This is likely because language uptake is most successful when "unrooted" to a pre-existing language framework. You're utilizing distinct memory/knowledge graph

This is likely because language uptake is most successful when "unrooted" to a pre-existing language framework. You're utilizing distinct memory/knowledge graphs. You sound like a bumbling fool because you're having to forge novel connections between the two recall graphs. It's why I look on awe on people like my high school Spanish teacher who fluently spoke 11 languages. When she got upset, you never knew which was going to come out!