Martin David Kelly: Speaking Through Time

After a long hiatus, I’m happy to announce that Legal-Phi is back. Our first new interview is with Dr Martin David Kelly. Martin completed his PhD in Law at the University of Edinburgh in 2021 and is now a Lecturer in Legal Theory there. His research focuses on the relationship between law and language, and especially on theories of legal content.

Hi Martin, welcome to Legal-Phi. To begin, can you tell us a bit about your trajectory? How did you end up in law? And what brought you to legal philosophy?

Martin: Thank you, Lucas! I certainly can, though the story is long and the route was circuitous. I went to university to do sciences (I wanted to be a physicist). The Natural Sciences degree at Cambridge makes you study a wide range of subjects, initially, and I chose History & Philosophy of Science as one of my second-year options. This turned out to be a life-changing decision, though its effects didn’t fully emerge for many years. I was taught philosophy by the late, great, Peter Lipton: whose lectures were brilliant and inspiring. But I was on rails: and I chose to do chemistry, in my final year, without really seriously considering whether to continue with HPS.

I soon decided that I didn’t want a career in science and I started looking at other options. The top law firms were competing for the best graduates and offering to fund students through law school (in return for you training as a lawyer with them). This seemed a great way to take on a new challenge while remaining a student for a couple more years. I accepted a contract (with Bates Wells) and I went to law school in York.

I enjoyed learning law, though I continued to read lots of philosophy. After law school, I moved to London and completed the training contract (which involved working in several areas of legal practice). On qualification, I moved to Farrer & Co — who are famous for being the lawyers to the Queen and the Royal Family — to work in their International Private Client team. I developed an interest in revenue law; and, to learn more about it, I completed a part-time LLM at Kings College, London. I found VAT law particularly interesting, and this led me to move to PwC (to specialise in VAT law). This also enabled me, in time, to relocate from London to Edinburgh — in order to be nearer my mum (for whom I had increasing caring duties).

Throughout my decade in legal practice I continued to read philosophy when I could. But I didn’t really know what I was doing: what to read or how to read it. I reached a point where — if I was going to scratch this philosophical itch properly — I had to learn it formally. So I left legal practice to do a Masters in philosophy at Edinburgh University. Initially, I was most interested in epistemology, but one of my options — David Levy’s course ‘The Philosophy of Wittgenstein’ — blew my mind. I was fascinated by the idea that philosophical problems are, fundamentally, problems of language. Under Dr Levy’s supervision, I wrote my Masters thesis on an aspect of Wittgenstein’s mature work. I then started thinking about how my improved understanding of language could be applied to law. And so I embarked on a PhD at Edinburgh.

In your doctoral thesis The Loquacious Legislature: are statutes ‘always speaking’? you address what you call the “temporal issue” in the meaning of laws. What is the issue?

Martin: The temporal issue is about whether the meaning of an utterance can change over time. (By ‘utterance’, I mean any use of language — including written uses.) There are two broad types of approach to the temporal issue. One is historical: meaning is determined when the utterance was originally made (this is known, especially in constitutional theory, as ‘originalism’). The second treats an utterance as if it had been made recently, and thus gives that utterance its current meaning. I call this second type of approach ‘currentism’ (in constitutional theory, it tends to be known as ‘living constitutionalism’). The temporal issue is most keenly debated for written constitutions (and it is especially contentious in the US). However, it arises not only for almost all legal provisions (including those in private law instruments, such as contracts, wills, and trusts) but also for all utterances — including non-legal ones — that are ‘always speaking’.

What is an ‘always speaking’ utterance?

Martin: A ‘No exit’ sign doesn’t just speak to those who happened to be present when it was installed: it speaks to everyone who considers leaving that way (even if many years later).  It is always speaking the words ‘No exit’, and so they can come to apply to people anew (on an ongoing basis). Of course, this is a metaphor: the sign is not literally always speaking. But imagine a building frequented by the visually impaired where, as well as a written ‘No exit’ sign at the relevant door, there is a sound system that plays an audio recording of someone saying ‘No exit’. To be effective, in instructing people (on an ongoing basis) not to leave that way, that recording would have to be played on a continuous loop. The sound system would need to be always speaking the words ‘No exit’.  

