Terminology
Linguists distinguish between two kinds of universals: absolute (opposite: statistical, often called tendencies) and implicational (opposite non-implicational). Absolute universals apply to every known language and are quite few in number; an example is All languages have pronouns. An implicational universal applies to languages with a particular feature that is always accompanied by another feature, such as If a language has trial grammatical number, it also has dual grammatical number, while non-implicational universals just state the existence (or non-existence) of one particular feature.
Also in contrast to absolute universals are tendencies, statements that may not be true for all languages, but nevertheless are far too common to be the result of chance. They also have implicational and non-implicational forms. An example of the latter would be The vast majority of languages have nasal consonants. However, most tendencies, like their universal counterparts, are implicational. For example, With overwhelmingly-greater-than-chance frequency, languages with normal SOV order are postpositional. Strictly speaking, a tendency is not a kind of universal, but exceptions to most statements called universals can be found. For example, Latin is an SOV language with prepositions. Often it turns out that these exceptional languages are undergoing a shift from one type of language to another. In the case of Latin, its descendant Romance languages switched to SVO, which is a much more common order among prepositional languages.
Universals may also be bidirectional or unidirectional. In a bidirectional universal two features each imply the existence of each other. For example, languages with postpositions usually have SOV order, and likewise SOV languages usually have postpositions. The implication works both ways, and thus the universal is bidirectional. By contrast, in a unidirectional universal the implication works only one way. Languages that place relative clauses before the noun they modify again usually have SOV order, so pre-nominal relative clauses imply SOV. On the other hand, SOV languages worldwide show little preference for pre-nominal relative clauses, and thus SOV implies little about the order of relative clauses. As the implication works only one way, the proposed universal is a unidirectional one.
Linguistic universals in syntax are sometimes held up as evidence for universal grammar (although epistemological arguments are more common). Other explanations for linguistic universals have been proposed, for example, that linguistic universals tend to be properties of language that aid communication. If a language were to lack one of these properties, it has been argued, it would probably soon evolve into a language having that property.
Michael Halliday has argued for a distinction between descriptive and theoretical categories in resolving the matter of the existence of linguistic universals, a distinction he takes from J.R. Firth and Louis Hjelmslev. He argues that "theoretical categories, and their inter-relations construe an abstract model of language...; they are interlocking and mutually defining". Descriptive categories, by contrast, are those set up to describe particular languages. He argues that "When people ask about "universals", they usually mean descriptive categories that are assumed to be found in all languages. The problem is there is no mechanism for deciding how much alike descriptive categories from different languages have to be before they are said to be "the same thing"
Read more about this topic: Linguistic Universal