Defining Language: Usage, Texts, Utterances, the Language Faculty, Competence and Languages

Introduction

In most sciences there is agreement about what the object is and what kind of science the actual science is. Linguistics has not yet reached this goal, although it is one of the oldest sciences, but then we have a specially diverse and complex object. According to Bickerton (2009: 4) language is «the greatest problem in science». Linguists have not been able to agree on whether language is mental or social, cultural or natural, an action or a product, an innate faculty or an acquired knowledge. Is linguistics a psychological science, a biological/natural one or a cultural one?

The question that has been most debated, is what kind of objects linguistic systems are and where they exist. There are three views on the question:

1) that they are something mental in the mind of each language user – competence, i.e. the knowledge of the user’s language. This view, usually called mentalism, is the dominant view in both generative and cognitive linguistics, and one of the few things on which the two theories agree. In that case linguistics belongs to psychology, and according to generative grammar, which thinks that competence is partially innate in the form of a universal grammar, also biology, making linguistics a natural science (Åfarli 2000: 138–139). According to Sinha (2009: 290) language is «intrinsic to human biology», and Johansson (2005: 5) identifies language with «the innate language endowments of children», i.e. the language faculty.

2) that they are something social, namely institutions consisting of conventions or norms in a language community. It is especially Saussure and European structuralism that represent this view. It implies that language is a cultural product and linguistics a cultural science belonging to the Humanities, where it has traditionally been placed.

3) since it’s difficult to disregard linguistic conventions entirely, some linguists have a third view, namely that language is both social and mental (Kofoed 1971: 13, Langacker 2016: 467).

In addition to a linguistic system language is also described as a behaviour (e.g. Bybee 2010: 6) or an acticity (e.g. Langacker 2010: 130) and identified with performance or usage. Croft (2009) describes «language» as no less than three things: a «cognitive ability» (p. 397), «joint action» (p. 398) and «a (largely) conventional coordination device» (p. 403, my emphasis in all quotes). An ability, an action and a device cannot possibly be the same thing. Calling all these things language can only lead to confusion. One reason for the ambiguous use of language may be the word itself – that everybody speak of language without articles, which makes the referent a rather abstract uncountable substance and can refer to anything linguistic. However, we are not dealing with a substance, but with individual languages or linguistic systems, e.g. English or a variety of English, e.g. Standard English or an «accent» of it, like Australian English. So I prefer to speak mostly of languages and a/the language in the singular. In my opinion, these linguistic systems, what Saussure named langue (i.e. language), are the only entities that deserve the name language or more precisely languages.

In everyday speech languages usually refer to the national languages used by ethnic groups (and often named after them): English, Norwegian, Finnish etc., which are usually counted as the «languages» of the world. This a sociolinguistic concept, defined by political criteria as well as linguistic ones. If Scandinavia had been one nation instead of three, the Scandinavian languages would just have been dialect groups. Dialects and sosiolects are only varieties of a «language» in this sense, but they are of course languages in the basic sense of ‘linguistic system’. Here language(s) is used only in this sense, and the purpose of the article is to delimit linguistic systems and the other entities mentioned in the title and discuss their place in a theory of language. They should all be described and related – just distinguishing between langue and parole (Saussure) or competence and performance (Chomsky) is far from sufficient.

According to Langacker (1987: 18), these oppositions (and most other binary oppositions) are «false dichotomies». But both competence and performance are real and important linguistic entities, and so are texts, utterances, the language faculty and languages. Languages must be distinguished from what Saussure called parole and Chomsky performance, i.e. usage. But usage, an action, should be distinguished from texts, the product of the sender’s action, and the utterances (including sentences) that texts consist of, which are the basic units of grammar (section 2). And languages must also be distinguished from the users’ language faculty (section 3) and their linguistic knowledge and skill, i.e. competence (section 4). Section 5 tries to delimit languages, especially from usage, utterances and competence, since languages are often equated with usage or competence and sometimes said to consist of utterances or sentences. The conclusion is that languages must be collective to function in communication and are institutions in language communities, more precisely sign systems, and these two entities are discussed.

My point of departure is a usagebased view of languages: that languages and usage are very close and that an account of languages, which are abstract and not directly observable, must be based on usage, texts and utterances. These are the only linguistic entities that are observable and therefore constitute the data of linguistics. I will also show that a consistent usage-based view is incompatible with mentalism, as usage consists of physical units like sounds or letters. If languages are usage-based, these units must also belong to the language that is used. This view does not preclude that parts of language are mental, namely meanings, but the units of expression are physical and cannot exist in the users’ minds.

2 The use of languages: usage, texts and utterances

Usage and Texts

Usage (latin usus) is the traditional or conventional use of a language (Langacker 1987: 65–66), i.e. use that is governed by conventions. It must be wellformed or grammatical, formed according to the conventions of the actual community, and varies between communities: What is right in one community, is wrong in another. Using something is an action, here the sender’s action: the production of «a symbolic expression assembled by a speaker» (Langacker 1987: 66). Like other actions, it has a purpose: to express a communicative act (speech act), e.g. make a statement or tell a story. In communication there is also a receiver, who in linguistic communication uses his/her knowledge of the conventions to understand the sender’s usage, with a little help from the situation and the previous context.¹ But one can hardly say that receivers use the actual language, only their knowledge of it (and of the world in general). The sender is normally one person, the receiver can also be everybody – the public (as in this article).

When using spoken or written language the sender’s action consists in producing a physical product: «a symbolic expression» or more precisely the conventional expressions of the «symbols» or signs of the language – a text, either speech consisting of speech sounds or writing consisting of letters and other characters.² Neither Saussure nor Chomsky nor Langacker distinguish between the action and its product, but describe parole, performance/E-language and usage as actions. For example, Saussure (1967: 30) describes parole as «an individual action which requires will and intelligence» – and purpose, one might add (my translation, here and in other quotes from other languages than English). So texts have no place in their theories. But in everyday life texts play a crucial role. We are ever more surrounded by texts on paper or screen, and many texts have had an enormous historical significance, e.g. the Bible, the Declaration of Human Rights, the Constitution or On the Origin of Species. Therefore texts should have a central place in linguistic theory.

The sender’s usage and the resulting text are closely connected, namely as cause and effect – formal or obscure usage results in a formal or obscure text. Still it is important to distinguish them (Lyons 1995: 18, 22), since one is a transient psycho-physical action and the other a more or less durable physical product. Compare Lamb’s (2000: 88) distinction between «The PROCESSES by which linguistic products are produced and/or interpreted» and «Actual PRODUCTIONS of speakers», or the distinction between speak and speech in English: Speech is the product of speaking. Texts can be lasting if speech is recorded or writing is hewn in stone, and can survive both usage, the user and even the language, e.g. a runic inscription or the Rosetta stone. They can therefore inform us of languages that no longer exist. If there are many texts in the actual language, the language can be at least partially reconstructed, as has been done with old Norse or Latin. So what we can preserve and study in practice, is texts, in speech recordings or transcriptions of them. For dead languages written texts are our only data. In practice, therefore, it is texts rather than usage that are the data of linguistics. «Corpus linguistics», which studies languages based on large electronic collections of texts, is therefore a better basis for linguistic analysis than the traditional method: to generalize on the basis of a few, accidental examples, selfmade or literary.

