1 Introduction

In his wonderfully thought-provoking article “Backwards Causation in Social Institutions”, Silver (2022) argues that if we take so-called “institutional facts” seriously, then we are led to conclude that backwards causation is not only possible, but commonplace. As such, Silver is not the first to argue that backwards causation is logically possible (others who have argued this include Dummett, 1954, 1964; Lewis, 1976; Roache, 2009; Tooley, 1999), nor is he the first to argue that backwards causation is actual (Dowe, 1997; Price, 1997; and Corry, 2015, for example, have suggested that backwards causation may account for the weirdness found in quantum mechanics), and nor is he the first to argue that backwards causation might occur amongst institutional facts (Barlassina & Del Prete, 2015; and Torrengo, 2018 suggest this also). What sets Silver apart is his claim that backwards causation is common in everyday contexts. I am not convinced. I will argue that although Silver is correct that backwards causation is a conceptual possibility, this possibility is not the best interpretation of what is going on in the cases he considers. Although my conclusion regarding Silver’s position is negative, I use his cases to develop positive theses about the nature of backwards causation and the nature of institutional facts.

2 Institutional Facts

Put briefly, an institutional fact is a fact that is made true by an appropriate institution declaring it to be so. For example, the fact that Charles is king of England, the fact that the Danube is the border between Romania and Bulgaria, and the fact that I am a lecturer in philosophy are all institutional facts.

The idea of institutional facts has its roots in Anscombe (1958) and was most influentially developed by Searle (1975, 1995, 2010). Searle argues that institutional reality is created by “Status Function Declarations”. As Searle says:

We can create boundaries, kings, and corporations by saying something equivalent to “Let this be a boundary!” “Let the oldest son be the king!” “Let there be a corporation!” […] in the nonlinguistic Status Function Declaration, we do more than represent: we create. (Searle, 2010: pp. 100, 114)

Of course, I cannot become king simply by declaring myself so. A successful status function declaration involves a “collective intentional” imposition of function on an entity (Searle, 1995, p. 41). Searle saw this imposition as involving systems of constitutive rules that are collectively agreed upon, and he uses the word “institution” to refer to such a system. Hence the term “institutional fact”.

Institutional facts have real consequences. The fact that Charles is king means that he can do things that I cannot. For example, as king, Charles can appoint and dismiss ministers, sign a bill of parliament into law, and even “claim ownership of all unmarked mute swans swimming in open waters throughout the country” (Royal.uk., 2018). So, when Charles became king of England this event made a real difference to the world. These thoughts motivate a realist attitude towards institutional facts according to which institutional facts are a real part of the world and can play a causal role in the world, both as causes and effects.Footnote 1 Searle was a realist about institutional facts, and realism is now the mainstream view. For the rest of this paper, I will be assuming such realism.

3 The Argument for Backwards Causation

If we accept realism about institutional facts, the route to backwards causation is straightforward. The basic argument can be reconstructed as follows:

  1. 1.

    Institutional facts are created when an appropriate institution declares them to be the case.

  2. 2.

    Institutional facts are real facts about the world.

  3. 3.

    Institutions can, and often do, declare facts about the past to be the case.

Therefore

  1. 4.

    Events at one time (institutional declarations, or events that cause such declarations) can, and often do, cause real changes in the world at an earlier time.

Premises 1 and 2 encapsulate realism about institutional facts. To motivate Premise 3, Silver (2022, p. 3) offers three putative examples of backwards causation of this type, which I repeat here in full:

Case 1: Retroactive Enrollment

Kevin made a mistake in graduate school. One semester, say in year X, he registered for PHIL755 rather than PHIL756. Both classes are open-ended, dissertation writing classes that one registers in as a formality, but he had already ‘taken’ PHIL755, and PHIL756 was next in the sequence. The mistake was not discovered until a year later (X + 1) when it was time to make sure his affairs were in order to graduate. After reviewing his transcript, recognizing his error, and becoming hysterical, he shared his mistake with the department administrator. Graciously, she assumed the bureaucratic headache of fixing it. Calls were made, emails exchanged, and eventually Kevin was retroactively enrolled in the course. By the end of the ordeal, actions taken by Kevin’s administrator in year X + 1 caused him to have technically been registered for and taken PHIL756 in year X.

Case 2: Forced Forfeit

The Tigers are a wrestling team that competes for and wins the National Championship in year X. Unfortunately, however, they cheated by using rigged weight scales. When this is uncovered in year X + 1, the league commissioner decides to strip the team of their title in the wake of the scandal. It is written down that the team forfeited their last match, and the title went to their final opponents. By the end of the ordeal, the announcement of the commissioner in year X + 1 causes the team to forfeit the match that occurred in year X.

