Competence/Performance

by Peter Ludlow, irift.net

Abstract: The competence/performance distinction has played a role in linguistic theorizing since 1965. The key idea is that one can distinguish between the grammar that an agent has and the performance errors that violate the posited rules or principles of the grammar. The first question to consider is what counts as a performance error and whether it can be defined in a way that avoids bailing out defective theories. The next question concerns what competence consists in — in what sense do we have linguistic rules or principles? One idea is that the rule plays a normative (guiding) role. Another idea is that we merely “cognize” or represent the rule. These lead to different conceptions of performance errors. We then turn to versions of the competence/performance distinction in other fields, including ethics and logic.

Key points:

· The competence/performance distinction suggests that there are rules of grammar that we have but which are nevertheless violated in practice.

· The applications of the competence/performance distinction must be handled with care to avoid letting defective theories of grammar off the hook.

· There are bith normative and computational ways of grounding the notion of competence.

· Extra-linguistic applications of the distinction include ethics and logic, the study of which can be insightful to linguists.

Introduction

One common idea in generative linguistics, dating to Chomsky (1965) is the distinction between competence and performance. While the distinction does a lot of work, there isn’t necessarily agreement on what it is, nor how it should be applied, nor what grounds it. After we consider these questions, we will look at similar distinctions in other sciences and reflect on whether they can shed light on applications of the distinction in linguistics.

What is the competence/performance distinction?

Current literature isn’t consistent on what the distinction comes to, and even when the distinction itself is clear, it isn’t entirely clear what grounds the distinction. In this article our aim is to get some clarity on ways to make sense of the distinction and the ways in which it might be grounded.

Linguists sometimes argue that agents have competence with respect to a certain grammatical rule, even though those agents do not always perform in accordance with the rule. Indeed, in some cases, they may rarely perform in accordance with the rule. What this means is that the rule is represented by the agent in some sense, and the rule has some controlling function, but the rule is nevertheless being violated. For example, I might have competence in a grammatical rule to the effect that allows so-called center embedding. Thus, if I can say “The dog barked”, I can also say “The dog the cat startled barked.” But once we try to embed once again, things get difficult. Thus, the following example is tough to recognize as acceptable English: “The dog the cat the mouse annoyed startled barked.” One view is to say that this example is unacceptable because it is ungrammatical, but another take would be to say that it is perfectly grammatical, but it is difficult to process for some reason or other (memory limitations or complexity limitations). The idea would then be that while we have a grammatical competence that includes unlimited center embedding, we have a performance error that short-circuits this competence.

There are of course any number of ways that our competence and performance might come apart. For example, our competence might allow us to have sentences of a million words or more (theoretically infinitely many), yet our performance is system is not up to that task. Take a sentence of a million words. Given our memory limitations we are apt to forget about the first part of the sentence before we get to the 100th word let alone the one millionth word. And even if our memories were up to this task, we are apt to get distracted while listening to such a sentence and quite likely would fall asleep at some point since it would take 55.6 hours to finish the sentence (at 300 words per minute). But even in simple sentences we stumble, cough, sneeze, loose our trains of thought etc. The question is, what does this difference in competence and performance come to? Why is it important and if it is important, what grounds it?

Application of the Competence/performance Distinction

The competence/performance distinction is a useful tool for linguists, but like any tool, it can be mishandled. The dangerous aspect of the tool is that it can be used to insulate a linguistic theory from criticism — the idea being that failures of the theory might be chalked up to failures of performance. One obviously needs to wield the tool with care. For the most part, we can recognize certain failures as non-linguistic issues — for example, the afore-mentioned coughs and sneezes, and a general recognition of memory limitations.

We’ve already spoken of one case where the competence/performance case has arisen — in the context of center-embedding judgments. But we can also recognize judgments as being deflected by pragmatic considerations. So, for example, one might find a sentence to be unacceptable because it is not viable in a context, or perhaps because it is offensive or impolitic or “too cringe” to say.

On the flip side of these considerations are judgments that a sentence is acceptable, when it should be ruled out by the proposed theory of grammar. For example, on many accounts of grammaticality it is perfectly fine to end a sentence with a preposition: “Who are you looking for?” But some people will argue that this is unacceptable and will insist that we use “For whom are you looking.” In this case, it can be argued that learned prescriptive rules have overridden the judgments of acceptability underwritten by the grammar.

Are there ways to say for sure whether something is unacceptable for reasons of competence or reasons of performance? Probably not, such judgments have to be made in the context of our theory of grammar in its entirety, alongside of our theories of human speech production and comprehension. There are certain easy cases — e.g. sentences of a million words — where we can say that whatever our theory of performance is, those cases will outrun our performance abilities. But there are other cases, such as the center embedding case, where we ultimately need to know more about the mechanisms by which embedding takes place and how memory and perceptual abilities intersect with that. Another way to put this is that these questions can only be answered in the context of theory development.

How is the competence/performance distinction grounded?

