Lexical Knowledge and Access in Spanish Heritage Speakers Silvina Montrul

1
Lexical Knowledge and Access in Spanish Heritage
SpeakersSilvina Montrul
3rd Heritage Language Summer Institute Urbana, IL
June 22-26, 2009
2
Research projects at UIUC

• Bilingual Past project in depth-comparison of
  proficiency-matched L2 learners and heritage
  language speakers
• Focus on form and reactivity to instruction in
  heritage language learners and heritage speakers
  (with Melissa Bowles)
• Comparative heritage languages Study of Spanish,
  Hindi and Romanian
• The role of the contact language in heritage
  language grammars (Ji-Hyes dissertation)

3
Acknowledgements

• University of Illinois Campus Research Board
  (Beckman Award to Silvina Montrul, Spring 2005)
• The Center for Advanced Study (UIUC)
• Research assistants and collaborators
• Rebecca Foote Alyssa Martoccio
• Silvia Perpiñán Lucia Alzaga
• Dan Thornhill Ben McMurry
• Susana Vidal Brad Dennison

4
Heritage Language Grammars

• Distinctive gaps in heritage speakers
  grammatical knowledge (Montrul, 2008 OGrady et
  al., 2001 Polinsky 2007 Rothman 2007).
• We know much less about lexical knowledge in
  heritage language grammars.
• What variables characterize heritage language
  speakers knowledge, retention and loss of words?

5
Theoretical significance

• Is there a relationship between lexicon and
  grammar (Bates et al., 1994, Thal et al. 1997,
  Polinsky 2005)?
• Polinsky (1997, 2007) found that lexical
  knowledge was correlated with grammatical
  knowledge in Russian heritage speakers.

6
Pedagogical Significance

• If there is a relationship between vocabulary
  size and grammatical knowledge, then vocabulary
  tests can be used as proficiency measures.
• Lexical decision tasks have been implemented as
  placement tests for language classes
• ESL (Meara Jones 1987, 1988)
• L2 acquisition of Spanish (Lam et al., 2003)
• Spanish heritage speakers (Fairclough, 2008)

7
Objective

• Discuss some results from a large-scale
  experimental study of Spanish L2 learners and
  Spanish heritage speakers.
• Several written and oral tasks testing knowledge
  of gender, cilices, tense, aspect, and mood.
• An on-line lexical decision task
• An on-line translation judgment task

8
(No Transcript)
9
Participants

• Baseline or control group
• 22 native speakers
• Experimental Groups
• 72 L2 learners of Spanish
• 69 Spanish heritage speakers
• All participants completed a language background
  questionnaire (6-page long for the heritage
  speakers)

10
L2 learners

• Age 21.91 (18-25)
• Native speakers of English
• Raised in English-speaking families
• Age of first exposure/acquisition of Spanish as a
  second language between the ages of 12-25 (high
  school, college)
• Enrolled in Spanish language classes at the
  University of Illinois at Urbana-Champaign
• Advanced speakers were graduate students and
  Spanish language instructors with very high (some
  near-native) command of Spanish.

11
Heritage speakers

• Age 22.64 (18-30)
• Born in the US to Mexican parents
• Exposed to English before age 5
• At least one of the parents had to be a first
  generation immigrant
• Schooled in the US
• Graduate and undergraduate students at the same
  university, some of them enrolled in the same
  classes as the L2 learners
• Some advanced speakers were graduate students and
  Spanish teaching assistants

12
Heritage Speakers Some Descriptive Stats

• First language Spanish (57), English (35),
  both (8)
• Parents both parents from Mexico (88)
• one parent from Mexico
  (12)
• Language used at home
• Only Spanish (44), Spanish and English (56)
• Languages parents spoke to participants
• Spanish (80), English (5), both (15)
• All participants had between 1-9 siblings and 20
  lived with a Spanish-speaking grandparent
• Language spoken with siblings
• Spanish (20), English (48), both (38)
• Relative strength of the languages
• 48 felt Spanish was like a native language, 52
  like a second language
• Self rated proficiency mean Spanish (3.9, range
  1-5)
• mean English (4.88, range 4-5).
• 100 wanted to improve their ability in Spanish
  for both professional and personal reasons

13
Spanish Proficiency Test

• Cloze part (fill in the blanks by selecting one
  of four possible responses) (DELE test) 30
  points
• Multiple choice vocabulary test (MLA) 20 points
• Maximum 50 points
• Has been widely used in many L2 acquisition
  studies

14
Proficiency Scores
Mean 48.5 SD 1.00 range 45-50
Mean 36.88 SD 8.17 range 15-48
Mean 35.34 SD 9.24 range 16-50
15
Participants
16
Research questions

• Polinsky (2005) found that low proficiency
  Russian heritage speakers had selective control
  of word classes, retaining verbs better than
  nouns and adjectives. Do Spanish heritage
  speakers also have selective control of verbs?
• Does Age of Acquisition (AoA) of word interact
  with age of acquisition of the target language in
  L2 learners and heritage language speakers?

