Do Scientific Truths Depend on Our Language?

“The limits of my language mean the limits of my world.“

—Ludwig Wittgenstein, Tractatus Logico-Philosophicus, 5.6

Editor’s Note: This is a really fascinating philosophical perspective, and your arguments address many of the same issues which Kant attempted to answer in his critical project. I have a pretty extensive background reading and writing about his main works, so I wanted to leave counter-arguments throughout the article to try and articulate how the Kantian model would view your ideas. Hopefully, by putting Kant’s ideas in dialogue with yours, we can get closer to understanding these metaphysical debates. —Bobby Becker, founder of The College Contemporary

Science is objective, right? Or, at the very least, if anything is objective, surely science is the best such candidate for being so. Electrons have a negative charge, water is composed of two hydrogen atoms and one oxygen atom covalently bonded, and when I drop a stone gravity causes it to fall to the ground. These just seem to be objective facts! We like to think that while our emotions, or some of our value judgments, or perhaps even truths about religion may be subjective, science somehow allows us to get a grip on objective reality; facts that cannot be debated or argued against depending on our individual, subjective minds. However, in the following, I hope to challenge that conception. While reality, the universe, etc. is plausibly objective (that is, there is something out there universal to all of us, independent of our individual minds), scientific truths may nonetheless be dependent on the language we use to formulate them.

Editor’s Note: Kant also addresses that there is something out there independent of our individual minds. He called this ‘Noumenon’, or ‘things-in-themselves’, as opposed to ‘phenomena’ which are how objects act according to our senses. Since all of our concepts about the world only apply to how objects behave within our perception, these ‘thing-in-themselves’, to Kant, can’t ever be described or even fully conceptualized. Nevertheless, he does assert that their existence is real (through a really complicated proof), and their existence plays a huge part in his metaphysical system.

Before I enter into my main point, and hopefully a good argument to back it up, I want to review the Sapir-Whorf hypothesis. That hypothesis concerns linguistic relativity, and whether language restricts the use of concepts. The strongest reading of the Sapir-Whorf hypothesis suggests that if an agent (a person, a mind, etc.) lacks a word which is representative of a concept, then they lack the concept altogether. Weaker readings of the hypothesis suggest, instead, that an agent may have a more difficult time using certain concepts if they are not properly represented in that agent’s language and grammar.

Editor’s Note: This is somewhat close to Wittgenstein’s position, which was the line of reasoning he used to critique Kant. To Wittgenstein, Kant’s entire project—along with most other metaphysical philosophers—are entirely limited by the natural language which they used to express their ideas.

Kant, in contrast, believed that while his rules didn’t describe reality in itself, they described how objects in our experience universally behave in our experience; thus, our scientific truths are ‘objective’ in a certain sense and run deeper than our language. However, this was not a critique that he was alive to address, so we are left to theorize how the Kantian model would respond.

A commonly cited, though controversial and perhaps fallacious, example is that of the many words for “snow” in the Inuit language. The Sapir-Whorf hypothesis suggests that because the Inuit have many names for what is “normally” considered a single concept, that of snow, they are able to reason more easily about what many would consider to be subtle, or even insignificant, differences between certain kinds of snow.

Editor’s Note: Another great example of this: certain languages have more precise words to describe a larger number of shades along the color spectrum. Empirical evidence shows that people who speak these languages can identify differences between shades more quickly.

There is also evidence which suggests that people who know Chinese can do basic math more quickly, since their language has a more consistent base-10 counting system (no irregularities like ‘eleven’ or twelve’). Some people also speculate that the reason why so many influential philosophers came from Germany is because of the complex grammatical structures and vocabulary of the German language, which made it possible to think and express more complex ideas.

Author’s Note: Pullam (1989) suggests that there may actually only be two roots for snow in the Inuit vocabulary, citing the “Dictionary of the West Greenlandic Eskimo Language” by Schultz-Lorentzen (1927). He argues that the proliferation of the idea of more words for snow comes from Western prejudicial views of the Inuit people. His claims, however, have been recently contested by Robson (2012) and others who argue that the Inuit undoubtedly have a rich vocabulary for many concepts related to snow and ice.