Case 3: Annulment

Billy and Suzy decide that after being friends for many years they must be meant for each other, so they get married (say, on day X). Unfortunately, they quickly realize that they should not be married; they’re both just too competitive. As it happens, they are also both Catholic, and both hope to remarry someone else someday. Luckily, they petition the church successfully and are granted an annulment of their marriage (on day X + 100). In the eyes of God, the marriage never occurred. So, actions of the diocese on day X + 100 cause the erasure of the marriage of Billy and Suzy on day X.

Since these cases are supposed to represent common scenarios in which institutions declare facts about the past, they are supposed to support the claim that institutions can, and often do, make such declarations.

4 Assessing the Cases

We might wonder why Silver presents three cases here, rather than just one. We know that Silver is arguing for more than the mere possibility of backwards causation, and so I suggest that there are two reasons why he might present three cases. The first possibility is that the three cases are supposed to represent three separate ways that we might get backwards causation amongst institutional facts. A second possibility is that presenting multiple cases is intended to support the claim that backwards causation is common. If we look at the cases closely, however, we will see that they do not achieve either of these purposes as well as might appear at first sight. To see this, let us look at them in reverse order.

As Silver notes, annulment in the Catholic Church does not actually work the way he describes in Case 3. To be granted an annulment, one must show that there is some reason why the marriage was never legitimate. For example, if Billy could show that he had never consented to be married to Suzy then the church might conclude that Billy and Suzy had never been legitimately married and so grant an annulment. The important point here is that the cause of the illegitimacy is Billy’s lack of consent on day X, not the Church’s granting of an annulment on day X + 100. Thus, there is no backwards causation.

Consider the case of the Alabama farmer Bill Wilson, who, in 1915, was convicted and sentenced to life imprisonment for the murder of his wife and child. Bill’s conviction was overturned in 1918 when his wife and child were discovered alive and well in Indiana (Wikipedia contributors, 2022). In this case, the overturning of his conviction in 1918 surely did not cause him to be innocent of murder in 1915. Rather, it was always the case that he was innocent of murder, and the pardon was simply a recognition of this fact, a recognition that was made possible by the discovery of his wife and child. Similarly, according to the Catholic Church, an annulment does not cause it to be the case that a marriage was never legitimate. Rather, an annulment is a recognition of the fact that the marriage was never legitimate in the first place.

This objection to Annulment hinges on the notion of legitimacy. I suggested that the marriage never actually happened because it did not satisfy the rules of the appropriate institution. This objection makes sense given Searle’s concept of an institution as constituted by a system of rules. However, Dörge and Holweger have suggested that the definition of institutional facts in terms of such institutions is too narrow. They argue that the way the concept of an institutional fact is actually used by Searle and others suggests that they take an institutional fact to simply be a fact that exists by virtue of a collective recognition of its existence (Dörge & Holweger, 2021, p. 4955). Given this broader understanding of institutional facts, the Annulment case is more interesting. It raises difficult questions—that I will not consider here—about what happens to institutional facts when a community of people believe that the rules of an institution have been satisfied where in fact they have not. What is interesting for our purposes is that a broader account of institutional facts might give us a way to save Silver’s interpretation of Annulment. I will not pursue this possibility further, however, for two reasons: (1) It relies on a separate debate about how broadly we should understand institutional declarations. (2) If we can save Silver’s interpretation in this way, it will have the same structure as the case of Retroactive Enrollment considered below.

We might think that Forced Forfeit faces the same objection as Annulment. Since the Tigers broke the rules by using rigged weight scales, we could argue, they never were legitimate winners of the national championships. So, when the Tigers were awarded the title in year X, this was simply a mistake—an understandable mistake because the league did not know about the rule breaking at the time. And when the league commissioner stripped the Tigers of their title in year X + 1, the commissioner is not retroactively causing the Tigers to have not been the winners in year X. The commissioner is simply recognising that the Tigers never were the legitimate winners.

As I noted above, however, the objection to Annulment hinges on what is legitimate, and the same is true in the case of Forced Forfeit. In particular, the objection only bites if the rules stipulate that the legitimacy of the win depends on some facts other than the declaration of the institution (such as the fact that the Tigers used rigged scales). In some sports, however, the legitimacy of a win will depend on what is declared by the referee at the time, regardless of what else actually occurred. A nice example of this is given by Barlassina and Del Prete (2015): in the 1986 FIFA world cup match between Argentina and England, Diago Maradona scored a goal using is hand, in violation of the rules of the game. Neither referee saw the violation, however, and the goal was declared legitimate. Crucially, the goal, and hence Argentina’s 2 − 1 win, is still regarded as legitimate by FIFA, even though there is compelling evidence that the ball came off Maradona’s hand.

