In May 1987, Kenneth Parks drove 14 miles to his in-laws’ home, entered the house using a key they had given him and bludgeoned his mother-in-law to death. Then he drove to a police station and handed himself in. His defence team argued that he was sleepwalking – technically, in a state of ‘non-insane automatism’ – and he was acquitted.

If we assume that he was sleepwalking, that may seem fair. Although he certainly caused his mother-in-law's death, it wouldn't be right – would it? – to hold him responsible for doing so, and convict him of murder, if he really was deeply asleep at the time. But why wouldn't it be right? Is it because he didn't kill her voluntarily? Because he didn't intend to kill her? Because he didn't know what he was doing? Because he couldn't understand at the time that killing her was wrong? Or because he wasn't in control?

The traditional answer goes back to Aristotle. Praise, blame, punishment, and reward, he taught, are all reserved for voluntary conduct. The American Model Penal Code agrees. It says that a defendant's guilt must always be due to ‘conduct which includes a voluntary act’ – or else to his failure to do something he was capable of doing, such as care for a child.

But many philosophers find the idea of a ‘voluntary act’ obscure. In early modern philosophy it became associated with the idea of a ‘will’ – a psychic mechanism that executes our desires by causing motion in our limbs. Psychologists since William James have regarded the will as a fiction, and neuroscientists have not discovered it in the brain.

Philosophers have therefore argued that what matters for responsibility is something quite different from a ‘voluntary act’. For example, one popular idea is that in order to be responsible, Parks would have needed to be ‘responsive to reasons’, in other words, capable of being guided by his knowledge that killing one's mother-in-law is wrong.

***

Philosophers have always relished the rich repository of examples in case law, such as R v Parks. But legal theory has something to offer philosophy as well. To see this, we need to take a step back from the puzzle about Parks to the ancient philosophical task of giving definitions, and ask what responsibility itself is.

Once we've answered that question, we can return to the question whether responsibility can only be for conduct that ‘includes a voluntary act’, and whether it requires the ability to respond to reasons, or intention, or control, or none of the above.

The Victorian philosopher F. H. Bradley defined responsibility as liability for punishment – in other words, responsibility for wrongdoing means that one can be legitimately punished for it. And philosophers have been guided by this definition ever since.

In the 1960s, when liberal ideas about punishment had a major impact on philosophy, it became more common to equate responsibility for wrongdoing with liability to blame or resentment instead of punishment. But the equation between responsibility and liability to (i.e. the legitimacy of) some kind of response persisted.

By contrast, legal theorists do not equate responsibility and liability. For example, a defendant's criminal responsibility for wrongful conduct – whether it is stealing an apple or killing one's mother-in-law – can be defined as his obligation to account for it, in other words, to explain the reasons why he did it. Liability follows later, if the explanation isn't good enough – in other words, if it doesn't avert liability, by providing an adequate excuse. At that point, the defendant becomes liable to conviction, and often punishment as well.

***

This distinction between responsibility and liability depends on thinking – in a rather abstract way – about the structure of a legal trial, so it may seem quite far removed from philosophy. But it transforms the problem of explaining why Kenneth Parks deserved to be acquitted, assuming he was deeply asleep at the time of the offence.

Suppose we are told either that he deserved to be acquitted because he didn't kill his mother-in-law voluntarily or because he wasn't able to understand and be guided by reasons. Since we are no longer equating responsibility and liability, we can now ask whether the defence is supposed to show that he wasn't responsible (i.e. obliged to account for his conduct), or, assuming that he was responsible, whether instead it is supposed to avert liability to conviction and punishment, by supplying an excuse. It may be a plausible proposal in one case, and implausible in the other.

For example, if we are concerned with the obligation to account for conduct, then the ability to understand and be guided by reasons is a plausible criterion. After all, accounting for conduct means identifying the reasons one was guided by at the time – ‘That's where the money is’, ‘I had a gun to my head’, and so on. So a responsible agent in this sense must be one who was capable of recognizing and being guided by reasons bearing on her situation and her conduct.

By contrast, if we are trying to define the basis of liability, voluntariness is a more plausible idea. Why so? Not because we should resurrect the early modern theory of the will, but because, as Aristotle explained, conduct counts as voluntary if it is not done through ignorance or under compulsion. And ignorance and compulsion are excuses – ways in which liability can be averted. (Compulsion here means force or severe threats.)

It is true that both kinds of excuses can be overridden. For example, as Aristotle and more recently Lord Bingham pointed out, a defendant cannot rely on the defence of duress if he voluntarily placed himself in a situation in which it was likely to occur, for example by associating with violent criminals. But in the absence of a defeating factor of this kind, liability can be averted with an excuse.

***

We began with the question why it might be wrong to hold Kenneth Parks responsible for killing his mother-in-law. But the distinction between responsibility and liability, which the analysis of criminal trials brought to light, revealed a complexity in the question, and it turned out that there wasn't a simple answer, such as ‘Because he didn't do it voluntarily’, or ‘Because he wasn't in control’.

In order to answer the question, we had to define and explain the connections between a number of distinct concepts: responsibility (the obligation to answer for or explain conduct), liability (the legitimacy of a response such as conviction, punishment or blame), responsiveness to reasons (the ability to be guided by knowledge), voluntary conduct (conduct that is not done through ignorance or under compulsion); and excuses (factors that avert liability for wrongdoing).

Philosophy consists in tracing the links in the web of concepts on which our thinking and reasoning depend, in this way. And we have only begun to discover the roots of responsibility in this little piece, since we introduced a number of difficult concepts that need explanation in their turn, such as blame, knowledge, and wrongdoing; and we ignored some equally difficult concepts, such as intention and control.

But even this brief exercise shows something about philosophical enquiry: the deepest problems are simple and familiar, and the answers are not difficult or obscure. But they depend on explaining and clarifying the main concepts or ideas we use in a domain of thought, and in discarding or modifying existing concepts, when they give rise to paradoxes or embody confusion – such as the early modern concept of a ‘will’.

In some parts of philosophy, such as ethics or philosophy of mind, the domain of thought is one we all inhabit, simply in virtue of being mature, socialized human beings; whereas in other parts of philosophy, such as philosophy of law or philosophy of physics, it is not. When it succeeds, this kind of philosophy makes us self-conscious thinkers, aware of the structure of our own systems of concepts and ideas.

It isn't immediately obvious that the question whether a man who killed an innocent woman should be acquitted can turn on a philosophical investigation of this kind. But we often need to clarify or refine our concepts in order to solve difficult intellectual problems, because concepts are the instruments of thought. We use them like surgical instruments, for dissecting; like scientific instruments, for observing and measuring; and like musical instruments, for inventing and imagining. All at once.