The Syntax of Verb Morphology in Children with SLI - PowerPoint

Document Sample
The Syntax of Verb Morphology in Children with SLI - PowerPoint Powered By Docstoc
					Linguistic Approaches to

         Dominik Rus

  GU Brain & Language Meeting
Instead of INTRO – SLI case study
Adult: This is Jim. Tell me a story about Jim.
Child: Him going fishing. Jim hold…water. And go fish. And
Adult: Ok. How many more do you think we have?
Child: I don’t know.
Adult: This is Kathy. Tell me a story.
Child: Kathy brush teeth. Her eat. And he get clothes on.
Adult: She must be getting ready to go to school, right?

(Leonard, Bartolini, Caselli, McGregor & Sabaddini 1992)
                     SLI – history - 1
  150 years of interest in language impairments/disorders

  First Terminology for SLI/SLI-like deficits
 congenital aphasia (Väisse 1866)
 hearing mutism (Coën 1886)
 congenital word deafness (McCall 1911)

  Gall 1922 (English translation 1935)

  There are many children…who do not speak to the same degree as
  other children although they understand well or are far from being
  idiotic. In these cases the trouble lies not in the vocal organs, as the
  ignorant sometimes insist, and still less in the apathetic state of the
  subject. Such children on the contrary, show great physical vivacity.
  They not only skip about but pass from one idea to another with
  great rapidity. If one holds them and pronounces a word in their ear,
  they repeat it distinctly.
  (Leonard 1998: 5-6 after Gall 1935: 24; emphasis mine)
                  SLI – history - 2
    Older Terminology on SLI/SLI-like deficits

   developmental aphasia (Kerr 1917) [dominating term until the 50s
   delayed speech development (Fröschels 1918)
   congenital auditory perception (Worster-Drought & Allen 1929)
   infantile aphasia (Gesell & Amatruda 1947)
   congenital verbal auditory agnosia (Karlin 1954)
   aphasia/dysphasia Weiner 1969)
   developmental aphasia/dysphasia (Clahsen 1989; Wyke 1978)
    [70s, 80s]

    PROBLEMS:           (a) SLI subsumed under LI
                        (b) different nature of deficit
                        (c) mainly physicians (bias?)
                        (d) aphasia/dysphasia – inappropriate term
                  SLI – history - 3
      SLI vs. other LI types
     Liebmann (1898): subtypes of children w/ language disorders
(a)   deficit motoric in nature
(b)   children who comprehend only single words
(c)   complete inability to comprehend language
      [only children w/ severe output limitations]

      80s and beyond
     developmental language impairment (Wolfus et al. 1980)
     specific language deficit (Stark & Tallal 1981)
     language/learning-disabled
     language-learning-impaired (Tallal, Ross & Curtiss 1989)
     specific language impairment (Leonard 1981)
     SLI (Fey & Leonard 1983)
          SLI today: major issues
(a)   expressive vs. receptive vs. expressive and receptive disorder
(b)   developmental aphasia/word deafness/developmental language
      disorder (US Department of Health and Human Services)

      clinical vs. non-clinical (e.g., educational/research) world
      approach to a study of developmental disorders
      Why study SLI?
      (a) clinical concerns (intervention)
      (b) educational concerns (assistance; special programs, etc.)
      (c) providing a type of baseline of language impairment
      (d) theories of language acquisition/development & language
          organization (domain-specificity; modularity; compensation;
          critical periods; maturation)
              SLI – quick facts
Disorder: significant limitation in language ability with no apparent
factors that accompany language learning problems (e.g., hearing
impairment, low nonverbal intelligence, neurological damage)
(Leonard 1998)
Onset: at birth
Peak: infanthood & childhood; continues into adulthood in a minority
of cases (BUT careful testing reveals disruptive linguistic knowledge
even w/ older adults; Leonard 1989)
Prevalence: about 7% (Leonard 1998; Tomblin 1996: 7.4%) [“pure
SLI” probably much less”; Tower (1979): 1.5%; American Psychiatric
Association: 5% for production; 3% production + comprehension]
Sex differences: more prevalent in males (usu 3:1) (Leonard 1998;
Tallal et al. 1989, 2001)      [BUT careful w/ these conclusions]
Role of genetic: SLI children more likely to have parents and
siblings w/ a history of language learning problems
                   Etiology of SLI
    Bishop et al 1995: 56
(a) “continuum of reproductive casualty” (perinatal hazard that
    damages higher cortical centers w/out leading to physical
    disability (Pasamanick & Knobloch 1960)
BUT …little evidence for this; may impair cognition but language is
    seldomly disproportionately affected (Bishop 1987)

