing the various nerves.
0:17:43	Here's the facial nerve,
0:17:46	which is cranial nerve pair number seven.
0:17:51	Here is (a) which is the trigeminal nerve
0:17:56	and perhaps the most
0:17:58	noticeable one
0:18:01	is this one that starts from around here in the base stand,
0:18:05	goes down all the way,
0:18:08	curls around the aorta of the heart
0:18:12	and goes back up
0:18:14	to control the larynx.
0:18:17	This is of course an effect of ascending up.
0:18:23	So five hairs of the twelve pairs of cranial nerves are involved in
0:18:28	controlling speech,
0:18:29	jaw movement, lip movement, larynx, tongue and so on.
0:18:40	So, probably
0:18:41	given both the
0:18:43	software in the head,
0:18:46	hardware in the throat and the mouth,
0:18:50	by about a hundred thousand years ago
0:18:54	our life
0:18:56	took a very important turn.
0:19:01	We left Africa.
0:19:06	Leaving Africa
0:19:08	is a hypothesis
0:19:10	that even Darwin mentioned
0:19:13	in the,
0:19:15	1871 book, Descent of Man.
0:19:19	But it's only within recent decade that the evidence become so strong.
0:19:25	The earliest evidence came in the 1980's
0:19:29	with mitochondrial DNA. Mitochondrial is only materially transmitted
0:19:36	and then came a whole spate of studies based on y-chromosome,
0:19:41	which is paternally transmitted
0:19:44	and now
0:19:46	they can do a huge amount of study
0:19:49	based on
0:19:50	single-nucleotide polymorphisms.
0:19:53	And it's largely
0:19:55	these genetic studies that leads to this picture
0:19:59	which was published in PNAS
0:20:02	a couple years ago
0:20:05	by a group of a geneticists at Stanford University.
0:20:12	So the scenario,
0:20:14	the best knowledge that we have for now,
0:20:18	is that humans
0:20:19	originated in Africa.
0:20:23	The earliest fossils take back to about hundred and fifty thousand years ago.
0:20:29	And
0:20:30	various genetic mutations took place
0:20:34	at around the same time. For instance
0:20:37	we have a particular version
0:20:39	of a gene called FOXP2,
0:20:42	okay.
0:20:43	Our version of the FOXP2 evolved probably
0:20:46	also around that time.
0:20:48	A whole bunch of things happened around that time
0:20:52	and
0:20:53	emigration started.
0:20:55	First to Asia,
0:20:57	back to Europe,
0:20:59	down to Oceania,
0:21:02	across the Bering Strait into the Americas.
0:21:05	And very quickly
0:21:06	all the way down to the, tip of South America, Tierra Del Fuego.
0:21:13	So here is an example of a genetic study,
0:21:16	done essentially by the same group at Stanford,
0:21:20	showing
0:21:21	how the
0:21:22	SNP, single-nucleotide
0:21:25	polymorphisms are related.
0:21:27	So here's the root of the tree
0:21:31	and in constructing these phylogenetic trees
0:21:34	often you have to take
0:21:35	outgroup.
0:21:37	Outgroup in this case is the chimpanzee.
0:21:40	Chimpanzee diverged from us about six million years ago.
0:21:45	So,
0:21:46	the chimpanzee provided a good
0:21:48	routing of the tree.
0:21:51	These are all Africans
0:21:54	and then step by step populate in the world.
0:21:59	Here's Asia,
0:22:02	here's the northeast,
0:22:06	here is
0:22:07	South America
0:22:09	and here are the
0:22:11	Pacific Islands.
0:22:17	What about language?
0:22:20	Obviously one people left.
0:22:23	They brought their languages with them.
0:22:26	And another
0:22:28	Stanford scholar professor Joseph Greenberg, was
0:22:31	actually my teacher,
0:22:36	classified
0:22:38	all of the seven thousand
0:22:41	or so languages of the world
0:22:44	into these major
0:22:46	superfamilies.
