Analysing interlanguage: how do we know what learners know?

ANALYSING
INTERLANGUAGE: HOW DO
WE KNOW WHAT
LEARNERS KNOW?
Usha Lakshmanan and Larry Selinker
1
Presented by: Wanna Phadyen M6020019, Presented to: Dr. Bussakorn

Outline
◦ Purpose of the study
◦ Brief discussion on interlanguage and the comparative
fallacy
◦ The problem involved in the analysis of spontaneous
speech
◦ Comparative fallacy overview
◦ Example of studies
◦ Conclusion
◦ Suggestion
2

Two major goals of second
language acquisition (SLA)
research
◦ 1) to determine the second language learner’s
L2 grammatical knowledge (i.e., interlanguage
competence); and
◦ 2) to explain how it develops over time from an
initial state to an end state and often a
fossilized state (learners do not achieve native-
like competence).
3

Purpose of the study
◦ To examine some of the problems
involved in the analysis of
spontaneous speech, with focus on L2
studies within the generative
framework
4
o Understand how the
comparative fallacy may have a
disastrous effect on
investigations of IL
o Revisit the issue of the
comparative fallacy in L2
research
(a real-time generated,
unplanned and non-
rehearsed type of
encoding linguistic
information)
(regards grammar as a
system of rules)
(the mistake of studying
the systematic character
of one language by
comparing it to another)

Brief discussion
◦Interlanguage
◦The comparative fallacy
5

Interlanguage
◦ An interlanguage is a type of language that has been
developed by a learner of a second language (or L2) which
preserves some features of their first language (or L1), and
can also overgeneralize some L2 writing and speaking rules
(Selinker, 1972).
◦ Interlanguage is a developing system with its interim
structure, rather than an imperfect imitation of the target
language.
6
Input Noticing Intake Interlanguage

Interlanguage Rules
◦ Has rules
◦ Rules are altered
◦ Rules are deleted
◦ Rules are added
◦ Moves towards L2, but may become fossilized
7
L1 IL L2
L1 IL L2

The processes that create
interlanguage
Interlanguage
Generalizations Transfer
Internal
sequences
8

interlanguage
interlanguage
Internal
sequences
9
• Learners group similar things, events
• make rules to predict how different
items will behave
• categorize what they hear and make
rules to those categories
• use those categories and rules in new
situations
• Overgeneralization occurs when they
make mistake because categories
have exceptions and they put
language in the wrong categories

interlanguage
interlanguage
Internal
sequences
10
• Error due to
transfer
• Pronunciation 
salad, earth, 19, 90
• Vocabulary  day
after day
• Speech acts
• Learners use their
knowledge of L1 to
understand and
organize L2
information
• When L1 and L2 are
difference, transfer
can lead learners to
make errors
• When L1 and L2 are
similar, transfer
helps learners

interlanguage
interlanguage
Internal
sequences
11
Learners
make use of
their
knowledge of
the L2
Learners
make use of
their
knowledge of
their L1 to
produce a L2

interlanguage
interlanguage
Internal
sequences
12
• Learners hear different language,
for example, in classrooms, their
environment
• Have different first language
• Expect that learners learn a L2 in
different ways

Research on Interlanguage
◦ Many researchers have found that learners who
receive grammar-based instruction still pass
through the same development sequences and
make the same types of errors of those who acquire
language in natural settings.
◦ Research also shows that L2 learners from different
L1 backgrounds often make the same kinds of
errors when learning the L2.
◦ Interlanguage is the evidence of the learners’ efforts to
discover the structure of the TL itself. Rather than
attempts to transfer patterns from their L1.
13

The comparative fallacy
◦ The comparative fallacy, first termed by Bley-Vroman (1983),
cautioned against the theoretically weak comparison of second
language (L2) learners’ linguistic systems to the target language
norm.
◦ L2 studies that employ analytical approach defined relative to the
target language are faulty in revealing the nature of IL and could
lead to incorrect conclusions.
◦ Reviewing some studies consider the comparative fallacy where the
bias is towards the native language
◦ The comparative fallacy in interlanguage studies can lead to the
underestimation and/or overestimation of the learners’ linguistic
competence
14

The problems involved in the
analysis of spontaneous speech
◦ The speech utterances in the earliest samples of speech
gathered may not have been collected early enough and may not
be reflective of the L2 initial state (starting point for L2
learners).
◦ The corpora (a collection of written text), particularly the
earliest samples gathered, may be limited and unusable.
◦ Language learners have been known to go through a ‘silent
period’, during which they do not produce any utterances in the
L2. As a result, an effective comparison of the development of
individual interlanguage grammars may be difficult to
accomplish.
◦ The researchers choose to include or exclude information for
the purpose of the analysis.
15

