Guided commentary on de Saussure's Course in Modern Linguistics

by Suzanne Kemmer

Introduction. Chapters 1-5, pp. 1-23

Chapter 1: Saussure's glance at the history of Linguistics

Saussure first takes a bird's eye view of the field of Linguistics. He sees the field as progressing through a number of stages, three of them older and one current. These stages are:

  1. Traditional grammar. Essentially, the logic-based analysis of languages, primarily for prescriptive purposes. Saussure includes under this rubric every European linguistic tradition from the Greeks to the end of 18th century, when French grammar was ascendent.
  2. Classical philology. The close study of written texts of antiquity to understand literature and culture of the classical civilizations of Greece and Rome.
  3. Comparative philology. The discovery and illumination of historical relationships by comparison; genetic classification of languages; and by the mid-19th century, reconstruction.
  4. Linguistics proper. Begins essentially with the Neogrammarians and was the contemporary modern-day approach in Saussure's time. (There were still representatives of all three of the earlier traditions around, however.)

As in all the preceding periods (although he doesn't mention it), the focus right up to Saussure's time was squarely on morphology. Syntactic patterns and problems were sometimes treated, as in Delbrück's volumes on the syntax of the Indo-European languages, but morphological patterns were considered the heart of language, and certainly of reconstruction, due to the structure of the Proto-Indo-European word.

Saussure states what he sees as the basic problems of comparative philology: The scholars, according to de Saussure, had no real historical sense. They had strange ideas about language change due to their fixed belief that languages 'progress' or 'decay'. Specifically, the early comparativists thought that ancient languages went through stages of development; and then after that, the languages began to decay, and modern languages are essentially degenerate versions of earlier ideal structures. (We saw the Neogrammarians in the Osthoff and Brugmann article making the same complaint about the same 'early period' of comparative linguistics.) These scholars failed to understand how language typological facts relate (or don't relate) to language classifications and the ways the languages got to the way they are now (the present systems need not directly correspond even though the etymological units do). In effect, they did not understand the relation between typological and genetic classifications and the historical basis of each. (Diachronic typology did not develop until Greenberg. )

Moving to Stage 4, the Neogrammarians were the first to "seek out the principles that govern the life of languages." They discovered that "similarities between languages are only one side of the linguistic phenomenon.[;]..comparison is only a means or method of reconstructing the facts." They put comparative studies in their "proper place", i.e. their correct historical perspective. They realized that language is not an organism that develops independently of people, but only as a product of the collective mind of linguistic groups.

There was still a focus on the grammatical forms of the parent and daughter languages, that is, specifically on morphological categories. Yet in this period, individual sounds suddenly take on a much greater importance. The interest in sound change began as an attempt to account for how the morphological forms and categories had changed so that the more recent languages had lost many of the distinct categories of the proto-language. But now, with the intense interest in regular sound laws, scholars began to think about the sounds of the language as though they were isolated elements rather than merely aspects of morphemes.

The study of phonetics began to get going in a very systematic way in the late 19th century. Phonetics and phonology were not distinguished then, but what was being studied was primarily the articulatory apparatus, so we would be more likely to call their object of study phonetics. The new ways of talking about the various classes of sounds in the fledgling field of phonetics, namely as distinguished by places and manners of articulation and by voicing, also helped researchers to think of the changes in the Indo-European languages in terms of what had happened to individual sounds in getting from the proto-language to the modern languages.

The result was an intense atomism or particularism at the height of Indo-European study at the turn of the 20th century, such that compendia of Indo-European knowledge could simply exhaustively list individual sounds and all the things that happened to them in the course of the histories of the various descendent branches and languages. Saussure does not mention this, but I believe it made the novelty of his proposal to treat language as a system, and not as a collection or inventory of individual pieces, even more striking than it would have been before the Neogrammarians.

Saussure wants to focus on a different set of dissatisfactions he has with the study of Linguistics, however. He makes it clear that Stage 4, although representing true Linguistics, is still in his view not a proper science of language. Despite their breakthroughs and move toward a more scientific methodology in reconstruction, the Neogrammarians never addressed the fundamental problem of what language IS. What follows in the Course is Saussure's contribution to understanding the nature of language and the kind of field Linguistics needs to be if it is to be scientific, realistic, and if it is to advance as a discipline.

The subject matter and aims of Linguistics; relations to other fields (Chs. 2 and 3)

Chapter 2: Saussure's view of what Linguistics should be

Here Saussure outlines what he sees as the proper scope of Linguistics. It includes:

  1. genetic classification, historical description, and reconstruction for all languages.
  2. discovery of universal 'forces' shaping language; generalization beyond historical peculiarities
  3. definition of linguistics as a field, delimitation of its subject matter, relation to other fields

The Neogrammarians, of course, focused their attention on the first goal. Osthoff and Brugmann (1878) had made the point that no languages were superior to others in bringing historical evidence to bear on Indo-European; the modern dialects should be studied as much as the ancient languages. Saussure's idea was just as uniformitarian, but was no longer pointing just towards the sameness of historical processes. His implication seems to be modern: no languages should be studied because we prefer their cultures, while others are left out, due to the lower social status of their speakers.

The second aim in the list points toward the search for generalization which is the hallmark of any science. The Neogrammarians certainly did attempt to generalize. Their aim was historically accurate generalizations of the processes of change that led from the parent language to the ancient written languages to (ideally, but hardly in practice) the modern languages. Furthermore, the founders, at least, made strong general claims about the exceptionless nature of sound change, and how it interacted with analogy.

Still, many linguists, looking back to Saussure's day, find that the state of Linguistics then seemed to have become hopelessly particularistic. As indicated above, it seemed like there were enormous lists of facts about what happened to this or that sound in Proto-Indo-European. In other words, scholars were tracing historically-particular paths and not going any further in generalization than what happened in a specific language and what order particular sound changes happened in.

Saussure, I think, wanted to leave this particularism behind by ditching the viewpoint of 'historical paths' of specific, unconnected elements like individual sounds, and turn toward something that seemed to him more important, and more real in speakers' minds, and furthermore, more connected with meaning, which certainly had gotten quite lost in the Neogrammarian program.

