Causal Talk Permeates Ordinary Language

Abstract

This chapter gives an overview of causal idiom in ordinary language and introduces some fundamental semantic distinctions. The main points are:

• The words ‘cause’ and ‘effect’ have quite a number of near synonyms.

• ‘Cause’ and ‘effect’ are relational terms.

• Causal relations relate three kinds of things: events, categories of events and variables used in statistical treatments of such categories.

• A mathematical relation between variables does not in itself tell us whether they are causally related or not.

1 Introduction

There is a great variety of causal expressions and even a very brief overview of ordinary language reveals how common expressions for causal relations are; some examples are ‘bring about’, ‘make happen’, ‘produce’, ‘do’, ‘perform’ ‘result in’ ‘effect of’, and ‘leads to’. Two questions immediately come to the fore: (i) is there any common meaning of these expressions, a core meaning of respectively ‘cause’ and ‘effect’, and (ii) what evidence is required as justification for causal claims?

Many different things have been related as cause and effect in ordinary as well as in philosophical and scientific discourse: events, states of affairs, properties, categories, quantities, processes, desires, beliefs and actions. This list is not complete and one may reasonably wonder if there is any core meaning at all for all the diverse uses of the terms ‘cause’, effect’ and other terms with a causal connotation. We believe there is, there are some necessary conditions for the correctness of sentences of the form ‘x causes y’, and its converse ‘y is the effect of x’, but in any particular context more conditions are needed and these differ in different contexts. There is a wide diversity of causal notions, which will be surveyed and analysed in this book.

Causal idiom and causal thinking is a basic trait of humans. Already in the second verse of Iliad (circa 800 BC), the first work of western literature, We are given a seemingly causal explanation of the hostility between Agamemnon and Achilles:

  And which of the gods was it that set them on to quarrel?

It was the son of Jove and Leto; for he was angry with the king

and sent a pestilence upon the host to plague the people,

because the son of Atreus had dishonoured Chryses his priest. (Transl. S. Butler).

Thus, the conflict between Agamemnon and Achilles is said to have been caused, indirectly, by the god Apollo (‘the son of Jove an Leto’) who is attributed a typical human psychology: Apollo was angry with the king and acted accordingly. The cause of the quarrel is an action of a sentient being.

In the epilogue to his book Causality: Models, Reasoning and Inference Judea Pearl does not mention this passage, but noticed that in ancient times questions about causes were questions about agents, their motives and desires:

The agents of the causal forces in the ancient world were either deities, who cause things to happen for a purpose, or human beings and animals, who possess free will for which they are punished and rewarded. This notion of causation was naive but clear and unproblematic.

The problem began, as usual, with engineering; when machines had to be constructed to do useful jobs. [....] And once people began to build multistage systems, an interesting thing happened to causality - physical objects began acquiring causal character. (Pearl, 2000, 333)

One might say that in ancient times the prototypical cause is an agent who acts for a purpose, whereas from the scientific revolution onwards the prototypical case is a ball colliding with another ball and changing the latter’s motion.

The ancient notion of causation—a cause is an action of an agent—is still very common in ordinary thinking and language, less so in scientific discourse, except perhaps in some social sciences. The latter are concerned with human actions, individual and collective, so in these disciplines there is plenty of talk about causal agents driven by beliefs and desires.

Having analysed the different aspects of causal discourse in both ordinary and several scientific contexts, we will at the end of the book focus on causal reasoning in complex SES. Analysing causal relations in such systems is particularly demanding because of the diversity of kinds of entities, properties and relations and the complex dynamics they generate.¹

2 Causal Phrases

As an illustration of the variety of expressions used for making causal claims, we may have a look at the beginning of (Lindegren et al., 2009):

Atlantic cod (Gadus morhua) is among the commercially most important fish species of the European waters. Many of the stocks have declined dramatically and still remain at historically low levels (1, 2). These collapses have largely resulted from overfishing (3, 4) and climate-driven declines in productivity (5, 6). The climate effect generally works through changes in the physical environment (e.g., temperature and salinity), but also through altered food supply for early life-history stages, eventually affecting recruitment (5, 6). In accordance with this effect, recruitment failure of Eastern Baltic cod was caused mainly by high egg and larval mortalities as a result of climate-induced hydrographic change (7, 8). In several areas the collapses of cod stocks were part or major drivers of large-scale reorganisations of ecosystems (9). These so-called regime shifts are frequently caused by climatic changes (9, 10) and/or over-exploitation resulting in cascading trophic interactions (11, 12). Similarly to other areas, the Baltic Sea underwent both regime shifts and trophic cascades (8, 13). Such alterations in ecosystem structure typically affect species interactions, eventually influencing food-web dynamics through both positive and negative feedback loops (14).

