In Causes and Conditions and Causation, J.L. Mackie and David Lewis respectively offer two differing accounts of causation. The regularity account of Mackie explains causation as a number of causes, one of them being necessary for an effect with the rest being sufficient, that together are both necessary and sufficient to bring about a given event. Lewis explains causation as a chain of events in which the last event is causally dependent on the previous such that it causally depends on the first one, which can be said to have caused the last one. Although both Mackie and Lewis successfully meet the challenge posed by pre-emption, the problem is not significant to the relative merits of the two accounts. Indeed, it is shown that Mackie is unable to distinguish between causal and non-causal relations, namely epiphenomenals, owing to his view of causation as a regularity. Lewis, however, offers an account under which causal and non-causal relations are differentiated, ultimately making his a more accurate explanation of causation.

Mackie’s account of causation is as follows: it can be said that A causes B iff the conditions as he sets out are met. It is required that A in conjunction with background conditions, X, constitutes a necessary and sufficient condition for B, together known as an insufficient but necessary part of an unnecessary but sufficient (INUS) condition. Also, it is required that both A and X are present when the event occurs, or that all three of A, X and B occur. Applied to an example, it can be said that a short circuit is the cause of a house fire iff there were background conditions enabling the short circuit to cause a fire, such as a poorly constructed house, no fireproof insulation and an unfortunately-placed stack of newspapers.

In the case of Lewis, it is said that A causes B iff there is a chain of events leading from A to B so that B causally depends on the last link in that chain, which in turn ultimately causally depends on A. An event causally depends on another iff the counterfactuals “if A hadn’t occurred, B wouldn’t have occurred” and “if A had occurred, B would have occurred” are both true. Applied to the earlier example, the fire is the culmination of a chain of events in which each constituent event depended causally on there being a short circuit.

With respect to the problem of pre-emption, Mackie cites the hypothetical example of two criminals, Smith and Jones, who were sent by a crime syndicate to commit a murder. If they had not committed the murder, the head of the organization would have sent two other criminals in their stead. Mackie shows that Smith and Jones were the cause of the crime because they meet the last of his conditions for a causal relation, but the other members of the criminal organization do not. In other words, while one can say that in the wider picture Smith and Jones, along with the relevant background conditions, were neither necessary (the other criminals would have done it) nor sufficient (the victim could have died of a heart attack). However, for the event as it happened, AX was both necessary and sufficient; the victim died as a result of Smith and Jones firing their guns and having been in the place he was in, not having been killed by a heart attack moments prior. Furthermore, the other criminals don’t satisfy the whole of the conditions Mackie sets out. Smith and Jones are said to have been necessary post factum, which the other criminals are not.

The account of causation offered by Lewis shows that Smith and Jones caused the crime because there is a causal chain from Jones and Smith firing their guns to the death of the victim. There is no causal dependence between Jones and Smith firing their guns and the death of the victim because a later event, such as the other criminals arriving a half-second later and firing their guns, could have killed the victim. Therefore, the death of the victim does not depend causally on Jones and Smith firing their gun. However, Jones and Smith can still have caused the death of the victim because that causal chain runs to completion; the causal chain running from the other criminals to the death of the victim does not run to conclusion since the victim would have been already killed.

Examples of pre-emption are significant as a test of the functionality of an account of causation, in that they account offers more than an explanation of a cause as being necessary and sufficient to bring about an effect. Both Mackie and Lewis pass this test by arguing what one would intuitively argue when presented with claims of pre-emption. Mackie argues that in the event, as it happened, A, along with the background conditions in X, was necessary and sufficient to bring about B and the pre-empted cause was not necessary because the effect happened without it. Similarly, Lewis argues that a causal chain exists between A and B that does not exist between B (the effect) and any of the pre-empted causes because the latter chain was interrupted when A caused B to happen. In essence, what both Mackie and Lewis put forth is that problems of pre-emption do not causation because A can be said to have caused B in the event as it actually happened. Given that both Mackie and Lewis are able to successfully integrate the problem of pre-emption into their respective accounts in rather similar terms, it cannot be said that examples of pre-emption are important to a comparison between these two accounts.

However, there are other, more significant terms of comparison. One of these is that the Mackie account cannot distinguish between what are causal and non-causal relations between events. For example, it would be consistent with Mackie’s INUS conditions to say that a tabletop being one metre above the ground causes the legs of the table to be one metre long. Although it is not the case that being one metre off the ground causes the legs to be a metre long, in this particular situation with the given background conditions, such a claim would be valid. Furthermore, Mackie’s account is also troubled by epiphenomenal relations where one cause produces two effects, such as flicking a light switch and both lighting a room and using electricity. It would be again consistent to say that lighting a room is an INUS condition for electricity consumption and therefore simply lighting a room causes electricity to be consumed.

The account of Lewis is not subject to this problem because it has stricter notions of what constitutes a causal relationship. Mackie defines causal relations in terms of laws, which in their application lead to a broad set of casual relations that in fact includes non-causal relations, such as lighting a room and electricity consumption, to which the INUS condition applied. The counterfactual account of Lewis does not allow non-causal relations to be considered causal relations. In the first example of the preceding paragraph, the chain of events by which the table leg came to be one metre long would include, perhaps, its being glued to the tabletop and its having been cut to the length of one metre. It would not, however, include the height of the tabletop above the ground. Further analyzed in counterfactual terms, namely that an event can be said to causally depend on another if the statement “if c had not happened, e would not have happened” and its converse are both true, it is clear that if the tabletop was not one metre above the ground, the legs would still be one metre long.

Both Mackie and Lewis, in offering accounts of what constitutes a causal relation, adeptly meet the challenge of pre-emption which undermines causation as defined in terms of simple necessity and sufficiency. However, pre-emption is not the sole barometer of a successful account of causation given that the solutions to pre-emption offered by both Mackie and Lewis are intuitive and rather similar. Examined on further terms, it is seen that Mackie’s INUS conditions are unable to distinguish between causal and non-causal relationships due to the widespread application of his law-like regularity. Conversely, Lewis’s strict requirements for what constitutes causation, as outlined by a causal chain, do not meet with this problem. Lewis’s account is able to differentiate between causal and non-causal relations on the basis of a causal chain that does not regard non-causal relationships such as epiphenomenals as being causal, as is the case with Mackie’s account.