naturalizedTruthAndPlantinga

Naturalized Truth and Plantinga’s Evolutionary Argument against Naturalism

Feng Ye*

 

Abstract   There are three major theses in Plantinga’s latest version of his evolutionary argument against naturalism. (1) Given materialism, the conditional probability of the reliability of human cognitive mechanisms produced by evolution is low; (2) the same conditional probability given reductive or non-reductive materialism is still low; (3) the most popular naturalistic theories of content and truth are not admissible for naturalism. I argue that Plantinga’s argument for (1) presupposes an anti-materialistic conception of content, and it therefore begs the question against materialism. To argue for (2), Plantinga claims that the adaptiveness of a belief is indifferent to its truth. I argue that this claim is unsupported and likely false unless it again assumes an anti-materialistic conception of content and truth. I further argue that Plantinga’s argument for (3) is not successful either, because an improved version of teleosemantics can meet his criticisms. Moreover, this improved teleosemantics shows that the truth of a belief is (probabilistically) positively related to its adaptiveness, at least for a class of simple beliefs. This contradicts Plantinga’s claim that adaptiveness is indifferent to truth.

 

Keywords  Plantinga, Evolutionary Argument, Naturalism, Materialism, Content, Truth

 

Introduction

Plantinga’s evolutionary argument against naturalism has undergone several revisions and refinements.[1] I will focus on his latest version, since it contains new refinements. Moreover, I will focus on the three theses that Plantinga defends. Let N, E, RM, and NRM denote naturalism, evolutionary theory, reductive materialism, and non-reductive materialism respectively; let NC be any naturalistic theory of content and truth; and let R be the proposition ‘human cognitive mechanisms are reliable (for producing true beliefs)’. Plantinga argues that

(1)    The conditional probability P(R|N&E) is low.

(2)    P(R|RM&N&E) and P(R|NRM&N&E) are still low.

(3)    [While P(R|NC&RM&N&E) may be high,] [2] the most popular proposals for NC (i.e., functionalism, teleosemantics, and the one proposed by Dretske) are not admissible for naturalism.

Plantinga admits that naturalism implies materialism. Therefore, I will simply understand N above as materialism.

Plantinga’s idea is that we come to believe in N&E by using our cognitive mechanisms. Therefore, given (1), if N&E is true, we will have little reason to believe that our own cognitive mechanisms are reliable, and then we will have little reason to believe that N&E is true. That is, N&E is self-defeating. (2) and (3) are meant to deal with some attempts by materialists to defeat this conclusion by accepting other beliefs in addition to N&E. For instance, if RM is consistent with materialism and P(R|RM&N&E) is not low, then materialists can defeat (1) by accepting RM. Plantinga considers three beliefs RM, NRM, and NC that materialists may accept in addition to N&E. He argues that P(R|RM&N&E) and P(R|NRM&N&E) are still low, but he argues that the most popular proposals for NC are not admissible for naturalism.

In a recent article, Xu argues that the proposition R formulated by Plantinga is not ‘a qualified explanandum crying out for scientific explanation’.[3] However, we can consider the following R*, a restricted version of R:

(4)    Human cognitive mechanisms are reliable for producing true simple beliefs about ordinary physical objects in human environments, such as the beliefs produced in recognizing and classifying foods and predators.

Now, Plantinga’s arguments do not discriminate between simple and complex beliefs. If his arguments for (1) and (2) are valid, they should still be valid when R is replaced by R*. This will imply that naturalistic evolution cannot explain R*. However, R* seems clear and specific enough and it does cry out for scientific explanation.[4] Moreover, if the probability P(R*|N&E) is low, we can similarly conclude that N&E is self-defeating, since the cognitive mechanisms mentioned in R* are the bases for the cognitive mechanisms needed for producing the more complex belief N&E. Therefore, the problem raised by Xu is perhaps not an essential problem in Plantinga’s argument.

There are other criticisms on Plantinga’s argument before Xu’s criticism. Some of them focus on issues other than the theses (1), (2), and (3) above, and others object to the conclusion (1) or (2) without looking into why Plantinga is wrong in reaching (1) or (2). Besides, none of them addresses the relevance of (3). I will not discuss these objections in this paper.[5] Apparently, in his latest presentation of the argument, Plantinga believes that none of these objections has revealed any essential problem in his argument.[6] I agree with his assessment to some extent, and I will briefly explain why in a moment, but this paper is intended to reveal what I think an essential problem in Plantinga’s argument.

In this paper, I will first argue that Plantinga presupposes an anti-materialistic conception of content in his argument for (1). At that point, I will not discuss what a materialistic conception of content should be, but I will explain why no serious materialists will accept the conception of content implied in Plantinga’s argument for (1) (no matter if materialists can in the end invent any viable materialistic conception of content). Therefore, Plantinga begs the question against materialism here. Then, for (2), I will focus on P(R|RM&N&E). Here, in order to show that this probability is low, Plantinga claims that the adaptiveness of a belief is indifferent to its truth. I will argue that this claim is unsupported and likely false unless it again assumes an anti-materialistic conception of content and truth. Therefore, Plantinga’s arguments against materialism based on (1) or (2) are either question-begging or inconclusive.

Now, Plantinga’s (3) is intended to imply that no materialistic theory of content and truth is viable. If (3) holds, the immediate consequence is that materialism has to hold eliminativism regarding belief, content, and truth. Plantinga never tries to refute eliminativism in the cited articles, but I personally agree that eliminativism has its own serious problems, and therefore I agree that, if Plantinga’s argument for (3) is successful, it will be a serious challenge against materialism. Note that this will be an independent argument against materialism, having nothing to do with the evolutionary argument, for it actually says that materialism is wrong because there are such things as belief, content, and truth but materialism cannot accommodate them. However, I will argue that Plantinga’s argument for (3) is not conclusive either, because there is a way to improve teleosemantics, one of the naturalistic theories of content and truth discussed by Plantinga, to reject Plantinga’s criticisms. I will briefly explain this improved teleosemantics.

Then, going back to Plantinga’s evolutionary argument, I will explain how, according to the naturalistic conception of truth offered by that improved teleosemantics, a belief is more likely to cause adaptive behaviors when it is true than when it is false, at least for simple beliefs about ordinary physical objects in human environments. This contradicts Plantinga’s main reason for supporting (2), namely, his claim that adaptiveness is indifferent to truth, since Plantinga does not discriminate between simple and complex beliefs in that claim. Moreover, this actually shows that the probability P(R*|RM&N&E) is high, assuming that RM takes that improved teleosemantics as its theory of content and truth. This may not contradict Plantinga’s thesis (2) directly, but it discredits Plantinga’s argument for (2), since his argument contains nothing that can discriminate between R and R*.[7]

Therefore, Plantinga’s evolutionary argument against materialism is either question-begging (by assuming an anti-materialistic conception of content and truth) or based on the wrong assumption that adaptiveness is indifferent to truth (when accepting the naturalistic conception of truth offered by that improved teleosemantics). Moreover, his independent argument against materialism in (3) is not successful either (because that improved teleosemantics can meet his criticisms).

The intuition behind Plantinga’s evolutionary argument seems to be that, under materialism, truth as a property of beliefs has to be (probabilistically) independent of adaptiveness. If that is true, then it is indeed very improbable that naturalistic evolution will produce human cognitive mechanisms reliable for producing true beliefs, since naturalistic evolution is driven by adaptiveness alone. There are already objections to the idea that false beliefs can be equally adaptive.[8] However, these objections are based on our intuitive observation that our true knowledge helps us survive in the world. Plantinga will deny that truth is positively related to adaptiveness in our world (which, as Plantinga believes, is not a result of naturalistic evolution). He only claims that it is hard to imagine how such a connection between truth and adaptiveness could come to be under materialism and naturalistic evolution.[9] These previous objections to Plantinga have not shown how, under materialism, truth can be positively related to adaptiveness. Therefore, they have not challenged the intuition behind Plantinga’s evolutionary argument. Now, as a naturalistic theory of content and truth, teleosemantics is exactly meant to characterize truth as a relation between brain states and environmental states by referring to evolution. Therefore, truth has to be closely related to adaptiveness. For instance, according to my improved version of teleosemantics, evolution selects simultaneously human cognitive mechanisms and truth as a correspondence relation between human brain states and environmental states. That is why evolution can produce human cognitive mechanisms reliable for producing truths.[10]  Therefore, as long as some version of teleosemantics is at least reasonable, Plantinga’s intuition behind his evolutionary argument is dubious. That is perhaps the real, essential problem in his argument.

 

Plantinga’s argument for (1)

In his argument for (1), Plantinga admits that materialism implies that a belief is a neural structure and that a belief causes behaviors by virtue of its neural-physiological properties (NP properties). This means that a neural structure can somehow possess content. Then, Plantinga claims,

 ‘It is easy to see how beliefs thus considered can enter the causal chain leading to behavior; …  It is exceedingly difficult to see, however, how they can enter that chain by virtue of their content; a given belief, it seems, would have had the same causal impact on behavior if it had had the same NP properties, but different content.’[11]

This will imply that the adaptiveness of a belief (or the behaviors caused by the belief) is indifferent to the content of the belief, which in turn implies that adaptiveness is indifferent to the truth of the belief, for the content of a belief exclusively determines its truth-value. This means that, given that a belief is adaptive, the conditional probability of its truth is about 50%. Then, as long as there are sufficiently many probabilistically independent beliefs, by the law of large numbers, the probability of the event ‘at least 75% of the beliefs selected by their adaptiveness are true’ will be very low. This will imply that the probabilities of R and R* are very low. All other things aside, this conclusion alone is already a very serious problem for materialism. It means that naturalistic evolution cannot even explain how humans come to have mostly true beliefs about their foods and predators. 

Now, what conception of content does Plantinga assume in this claim? He has to assume some conception of content, since he heavily relies on the notion of content here. However, it cannot be any materialistic conception of content. An easy way to see this is to look ahead a little to note that, in arguing for (2), Plantinga admits that according to both reductive and non-reductive materialism the content of a neural structure (as a belief) supervenes on its NP properties (as well as the environmental states, since content may be broad). Then, according to both reductive and non-reductive materialism, a neural structure could not have had the same NP properties but different content. Now, materialism is just either reductive materialism or non-reductive materialism. If both reductive and non-reductive materialism hold that a neural structure could not have had the same NP properties but different content, how can Plantinga claim here in arguing for (1) that, given materialism in general, a neural structure could have had the same NP properties but different content? He seems to be contradicting himself.

Therefore, here Plantinga must have assumed a conception of content that no materialists (neither reductive nor non-reductive materialists) accept. This conception of content implies that the content of a neural structure does not supervene on the materialistic (physical, chemical, or neural-physiological) properties of material things. Then, the conclusion (1) is unsurprising. Naturalistic evolution can affect only material things and their materialistic properties. If the content of a neural structure does not supervene on materialistic properties,[12] then it will certainly be a huge coincidence that naturalistic evolution selects mostly neural structures (as beliefs) with true content. Of course, if Plantinga’s argument for (2) is successful, this is not a serious problem. He can disregard (1). However, a careful analysis of his fallacy in arguing for (1) is helpful, because we will see that his argument for (2) has a similar problem.

One might think that the content of a belief is an abstract entity, such as a proposition, which is not a material thing and does not have any causal power. This may be true. However, for both reductive and non-reductive materialism, a belief can still be adaptive or maladaptive in an environment by virtue of its content, which means that the belief can be adaptive or maladaptive by virtue of its NP properties and its environmental states, which together metaphysically determine the content of the belief by the supervenience relation. What can have causal power are content properties in the format ‘having so and so content’ of a neural structure, which supervene on NP properties and environmental states, not a content as an abstract entity. If such content properties can have causal power, then naturalistic evolution can potentially select neural structures with true content.

At this point, I am not saying that there is a viable materialistic conception of content. My point is merely that assuming an anti-materialistic notion of content makes Plantinga’s argument for (1) question-begging and irrelevant. Moreover, we will see that Plantinga’s argument for (2) has a similar problem. Then, the idea of an evolutionary argument against materialism fails. The only relevant part of his entire argument will be (3), which, if successful, will actually be an independent argument against materialism.

 

Plantinga’s argument for (2)

I will focus on his argument for the thesis that P(R|RM&N&E) is low. Plantinga admits that reductive materialism implies that a belief is a neural structure and the property ‘having so and so content’ of a belief is a neural-physiological property (NP property) P (relative to some environmental state, since content can be broad). Then, Plantinga makes a similar claim:

 ‘Now we may assume that having P is adaptive in that it helps to cause adaptive behavior. But (given no more than N&E&RM), we have no reason at all to suppose that this content, the proposition q such that P is the property having q as content, is true; it might equally well be false.’[13]

This means that adaptiveness is indifferent to truth (if not indifferent to content). Then, by the same probabilistic reasoning, it easily follows that P(R|RM&N&E) is low.

