Schedule | Programme

 Since approximately 2016, Ash Asudeh and I have been developing a model called Lexical Realizational Functional Grammar, which divorces Distributed Morphology (as a theory of morphology) from Minimalist Syntax and instead marries it to Lexical Functional Grammar.  The chief gain for DM from this unlikely marriage is that LRFG can use the tools of DM applied to the multi-modular framework of LFG to successfully model all the complex interfaces required for complete modeling of exponence.  In this talk, I will briefly overview the motivations for LRFG and what key principles of DM it maintains such that it remains a "daughter" of DM.  I will then showcase the tools LRFG provides with our model of the English Comparative. The English comparative -er is a particular challenge for contemporary morphological analysis (see, among others, Lindquist 2000, Mondorf 2003, 2007, Hilpert 2008, Matushansky 2013, Dunbar and Wellwood 2016).  Embick 2007 models what we consider to be the standard DM approach.

1. The comparative and superlative in English are in an ABB suppletion relationship (good, better, best; bad, worse, worst), which strongly suggests a containment relationship (Bobaljik 2012). This in turn suggests that -er and -est are in competition with each other; i.e., there is a common set of features that is a subset of the features they expone (e.g., COMP +, given Bobaljik 2012) and they expone a shared syntactic position.

2. Additionally, more and -er are in (mostly) complementary distribution, suggesting that they are allomorphs. This again suggests that they are in competition with each other for the same position of exponence.

3. The complementary of -er and more seems to be such that phonologically-conditioned: monosyllabic stems get -er and trisyllabic+ stems get more (*beautifuller vs prettier), so this competition has to be sensitive to phonology.

4. The blocking of -er is not only triggered by phonology, but also by syntactic triggers, as in (1), and semantic triggers, as in (2).

1. the adornment is more pretty than practical ̸= the adornment is prettier than practical

2. De’Aaron Fox was more clutch/*clutcher than any other player last year

5. Finally, sometimes pure complementarity fails and both more and -er are licit (I am even madder and I am even more mad), but it does so in semantically predictable ways (in contrast to true optionality).

The net of all these properties is that the appearance of -er is the result of a complex competition involving three competitors (more, -er, -est) and phonological, semantic, and syntactic conditions restricting their distributions. The complex nature of this competition, which draws on mappings to multiple distinct representations, lends itself to a constraint-based, modular framework such as LRFG. Also, given the complexity of the competitions, -er represents the ideal morphological phenomenon to showcase all the different aspects of analysis in LRFG and to provide the basis for a ‘soup-to-nuts’ demonstration of the framework.

This talk looks at verb doubling phenomena from an architectural perspective, focusing on cases of doubling apparently motivated as “repairs” in Ingush (Nakh-Dagestanian), and comparing potential analyses in two realizational morphological frameworks: Minimalist Distributed Morphology (DM) and Lexical-Realizational Functional Grammar (LRFG). 

Verb doubling occurs in Ingush as a form of emphasis (Nichols 2011: 374); it differs from ordinary reduplication in that the two copies of the verb are separated by a clitic or negative particle, and in that the two copies of the verb can be realized via different suppletive allomorphs. This can be seen in (1), where the verb lezh (`die’, a suppletive form used in clause-chaining environments) is doubled as vala (the default form of the verb stem).

(1) Vala cy lezh     vyssaav     yz?

die NEG die.CONVERB             V.remain.NW.V             3s

“He didn't die?” ('He remained not dying?) (Nichols 2011: 374 ex. 223)

Doubling also occurs in Ingush to provide the clause chaining clitic =ʔa with a host—but in this environment doubling has a “last resort” character, licit only in chained clauses that lack any other potential host (Peterson 2001). 

(2) Je kinashjka     diisha='a     diishaa

this book             D.RED=&         D.read.CONVERB 

“(s/he) read this book and …” (Nichols 2011: 374 ex. 219)

As both Bjorkman (2022) and Asudeh et al (2024) observe, doubling of the type found in Ingush is naturally accounted for in a realizational theory of morphology, where there is a level of representation (syntax) in which the two copies are indeed identical. But the architectural differences between DM and LRFG lead to important differences between the analyses they make possible: in DM doubling of the type seen in Ingush is best accounted for as part of post-syntactic linearization, while in the non-derivational theory of LRFG it is best modelled syntax-internally. This talk compares the two analyses, using this as a framework for discussing the ways in which DM and LRFG can be understood as notational variants of one another, and the ways in which they make meaningfully different predictions.

