How native are heritage speakers?
Silvina MontrulUniversity of Illinois at Urbana
Champaign
Heritage Language Summer InstituteUniversity of California, Los AngelesJune 18-22, 2012
Who is a native speaker?
• We all have an intuitive notion of what a native speaker is or should be (Sharwood-Smith 2011)
• Precise definitions are elusive (Davis 2003)
Stability vs. constrained variability
• Abstract linguistic knowledge (Chomsky)• Sociolinguistic variation (Labov)• Psycholinguistic variation (working
memory, executive control, aptitude, other individual variables)
Knowledge of language
Birth 4 yrs 12 yrs 18 yrs 40s
Knowledge of language
complete, stable?
FULLY FLUENTNATIVE SPEAKER
DEVELOPINGNATIVE SPEAKER
What develops?
• A phonological system• A lexicon• A set of grammatical rules and principles• Morphological expressions of forms and
meanings• Sentence structure (syntax)• communicative competence• Sociolinguistic competence
Examples of complete, fluent knowledge of language
Mature educated native speakers• Pronounce the sounds of their language well • Do not make morphological errors of omission or commission.
They are more than 90% accurate on the use of morphology in obligatory contexts.
• Know how to conjugate their verbs and make agreement in phrases
• Know many words in their language• Speak and write in grammatical sentences.• Understand different meanings of words and phrases• Know how to use language in different sociolinguistic contexts• Have pragmatic competence• Show consistent ceiling performance in tasks of grammatical
ability regardless of modality of task
Key variables that define and affect the developing native speaker
• Exposure to the language from birth• Use of the language at home• Schooling in the language• Socialization beyond the home in the language• Consistent exposure and language use in a variety of
contexts and communicative situations until about early adulthood
• Abnormal language development (SLI, down syndrome, autism, etc.)
Types of Native Speakers
Monolingual native speakers
Bilingual native speakers
Monolingual native speakers vary in
SES: low, mid, high SESLiteracy: literate, semiliterate, illiteratePathology: healthy vs. language impairedOther
How these variables affect linguistic competence is a matter of debate (see Dabrowska, 2012 and commentaries)
Bilingual native speakers vary in
All dimensions of monolingual speakers as well as in
• age of acquisition of the 2 languagessimultaneous bilingualssequential bilinguals
• degree of use of the language/sfluent vs. non-fluent
Ultimate attainment in monolingual native speakers
Birth 4 yrs 12 yrs 18 yrs 40s
Knowledge of language
complete, stable?
FULLY FLUENTNATIVE SPEAKER
DEVELOPINGNATIVE SPEAKER
beginning middle end
What is ultimate attainment?
The end state of the acquisition/language development process.
Is ultimate attainment always “native” level in bilingual native speakers?
NO• It can be fully native, as in monolingual native
speakers• It can be near-native, as in some L2 learners• It can be non-native, as in most L2 learners
Typology of Bilingual Native Speakers
• The fully fluent native speaker • The interrupted native speaker (heritage
speakers, international adoptee)• The attrited native speaker• The bilingual aphasic native speaker• Other?
“Native” ability
• Can also be dissociated in bilinguals• E.g., native or near-native in phonology and
non-native in morphosyntax (heritage speakers in Au et al. 2002)
• Or native/near native in morphosyntax and non-native in phonology (near-natives in White & Genesee 1996)
• Very few L2 speakers are “native” on all linguistic dimensions (Abrahamsson & Hyltenstam 2009)
Purpose of this presentation
To show that despite exhibiting high degree of variability in degree of ultimate attainment like L2 learners, heritage speakers show a much higher incidence of native ability in morphosyntactic and lexical aspects of language that are extremely hard for L2 learners to master at native levels, even after significant amounts of input.
Heritage speakers and L2 learners
If we control for proficiency, does early language experience bring advantages to Spanish heritage speakers in their knowledge of early acquired aspects of morphosyntax when compared to late L2 learners of Spanish?
Advantage = more native-like performance
17
Phonology (Au et al. 2002)
18
Morphosyntax (Au et al. 2002)
Gender Agreement
• Mastered by monolingual Spanish-speaking children error free (100% accuracy) by age 3 (Montrul 2004) in oral production.
• Yet, full mastery of gender agreement in production is highly unlikely in L2 acquisition, even in so called near natives, with the highest amount of exposure in the language for several years and proficiency scores on global measures that fall within the range of variation of native speakers.
Examples
Franceschina (2001): Case study of Martin (British guy who had been living in Argentina for more than 30 years)
Almost 10% of gender agreement errors in production, especially with adjectives, articles and demonstratives.
Grüter, Lew-Williams & Fernald 2012: 19 near native speakers of Spanish exhibited 20% errors in an oral production task (17.2% assignment, 2.8% agreement)
Near native ability?• Studies that have found that non-native speakers
do not differ from native speakers have used tasks focusing exclusively on regular or canonical ending nouns: ending in –a if feminine and in –o if masculine (e.g., White et al. 2004).
• Several studies have shown that gender assignment and agreement with non-canonical or non-transparent nouns take longer to learn and to process (Bates et al.1995, 1996; Taraban & Kempe 1999, Taraban & Roark 1996).
Spanish Masculine Nouns
canonical
-o
non-canonical-e
non-canonical-cons
non-canonical-a
libro ‘book’
puente‘bridge’
lápiz‘pencil’
problema‘problem’
ojo ‘eye’
coche‘car’
mantel‘tablecloth’
mapa‘map’
suelo ‘floor ’
cable‘wire’
reloj‘clock’
planeta‘planet’
Spanish Feminine Nounscanonical
-a
non-canonical-e
non-canonical-cons
non-canonical-o
casa‘house’
leche‘milk’
nariz‘nose’
mano‘hand’
mesa‘table’
fuente‘fountain’
piel‘skin’
foto‘picture’
manzana‘apple’
llave‘key’
canción‘song’
moto‘motorcycle’
Montrul, Foote & Perpiñán (2008)
• 140 Spanish L2 learners and heritage speakers ranging from low to advanced levels of proficiency.
• Both L2 learners and heritage speakers made gender errors, especially with non-canonical ending nouns.
• Advantages for heritage speakers on gender agreement depending on task.
• L2 learners performed better than heritage speakers in highly metalinguistic written tasks.
• Heritage speakers performed more accurately than L2 learners in oral production tasks.
Task Effects
50
55
60
65
70
75
80
85
90
95
100
native speakers heritage speakers L2 learners
Written PIT
Written RT
Oral PDT
Revisiting knowledge of gender agreement (Montrul, Davidson de la Fuente and Foote)
Control for modality and avoid use of written language (literacy effect).
Investigate both regular and irregular nouns.4 aural/oral experiments that vary on the
implicit/explicit dimension 1. Timed grammaticality judgment task (GJT)2. Timed aural gender monitoring task (GMT)3. Timed oral word repetition task (WRT) 4. Elicited production task (EPT)
Participants• 24 Spanish native speakers (control group)• 29 Spanish heritage speakers (acquired
Spanish at birth and English before age 6)• 37 English-speaking L2 learners of Spanish
(acquired Spanish after age 12)Heritage speakers and L2 learners ranged from
intermediate to advanced based on a written proficiency test.
Picture-Naming Task
Rationale: proficiency measure based on oral production
Stimuli:48 inanimate object nouns– frequency of 3 or higher in Spanish (Alameda &
Cuetos, 1995)– 24 canonical endings: 12 masculine -o, 12 feminine –a– 24 non-canonical endings: 6 masculine -cons, 6
feminine -cons, 6 masculine -e, 6 feminine -e
Picture-Naming Tasks• Participants completed the task both in Spanish and
in English (only HS and L2ers); native speakers only in Spanish.
• They were asked to view a series of black and white images and to name them as quickly as possible after hearing the audio prompt “diga” / “say” (recorded by a female Spanish native speaker)
• Items in both tasks were presented in random order• Naming accuracy and reaction times (after the onset
of the prompt) were measured
Picture-Naming TasksImage samples
Libro “book”
Sobre “envelope”
Corazón “heart”
Casa “house”
Llave “key”
Flor “flower”
Picture-Naming Tasks: ResultsEnglish
Picture-Naming Tasks: ResultsSpanish
****
Word-Recognition Experiments
1. Timed grammaticality judgment task (GJT)2. Timed aural gender monitoring task (GMT)3. Timed word repetition task (WRT)
(Bates et al. 1995, 1996, Guillelmon & Grosjean 2001)
Experimental Design• For all three tasks, 300 determiner-noun-
adjective phrases (half target, half fillers) were constructed with 150 nouns, 3 determiners (masculine el, feminine la, neutral su) and 7 adjectives.
• All nouns were inanimate (half feminine, half masculine) with canonical and non-canonical endings, controlled for syllable length, stress, and frequency.
