You are here : Home Page / Theories / Jakobson / The Functions of Language

Available Languages

Follow us on

SignoSemio's Facebook SignoSemio's Twitter

The Functions of Language

By Louis Hébert

Professor, Université du Québec à Rimouski



Roman Jakobson

Jakobson's model of the functions of language distinguishes six elements, or factors of communication, that are necessary for communication to occur: (1) context, (2) addresser (sender), (3) addressee (receiver), (4) contact, (5) common code and (6) message. Each factor is the focal point of a relation, or function, that operates between the message and the factor. The functions are the following, in order: (1) referential ("The Earth is round"), (2) emotive ("Yuck!"), (3) conative ("Come here"), (4) phatic ("Hello?"), (5) metalingual ("What do you mean by 'krill'?"), and (6) poetic ("Smurf"). When we analyze the functions of language for a given unit (such as a word, a text or an image), we specify to which class or type it belongs (e.g., a textual or pictorial genre), which functions are present/absent, and the characteristics of these functions, including the hierarchical relations and any other relations that may operate between them.

This text may be reproduced for non-commercial purposes, provided the complete reference is given:
Louis Hébert (2011), « The Functions of Language », in Louis Hébert (dir.), Signo [online], Rimouski (Quebec),

An updated and extended version of this chapter can be found in Louis Hébert, An Introduction to Applied Semiotics: Tools for Text and Image Analysis (Routledge, 2019,


The well-known model of the functions of language introduced by the Russian-American linguist, Roman Jakobson (1960, pp. 350-377), can be disputed on several grounds from a theoretical standpoint. Our purpose in this chapter is simply to suggest a few ways of exploiting the analytical potential of this device. When we analyze the functions of language for a given unit (such as a word, a text or an image), we specify to which class or type it belongs (e.g., a textual or pictorial genre), which functions are present/absent, and the characteristics of the functions, including the hierarchical relations and any other relations that may operate between them.


We will mention just one point of controversy here, which is the number of factors (terms) and functions (relations between the terms) the model contains and the possible subtypes of any factor or function. Rastier (1997, p. 25) sees the metalingual function simply as a specific subtype of the referential function. Arcand and Bourbeau (1995, pp. 27-28) believe that there are two forms of the appellative function (conative function): in a "directive-appellative discourse, the sender leads others to act without justifying his will with arguments of any kind. In an argumentative-appellative discourse, the prompting [...] takes the form of an argument. The sender can give the pros and cons, defend his ideas and counter other people's ideas." (trans. of Arcand and Bourbeau, 1995, p. 28)


According to Jakobson, any act of verbal communication is composed of six elements, or factors (the terms of the model): (1) a context (the co-text, that is, the other verbal signs in the same message, and the world in which the message takes place), (2) an addresser (a sender, or enunciator ), (3) an addressee (a receiver, or enunciatee), (4) a contact between an addresser and addressee, (5) a common code and (6) a message.

Each factor is the focal point of an oriented relation, or function, that operates between the message and the factor. This yields six functions:

Factors of communication and functions of language

Target factor and
function no.




























Briefly, these six functions can be described as follows:

"(1) the referential function is oriented toward the context (the dominant function in a message like 'Water boils at 100 degrees'); (2) the emotive function is oriented toward the addresser (as in the interjections 'Bah!' and 'Oh!'); (3) the conative function is oriented toward the addressee (imperatives and apostrophes); (4) the phatic function serves to establish, prolong or discontinue communication [or confirm whether the contact is still there] (as in 'Hello?'); (5) the metalingual function is used to establish mutual agreement on the code (for example, a definition); (6) the poetic function (e.g., 'Smurf'), puts 'the focus on the message for its own sake' [(Jakobson, 1960, p. 356)]" (trans. of Tritsmans, 1987, p. 19).


Several competing names have been proposed for the "same" factors and functions. (A different name often indicates, insists on, reveals, hides, or even results in an important conceptual difference.) Some other names for the factors are (numbers refer to the table above): 1. referent, 2. sender or enunciator, 3. receiver or enunciatee, 4. channel. Some other names for the functions are: 1. denotative, cognitive, representative, informative, 2. expressive, 3. appellative, imperative, directive, 4. relational or contact, 5. metasemiotic (in order to extend the function to any semiotic act, such as an image), 6. esthetic or rhetorical.