We are surrounded by ‘always speaking’ utterances. A ‘Jesus loves you’ billboard speaks to everyone who passes it, not just to those who happened to watch it being posted. The same is true of virtually all road signs, etc. And, as I show in my thesis, it is also true of most laws: if they are to achieve their purposes, they need to apply to people anew (on an ongoing basis). That is, they must be treated as if they are ‘always speaking’.

What is the main connection between ‘always speaking’ utterances and the temporal issue?

Martin: The temporal issue only really arises for ‘always speaking’ utterances. Most of our utterances ‘speak’ only once: at the time when they are made. And so they are inevitably to be understood historically: to be given their original meaning. Scalia and Garner gave a vivid example of this: Queen Anne saying of the newly-built St Paul’s Cathedral, in 1711, that it is “awful, artificial, and amusing”. To understand what she meant by this, we must take ourselves back to 1711 — when those words meant awe-inspiring, artistic, and thought-provoking. If we gave Queen Anne’s words their current (2022) meaning, we would misunderstand her: we would take her to be criticising the Cathedral, when she was actually praising it.

Scalia and Garner use this example as part of their argument for originalism in statutory interpretation. But there are two problems with doing so. The first is that this is not an ‘always speaking’ utterance: Queen Anne did not intend that it should apply on an ongoing basis (unlike, say, a standing permission or instruction). So the temporal issue doesn’t really arise for it: utterances that ‘speak’ only once are inevitably intended to be understood historically. So this example (and many others like it) doesn’t really tell us how to resolve the temporal issue.

When we analyse how ‘always speaking’ utterances are intended to be understood over time, a much more complex picture emerges. In particular, we need to distinguish different notions of meaning. And this leads us to the second problem with Scalia and Garner’s argument.

What are the notions of meaning that we need to distinguish?

Martin: A common response is to distinguish meaning from application, and to say that application may change over time but that meaning cannot change. The application of an ‘always speaking’ utterance — the people and circumstances that it applies to — must be mobile: it must be able to change over time. Otherwise it could not fulfil its function: of influencing conduct on an ongoing basis. But — under this fairly standard response — the meaning of an ‘always speaking’ utterance is static: it is fixed at the time when it was originally made.

In my thesis, I illustrate this with a sign on a woodland trail that says ‘You must wear orange during hunting season’. This is intended to make walkers visible to the hunters: to reduce the risk of their being accidentally shot. It is clearly an ‘always speaking’ utterance: it is intended to apply to people anew, not just those who may have been present when it was originally posted. Otherwise, it couldn’t perform its function: of protecting walkers from injury on an ongoing basis.

Now imagine that, some time after this sign was posted, the phrase ‘hunting season’ acquires a new meaning. It comes to mean That part of the year during which singles tend to seek a settled romantic relationship (this is actually called ‘cuffing season’). So the phrase ‘hunting season’ has become ambiguous: it now has two meanings. But it seems that, to understand what is meant by ‘hunting season’ as it is appears on this sign, we must read it with its original meaning. And this would be true even if we completely stopped using the phrase ‘hunting season’ to mean hunting animals (imagine that, instead, we now only call this ‘game season’). Even in that scenario, this sign would still be about hunting animals — not about hunting for a romantic partner. So we must treat its meaning as being static.

Do you think this meaning/application response to the temporal issue works?

