Language Production

1
Language Production
ABRALIN 21-FEB-05

• Eva M. Fernández
• Queens College Graduate Center
• CUNY

2
Second Week, Proposed Schedule

• Segunda-feira PRODUCTION
• Terça-feira PERCEPTION
• Quarta-feira METHODS
• Quinta- e Sexta-feira BILINGUALS

3
General Domain of Psycholinguistics

• How does linguistic competence develop?
• How is linguistic competence used, in real time?
• Linguistic performance
• language acquisition
• language processing
• ? production
• ? perception

4
Language Is
COMPETENCE PERFORMANCE
SIGNAL
MEANING
5
Language Is

PERCEPTION
grammar lexicon
SIGNAL
MEANING
PRODUCTION
knowledge about the real world
logic, math
6
A Model of Speech Production
MEANING
MEANING lemma
SEMANTICENCODING
LEXICAL RETRIEVALSTRUCTURAL ASSIGNMENT
FORM lexeme
APPLICATION OF PHONOLOGICAL MORPHOLOGICAL RULES
PHONOLOGICAL REPRESENTATION
instructions to articulatorsproduction of
speech
SIGNAL
7
Evidence for this Model?

• The model seems to indicate theres a time-course
  of processes in sentence production
• some come earlier, others later
• Do people plan their speech before they speak?
  (despite the fact that they may not plan their
  ideas)
• Does this take long?
• Speech error corpora
• Inducing tip-of-the-tongue (TOT) states
• Eliciting production

8
Two Landmark Historical Pieces

• Al Kitaab (12th Century), Arab scholar
• Errors of the Populace
• analysis of production errors, to demonstrate
  phenomena in diachronic language change
• Sigmund Freud, Vienese psychoanalyst
• The Psychopathology of Everyday Life (1901)
• lapsus linguae (slips of the tongue), lapsus
  calami (slips of the pen), or in general,
  Fehleistung ( faulty action, parapraxis,
  Freudian slip)
• repressed emotions cause errors (id over ego)

9
Freudian Slips Examples

• President of the Austrian Parliament says I
  take notice that a full quorum of members is
  present and herewith declare the sitting closed!
• Hotel boy knocks at bishops door, bishop says
  Who is it? and boy nervously replies The
  Lord, my boy!
• Member of the House of Commons refers to another
  honorable member as Central Hell (rather than
  Hull)
• Professor says In the case of the female
  genital, in spite of the tempting I mean, the
  attempted
• Lady of well-determined character says My
  husband asked his doctor what sort of diet ought
  to be provided for him. But the doctor said he
  needed no special diet, he could eat and drink
  whatever I choose.

10
Malapropisms ? speech errors

• Mrs. Malaprop, a character in Sheridans The
  Rivals (1775)
• using words mal à propos ( out of place)
• Shes as headstrong as an allegory on the banks
  of the nile. (alligator)
• Comparisons are odorous. (odious)
• He is the very pineapple of politeness.
  (pinnacle)
• I want to remind you all that in order to fight
  and win the war, it requires an expenditure of
  money that is commiserate with keeping a promise
  to our troops to make sure that theyre
  well-paid, well-trained, well equipped
  (comensurate)
• Even Napoleon had his Watergate! (Nixon or
  Waterloo)

11
Code-switches borrowings ? speech errors

• Llegó el big chief.
• Estaba training para pelear.
• Vivo en ese building, el del rufo verde.
• Vamos a lonchar?
• Está flipando.
• Quiere correr para mayor.
• Se manchó la carpeta.

CODE-SWITCHES
BORROWINGS, LOANS (ESTABLISHED,
NONCE)
CALQUES, LOAN TRANSLATIONS
12
Spoonerisms ? speech errors

• Reverend Spooner, 19th Cent Oxford professor
• Work is the curse of the drinking classes!
  (drink working)
• queer old dean (dear old queen, i.e., Queen
  Victoria)
• noble tons of soil (sons of toil)
• You have tasted the whole worm. (wasted term)
• I have in my bosom a half-warmed fish
  (half-formed wish)

13
Why do Peoples Tongues Slip?

• Freud
• slips of the tongue, and other such phenomena
  caused by repressed emotions
• More modern approaches
• things go wrong at one or another phase as the
  representation is built, during language
  production

