Research in Code Mixing has been increasing in the recent years, though much of the work is on mixing of Spanish and English in the United States and Indian Languages (Hindi and Tamil) and English in India .  All the researchers agree that mixing is rule governed and that there are some universal constraints on mixing.  Quite a few such constraints have been proposed in the literature and counter examples to them have also been cited.  I will not go into this issue in this paper, but will approach the problem from the other end by showing that three variables must be taken into account in this question.  One is the typology of the languages involved in mixing; the second is the psychological relationship of the speaker with the languages involved and the third the differences in the situations in which the languages are mixed.  The point of the paper therefore is that the linguistic, psycholinguistic and sociolinguistic variables play a role in determining the constraints on mixing.

       To understand the linguistic variable fully, it would be useful to study the mixing of more number of languages with different typologies.  Nevertheless, I shall be cite one example from the literature whether the difference in mixing may be attributed to the typological property of the language.  When English is mixed with Spanish, the mixing of verb is very severely restricted, but when it is mixed with an Indian language, the mixing of verb using the dummy very `to do' is very common.  This is probably attributable to the verb final structure of the Indian languages.  Moreover, in the case of Spanish, when the verb is mixed no dummy verb is added to take on the inflectional suffixes; the verbal root is English and the suffixes are Spanish.

       A balanced bilingual mixes more in his MT than in his OT. (Pfaff 1979).  The mixing of English when speaking Indian languages by Indian bilinguals is a well attested fact.  This may be due to the prestige attached to OT and the formal situations in which it is used.  It would be interesting to see whether the same difference exists in the case of persons bilingual in two Indian languages which are not different in prestige or formality.  There are, however, some studies of mixing OT (English) with MT (Spanish) and the elements which have relatively high frequency in this mixing are clauses and simple nouns.  The verbs and VPs have zero frequency and the NP or any single constituent of it has extremely low frequency.  This seems to be true of UT-MT mix in the Indian situation also.  The sentential mixing is relatively more frequent than clauses which are more frequent than nouns; the verbs are not mixed at all.

1.    a.  I saw a girl in the bus; onne maadiriyee irundaa

                                                (she) was just like you

b.       If you pray to God, onakku oru kaSTamum varaadu

                                         You will not have any hardship

c.       I bought a puusanikka in the market

                                    Pumpkin

d.       I went to my periyappa's house

                                    Paternal uncle's

e.       *She {taaLiccaa} the sauce

                      {taaLied   }

                      {seasoned }

            *        {taaLiccaa}

                      {taaLied   } ko Zambe

                      {seasoned } the sauce

       The differences in the constraints in mixing between MT-OT cannot be attributed to the level of competence in the languages involved because these are cases of mixing when the bilingualism is additive and stable.  They may not be attributed to the structural types of the languages involved because similar constraints seem to apply in the case of English and Spanish, which belong to the same structural type and in the case English and Tamil, which do not belong to the same structural type.  They may not be attitudinal differences.  This is perhaps due to yet unknown psycholinguistic distinction which people perceive between MT and OT.

       Codes are mixed not only in case of stable or balanced bilingualism mentioned above, but also in the case of developing bilingualism as in the language learning situation, both formal and informal.  A characteristic feature of code mixing in developing bilingualism is that mixing is not bi-directional and only the language being learnt is mixed; that is, the mixing is OT-MT and not MT-OT.  Mixing is more common in the acquisition of bilingualism through informal or untutored learning than through formal or tutored learning, because of constant correction by the tutor in the latter.  This mixing is one aspect of the phenomenon called language interference in the language teaching literature, which cover both mixing and convergence.  Research on code mixing has not so far included under its purview the data from language teaching-learning situation.

       Poplack (1980) studies code mixing among imbalanced (but stable) bilinguals, whose competence, in selfrating, in MT (Spanish) is higher than in OT (English) and compared with codemixing among balanced bilinguals.  There is systematic difference between the two and the generalizes on the basis of her three types of code mixing hierarchically organized as Tag like (interjections), sentential and intra-sentential that the frequency of mixing decreases in the above order for the imbalanced bilinguals.  She discourages only of mixing of OT with MT (MT-OT) and attributes this pattern of mixing to the relative lack of competence in OT.  She hypothesized that the need for the knowledge of the grammar of OT increases progressively for the three types of mixing in the above order and relates this to the progressive decrease in frequency.

       One may notice the parallel between MT-OT in imbalanced bilingualism and OT-MT in balanced bilingualism discussed above.  If the constraints in both situations have the same explanation and the difference in competence cannot be alluded to explain the situation with balanced bilingualism, some other explanation will be required.  Further research is necessary, Data from imbalanced bilinguals, who claim higher competence in OT than MT as the Tamil-Kannada bilinguals in Mysore , may throw some light.  Similarly, data from imbalanced bilingualism which is unstable and is either additive as in language acquisition or substractive as in language loss will be useful.

       One cannot rule out the role of competence in determining the types of mixing.  I have observed in mixing English with Tamil, the mixing by the persons more competent in English follows more commonly the English syntactic patterns and the mixing by the person less competent in English has more of Tamil syntactic pattern.  In descriptive predicate, English uses the participle and Tamil uses the noun and the adverbial suffix.  The person more competent in English uses the English participle in mixing and the person less competent in English uses the English noun.

2.    a.  ellaam confused – aa irundadu.

b.       ellaam confusion – aa irundadu

          everything was confused.

3.    a.  inda kade interesting – aa irukku

b.       inda kade interest – aa irukku

         This story is interesting.

       Code mixing is attested among the bilingual children also.  One study (Reolinger and Park 1980) of children bilingual in English and Spanish shows that "the most frequent part of speech occurring overall as a substituted element in mixed utterance was the noun followed by the adverb, article, pronoun, verb, adjective, preposition and conjunction.  Phrasal mixtures constituted only a small number of the total, mixed utterances.  This seems to follow generally the pattern of mixing by adult bilinguals.  The order is likely to be different for children bilingual in an Indian language and English and the verb is likely to figure higher in the list.

       There is a crucial difference between mixing by the children and the adults.  The children do not perceive the codes as distinct at the early stages.  This is evidenced by the declining rate of mixing as the children's linguistic competence grows as reported in the above study.  According to that study, at stage I mixing levels were between 12% and 20%, at stage III levels between 6% and 12% and at stages IV and V between 2% and 6%.

       I have also observed that in the mixing of Kannada and Tamil by my child, the constraint in adult mixing that the bilingually inflected forms are not allowed is not maintained.

       3.  naan jaaruneen

            I slipped

       This paper, I think, has shown that the data base of studies of code mixing should be extended to cover languages of different types and different situations of language contact. This will give greater insight into the rules that govern mixing.