(b) inadequate language stimulation (the “it is the parents” view)
BUT …affected & unaffected kids found w/in same families

(c) genetics
…most likely
                    SLI criteria - 1
   based on inclusion & exclusion

   language criterion least problematic
   BUT language is disruptive in a lot of other disorders

 no attentional deficits (eye gaze; JVA)  no theory of mind-related
 though language emerges late(r), there is a lot of variability among
  children (even among healthy-developing children)
 language problems may vary considerably (Bishop 1997; Guasti
                        SLI criteria - 2
 areas considered before SLI applies (Leonard 1998: 10)
       FACTOR                            CRITERION
Language ability         Language test scores of -1.25 SD or lower; at risk
                         for social devalue
Nonverbal IQ             Performance IQ of 85 or higher
Hearing                  Pass screening at conventional levels
Otitis media effusion    No recent episodes
Neurological             No evidence of seizure disorders, cerebral palsy,
Dysfunction              brain lesions; not under medication for control of
Oral structure           No structural anomalies
Oral motor function      Pass screening using developmentally
                         appropriate items
Physical and social      No symptoms of impaired reciprocal social
interactions             interaction or restriction of activities
                  SLI criteria - 3
in general,
NO perceptual-motor deficits, like hearing loss
NO neurological dysfunction
NO intellectual or socio-emotional deficits
(Bishop 1997; Guasti 2002)


“It is evident that these criteria are very loose, for they exclude only
populations with mental retardation and populations with sensory
deficits.” (Guasti 2002: 376)
SLI & Language Assessment -1
      Language Tests
…do not do justice to SLI children’s language problems (Muma 1986),
but are a good starting point in the diagnostic procedure

(a)   (too) wide (of a) range; different modalities (comprehension vs.
(b)   cover different areas/domains of language (e.g., phonology,
      morphology, syntax; or a combination of two or more)
(c)   serve only as the starting point
(d)   different criteria both modality- as well as score-wise (e.g., only
      production or both production & comprehension?; “language” age
      or standard score, or both?)          Rice & Co.: 1 SD below mean
                                            Tomblin & Co.: 1.25….
      Spontaneous Speech (typically in conjunction w/ Lang Tests)
(a)   MLU (morphemes – norm: Miller & Chapman 1981; words – norm:
      Templin 1957)
SLI & Language Assessment -2
TOLD-P:2 (Test of Language Development-Primary:2) (Newcomer &
          Hammill 1991)

SLI: usually 81 or lower (1.25 SD below mean)

 appropriate for kids aged 4;0 – 7;11 (Leonard 1998)
 3 “grammatical” subtests (2 production, 1 comprehension); 2 lexical
  tests (1 production, 1 comprehension); 2 phonological tests (1
  production, 1 comprehension)
      SLI & Nonverbal Intelligence
      score on nonverbal IQ one of the most fundamental criteria
      usual norm: IQ of at least 85, or less than 1 SD below mean (but
      this may vary; Bishop et al. 1999: 80 or more
      discrepancy btw/ IQ & language (the more, the better; usually at
      least 15)
      discrepancy btw/ mental age & language age of at least 1
      most common IQ tests:
(a)   Raven’s matrices (Raven, Court & Raven 1986)
(b)   Wechsler Intelligence Scale for Children-III (WISC-III) (Wechsler
(c)   short form Performance IQ (PIQ) on the WISC-III
       SLI – How specific? - 1
“Specific language impairment is a condition in which language is
impaired but other cognitive functions are normal.” (Guasti 2002:
376, emphasis mine)

“Children with this condition demonstrate a delayed emergence of
grammar at the same time that prerequisite levels of hearing acuity,
general cogntive ability, and social development meet normative
expectations.” (Rice & Wexler 1995: 451, emphasis mine)

“SLI children are characterized by severe problems in the
development of language comprehension and expression but do not
have an impairment in nonlinguistic cognitive or motor development,
hearing, or emotional-social behavior…” (van der Lely 1996; partly
citing Benton 1964, emphasis mine)
       SLI – How specific? - 2
subgroups within SLI?

