In General Linguistics Ferdinand De Saussure

What de Saussure Actually Said and Why It Still Matters

Ferdinand de Saussure didn't publish much during his lifetime. He taught at the University of Geneva in the early 1900s and his students compiled his lecture notes into the Cours de linguistique générale after he died in 1913. That book became one of the most influential texts in 20th-century linguistics, but it was never edited by him, and it contains contradictions his students apparently couldn't reconcile. The core idea most people encounter first is the sign. A linguistic sign isn't a word pointing to a thing. It's a two-sided psychological entity where the signifier is the sound pattern and the signified is the concept. The connection between them is arbitrary. There's no natural reason the sequence of sounds /k/a/t should mean the animal with whiskers and a tail. Different languages arrange the same arbitrariness differently, which is why the concept exists independently of any particular language's inventory.

In General Linguistics Ferdinand De Saussure

When you're reading Saussure, you'll hit the distinction between langue and parole almost immediately. Langue is the system of conventions shared by a speech community. Parole is individual utterance. This matters because Saussure argued that linguistics should study langue, not parole. Language as a social institution exists before any single person speaks. An individual speaker can't change it alone. That's a methodological claim, not just a definition. I ran into a practical problem when I was teaching a seminar on structuralist linguistics and a student brought up the syntactic structures in Chomsky's work. They assumed Chomsky inherited the langue/parole distinction intact from Saussure. He didn't. Chomsky explicitly rejected the social nature of langue and rebuilt the competence/performance dichotomy around individual cognition. The overlap is surface-level. If you're citing Saussure to support a Chomskyan position, you're misreading both of them. I had to walk three students back from that confusion mid-semester. Another distinction that gets misused constantly is synchronic versus diachronic linguistics. Synchronic analysis examines a language at a single point in time. Diachronic analysis traces how it changes across time. Saussure wanted synchronic study to take priority because the system's internal relations are what give language its structure at any given moment. A pronunciation shift doesn't affect other sounds because of its history. It affects them because of its position in the current system. That's a subtle but important point that most textbooks gloss over.

Here's the part beginners miss: Saussure's value theory. He argued that in a language system, elements have value only through their difference from other elements. "Sheep" in English and "mouton" in French don't simply map onto each other because the English lexical field includes "ewe" and "lamb," which split the semantic space differently than French does. The signified of each word is determined by what it is not within that particular system. This is why translation is never a simple substitution operation. You're moving between different value systems, not between labels and referents. The arbitrariness principle has a limitation that Saussure himself acknowledged but didn't resolve well. Onomatopoeia and interjections seem to contradict it. The word "bang" sounds like what it describes. Saussure dismissed these as marginal cases, but they're not marginal in actual language use. Children acquire them early. They appear in creative writing constantly. Modern linguistics treats iconicity as a real force in language, something Saussure's framework barely accommodated. If you're building a model based on his work, you'll need to add something for it. I remember working through a corpus analysis once and trying to apply Saussurean structural analysis to a set of dialect variations in northern England. The system was supposed to be closed and self-determining. Dialect continua aren't closed systems. They bleed into each other across geographic boundaries. The binary oppositions Saussure described work cleanly on abstract models but fall apart when you look at actual speech communities. I ended up supplementing his framework with Labov's variationist approach because neither method alone could account for what the data showed. That's the honest takeaway from reading Saussure: his model is necessary but not sufficient for real linguistic analysis.

The structuralist tradition he inspired stretched far beyond linguistics into anthropology, literary theory, and psychology. Lévi-Strauss applied his methods to kinship systems. Barthes used them for cultural analysis. Lacan reworked them for psychoanalysis. Each adaptation changed the original significantly. Saussure himself would probably have objected to how his students treated langue as a totalizing structure when he emphasized its empirical grounding. The Cours was already a compromised text before the post-Saussureans got hold of it. If you want to read the primary material, the Norton Critical Edition is the standard reference. It includes Bally and Sechehaye's original compilation alongside revisions proposed by Albert Riedlinger, who had access to Saussure's manuscripts after the fact. The differences matter. Riedlinger's version reorganizes several sections and adds material the original editors left out. It's not a definitive edition, but it's closer to what Saussure actually wrote than the Cours as most people encounter it. The biggest practical limitation of Saussurean linguistics is that it treats the language system as static and coherent. Real languages are messy, contested, and constantly negotiated. Code-switching, language contact, and sociolinguistic variation all exist outside the model's capacity to describe them. Saussure's framework is excellent for analyzing the internal structure of a single language at a single point. It's nearly useless for explaining why that language changes or how speakers actually use it in interaction. That doesn't make it wrong. It makes it incomplete. Every serious linguist I know keeps it in the toolkit and reaches for something else when the problem demands it.