So-called grounding questions aim to get at the underlying principles — typically metaphysical principles — that anchor an object of inquiry. If there is a competence/performance distinction, and if it plays an important role in linguistic inquiry, it is natural to ask what grounds that principle. For example, Chomsky (1980, 1986) spoke of our “knowledge of grammar.” This suggests that our competence would consist in our knowledge of certain rules or principles. A performance failure could consist in the failure to follow a rule — perhaps by choice. In his later work he has spoken of our “cognizing” these rules or principles. This suggests that the rules are represented but that acting in accordance with those rules might constitute some sort of breakdown in execution of the instructions by the performance system — like slipping gear in a machine, or physical limits of the machine.

Let’s start with the case of knowing linguistic rules. One idea, pursued in Ludlow (2013) is that we could think of such rules as playing the role of individual norms — not prescriptive norms but norms that are a function of the psychology of the agent. This could involve the idea that there are normative rules that an agent has, but which are not consciously accessible to the agent. They just set off an alarm bell (“that doesn’t sound right”) when they are violated, even though the alarm bell may be drowned out by other factors (thus giving rise to performance failures). We will return to this idea in the context of the competence/performance distinction in ethics in the next section.

Alternatively, as noted, one might avoid talk of rules that an agent “knows” and opt for a more conservative characterization of the agent “cognizing” certain rules. What might this come to? It might be that there are computational data structures that represent the relevant rules, and these rules thus play some causal role in an agent’s linguistic performance, albeit a causal role that can be short-circuited my memory limitations etc. This idea would bear some similarity to the “computational level” in the theory of vision articulated by Mar (1982).

Extra-linguistic uses of the competence/performance distinction

One domain where the competence/performance distinction has played an important role is in the psychology of human reasoning. This is a field which aims to treat logic, or at least part of logic, as a domain of human psychology, in the sense that persons have the competence to reason logically, but that they do not do so (as every logic teacher can attest). Work by Rips (1994) explored the psychology of human logic reasoning, and one of the headline ideas of that work is that persons do have a competence in logic — albeit one that can be short-circuited by certain kinds of complexities. Ludlow and Zivanovic (2023) explored this idea in the context of “natural logic” — an account of logic built on the logical forms of natural language, again relying on a competence/performance distinction to explain why failures in logical reasoning nevertheless happen.

Similar considerations have arisen in ethics. Arpaly (2003) has argued that persons have ethical principles or rules are not accessible to consciousness, but which can be normatively guiding, and which can also be short-circuited. To draw on one of Arpaly’s examples, in the Mark Twain book Huck Finn, Huck is travelling on the Mississippi river with an escaped slave named Jim. At one point Huck is asked if he is travelling with an escaped slave, and he says “no” thinking he will go to hell for lying, but having the sense that it just wouldn’t sit right to turn in Jim. Railton (2006) develops this idea, drawing comparisons to the idea of competence in linguistics. On Railton’s reconstruction of the idea, persons have unconscious representations of rules which set off signals (“feelings of discomfit”) when they are violated. Of course, those feelings of discomfit can be washed out by training or distractions or failures of concentration, suggesting that there can be failures of performance with respect to the ethical rules that one unconsciously represents.

These cases suggest that the competence/performance distinction may be in play in many aspects of human cognition — perhaps any element of human cognition that has a normative component. Whether exploring these extra-linguistic investigations of the competence/performance distinction prove fruitful is yet to be determined, however it is certainly clear that there are deep similarities.

Conclusion

The competence/performance distinction in linguistics has been deployed throughout the history of modern linguistics. As noted, the distinction must be handled with some care, as it can be used to rescue defective theories of grammar, but at the same time it can help to clarify the distinction between the core theory of grammar and other mechanisms that play a role in linguistic performance. Does linguistic competence consist merely in the representation of rules that have some causal role in our performance, or does it involve some notion of unconscious rule following? All of these questions are deserving of further investigation.

Bibliography

Arpaly N (2003). Unprincipled Virtue: An Inquiry into Moral Agency. Cambridge: Cambridge University Press.

Chomsky N (1965). Aspects of the theory of syntax. Cambridge: MIT

Chomsky N (1980). Rules and Representations. New York: Columbia University Press.

Chomsky N (1986). Knowledge of Language. New York: Praeger.

Ludlow P (2013). The Philosophy of Generative Linguistics. Oxford: Oxford University Press.

Ludlow P & Zivanovic S (2023). Language, Form, and Logic In Pursuit of Natural Logic’s Holy Grail. Oxford: Oxford University Press

Marr D (1982). Vision. MIT Press.

Railton P (2006). ´Normative Guidance.’ In Schaefer-Landau R (ed.), Oxford Studies in Metaethics, vol. 1, Oxford: Oxford University Press, 3–34.

Rips L (1994). The Psychology of Proof: Deductive Reasoning in Human thinking. Cambridge, MA, and London: The MIT Press.

from 2025, International Encyclopedia of Language and Linguistics