A key point in Saussure's new way of looking at language is that "Everything in language is basically psychological." (p. 6) So instead of the 'blind' articulatory processes that were assumed to constitute sound change (contrasting with the 'psychological' but irregular and heterogenous processes of analogy), he says we HAVE to focus on the psychological in language because that is the crucial aspect of language.

Moreover, "The thing that constitutes language is...unrelated to the phonic character of the linguistic sign." This was a very surprising claim given that right back to Greek philosophy, scholars had assumed (as in any speaker's folk idea of language) that the forms of words were a crucial element in language. In contrast to this common assumption, Saussure tells us later in the section that the only relevant aspect of sound is the 'sound image', an essentially psychological and not physiological/articulatory element. His purpose here is not to tell us that phonetics is useless or unimportant as an area of study, but that it is only indirectly related to the symbolic function of language--and that function is carried by the linguistic sign, an essentially psychological (now we would say cognitive) unit. The precise sounds a given sign happens to have as its 'sound image' part are unimportant--all that matters is that it have SOME phonetic form that people can learn and they they can call up for use in communicating.

The next point Saussure addresses is the importance of Linguistics. Linguistics, says Saussure, because of the universal centrality of language to human life, should be of interest and importance to everyone. But no other field, he points out, is so full of mirages, fictions, and incorrect folk beliefs. (Of course we can add that this is still true today in the 21st century; everyone considers him or herself an expert on language by virtue of knowing one.) Saussure calls for linguists to try to get rid of these misconceptions and disseminate a scientically correct view of language. This, he says, is one of the important aims of Linguistics.

Many linguists have taken this aim to heart. Much of the mission of modern undergraduate Linguistics programs is to 'debunk' myths about language. Bringing correct knowledge about language to ordinary people was begun before Saussure by Max Muller and William Dwight Whitney in the 19th century, but in the 20th century Saussure's influence in the work of most linguists writing for a general public is clear. Edgar Sturtevant, Edward Sapir, Leonard Bloomfield, and Charles Hockett all aimed to correct mistaken ideas about the nature of language on the part of ordinary people. In the modern world, we have Steve Pinker, who in fact reiterates many of the same observations made by these others, right back to Saussure.

Chapter 3: The Object of Linguistics


1. Defining Language

First Saussure tells us why defining and studying language is so hard. One reason is, that the objects of study in language are not given in advance. In other sciences, the facts, the objects of study, are entities already existing prior to their observation. These pre-existing facts can therefore be considered from different points of view. In Linguistics, however, the objects don't antedate the viewpoint: "it would seem that it is the viewpoint that creates the object" (p.8). Worse, there's no objective way to determine which point of view, if any, is 'correct'.

No matter the viewpoint we take, he stresses, the linguistic phenomenon always has two interdependent sides, like a piece of paper. No matter what aspect of language we consider, there is some opposition of two aspects, each of which 'deriv[es] its values from the other' (p. 8). One example of this two-sidedness is the case of acoustic perceptions and articulatory productions. An articulated [n] by itself, unperceived, is nothing. We can't reduce the unit to just its sound, or just its articulation. Both are required for it to exist as the kind of thing it is.

But let's say we recognize the two-sidedness of the sound units. It turns out, he shows us, that even when you put together sound and speech as an 'acoustic-vocal unit', it still isn't language. There is something essential missing. We have to combine the complex sound unit with an idea, a unit of thought, to form a "complex physiological/psychological unit". (p. 8)

Yet even this is still not enough to characterize the basic objects in language. We have to consider that language has both an individual and a social dimension. Neither can exist without the other.

Then, there are still two more planes or dimensions: the synchronic and diachronic. "Speech always implies both an established system and an evolution; at every moment it is an existing institution and a product of the past."(p.8) It is not easy in fact to disentangle these two things.

For all these reasons, it is hard to find the 'true object' of linguistic study, without risking ignoring something fundamental.

What's the solution to this conundrum? Saussure sees only one way out: Forget trying to find some all embracing science that includes all of these dimensions at once. How can we do ethnography, sociology, phonetics, philosopy, and psychology simultaneously?

Instead, he says, let's study a carefully delimited thing called a Language [langue], separating this essential core away from everything else relating to speech. Aspects of speech are merely derived from language in some way. Language is primary, speech is secondary. We have to always look at Langue instead of langage.

Langage : Saussure's first English translator, Baskin, translates this as "human speech". But other Saussure scholars, including Harris, suggest it should be translated as "human language" in general. Langage in this generalized sense, according to Saussure, is many-sided and heterogenous; it straddles many fields and kinds of phenomena--physical, physiological, psychological; both individual and social. Trying to study langage or language in general, in one all-encompassing discipline which simultaneously has to deal with all these disparate types of phenomena with no overarching aim would, in Saussure's view, lead to chaos; and indeed it did at that time, given what he saw as a confusing maze of studies that did not fit together comfortably.

Langue: This is best translated as "a Language". I capitalize Language to show that Saussure means this in a technical sense; it is not a language as ordinary people normally think of languages, for example in terms of the folk theory of a language as a collection of names pointing to things, or considering the facts of articulation as central, which most people do. He means by Langue a language in the sense of a system of linguistic social conventions. A Language is unitary. It is self-contained, a whole. It is a principle of classification (in modern terms, a categorization system.) It is a social product of human speech and a collection of conventions adopted by a social body to facilitate speech. It is acquired and conventional, and not strictly instinctual or natural. Humans have to learn it, unlike most of the animal communication systems that we know about.

Because a Language is a convention, Saussure points out, the exact nature of what is agreed on doesn't matter. This means that vocal speech and auditory linguistic perception are secondary. This was a revolutionary view at the time but is now accepted by all linguists. The fundamentally identical linguistic nature of signed languages such as American Sign Language, British Sign Language, French Sign language, etc., supports the idea. All these signed languages are conventional systems with conventional units, just like spoken languages except for the fact that instead of a 'sound image' there is a 'gesture image'.

