Post on 23-Jun-2020
Language Development
Specific goal 1: What is human language?
A complex system made up multiple subsystems
Specific goal 2: Why asking “How do children learnlanguage?” is an ill-formed question.
Given the answer to 1, it is only useful to ask how childrenlearn all the subsystems and how they manage tocoordinate them into a single whole integrated system.
Language Development
Today: Specific goal 3 - What perspectives
have researchers taken on this question?
How nature/nurture/interactionist options
play out in (psycho)linguistics.
Focus on the learning of morphology (word
patterns), syntax (patterns of word
combinations), understanding the
intentions of others.
The Central Mystery
How do children acquire the discretely organized
subsystems that make up Language?
How do children acquire the sound system, word
shape system, word combination system,
word meaning (and sentence meaning)
system, use system -- and come to coordinate
all of these systems together?
Necessary Questions
What is Language? In what ways are all natural
languages alike? What ways are they different?
What distinguishes natural languages from animal
communication systems and artificial languages
and even programming languages? How can we
characterize the adult’s knowledge of his/her
native language(s)?
This is the domain of LINGUISTS.
More Necessary Questions
What is learning? How do children develop
mastery in non-linguistic domains such as facial
recognition or object recognition or concept
formation? What is the time course of learning
and are there correlations between learning in
different domains? What are the mechanisms or
processes that facilitate or impede learning?
This is the domain of DEVELOPMENTAL or
COGNITIVE PSYCHOLOGISTS.
Resulting Interdisciplinary
QuestionsWhat is language learning? How do children develop
mastery of their native language(s)? Do they rely on thesame operations as in non-linguistic skills? What are thebiological bases and the actual learning patterns of thelanguage development process? How does learning innormal and special populations differ and how is it similarto language learning and learning in other cognitive and
social domains? What is the relation between the adult’sknowledge and the child’s knowledge, I.e., what is therelation between the infant startstate and the adultendstate?
This is the domain of DevelopmentalPSYCHOLINGUISTS.
Alternatively put…
“The most fundamental question in the study of thehuman language faculty is its place in the naturalworld: what kind of biological system it is, andhow it relates to other systems in our species andothers.
A second question is what parts of a person’slanguage ability (learned or built-in) are specific tolanguage and what parts belong to more generalabilities.”
- Jackendoff & Pinker 2005
Alternatively put…
The third question is which aspects of the language capacityare uniquely human and which are shared with othergroups of animals, either homologously, by inheritancefrom a common ancestor, or analogously, by adaptation toa common function…As with the first two questions,answers will seldom be dichotomous. They will oftenspecify mixtures of shared and unique attributes, reflectingthe evolutionary process in which an ancestral primatedesign was retained, modified, augmented or lost in thehuman lineage.” 2005:3
The “answer”
Nobody knows...but we have gotten increasingly
interested in:
• The types of methods used to explore key areas
of language,
• The types of models proposed by researchers,
• The types of questions that need to be asked to
ultimately arrive at satisfying and compelling
“answers”.
Some Consequences of
AnswersHelp us to understand the nature of the human
mind.
Help us to understand the relation between humanbehavior in relation to the behaviors of non-human primates and other creatures: What’s ourplace in the biological world?
Help to understand the role of language in humanculture
Help to address issues in special populations, I.e.,deaf, neurogenetic disorders, etc.
What we want in our
theories (Pinker 1984)
• Learnability: the theory must assume mechanisms which areadequate to acquire a natural language and which, thereby, facilitatethe acquisition of language in the first place.
• Equipotentiality: the theory should not postulate mechanismswhich solely facilitate the acquisition of a favorite grammar, but thetheory should account for the acquisition of all natural languages.
• Time: the theory should account for acquisition in real time.
• Input: the mechanisms invoked by the theory should operate ondemonstrably plausible data.
• Development: the theory should predict stages of attesteddevelopment.
• Cognitive: the mechanisms appealed to for the acquisition ofLanguage (or language) should not be incompatible with the generalmechanisms of cognition.
How is language learned without explicit instruction?
