The final category we consider is linguistic descriptions, such as field notes and grammars. In the usual case, the language described is specified with Subject.language while the language of the commentary is specified with Language. One interesting case is where a third language is used for elicitation. For example, a sentence from Amis may have been elicited using a sentence from Chinese, and both sentences may have been entered in the field notes. Next, commentary in the language of the linguist, such as English, may have been added. In this case, we would say that the field notes include bitext, and the languages of the bitext would be described in the usual way. The audience language of the field notes would also be specified. Using OLAC's flat metadata, we would specify the languages as follows:
We believe this is perfectly adequate for the majority of retrieval purposes. If it were necessary to represent the structure more accurately, two OLAC records could be associated with the same resource, one describing the field notes as a whole (with the above language designations), and one describing the bitext content (with just the AIS and CFR designations). The different linguistic types would be expressed using the Type.linguistic element, and the two records could refer to each other using the Relation element: