 I'll speak in the central seminar that today is James Finley from the University of Oxford. And I'm going to take out the pseudo of the pseudo passive. Right. So yeah, OK, what I want to talk today about is what some people call the pseudo passive and some people call the repositional passive, which is essentially what you see in one. So it's this kind of construction where the subject corresponds to the object of a preposition in the active rather than the object of a verb. So you've got a verb like rely, which takes a PP complement. Scott relies on Logan. And you can pacifize this to Logan is relied on by Scott. Right. Now people call this the pseudo passive because, well, this has implications, right? It suggests that it's not really a good passive. It's defective in some way. It's kind of pretending. I don't think that's really the case. So I'm going to call it the prepositional passive. But terminology kind of waivers. So what I want to do is give you a little bit more background about what the construction is. Kind of look at some of the data. And then kind of an overview of what the properties of the regular passive are, in English at least, and how the prepositional passive looks a lot like that. Then we'll think about ways in which it might look different, or it might look special, and that have kind of troubled people, made them think it's not quite such a good example. I'm going to give a sort of brief LFG analysis, not getting into the nitty gritty too much. And then we'll finish. So that'll be good. Right, so starting off, if you think about verbs like rely on these kind of prepositional verbs, it's maybe not that surprising that they have this passive alternation, right? Because you have this idiosyncratically selected preposition that's semantically inert. So it's not really doing anything. The complex is operating as a whole. And as a whole, it has this kind of unified transitive meaning. So if you were to write the meaning of rely in some kind of logical language, you probably have a two-place predicate rely on or something, because it's just kind of a unified semantic whole. And given that it has a kind of transitive meaning, it's not surprising that it participates in the passive, because transitivity and passive go together. And if this was all there was to it, there maybe wouldn't be so much to say. But the preposition doesn't need to be specified by the verb. So we get very sort of semantically contentful prepositions. So like in two, we have sort of contentful argument PPs, where Scott spoke to Gene, it's OK, Scott spoke about Gene, and both have good prepositional passives, right? Gene was spoken to by Scott, Gene was spoken about. And OK, maybe these aren't contributing kind of full truth conditional value. They're just doing some kind of thematic role alignment or something. But the point is that they are contentful. They do do something. And if you change them one word, you change the meaning of the expression. So that word is meaningful in some sense. But what's more telling, or maybe worse, depending on your point of view, is that you get these with what look like adjunct PPs as well, where the preposition clearly seems to be contentful, and it clearly has some meaning. So there's some nice real world examples here. In three, you have, when I'm on the bus, I don't like being sat next to, right? Preferably good, passive, and it's got this locative preposition that the N stranded. Similarly, with Charles Dickens' quill pen has been written with by me, which is instrumental. And in five, which is from The Voice, which I'm assuming you're all desperately familiar with, this sort of talent show where the judges face away from the singers so that they can only judge them on their voice. And then if they like them, they turn around. There was a cadet who came on. Nobody turned around. She'd been on before. And one of the judges said, to come back and not get turned around for, she's trailed off into some tragedy. But is there some kind of benefactive for phrase? Maybe, you know, so there's all sorts of things that, look, or people have at various times thought of as adjuncts that seem to take part in this prepositional passive. So huddle, fill, and pull them in the Cambridge grammar of the English language, divide up the prepositional passive between these two kinds, basically, between the ones where the preposition is selected, so type one, and those where it's contentful. It has some kind of meaning, so type two. And if there is this kind of subtype that deals with adjuncts rather than arguments, then this is, you know, we might start to see where the pseudo is coming in, right? Because this looks problematic. If we think of the passive as an argument alternation, it shouldn't be involving adjuncts. Another thing about the prepositional passive is just, you know, as an aside, it's very productive. You know, you might think, oh, it's just been kind of lexicalized, or there's certain expressions that kind of come together quite often. But you get it with very recent neologisms. It seems like as soon as you have a verb, you can productively make prepositional passives out of it. So in six, we just have a few kind of tech examples, right? We can't bring you everything that's being blogged about. Sean was tweeted at by Molly Mesnick. I presume that's exciting if you know who Molly Mesnick is. This will definitely be Facebooked on. All these kind of things, clearly productive, clearly just part of the grammar. So it's my contention, really, that I agree with Alcina that the prepositional passive has all the features of a canonical passive construction except for