Saussure says that what is natural to mankind is not oral speech but the faculty of constructing a language--in other words, a system of distinct signs corresponding to distinct ideas. He says, ..."Beyond the functioning of the various organs there exists a more general faculty which governs signs and which would be the linguistic faculty proper." (my emphasis).

So, the Saussurean Sign is the basis of what for him should be the new field. It is the true object of study of the field of Linguistics.

Next Saussure goes on to try to define the Sign as clearly as possible. To do this he needs to separate the true realm of the Sign, so to speak, from the extraneous aspects commonly thought of as essential to the nature of language, but in his view are really not. That true realm is a Language (Langue)--a shared system of linguistic signs.

2. Separating Langue (a Language) from Parole (Speech)

Here Saussure presents an analysis of how a Language relates to Speech in a communicative act. First he examines an "individual act" of Speech. This is an act of a complete transfer of communication between two interlocutors. It apparently consists of more than Langacker's notion of "usage event" which has some things in common with the act of Speech--for example, it is situated in a speech context--but the Langackerian concept lacks an explicit social dimension.

Next Saussure describes "the speaking circuit" in a conversation. It includes actions involving the following body parts (where S and H are Speaker and Hearer): (S) brain, (S) mouth, (H) ear, (H) brain, (H) mouth, (S now H) ear, (S now H) brain. To be more technically precise, 'mouth' really refers to the whole speech-producing vocal tract.

  1. A concept evokes a sound image (linguistic acoustic image) in the brain. (Psychological)
  2. The brain transmits an impulse "corresponding to the image" to the articulators. (Physiological) (Here we have a close connection betweeen perceptual and motor cortex, anticipating modern neurology.)
  3. Sound waves produced by articulators travel from mouth to ear. (Physical)

Then the "reverse process" takes place:

  1. The ear picks up the sound. (Physical) (de Saussure doesn't actually mention this explicitly)
  2. The sound (air vibrations) physically acts on the parts of the inner ear so as to produce activations of the sensory nerves, which then arrive at the brain (Physiological) and create a perceptual sensation --the 'sound image'. (Psychological)
  3. The sound image evokes its associated concept. (Psychological)

Saussure notes that the sound image is a complex psychological thing, and can actually be split down further into the following components:

  1. A pure acoustic sensation (a Perception); this must occur in the auditory cortex somewhere
  2. The identification or linking of this acoustic perception with "the latent sound image" (in more modern terms, a memory trace of a linguistic event (modern Psychology); an entrenched unit (modern Cognitive Linguistics).
  3. "muscular image" of phonation. We could call this a "motor image". It would involve the entrenched motor routines of speech production that are psychologically linked (associated) with the perceptions of specific sounds. The psychological association lets us recognize specific sounds that are produced by specific articulations of the vocal tract.

Recent neurological findings show that motor neurons linked to the receptive sensory cortex are activated simultaneously with analogous perceptual neurons. That is, if you see a particular action, there is some activation of the parts of your motor cortex that would be activated if you carried out the action. If this is true for the sensory and motor cortex involved in language processing (and it is hard to see why it would not be--language processing just uses subareas of the general sensory and motor cortex, linked to areas more centrally involved in language processing), then Saussure's theory makes perfect sense in terms of modern neurology.

Like Langacker, de Saussure uses "image" much more generally than just for visual perception. It is for any kind of perceptual 'image' (really a psychological gestalt, arising ultimately from perceptual stimuli in learning, but at some point acquiring some existence apart from the original perceptual instigation).

Saussure stresses the separation of the perception (which is physiological or biophysical) from the purely psychological sound and motor images. These images, he insists, are as psychological as concepts. In modern terms, the sound and motor images are cognitive phenomena; they are constituted by cognitive processing. They belong to what Langacker calls "phonological space". The link to the concept it evokes is a psychological association. The concept belongs in what Langacker terms "conceptual space".

Going beyond isolated sounds, we also need to invoke some way of putting them together: this requires what Saussure calls the "associative and co-ordinating faculty". These are similar to Langacker's notions of 'correspondence' and 'composition': the mechanisms we have for coordinating items which fill different roles, and mechanisms for putting together simpler units into more complex units.

The associative and coordinating faculty is important for the social side of the act (we don't generally communicate in single signs; we typically make more complex messages), but Saussure doesn't say much more about it than that.

There must be some kind of average or common sign-concept relations that individuals in a community share.

Next, Saussure asks himself, which parts of the speech circuit are most involved in the "social crystallization of language"? "Social crystallization" seems to be another way of expressing the idea of the inventory of shared signs --they are social, and not individual. Saussure answers the question as follows.

Much of the speech circuit is not in fact part of a Language (i.e. linguistic system or Langue). The sound part of it is surely excluded from the Language, because the sounds of any foreign language don't give us any information; they don't 'unlock' the sign for us, i.e. we are not brought into the social aspect of language, because we can't understand them. And also the executive, or actively producing, side of the circuit is not part of the sign system, either, because this execution stage happens ONLY at an individual level. This is Parole par excellence, and not Langue at all. Langue belongs to the level of the speech community, not the individual level.

Psychologically speaking, says Saussure, the individual's faculties build up "impressions that are perceptibly the same for all". Harris' translation renders these psychological elements as a "stock of imprints" of which each person has his/her own collection, essentially the same as the next person's. Notice that inside the individual, we don't have the Signs, we just have the individual imprints of them--they are just word patterns, not signs. The SOCIAL product constituting Langue is something else. Saussure asks, "how may we envision this?" Here's how he does so, in a key passage:

If we could embrace the sum of word images stored in the minds of all individuals, we could identify the social bond that constitutes Language (Langue). It is a storehouse filled by the members of a given community through their active use of speaking, a grammatical system that has a potential existence in each brain, or more specifically, in the brains of a group of individuals. For Language is not complete in any speaker; it exists perfectly only within a collectivity. (p. 13) (emphasis mine)

The Harris translation makes it perhaps come out even a bit more clearly what Saussure has in mind:

If we could collect the totality of word-patterns stored in all those individuals, we should have the social bond which consitutes their language (Langue). It is a fund accumulated by the members of the community through the practice of speech, a grammatical system existing potentially in every brain, or more exactly in the brains of a group of individuals; for the language is never complete in any single individual, but exists perfectly only in the collectivity.