If the rules governing the competition in which the Tigers were playing were like the rules of the FIFA world cup, then the legitimacy of their title in year X will depend on what the umpire declared at the time, and hence the Tigers would have legitimately won even if it was later found that they broke the rules, and the case will not be like that of Annulment. So, if the league commissioner later declares that the Tigers did not win, then we might be able to view this as backwards causation as Silver intends, rather than discovering a mistake. Note, however, that if the rules of the sport state that a win is legitimate if declared to be so by the umpire at the time, then it is hard to see why the rules would also allow for a later review of this win. Rules that did allow this would seem to involve a kind of internal contradiction (we will revisit this contradiction later). Presumably, this is why FIFA have not reviewed Maradona’s goal. Footnote 2

Let us say that a fact is purely institutional iff the relevant institution has the discretion to make declarations about this fact regardless of what else may be the case. So, for example, Billy and Suzy’s marriage (or otherwise) is not purely institutional, according to the Catholic church, because a declaration about whether someone is married will only be legitimate if other facts obtain (such as consent). Whether or not Maradona scored a goal, on the other hand, is purely institutional since the referee’s declaration that it was a goal is sufficient for it to count as a goal, regardless of other facts (such as the fact that the ball came off Maradona’s hand).

Our considerations of Annulment and Forced Forfeit suggest that potential cases of backwards causation of institutional facts will require that the facts involved are purely institutional. Annulment in the Catholic church does not involve purely institutional facts and so the example does not support the claim that backwards causation is actual. Some sports do involve purely institutional facts, and so cases like that in Forced Forfeit might occur. However, sports with purely institutional facts that also have reviews of those facts are likely to be uncommon since they would seem to involve an internal contradiction (and if the rules truly are contradictory, then I am not sure how we should interpret the legitimacy of an institutional declaration). So even if Annulment and Forced Forfeit support the claim that backwards causation is possible, they do not support the claim that it is actual and common, and hence the cases fail to support Premise 3 of the argument for backwards causation.

Retroactive Enrollment is more promising. In this case the administrator in year X + 1 does not discover some earlier fact which made it true that Kevin was enrolled in PHIL756 all along. Instead, the administrator’s actions in year X + 1 seem to cause Kevin’s enrollment in PHIL756. So, if it is true that, in year X, Kevin was enrolled PHIL756, then this does look like backwards causation.

As I see it, the crucial characteristics of Retroactive Enrollment are the following:

  1. A.

    Kevin’s university has, by its own rules, authority and discretion regarding enrollment, both past and present (so that facts about enrollment, past and present, are purely institutional).

  2. B.

    In year X + 1, the administrator made a phone call which resulted in the university declaring that Kevin was enrolled in PHIL756 for year X.

According to the realist about institutional facts, such facts can be made true by appropriate declarations and so it would seem to follow from A and B that Kevin was, indeed enrolled in PHIL756 at time X. Since the declaration upon which this fact depends was caused by the administrator’s phone call at time X + 1, it looks like an event at time X was caused by an event at time X + 1, and so it looks like we have backwards causation.

One might, at this point, object that after the administrator’s phone call the university has simply decided to pretend that Kevin was enrolled in PHIL756 in year X. Silver agrees that this is something the university could do, but he argues that the university could also declare that Kevin was actually enrolled in PHIL756 in year X. Since the university has the power to make facts about enrollment true through such declarations, Kevin’s enrollment will be legitimate, not mere pretence. This is the situation we are to imagine here.

5 Changing the Past vs. Influencing the Past

Silver suggests that the best interpretation of Retroactive Enrollment is as follows:

Originally, in year X, there is an event (a) ‘Kevin’s taking PHIL755’… In year X + 1, we saw that there is a second event (b) ‘The administrator’s phone call’. The suggestion is that this event is a cause of a third event, (c) ‘Kevin’s taking PHIL756’, which occurs back in year X. Once the administrator has made the phone call, there is no more event (a) of ‘Kevin’s taking PHIL755’ in year X. (Silver, 2022, p. 6)

So, Silver thinks the backwards causation involved in Retroactive Enrollment works by changing the past. Before the phone call the past is one way, after the phone call the past is some other way.

Now, there is a long list of philosophers who have argued that talk of “changing the past” is incoherent. As Swinburn puts it, “To change something is to make it different from what it was at another temporal instant. I cannot make a thing at t1 different from what it is, was, or will be at t1— for this would imply that it both did and did not have some property at t1” (Swinburne, 1968, p. 161).