Martin: No. It doesn’t work because ‘meaning’ conflates two notions that differ in their mobility over time. To see this, think about what would happen if the dates of hunting season change. Let’s say that, when the hunting sign was first posted, hunting season began on 7 October each year and ended on 12 April the following year; but those dates have changed to 11 October and 17 April. Now imagine that you go for a walk in those woods on 15 April. Do you need to wear orange? The answer is clearly ‘yes’: we are intended to understand ‘hunting season’ according to the current dates, not the original dates. This just follows from the purpose of the sign: to make you more visible at times when you are actually at risk of being shot, not when you would have been at risk of being shot. To understand the sign in any other way would just be absurd.

This shows that there are three relevant notions of meaning. We must distinguish (linguistic) meaning not just from application but also from what philosophers call ‘content’. We can think of content as application-conditions: the criteria (or set of conditions) that determine whether or not an utterance applies to a situation. We are familiar with thinking of adjudication as a two-stage process: first we find the relevant facts and then we determine whether those facts fall within the legal norm. The content of an utterance is what tells us, for a given set of facts, whether the utterance applies to those facts.

In our example, the content of ‘hunting season’ tells us, given the relevant facts (i.e. the date of our walk), whether it applies: whether we are required to wear orange. And so this example shows that the content of ‘hunting season’ (as used on this sign) is intended to be mobile, even though its linguistic meaning is intended to be static. (In the thesis, I use a technical term — ‘character’ — for linguistic meaning.) So, by lumping together both content and linguistic meaning (character) into a single notion — ‘meaning’ — we miss the fact that content and character often (in fact, usually) differ in whether they are intended to change over time. We miss that content may be mobile even if linguistic meaning (character) is static.

Can you give a legal example of this?

Martin: Yes, there are lots: here’s one. In 1969, the UK Parliament first enacted legislation that allows divorce if your spouse’s behaviour is such that you “cannot reasonably be expected to live with” them. Now the linguistic meaning (character) of ‘reasonably’ has not changed since 1969: ‘reasonable’ meant (and still means) in accordance with reason. Even if this had changed, we would read this provision with its original linguistic meaning. We also use the word ‘reasonable’ to mean inexpensive. But, even if this became the sole meaning of ‘reasonable’, the 1969 Act would still speak “reasonably” with its original character. It would not allow divorce on the basis that you cannot inexpensively be expected to live with your spouse. So the linguistic meaning (character) of ‘reasonably’ — as it is used in the 1969 Act — is clearly intended to be static.

But what counts as being reasonable marital behaviour has changed, out of all recognition, since 1969. In particular, a wife in 1969 was expected to tolerate behaviour from her husband that she would not be expected to tolerate in 2022. For example, the UK High Court held — in Priday (1970) — that a husband was “not in any sense culpable” when he used force against his wife in attempting to have sexual intercourse with her: a decision that the UK Supreme Court in Owens (2018) described as “inconceivable” today. The ‘reasonable behaviour’ divorce rule is still in force today and, when we apply it, we judge a spouse’s behaviour by current standards, not by the standards of yesteryear. That is, we treat the word “reasonably” — as used in this legislation — as having a mobile content. If we had to treat its content as static, then we would have to evaluate whether a wife could divorce her husband in 2022 according to whether it would have been reasonable in 1969 to expect her to live with him. But that would just be absurd.

So are all legal provisions intended to have a mobile content?

Martin: No, for two reasons. First, some legal provisions are not ‘always speaking’: they are not intended to apply on an ongoing basis. Instead, they are intended to make a one-off change in the legal status of a person (or a thing) and, once they’ve done so, their effects are exhausted. For example, it used to be the law (in the UK) that a divorce could only be effected by Act of Parliament. The provisions of such Divorce Acts ‘spoke’ only once — when they came into force — and their legal effects were exhausted at that time. Similarly, the provisions of Acts of Attainder — which convict a named person of a specified crime — are not ‘always speaking’. A more mundane set of examples is those statutory provisions that amend other statutory provisions. Once they have come into force — and the specified amendment has been made — their effects are exhausted (though the amended provision may itself be ‘always speaking’).

