Can neuroscience contribute to theories of linguistic competence?
Noam Chomsky and fellow Essentialists see our knowledge of language as a state of the mind and assert that linguistics should study this state of the mind independently of external factors, such as socio-cultural influences and psychological states. In Chomsky’s view, a linguistic theory should be concerned with the principles of the knowledge of language (competence), and not with the ways in which this knowledge is influenced by other factors and expressed in concrete situations (performance).
However, according to the neuroscientist David Marr, the concrete execution of language – i.e., the neuro/psycho/sociological mechanisms that enable us to learn and produce language – may be “all there is to explain” about humans‘ knowledge of language. If Marr is right, neuroscience will play a key role in the development of new linguistic theories. In other words, if competence can be ultimately reduced to performance, linguistic structures may be explainable in terms of patterns of neural activity. On the other hand, even if we consider linguistic competence as an independent and irreducible mental state, the structure of language may still be reflected in the activity of specific neurons.
In this article, after explaining some fundamental linguistic terminology and summarising the main differences between the Essentialist (“Chomskyan”) approach and two other important tendencies in linguistics, I will argue that, in addition to the study of performance, neuroscience can also contribute to the development of theories of competence, on the basis of recent studies which show a relationship between abstract, “higher” linguistic structures and concrete, “lower” electrical behaviours of neurons.
Part 1) The Essentialist Approach to Linguistics
Two Chomskyan distinctions
In Aspects of the Theory of Syntax, Noam Chomsky  defines competence as, roughly, “what a speaker-hearer knows about his/her language”, whereas performance is “how the speaker-hearer actually uses this knowledge in his/her everyday life”. According to Chomsky, a linguistic theory should be a theory of competence, whose object of study is not the use of language in concrete situations. “Memory limitations, distractions, shifts of attention and interest, and errors”  all belong to the study of performance, not to that of competence, and therefore they should not be part of a linguist’s field of study.
To solve the apparent contradiction of excluding the concrete use of language from linguistic theories, it is important to stress that Chomsky and his followers see our knowledge of language as a mental state, which, even if related to other states of the mind/brain, is essentially independent of external influences. In Chomsky’s view, the abstract properties of this mental state should be the subject matter of a theory of linguistic competence, i.e. a theory which, starting from the detection of syntactic structures which are common to specific languages, suggests a set of universal principles to explain these properties. For example, languages such as Italian, Portuguese and Spanish are “pro-drop languages” (i.e. they accept null subjects), while English and French are not . To give an example, the sentence Sono andato a scuola (literally Went to school) is grammatical in Italian, but not in English. A theory of linguistic competence may attempt to explain the abstract principles that make this sentence grammatical in Italian but not in English, setting aside the fact that Went to school is perfectly understandable by English speakers and used quite frequently in informal contexts. In synthesis, a theory of competence is concerned with the abstract structure of language, and not with its concrete execution. Neuropsychological processes, cultural influences and social and historical contingencies are not considered in theories of linguistic competence .
To further clarify what a theory of competence should be about, I will briefly mention another distinction made by Chomsky: the distinction between E-language and I-language. Examples of E-languages are “Arabic”, “Chinese” and “English”, which, according to Chomsky, are fictional entities, in the sense that they are historical/political constructs. To say that all Chinese people speak the same language is wrong, not only because there are hundreds of different variants of Chinese, but also because every single Chinese speaker uses a language that is slightly different from that of all the others. This individual language, unique to every human being, is the I-language [3, 4].
To summarise, Chomsky thinks that a linguistic theory should be a theory of competence (i.e. a theory about the mental state which constitutes our knowledge of language), with common features among specific I-languages as object of study.
Prewired language learners
The primacy of competence over performance and of I-language over E-language, theorised by Chomsky, is at the basis of the Essentialist approach to linguistics. By pointing at linguistic properties which are common to every human language, the Essentialists postulate the existence of a “generative grammar”, i.e. universal linguistic structures which allow each of us to rapidly acquire knowledge and mastery of the language(s) we are exposed to from birth .