one, namely the definitional one, that it involves alternation with the object of a preposition rather than the object of a verb. And so in this section, I kind of want to go over a few of the properties of the regular passive that it shares and just kind of try and convince you that these are really doing the same thing in some way, shape, or form. So the first thing, and it might seem kind of trivial, but it's worth mentioning, is that the Morpheus syntax is the same, right? We have the same morphology. So the passive in English, you have this perfect participle that gets used for the passive. And in the prepositional passive, it's exactly the same, right? You have the same perfect participle used. It's not a different form. Similarly, you have the same auxiliary choices, right? So in English, you can have B passives and you can have get passives. And you get this with the prepositional passive too, right? So a new experience I got spat on today, but this is OK with B as well. So this might seem trivial, but if they really were kind of different animals, there's no reason why you couldn't have different morphological constraints, different sort of syntactic selectional constraints with the prepositional passive. But you don't. More interesting, perhaps, are the kind of semantic pragmatic constraints that you get on the passive. So there's been a lot of literature about the passive and the prepositional passive, kind of trying to categorize properties that the subject has. And it's kind of not surprising because the subject is the privileged thing that we've promoted. The reason we're doing the passive is so we can do something with this subject. And it seems like certain properties are predicated of it. And basically, these are much the same between the two kinds of passive as well. So the first and the kind of big one that people talk about a lot is what Bollinger called the effectiveness condition, where he says the subject in a passive construction is conceived to be a true patient, i.e. to be genuinely affected by the action of the verb. So he has this nice minimal pair in 8 where we have I was approached by the train. It's supposed to be bad. Whereas I was approached by the stranger is OK. I think I share that judgment. And the claim is that I was approached by the train is bad because in this case, the subject is not actually a patient. It's some kind of, well, Bollinger calls it the terminus of a path or something. But the train comes towards me, but it doesn't interact with me in any way and not actually affected by this train. Whereas if I'm approached by a stranger, the implication is that they're talking to me. They're propositioning me. They're panhandling, whatever. And so I'm affected. I'm a patient in this interaction. And exactly the same thing is true of the propositional passive. So we have pairs like 9 where we can't say soul was slept in by the businessman last night. But we can say the bed was surely slept in by a huge guy last night. The argumentation goes along similar lines. Cities aren't affected by having one person sleep in them. They're relevant important properties are changed. But beds are affected by being slept in, especially by huge guys or sort of laboring the point a bit. But that's the idea. As a kind of criterion, though, this is not very formally explicit. And actually, it might end up being so broad as to be basically meaningless. So I was sort of pointing to a little back and forth in the literature about this. But I think a bigger problem is that it's not actually accurate. So it's not the whole story, at least. So Riddle and Chine took this to task and come up with a number of examples from literature and from elsewhere, which seem to A, not satisfy the effectiveness condition, and B, actually kind of go against it, go near the inverse. So in 10, we have the example, such a dress can't be sat down in, which seems fine to me. But in this case, it's not the dress that's being affected by the passive predicate. If anything, it's nothing doing the effecting. It's kind of stopping somebody sitting down. And similarly, this is sort of literary example. There, the mistakes were in their houses, pervading their lives, having to be sat with at every meal and slept with every night. Well, again, it's not the mistakes that are being affected. It's the people that they're tormenting, whatever. So the subject isn't affected. And if anything, it's affecting. So what Riddle and Chine took suggest is actually, if not effectiveness that's relevant here, it's role prominence. So they say that all and only NPs who's referenced the speaker views as being role prominent in the situation described by the passive clause occur as subjects of passive verbs. Okay? But this, okay, now we want to know what role prominent means. So they get the concept from Schachter, who says that role prominence belongs to the MP who's referenced the speaker views as being at the center of events. Okay, but again, this is pretty vague. It seems intuitively to go some way to explaining 10, right? So, okay, the dress and the mistakes aren't affected by these predicates, but the sentences are about them in some sense, right? They are kind of, maybe they're topical, I don't know, but something they are kind of centered in some way, prominent in some way, right? And as Riddle and Chine took point out, this might offer some explanation for behavior in the regular passive as well. So it seems that passes with first person by phrases are generally pretty odd. So the cake has been eaten is all right, but the cake has been eaten by me is distinctly peculiar. The argument being that, well, if I'm the