YT YouTube2Text

YouTube2Text → Transcripts → Words and Rules: The Ingredients of Language with Dr. Steven Pinker (1999)

Words and Rules: The Ingredients of Language with Dr. Steven Pinker (1999) — Transcript

by The 92nd Street Y, New York · 15,215 words · 2,186 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:03the preservation of and increased access
0:05to the 92nd Street y Humanities audio
0:08archives is generously funded by the
0:10National Endowment for the Humanities
0:12good evening I'm Jennifer Juna howler
0:15from the Charles Simon Center for adult
0:17life and learning it's my pleasure to
0:19welcome you tonight and to introduce our
0:21speaker Dr Steph
0:23Pinker Dr Pinker will give a lecture and
0:26we've set up a microphone in the
0:28audience for you to ask questions after
0:30after the
0:30lecture after the program there will be
0:33a book selling and signing in the
0:34adjacent
0:36Hall we're so excited about our spring
0:38season and I just wanted to take a
0:40minute to share some of that with you
0:43before I introduce our speaker we have a
0:46wonderful Millennium series coming up uh
0:49it's featuring people like K Roberts
0:52William Sapphire Norman Lear Katie coric
0:56Gloria steinm Sylvia Earl Joyce Carol
1:00oats Mel caresan Alan dotz Susan sag the
1:06Reverend Jesse Jackson Peter Max stevenh
1:08J gold and many others so please keep an
1:11eye out for our new spring brochure
1:13which is coming out very soon and be
1:16sure to order tickets early for the
1:17programs that you're sure you don't want
1:19to
1:20miss our speaker tonight is a man who
1:23thinks deeply about thinking and we have
1:25been very fortunate to have him on our
1:28stage at the Y several times to share
1:30his fascinating insights with
1:33us he is Professor of psychology and
1:36director of the Center for Cognitive
1:38neuroscience at the Massachusetts
1:40Institute of
1:42Technology he's also a best-selling
1:44author who has already shed light on the
1:46workings of the brain and how it
1:48processes language in his book how the
1:50mind works which was nominated for the
1:53pet Sur prize and the national book
1:55award and was the number one bestselling
1:58book for Amazon . in
2:021997 his previous book The Language
2:05Instinct was also a bestseller and was
2:08named one of the best 10 books of the
2:10Year by the New York Times book
2:13review his new book which will be
2:16available to sign copies of after the
2:18lecture is words and rules the
2:20ingredients of language it explores the
2:22inner workings of the human brain as it
2:25moves from thought to sound to language
2:28please welcome Dr Steve
2:40Pinker thank you very
2:43much I'd like to begin by acknowledging
2:47uh some of the people who collaborated
2:49on the ideas and experiments I'm going
2:51to talk about this evening uh a large
2:53number of uh colleagues former students
2:56collaborators at MIT and in particular
2:58two who are in the audience tonight
3:00Professor Gary Marcus of NYU and
3:03Professor Annie sanghas of Barnard
3:07College language comes so naturally to
3:09us that we're apt to forget what a
3:11strange and miraculous gift it is but
3:14think about what we're doing right now
3:16for the next hour you're going to sit
3:18patiently listening to a man make noise
3:20as he exhales now why would you do
3:22something like that not because I can
3:25claim that the sounds are particularly
3:27mifl but because there's information
3:30that's coded into the precise sequence
3:33of hisses and hums and squeaks and Pops
3:36that I'll make that you can recover
3:38allowing us to share ideas now over the
3:42next hour the ideas that I'm going to
3:46cause you to think our ideas about this
3:48gift language itself but with a slightly
3:52different set of noises I could be
3:53causing you to think thoughts about
3:56anything from theories of the origin of
3:58the universe to the latest twists and
4:01turns of your favorite soap opera plot
4:04uh this is the miracle of language its
4:06vast expressive power and I think the
4:09first question that we have to ask in
4:12understanding language is what is the
4:14trick what's the trick behind our
4:16ability to fill each other's minds with
4:19so many different kinds of ideas the
4:22point of the book words and rules and of
4:24the talk this evening is that there's
4:26not one trick but two of them and they
4:28were both ident ified in the 19th
4:30century by European linguists the first
4:33is the principle of the memorized word
4:37what fand do soer called the arbitrary
4:40sign now the word duck for example
4:43doesn't look like a duck or walk like a
4:46duck or quack like a duck but I can use
4:48it to get you to think the thought of a
4:50duck because all of us at some point in
4:52our lives have memorized an association
4:55between that sound and that meaning that
4:58means that there has to be some
5:00information stored in the mind that
5:02captures that Association which in
5:04simplified form might look something
5:05like
5:06this there is a symbol for the word
5:09itself there is an uh entry for its
5:12sound duck and then some kind of entry
5:15for its
5:16meaning now words simple though they are
5:20convey a number of
5:22advantages uh for one because human
5:24memory is capacious there are a large
5:26number of Concepts that we can express
5:28with words it's been estimated that a
5:30typical high school graduate can uh has
5:34acquired on the order of 60,000 Words
5:37which works out to a rate of learning
5:40words of one every two waking hours
5:43starting at the age of one also the
5:46brain is well equipped to produce and
5:49understand words it takes less than a
5:50quarter of a second for the brain to uh
5:53come up with the meaning behind a word
5:56after hearing it or and also less than a
5:58quarter of a second uh given the thought
6:00to recover the information on how to
6:03pronounce it on the other hand there are
6:05some disadvantages to words you're stuck
6:08with a finite number of predetermined
6:10Concepts those for which your language
6:13already has uh sound meaning pairings
6:17and words are slaves to memory everyone
6:20has to have memorized a word for it to
6:22be useful if you're the only one who
6:23knows a word you may as well not know it
6:25at all that brings me to the second
6:28trick behind language combinatorial
6:30grammar what vilhelm Von humble called
6:33the infinite use of finite media the
6:36ability
6:37to uh produce strings of words in which
6:40the meaning of the string can be
6:42computed from the meanings of the
6:44individual words and the way that
6:46they're combined in order for a speaker
6:49and a hearer to uh convey information by
6:53ordering words they both have to share
6:55some kind of mental algorithm or recipe
6:58or software that uh captures this
7:01information and that's what uh linguists
7:03call a grammar again in simplified form
7:06a couple of rules of grammar that we use
7:09to express our thoughts are these a uh
7:12sentence is composed of a noun phrase in
7:15English followed by a verb phrase that
7:17is the subject followed by the predicate
7:19the verb phrase consists of a verb
7:21followed by a noun phrase the object
7:23followed by a uh sentence the compliment
7:27the power of rules comes from the fact
7:29that they're built out of symbols so for
7:31example the verb phrase symbol that's
7:34the name for the right hand rule is also
7:38listed in the leftand rule so that uh
7:41one can be inserted inside the
7:44other since the words as well bear
7:47symbols indicating how they can be
7:49slaughtered into sentences we have a way
7:52of uh with a finite set of rules of
7:54generating brand new structures on the
7:57fly now among the advantages of
8:00combinatorial grammar are that by
8:02combining symbols new ideas can be
8:05expressed there's a cliche in journalism
8:07that when a dog bites a man that isn't
8:09news but when a man bites a dog that is
8:12news the beauty of grammar is that it
8:14allows us to convey news by rearranging
8:18words in new
8:19orders also because grammar is couched
8:23in abstract symbols like noun and verb
8:26the same rule that allows us to convey
8:29uh dog bites man can also allow us to
8:31convey uh a big bang creates the
8:34universe or any other combination of
8:37ideas in any combinatorial system the
8:40number of combinations grows
8:42exponentially with the length of the
8:44sentence so if there are say 10,000
8:47nouns with which to begin a sentence and
8:494,000 verbs with which to continue it
8:52you already have 40 million uh ways of
8:55beginning a sentence and you're only two
8:57words into it and the number EXP
8:59explodes from there the psychologist
9:01George Miller once estimated that a
9:04person is capable of producing or
9:06understanding 100 million trillion
9:09different sentences 20 words long or
9:11shorter as uh the late Carl Sean might
9:14have said billions and billions of
9:18sentences also there's a sense in which
9:21the number of thoughts that we can
9:22express through language is not just
9:24humongous but it is technically infinite
9:27thanks to a trick of language called
9:30recursion recall that a uh sentence can
9:34contain a verb phrase and a verb phrase
9:37can contain a sentence that means that
9:39you can insert a sentence inside a
9:42sentence and a sentence inside a
9:43sentence inside a sentence therefore
9:46generating sentences of any length and
9:48hence an infinite number of sentences
9:50there is no longest sentence in
9:53principle now it turns out that this is
9:55not a banal uh observation because there
9:58has been a claim that there is a longest
10:01sentence in English where would you find
10:03such a claim well where else the
10:05Guinness Book of World Records world's
10:07longest sentence you could look it up
10:09it's 1,300 words long it comes from a
10:12novel by William Falkner which should
10:15come as no surprise to those of you
10:16who've read the novels of William
10:18Falkner uh I won't reproduce the
10:20sentence in its entirety but here's the
10:22way it begins they both bore it as
10:24though in deliberate flagellant
10:26exaltation and it runs on from there
10:30however I have an announcement to make
10:32this evening this is not in fact the
10:34world's longest sentence and I've been
10:37tempted to attain immortality in
10:39Guinness by submitting the following
10:41recordbreaker faler wrote they both
10:44bored as though in deliberate flatulent
10:48exaltation but then I realized that this
10:50would not be immortality after all but
10:52only the proverbial 15 minutes of fame
10:55because any one of you could then submit
10:57Guinness noted that f wrote they both
10:59bore it Pinker mentioned that Guinness
11:01noted that Falon wrote who cares that
11:03Pinker mentioned that Guinness noted
11:04that Falon wrote dot dot dot well
11:08grammar is a wonderful thing uh allowing
11:10us to convey an infinite number of
11:12Thoughts by rearranging a finite number
11:14of words are there any disadvantages to
11:17grammar well in fact there are and
11:19they're best appreciated by trying to
11:22imagine what a language would look like
11:24if it had nothing but grammar no words
11:28well we don't have to imagine it it's
11:29been imagined for us because in the uh
11:3316th and 17th and 18th century there
11:35were a number of proposals for perfect
11:38languages that would do away with this
11:41ugly process of wrote memorizing
11:45thousands of arbitrary sound meaning
11:47pairings which were considered an
11:49affront to Enlightenment sensibilities
11:52of rational design the most famous is
11:55the analytical language of John Wilkins
11:58the subject of a wonderful essay by