According to the Essentialist approach, human new-borns are prewired with linguistic structures that allow them to rapidly learn one or more languages, independently of intelligence, socioeconomic status and type of education. If we consider that early-adopted children often forget the language of their biological parents to learn a new one when they start living with their adoptive family, it seems evident that humans are, at least during childhood, predisposed to language-learning independently of the identity of the language.
But it is not just as evident if this predisposition – generative grammar – persists during adolescence and adulthood: if a French grown-up woman moves to Barcelona, she may learn Catalan in more or less one year just by speaking with the locals, but the same would not happen to a French who moves to Beijing. This suggests that our knowledge of languages is indeed, if not determined, at least strongly influenced by factors other than generative grammar, such as the plasticity of our brains – which worsens with age – and the similarities and differences between our mother tongue and the new languages we try to learn.
Neuroscience can help us identify and understand the neural correlates of our predisposition to language learning, and why this predisposition declines with age – i.e. if we can only speak about generative grammar when referring to children (whose brain is plastic enough to grasp any kind of human language), but not when we consider the way in which adults (whose brain is already wired towards the structure of their mother tongue) learn new languages.
Part 2) Other Approaches to Linguistics
Linguists have argued – and continue to argue – extensively both in favour of and against the Essentialist approach. Even supposing that Chomsky and fellow Essentialists are right, and that all humans have an internalized generative grammar which enables us to rapidly learn whatever language we are exposed to as children, one can still be interested in studying the concrete execution of language, rather than its abstract structure, and the ways in which states of the mind devoted to the grasp and production of language are influenced by other states of the mind, and by the environment – in other words, how linguistic competence relates to performance. In the next paragraphs, I will mention a „dispute“ between Jerry A. Fodor and John Collins regarding the Chomskyan definitions of competence and performance, as it shows that these definitions are still being debated and re-defined. I will also give an outline of two – Externalist and Emergentist – important non-chomskyan approaches to linguistics, to clarify that performance is being investigated by linguists as well as by philosophers and neuropsychologists.
Competence vs. performance: a dispute
According to Fodor , the only difference between competence and performance is that theories of linguistic competence account for the knowledge, abilities and behaviours of a speaker-hearer only by reference to her internalized grammar, whereas theories of performance are concerned with the interactions between the internalized grammar and other aspects of the speaker-hearer’s psychology (“internalized grammar”, as used by Fodor, means a grammar that is internally represented to be exploited by the speaker-hearer whenever she talks and listens). Therefore, in Fodor’s view, the difference between competence and performance is essentially a difference in the level of abstraction, with theories of competence being comparable to the mathematical model of a pandemic, and theories of performance being more like the concrete observation of how the pandemic develops in specific countries with different policies – to use a COVID-19 analogy.
Fodor is argued against by Collins , who reiterates that a generative grammar, in the Chomskyan definition, is not the internal representation of a set of abstract rules and propositions, but rather a state of the mind. What Collins wants to point out is that the language faculty is not a container of grammatical rules imprinted in our brains, to draw from when we formulate or hear a sentence, but rather a property of the mind which enables human babies to grasp a language without having to study its grammatical rules, and which persists in our adult life as the basis of all the utterances that we produce – similarly to the occipital cortex, which learns how to process visual information in sighted children and keeps doing it throughout adult life, without the need for any explanation of the “rules of vision”. According to this view, as we can describe the formation of visual images abstracting from defects of the peripheral visual pathway, or from the influence of neuropsychological states on visual imagery, we should also aim to describe linguistic properties of the mind irrespective of their concrete execution, and of external influencing factors.
Externalists and Emergentists
Some linguists would rebut the parallel between the visual and the linguistic systems by pointing out that, while visual imagery is essentially an individual experience (even though it can be communicated through means such as the figurative arts), our knowledge and use of language has a fundamentally social function; language should therefore be studied in its concrete manifestations, taking socio-cultural factors into account.