Root and pattern morphology is a linguistic phenomenon where the lexical root of a word appears to consist just of consonants (the “consonantal root”), while functional content is contributed by a series of vowels (the “vocalic melody”), which appear to interleave with the root consonants in a particular skeletal sequence of Cs and Vs (the “template”). Particularly in the Semitic context, root and pattern morphology—with its myriad quirks at every level of description—has long puzzled phonologists, morphologists, syntacticians, and semanticists alike. Some interconnected questions that arise include:

i. Is the root morpheme purely consonantal, or is it stored with an (unmarked) vocalic melody and template?

ii. Putting the root aside, is the template (Cs/Vs) separable from the vocalic melody, and if so, is the template on its own a linguistic object (i.e., a morpheme), or is it epiphenomenal, arising from purely phonological constraints?

iii. How exactly does the vocalic melody end up interleaved with the root, and is this a matter purely of phonology, or does it belong to the domain of morphology?

iv. Is there syntactic and semantic regularity to be found in the melody and/or template, or do these correspond more to something like theme vowels, stemming from arbitrary lexical classes?

While the above questions have all received a fair bit of attention in the literature, a question that has received rather little attention is the following:

v. Is the vocalic melody itself decomposable into multiple morphemes? In other words, do the first vowel and the second vowel correspond to two different morphemes?

Although a logical further decomposition of root and pattern morphology, nevertheless this move—decomposing the melody itself into multiple morphemes—has only been considered in one work to my knowledge, Faust 2012. Most work on root and pattern morphology simply assumes the vocalic melody is not decomposable into multiple morphemes (e.g., McCarthy 1981, Aronoff 1994, Doron 2003, Arad 2005, Borer 2013, Kastner 2019, 2020). 

In this talk, I use a case study of Modern Hebrew verbs to explore the (perhaps very out-there) idea that in fact the vocalic melody IS decomposable, such that the multiple vowels in a given melody each correspond to different morphemes. Building on and extending ideas in Faust 2012, I consider the empirical motivation for this decomposi;on and propose a particular theoretical implementation, based in part on Kastner’s (2019, 2020) syntactic decomposition of Hebrew verbs. I then consider the empirical and theoretical consequences of this implementation, in particular with respect to the possibility that infixation plays a central role in root and pattern morphology.

This work is in its very early stages, and all feedback is welcome.

Akuzipik (aka St. Lawrence Island Yupik ISO 639-3: ess) is an endangered language spoken primarily on St. Lawrence Island, Alaska by fewer than 1000 people. Akuzipik is on the Yupik branch of the Inuit-Yupik-Unangan language family, a family populated by languages that have variously been called ‘tenseless’. In this talk I present in-progress semantic fieldwork on several strategies for temporal reference in the language and discuss the difficulty of accounting for the data within previous accounts. Akuzipik is polysynthetic and exclusively suffixing. Unmarked stative verbs and, in some situations, eventive verbs, have a present interpretation. A number of overt morphemes can be involved in past reference, including -ma (~ completive aspect), -kaa (~ distant past), and -aqe (~ a stativity transposer that yields past interpretation on stative verbs); zero-marked eventive verbs are also interpretable as past-referring. Several overt morphemes lead to future interpretations, including -lleq (~ future), -naagh (~ distant? future), and -naqe (~ prospective).

I further discuss this fieldwork within the context of the ongoing revitalization efforts on St. Lawrence Island and my work with speakers of the language as community linguists. Many of the temporal contrasts under scrutiny here are both under-documented and fairly low-frequency, making them easy targets for attrition in a language endangerment situation. Finally, I briefly describe how I am working to integrate active fieldwork by students into my program’s core courses.

Gast, Putnam, and Jóhannsdóttir (2025) discuss two ways of describing events carried out ‘in excess’ in Icelandic. Consider examples (1) and (2):

(1)  Sara            bakaði     yfir          sig.                     (2)  Sara      of-bakaði          köku-na.

       Sara             baked     OVER     ANPH                       Sara     OVER-baked   cake-DEF

      ‘Sara baked too much.’                                               ‘Sara overbaked the cake.’

In example (1), the Agent (Sara) is portrayed as having carried out baking events in excess. The implication is that she is exhausted or fed up with baking. This is indicated by the prepositional phrase yfir sig, lit. ‘over her/himself’. By contrast, in (2), where the verb carries an of-prefix, the baking event is portrayed as excessive relative to the cake, not the Agent. Gast et al. (2025) argue that both types of expressions share a common core as they express excess, and that the difference between them only concerns the standard of comparison, which is the Agent in the case of yfir sig, and the Theme in the case of the of-prefix.