In this introduction we observe at least seven expressions indicating causal relations:

• 1. ‘These collapses have largely resulted from overfishing.’

• 2. ‘...climate driven declines in productivity...’

• 3. ‘The climate effect generally works through changes in the physical environment.’

• 4. ‘...failure of Eastern Baltic cod was caused mainly by high egg and larval mortalities.’

• 5. ‘....the collapses of cod stocks were part or major drivers of large-scale reorganisations of ecosystems.’

• 6. ‘These so-called regime shifts are frequently caused by climatic changes.’

• 7. ‘Such alterations in ecosystem structure typically affect species interactions, eventually influencing food-web dynamics through both positive and negative feedback loops.’

All these sentences convey information having the form that someone or something makes something else happen. In neither case could the researchers directly have observed these events, so one immediately wonders how the authors know that the relations expressed by ‘cause’, ‘make happen’, ‘drives’, ‘affect’, etc, really obtain, i.e. what evidence they have collected. This question in turn triggers the question what these expressions really mean, for we cannot decide what evidence we require for a certain statement if we do not know what it means. Questions about evidence for causal claims and questions about the meaning of these claims are thus deeply and intimately related; the meaning of an expression determines what kinds of evidence there might be for a sentence containing this expression.

The diversity of ideas about causation is not only a matter of methods, there are also different ideas of what causation is and what it entails. For instance, people disagree about (i) generalisability of singular causal relations, (ii) origins of causal powers, and (iii) which causes are more important than others.

3 Some Remarks on the Semantics of ‘Cause’, ‘Effect’ and Their Cognates

3.1 Causal Relations Between Events/States of Affairs

Our most basic use of causal idiom consists in relating particular events or states of affairs by a two-place predicate ‘x caused y’, ‘x was the effect of y’, ‘x leads to y’, or some other expression with clear causal meaning.² In other words, things causally related to each other are singular events or states of affairs that occur at particular times and places. This is the fundamental use of causal idiom. But use of words for causal relations is wider: with the development of modern science, causal talk has been extended to cover also relations between categories and quantities.³ By abstracting from individual cases, we simply say that one attribute is the cause of another attribute. For example, overweight is said to be a cause (not the only one!) of high blood pressure.

3.2 Causal Relations Between Categories

The pandemic Covid-19 was caused by the virus SARS-CoV-2. This is a relation between two categories: the cause is a virus of a certain type or category, and the effect is a disease of another category; being exposed to particles belonging to this virus type increases the probability of getting the disease Covid-19.

A necessary condition for this causal relation to occur is that the conditional probability to attract Covid-19 when exposed to the virus is higher than the marginal probability to get the disease. In mathematical notation:

$$\displaystyle \begin{aligned} prob(\vert \text{Covid-19 | being exposed to SARS-CoV-2}) > prob( \text{Covid-19}). \end{aligned} $$

(2.1)

This means that Covid-19 and SARS-CoV-2 are correlated. Still, this statistical relation is not sufficient evidence for there being a causal relation; more evidence is needed, see Sect. 2.4 and Chap. 6. In this case we know that the virus is the cause of the infection based on experimental evidence, not just on statistical correlations. More carefully expressed: we know that in any individual case of someone having the Covid-19 infection, this event was caused by him being infected by the SARS-CoV-2 virus. So we generalise and say that the virus causes this disease, thus saying that one category causes another category. We have thus extended the possible relata of the relation ... causes.... to include categories. (In this case the disease is identified by its cause, so the probability of getting the disease without being exposed to SARS-CoV-2 is zero.)

From a purely grammatical point of view, the terms ‘SARS-CoV-2’ and ‘Covid-19’ are singular terms (see textbox below). That does not conflict with their referents being categories, i.e., classes of things. A category, when given a name, is treated, grammatically and from a logical point of view, as an entity, a particular thing. (Whether this should be understood as that there exists categories, i.e., properties and relations, over and above the individual cases, is a perennial dispute in metaphysics. Luckily, for the purpose of this book, we need not take any stance in this debate.)

Singular and General Terms

From a logical point of view, the simplest cases of complete declarative sentences consist of a general term, (a predicate phrase) and one or several singular terms. Singular terms are proper names, personal pronouns, definite descriptions, or variables standing for such things.

A singular term cannot function as predicate, while general terms can occur both as nouns and predicates.

An example with one singular term is

• #1. The oldest person in Sweden is more than 100 years.

Here, ’The oldest person in Sweden’ is the singular term, it is a definite description of one distinct entity. The rest of the sentence is the predicate phrase ‘is more than 100 years’; this is a general term, which means that it can be predicated about many things.

A sentence with two singular terms is

• #2. Sweden has a smaller population than Germany.