Thus the Language (Langue) is larger than any given speaker; it is a conventional system, in effect bonding together speakers in a community. It exists and is real, but it exists OUTSIDE the speaker. It is a system of linguistic potentials that can be accessed by actual brains in speaking, but is not actual itself AS A SYSTEM OF SIGNS in individual speakers. Whatever the speakers have in their heads is not the system of signs Saussure posits as the object of the scientific study of language.

The view of the sign system by most linguists since Chomsky, including the Cognitive Linguists, is quite different. For Chomsky AND Langacker, the sign system resides in a speaker's mind and nowhere else. It is indeed a system of linguistic potentials, and both Chomsky and Langacker, in different ways, are interested in the linguistic potential, i.e. what a person COULD say that would be licensed by their knowledge of the patterns of language (in Chomsky's terms, "rules", in Langacker's, "schemas").

These two more "mentalist" views diverge in that for Chomsky, ALL you need to do is study the potentials, which have a very indirect relation to use--and language use itself can essentially be ignored. Langacker, on the other hand, insists that the linguistic potentials, the possible productions of language, must be studied in co-occurrence with their use. There is no sharp dividing line between a speaker's knowledge of language and his or her use of it, according to Langacker and others. They feed each other, in that the speaker's system of language produces the productions, and the productions, of the individual AND all of the members of the speech community in turn shape the system--fairly quickly in the process of language acquisition where there is no system there to start with, and more slowly as the connections are made and entrenched via use of the system in production and comprehension.

Some other modern linguists, however, DO agree with the idea of language as existing independently of any given speaker. The linguistic system is seen as large and overarching, and each speaker has a subset of the overall system, plus some individual idiosyncrasies. The large amount of overlap in what patterns the speakers know allows communication. This is how Anna Wierzbicka seems to view the linguistic system, for example.

For Sydney Lamb at least, and possibly for Langacker, who always studies the linguistic system as internalized by a language user, the 'reality' of such a supra-system would be open to doubt. An entirely social Langue beyond the speaker would be seen as an abstraction, perhaps a useful one (or perhaps, some would think, a pernicious and distorting one). The only reality for Lamb and for Langacker is ultimately inside the head, as mental processing takes place. Language processing and the linguistic system are not seen as ultimately separable, because the linguistic system, latent in entrenched neural connections, is ACTIVATED during linguistic processing, and thus can be thought of as coming into real existence only during acts of linguistic production and perception. In this sense linguistic patterns are completely emergent. The linguistic patterns are not stored as patterns, but as a very complex system of latent neurocognitive connections. The patterns emerge only in deployment, i.e. via activation in use. Both Langacker and Lamb agree on this. Unlike Lamb's, Langacker's theory does not focus on connections at lower levels of organization such as types and groups of neurons and where they reside in the brain and what other parts of the brain they connect to; but he has written about how he envisions the ultimate "wetware" implementation of his cognitive linguistic network system (a language is a "structured inventory of conventional units") as a connectionist network with patterns emerging in activation.

I agree with Langacker and Lamb that linguistic systems take the form of systems of connections among nodes that ultimately reach down to the neuron level. However, I don't have any problem also being able to look at a linguistic system as a group phenomenon, rather than an individual phenomenon. I think analysts should be able to abstract any entity they need or find appropriate to find patterns at any level which will be fruitful for understanding human behavior. Sociologists look at societal patterns, not individuals' patterns of behavior, because they want to understand society as a phenomenon. If they do observe individual behaviour, they then re-form the data into groups, averaging any quantificational data, to see larger patterns in behavior of the groups observed.

Obviously society does not reside in people's heads, and if we confine ourselves to the individual we would be saying there is no object of study at all in sociology. From Durkheim onwards, the study of societies and subsegments of them, in the abstract, has been extraordinarily useful for drawing generalizations about how people act in the aggregate. Those who denigrate scholars who invoke "society" or "social groups" because these are abstractions and not "real" seem to me to be missing the mark. All human concepts are abstractions at some level. No concept is a fully concrete entity, even if some concepts relate more obviously than others to things observable with the perceptions. If we confine ourselves to the study of things that have concrete, in-the-world referents, we will never understand the mind at all, or much of anything else.

[A side note: There is generally an ideological basis to the rejection of society as an entity that can be studied at a level higher than that of an individual. The basic claim in this view is that individuals act individually and not by virtue of some abstract entity called society which can exert "societal pressures". To say that "society" can have an influence on individuals, in this view, is to take away personal responsibility. This is the Thatcherian conservative view and has a populist basis of support among people who deny that social inequalities can be a causal factor in crime and other anti-social behavior. Such people typically want punishment for criminals that does not take into account factors such as social deprivation, childhood abuse coming from the social milieu, etc. In the United States this view is part of social conservative worldview, in which personal choice is the only force influencing behavior that can possibly count. ]

For language, I think it is indeed useful to posit a type of linguistic system not limited to what is in speakers' individual heads, and it can certainly can be, and has been, studied as an entity in its own right, belonging to the speech community as a whole (however one defines that). If this is an abstraction, and I think it is, I have no problem with that.

Observing representative linguistic corpora can get at the structure of this larger entity more accurately than they can get at the system of an individual speaker, unless the analyst confines him or herself to looking at the productions of a single speaker (cf. Barlow and Kemmer 2004). There are purposes for which examining single-speaker productions is a very useful methodology (although nobody has figured out how to isolate single-speaker input!!). But for studying what speakers have in common it obviously leaves much to be desired, since we cannot then easily factor out what is idiosyncratic to the speaker.