The second reason is that, in some ‘always speaking’ legal provisions, a key term may be intended to have a static content. This is relatively rare: in the thesis, I analyse all the reported UK cases that I could find in which a judge decided that a legislative term has a static content. In some such cases, the judges appeal to what I call an ‘original-content justification’: a special reason that, exceptionally, justifies the court in giving a term its original content. These cases often involve provisions which create serious criminal offences and where, if the content of a key term is mobile, the scope of the offence would have expanded. So considerations of fair notice might justify ascribing to the original enacting Parliament an intention that the term should, exceptionally, have a static content.

An intention that the content of term should be static may also be inferred from other features of the context. This is also true of some non-legal utterances that are ‘always speaking’. A good example emerged from a Twitter exchange with Barry Smith (who is a leading philosopher of language but also an expert on wine). Barry said that the old adage ‘serve red wine at room temperature’ is now problematic: because the typical temperature of our rooms has changed. When this instruction was originally conceived, room temperatures were typically around 15-16oC (60oF): which is the optimal temperature for serving most red wines. But typical room temperatures have gradually increased and they are now around 20-21oC (70oF). This has misled us into serving red wines too warm: because we’re wrongly giving ‘room temperature’ its current content — we’ve missed that its content is intended to be static.

A lot seems to turn on this distinction between linguistic meaning (character) and content. Can you say more about it?

Martin: I give a full account of this in my thesis, but the distinction is fairly intuitive. When we use language to say something we express a content. As I said earlier, the content of an utterance is a set of application-conditions. We can summarise this in a simple rule: content + facts = application. But most of our terms are context-sensitive: we can use them to express different contents in different contexts. The character of a term captures its context-sensitivity: the myriad ways in which it can be used, in various contexts, to say different things. So we can also summarise this relationship in a simple rule: character + context = content.

Can you give an example of this context-sensitivity?

Martin: Recently, my wife Rachel wanted to take a photo of our dog Preston — but my feet were in the way. So she said ‘move your feet’. Now if I had just wiggled my feet, she would not have been impressed. Clearly, she meant that I should move my feet out of the line of sight of the camera. But I only understood this from the context: the words ‘move your feet’ themselves don’t tell me how I am to move my feet. And if the context is different then what counts as moving my feet may be different.

If I’m vacuuming the rug and Rachel is standing on it and I say ‘move your feet’ then, clearly, I mean move them off the rug. If we’re dancing at a wedding and I’m doing a lazy upper-body dance and Rachel says ‘move your feet’ then she obviously means move them in a dance-like way. And if I’m lying in a hospital bed, having just come round from spinal surgery, and Rachel says ‘move your feet’ then — in that context — she means move them in any way that you can. To understand the character of the phrase ‘move your feet’ is to understand how we can use that phrase in these various contexts (and many more) to say different things: to express different contents. 

In your thesis, the distinction between context-sensitivity and ambiguity plays a crucial role in your solution to the temporal issue. How do you draw that distinction?

Martin: Right. So the word ‘ambiguity’ is often used loosely: to mean any kind of uncertainty in applying an utterance (including the context-sensitivity I’ve just outlined). But I use ‘ambiguity’ in its strict, linguistic sense: a word etc. is ambiguous if it has more than one linguistic meaning (character). Each of the characters of a word should be listed as a separate entry in a dictionary. For example, we mostly use the word ‘green’ to name a colour; but it also has environmental, sickness, parkland, money, and inexperience characters.

If someone utters an ambiguous word etc., we often use the context of utterance to determine which character they meant. If you say ‘‘bring me all the green things in this building”, I would need to disambiguate your instruction to be able to comply with it: to know whether I need to bring (say) all of the energy-saving lightbulbs, or all of the new interns. But this disambiguation-by-context differs from the (mere) context-sensitivity that I illustrated with the ‘move your feet’ example. Each of the different characters of an ambiguous word may itself be context-sensitive: for example, different kinds of thing count as being green (qua colour) in different ways. If you hire me to paint your car green, I would have misunderstood (in fact, breached) my contractual obligation to you if I painted its tyres green, or its dashboard, or its windscreen.

