 So, thank you for inviting me. I'm going to talk about reconstructing derivational morphology. So yesterday we've had a number of talks about phonological reconstruction, and then in the end it was sort of pointed out that we also need to have methods for morphological reconstruction and we maybe don't feel quite as confident about morphological reconstruction as we feel about phonological reconstruction. So this talk sort of goes in that direction and tries to figure out what kind of methods or what kind of methodology we need or we should use to do derivational morphology in reconstruction. So when we do reconstruct derivational morphology there are some problems that we're immediately confronted with that we need to worry about. We need to think about the properties of verbal derivational morphemes that we want to reconstruct or we need to worry about which ones we can reconstruct, which ones we should reconstruct and those may differ from the ones that are regularly reconstructed. So we need to decide which properties of the historically attested derivational morphology that we see belong to the reconstructed proto-language and which properties are independent innovations. And then we need to decide on how to formalize the properties of these morphemes and I think that's where there's the most insecurity because we're not always sure whether we want to do that and how we want to do that and I think that's actually an important step that we need to take in order to help us generalize and also compare diachronic developments within and across languages. So this is going to help us to explain how morphosyntactic properties of derivational morphology changes. Okay, so to give you a concrete example from the Indo-European so this is from the lexicon of the Indo-European verbs, the LID. And this is the entry for the Indo-European Gwen to HIT and now we are confronted with a frequent puzzle that we have when we use the LID mainly how many primary verbal stems to be reconstructed for this route. So here the LID editors in this blue box tell us that probably the European had three primary or inherited verbal stems. The oldest one is an old asymptomatic present that we have attested in HIT page and in the Iranian so this is actually attested in some languages and then we have a duplicated present and a scale present and this is attested in HIT and in the Iranian respectively. And then in this orange box we have a number of verbs in different languages so meaning with that old Irish and so on with this square bracket here and this is how the editors tell us that these are innovations so the stem from morphology in these verbs is not directly inherited, it's not the same as these verbs, it's an innovation. So this actually raises a number of questions so why do we have three primary present stems and usually people say oh they have different functions originally or maybe they're different diachronic layers and then the reflexes of some of the stems vary as we've seen so we have as reflexes of this oldest stem we actually have thematic stems or yeah or presents in the Indo-European languages. So the question here is how do we decide whether any stem an equivalent stem that we see in the Indo-European other languages is inherited or a brand specific innovation. So what makes the editors of LIE decide that this thing is an innovation but these two that are also only attested in one branch are inherited. So this is a common problem and of course LIE is a great resource and the editors usually do tell us why they decided one way or the other in their footnotes and so on but this is just a problem that you have for basically every entry in LIE this is a very common problem. So the question is is this just a problem that weird in the Europeanists have but I think this is a much more common problem that you have immediately when you reconstruct the relational morphology in any family. So just illustrating to try to find examples from other families in old Chinese backstand saga talk about all these reconstructed verbal morphemes and then they say about this end prefix this typically derived state of intransitive verbs often from transitive verbs. And then one of the examples they give to me looks like a straightforward causative alternation where you have a transitive causative and then you derive the intransitive anti-causative to detransitive. And that's all nice and well and then they talk about the S prefix and they say this increases valence in a verb and then one of their examples also looks like a causative alternation verb but in this case the anti-causative variant is basic and the causative variant is the derived one. So there are two ways of treating causative alternation verbs and now we could ask which ones are or maybe there are more ways so I'll just pick this example so now we could ask which ones are primary causatives which ones are primary anti-causatives how can we tell and which type of verbs can be selected by these prefixes and can they be combined and so on. So I think this is a more general problem or another very frequent problem is a situation that we see here in proto-algonquian in the verbal system we have this formative morphine that's lost with question marks here in the independent inflection where it's not quite clear what it's doing there it seems to be some sort of verbal stem-forming suffix maybe but proto-algonquian also had a nominalizer that looked very single and that you could use to form excretions from verbal roots or stems and these are both reconstructed for proto-algonquian so it looks like some sort of reanalysis happened and a nominalizer became part of the verbal stem-forming suffixes so this is also something that we see happening very often and the question is how did this happen why did this happen and how did the properties of this thing change from forming nouns to forming verbs so these are the general questions that we're going to be confronted with when we do a reconstruction of verbal derivational morphology so the goal today is to discuss some theoretical background and methods that I think might be useful in reconstructing derivational morphology especially with respect to understanding the diacrony of derivational morphings in the verbal system and we're going to look at three case studies from Indo-European hopefully we get to look at all three and then conclude and try to point out some generalizations that