Understanding Functionalism: Varieties, Formalizations, and Philosophical Disagreements

The Varieties and Origins of Functionalism

Functionalism represents a diverse set of doctrines across various fields, including anthropology, literary criticism, psychology, and the philosophy of psychology. While these applications might share little more than a name, the versions of functionalism found within the philosophy of psychology form a closely-knit group with a common historical origin in the works of Aristotle. Three primary forms of functionalism have significantly influenced the fields of philosophy of mind and psychology: Functional Analysis, Computation-Representation Functionalism, and Metaphysical Functionalism. Each of these frameworks offers a distinct perspective on how to understand, explain, or define mental processes and states.

Functional Analysis as a Research Strategy

Functional analysis is primarily defined as a type of explanation and a corresponding research strategy. This approach seeks to explain the working of a complex system by decomposing it into its component parts. The explanation is grounded in the specific capacities of these parts and the way they are integrated with one another to produce the overall behavior of the system. Robert Cummins (1975) provides a definitive description of this sense of functionalism, which is also supported by the works of Fodor (1965, 1968a, 1968b) and Dennett (1975). A practical example is the production of refrigerators in a factory; one explains the output by appealing to the capacities of various assembly lines, individual workers, heavy machinery, and the overarching organization of these components.

Computation-Representation Functionalism and the Language of Thought

Computation-representation functionalism is a specialized application of functional analysis that views psychological explanation as analogous to providing a computer program for the mind. This doctrine suggests that the perceived mystery of mental life is dissolved when mental processes are analyzed as being composed of computations as mechanical as the primitive operations of a digital computer. These underlying processes are described as "stupid" or mechanical to ensure that the psychological explanation avoids question-begging. The foundational concepts here are representation and computation. Psychological states are viewed as systematically representing the world through a language of thought, and psychological processes are treated as computations that involve these internal representations.

Metaphysical Functionalism: The Nature of Mental States

Metaphysical functionalism is a theory regarding the nature of the mind rather than a theory of psychological explanation. It asks "What are mental states?" rather than how they account for behavior. The answer provided is that mental states are functional states, often leading to the description of these theses as functional state identity theses. This doctrine shares the same concerns as behaviorism and physicalism, addressing questions such as "What is pain?" or seeking the commonality between all instances of pain that makes them pains. Significantly, metaphysical functionalism focuses on mental state types rather than tokens. While a functionalist might concede that every particular token of pain is a physical state (token physicalism), they argue that what makes those tokens "pain" is a shared functional property rather than a shared physical property.

Characterizing Mental States via Causal Roles

Metaphysical functionalists characterize mental states by their causal roles, specifically their relations to sensory stimulations, behavioral outputs, and other mental states. For instance, a theory of pain might define it by its tendency to be caused by tissue damage ($Input$), its tendency to cause a desire to be rid of the sensation ($Other mental state$), and its tendency to produce actions such as flinching or withdrawl ($Output$). The specific disagreements among functionalists regarding these characterizations prevent an easy single definition of the doctrine, yet all share a commitment to the causal role framework.

The Role of Turing Machines in Functionalist Thought

All three described branches of functionalism share a connection to the concept of the Turing machine. Metaphysical functionalism often identifies mental states with the "table states" of a Turing machine. Computation-representation functionalism aims to provide a functional analysis where mental capacities are broken into mechanical, algorithmic processes. As per the Church-Turing thesis, if these processes are algorithmic, they are also Turing-computable. However, if mental representations are analog (as suggested by experiments on mental imagery), they might not satisfy the requirements of the Church-Turing thesis. A Turing machine is formally defined by two functions: one mapping inputs and states to outputs, and one mapping inputs and states to subsequent states. This is often represented as a machine table comprising conditionals: If the machine is in state $S$ and receives input $I$, it will emit output $O$ and transition to state $S'$.

