The Weak Version
As a reaction to the criticism of the strong version of the CAH,
Wardhaugh offered the ‘weak version’.
The weak version does not imply the prior prediction of certain fine
degrees of difficulty. It recognizes the significance of interference
across languages, the fact that such interference does exist and can
explain difficulties, but it also recognizes that linguistic difficulties
can be more profitably explained after they have been noticed.
(Brown 1980: 157).
Thus it has rather explanatory power, helping the teachers of foreign
languages understand their students’ sources of errors.
The starting point in the contrast is provided by actual evidence from
... learning difficulties ... and reference is made to the two systems
only in order to explain actually observed interference phenomenon.
i) Positive transfer - where features of the L1 and the L2 match,
acquisition of the L2 is facilitated.
ii) Negative transfer (L1 interference) - acquisition hindered
where L1 and L2 differ.
• A mismatch in L1 and L2 surface structure features will not necessarily cause
interference in learning (e.g. basic word order, morphological structure) i.e.
not all differences cause problems.
• Some errors occur in L2 learners regardless of L1 background (e.g. causative
constructions: I make him to leave) i.e. not all difficulties result from language
Empirical studies have shown that foreign language learners made numerous
mistakes that were not at all predicted by contrastive studies. On the other hand,
mistakes that were predicted were hardly ever made by learners. This applies, in
particular, to grammar, but also – to a lesser extent – to phonetics and
phonology. Furthermore, only about 50% of all mistakes are due to interference,
which shows that there is a variety of factors which are responsible for learning
The Moderate Version
In the 1970s, Oller and Ziahosseiny proposed a compromise between
the two versions of the CAH and called it a ‘moderate version’. Their
theory was based on their research of spelling errors in learners of
English as L2 which showed that spelling errors were more common
among those learners who used a Roman script in their native language
(e.g. Spanish or French) than among those who used a non-Roman
script (e.g. Arabic or Chinese). However, the strong version of the
CAH would predict the contrary, i.e. more difficulties on the part of the
learners who had to acquire a new writing system (Brown 1980).
Brown (1980: 159) concludes that interference is more likely to occur
when there is similarity between the items to be learned and already
known items than in the case of learning items which are entirely new
to the learner. He also points to the fact that most of the errors
committed by L2 learners are ‘intralingual’ errors, i.e. errors which
result from L2 itself and not from L1.
Whitman and Jackson carried out a study in which predictions made
in four separate contrastive analyses by different linguists were used
to design a test of English grammar which was given to 2,500
Japanese learners of English as L2. After comparing the results of the
test to the predictions based on the four contrastive analyses,
Whitman and Jackson found out that they differed a lot. They came
to the conclusion that “contrastive analysis, as represented by the
four analyses tested in this project, is inadequate, theoretically and
practically, to predict the interference problems of a language
learner” (Whitman and Jackson 1972 cited in Brown 1980: 158).
Besides the problem of inappropriate predictions, Towel and Hawkins
(1994: 18-19) state two other problems. One of them is that “not all
areas of similarity between an L1 and an L2 lead to immediate
positive transfer” (1994: 19). Towel and Hawkins support this
argument by the findings of Odlin’s study in which L1 Spanish
learners of L2 English omitted the copula ‘be’ at the early stages of
learning regardless the fact that Spanish also has a copula verb
adequate to English ‘be’ and thus the positive transfer was possible.
However, it didn’t happen. The other problem, they argue, is that only
a small number of errors committed by L2 learners could be
unambiguously attributed to transfer from L1.
Thus, the strong version of the CAH has been proved inadequate,
except for the phonological component of language, where it is quite
successful in predicting the interference between the L1 and L2 in
pronunciation in the early stages of L2 acquisition.
The weak version is not satisfactory because it is only able to offer an
explanation for certain errors. The only version which remains
acceptable is the moderate version. However, its findings as presented
by Oller and Ziahosseiny are in contradiction with Lado’s original
This doesn’t mean that the idea of L1 interference was completely
rejected, but the CAH is applicable in practice only as a part of Error
Linguistic Levels of Analysis
Contrastive analysis includes all fields of linguistics such as
phonology, semantics, syntax, morphology and pragmatics. It even
seems that contrastive studies should rather be regarded as an
approach, not as a branch of general linguistics.