Here, ‘Sweden’ and ‘Germany’ are the two singular terms, they are each a name of a political and geographical entity. (Thus in predicate logic we do not distinguish between noun and direct object.)

A sentence with three singular terms is

• #3. The judge in a civil case determines whether the plaintiff or the defendant wins the case.

Here, ‘The judge in a civil case’, ‘the plaintiff’ and ‘the defendant’ are the three singular terms.

So singular and general terms are the logical constituents of the simplest complete declarative sentences. Singular terms may refer to things, objects, events, etc., i.e., entities that can be identified as individuals, i.e., thought of as one.

A singular term need not refer to anything. ‘The king of France’ is a singular term, but there is no king of France.

Things referred to by singular terms, i.e., individuals, need not be individuals in any ordinary sense. In the sentence ‘Manchester United won The Premier League 2012–2013’, the name ‘Manchester United’, is a name for an individual entity, a football club. When talking about Manchester United we treat it as one object, disregarding whether or not it consists of a number of players and other members.

3.3 Causal Relations Between Quantitative Variables

A common scientific question is whether a certain variable is the cause of another variable and huge efforts are often made in order to answer such questions. The starting point when asking about a possible causal relation between two variables is to see whether they are correlated or not. Suppose the answer is yes. That is not sufficient for inferring that they are causally related, since a correlation can occur without there being a causal connection. But if there is a causal relation between two variables, they are correlated, when other variables are controlled for. So observing a correlation is a reason for further inquiry to see whether there is a causal relation or not.

As a starter, we observe that an expression of the form ‘variable X is a cause of variable Y’ means that a change in the value of variable X attributed to some object causes a change in the value of variable Y attributed to the same or another object. In other words, causal relations between variables are based on causal relations between ordered pairs of individual events. But neither in ordinary nor in scientific contexts is this explicitly stated; the common expression is that a certain variable causes another one.

Relations between two quantitative variables are often expressed as functions of the form $y=f(x)$. Such an expression does not contain any information about any causal relation, because a function by itself does not express any relation between events in the real world. But even if we add an interpretation to the effect that variable values represent events in the world that are causally related, the form of this expression does not distinguish between cause and effect. The reason is that the causal relation is asymmetric, it is ‘directed’ from the cause to the effect, while a mathematical function only expresses a numerical relation between the values of the two variables.

In many cases, when the function $y=f(x)$ has an inverse, this is immediately clear, since the equations $y=f(x)$ and $x=f^{-1}(y)$ are logically equivalent, they are two different expressions for the same fact of the matter. Thus, the mere syntactic form of the equation $y=f(x)$ does not tell us anything about which is the cause and which is the effect, or whether there is any causal relation at all between these variables. We will discuss this more thoroughly in the next section. The formalism of structural equations is another matter, to be discussed in Chap. 7.

Most often a causal interpretation of an equation comes from an intuitive and tacit judgement about which variable we naturally, or most easily, can manipulate. Hence, the causal interpretation of a mathematical relation between two variables comes from our agency perspective, to be discussed in Chap. 3. An equation by itself does not say anything at all about causal relations.

3.4 Common Causes

The fact that a correlation between two variables is very strong does not by itself say anything about the probability for there being a causal connection. There are many well-known cases of correlations between two variables that no one would think of as causally related. The correlation is in such a case explained by being produced by a common cause, often called a confounder. If variable A is a cause of variable B via one causal mechanism and a cause of another variable C via another causal mechanism, we may observe a correlation between B and C without there being any causal link between them. Here is one example of a strong correlation which most plausibly, given even a very limited background knowledge, is the outcome of a common cause. Tyler Vigen, from whose home page tylervigen.com/spurious/correlation the figure is taken, discusses some possibilities (Fig. 2.1).

Fig. 2.1

The number of bachelor’s degrees in physical science strongly correlates with the distance between Saturn and the moon during the period 2012–2021, see Tyler Vigen, Spurious Correlations, available at https://tylervigen.com/spurious-correlations

The expression ‘confounding cause’ is often used when referring to a common cause. This term is a bit misleading since it suggests that the confounding cause is not a real cause, while in fact it is the cause of the correlation. What the confounder can do is to mislead us into thinking that there is a causal link between two observed and correlated variables, while there is not.

How to decide, by empirical means, whether two observed variables really are causally related or whether there is a common cause will be discussed in Chap. 6.

4 Causal Powers

When we talk about the cause of a certain effect, we are inclined to describe the situation as that some entity, a person, a physical object or a machine, has a certain causal power, which under certain conditions is manifested by bringing about the effect. It is, for example, common to say that the Earth has the power to attract bodies and that this explains why a stone falls to the ground. This causal power is the Earth’s gravitation and this explanation appears to most people satisfying.