Formalizing Functionalism: Ramsey Sentences and Logic

To achieve a precise formulation of functionalism, philosophers use the Ramsey sentence method. Let $T$ be a psychological theory describing the relations between pain, other mental states ($s_1, ..., s_n$), sensory inputs, and behavioral outputs. The theory is written as a conjunctive sentence $T(s_1, ..., s_n)$. By replacing mental state terms with variables ($x_1, ..., x_n$) and prefixing existential quantifiers, we form the Ramsey sentence: $Ex_1 ... x_n T(x_1 ... x_n)$. If $x_i$ replaces "pain," then the property of having pain is defined as: $y$ has pain if and only if $Ex_1 ... x_n [T(x_1 ... x_n) \,\&\, y \text{ has } x_i]$. Using the notation $\%x Fx$ to mean "the property of being an $x$ such that $x$ is $F$," pain is identified as the property: $\text{pain} = \%y Ex_1 ... x_n [T(x_1 ... x_n) \,\&\, y \text{ has } x_i]$. This identifies pain with an abstract causal property tied to the world only via its relational structure to inputs and outputs.

Functionalism versus Behaviorism

Functionalism is frequently seen as a descendant of behaviorism, yet it avoids the major pitfalls of that doctrine. Behaviorists attempted to define mental states as pure dispositions ($\text{e.g.,}$ Ryle defined brittleness not as a cause of breaking, but as the fact of breaking easily). However, functionalists (like Lewis and Armstrong) emphasize that mental states are real, causal entities. There are two main differences: first, functionalists allow mental states to be the effects of stimuli and the causes of behavior; second, functionalists emphasize the connections between one mental state and other mental states. Behaviorism fails to account for the fact that a disposition to reach for an ice-cream cone depends on knowledge and other desires ($\text{ceteris paribus}$ or "other things being equal"). Furthermore, the "perfect actor" counterexample (Putnam, 1963) shows that one can have the behavioral disposition of pain-absence while actually being in pain, proving behavioral dispositions are neither necessary nor sufficient for mental states.

Multiple Realizability and the Critique of Physicalism

One of the most significant insights of functionalism is multiple realizability. A functional state, like a state in a "Coke machine" automaton, can be realized by many different physical systems (electronic, hydraulic, or mechanical). Because the same machine table could describe a human brain and a series of radio-linked objects, it is argued that mental states cannot be identical to specific brain states. This led Source I functionalists (Putnam and Fodor) to argue that physicalism is likely false. If pain is a machine table state, and that table can be realized by non-brain systems, then pain is not a brain state. Even if every system that has pain is physical (token physicalism), the property that makes them all "pain" is functional, not physical.

Functional State Identity versus Functional Specification

A major disagreement exists between functional state identity theorists (Source I: Putnam, Fodor) and functional specifiers (Source II: Lewis, Armstrong, Smart). For identity theorists, pain is the functional property itself ($\text{pain} = \text{the property of having causal role } R$). For functional specifiers, pain is the physical state that happens to occupy the causal role ($\text{pain} = \text{the } x \text{ that occupies role } R$). Lewis and Armstrong use this to argue for the truth of physicalism: conceptual analysis shows pain is the occupant of role $R$, and science shows a neural state occupies role $R$, therefore pain is that neural state. However, the identity theorist responds that what all pains (in humans, dogs, or Martians) share is the role $R$ itself. Characterizing pain as a brain state in humans and a different state in dogs (species-specific identities) sidesteps the metaphysical question of what commonality makes both sensations "pain."

The Reductionist Nature of Functionalism

Functionalism provides a form of reduction without elimination. While it reduces mental terms to input-output structures (and the relations between states), it does not treat mental states as fictions. By using existential quantification in Ramsey sentences, the doctrine commits to the existence of internal states that satisfy the specified causal roles. As Armstrong (1977) and Shoemaker (1975) indicate, mental concepts are introduced together as a system, reflecting their logical interdependence. Ultimately, despite the bizzare disagreements between the various camps, all metaphysical functionalists are committed to the insight that mental states are defined by their position within a complex causal network rather than their internal physical composition.