In this presentation, we build upon the analysis put forth by Gast et al. (2025) and extend their treatment of excessive se mantics to the domain of adjectives. Like transitive and unergative predicates, Experiencer adjectives can combine with yfir sig. See the examples in (3):

(3)     a. yfir sig ánægður ‘overly happy’

         b. yfir sig ástfanginn ‘overly in love’

         c. yfir sig gáttaður ‘overly astonished’

         d. yfir sig heillaður ‘overly fascinated’

Although the excessive adjectives in (3) are related to the excessive predicates in (1), they differ from them in a crucial way: they do not carry a negative connotation, but rather express a high degree on the scale established by the relevant predicate.

Analysis. We offer a semantic and syntactic analysis of yfir sig combining with adjectives. We adopt a decompositional, late-insertion approach to the syntax of lexical items (i.e., Distributed Morphology, DM). We assume that as an adjectival modifier yfir sig exhibits the same syntactic structure as when used as a verbal modifier, see the structure in (4) corresponding to (3-a).

(4)     [sig [ yfir ánægður ]]

We argue that the difference in interpretation stems from the different types of scales associated with adjectives (linear) and events (Gaussian). One attribute that unifies these deverbal excessive adjectives in Icelandic is the fact that they are formed from state/psych-predicates (e.g., being happy, being in love, being fascinated, etc.) (Rothmayr, 2009). Syntactically, we advocate for the presence of a pro argument that appears in Spec,vP, which is responsible for binding the reflexive sig (Landau, 2010). Finally, we provide preliminary thoughts on the lack of similar types of adjectives in related languages (e.g., English & German). Mitrović  and Panagiotidis (2020) challenge the notion of a primitive status as functional heads that exclusively function as ‘adjectivizers’. Rather, they advocate for a bicategorial structure consisting of (i) a √root, (ii) a nominal domain, and (iii) a verbal domain (in this order of dominance for Indo-European adjectives). We adopt this bicategorial approach to the decomposition of adjectives, interpreting √roots as properties, and categories as kinds (Mitrović, 2022).

References

Gast, V., Putnam, M. T., & Jóhannsdóttir, K. (2025). Excessive events: The syntax and semantics of OVER-modification in Icelandic. Glossa, 10(1), 1–33.

Landau, I. (2010). The locative syntax of experiencers. Cambridge, MA: MIT Press.

Mitrović, M. (2022). First-phase semantics. In P. Panagiotidis & M. Mitrović (Eds.), A0- The lexical status of adjectives (pp. 257–292). Amsterdam: John Benjamins.

Mitrović, M., & Panagiotidis, P. (2020). Adjectives exist, adjectivisers do not: a bicategorical typology. Glossa, 5(1), 1-28.

Rothmayr, A. (2009). The structure of stative verbs. Amsterdam: John Benjamins.

In this talk, I discuss the structure and interpretation of noun-adjective (N+A) compounds like razor sharp, with a particular focus on how they condition allosemy on roots and functional heads. N+A compounds are understudied compared to other kinds of compounds, but turn out to exhibit a mix of compositional and idiosyncratic characteristics, displaying both productivity and apparent lexicalization. Taking a Distributed Morphology (DM) perspective, I argue that N+A compounds are actually in some sense not true compounds but rather involve a phrasal structure wherein a categorized noun is introduced in a DegP (degree phrase) structure. This accounts for various properties of N+A compounds, such as their resistance to nominalization (*razor sharpness, *rock hardness), their phrasal stress (rather than compound stress), and constraints on modification (?? very razor sharp, *razor sharper). Moreover, the interpretation of this structure determines the semantic contribution of the roots, but the roots themselves do not interact with each other directly. I focus primarily on elative/intensifying compounds—such as bone dry, razor sharp, and stone cold. I show that the “high degree” interpretation of the adjective is not predictable from the noun or adjective alone, and propose that it arises from an alloseme of the Deg head. Second, and independently, the noun within these compounds undergoes a process of semantic coercion or metonymy, which allows it to bear a contextually-determined relation to a degree on the relevant scale. I then begin to extend my account to other N+A compounds, specifically measurement-based compounds (knee deep, nationwide), color compounds (ruby red, baby blue), and more idiosyncratic cases (pitch black, brand new). The idiosyncratic cases will follow directly from the compositional account, while the other two differ only in the alloseme of Deg that is triggered in the context of those adjectives. More broadly, this study of N+A compounds shows how cases of apparent idiosyncratic interaction between roots actually stems from a fairly ordinary kind of allosemy that is familiar from the verbal and nominal domains. Once this is recognized, it paves the way for a more precise understanding of the locality conditions on allosemy more generally, in a way that can be tested more widely in other kinds of compounds, and indeed other kinds of syntactic structures altogether.