Texts consist of the expressions of signs chosen from the language that is used. This text consists of the expressions of some signs of written English – letters and other characters, like punctuation marks. They express and are understood as signs of diverse types: morphemes, word forms³ (separated by spaces), sentences (marked by punctuation marks) etc. But this is something the receivers contribute when they hear or read the text and try to understand it. It’s when we hear or see a text in an unknown language that the text appears as what it objectively is: a stream of strange sounds or characters.

Also texts have a purpose and can be divided in many text types accordingly. This text is an example of public writing, but there are also personal written texts like letters or e-mail. After the invention of writing we got a series of new text types, like letters, novels and dissertations, with e-mail, cell phone messages and discussions on the internet as the newest. In speech texts are mostly conversations, which is the original type of texts, and still the basic one in everyday life. But there are also public and formal spoken texts, like speeches, lectures, poems, fairy tales, riddles etc.

Spoken texts are normally improvised and therefore usually quite messy, with unfinished utterances, corrections and other imperfections. Written texts are normally edited and mark word forms (with spaces) and utterances and their type (with capitals and punctuation marks). And they can be divided according to their content in paragraphs, marked by indentation, and chapters, sections and subsections with headings that suggest the content. This makes written texts far more structured than spoken ones, and makes it possible to structure large texts hierarchically, which is especially useful in science. A spoken dissertation would hardly be possible. This, as well as permanence, is the greatest advantage of writing.

Utterances and Sentences

Grammatically, texts consist of one or many utterances: the smallest part of a text that is grammatically independent in the sense of not being a constituent in a syntagm and having a communicative purpose like expressing a statement, a question, an answer to a question, an exclamation, a greeting, a curse etc. Therefore an utterance can constitute a text by itself – a minimal text (Lyons 1995: 35) – e.g. on signposts and posters: Don’t walk on the lawn (a prohibition), Exit or Fresh tomatoes (information). But most texts consist of many utterances which are semantically connected, making the text coherent. Although there are semantic relations between them, potentially expressed by words like but, however, therefore, they are grammatically «…in a relation of H[jelmslev] constellation only: none of them presupposes another, since each is potentially at least a self-contained unit» (Garvin 1954: 85).

In writing utterances are usually marked by capitals and «big» punctuation marks (excluding comma), which Faarlund et al. (1997: 923) use as a definition. Therefore they consider the sequence Ja, det stemmer (Yes, that’s right) as one utterance. This corresponds to the definition of utterances in speech by intonation and pauses, as in Crookes (1990: 186 f). But neither intonation and pauses nor punctuation are very consistent. In speech people often make pauses in the middle of a sentence while they are planning how to finish it or searching for the right word, and in writing utterances can also be marked by comma and a corresponding intonation in speech if they are short and closely connected semantically, as in Faarlund et al.’s example. Here Ja is not grammatically integrated in the following sentence and functions as an utterance by itself – a confirming answer. The sequence therefore consists of two utterances with the same semantic function (confirming), and the last one repeats the content of the first. The defining quality must therefore be grammatical independence, not punctuation or intonation. At most, the possibility of using big punctuation marks can be used as an indicator of utterance status: The possibility of writing the example as Ja! Det stemmer (and pronounce it with a pause) shows that we have two utterances.

Most utterances are sentences, which is why sentences are such an important category in grammar. Most linguists distinguish between utterances and sentences, but in different ways. There seems to be agreement that utterances are the largest unit of grammar (Lyons 1968: 172), and the same goes for sentences (Garvin 1954: 84). There is also agreement that both utterances and sentences are grammatically independent, compare Bloomfield’s (1933: 170) definition of sentences as «…an independent linguistic form, not included by virtue of any grammatical construction in any larger linguistic form». But according to Lyons (1968: 52) sentences are «units of langue», utterances «units of parole». However, also individual sentences like this one are part of parole, i.e. usage and texts. They are both utterances and sentences, and neither of them are part of the actual language (see section 5.2).

Others regard sentences as products, utterances as actions, e.g. Hurford & Heasley (1983: 15): «An utterance is the USE by a particular speaker, on a particular occasion, of a piece of language...». And according to Dyvik (1998: 15) the Norwegian sentence Det har snedd i natt (It has been snowing tonight) uttered on two different occasions are two different utterances (but the same sentence). But as many utterances are sentences and both occur in texts, also utterances should be regarded as products, «a piece of language». Like sentences they have a certain structure – e.g. they can be interjections or sentences, consist of so and so many words etc. In that case Dyvik’s sentence is the same utterance (and sentence) every time one utters it, and one can use a certain utterance many times, as we do with utterances like Hello! or How are you?

Bloomfield’s sentence definition above would include units like Hello, Yes, Wow or You fool, which are usually not regarded as sentences. However, they are utterances – a wider category that includes sentences, which have a certain structure: They must contain certain constituents, at least a finite verbal, usually also a subject and other sentence constituents, alternatively a subject (NP) and a predicate (VP). They normally function as utterances and express a speech act. We have sentence types for the most important speech acts, such as informing, asking for information or requesting actions. However, they may be elliptical, especially in speech. In the sequence What’s your name? – John Smith, the answer can be regarded as an elliptical sentence with only the new information mentioned, short for My name is John Smith.

Also utterances have a structure. Most of them are sentences, but they may also consist of syntagms like You fool, Good morning, Down with X or single words, either interjections or yes/no – words that function as complete utterances by themselves and cannot be used as contituents. There are «rules» or conventions also for such utterances. So utterances are the largest units of grammar, as Croft & Cruse (2004: 291) point out when they say that grammatical knowledge represents utterances. That’s why we mark them in writing.

The Faculty of Language

3.1 Faculties, knowledge and skills

Language is sometimes described a knowledge and sometimes as a faculty, capacity or ability, but faculties, knowledge and abilities or skills are different things. Faculties or capacities are innate and universal, a quality of the species, e.g. man. The reason is that they depend on our organs and are functions of them: We have the capacity to walk and run because we have legs, but not the capacity to fly because we don’t have wings. And we have all the same organs and therefore the same faculties, although in different degrees – som are better runners or communicators than others. Knowledge and skills are acquired or learnt from experience and individual – we don’t have the same experience with and knowledge of languages and other objects. Some people have no knowledge of e.g. English while others have an almost perfect knowledge of it and are skilled users of it. But everybody have the capacity to speak English.

The faculty of language is the capacity to make, learn and use languages. Using a language requires certain organs, both the speech organs or the hands for writing and signing and a well developed brain to govern them. We must produce and perceive complex sound sequences at great speed, combine word forms hierarchically to complex utterances and interpret such utterances on the spot, as they are produced. Only man has these capacities in a suffiently high degree and is the only species with an open sign system with a large lexicon, grammar and an expression system like sound systems or alphabets, making it possible to express any message. Other animals have at most small, closed systems of signs without grammar which can only express certain messages, e.g. a set of warning cries for different predators among vervet monkeys (Seyfarth et al. 1980).

But we must of course distinguish our faculties from the products of these faculties, from languages to atomic bombs. When Bouchard (2013: 333) says that «…language is a biologically determined, species-specific, genetically transmitted capacity», he describes the language faculty, not languages. A language, e.g. English, is not a capacity, but a product of our language capacity. Unlike faculties, languages are transmitted not genetically, but culturally, by learning. Therefore we must distinguish, like Fitch (2010: 381), between «glossogeny» or the evolution of languages and «phylogeny», the evolution of the language faculty, which must have evolved in parallel, under mutual influence: A more complex language required a better language faculty, and a better faculty resulted in a more complex and better language, which resulted in better survival for those with the best language and language ability.

