One aspect of Arabic morphology that has been a focus of attention for both grammarians and linguists is the phenomenon described by modern Arabists as the biradical hypothesis (for some even the “uniradical” hypothesis), or within the tradition as al-ishtiqaaq al-kabiir, from Ibn Jinni’s initial formulation of it: namely, that Arabic triconsonantal roots may be categorized based upon similarity in only two (or one) of the root consonants, which leads to the claim that these two radicals together (or even single radicals) have an inherent meaning which explains these types of relationships. While some trace this phenomenon to an earlier proto-Semitic stage in which roots consisted of primarily two root consonants or radicals, others view it as a real and persistent synchronic feature of Arabic. This paper will critically examine these previous analyses and offer alternative ones which take into account a wide array of possible sources of this phenomenon The first step to this alternative analysis is to recognize a distinction between different types of morphological processes in Arabic: between paradigmatic and non-paradigmatic or associative processes. The first type of process is the most commonly recognized type of process described in linguistic derivation and inflectional morphology, while the second is a freer, less rule defined relationship between words which share acoustic features. The groups of radicals and lexical items delimited by the biradical hypothesis fall into the latter type of process. A second step is to recognize that there may be different sources for the associative word classes that the biradical hypothesis defines. First, the classes may be defined within a diachronic variationist framework: the classes are due to an amalgamation of different dialectal variants of Arabic (or related languages) into the linguistic koine that became Literary Arabic. Second, the word classes may be defined within an associative morphology viewpoint, which can be both diachronic as well as synchronic in scope. An example of such an associative word class are blend words (in English: smoke + fog -> smog; in Arabic xalaT + xabaT -> laxbaT), which are or may be productive in a language at any point in its history, and which can lead to neologisms based upon acoustic or articulatory similarities in the sounds of different words. These associative word classes will also be examined from a cognitive linguistic perspective: there is a natural tendency for language users to search for a common meaning for similar sounding words.