Such a weak interpretation of Sapir-Whorf seems intuitive to get behind. Certainly we are all better at reasoning, at least formally, about some concepts once we have words attached to them. How might you distinguish gravity and propulsion if you failed to have access to those words? You’d have to use long chains of descriptives. For example, instead of “gravity” you would have to say “the force which causes objects to fall to the ground, the Earth to revolve around the sun, the sun to produce helium from hydrogen due to fusion, etc.”. And instead of “propulsion”, you’d say “the force which causes objects to move when combustion occurs, or when a fluid is manipulated, or when an elastic potential is released, etc.”

Of course, without these more fundamental concepts, the description would become even more complicated, i.e., if we further had to reduce concepts like helium and hydrogen to descriptions of protons and electrons, and eventually those to the fundamental fields. In this way, our use of language allows us to simplify long chains of more fundamental descriptive properties which reduce the cognitive load necessary for thinking about such concepts. Thus, we are able to reason more effectively about them.

Nonetheless, while the weak reading of Sapir-Whorf may hold, the strong reading is a bit dubious. It’s not as though societies which lacked a word for “addition” failed to understand that two pieces of fruit and two more pieces of that fruit made four pieces of fruit. We all understood the strange feeling of pleasure when witnessing the minor suffering of another before learning the German word “schadenfreude”. And we don’t think that children fail to see colors before learning the words “red”, “green”, “blue”, etc. Language may reduce our cognitive load, but it doesn’t give us entirely new concepts, at least not all on its own.

… or does it?

In what follows, I aim to suggest that language does, indeed, provide us with novel concepts, even scientific concepts, concepts which may not really exist as an objective part of reality. To build up to that point, however, I need to first introduce a philosophical problem raised by the Scottish Enlightenment philosopher David Hume: the problem of induction.

Editor’s Note: There are some similarities between this perspective and Kant’s system: He also believed that our scientific truths do not describe objective reality (in a certain sense of the word ‘objective’, at least). Rather, like many other idealists, he believed that scientific truths describe objects as they consistently behave within our experience of them, which makes them capable of having predictive power.

However, unlike this viewpoint that you’re positing, he believed that they derive from something deeper than language—even if they don’t derive from reality-in-itself. To Kant, objects in our experience act in accordance to a priori rules which our mind uses to structure how they behave.

Kant thought he arrived at the exact number of these rules which our mind uses (12 in total, which he called the ‘Categories of Understanding’, along with 12 additional corresponding rules which we use to exercise our reason, which he called the ‘Categories of Judgements’). While even the most generous Kant scholars are skeptical of his claim that he definitively found the exact number, the idea that there are a discrete number of ways that our mind uses to structure our experiences is still a respected and influential philosophical position today. Just like how we have a set amount of axioms and rules of inferences to make claims in formal logic, we may use a limited number of possible rules to make judgements about the world.

Let’s suppose you believe that the sun will rise tomorrow morning. Unless you’re a radical skeptic, you presumably do. I certainly do. But why do you believe that? The basic argument formulated by your mind probably goes something like this:

(P1) The sun has always risen before.

(P2) Events in the future which have been entirely dependent on the laws of physics in the past are predictable.

(P3) The rising of the sun has been entirely dependent on the laws of physics

Therefore:

(C) The sun will rise tomorrow.

Now I’m sure everyone accepts (P1); the sun has indeed always risen before, at least since the sun and the Earth have existed. And nearly everyone accepts (P3), unless they have some radical religious views (e.g., maybe Helios needs to ride his chariot across the sky to make the sun rise, rather than the laws of physics merely needing to stay their course). But why should we believe (P2)? This was the question raised by Hume.

David Hume argued that all kinds of knowledge fall into two categories: relations of ideas and matters of fact. Relations of ideas are truths which follow from the definitions of our concepts, themselves, while matters of fact are truths which we can only know are true by investigating the world. Consider the difference between the two truths “Barack Obama was elected president of the United States in 2008” and “Every bachelor is unmarried”. On the one hand, we cannot know that Barack Obama was president of the United States simply by analyzing the concepts “Barack Obama” and “president of the United States”, but we can know that every bachelor is unmarried just by using the definition of the relevant terms. Hence the former is a matter of fact and the latter a relation of ideas.

Recall, then, statements (P1), (P2), and (P3). Certainly both (P1) and (P3) are matters of fact. We have observed the sun rise every morning since the dawn of humankind, giving us reason to believe (P1). And though for a while it wasn’t known that the rising of the sun was entirely dependent on the laws of physics, through careful scientific study and theorizing we now know that every time it has risen in the past it was due to the rotation of the Earth and the sun’s constant location relative to it. That gives us reason to believe (P3). But why should we believe (P2)?