That we make, learn and use languages, shows that we have the ability to do so. What one can discuss, is what this ability consists of, and especially if it is a specifically linguistic ability or more general abilities like high intelligence and a good memory. Most likely, it is both (Lieberman 1984: 1). Using a language obviously requires high intelligence and a good memory. We must be able to generalize and establish categories, and we need both a good long time memory and a good working memory: We must remember tens of thousands of words a whole lifetime, and the beginning of a sentence until we have heard the end of it. But within certain limits the language faculty is relatively independent of general intelligence (Lenneberg 1967: 229, Bybee 1996: 38 f), and the organs we use for speech are adapted to speech (Lieberman 1984: 271). Furthermore, there is a critical period for language acquisition and language centers in the brain, usually in the left hemisphere.

This shows that using language also requires specifically linguistic faculties, e.g. for articulation and for constructing and understanding sentences. Also that our nearest relatives among the animals, the chimpanzees, have a limited language ability and hardly any speech ability at all, shows that we have developed language abilities, at least speech abilities, since we parted from them. According to Fromkin et al. (2019: 453) «…language is a separate cognitive module», but many parts of the brain contribute to the use of language, also the right hemisphere (Fromkin et al. 2019: 454, Lobben 2023: 112, 153). And according to Lieberman (2006: 374), the parts of the brain that govern language, also govern other activities, like cognition and movement. So there is no specific «language organ». But we have used speech for a so long time that we are adapted to speech, both physically and mentally, and therefore spoken language is the only universal sign system, used by all normal individuals in all communities. So speech must have been our first language, and the only one for a long time.

However, today we also have sign systems with others expressions, first and foremost writing and gestures. The ability to communicate by sounds can obviously be transferred to signs with other expressions, so we have a general language faculty, not limited to speech. Although Saussure himself equates language and speech, he declares (1967: 26) that Whitney is right to say that the nature of the sign is inessential and that what is specific to man is «…the ability to create a language, i.e. a system of distinct signs corresponding to distinct concepts». And Whitney is right. According to Fromkin et al. (2019: 453), all «modalities» are controlled by the left hemisphere and organized in the brain in the same way. Different sign systems require different abilities, but they have basic common qualities, as they all consist of signs, both minimal signs (morphemes) and complex signs like sentences, and have both a lexicon and a grammar. So we have both an innate and general faculty to use sign systems and special abilities or skills to use specific sign systems. Ability to speak doesn’t entail ability to write or use gestural languages, and people may have a reduced ability to read and write (dyslexia) without effects on the ability to speak. This shows that these are distinct linguistic abilities or competences, which must be learnt.

Competence

4.1 Knowledge and skill

That we have the capacity to use any language, doesn’t mean that we have the ability (are able) to use any language. To be able to use specific languages, we must be competent in using them. We need both a theoretical knowledge of the language (knowing that) and a practical skill (knowing how) in using it rightly as sender, which requires practice. For example, it is not enough to know what case German prepositions govern and the case forms of all case-inflected words – one must also be able to put any noun or pronoun with potential modifiers in the right case on the spot, without reflection. So competence is partly a knowledge: a more or less complete mental model of the actual language in both senders and receivers, and partly a skill: the ability to use that knowledge to make and understand utterances in the language. And we must know both the structural conventions – lexical, grammatical, phonological and graphological – and the functional, i.e. pragmatic or communicative, conventions (see section 5.4.3), e.g. how to speak to strangers or children, ask people to do things, tell a joke or write an article.

Some linguists, e.g. Saussure (s. 1967: 112) and Hockett (1958: 137), describe languages as habits. But habits are relevant for skills, not languages. A convention, e.g. the English do construction in questions and negative statements, is not a habit, but to use the convention becomes a habit when one has done it a sufficient number of times. Skills are acquired by practice and must be entrenched – a habit. Linguistic skill is partly automatic – according to Lieberman (1984: 59) automatization is «one of the essential elements of human language». A clear example is articulation, which involves rapid, coordinated movements of several organs without any reflection or conscious effort whatever.⁴

Linguists sometimes call competence «internalized norms», e.g. Vikør (2007: 72). But competence cannot be norms because it is individual and governs only the usage of the actual user, not of a language community. Norms or conventions are collective. Competence is knowledge of norms, and like all other knowledge, it varies. It may be incomplete, especially in users like children or foreigners. And it may wrong; like all knowledge, linguistic knowledge is fallible and may not be accordance with usage, resulting in faulty use. If the sender’s competence is imperfect, e.g. a word is not spelled like the sender thinks it is, his/her use of it will also be imperfect.

4.2 Active and passive competence

Since competence is used both to produce and to understand utterances, we have to distinguish between the sender’s and the receiver’s competence, or active and passive competence. It seems that active competence is more demanding than passive competence. People usually have a larger passive than active vocabulary (Lamb 2000: 112), i.e. they can understand more than they can say themselves, and one may be able to hear a phonetic difference without being able to produce it. And in language acquisition comprehension comes before production (Burling 2000: 27, Johansson 2005: 181). Even many animals, like dogs and horses, can understand quite a few words and utterances without being able to produce them.⁵

The reason for this imbalance must be that the sender has to plan and produce utterances, which requires skill and practice, while the receiver merely has to understand them, which only requires intelligence and knowledge of the language. We learn language primarily as receivers, by observing and gradually understanding the utterances of others. Therefore, active competence presupposes passive competence: First one must understand what other people say, e.g. the words they use, and then one can use the words oneself, after some practice. According to Lamb (2000: 112) we should «...impart some degree of primacy to the receptive side of the linguistic system».⁶ In addition, the sender goes from meaning to expression while the receiver goes the opposite way, and, as Lamb points out, the first is more difficult than the second. It is harder to remember how a certain meaning is expressed, i.e. to «find» or activate the expression, than to remember what an expression means. This is especially wellknown for proper names: Sometimes we cannot remember what a known person is called, but as soon as we hear the name, we recognize it and know who it refers to. According to Jackendoff (2002: 213, note 9) the reason is that receiving just requires «recognition memory» while sending requires «the much more difficult recall memory».

4.3 Innate competence and universal grammar?

According to Fromkin et al. (2019: 367) spoken and sign languages, unlike writing, are «biologically innate and need not be purposefully learned». If they are innate, why do they vary? Why can’t everybody sign, and why must language acquisition «…be triggered by input from the environment» (Fromkin et al. 2019: 462)? Purposefully or not, they must all be learnt. The language faculty is innate, but not languages and competence. Conventions are man-made and must be learnt by ordinary cognitive processes like abstraction and generalization, and linguistic conventions are learnt the same way as other conventions: by observing and imitating others who know them and follow them, i.e. others’ usage and utterances.

As we learn our first language without systematic instruction, competence in our native language is to a large degree unconscious or intuitive and inaccessible for introspection – with the exception of linguists. Every day we use words whose meaning we cannot define and follow grammatical rules that few of us can describe. This largely unconscious knowledge also makes us able to hear whether a sentence is grammatical or ungrammatical, usually without being able to explain why, which presupposes grammatical terms and concepts.

