Human language is an incredibly vast topic, covering its social implications, usage in literature and other arts, history, psychology, neurology, its acquisition, emergence in human evolution, and computational processing, to cite just a few areas. Like many linguists, I’m interested in the grammar of human language, in how sentences are assembled in our brain. Furthermore, I’m interested in the counter-intuitive hypothesis that all human languages share a universal grammar that is innate, part of our shared genetic code, as proposed by the famous linguist Noam Chomsky.
When hearing the word ‘grammar’ many people think — with horror or fascination — of a heavy grammar rulebook. A grammar can indeed be described as a set of rules that combine sounds or signs into words, and words into sentences. But the rules that interest linguists are those at work in our heads, not the prescriptive rules taught at school. Take rules like don’t split infinitives (such as don’t say ‘to boldly go where no one has gone before’) or use nominative subjects (for example, when announcing yourself, say ‘it is I’, not ‘it’s me’). They tell us nothing about how our brain builds sentences. In fact, these two rules historically emerged from a desire to bring English closer to Latin, whose grammar was considered worth aspiring to. They do not belong to the core mental grammar of native speakers of English.
We know such mental grammar exists because the world has always been full of people who successfully speak or sign despite having been denied an education, and investigating human language reveals a more complex and unexpected picture than prescriptive grammarians would have. Consider this sentence from a recent talk at University College London by an American professor:
I am the only one here who takes care of her son.
Under the interpretation where ‘her son’ includes the speaker’s son (as in ‘I take care of my son, but other parents do not take care of their children’), the sentence is perfectly acceptable. What this shows is that even very intuitive and robust rules such as match possessive pronouns with their possessor (in this sentence they were ‘her’ and ‘I’) may systematically fail to apply in specific linguistic contexts.
As native speakers we have this wonderful ability to intuitively judge whether a sentence is acceptable in our native language or whether something is not quite right with it. For example, while ‘John likes himself’ is fine for most English speakers, ‘John thinks that his mother likes himself’ never is. Unacceptable sentences feel odd because they violate one or more rules in our mental grammar. This enables linguists to painstakingly assemble together a model of our mental grammar by examining ungrammatical sentences across all of the world’s languages, then proposing hypotheses about exactly what rules are violated, and finally testing those hypotheses.
To illustrate how this process of proposing hypotheses and testing them works, consider the hypothesis that reflexive pronouns like ‘himself’, ‘herself’, and so on, refer back to an immediately precedent subject. In the sentence ‘John thinks that his mother likes himself’, the subject would be ‘his mother’, which has the wrong gender, potentially explaining why this sentence is unacceptable. Yet further testing would soon find our initial hypothesis wanting. For example, ‘John thinks that himself is intelligent’ is unacceptable even if ‘John’ is a subject and immediately precedes ‘himself’. In fact, the rules governing reflexives are quite complex. As a first and very rough approximation, we could say that the reflexive must refer back to an item within the same clause. In the first example this subject would again be ‘his mother’, which has the wrong gender, whereas in the last example there would be no item to refer to, as the relevant clause would be ‘that himself is intelligent’. This, too, makes the sentence unacceptable.
What type of questions can linguistics hope to address with a model of our mental grammar? Let’s briefly consider just two. First, we may ask what the main components of human grammar are and also why them and not others. Typically, they would at least include syntax and morphology (how sentences and words are formed), phonology (how words and sentences are put into sounds/signs), and semantics (the meaning of words and sentences). Many linguistic phenomena concern how these components exchange information with one another. The mismatch of the pronouns in the American professor’s sentence, for example, is acceptable under the interpretation described but not in simpler sentences: we cannot say ‘I take care of herself’ to convey the meaning ‘I take care of myself’, hence syntax and semantics are both relevant when it comes to choosing pronouns. A major research programme proposed by Noam Chomsky in the 90s, called Minimalism, pushes this view to the limit by proposing that syntax is minimal — in other words, that it only exists to serve the demands of other components of grammar, such as semantics and phonology.
Secondly, we may also ask what makes a sentence acceptable to a native speaker, or, in more formal terms, how grammaticality is defined. Historically, linguists working in the Chomskyan tradition have assumed that sentences are acceptable — or, as linguists would say, ‘grammatical’ — when they satisfy all the conditions of universal grammar plus any other condition specific to the individual language being tested. In the early 90s, however, Alan Prince and Paul Smolensky argued for a new model, called Optimality Theory, where universal conditions might conflict with each other in specific linguistic contexts. Under their model, individual languages rank these conditions, and when a conflict arises each language chooses the higher ranked conditions. This new conception of universal grammar explains the variation observed across different languages in terms of the different rankings associated with each language, thus potentially eliminating the need for positing conditions specific to each individual language.
Under this model, linguistic variation itself is rooted in universal grammar. For example, Italian and English both normally place subjects at the beginning of a sentence and the main intonational stress — that is, the main intonational prominence or accent in the entire sentence — at the end. Simplifying quite a bit, let’s say that this behaviour reflects two universal conditions: SUBJECT-FIRST and STRESS-LAST. There are occasions, however, where we want to use main stress to emphasize a subject, as in ‘JOHN stole it, not Bill’ [main stress shown in capitals]. How can we satisfy both rules in this case? We can’t, because subject and stress are linked together and hence will share the same position at the beginning or the end of the sentence. Instead, each language determines which rule takes priority. English ranks SUBJECT-FIRST higher; thus, it keeps subjects first and allows stress to be non-final. In Italian, instead, STRESS-LAST may trump SUBJECT-FIRST, enabling sentences ending with stressed subjects, as in ‘Lo ha rubato GIANNI, non Bill’ (it have stolen JOHN, not Bill). Linguistic variation thus emerges from the ranking of universal conditions rather than from conditions specific to each language.
The grammar of human language can be complex and intricate. But one thing is clear: the rules in our brain are very different from those found in the heavy grammar rulebooks sometimes used at school.