We report on the rapid birth of a new language in Australia, Gurindji Kriol, from the admixture of Gurindji and Kriol. This study is the first investigation of contact-induced change within a single speaker population that uses multiple variants. It also represents an innovative modification of the Wright-Fisher population genetics model to investigate temporal change in linguistic data. We track changes in lexicon and grammar over three generations of Gurindji people, using data from seventy-eight speakers coded for their use of Gurindji, Kriol, and innovative variants across 120 variables (with 292 variants). We show that the adoption of variants into Gurindji Kriol was not random, but biased toward Kriol variants and innovations. This bias is not explained by simplification, as is often claimed for contact-induced change. There is no preferential adoption of less complex variants, and, in fact, complex Kriol variants are more likely to be adopted over simpler Gurindji variants.