11:59Jorge Luis bores and also of a book by
12:03Umberto Eko in search of the perfect
12:05language Wilkins wrote we should by
12:07learning the names of things be
12:09instructed likewise in their natures and
12:12here's in uh brief how a system worked
12:15Zita in his artificial language meant
12:18dog but you didn't have to memorize that
12:21you could deduce it
12:23because Wilkins divided the universe
12:26into 40 ontological categories and
12:29assigned each one of them a sound so Z
12:32for example stood for animal I for
12:35quadrip t for rapacious terrestrial
12:37European canine and then a pinpointed
12:40the species similarly deba is a portion
12:44of the first terrestrial element uh AKA
12:48flame and Koba is consanguinous relation
12:51of direct ascendant uh to it father now
12:55needless to say this scheme did not
12:57catch on and here we are are centuries
13:00later still memorizing arbitrary words
13:04every two hours of our waking lives and
13:07it's easy to see why it didn't catch on
13:09for one thing you need complex mental
13:12computation for every word every time
13:15someone opens his mouth and blurts out a
13:16word you've got to play a game of 20
13:18questions to figure out what he's
13:20talking about and also the combinatorial
13:23power is overkill for most everyday
13:26Concepts we generally don't have to talk
13:28about a portion of a father or a fire
13:31with four legs and so a scheme that gave
13:34us the ability to do that is really more
13:37combinatorial power than we need it
13:40suggests that the basic design of human
13:42language tries to take advantage of the
13:45strengths of both uh tricks we have
13:48words for common entities for the dogs
13:52and ducks and men for which the
13:54psychological mechanism is memory and we
13:57have grammatical rules for novel
13:59combinations of entities for men biting
14:02dogs and big bangs creating universes
14:05for which the mental mechanism is some
14:07kind of online
14:08computation now how would you test an
14:11idea like this well ideally you should
14:13find a case in which words and rules
14:15express the same ideas but would still
14:18be psychologically and perhaps even
14:20neurologically
14:22distinguishable the point of words and
14:23rules is that we do have such a test
14:26case they're what we uh remember with
14:29Dread as regular and irregular forms
14:32just a little brush up uh regular verbs
14:35uh are uh verbs like walk walked jog
14:39jogged kiss kissed words that form their
14:41past tense in uh predictable ways for
14:44verb after verb all you have to do is
14:47add uh Ed and you have the past tense
14:49form the regular verbs are open-ended
14:52there are thousands of existing regular
14:54verbs in English and new ones are being
14:56added to the language all the time when
14:59the verb to facts came into common
15:00parlament about 15 years ago you didn't
15:03have to run to the dictionary to look up
15:05its past tense form you could figure out
15:07that it was faxed likewise when uh we
15:10first learned spam snarf M mosh and dis
15:13we knew that the past tenses were
15:15spammed snarfed munged moshed and
15:18dissed even uh children can do this if
15:21you bring a 5-year-old child into a lab
15:24and say uh here is a man who knows how
15:26to wug uh the child will easily say
15:30yesterday he wagged a form that they
15:32couldn't have memorized as a word
15:35because the experimenter had made it up
15:37beforehand but uh proof that children
15:40can uh extend the pattern to novel verbs
15:44and there's a sense in which all
15:46children show off this kind of
15:48creativity because children go through a
15:50a stage in which they make errors like
15:53ComEd and bringed and holded uh which
15:56they also couldn't have memorized
15:58directly from their parents but must
15:59have created in their own minds this
16:02brings me to the second kind of verb in
16:04English the irregular
16:07verbs verbs like bring brought hit hit
16:10go went Sing sang sleep slept make made
16:13ring rang fly flu which contrast with
16:16the regulars in just about every way uh
16:19in contrast to the monotonous
16:21predictability of the Irregulars verb
16:23after verb forming its past tense in the
16:25same way the Irregulars are quirky and
16:28ID syncratic the past of syn is sank but
16:31the past of cling is not clang but clung
16:34the past of think is neither than nor
16:36thunk but thought and the past of blink
16:39is not blank or blunk or blot but is
16:42regular
16:43blinked also it's a closed class there
16:46are about 165 uh irregular verbs in uh
16:49English and there haven't been any
16:51recent new ones well this leads to a
16:53simple theory that irregular forms are
16:56words that we learn in the same way that
16:57we learn uh Dog and Duck and man and so
17:01on uh in this case though one of the
17:03words that we learn brought is the past
17:05tense of another one and so that the two
17:07are connected but are otherwise separate
17:10entries in memory corresponding to
17:12separate words regular forms on the
17:14other hand don't have to be stored
17:16because they can be generated whenever
17:18you need them by a rule a rule exactly
17:22analogous to the ones that form
17:24sentences out of subjects and predicates
17:26and so on also their interaction is
17:28straightforward if a word can provide
17:30its own past tense form for memory the
17:32rule is blocked that's why we adults
17:34don't say things like bringed and hold
17:36it and go elsewhere or by default the
17:39rule applies that's why both children
17:42and adults uh can create a past tense
17:47form even if they don't have one in
17:48memory and that's how come we can say
17:50things like wagged and faxed and spammed
17:52and mosed and dissed well I could end
17:55the talk here and words and rules would
17:58end after chapter 1 uh were it not for a
18:01complication to this neat dichotomy and
18:04that is that the irregular verbs are
18:06shot through with patterns you find
18:08families of similar words like keep kept
18:11sleep slept feel felt dream dreamt wear
18:14War bear bore tear tour swear swore
18:16string strung swing swung sting stung
18:18fling
18:19flung also these patterns are not just
18:23inert redundancies in memory but they
18:26are occasionally generalized by Flesh
18:28and Blood uh speakers of the language
18:31every once in a while children will come
18:34out with an error like brang as the past
18:36of bring or uh bite boat wipe in the
18:40history of the language every once in a
18:43while one of these Innovative forms will
18:44gain a tow hold and uh stay in the
18:47language thereafter quit and caught for
18:50example are only a couple of hundred
18:52years old Jane Austin used quitted and
18:55George Washington wrote catched and uh
18:57the most recent irregular snuck sneaked
19:00into the language only about a 100 years
19:02ago and is now pretty unexceptionable
19:06among people under the age of 40 uh but
19:08uh still sounds uh weird and new to
19:11people over the age of
19:1440 this is especially obvious when you
19:16compare dialects of English which have
19:18uh preserved their separate patterns of
19:20Innovations in various rural American
19:23and British dialects you find forms like
19:25climb Clum uh drag drug or one that was
19:28used by President Clinton a couple of
19:30months ago he swole up and
19:33died and um honest he really said it uh
19:38and as a psychologist I've been trained
19:40not to accept anything as a fact until
19:42it's been tested in the lab on either
19:44rats or sophomores now to to my
19:48knowledge no one has studied irregular
19:49verbs in rats but if you bring
19:51sophomores into the lab and you say what
19:53is the past tense of to spling then
19:56about 80% of them will say Spang or or
19:58spung now this process was brought to a
20:01high art by the linguist and uh baseball
20:04pitcher and radio announcer Dizzy Dean
20:07who once uh gave the following play
20:09byplay the pitcher wound up and flang
20:12the ball at the batter the batter Swang
20:14and missed the pitcher flanged the ball
20:17again and this time the batter connected
20:18he hit a high fly right to the center
20:20fielder the center fielder was all set
20:22to catch the ball but at the last minute
20:24his eyes were blowned by the Sun and he
20:26dropped
20:27it
20:28so how do we account for the Dizzy Dean
20:31that's present in all of us well there
20:33have been two alternative theories Each
20:36of which tries to take one of these two
20:39tricks words or rules and stretch it to
20:42account for the other uh phenomenon
20:45according to the theory of generative
20:47phenology from the linguists Noom
20:49Chomsky and Morris Holly rules rule not
20:53only do you have a rule for the regular
20:55past tense aded but there is a family of
20:58rules that fiddle with vowels and
21:00consonants such as changing an i to a
21:03for cling
21:05clung
21:06the problem with the theory is how you
21:09target the right verbs how do you get
21:12the say an rule to apply to string sting
21:15and fling but not to apply to wish
21:19giving wash or FIB giving fub or fish
21:22giving fush and so on there are two ways
21:25in which you might do it and both of
21:26them are problematic you could stipulate
21:29them in a list which is what uh Chomsky
21:31and Holly did say that the rule changing
21:33it a applies to these 17 verbs in these
21:3717 verbs alone but then it's a an
21:40unexplained coincidence why the words in
21:43the list are so similar why you have
21:45string and sting and fling and not say
21:48FIB or uh wish on the other hand if you
21:51try to find the common pattern distill
21:54it out and append it as a condition to
21:56the rule such as
21:58consonant consonant the I vowel and then
22:01the um NG or sound the V or nasal it
22:05doesn't work it fail makes mistakes both
22:08of omission and commission it falsely
22:11includes bring whose past is not uh
22:13brung but brought and spring whose past
22:15is not sprung but sprang and it excludes
22:19stick which should be in because its
22:21past is stuck but doesn't uh fit the
22:24pattern with the sound at the end and
22:26also fails to include spin which really
22:29ought to be in there but also misses the
22:31condition of the rule by a whisker well
22:34this leads to something completely
22:36different the theory of artificial
22:38neural networks or connectionism uh
22:41proposed by David rumelhart and James
22:43mcland and 25 successor models According
22:46to which there are no rules they're just
22:49analogies from memory if you beef up
22:52memory in a certain way in particular in
22:55a pattern associator memory a kind of
22:58neural network you link sounds to sounds
23:01instead of linking words to words so in
23:04this kind of memory you've got a set of
23:07uh units Each of which stands for a bit
23:10of sound in the verb another set of
23:12units Each of which stands for a bit of
23:14sound in the past tense form every input
23:17is connected to every output the memory
23:19is trained by giving it walk walked and
23:22bring brought and Sing sang and play
23:25played and so on it forms associations
23:27between between sounds in the verb and
23:29sounds in its past tense such as in to
23:32Ang or o to O it superimposes these
23:36patterns across different verbs and
23:38therefore it can generalize to new verbs
23:41be if they overlap in sound with old
23:44verbs the reason that it does that is
23:46that similar words literally share
23:49representational real estate so anything
23:51connected to a sound from one word is
23:53automatically connected to that same
23:55sound for all similar words