0:22:50	So, for instance, again just as with people,
0:22:55	languages have the greatest diversity in Africa.
0:23:01	A very successful
0:23:04	family of languages
0:23:05	is called Eurasiatic.
0:23:11	Europe .. Asia. Eurasiatic.
0:23:16	And within this family
0:23:18	there's one sub-family that's
0:23:21	extremely
0:23:22	dominant. That's called Indo-European.
0:23:27	Indo-European because at one end you have India
0:23:31	at the other end
0:23:32	you have Europe extending all the way to Iceland.
0:23:37	So English
0:23:40	is Eurasiatic,
0:23:42	because English
0:23:44	is West Germanic. Germanic .. Indo-European ..
0:23:47	Eurasiatic.
0:23:49	Going up the family tree.
0:23:53	Another language here .. Tamil.
0:23:56	Tamil is not related to any of the Eurasiatic languages.
0:24:00	Tamil is a member of Dravidian family.
0:24:06	Earliest forms of Tamil actually
0:24:08	were found
0:24:10	much more
0:24:12	northwestern.
0:24:14	Where it's not spoken.
0:24:16	It was founded in the Indus Valley.
0:24:19	With the coming of the Indo-Europeans
0:24:21	they were push further and further south, so that
0:24:25	Tamil is now in this small region here.
0:24:31	Malay as we saw in the last slide
0:24:34	is an Oceanian language
0:24:37	that
0:24:38	has a relatively recent history,
0:24:41	but what about Chinese?
0:24:45	Chinese belongs to a family called Sino-Tibetan.
0:24:49	Chinese is related to Tibetan.
0:24:52	Well, recognized
0:24:53	over a hundred years ago.
0:24:56	And
0:24:58	more recently
0:24:59	people like Greenberg say Chinese
0:25:02	is Dene-Caucasian.
0:25:05	Chinese is here in red,
0:25:07	but there are spots of red
0:25:11	all over the world.
0:25:14	According to several very eminent linguists, Greenberg's one of them,
0:25:18	Sergei Starostin in Moscow is another one, okay,
0:25:23	say all these languages are related.
0:25:26	Chinese is related
0:25:27	not only to Tibetan,
0:25:29	but also to the Yeniseian languages.
0:25:33	But not only to the Yeniseian languages,
0:25:35	but all the way across Bering Strait
0:25:38	the Athabaskan languages.
0:25:40	Including for instance
0:25:42	Navajo.
0:25:46	So,
0:25:46	together with genes, archaeology and so on
0:25:50	we are getting a better notion
0:25:54	of what we come from
0:25:56	and speech plays a very important role here.
0:26:02	Given all these languages,
0:26:04	how do we classify them?
0:26:07	Well, one way to classify is very straightforward, you know, where is it spoken?
0:26:12	Geographical.
0:26:14	This gives us a geographical distribution.
0:26:19	Another way is
0:26:21	typological.
0:26:22	What kind of structure does it have?
0:26:26	Does it have a gender system?
0:26:30	Is indirect object before the direct object?
0:26:34	Or is the other way?
0:26:36	Does it have tongues?
0:26:39	These are typological features.
0:26:40	So, this is a second way of classifying languages.
0:26:44	And the third way of classifying languages,
0:26:46	what is it related to?
0:26:49	Is Chinese related to Tibetan?
0:26:51	Is Tamil related to Telugu?
0:26:54	Which it is.
0:26:55	So on.
0:27:00	Sometimes
0:27:02	genetic or
0:27:03	historical relationship
0:27:06	and typological relationship
0:27:10	classified differently.
0:27:13	English for instance, as I said a few minutes ago,
0:27:16	is the Germanic language.
0:27:18	But it doesn't look anything like German.
0:27:21	English does not have chen (suffix),
0:27:23	a gender system.
0:27:24	German has three genders: masculine, feminine, neuter.
0:27:30	In English the primary
0:27:32	order for declarative sentences .. you put the verb in the middle,
0:27:36	I saw John.