17
Research questions

• Does accuracy in a lexical decision task
  correlate with accuracy on a written proficiency
  and other written measures of grammatical
  competence? Is the lexical decision task a
  reliable placement tool for both L2 learners and
  heritage language learners?

18
Questions in Bilingual lexical processing

• What is the relationship between words and
  concepts in the bilingual lexicon?
• What is the architecture (organization) of the
  bilingual lexicon?
• What factors influence lexical access and speed
  during lexical processing by bilingual
  individuals?

19
Some factors that influence speed of lexical
processing

• Word frequency
• Phonology
• Morphological complexity
• Syntactic category
• Semantic priming
• Lexical ambiguity
• Imageability
• Age of acquisition

20
Lexical Access

• In L1 attrition, lexical access is assumed to be
  the aspect of language most susceptible to
  language loss (de Bot 1996, 1998 Weltens
  Grendell 1993).
• Speakers encounter lexical retrieval difficulties
  in the L1 due to low level of activation and
  reduced proficiency.

21
Frequency Effects in L1 Attrition

• Hulsen (2000)
• Study of lexical access in 3 generations of Dutch
  immigrants to Australia
• Picture naming and picture matching tasks
  (production)
• More frequent and cognate nouns are retained and
  accessed faster than less frequent nouns in 1st,
  2nd and 3rd generation speakers.
• Accuracy decreases and reaction times increases
  by generation 1gt2 gt3.

22
Lexical Categories or Grammatical Class

• Distinction between NOUN, VERB, ADJECTIVE, etc.
• NOUNS are referential, VERBS are relational,
  ADJECTIVES are neither (Baker 2004)
• NOUN-VERB distinction figures prominently in
  normal L1 acquisition In many languages,
  including Spanish and English, NOUNS are acquired
  before VERBS and ADJECTIVES (Clark 1993). In
  Chinese and Korean, VERBS are acquired before
  NOUNS (Choi 1998, Choi Gopnik 1995)
• NOUNS and VERBS are selectively impaired in
  aphasia (Shapiro Caramazza 2002)

23
Polinsky (2005)

• Do incomplete learners of Russian (i.e., Russian
  heritage speakers) differ from complete speakers
  in their access to words in Russian?
• In L1 acquisition of Russian, NOUNS are acquired
  before VERBS. Is lexical access selective by
  lexical class in incomplete acquisition of
  Russian?

24
Polinsky (2005)

• Study of 5 incomplete learners of Russian and 4
  Russian native speakers
• Stimuli VERBS, NOUNS and ADJECTIVES of low, mid
  and high frequency ranges (11 items per frequency
  range for each class)
• Cognates and latinate words were avoided
• Experiment 1 lexical recognition task
• Experiment 2 translation task

25
Polinskys Findings

• Different control of word classes in the heritage
  speakers.
• Native speakers had balanced control of NOUNS,
  VERBS and ADJECTIVES
• Incomplete learners had faster reaction times in
  Experiment 1 and higher accuracy in Experiment 2
  for VERBS.
• Primacy for VERBS in lexical retrieval
• Explanation semantic density of VERBS

26
Vocabulary Recognition
27
Translation
28
Motivations for our Study

• Examine the effects of lexical class and AoA (and
  of frequency) in Heritage speakers (a case of
  incomplete L1 acquisition) and late L2 learners
  of Spanish.
• Practical implications Many colleges and
  universities in the United States are developing
  lexical decision proficiency tests in order to
  place L2 learners and Heritage speakers into
  different proficiency levels in language
  programs.
• Assumption size of vocabulary correlates with
  grammatical development.