Example of research study: clause
structure in the L2 initial state
◦ Lakshmanan (1993/94; 1998) has reported, the
verbless utterances in child L2 grammars are not
realized as binominal expressions. Although the
transitive verb is omitted, the verbless utterances
contain elements such as for, with and and.
16
an expression
containing two
words which are
joined by a
conjunction (i.e.
give or take)
o Verbless utterances have also been reported for child L1
grammars of English (Bloom, 1970). The verbless
utterances shown in child L1 acquisition are typically
binominal expressions such as “Kendall bath” =
Kendall takes a bath.

◦ a. Marta (L1: Spanish):
◦ [ Say hello]  For Hello
◦ [picture of boy eating cookies]  This is the boy for the cookies
◦ [The girl is playing the tambourine]  The girl for tamboron
◦ [in response to: What’s cookie monster going to do?]  For the head the
little girl
◦ b. Cheo (L1: Spanish):
◦ [picture of boy eating cookies]  The boy and the cookie
◦ [picture of boy pouring milk into a glass]  The boy with the milk
◦ c. Xi-Xi (L1: Chinese)
◦ [The boy is kicking the ball]  Boy and ball
◦ [The cat is driving a car]  This cat and car
◦ [The pig is building a house]  And house
17

◦ Lakshmanan (1993/94; 1998) proposed
that there is an implicit verb (unspoken) in
these utterances which heads the VP
projection; as the verb does not have a
proper linguistic antecedent, it cannot
assign accusative case to its object.
18
(an expression
(word, phrase,
clause,
sentence, etc.)
a noun is the
grammatical case
used to mark the
direct object of a
transitive verb
(them, her)
o She argued that elements such as for and
and in these verbless utterances are
functional morphemes generated in
inflectional position.

◦ However, in 2000 she argued that although it
seems that an implicit verb may not present and
a VP may not be projected, under the
alternative analysis, the elements for, and and
within the verbless utterances may be viewed as
‘minimal’ maximal projections of a lexico-
functional category bearing the categorial
features [–N, –V] and the functional feature F.
19
Lexical items
that do not
project

Example of research study: the
phenomenon of root infinitives
◦ L2 learners, like child L1 learners, also produce
root clauses with a finite verb form (main
verbs) as well as root clauses with a non-finite
verb form (gerund, infinitive and participles)
◦ These studies assume that the root infinitives
(default verb form) produced by language
learners are ungrammatical from the
standpoint of the target grammar.
◦ But what about from the standpoint of the
independent interlanguage grammar?
◦ Lasser (1998) systematically compared adult
root infinitives (default verb form) and child root
infinitives. Interestingly, she found that there are
qualitative similarities, both structural and
interpretive, between child root infinitives and
adult root infinitives. 20
root clauses differ
from embedded
clauses in that the
finite verb is in
second position

Revisited - Comparative fallacy
◦ Comparative fallacy is serious obstacle to the investigation of
crucial questions about the nature of learners’ languages.
◦ ‘Obligatory Contexts’(a place in the sentence that requires the
morpheme to make the sentence grammatically correct) has been
widely used in both L2 and L1 acquisition research to compare
the relevant interlanguage utterances with the related speech
utterances in adult native-speaker spoken discourse .
◦ The difficult task is to determine what the language learner
meant to say, although the discourse context in which the
learner’s utterance occurred may provide useful cues for
determining the meaning of the utterance.
◦ So many L2 researchers tended to restrict definition of
obligatory context to linguistic context without analysing the
discourse context in which the sentence occurred.
21

◦ Furthermore, the identification of obligatory contexts for the
suppliance of grammatical morphemes has been carried out
from the standpoint of the target grammar.
◦ Thus, setting up obligatory contexts solely in terms of the
target language is likely to lead to an underestimation of the
knowledge of the learner.
◦ A related problem concerns the criterion level adopted to
determine whether a particular functional morpheme has been
acquired, which is 90% level (Brown’s, 1973).
◦ Hyams and Safir (1991) have stated in their review of
Lakshmanan (1991), the failure to reach high criterion levels for
acquisition (e.g.,90% criterion level) probably stems from lack
of mastery and not lack of knowledge.
22