Structural conventions are learnt by observing regularities in the utterances, factors that recur. According to Langacker (2000: 30) language acquisition «...in essence [...] reduces to reinforcement of the commonalities inherent in expressions that actually occur», and Ono & Thompson (1995: 219) say that spoken utterances show that we use conventional «constructional schemas» which are «…distilled from large numbers of speech events». In other words, we construct hypotheses about the language on the basis of the utterances that we hear, which we test in our own utterances. Generalization (induction) is fallible because the data base, in this case observed utterances in the language, is small in the beginning – but it increases every day, and hypotheses that are not confirmed, can be revised. The method is called trial and error, which is the standard method of learning. It consists of both induction and deduction: One goes from incidences (tokens) to a general rule (types), and from the rule to new incidences – an inference that can give a wrong result, e.g. analogical forms like gived for gave, from the wrong hypothesis that give is a regular verb. This hypothesis is falsified when one has heard gave a sufficient number of times. But forms like gived show that the user has induced the rule for the past of regular verbs, even if the deduction to gived is wrong because the rule was applied to a verb where it doesn’t apply according to the conventions of English.

Such analogical forms show that grammar is learnt like the rest of language. According to Fromkin et al. (2019: 385), children “…know that sentences are organized hierarchically, and this is not information that is provided in the input”. But it is provided. As ordinary syntactic analysis show, most sentences are hierarchical, with constituents within constituents, and one cannot understand them without realizing this. To understand who went home in The old man went home, one must analyse The old man as one constituent (subject), referring to the agent.

Fromkin et al. also point out (2019: 385) that “children are not provided with information about structure dependency, constituent structure or indeed any abstract property of grammar”. No, but they are provided with intelligence and the ability to analyse utterances in constituents and discover their structure. The authors admit (p. 386) that word (or constituent) order is learnt from the input, so according to them grammar is partly innate and partly learnt. But if children can learn constituent order, they must be able to analyse sentences in constituents. The authors also admit (p. 422) that children “…extract the rules […] from the language(s) spoken around them” and (p. 439 f) that listeners must determine the syntactic relations between constituents and solve syntactic ambiguities. When people can understand the difference between They have seen you and You have seen them, they must intuitively have analysed the sentences grammatically and understood that the pronouns have different syntactic functions in the sentences and different semantic functions in the propositions they express. That “the human brain is specially equipped for acquisition of human language grammars” (Fromkin et al. 2019: 386) doesn’t mean that we are born with a grammar, just with the ability to “extract” it from utterances.

So grammar is just as conventional as the lexicon, and there is no innate knowledge of language and no universal grammar, no more than a universal lexicon. Like words, grammar varies from language to language and must be labouriously learnt. Even closely related languages like English, German and Scandinavian have quite different grammars. Moreover, lexicon and grammar overlap since there are grammatical words and word classes are regarded as grammatical organization and treated in morphology. And there are fixed expressions with a syntactic structure but a holistic meaning, e.g. figure out or put up with, which must be regarded as both grammatical and lexical units. So grammar and lexicon are closely related and are learnt together in the same way.

Languages

5.1 Languages and usage

100 years after Saussure many linguist still do not distinguish between languages and the use of them. Bickerton (1995: 11) rightly complains over «…the persistent confusion between a thing and the uses of a thing». Bybee calls «language» both «…one of the most systematic and complex forms of human behaviour» (2010: 6) and «a conventional cultural object» (2010: 119, my emphasis). Langacker (1987: 18) rejects the whole distinction and calls «language» both «a structured inventory of conventional linguistic units» and «a processing activity» (2010: 130, my emphasis). Also Kemmer & Barlow (2000: XI) reject the distinction «...since performance is itself part of a speaker’s competence» and «...the primary object of study is the language people actually produce and understand» (p. XV). But performance (usage) cannot be a part of competence since performance is an action while competence is knowledge of the system used and the ability to perform the action, and what people produce and understand, is texts and utterances, which are not the primary object of linguistics. The primary object of linguistics cannot be anything else than languages, which are not actions but systems that may be used to produce texts and utterances in the actual language.

However, languages and usage are closely related. Cognitive linguistics is usagebased, which means that languages and usage are supposed to be entities of the same kind and influence each other: «…structure [...] is not independent of usage or radically different in nature. Rather, structure emerges from usage, is immanent in usage, and is influenced by usage on an ongoing basis» (Langacker 2010: 109). According to Bybee «…linguistic structure emerges through language use» (2010: 204), and «Usage feeds into the creation of grammar just as much as grammar determines the shape of usage» (2006: 730). Kemmer & Barlow (2000: IX) say that usage events «...both result from, and also shape, the linguistic system itself in a kind of feedback loop» (my emphasis). Saussure (1967: 37) makes the same point when he says that languages (langue) and usage (parole) presuppose each other and calls languages both the instrument in and the product of usage. Newmeyer (2003: 682) calls usage-based theories «anti-Saussurean», but they are definitely not. One could even call Saussure implicitly usage-based. That languages are usage-based does not mean that usage of a language is the language, even if some reject a sharp distinction between the two. The present author is not among them.

Usage-based means that usage creates and defines a language, something linguists acknowledged long ago by the proverb Usus (est) tyrannus, usage decides (what is right). According to Langacker (1998: 35),

...all structures ascribed to a linguistic system must either be directly encountered as parts of occurring expressions, or else capable of emerging from them through the basic processes of abstraction and categorization.

Compare Langacker’s (1987: 53–54) «content requirement»: that the only linguistic units are «phonological, semantic, and symbolic structures that actually occur in linguistic expressions». Langacker (1991: 289–290) underlines that this excludes «arbitrary descriptive devices» that lack expression and content, and Kemmer & Barlow (2000: IX) point out that units that don’t occur in usage, simply don’t exist. There are no «underlying» structures and no «transformations» (movements).

And usage is not opposed to grammar. There is grammar also in usage: suffixes, grammatical words, phrases, clauses, sentences etc. – the same units as in the actual language. The only difference between a language and usage, texts and utterances in the language is that the units of a language are abstract types (classes) while the units of usage, texts and utterances are concrete, physical tokens (exemplars) or occurrences of those types, used for a communicative purpose. In English usage and texts there is an unlimited number of the article the, but they count as one in the English language. In languages the units are potential – possiblities that one can choose and use – in usage, texts and utterances they are actual, chosen or used possibilities. Therefore we can observe usage, texts and utterances while languages are abstract. For example, in languages sentence constituents like subjects and objects have a certain distribution or possible positions, in usage they have an actual position, one of the possible ones. And we can only know the distribution of the subject in a language from its positions in usage and texts.

So the grammar of the language can only be known through the grammar of usage and texts in the language. To «characterize grammar independently of usage» (Newmeyer 2003: 702) is impossible. If we want to know how interrogative sentences are made in English, we must examine the interrogative sentences of English usage and texts to see what they have in common that distinguishes them from other sentence types. There is no other way, as Hjelmslev (1973: 127) points out: «…it is only through the analysis of the text that we learn to know the language». Usage, texts and utterances show what the units of the actual language are, the linguistic system or the language that is used. That’s why we can learn a language from usage, texts and utterances, by generalizing from tokens to types. And that is why a runic inscription can tell us something about Proto-nordic, and why grammarians support their rules with examples. We demand examples of a category, preferably several, before we accept it. This makes linguistics an empirical science with an observable basis that can falsify any theory.

5.2 Languages and utterances

Since utterances are the largest unit of grammar, some linguists regard languages as collections of utterances or sentences. Chomsky (1957: 13) defines a language as a set of sentences, and Croft (2000: 26) defines it as «a population of utterances in a speech community». Dyvik (1998: 14–16) rejects that a language can be a collection of actual utterances – that would be a text or a corpus – but accepts that it is a collection of potential utterances. Compare Bloomfield’s (1926: 155) definition of a language as «The totality of utterances that can be made in a speech-community...» (my emphasis).