follow from this okay, so for the background I tried to keep this very short but I think there are some concepts that are useful and that I can take for granted some of them are framework specific the uniformitarian principle we all know and love the Borochomsky conjecture a non-lexicalist approach to derivational morphology and then what exactly we mean by selection and what exactly we mean by reanalysis and some of these are probably familiar to you but just to make sure so the uniformitarian principle you probably all know and use in some version that's similar to this quote by Hoppe here so the general processes and principles which can be noticed in observable history are applicable in all stages of language history and I think we all subscribe to that because otherwise there wouldn't be much point in doing reconstruction what another way of formulating this in generative approaches to linguistic theory is that this uniform behavior of human language follows from the fact that there is an innate human capacity for language or for acquiring a language that's also called universal grammar sometimes also a shared genetic endowment that makes it possible for humans to acquire language but also limits possible synchronic languages that can be acquired in possible diachronic development so in a way this sort of makes uniformitarians a little bit more restricted and this is sometimes called syntax or core computational component as in this quote by Mark Hale here who says the computational component of the syntactic module of human grammar is universal and invariant variation is limited to the lexicon and what has traditionally been considered syntactic change is to be taken instead as representing a change in lexical features in particular those lexical features that are syntactic relevance so he talks about syntactic change but all of this is also applicable to the kind of morphological changes that we're going to look at for specific reasons so this is known as the Boroyachomsky conjecture all parameters of variation are attributable to the features of particular items or functional heads in the lexicon and then Hale says if we can reconstruct the morphosyntactic features of lexical items and functional heads and we assume the syntactic computational system is universal and invariant we can reconstruct specific sentences for a proto-language and again this is relevant for us because we're going to assume a non-lexicalist position at least for the purposes of this talk where word formation is not distinct from syntax from phrase formation you don't have to do this outside of this talk I just decided a while ago that there are good arguments against lexicalist approaches so we're going to use this non-lexicalist approach and this is what it looks like so in this approach we're going to treat derivational or stem-forming morphology as functional heads or syntactic terminals which have particular morphosyntactic features so this is a simple example from the end of 2015 a structure for the English word cats after insertion so we have a core a root with the core lexicals meaning a categorizing head these are the functional heads that we're interested in so a categorizer that says this is a noun that can have particular features and that has a particular realization in this case it's zero because English doesn't do much in terms of nominal stem formation and then a separate head that tells us that it's plural so corresponding to derivational and inflectional morphology more or less in the traditional terminology so these heads and their features are stored in the mental lexicon of speakers and they vary from language to language which isn't shocking because we know that the plural in English looks different from the nominal plural in Swahili or Old-Head German but this is essentially the Boria-Chomsky conjecture if there is variation it's in these functional heads and of course also in these roots so we're going to use this we're going to treat verbal stem from morphology exactly the same way so parallel to little n we're going to have a little b here and this little b can have different features different morphosyntactic features or values it can be causative or incoative or express a state or an activity there are different ways of formalizing this this is basically terminology following work by Heidi Harley and others and these categorizes select lexical roots or other categories with particular features that have to be compatible and again this is this is just formalizing empirical observations because we know not every verb can behave as a causative alternation verb not every English verb can undergo the causative alternation and not every verb acts as an unaccusative or expresses a state so this is just a formalization of these observations and the selection of properties and features of these categorizes can change over time and this is relevant for us this happens through reanalysis defined here as a surface stream in a particular grammar G2 receives a different underlying interpretation representation than in the input grammar G1 during language acquisition this is an acquisition based approach to reanalysis for our purposes it's just important that this is not structural simplification because we're going to see cases where we don't actually lose structure we gain structure or we just have the same amount of functional heads so that's why like George Walkton's definition of reanalysis or description of reanalysis as a process whereby the here are assigned the parts to the input that does not match the structure assigned by the speaker so a reanalysis is a mechanism and it's not causal so it doesn't cause syntactic change it is syntactic change it's a very nice way of putting this and the same holds for morphological change because we've just said we're going to have a non-lexicalist position so these observations also apply to morphological reanalysis in our approach okay and I think that's all we need to turn to our first case study that's put in the European NP and fortunately Hannes already talked about this yesterday so I can go over this fairly quickly and go to the analysis but just as a reminder this is the suffix in Proto-Indo-European that looks like a symphonic active participle suffix in most of the older Indo-European languages and in the Iranian, Greek, Italian, Germanic Tuqueria and so on and in these languages what's important is whether