agent, so I'm the one doing the action, it's unlikely for me to view something else as more prominent than me, because I'm always going to be quite prominent to me anyway. And if I'm agentive, then I'm even more sort of prominent. So it's going to be very odd for something else to take that position as the subject of the passive. So again, it looks like this kind of constraint applies to the regular passive as much as it does to the provisional passive. But again, this is kind of no less vague than the effectiveness condition. So if we want to kind of formalize this, we're going to run into trouble. And Riddle and Chine took point out that there's no, well, they say there's no algorithm for determining role prominence. And this is no problem, but it seems to me like that should be a problem. If you wanted to have any predictive power, then you need to have a way of deciding whether something's role prominent ahead of time, right? One final characteristic is what people are called characterization. So we get these contrasts in the prepositional passive like in 12 and 13, where again, you can't say Saul was walked around by his father, by somebody or whatever, presumably because of effectiveness, right? It's not going to affect the city. But you can say Saul can be walked around in a day. Similarly, you can't say this statue was stood beside by John, but you can say no statue should be stood beside in this park. And it seems problematic for the other two, because I mean, A, Saul doesn't seem to be affected in either case, right? And B, it seems pretty role prominent in both cases, or it seems like it could be role prominent in both cases. Similarly in 13, you know, the statue isn't affected, okay? But no statue isn't affected either. It's not really clear sort of how the role prominence thing would play in with kind of these negative existentials. So what Kim and others say is that actually, what matters is that the VP gives a kind of general or characteristic property as a subject, or characterizes it in some way. And you know, this is a property of Saul is that it can be walked around in a day. And again, this has a reflex in the regular passive. So I think this might go some way to explaining capacities like 14, which come from verbs where the active voice object is not a patient. So it can't be affected, right? So in spiders are feared by many people. Spiders aren't affected by being feared. I mean, they might be squished occasionally, but in the sort of abstract, they're not affected by being feared, but they are characterized by it, right? So this might go some way to explaining that, the existence of passives like this. So there's kind of, it's a whistle stop tour really, but of a few of the kind of semantic pragmatic constraints that people have talked about in terms of passives. And they seem to apply equally well to prepositional passive and the regular passive. It seems to be properties of both. So now let's look at kind of why we might think that the prepositional passive is actually a bit more special. It's from kind of the concerns that people have had that make them think, well, there's something else going on here, right? We can't just say this is a passive with a little bit of sort of syntactic sugar is actually something special is happening. Well, one thing that's going to be different, well, one thing that might well be different is that with the prepositional passive, we have a preposition to account for. And so we can ask questions about what's the relationship between this preposition and the verb. And one common observation is that the verb and the preposition have to be adjacent or at least they have some kind of closeness that they don't have in the active. So we get contrast like 15. We rely increasingly on David. That's fine. But David is relied increasingly on, it's supposed to be pretty bad. But this isn't as straightforward as it looks. So others have pointed out, this is inadequate. You can get other things intervening. So you can get things like PP specifiers, these kind of small little words like right, straight or clear that appear with certain prepositions. You can get these intervening. So you can have I stood there heavily pregnant and obviously so and then look straight through. That's okay. Well, you've been walked right by these cases where there's something intervening between the verb and the preposition. That's supposed to be bad, but it seems to be okay in these cases. And actually, if not just these kind of specified type things, you can get all sorts of adverbs here. So 17 is just a few examples with rely on. 17A, the other thing that he sees in wintertime is a lot of the services are relied hugely on by other people in need over the festive season. I mean, it seems all right. This is from an Irish newspaper. Maybe it's a feature of Irish English, but I don't think so. It just seems perfectly normal. 