Old fish and young fish meet somewhere in the ocean. Old fish: How is the water? Young fish: What is water? 

In the scientific community, we usually establish correlations, but ideally, we want to establish causation. Correlation is not causation, but we attempt to develop different approaches to be able to speak about causes. However, the notion of causation seems so self-evident and obvious that nobody really discusses what it is exactly that we want to uncover. I learned that this impression is misleading, and causation is not obvious. The book that I read offers to engage in deeper contemplation. It is “Getting Causes from Powers” by Rani Lill Anjum and Stephen Mumford. 

There are objects, and there are events. For example, objects are cats and vases. An event is a cat breaking a vase. The authors suggest that objects possess properties (=powers). These properties are powerful in the sense that they may perform causal work. Causal effects are when causal powers manifest themselves (or when dispositions become realized). 

This proposal challenges David Hume’s idea (also called the regularity theory or two-event model). For Hume, causation is the relation between two events that come in succession (close together in space and time): the cause occurs, followed by the outcome. In regularity theory, causation is not about underlying forces or vectors, but about the regular association termed constant conjunction (cause is constantly followed by the effect). The authors reject the regularity theory. 

Main theses of the authors’ proposal. 

Causation may be described with a vector model. The vase-breaking causes include those that dispose towards breaking and those that dispose away from breaking. The cat jump is a vector in the direction towards breaking. Similarly, vase composition (fragility) and vase placement (at the edge of the table) dispose towards breaking. In contrast, the presence of cat toys in the room or the light weight of the cat disposes away from breaking. Effects are produced by many causal powers working together, which is termed a “poligenic” nature of causation. 

There are many ways for powers to combine to produce an effect. Because many objects/events combine in the vector model, there could be multiple combinations of them when it comes to bringing about the effect. A heavier cat will increase the vector toward breaking, even if the vase is stable in its placement. Conversely, a lighter cat may be offset by a more fragile vase. Additionally, the possibility of multiple combinations implies the possibility of overdetermination. A heavy weight of the cat may alone determine the vase breaking, and the vase position at the edge of the table may alone determine the vase breaking, but also the heavy cat and vase position together determine the vase breaking. 

Properties are dispositions. The vector model suggests that properties that objects possess may combine in a linear (or non-linear) way to produce an effect. It means that a single property (=causal power) of a single object merely disposes towards the outcome. The fragile vase doesn’t break simply because it is fragile, but we may agree that being fragile in the presence of a cat increases the chance of an unfavorable outcome. The authors are against necessitation, so no single causal power is necessary to produce an effect. Instead, they advocate for a threshold. The sum (or some non-linear combination) of powers crosses the threshold, and the outcome is produced. However, it doesn’t mean that before the outcome, there is no causation. Causal powers operate all the time, including before the crossing of the threshold. 

Causation is simultaneous. Causation is a process in which the cause turns into the effect (almost as if). In fact, the authors reject the two-event model, stating that there are no causes or effects, but rather a single causal process. It's not that the cat caused the vase to break. The cat and the vase with their dispositions are parts of the same process that evolved from one state to another. In the same vein, two causal processes may be linked but not in a sense that one causes the other, but that one provides powers (vectors) into a set of powers that already act within another. 

Causation can be perceived. We are acquainted well with causation because we experience it in our bodies. When something is touching our skin or when we make voluntary movements, we “experience” the causation. 

The vector model and dispositionalism sound reasonable to me. However, if we accept that, the problem for empirical research arises: identifying disposing factors and their weighted contributions to the vector configuration, as well as transformations (linear and non-linear) from disposing factors to the outcome, seems to be nearly impossible. Also, if we accept the proposal, the common standard approach "Variable A causes Variable B” used in most labs is technically a massive oversimplification. It's not clear what to do about it, but having conversations may be a starting point.

Simultaneity of causation strikes me as contrintuitive. The authors provide an example of raising the hand. They say that when you want to raise your hand, it rises right away, not within some time. But actually, it takes time. The signal that needs to travel from the associative cortex to the motor cortex, then to the alpha motor neuron, and then to the muscle fibres in the hand. The authors note that simultaneous doesn’t mean instantaneous. So it seems that the whole signal transmission can be interpreted as a process of raising the hand. Still, it feels as if first I decide to raise a hand, and then the hand rises; it doesn’t feel simultaneous. 

Similarly, it’s hard to agree with the idea that causation is directly perceived. The authors describe that something touching our skin evokes the sensation of causation (=perception). However, if something is a perception, there should be perceptual organs. It’s not clear whether Pacinian corpuscles (skin receptors detecting pressure) can detect causation. Additionally, from a subjective perspective, it feels that a lot of causal relationships are learned from previous experience or inferred via observation. It may be argued that inference is possible only because there is lower-level perception, but I don’t find it convincing. It appears more convincing that causation is understood through the activities of one’s own body, such as raising the hand. However, in this case, the brain “knows” about the movement it performs due to the efferent copy sent to sensory regions. It is not directly perception, but the sensory cortices are involved. Still, even people with disorders, such as alien limb syndrome (when patients report that the arm has a mind of its own), have no problems attributing causation to the movements of the alienated arm. For me, the claim that causation is perceived is too strong.

Overall, the book was stimulating. It may be an interesting read for empirical researchers. 

Favourite quote: 

“we use causal verbs all the time, which might be taken to indicate that we acquire causal concepts easily and use and apply them routinely”

March, 2026