23:58the problem with this theory is
24:00generalizing
24:02properly pattern associators generalize
24:04to new words that are similar to Old
24:06words and that that works pretty well
24:08for irregular patterns where people do
24:10the same thing both people and a pattern
24:13associator will Having learned that the
24:15past of cling is clung will generalize
24:18that the past of spling is spung it also
24:21works for the regular pattern if the new
24:24word is similar to Old regular verbs so
24:27for example if you um test people and a
24:31pattern Associated Mo model on a novel
24:33verb like tolip well that sounds a lot
24:36like clip flip trip slip strip nip Whip
24:39and so on and so the model like people
24:42can generalize to clipped the problem is
24:45it doesn't work with new words that are
24:47not similar to Old words such as to
24:49plump which doesn't rhyme with anything
24:51in English in fact it's kind of violates
24:54the sound pattern of English here though
24:56people can generalize the regular
24:58pattern to any new verb regardless of
25:00its sound but pattern associators can't
25:03do it if you compare them you find for
25:05example that people will say the past of
25:07plump is plumped but the best that the
25:09model could come up with was bro
25:11likewise uh people say trilled the model
25:14comes up with tree LT people uh say
25:17smeed the model came up with Leaf log
25:20people come up with friled the model
25:22came up with freezel uh and the reason
25:24is instructive that the a pattern of
25:27associator memory lacks the simple
25:29computational device known as a symbol
25:32or a variable like an x in algebra or
25:35verb in grammar it doesn't have
25:38something that can embrace all verbs
25:40regardless of its sound but instead it
25:42forms associations to the individual
25:44sounds if a new verb comes along that
25:47doesn't share any sounds with the ones
25:49in the training set the best that the
25:51model can do is cough up a hairball of
25:53bits and pieces that are the closest
25:55things it's Associated to the sounds
25:57that that are in the new uh word well
26:00this leads to a um modification of the
26:03words and rules theory that takes into
26:06account what these two systems are good
26:08and bad at um I will continue to argue
26:11that Irregulars are words stored in
26:13memory but here's the twist that memory
26:16is not a list of slots that's just not
26:18the way human memory works but is partly
26:21associative much like a artificial
26:24neural network pattern associator memory
26:27namely we connect sounds to sounds as
26:30well as words to words and that explains
26:33why people generalize irregular patterns
26:35to similar words as in Dizzy Dean and
26:38spling
26:39spung on the other hand I'll also argue
26:41that we still need a rule for the
26:43regulars and that explains how people
26:45can generalize the regular pattern to
26:47any word whether it's similar or
26:49dissimilar familiar or
26:51unfamiliar the prediction is that people
26:54uh can apply the regular pattern
26:56whenever memory fail fails for any
26:58reason that memory fails and this simple
27:00prediction helps to explain a wide
27:03variety of facts about language and
27:06people such as where regular and
27:08irregular forms came from the logic
27:11behind apparent grammatical quirks the
27:13use and misuse of Rules by children
27:16where language resides in the brain and
27:18how and why languages differ let me
27:21begin with where regular forms come from
27:24um a lot of information can be found
27:26simply in these statistical structure of
27:28the English language as we find it today
27:31um here's a top 10 list top 10 verbs in
27:35English in order of frequency that is
27:37number of occurrences in a million words
27:39of text B have do say make go take come
27:47see get notice that all 10 of these are
27:51irregular B was have had do did say said
27:54make made go went take took come came
27:56seesaw get got
27:58now there can't be a bottom 10 list uh
28:00for a frequency in the language because
28:03there's a 788 way tie for last place
28:07last place being one in a million which
28:09is the lowest frequency that you can
28:11measure in a million word Corpus but
28:13I'll give you the first 10 words from
28:15that list in alphabetical order here
28:18they are aate abbreviate aore ablate a
28:21bridge abgate a climatize acculturate ad
28:25miix adulterate notice that all 10 are
28:28regular Abate abated abbreviate
28:30abbreviated abore abort and so on
28:33there's a massive correlation in English
28:35and all other languages between
28:37frequency and irregularity the irregular
28:40words are the most commonly used and
28:42there's a simple explanation irregular
28:45forms depend on memory because they're
28:47idiosyncratic and have to be
28:48individually learned memory depends on
28:51frequency the more often you hear
28:53something the easier it is to remember
28:55if a word declines in popular ity uh it
28:58may not be uniformly memorized by a
29:01generation of children uh when it when
29:04the they aren't uniformly memorized the
29:06speakers will default to the regular Ed
29:08converting a verb from irregular to
29:10regular for them and for all subsequent
29:14generations and in fact we can see this
29:16in the history of the language in Old
29:18English and Middle English there were
29:20about twice as many irregular verbs as
29:22we find today if Cher were here right
29:25now for example he would tell you that
29:27the past of to cleave is clove likewise
29:30abide Abode chai chid cow crew gripe
29:34grope R Ro uh and so on the common ones
29:38in Cher's time have stayed irregular to
29:40the present day the rarer ones have
29:43defected to the regular side you can
29:46even feel that force of History acting
29:48today because the remaining infrequent
29:51Irregulars are already starting to sound
29:53a bit strange as if they're slipping out
29:55of the language complete this sequence
29:58for example I stride I stroe I
30:01have STD uh I guess STD but it doesn't
30:05sound completely natural similarly Smite
30:08SM slay slooh bid bade forsake forsook
30:12have a kind of stilted or uh stuffy or
30:15rusher sha feel to them something that
30:17you recognize from writing but that you
30:19wouldn't use in everyday conversation
30:22and you can predict that in a 100 years
30:24these will have gone the way of uh abode
30:27and roow and clove and so on in contrast
30:31low frequency regulars always sound fine
30:34if I ask you to complete this sequence I
30:36abgate I abgate I have abreg there's no
30:40problem uh because even though the
30:43frequency of abreg is the same as the
30:45frequency of Strode and
30:48slooh now you might say um okay well
30:51granted uh I don't use smote all the
30:54time but it's not as if I use the verb
30:56to Smite in uh every conversation either
30:59but the theory predicts that an
31:00irregular can be um an infrequent
31:03irregular can sound strange even when
31:05the verb itself is familiar which is
31:08what you would expect if in an irregular
31:10the verb and its past tense form are
31:13separate entries in memory one of which
31:15can be strong while the other one is
31:17weak take for example the the verb to
31:20forego which is reasonably uh common
31:24especially in sarcastic expressions like
31:26uh to for go the pleasure of as in the
31:28following sentence you'll excuse me if I
31:31forgo the pleasure of watching the video
31:32of your wife giving
31:34birth and now if you take this
31:36expression and you force it into the
31:38past tense you get last night I for went
31:41the pleasure of watching herbs vacation
31:43slides which is rather peculiar
31:46similarly It's Perfectly Natural to say
31:48I don't know how she bears that guy but
31:51I don't know how she bore that guy is
31:53almost
31:55unintelligible you might say I dig the
31:57doors man but you probably wouldn't say
31:59in the 60s your mother and I dug the
32:01doors son uh and here's an example that
32:04I recently clipped out of the newspaper
32:06from the comic strip called zits there
32:08two teenagers one says to the other this
32:10week totally bit bit okay this week
32:15bited that can't be right either this
32:17week has totally bitten almost not quite
32:21back on Monday I didn't know this week
32:23was going to bite so bad I hate
32:25conjugating irregular vulgarity
32:29note however that there's no problem
32:31conjugating regular vulgarities if the
32:33teenager said this week totally sucked
32:36then there would be no comic strip
32:38because there's no problem with the past
32:39tense form and the straightforward uh oh
32:43and um likewise if you choose other
32:46regular verbs that are common in the
32:49present tense but rare in the past tense
32:52um they're still perfectly fine I don't
32:53know how she copes with him I don't know
32:55how she coped with him no no problem she
32:58doesn't suffer fools gladly a uh cliche
33:01or an idiom none of them ever suffered
33:03fools gladly uh no problem again the
33:06explanation is that with irregular forms
33:08you've got two different entries in
33:10memory which can part company in terms
33:12of their familiarity one of them can be
33:15strong while the other one is faint
33:18whereas for an irregular you don't have
33:20to store a separate past tense form you
33:22just create it whenever you need it and
33:24so the past tense form inherits whatever
33:26sense of familiarity or unfamiliarity
33:30had accured to the verb itself well
33:33that's if that's a an explanation of how
33:35we tend to get regulars where do the
33:37Irregulars come from the general answer
33:40is that irregular forms are the fossils
33:42of dead rules Now by symmetrical logic
33:46if uh irregular forms depend on memory
33:50and are uh vulnerable in those
33:52circumstances in which memory is fragile
33:55then if regular forms depend on rule on
33:57a rule they should be vulnerable
33:59whenever a rule is fragile and a rule
34:02can be fragile when they it becomes too
34:05complicated or obscure for a generation
34:07of children to figure out um and indeed
34:10most of our irregular vowel change
34:13patterns come from rules that were
34:15active in the minds of people who lived
34:175,500 years ago the uh Proto
34:21indo-europeans For Whom the past tense
34:23of s was uh s giving us sink sign today
34:27and likewise bear bore was a rule
34:30governed form in their language and now
34:32is memorized uh today well what causes
34:36such a transition why would a perfectly
34:38good rule die out let me give you an
34:40illustration from a more recent case and
34:42one that's a little bit easier to follow
34:45um in the Middle English period uh the
34:49past tense rule was alive and well but
34:51you got to keep in mind that in those
34:53days people were into phonetic spelling
34:55and the spelling of English words
34:57reflected their pronunciation much more
34:59than it does today so ke P was actually
35:03pronounced uh Cape with a long a sound
35:06uh and the rule that added Ed it was
35:09then spelled with a t but it's the same
35:11rule that we have today there was a
35:13second rule in the language a rule that
35:15would shorten the vowel whenever you
35:18glommed on some extra stuff at the end
35:20of a syllable so caped would then get
35:23shortened to kept there this is a common
35:26kind of rule in languages because a
35:28syllable is a unit of timing in speech
35:31if you add a whole bunch of extra junk
35:33to the end of a syllable you tend to
35:35shrink or compress the vowel to keep the
35:38syllable within one tick of the speech
35:41metronome and that's why we have