• All tasks used the same stimuli but with different distribution of fillers and targets in 3 conditions.
Conditions used in the three tasks
Conditions Gender Noun endingcanonical Non-canonical
grammatical feminine la gran casa la gran calle
masculine el peor texto el peor viaje
ungrammatical feminine *el gran casa *el gran calle
masculine *la peor texto *la peor viaje
neutral (control)* feminine su gran casa su gran calle
masculine su peor texto su peor viaje
NOTE: only the GMT and the WRT had a neutral condition, the GJT did not.
Procedures• The GMT required participants to listen to the noun
phrases and push one of two buttons on the keyboard (one for feminine, one for masculine), depending on the gender of the noun. (VERY EXPLICIT FOCUS ON GENDER)
• In the GJT, participants listened to the noun phrases and pushed one of two buttons to indicate whether the phrase was grammatical or ungrammatical. (INDIRECT ATTENTION TO GENDER)
• In the WRT, participants heard the noun phrases and were asked to repeat the last word in each phrase as quickly and accurately as possible. (NO ATTENTION TO GENDER)
Predictions
Task Type of response
Degree of explicitness
Advantages for HS over L2 learners?
GMT Decide whether a noun is feminine or masculine
very explicit focus on gender
no
GJT Decide whether a noun phrase is grammatical or ungrammatical
Explicit, but indirect focus on gender
no
WRT Repeat the last word in the phrase
implicit yes
Gender Monitoring Task: Accuracy
Grammaticality effect for all three groupsNative speakers > [heritage speakers = L2 learners]
Canonicity effect for L2 learners and HS.
Summary Effects GMT(difference % between ungrammatical and grammatical sentences)
Gender Monitoring Task RTs
Grammaticality effect for all three groupsNative speakers > [heritage speakers = L2 learners]
Canonicity effect.
Summary Speed Effects GMT(grammatical - ungrammatical RTs)
Grammaticality Judgment Task Accuracy
Grammaticality effect for all three groups.Native speakers > [heritage speakers = L2 learners]
Summary Accuracy Effects GJT
Grammaticality Judgment Task RTsGrammaticality effect for all three groupsNative speakers > [heritage speakers = L2 learners]
Summary Speed Effect GJT
Word Repetition Task--RTNo grammaticality effect for L2 learners.[native speakers = heritage speakers] >L2 learners
Summary Speed Effect WRT
Summary• Canonical and noncanonical nouns are processed differently.
Noun ending did not affect the native speakers to the same extent as the two experimental groups.
• HS and L2 learners were slower and less accurate on non-canonical ending nouns than on canonical ending nouns.
• The results of the GJT and the GMT revealed significant grammaticality effects for all groups. – They use gender cues on determiners in noun recognition.
• In the WRT, the NS and the HS showed a grammaticality effect, while the L2 learners did not. – L2 learners may not have the same type of implicit knowledge of
gender tested by this type of task.
Task effectTask Response Degree of explicitness
GMT Masculine/feminine Very explicit
GJT Grammatical/ungrammatical explicit
WRT repeat implicit
Favors heritage speakersHelps L2 learners
Conclusion
• These results confirm that HS have an advantage (i.e., show native-like patterns) over L2 learners in tasks tapping implicit knowledge. – Although this advantage could be due to age of
onset of bilingualism (early vs. late) (Guillelmon & Grosjean 2001), it may also be related to context of acquisition (naturalistic vs. instructed) and experience with oral production.
The Role of Experience
• Elicited Oral Production Task (untimed)• Elicitation of simplex and diminutive nouns with
gender agreement. elefante elefantito
simplex diminutive
Why diminutives?
• Hallmark of Child Directed Speech in early language development
• Highly productive morphological mechanism• Appear to facilitate the acquisition of
declensional noun endings in many languages (Savickienė & Dressler 2007): Lithuanian, Russian, Greek, Italian, Spanish, Dutch, Austrian German, Hungarian, Finnish, Hebrew
Diminutives in Child Language
• They are acquired early due both to their frequency in the input (in Child Directed Speech) and their morphological characteristics in many languages.
• In Spanish, they are acquired/used productively between the ages of 1;9 and 1;10 (Marrero, Aguirre and Albalá, 2007)
Spanish
• Aguirre, Marrero & Albalá (2007) claim that the frequency of usage of diminutives is not very high in Standard Spanish.
• Yet, Spanish-acquiring children use 13 times more diminutives than Spanish-speaking adults, and adults addressing children use them as much, if not more than the children themselves (Marrero, Albalá & Moreno, 2002: 155).