14
Collecting Data on Speech Errors

• Corpora, built up by annotated observed errors
• even you can do this write down what the speaker
  said and as much context as possible, as well as
  what the speaker meant to say
• Sometimes, induced in lab
• I say fuzzy, you say duck experiment
• bias target pairs paradigm
• flip Ns in NP paradigm

dart board dust bin duck bill barn door
(darn bore?)
una niña de los gatos unos gatos de la niña
(unas gatas del niño? un gato de las niñas? )
15
Three Broad Categories

• PERSEVERATION
• ANTICIPATION
• EXCHANGE
• Useful hint in the next three slides, check the
  direction of the arrow, which points to the
  location of the error.

16
PERSEVERATION

• an earlier segment/word perseveres
• I cant cook worth a cam.

c
c
damn
17
ANTICIPATORY

• intrusion of a segment/word thats coming up,
  thats being anticipated
• taddle tennis leading list
• leaders of Lebanon

t
t
l
l
paddle
reading
l
L
cedars
18
EXCHANGE

• switch of two segments/words
• hass or grash
• you have tasted the whole worm

sh
ss
hash or grass
t
w
wasted
term
19
Whats persevering, being anticipated, being
exchanged?
INTENDED STRING ? ERROR (whoops!)
pink stems tink spems
find wit fide wint
cedars of Lebanon cedars of lemadon
Terry and Julia derry and chulia t?ul??
big and fat pig and vat
clear blue sky glear plue sky
pedestrian tebestrian
scatterbrain spattergrain
20
Speech Errors Constrained

• Certain things just dont happen.
• Why not?
• What constrains the behavior?
• Impossible forms never created
• no error ever violates phonotactic constraints
• no error ever violates phonological /
  morphological rules
• slips of the tongue ? tips of the slung
• tlips
  of the sung

21
More Constraints

• Similar elements are always involved in
  substitutions and exchanges
• Cs Vs dont exchange, but Cs Cs do, so
  do Vs Vs
• Content Content words, yes Content Function
  words, no!
• Large majority of errors occur within clauses
• This suggests speech is planned / organized in
  clause-sized bundles

22
More Constraints

• Errors respect syntactic structure
• Theres an island on the small restaurant
• (Intended Theres a small restaurant on the
  island)
• (Never Theres island restaurant on the a small)
• Exchange involved NPs, not just Ns
• Resulting sentences are always grammatical!

23
Details of When Rules Apply

• Consider the following error
• You ordered up ending
• (Intended You ended up ordering)
• (And never You ordering up ended)
• -ing morpheme didnt move up with order why not?
• morphemes are perhaps part of the derivation, by
  themselves?
• morphemes get stranded

24
More Such Examples

• We roasted a cook
• (Intended We cooked a roast)
• If you give the nipple an infant
• (Intended If you give the infant a nipple)
• You have a lot of churches in our minister.
• (Intended You have a lot of ministers in our
  church.)

25
And Yet More Examples

• Esas bocas no han salido de mi palabra.
• (Intended Esas palabras no han salido de mi
  boca.)
• No quiero que crea.
• (Intended No creo que quiera.)
• Las manecillas sin reloj
• (Intended El reloj sin manecillas.)

26
Stress (and Sentence-Level Prosody)

• Sentence-level stress is applied based on the
  structure of the clause (its not just something
  associated with a particular lexical item), so
  stress wont move with lexical exchange error
• When the PAPER hits the story
• Intended When the STORY hits the paper
• (Never When the paper hits the STORY)
• Stop beating your BRICK against a head wall
• Intended Stop beating your HEAD against a brick
  wall
• (Never Stop beating your brick against a HEAD
  wall)

27
Speech Errors Constrained

• Errors never generate impossible forms or
  ungrammatical sentences
• Phonotactic constraints are always obeyed
• Syntactic constraints are never violated /
  syntactic structure is never misconfigured
• Errors typically involve similar elements (same
  lexical category, same segmental category)
• Errors usually occur within a clause
• Errors offer interesting data regarding the steps
  (modules) involved in planning speech

28
Inducing Freudian Slips

• Motley (1980) bias / target word-pair reading
  task
• e.g. dart board, dust bin, duck bill barn door
  ? darn bore
• 10-15 errors produced on target pairs
• errors more likely if
• results are real words barn door ? darn bore
  (cf. born dancer ? dorn bancer)
• phonologically similar target words (? vowels)
  left hemisphere ? heft lemisphere (cf. right
  hemisphere ? hight remisphere)
• manipulations of context electrical / sexual
• shad bock ? bad shock
• tool kits ?