Some insights
“In the process of naturally mastering one’s mother-tongue no rule, as such, isgiven, but only a number of examples. We hear gradually a number ofsentences which are connected together in the same way, and which henceassociate themselves into one group. The recollection of the special contentsof the single sentences may grow less and less distinct in the process; thecommon element is always strengthened anew by repetition, and thus it comesabout the rule is abstracted is unconsciously abstracted from the examples. Itis precisely because no abstract rule is laid down that no single examplesuffices, but only a group of examples whose special contents appear a matterof indifference.” Paul 1970:98. (originally written over 150 years ago)
“We would like to reconstruct this ability [the remarkable ability of any speaker ofa language to produce utterances which are new both to him and to otherspeakers] within linguistic theory by developing a method that will enable usto abstract from a corpus of sentences a certain structural pattern, and toconstruct, from the old materials, new sentences conforming to this pattern,just as the speaker does.” Chomsky 1955:131
A basic question(Adapting Tomasello 2001:169)
All human languages consist of abstract patterns.
How do children acquire the syntactic structure andthe word structure patterns of their language whenthey hear those around them only utteringindividual sentences and words, not abstractsyntactic categories such as Noun and Verb,sentential schemas such as Subject Verb Object,word schemas such Noun+plural?
What we don’t know we know, but do.(following Pinker, Word and Rules 1998)
Consider when we use irregular, not regular forms:
Prefixing: Irregular Regular
overate *overeated
overshot *overshooted
preshrank *pre-shrinked.
Given a word bleet and its past tense blate, what is theform of the past for overbleet?
What we don’t know we know, but do,
somehow(following Pinker, Word and Rules 1998)
Consider when we use irregular, not regular forms:
Compounding : Irregular Regular
workmen *workmans
superwomen *superwomans
stepchildren *stepchilds
snowmen *snowmans
But, sometimes the expected irregular
form fails to appear
The Toronto Maple Leafs/*Leaves (a hockey team namedafter Canada’s national symbol, The Maple Leaf).
Renault Elfs/*Elves (cars).
Michael Keaton starred in both Batmans/*Batmen (movietitles).
We’re having Julia Child and her husband over for dinner.You know, the Childs/*Children are really great cooks.
More words that only display regular
plural marking -s or regular past tense
marking -edAll my daughter’s friends are low-lifes (*low-lives).
I’m sick of dealing with all the Mickey Mouses in thisadministration (*Mickey Mice).
Boggs has singled, tripled, and flied out (*flown out) in thegame so far.
How would you pluralize sisterwife?
What is the past tense of grandstand?
Are these random behaviors across speakers, or do nativespeakers generally agree, even if they have not previouslyheard the singular or the present tense?
What else don’t we know we
know? CausativesBlack Black-en
Red Redd-en
White Whit-en
Green ?
Dark ?
Light ?
Highlight ?
Grue ?
Drick ?
Quiet Quieten (Guardian Unlimited 9/26/05
A question
Do we simply remember and repeat previouslyencountered specific words and phrases or do weabstract patterns that complex words and phrasesparticipate in? What’s the evidence that we don’t?
If we don’t just repeat, how do we identify thecorrect patterns into which we assign words andsentences?
Wild (Feral) Children
Language is irrepressible in humans: it ariseswherever normal humans congregate.
Tall tales: Tarzan, Mowgli (The Jungle Book),Nell,,,,
True tales: The Wild Child - 18th century (FrancoisTruffaut), The Enigma of Kaspar Hauser - 19thcentury (Werner Herzog)
A Sad Case
Genie (1970s Los Angeles, Ca.; 13 yrs. old)
Isolated at home by an abusive father; likely braindamage.
Learned vocabulary, but limited (if any) syntax, I.e.,didn’t seem able to gain command over theprinciples for combining words with one another.
Language processed in the right, rather than the lefthemisphere.
Happy cases
The development of sign languages inNicaragua and the Negev desert.
The development of creole languages from thebits and pieces of pidgins langauges used forsimple communication.
Enculturated creatures: not
linguistically irrepressible, but very
clever…
Kanzi (pygmy/bonobo
chimp)
Animal Communication
Systems
A goal of some researchers:
Establish that there is less of acategorical/radical difference betweenhumans and other creatures.
A hypothesized distance is lessened bydemonstrating that some property, contraryto belief, is not unique to humans.
Lexigrams: “words”
Impressive acquisition of
lexigrams, but inability
to produce small set of
consistent patterns for
combining these
“words”.
A Bi-lingual Bonobo: English
Comprehension by Child vs. Chimp(Savage- Rumbaugh et. al. 1993)
Task: Compare language development in a normal child (Alia 2;0) andnormal bonobo (Kanzi 8;0), based on responses to 660 spokeninstructions.
Kanzi: Exposed to spoken English and lexigrams from 6;0 mo.; exhibitedspeech comprehension at 2;0 and spontaneous use of lexigrams at 2;5.
Alia: Exposed to spoken English from birth and lexigrams from 3 mo.:comprehension of 32 spoken words at 13 mo. and spontaneous use oflexigrams at 11 mo.