• In Spanish, most diminutives appear with nouns (but also with adjectives and adverbs).
Regularizing Gender Suffixes with Diminutives
-ito , -ita (among other dialectal variants)
Canonical ending nouns Non-canonical ending nouns
simplex diminutive simplex diminutive
Feminine la casa la casita la nariz la naricita
Masculine el auto el autito el coche el cochecito
Diminutives and gender• Seva et al. (2007)
Experimental study of 2 and 3 year old Russian and Serbo-Croatian children
• children were given simplex and diminutive forms of nouns and were asked to produce noun phrases with adjectives showing gender
• the toddlers were more accurate at gender agreement with diminutive nouns than with simplex forms. -> DIMINUTIVE ADVANTAGE
Research Question
Does the diminutive advantage carry over into adulthood?
Hypotheses• Heritage speakers should know more (about)
diminutives (i.e. their form and how to use them) than L2 learners.
Hypotheses• Heritage speakers should be more accurate
at producing gender agreement in general and with noncanonical nouns in particular than L2 learners of Spanish.– Diminutives help regularize noncanonical nouns– Heritage speakers were exposed to child directed
speech in Spanish; adult L2 learners were exposed to more “adult” input in Spanish, which contains very few diminutives.
Elicited Oral Production Task
• Stimuli came from the noun images used in the picture-naming tasks
• Subjects were now asked to produce utterances containing a determiner, a noun and a color adjective
• 8 color adjectives used: 4 explicit (rojo/a), 4 non-explicit (azul)
• Target nouns were randomly assigned an explicit color adjective, distracters were randomly assigned a non-explicit color adjective
Design and Procedure• Participants saw a total of 96 images (both
simplex and diminutive forms were requested)• Order of presentation was randomized and held
constant across participants• Participants’ responses were monitored by a
researcher with a check-list• Check-lists and audio recordings were analyzed
by at least two different raters
Elicited Oral Production TaskStimulus Samples
un pan grisun pancito gris
“a gray bread(dim.)”
una cruz negrauna crucecita negra
“a black cross(dim.)”
62
Gender agreement in oral production
Results: Canonicity by Group
Gender by domain of agreement
Canonical Nouns in simplex and diminutive forms
Non-canonical nouns in simplex and diminutive forms
Summary– Native speakers are at ceiling in everything: no
effect for diminutive, gender or canonicity.– Both HS and L2ers perform almost at ceiling
with canonical-ending Ns regardless of whether they are in the simplex or in the diminutive form
–With non-canonical Ns, HS show a clear quantitative advantage over L2ers.
– No Diminutive Advantage for heritage speakers.
Error Analysis by subjects
Errors with diminutives1. produce simplex instead of diminutive form2. wrong form of diminutive (augmentative or other)
Errors with gender1. lexical assignment errors =*el serpiente(F) negro(M)
2. agreement error = *la serpiente(F) negro(M)
3. ambiguous = *el serpiente(F) negra(F)
Percentage of errors by group
Type of errors with diminutives
Percentage of individuals who made gender errors in each group
Types of errors with gender
Summary
• Native speakers do not make errors with gender in production
• Many heritage speakers did not make errors either (19/29)
• All the L2 learners made gender errors• The most frequent error (~80%) in both L2
learners and heritage speakers is the one of lexical assignment.*El serpiente(F) negro
Conclusions
• Early language experience confers some advantages to heritage language learners with early acquired aspects of language and in oral production, like gender agreement.
Are there differences between L2 learners and heritage speakers?
YES, but it depends on structures and tasksMetalinguistic tasks favor L2 learnersLess metalinguistic tasks favor heritage
speakers (see Bowles, 2011)Written tasks favor L2 learnersOral tasks favor heritage speakers (proficiency
matters, cf. Au et al. 2002)
Tasks oral/aural written/visual
implicit heritage speakers
explicit L2 learners
Why?
Learning experienceHeritage speakers are child learnersAural input
L2 learners are adult learnersVisual and aural input
Different input and input processing experience.
Gender processing in L1 acquisitionGender is in the lexicon.Children hear sequences of determiners and
nouns in the acoustic input and must identify nouns in the speech stream (through computations or transitional probabilities)
Spanish-speaking children in the one-word stage produce prenominal vowels that coincide with vowels found in determiners (a fó, e pe, ua queca) (López Ornat 2007, Lleó 2001).
