The paper deals with the level of syntactic complexity of subordinate clauses in argument texts spontaneously produced in hebrew by Arab female freshmen specializing in the teaching of Hebrew at Academic College of Education in Israel. Syntactic complexity is examined by means of the relationships between main clauses and various types of subordinate clauses; by categorizing types of logical connections encoded; and by determining the complexity of the subordinate clause itself. Our research revealed three categories of subordinate clauses arranged by their level of syntactic complexity: a. content clauses indicating a low level of complexity due to their role as mere providers of necessary information; b. Descriptive clauses indicating a high complexity level due to their free main clause placement; c. relative clauses expanding the nominal phrase and creating a high degree of compression. We found that the types of logical connections encoded by the clauses are few, unvaried and at times lexically wrong or completely absent due to first language interference, or are repeated so as to validate the addressor's position in an argument text. Furthermore their subordinate clauses contained many contents units pointing to undeveloped segments of thought: a kind of brain storm the writer conducts with himself. This may be the beginning of understanding the differences between everyday speech (verb-based, syntactically complex, lexically sparse) and academic writing (noun-based, syntactically relatively simple, but lexically complex and dense/compact).