29
Method

• Experiment 1 Visual lexical decision task
• Experiment 2 Visual translation judgment task
• Dependent variables accuracy and reaction times

30
Stimuli in each experiment

• 108 Spanish words
• (non-cognate)

Matched for frequency
31

• Words were matched for frequency and syllable
  length across the three lexical classes
• Only 1/3 of words appeared in the two experiments
• AoA was decided by consulting the Spanish version
  of the MacArthur Communicative Development
  Inventory (CDI) for L1 acquisition and first year
  Spanish textbooks for L2 acquisition
• 108 filler items (36 nouns, 36 verbs, 36
  adjectives)
• Equal number of Non-words in each experiment

32
Lexical Decision Task

• Subjects saw a string of letters in the center of
  a computer screen and had to indicate whether the
  string of letters formed a real word of Spanish
  or not.
• pañal coler
• SI NO SI
  NO

33
Translation Judgment Task

• Spanish words were presented on a computer screen
  followed by an English word. Subjects were asked
  to decide as fast as possible whether the English
  word was an accurate translation of the Spanish
  word, by pressing YES or NO keys.
• Pañal Diaper
• SI NO

34
Research Question 1

• Do Spanish heritage speakers also have selective
  control of verbs?

35
Lexical Decision Task

• 60 heritage speakers
• 20 native speakers
• Heritage Speakers Proficiency scores
• 29 advanced
• 21 intermediate
• 10 low

36
Results
Main effect for
frequency F(1,19) 8.76, p 0.008
37
Main effect for word class, for frequency and
word class X frequency interaction
38
Results
Main effect for
frequency F(1,19) 19.75, p lt 0.01
39
Main effect for
frequency F(1,59) 72.1, p lt 0.01
40
Results

• Both native speakers and heritage speakers were
  more accurate with and responded faster to high
  frequency than with low frequency words.
• There was an effect of word class in the accuracy
  analysis only for the heritage speakers, as well
  as a class by frequency interaction.
• Heritage speakers were more accurate on high and
  low frequency nouns and least accurate with low
  frequency verbs.
• No effect of word class in reaction times.

41
Heritage speakers Proficiency analysis

• Main effect by level in both accuracy and RT.
• Main effect for frequency in accuracy and RT
• No effect for word class

42
Primacy for verbs in accuracy for low and
intermediate groups, but ns
43
No effect for Word class in RTs
44
Heritage speakers Proficiency analysis
45
Heritage speakers Proficiency analysis
46
Lexical Decision Summary of results

• Main effect for frequency
• Main effect for proficiency level
• Advantage for nouns in Accuracy
• Slowest on verbs and adjectives in reaction times

47
Results Translation Judgment Task
Main effect for word class, for frequency and
word class X frequency interaction
48
Results Translation Judgment Task
Main effect for word class and for frequency
49
Summary

• If word class advantage, it is a noun advantage
• In the accuracy analysis, the heritage speakers
  were more accurate on high and low frequency
  nouns.
• In the speed analysis, heritage speakers were
  faster with high and low frequency nouns.
• They are slowest and least accurate with low
  frequency verbs and adjectives

50
Translation Judgment Proficiency Analysis
51
Translation Judgment Proficiency Analysis
52
Primacy of Nouns for Adv. and Interm.
53
Slower on Verbs than on Nouns and Adjectives
54
Translation Judgment Summary of Results

• Accuracy Advantage for Nouns
• RT slower on verbs
• Proficiency effect

55
Conclusion

• Advantage for verbs may be possible at lowest
  levels of proficiency.
• In general, this study found an advantage for
  high frequency words and NOUNS.
• Difference between Spanish and Russian verbs and
  nouns?

56
Research Question 2

• Does Age of Acquisition (AoA) of word interact
  with age of acquisition of the target language in
  L2 learners and heritage language speakers?

57
Age of Acquisition effects in monolingual lexical
processing

• Bonin, Barry, Méot Calard (2004)
• Age of acquisition (AoA) in these studies refers
  to the age at which words are first learned in
  their spoken and written form.
• AoA effect words acquired early in life are
  processed faster and more accurately than those
  acquired later.

58
Locus of AoA effects

• Phonological
• Semantic
• Process of mapping between different lexical
  representations (orthographic, semantic,
  phonological)

59
Bilingual processing

• AoA also affects lexical processing in a second
  language
• Izura Ellis (2004) found that the lexical
  decision speed of words in Spanish (the L1 of the
  Spanish-English bilinguals tested) was predicted
  by AoA of words in Spanish.
• In a translation judgment task, speed in English
  (the L2) was predicted by AoA of English words.
• Support for Mapping Hypothesis (Ellis
  Lambon-Ralph, 2000)

60
Izura Ellis (2004)

• Words used were 80 NOUNS, the rest were
  ADJECTIVES and VERBS
• Age of acquisition of bilinguals ranged from 6-24
• AoA of words was decided by asking subjects when
  they thought they had acquired words

61
Age of Acquisition in Second Language Acquisition
and Bilingualism

• Age of Acquisition age at which the L1 and the
  L2 of bilinguals/second language learners were
  acquired.
• Early bilinguals L1 and L2 acquisition takes
  place simultaneously or sequentially, before the
  onset of puberty (within the Critical Period)
• Late bilinguals L2 acquisition takes place after
  puberty, after the foundations of the L1 are in
  place.