Existing analyses of singular they (e.g. Bjorkman 2017, Conrod 2019, Konnelly & Cowper 2020) don’t provide an account of the fact that it obligatorily triggers plural verbal agreement, despite having singular reference (e.g. Kelly(i) thinks they(i) {deserve/*deserves} first place). These accounts claim that the plurality of singular they is just a matter of exponence, due to underspecification. The agreement facts, however, point to singular they being morphosyntactically plural. We propose that singular they belongs to a class of pronouns we term townhouse pronouns (or simply townhouses), which include editorial we and nurse we, among others, all of which display the same type of verbal agreement mismatch. Under our analysis, what's special about singular they and other townhouses has to do with pronouns and their structure, not with gender per se. More specifically, we provide a unified analysis of townhouses and ordinary pronouns:  Like ordinary pronouns (Postal 1966, Elbourne 2001), townhouses are D heads with a null complement, but unlike them, their complement is a DP. In the specific case of singular they, the nominal has an outer plural shell DP headed by overt they, and a covert inner singular core DP headed by pron[-g], an otherwise ineffable non-masculine, non-feminine, animate pronoun. We provide evidence for all core aspects of the proposal. Finally, we account for restrictions on possible townhouses that may shed light on the inventory and representation of phi-features, based on the hypothesis that the phi-features of the core DP must be a subset of the phi-features of the shell DP. If correct, the analysis provides evidence for a mixed theory of phi-features in which some are privative (e.g. [pl] vs. [  ] for plural vs. singular) and others are binary (e.g. [+/-participant] for 1st/2nd vs. 3rd person).

The language Jarawara (Arawá, spoken in Brazil) exhibits passive-like properties in what is referred to as the ‘O-Construction’, in contrast to the active ‘A-Construction’ (Dixon 2004, Adamson and Kramer to appear). The two constructions differ in several ways, including word order preference; which argument verbal agreement indexes; and prefixation of hi- to the verb. Strikingly, however, O-Constructions exhibit varied behavior across different person combinations of arguments in: i) the presence of the hi- prefix, ii) the choice of controller for T and C agreement, iii) whether there is accusative marking on the internal argument, and iv) whether T and C must be expressed overtly. We propose that O-Constructions are uniformly ‘non-canonical’ passives, in that they promote the internal argument of a transitive but still project an external argument (cf. Legate 2021), and that their person-based variation in agreement and overtness stems from the following: i) the T probe in the O-Construction searches specifically for third-person arguments (cf. Grishin 2023 for Algonquian), and ii) speech act participants in the language need to be licensed via relations with heads in the clausal spine (Kalin 2018 on DOM in Senaya; cf. licensing accounts of PCC effects such as Bejar and Rezac 2003). Our analysis is supported by cross-linguistic evidence within the Arawá family, and our findings support both the view that a subset of marked nominals can require special grammatical licensing and that ‘third person’ is not the absence of person features altogether.

In this talk, I lay out the case for representing grammatical gender with diacritic — but nevertheless syntactically active — features. In essence, I treat gender like declension class at a distance. I propose a learning approach in which language acquirers posit features that are nothing more than diacritic triggers to specific agreement exponents. For example, consider the Italian singular (i.e. [+ATOMIC]) agreement VI -a found on the definite article and adjective in (1a), pretheoretically labeled as feminine. Most featural approaches to gender posit a unitary representation (in (2a), it is the [+FEM] feature) — that is constant across all other agreement features (in Italian, only number). In contrast, I posit that the singular VI insertion rule in (2b) labeled a_2 contains a reference to itself in its insertion context; likewise for the plural VI insertion rule labeled e_2. Learners then bundle these syntactic diacritic features on the nominalizing head n; the feminine gender of Italian, then, would simply be the combination of the a_2 diacritic feature and the e_2 diacritic feature. This recalls the usually separate treatment of, e.g., singular vs. plural noun class agreement in descriptions of Bantu (as in the traditional Bleek-Meinhoff numbering; see also Msaka 2019; cf. Carstens 1991) and other Niger-Congo languages (McLaughlin 1997; cf. Babou & Loporcaro 2016).

