Do Animals have Belief

July 10, 2025 | 1,541 words | 8min read

Paper Title: Do Animals have Beliefs?
Link to Paper: https://philpapers.org/rec/STIDAH
Date: 1979
Paper Type: Philosophy, Action Theory
Short Abstract:
Stephen Stich presents arguments on both sides of the debate about whether animals have beliefs, but ultimately concludes that the question is moot, since the concept of belief cannot be meaningfully applied to animals at all.

1. Introduction

Do animals have beliefs? Many philosophers have considered this question, and most think the answer is rather obvious. The problem is that some think the obvious answer is yes, while others say no.

The author argues that both sides are mistaken, because the question itself is problematic: the concept of belief does not neatly map onto animals. The issue isn’t that we lack sufficient information about animal psychology, but rather that the problem is conceptual in nature.

2. The Argument in Favor of Animal Belief

The basic argument goes something like this: humans act, and most agree that an action is explained by the beliefs and desires behind it. Beliefs tell us how the world is, and desires guide how we want to change the world through our actions. Together, these states provide an adequate explanation for behavior.

If this is true, there seems to be no good reason not to apply the same model to animals, especially since humans and animals are not so different from each other. The belief-desire model of action doesn’t just work for humans but, arguably, also for animals. Importantly, the reasons for applying this model to animals are the same reasons we apply it to humans.

Additionally, if we were to propose an entirely different theory of action for animals, we’d also have to explain why that theory wouldn’t apply to humans as well.

3. The Argument Against Animal Belief

(This argument is presented in more detail in the paper or in the article Rational Animal.)

The basic argument against attributing belief to animals is this: in order to have a belief, an animal would need to hold certain propositions that support that belief. That is, a belief never exists in isolation, it always comes within a network of associated beliefs. But it seems unlikely that animals possess this required web of beliefs.

For example, a dog may be able to recognize a bone, but that doesn’t mean it knows what a bone is in the conceptual sense. It wouldn’t be able to recognize an atypical bone, such as one from a whale, nor would it understand the difference between a real bone and a fake plastic one made from soy protein but tasting the same. So, although a dog may recognize a bone, it lacks the broader beliefs we associate with knowing what a bone is.

This kind of argument can be repeated with other objects, like meat. Thus, the conclusion is that dogs—and by extension animals in general, do not have beliefs about the world.

In other words, we cannot explain a dog’s behavior in terms of belief, because we cannot meaningfully identify what those beliefs would even be.

4. Armstrong’s Way Out

We are now left with the following dilemma:

• The belief-desire theory of action explains both human and animal behavior.
• But on the other hand, we cannot say anything specific about what these supposed beliefs in animals actually are.

Armstrong proposes a way out by suggesting a different kind of reading of this problem: he invokes Quine’s concept of referentially transparent propositions for belief.

What is a referentially transparent reading?

This type of reading allows the substitution of co-referential expressions in a sentence without altering its truth value.

For example, consider the following two statements:

1. Oedipus married his mother.
2. Oedipus married Jocasta.

“Mother” and “Jocasta” are co-referential, they refer to the same person. Although from an external perspective both statements are true, from Oedipus’s point of view only the second is true: he believed he was marrying Jocasta, not his mother.

If we adopt a referentially transparent reading of belief, then we are allowed to substitute “mother” with “Jocasta.” But under the more standard, referentially opaque reading (which is typically used for beliefs), such substitution is not allowed because it changes the meaning from the subject’s point of view.

Armstrong applies this same logic to animals. He argues that even if we don’t know the exact content of a dog’s belief, we can still talk about that belief by substituting it with our own co-referential expression. In other words, we might not know what exactly a dog believes about a bone, but we know what we mean by “bone,” so we can substitute the content of the dog’s belief with our own conceptual framework.

The Problem with This Strategy

However, there’s an issue: there is no idiom in English or German that permits this kind of substitution within belief contexts.

Consider the following example in English:

1. Jones believes that the principal will inspect his teaching methods during class.

From this, we cannot validly infer:

2. Jones believes that the principal experienced the birth of a giant panda in prison,

even if we were to propose a universal substitution rule like:

3. \(\forall x\), x will inspect Jones during class, where \(x =\) the one who experienced the birth of a giant panda in prison.

This shows that belief contexts resist substitution of co-referential expressions, and thus a referentially transparent reading of belief is linguistically and conceptually problematic.

Armstrong’s Response

One possible response Armstrong could give is: if such an idiom or linguistic structure doesn’t currently exist, we should simply create one. That is, we redefine belief for animals by conducting psychological investigations to determine which attributes or properties make up their “belief” about a bone.

For example, we could define belief in this way:

A dog believes that something is a bone if it has properties \(a_1, a_2, ..., a_n\).

We would then replace the term “belief” with this new, behaviorally grounded definition:

The dog believes (in the sense defined above) that something is a bone if it has \(a_1, a_2, ..., a_n\).

But Does This Solve the Problem?

If we step back and examine what we’ve done, it becomes clear that we haven’t simplified the issue, we’ve only complicated it.

Suppose that for a dog, something counts as a bone if it has a particular chemical composition. But the dog has no concept of chemistry, nor of chemical compounds. So, while we’ve replaced the term “belief” with a list of properties, we still haven’t identified what the dog actually believes.

In short, we began with the goal of substituting the original concept of belief with the content the dog understands. But in the end, we just substituted it with a list of attributes we understand, and still have no clear idea of what the dog believes under that concept.

5. Belief and Content

The previous section has shown that, using Armstrong’s approach, we still cannot ascribe content to the beliefs of a dog. This means we don’t know what we mean when we say that a dog has a belief.

Therefore, if we assume that a belief must have some identifiable content in order to count as a belief, we are forced to conclude that animals do not have beliefs. And we can only attribute content to beliefs if the beings in question are sufficiently similar to us. The reason we can’t attribute content to a dog’s belief is that dogs experience the world so differently from us that we cannot describe what their beliefs would even entail.

In other words, to attribute content to a belief-state, that state must be part of a system of belief-like states that overlaps meaningfully with our own.

The Problem of Similarity

But if we say that a belief is not a real belief unless it has content, we run into problems in other cases as well. For example, people with cognitive disabilities, aliens, ancient humans from thousands of years ago, people from vastly different cultures, or even future humans might be so different from us that their conception of belief would no longer resemble ours.

We use the word belief comfortably in the context of our everyday interactions with fellow human beings. But when we try to apply the concept of belief to more “exotic” subjects, the idea breaks down—our understanding of what it means to believe becomes too different. (This echoes Heidegger’s view that meaning arises only within specific modes of being.)

A Conceptual Trade-Off

Thus, we are faced with a conceptual trade-off:

• We can require content for belief, but then we may no longer be able to ascribe beliefs to beings too dissimilar from us.
• Or we can drop the requirement for content, but then the phrase “X believes that ___” becomes hollow—it loses any specific meaning or explanatory power.

Stephen Stich himself suggests that, if we want the concept of belief to be more useful for scientific psychology, then the first thing we should remove from it is the notion of content.

reply via email