17B is from an academic conference about agriculture, I think. Yerba Mata is consumed in nearly 100% of Argentine households. It's considered a staple food and is relied especially on by poor Argentines when food is scarce. That seems all right to me. And 17C is from a peer-reviewed academic journal. It's a chemistry journal. It's called Combustion and Flame, which is a much cooler name than I think any linguistics journal has ever had, but there's gas turbines have rapid transient response capabilities and thus will be relied increasingly on in markets with large intermittent sources. And this is exactly what we said was bad in 15B. We had David as relied increasingly on, now we've got a pretty natural example that seems to be fine. I'm much happier with 17C than I am with 15B. So it's clearly not an outright ban on things appearing in between the verb and the preposition. One thing to note is that in 17, all of them have material after the preposition. So in the first two, it's the by-phrase. And in the third one, we have this sort of in markets. And I think that might sort of point towards something that's going on here. Maybe there's a prosodic sort of issue at hand. Maybe there's some kind of discourse, information structural thing happening. I don't really have a story for this, but the point is that it's not actually a sort of syntactic constraint on the prepositional passive as some people have proposed. And I think really the problem is that context is important. There's clearly something going on here and there's some need to get an acceptable context for these things to work. And so, you know, if you're an armchair linguist sitting thinking up examples, you think of examples like 15 and then you say they're bad. But really, there are examples that make sense. So adverbs seem to be okay. We can get those. The question there is, well, what about direct objects? Because you can get these kind of, you ought to be able to get constructions where you have direct objects intervening. So these are generally pretty bad. Again, along the same lines as we've said before, yeah, along the lines of the adjacency constraint. You say they put some books on the table, that's okay, but the prepositional passive, the table was put some books on is, you know, garbage. But again, it's kind of leaky, this rule, right? So non-thematic things can appear there. So idiom chunks or bits of like verb constructions. So you can say you've been taken advantage of or Russia was declared war on or I've been made a fool of, these kind of things. These are perfectly acceptable prepositional passives with intervening objects. And again, it might be the case that really context is the key. So Bola just suggests that the only real restrictions are clarity and intent, which are definitely non-synaptic features. But, you know, you get examples like in 20, so that city has been fought many a battle over. We have quite a complex MP appearing between the verb and the preposition here. Or he's been burned, stuck pins in, beheaded, all in effigy of course. To be whispered such dirty innuendos about was enough to break any girl's heart. This one doesn't seem so good to me, but this tool has never been used for its main purpose. In fact, it's never been done anything with at all. Or, you know, simply I don't like to be told lies about. So your mileage may vary with some of these examples. There are definitely ones I think are better than others. And maybe some of them are more kind of light, verby, idiom-y than others. But I think the point is, it's kind of clear that there are things that kind of appear here, quite complex things sometimes, which seem to be real objects that seem to be okay. But again, it doesn't seem to be an outright ban. So I'm gonna assume that there's kind of no actual syntactic principle stopping things appearing here. There's clearly something kind of going on, but it's not ruled out sort of too cool. It's not a complete ban. But even if there were something syntactic going on, there's a kind of get out, which is that the same thing happens with the regular passive, it's just harder to see. So Trusswell pointed out that there's a certain parallel here with the fact that English is what people call an asymmetric language when it comes to double object constructions. So you have examples like Kim gave Sandy a book, 21a, and you can get the passive Sandy was given a book where you sort of promote the nearer argument, the closer object to the verb, but you can't have a book was given Sandy, at least in standard English, standard British English. And this is kind of parallel to the fact that you get this thing with the prepositional passive as well, where if you promote the nearer argument, so a book was given to Sandy, you're okay, but if you try and promote the further argument Sandy was given a book to, it's no good. So this is kind of suggestive, that maybe something similar is happening even in the non-prepositional passive. But what's, I think, more interesting is that in symmetric languages, where this contrast doesn't exist in the regular double object passives, the contrast doesn't exist in the prepositional passive as well, and the things that are bad in English are perfectly fine. So there are languages like Norwegian where you can promote either objects in a double object construction, right? So you can have John was given a fiddle or a fiddle was given John, both meaning that John's receiving the fiddle. And correspondingly, you get prepositional passives where you have an intervening direct object. So the letter was pasted stamps on or the answer was given a watch to yesterday. Well, so the various authors hedged their bets on these a little bit and some say some people find them awkward, some people don't like them so much, but they're certainly not as bad as the English versions seem to be. So even if there is something syntactic going on here, it seems to be the same syntactic thing as is going on in the regular passive. So it's not, again, not a special property of the prepositional passive, really it's something about the passive in general. Okay, so now we come to kind of what's maybe the more, well, troubling thing, I guess, for some people, is that you get these, suppose it adjuncts taking part in the argument alternation, in the passive. So we get, especially locatives and instrumentals, right? So he's had some examples