In linguistics, formalism is a theoretical approach characterized by the idea that human language can be defined as a formal language like the language of mathematics and programming languages. It is contrasted with linguistic functionalism approaches like cognitive linguistics and usage-based linguistics.
Prominent figures in this school of thought are Wilhelm von Humboldt, Ferdinand de Saussure (founder of structuralism), and Noam Chomsky (author of generativism). Louis Hjelmslev can also be seen as a forerunner of Chomsky's generative grammar, and Chomsky derived many of his ideas from him. De Saussure was in turn influenced by the ideas of 4th century BCE grammarian Pini, who wrote a rule-based grammar of Sanskrit.
Ferdinand de Saussure's 1916 work heavily influenced approaches that attempted to describe human language as a strictly formal system. De Saussure's approach became known as structuralism, and from it spawned the two contrasting approaches of formalism and functionalism. The Prague linguistic circle, with a functionalist approach, was founded in 1926. Roman Jakobson was a member. The Copenhagen School of linguistics was founded by Louis Hjelmslev and a group of colleagues in 1931. Hjelmslev developed the theory known as Glossematics.
In the United States, the linguistic approach of distributionalism originated from the work of Leonard Bloomfield in the 1930s and 1940s, and was further formalized by Zellig S. Harris from 1951.
Distributionalism was one of the influences for Noam Chomsky's 1957 work Syntactic Structures. It proposed an influential systematic formalization of the syntax of a human language, and started the approach of generative linguistics, an approach that has been dominant in linguistics for decades. Hjelmslev's transformational grammar has also been reworked and included in the works of Harris and Chomsky. In the late 1960s and early 1970s, a number of Chomsky's students broke with the generative idea that semantics is computed on the basis of syntax, proposing a new framework called generative semantics in which syntax was computed on the basis of semantics. The often-acrimonious conflict between these two approaches is known as the linguistic wars. In the aftermath of the linguistics wars, a number of generative semanticists such as George Lakoff abandoned formalism altogether, going on to establish a form of functionalism which was later called cognitive linguistics. Around the same time, the philosopher Richard Montague wrote a series of papers proposing a compositional model theoretic approach to linguistic meaning known as Montague grammar. Montague's work was initially poorly received by linguists in general and Chomsky in particular, leading Montague to tell Barbara Partee that she was "the only linguist who it is not the case that I can't talk to". However, after subsequent work by Partee and others, Montague Grammar grew into the broader framework of formal semantics which is now the primary formal approach to grammatical meaning.
Chomsky has revised his approach multiple times, replacing transformational grammar with the principles and parameters framework and later with the minimalist program. These approaches differ significantly in their details, while sharing the basic premise that the analysis of syntactic structure requires a way to specify recursive constituency. Since the 1960s, others have proposed more radical breaks with the transformational approach to formal syntax, including HPSG and LFG.
Models of this sort are largely ignored in branches of computational linguistics which seek to build parsers for naturally occurring sentences. However, they have been influential in branches of theoretical computer science such as formal language theory where they form part of the basis for the mathematical study of programming languages.
A central assumption of linguistic formalism, and of generative linguistics in particular, is called the autonomy of syntax, according to which syntactic structures are built by operations which make no reference to meaning, discourse, or use. In one formulation, this notion is defined as syntax being arbitrary and self-contained with respect to meaning, semantics, pragmatics, and other factors external to language. Because of this, those approaches that adopt that assumption have also been called autonomist linguistics. The assumption of the autonomy of syntax is what most prominently distinguishes linguistic formalism from linguistic functionalism, and it is at the core of the debate between the two. Over the decades, multiple instances have been found of cases in which syntactic structures are actually determined or influenced by semantic traits, and some formalists and generativits have reacted to that by shrinking those parts of semantics that they consider autonomous. Over the decades, in the changes that Noam Chomsky has made to his generative formulation, there has been a shift from a claim of the autonomy of the syntax to that of an autonomy of grammar.
Another central idea of linguistic formalism is that human language can be defined as a formal language like the language of mathematics and programming languages. Additionally, formal rules can be applied outside of logic or mathematics to human language, treating it as a mathematical formal system with a formal grammar.
A characteristic stance of formalist approaches is the primacy of form (like syntax), and the conception of language as a system in isolation from the outer world. An example of this is de Saussure's principle of arbitrariness of sign, according to which there is no intrinsic relationship between a signifier (a word) and the signified (concept) to which it refers. This is contrasted by the principle of iconicity, according to which a sign, like a word, can be influenced by its usage and by the concepts it refers to. The principle of iconicity is shared by functionalist approaches, like cognitive linguistics and usage-based linguistics, and also by linguistic typology.
Generative linguistics has been characterized, and parodied, as the view that a dictionary and a grammar textbook adequately describe a language. The increasingly abstract way in which syntactic rules have been defined in generative approaches has been criticized by cognitive linguistics as having little regard for the cognitive reality of how language is actually represented in the human mind. Another criticism is directed toward the principle of autonomy of syntax and encapsulation of the language system, pointing out that "structural aspects of language have been shaped by the functions it needs to perform," which is also an argument in favor of the opposite principle of iconicity.