But as the number of possible utterances and sentences is unlimited, this would imply that a language is unlimited and therefore cannot be described or learnt completely, and that using a language consists of choosing suitable utterances (which then become actual). But Atkinson et al. (1991: 38) point out that it is «...conclusively demonstrated that the number of English sentences is infinite and, therefore, that English cannot be equated with any corpus no matter how large». A language is something limited that can be described and learnt almost completely, and most utterances are not chosen but constructed. It is not difficult to construct utterances which are original, like Chomsky’s famous Colorless green ideas sleep furiously. You just have to say something sufficiently strange or silly. Exceptions are fixed phrases that are used every day, like Good morning! or How do you do? Here we can speak of choosing from a collection. As Croft (2010: 328) points out: «Clearly, some utterances are computed, because they are novel. Equally clearly, many utterances, or large chunks of structure in those utterances, are simply retrieved, chiefly because they occur frequently enough to be stored autonomously». In addition, utterances are individual products consisting of concrete units (tokens) like individual sounds while languages are collective and consist of general types of units like sound types. It is exactly because languages consist of types that they are limited and can be described and learnt, in theory completely.

So a language is more abstract than a collection of potential utterances, namely the system or the conventions that the utterances realize or follow. Dyvik (1998: 16) concedes that «All sentences [that are possible in Norwegian] must be structured according to certain principles or norms for how a Norwegian sentence is composed, and there must be a limited number of such principles or norms». And the principles we know from previous utterances that we have observed and unconsciously analysed. We don’t learn utterances, but their general structure, i.e. utterance types, mostly sentence types, e.g. interrogative, passive or relative sentences. Interrogative sentences in general, as a sentence type or construction, are a part of English, but not the individual interrogative sentences in usage and texts. We have to learn grammatical units like question words and rules like inversion (finite verbal + subject), e.g. (Why) have you done it? and the use of do as auxiliary when there is no other auxiliary, e.g. (Why) did you do it?, with exceptions like (Why) are you sick?, but fortunately not the unlimited number of individual interrogative sentences like these examples. They are not part of English, but their structure is.

5.3 Languages and competence

5.3.1 Individual languages?

Mentalism equates languages and competence, which means that languages exist as linguistic knowledge and skill in the brains or minds of each user of the language: The brain is «where human language resides» (Fromkin et al. 2019: 447), «Language is an integral part of human cognition» (Langacker 1987: 12), the goal of cognitive linguistics is «to model linguistic knowledge correctly» (Langacker 2009: 251), and «...symbolic relationships between form and meaning are not to be found ‘out there’, but are only established in the minds of speakers» (Langacker 2010: 128). However, Langacker also underlines that languages are conventional: He calls language «a structured inventory of conventional linguistic units» (2013: 222) and bases linguistics on usage, the conventional use of languages. So cognitive linguists must be counted among those who think that languages are both mental and social.

But what exists in the brain, must be individual, since there are only individual brains. This way languages are reduced to individual systems or idiolects: «The only scientifically genuine entities are individual grammars situated in the heads of individual speakers» (Pinker and Bloom 1990: 721). As Faarlund (1990: 32) points out, this means that there are as many «languages» as there are speakers, and one cannot speak of language varieties like national languages (e.g. English), standard languages (e.g. Standard English), dialects and sociolects, which all are collective entities. Such varieties are what linguists usually are interested in and are described in dictionaries and grammars, and it is generally estimated that there are about 6 thousand (national) languages on Earth, not 8 billion. And we usually think that languages are passed down between generations and have a history of thousands of years, as Bybee (2010: 119) underlines: «...language is a conventional cultural object that has evolved over time and continues to evolve». Faarlund (1990: 32) points out that this is incompatible with a view of languages as individual knowledge, because such knowledge dies with the user and lasts only a lifetime.

To be «internalized» or «established in the minds of speakers», linguistic units must exist «out there», outside their minds. One can only «internalize» or learn something that exists externally, by making a mental copy of it, i.e. knowledge. Namely in society – partly as individual tokens (occurrences) in usage and texts and partly as the general types of units that usage represents. The article the definitely exists also in the outer world, partly as occurrences that can be observed in speech and on paper or screen, and partly as a type in the English language. It is only the expression – three letters – that the readers can see, but if they know the conventions of English writing which connect these letters to their meaning and use, they recognize the letters as the definite article. This text consists of the expressions of the signs of English writing – morphemes, word forms, sentences etc. – and readers who know English writing, will know the meanings of the signs and can hopefully infer what I mean by these expressions.

Nor can a mentalistic view of language explain grammaticality – that some utterances are grammatical or conventional while others are not. According to Dyvik (1998: 21) it’s competence that decides what is correct usage. But competence is individual and governs only the actual user’s own usage, not the usage of others, and it may be incomplete or simply wrong. If someone should believe that might in English is spelled mite, this does not make it right. The convention and usage of English writing is the spelling might, and that is an objective fact which is independent of what individuals think and know. The large majority of users of English have agreed on this spelling and show it by using it and accepting it in others’ usage while condemning other spellings. Langacker (2000: 10) underlines that wellformed means the same as conventional, i.e. in accordance with the collective norms of usage, not individual competences.

If languages are conventional, they cannot be individual. There is agreement that conventions are collective and exist in a group (Itkonen 1978: 111, Dyvik 1995: 24). Convention means ‘agreement’, and one cannot agree with onseself. And nothing can be both individual and collective, so here one has to choose. The contradiction comes out clearly when Dabrowska (2015: 222) says that «...although languages live in the mind, they belong to communities, not to individuals». If languages live in the mind, they cannot belong to communities, which have no mind. What exists in the mind and what exists in society must be two different things, and only one of them can be a language. Sinha (2009: 306) underlines that «Language as a social institution has its own proper structure which necessitates, but is irreducible to, the intentionality of its users», and Geeraerts (2016: 529) says that «Language as a communicative instrument transcends the individual».

Harder (2003: 67) points out that meaning is not «private property, but something shared with others». A private language would be useless, because it could not be used for communication. To be used for communication a language must be known and used by a group, as Saussure pointed out (1967: 112).⁷ That’s why nobody can decide over a language (if you are not a dictator), but everybody have to accept it as it is, as Saussure also pointed out (1967: 101, 104), if they want to communicate. At best, one can try to change it by one’s own usage. If one is wellknown and respected, one might succeed.

Especially standard languages pose a problem for mentalistic linguists. According to Leech (1981: 205) «...culturally institutionalized languages such as English acquire what might be called a ‘corporate lexical competence’, greater than the lexical competence of any one of its users». But a language cannot have (e.g. lexical) competence – that is for language users only. Instead it can have a greater lexicon than any single user of the language, and that is what we usually find in a dictionary. Hudson (1984: 31–32) thinks that language is «…a property of the individual», but still accepts that standard languages are collective because they are regulated and codified in dictionaries and grammars. But not all standard languages are formally regulated by a language academy, e.g. standard English. They are just accepted as they are used, with an enormous authority and prestige, and everybody try to write and even speak «correctly», according to standard usage. As to codification, all languages can be codified or described in dictionaries and grammars, and also many dialects are exhaustively described. The greater need for regulation and codification of standard languages is simply due to the fact that they are used in writing and taught in schools.