How does this help us to solve the temporal issue?

Martin: Distinguishing character from content allows us to see that the content of an ‘always speaking’ utterance may be mobile even though its character is static. And, if it is, then we are to disambiguate it by reference to the original time; but we are to resolve any context-sensitivity in it by reference to the current time. So if ‘hunting season’ on our sign originally meant hunting qua animals then it cannot later come to mean hunting for a romantic partner season simply because our use of that phrase has changed since the sign was posted. That’s because it involves an ambiguity (in the strict sense). But ‘hunting season’ (qua animals) is clearly context-sensitive: what counts as hunting season varies from place to place and, in any given place, it can change over time. If we want to know whether our proposed walk would take place during hunting season then that involves a (mere) context-sensitivity. And so, if the content is intended to be mobile, we are to resolve that context-sensitivity by reference to the current time.

This highlights the second problem with Scalia and Garner’s use of the Queen Anne example in their argument for originalism: each of the words — ‘awful’, ‘artificial’, and ‘amusing’ — is ambiguous (in the strict sense). The first problem was that Queen Anne’s utterance is not ‘always speaking’ and so it must be understood historically: both its character and the context are static, and so its content must also be static (because character + context = content). But imagine that she gave a standing instruction: “The English must always strive to be amusing”. Clearly, we must disambiguate ‘amusing’ historically: her utterance could not later come to mean The English must always strive to be funny simply because we came to use ‘amusing’ to mean funny instead of thought-provoking. But ‘amusing’ (qua thought-provoking) is context-sensitive: whether something counts as being thought-provoking can vary from place to place, and it can change over time. So the second problem with Scalia and Garner’s argument is that their example involves an ambiguity, not merely a context-sensitivity. And so it only supports treating character as static. But this is entirely consistent with content being mobile.

In your thesis, you use the character/content distinction to present a taxonomy of responses to the temporal issue. What are the two key approaches that you identify?

Martin: I use the label ‘Character Originalist’ for the approach that treats the character of an ‘always speaking’ utterance as static but treats its content as mobile. This is how they are generally (though, as the wine example shows, not exclusively) intended to be understood. I show, in Part 2 of my thesis, that the UK courts presume that legislation should be understood in a Character Originalist way (although they do not say it in those terms). So my theoretical framework explains the vast majority of the reported UK cases which have turned on the temporal issue. It also enables us to argue that if the court has treated legislative content as static (without appealing to an original-content justification) then its decision is incorrect.

I call its main rival — the approach that treats both character and content as static — a ‘Content Originalist’ approach. Almost every theorist who self-identifies as an originalist treats ‘meaning’ (which includes both character and content) as static. But their most telling arguments for that position — such as the ‘linguistic drift’ argument (illustrated by the Queen Anne example) — only support a Character Originalist approach. And Content Originalism requires giving context-sensitive legal terms (e.g. ‘reasonable’, just’, ‘safe’, ‘fair’, ‘indecent’, ‘offensive’, ‘cruel’, and ‘welfare’) their original content, which is problematic. A Character Originalist approach, on the other hand, allows legislative content to evolve but only within the constraints imposed by character. So it prevents ‘meaning’ changing so radically that it crosses an ambiguity threshold.

This sounds a bit like Balkin’s ‘Living Originalism. How does it differ from it?

Martin: Jack Balkin is the exception: he self-identifies as an originalist, but he distinguishes between ‘thin’ and ‘thick’ meaning and holds that only thin meaning is fixed historically. So his theory is close to mine in several respects. But there are two key differences. First, his thin/thick meaning distinction is rather unclear, whereas the character/content distinction is well-established in modern philosophy of language. Secondly, and more importantly, Balkin does not recognise the key role played by the ‘always speaking’ aspect of laws. This leads him to miss the constraint imposed by the current content of the law. Balkin claims that judges are constrained by the thin meaning of the Constitution but otherwise have discretion as to its content. But my account shows how judges may be constrained by both the original character and the current content of a law. So the most telling objection against Balkin’s ‘Living Originalism’ — that it inadequately constrains judges — does not apply to a Character Originalist approach.