We then need to understand whether (P2) is a relation of ideas or a matter of fact. If it’s the former, then it somehow follows from the definition of “events of the future”, “predictability”, “events of the past”, and “the laws of physics”. But that doesn’t seem to be true, at all. Certainly there have been events in the past which have seemed to be predictable but which eventually failed, e.g., Newtonian mechanics being used to predict the location of electrons in the double-slit experiment. So that can’t just be what we mean by the relevant terms. Therefore, (P2) must be a matter of fact. That is, we may only have reason to believe it based on observation. But (P2) makes a claim about the future, which we cannot observe. Therefore, we have no good reason to believe (P2): that we may use past events to predict future events. (The type of reasoning afforded by (P2) is called inductive reasoning, and hence the problem raised by Hume is called the problem of induction.)

We then find ourselves at a crossroads. Do we side with Hume, agreeing that we have no good reason to believe in the predictability of the future based on the events of the past, and reject belief in any certain knowledge of the universal laws of nature? Or do we reject his reasoning based on some prior intuition about the uniformity of nature? Philosophers who belong to the former camp are called “Humean reductionists”, while those belonging to the latter camp are often called “scientific realists”. To get a better handle on what these groups of philosophers believe, let me briefly summarize their prevailing views, beginning with the latter.

Author’s Note: Kant (1781) offered a counterargument to Hume along these lines.

Editor’s Response: To be clear, I don’t think Kant fits neatly into the scientific realist camp. His critical project was, in fact, an attempt to resolve the arguments between these Humean reductionists and scientific realists, which he viewed as a component of the larger split between rationalists and empiricists.

Kant held strong admiration for the problems which Hume raised about ‘causality,’ and he believed that these problems could not be properly addressed by scientific realists and, in fact, it was this very problem he credited for sparking his formulation of transcendental idealism; in his system, ‘causality’ is not something which we can empirically observe. It is, in fact, one of the rules which our mind uses to make judgements about and perceive objects within our experiences—more of a ‘lens’ that structures our experience a priori rather than object which we perceive with our experience. Thus, while ‘causality’ itself cannot be observed, the concept is necessary for shaping our scientific observations.

Author’s Response: You’re right that Kant would not fit into the same camp as Maudlin and the “strict” scientific realists. Nonetheless, he is a realist about laws of nature (which to him might look more like the basic rules of thought) and argues for them based on a kind of prior intuition we (according to Kant, necessarily) have for them. The Humean response, as I understand it, would essentially be this: causality is not necessary for shaping our scientific observations. Correlation is, but causality isn’t. Nonetheless, causation is advantageous to shaping our scientific observations (see my later discussion).

Tim Maudlin is perhaps the pre-eminent scientific realist, and we can roughly characterize his view as the standard, common-sense, physical view of reality. To Maudlin and the realists, there is the Big Bang on one temporal side of the universe, the heat death on the other, and a bunch of junk in between, all riding along a four-dimensional block of spacetime. The “physical stuff”, e.g., electrons, photons, molecules, your body and mine, is all situated exactly where physics tells us it is on that spacetime block, and there’s really nothing more to the picture. Science, of course, is then just giving us a very good description of that block. So good, in fact, that it allows us to predict future events based on past and present events. The block has rules; it must have rules, due somehow to its “nature”, and they are those which are being discovered by physicists. There is, on this reading, no good reason to doubt (P2), so we are a priori justified in believing it.

On the other hand, we have the Humeans, David Lewis and his philosophical kin, who importantly would not consider themselves “scientific antirealists”. They, just like the realists, really do believe that the laws of physics, as universal laws of nature, exist. They merely hold a different view about what, precisely, those laws are. Instead of beginning their metaphysics with the four-dimensional block of spacetime on which physical bodies are all placed and operating, the Humeans begin with constructed world of experience, what they call the “Humean mosaic”, and deduce from that mosaic the best explanation for what they are experiencing.

Author’s Note: Importantly, this is distinct from the “idealism” vs. “materialism/physicalism” debate, as many Humeans would still consider themselves materialists/physicalists. The distinction between them and the scientific realists given above is much more subtle. They really do believe that, fundamentally, everything is physical in nature. But to them, what it means to be “fundamental” is nothing more than being quantified over in the best metaphysical theories, with the best metaphysical theories being merely descriptive of reality. Thus, what is “physical”, “mental”, etc., is theory-laden.