5.3.2 Mentalism and usage-based

Kemmer & Barlow (2000: VIII) think that a usage-based theory is compatible with a view of language both as «...structures derived by the analyst from observation of linguistic data...» («the ‘external’ linguistic system») and as «...structures posited by the analyst as a claim about mental structure and operation» («the ‘internal’ linguistic system»), even if the last view is the usual one. But usage consists of physical units like sounds or letters, and if usage and language are not «radically different in nature», these units must also belong to the language that is used. If languages were mental while usage is physical, they would be radically different in nature, and usage could not be the use of a language. A tool does not change nature when it is used. So languages cannot be purely mental, but include the physical units of usage as types. Spoken languages include sound types, and they are certainly not mental objects. That is probably why phonetics often is not regarded as part of linguistics (Matthews 2001: 41).

Of course we need a mental model or representation of a sign’s expression as well as its meaning (Saussure’s image acoustique), but that is competence, not language. A sound picture cannot explain communication, because nobody can hear it. An expression must be physical to communicate and be learnt (Papazian 2025 b: 68). Atkinson et al. (1991: 59) rightly point out that “…if language is to have public status it must be encoded in a medium accessible to the senses”. Signs and languages are partly physical (the expression) and partly mental (the meaning). The whole point of a sign must be to connect by convention ideas, which cannot be observed, to something physical, which can be observed. So usage-based is incompatible with mentalism, and implies that languages are institutions.

5.3.3 Knowledge and its object

Harder (2015: 239) mentions «a basic theoretical lacuna in mainstream thinking about language [...]: the absence of an account of the relation between language in the mind and language in society». In his own contribution to the filling of this lacuna (Harder 2003), the conclusion (p. 66) seems to be that «…language straddles the internal/external divide», i.e. that languages are both internal (mental) and external (social), which is impossible: The same object cannot be both internal and external, so there must be two objects, one internal and one external. The lacuna should be filled simply by distinguishing between internal knowledge (in the mind) and its external object (in society) – what one knows, which may be a language. Knowledge always has an object and is knowledge of something, e.g. a language. There must be something to know. As Jackendoff (2002: 298, note 4) points out, the expression knowledge of language implies «an external entity, ‘language’, that is known».

When Bybee (2001: 7) says that «Mental representations of linguistic objects have the same properties as mental representations of other objects», she implies that we have to do with two things: «linguistic (and other) objects» and «mental representations» of these objects. The history of England and someone’s knowledge of the history of England are two different things, and so is English and someone’s knowledge of English. The first is something social, collective and objective, the second something mental, individual and subjective. When Langacker (1991: 286) says that the goal of linguistics is to describe «…those structures and abilities that constitute a speaker’s grasp of linguistic convention» and Atkinson et al. (1991: 40) that it is to describe «...the native speaker’s knowledge of his language» (my emphasis in both quotes), one could ask if it’s not rather to describe «linguistic convention» and «his language», preferably more completely and precisely than an ordinary speaker’s knowledge of these objects.

If languages are knowledge, we must ask whose knowledge it is, because knowledge is individual and variable. Chomsky’s answer to the question (1965: 3) is that it is the competence of an “ideal” language user with “perfect” knowledge of the actual language. But no one knows a language perfectly, e.g. all the words that are used in the actual speech community. And how would we know that a user’s knowledge is perfect? This presupposes a standard to compare the user’s knowledge with, and that cannot be anything else than usage and the language itself, the object of knowledge. Knowledge is considered perfect or right if it agrees completely with its object and imperfect or false if it doesn’t. In other words, knowledge is true belief of the object.⁸ That it is meaningful to speak of knowledge of languages, shows that languages cannot be knowledge, but, as Dyvik (1998: 21), points out, something we can have knowledge of (and be right or wrong about), a possible object of knowledge. Namely institutions, more precisely sign systems created by our forebears that we recreate ourselves today.

5.4 Languages as institutions

5.4.1 Conventions

Institutions exist as conventions in communities, in this case those that govern the usage of a language. This view has been advocated mainly by Saussure, who calls language (langue) «a social product of the language faculty and a collection of necessary conventions» (1967: 25), «an institution» (p. 33) and «a kind of contract between the members of society» which is «outside the individual» (p. 31), «exists completely only in the collective» (p. 30) and presupposes «a speaking group» (p. 112).⁹ This means that languages are creations or cultural products. According to Bickerton (2009: 92), «…no serious scholars nowadays doubt that language is, at bottom, biological rather than cultural, and therefore was not created, but somehow evolved». But creation is not opposed to evolution – also cultural products and society itself evolve. Linguistic units like sounds, words and constructions don’t just pop up by themselves, they must have been created or invented by someone who needed them. That languages are created, is especially obvious in «artificial» languages like Esperanto or Nynorsk (New Norwegian), which have been created by a single, known person. «Natural» languages have been created gradually by their users, not deliberately and consciously, but in the same way as they are altered or recreated in language change: through usage. We continue to create new sounds, words and constructions, i.e. new languages. So all languages are artificial in the sense of being man-made.

Conventions or norms, in everyday speech «traditions», «customs» or «rules», are agreements between people in an organized group. They are agreements in practice, manifested by the same behaviour or «a regularity in behavior» (Croft 2009: 401), e.g. the custom of calling a cat a cat. We are speaking of collective and not individual behaviour – that one person drives on the right side of the road might be a coincidence and shows nothing. But if all or most people drive on the right side, they must have agreed to do so, at least tacitly: People just do like others do, and nothing needs to be said about it. Speakers of English have agreed to call a cat a cat by doing so regularly.

The purpose of conventions is to coordinate behaviour in a certain type of situations (Lewis 1974: 5, Bartsch 1987: 174) by deciding what is right behaviour in these situations and making people behave in the same way*.* For example, it is practical to agree on driving on either the right or the left side of the road, instead of having to negotiate every time you meet somebody. Or on what a certain word means and how it is pronounced and spelled. Communication is obviously a situation that calls for coordination and agreement: «...there must, of course, be agreement between producers and receivers if people are to understand each other» (Corballis 2002: 112). To communicate well we have to use the same linguistic units, e.g. the same words. The more similar our usage is, the better we understand each other.

Conventions may be formulated, usually in writing, in laws and regulations like the constitution, the ten commandments or traffic regulations, and there may be fixed penalties for not observing them. Regulations for language by authorities or academies usually regulate only parts of written language, especially the spelling of word forms and thereby the inflectional system. Spoken languages are regulated only tacitly by the users.¹⁰ Such regulations are sometimes called «established norms», e.g. by Vikør (2007: 74), but if norms are regularities in behaviour, they cannot be norms, but are just prescriptions, trying to establish (or change) norms. There is agreement that norms must be followed and govern usage to be valid (Haas 1982: 5, Bartsch 1987: 177, Gullvåg 1991: 18). They must be what Dyvik (1998: 21) calls «operative» and exist «…by virtue of regulating people’s usage». Prescriptions are not always followed in usage, so we can get a contrast between a language as it should be and as it is. This is often the case in a regulated standard language like Norwegian Bokmål.