Linguistics have differed in the degree of separateness/integration of
the mentioned levels. For instance, while Chomsky once argued that
grammar is autonomous and independent of semantics (1957),
another tradition initiated by Firth claims that there is no boundary
between lexis and grammar. He claims that lexis and grammar are in
The emphasis given to various linguistic levels has not been the
same in different linguistic theories. For instance, while the main
focus in Generative Transformational Grammar is syntax, the
Communicative Theory is more concerned with the pragmatic uses
Procedures for Comparing Languages
Contrastive analysis is based on the assumption that languages can be
compared and contrasted. This comparison is done by linguistics to render
descriptive accounts of the learner’s NL and the TL on various linguistic
levels (phonology, morphology, syntax, lexis, and pragmatics).
Doing a CA involves two steps: description & comparison, (James, 1980)
5 different steps have been mentioned in the literature for CA ( comparing &
contrasting two languages):
Comprehensive comparison of two languages for pedagogical
purposes is neither feasible nor necessary.
Certain areas of difficulty in TL are selected based on analyst’s prior teaching
experience & bilingual intuition or prior analysis of errors of the learners ( same
Compared and contrasted with parallel features in the learner’s NL
After the selection of certain linguistic items, structures or rules, the
two languages are explicitly described in question.
Scientific parallel description of the two languages has always been the core of
The two languages should be described through the same linguistic model or
framework. For example, if certain aspects of the grammar of L1 are described
through Generative-Transformational Grammar, the same model for the
description of L2 should be applied as well.
The subsystems of the two languages are juxtaposed in order to find
similarities and differences between them.
Linguistic features of the two languages are compared on three
levels: form, meaning, and distribution of forms.
Predictions are made about difficulties learners may come across in
acquiring the L2.
Similarities and differences found through the comparison of the two
languages should be judged to see if they are problematic for the
learners or not.
Predictions are made through the formulation of a hierarchy of
difficulty (section 1.8).
Finding out whether the predictions made about errors and
difficulties actually materialize or not.
Hierarchy of Difficulty
Several attempts were made to formalize the prediction stage of
contrastive analysis to avoid the subjectivity involved in CA. One of
the best known was a hierarchy of difficulty (Stockwell, Bowen,
and Martin 1965) by means of which it was possible to make a
prediction of the relative difficulty of a given aspect of the second
language. The first such hierarchy was devised for English and
Spanish, but it was claimed to be universally applicable.
Stockwell and his associates suggested 8 possible degrees of
difficulty for phonological systems.
These degrees were based upon the notions of:
- Transfer (positive, negative, and zero)
- Optional and obligatory choices of certain phonemes in the languages in
Applied linguists were able to derive a reasonably accurate inventory of
phonological difficulties that a second language learner would encounter.
Stockwell and his colleagues also constructed a hierarchy of
difficulty for grammatical structures of two languages in contrast. It
included 16 levels of difficulty, based on the same notions in
addition to structural correspondence and functional/semantic
Clifford Prator captured the essence of this grammatical hierarchy in
six categories of difficulty.
Prator’s hierarchy was applicable to both grammatical and
phonological features of language.
Prator’s six categories in ascending order of difficulty
Level 0 – Transfer: no difference or contrast is present between the two
languages; the learner can simply transfer (positively) a sound, structure,
or lexical item from NL to TL (no difficulty, hence the label of “level
Example: many vowels and consonants of Persian and English which are
the same in both languges.
Level 1 – Coalescence: two items in NL become coalesced (i.e.
merge) into essentially one in TL; learners need to learn to overlook
a distinction they are used to.
Example: the Persian learner of English must overlook the
distinction between daneshju & daneshamuz, and just use the
English word student.
Level 2 – Underdifferentiation: an item in NL is absent in TL; the
learner must avoid (forget) the item.
Example: phonemes /x/ and /q/ or expressions such as chakeretam
and nokaretam since in English for the same meaning no one says I
am your servant.
Level 3 – Reinterpretation: an item that exists in NL is given a new shape or
distribution in TL; Prator claimed that in some cases items in TL are perceived
as reinterpreted NL items;
Example: the phoneme /l/ in Persian is mainly clear /l/; whereas, in English
depending on the phonological environment, it may be clear or dark.
Level 4 – Overdifferentiation: An entirely new item needs to be
learned in L2 because of little or no similarity to the native language
Level 5 – Split: one item in NL becomes two or more in the target
language; requiring the learner to make a new distinction.
(counterpart of coalescence).
Example: desk and table in English become one single word in
Persian ( miz).