In a proper analysis, we start by determining whether each of the functions of language is present or absent. Each factor must be present and concordant in order for communication to succeed. Consequently, relations are established between all of the factors, particularly between the message and the other factors. But here, we are interested in particular relations or functions. We will assume that while one or more – or even all – of the functions of language may be absent in short units (such as an isolated sign), lengthy units can activate all of them. Where more than one function is present, we will establish either: (1) a simple hierarchy, by identifying the dominant function and not ranking the other functions, or (2) a complex hierarchy, by specifying the degree of presence of some or all of the functions.


Various criteria can be used to establish the functional hierarchy. For example, Arcand and Bourbeau (trans. of 1995, p. 35) use an intention-based criterion: "The dominant function is the one that answers the question, 'With what intention was this message transmitted?' and [...] the secondary functions are there to support it." We must distinguish the intention associated with each fragment from the overall intention, which is "a sentence or series of sentences that corresponds to an intention" (1995, p. 27). Since the intention can be hidden, the function that is dominant in terms of overt degree of presence may not be dominant in terms of intention. Arcand and Bourbeau also distinguish between direct and indirect manifestations of intention, which correlate to the opposition between actual and overt functions. The appellative (conative) function is manifested directly in "Go answer the door" and indirectly in "The doorbell rang" (which is equivalent to "Go answer the door"), where the overt function is the referential (or informative) function (1995, pp. 30-33). In addition, we need to distinguish between cause and effect functions, as well as ends and means functions (the ends being the effect that is sought). For example, when the phatic function (cause) is overactivated, it can trigger the poetic function (effect); overactivation can be used for esthetic ends, and in this case the poetic function is an end and the phatic function is a means.


The functions of language can be linked to the various possible enunciative agents. In a literary text, for example, these agents are as follows: the empirical (real) author, the implied author (our impression of the author from reading his text), the narrator, the character, the narratee, the implied reader and the empirical (real) reader. (For more details, see the chapter on dialogics.) To take a simple example, in a disconnected interaction between characters, the disintegration of the phatic function (as when dialogue degenerates into parallel monologues) might correspond to a) symbolically, a phatic dysfunction between the empirical author and reader, and b) the poetic function being activated through the dysfunction between characters. In this case the phatic function is thematized, and it is fictional (it is operating between characters), and the poetic function is "real" (it originates from the real author and is meant to be perceived by the real reader). This thematized, fictional phatic function is thus a way of activating the poetic function in reality.


The presence/absence of the functions and their hierarchical structure can be used not only to describe units, but also classes or types of units (e.g., textual or pictorial genres). For Jakobson, what characterizes poetry and distinguishes it from other genres (literary and textual in general) is not so much the presence of the poetic function as its dominance. By identifying the functional configuration (e.g., by specifying the secondary dominant function) we can create a typology. Jakobson recognizes that epic poetry – focused on the third person, as opposed to lyric poetry (first-person) or poetry of the second person – "strongly involves the referential function of language" (Jakobson, 1960, p. 357)


Without going into all of the details, let us posit an "energy-based model" to describe the dynamics (in the literal sense) of the functions of language. (Dynamics is defined as "the set of interacting and opposing forces in a phenomenon or structure" [trans. from Le Petit Robert].) The strength of a function in a particular configuration (the target configuration) may be interpreted dynamically (and metaphorically) as resulting from an upward or downward "thrust" applied to the function as it appeared in a configuration (the source configuration) considered as the source of this particular configuration.

If we are going to analyze changes in the functional balance, this implies a comparison between two models, a source and a target. For example, in epic poetry (or narrative poetry) the referential function is intensified as compared to lyric poetry, while the emotive function is diminished.

This dynamic model seems useful from a descriptive standpoint. Certain works and certain genres appear to be based on emphasizing and/or deemphasizing (or even neutralizing) one or more functions. For example, hyper-realism in painting is an exaggeration of the referential function; pictorial abstraction and, in literature, the "destruction" of the Balzacian universe by Robbe-Grillet and Kafka are attempts to neutralize the referential function.