Now, on what ground can Plantinga claim that this content ‘might equally well be false’? One way to see the problem in this claim is also to look ahead a little. In arguing for (3) Plantinga appears to admit that, if any naturalistic theory of content and truth (denoted as NC) is viable, then the conditional probability P(R|NC&RM&N&E) can be high, which means that the truth or falsehood of a belief can affect its adaptiveness. Plantinga did not say this explicitly, but he did not argue again that the probability P(R|NC&RM&N&E) is low. Instead, he argued that the most popular proposals for NC were not admissible for materialism. Therefore, we can reasonably infer that Plantinga may admit that, as long as some version of NC is admissible for materialism, P(R|NC&RM&N&E) will be high. However, reductive materialism is the idea that all properties are in principle reducible to materialistic properties. These should include the semantic properties ‘having so and so content’, ‘true’, and ‘false’ of a belief (as a neural structure), unless a materialist holds eliminativism on semantic notions, which I will not consider here. Therefore, reductive materialism (RM) has to contain some coherent naturalistic theory of content and truth (NC) as a part of it, as long as it is coherent (and is not eliminativism). Then, given reductive materialism, one should not hastily claim that adaptiveness is indifferent to truth and that the probability P(R|RM&N&E) is low, as long as one admits that P(R|NC&RM&N&E) can be high.

Note that Plantinga did not give any further reason to support his claim that adaptiveness is indifferent to truth. His reason seems to be merely that he himself cannot imagine how truth is related to adaptiveness under reductive materialism. This alone is not a sufficient reason. Since he is evaluating the conditional probability P(R|RM&N&E), where the background condition includes reductive materialism, he has to take reductive materialism seriously and has to assume the full strength of reductive materialism. His qualification ‘given no more than N&E&RM’ reveals that he may take reductive materialism too lightly here. He ignores the fact that reductive materialism is supposed to be a comprehensive worldview, including views about consciousness, intentionality, content, truth, and everything. Again, I am not assuming here that reductive materialism is coherent. If Plantinga can successfully argue for (3), then he will have an independent argument against the coherence of reductive materialism. The concern here is merely whether his evolutionary argument contains any new challenge against materialism. It is perhaps common knowledge that materialism (or physicalism) has to meet the challenge of offering a coherent account of content, truth, phenomenal consciousness, and so on and so forth.

Another way to see the problem in Plantinga’s claim is to ask what truth ought to be according to reductive materialism. I will consider only beliefs about ordinary physical objects in human environments. First, note that content can be broad. That is, we have to say ‘the content of a belief with respect to an environment’, not just ‘the content of a belief’. In other words, ‘having so and so content’ is a property of beliefs with respect to their environments. ‘True’ is a property of content. It distinguishes true content from false content. Therefore, ‘true’ is also a property of beliefs with respect to their environments. That is, it distinguishes beliefs that are true in their environments from those that are false. Now, as Plantinga admits, under reductive materialism, the property ‘having so and so content’ is essentially a materialistic property of neural structures with respect to their environments. Therefore, truth should also be a materialistic property of neural structures with respect to their environments under reductive materialism. Then, since Plantinga admits that, under reductive materialism, content properties as materialistic properties can be adaptive or maladaptive, why does he think that truth, another materialistic property of neural structures, is different? Moreover, note that adaptiveness is also a materialistic property of neural structures with respect to their environments. Then, there is a chance that truth and adaptiveness (as two materialistic properties) are related to each other probabilistically. That is, it may happen that when the property truth exists for a neural structure with respect to its environment, the property adaptiveness is more likely to exist than not. In that case, evolution will favor neural structures that are true.

Intuitively, we naturally think that this ought to be the case. To see this, let us consider an intuitive example by Draper.[14] Draper wants to take a bath and there is an alligator in Draper’s tub; true beliefs such as ‘there is an alligator in my tub’ will save Draper’s life, but false beliefs such as ‘there is nothing in my tub’ will cause maladaptive behaviors. Let us examine this under reductive materialism more carefully. A neural structure S as a belief causes Draper’s behavior (jumping away or stepping into the tub) at that moment. There are two possible cases regarding the property ‘true’ for that neural structure S and its potential environments. Case (i): S has the property ‘true’ with respect to the environments in which there is an alligator in the tub, but S does not have the property ‘true’ with respect to other environments. Case (ii): S does not have the property ‘true’ with respect to the environments in which there is an alligator in the tub, but S may have the property ‘true’ with respect to some other environments. Intuitively, we naturally think that a neural structure in case (i) is more likely to cause the adaptive behavior (i.e., jumping away) and a neural structure in case (ii) is more likely to cause the maladaptive behavior (i.e., stepping into the tub). Then, evolution will favor the cases where the involved neural structure is true in the corresponding environment. This is the commonsense view that knowledge is power. I admit that at this point it is not very clear why a neural structure in case (i) is more likely to cause adaptive behaviors. I will explain this when I propose amendments to teleosemantics later in this paper. However, at this point, we can at least say that as long as truth is a materialistic property of neural structures with respect to their environments, without further arguments, we have no reason to claim that adaptiveness is indifferent to truth. That is, at least we can say that Plantinga’s argument for (2) is unsupported.

Plantinga’s response to this example claims that the objector conflates beliefs with ‘indicator representations’.[15] An indicator is a neural structure that represents other states by causal or other connections as a matter of natural regularity. It is neither true nor false. For Draper’s example of alligator in his tub,  

‘Suppose m holds false beliefs, believing at the time in question that the alligator is a mermaid, or even that he’s sitting under a tree eating mangoes. Will that adversely affect his fitness? Not just by itself. Not if m has indicators and other neural structures that send the right messages that cause his muscles …. It’s having the right neurophysiology and the right muscular activity that counts.’[16]

However, this is very puzzling from a materialist’s point of view. Why does Plantinga think that all those irrelevant beliefs as neural structures can also cause one to jump away from the alligator? Some specific belief causes Draper to jump away at that particular moment, and we have good intuitive reason to think that the belief has the content ‘there is an alligator in my tub’. Now, under reductive materialism, this belief is just a neural structure. From a materialist’s point of view, this means that the neural structure that causes Draper to jump away is exactly the belief ‘there is an alligator in my tub’. The neural structure that is the belief ‘I am sitting under a tree eating mangoes’ should not cause one to jump away. It may cause one to put something in one’s hands into one’s mouth. It seems that in this response Plantinga once again imagines that an immaterial mind can grasp content and freely assign content to a sentence or neural structure. Then, we can indeed assign a very different content to the neural structure that causes Draper to jump away at that moment. For instance, we can assign the content ‘I am sitting under a tree eating mangoes’ to it. Otherwise, if you seriously think that this latter content property is an NP property, you will naturally think that this NP property may cause one to put something in one’s hands into one’s mouth, but it will not cause one to jump away.