Swinburn may be overstating the problem. Put more carefully, the worry is that changing the past is inconsistent with the claim that the past consists of a single consistent history. We can avoid the problem if we deny this claim. In this vein, a number of authors have suggested that we can make sense of “changing the past” if we inflate our ontology. So, for example, Meiland (1974), argues that we could change the past if time were two-dimensional, Goddu (2003) and van Inwagen (2010) add a higher-order temporal dimension (“hyper-time”), Abbruzzese (2001) suggests a multiverse will do the trick, and Loss (2015) suggests a single time-line in which history repeats itself with minor variations.Footnote 3

Silver’s conception of backwards causation, therefore, requires some serious metaphysical commitments, and I do not believe that enough has been done to justify these commitments. It would seem odd, to say the least, if considerations of our social practices alone were sufficient reason to believe in one of these inflated metaphysical views. What is more, we would need some story about how, and why, it is that institutional declarations interact with these new metaphysical degrees of freedom. Without an independent motivation for a metaphysical model that allows changing the past, and without a story about how institutional declarations interact with this model, I, for one, am going to get off the boat at this point.

At this stage, then, it might seem that we have an argument against the possibility of backwards causation in general, not just among institutional facts:

  1. 5.

    Backwards causation involves changing the past.

  2. 6.

    Changing the past is either impossible or requires an unlikely metaphysics.

Therefore

  1. 7.

    Backwards causation is either impossible or unlikely.

Although there are versions of this argument in the literature (Flew, 1954, for example), it is too soon to conclude that backwards causation is impossible, because Premise 5, above, is false.

Forwards causation makes the future what it is, rather than changing the future from one state to another. Consider, for example, a situation in which Suzy throws a brick through a window at time t, causing the window to break. In this situation, Suzy’s throw changes the state of the window from being whole before t to being broken after t. What Suzy’s throw does not do is change the state of the future from containing a whole window after t to containing a broken window after t. Thus, the only sense in which we “change the future” is that we make it different to what it would have been if we had not acted. Following Horwich (1975, pp. 435–436) let us say that we don’t change the future, we influence it to make it what it will be.

If forwards causation works by influencing the future rather than changing it, backwards causation should work the same way.Footnote 4 If the administrator’s phone call caused Kevin to be enrolled in PHIL756 for year X, then it was always true that Kevin was enrolled in PHIL756 for year X. The phone call does not change the past from not containing such an enrollment in year X to containing this enrollment in year X. The past always contained Kevin’s enrollment, but this enrollment was made the case by the influence of the future phone call. If this is the correct story, we have backwards causation with no changing the past and so no incoherence and no requirement of an inflated metaphysics.

Now, we might wonder why Kevin would have got into such a panic if he was enrolled in PHIL756 all along. And we might note that if he had not panicked and talked with the administrator then he would not have been retroactively enrolled in PHIL756. Something like a paradox seems to threaten.

But there is no paradox. The reason Kevin got into a panic is that he did not know that he was enrolled in PHIL756. This is understandable, since Kevin had filled out the paperwork to enroll in PHIL755 not PHIL756, and all the records at the time said (falsely) that he was enrolled in PHIL755. The events that caused him to be enrolled in PHIL756 had not happened yet, so it is not surprising that he did not know about these events.

6 Conflicting Declarations

Let us pause for a moment to take stock. I have argued that backwards causation may not be as common as Silver’s cases suggest, but I have rejected the claim that backwards causation is impossible or unlikely, on the grounds that it requires changing the past. So far, then, it would seem that there is still room for backwards causation in general, and for backwards causation of institutional facts in circumstances like that described in Retroactive Enrollment.

However, there is a problem with backwards causation of institutional facts that is not shared by backwards causation in general. We have seen that the cases Silver considers only raise the possibility of backwards causation if the institutional facts in question are purely institutional. In other words, we only get the possibility of backwards causation if the declaration of the fact in question by the appropriate institution is sufficient for that fact being the case. But if a declaration is sufficient to make a fact the case, then we run into problems if there can be multiple, conflicting, declarations about the same situation at the same time.