29
LEXICAL RETRIEVAL

• Security is the essential roadblock to
  achieving the road map to peace. (GWB,
  Washington, DC, July 25, 2003)
• The law I sign today directs new funds and new
  focus to the task of collecting vital
  intelligence on terrorist threats and on
  weapons of mass production. (GWB, Washington,
  DC, November 27, 2002)
• Source The Complete Bushisms, Jacob Weisberg,
  http//slate.msn.com/id/76886/

30
Among the Many Words You Know

• How do you go about selecting the ones you wind
  up using in your sentences?
• How is a lexical item retrieved?
• FORM MEANING and FREQUENCY
• SPEAKER v. HEARER
• speaker uses meaning to retrieve words, but does
  form ever help choose?
• uses form to retrieve words, but does meaning
  ever bias?

31
How is the Lexicon Organized?

• Frequency
• Meaning close meaning associates to a word are
  stored nearby
• I hate uh, I meant LOVE dancing with you!
• I just feel like whipped cream and mushrooms
• (strawberries)
• All I want is something for my shoulders
• (elbows)
• Sound / Form words with similar form are also
  stored nearby
• If you can find a gargle around the house
• (garlic)

32
How is the Lexicon Organized?In addition to
frequency

• this restaurant hasnt been awake very long
• (open)
• Put the oven on at a very low speed.
• (temperature)
• The picture of the whale from Jaws
• (shark)
• Dont forget to wash your head tonight.
• (hair)
• Youll earn her eternal grapefruit.
• (gratitude)
• I give you my undevoted attention.
• (undivided)

MEANING lemma
FORM lexeme
33
How is the Lexicon Organized?Evidence TOT States

• TOT Tip of the Tongue
• You know the word you need, but you cant quite
  get it
• But you always knowsomething about it
• number of syllables
• initial or final sounds/letters
• location of primary stress
• Whats the item on the right?

34
A Model of Speech Production
MEANING
MEANING lemma
SEMANTICENCODING
LEXICAL RETRIEVALSTRUCTURAL ASSIGNMENT
FORM lexeme
APPLICATION OF PHONOLOGICAL MORPHOLOGICAL RULES
PHONOLOGICAL REPRESENTATION
instructions to articulatorsproduction of
speech
SIGNAL
35
What about features?

• Null hypothesis retrieval and assignment of
  inflections is the same for all sorts of
  inflectional features
• Gender and number agreement errors in Spanish
• Su rendimiento en esta prueba no está asociada?
  con...
• Es el que más pinto? más pinta de chino tiene.
• Lleva once años casados? con
• A tan sólo unos centenares de metro?
• Igoa, J. M., García-Albea, J. E.,
  Sánchez-Casas, R. (1999). Gender-number
  dissociations in sentence production in Spanish.
  Rivista di Linguistica, 11.1, 163-196.

36
Agreement in Spanish

• person and number in agreement relations
  between a verb and its subject
• La niña ha comido pan.
• Las niñas han comido pan.
• gender and number in agreement relations
  between a noun and (i) the constituents within
  the NP or (ii) an adjunct predicate of the NP
• La profesora alemana
• Las profesoras alemanas
• El niño jugaba contento.
• Los niños jugaban contentos.

37
Levels of Representation

• morphological (inflectional suffixes)
• -a (gender)
• -s (number)
• syntactic
• gender/number marking of agreeing constituents
• conceptual/semantic
• sex (gender)
• numerosity / notional number (number)

38
A Difference in Representation

• GENDER can be determined by the semantics of
  words, or by their morphology
• mujer / hombre
• silla / zapato
• it must be lexically specified (for nouns)!
• it is arbitrary
• NUMBER not predictable from the semantics of
  phrases, or of words
• Abunda el polvo en mi apartamento.
• La policía llegó.
• Todo el mundo es bueno.
• it isnt lexically specified!

