Concept

Intentionality

Intro

Think about Paris.

Whatever just happened in your head, it was about Paris. Not near Paris, not caused by Paris, not shaped like Paris. About it. Paris was the thing your thought reached for, and it reached across two thousand miles and did not need to move an inch.

Now look at anything purely physical and try to find that property. A rock is not about anything. A river is not about anything. A pattern of neurons firing is, considered purely as physics, a pattern of neurons firing. It has a shape, a location, a rate, an energy. Nowhere in that list is what it is about.

Philosophers call this property intentionality, and the word is confusing because it has almost nothing to do with intending or intentions. It just means aboutness, directedness, being of something. Beliefs are about things. Fears are of things. Words mean things. Maps represent places.

This is not a curiosity. It is arguably the single hardest problem for the view that everything real is physical. Not because it is mysterious in a foggy way, but because it is very precise: physical descriptions are complete without ever mentioning what a thing is about, and mental descriptions are impossible without it.

◎ See this page's graph and discover linked topics →

In full

Intentionality is the property of a state, act, or item of being directed toward, of, or about something other than itself. Franz Brentano reintroduced the scholastic term in Psychology from an Empirical Standpoint (1874), proposing that "intentional inexistence," the containment of an object within a mental act, is the mark distinguishing mental phenomena from physical ones: every mental phenomenon exhibits it, and no physical phenomenon does. This is Brentano's thesis.

Edmund Husserl took the thesis as the organizing structure of consciousness and built phenomenology around its analysis, distinguishing the noesis, the act of consciousness, from the noema, the object precisely as intended in that act. The distinction allows the same object to be intended under different aspects (the same building perceived, remembered, doubted, feared) and allows the object to be intended when it does not exist at all, which is the feature that makes intentionality resistant to any relational analysis modeled on physical relations.

Contemporary discussion divides sharply on whether intentionality can be naturalized, that is, given a reductive account in terms acceptable to physicalism. The principal programmes are informational or indicator semantics (Fred Dretske), teleosemantics (Ruth Millikan, David Papineau), and asymmetric-dependence accounts (Jerry Fodor). Each faces a well-known cluster of objections concerning misrepresentation, the disjunction problem, and the determinacy of content.

The apologetic weight of intentionality is that it is the load-bearing premise under several distinct arguments in the codex. It is what makes physical states unable to be true or false, hence the Argument from Reason. It is what a merely causal account of perception cannot supply, hence Phenomenology of Perception and Argument from Perception. And it is what makes eliminativism about the mental self-defeating, since the eliminativist's own position is an intentional state offered as being about the world and being correct.

Original and derived intentionality

The most useful distinction in this area, drawn sharply by John Searle, separates two kinds of aboutness.

Derived intentionality is aboutness a thing has because minds confer it. The word "cat" is about cats. Ink is not about anything; the marks are about cats because a community of speakers treats them so. A red octagon means stop. A thermostat "detects" temperature. A chess program "knows" the position. A photograph is "of" your grandmother. In every case the aboutness is on loan, and it evaporates the moment there is no mind left to interpret.

Original intentionality is aboutness a thing has intrinsically, not by anyone's assignment. Your thought about Paris is about Paris whether or not anyone interprets you, and no interpretive convention could make it be about Vienna instead.

The distinction matters because almost every proposed naturalistic model of aboutness turns out, on inspection, to be a model of the derived kind. Tree rings "carry information" about a tree's age, but only in the sense that an observer who knows the correlation can infer the age. The rings are not about anything on their own; they just are what they are. Point the same analysis at a brain state and the same result follows: the state correlates with things, and correlation plus an interpreter yields aboutness, while correlation alone does not.

This generates the naturalist's dilemma. If all intentionality is derived, then it is derived from something, and the regress needs a terminus in original intentionality. If some intentionality is original, then there is something in the world with a property no physical description contains.

Why the aboutness problem is hard

Syntax is not semantics

Searle's Chinese Room thought experiment (1980) makes the point compactly. A person who knows no Chinese sits in a room with an exhaustive rulebook for manipulating Chinese characters. Questions come in, the rules are followed, correct answers go out. From outside the room, the system appears to understand Chinese. Inside it, nobody understands anything. Symbols are being shuffled by shape.

The generalization: formal symbol manipulation is defined over the shapes of symbols, not their meanings, and no amount of shape-manipulation adds meaning. Computation is syntax. Understanding is semantics. The gap is not a matter of insufficient speed or scale. See Access and Phenomenal Consciousness for the parallel point that integration and reportability never add up to felt experience.