Consider Retroactive Enrollment again. In year X it was declared that Kevin was enrolled in PHIL755 for year X. Then, in year X + 1, there is a declaration that Kevin was, in fact, enrolled in PHIL756 for year X, not PHIL755. If each of these declarations is sufficient to bring about their associated fact, then we are led to an apparent contradiction: it is both true and false that Kevin was enrolled in PHIL755 for year X. And if institutional facts are real facts, then this contradiction is not merely apparent.Footnote 5

Note that we cannot avoid this problem by imagining a case where there is no explicit declaration about Kevin’s enrollment for year X before the phone call is made in year X + 1. Suppose, for example, that instead of enrolling in the wrong class in year X, Kevin had forgotten to enroll in any class. In this situation there would still be an implicit declaration in year X that Kevin is not enrolled in PHIL756 for year X. For students are enrolled in a class if, and only if¸ the university declares them to be so enrolled. This is why, for example, I am not currently enrolled in all the classes on offer at the University of Oxford. So, the university’s explicit declaration in year X + 1 about Kevin’s enrollment in year X will contradict the implicit declaration in year X. This problem generalises to all institutional facts, since an institutional fact is true if, and only if, it is declared to be so by an appropriate institution.

It is most clear that conflicting declarations lead to contradiction in Retroactive Enrollment if we assume the simple ontology of a single consistent history. For the conflicting declarations disagree about what this history is. We might hope we can avoid contradiction, though, if we assume one of the inflated ontologies mentioned above (presumably, this is why Silver talks of “changing the past” in this situation). Now, I argued above that we should not accept any of these inflated ontologies without good independent reasons. However, even if one does accept one of these inflated ontologies, I am not sure that they will help avoid contradiction here.

Suppose, for example, that we accept an ontology of multiple universes. In this case we might hope to avoid contradiction by claiming that Kevin is enrolled in PHIL755 for year X in one universe, and not enrolled in PHIL755 for year X in another. But this strategy will not work. Consider the universe in which the administrator makes the phone call in year X + 1 that leads to a declaration that Kevin is enrolled in PHIL756 not PHIL755 for year X. This universe contains an earlier declaration that Kevin was enrolled in PHIL755 for year X, for this is why Kevin panicked and asked the administrator to make the phone call. Thus, the conflicting declarations cannot be separated into different universes, and so neither can the contradictory institutional facts.

Or suppose that we accept a two-dimensional view of time. We might then hope to separate the conflicting facts along the second temporal direction. Let us label the two temporal dimensions T1 and T2. There is no contradiction if Kevin is both enrolled and not enrolled in PHIL755 at T1 = X, so long as these facts differ in their second time coordinate. So, for example, it might be that Kevin is enrolled in PHIL755 for year (T1 = X, T2 = Y) and Kevin is not enrolled in PHIL755 for year (T1 = X, T2 = Z). Whether such a situation makes sense will depend on how we interpret the two time dimensions, and how institutional declarations interact with these dimensions. Without such details, I have no idea how to interpret these claims about Kevin’s enrollment. Note, however, that if this strategy is to solve the problem of multiple declarations, then it must satisfy certain constraints. First, the two-dimensional view cannot be closely analogous to the multiverse view, since the multiverse view does not solve the problem. So, for example, we cannot interpret the set of events with a constant T2 as making up a “universe”. Second, we would need to somehow rule out the possibility that the university might declare that Kevin is enrolled in PHIL755 for year (T1 = X, T2 = Y), and then declare that Kevin is not enrolled in PHIL755 for year (T1 = X, T2 = Y). Perhaps it is possible to tell a two-dimensional time story that allows us to avoid contradiction, but it will not be straightforward, and I have little faith that it would be plausible. For the remainder of this paper, then, I will investigate whether it is possible to solve the problem of conflicting declarations without requiring an inflated ontology.

Another strategy for avoiding contradiction is to insist that when we get conflicting declarations like those in Retroactive Enrollment, one of the declarations must fail. So, for example, we might argue that if the university rules grant administrators the power to retroactively enroll and unenroll students, then it is implicit in these rules that any enrollment declaration is legitimate only if it is not later countermanded. Thus, the declaration that Kevin was enrolled in PHIL755 for year X was not legitimate because it was later countermanded by the administrator.

This solution is unsatisfying for at least two reasons. First, this solution does not respect the intuitions that motivate realism about institutional facts. For, in year X, Kevin has all the powers that someone enrolled in PHIL755 has, and none of the powers that someone enrolled in PHIL756 has. For example, Kevin would be allowed to attend an event that is exclusively for students in PHIL755 and would not be allowed to attend an event for students in PHIL756. If realism about institutional facts is motivated by the real powers that these facts grant, then it seems perverse to deny that Kevin is enrolled in PHIL755 in year X.

Second, if the legitimacy of a declaration is dependent on the future, then, given our general ignorance about the future, we can never be sure that any declaration is actually legitimate, and so cannot be sure that the institutional facts are what we think they are. Such epistemic humility is not necessarily a bad thing, but it does seem to go against the idea that institutional facts are made true by the fact that the relevant people believe them to be true.