Scholars such as Doug Biber, a corpus linguist, are studying, via large corpora, what they view as the linguistic system at large, and Biber at least does not view this as a psychological entity. His view and method are extremely useful for discovering patterned linguistic behavior. In particular, Biber has identified the "linguistic markers" of particular genres of language and that is in itself valuable for creating linguistic reference works, for advanced language teaching, and even, ultimately, as a basis for a representative input that could give us insight into how individual speakers can draw the linguistic generalizations they do when creating and modifying their individual linguistic systems.

Saussure adds a few details about the nature of Language before leaving it for a while, until the discussion of the nature of the Sign. He says a Language is "passively assimilated" by the speaker. Premeditation, or intentionality, are not part of the system. This relates to the tenet of later linguistics that the speaker does not have access to the understanding of the sytem; speakers can't and don't analyze it, they just use it. That is why it is so hard to come up with accurate descriptions of language: we are too busy using it to be able to reflect on how it works; linguists can make this reflection, but then we are not using it right at that moment, so its nature is necessarily distorted.

Speech, or parole, on the other hand, for Saussure involves conscious choices by the speaker. Volition and intellect are both brought to bear.

In the rest of this second section, Saussure sums up the crucial elements of what he has laid out as a well-defined object of study, the CRUCIAL aspects of language that he defines as Langue, i.e. a particular linguistic system. He separates the Langue from everything extraneous to it that would lead to muddling in things that are less important to the nature of language. A Language (Langue) boils down to the tiny part of the speech circuit that unites the auditory image with its associated concept--a specific type of psychological association that is SOCIAL in nature.

Saussure states on p. 15 in the Harris translation:

Linguistic structure is no less real than speech, and no less amenable to study. Linguistic signs, although essentially psychological, are not abstractions. The associations, ratified by collective agreement [or "associations which bear the stamp of collective approval", in the Baskin translation], which go to make up the language are realities localised in the brain [or in Baskin's more literal translation 'which have their seat in the brain']. [my emphasis]

From the modern point of view it is difficult to grasp this apparent contradiction. Associations as conceived in modern psychology and linguistics ARE psychological and ARE localized in the brain, as Saussure says, but what analogue is there in the brain to their being 'ratified by collective approval'? This part sounds like something outside the mind, out in the abstract realm of intersubjectivity. He goes on to say that linguistic signs are 'so to speak' tangible, because they can be reduced to static representations like written linguistic symbols. But obviously the tangible representations are not the signs themselves, a point he makes himself elsewhere in the Course.

Cognitive linguists see psychological associations (ultimately, neurocognitive associations, as described above in connection with Lamb and Langackers' views) as the basis of linguistic signs. These associations are understood as ultimately reducible to a concrete reality in the brain, namely patterns of neural activations. Linguistic signs would then be the cognitively entrenched paths of neural connections that are activated in speakers' use of language. They are entrenched precisely because of the frequent encounters of these signs in the social context. They are the psychological, internal RESULT of social interaction -- and at the same time the means by which MORE linguistic interaction can take place.

If the signs are, as Saussure describes, shared in a conventional, i.e. inter-speaker system, it would seem impossible for them to at the same time reside in the individual psychological or cognitive system. The precise nature of the Langue SEEMS, then, to float above individual speakers, in what to me is an abstract conceptual space, a kind of analyst's reification, of the collectivity of individuals' systems of Signs.

The only way I can try to resolve the paradox in my own mind is to say that the signs in the collectivity ARE all in minds--but they are distributed across various speakers' minds, with minimal divergence so we can say they are effectively shared. The signs themselves then would be concrete, mental entities. But the collectivity--the sum total of the all the speakers' signs--would constitute an interpersonal system of Langue that could be studied independently of the individual speakers and their own personal manifestation of the signs. Even with this way of thinking about it, to me, the notion of the Language, Langue, as a system of shared signs of a collectivity of speakers, still looks like an abstraction--an abstract generalization over the particular systems of individual speakers in individual heads.

To be fair, modern linguists, in describing linguistic systems, are still driven to expedients that seem to involve abstraction, that is, mentally isolating out a set of shared commonalities that do not necessarily reside in a single individual, whether explicitly or implicitly. For example, Chomsky unapologetically abstracts out the "ideal speaker-hearer"--an idealized creature not existing in the real world, because all of the real-world cognitive and perceptual influences on the system, like memory, attention, emotion (outside what the linguistic message explicitly makes reference to), etc. are deliberately left out of his object of study. The idealization also gets rid of all inter-speaker variation, a thorn in the side of the type of rule system he proposes. Intra-speaker variation of the sort in which a speaker will decide that an utterance is grammatical one time but ungrammatical another time, will also be filtered out by this means, because such indecision would be claimed to be unrelated to a speaker's real grammar, or competence, but instead a function of performance factors that interfere with the idealized grammar or competence.

Sydney Lamb takes a different tack. He says the object of study is the linguistic system "of a representative speaker". This system is undoubtedly in a speaker's head. The representative speaker is an actual speaker, but one having a system that is maximally similar to what all the other speakers have. The question is, is there, or even can there be, such an average actual speaker in the real world? Or is the notion really a theoretical abstraction of some 'average' type? If there is one, or a range, of speakers who have a system that is within some limits about the same as most speakers' grammars and therefore should be studied, how exactly do we know which speakers those are? Or how many of them there are--what is the 'normal variation', given the usual kinds of dialectal and social stratification that the sociolinguists have been observing? Given these factors, it seems difficult to operationalize the notion of 'representativeness'. And if we get a speaker out of the ideal range of variation for a particular group, without knowing it, someone who is in fact NOT representative, then aren't we studying a system which is NOT the correct object of study?

The climax of Saussure's explication of Langue in Chapter 3 is the claim that we CAN study this newly differentiated and defined object of study. The fact that we can reduce signs to representations (like written forms, and visual images), he says, means we can study it with far more exactness than other shared knowledge systems.