0:27:39	In most of other Germanic languages, including German,
0:27:43	in declarative sentences
0:27:45	the verb is at the end.
0:27:47	All these structural differences.
0:27:50	So, when we want to use corpora of languages
0:27:54	to
0:27:55	use cross-linguistically
0:27:57	it's primarily the typological
0:28:00	classification that's important.
0:28:02	Not the genetic relation that's important.
0:28:08	Well, let's ask the question,
0:28:11	How did all this
0:28:12	diversity arise?,
0:28:15	Where did all this diversity come from?
0:28:21	For instance
0:28:22	here's a distribution by
0:28:25	area. This is the geographical distribution.
0:28:30	This was shown by an earlier speaker too.
0:28:35	According to the Ethnologue which is
0:28:38	essentially
0:28:39	missionary organisation where the team of thousands of linguists
0:28:44	travelling far-flung places of the world.
0:28:48	With one of their missions to translate The Bible
0:28:52	and in this process
0:28:54	they live there, they know the region well, they can report on the language and
0:28:58	according to the
0:28:59	Ethnologue there are seven thousand or so languages.
0:29:06	We shouldn't take these numbers very seriously. They are very good
0:29:11	rough guide.
0:29:13	For instance
0:29:16	when I went to Hong Kong, I couldn't understand the
0:29:20	word of the Chinese that they spoke there .. Cantonese.
0:29:24	Cantonese and Mandarin are supposed to be Chinese.
0:29:28	One language.
0:29:31	If this is so, then the
0:29:33	dozens of dialects in China only count once, even though they're mutually unintelligible.
0:29:40	On the other hand one year I was teaching in Stockholm.
0:29:44	Bought myself a little Volvo,
0:29:46	drove across from Stockholm to Oslo.
0:29:49	Swedish .. Norwegian.
0:29:50	No problem at all communicating.
0:29:54	Yet there we have two languages, okay.
0:29:56	So these numbers
0:29:58	just can't be taken too seriously,
0:30:00	but it's good to have kind of as a
0:30:03	guideline.
0:30:09	Was it
0:30:11	also this complicated at the beginning when we first heard language?
0:30:16	What does this diversity come from?
0:30:20	Well, linguists have two ways of talking about the origin.
0:30:26	Either
0:30:27	monogenesis,
0:30:29	all of the languages in the world
0:30:32	came from a single source
0:30:35	or at the beginning
0:30:38	there were various tribes in different regions
0:30:41	of the world maybe,
0:30:43	of Africa maybe,
0:30:45	and each of them came upon the idea of inventing language.
0:30:51	And some years back
0:30:54	I collaborated with the
0:30:55	mathematician, friend of mine at Berkeley, David Freedman.
0:30:59	And we did some statistical modeling
0:31:02	and
0:31:04	our thought
0:31:06	that is
0:31:07	much more likely,
0:31:09	given you have
0:31:11	a large number of sites,
0:31:13	so much more likely
0:31:15	the language arose
0:31:16	polygenetically.
0:31:19	This means, if this is right,
0:31:22	that some diversity was there
0:31:24	right at the beginning.
0:31:27	And this is not completely unreasonable because
0:31:32	you think about the invention of fire
0:31:35	that's had terrific consequences.
0:31:38	That was polygenetically invented.
0:31:41	Many people invented them independently.
0:31:45	You think about pottery.
0:31:47	Partly another major
0:31:49	prehistoric invention.
0:31:51	Independent.
0:31:52	More polygenesis.
0:31:54	You think about written language.
0:31:57	Many cultures independently invented written language.
0:32:01	So, spoken language, you know, not in the nuanced
0:32:04	powerful sense of spoken language now.
0:32:08	Spoken language
0:32:10	hundred thousand years ago
0:32:12	could have been invented polygenetically.
0:32:18	One person who took the monogenetic
0:32:21	approach
0:32:22	was Quentin Atkinson
0:32:25	who published this paper in Science few years back.