Understandably, Plantinga may find it very difficult to imagine how a neural structure can somehow have the specific content ‘there is an alligator in my tub’. However, since he is evaluating the conditional probability P(R|RM&N&E), he has to put himself in materialists’ shoes. Any viable materialistic theory of content has to be able to predict that, in the example here, the neural structure that causes Draper to jump away very likely has the content ‘there is an alligator in my tub’. Otherwise, that theory of content cannot be a viable one, because it fails to predict intuitively obvious phenomena. Again, Plantinga can try to argue that no materialistic theory of content can succeed in doing this, but that will be an independent argument against materialism. My concern here is merely whether his evolutionary argument contains any new challenge against materialism, beyond the well-known challenge that materialism has to offer a coherent account of belief, content, and truth. If you assume an anti-materialistic conception of content and truth, then it is perhaps trivial to conclude that any kind of completely materialistic evolution cannot select true beliefs reliably.

 

 

Plantinga’s argument for (3)

Now, Plantinga does consider several naturalistic theories of content and truth in his latest article on the subject.[17] I will focus on his criticisms on teleosemantics. Teleosemantics (or a version of it) suggests that a belief is associated with some biologically normal responses to tokening the belief and the belief represents the environmental conditions that make those responses realizing a proper biological function of the belief.[18] Plantinga gives two major objections against teleosemantics. First, many complex beliefs, including the belief of naturalism, do not seem to have any biologically normal responses associated, and it is hard to imagine what the biological functions of those beliefs could be. Second, for a belief to have content, teleosemantics requires that it carry information about the environment, but then universally true and universally false beliefs, including the belief of naturalism again, do not have content, because they do not carry any specific information about the environment.

The first objection is based on the assumption that the teleological account of content should apply to all semantic representations, including complex beliefs about things and states of affairs without any direct connection with evolution on the earth. This makes teleosemantics sound absurd, since the idea of teleosemantics is to refer to evolution on the earth to characterize content. I suppose that no supporters of teleosemantics intend to hold that view (although some of their writings may give that impression). For instance, no one would try to characterize the content of the belief ‘there was a Big Bang’ or ‘there is an odd number of stars in the galaxy’ by referring to the conditions of the environment that make the belief realizing its biological function. The conditions represented by such beliefs cannot have any influence on what is happening on the earth. However, the teleosemantic approaches to naturalizing content do not have to assume that the teleological account of content must apply directly to all semantic representations. If we artificially construct a Boolean combination of a hundred atomic sentences, then it seems obvious that the content of that composite sentence as a belief should be characterized compositionally, based on the content of its atomic components and based on its logical compositional structure. Similarly, the truth-value of such an artificially constructed belief should be determined recursively in the Tarskian style. Another way of determining content and truth-value is quite inconceivable for such an artificially constructed logically composite belief. Then, when we consider a more natural logically composite belief, obviously, we should also determine its content and truth-value compositionally and recursively, as long as that is doable.

On the other side, for very simple representations, a teleological account of content is intuitively very appealing. The basic idea of a teleological account of content, as I understand it, is as follows.[19] A neural structure is constantly associated with some environmental feature as a matter of natural regularity, but it allows exceptions. These exceptions become semantic misrepresentations,[20] and therefore the semantic norm emerges, when we consider the proper biological function of that neural structure. This then reduces the semantic norm to the biological norm and transforms an indicator in Plantinga’s sense into a semantic representation with the possibility of misrepresentation[21]. For instance, imagine that a neural structure is constantly associated with a spatial location relative to one’s body and reachable by the hands, in the sense that the neural structure is constantly generated by seeing something at that location (relative to the body) and it participates in controlling the hands to stretch out and grasp something at that location. There can be exceptions to these constant associations because of a biological malfunction in the vision, hands, or brain, or because of a biological abnormality in the environment. For instance, sometimes there is an illusion caused by some abnormal environmental condition, or sometimes the brain is influenced by alcohol and it cannot control the hands properly. Therefore, the spatial location (relative to the body) that the neural structure really represents (i.e., semantically represents vs. causally indicates) should be determined by the situations in which everything is biologically normal. That is, the neural structure semantically represents the location accessed by the hands, controlled by that neural structure activated by seeing something at the location, when the brain, body, and environment are all biologically normal. Similarly, there may be constant associations between looking at a color and creating a neural state with possible exceptions, and what color that neural state semantically represents should be determined by biologically normal situations. For these extremely simple representations, it is intuitively reasonable to assume that the semantic norm coincides with the biological norm and that semantic misrepresentations can occur only in biologically abnormal situations. If you look into a brain and try to find out which color a neural state semantically represents (with the possibility of misrepresentation), it seems natural to look for the color that regularly activates that neural state in biologically normal situations and take aberrations that occur in abnormal situations as misrepresentations. Otherwise, you have to assume that there is an immaterial mind behind the brain that can somehow ‘intend’ that neural state to represent some particular color, regardless of what actually happen in biologically normal and abnormal situations.

It only becomes funny when one applies a similar characterization of content to a complex semantic representation, for instance, the belief ‘there is an odd number of stars in the galaxy’. No environmental condition regularly activates that belief in biologically normal situations, and the belief does not regularly participate in controlling any specific motor actions in biologically normal situations. We certainly make semantic mistakes in biologically normal situations for such beliefs. That is, it is obvious that the semantic norm does not coincide with the biological norm for such complex representations.

Therefore, very naturally, the teleological account of content should apply only to some simple and primitive semantic representations. Remember that Tarski’s theory of truth lacks an account of how primitive names and predicates get referents.[22] A natural idea is that a teleological theory can supplement the Tarskian theory and determine the broad content (i.e., referents) of those primitive names and predicates. That will be a structural-teleological theory of content and truth. The real situation is actually much more complex, because names and predicates as concepts are still composite representations, as I will briefly explain in the next section. However, this idea can already show that this objection to teleosemantics by Plantinga is not fatal.