earlier, there's a couple more examples here. This broken shovel has clearly been shoveled with rather too enthusiastically. The room had not been cleaned, the bed was slept in, but towels were everywhere, et cetera. So yeah, we definitely get these things that look like adjuncts and, you know, this is a problem. First of all though, note that it isn't all adjuncts, it's not that this is completely leaky and sort of open, it's not anything goes when it comes to these kind of things, there are still bad examples, so things like 27 and 28. The Jones is left after dinner is okay, but not dinner was left after. And Lincoln died in this theater, okay, but this theater was died in, that's just terrible. So part of this contrast seems to be a lexical thing, right? And this isn't surprising if we think that the passive is lexical, then, you know, there are gonna be lexical exceptions to it. And so people have pointed out that there's a difference between unaccusative and un-negative verbs in this respect. So this would explain why dye is bad. Unaccusative verbs are supposed to not like the passive, right? And dye is unaccusative, so, you know, this is just bad for those reasons. And this contrast is supposed to come up in little minimal pairs like 29, right, where the desk was sat on by the gorilla is okay, because it's sort of volitional, right? Whereas the desk was sat on by the lamp is bad. And this is supposed to kind of pick up the un-negative, unaccusative distinction, something approximating it. And there may be a sense in which this applies to regular passives as well. So things like Kim resembles Sandy is okay, but Sandy is resembled by Kim. It's not great. This might be an effectiveness thing or something else going on here as well. Not so sure this is quite so strongly the same effect. But I think a bigger problem here is the kind of issue of delimiting arguments from adjuncts in the first place. So the force of this objection is basically, well, there are these two categories, arguments and adjuncts. Passive targets, this group, arguments are not these. The prepositional passive seems to target both, therefore it's not the same thing, right? But we can take issue with the very first premise that there are these things called arguments and adjuncts. And, you know, many people have. There's been a lot of ink spilled on trying to find ways to successfully delimit the two, to find tests that reliably work for the two, and you can always pick out the right thing. And, you know, to no great success. So it seems really, I mean, the conclusion has always been that it's kind of gradient thing, but really it seems like there aren't good reasons for thinking of these as two separate categories. And it's been, well, there's been a lot of work on kind of, you know, either we end up inventing extra categories to kind of put in the middle or we just do away with the notion altogether. It's easy to say that there aren't kind of behaviors that correlate with these kind of properties, right? So there are clearly more argument-like things and more adjunct-like things. And perhaps, you know, the difference between a direct object and a manner adverb are quite clear. So I ate the cake versus I ate happily or something. There's no one's going to be confused about whether happily is an argument or not. And these have syntactic reflexes because, you know, there's an ordering thing going on here. You can't say I ate happily the cake, but you can say I ate the cake happily, right? So the direct object has to appear. The argument has to appear before the adjunct. This is kind of what we're all supposed to be happy with. But, you know, particularly when it comes to obliques, particularly when it comes to PP kind of complements, the distinction is much more leaky, right? So the ordering thing doesn't work anymore, right? So you can speak happily to John or you can speak to John happily. And these kind of seem to permute quite readily. And, you know, there's been various proposals to kind of collapse certain kinds of obliques into adjuncts or vice versa, or to collapse everything into one kind of thing. So Shepiokovsky, the citation there, is suggesting that really we want an HPSG style thing where we have like a list of all the dependents and we don't really care too much which ones are arguments, which ones are adjuncts. There's just a kind of one type of thing. So if this is true, well, so this will affect how you look at the problem, right? If you're someone who's coming at this saying arguments and adjuncts are definitely different things and we need to respect that in our grammar and, you know, the passive is very clearly an argument targeting thing, then you're going to see the fact that the prepositional passive can kind of cross borders as a problem or as a reason to think of it as different from the passive at least, right? So not as the same kind of thing. But if you are coming at it from the perspective of, well, there's a kind of, you know, it's kind of messy to have this distinction in the first place and maybe it's better if we don't, then really this is kind of more grist to the mill. You're just saying, well, there's another example of why there's no distinction between arguments and adjuncts. The prepositional passive doesn't care either. This kind of operation doesn't distinguish between them. Yet another instance in the grammar of this distinction not mattering. So I think to some extent it depends on where you're coming from. You might object then, well, why don't we get regular passives from adjuncts, right? And I think probably it is a kind of categorical issue here where most of