Does your theoretical framework have any wider legal implications?

Martin: Yes, several. I’m using my framework (in a paper that I’m writing now) to solve the problem of legislative intention. But I’ve already gone on too long, and it’s not possible to explain my solution in a brief summary. I’m presenting the paper in Edinburgh in March and in Maastricht in April, and I’d be happy to send the draft paper to those who are interested.

It also has vital implications for the doctrine of precedent, which I touch on — but do not fully develop — in my thesis. Precedent operates in two different realms (which legal theorists rarely distinguish) in Anglo-American legal systems. One — which is the main focus of theories of precedent — is the common law. The other — which is relatively neglected — is the interpretation of legal texts. This is where an appellate court adopts a particular interpretation of a legal text and then a later court takes itself to be bound by that interpretation. But if a jurisdiction takes a general Character Originalist approach to legal texts then the later court will often be wrong to do so. And so precedent, in statutory (and constitutional etc.) interpretation, is actually much less important than we’ve generally assumed.

This is because a Character Originalist approach treats content as mobile: so a decision on the content of a legal text at a given time (t1) is not determinative of its content at any later time (t2). For example, the court’s decision in Stanton (1920) that a house which was overrun with rats was “reasonably fit for human habitation” would not bind a court in 2022 to make the same decision on similar facts. And the court’s decision in Gammans (1950) that a long-term unmarried couple are not capable of being members of each other’s “family” would not require a similar decision in 2022 (and, in fact, the court in Dyson (1975) reached the opposite decision on similar facts).

The character/content distinction explains the limited scope of precedents in interpreting legal texts: they are binding as to the character of the text but not as to its content. So if an appellate court has resolved an ambiguity (in the strict sense) then later courts are bound by that interpretation: they must disambiguate the text in the same way. But if the earlier court merely resolved a context-sensitivity then later courts are not bound by that interpretation. The most we could argue is that the later court must assume that the earlier court was correct about the content of the provision as at t1. That would impose some constraint on the later court: if it wishes to reach a different result on similar facts, it must explain how the context has changed since t1 such that the provision now expresses a different content at t2. But, in many cases, it will be possible to explain this: and so the court may ‘depart’ from the ‘precedent’.

What about implications for uses of language outside of law?

Martin: My thesis contributes to our understanding of language: by giving a detailed account of a type of utterance (the ‘always speaking’ utterance) that has received very little attention from philosophers. Modern pragmatics is largely focused on explaining language use in conservational contexts, and on verbal uses of language (rather than written ones). Utterances that are ‘always speaking’ — e.g. standing permissions or instructions — tend to be written (rather than spoken) and tend not to occur in conversations. There is a philosophical debate over the so-called ‘answering-machine paradox’: how can it make sense to say ‘I’m not here right now’? But, though answer machines ‘speak’ more than once, they are not always speaking (Andy Egan calls them “utterance bombs”).

But we do sometimes make ‘always speaking’ utterances in conversational contexts. Imagine that, in 1985, my grandmother gave me some advice: ‘throughout your life, treat people with respect’. I would apply this advice 40 years later, in 2025, by treating people according to what counts as being respectful in 2025 — and not, if different, what counted as being respectful back in 1985. But if she had added ‘… and never be proud’ then I would not take her to mean that I should never take pleasure in my achievements (or in those of my friends and family). The word ‘proud’ is ambiguous, in the strict sense, between (1) feeling satisfied with an achievement and (2) having an excessively high opinion of oneself — and my grandmother clearly meant it in the latter sense (i.e. pride as one of the seven deadly sins), even if the former usage is now more common. My framework explains this: I understand my grandmother’s advice as intended to have a mobile content but a static character — I take a Character Originalist approach to understanding it.  