When relations between functions are studied, most analyses are limited to establishing a hierarchy. We would like to go a step beyond that. Let us posit that two kinds of correlations can be shown to exist between two functions. The correlation is said to be converse, or direct, if (1) an intensification of one of the two functions is accompanied by an intensification of the other and (2) a decline in one function causes a decline in the other. The correlation is said to be inverse if an intensification of one of the two functions is accompanied by a decline in the other, and vice versa.

We will sketch out a brief analysis of this type below. Generally speaking, when one function is accentuated, it tends to diminish the importance of all the others, and the opposite happens when the function is deemphasized. But we will also postulate that some functions are generally paired in an even more definite inverse relation. The most obvious pairings are the expressive and conative functions and the referential and poetic functions.


It is difficult to draw the line between interaction and merging of functions. Klinkenberg asks the question directly (trans. of 1996, p. 61): "Are functions actually distinct from one another?" It remains to be determined just how much is interaction and how much is commingling in each possible pairing of functions. Klinkenberg describes some of these pairings (1996, pp. 61-62). Let's look at one of them: the referential and conative functions. "Any information – the referential function – changes the receiver's knowledge stock; we can thus say that it acts on the receiver: that is the conative function. Moreover, a lot of so-called information leads to a behaviour as its final result. The sign 'falling rock' is meant not just to convey information, but most of all to elicit a certain attitude in the driver" (trans. of Klinkenberg, 1996, p. 61). The second interaction Klinkenberg mentions, as we will show, involves the opposition actual vs. overt function – in this case, the conative and referential functions, respectively.


The perfect addresser-message equivalence would have to be the spontaneous cry of pain. Even though the cry may be "addressed" to a receiver, it is associated almost consubstantially with the addresser, thereby leaving the conative function empty, so to speak. Conversely, an educational message is intended for the addressee, and generally entails an attenuation of the emotive function (when the emotive and conative functions are incompatible, at any rate).


Jakobson appears to recognize the relation between the poetic and the referential functions, since he places them in a sort of battle for supremacy (1960, pp. 370-371):

"Ambiguity is an intrinsic, inalienable character of any self-focused message, briefly a corollary feature of poetry [...] The supremacy of poetic function over referential function does not obliterate the reference but makes it ambiguous. The double-sensed message finds correspondence in a split addresser, in a split addressee, and besides in a split reference, as it is cogently exposed in the preambles to fairy tales of various peoples, for instance, in the usual exordium of the Majorca storytellers: 'Axio era y no era' ('It was and it was not')."

We will call this relation an inverse correlation. The more the message "talks" about itself and refers to itself (the poetic function), the less it talks about the context and refers to it (the referential function) and vice versa.


The functions (and the factors) do not necessarily all operate on the same analytical level. The poetic function in particular can operate at least partially on a second level, as the beneficiary of certain transformations in the functional balance, especially if they are marked (Klinkenberg, 1996, p. 58). In this case, the poetic function is linked to the other functions by a non-symmetrical relation. We will call it the ascending correlation: The poetic function is intensified as a result of a significant and selective strengthening or weakening of any other language function, but the reverse is not necessarily true (e.g., the emotive function is not necessarily strengthened by intensifying the poetic function). It is hard to imagine that a marked emphasis or attenuation in one function would not draw attention to the message itself, at least in some cases. This is what happens when the phatic function breaks down in Ionesco's The Bald Soprano, developing into parallel monologues as opposed to real dialogues. This contributes to the poetic effect of the play. It remains to be seen whether all variations in the poetic function necessarily result from a change, either qualitative or quantitative, in one or more other functions. Moreover, the poetic function is not necessarily the only one affected in cause-and-effect relations with one or more other functions. For example, when the poetic function is suddenly accentuated, the result and/or effect may be to keep the addressee's attention at a time when it was beginning to wander (the phatic function).



Let us incorporate Klinkenberg's suggestion to extend the range of the emotive function (1996, p. 53): "The expression of 'emotive function' (which could more aptly be called the 'expressive function') should not be understood in the usual sense, as referring to human affect. It actually has nothing to do with emotion. Any message, including the most neutral, reveals the condition of its sender." Even the crackling of a defective electrical part on a stereo system reflects this function, since it indicates the stereo's poor condition.