This applies to Plantinga’s second objection as well. The teleological account should be applied only to representations that do indicate specific information about human environments, that is, representations that can make a difference in human evolutionary history. As for universally true beliefs, they are complex and composite representations, and some Tarskian recursive semantic rules, together with the referents and truth-values of the constituents of those beliefs, can determine that those beliefs are universally true.

 

Outlining a structural-teleological account of content and truth

I have argued that Plantinga’s criticisms on teleosemantics are not fatal and that a structural-teleological theory of content and truth can meet his criticisms, but so far it is still unclear what a structural-teleological theory of content and truth will look like. I proposed such a theory in two previous articles[23]. In this section, I will briefly introduce the ideas in those papers.

This structural-teleological theory assumes that a simple belief is a composition of concepts and a concept is a composition of an even more primitive kind of inner representations called ‘inner maps’. As a materialistic theory, inner maps, concepts, and beliefs are supposedly neural structures in brains. The theory so far addresses only concepts and beliefs representing ordinary physical entities and their states in human environments,[24] and it treats only broad content. It means that the critical task is to characterize, in naturalistic terms, the semantic representation relation between inner representations and external physical things, their properties, or their states of affairs. For a concept, this means a naturalistic characterization of what entities (or which property) the concept semantically represents. For a belief, this means a naturalistic characterization of the environmental states in which the belief is true.

An inner map is essentially a perceptual mental image in one’s memory. It typically represents a concrete physical object that one experienced before. Note that an inner map typically represents a particular object (e.g., the rabbit that I saw yesterday), not a class of physical objects (e.g., the class of all rabbits).[25] An inner map is still a composite representation. The most primitive inner representations are the components that make up an inner map, called ‘map attributes’. A map attribute can represent the color, brightness, or other perceptual appearance property of a very small point on the surface of a physical object, and it can represent the spatial location of a particular object (or a small part of it) that one experienced before, relative to one’s actual body location when experiencing the object. The teleological account of content suggested in the last section applies to a map attribute and determines which property or spatial location the map attribute represents. Then, some structural semantic rule decides which object an inner map composed of map attributes represents. There are many details on the structure of an inner map and on how an inner map represents a particular physical object. I have to skip them here.[26]

Concepts belong to a higher level in the hierarchy of inner representations. A basic concept is a composition of inner maps, and basic concepts are in turn components of composite concepts. Essentialist concepts are a common type of concepts; they represent natural classes. A typical essentialist concept may contain some inner maps as exemplars, which represent object instances to which one applied the concept before (in learning or using the concept). The semantic rule for an essentialist concept says that an entity is represented by the concept if it shares the same internal structure as those object instances represented by the exemplars of the concept. For instance, if my concept RABBIT contains some exemplars representing some rabbit instances to which I applied the concept before (in learning or using the concept), then an entity is represented by my concept RABBIT if it has the same internal structure as those instances. There are again many details regarding the structure and content of a concept. In particular, this theory tries to integrate the summary feature list theory, exemplar theory, and theory theory of concepts invented by cognitive psychologists.[27] I have to skip the details here.

Simple thoughts are composed of concepts, and logically composite thoughts are composed of simple thoughts and logical concepts. Thoughts are truth bearers. A belief is a thought that a brain puts in its ‘belief box’. There are many composition patterns for composing simple thoughts from concepts, and one of the simplest patterns is a composition like <THAT is a RABBIT> corresponding to the sentence ‘that is a rabbit’. Here, THAT is a singular concept representing an object located at some location relative to one’s body, for instance, the focus point of one’s eyes. The truth-value of a simple thought is determined by what its constituent concepts represent and by some semantic rule depending on the composition pattern for composing the thought (out of its constituent concepts). For instance, the thought <THAT is a RABBIT > is true just in case the object represented by the singular concept THAT is among the objects represented by the concept RABBIT. The truth-value of a logically composite thought is determined recursively by Tarskian truth definition rules.

 

How truth is positively related to adaptiveness

Now, consider how this theory predicts the relation between adaptiveness and truth and the reliability of human cognitive mechanisms. I will consider simple beliefs like <THAT is a RABBIT>. Plantinga does not discriminate between simple and complex beliefs in his argument for (2). If we can show that, for a simple belief, the behaviors it normally causes are more likely to be adaptive when it is true than when it is false, according to our structural-teleological characterization of truth, then this should be sufficient to reject Plantinga’s claim that adaptiveness is indifferent to truth, which is his main reason for supporting (2). This will also show that evolution favors more reliable human cognitive mechanisms for producing true simple beliefs, which sheds some doubts on the conclusion (2) itself (although this has not shown that evolution favors more reliable cognitive mechanisms for producing true complex beliefs as well).

First, I will assume that evolution selects several things simultaneously. I will use the concept RABBIT as an example. Evolution determines a class of entities, namely, rabbits, as a natural class relative to human cognitive mechanisms. It simultaneously selects human cognitive mechanisms that can recognize rabbits consistently. It also determines a class of environments in which human cognitive mechanisms can recognize rabbits stably. I will call these ‘optimal environments (for recognizing rabbits)’. In other words, human cognitive mechanisms are adapted to human environments, so that physical objects in the optimal environments (for recognizing a particular natural class) with similar appearances, relative to human pattern recognition mechanisms, mostly have the same internal structure as well and therefore constitute a natural class. This is the basis for developing essentialist concepts in human brains. The evolutionary value of such cognitive mechanisms adapted to natural classes is obvious. It allows human ancestors to reliably recognize and classify foods and predators in their (optimal) environments. Note that I did not take the reliability of human cognitive mechanisms for producing true beliefs as an assumption here. The claim that human cognitive mechanisms can reliably recognize rabbits in the optimal environments is merely a claim about some natural regularity in the interactions between neural networks and their environments. The notion of semantic representation and truth and the issue of reliability of human cognitive mechanisms for producing truths have not come up yet. You can take the neural structures involved here to be indicators in Plantinga’s sense.