My own observation on this is that yes, language does appear to have more to 'grab' and analyze precisely than other systems of social meaning like fashion, or art, or other non-linguistic semiotic systems that people use to signal each other. The linguistic system is huge and complex, but somehow, more precise and specific than signifiants like the shape or length of a tie or jacket or skirt. Chris Taylor (in 2006 class) pointed out that shoes, etc. are concrete physical objects and we can manipulate them physically. I agree that linguistic signs do not seem at all that concrete and tangible as Saussure suggests. Just because we can make a representation that we can then analyze (like the written representations in a phonology problem) doesn't mean that the phenomenon of the sign itself is concrete.

In the final section of Chapter 3, Saussure outlines a way in which his method of focusing on the link between signifier and signified can apply to other human sign-like systems besides language. He envisions a new science of semiology or semiotics not knowing quite what form it should take, but seeing that it is a logical outgrowth of his way of approaching language.

Later scholars took up his idea of having a general theory of signs that includes language but only as a special case alongside other meaningful systems, and it became the basis of a very wide field. Anthropology first, then literary criticism, then cultural studies (essentially the intersection of the last two), not to mention Art and Architecture, applied the notion of the Sign to a wide variety of human representational systems, including rites, customs, and all sorts of social and cultural behavior.

Semiology still exists, but in linguistics it has not remained a field of interest for most scholars. In effect, it has been replaced by the idea of Cognitive Science--the large, sprawling discipline that studies human cognition and treats language as only a subpart of this. In Cognitive Science, social structures are sometimes an afterthought, but in recent years a field called Social Cognition has emerged, which aims to study the cognitive processes that occur in group interactions.

Cognitive Linguistics focuses on mind-internal, individual processes, but the Cognitive Linguistic theoretical concept of entrenchment provides a way of understanding the degree of social conventionalization of linguistic units. The more conventionalized across a group a unit is, the more entrenched it will be in the minds of members of the group. Cognitive Linguistics hopes to be able (at least some day) to incorporate social structures and social interaction in an integral way, but the number of studies focusing on the social dimension is relatively small. (I wrote a co-authored paper myself once on a sociolinguistic variation pattern which we analyzed in Cognitive Grammar terms.)

The Nature of the Sign

The fundamental arbitrariness of the connection between the sound image (signifiant, often called in English the signifier) and the concept (the signifié, often termed the signified) was for Saussure the key to the essence of human language.

Saussure discusses some apparent departures from arbitrariness, like onomotopoeia, and tells us why they do not vitiate the essential arbitrary nature of language. (See Sapir 1949 for an investigation of the question as to whether there is an onomotopoeic basis to word forms designating concepts of 'large' and 'small' in the languages of the world. Sapir claims to have found a genuine, iconically motivated form-meaning relation. Still, he himself agreed with the notion of the basic arbitrariness of the form-meaning link.)

Arbitrariness is now accepted pretty universally by linguists as a fundamental property of language. Still, modern scholars have found a number of ways and areas in which non-arbitariness, or motivation, is found in linguistic structure. Some have maintained that Saussure went too far in emphasizing arbitrariness. But he does qualify it somewhat in saying that some aspects of language are partially motivated from within the system. (Motivation from within the system is an important aspect of Cognitive Grammar, which also allows for different degrees of motivation.)

Signs as nodes in a network of contrasts


An important aspect of Saussure's linguistic Signs, those conventional (socially shared), and at the same time psychological, units, is that they are defined by their position in the entire network of signs. So far this is similar to Cognitive Grammar's view of language as a "structured inventory of form-meaning units" (Langacker 1987). But Saussure also saw the essence of each sign as ONLY its contrast with other related units. The European structuralists of the 1920s and 1930s, who were Saussure's followers, certainly treated this as a basic assumption. The implication is that a sign would seem to have no content at all, independent of its specific set of contrasting relations with other signs.

For example, the concept associated with the English preposition on would be defined by the properties in which it directly contrasts with the concepts belonging to the word forms off, up, in, and any other prepositions with which on shares semantic properties. This means that there is no room in the semantic characterization of a unit for features of meaning that cannot be stated in terms of a dimension in which there is a contrast with another related sign. So, if there were no other prepositions in the language that referred to a property like, say, 'contact between an object and another object', we could not posit that property for on. However, we CAN posit this property for English on, because there are contrasts in various parts of the language, such as that between above and over, which specifically make reference to this dimension of 'contact vs. no contact'.

A vast amount of productive descriptive analysis of various grammatical systems has resulted from this insight about the importance of contrast in a linguistic system. It seems to work well for grammatical systems, because grammatical units, the more they have undergone the diachronic process of grammaticalization, or subsumption into the grammatical system, occur in relatively closed sets of elements contrasting in regard to particular grammatical and conceptual categories: singular/plural, accusative vs. other cases, 'reference to container object' vs. 'reference to non-container object' (found in in vs. on, the verbs open vs. close, and adjectives deep and shallow).

For non-grammatical concepts, though, the structuralist method of analysis solely by contrasts doesn't work so well. Is the concept of 'cup' or 'cat' simply nothing more than its semantic contrasts with our concepts for mugs and bowls or dogs and mice? Modern theories of cognitive semantics do not restrict themselves to this criterion of 'contrasts only' in the semantic properties they posit in their semantic analyses. Concepts can be characterized positively as well as in terms of features they happen to lack when contrasted with other units.

It is hard to know whether Saussure would approve, looking at how semantics is now analyzed in cognitive linguistic theories. I would hope that he would recognize the need for recognizing more to the semantic side of the sign once we get past the simplest grammatical/semantic oppositions.

Mutability and Immutability of Language; Synchrony and Diachrony as two dimensions of Langue

Saussure says that William Dwight Whitney was the one who first noticed the importance of the linguistic Sign. But Saussure was apparently the first in linguistics, and perhaps in any social science, to claim that the structure of a semiotic system and the relations of its component parts are paramount, and that individual elements therefore take a back seat to the system as a whole. For linguistics, the effect was to make historical-comparative linguistics, at that time essentially the only thing regarded as linguistics, look incredibly particularistic, because as knowledge of the history of sound changes in the various branches of Indo-European advanced, and problems of reconstruction and change were solved, the tendency was to focus more and more on individual sounds and morphemes and their trajectories through time. I believe that this focus was largely a practical matter, as it was an efficient way to organize a great deal of knowledge; but it certainly left something to be desired if one wanted to look at the forest rather than tree bark of a number of different trees under a microscope.