Rather than using the term "expressive function", we could approach this function in terms of symptoms and indices. As Rastier points out,

"The main contemporary representations of linguistic functions are based on the sign model presented by Karl Bühler. The sign functions as such through its relations with the sender, the receiver (Empfänger), and the referent (Gegenständen und Sachverhalten). Relative to each of these three poles, the sign pertains to a different semiotic type: it is a symptom [an index] in relation to the sender, a signal in relation to the receiver, and a symbol in relation to the referent" (Rastier, 1997, pp. 24-25).

Any semiotic act, then, is indexical in relation to its producer (the expressive function) and a means of signalling to its receiver (the conative function). We could add that it is also an index of the state of the other factors and of the mental image that the message's producer makes, rightly or wrongly, consciously or unconsciously. For example, a written message containing the word "loose" instead of "lose" (as in "Did you loose your keys again?") – a common lexical error – results from a warped image of the language code. We will add that while the relation between the sign and the referent is indeed symbolic in nature, at least from Bühler’s perspective, this symbol can function as an index (someone – the sender – gives us a garment belonging to the loved one), an icon (someone gives us a photo of the loved one), or a symbol (someone gives us a text describing the loved one). (See the chapter on Peirce’s semiotics.)


Jakobson remarks that the context is what is known as the " 'referent' in another, somewhat ambiguous, nomenclature" (1960, p. 353). Amazingly, this does not stop him from using the term "referential" for the function whose target factor is the context. Moreover, the term "context" is no less ambiguous, both in general and in this particular case. Jakobson says that the context is "either verbal or capable of being verbalized". As for the referential function, Jakobson gives the synonyms "denotative" and "cognitive" (1960, p. 353), but unlike all the other functions, this one is not presented in detail, and seems to be taken for granted. We believe that there are two main ways of interpreting this function in the work of Jakobson and those who use his model.

1. The referential function relates to the thing "spoken of" (Jakobson, 1960, p. 355).

2. The second way of viewing the referential function seems more useful and operative than the first. The referential function is associated with an element whose truth value (true or false status) is being affirmed (or questioned), particularly when this truth value is identical in the real universe and in the assumptive or reference universe that is taking it on.

This calls for some explanation (for more details, see the chapter on dialogics). A universe of assumption (such as the universe of a character in a literary work) may be reinforced or contradicted by the universe of reference (as defined by the omniscient narrator, for example), which stipulates what is ultimately true or false (or undecidable) in the more or less "realistic" universe constructed by the semiotic act. So the statement "the sun rises in the East" – which is true in reality and in a realistic text – would be more of a referential assertion than "the sun rises in the West", which would be perceived as somewhat poetic, in that the incongruity draws attention to the message (even if the utterance is true in the universe of reference, say, of a science-fiction novel).


Jakobson says that unlike declarative sentences, imperatives (linked to the conative function) cannot be tested for their truth value. The imperative "Drink!" "cannot be challenged by the question 'is it true or not?' which may be, however, perfectly well asked after such sentences as 'one drank', 'one will drink', 'one would drink'." (Jakobson, 1960, p. 355) Considering that declarative sentences clearly activate the referential function, then the "truth value test" becomes a test we can use to identify the referential function.


Firstly, we propose making the metalingual function into a more general "metacode" (or "metasystem") function. This will allow us to apply it to non-linguistic "messages". Secondly, we propose recognizing any normed and norming system as a code, and not restricting ourselves to the language code where text is concerned. Rastier takes the view that a text is the result of three systems interacting (1994, pp. 222 and 224 and 1997, pp. 27-29): (1) the dialect (the language system), (2) the sociolect (the particular usage of a dialect specific to a differentiated social practice with its own discourse organized through genres), and the idolect (a given author's individual usage of a language and a sociolect). In this case, utterances like "A Sonnet has 14 lines" (which deals with a genre, or sociolectal phenomenon) and "Baudelaire liked antitheses" (which deals with an individual's style, or idiolectal phenomenon) are as much about a code as "How do you spell 'surreptitiously'?" (which deals with the language system).

The examples given above use a thematized metacode function, embedded in the signified, or content. However, we should expect that non-thematized metacode functions may also exist. When the code norm is transgressed, attention is directed indexically, but clearly, to the code, as in: "The wind, he blow." Or if a given sonnet simply deviates from the norm, doesn't this evoke the model sonnet by contrast, and thus the system that defines it?