or not you have a formally active finite paradigm if you have a formally active finite paradigm you can make an active non-participle so valency isn't important we saw a transitive an irritative and an unattuitive valency isn't important what matters is the presence of a finite active paradigm and here are also examples to show you that this is really syntactically active so we have participles with accusative internal arguments and subject agreement and so on so this is usually described as subject oriented and then we already saw that the heated and participle forms adjectival passives or looks like an adjectival passive it does not care about voice morphology so it can be found to morphologically active or non-active finite verbs but it does care about valency so here we have some un-participles to formally active verbs and here we have some un-participles to formally non-active or middle verbs and they are they behave syntactically the same way so they look like passive participles but so theme oriented is maybe a little more accurate so what's important is that you have an internal argument or theme and if this were an in-european language this would mean ceasing, giving and going and not cease, given and gone so theme orientedness instead of subject orientedness and this can also so this really behaves synchronically like a passive participle it occurs in periferastic passive with demoted agents and it's generally described it does passive or is theme oriented participle ok so the problem that we already saw yesterday is that the heated reflexes of nt are passive or maybe more accurately described as theme oriented or in some state and the same thing is syntactically active in Tokarian and Indo-Iranian and Greek and so on so this tree again shows you the function of nt correlated with the commonly assumed split of dates so here Tokarian still looks like the second to branch off so this should probably be a little lower in the tree but the important thing is that everything after Anatolian has something that looks like active nt right so the question is what should we reconstruct and this actually leads us back to Eugen's question from yesterday so how do we know which one is older should we reconstruct active so that makes it very difficult to get the Anatolian use from that so it's difficult to imagine can I go back here so if we have active input structures like this it's difficult to imagine how that could be re-analyzed as passive so going from active to passive is going to be very difficult even if you take an unaccusative form and let go it's difficult to see how that's going to be re-analyzed so so this is relatively difficult we could start with the theme-oriented situation in Anatolian but then we have to explain how this changed into the inner European active participle and then the solution is going to be what actually looks like a complication the evidence for the nominal nt in the European branches that we've already seen yesterday so again some examples we've seen that this nt also derives adjectives from nouns with nice examples from Hittite and from Indo-Iranian some of these look very old so probably inherited from Proto-Indo-European and even some synchronic ones in Sanskrit that look like participle so Sahand and Shuchan but they can't be synchronic participle because there's no synchronic finite verbals or in this case this is even worse if there is one then it's non-active and we see the same in some other Indo-European languages where we have these mismatches where things that look like active participle don't have a synchronic finite paradigm or if they do it's not an active one so this looks like it should go with an active participle but there's no such thing or creole these don't exist and even worse this one looks like an active participle but the corresponding finite verb is always non-active this is a deponent verb so these have long been interpreted as archaisms and it's been suspected that this participle nt was actually originally a denominal adjective there are also typological parallels for this kind of development so for our purposes what this means is that this nt must have acquired more verbal functional structure it must have been able to select verbal functional heads at some point in the course of its development and this means its selectional properties must have changed from selecting roots to selecting a particular type of v so originally a stative v and then eventually also eventive v and this is what we need to get to the inner and the European stage so this is our first attempt at formalizing this so in probably in the European and probably in the Italian we must have had this stage so this thing selects nouns or roots we must be able to select nouns because we've seen examples from Hittite where we see this nominal stem from in morphology and then from the reanalysis so these adjectives are going to express states so to go from this kind of state to a verbal state should be a fairly trivial reanalysis so we get to this stage and this initially would have had a morphosyntactic feature state or stative so we get from this to the next stage well states are ambiguous there's a lot of literature in that it can be a resultant state or a target state or different kinds of states and some of these are more amenable to being reanalysed as eventive and some as processual which Hannes already talked about yesterday and then we get to a situation where there's even more verbal functional structure and this thing here that we don't need to worry about too much is what gives us the subject orientedness so now we select something that actually has an external argument so this is an Italian state where we only have the internal argument that this is the European stage where we select something that also has an agent okay so this is the first case study the or this is an attempted formalizing this analysis the second one is similar here we also go from a nominal state or a nominal morpheme to a verbal morpheme this is probably in the European e-larange one and here we have a similar puzzle we see in almost all branches of Indo-European a verbal standpoint that looks like a long e from e-larange one but the distribution is weird so it makes a denominality verbal present but also hei-rists so it doesn't seem to know whether it's a present or an hei-rist stem and this is unusual for a verbal stem from morphology imported in