Multiple realizability

For an example of parallel claims about multiple realizability in psychology and its consequences, see Bechtel and Mundale (1999). As Van Regenmortel (1999) has remarked with respect to structure-function relations ... 

Bechtel, W. and Mundale, J. (1999), Multiple realizability revisited linking cognitive and neural states , Philosophy of Science, 66, 175-207. 

Figure 6.8 Observed growth in tracer spatial variance agrees with theoretical prediction. Velocity is 0.9 cm s-1 (x), 2.0 cm s 1 (o), 3.2 cm s-1 (A), 4.1 cm s-1 ( ). Vertical bars indicate standard deviation of multiple realizations for same conditions. Predicted growth based on equation (6.29) (solid line) and equation (6.30) (dashed line).

The realization is described by a class diagram containing a class for the respective task net as well as classes for the subtasks. Although multiple realizations may be defined, we discuss only a simulation based realization (SimulationBased ). The inputs and outputs attached to SimulationBased are internal parameters which are used for vertical communication with elements of the refining task net. 

Multiple realizations of conductivities at the small scale of 5 m by 5 m cells are generated. For each realization, the numerical modelling described in the previous subsection is applied. The realizations are generated independently for each formation type and for the vertical fracture in each formation type by sequential simulation using the code gcosmSd (Gomez-Hemandez, 1991). For each formation the same spatial correlation (10 m range... 

In regression analysis, the term multiple realizations of an x value is common instead of factor combinations. 

This observation has been thought by some to constitute an outright refutation of type-identity materialism. However, reductionists have employed various strategies for defending reductionism in light of observations about multiple realizability. I will not discuss the merits of these attempts in this dissertation. I will say, however, that Kim s... 

My criticisms might not apply to a view that is less reductive than Kim s. Kim s view seems to have an advantage over more reductive type-identity materialist views, because Kim s functional reduction seems to be able to account for the multiple-realizability of higher-level properties in a way that type-identity materialism cannot. However, Kim s theory of functional reduction ultimately fails, because it cannot account for the context dependence of many higher-level properties. If there were some version of reductive materialism that could somehow account for the context dependence of higher-level properties, then this view might be immune to many of my criticisms. However, we might wonder how a view that allows for context dependence could even be considered reductionist at all. 

The lower-level-higher-level distinction is really just about multiple realizability. To say a property is a higher-level property just means that it is multiply realizable with respect to its lower-level instantiation. So, for instance, macroproperties like size can be higher-level properties, since size is multiply realizable. We could, for instance, have five balls that are the exact same size, but each ball could have an entirely different microstructure (e.g., one is made of plastic, one of metal, one of wood, etc.). 

Because of the difficulties associated with Nagel s bridge-law model of reduction and the success of multiple realization arguments against type-identity materialism, Kim has come to reject type-identity materialism in favor of what he ealls functional reduction. The functional reduction of a property requires three steps ... 

Kim now realizes that because of multiple realizability, we cannot expect to find... 

Ron McClamrock (1995a, ch. 3) has argued that we can use the ideas of multiple realizability and context dependence to pick out preferred levels of causation and explanation. When we have token identity at different levels, we can use the ideas of multiple realizability and context dependence to hold properties at one level fixed while varying properties at other levels, in order to try to see at which level the causal mechanism occurs. In the ball-sorter case, we see that the causation occurs in virtue of size and that the macro-level of size is the right level of explanation. 

We can use multiple realizability and context dependence to help us try to isolate the properties in virtue of which the event is caused without being forced to exclude properties at other levels as also causal. 

I have argued that if we take a closer look at causation, we see that there is no plausible reason for accepting the exclusion principle. We can use the ideas of multiple realizability and context dependence to help us isolate which properties are causal properties. When we do this, we can see that there is nothing about causation that ought to lead us to think that sufficiency at one level will rule out causation at other levels. 

Aecepting nonreductive materialism gets us what we want and need out of a theory of mental causation. We can have sufficient causation at the physical level and causation in virtue of irreducible mental properties, without any sort of problematic overdetermination. We can account for both the multiple realizability and the context dependence of many mental properties. And we can account for why mental explanations often seem to offer us better predictive and explanatory power than lower level physical ones. Mental explanations work because menial properties are causal properties. 

It is clear that mental properties are multiply realizable, and there also seem to be good reasons to think that they are context dependent as well. For this reason, it seems that nonreductive materialism has the most initial plausibility, because it can account for the multiple realizability and context dependence of mental properties. Although we can argue over whether mental properties are context dependent, we can t deny that they are multiply realizable. Since type-identity materialism has a hard time dealing with multiple realizability, it seems to be at a clear disadvantage. Kim s theory of ftmctional reduction has an advantage over type-identity materialism because it does have a way of accounting for the multiple realizability of mental properties. So the real debate here, as I see it, is between nonreductive materialism and functional reduction. 

The features are (a) multiple realizability, (b) externalism, and (c) the existence of an explanatory gap between mental and physical facts. 

At this point it may be useful to remind ourselves what is at stake in the exclusion argument. Various considerations (functionalism, multiple realization, externalism, explanatory gap) make M P persuasive. Also, there are scientifically compelling reasons to accept physicalism and Closure. Giving up mental causation is a last resort. So, Kim s argument is a paradox. Each o] M P, Closure, Mental Causation, and Exclusion is plausible, but together they are inconsistent. We proponents of NRP accept M P, Closure, and Mental Causation, so we have to reject Exclusion. 

Many suppose that it is conceivable that there be a realization of human functional organization that is mentally different from ours, for example, inverted or absent qualia. The argument just given provides a case for multiple realization of even the lowest level of physics. 

Kim, J. (1992). Multiple realization and the metaphysics of reduction. Phibsophy and Phenomenobgical Research, 52,1—26. 

---