Editor’s Response:From the perspective of transcendental idealism, the very idea of ‘physical’ in nature is a concept that we can only assign to objects in our experience. The world-in-itself cannot be described or conceptualized, since all of our adjectives only describe objects as they appear in our experience.

This is why Kant isn’t a strict ‘idealist,’ since he believes that things do exist outside of our experience of them. But, he thinks that this world has no relation to what is typically conceptualized as ‘physical’ or ‘corporeal.’ Those concepts are just useful for formulating predictive physical models of the world.

That is, they only believe in electrons, photons, and molecules because they best explain the patterns of experience you and I are having. Instead of the “fundamental particles” being fundamental, they are derivative from experiences. Nonetheless, the Humeans believe that science is doing a very good job of developing theories which explain our experiences. As such, they believe in fundamental laws of nature. The laws of nature, on this reading, are just the universal truths of the best theories which describe all our experiences. So they are, indeed, realists about the laws, but the laws play no fundamental role in their metaphysics.

Editor’s Note: To Kant, the correct ‘laws of nature’ have objective validity; they come from when we make conclusions about the world that are consistent with our ‘reason,’ which is a separate faculty that we have that unifies all of our judgments in an overarching model of reality. With our reason, when can identify when a new judgement contradicts another valid judgement in our model of reality, which allows us to then determine that judgement as false.

For example, if we say that ‘water boils at 212 degrees Fahrenheit’ and we find that this observation can be consistently empirically observed, we can say that such a judgment is objectively correct. If, in fact, the boiling point of water was 150 degrees, our original claim would contradict our empirical observations, and would thus be false. To Kant, ‘correct’ observations can be consistently concluded by all rational beings, which gives them an ‘objective’ validity even if they do not describe reality in-itself.

Author’s Response: The Humean could concede much of this to Kant, but still disagree with the conclusion. That is, they might agree that all it means for a sentence to be ‘objectively’ valid is to hold true to all rational beings, but would add the caveat that the sentence is nonetheless formulated in some language. Thus to be ‘objectively’ valid is language-dependent, and the rational beings over which objectivity is evaluated must be competent users of that language. In the case of causality, in a linguistic community who lack a word for ‘causality’, the sentence ‘Gravity causes the Earth to revolve around the sun’ has no meaning (or translation), so is not objectively valid. Nonetheless it is, even to the Humeans, objectively valid in our linguistic community.

Understanding these positions, I want to return to my main point: that our language may structure our science. If the scientific realists are correct, then language plays no role in determining the scientific truths, themselves, just our understanding of them. But if the Humeans are correct, then language plays a foundational role in scientific truths, since the laws of nature, and thus the truths of physics, chemistry, biology, etc. are theory-laden, dependent on the best theories which describe experience. And those theories, themselves, are inevitably formulated in some language. In this way, the language we use to describe reality, in a very literal sense, defines our reality, rather than merely influencing our perceptions of it.

Editor’s Note: To Kant, the way we formulate them in our language reflects fundamental, a priori rules which structure how we perceive the world and make judgements about it. It is, thus, deeper than our language.

Author’s Response: The Humeans would agree that there are rules which structure how we perceive the world, but suggest that they are nonetheless a posteriori. The ‘rules’ of experience, then, are just the universal regularities in our ‘best’ description of that experience.

To see how this occurs, allow me to introduce a toy case with the concept of “causality”. Suppose the Humean point-of-view is correct: that laws are nothing more than descriptions of universal truths in the best theories which describe our collective experiences. Now consider two theories of science (broadly-construed): one which includes a word for both “correlation” and “causation” (like our current scientific theories), and one which only includes a word for “correlation”.

Author’s Note: I’m using ‘force’ here in the classical sense, for simplicity. In general relativity, however, gravity itself is caused by the warping of spacetime, and thus the term “force” is often replaced with ‘interaction’.

