The present paper deals with contemporary Kazakh-Russian code-mixing, the alternation of two languages within a single clause. Kazakh, a Turkic language, has been in intense contact with Russian, a Slavic language, for centuries. The main purpose of the paper is to demonstrate how Russian adjectives are embedded in Kazakh clauses. The analysis focuses on synchronic code-mixing and is based on eleven tape recordings (38,636 tokens) of casual speech by bilinguals, which were made in Almaty, Kazakhstan. Russian adjectives will be examined as a part of (1) Russian attributive expanded NPs, (2) mixed Kazakh-Russian NPs and (3) as predicative adjectives. The Matrix Language Frame Model (with 4-M Model and Abstract Level Model) provides the theoretical framework for this paper. Particular attention will be given to the concept of Embedded Language (EL) Islands, i. e. constructions consisting of at least two well-formed content morphemes of the EL. Based on the analysis of my empirical corpus, it will be shown that EL-Islands consisting of Russian complex NPs are often strongly dominated by the morphosyntax of Kazakh, the Matrix Language.Russian adjectives often take the unmarked form of the nominative, singular, masculine, despite the feminine or neutral Russian nouns they precede. Our findings will be compared with other data sets from Turkic languages in contact with Russian. The concept of EL-Islands will finally be shown to be invalid.