0:32:30	He
0:32:31	use the
0:32:32	corpus
0:32:33	world atlas of language structures which is maintained in Leipzig in Germany.
0:32:40	And looked at five hundred and four languages
0:32:43	in terms of their phoneme inventories.
0:32:47	How many consonants, how many vowels and so on.
0:32:53	And you can tell from the title of his paper
0:32:56	Phonemic Diversity Supports
0:32:59	a Serial Founder Effect Model
0:33:02	of Language Expansion.
0:33:05	In another words
0:33:06	he's trying to
0:33:08	construct the picture
0:33:11	kind of like this that we saw earlier for genes.
0:33:15	A serial founding effect.
0:33:17	He wants to do it
0:33:20	for languages and he does it
0:33:23	with this diagram
0:33:24	that you see on the
0:33:27	left side of the slide
0:33:30	from where you seti.
0:33:35	Without going into great detail of Atkinson's methodology,
0:33:40	I should mention that
0:33:43	his proposal was met with several very severe
0:33:47	criticisms
0:33:49	from three groups of scholars.
0:33:52	Some in Europe, some in America and so on.
0:33:57	And they are listed on the slide.
0:34:01	For instance
0:34:04	Cysouw
0:34:05	Dediu and Moran,
0:34:07	in the later issue of Science,
0:34:11	contested
0:34:13	both their databases
0:34:15	and their method
0:34:17	and came up with a very different picture
0:34:20	based on phoneme inventories.
0:34:24	Atkinson's
0:34:25	hypothesis would put
0:34:28	origin of language here, but actually there are many other sites.
0:34:37	In order to give us a clearer idea of how sound patterns actually move
0:34:43	I will now move to
0:34:44	three case studies.
0:34:50	Atkinson took the seven thousand languages of
0:34:54	global perspective, all the languages of the world.
0:34:58	I think perhaps we're not quite ready
0:35:00	to be that global.
0:35:03	It might be safer to stick with one family on which there is a very
0:35:10	detailed meticulous scholarship
0:35:11	that has been going on
0:35:13	since
0:35:14	the
0:35:15	eighteen century, latter part of the eighteenth century.
0:35:19	So, in Science 2004
0:35:22	we have a
0:35:24	picture of the Indo-European languages.
0:35:28	I'm not sure
0:35:29	whether you can catch everything
0:35:32	but there is not an English.
0:35:37	Traces to
0:35:39	Middle English of Chaucer,
0:35:41	Old English of King Alfred
0:35:44	and all the way up to Indo-European.
0:35:49	Indo-European also includes
0:35:52	Tocharian.
0:35:53	Which is an language that's become extinct.
0:35:57	You find remnants of it in western in China, in Xinjiang.
0:36:03	And you have
0:36:05	the
0:36:06	Indo-Iranian branch. Within the Indo-Iranian branch
0:36:10	you see Kurdish,
0:36:13	which is very much in the news these days. They're the ones fighting in Iraq.
0:36:18	So, Kurdish is actually an Iranian language belonging to the Proto-Indo-Iranian family.
0:36:26	This gives you an idea of the membership
0:36:29	doesn't tell us very much about the time scale.
0:36:34	In Nature in 2003
0:36:37	Russell Grey
0:36:39	and Atkinson
0:36:41	did an extensive
0:36:44	statistical analysis based on vocabulary.
0:36:49	They took almost two thousand five hundred words
0:36:52	from eighty seven languages,
0:36:55	all of from Indo-European and constructed this tree.
0:37:01	Along this tree
0:37:03	you see estimated time of death.
0:37:06	So, according to these people published in Nature in
0:37:11	2003
0:37:13	Indo-European
0:37:15	is about nine thousand years old.
0:37:21	As it gradually split
0:37:23	into
0:37:25	more language families
0:37:28	based on innovations,
0:37:30	okay.
0:37:32	A few thousand years ago
0:37:34	it
0:37:36	led to Germanic languages.
0:37:39	Eventually to English.