Some Results(Savage- Rumbaugh 1998:71)
“Rarely did either Kanzi or Alia make mistakes that indicateda lack of understanding of the basic grammatical structureof the sentences. Both of them readily differentiatedbetween requests to retrieve objects from locations (Go tolocation X and get object Y) and requests to take objects tolocations (Take object X to location Y). They also understoodthe difference between sentences that required them tomove through space in addition to acting on objects andsentences that required them to act on objects withoutmoving about.”
Interpretation(Savage- Rumbaugh et. al. 1993)
“… under relatively similar rearing conditions andvirtually identical testing conditions, they couldcomprehend both the semantics and the syntax ofquite unusual English sentences.”
So, bonobos appear to perform some extraordinary
“language” feats - seeming to be on par with ahuman infant, for a short time.
But, bonobos reach a barrier very early on, while thechild keeps developing.
A View From a Primate
Researcher(Sue Savage-Rumbaugh 1996 Discover Magazine)
“Now I understand in ways that I cannot fully describe that language isn’ta matter of learning little building blocks like words and stringing themtogether in some kind of hierarchical structure and then going out andkind of throwing these out to the rest of the world so that ideas jumpfrom my mind to yours. Language is a matter of me learning tocoordinate my behavior with all of the other individuals in the worldaround me and that much of this initial coordination is throughglances, through patternings of behavior together, through jointunderstandings of how the world works, and joint constructions of howwe’re going to operate in this world together.”
Language as essentially a communicative system, rather thana system by which communication is conveyed.
A View From a Developmental
Psychologist(Annette Karmiloff-Smith 1992:63)
Child: “What’s that?”
Mother: “A typewriter.”
Child: “No, you’re a typewriter, that’s a typewrite.” (Yara, 4.0)
“Thus, even if the chimpanzee were to have an innately specified linguisticbase, I speculate that it would still never go as far as the human child.It would never wonder why “typewriter” isn’t used to refer to people. Itwould simply repeat the linguistic labels that it was given. But childrendoi not simply reach efficient usage; they subsequently develop explicitrepresentations which allow them to reflect on the component parts ofwords to progressively build linguistic theories.”
“… a crucial difference shows up when we look at what happens beyondsuccessful mastery. Chimpanzees do not go beyond behavioral mastery.”Karmiloff-Smith 1995.
A basic question, again(Adapting Tomasello 2001:169)
All human languages consist of abstract patterns.
How do children acquire the syntactic structure andthe word structure patterns of their language whenthey hear those around them only utteringindividual sentences and words, not abstractsyntactic categories such as noun and verb orschemas such as Subject Verb Object orNoun+plural?
Two competing answers(Tomasello 2001:169)
H1: Representational Nativism (Chomsky)
“children do not have to learn or construct abstract representations at all,but rather they already possess them as part of their innate languagefaculty. This so-called continuity hypothesis justifies the use of adultlike formal grammars to describe children’s early language.
H2: Constructivist
“children’s early utterances are organized around concrete and particularwords and phrases, not around any systemwide syntactic categories orschemas. Abstract and adult-like syntactic categories and schemas areobserved to emerge only gradually and in piecemeal fashion during the
preshcool years. Discontinuity hypothesis in that the child’s earlygrammar and the adult’s eventual grammar are not necessarily madeof the same stuff.
(Dis)Continuous Knowledge(Aitchison1976:127)
Chomskyan: language
specific structures and
categories of adult
endstate are refined
versions what’s at the
infant startstate.
Constructivist: language
specific structures and
categories of adult
endstate are possibly
radically different from
those infant startstate.
What is there, when so little is evident?
Bever’s Syllogism:
Representational Nativisn
To be proven: Language is innate.
1. The essence of Language has property Pi
(Coordinate Structure Constraint).
2. Pi cannot be learned by any (known)(conceivable) theory of learning.
3. Therefore, Pi is innate.
4. Therefore, the essence of Language isinnate (and caused thereby).
Bever’s Syllogism:
Constructivism
To be proven: Language is learned.
1. The essence of Language has property Pi
(Coordinate Structure Constraint).
2. Pi cannot be transmitted by any (known)(conceivable) genetic mechanism.
3. Therefore, Pi is learned.
4. Therefore, the essence of Language islearned (and caused by how it is learned).
Noun Coordination and Questions(Doctor Doolittle’s Dilemma Stephen Anderson 2004:224)
3) Pat is majoring in [Linguistics and Philosophy].