Let’s try and formulate a basic definition of gravity in both theories. In the first theory, we can say that gravity is that force which causes stones to fall when dropped, and the Earth to revolve around the sun, and spacetime to get warped into black holes, etc. In the second theory, we can say that gravity is the falling of stones when dropped, and the Earth’s revolution around the sun, and the warping of spacetime as black holes, etc. The former theory suggests that gravity is somehow the cause of all the events which are merely present as correlations in the second theory. But what work is this causal relation playing in the theory’s ability to explain our experiences? Is there really any deep, metaphysical, distinction between the two theories? To investigate this, we need to better define the role causation is playing in the former theory.

Suppose you stand above the Earth, hold out a stone, let it go, and see it fall to the ground. Why do you believe gravity caused the stone to fall? Presumably, you believe something like this: if gravity had not existed, then the stone would not have fallen to the ground. But why do you believe this? Let’s return to Hume’s fork, that knowledge comes from relations of ideas or matters of fact. Is it a relation of ideas that if gravity had not existed, then the stone would not have fallen? Certainly not, for some other force could have acted upon the stone to cause it to fall. Perhaps it is magnetic, along with the ground below you, or perhaps an invisible hand could arrive and push it downward. (These are strange scenarios, indeed, but not impossible. And definitely not impossible in worlds where gravity fails to exist!)

Thus it must be a matter of fact that the ball would have failed to fall to the ground in a world with no gravity. But to verify such a matter of fact, you’d have to observe it. And you cannot observe the world in which gravity fails to exist because gravity, of course, does exist. So how can you justifiably believe that if there had been no gravity then the stone would not have fallen? It seems that you have no good reason to do so.

Editor’s Note: Notably, though, correlation—or more broadly, the concept of ‘relation’—is essential for any natural model of the world. In my Kant-informed perspective, even if we give language its most generous possible role in shaping our observations, some concepts must proceed it. However, contrary to Kant, I think you make a pretty good case that ‘causation’ isn’t a metaphysically necessary concept, and that we could develop a robust and accurate scientific model of the world without it.

Returning, then, to what role causation was playing in the earlier theory, it seems it may have been playing none at all, at least not metaphysically. Notably, the force of gravity exists in both theories. In the former, it is “causal”, somehow determining events in unactualized possible worlds, while in the latter, it is merely a collection of correlative events. So why do we prefer the first theory to the second? Why have we included the concept of causation in all our scientific theories? This is where the Sapir-Whorf hypothesis makes its unexpected return.

Editor’s Note: I would argue that, for our concept of ‘causation’ to make our language more efficient, it has to represent something distinct from other forms of correlation. A ‘causal’ relationship usually represents that, after an event occurs, another event necessarily or is more likely to follow, and that this relationship between those two events is directional.

There are correlated events which give the illusion of causality—eating ice cream, for example, is positively correlated with getting a sunburn, but claiming they have a causal connection would be false since, in fact, they are both caused by sunny weather… but, then again, this relationship can be described purely with the concept of correlation: we can say that ice cream and sunburns are indirectly correlated since they are both directly correlated with sunny weather.

So, maybe you’re right, and that causation is simply a more efficient way to describe a direct correlative relationship that doesn’t add anything new. But, it’s still possible that ‘causation’ and ‘correlation’ are metaphysically distinct concepts that we use to make judgments, with or without language. But in either case, the concept of ‘correlation’ seems to be more fundamental and necessary than our concept of ‘causation.’

Author’s Response:To be clear, ‘correlation’ and ‘causation’ are metaphysically distinct concepts, since correlation is only dependent on the actual world while causation is dependent on other ‘nearby’ possible worlds. Nonetheless, while the Humeans will agree that there are objective (in the external, non-language dependent sense) facts about correlation, there may be no similar such facts about causation.

Recall that the weak formulation of Sapir-Whorf suggests that the inclusion of certain words into a language can reduce the cognitive load attached to processing tasks performed in that language. Such theories, in philosophy talk, are epistemically virtuous; they allow us to reason efficiently and soundly. Then maybe that is all “causation” is really doing in our scientific theories, allowing us to reason more efficiently about the vast collection of events we view on a day-to-day basis. If that’s true, then there’s nothing deeper to causation than that. In that case, some basic scientific concepts, like causation, are no more metaphysically deep than descriptive differences between the many different kinds of ice and snow.

Caleb Camrud is a Ph.D. student studying philosophy at Brown University, with a focus on metaphysics, philosophy of language, and philosophy of mind. Editor’s Notes were written by Bobby Becker, a Master’s student at Tulane University and the founder of The College Contemporary.