(1) a.      l-a                                      virtù                            platonic-a 

DEF-SG.feminine   virtue(feminine)   platonic-SG.feminine

                 ‘the platonic virtue’

      b.       l-e                                    virtù                            platonich-e

DEF-PL.feminine   virtue(feminine)   platonic-PL.feminine

‘the platonic virtues’

(2) Some Italian VI insertion rules for feminine gender agreement:

      a.    gender as [±FEM] b. gender as diacritics to agreement VIs

                        (traditional account)                                   (my proposal)

Agrº: [+FEM, +ATOMIC] ↔ /-a/ a_2)          Agr°:[a_2, +ATOMIC] ↔ /-a/ 

Agrº: [+FEM, −ATOMIC] ↔ /-e/ e_2)          Agr°:[e_2, −ATOMIC] ↔ /-e/ 

Nothing featural links a gender’s value in the singular to its value in the plural; learners have to identify that pairing themselves. This is much the same way that declension class just points to language-specific inflectional allomorph exponents; the only difference is the syntactic nature of the gender diacritic representations versus the morphophonological, post-syntactic nature of the declension diacritic representations. I discuss some consequences of this representational independence of each syntactic trigger, from accounting for well-studied cases of ambigenerics (e.g., Romanian, Guébie) to novel examples of gender mismatch across multiple relative markers in Wolof.

Left edges tend to be strong across languages, by licensing more contrasts, by requiring low sonority segments, and more (Beckman 1997; Lombardi 1999; Steriade 1997, 2008; Zoll 1998). It is therefore striking that every stem-initial nasal consonant in Blackfoot deletes after a prefix. This process could be viewed as extreme lenition to zero, and it shares some characteristics with other lenition processes that increase sonority within a phonological domain (Harris 2003; Katz 2016; Keating 2006; Kingston 2008; Ségéral & Scheer 2008). For example, while nasals delete in this position, higher-sonority segments like glides do not delete, supporting an analysis where low-sonority segments are targeted for lenition. There are three problems with such a lenition analysis:

1. Lower-sonority segments obstruents do not delete. Instead, a vowel [i] is epenthesized to the left of a stem-initial obstruent, which is not predicted at all by sonority-increasing lenition.
   

2. Deletion and epenthesis neutralize contrasts,which is uncommon for this type of lenition (Katz2016). While stems may begin in obstruents, nasals, or vowels at the left edge of a word, these processes ensure that stems only begin in vowels after a prefix.
   

3. Deletion and epenthesis only affect left edge consonants, whereas sonority-increasing lenition typically affects all consonants within a phonological domain.
   

I argue that left-edge deletion and epenthesis in Blackfoot are a type of initial consonant mutation (Iosad 2010). Under my analysis, a low-sonority consonant in a morphologically prominent position (e.g., at the left edge of a morpheme) must be licensed by a prosodically prominent position (e.g., a left edge of a prosodic constituent). This is enforced via a prosodic licensing constraint. When prosodic correspondence or marked- ness constraints prevent a prosodic left edge from aligning to the morpheme, then the consonant may be subject to lenition.

In Blackfoot, this is precisely what happens: although prefixes tend to be parsed to a separate prosodic word (Elkins 2020), prefixes in Blackfoot are parsed into the same prosodic word as the following stem. There is no evidence from right edge restrictions, stress, or minimal size constraints that prefixes correspond to separate prosodic domains (Weber 2022). The processes of deletion and epenthesis remove the consonant from the left edge of the morpheme in order to avoid violations of the licensing constraint.

For the purposes of the workshop, I am mainly interested in exploring what this analysis requires for the syntax-phonology interface in Blackfoot and in general. I would also like to explore how this analysis extends to other languages with initial consonant mutation.

Norwegian makes a two-way distinction in the realization of stressed syllables, known as tonal accent (Kristoffersen 2000). Dialects vary markedly in the specific melodies associated with Accent 1 (A1) and Accent 2 (A2), yet the core distributional patterns are remarkably consistent. All monosyllabic words are A1, and A2 requires at least two syllables; however, A1 can also occur in disyllabic environments, leading to some surface contrasts like A1 1tank-en ‘the tank’ and A2 2tanke-n ‘the thought’, where “minimal pairs” often denote morphophonological differences (Morén-Duolljá 2013), here in the sizes of the stems (tank vs. tanke).