Most Recent Experimental Evidence
• Noun-gender associations are strong in the L1 lexicon.
• Lew-Williams and Fernald (2007, 2010): native Spanish speakers and 3-4 year old Spanish speaking children use gender information in determiners to predict nouns (Visual world paradigm).
L2 learners are different• Already know about determiners and nouns from
their L1.• Visual input gives information about word
boundaries and L2 learners do not need to rely as much on distributional properties and transitional probabilities to segment the acoustic stream.
• The association between noun-determiners and noun-gender in the lexicon is not very strong in the L2 (Grüter, Lew-Williams and Fernald, 2010).
• Input modality matters for language processing and production.
How about heritage speakers?
• They are like L1 learners• Their noun-gender lexical associations are
stronger than in L2 learners but weaker than in mature native speakers.
Weaker Links Hypothesis (Gollan et al. 2008)
Why are heritage speakers quantitatively different from native speakers?
Noun-gender associations are part of lexical acquisition.
Reduced input and use of the minority language throughout the school-age period leads to reduced frequency of use of nouns and their associated genders.
Links might have been stronger in childhood but progressively weakened as the first language became the secondary language.
Effects of Weaker Links
• Gender assignment errors• Slower retrieval of nouns in the lexicon• Slower insertion of nouns in the syntax• Reduced speed at computing syntactic
dependencies (concord with determiner and noun)
• Gender agreement errors
Differences in heritage speakers
19 HS did not make any gender errors in the Oral Production task
10 HS didRTs in Spanish PNTHS who made errors 1262 msHS who made no errors 1030 ms
232 (t(29)= 8.54, p < 0.0001)
Why does canonicity (noun ending) matter for L2 learners and Heritage speakers?
Dual mechanism model? (Pinker & Prince 1994, Pinker 1999, Ullman 2002)
Canonical nouns are stored in procedural memory and handled by rule (implicitly acquired in childhood by heritage speakers and learned later but automatized in L2 learners).
(fem –a and masc –o word markers are “regular”, other markers are irregular Harris 1991)
Non-canonical nouns need to be memorized.
Native speakers are not affected
• Reduced input in L2 learners and heritage speaker affects storage in declarative memory.
• Mature native speakers whose primary language is Spanish, do not exhibit gaps with declarative memory because they use the language more frequently and the lexical-association links remain strong for both canonical and non-canonical ending nouns.
Prediction
Native speakers undergoing L1 attrition will make gender errors with non-canonical ending nouns.
Montrul (2011): Alicia, Guatemalan adoptee
Accuracy with Gender in Alicia
Spanish native speakers
Alicia
Oral production 100% 84%
Morphology recognition task
100% 82.5%
Picture identification task
100% 71.8%
Grammaticality judgment task
100% 60%
Alicia’s errors
• Most errors were with non-canonical ending nouns, especially on the picture description task: atleta, planeta, mapa, dentista, país, nuez, luz.
• 7 of 9 errors in the morphology recognition task were with non-canonical ending nouns
Another possibility• Single associative model (canonical and
noncanonical nouns)• Lexical links in L2 learners and heritage speakers
are weaker than in native speakers• Morphophonological cues in Spanish help
activate/access canonical nouns faster and more accurately by virtue of phonological regularity.
• Noncanonical nouns do not have cues• Differences in cues affect strength of lexical links.
Conclusion• Age effects may explain important differences
between L2 learners and heritage speakers• But it is also important to evaluate the
contributing role of experience with the language, as revealed by different tasks.
• Experience affects how input is processed more generally and the type and size of vocabulary L2 learners and heritage speakers may possess.
Conclusion
• Heritage speakers benefited from CDS in Spanish and have retained some of these features (i.e., diminutives)
• Many of the perennial problems observed with gender in advanced L2 learners and heritage speakers seem to be related to issues of lexical assignment rather than of actual syntactic agreement.
Conclusion
• When it comes to gender agreement, a grammatical area that is very difficult for L2 learners to master, heritages speakers show remarkable “native” abilities.
• Differences between native speakers and heritage speakers are due to cumulative input and use of the language throughout the lifespan.
Looking ahead
• Heritage speakers can be really advanced speakers of the language as well, not just “incomplete native speakers” .
• A lot remains to be done to understand the individual differences in language proficiency achieved throughout the lifetime by heritage speakers.
Heritage speakers as native speakers
We also need to focus on the higher end of proficiency and stress that many of these abilities develop in childhood.
Thank you very much!