Evolution also selects a human cognitive mechanism that recruits its perceptual memories (i.e., inner maps) of rabbits encountered in the optimal environments as the constituents (i.e., exemplars) of one’s essentialist concept RABBIT. Recall that, according to the semantic rule for essentialist concepts suggested by this structural-teleological theory of content, one’s concept RABBIT then semantically represents anything with the same internal structure as the object instances represented by those perceptual memories (as the exemplars of one’s concept RABBIT). Here I assume that only clear perceptual memories of rabbits obtained in the optimal environments for seeing rabbits are included in one’s concept RABBIT as the exemplars. In particular, unclear memories that one obtained in non-optimal environments, such as the memory of a rabbit in bushes, will not be included in one’s concept. This strategy has its adaptive value as well. An optimal environment for seeing something is an environment in which one can identify that thing more stably. This strategy allows a human concept to represent things in their corresponding optimal environments more stably. Besides, these memories of rabbit instances in the optimal environments (as the constituents of one’s concept RABBIT) are associated with the memories of past adaptive behaviors of dealing with rabbits, for instance, chasing, catching, roasting, and eating rabbits in some specific manner. This means that the belief <THAT is a RABBIT> tends to cause the behaviors of hunting, roasting, and eating the object represented by the singular concept THAT in some specific manner. These are the biologically normal behaviors associated with the belief. This connection between the belief <THAT is a RABBIT> and the behaviors is selected for its adaptiveness in dealing with rabbits in the optimal environments in the past. The adaptiveness of this belief-behavior association also owes to the fact that, in the optimal environments, when one identifies something as an instance of RABBIT and produces the belief <THAT is a RABBIT>, there is a high probability that the object is a rabbit and can therefore be hunted, roasted and eaten in that specific manner.

In other words, evolution simultaneously determines (1) rabbits as a natural class relative to human cognitive mechanisms, (2) a class of environments as the optimal environments for recognizing rabbits, (3) human cognitive mechanisms that can recognize rabbits in the optimal environments stably, (4) human cognitive mechanisms that keep the perceptual memories of rabbits obtained in the optimal environments as the exemplars of their essentialist concept RABBIT, and (5) human cognitive mechanisms that associate the belief <THAT is a RABBIT> with the adaptive behaviors dealing with rabbits in the optimal environments.

Then, when one sees a new rabbit instance in an optimal environment, one’s cognitive mechanism performs visual pattern recognition. It compares the new visual image received with the visual images in memory and decides that the new visual image is similar to some of the visual images that are the constituents of one’s concept RABBIT. This causes one to accept <THAT is a RABBIT> as one’s belief. The reliability of human cognitive mechanisms here is guaranteed by the fact that objects in the optimal environments with a similar appearance (relative to human pattern recognition mechanisms) mostly have the same internal structure as well. That is, if one recognizes that a new object has a similar appearance with those rabbit instances that one encountered before and kept in memory (as the exemplars of one’s concept RABBIT), then it is highly probable that that new object is a rabbit.

In non-optimal environments, the cognitive processes leading to a belief like <THAT is a RABBIT> are more complex. They require a more sophisticated human cognitive mechanism that can perform cognitive tasks such as deduction, induction, abduction, and various kinds of heuristic reasoning and memory association. However, evolution favors a more reliable cognitive mechanism in non-optimal environments as well. To see this, imagine that a hunter sees some object X moving in bushes. This is a non-optimal environment for identifying rabbits. Suppose that, for whatever reason, the hunter’s cognitive mechanism causes the hunter’s brain to put the thought <THAT is a RABBIT> into its ‘belief box’ as a belief (after some complex cognitive process). As I have explained above, this belief tends to cause the immediate behavior of chasing the moving object, a connection established in the optimal environments for seeing rabbits because of its adaptiveness. Apparently, the behavior is more adaptive (in this non-optimal environment) when the belief is true, that is, when X is a rabbit, than when it is false, for instance, when X is a tiger or a pheasant (which requires a different hunting method). This should not be very surprising. Recall that one’s belief <THAT is a RABBIT> produced in the optimal environments has a high probability to be true. That is, a rabbit in the optimal environments has a high probability to cause one to produce the belief. Then, when moving to a new and non-optimal environment, the truth of the belief <THAT is a RABBIT> means exactly that the new environment is similar to those optimal environments in a relevant aspect. That is, the relevant object belongs to the same kind as those objects that are very likely to cause the same belief in the optimal environments. This then means that the behaviors that are associated with the belief and are adaptive in the optimal environments will also be adaptive in the new environment. That is how the truth of the belief <THAT is a RABBIT> in a non-optimal environment confers a high probability on the adaptiveness of the behaviors caused by the belief. That is how evolution favors more reliable cognitive mechanisms for producing true simple beliefs.

However, note that truth is not identical with adaptiveness. In the example above, if the rabbit is contaminated with some dangerous virus, then the behavior caused by the belief may turn out maladaptive, although the belief is still true. Such cases should not happen too often. Otherwise, the behaviors of hunting, roasting, and eating rabbits will not be adaptive and evolution will not select them as the biologically normal behaviors associated with the belief <THAT is a RABBIT> in the first place. However, as long as such cases do happen occasionally, it is sufficient to show that truth is not identical with adaptiveness.

 

Responses to some potential objections

Finally, consider some potential objections. First, Plantinga may object that inner maps, concepts, and beliefs described here are merely indicator representations, not semantic representations.[28] Here we must note that, under materialism, whether something is a semantic representation is determined by its naturalistic characteristics, not by any subjective intention that a non-physical mind puts on it. A neural structure is a semantic representation if it has some special naturalistic characteristics or if it performs some special cognitive functions. First, it must be a representation. That is, there must be some regular correlation between it and other things, properties, or states. Second, a major characteristic that distinguishes a semantic representation from an indicator representation is that the former allows misrepresentations. There may be other requirements for a semantic representation. I will not try to explore here. In naturalizing a semantic representation relation, we define a relation between neural structures and other things or states using naturalistic terms alone, that is, without using any intentional or semantic terms such as ‘mean’, ‘represent’, ‘belief’, and so on. We describe the environments in which that relation exists and the environments in which that relation does not exist. The latter are the environments in which there is a misrepresentation. We then compare that naturalistically defined relation with our intuitive understanding of the semantic representation relation. If they fit each other, that is, if they exist in the same environments, we claim that the relation defined using naturalistic terms is the semantic representation relation. Now, our definition of truth for beliefs like <THAT is a RABBIT> does allow misrepresentations and does seem to fit our intuitive understanding of truth. Therefore, under materialism, we claim that it is truth and that the beliefs described here are semantic representations, not merely indicators.