In particular, the Externalist tendency studies corpora of actual utterances (e.g. all the comments written on YouTube from its foundation in 2005 to the present day) to infer the structure of linguistic expressions, and to figure out which expressions are grammatical and which are not . The Emergentists, instead, see language as a “cultural or social product” , and base their explanations of linguistic phenomena on cognitive processes, socio-cultural influences, historical events, and evolutionary theories . While the Essentialists are interested in common features of different languages, which point to the existence of universal linguistic structures, the Emergentists are concerned with differences between languages, and with the possible causes of these differences – such as social class .
To summarise, we could say that the Essentialist see abstract linguistic structures as prescinding from and preceding the use of language in everyday life, while the Externalists and Emergentists start from the collection, analysis and observation of empirical data (the concrete use of language) to infer linguistic properties.
Part 3) Neurolinguistics: between Competence and Performance
In this complex panorama, what is the role of neuroscientific research? Can neuroscience explain language processing only at the level of performance – e.g. which neurons are responsible for language processing, and why specific brain lesions lead to linguistic impairment, or can it contribute to abstract theories of linguistic competence as well? I will attempt to answer this question in the next paragraphs.
Levels of description
Chomsky’s theory of linguistic competence was put forward by the late neuroscientist David Marr as a possible example of computational (“type 1”) theory – i.e. a theory which is concerned with the principles of an area of knowledge (competence), and not with the ways in which this knowledge is expressed (performance). But Marr is not sure if our knowledge of language can be described by a “type 1” theory, whose abstract principles are not reducible to their concrete execution, or rather by a “type 2” theory, in which concrete execution (performance) is all there is to explain . In this second case, studying the activity of neural networks involved in language processing may be the key to understand our language faculty.
Neeleman and Van de Koot  resume and slightly modify Marr’s arguments to describe language on three different levels: the first – “cognitive level” – is an abstract description of the logical, universal structures of language; the second – “computational level” – is a description of the algorithm which translates linguistic inputs into logical structures of the first level, and these logical structures into linguistic outputs; the third – “physical level” – is a description of the neurology associated with linguistic inputs and outputs. Similarly to Fodor , Neeleman and Van de Koot  see theories of competence and performance as descriptions of the same object (language), but at different levels of abstraction: competence – alias grammar – theories describe language at the first and most abstract “cognitive” level; performance theories at the second, “algorithmic/computational” one.
If Neeleman and Van De Koot’s model is correct, shall we conclude that neurolinguistic studies can only offer theories of language at the third, “physical” level? It depends on the aim of the study. If a neuroscientist wants to describe the electrical behaviour of a specific population of neurons in Broca’s area, all the better. But to formulate a thorough theory of human language, it is necessary to integrate information about language processing at different levels. Neuroscientists need to be acquainted with up-to-date theories of linguistic competence (“cognitive-level” type of descriptions) to design experiments which can shed light on the correlations between neural activity and higher-level language processing. In the last two paragraphs, I will review two examples of neuroscientific studies of linguistic performance, and two of linguistic competence.
Neurolinguistic of performance
As reported in França , an influential positron emission tomography (PET) study by Damasio, Grabowski  shows that stored lexical items tend to be organized in the temporal lobe according to the type of information they represent. In Damasio’s study, healthy participants showed increased neural activity in the anterior part of the left temporal lobe when having to name celebrities, in the inferotemporal region when naming animals, and in the posterior part of the inferotemporal lobe as well as in the anterior part of the left occipital region when naming tools; these results were partially confirmed by PET scans of patients with different types of temporal lesions. This study is very interesting since it shows a correspondence between retrieval of words representing a similar type of information and specific brain areas, but it does not go beyond the third, “physical” level – to use Neeleman and Van De Koot’s terminology.
An example of recent neuroscientific research at the border between competence and performance is offered by Ding, Melloni , who demonstrate that neurons involved in phrasal and sentential comprehension are of a different kind from those merely responding to the acoustic features of syllables in spoken language. This and similar studies seem to confirm the existence of a hierarchical structure for language processing, which may be reflected in the electrical activity of neurons, but do not bring evidence in favour of or against any specific theory of linguistic competence.