39
A Typology for Spanish Gender

• Elías-Cintrón (1994)

A/B differ in biological sex
? variable feature
Type Example Mrph Synt Sem
A el vecino / la vecina ?? ? ?
B el estudiante /la estudiante ? ?
C la víctima ?
D el puerto / la puerta
ANIMATE
INANIMATE
A only type where gender is derived by rule
elsewhere, its lexical
D false pair
40
A Typology for Spanish Number
Elías-Cintrón (1994)
? variable feature
Type Example Mrph Synt Sem
W el vecino / los vecinos ?? ? ?
X el aguafiestas /los aguafiestas ? ?
Y la policía ?
Z nadie
41
Gender Number Same Processing Loci?

• Null hypothesis retrieval and assignment of
  inflections is the same for all sorts of
  inflectional features
• Dissociation hypothesis gender features are
  retrieved and assigned from lemmas, while number
  features are independently assigned to lemmas and
  later to word forms through phrase structure
  building operations

42
Corpus Analysis Data

• 725 errors involving gender and number (from
  Corpus of Spanish Slips of the Tongue, Del Viso
  et al., 1987)
• SUBSTITUTIONS
• EXCHANGES
• MOVEMENT ERRORS WITH STRANDING
• NONCONTEXTUAL ERRORS

43
Substitutions

• Gender
• Feminine for masculine
• Su rendimiento en esta prueba no está asociada
  con
• Masculine for feminine
• Es el que más pinto de chino tiene
• Number
• Plural for singular
• Lleva once años casados con
• Singular for plural
• A tan sólo unos centenares de metro

44
Exchanges

• Gender
• He cantado líneo y binga.(línea y bingo)
• Number
• Y pagamos a media las cuotas.(a medias la cuota)

45
Movement Errors with Stranding

• Gender
• Una cuera de suelo(Una suela de cuero)
• Number
• Esas bocas no han salido de mi palabra.(Esas
  palabras no han salido de mi boca.)

46
Noncontextual Errors

• Form-based word substitutions
• Hay dos apóstoles. (epístolas)
• Meaning-based word substitutions
• Toma sólo tres tenedores. (cucharadas)
• Context-based word substitutions
• El estómago de las uñas. (rumiantes)
• Word blends
• Hubo un confrontamiento. (confrontación/enfrentam
  iento)

also form-based, no?
47
Corpus Analysis Results
Gender Number Total
Substitutions F(M) 30 M(F) 24 -e 4 P(S) 32 S(P) 9 99
Exchanges 12 (2 with Ns) 22 (13 with Ns) 34
Movement errors with stranding 40 56 131
Noncontextual Errors (next page) (next page) 460
48
Corpus Analysis Results
Noncontextual Errors (lexical errors, par
excellence)
Gender Number Neither
Form-based substitutions 4 0 96
Meaning-based substitutions 17 2 81
Context-based substitutions 17 6 77
Blends 26 0 74
49
Elicited Production Experiment

• To induce stranding of gender / number
  inflections
• Three manipulations
• mismatch, types of gender, plausibility
• Procedure
• ? una niña de los gatos
• ? unos gatos de la niña (no stranding)
• ? unas gatas del niño (gender stranding)
• ? un gato de las niñas (number stranding)
• ? una gata de los niños (gender number)

50
Three Manipulations

• MISMATCH GDR NMBR
• un hermano del abogado ? ?
• un hermano de la abogada ? ?
• un hermano de los abogados ? ?
• un hermano de las abogadas ? ?
• GENDER TYPE MRPH SYNT SEM
• A el niño / la niña ? ? ?
• B el amante / la amante ? ?
• D el libro / la libra
• PLAUSIBILITY
• Plausible una prima de la camarera
• Implausible una camarera de la prima

51
Elicited Production Results
lt
gt


52
Elicited Production Results
Total Stranding
No Stranding
A B
B lt A lt D
lt D
53
Elicited Production Results
Examples (and distribution of errors) un
maestro de la estudiante ? 78 un estudiante del
maestro (N1) or 22 una estudiante de la
maestra (N2) unas monas de la rama ? 73 unas
ramas de las monas (N1) or 27 una rama del
mono (N2)
gt

lt