Grammatical SLI (G-SLI) (van der Lely 1996, 2004a,b,c, 2005;
van der Lely & Stolwerck 1996)

“I do not claim that Grammatical SLI is, necessarily, an autonomous
sub-group from all other SLI sub-groups, but I propose that the
Grammatical SLI children are homogeneous within the sub-group.”
(van der Lely 1996: 189, emphasis mine)

…although poor sensory or non-verbal abilities often co-occur with
SLI, there is no evidence that these impairments cause the
grammatical deficits found in SLI. Moreover, evidence suggests that
impairment in at least one subgroup is specific to grammar.” (van
der Lely 2005: 53, emphasis mine)
         SLI – How specific? - 3
  “There are some serious flaws in the argument presented by van der
  Lely and Howard (1993) that children with specific language
  impairment do not have deficits in verbal short-term memory,
  despite earlier evidence that they do (Gathercole & Baddeley
  1990).” (Gathercole & Baddeley 1994, emphasis mine)

  “Specific Language Impairment in Children is Not Due to a Short-
  Term Memory Deficit: Response to Gathercole & Baddeley” (van der
  Lely & Howard 1994)

  …despite the standard use of exclusionary criteria to diagnose SLI,
  the disorder is clearly not limited to language.” (Ullman & Pierpont
  2005: 399, emphasis mine)

               Approaches to SLI
   Linguistic (Competence)

   (Non-linguistic) Processing (Performance)

   Neurocognitive/Neuropsychological (Brain + Behavior/Function)

attempt to characterize the deficit (of some SLI
populations) by using linguistically motivated criteria

 language is disrupted
 “learnability problem” at the heart of the language acquisition
 test linguistic theories against data from language acquisition
  (impaired + healthy-developing kids)
  SLI: Common Linguistic Features - 1

   language emerges late(r)

  language shows unexpected patterns of phonology, morphology,
  and syntax, remaining below age expectations

  one of the most affected domains is that of verbal inflectional
  morphology (specifically 3 SG PRES TNS –s and PAST TNS –ed

  language impairments usually continue well into elementary school
  years and even persist into adulthood
  SLI: Common Linguistic Features - 2

  pluralization of nouns (accuracy 90% and above, Oetting & Rice
  1993) [interestingly, SLI subjects overregularize regular –s (*foots,
  *mans) evidence contra rote learning/memorization; Oetting & Rice

  BUT Leonard et al. (1992): mean correct use of plural –s
  68.6% only (WHY? Statistical analysis with SD: 34%?,
  different population?)

  prepositions (Rice & Wexler 1996)

  when morphemes are supplied, they show correct forms (features)
  (*She are here)
 Linguistic Heterogeneity in SLI

  not every aspect of inflectional morphology is equally impaired
  areas other than inflectional morphology may be impaired (e.g.,
  passive voice, etc.) (van der Lely 1996, 1998; van der Lely &
  Harris 1990)
  the acquisition of words (esp. verbs) is vulnerable (Rice et al.
  1994; Oetting, Rice & Swank 1995)
  word retrieval problem sometimes occurs w/ SLI (Leonard 1998)
  (mild) phonological deficits can be observed in some subjects
  the disorder may be receptive and/or expressive
  in some cases language impairments disappear in elementary
  school years or not before adulthood
Sources of Linguistic Heterogeneity
              in SLI
 methodology (different probing of grammatical knowledge; different
 representation and/or interpretation of data)

 age of SLI subjects (older SLI subjects usually better than younger
 ones; compensation?, maturation?)

 comorbidity with other developmental disorders (SLI a lot of times
 used as a cover term for possibly a whole range of language
 disorders; Bishop 1997)
Linguistic Approaches to SLI
deficit alters local aspects of grammar

deficit in establishing agreement relations (particularly,
subject-verb agreement relation)

deficit in computing structure-dependent relations

deficit targets only particular grammatical (morphosyntactic)
features, such as [TNS] and [AGR] or [PLURAL])