Neurolinguistics of competence
Let’s consider the sentences He drank the coffee and She ate the curtains. In the first one, the transition between the verb and the direct object (“verb-complement merge”) leads to a congruous reading, while the second does not . Thanks to electroencephalographic (EEG) techniques, the act of reading incongruous sentences like She ate the curtains has been shown to be associated with the N400 (a wave of negative polarity detectable 400 ms post-stimulus) . These findings show a correspondence between processing of sentences that are incongruous at the first, “cognitive” level (competence) and specific brain waves at the third, “physical” level.
In another EEG experiment, Franca, Lemle  measured the brain activity of Brazilian Portuguese speakers while these were reading three series of congruous and incongruous sentences which all contained a verb-complement merge operation; all sentences in each series shared a specific syntactic configuration. In addition to showing that congruous and incongruous sentences are associated with different event-related potentials (ERPs), Franca, Lemle  observed that each series was correlated with a specific brain wave morphology. In linguistics, the verb-complement merge operation is divided in different sub-modules according to the specific syntax of the sentence containing this operation. Therefore, by showing that three of these different sub-modules are each paired with a specific brain wave morphology, Franca, Lemle  add new neurophysiological evidence to theories of linguistic competence.
Language processing can be analysed at different levels, and each of these levels can be studied independently of the others. However, in this article I stressed the importance of integrating approaches at different levels to gain a broader understanding of the human faculty of language. As demonstrated with some examples at the end of this article, while studying the lower, “physical” level – i.e. neurons and brain areas involved in language processing, neuroscience can also contribute to “higher” theories of linguistic competence. To design experiments which could shed light on the correlations between neural activity and linguistic structures, it is essential for neuroscientists to refer to theories of linguistic competence, and to incorporate abstract linguistic concepts in their research and terminology.
I would like to thank Corrado Bertani and Matteo Maspoli for their philosophical suggestions, Francesca Geymonat for her help with the linguistic aspects of this essay, and Aneirin Rhys Potter for double-checking my bad English.
1. Chomsky, N., Aspects of the theory of syntax. 1965, Cambridge: M.I.T. Press.
2. França, A.I. Introduction to Neurolinguistics. 2004.
3. Scholz, B.C.a.P., Francis Jeffry and Pullum, Geoffrey K., Philosophy of Linguistics, in The Stanford Encyclopedia of Philosophy, E.N. Zalta, Editor. 2020, Metaphysics Research Lab, Stanford University.
4. Chomsky, N., Knowledge of Language: Its Nature, Origin, and Use. 1986: Praeger.
5. Fodor, J. Introduction: Some Notes on What Linguistics Is About. 1981.
6. Collins, J., Faculty Disputes. Mind & Language, 2004. 19(5): p. 503-533.
7. Sapir, E., The Status of Linguistics as a Science. Language, 1929. 5(4): p. 207-214.
8. Labov, W., The social stratification of English in New York City. 1966, [Washington: Center for Applied Linguistics.
9. Marr, D., Artificial intelligence—A personal view. Artificial Intelligence, 1977. 9(1): p. 37-48.
10. Neeleman, A. and H. Van de Koot, Theoretical Validity and Psychological Reality of the Grammatical Code. 2010. p. 183-212.
11. Damasio, H., et al., A neural basis for lexical retrieval. Nature, 1996. 380(6574): p. 499-505.
12. Ding, N., et al., Cortical tracking of hierarchical linguistic structures in connected speech. Nature Neuroscience, 2016. 19(1): p. 158-164.
13. Osterhout, L. and P.J. Holcomb, Event related potentials and language comprehension, in Electrophysiology of mind: Event-related brain potentials and cognition. 1995, Oxford University Press: New York, NY, US. p. 171-215.
14. Franca, A.I., et al., Discriminating among different types of verb-complement merge in Brazilian Portuguese: an ERP study of morpho-syntactic sub-processes. Journal of Neurolinguistics, 2004. 17(6): p. 425-437.