35:43families of forms generated by these old
35:45rules like sleep slept creep crept feel
35:47felt and mean meant and it applied
35:49across the language and that's why we
35:51have forms like sheep Shepherd Christ
35:53Christmas deep depth at 55 and so on but
35:57then something happened in the 15th
35:59century in a revolution called the Great
36:02Vowel Shift people started to scramble
36:04their pronunciation of the long vowels
36:07what used to be pronounced a was now e
36:10what used to be e was now I and so on
36:13the result was that two forms that used
36:15to be perfectly obviously related kep
36:19and kept no longer were related but had
36:22qualitatively different vowels e and e
36:27uh the shortening rule died because
36:29children could no longer figure out how
36:30to derive one from the other because now
36:32they were completely different in kind
36:35and how we know that is that new words
36:37that entered the language after the
36:39Great Vowel Shift no longer underwent
36:41shortening rule we have perfectly
36:44regular forms like peep peeped from 1460
36:47seep seeped from 1790 leaving at behind
36:51as fossils all of these forms that
36:53formerly were generated in a predictable
36:56Way by rule
36:58well this leads to a more general
36:59question of the nature of language which
37:02is why irregular words sometimes show up
37:05with regular forms a uh puzzle that has
37:08long preoccupied the various language
37:11mavens uh word Watchers verb ofor uh
37:15guardians of proper style and so on for
37:18example you might say all my daughter's
37:20friends are low lives you wouldn't say
37:23all my daughter's friends are low lives
37:25even though the ordinary irregular
37:27plural of life is lives one might say
37:30I'm sick of dealing with all the Mickey
37:31Mouses in this organization not the
37:34Mickey
37:35mice um in the 20 years since the Sony
37:38Walkman uh was introduced no one really
37:42feels comfortable uh with the plural
37:44Walkman and uh some people like the
37:46owner of this Electronics shop actually
37:49bent the regular plural Walkman into the
37:51tubing of a neon sign let me give you
37:55one other example of this class of
37:56puzzles that's illustrated by two little
37:59Leaguers who have a lot of time on their
38:01hands as they're sitting on the bench
38:03why is home plate called that when it
38:05doesn't even resemble a plate why do
38:07they call it a strike when you've
38:09actually missed the ball and haven't hit
38:10anything at all plus why does a baseball
38:13manager wear a uniform when he never
38:15plays and why do they say a batter flyed
38:18out instead of flew out indeed no mere
38:21mortal has ever flown out to center
38:25field well
38:27the first thing that this phenomenon
38:29shows is that sound alone can't be the
38:32input to the inflection process because
38:35the same sound can go in one end of the
38:37Box such as fly and come out the other
38:40other end of the Box either as flyed or
38:43flew depending on some extra input and
38:46the question is what is that extra input
38:50the most popular theory is the extra
38:52input is meaning that the problem with
38:54low life and Mickey Mouse and Walkman
38:57and fly fly out is that you've stretched
39:00the original meaning of a word and when
39:02you do you weaken the associations to
39:05any irregular form stored in memory and
39:07you switch over to the
39:09regular the um problem with a theory is
39:13that
39:14um not to put too fine a point on it but
39:17that it's wrong that 99% of the time in
39:21which you stretch a word's meaning The
39:24Irregular form comes along unscathed bed
39:27for example in metaphors uh where you
39:30obviously by definition stretch the
39:31meaning of a word you never by making a
39:35metaphor alone lose an irregular man men
39:38chess man Chessmen not chess man's
39:42similarly Straw Men snowmen saw teeth
39:45God's children and so on in idioms uh
39:48say to catch a cold is a considerable
39:50stretch of meaning from the ordinary
39:52sense of the word catch but nonetheless
39:54the past tense is not catch to cold but
39:56caught a cold despite that tenuous
39:58threat of meaning similarly we get cut a
40:01deal not cut it a deal took a leak
40:03bought the farm hit the fan blew them
40:04off came off well put out went nuts got
40:07a life had a cow literally hundreds and
40:09hundreds of them all of which are
40:11perfectly happy to bring their irregular
40:14form along the a better theory is that
40:17the missing input is grammatical
40:19structure now what does that mean well
40:21remember that the point of grammar is to
40:23allow people to compute the properties
40:25of a new word from the properties of old
40:27words in the way that they're arranged
40:30the problem with regularized Irregulars
40:32is that they have to opt out of this
40:34scheme they have a structure that
40:36doesn't allow information about an old
40:38word in memory to apply to the new word
40:41let me uh now explain what I meant by
40:44that here's how we ordinarily build
40:46words according to grammatical
40:48Rules start off for example with the uh
40:51ordinary verb uh to eat with its uh
40:55sound its meaning and its irregular past
40:58tense form We join it with a prefix over
41:01to get a new word overeat and the
41:03question is how do we know how to use
41:05the word overeat or in the notation what
41:08do we put up here in this slot based on
41:11everything that goes down here well the
41:13answer is easy
41:15the properties of the new word come from
41:18the properties of a special entity
41:21called the head which is the rightmost
41:24uh morphe inside the word and the
41:28information stored with the rightmost
41:30word percolates up along this pathway
41:32and it applies to the word as a whole so
41:35for example what is the part of speech
41:37of overeat what kind of word is it well
41:39it's a verb because eat is a verb and
41:41that information floats upstairs what's
41:44the meaning of overeat it's a kind of
41:46eating eating too much because the
41:48meaning of eat is e finally what's the
41:52past tense of ovat it's over eight
41:55because the past tense of eat is eight
41:57and that information percolates uh
41:59upwards let me walk you through another
42:01example start out with the ordinary noun
42:04uh man with its sound its meaning and
42:06its irregular plural men join it to the
42:09verb work to form a new word Workman how
42:12do we know how to use the word uh
42:14Workman well we get the information from
42:17the rightmost word inside it the head
42:19man and it transfers upstairs along the
42:22right hand Edge so a Workman is a noun
42:25because man is a noun the meaning of
42:28workman is a kind of man a man who works
42:32because the meaning of Manan is man and
42:34the plural of workman is workmen because
42:37the plural of man is men and that
42:39information uh gets transferred
42:42upstairs but some words can't be allowed
42:45to get their properties from their
42:47rightmost element if they're going to
42:48work the way the speaker wants them to
42:50work the usual pathway has to be blocked
42:55when it is blocked
42:57the any irregular form is trapped in
43:00memory and can't be passed upstairs to
43:03the whole word the regular rule steps in
43:06as the default okay and this explains
43:09all of the funny examples that I've
43:11mentioned so far plus many others let's
43:14start out with low lives the thing about
43:16low life is that
43:18it's not a kind of life the way a
43:22Workman is a kind of man or overeating
43:24is a kind of eating a low life is a kind
43:27of person namely a person who has or
43:30leads a low life so when you form it
43:34start out with the ordinary noun life
43:36join it to the adjective low to form low
43:38life how do you know how to use it well
43:41you can't apply the usual mechanism
43:44because then a low life would be a kind
43:46of life which is not what you want it to
43:47mean so you have to erect a barrier the
43:50irregular plural stored in memory with
43:53life has no way of bubbling up to the
43:56uh topmost node and so you default to
43:59the uh regular and you get low lives
44:01instead of low lives and that happens
44:04whenever you have a compound that works
44:06in this way a still life for example is
44:08also not a kind of life but a kind of
44:10painting and that's why we say still
44:12lives not still lives sabertooths not a
44:15kind of tooth but a kind of cat flat
44:19foot not a kind of foot but a h slang
44:21term for a policeman hence uh plural
44:24flat foots and that explains uh the
44:27Walkman uh which is not a kind of man in
44:30the way that a Workman is a kind of man
44:33incidentally the Sony Corporation has an
44:35official answer to the question of
44:36what's the correct plural of wman the
44:39answer is wman brand personal stereos uh
44:42in the interests of trademark protection
44:44they don't want the term to be uh
44:46pluralized uh which would imply that
44:48it's generic and hence usable by
44:51competitors um now Mickey Mouse is uh
44:55also has to opt out of the usual scheme
44:57because it's a noun based on a name or
45:00an eponym start out with the English
45:02word for mus musculus namely Mouse Walt
45:05Disney turned that noun into a name by
45:08concatenating it with the name Mickey
45:10then an ordinary slang that got
45:13converted back into a noun but this time
45:15a noun that refers to a fool or a
45:18simpleton uh the that means that the
45:20usual pathway has to be blocked once
45:23when Walt Disney turned a noun into a
45:25name because ordinarily nouns give rise
45:27to new nouns not to something completely
45:30different like a name and again it had
45:32to be erected when in ordinary slang the
45:35name was converted back into a noun
45:37because ordinarily new names uh old
45:39names give you new names with the
45:42pathway block there's no way for the
45:44irregular plural mice to travel upward
45:47and hence you default to the regular and
45:50get Mickey Mouses again this is quite
45:52General whenever you pluralize a name in
45:54Toronto a uh hockey player is called a
45:57maple leaf and when six of them skate
45:59onto the ice they are the Maple Leafs
46:01not the Maple Leaves Rena used to export
46:05a car called the elf and a bunch of them
46:07in a lot was not a horde of elves but a
46:10bunch of elves you might say Michael
46:12Keaton starred in the first two uh
46:14Batmans not the first two
46:16Batman and you might say we're having
46:19Julia Child and her husband over for
46:21dinner you know the Childs are really
46:23great Cooks not the children are really
46:25grape
46:27cooks and finally we get to flying out
46:30where here the principle is uh well
46:33explained in the following dialogue
46:35between uh Calvin and his toy tiger
46:38Hobbs uh Calvin says I like to verb
46:41words what I take nouns and adjectives
46:45and use them as verbs remember when
46:47access was a thing now it's something
46:49you do it got verbed verbing weirds
46:52language maybe we can event eventually
46:55make language a complete impediment to
46:58understanding now in fact verbing uh
47:00doesn't make language a complete
47:02impediment to understanding but follows
47:04the scheme I've explained uh perfectly
47:06um the thing about flying uh out in
47:09baseball is that it means to hit a fly
47:12it's a verb that comes from a noun start
47:15out with the regular verb fly to slip
47:18the sirly bonds of Earth and dance the
47:20sky on laughter silvered Wings in
47:22baseball that c converted into a noun a
47:25fly as in a Fly ball or a pop fly
47:28meaning a ball hit on a uh conspicuously