The comparative historical approach also made Linguistics up to that point look like it was missing the essential heart of language, namely, meaning -- which is of course exactly what Saussure was arguing.

Comparative linguistics as developed by the Neogrammarians, and also the new fields of dialect geography whose development was encouraged in Osthoff and Brugmann's "Neogrammarian Manifesto", both focused on the sounds of language--the form, rather than the content, of linguistic signs. The complex morphology found in Indo-European languages was, to be sure, investigated, but few linguists seemed to care much about the meaning side of the morphemes, perhaps because it was harder to state and to draw generalizations about than about the form. (Both Brugmann and Delbrück did, however, discuss meaning of various grammatical categories and constructions in their work, and Brugmann in particular seems to have had very good semantic instincts. But other comparative linguists of the time tended to assume a very tight relation between form and meaning, with form expected to transparently reflect meaning; and to shy away from discussing meaning if it was not very concrete. )

The consequence of the idea that a linguistic system is simultaneously shared by speakers who communicate with one another, and structured by contrasts recognized by those speakers is the following: that in order to properly study language (in its non-technical sense), two dimensions must be sharply distinguished: Diachrony, or evolutionary/historical linguistics, and Synchrony, or static linguistics. As Roy Harris puts it, in his translator's introduction to the Course (p. x):

[W]ords, sounds, and constructions connected solely by processes of historical development over time cannot possibly, according to Saussure's analysis, enter into structural relations with one another, any more than Napoleon's France and Caesar's Rome can be structurally united under one and the same political system.

In the Course Saussure sometimes presents synchrony and diachrony as two equally important perspectives or dimensions in language; but elsewhere in his chapters it is clear that he privileges synchrony over diachrony for understanding the nature of language and of how speakers communicate.

Saussure's structuralist followers seem to have paid more attention to such passages, and elevated synchrony to its current position as the basic dimension on which language is thought to be most appropriately studied. Diachrony was thus left as the derived, and less important, dimension, and the study of language change came to be referred to in English as Comparative Linguistics, Comparative-Historical Linguistics, or Historical Linguistics, rather than simply Linguistics. Even the word Philology, the older English word for Linguistics, fell out of use, as possibly because it was so strongly associated with the historical dimension. In the twentieth century, the study of the synchronic structural dimension of language that Saussure advocated, on the other hand, acquired the simple, unmarked name Linguistics.

The basic argument for the priority of studying synchronic linguistics was (and still is) that the speaker does not, in general, know the history of his or her language, nor have access to it while speaking. Saussure was the first, to my knowledge, to give this as an argument for taking a synchronic perspective. Because of these undeniable facts, he maintained that in consequence, historical information should never be mixed, consciously or unconsciously, into any linguist's analysis of what the speaker knows about the structure of the system, because if it is, it will fatally distort the object of study.

And studying what the speaker knows (as evidenced by how the speaker uses language) is what we should be studying, says Saussure, because it will give us insight into the nature of the human being as a species that creates and uses semiotic systems--and especially, as emphasized by later scholars, into the nature of the human mind: the language-user's system of knowledge, the cognitive operations by which it functions, and the interactions of the language user with other langauge users/conceptualizers.

Obviously this view is directly contrary to the Neogrammarians' take on what the scientific study of language must be; they were deeply and fundamentally historically-oriented, and either never thought of the idea of a synchronic linguistics, or if they did, denied even the possibility of it.

Below is a famous quote from Hermann Paul's Prinzipien der Sprachgeschichte (Principles of Linguistic Science), in which Paul reacts to claims that there can be a scientific, yet non-historical, approach to language:

Es ist eingewendet, dass es noch eine andere wissenschaftliche Betrachtung der Sprache gäbe, als die geschichtliche. Ich muss das in Abrede stellen. Was man für eine nichtgeschichtliche und doch wissenschaftliche Betrachtung der Sprache erklärt, ist im Grunde nichts als eine unvollkommen geschichtliche, unvollkommen teils durch Schuld des Betrachters, teils durch Schuld des Beobachtungsmaterials. Sobald man über das blosse Konstatieren von Einzelheiten hinausgeht, sobald man versucht den Zusammenhang zu erfassen, die Erscheinungen zu begreifen, so betritt man auch den geschichtlichen Boden, wenn auch vielleicht ohne sich klar darüber zu sein. (Paul 1884 p. 20).

It has been objected that there exists another scientific view of language besides the historical one. I must dispute that. What is declared to be a non-historical and yet scientific view of language is in essence nothing but an imperfectly historical one. The flawed and incomplete character of this view is in part the fault of the observer, and in part attributable to the nature of the data under observation. As soon as one goes beyond the mere statement of facts, as soon as one attempts to capture the coherence of the relationships, to truly understand the phenomena, one steps onto historical ground, even if one is not necessarily aware of it.

It is certainly true that in Saussure's analogy of the chess game, only the current position of the board matters. This state, along with the accepted possibilities for future movement of each piece on the board, tell us what is happening.

Further comment on this metaphor would be appropriate. It bears on the relation of the two views of language, synchronic and diachronic. Is this dichotomy a resurrection of the old Heraclitus/Parmenides debate about what is "real"? Or do we have two mutually compatible or even complementary views? Should synchrony have preeminence over diachrony in linguistics today?

Other issues

You might want to consider some of the following issues that relate to specific concepts and distinctions drawn by de Saussure:

* the innovation/spread distinction. Does innovation belong only to Parole, and then only later (if at all) spread to Langue? How exactly does the specific change, if an innovation counts as a change, 'cross the line' into Langue?

* the possible difference in types of linguistic changes, in regard to how they are innovated and/or how they spread. Difference in lexical vs. grammatical change? Syntactic vs. morphological?