the European so this is immediately this makes it immediately suspicious in Anatolian for example we have presents with this long e from adjectives so Marcia falls Marcia it becomes falls in Latin we have primary the verbal e-forms and the adjective ones and in Germanic the same way so these languages all have presents but then in Greek and Slavic and Baltic it looks like this thing makes primary ae-rists so in Greek this is called the passive ae-rist but it's usually in the older stage it's not necessarily passive but in co-etive so the solution for this that LIV proposes is to reconstruct two different primary verbal suffixes for Proto-Indo-European one that they call fientif and this is this ae-laryngeal one that gives ae-rists in the older languages and one that they call passive and that's the laryngeal of the suffix plus another suffix so this is a composite suffix actually and that results in presence in the older languages and to give you an example for the entry of this root tells to become dry here you see they reconstruct the primary ae-rists, the primary present and then a fientif and an ae-sif and to be fair they have some question marks here for some of these forms but it feels a little redundant to have fientif and ae-sif in addition to reconstructed primary ae-rists and present stems and then the question is why do we assume that branch specific things that look like branch specific developments are actually inherited and they also say that well maybe this is actually derived from that and so on so this doesn't feel like the ideal solution there are also phonological problems associated with some of the reflexes of these suffixes so an alternative that's been proposed is that this long ae actually originated as a nominal suffix we see many old ae-verbs associated with so-called adjectival roots or roots expressing a state that belong to the so-called Kalan system so primary adjectives like red or dry and so on and it's been suggested that this verbal ae was originally identical to the instrumental singular ending on abstract nouns on adjectival abstracts and we still see this in some instructions in Vedic and Latin where we have things with a long a or a long ae plus some auxiliary that express a state or a change of state so become hidden or become hot or make hot or something like that so you can also make factitives so this suggests again that this thing that looks like a verbal suffix was originally denominal or adjectival and developed into a state if in co-authiff verbal stem-forming suffix and we still see this stage in Greek actually because as we said the passive ae-reast is actually not passive at the oldest stage it's in co-authiff it's in complementary distribution with verbal stem-forming morphology so it looks like lexical aspect originally and not syntactic aspect and it originally did not express a voice a voice distinction so the passive use developed later and also suggests that this suffix changed in the history of Greek and acquired more functional categories so again this is my attempt to formalize this a little bit we start with a noun with particular features this probably should be a separate functional projection but we have to formalize these instrumental features somehow this seems to have meant with or maybe possession and it looks like these types of features when they're re-analyzed as part of the verbal domain express states and that's not surprising because yeah this thing with or possessing something also expresses a state so this feature was re-analyzed as part of the verbal domain and conflating little b and aspect here which I probably shouldn't do but originally this must have become re-analyzed as a feature of little b and then if we do that we can also say in the European languages then categorize this as perfective or imperfective depending on exactly what this little b expressed so what kind of state it is or is it entry into a state then you're going to get a perfective verb or just a plain state and then it's going to be a present stem so not all of these stages are illustrated here but we basically go from a nominal suffix to a state of little b to an inchoate of little b this is what we see in Greek and then an event of little b this is also what we see in Greek when we get the passive use okay and this actually has some nice consequences because this means we can take this whole thing and also this Vedic form up there and conflate them into one which we're going to skip now because I'm out of time I also wanted to talk about verbal diminutives in German and in Germanic this is the third case study that's actually the most exciting one I think I'll just show you maybe the last slide so diminutives verbs in German are very productive they are formed with these L suffix that seems to be the same that we have nominal diminutives so the idea is here that we originally form a nominal diminutive with these suffix and verbalize it with a verbal category changing head and then this feature becomes re-analyzed as part of the verbal domain and interestingly diminutive in the verbal domain seems to mean iterativity so these things are plurational iterative activity verbs okay so to conclude very quickly so this case study here is that we see regularities in the way verbal derivational morphology develops surprisingly often we started with something nominal that becomes re-analyzed in the verbal domain we go from selecting nominal heads to selecting verbal heads we very often go from stative to eventive so adjectival state, verbal state, event and there are good pathological parallels for all these developments especially the development of adjectival morphology to participle morphology as Hans already mentioned has a lot of parallels and I think when we start to generalize and formalize the morphosyntactic features of these formations we can identify we can better identify regularities right so we see possessive means state when it's re-analyzed into the verbal domain and so on and this works even for developments that do not display traditional grammaticalization characteristics so most of these things I think would probably not be analyzed as grammaticalization you don't go from less grammatical to more grammatical or from lexical category to functional category that doesn't seem to be happening but there's still a re-analysis going on and I think we can find out more interesting things about that thank you