47:32parabolic trajectory that got converted
47:35back into a verb to make an out by
47:37hitting a fly and so we have a verb from
47:40a noun from a verb the ordinary pathway
47:43has to be blocked had to have been
47:45blocked when a fly was created because
47:48ordinary ver ordinarily verbs give rise
47:50to verbs not nouns had to be blocked a
47:52second time when the noun got converted
47:54back into a verb because ordinarily
47:56nouns be nouns that means that flu and
48:00flown are trapped here in memory and
48:02have no way of percolating up and
48:04speakers default to the regular giving
48:07flly again that happens whenever you
48:09have a verb from a noun if Us in hockey
48:13someone uh decapitates an opposing
48:15player and is sent to the Penalty Box
48:17for 2 minutes for high sticking then we
48:20say that he High sticked the goalie not
48:22that he High stuck him because it comes
48:24from the noun a stick you'd say Ventura
48:26Grand standed to the audience played to
48:29the grandstand a noun not Grand stood
48:31and Powell ringed the city with
48:33artillery not rang or rung because this
48:36version of the verb uh to ring means to
48:38form a ring around again a noun has been
48:41verbed to give rise to it let me turn
48:45now to Childhood and to the phenomenon
48:46that I mentioned way at the beginning of
48:48the talk that's Illustrated in the
48:49following cartoon mommy dolly hitted me
48:53Dolly hit me you too boy she's in
48:56trouble uh this is the phenomenon of
48:59children
49:00overgeneralizing uh the Ed to forms that
49:03don't ordinarily take it to produce
49:05errors such as we holded the baby
49:07rabbits the alligator go klunk Horton
49:10here to who or my brother got sick and
49:13puked
49:14it now I think there's a simple
49:16explanation that children's memory
49:18retrieval is less reliable than adults
49:21not because children have a bad memory
49:23but because they haven't lived as long
49:25that's what being a child means if
49:28you're only 3 years old you haven't
49:29heard forms like uh hold held and come
49:32came as often as an adult in cases of
49:35where the child fails to retrieve an
49:37irregular because it's uncommon and the
49:40child has the rule the child applies the
49:42rule to The Irregular to fill the Gap
49:45resulting in an error like here or go
49:48what's the evidence well the evidence
49:51that weak memory is a factor comes from
49:53the discovery uh made by Gary Mar
49:56and uh uh myself that the less often a
49:59parent uses an irregular form the more
50:01often the child will make an error on it
50:03this was true of all 19 of the children
50:06that We examined evidence that the uh
50:09rule has been acquired and that that's a
50:11crucial factor is comes from the fact
50:14long known in child psychology that
50:16errors with the irregular forms don't
50:18appear right away that children go
50:20through a phase in which they don't
50:23produce these errors uh before they pop
50:25out with their first one here's an
50:27example from uh one boy where we have
50:30plotted the percent of irregular past
50:33tense forms that are correct that is
50:35what proportion are uh heard and came
50:38and went as opposed to heared and ComEd
50:40and goed each dot is a month and for
50:43eight months in a row this boy never
50:45produced an error before he came out
50:47with his first one shortly shortly
50:49before his third birthday at which point
50:51he continued to make them at a rate of
50:53about 5% for several years uh afterward
50:56well what happened at this point here
50:58why does a child wake up one morning and
51:00start to say uh Horton here to who um
51:03the answer comes from when you look at
51:05what the child was doing with the
51:06regular verbs in his vocabulary uh
51:09whether he was saying yesterday we walk
51:12uh or yesterday we walked and if you
51:15look at that you see that the boy goes
51:19through a transition from leaving the
51:21regular forms unmarked most of the time
51:23to supplying them more often the not
51:26that is going from saying yesterday we
51:28walked to yesterday we walked and the
51:30transition is exactly the point at which
51:32he makes his first error like ComEd and
51:35heared so this is in a sense a picture
51:38of a that catches the child in an act in
51:40the act of acquiring a rule before he
51:43had the rule when memory came up
51:45empty-handed and he couldn't retrieve
51:48the past tense of say break he had no
51:51choice but to say yesterday we break it
51:54after he has the rule then he it's
51:56possible to say yesterday we breaked it
51:59and so the graph shows the rule coming
52:01into existence both in the child
52:04applying it correctly where it does
52:06belong and applying it incorrectly where
52:08it doesn't belong let me switch now to a
52:11completely different kind of evidence uh
52:13cases where the word system or the rule
52:15system in the brain are directly damaged
52:18by disease or injury the prediction is
52:21that damage to memory for Words should
52:23hurt the Irregulars damage to the brain
52:27system for grammatical computation
52:29should hurt the regulars and uh Michael
52:32Alman and I looked at a uh set of
52:34neological patients and we began with
52:36patients with damage to the this part of
52:39the brain the uh left hemisphere uh on
52:42the side in the parasylvian area towards
52:46the rear which gives rise to a symptom
52:49called anomia typically where the
52:52patient has difficulty retrieving words
52:54they'll be unable to to name objects
52:56they will be um uh use a lot of generic
53:00words like thing and guy and stuff but
53:02nonetheless can be fluent and can speak
53:05in more or less grammatical sentences
53:08but without the words slotted in on time
53:11we found that irregular forms for this
53:13kind of patient were harder than
53:15regulars which is exactly what we would
53:16expect if Irregulars depend on word
53:18memory that the patients make a large
53:21number of Errors like children swimmed
53:23and breaked and and that they are
53:26pretty good at doing a wug test today I
53:28wug yesterday I waged uh because that
53:31does not depend on memory which is what
53:33is impaired we compared them with
53:35patients with damage to the front half
53:37of the brain in the same hemisphere
53:39where uh a stroke will often leave the
53:42patient with a symptom called a
53:44grammatis where the patient will have
53:46difficulty assembling words into
53:48sentences and uh speaking in fluent
53:52grammatical utterances often the
53:54patients will have less severe
53:56difficulty in coughing up individual
53:58words we found as predicted that this
54:00time the Irregulars are easier than
54:03regulars because the easier the
54:04Irregulars depend on word memory which
54:07is more preserved in these patients
54:09there were few or no regularization
54:12errors like swimmed which depend on the
54:14operation of a rule and they could not
54:16do a wag test uh which also depends on
54:18the operation of a rule we got a similar
54:21dissociation between Alzheimer's disease
54:24patients who of course have a severe
54:27impairment in memory often including
54:29memory for words and who therefore find
54:31irregular forms harder as compared to
54:34Parkinson's disease patients who have
54:36difficulty in planning and sequencing
54:38speech and therefore have more problems
54:41with regular verbs which depend on
54:43assembling things in real time well um
54:47before I conclude I just want to mention
54:49one possible confounding uh
54:51factor and that is that regular verbs
54:54are the majority in English um I
54:57mentioned briefly at the very beginning
54:59of the talk that English has uh
55:02thousands and thousands of regular verbs
55:04but only about 165
55:06Irregulars perhaps the regular pattern
55:09is generalized more freely just because
55:11it's most frequent not because it's
55:14computed by a separate system in the
55:16brain um in order to rule out that
55:20possibility you would need to find a
55:22language where a regular or default rule
55:25rule applies to a minority of forms now
55:28note that this at first might look like
55:30an oxymoron because when you learn a
55:32language often the teacher will simply
55:34equate the regular form with the one
55:37that's most common in the language but
55:40in the way that I've been using regular
55:42it doesn't apply to the statistics of
55:44words in the language but to a bit of
55:46psychology namely whatever operation the
55:49speaker applies when memory comes up
55:52empty-handed and so in theory there
55:55could be a language in which a regular
55:57rule only applied to a minority of forms
56:00the question is could such a language
56:02exist in reality could there be a
56:04language that's so malicious so Twisted
56:10so perverse that would actually Force
56:12its speakers to memorize by rote a
56:16majority of the conjugational and
56:18declensional forms in the language well
56:21I going to read you now a quote from an
56:23essay by Mark Twain
56:27called the horrors of the German
56:30language a person who has not studied
56:33German can form no idea of what a
56:34perplexing language it is one is one has
56:37washed about in it hither and thither in
56:39the most helpless way and when at last
56:41he thinks he has captured a rule which
56:43offers firm ground to take a rest on
56:45amid the general rage and turmoil of the
56:4810 parts of speech he turns over the
56:50page and reads let the people make
56:52careful note of the following exceptions
56:54he runs his eye down and finds there are
56:57more exceptions to the rule than
56:59instances of it perfect and indeed uh
57:03German has uh three kinds of participles
57:06that they it forces its Learners to cope
57:09with ones with a vowel change and en
57:11which are uh the vowel changes are
57:14unpredictable hence irregular ones with
57:17an unpredictable vowel change and T the
57:19so-called mixed verbs also irregular and
57:22finally participles that simply add the
57:24T which are regular the uh plural system
57:28is even more horrifying with eight kinds
57:31of plurals four different suffixes uh e
57:35r en and S some plurals that don't have
57:38a suffix and then just to keep Learners
57:39on their toes uh some of them can
57:42unpredictably change the vowel in the
57:44process called umlout um nonetheless
57:47this is what uh Gary Marcus and I call
57:50the exception that proves the rule
57:53because despite a huge difference in the
57:56statistics of English versus German Ed
57:59being uh applying to 86% of the top
58:03thousand verbs in English and only 45%
58:05of the top thousand verbs in German
58:08psychologically they behave in virtually
58:10the same way so we use the Ed for rare
58:14verbs like ablated where we couldn't
58:16have memorized the past tense form
58:18German speakers use T for rare verbs
58:21such as gtit where they couldn't have
58:22memorized the past tense form soldered
58:26we use it for weird verbs that uh can't
58:28be analogized from something similar in
58:30memory German speakers use it for weird
58:33verbs as in a g PL PL by the way is
58:36German for
58:40plum in English we verb things and when
58:43we do we use the Ed as in fed out in
58:46German you can also verb things and they
58:49use it for uh the T for verbs from nouns
58:52as in gous housed and English-speaking
58:55children say singed German speaking
58:58children say uh gasin the plural system
59:01offers an even more dramatic contrast uh
59:0599.6% of English nouns take s only 7% of
59:09German nouns take s nonetheless we have
59:12plumps they have plows we have the Julia
59:15Childs they have uh Thomas man and his