I also note that this solution may not be available if we move beyond Searle’s concept of an institution. For example, the solution will not work if we accept Dörge and Holweger’s broader notion of institutional facts. There is collective recognition in year X of Kevin’s enrollment in PHIL755 and collective recognition in year X + 1 of his non-enrollment in PHIL755. Thus, it would seem to follow from Dörge and Holweger’s theory that Kevin both is and is not enrolled in PHIL755 for year X.

7 Relational Properties

How then should we understand what is going on in cases like Retroactive Enrollment? The problem with this case is not that it may involve backwards causation, but, rather, that there seem to be two contradictory declarations regarding Kevin’s enrollment for year X and these contradictory declarations lead to contradictory facts. Given that cases like this actually occur, we need to make sense of what is going on when we have conflicting declarations, regardless of whether there is backwards causation involved. It will be helpful to note, therefore, that these cases of apparent backwards causation are not the only situations in which we get conflicting declarations.

Consider Mary and Hannah, a lesbian couple who had been living together for many years in Sydney. When Australia changed the law in 2017 to allow same-sex marriage, Mary and Hannah jumped at the chance to formalise their relationship and got married. In 2020, they moved to Singapore and have lived there since. Now, Singapore does not currently recognise same-sex marriage, so it seems that there are conflicting declarations about Mary and Hannah. Australia has declared that Mary and Hannah are married, but Singapore has declared that same-sex couples cannot be married. If declarations are sufficient for marriage (and non-marriage), then it seems like we have the contradictory conclusion that Mary and Hannah are both married and not married.

At first sight, there seems to be a straightforward way to avoid contradiction in the case of Mary and Hannah. We can simply say that they are married according to the Australian government and are not married according to the Singapore government. By shifting the ascription into an intensional context like this we avoid contradiction in the same way that there is no contradiction between “John believes aliens have visited” and “Mark believes aliens have not visited”. The problem with this solution is that it fails if we assume realism about institutional facts. According to realism, an institutional fact is true if and only if an appropriate institution declares it to be the case. In other words, it is true if and only if it is true according to the appropriate institution. So, for example, Mary and Hannah are married if and only if they are married according to Australia. Realism, therefore, shifts the descriptions back out of their intensional contexts, producing the contradictory statements “Mary and Hannah are married” and “Mary and Hannah are not married”.

Although the intensional solution above fails, the motivation behind it is on track. The thought is that we have two different institutions, and Mary and Hannah are married with respect to one and are not married with respect to the other. To make this solution work, we simply need to make sure the relationship between institution and marriage is not within an intensional context. In other words, instead of “Mary and Hannah are married according to Australia”, we should say “Mary and Hannah are married relative to Australia” where the relation in question is not intensional. We thereby avoid contradiction because there is nothing inconsistent in the claims that “Mary and Hannah are married relative to Australia” and “Mary and Hannah are not married relative to Singapore”. Realism about institutional facts does not cause a problem, because instead of “Mary and Hannah are married if, and only if, they are married according to Australia”, realism tells us that “Mary and Hannah are married relative to Australia if, and only if, they are married according to Australia”.

My suggestion, then, is that institutional facts in general are relational facts. To say that Charles is king is really to say that he is king relative to the institution of the British monarchy. To say that the Danube is the border between Romania and Bulgaria is to say that it is the border relative to some institution of international relations. And to say that I am a lecturer in philosophy is to say that I am a lecturer relative to my university.Footnote 6

Statements of institutional facts may not wear their relational status on their sleeve, but my suggestion is not too radical. After all, what sets institutional facts apart from other facts is that they rely on the existence of appropriate institutions. Thus, it should be no surprise if the appropriate institution, itself, is an important subject in any proposition involving institutional facts.

The nature of the relation involved in institutional facts will depend on the details of the correct account of such facts. So, for example, if Searle is right, then Charles is king relative to the British monarchy if, and only if, the rules of the British monarchy declare that Charles is king (if you are worried about circularity, we could replace this with “Charles is king relative to the British monarchy if, and only if, the rules of the British monarchy declare that Charles has the following rights and responsibilities…”). If Dörge and Holweger are right, on the other hand, then Charles is king relative to the British monarchy if, and only if, the people making up the institution of the British monarchy (i.e., the citizens of Britain) believe that Charles is King. For our purposes, these details are not important. All that is required to solve the problem of conflicting declarations is the claim that institutional facts are relational in some way or other.