Second, consider the problem that a belief-desire pair with a false belief and a harmful desire can cause the same behavior as a pair with a true belief and a healthy desire does.[29] Another way to put the problem is the following. Suppose that H is an adaptive behavior in some environment. Then, for any desire D, the belief-desire pair <if H then D, D> will cause the same behavior H, no matter if D is healthy or harmful, and no matter if the belief ‘if H then D’ is true or false. Suppose that the way evolution works is to select such belief-desire pairs from a pool of many available pairs based on their adaptiveness. Then, there is indeed no guarantee that evolution will tend to select true beliefs. However, apparently evolution does not work that way. In fact, evolution does not select individual beliefs (or desires or belief-desire pairs) directly. Evolution selects genes that determine human cognitive mechanisms with various traits. Simple beliefs are produced by human cognitive mechanisms reacting to environmental stimuli. In my version of teleosemantics, I suggest that human cognitive mechanisms allow humans to identify natural classes consistently in optimal environments and develop concepts representing natural classes. Human cognitive mechanisms and the optimal environments also determine what a concept represents and when a simple thought is true. Moreover, a simple belief produced by a human cognitive mechanism in optimal environments is normally true, and in non-optimal environments, evolution favors a more reliable cognitive mechanism. I cannot guarantee that this must be how things go in evolution, but it seems obvious that it never happens that some random factors produce many belief-desire pairs and then evolution selects adaptive belief-desire pairs. This is a wrong model of how human cognitive mechanisms work.

Since the advent of human language, humans can indeed produce complex beliefs such as ‘if H then D’ using language and they can consciously consider many options for their beliefs and desires. However, it is still not true that evolution (or cognitive mechanisms produced by evolution) selects complex belief-desire pairs directly based on their adaptiveness. Complex beliefs are produced either by the same mechanism that produces simple true beliefs reliably, or by some extension of it. I cannot discuss here how evolution (including cultural evolution) still favors reliable extensions. However, it seems at least clear that this example of belief-desire pairs with false beliefs does not affect the conclusion that evolution favors human cognitive mechanisms reliable for producing true simple beliefs, because this example is based on a wrong model of how evolution works and how human cognitive mechanisms produce beliefs.

One may wonder if we can reproduce this problem in the model described in the last section. A potential example might be the following. Upon seeing some rabbit moving in bushes, a hunter’s cognitive mechanism produces the false belief ‘that is a tiger’, but the hunter somehow has the dangerous desire of catching and eating tigers, and therefore the hunter chases the moving object. This behavior turns out to be adaptive, because the hunter in the end catches and eats the rabbit, although his belief is false. However, this example does not work. As in the case of the concept RABBIT, the biologically normal behaviors associated with one’s belief <THAT is a TIGER> are the adaptive behaviors when the belief is produced in the optimal environments for seeing tigers. In such environments, the adaptive behaviors can only be running way, not chasing the object, because when one produces the belief <THAT is a TIGER> in the optimal environments for seeing tigers, there is a high probability that the object is a tiger. It means that if a hunter’s behavior caused by a belief produced by seeing something moving in bushes is chasing the object, then the belief is much more likely to be <THAT is a RABBIT>, not <THAT is a TIGER>.

Note that under physicalism we do not ask which belief the hunter subjectively intends for a neural structure in his brain to mean. Which belief with a specific content the hunter actually has in his brain at a particular moment depends on the natural facts about the neural structure (as a belief) that causes his behaviors at that moment. If the behavior caused is chasing the object in a specific manner, like hunting a rabbit, we have reason to conclude that it has a high probability that the neural structure is a belief with the specific content ‘that is a rabbit’, given how the belief-behavior associations are developed. That is, the neural structure is more likely to be the belief (i.e., the same neural structure) that is normally produced upon seeing a rabbit in the optimal environments for seeing rabbits, not the belief that is normally produced upon seeing a tiger in the optimal environments for seeing tigers. A case like this example may happen, but it will have a low probability. Otherwise, it would mean that chasing a tiger is a biologically normal behavior upon seeing a tiger in the optimal environments for seeing tigers. That cannot be true. Therefore, such low probability cases do not affect our conclusion that truth is positively related to adaptiveness.

Third, consider Plantinga’s example of a God-believer tribe whose people always express their beliefs in the form ‘this apple-creature (meaning ‘this apple created by God’) is good to eat’.[30] All such beliefs are false according to atheists, but they are adaptive. This might support the idea that adaptiveness is indifferent to truth. The problem with this example is that those people’s cognitive mechanisms actually appear to be quite reliable according our intuitive understanding of the reliability of a cognitive mechanism. To see this let us dramatize the example. Suppose that people in that tribe have all the scientific knowledge we have today, from fundamental physics to biology and cognitive sciences. The only difference between them and us is that they always use predicates such as ‘lepton-creature’, ‘black-hole-creature’, and ‘DNA-creature’, and they always express their beliefs using these predicates. Then, an atheist will naturally think that people in that tribe have only one clearly false belief, that is, the belief that everything is created by God, and except for that single belief, all their other beliefs are actually the same as our scientific beliefs. Their cognitive mechanisms are thus as reliable as ours are, since that single false belief is quite insignificant. Counting the number of true beliefs can be tricky. Suppose that I state my beliefs as {A&B, A&C, A&D}, and suppose that A is false and B, C, and D are true. Do I have three false beliefs? or do I have one false belief and three true beliefs?  It may be difficult to invent a theory on how we should count the number of true beliefs for evaluating the reliability of a cognitive mechanism, but our intuition seems to be that in this example those people’s cognitive mechanisms are in fact as reliable as ours are.

The structural-teleological theory of content and truth introduced above actually supports this intuition. Although those people explicitly use predicates like ‘apple-creature’ exclusively, we have reasons to believe that inside their brains they have simple concepts such as APPLE as well. Such simple concepts were developed in an early stage of their evolution, and their cognitive mechanisms developed in the early stage may be the same as ours, and they may be as reliable as ours are for producing true simple beliefs. Their predicate ‘apple-creature’ expresses a composite concept, although they invent a simple name in their language for it. Their notions of God, Creation, and apple-creature are developed much later (and are the products of their cultural evolution, not biological evolution). Therefore, this kind of examples does not affect the conclusion that evolution favors a human cognitive mechanism reliable for producing true simple beliefs.[31]

 

References

Adams, F. (2003). Thoughts and their contents: naturalized semantics. In S. Stich & F. Wafield (eds.), The Blackwell guide to the philosophy of mind. Oxford: Basil Blackwell.

Beilby, J. (ed.) (200). Naturalism Defeated? Essays on Plantinga’s Evolutionary Argument against Naturalism. Cornell University Press.

Draper, P. (2007). ‘In Defense of Sensible Naturalism’, http://www.infidels.org/library/modern/paul_draper/naturalism.html

Dretske, F. (1986). Misrepresentation. In R. Bogdan (ed.). Belief: Form, Content and Function. Oxford: Oxford University Press.

