There are three processes, (among others) involved in learning English through conversation:
1. By having conversations, we get more exposure to the target language. New items and structures then become available for use. You can read more about the Input Hypothesis / Input Processing Theory by reading up on works by S. Krashen and Bill VanPatten.
“The goal of language is for communication and the most effective way to communicate is through speech.” In: The Effect of Input-based Instruction on the Speaking Ability of Iranian EFL Learners
See this link for the article in pdf format.
2. Practice makes perfect: when talking to other people, the psychological processes involved in retrieving information and creating sentences become more automatised. You may want to look up on Automaticity and Anderson’s ACT* theory.
“It would be the exception rather than the rule that we would find in a scientific theory mechanisms that directly correspond to natural language concepts. The critical question is whether ACT* can give an account of the phenomena associated with the term automaticity.”
To read the Abstract and request the full text see this link on ResearchGate
3. It seems that when conversing, learners become more aware of gaps between what they can do, and what they are not yet able to do. This appears to help students with Noticing, which is a requirement for learning. Merrill Swain’s Comprehensible Output Hypothesis and work by Schmidt on Noticing mention this.
“In my situation of self-directed learning, I have to ‘push’ myself to comprehensible output, that is, to meet the social and pragmatic needs in this target language environment.”
To read more go to this link for a Learner’s perspective in pdf format.
Here is a pdf about: Roles of Noticing in English Language Learning : A Literature Review
These three points are all good reasons to use this method.