* The role of what Saussure calls "premeditation", or similar notions like intentionality and purpose. Do these necessarily belong outside Langue, only in Parole, as he claims? Or is there some role for intentionality within Langue? Or does this lead to contradictions so that we would have to have a slightly different notion of Langue than Saussure?

* the relation of linguistic signs to signs in other semiotic systems: the similarities/ differences between language and fashion, for example, or language and musical notation, or language and "biosemiological systems" such as insect mating signals/parts of display rituals. What other semiotic systems of note can we compare to language?

Some larger issues


Why is Saussure's conception of language seen as so revolutionary? For one thing it seems to answer an ancient question about how language is to be viewed, in a novel way.

One ancient Greek school held that the forms of words ('a form of a word' here means a pronounced word) were intrinsically connected to the things they designated. In other words, a dog was called dog (well, kynos, because these were Greek-speaking philosophers) because there was some essence to the sound of the word that naturally 'belonged' with its meaning. That was essentially taken as an explanation of why things had particular word forms attached to them. Essentially, we could say that this view represented a still deeply-ingrained folk theory of language. This is exemplified by the anecdote of the West Country English farmer, who, surveying his pigs eating greedily and messily, mused, "Aye, rightly is they called pigs!" This theory can only work of course, if there is only one language in the world. (And then only POSSIBLY work.) The concepts, as well as the sounds intrinsically belonging to them, on this view, belong to external reality.

Plato, on the other hand, thought that for language, concepts were primary, rather than the sound-forms of the words. His notion of concepts essentially boils down to the notion of idealized, perfect ideas, and fully independent of human experience. These are called "Platonic concepts" and there are still strong echoes of this view in modern Philosophy. The 'words' (word forms) that we associate with the concepts, however, are learned, and are therefore essentially arbitrary labels, conventionally given to the concepts. Plato didn't focus on the arbitrariness, but on the nature and inventory of the Platonic concepts. His theory at least could account for the existence of different languages, and for the phenomenon of language change. (I have no idea if Plato actually considered these points as advantages to his theory, but I think they are advantages over the folk theory). Plato's theory also seems to entail a universal language of thought, made up of concepts. (There is apparently be a much later disagreement over whether the concepts belong only to the human mind, or to external reality as above. I would need a philosopher to clarify this for us. After looking around for explanations of these controversies, I am sure glad I am a linguist.)

So where does Saussure stand on the basic issue here? It should be obvious that his answer to the question of "Is there an intrinsic connection between words and things? Or are arbitrary sounds assigned to the REAL things?" was, "Neither".

Basically, he didn't see the relation between the forms of words and their meanings in remotely the same way as the ancients. Right at the outset he warns us that the forms of words (the signifiants) are only incidentally sound-based. The sound isn't crucial to the important part. Furthermore, the concept part is not at all to be understood as some kind of immutable Platonic concept. Saussure would have certainly agreed with the "convention" idea, but the similarity ends there. The concepts themselves are affected, indeed, created, by the social use of linguistic signs. Roy Harris, in his introduction to his 1983 translation of the Cours, puts it like this:

Words are not vocal labels which have come to be attached to things and qualities already given in advance by Nature; or to ideas already grasped independently by the human mind. On the contrary, languages themselves, collective products of social interaction, supply the essential conceptual frameworks for men's analysis of reality and simultaneously, the verbal equipment for their description of it. The concepts we use are the creations of the language we speak. (Harris, ix, emphasis mine)

To this I would add, by 'creations of the language' what is meant here is social creations, not individual creations; specifically, the creation of conventions. The creation process is at a higher level than the creative acts of individuals, which in Saussure's framework belong only to Parole.

I might remark that the phrase 'higher level' that I just used seems to imply an abstract level, involving commonalities not actually residing in real people. This is a reflection of my interpretation above that Saussure's Langue is what I would call an abstraction, similar to the way the concept of "society" as used by ordinary people or by sociologists is an abstraction. Unlike Margaret Thatcher, however, I think that abstract entities like 'society' do exist inasfar as they can be observed to have consequences and affect people's behavior. Abstractions, in other words, exist to the extent people act as though they do.

It seems obvious to me that not just academics, but all language users have abstract concepts in their arsenal of concepts, and that these can certainly influence their behavior and therefore the world. Linguistic concepts, surely, are of this nature. If we have a concept of a 'sister' that includes typical relations of sisters (culture-specific or not), then people who share this concept will likely treat their sisters as sisters and not as strangers. Similarly for any other categories of humans ('child', 'American', 'Democrat', 'sheriff', 'teacher', 'killer', etc.), for animals and inanimates ('horse', 'desk'), and for abstract concepts ('sincerity', 'liberty', 'hatred'). This is not to say that everyone acts or thinks the same way regarding members or things characterized by the conceived categories, but that our understanding that things fall into these categories affects our behavior in regard to the members.

When things (such as categories) have consequences, this makes me say they are perfectly real things, albeit not concrete things. Construals of external reality, in other words, constitute humans' internal reality. In saying that, it is clear to me that I am a deep-dyed Saussurian in this particular regard.

A crucial thing here is that Saussure's idea provides the seeds of a view that has affected all modern linguists to some degree or other. It is that our reality, whether viewed as an internal linguistic/cognitive system or an external linguistic/social system, is a created reality--a humanly-created reality.

This idea, taken to an extreme, emerged as the common mid-twentieth century view that linguistic categories in a particular language, a Langue, are in effect the straightjacket of thought. As Ludwig Wittgenstein (1889-1951) said later, "The limits of my language mean the limits of my world."

In Linguistics, Saussure's basic idea of Langue as a humanly-created and shared system of Signs--linguistic units which categorize what we know--takes us to straight to Boas, Sapir, and Whorf, each of whom had his own precise take on the issue of the relation of categories of language to categories of thought. These ideas are sometimes portrayed far more extremely and less subtly than the original thinkers laid them out, so we will now begin to examine exactly what the latter three said about the relation of language and thought.

[ jump to top ]