62
AoA is correlated with bilingual outcomes or
ultimate attainment

• L2 Acquisition
• The earlier the AoA of an L2 the more native-like
  the bilingual is likely to become in the L2,
  especially in phonology and morphosyntax.
• L1 attrition and/or incomplete acquisition
• The earlier the AoA of an L2 the less nativelike
  the bilingual is likely to become in the L1,
  especially in phonology and morphosyntax.

63
Method

• Same tasks
• Subset of heritage speakers and L2 learners
  matched for proficiency (n 28 per group)

64
Words and main variable manipulated
within group variables
Between group variables
65
Example words for each word AoA category
66
Hypotheses between groups

• The heritage speakers may show an advantage over
  the L2 learners in both accuracy and reaction
  times with Spanish words acquired early in L1
  acquisition, but late in L2 acquisition (Early
  L1-Late L2) and with words acquired early in L1
  acquisition and early in L2 acquisition (Early
  L1-Early L2), since the heritage speakers will
  have acquired all of these early words at a much
  younger age than the late L2 learners.
• The heritage speakers may also show an advantage
  over the late L2 learners with words acquired
  late in L1 acquisition but early in L2
  acquisition (Late L1-Early L2), or the two groups
  may show similar results, depending on when each
  group acquired these words. 

67
Hypotheses Within groups

• The heritage speakers will be faster and more
  accurate in their responses to Early L1-Late L2
  and Early L1-Early L2 words in comparison to Late
  L1-Early L2 words.
• The L2 learners will pattern in the opposite
  direction, with an accuracy and reaction time
  advantage for the Late L1-Early L2 and the Early
  L1-Early L2 words over the Early L1-Late L2 words.

68
Results Lexical Decision Task
69
Summary of Results Accuracy

• Main effect for word AoA and word AoA by group
  interaction
• Overall accuracy rates in the two participant
  groups were similar (no main effect for group).
• The heritage speakers showed the predicted
  accuracy advantage over the L2 learners for Early
  L1-Late L2 words.
• However, they were not more accurate than L2
  learners in either of the other two word AoA
  categories.
• The L2 learners were less accurate with Early
  L1-Late L2 words (mean 91) than with Late
  L1-Early L2 words (mean 96)

70
Results Lexical Decision Task Accuracy
71
Summary of Results Reaction times

• Main effect for AoA and AoA by group interaction.
• The L2 learners were faster than the heritage
  speakers in all three word AoA conditions (but
  non-significant).
• The predicted speed advantage for the heritage
  speakers over the L2 learners was not borne out.

72
Results Lexical Decision Task RT
73

• Within groups, the heritage speakers were fastest
  to respond to Early L1-Early L2 words.
• They were next fastest to respond to Early
  L1-Late L2 words and slowest to respond to Late
  L1-Early L2 words, as predicted.
• The L2 learners were also fastest to respond to
  Early L1-Early L2 words, but patterned
  differently with the other word AoA categories,
  responding next fastest to Late L1-Early L2
  words, and slowest to Early L1-Late L2 words, as
  predicted.

74
Translation Judgment Task
75
Results Translation Judgment Task
76
Summary of Results

• The heritage speakers showed an accuracy
  advantage over the L2 learners for the Early
  L1-Late L2 words, but the L2 learners were faster
  than the heritage speakers in all three word AoA
  conditions (although non-significant).

77
Within-group differences results patterned as
expected

• heritage speakers showed a speed advantage for
  both Early L1-Early L2 and Early L1-Late L2
  translation pairs over Late L1-Early L2
  translation pairs
• L2 learners showed an accuracy advantage for
  Early L1-Early L2 pairs and Late L1-Early L2 word
  pairs over Early L1-Late L2 pairs, and a speed
  advantage for Early L1-Early L2 pairs over Early
  L1-Late L2 pairs.

78

• However, as in Experiment 1, heritage speakers
  showed a speed and accuracy advantage for Early
  L1-Early L2 word pairs over Early L1-Late L2 word
  pairs.
• This is in contrast to our predictions, based on
  the assumption that both of these word AoA
  categories were acquired early in Spanish and in
  English for these participants.