the, you want a nominal, right? You want an NP to take part in the passive and most of the adverbial MPs I can think of are sort of time things or maybe I ate yesterday or something. And that's going to be a bad passive subject for all sorts of reasons because it's not going to be affected by things. It's not going to characterize it very likely. It's not going to satisfy a lot of the other constraints in the passive. So I think probably the reason you don't get those not a lot of options and the ones that there are are just a bad passives. Whereas when you go into the world of obliques, you know, like I say, the boundaries between the obliques and adjuncts is much blurrier anyway. And there's a lot of kind of a lot more nominal type things in prepositional phrases, right? We all have this nominal object, which is available to kind of get used. So this is my kind of hand wavy way of saying that I guess it doesn't matter. If there's no such thing, if there's no distinction in arguments and adjuncts, then this kind of objection doesn't get off the ground, right? If it doesn't matter that the prepositional passive doesn't care because this distinction doesn't exist, so it shouldn't care. I have a slight story to tell later on about what it is that's kind of relevant. It's not really argument adjuncthood. I think it's sort of to do with the semantic pragmatic stuff and to do with kind of more meta-linguistic things. But I will come back to that. So for the moment, I'm sort of asking you to take it on faith that it's OK to collapse these things. Let's not worry too much about the fact that this seems to kind of cross borders here. So OK, in that case, I hope to have shown you that there's not that much going on in the prepositional passive, which is different. Really, the thing that's different is the mapping between what the syntax is doing and what the semantics is doing. So the difference is that the subject corresponds to the objective preposition and the active buses of verb. What that means for the analysis is that the subject of the clause has to correspond to the internal argument of a preposition in the semantics. So in the bed we're slept in, we want slept to go within in the semantics. So we want bed to go within in the semantics. And given that I'm claiming that there's not that much difference between the regular passive and the prepositional passive, what would be nice is if we had an analysis of the regular passive that we could just kind of tweak a little bit. That's what I'm sort of proposing to do. So I'm kind of taking a version of an LFG analysis of the passive and just adding a little bit of information. And I think this suffices. So I think LFG kind of has an easy time describing this kind of thing because of the kind of modularity. So I'm assuming everyone sort of knows a little bit about LFG, but I'll go over some kind of basic stuff. So because we have this kind of very modular grammar, it's kind of easy to get what it's easy to describe these kind of relationships between the different levels. So in figure one, you've got this kind of baby version of some part of the correspondence architecture. The kind of classic LFG is the constituent structure and the functional structure. You have the tree, the constituent structure, which tells you about hierarchy, tells you about constituency, linear order, this kind of stuff. And then you have the functional structure that tells you about more abstract things like grammatical functions, like long-distance dependencies, binding, that kind of thing. And they're connected by this function phi that kind of protects from the C structure to the F structure. Over the years, there's been all sorts of other structures added, proposed, whatever. And kind of one of the more venerable ones is the S structure, the semantic structure, which kind of changed shape various times over the years. But I'm kind of taking a modern version of it where it's kind of seen as this connected kind of attribute value matrix that describes kind of semantic relations. So figure two kind of gives you a little example of this. On the left, we have the S structure for the bed was slept in, where I just have, you know, this predicate sleep, the subject, the bed. It still has this kind of oblique hanging around within it, and it's in the passive. And then on the right-hand side, you have this semantic structure, where there's a sleep relation that has an event in an argument. And it has a first argument, an arg1, which is just the kind of the highest argument, whatever you want to call it. And then this location argument, which has its own kind of internal structure, where there's this in relation and its argument, which is called PRG, for the sake of kind of brevity. That's just the internal argument of the preposition. So really, what we want to describe is the mapping from the F structure to the semantic structure. And the way that I'm going to do this is based on some sort of recent work by Vasude and Giorgolo, and me, and some other people, where basically we have some version of a theory of lexical mapping theory, which is a family of theories, I guess, that basically describe these kinds of relations between grammatical functional things and semantic arguments, wherever they may be. And they provide equations a bit like 31. So 31, the arrow just refers to the F structure that you're in. The little sigma tells you to go follow the sigma function from wherever you are. So it says, OK, the F structure that I'm in, go to the semantic structure that corresponds to that. That arg1 is the same as the sigma projection from the subject or the oblique agent of the F