So linguistic conventions exist just as usage, and languages are simply the units and rules that are used regularly – once or twice does not make a convention. As Bybee (2010: 214) points out, «...high frequency of usage leads to conventionalization». Croft & Cruse (2004: 292) explain this by pointing out that use of a unit activates connections in the brain, and frequent activation leads to «...its ultimate storage as a conventional grammatical unit». That’s why frequency is important both in languages: Frequent grammatical words are usually short, often also irregular, and in language acquisition and change: Frequent words are learnt earlier and are more resistant to change. That is also why usage is important: Every time we use a language or a linguistic unit we strengthen not only our own competence, but also that of others, and thereby the actual language or linguistic unit. Reduced use of a language will eventually lead to its death, so usage is a question of life and death for languages.

5.4.2 Linguistic variation

As conventions are valid in a certain community, linguistic conventions may vary, and usually do vary, between language communities that have little or no contact and communication. But even in the same language community there is variation and varieties. Most communities consist of local or social subgroups with somewhat varying languages, e.g. between older and younger members. And even in the same subgroup there is variation. We often disagree on what words mean, and we don’t have exactly the same lexicon: There are technical words that only some users know, like fricative or modifier. There may even be competing conventions for the same concept. An example is the many alternative word forms in the Norwegian written standards, sometimes between whole grammatical categories, such as Bokmål having either female gender with special articles (ei bok, bok-a, a/the book) or common gender with the same articles in feminines as in masculines (en bok, bok-en).¹¹ So collective does not mean that all members of a language community agree completely and that their usage is exactly the same, only that most of it is common. Linguistic communication can tolerate quite a bit of variation. Scandinavians can communicate fairly well by using their own languages, at least with some training.

As conventions have to be passed on through generations and different language communities and languages influence one another, conventions can be broken and changed. So we also have variation in time or language change if sufficiently many users break them in the same way. Boye & Harder (2012: 8) point out that linguistic conventions are «dependent on usage» and «can remain stable only if they continue to be reflected in actual usage», and may «change as a result of changes in usage». And conventions that go out of usage, also go out of the actual language: We have a «dead» convention or even a whole «dead» language. So linguistic conventions are «a plural and multiple entity» (Harder 2010: 278) and «…a natural reservoir of variation, with some features becoming less diverse and others becoming more diverse over time» (Deacon 1997: 114). This makes it difficult to delimit languages and e.g. say how many dialects there are in Norway or how many national languages in the world. Is Serbian and Croatian one language or two? It is social and political rather than linguistic facts that determine such questions.

5.4.3 Structural and functional conventions

Searle (1969: 33 f) and Dyvik (1993: 168 f) point out that there are two kinds of linguistic conventions: constitutive or structural, which constitute the language and are mandatory, and regulative or functional, which govern usage and are optional. Compare Vikør’s (2007: 72) distinction between grammatical norms and usage norms and Dik’s (1997: 3) between rules that govern «the constitution of linguistic expressions» and those that govern «the patterns of verbal interaction in which these linguistic expressions are used». We have the same distinction in chess and other games: Some rules constitute the game and must be followed, e.g. how the chess pieces can move. But where you put them, how you open the game, whether you play defensively or offensively etc., is up to you, as long as you follow the constitutive rules. One thing is to play chess correctly, another is to play chess well. The same goes for language. Some are better communicators than others, and most communities appreciate eloquence.

The structural conventions are the lexical, grammatical and phonological or graphological conventions that govern utterances in the language. They are unmotivated and may therefore vary between languages. The functional conventions concern how to communicate well in different situations, e.g. how to tell a joke, make a speech or write a scientific article. They are motivated by communicative considerations like Grice’s principles, and more or less universal, since usage has the same goal everywhere: to be understood and listened to. One has to express oneself clearly, unambiguously and truthfully, assess the receivers’ knowledge and not tell them something they know or use words they won’t understand, follow politeness conventions etc.

According to Dixon (2012: 442) all languages have conventions for addressing different types of receivers – strangers, superiors, small children etc., which may require special pronouns like the German and French plurals Sie and vous to adress one person. In conversation turntaking is essential and it’s considered impolite to interrupt others, in scientific texts one is expected to use references and list them at the end. There are norms for ellipsis, e.g. that we usually don’t repeat the subject in successive sentences with the same subject: I came, saw and won. We are supposed to be truthful, but there are «white lies» to spare the receiver for unpleasant truths. We can be ironic and say the opposite of what we mean, and hope the receiver will understand. One can break every one of these conventions and still speak or write grammatically, but perhaps not very comprehensibly, sensibly or politely.

5.5 Languages as sign systems

5.5.1 The primary function of languages

We have institutions for a reason. If languages are institutions, what is their purpose? That it is meaningful to speak of use of languages, shows that languages are tools, something that we use for a certain purpose. This implies a functionalistic view – that a language like other tools is «...determined by the functions that it is intended to perform» (Croft 1991: 1). Compare Halliday (1974: 13): «...we try to explain the nature of language, its internal organization and patterning, in terms of the functions that it has evolved to serve». The primary function of languages must be what we normally use them for: communication, more precisely linguistic communication, communication by signs (we also have non-linguistic communication, by both animals and people). And we communicate for many different purposes, compare the different «speech acts» of Austen and different uses of written language, from personal letters to scientific treatises.

This doesn’t mean that communication can explain everything in languages. It can only explain universal traits, like sounds, morphemes, words, grammar etc., but not which sounds, morphemes etc. languages consist of. That is decided by convention and history, e.g. that English was influenced by French after the Norman invasion. Nor does it preclude secondary functions. When we speak to ourselves or swear, we don’t communicate, but concentrate or give vent to feelings. Writing is indispensable as a memory support, as when preparing a speech, taking notes from lectures or making shopping lists. We can also think in language, e.g. in «imaginary conversations», and also have to think linguistically when we plan utterances. Language can also be used to «…store information or carry out thought processes» (Bickerton 1995: 11) and «…serves both to organize and to communicate thoughts» (Chafe 1998: 96).

Some linguists hold that language is primarily for thinking and that communication is «just one use to which language can be put» (Bickerton 1995: 40). However, we don’t need languages to think, as animals and small children show. But we need them for advanced or systematic communication, as anyone knows who has been in a strange country and tried to ask people the way to the airport. Communication is clearly the primary purpose, but signs can also help us concentrate and clear our thoughts because they have an expression. For example, it is easier to remember a large number if you say it, or even think its expression.

For systematic or linguistic communication we need signs: conventional combinations of a mental content or meaning – conceptions or ideas – and a physical expression like speech sounds, written characters or gestures. We need both minimal signs or morphemes like bird to express single ideas and a grammar to make complex signs expressing complex ideas by combining morphemes into complex words like blackbird and word forms like birds (morphology), and word forms into syntagms like a/the bird (syntax). And as signs need an expression, we also need an expression system that may be independent of the sign system, e.g. a sound system or an alphabet – double articulation (Martinet 1965: 2) or duality of patterning (Hockett 1960: 90). However, there is no independent meaning system – meanings are just part of signs, from morphemes to sentences (Papazian 2025 a: 13). So languages are sign systems consisting of subsystems: a lexicon, a grammar and a sound or writing or gestural system.

5.5.2 Utterances as the maximal sign and data for linguistics and language acquisition

Utterances, including sentences, must be regarded as the maximal sign of language because there are mandatory (constitutive) conventions for their structure: They must have a certain structure and may be (un)grammatical, (not) in accordance with the structural conventions of the language. According to Trager (1974: 84), «…linguistic analysis in the strict sense stops at the sentence boundary», and Crystal (2007: 249) defines sentence as «...the largest unit to which syntactic rules apply». Knowing a language means knowing how to make grammatical utterances in the language. For larger texts there are no constitutive norms; they are governed only by functional – pragmatic and rhetorical – norms. Compare Matthews (2001: 28): «Within sentences relations are part of the language system. Across sentences they are not». Such texts cannot be ungrammatical, but only semantically inadequate – incoherent, incomprehensible, boring, silly, false etc., and are not conventional signs.

