“Sense” is the distinctive central notion in theories of thought and language inspired by the later work of Gottlob Frege (“sense” translates Frege’s Sinn). For Frege what we think (not the act of thinking it) is a thought, an abstract object. Thoughts have quasi-syntactic structure.
Any simple or complex constituent of a thought, even the thought itself, is a sense; thus, senses are abstract. Frege assumes that it is irrational to assent to a thought and simultaneously dissent from it. Since someone misled about astronomy may rationally combine assent to the thought that Hesperus is Hesperus with dissent from the thought that Hesperus is Phosphorus, the thoughts are distinct.
Although the names “Hesperus” and “Phosphorus” have the same reference, they express different sense, two modes of presentation of one planet. The role of a sense is to present the thinker with a reference—that is, something on which the truth-value (truth or falsity) of the thought depends; if the sense fails to present a reference, the thought lacks a truth-value.
For Frege the truth-value of a thought is independent of where, when, and by whom it is thought. Thus, what reference a constituent sense presents is independent of when, where, and by whom it is thought. Sense determines reference, not vice versa.
Frege used his notion of sense to analyze the semantics of thought attributions in natural language, as in the sentence “Someone doubts that Hesperus is Phosphorus.” On Frege’s account expression within such “that” clauses refer to their customary senses.
This explains the presumed failure of the inference from that sentence and “Hesperus is Phosphorus” to “Someone doubts that Hesperus is Hesperus”: The two names have different references within “that” clauses, for their customary senses are different.
If sense determines reference, then the sense of “Hesperus” in “Someone doubts that Hesperus is Phosphorus” defers from its sense in “Hesperus is Phosphorus,” since the reference differs. By appeal to iterated attributions such as “He doubts that she doubts that Hesperus is Phosphorus,” it can be argued that Frege is committed to an infinite hierarchy of senses. His account involves the assignment of senses to natural-language expressions.
However, in order to understand many words (e.g., proper names and natural-kind terms), there is arguably no particular way in which one must think of their reference; they do not express senses common to all competent speakers. Fregeans therefore distinguish sense from linguistic meaning but in doing so sacrifice Frege’s original account of thought attributions.
Sense must also be distinguished from linguistic meaning for context-dependent expressions such as “I.” Two people may think “I am falling” and each refer to themselves, not the other. Since the references are distinct and sense determines reference, the senses are distinct, even though the mode of presentation is the same. Others cannot think the sense that one expresses with “I”; they can only think about it.
Communication here does not amount to the sharing of thoughts, and “You think that I am falling” does not attribute to the hearer the thought that the speaker expresses with “I am falling.” In contrast, the linguistic meaning of “I” is the same for everyone; it consists in the rule that each token of “I” refers to its producer. Unlike a sense, the rule determines reference only relative to context.
Such cases reveal tensions within Frege’s conception of sense. Sense cannot be both what determines reference and how it is determined. Since senses can be qualitatively identical but numerically distinct, they are not purely abstract objects, if qualitatively identical purely abstract objects must be numerically identical.
Although Fregeans distinguish sense from linguistic meaning, they still treat a given speaker on a given occasion as expressing senses in words. Frege gave the impression that the sense expressed by a word was a bundle of descriptions that the speaker associated with it: the word refers to whatever best fits the descriptions.
However, this descriptive model of reference has fared badly for proper names and natural-kind terms. Nondescriptive models may also allow different routes to the same reference, but that is a difference in sense only if it is a difference in presentation to the thinker.
In spite of these problems a role for something like sense remains. An account is needed of the deductions that thinkers are in a position to make. When, for example, is one in a position to deduce “Something is black and noisy” from “That is black” and “That is noisy”?
It is necessary but not sufficient that the two tokens of “that” refer to the same thing, for, even if they do, the thinker may lack evidence to that effect: Perhaps one refers through sight, the other through hearing. What is needed is more like identity of sense than identity of reference.
Thus, the theory of rational inference may still require a notion of sense. It does not follow that thinkers are always in a position to know whether given senses are identical, for it is not obvious that they are always in a position to know what deductions they are in a position to make.