* * *

[If you drink, then drive, you’re a bloody idiot]

If you drink, then drive, you’re a bloody idiot

* * *

The poster advertisement [If you drink, then drive, you’re a bloody idiot] is part of a series of advertising campaigns based on the same slogan and launched in 1989 by the Australian Transport Accident Commission (TAC). The image was reproduced in Touring (2002), Laval (Québec), 80, 2, summer, p. 33). 

Generally speaking, the advertising message has to accomplish the following, in three successive stages: (1) attract attention (the phatic function), (2) convince (the conative function), by appealing to reason (the referential function) or emotion (the emotive function), and (3) get people to act (the conative and referential functions). The third objective is clearly the most important, and the others are subordinate.

Two actions – drinking and driving – are combined into sequences: this is an attack, not on drinking-and-driving, but on the act of drinking-then-driving, which is more commonplace. Three possible sequences (the referential function) are open to the addressee (the conative function): (1) not drinking, then driving, (2) drinking, then not driving, (3) drinking, then driving. While neither action is good or bad in itself (with a possible nuance for excessive consumption of alcohol), all of their possible sequencings are given a moral value: the first two scenarios fall under good behaviour, and the third comes under bad behaviour. The advertising message clearly takes aim at the third scenario. It does this by showing the possible dire consequences – the addressee's death – in a very striking way (the emotive and poetic functions). This is not the death of some other person, be they a stranger or a loved one (these two scenarios, which appeal to the drunk driver's sense of guilt, appear in other messages by the same organization); this is the worst possible death: yours (the conative function). In other words, this is not a referential third-person death, but a conative second-person death. This death is the concrete (pragmatic, in Greimas' terms) punishment for – or at least the consequence of – not toeing the mark, not keeping the contract contained in this ad. Likewise, the symbolic (cognitive, in Greimas' terms) punishment is being called an "idiot".

This death is presented as being highly avoidable, since it is reserved for the "bloody idiots" with whom no addressee with any glimmer of intelligence would want to associate. The word "bloody" indicates the level of idiocy within the class of idiots, and at the same time it demonstrates the intensity of the addresser's emotion (the emotive function); note that there is no exclamation point, which would have emphasized the expressive function. Perhaps the addresser is highly concerned about what could happen to us (the conative function), or perhaps his utterance merely expresses a coldly objective truth (the referential function) along with an unsympathetic "too-bad-for-you" attitude. In addition to the standard meaning, indicating intensity (the expressive function), possible concern (the expressive function) and familiarity (the conative function), "bloody" happens to be a polysemic word, and thereby draws attention to itself (the poetic function). It alludes to blood – the blood we will shed, but also the blood that shows our blood alcohol level. Speakers of English no longer make the connection to blood when they say "bloody", just as speakers of French (in France) no longer make the connection to a hooker when they use "putain" as an interjection. By re-actualizing the original content, the slogan de-automates the use of this word, drawing our attention to an otherwise innocuous, transparent word. Moreover, "bloody" is a term used in the names of drinks like "bloody Mary" and "bloody Caesar". It stands in opposition to "virgin" (virgin Mary, virgin Caesar). "Bloody" indicates an alcoholic drink; "virgin" indicates a non-alcoholic drink. So "bloody idiot" roguishly suggests a new kind of alcoholic drink.


ARCAND, R. and N. BOURBEAU, La communication efficace. De l'intention aux moyens d'expression, Anjou (Québec): CEC, 1995.
JAKOBSON, R., "Linguistics and Poetics", in T. Sebeok, ed., Style in Language, Cambridge, MA: M.I.T. Press, 1960, pp. 350-377.
KLlNKENBERG, J.-M., Précis de sémiotique générale, Paris: Seuil, 1996.
RASTIER, F., Meaning and Textuality, trans. Frank Collins and Paul Perron, Toronto: University of Toronto Press, 1997 [1989].
TRITSMANS, B., "Poétique", in M. Delcroix and F. Hallyn (dir.), Méthodes du texte. Introduction aux études littéraires, Paris: Duculot, 1987, pp. 11-28.


Which functions of language are activated in the following text?

"This text you gave me to correct is a bunch of rubbish! Listen to this, you've got several verbs with no subject, you state the obvious ('a day lasts 24 hours'!), then – are you still following me? – you use obscure metaphors ('work is the drop hammer of life') and stupid malapropisms ('You are the suntan of my life').”

Share this page