0:37:44	Those are
0:37:47	the,
0:37:48	the slide is kind of hard to see,
0:37:50	so I will take this section here
0:37:54	and blow it up
0:37:55	and we have this.
0:37:59	So, here we have English,
0:38:01	here we have German,
0:38:04	and we have the Romance languages,
0:38:08	Italian, Spanish,
0:38:11	and a separation here seems to be about
0:38:14	five thousand years ago.
0:38:21	What I'd like to
0:38:23	discuss now
0:38:25	is what other innovations
0:38:29	that led to just
0:38:31	this one little group.
0:38:35	How did
0:38:37	Indo-European
0:38:39	become Germanic
0:38:41	in a part of the world?
0:38:46	And this is how linguists operate, okay.
0:38:49	Here's some data.
0:38:52	Colin Renfrew is very esteemed archaeologist at Cambridge University
0:38:59	and he gave as a bunch of words.
0:39:01	The integers one to ten
0:39:04	in a bunch of languages.
0:39:11	English one, two, three, four, up to ten.
0:39:15	Japanese
0:39:16	hitotsu, futatsu, mittsu, yottsu.
0:39:19	Doesn't look anything like the rest of them.
0:39:23	So, Japanese
0:39:25	it's not an Indo-European language.
0:39:28	But the rest.
0:39:31	English
0:39:33	two
0:39:35	corresponds to Gothic.
0:39:37	Gothic is an extinct language but we know it because there's lot of textual material
0:39:42	preserved.
0:39:44	Gothic has twai.
0:39:46	Two ..
0:39:47	twai.
0:39:49	Why isn't the w pronounced in two
0:39:53	but is still written?
0:39:56	In Latin is duo.
0:40:00	In Greek
0:40:01	is duo,
0:40:03	in Sanskrit
0:40:05	it's dva.
0:40:07	In Russian is also dva.
0:40:12	Obviously there's similarity in there.
0:40:15	But is there something systematic that we can extract?
0:40:20	Yes, the systematic aspect is that
0:40:23	Germanic,
0:40:25	English and Gothic
0:40:27	T
0:40:29	corresponds
0:40:32	to other Indo-European languages'
0:40:36	D.
0:40:40	Let's take the next one.
0:40:43	Three
0:40:44	th is just a abbreviation
0:40:47	referring to the sound that Anne Cutler was talking about interdental fricative θ.
0:40:53	Three that's a interdental fricative.
0:40:56	So, here's a fricative, here's a fricative in Germanic
0:41:00	and these fricatives correspond to T here, T here,
0:41:05	T here,
0:41:07	systematically
0:41:09	as a pattern.
0:41:11	Without going through many more
0:41:16	this is the correspondence that we can extract
0:41:23	and
0:41:24	here's the great man
0:41:25	who formulated the law
0:41:28	Jacob Grimm.
0:41:30	Older brother of Wilhelm Grimm.
0:41:33	The brothers of the Grimm's Fairy Tales, okay. Jacob was also a great linguist.
0:41:38	In 1822
0:41:40	he published what's called the Grimm's Law.
0:41:43	He said
0:41:45	these
0:41:46	words
0:41:48	in Indo-European
0:41:51	pronounced with a
0:41:52	B(h)
0:41:53	D(h), G(h) type of sound,
0:41:57	became in Germanic
0:42:00	P,
0:42:00	D and G.
0:42:04	Sounds which were B, D, G in Indo-European
0:42:09	became P, T, K
0:42:13	in Germanic.
0:42:15	And sounds which where P, T, K in Germanic,
0:42:20	in Indo-European, sorry. Became F, θ, H
0:42:25	in Germanic.
0:42:27	Very regular.
0:42:30	We are able to look at some of these in comparison because
0:42:36	after the Norman conquest, about a thousand years ago,
0:42:40	English powered a lot of romance words.
0:42:44	French words, Latin words.
0:42:46	So we actually have
0:42:48	Latin and Romance words sitting side by side with the native Germanic words.
0:42:53	So take a word like
0:42:55	ped,
0:42:55	the pedal
0:42:57	on a bicycle.