structure that I'm in. So it basically just says, this will occur in a verb, lexical entry. And it'll just say, well, my arg1 is either a subject or it's an agentive byphrase. So it's underspecified in this way. And this is to allow for argument alternations. It's slightly more complicated than that, because in the short passage, for example, it doesn't get realized at all. So if you care about the sort of details, I've put some stuff in the footnote, but it all gets a bit weird and it's too many brackets. So I just wanted to simplify. But assume that 31 also accounts for the fact that you cannot realize this at all. Then what the passive has to do basically is constrain these possibilities. So the normal mapping thing says, OK, you can have one of two grammatical functions. And the passive says, no, actually, you can only have one. So 32 is a template, which is just a kind of named bundle of information which is supposed to correspond to basically what the passive does. So if a verb is in the passive, it has this information in it. So it says, OK, my voice is passive, sure. Then you have the mapping thing that basically constrain is 31. It's just a sort of non-open version of 31, which says, no, my arg1 is my agentive oblique. It's not a subject. And then we have this optional meaning constructor, which you don't need to worry about too much if you're not into glue semantics, but it's basically just a way of existentially closing the first argument. So in the short passive, you don't get anything that corresponds to the arg1, but you still need to do something with the meaning of the predicate in the semantics. So you just kind of existentially close it and say, there is some x that's doing whatever you want arg1 to do. That's fine. So this is kind of what the basic passive analysis is. So the question then is, well, what do we need to add to this to get the prepositional passive if it's really just the same, but with a little bit of a tweak? Well, really, we just need to add the mapping information. So it's just the question of, in Figure 2, the kind of bottom arrow. That's what we're worried about. The rest of it will follow from more general stuff, but we want to get the subject into the PR. It has to be the internal argument of the preposition and the meaning. So essentially, we just kind of stick this in. So 33 is just an equation that does exactly that. It says, OK, find the subject. Follow the sigma projection. You are at the PR of your oblique. Again, this is slightly more complicated because it's kind of ordering things that we have to figure out, but don't worry too much about that unless you really care about LFG architecture. So then, really, we just bung this in the passive template. We get what we see in 34. So we just have this as the third line. We put it in brackets because it's optional, because in the regular passive, you're not going to need it. In the prepositional passive, it'll have to appear because otherwise you'll end up, you won't be able to get a good proof of the meaning, because you won't have enough resources, you won't have enough things linked up. So basically, it appears when you need it and doesn't when you don't. So this is kind of what I'm suggesting is going on. So how does this kind of play out? Well, for the type ones, the things where we have this kind of esoterically selected preposition, what I suggest is that really, we just have that order. We have on page nine, wherever that may be. It's not where I thought it was. There we go. We have in 35, where we have the kind of excal entry for a rely. The meaning is the bottom line, just like any other transitive verb, basically. It's just some predicate rely on. But the second line just equates its second argument with the internal argument of on. This is OK. Well, there's some oval on. And that thing's PR, that thing's internal argument is my arg2. Now, in the normal case, that will be the object of the preposition. That's just how this kind of normally works. It just passes the meaning from the preposition to the main predicate. But in the prepositional passive case, we have this kind of mapping thing that says, OK, the subject is actually the PR. And that just ends up saying that the subject is the arg2, which is kind of what you want. And that's fine, basically. For the type twos, it's a little bit more complicated. Well, if the type twos are obliques or if we can collapse everything into one category or whatever it is I wanted us to do, then this is fine. This kind of carries over just the same. Because we're doing exactly what we need to do. We're just telling the semantics where to find the argument it needs. So if it needs it, it'll find it. The question is, well, if they're not obliques or if there is a distinction between obliques and adjuncts, then what's going on here? What actually is motivating what can take part in these things? And what it seems to me is that, well, the semantic pragmatic constraints really are the kind of ultimate arbiters of what's good and what's not. So 36 is a classic contrast that this bed has been slept in is OK, but this bed has been slept under is not OK in a kind of null context. Out of the blue, this contrast is supposed to hold because of effectiveness, presumably. But passes with slept under are fine. It's not that slept under can never express a good passive predicate or something because you can get things like 37, right? This heavy sheet is designed to be slept under where I know we've got some kind of characterization going on, perhaps. So given an appropriate context, you can rescue a lot of these things. And really, if it's about, well, even if