The structural conventions govern utterances by defining the structure of utterance types or constructions, e.g. interrogative sentences. Thereby they make an unlimited number of utterances – examples of the constructions – possible or grammatical. When we know them, we can decide if an utterance is grammatical, in accordance with the conventions, i.e. if it is a possible utterance in the actual language. Therefore, utterances are the decisive data of linguistics, showing what the constitutive norms of the actual language are: «Any sound linguistic theory must be based on concrete utterances of speech» (Vachek 1989: 2). They are also data for those who are trying to learn the language. In addition, there are dictionaries and grammars, and teaching of language in schools. But when we learn our first language, we have no other data than others’ utterances, and have to generalize from them to the structural conventions that the utterances follow and the constructions they are examples of.

But as individual products utterances may be ungrammatical, either because the sender has made a slip or because (s)he has imperfect competence in the actual language. This is especially the case in speech, which is usually improvised and full of unfinished utterances, repetitions, corrections or plain mistakes, e.g. by foreigners or children. Therefore we must restrict ourselves to grammatical utterances: «The corpus […] must be ‘cleaned up’» (Atkinson et al. 1991: 35). The use of languages must be restricted to the conventional use, i.e. usage, exactly to avoid individual deviations from the structural norms.

According to Atkinson et al. (1991: 38), this means that intuitions are data. But intuitions are unobservable and cannot be data for others. Dyvik (1995: 26) says that data for linguists are the informants’ report of their intuitions. But intuitions are «tacit» or unconscious knowledge (Atkinson et al. 1991: 43), and such reports may be wrong and deviate from the informants’ usage, as Dyvik admits (1995: 27). In that case, we trust usage and think that the informant reported what (s)he wants to use or has learnt is «correct». So there is reason to be sceptical to informants’ report of their own usage, and grammatical utterances are the only reliable data for a language.

But how do we know that an utterance is grammatical? Here we depend on informants’ judgements. Dyvik (1995: 27) points out that even if informants seldom have trustworthy intuitions about the norms themselves, they usually have more trustworthy intuitions about «the norms’ consequences for individual sentences», i.e. they can intuitively decide if an utterance «sounds right» or is in accordance with the norms or not, usually without being able to explain why. That requires conscious knowledge and is the linguist’s task. A grammatical utterance must be an utterance which (a selection of) the language users accept as right or possible in the actual language. So informants’ judgements can delimit valid data – grammatical utterances which follow the structural conventions. And they govern only the structure of utterances, both the units that occur in them (like words or morphemes) and how they can be combined in the utterance types of the language. When we know these units and rules, we know which utterances are grammatical in the actual language and can make an unlimited amount of grammatical utterances.

5.5.3 One or several sign systems?

One of Saussures bestknown claims is that language is «form» (1967: 157), i.e. relations or functions, not «substance» or objects with qualities, and that in language «there are only differences without positive terms» (1967: 166). The expression is «not constituted by its material substance, but only by the difference between it and other sound images» (1967: 164). None the less he insisted that languages are spoken and described the expression of signs as a «sound image» (image acoustique). So signs have (phonic) qualities after all. As Vachek (1989: 106) points out, there is a contradiction here.

But Hjelmslev (1966: 93) took the consequence of Saussure’s claim by saying that the expression is neither phonic nor graphic (nor has any other material substance), but may be realized in any substance. Also Uldall (1995: 214) sees speech and writing as «expressions of one and the same language». This view seems to be quite widespread today, and many include gestural languages. According to Bickerton (2009: 3), «Speech is simply one vehicle for language. Another is manual sign». Also Bolhuis et al. (2014) regard speech and gesture as different «externalizations» of «language». According to Bouchard (2013: 146), the sign is «modality-independent», and Hurford (2014: 106) says that «…underlying the medium in which language is expressed, whether signed or spoken, is a system which is independent of the medium».

Here spoken, written and gestural usage are just «modalities» or expressions of the same language, which is neither spoken nor written nor gestural. In other words we have an abstract language, consisting of signs with an abstract expression. However, the claim that languages consist of differences and do not have any qualities, is untenable. Nothing can consist of functions or relations without qualities – there must be something, with specific qualities, that has a function and relations between entities with specific qualities. Differences can only mean ‘different qualities’, and relations presuppose entities with qualities, as Harder (1996: 26) points out (Hjelmslev’s view is critized also in Garvin 1954: 91 f, Eco 1984: 23 and Vachek 1989: 109).

The multimodal view also implies a favouring of the content: It is the content which is constant and defines the sign and the language, while the expression varies. According to Uldall (1995: 215), the fact that the pronunciation and the spelling of a word have the same meaning «…proves that they are expressions of the same language». But the same meaning does not imply the same sign: There are synonymous expressions, either in different languages, like German Ochs and French bœuf, or in the same language, like English ox and bull. Synonyms are different words though they have the same meaning, and signs are defined by both expression and content. They have a specific expression as well as a specific content – not any expression.

If so, also signs with different types of expression, like speech, writing or gestures, must be different signs. That gestural languages are sign systems of their own, may be obvious, but «verbal», i.e. spoken and written, languages, are parallel systems and therefore more easily seen as the same system. Lexico-grammatically, standard, orthographic writing and standard speech are close to identical, and grapho-phonologically they correspond more or less closely. The reason is that speech has been the model for writing, lexico-grammatically and in alphabetic writing also grapho-phonologically, with alphabets copying sound systems. And speech and writing have influenced each other, especially in standard speech and writing.

However, there are written signs that have no correspondence in speech and cannot be pronounced, like parentheses or quotes, and signs that are independent of speech and can be pronounced in any spoken language, like %, 9 or 1984. The alternative to Hjelmslev’s view, which especially Vachek (1989) has argued for (see also Coulmas 2003: 16–18), is to regard spoken and written languages as different sign systems. This means that to read aloud a written text or write down a spoken one is a translation, as Haas (1970) underlines. Although we have a general language faculty, there is no general language, only specific languages with different types of expression.

Summary

Usage is the conventional use of a language as sender. Using spoken or written language results in a text, either speech consisting of speech sounds or writing consisting of written characters. These units express the signs of the language, most importantly utterances, chiefly sentences, that are grammatically independent and have a communicative purpose or express a certain speech act.

We have a general language faculty, not limited to speech: the capacity to make, learn and use sign systems. It has evolved to use spoken language, but can be transferred to sign systems with other expressions, like writing and gestures. But to use specific languages we also need competence in the language: knowledge of its structure and skill in using it properly and purposefully, which must be learnt. We need both active competence to produce grammatical utterances and passive competence to understand them.

A usage-based approach to language means that languages cannot be mental and individual, as usage is physical. Expressions have to physical and signs have to be collective to function in communication, which is their purpose. So languages are cultural products: institutions consisting of conventions, both constituive or structural and regulative or functional. More precisely, they are open sign systems with a grammar for making complex signs. Utterances are the maximal sign, and grammatical utterances are the only valid data for the language. Signs are partly physical (the expression) and partly mental (the content) and have a specific expression. So there is no multimodal language, just different sign systems with different expressions.