◦ This suggests that instead of criterion levels for
acquisition, alternative diagnostics may need to be
used for determining learner’s competence.
◦ The following diagnostics are better indicators of
a learner’s knowledge of a particular grammatical
element/process:
◦ • overregularizations: e.g., *goed;
◦ • the absence of form/class errors: e.g., *mays; and
◦ • the absence of agreement errors: e.g., *I goes.
23

Example: How obligatory contexts –
underestimated of learners’ knowledge
◦ Lardiere (1998) - Tense and Case in the fossilized steady state
of Patty,
◦ Patty - an Indonesian-born, Mandarin speaking and Hokkein-
speaking adult non-native speaker of English living in the
USA.
◦ Lardiere collected data over a period of eight years.
◦ • 1st - after Patty had been exposed to English in the USA
for 10 years.
◦ • 2nd and 3rd (two months apart) - 8.5 years later
◦ • Formal interview about her language background and
life history, narratives of past events in her life and discussion
of her philosophical and religious views.
24

◦ To determine Patty’s suppliance rates for past tense marking, each
verb in a finite past context was checked carefully.
◦ Certain obligatory finite past contexts in which past tense marking was
not supplied were eliminated from the count.
◦ Lardiere provides detailed information about the contexts that were
excluded, some examples of these are as follows:
◦ a past situation context where the situation still holds true in the present
and therefore a present tense temporal reference is equally possible; e.g.,
She’s maybe ten years older;
◦ formulaic utterances (like... Err.. Umm.., idioms and proverbs) ;
◦ ambiguously finite contexts could be interpreted as either quotations or
reported speech;
◦ contexts in which past tense inflection would be adjacent to homophonic
stops; (e.g., We exchanged diary. I stopped talking.)
◦ utterances followed immediately by spontaneous self-corrections, etc. 25

◦ Lardiere’s analytical method was to count past tense
marking as supplied if the main verb or the auxiliary was
modified. Importantly, past tense was considered supplied
even if agreement marking was incorrect (e.g., was instead
of were) or if the accompanying elements of the VP were
incorrectly inflected (e.g., did went instead of did go).
◦ The results shows the low suppliance rate in relation to
past tense marking was 34%.
◦ subjects pronoun in finite contexts was perfect 100%
◦ the distribution of nominative and accusative pronouns
in Patty’s data was also found to match the target grammar
perfectly. 26

◦ Based on her findings, Lardiere argues that the relationship
between the acquisition of overt inflectional morphology and
the development of the syntactic representation of
functional categories, projections and features is not a direct
one.
◦ Patty’s low rate of suppliance of past tense marking does not
constitute evidence for the absence of the functional category of
Inflection INFL, its projection and the abstract
TENSE/Finiteness feature associated with it.
◦ Patty’s low percentage of suppliance of past tense marking due
to Lardiere’s difficulties in mapping from the abstract
functional feature TENSE to the Phonetic Form.
◦ Unlike pronominal case marking, which is simple and invariant
in English, the mapping from feature to form in relation to tense
morphology is complex and variable.
27

◦ The researcher claimed that Patty’s suppliance
of past tense marking may in fact have been
underestimated since the setting up of the
obligatory finite past contexts was done in
terms of the target language and not in terms
of Patty’s interlanguage grammar.
28

Comparative fallacy extended
◦ In order to characterize the
language learner’s linguistic
competence in the L2
accurately.
29
Interlangu
age must
be
analysed
in its own
term
Interlanguage -
Occurs
Native language -
blocked
o The native language
influences SLA, the exact
nature of its influence still
remains “unclear”o When investigating L1 transfer
using longitudinal spontaneous
speech data.
Focus on the
comparative
fallacyo It results from a bias towards
the native language.

The case of null subjects from
the Spanish L1
◦ Hilles (1991) examined the longitudinal spontaneous
speech data from 6 Spanish speakers acquiring English as
L2
◦ Let’s focus on Hilles’ method for calculating the rate of
omission of subjects from tensed clauses.
◦ In order to calculate the percentage of null subjects for
each learner, Hilles used the following method:
30
the number
of actual
instances of
null subjects
the number of
instances
where null
subjects could
have occurred
but did not

Example: null subjects from the
Spanish L1
◦ Hilles assumed that in all cases the Spanish speakers would
transfer their discourse rules concerning the use of null
subjects from their Spanish L1 to their English interlanguage.
◦ An outcome of Hilles’ method is that the value the sum of Y + X
is lower than would have been obtained, if the Y cases had
been included all clausal utterances, it would be clearly
ungrammatical in the target language to omit the subject.
◦ If this alternative method had been used to calculate the Y
cases, it would have increased the value of the sum of X + Y,
and the percentage of null subjects would have been lower
than what Hilles found.
31