I have argued that we can solve the problem of conflicting declarations by different institutions if we accept that institutional facts are relational. But how does this help in the cases of apparent backwards causation above? The case of Mary and Hannah involves conflicting declarations by different institutions, so we solve the problem by relativising to these different institutions. Silver’s backwards causation cases, on the other hand, involve only a single institution, and so it seems that the same strategy is unavailable.

There is a simple fix: we simply assert that institutional facts concern relations to an institution at a time.Footnote 7 Institutions are not static: the people making up the institution, the facts endorsed by the institution, and even the rules of the institution can change over time. So, it should not be surprising to find that institutional facts contain a temporal relation. Mary and Hannah are married relative to Australia today, but they are not married relative to Australia before their wedding took place in 2017.

If my suggestion is right, then statements of institutional facts are actually statements about three place relations. So, for example, if we let K stand for the relation “is king relative to”, c stand for Charles, b stand for the institution of the British monarchy, and t stand for the present moment, then “Charles is King” can be understood as asserting that K(c,b,t). Note that K(c,b,t) is not asserting that Charles is king of Britain now, but rather that Charles is declared king by Britain now.

Having laid out my strategy, let us take another look at Retroactive Enrollment. The university in year X did not regard Kevin as enrolled in PHIL756 for year X. For Kevin had not performed the actions that the university recognised as enrolling in this class, and the university records did not indicate that he was so enrolled. If you had asked the appropriate officials at the university, they would have told you that, no, Kevin was not so enrolled. And the university would not have seen any reason to award Kevin a grade for PHIL756. Let R(y,z,t) be the claim that there is an enrolled-in-PHIL756-for-year-X relation between person y, institution z, and time t, and let k stand for Kevin and u stand for his university. Then R(k,u,X) is false.

On the other hand, the University in year X + 1 does regard Kevin as enrolled in PHIL756 for year X. Appropriate enrollment procedures have been performed, the university records indicate that Kevin was enrolled, university officials will state that he was enrolled, and the university will recognise a need to award Kevin a grade for this class. So, R(k,u,X + 1) is true.

Finally, there is no contradiction between R(k,u,X) being false and R(k,u,X + 1) being true, and so we have a coherent interpretation of what is going on in the case of Retroactive Enrollment.

8 Against Backwards Causation

Is there backwards causation given my interpretation of Retroactive Enrollment? I think not, at least, not in any interesting sense. Note, first, that the relevant effect of the phone call in year X + 1 is to cause R(k,u,X + 1) to be true (so, in particular, the phone call does not affect the falsity of R(k,u,X)).

Second, the phone call affects the relationship between Kevin, the university, and year X + 1 by changing certain intrinsic properties of the university in year X + 1. To see this, without begging any questions regarding backwards causation, consider a more normal case of enrollment. Suppose that, at the beginning of year X, Lauren asks the administrator to enroll her in PHIL756 for year X. The administrator makes a phone call and as a result it is made true that Lauren has the enrolled-in-PHIL756-for-year-X relation to the university and year X. The phone call does not cause any changes to Lauren’s intrinsic properties, but it does cause changes to intrinsic properties of the university (for example, the university database lists Lauren as enrolled in PHIL756). And these changes occur after the phone call. In normal cases, then, the enrollment relation is made true by changing intrinsic properties of the university after the enrollment process has begun, and hence this is the most plausible interpretation of what is going on in Retroactive Enrollment as well.

Putting these two thoughts together, we see that the phone call in year X + 1 makes R(k,u,X + 1) true by changing certain properties of the university in year X + 1. Since the relevant properties of the university in year X + 1 were changed after the phone call, the causal arrow is pointing forwards in time and there is no backwards causation going on.

The appearance of backwards causation comes from the fact that we are talking about Kevin’s enrollment for year X (this is built into our definition of R). Further, we might argue that the university is not saying that Kevin now is enrolled in PHIL756 for year X. Rather, it is Kevin back then in year X who is declared to have been enrolled in PHIL756 for year X. To make this clear, let us replace talk of Kevin with talk of Kevin-in-year-X.Footnote 8 Now, consider the monadic predicate “is enrolled in PHIL756 for year X according to the university at time X + 1” (this is the predicate we get if we remove k from the proposition R(k,u,X + 1)). The administrator’s phone call in year X + 1 causes Kevin-in-year-X to satisfy this predicate. And this looks like a clear case of backwards causation.

