Compounding is one of the most common and also very productive word formation processes among languages in the world. Because of its productivity and high frequency, it has precisely been studied by many Iranian and non-Iranian linguists. However, this productive process has not yet been studied in Mazandarani. In the present study, first Mazandarani compound words taken from Tabari dictionary, have been analysed in the form of a table and based on the criteria such as syntactic category, semantic classification of compound types (endocentric, exocentric and dvandva), the position of head and being verbal or nonverbal. By taking into consideration the above mentioned criteria and based on three fundamental concepts, namely, head, dependent (non-head) and feature percolation, the resulting compounds have been analysed separately. At the end of the article, after studying and evaluating the hypotheses, appropriate explanations have been presented. The high frequency of “N+ N → compound noun construction” refers to their predictability in meaning compared to other constructions. It means that there is a direct relationship between high frequency of a word formation process and its semantic predictability. Also, the reason why endocentric compounds are more frequent than others is due to the fact that both constituents of the compound are used in their original meanings. Hence, endocentric compounds are unmarked elements and are easier to learn in the process of language learning. The above mentioned findings are in line with markedness theory. That is why the frequency of occurrence of unmarked elements is higher than marked ones which is related to language universals.