3’) What is Pat majoring in [Ø and Philosophy]?
3”) What is Pat majoring in [ Ø ]
4) Pat is majoring in Linguistics [along with Philosophy].
4’) What is Pat majoring in Ø [along with Philosophy]?
Observation
You cannot form a question about only one conjunctin a coordinated phrase, but you can form aquestion from a single element when it’s not in acoordinate phrase.
A sentence with a similar meaning, but differentstructure permits you to question a single element,so the difference in behavior between thesequestions isn’t likely to be sensitive to meaning.
Instead, it’s likely that explanation for the differenceconcerns structure of the sentences.
Empirical research and an
interpretation“This kind of knowledge could not plausibly have been
acquired on the basis of experience. Therefore it seemslikely that these aspects of syntactic organization are asmuch a part of the biologically determined humanlanguage faculty as the structure of the vervet monkey callsis specific to animals of that species.” SA 2004:229
A (proposed) universal
property of grammar
Coordinate Structure Constraint
A single conjunct in a coordinate structure
cannot be questioned alone.
XP
XP1 and XP2
Poverty of the Stimulus:
Core properties of grammar“As children, we came to know the generalizations
and their exceptions, and we came to thisknowledge quickly and uniformly. Yet ourlinguistic experience was not rich enough todetermine the limits to the generalizations. Wecall the the problem of POVERTY OF THESTIMULUS. Anderson & Lightfoot 2002:20
But, is the stimulus really so impoverished?
Constructivism/
Emergentism (Piaget)
“Language structure is an “inevitable” emergent solution to a
series of interactions. Because that structure is inevitable, it
does not have to be innate. There is no reason for nature
to waste perfectly good genes on an outcome that is going
to happen anyway. Applied to language, this approach
suggests that the semantic and grammatical structures of
language are the inevitable set of solutions to the problem
of mapping certain non-linguistic, cognitive meanings and
social interactions onto the highly constrained linguistic
channel.” Bates and Snyder 1985
Developmental Hypothesis
“ …Human beings as well as other primates are atbirth relatively unspecialized. Infants possess atbirth very few ready-to-use intellectual or physicalabilities… for humans, evolutionary specializationmanifests itself by a relative lack of specialization atbirth. If this is the evolutionary solution for manyhuman physical and cognitive skills, in our view itis also the most likely solution for language.Karmiloff-Smith and Karmiloff 2001:224
A developmental hypothesis abouthuman language
Language is constructed by the child using inbornmental equipment provided to us as a species ratherthan to inchworms, and, while operating oninformation provided by the environment, much of itreflecting stored knowledge of grammar emerges asa product of this interaction.
Distributional/Correlational Learning:
German gender
Distributional/Correlational Learning
Distributional/Correlational Learning
Distributional/Correlational Learning: English
lexical categories
A Verb is an entity that
exhibits a particular
distribution and this
becomes clear with
experience.
Even if Verb was innate,
one would still have to
learn how it behaves in
a given language.
Interim summary
Individuals learn a lot of subtle stuff for which there is noobvious stimulus in the environment, until you carefullyexamine many potential source of information.
Importantly, if your basic assumption is that numerousaspects of language are acquired in the absence ofenvironmental cues and clues, then you foreclose lookingmore carefully into the environment and create a situationwhere you never critically examine your nativistassumptions.
(Recall the pecking “instinct” of domesticated chicks)
So, what does the
environment contain?The formulation of Plato’s problem presumes that the
environment cannot provide the infant with necessaryinformation to arrive at adult knowledge.
Reliance on transitional probabilities to learn what sorts ofsyllables cohere into “words” seems a surprising synthesisbetween what the world supplies and our (non-linguistic)talent to make sense of it. Likewise for we are apparentlyskilled at readings others’ intentions. We are likewisetalented in abstract pattern identification and patterns ofinflected words in Estonian.
Will all areas of language learning be similarly surprising withrespect to what the environment may supply and what ourtrue abilities may be?
A probable moral
There is both more information in the stimulus andmore ability in infants to use this information thanresearchers previously supposed.
This should make us cautious about assuming whatis in the input and what the infants is capable ofdoing with it, I.e., it should make us cautiousabout how we interpret Poverty of the Stimulusarguments.
In summary
“The most fundamental question in the study of thehuman language faculty is its place in the naturalworld: what kind of biological system it is, andhow it relates to other systems in our species andothers.
A second question is what parts of a person’slanguage ability (learned or built-in) are specific tolanguage and what parts belong to more generalabilities.”
- Jackendoff & Pinker 2005