59:18wife uh Mrs man D Thomas mans not D
59:21Thomas
59:23Menor we enjoy
59:25uh films about the cape Crusader
59:27Batman's they enjoy films about uh The
59:31Alchemist F uh
59:34FS we import eles they export
59:38Cadets English-speaking children say
59:40men's germane speaking children say mans
59:43despite the fact that they've only heard
59:45a smattering of nouns with s on it this
59:48leads me to the final point which is why
59:51English and German are similar in
59:52grammar but different in number numbers
59:55and for that you need a mixture of
59:57history and
59:58psychology in protog Germanic the
1:00:01ancestor of English and German spoken
1:00:03more than 2,000 years ago a majority of
1:00:05verbs are what we would today call
1:00:08irregular but even then an ancestor of
1:00:11the Ed or t suffix was used to for words
1:00:14that were borrowed from other languages
1:00:16and hence had strange sounds and for
1:00:20derivations from other grammatical
1:00:22categories it just so happens that an
1:00:24English the major growth areas for verbs
1:00:27in the next couple of Millennia were
1:00:30loans from French because of 1066 and
1:00:33all that when William the Conqueror
1:00:35invaded and forced the French language
1:00:37on the British Isles for the next couple
1:00:39of centuries and Latin because of the
1:00:42influence of the church and later
1:00:44Renaissance Scholars I've estimated that
1:00:46about 60% of English verbs came into the
1:00:49language from French and Latin and also
1:00:52since we like to verb things uh we find
1:00:55that another 20% of our verbs were uh
1:00:58originally nouns or adjectives note that
1:01:01both kinds of verbs had to be regular
1:01:04according to the logic of grammar that
1:01:06I've been sketching out throughout this
1:01:07talk so it's not that a majority of
1:01:11verbs are regular and that pounds the
1:01:13regular pattern into people's heads
1:01:15leading them to use the regular as the
1:01:18default as the uh pattern associator
1:01:21Theory would have it but rather the
1:01:23history tells us that that's exactly
1:01:25backwards that instead speakers have
1:01:28been using the regular as the default
1:01:30literally for thousands of years and
1:01:32that's why the majority of verbs uh in
1:01:35English are regular it's because the
1:01:37regular rule got first dibs on almost
1:01:39all of the new arrivals of the last two
1:01:42Millennia so to sum up despite the
1:01:45identical function of regular and
1:01:46irregular forms just two ways of talking
1:01:49about something that happened in the
1:01:51past or more than one of something we
1:01:53find that people avoid Irregulars and
1:01:56apply the regular suffix in a variety of
1:01:58circumstances that have nothing in
1:02:00common except a failure of access to
1:02:02information in memory these include rare
1:02:06words like abrogated and chided
1:02:09difficult to analogize words like
1:02:11plumped and friled words with
1:02:13inaccessible Roots like low lifes and
1:02:15flyed out words poorly recalled by
1:02:17children as in break and holded and
1:02:19words poorly recalled by patients with
1:02:21disorders of word retrieval such as a IA
1:02:24and Alzheimer's
1:02:26disease it suggests that regular
1:02:28inflection is computed by a mental
1:02:30operation that does not need access to
1:02:33the contents of memory namely a symbol
1:02:35combination rule which can step in as
1:02:37the default rules therefore free us from
1:02:40the constraints of
1:02:42memory therefore I'd like to suggest
1:02:45that we've identified distinct mental
1:02:47mechanisms that implement the two
1:02:49principles of language responsible for
1:02:51its vast expressive power memory for the
1:02:54arbitrary sign underlying the word and
1:02:57symbol combination for the infinite use
1:02:59of finite media underlying grammatical
1:03:02rules thank you very
1:03:15much and are there any questions I think
1:03:18a microphone has been set up and I've
1:03:20been asked to ask you to uh step up to
1:03:23the microphone so so that everyone else
1:03:25can hear your
1:03:33question has anyone studied the
1:03:35different effects of the injury to two
1:03:38parts of the brain on people who speak
1:03:40German people who speak German
1:03:46um let's see my our collaborators in
1:03:49Germany Harold Clauson and Richard Visa
1:03:52are in the process of doing that um
1:03:54but I they have not published their
1:03:56results and I can't um I can't tell you
1:03:59what they found although it's an
1:04:01excellent question and that one would
1:04:03expect similar patterns that you find in
1:04:06English again despite the vast
1:04:08differences in the statistics so that's
1:04:10exactly the right kind of question to
1:04:12further test the theory and I only wish
1:04:14I could give you the
1:04:15answer yes thank you so with my limited
1:04:18thought process I'm trying to frame an
1:04:20intelligent question here um I was
1:04:23thinking oh it's a two-part question
1:04:25it's probably not even connected but
1:04:29number one is there any sense that you
1:04:30can extrapolate to all human speech and
1:04:34all the processes and all the languages
1:04:35because I was thinking for example about
1:04:38languages with inflections like Chinese
1:04:41but more to the point I was thinking of
1:04:44a
1:04:45language that maybe
1:04:47Works in a much simpler and more logical
1:04:50way Hebrew where n over 90% of the words
1:04:53are formed from trilet Roots where
1:04:55there's a limited number of conjugations
1:04:57where there are no tricks of spelling
1:04:59where you can conjugate out uh just
1:05:02using uh it's almost mathematical as you
1:05:05know yeah so I wonder if you'd comment
1:05:07on that yes and in fact I I recently
1:05:09published a paper with uh the um
1:05:12psychologist Iris barent applying these
1:05:14ideas to Hebrew and it works uh in a
1:05:17number of ways it works quite well um in
1:05:19particular uh while uh the English the
1:05:23Hebrew plural system again has some
1:05:25rules such as masculine uh nouns take IM
1:05:29uh feminine nouns take but there are a
1:05:32handful of exceptions and the exceptions
1:05:36therefore the Irregulars which have to
1:05:37be memorized behave very much like the
1:05:40English ones so for example uh if I can
1:05:42remember one of our um instances uh kir
1:05:46means wall and ordinarily uh you'd
1:05:49expect the um plural to be uh Kim but
1:05:54it's an irregular form in the plural is
1:05:56kirot however if I if I have the example
1:05:59right if there are any Hebrew speakers
1:06:00you can in the audience you can correct
1:06:01me but if you change it to say Mr and
1:06:05Mrs Kier you create turn the noun into a
1:06:08name now you refer to them as the Kim
1:06:12not the kirot uh likewise if you have
1:06:15the word now as a lone word from another
1:06:18language and you say you know the French
1:06:19they put creme Cassis into champagne and
1:06:22they call it a kir would you like a
1:06:24couple of them then Hebrew speakers
1:06:25again say a Kim exactly like the flly
1:06:28out in the low lives uh in English so
1:06:31despite the fact as you point out that
1:06:33there are that Hebrew has no well
1:06:36documented historical connection to
1:06:38English it's not a Germanic or even an
1:06:40Indo-European language despite the fact
1:06:43that you assemble words in Hebrew not by
1:06:45joining things end to end by but by
1:06:48sticking vowels inside of slots inside
1:06:51of uh Roots nonetheless the logic of
1:06:54regularity and irregularity seems to be
1:06:56quite comparable and the even nicer
1:06:58punchline is that um in Hebrew the facts
1:07:02that I mentioned about um how German has
1:07:05a very different statistical pattern to
1:07:07English but nonetheless obeys the same
1:07:09logic that is true to an even greater
1:07:12extent in Hebrew where a lot of the
1:07:14sound patterning that we see in English
1:07:16like ring rang Sing sang drink drank
1:07:19doesn't apply to Hebrew but nonetheless
1:07:22you get the same interaction between the
1:07:24rule governed in the memorized
1:07:27forms has anything that uh in in your
1:07:30studies caused you to uh form a view
1:07:33about the way we teach language to
1:07:35children and whether it's being taught
1:07:37properly um certainly nothing that I've
1:07:39talked about today would be directly
1:07:41applicable you you would certainly
1:07:43expect that um since there is some
1:07:46evidence that the ability to acquire
1:07:48language instinctively um declines with
1:07:50age particularly um in uh late childhood
1:07:54and early adulthood uh after puberty it
1:07:56becomes increasingly difficult to
1:07:58acquire the syntax and sound pattern of
1:08:01a foreign language uh effortlessly that
1:08:04a lot of what ordinary kids um have to
1:08:07do with irregular forms adults might
1:08:10even have to do with regular forms that
1:08:12is there's a lot more memorization of
1:08:15cases that are rule governed but when
1:08:17you're a child it's easier to discern
1:08:19the rule when you're an adult you might
1:08:22have to do a lot more uh memorizing and
1:08:25and um adults often will Mis memorize
1:08:28forms I have a colleague a distinguished
1:08:31professor at MIT born in Italy who uh
1:08:34consistently says uh he's choosing it uh
1:08:36for example an irregular form but not
1:08:38the right
1:08:42one I was wondering if you would um
1:08:45comment on also that some were many many
1:08:48sounds become multiple meanings and duck
1:08:50is a pretty good example that you were
1:08:51using at the beginning and how this
1:08:54comes about and then your mind has to
1:08:56separate out the appropriate one yeah
1:08:58this it it's true that the um most
1:09:01common words in a language tend to have
1:09:04multiple meanings if you look up a word
1:09:06like set in the dictionary there can be
1:09:09up to a 100 different meanings listed
1:09:11for it whereas a low lower Frequency
1:09:13Word like Matrix or triangulate might
1:09:16have only one or two meanings this is a
1:09:19one of the reasons that computer
1:09:21language understanding systems are still
1:09:22so primitive um that we apply our common
1:09:27sense to winnow out all of the
1:09:30inappropriate meanings whereas computers
1:09:32which are bereft of Common Sense uh will
1:09:36app literally not know which of the many
1:09:38meanings to uh to use fortunately um as
1:09:42I mentioned in this particular
1:09:44circumscribed domain meaning has a uh
1:09:47very narrow role in general when you
1:09:50only change the meaning of a word
1:09:53without changing ing its grammatical
1:09:54structure all of the Irregulars come
1:09:57along for the ride so of
1:10:00the 20 or however many uh definitions
1:10:04there are of set in the dictionary all
1:10:07120 of them are irregular today I set
1:10:10yesterday I set um the only case in
1:10:13which a meaning change interacts with
1:10:14anything that I've said is when the
1:10:16meaning change also changes the um
1:10:19grammatical structure such as a name to
1:10:21a noun a noun to a name a verb to a noun
1:10:24and so
1:10:25on yes uh with words that rhyme and have
1:10:31similar meanings like Dash crash
1:10:35smash whether it's a violent um is there
1:10:39anything you can say about those or
1:10:40there any rules or just whatever you
1:10:43could say yeah uh forms like automat PF
1:10:47like Dash crash smash uh and so on or
1:10:50phonetic symbolism sound symbolism cases
1:10:53where you're not actually talking about
1:10:55a noise but nonetheless there's
1:10:57something about the word that feels
1:10:59right G given what it means like flutter