Physical states cannot be false

A belief can be false. A pattern of neural activation cannot. Falsity is a semantic property that requires content, and content is what a purely physical description omits. This is the same normative point that appears in Phenomenology of Perception about perceptual error, and it is the premise on which the Argument from Reason turns: an inference is valid or invalid in virtue of the contents of its premises and conclusion, and if there are no contents, there is no inference, only succession.

Indeterminacy

The deepest version of the problem, and the least often noticed. Suppose a brain state reliably tokens when cows are present. Is it about cows? Or about cows or dark shapes at dusk, given that it tokens for those too? Or about mammals of a certain size? The causal facts underdetermine which. Fodor's disjunction problem, Quine's inscrutability of reference (see W.V.O. Quine), and Kripke's rule-following puzzle are three routes to the same result: physical facts about correlation do not fix a unique content.

James Ross pressed this into a direct argument in "Immaterial Aspects of Thought" (1992). When you add two numbers, you are performing addition, determinately, and not the mathematically distinct function that agrees with addition on every case you will ever compute and diverges thereafter. But no physical process determines which of those functions it is realizing, since a physical process has only a finite actual history and infinitely many functions fit any finite set of cases. Therefore, Ross argues, determinate formal thinking is not a purely physical process. Edward Feser has developed the argument at length.

The argument is disputed. It is also considerably harder to answer than its brevity suggests, and it is worth understanding before deploying, because a botched version invites an easy reply.

What naturalists propose, steel-manned

The position is not helpless, and treating it as helpless is how apologists lose this exchange.

1. Informational semantics. A state represents whatever it reliably indicates under normal conditions. Smoke means fire because smoke reliably indicates fire. Best objection it faces: misrepresentation. If content is fixed by what a state reliably indicates, a state that fires wrongly is not misrepresenting; it is correctly indicating a wider disjunction. The theory has trouble making room for error, which is the thing it most needs to explain.
2. Teleosemantics. A state represents what it was selected by evolution to track. The frog's detector is about flies, not about small dark moving things, because tracking flies is what it was selected for. Best objection: selection history is as indeterminate as causal correlation. Selection favored the frog's eating of nutritious moving specks, and "fly," "nutritious moving speck," and "food" are all consistent with the selection facts. Swampman cases press the same point from the other side.
3. Functional-role semantics. Content is fixed by a state's inferential role in the whole system. Best objection: holism. If content depends on the total network, then no two people ever share a belief, and beliefs change content whenever anything else does.

None of these is absurd, and each captures something. What none has yet delivered is a determinate, error-permitting, non-derived content out of purely physical resources. That is the honest state of play, and it should be stated that way rather than overstated.

The theistic account

On classical theism the terminus of the regress is not a puzzle to be solved but a doctrine already held. God is not a physical system whose states happen to correlate with things. God knows, and divine knowing is not derived from anyone's interpretation. Original intentionality has a home.

Created minds, on this account, are not machines that somehow bootstrapped semantics out of syntax. They are made in the image of a knowing God (see Imago Dei) and are intentional derivatively in the ontological sense, dependent on God for their existence and capacities, while intentional originally in the semantic sense, since their thoughts are about what they are about without a further interpreter standing behind them.

The Johannine identification of the Logos as both divine and creative (John 1:1 to 3) sits naturally with this: intelligibility and aboutness are not late arrivals in a fundamentally mindless universe but belong to what is most basic. See Argument from Intelligibility.

Two cautions. First, this is an explanatory advantage, not a demonstration. A naturalist may reasonably hold that the problem is unsolved rather than unsolvable, and the honest form of the argument concedes that and then presses on comparative explanatory adequacy. Second, do not claim that theism explains the mechanism of intentionality. It does not offer a mechanism. It offers an ontology in which aboutness is fundamental rather than anomalous.

See also

• Phenomenology, the method built around the analysis of intentional structure
• Phenomenology of Perception, where intentionality does the work in a specific case
• Argument from Perception, the apologetic argument that depends on it
• Argument from Reason, which turns on physical states being unable to be true or false
• Argument from Consciousness, the neighboring hard problem
• Access and Phenomenal Consciousness, the parallel point that function never adds up to feeling
• Materialism, the position the aboutness problem presses against
• Naturalism, and the naturalizing programmes surveyed above
• Self-refutation, the structure that catches eliminativism about the mental
• Edmund Husserl, who made intentionality the organizing structure of consciousness
• W.V.O. Quine, whose inscrutability-of-reference argument is one route to the indeterminacy result
• Edward Feser, who has developed the Ross argument at length
• Argument from Intelligibility, the wider claim that mind is basic rather than derivative
• Universals, the related question of how thought grasps general kinds
• Theories of Truth, since truth-bearing presupposes content