Example: null subjects from the
Spanish L1
◦ Hilles’ method for the calculation of null subjects is biased in
favour of her hypothesis that Spanish speakers will transfer the
property of null subjects to their English interlanguage.
◦ Lakshmanan (1994) used the alternative method (slide 23) in
order to calculate the percentage of null subjects in the English
interlanguage of child L2 learners of English.
◦ Two of the learners (Marta and Cheo) were the same
individuals considered in Hilles’ study.
◦ The percentage of null subjects reported by Lakshmanan for
these two subjects was considerably lower than what was
reported by Hilles. It is not clear how one can resolve this
difference.
32

Example: the use of head final VPs in
the English interlanguage of Erdem
◦ Haznedar (1997) reported the use of head final VPs in the English
interlanguage of Erdem, a four-year-old native speaker of Turkish,
who was acquiring English in the UK.
◦ Erdem was observed approximately three times a month over a period
of 18 months.
◦ Erdem had been exposed regularly to English for approximately one-
and-a-half months in nursery school.
◦ Prior to attending nursery school, Erdem was always at home with his
Turkish-speaking parents.
◦ Haznedar does not state whether Erdem’s parents knew English, but
assuming that they probably knew some English.
◦ These data are important in that they are original data, which were
collected by Haznedar who is bilingual in Turkish and English. 33

The case the use of head final VPs in
◦ Haznedar classified Erdem’s utterances that contained a verb as either
XV or VX, where X represents ‘other VP material (e.g. a direct object or
an adverbial)’
◦ The examples of the XV utterances:
◦ a. Yes # # ball playing. - first 2 months of data collection
◦ b. Would you like to outside ball playing
◦ c. I something eating. – first 2 months of data collection
◦ d. Yes, toys play. [in response to: Shall we play with your toys?]
◦ e. Newcastle going [in response to: Where are we going now?]
◦ f. fast push [context: on the swing]
34

◦ The researchers believe that Haznedar’s conclusion in relation to
L1 transfer is problematic for the several reasons.
◦ • First, during the initial period when XV utterances occur,
the overall number of verbal utterances produced by Erdem are
scarce (only 23 verbal utterances in all).
◦ • Secondly, Haznedar categorized the verbal utterances only
into two types, XV or VX, where X represents VP material such as
direct object or adverbial.
◦ The XV utterances should have been further categorized to
separate those cases where the X element was clearly a direct
object (i.e., a thematic argument) of a transitive verb from other
cases where the X element was. 35

◦ • third problem is ambiguous data, which are open to alternative
analyses.
◦ ‘ball playing’ could be a nominal,
◦ ‘toys play’, ‘fast push’ and ‘Newcastle going’ may be topicalized
constructions where the X element is topicalized.
◦ ‘I something eating’ is a clearer example of XV order.
◦ An additional problem is the nature of the speech situation when data
were collected. (Erdem’s mother was also present when Erdem was
observed by Haznedar). codeswitching/ code-mixing, the data may be
‘contaminated’
◦ The point is that if the theory predicts that syntactic properties
transfer from the L1, then making decision of what to include in the
analysis will have to be made against the theory being tested and not in
the favour of it.
36

Conclusion
◦ The researchers investigated the role of learner
spontaneous speech in the questions raised in the
introduction to this issue,
◦ refining the general questions to the specific one of
how we as researchers know what learners know
when we use this type of data
◦ In dealing with the detail of the comparative
fallacy – both as originally proposed and in its
suggested extended state
◦ Having delineated dangers of biases of several
types
38

Suggestion
◦ One suggestion was proposed by Corder (1981) that
◦ The use of ‘bilingual researchers’ has not been seriously exploited
(make full use of and derive benefit from) in the field; we intend the
word ‘bilingual’ in a special sense: knowing both the advanced
interlanguage and more earlier stages of that interlanguage
◦ Corder claimed, that in terms of understanding interlanguage
competence, such bilingual researchers might be those who are
closest to the native speaker in the classical Chomsky sense.
◦ Such researchers would know several relevant linguistic systems:
◦ * the native language of the learner,
◦ * a very advanced state of the interlanguage,
◦ * earlier interlanguage systems and the target language itself,
◦ * at least in a declarative sense.
39

Analysing interlanguage: how do we know what learners know?

More Related Content

What's hot (20)

Similar to Analysing interlanguage: how do we know what learners know? (20)

More from Pun Yanut (6)

Recently uploaded (20)

Analysing interlanguage: how do we know what learners know?

Editor's Notes