Considering the monadic predicate above, we might be tempted to say that the administrator’s phone call changes the enrollment status of Kevin a year earlier.Footnote 9 But the change here is what is known as “mere Cambridge change”—a change in a monadic property of an object that is constituted by changing an intrinsic property of something else. In this case we change a monadic property of Kevin-in-year-X by changing the intrinsic properties of the university in year X + 1. Geach, who coined the term, does not regard mere Cambridge changes as real changes (Geach, 1969, p. 72), and if he is right these changes cannot be regarded as effects. But even if there is a sense in which mere Cambridge change is real, we should not regard such changes as effects in the sense relevant to our discussion, lest backwards causation turn out to be even more common than Silver suggests. For example, back in 1986 I grew taller than 168 cm and so changes in my body in 1986 caused Napoleon in 1800 to be shorter than me. I doubt anyone would regard this as backwards causation in any interesting sense.

Silver considers and rejects the objection that his putative cases of backwards causation involve mere Cambridge change. He says that, unlike mere Cambridge changes, properties are not merely ascribed to things in the past, these properties are instantiated by those things in the past. Furthermore, he continues, “the changes in question are not extrinsic or relational in the way that Cambridge changes often are” (Silver, 2022, p. 11). I agree that in the cases considered properties are not “merely ascribed to things in the past” but Silver is wrong when he says that the changes in question are not relational. If I am right, institutional facts are only true relative to an institution and a time.

Before finishing, I would like to consider one final worry that was suggested by an anonymous referee. If the changes to Kevin’s enrollment are mere Cambridge changes, and Cambridge changes are not real enough to be regarded as true effects, then doesn’t this undermine realism about institutional facts? More generally, if institutional facts are relational in the sense sketched here, then isn’t the result that institutional properties do not merely fail to backwards-cause, but they fail to cause full-stop?

Fortunately, my position does not have this consequence. I have argued that “changes” in Kevin’s enrollment are mere Cambridge changes, but this does not mean that there is anything suspicious or ineffectual about the relations involved. Consider, for example, a case in which Alex is taller than Sam in year X, but then Sam grows and becomes taller than Alex in year X + 1. In this situation Alex has changed from being taller than Sam to being shorter than Sam. This is a mere Cambridge change in Alex since it is made true by changes to the intrinsic properties of someone else. As such it does not seem right to say that Sam’s growth (directly) causes a change in Alex. But Sam’s growth does cause a change in the relation between their heights, and this relation can have real effects (it can cause Alex grief, for example). Similarly, the administrator’s phone call causes a change in Kevin-in-year-X’s relation to the university in year X + 1. Since this change is produced by changing the intrinsic properties of the university in year X + 1, it is not a “real” change in Kevin-in-year-X. But the new relation can have real effects. It can, for example, be part of the cause of the university awarding Kevin a degree. In short, institutional facts concern relations, and these relations can be perfectly real, even if the mere Cambridge changes that such relations allow are not real. Thus, denying the reality, or causal efficacy of Cambridge changes does not impugn the reality or causal efficacy of the institutional relations that makes these Cambridge changes possible.

9 Conclusion

Silver has argued that institutional declarations about the past literally change the past and thereby constitute cases of backwards causation. Many of us will not be convinced by this because the concept of changing the past is problematic. But I have shown that this worry about the metaphysical possibility of changing the past does not rule out backwards causation per se, as we can (and should) understand backwards causation as influencing the past rather than changing it. Influencing the past, like influencing the future, involves no incoherence, even in a simple ontology containing a single consistent history.

Backwards causation, in general, then, is metaphysically possible. However, there is a problem that is specific to backwards causation among institutional facts. We certainly face this problem if we interpret backwards causation as influencing the past, and I suggest that we probably face it even if we interpret backwards causation as changing the past. The problem is that backwards causation of institutional facts involves conflicting declarations. Given realism about institutional facts, such conflicting declarations seem to lead to contradiction.

The problem of conflicting declarations is not limited to cases of putative backwards causation and is a problem that needs to be solved by any realist about institutional facts. I have argued that to solve the problem, we must view institutional facts as relational facts. So, for example, being king, being a border, and being a lecturer are all relational properties. They are properties that something has in virtue of their relation to the appropriate institution at a time.

Finally, once we have adopted the view that institutional facts are relational facts, we find that Silver’s examples of putative backwards causation are better interpreted as cases in which all the causal arrows point forwards. The appearance of backwards causation is due to mere Cambridge changes in the past. Hansson Wahlberg (2021) has provided an independent argument for a similar conclusion, arguing that, in order to make institutional facts consistent with the theory of special relativity, we must regard changes in the relevant properties to be mere Cambridge changes. Note, however, that this conclusion is not a denial of realism about institutional facts. These facts are just as real as the fact that I am taller than Napoleon, and the relational properties involved are just as much a part of reality as the relation of being one metre apart.