0:42:59	It's foot.
0:43:01	So the P corresponds to the F.
0:43:04	P corresponds to the F.
0:43:09	The pad, the D
0:43:11	corresponds to the T.
0:43:13	D corresponds to the T.
0:43:20	So, the point that I wanna make with this slide,
0:43:24	I better move faster,
0:43:26	is that
0:43:27	it's a whole natural class that's displaced.
0:43:31	It's displaced not only as total patterns
0:43:35	but it's usually displaced
0:43:38	by single features.
0:43:40	As you go from here to here
0:43:43	you're going from
0:43:44	voice task ??
0:43:46	to voice ??.
0:43:49	As you go from here to here
0:43:50	you're going from voiced to unvoiced.
0:43:53	As you go from here to here you're going from stop to fricative and so
0:43:58	on.
0:43:59	It's
0:44:00	beautifully symmetric.
0:44:03	So, let's get closer to home.
0:44:07	Here's the sound change that took place.
0:44:11	It sided shortly after Chaucer's death
0:44:16	and it kept on going after Shakespeare.
0:44:19	It's called the Great Vowel Shift.
0:44:22	It's been studied in ??Greatstepft by Danish linguistic called
0:44:26	Otto Jespersen.
0:44:28	And some years back I also
0:44:30	followed up that study.
0:44:36	Why is it that we say
0:44:38	sanity,
0:44:41	but sane.
0:44:44	Why is it that in French
0:44:47	China's called Chine?
0:44:50	Chine versus China.
0:44:54	Christmas
0:44:55	versus Christ.
0:44:58	Why are there all these alternations?
0:45:02	It's because of the Great Vowel Shift.
0:45:06	And to hear one example.
0:45:11	The original vowel quality was actually a.
0:45:16	It used to be sanity.
0:45:20	But if you take away the suffix,
0:45:23	which protects this vowel,
0:45:25	the vowel itself raises.
0:45:27	So, it became sane.
0:45:29	So, we still say sanity
0:45:31	but we no longer say, he's insane.
0:45:35	We have to say insane.
0:45:38	As wind shift
0:45:42	we can say serenity
0:45:44	but we cannot say, this man is very serene.
0:45:48	We have to say serene
0:45:50	A has become E, E has become I. I can't go any higher.
0:45:55	So,
0:45:56	Christmas, the I,
0:45:59	has become Christ.
0:46:02	Chine
0:46:03	in French has become China in English, because of this vowel shift.
0:46:09	Again it's a systematic shift in patterns.
0:46:15	Okay, let's
0:46:16	spend a few minutes on
0:46:19	perhaps less familiar territory.
0:46:22	You build words not only with consonants and vowels
0:46:26	but also with tones.
0:46:29	People think of tones they think of Chinese
0:46:32	but there are thousands of tone languages in the world.
0:46:35	Here's a tone language called Trique,
0:46:37	in Mexico.
0:46:39	And in Trique
0:46:41	there are five levels of tones.
0:46:45	So, we have,
0:46:46	this is,
0:46:48	no,
0:46:49	gu du we ku.
0:46:51	Gu du we jo.
0:46:53	And so on.
0:46:57	Each with a very different meaning.
0:47:00	So, ku is bones.
0:47:03	Jo, ka, ?a, za.
0:47:07	Five different level tones.
0:47:09	With five different words.
0:47:19	But is it the case that
0:47:21	once you have a tone language always a tone language? No.
0:47:26	The paradigm tone language Chinese. People thought, Well, Chinese is a tone language.
0:47:31	Was it always a tone language? No.
0:47:35	Chinese became a tone language probably about
0:47:38	two thousand
0:47:40	twenty five hundred years ago.
0:47:42	And this is the result of
0:47:44	the loss of consonant clusters.
0:47:48	In Enghlish you have a lot of consonant clusters like
0:47:51	speech, play, spring, these all begin with consonant clusters.