Surface Hypothesis
deficit alters local aspects of grammar - 2
[researchers: Rice, Wexler & co.: Cleave, Grimm, Noll, Oetting, Schütze]

   absence of tense and/or agreement [features] or finiteness in
   Wexler’s (and Rice’s) Extended Optional Infinitive (EOI) Stage
   (developed from Optional Infinitive (OI) Stage posited for normally
   developing kids)

   TNS feature is optionally missing and TNS morphemes are thus
   optionally omitted
   other inflectional morphemes (e.g., plural marker on nouns) and
   prepositions are NOT omitted
   when children choose the TNS feature, they respect the
   morphosyntactic properties (e.g., word order (V placement as a
   consequence of V movement), etc.)
   globally, grammar is intact: same categories (e.g., Tense), same
   movement properties (e.g., verb movement) (Rice et al. 1995; Rice
   & Wexler 1996; Wexler et al. 1998; Wexler & Schütze 1999)
deficit alters local aspects of grammar - 2

 the “Truncation” hypothesis; similar to EOI but different mechanism
 Deficit to unoperational mechanisms of grammar (Rizzi 1993/4)
 possibly also due to processing limitations; Rizzi 2002)




deficit alters local aspects of grammar - 3

      underspecification in early systems: NOT meaning that certain
      parameters are not fixed but rather they are NOT OPERATIVE
      initially (maturation, as in Borer & Wexler 1987)

      SLI grammars are underlyingly the same as grammars of normally-
      developing children

      if the maturational approach is correct, we’d expect not only
      (a) structures attested somewhere else in actual or potential natural
      languages, but also structures that NO ADULT NATURAL LANG will
      exhibit – a potential candidate: ROOT INFINITIVES (RIs)

[1]        Mama throw ball
[2]        Quetto qui mangiare chellini (Italian: Martina, 1;11)
           this   here eat-inf piglets (Guasti 1993/4)
deficit in establishing agreement relations (particularly,
subject-verb agreement relation)

[researchers: Clahsen & co.: Bartke, Göllner, Rothweiler; Tsimpli &

(a)     SLI children have trouble with AGR morphemes
(b)     SLI children do not have trouble with other inflectional morphemes
(Clahsen 1986, 1991; Rothweiler & Clahsen 1993; Clahsen, Bartke &
Göllner 1997)

German-speaking SLI children are proficient in using TNS morphemes, but
not AGR morphemes (Rothweiler & Clahsen 1993)

TNS is supplied perfectly (99% corr on lex Vs & 100% o SEIN), but AGR is
not (67% on lex Vs & 93% corr. on SEIN) (Clahsen, Bartke & Göllner
           deficit in computing structure-
                 dependent relations
      [researchers: van der Lely, Stollwreck & co.; Connell & Franks]

   Subject-agreement relation
   Binding theory (interpretation of pronouns and anaphors)
   Active vs. Passive

difficulties with structure-dependent representations


“This claim might be too strong. Most linguistic knowledge is couched in
   structure-dependent terms.: (Guasti 2002: 391)
    deficit targets only particular grammatical (morphosyntactic)
           features, such as [TNS] and [AGR] or [PLURAL])
                 [researchers: Crado & co., Gopnik & co.]
    Missing grammatical features such as [+PLURAL] or [+PAST]
    Evidence from several languages (Canadian & British English,
    Greek, Japanese, Quebec French)
    Children cannot construct implicit rules governing morphological &
    phonological processes in grammar (Gopnik et al. 1997)
    Children cannot see the internal structure of inflected words & are
    not able to build implicit ruels for handling inflectional morphology
    (een regular morphology is rote-learned & stored in the lexicon as
    unalyzed chunks: Gopnik et al. 1997; Guasti 2002: 392)
(a) frequency effects for regular & irregular words
(b) difficulties w/ the inflection of novel words
(c) incorrect segmental & prosodic features of inflected words
[a la Pinker’s word-and-rule theory]
      SLI & word-and-rule theory
      SLI children:
(a)   bad on the wug test (usually /wags/)