1:11:03where flutter you can't hear something
1:11:06fluttering but nonetheless the motion of
1:11:09your tongue calls to mind the actual
1:11:11meaning or
1:11:12cantankerous um sneer snob snoody uh and
1:11:17so on uh listless lazy lethargic they're
1:11:22often these families of words
1:11:24first of all they are these families are
1:11:27counter examples to s's principle of the
1:11:30arbitrary sign because there's at least
1:11:32a little bit of the sound that isn't
1:11:35completely arbitrarily paired with the
1:11:37meaning um also these families tend not
1:11:40to be governed by a rule there isn't a
1:11:43rule in English that says if a word
1:11:44begins with SN then it refers to
1:11:48snobbishness or if it begins with uh L
1:11:51it refers to um
1:11:54low energy um it's more statistical and
1:11:57the generalizations are made in a uh
1:12:01probabilistic manner much like um spling
1:12:05Spang dizzy Deans blind blound and so on
1:12:08and so I suspect that the generalization
1:12:10of a sound to a word with a similar
1:12:13meaning uh might be due to the same
1:12:17property in memory that allows us to
1:12:19generalize from uh drink drank to spling
1:12:22Spang namely the fact that similar
1:12:24sounds literally overlap in the brain
1:12:27and that which is associated with One
1:12:28automatically transfers to the other the
1:12:30pattern associating property of
1:12:34memory do German children discern the
1:12:37rule from the speech they hear around
1:12:38them or are they taught it explicitly
1:12:40and if the first how do they discern it
1:12:42if there's not a statistical yeah basis
1:12:44for it um they must discern it obviously
1:12:46not explicitly by being taught um
1:12:50likewise as in in English children of
1:12:52that age aren't given given uh grammar
1:12:54lessons um we don't know for sure how
1:12:58children do it it can't really be
1:13:00statistically in the sense of sheer
1:13:02number of words that exemplify a pattern
1:13:04because in German the words the nouns
1:13:07with s are a minority nonetheless it's
1:13:10one of the two patterns that the German
1:13:12children are willing to generalize um we
1:13:16suspect that um it may be that children
1:13:18use the kinds of information that are a
1:13:20little bit similar to the ones that a
1:13:21linguist would use in deciding whether
1:13:24something is truly regular namely if
1:13:27you're a German speaking child trying to
1:13:29decide whether say s or e r is the uh
1:13:33regular a couple of the tip offs are the
1:13:36fact that s applies to weird sounding
1:13:38words like Hobby and outto um and
1:13:41therefore are unlikely to be stored
1:13:45together with the ending As Natural
1:13:47words in the language it applies to
1:13:49aatopia and if you recognize the sound
1:13:53in a word then it's got some extra stuff
1:13:55stock onto the end you can infer that
1:13:58that stuff was added to the word rather
1:14:00than being a part of the word um it
1:14:02could be due to words that are um uh
1:14:07heterogeneous in sound that all take the
1:14:10same ending and so uh that is evidence
1:14:14that that ending applies regardless of
1:14:16sound and our uh we uh Gary Marcus and I
1:14:20have looked at at least in the case of
1:14:21English what kind of words do children
1:14:24of that age have in their vocabulary do
1:14:27they have automa PE words do they have
1:14:30words with unusual sounds with po
1:14:32multiple syllables and so on and the
1:14:34answer is that they do so we don't know
1:14:37that children use the structure of words
1:14:39as evidence for an ending being applied
1:14:43by rule but if they're if they were
1:14:45equipped to do so the information would
1:14:48be there they wouldn't need Clairvoyance
1:14:50or ESP they could figure it out
1:14:56are some of the rules that you mentioned
1:14:58uh like the tendency to use regular
1:15:00endings for loaned words common to all
1:15:03languages or virtually all if so why are
1:15:06they for example in our genome these
1:15:09rules yeah um I I would say that that's
1:15:13kind of more or less true I wouldn't say
1:15:15it's exactly true but in General Lone
1:15:18word in in many languages in fact in all
1:15:21the languages that I've looked at and
1:15:23I've tried to sample ones that were from
1:15:25very historically and geographically
1:15:27unrelated language families a lone word
1:15:31from another language will take the
1:15:32regular ending a uh word converted from
1:15:36another category very often will take
1:15:38the regular ending U and and so on so
1:15:41the pattern that is the fact that it's
1:15:44true in eight languages doesn't mean it
1:15:46applies to all 6,000 languages but it's
1:15:49interesting that in languages as far
1:15:51flung as Hebrew Chinese um ninian
1:15:55languages this pattern more or less
1:15:57seems to hold and it wouldn't be
1:16:00directly coded in the genome in the
1:16:01sense that there'd be a a gene for the
1:16:03regular past tense but it could be that
1:16:06the way that the genome organizes the
1:16:10language system in the brain as the
1:16:11brain develops um gives rise to a
1:16:14certain kind of rooting diagram sort of
1:16:16the how the boxes and arrows are set up
1:16:19might be coded in the genome and it's a
1:16:22consequence of what information is fed
1:16:25into which process that gives rise to
1:16:28these quirks and phenomena for example
1:16:31the fact that the um Regular seems to be
1:16:35added kind of as the last step in the
1:16:38assembly line of building a word so that
1:16:40you build a word you add stuff you
1:16:42fiddle with it and when you're just done
1:16:45and ready to stick it into a sentence
1:16:46that's when the Ed or the S gets added
1:16:48on that kind of information processing
1:16:52architecture
1:16:53is what we're guessing is coded in the
1:16:55in the genome certainly nothing as
1:16:57specific as a particular suffix a
1:16:59particular rule a particular process in
1:17:01in the
1:17:06language there are no verb conjugations
1:17:09or plurals of verbs in Vietnamese and
1:17:11I'm wondering if that's highly unusual
1:17:13in the
1:17:14language um not necessarily um that in
1:17:18Vietnamese and Chinese as well there is
1:17:20no inflection and um the
1:17:25um and languages have a mixture of um
1:17:31Reliance on syntax as you have in
1:17:33English and Chinese where most of the
1:17:35action consists of stringing words
1:17:37together in phrases and Reliance on
1:17:40morphology or building words where most
1:17:42of the action consists of stringing
1:17:44together lots of prefixes and suffixes
1:17:46in a word English is mixed we have some
1:17:49inflection adding Ed adding s iing but
1:17:52we rely more on syntax the claim um of
1:17:56the the theory that I'm making today is
1:17:58that the what all languages do have is a
1:18:02storage mechanism and a combinatorial
1:18:05mechanism and that there's a close
1:18:07relationship between the combinatorial
1:18:09mechanism for words in phrases which is
1:18:12what Chinese goes crazy with and the
1:18:14mechanism for assembling prefixes and
1:18:17suffixes inside a word which is what a
1:18:19language like Latin or many African
1:18:22languages go crazy with uh but that
1:18:24these at the level of psychology and
1:18:26perhaps at the level of Neurology the
1:18:28main distinction is combinatorics versus
1:18:31storage and and uh retrieval is it odd
1:18:34though Vietnamese is written with
1:18:36letters that are similar to the English
1:18:38language and yet pronounced more like
1:18:40Chinese so there's sort of a contrast
1:18:42between which way the grammar develops
1:18:45in general it's an interesting point
1:18:47because it shows that the written form
1:18:48of the language is really more a matter
1:18:51of historical ACC accident than of the
1:18:54inherent logic of the language and
1:18:57writing systems are rare they're recent
1:19:01in history they tend to be um evolve a
1:19:05small number of times from very complex
1:19:07civilizations and then tinkered with and
1:19:09improved over hundreds of years um and
1:19:13they're in a sense psychologically very
1:19:14unnatural they're almost like the
1:19:16internal combustion engine or the you
1:19:19know the cotton gin their inventions and
1:19:21so it just so happens that Vietnamese
1:19:24didn't have an indigenous writing system
1:19:27very late in its history instead of
1:19:29going to the trouble of involving one
1:19:30from scratch it just applied readymade
1:19:33the Roman alphabet but it could have
1:19:35been cilic it could have been the
1:19:38Chinese characters which is what the
1:19:39Japanese uh began with it's kind of
1:19:42whatever was handy depending on which
1:19:45missionaries were around which
1:19:46conquerors which neighbors uh and so on
1:19:49but it there is doesn't necessarily have
1:19:51to be a match between the way the
1:19:53language works and whatever writing
1:19:55system you apply to it thank
1:19:58you you've mentioned the distinct
1:20:02mechanisms for the two different
1:20:05principles of language you've identified
1:20:07does that require still another
1:20:09mechanism to determine which of these
1:20:12two mechanisms to invoke yeah I very um
1:20:16indeed if you have two mechanisms you
1:20:17need something that coordinates them
1:20:19chooses selects or chooses which two
1:20:22utilize well basically um you don't need
1:20:26anything very complicated to choose
1:20:28between them as far as as we can see the
1:20:30idea which I mentioned very briefly and
1:20:32and didn't return to is that in general
1:20:35when you have something in memory it
1:20:37turns off the rule that would generate a
1:20:39corresponding form so if you come up
1:20:42with brought as the past tense of bring
1:20:44that turns off the aded process and so
1:20:48to coordinate these two basically you
1:20:51just need kind of a oneway inhibitory
1:20:54linkage as the memory system says I
1:20:56found something that sends a signal uh
1:20:59turning off the rule box to learn it and
1:21:03to decide where to store it and what
1:21:05mechanism to use is there still
1:21:09another mechanism behind that that is
1:21:12required um no I don't learning process
1:21:14itself um you decide what to invoke
1:21:17whether it will be a rule or before it
1:21:21will be either
1:21:24before it will be used either stored
1:21:28either at according to the rule or the
1:21:30word mechanism how would the brain
1:21:34decide which of these abilities to make
1:21:37use of in the learning process yeah no
1:21:40it's a good question and um as far as we
1:21:42can tell to begin with a child really
1:21:46has no choice but to memorize everything
1:21:47as if it's a word because the rules
1:21:50couldn't possibly be coded in the genome
1:21:52they they differ depending on what
1:21:54language you speak added to form the
1:21:57past tense is not something you could be
1:21:59born with so to begin with the idea is
1:22:02that kids memorize walked the same way
1:22:04they memorize uh uh BR or rang um and in
1:22:09fact we have evidence that that's the
1:22:11case that the that early on at the point
1:22:14I showed you in the graph before the
1:22:15child is generalizing the rule there
1:22:17seems to be very infrequent and sporadic
1:22:20use of regular forms like like played
1:22:23and walked
1:22:25and the uh the hunch is that the child
1:22:29juxtaposes individually memorized pairs
1:22:32like walk walked play played uh talk
1:22:35talked explain explained and so on
1:22:38notices that words come in two ever so
1:22:41slightly different forms is equipped
1:22:44with some kind of algorithm that doesn't
1:22:46just write it off as variety for the
1:22:49spice of life or you know oh Mom she