79
Results Translation Judgment Task Accuracy
80
Results Translation Judgment Task RT
81
Conclusion

• Predictions based on AoA of L1 and L2 were
  generally not borne out in the results.
• No age effects of language in the acquisition of
  words, unlike morphosyntax and phonology.
• Word AoA was found to be significant within each
  group in the two experiments, confirming previous
  findings in the psycholinguistics literature
  (Carroll White 1973a,b Ellis Morrison, 1998
  Ellis Lambon Ralph, 2000 Izura Ellis 2002,
  2004).

82
Conclusion

• Age of language acquisition does not confer an
  overall speed advantage in lexical access, though
  it may confer an accuracy advantage, at least for
  words that are learned later in the course of L2
  acquisition, but early in L1 acquisition.
• This supports the idea that there is no critical
  period for the acquisition of lexical items,
  though it must be kept in mind that we only
  investigated lexical access in a visual
  comprehension task.

83
Research questions 3

• Does accuracy in a lexical decision task
  correlate with accuracy on a written proficiency
  and other written measures of grammatical
  competence?
• Is the lexical decision task a reliable placement
  tool for both L2 learners and heritage language
  learners?

84
Vocabulary and Proficiency in L2 acquisition

• Meara Jones (1988) and Meara Buxton (1987)
  found correlations for L2 learners of English
  between a lexical decision task and the Cambridge
  Proficiency Exam.
• Lam, Pérez-Leroux, Ramírez (2003) found a
  correlation between knowledge of vocabulary and a
  proficiency test in Spanish L2 (Canadian
  University)
• Assumption more words, more exposure, better
  proficiency skills

85
Vocabulary and syntactic development in heritage
grammars

• Polinsky (2007) found that vocabulary proficiency
  correlated positively with structural accuracy in
  Russian heritage speakers
• Those speakers who knew more basic words from a
  list of 200 items exhibited better control of
  agreement, case markers, and subordination in
  spontaneous speech.

86
Fairclough (2008)

• Investigated whether the lexical decision task
  was a good tool for language placement for both
  L2 learners of Spanish and Spanish heritage
  speakers.
• She found high positive correlations (r above .7)
  between a cloze test and accuracy on a lexical
  decision task in both groups.

87
Faircloughs results
88
Fairclough (2008)
89
Our Study

• We had a total of 108 words selected from Léxico
  Informatizado del Español (LEXEP, 2000).
• But our results are very similar to those
  reported by Fairclough (2008).

90
Heritage Speakers Proficiency Test
91
Heritage Speakers Lexical Decision Task
92
Heritage speakers Correlation
Significant positive correlation between two
scores r 0.647 p lt 0.001
93
L2 learners Proficiency Test
94
L2 learners Lexical Decision Task
95
L2 learners
Significant positive correlation between two
scores r 0.678 p lt 0.001
96
Conclusion

• Although we have not used our results as a
  placement measure, they do confirm that accuracy
  in lexical access and overall grammatical
  proficiency are correlated in the two
  populations.
• Many people have raised concerns about our use of
  a written proficiency measure originally
  developed for L2 students with heritage language
  students.
• Our research shows that the proficiency test we
  use is not only reliable (Cronbach alpha above
  .80) but also suitable for heritage language
  learners.
• This does not mean that the ACTFL OPI (A measure
  of oral proficiency as demonstrated by Valdés
  1997) will have the same results with the two
  groups.

97
BIG QUESTION

• Theoretical relationship between grammar and the
  lexicon.
• Nativist position word learning and grammatical
  development proceed in a different way and at
  different pace because grammar and the lexicon
  are separate.
• Emergentism There is no separate grammar. It is
  learned with the same general learning principles
  as the lexicon.

98
Minimalism

• Lexical, grammatical and abstract features all
  form part of the lexicon.
• Functional categories (grammatical words) are
  part of the lexicon, so the fact that we find a
  correlation between grammatical functors and
  grammar is not surprising at all (we are
  correlating the same thing) (Bates Goodman,
  1997).

99

• What is interesting is that we are finding a
  correlation between knowledge of content words
  (nouns, verbs and adjectives) and GRAMMAR, which
  are assumed to be learned very differently and to
  be handled by different mechanisms.
• No one has proposed that grammar can begin in the
  absence of lexicon.
• Computational mechanisms for grammar must be in
  place in order for the grammar to use the lexicon

100
Muchas gracias