(a)   Search for phonologically similar item w/ novel Vs (brom-?:

(a)   generally 30-50% correct on regular infl morphology (rote-
      learning, apparently)

(a)   retrieval of irregular Vs subject to frequency effects

(b)   BUT Rice & Oetting (1993): SLI kids process regular & irregular
      plurals distinctively
 Empirical Studies of SLI: Methodology

   usually 2 control groups: age-matched (AM) + language matched
   (=MLU-matched) (LA)

   usual lag of 2 years (e.g., SLI5, N5, N3: 5 year-old SLI kids vs. 5
   year-old normals matched on age + 3 year-old MLU-matched

   spontaneous speech + probe (elicitation tasks)

   Example of PROBES:
This is Mary. She sings. /…/ What did she do? (Rice ?)
Every day I spling around London. Just like every day, yesterday I
   _________around London ((van der Lely & Ullman 1996)
Empirical Studies of SLI: Shortcomings
   Shortcomings of Empirical Studies

(a) SLI subjects of different ages
(b) different methodology
(c) spontaneous speech data reported generally from the same corpus
    (Kansas Language database…)
(d) Vs not controlled for syllable length and/or frequency
(e) usually no intra-subject data (i.e., performance on each morpheme
    within the subject)
(f) studies have never compared all morphemes at the same time (-s-, -
    ed-, -ing, stem) within the same data corpus
(g) most studies have not looked at the correlation between a null
    subject and an uninflected form, as observed in normally developing
(Pretty) Unexplored Areas/Topics

neurocognitive accounts (the brain has been largely
ignored; Ullman & Pierpont 2005)

cross-linguistic evidence (part. morphologically complex

bridging linguistics and neuropsychology/neuroscience
(The Hierarchy Hypothesis?)
                         HCH -1
Hierarchy Complexity Hypothesis - Merge Impairment
(combinatory mechanism that merges categories)
evidence from the neurolinguistics of aphasia (Broca's
type/agrammatic aphasia)
anterior aphasics vs. posterior aphasics (Izvorski & Ullman)
“the more you merge, the worse you are”
Hierarchy: CP-AgrP/TP-NegP-AsP-VP
difference from the pruning hypothesis (PH)? the PH does not allow
for deficits to be graded across functional categories w/in an
individual subject
not all functional categories are fully and equally impaired
The relative height of a given category in the hierarchy predicts its
severity of impairment (Pancheva & Ullman)
“Linguistic forms dependent upon higher categories should be
probabilistically, not categorically, more problematic than those
dependent on lower categories.” (Pancheva & Ullman)
w/in and across subject differences
                              HCH -2
(a)   possibly (?) related to Brown’s order of morpheme acquisition

      ing > irregular past > uncontractible COPBE > regular past > -s (reg)
      > -s (irreg) > uncontractible AUX > contractible COPBE >
      contractible AUXBE

BUT        Puzzles/Problem: why is uncontractible COPBE acquired
           before –ed or –s? Status of (un)contractible AUX &
           contractible COPBE?

(b)        Why patterns of CP-VP w/ no intervening TP/AGrP in healthy-
           developing kids as well as second language learners
                             HCH -3
(c) different errors depending on whether verbs come with inflection &
    are licensed by some checking procedure vs. inflectional
    morphology is in the syntax??? [MERGE (Chomsky 1995) IS

(d) Is all that is at stake the fact that this sequence of FCs might matter
   in processing?  performance model rather than competence

(e) Should HCH explain all patterns of linguistic behavior (i.e., normally
   developing children as well as SLIs, aphasics, etc.?)

(f) HCH & Ullman’s DP model (frequency effects, reg vs. irreg)

(g) How to explain deviations from HCH’s predictions in SLI literature?
   (e.g., better performance at –s than –ed)?
  Future Research/Issues to be
Hierarchy Hypothesis & SLI: Hierarchy
Hypothesis and cross-linguistic evidence
(Spanish, Slovenian?)

going beyond English SLI in general: more
cross-linguistic data

compatibility of data from healthy-developing
children with SLI children […working on it ]