Fales, E. (2002). Darwin’s Doubt, Calvin’s Cavalry. In Beilby (2002), p. 43-58.

Field, H. (1972). Tarski’s Theory of Truth. Journal of Philosophy, 69, 347-375,

Fitelson, B. and E. Sober (1998). Plantinga’s Probability Argument against Evolutionary Naturalism. Pacific Philosophical Quarterly, 79, 115-129.

Laurence S. and E. Margolis (1999). Concepts and cognitive science. In E. Margolis & S. Laurence (eds.). Concepts: core readings. Cambridge, MA.: MIT Press.

Macdonald, G. and D. Papineau (2006): Teleosemantics: New Philosophical Essays, Oxford University Press.

Millikan, R. (2004). Varieties of Meaning. Cambridge, MA.: MIT Press.

Murphy, G. (2002). The Big Book of Concepts. Cambridge, MA.: MIT Press.

Neander, K. (2004). Teleological Theories of Content. In Stanford Encyclopedia of Philosophy. E. N. Zalta (ed.). http://plato.stanford.edu/entries/content-teleological/

Papineau, D. (1993). Philosophical Naturalism. Oxford: Basil Blackwell.

Plantinga, A. (1993). Warrant and Proper Function. New York: Oxford University Press.

Plantinga, A. (2002a). Introduction: The Evolutionary Argument against Naturalism. In Beilby (2002), p. 1-12.

Plantinga, A. (2002b). Reply to Beilby’s Cohorts. In Beilby (2002), p. 204-275.

Plantinga, A. (2003). Probability and Defeaters. Pacific Philosophical Quarterly, 84, 291-298.

Plantinga, A. (2007a). Naturalism vs. Evolution: A Religion/Science Conflict? http://www.infidels.org/library/modern/alvin_plantinga/conflict.html

Plantinga, A. (2007b). Against “sensible” Naturalism. http://www.infidels.org/library/modern/alvin_plantinga/against-naturalism.html

Plantinga, A. (2007c). Region and Science. In Stanford Encyclopedia of Philosophy, http://plato.stanford.edu/entries/religion-science/

Plantinga, A. (forthcoming). Content and Natural Selection. Forthcoming in Philosophy and Phenomenological Research.

Ramsey, W. (2002). Naturalism Defended. In Beilby (2002), p. 15-29.

Xu, Y. (forthcoming). The troublesome explanandum in Plantinga’s argument against naturalism. Forthcoming in International Journal for Philosophy of Religion.

Ye, F. (online-a). A Structural Theory of Content Naturalization. Available online at http://sites.google.com/site/fengye63/

Ye, F. (online-b). Truth and Serving the Biological Purpose. Ibid.



*  Feng Ye

Department of Philosophy, Peking University, Beijing 100871, P. R. China

Email: fengye63@gmail.com

[1] Plantinga (1993, 2002a, 2002b, 2007a, 2007b, 2007c, forthcoming).

[2] Plantinga did not say this explicitly.

[3] Xu (forthcoming).

[4] I cannot go into any detail here, but such a specific reliability claim should be able to meet Xu’s criteria for a qualified explanandum if any reliability claim can.

[5] See Fitelson & Sober (1998), Beilby (2002), Draper (2007), and Plantinga (2003, 2007b).

[6] Plantinga (forthcoming).

[7] This may leave room for revising Plantinga’s argument to show that, given materialism, the conditional probability for the reliability of human cognitive mechanisms for producing complex scientific beliefs is low. I cannot discuss this issue here, but I like to mention that the evolution of our scientific methodologies is cultural evolution, not genetic evolution, and it is not driven by adaptiveness alone.

[8] See, for instance, Ramsey (2002) and Draper (2007).

[9] See Plantinga (2002b, 2007b, forthcoming). Plantinga did not say it explicitly, but we can naturally infer that Plantinga’s believes that a divine guidance is necessary for connecting truth with adaptiveness.

[10] See the last three sections of this article for the details. Note that truth is not identical with adaptiveness for teleosemantics. They are merely positively related.

[11] Plantinga (forthcoming). Similar claims are in other presentations of his argument.

[12] Such a conception of content must assume that there are immaterial minds that can grasp content and freely assign content to sentences or neural structures.

[13] Plantinga (forthcoming).

[14] See Draper (2007). False (2002) and Ramsey (2002) raise similar objections against Plantinga.

[15] Plantinga (2007b).

[16] Plantinga (2007b).

[17] Plantinga (forthcoming).

[18] See Neander (2004) for a survey and see Macdonald and Papineau (2006) for some recent debates.

[19] Cf. Papineau (1993), Millikan (2004), Neander (2004).

[20] One of the major tasks for naturalizing content is to explain, in naturalistic terms, how semantic misrepresentations are possible. See Adams (2003), Dretske (1988).

[21] See Plantinga (2002b, 2007b, forthcoming) for his characterization of indicators, and see the last section of this paper for a response to the objection that such a neural structure is still an indicator, not a semantic representation.

[22] Field (1972).

[23] Ye (online-a, online-b).

[24] It can potentially extend to more complex scientific beliefs. As for mathematical concepts and thoughts, I personally hold nominalism (for alleged mathematical entities) and instrumentalism (for mathematical theorems), and therefore this account does not apply to mathematical concepts and thoughts directly.

[25] An inner map can determine a spatial-temporal location relative to one’s body and the moment of retrieving the inner map from one’s memory, and that is why it can represent a particular object instance that one experienced before, not necessarily a class of similar-looking objects.

[26] See Ye (online-a) for more details.

[27] Murphy (2002), Laurence and Margolis (1999).

[28] See Plantinga’s response to Draper quoted above (Plantinga 2007b).

[29] Plantinga gives an example in which someone wants to be eaten by tigers but falsely believes that running away will increase the chance to be eaten (Plantinga 2002a).

[30] See Plantinga (2002b). Ramsey (2002) criticizes some similar examples by Plantinga (2002a). This example is from Plantinga’s reply to Ramsey and it is supposed to reject Ramsey’s criticisms. My response to this example below follows a similar strategy as Ramsey’s earlier criticism, but we will see that this structural-teleological theory of content makes the problem in Plantinga’s example clearer. I would like to thank a referee for a comment that leads to this clarification.

[31] An early version of this article was presented at 2009 Beijing Conference on Science, Philosophy, and Belief. I would like to thank the participants for their comments. I would also like to thank Professor Alvin Plantinga for several discussions on the topic during and after the conference.

Comments