0:47:57	But if you look at the Chinese dialects now, any dialect you want look at,
0:48:01	no consonant clusters.
0:48:04	So, how do we know that there was consonant clusters?
0:48:08	Chinese is not written alphabetically. How can you tell what it sounded like?
0:48:13	Well, it's true Chinese is not written alphabetically but a large part is written
0:48:19	phonetically.
0:48:22	It's not in terms of consonants and vowels. It's in terms of syllables.
0:48:27	So, for instance
0:48:30	here's,
0:48:33	I call these things sinograms
0:48:35	for Chinese characters. Here's a sinogram
0:48:39	that serves
0:48:41	as the phonetic
0:48:42	of this more complex sinogram.
0:48:46	Here's a sinogram
0:48:48	that serves as a phonetic.
0:48:51	Sinogram that serves as a phonetic and so on.
0:48:56	But then you ask me,
0:48:58	How is that phonetic?
0:49:01	They're different.
0:49:03	These are
0:49:05	spelled with a letter G but actually there's K sounds there.
0:49:09	Unvoiced un-aspirated velar stops,
0:49:12	okay.
0:49:13	These are all L.
0:49:18	Well, that's one of the
0:49:20	evidence
0:49:21	that they were consonant clusters.
0:49:24	If you take these words,
0:49:26	compare them with Tibetan.
0:49:28	Many languages in Southeast Asia
0:49:31	they also have constant clusters.
0:49:34	So, because these consonant clusters were lost,
0:49:38	tones arose.
0:49:39	Otherwise you'd just have too much ??.
0:49:45	And they arose in different ways.
0:49:49	So, here's Beijing,
0:49:51	Xiang, Xi'an, ??Hankou\Hong Kong.
0:49:54	and so on.
0:49:57	Each one
0:49:58	with the different number of tones.
0:50:01	In Mandarin they're four.
0:50:03	In Cantonese there are none.
0:50:09	So, this illustrates the four tones of
0:50:12	Mandarin.
0:50:14	This illustrates the nine tones of Hong Kong Cantonese.
0:50:19	Notice in Mandarin
0:50:21	because of the merger
0:50:24	??
0:50:26	??
0:50:27	??
0:50:29	have all become homophones.
0:50:31	Exactly the same pronunciation.
0:50:34	But in Cantonese
0:50:36	one is mid-level
0:50:39	??
0:50:40	??
0:50:41	??
0:50:43	three different tones
0:50:45	they kept.
0:50:48	And if you
0:50:49	did F0 analysis to extract the pitch.
0:50:54	This is actually.
0:50:55	my own voice
0:50:58	extracted from a PDP computer. Many of you may not even remember what PDP computers
0:51:04	are.
0:51:05	Back in 1973.
0:51:09	So, it is my voice saying
0:51:11	??
0:51:13	??
0:51:15	And here are the nine tones
0:51:17	from Cantonese.
0:51:23	You could plot these
0:51:26	on a
0:51:27	two-dimensional graph
0:51:29	using slope
0:51:31	and normalized height.
0:51:34	And you can see that the four tones in Mandarin
0:51:37	??
0:51:39	are quite evenly distributed.
0:51:43	On the other hand in Cantonese
0:51:47	it's kind of a mess.
0:51:50	There's a lot of overlapping
0:51:53	and this happens a lot whenever
0:51:55	a metropolitan centre
0:51:57	suddenly takes in a lot of influx of people
0:52:00	speak different dialects, different languages. The language will change faster.
0:52:05	So, as a result
0:52:07	in Cantonese tone two
0:52:09	and tone five
0:52:11	are merging.
0:52:13	Many of the words
0:52:15	are no longer distinct.
0:52:17	Tone three and tone six
0:52:19	are merging.
0:52:20	Many of them are no longer distinct.
0:52:30	I see I've ran out of time, Haizhou.
0:52:39	The last point on this tone part,
0:52:43	I'll just mentioned very quickly,
0:52:45	is that we saw a little bit earlier the vowel shift changed the vowels chasing
0:52:50	each other around, okay.