One source of ‘always speaking’ utterances that many consider to be very important are religious texts. For example, the Ten Commandments are clearly ‘always speaking’. An illustration that I discuss in my thesis is the Fourth Commandment: “honour thy father and thy mother”. This is presumably intended to have a mobile content: it commands us to do now what currently counts as honouring our parents, and not (if different) what counted as doing so in Moses’ day. But, arguably, we’ve also taken its character to be mobile. Nowadays, it tells us to treat our parents with respect; but the ancient near-Eastern term that was translated into English as “honour” involves conducting oneself in a way that reflects honourably on one’s parents. And these can clearly come apart: one can live an outwardly honourable life despite privately disrespecting one’s parents (or vice versa). The word ‘honour’ is strictly ambiguous between these two characters; so we seem to be taking a full-blown currentist approach to this Commandment.

What was the biggest challenge during your doctorate? Any advice to current graduate students?

Martin: The biggest challenge, for me, was settling on a project that was PhD sized and shaped. The law/language relationship involves a set of deeply intertwined issues, and my various attempts to cabin them invariably broadened into a project of unmanageable size. The one that eventually became my thesis was, I think, my 6th or 7th attempt to find a topic discrete enough to say something original and interesting about within the confines of a PhD. Even then, much was left on the cutting room floor — and I still needed a sizeable word limit extension (my PhD thesis was the longest in the current institutional memory of Edinburgh Law School). The other main challenge was structural: how can I say anything when, in order to say something properly, I really need to have already said everything? In other words, the problem of scaffolding.

My main advice to current PhD candidates is: don’t try to write a PhD thesis like I did! Especially given that, nowadays, a good publication or two is close to being a pre-requisite for getting a first academic job. A much better route is to write a thesis which is essentially a collection of essays on a common theme. So each chapter is potentially publishable as a journal article; and, indeed, you might submit some of them for publication before finalising your thesis. Your PhD thesis doesn’t have to be — and almost certainly won’t be — your magnum opus.

What are your other interests?


Preston and Martin

My main passion is dogs. I absolutely love dogs and, in 2015, I was finally able to get my own wee dog. He’s called Preston and he’s my best ever friend. In the little spare time that I get, I like to walk Preston. We’re fortunate in that, where we live in Edinburgh, we’re very close to some nice parks, woodlands, hills, and beaches. So we go on lots of lovely walks together.

I’ve also recently started playing chess again. I played lots as a schoolboy, to a relatively high level, but I haven’t played properly in about three decades. On a slightly more esoteric note, I’m a member of the Cloud Appreciation Society. This requires me to “seek to persuade all who’ll listen of the wonder and beauty of clouds” (which you’ll now recognise as an ‘always speaking’ utterance). So consider yourselves proselytised!

What books or papers influenced your philosophical views the most?

Martin: The book that has had the greatest influence on my philosophical views, by far, is Wittgenstein’s Philosophical Investigations. But I wouldn’t recommend reading it without first immersing yourself in the issues that Wittgenstein was tackling. Far better to start with a good introduction to Wittgenstein’s philosophy. Or, if you want to read a primary source, start with his  Blue Book (though his views changed a little between the Blue Book and Philosophical Investigations).

To close the interview, what is your favourite book outside law and philosophy?

Martin: I read so much for work that I don’t read very much for pleasure now. Most of the fiction I read is fairly low-brow: to help me to switch off my mind at bedtime. My favourite work of literature is probably Gulliver’s Travels. When I was a boy, my favourite novel was Frank Herbert’s Dune. (I’ve seen the new movie and, while it’s a major improvement on the David Lynch version, I still think that the book is unfilmable.) I’m also very fond of Douglas Adams’ books: the Hitchhiker’s trilogy contains much wisdom, as well as being wickedly funny.