Another example: some persons are charming, which is a dispositional property of having the ability to charm other persons, thus sometimes causing certain behaviours in other persons. For example, one might explain a person’s foolish behaviour by saying that he/she was charmed by another person, and the latter is well-known for their charming capacity. Thus, the terms ‘dispositional property’ and ‘capacity’ are often believed to express causal powers attributed to things.

One may first observe that a causal power is not the same as the cause of a particular event or action. The cause of a particular event is another event. But when we observe a regular connection between two types of events, for example that ponderable bodies fall to the ground when the support is removed, or that many people are charmed by a certain person, we are inclined to explain such a regularity by postulating a causal power in the entity thought to be the cause of this regularity. Thus, causal powers are invoked in causal explanations of recurring features and behaviour, although strictly speaking the causal power is not the direct cause of a singular effect.

Saying that a certain thing has a certain power is to ascribe to this thing a dispositional property, i.e., a property that only under certain conditions is manifested as an event that can be observed; but causal powers themselves are not observable, although thought to be permanent properties of things.

Postulating that an object has a certain causal power may be regarded as the proper explanation of our observations of some regular and recurrent behaviours, but it does not increase our ability to manipulate things or predict the future. This is so because any inference to future events is based on observed regularities, and these observed regularities are exactly the very reasons we have for inferring a causal power. Postulating a causal power does not increase the probability for a certain future event to occur. Hence, philosophers with an empiricist mind-set are sceptical about causal powers. The argument is Ockham’s razor: ‘do not without necessity postulate an entity.’ (There are several formulations of this principle; the Latin version is ‘Entia praeter necessitatem non esse multiplicanda.’) We can make exactly the same inferences, with the same degree of certainty, to future events without causal powers.

Reflecting on Ockham’s razor, one might ask ‘necessary for what?’ and it is pretty obvious that the tacit assumption is that the goal is to make predictions about events and states of affairs not observed when the utterance is made.

But causal powers are often held to have explanatory force, they give us understanding of recurrent events, and one might be tempted to infer that understanding is a prerequisite for successful predictions about the future.

The validity of this line of thought depends crucially on the criteria for understanding a phenomenon. A prediction either succeeds or fails and that can be determined by observations. But what are the criteria for understanding? They seem to be crucially dependent on background knowledge had by those being given the explanation. We will discuss this topic further in Chap. 8.

Causal powers are unobservable, but that is not the relevant epistemological point. There are, for certain, many cases in the history of science where unobserved entities are postulated in order to explain observed phenomena. The crucial point is that such a postulated entity is accepted as real only when there is independent evidence for its existence. In the case of causal powers, there cannot be any such independent evidence; a causal power only manifests itself as a certain observed regularity, which is exactly the same as what is needed as the empirical basis for inferences about unobserved events.

Summarising, there is no empirical evidence for causal powers being responsible for the observed events in nature or for human behaviour. Nevertheless, many people hold that causal powers explain observable events. Whether that is so depends on our criteria for causal explanations.

5 Summary

Causal idiom is a basic feature of natural language, just as words for e.g., animals, colours, people, activities and events. The difference between on the one side words for these things, and on the other hand causal expressions, is that the latter concern relations between pairs of entities, while the former are talk about singular entities. In the second case, one observes two events and under certain conditions infers that they are related as cause and effect. Children learn causal expressions directly, in interactions with parents and other care-takers, not by being taught verbal definitions.

The notion of cause (and effect) can be expressed by quite a number of different words and expressions, for example, ‘bring about’, ‘make happen’, ‘produce’ and ‘do’. None is more fundamental than the other.

When we ask for a cause, we have tacitly a certain event in mind, we ask for the cause of a particular event. The question has the form ‘What is the cause of E?’ Hence, the terms ‘cause’ and ‘effect’ are relational words; the effect (or the cause, if the effect is asked about) is often not mentioned as being obvious in the context at hand.

Questions about causes and effects are basically questions about relations between events and, secondarily, relations between types of events and their representations, quantitative and category variables. Casual relations between quantitative and category variables depend on causal relations between individual instances of these relations.

Discussion Questions

• 1. Is it possible to observe the causing relation, that a particular event causes another event? If your answer is yes, how, then, to distinguish between the mere succession in time of two events and them also being causally related?

• 2. Could a singular event be the cause of a category or vice versa?

• 3. Every day the sunrise is followed by sunset (provided you are not living in polar areas) without exception. Why not say that the sunset is caused by sunrise?

• 4. Many laws in physics are expressed as mathematical equations relating quantitative variables. Such laws, for example Newton’s second law, (${force = mass \cdot } acceleration$) is usually thought of as stating a causal relation. What is the justification for such a reading?