0:52:52	Low become middle, middle becomes high, high becomes
0:52:56	??deflunguised
0:52:58	Well, you can't say things in tones.
0:53:02	In Taiwanese
0:53:04	okay, here's
0:53:05	database on the Ministry of Education in Taiwan,
0:53:09	if you take each tones
0:53:11	and notice how they change.
0:53:14	They actually change in circle.
0:53:19	And detailed study of this
0:53:22	is available
0:53:25	in the literature.
0:53:30	The last point that
0:53:32	I wanted to discuss, but I can discuss with you individually,
0:53:37	is that many people think that language actually arose
0:53:41	from music.
0:53:44	The proto-language of the humankind
0:53:47	was actually singing.
0:53:50	Darwin himself
0:53:52	made many remarks to this
0:53:54	point.
0:53:57	So, in English, archaeologist Steven Mithen,
0:54:01	wrote the Singing Neanderthals
0:54:03	Origins of Music, Language, Mind, and Body.
0:54:08	And the linguist Fitch at University of Vienna
0:54:12	in his book Evolution of Language,
0:54:15	chapter fourteen
0:54:16	is all about
0:54:18	proto-language.
0:54:26	And here's a slide prepared by a graduate student at The Chinese University
0:54:31	showing
0:54:32	the
0:54:34	grammatical operators,
0:54:36	like drawing this type of trees,
0:54:38	applies very well
0:54:40	to music.
0:54:41	So, here's one of Bach's Menuets that
0:54:47	has
0:54:49	syntactic structures very much like
0:54:52	real sentences.
0:54:55	And this last slide shows that
0:54:58	actually
0:55:00	if you look at it just from the point of view region of interest with
0:55:05	fMRI
0:55:06	they're actually
0:55:07	distinct regions from music and language.
0:55:11	All this is very interesting, not too surprising
0:55:15	but we really need to go further
0:55:17	and look at the fiber tracks
0:55:19	not just the regions.
0:55:21	But how these regions are connected.
0:55:25	Okay, I have two more slides.
0:55:29	So, the roots of language then,
0:55:31	as a summary,
0:55:33	reach back over three million years,
0:55:37	when our remote ancestors transitioned
0:55:40	to bipedal posture,
0:55:43	restructuring the hands,
0:55:45	the vocal tract,
0:55:47	and the brain.
0:55:50	Speech
0:55:51	with its building blocks of syllables,
0:55:54	vowels,
0:55:55	consonants,
0:55:57	is a powerful vehicle for language,
0:56:00	emerged over a hundred thousand years ago.
0:56:04	Language and music are both universal to our species
0:56:09	and share evolutionary roots.
0:56:12	They have similar functions of communication,
0:56:16	similar principles of organisation.
0:56:21	Diversity in language is the cumulative product
0:56:25	of culturally selected innovations
0:56:28	made by numerous generations of speakers.
0:56:33	Spoken language has spawned
0:56:35	various auxiliary forms,
0:56:37	such as written language,
0:56:40	signed language, there is electronic media providing
0:56:44	additional windows for studying how we communicate.
0:56:49	And ever more powerful
0:56:51	technology of brain imaging and
0:56:53	computer analysis for spoken language and music
0:56:57	is already shedding much light
0:56:59	on our mind
0:57:01	and promises
0:57:03	to reveal
0:57:04	much more.
0:57:06	Thank you.
0:57:23	As much as we are a bit overtime I still welcome very short comments
0:57:27	after a mind-boggling talk that perhaps
0:57:32	bring back
0:57:34	speech and language
0:57:36	all the way to the roots.
0:57:43	If not
0:57:45	on behalf of Interspeech
0:57:48	we have a token of appreciation
0:57:52	that we would like to ask
0:57:54	the conference chair Haizhou Li to present to professor
0:57:59	William Wong.

Sound Patterns in Language

Keynotes

William Shi-Yuan Wang, Chinese University of Hong Kong, Hong Kong