it's about effectiveness, it's certainly about context. And it's really about kind of non-linguistic stuff or at least not just linguistic stuff. So the argument for 36 is that, OK, beds are affected by being slept in, but not by being slept under, right? So that's why 36A is OK, but 36B is not. But that's not really lexical knowledge, right? That's knowledge about the world. That's knowledge about what happens to beds when people sleep in them and what sleeping in a bed involves. But it seems to have grammatical effects, grammatical reflexes, right? It affects whether things are acceptable or not. And this is quite reminiscent of what James Pistoyovsky has talked about in relation to what he calls qualia structure, where he argues that there are lots of things like this, where there's kind of real world knowledge that has to be baked into the lexicon somehow in order for us to interpret certain things. So his classic example is he says, well, OK, the noun book has to include in its qualia structure the fact that what you normally do with books is read them or write them. And this is how we interpret sentences like John began the book, because really, we take that to mean John began to read or write the book, basically. The claim is that began kind of needs an event or some kind of thing, and you begin doing something. And so it looks inside the lexical entry for book and finds these kind of events, these reading and writing events. And so it must be one of those then that we're beginning. And that's how that works. And I recently found out about this nice example in German where a similar kind of thing happens where when a PP argument is separated from the noun that selects it, then the acceptability is determined on whether the meaning is kind of contextually unmarked or not. So in 38, I'm not going to attempt the German, because I'll embarrass myself, but we have about syntax. Has Hans a book borrowed? So it's the book about syntax that he's borrowed, but the about syntax has been sort of fronted. And this is OK. But if you have the same sentence, but with stolen instead, so about syntax has Hans a book stolen, this is supposed to be bad. I don't know if there are any German native speakers who can verify this judgment or not, but this is supposed to be the case. And again, this seems to be an example of real world type stuff affecting grammatical judgments. It's not that this is exactly the same syntactic structure. It's just that one meaning is less plausible or something. And therefore, this seems to rule it out grammatically. And the fact that context is important comes back to the contrast in 36, because it's not just that we have to have a linguistic context. If we get a good enough real world context, then 36B seems all right too. So if you imagine a situation where we're all sleeping in a dormitory, but it's really, really busy. And so some of us have started sleeping under the beds as well, because it's just not enough room. And then someone else comes in and looking for somewhere to bed down. They want to go sleep under a bed. You could quite happily say 39, I think. That bed is being slept under already, or that bed is already being slept under or something. And then it's fine. So really, it's about the context. But the point is that 39 requires a more marked context to come off successfully. You don't get it in the sort of null context. And so what it seems like is really that, well, relations that are available to the prepositional passive are simply those that are contextually relevant. The point is just that some of those relations are always relevant, or at least are available in the null context for want of a better expression. And so then you can use them without this richer background. And so what the, well, this is my more speculative Mickey think message, is that, well, actually, all sorts of things like, so all these sort of somatic-pragmatic conditions like effectiveness or prominence, and maybe even this argument adjunct distinction in some sense. So the things that are available in the null context are sort of more argument-like in that way, in that they are, if we go for a kind of qualia structure thing, then they are specified in the lexicon. Sleep does select for in, in some sense, because what we know about sleeping is that we usually do it in things. What we know about beds is that we normally sleep in them or something. So in some sense, they do select for each other. So it is more argument-like. And perhaps these kind of things are all actually just epiphenomena, and they may really be the result of some kind of Greithian-type inference interacting with knowledge about the world and its relationship with word meetings and things. And so actually, that's kind of the bigger explanation for what's going on with some of this stuff. So, OK, just take it, take us home. I hope to have convinced you, at least, that the propositional passive is only minimally different from the regular passive. And if that's the case, then a minimally satisfying analysis really is just the same analysis as we have for the passive, but tweet a little bit. And really, we can bake in these semantic pragmatic constraints if we want to. We can stick extra constraints into the passive template. We can put extra meanings in there that say things about the subject. So that's not a problem. But really, it would be nice if they were reducible to something more general or more basic. And I've tried to sort of gesture towards what that might be. But as ever, if it is to do with kind of relevance and more metalinguistic stuff, then a good formalism is going to be hard to come by. Thank you.