1:22:52sometimes sticks s's at the end for no
1:22:54possible reason um but rather looks for
1:22:57a rationale for systematic variation why
1:23:00is Mom sometimes adding s let me see
1:23:03what the difference is between when she
1:23:05says dog and when she says dogs aha
1:23:08there are a bunch of them now and the S
1:23:10is there therefore the S is used to
1:23:12indicate a bunch of that gets enshrined
1:23:16as a rule peeled off the individual
1:23:18words then it can be then the child can
1:23:21run a muck with it and apply it every
1:23:23which way that's the general
1:23:28idea can the process of applying the
1:23:31regular
1:23:32form to words that have not or forms
1:23:38rather that have not been memorized or
1:23:40that or where memory fails can can that
1:23:44be applied to the spelling of
1:23:47dyslexics um where I've seen for example
1:23:52went spelled very very often wh NT is
1:23:58that is that an application of the
1:24:01regular pattern of all of the question
1:24:03words that are learned early on that
1:24:05begin with wh and because the sounds is
1:24:08what is that why um and there are many
1:24:11many spelling
1:24:13failures um that seem to follow regular
1:24:16patterns and if that is so um is is
1:24:20there um a a a training technique that
1:24:25you could suggest that could um reduce
1:24:29that yeah um in fact there are many
1:24:32parallels between spelling and um and
1:24:34grammar there are um something like 87%
1:24:38of English words are spelled according
1:24:40to regular rules but then there's this
1:24:42133% of exceptions like o f for of and
1:24:46ey for I and so on the Irregulars tend
1:24:49to be highest in frequency because they
1:24:52have to to be filtered through the
1:24:53memorization of subsequent gener
1:24:55Generations when you have damage to
1:24:57memory you tend to get overgeneral
1:24:59spellings like
1:25:02um um like the one one that you
1:25:04mentioned or say um i e for instead of
1:25:08aisle um and uh conversely when there's
1:25:11a problem in grammatical combination you
1:25:15often have an inability to spell new
1:25:17words like to plam or to Ginge um so
1:25:21there are a number of parallels um and
1:25:25and there must be some psychological
1:25:28commonality um I wouldn't be prepared to
1:25:31say simply because of my own ignorance
1:25:33how you would translate that into a
1:25:35curriculum other than being aware of the
1:25:37fact that a student could spell a word
1:25:39correctly for two very different reasons
1:25:42either by applying a rule or by
1:25:44memorizing it and that you have to sort
1:25:45of diagnose which of the two a student
1:25:48is using moreover that um the the these
1:25:52can be that that uh in cases where
1:25:54there's a misspelling it may not just be
1:25:56random garbage or noise but might be a
1:25:59case of a child over applying a rule to
1:26:02an irregular case so it may not be
1:26:05stupidity laziness sloppiness but one
1:26:08system applying where the other one
1:26:09ordinarily
1:26:11does thank you
1:26:13thank uh this uh question is probably a
1:26:16little bit less technical um I'm raising
1:26:19um um a 2-year-old daughter English and
1:26:21German bilingually and I'm you know
1:26:23hearing your talk I I was like
1:26:25struggling to find some kind of a
1:26:27practical implication on how to best
1:26:30Riser bilingually English and German for
1:26:33example I've noticed it's I'm witnessing
1:26:35firsthand how uh how much easier it is
1:26:38for her to to learn English words for
1:26:40example like one of her first words was
1:26:42bubble and the German word is cphon
1:26:44blaza it's so much more cumbersome is
1:26:47and there are many examples like that
1:26:49and her German really is is totally
1:26:51lagging behind so I was I was struggling
1:26:53to find some kind of a practical
1:26:55implication since you mentioned you know
1:26:57English and German in your talk how to
1:26:59best raise a daughter in you know
1:27:01bilingual in English and German yeah
1:27:04well despite Twain's uh Mark Twain's
1:27:07complaint about the awful German
1:27:08language it's really awful only to
1:27:11adults kids obviously must be doing fine
1:27:14with German otherwise it wouldn't have
1:27:15survived to the present day with all of
1:27:17these multiple plural forms and so on um
1:27:20you know Twain also said i' I'd rather
1:27:22um I'd rather decline two drinks than
1:27:25one adjective in German
1:27:28and but this is two more question two
1:27:31more questions um but the thing is that
1:27:33if kids continue to hear it continue to
1:27:35use it uh then then they they will
1:27:37Master it as they have for you know
1:27:40Generations the trick is uh it's
1:27:43generally kids are a bit faster in
1:27:44coping with two languages when they come
1:27:46from different speakers when Mom speaks
1:27:48one language and dad speaks another that
1:27:50gives them a a little bit of help and
1:27:53they'll generally retain it if they have
1:27:55to use the other language with their
1:27:57peers that if it's just Mom and Dad
1:28:01especially if Mom and Dad are bilingual
1:28:03the child will force the parents to use
1:28:06the dominant language say English
1:28:08whereas if the child goes to summer camp
1:28:11plays with cousins every summer uh has
1:28:13reason to use it with other kids that's
1:28:16a better uh portent of of keeping both
1:28:19languages makes sense thank you okay and
1:28:22and just these these last two and I've
1:28:23been uh asked to bring the proceedings
1:28:26to a close after these two okay I have
1:28:28three unrelated questions they're all
1:28:33brief the first is do you have any idea
1:28:36of what prompted the big vowel reduction
1:28:39in the 15th century and the second is um
1:28:44how do you feel about uh the cobbling
1:28:48together of prefixes suffixes and words
1:28:51the you called it morphol morphology
1:28:54morphology it seems as if it's happening
1:28:57more and more words like dequ or
1:29:03self-actualize um and the third is uh
1:29:08Japanese
1:29:09sounds um is spelled exactly the way
1:29:13it's pronounced in English is that
1:29:15because the language was isolated for so
1:29:19long and when the English decided
1:29:22to translate it it it happened very
1:29:25systematically or you're talking about
1:29:27the the transliteration of Japanese into
1:29:29Roman characters yes oh yeah okay first
1:29:33no one really knows what caused the
1:29:35Great Vowel Shift the two main theories
1:29:37are that it was a kind of uh uh sequence
1:29:41of enhancements that as
1:29:44um because around that time people were
1:29:47swallowing the second syllables of words
1:29:51um and were um the reason that we have
1:29:54all of these crazy silent vowels at the
1:29:57end of words like make M A ke is that it
1:30:00used to be Maka and then the e at the
1:30:03end eventually got dropped alt together
1:30:05because words were becoming more and
1:30:07more similar because of this uh slurring
1:30:09of the second syllable people started to
1:30:11exaggerate differences among vowels to
1:30:14compensate and they started off say with
1:30:16one vowel contrast they exaggerated the
1:30:19difference between them but then the
1:30:21vowel that they exaggerated started to
1:30:22bump up against some second vowel so
1:30:24they had to exaggerate that in a
1:30:26different direction to differentiate it
1:30:28and you got kind of a billiard game
1:30:30where the different vowels got knocked
1:30:31around in space and when all was said
1:30:34and done no one was any better off than
1:30:37when they began but it was all done one
1:30:38vowel at a time the other is that it was
1:30:41kind of social differentiation that as
1:30:44French had vanished as a means of
1:30:46distinguishing the um the beautiful
1:30:49people from the hooy uh you needed some
1:30:52fussy way of speaking in order to show
1:30:54that you were an aristocrat and so
1:30:56people started to uh exaggerate their
1:30:59speech in fancy ways and it could be
1:31:01that both of those theories are right in
1:31:03terms of um Mor uh morphology is there
1:31:07more and more of it um it's something
1:31:09that English has been doing for an
1:31:10awfully long time there are cases in
1:31:12which you know things like you know to
1:31:15prioritize and to you know accessorize
1:31:17and so on can kind of grade on the ears
1:31:20because it's something that Bureau RS
1:31:22and um uh sort of the Dilbert business
1:31:25crowd uh tends to do to impress people
1:31:28with how scientific their approaches uh
1:31:31often though it does add to the
1:31:34expressive power of a language Strunk
1:31:36and white for example were upset about
1:31:38the word to moisturize saying well why
1:31:40can't you just say to moisten but of
1:31:42course anyone who uses a moisturizer
1:31:44knows that moisturizing and moistening
1:31:46aren't the same thing I mean if you
1:31:47moisten your skin you can just use a
1:31:49water pistol uh but moisturiz in means
1:31:52just it's a a completely different
1:31:54concept and so even though some of these
1:31:56forms have lots of syllables they're
1:31:58kind of ugly they do allow us to convey
1:32:01ideas and that's I think one of the
1:32:02reasons they're indispensable and in
1:32:05terms of Japanese the transliter
1:32:07transliterations where you apply one
1:32:10writing system to another language for
1:32:12the benefit of the speakers of the first
1:32:16language they're generally regular
1:32:18simply because that's the way you
1:32:20concoct that's the way you apply the
1:32:21system uh to begin with the irregularity
1:32:24tends to accumulate over centuries as a
1:32:27system is entrenched and doesn't track
1:32:30the changes in pronunciation the
1:32:32language is kind of a the speech is a
1:32:34moving Target but the spelling stays put
1:32:37and so the spelling over centuries
1:32:40becomes more and more irregular compared
1:32:42to the sounds and so that doesn't happen
1:32:44when you artificially extend uh a one
1:32:47writing system to another
1:32:49language and what last question
1:32:52uh the questions related to the great V
1:32:54shift um what do you think are the most
1:32:56powerful forces that change
1:32:59language um and the great Val shift that
1:33:03obviously has the name the great Val
1:33:05shift can you uh give us uh any current
1:33:08examples of something that's significant
1:33:10or any changes in language um from
1:33:13recent times that you can equate to that
1:33:16sure dude um the uh the kind of Valley
1:33:20Valley Girl speak for example uh where
1:33:23we recognize fairly recently a shift in
1:33:27pronunciation or um if you've ever
1:33:30listened to a a person in their 20s uh
1:33:32whenever they use a sentence they
1:33:35pronounce it as if it's a question uh
1:33:37and this is kind of recent uh that would
1:33:41be another example and for all we know
1:33:42that could be put a permanent stamp on
1:33:45the English language as people like me
1:33:47uh die and people who
1:33:50are uh and people who are now in their
1:33:5220s uh out last I understand what you're
1:33:58saying thank you very much thanks to all
1:34:01of
1:34:02[Music]
1:34:05you thanks for listening for more
1:34:07information on the 92nd Street y New
1:34:09York and all of our programs please
1:34:11visit us at
1:34:1392.org

About this transcript

This page contains the full transcript of Words and Rules: The Ingredients of Language with Dr. Steven Pinker (1999) by The 92nd Street Y, New York, generated from the public captions YouTube serves with the video. The transcript has 15,215 words across 2,186 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.