Because Norwegian tonal accent is a part of the stress system (Kristoffersen 2000), its distribution is at least partly dependent upon the output of stress calculations. Here, I follow Kristoffersen (2000:159–160ff) in assuming that stress is assigned from the end of the word (R to L), targeting the penult unless the last syllable is heavy (final stress) or extrametrical (antepenultimate stress): 2skri.ve ‘write’, 1skri.vet ‘the missive’, be.1skri.ve ‘describe’,  tra.1fikk ‘traffic’, 2hel.ve.<te> ‘hell’. I propose that rhythm, however, is calculated after stress (van der Hulst 2014) from L to R, with left-hand prominence, in Norwegian. On this view, A2 results when the heads of the stressed foot and the rhythmic foot are aligned (cf. 2skrive, 2hake, and 2helvete). When there is a misalignment (cf., tra1fikk, and be1skrive), A1 occurs. A1 will also occur when constructing a rhythmic foot is not possible, e.g., in monosyllabic words and instances like 1skriv-et, where -et DEF is affixed after both stress and rhythm are calculated.

This metrical proposal departs from analyses that rely on tones (Kristoffersen 2000), A1 specification (Wetterlin 2010), and structural relationships in the prosodic hierarchy (Morén-Duolljá 2013) to model tonal accent patterns. Rather, I show that tonal accent is predictable from an interplay between primary accent and rhythm and that instances of unpredictable tonal accent can be reduced to representations that are independently required for unpredictable stress.

References

Kristoffersen, Gjert. 2000. The phonology of Norwegian. Oxford: Oxford University Press.

Morén-Duolljá, Bruce. 2013. The prosody of Swedish underived nouns: No lexical tones required. Nordlyd 40(1): 196–248.

van der Hulst, Harry. 2014. The study of word accent and stress: Past, present, and future. In Harry van der Hulst (ed.), Word Stress: Theoretical and typological issues, 3–55. Cambridge: Cambridge University Press.

Wetterlin, Allison. 2010. Tonal accents in Norwegian: Phonology, morphology and lexical specification. Berlin: De Gruyter.

Many Romance languages exhibit a morphomic pattern dubbed PYTA (for {perfecto/pretérito} y tiempos afines—see especially Maiden 2018: Ch 4).  Spanish exhibits a striking instance of this phenomenon. No matter how irregular a Spanish verb is otherwise, it always has the same stem for both imperfect subjunctive paradigms and for the future subjunctive (which is defunct in the modern spoken language, but was in active use a few centuries ago), and this stem can be unerringly arrived at by taking the 3rd person plural preterite indicative form of the verb and subtracting the sequence -ron.  This generalization is exceptionless, but it is not directly formulable in a piece-based view of morphology.  For this reason, the PYTA morphome has been suggested to require a paradigm-based view of morphology (see for instance O’Neill 2014).  

In this talk, I will argue that the PYTA morphome, at least as it arises in Spanish, is actually two independent generalizations wearing a trench coat. One of these generalizations involves morphosyntactic conditioning of root suppletion by a feature [-Pres], the other is phonological and concerns the circumstances under which the theme vowel of 2nd and 3rd conjugation verbs diphthongizes.  Once these generalizations are independently explained (as they have to be, since they each affect different parts of the paradigm which extend beyond the claimed morphomic pattern), there is no work left for a direct statement of the morphomic pattern to do, and thus no threat to piece-based models of morphology from this quarter.

In my presentation I will provide an overview of several key questions being addressed in current theories of the syntax/morphology interface. My particular focus will be narrowed down to approaches that share two key properties: first, that there is Late Insertion at the PF interface; and second, that insertion takes place in syntactic structures.

Within approaches with these properties, there are apparent differences with respect to how much substance there is to the interface. By this I mean that- on the assumption that all such approaches require something like Vocabulary Insertion- there are differences with respect to what is posited at PF beyond this. Taking 'what happens at PF' to be morphology in the broad sense, current approaches thus differ in terms of how morphological they are about the interface between syntax and form.

The issues that I will address fall under two headings that are logically independent of one another. The first concerns the operations that might apply at PF, postsyntactically but prior to Insertion. The second concerns morphology after Insertion: what is often called morphophonology, where the question is whether the phonological grammar (broadly construed) has access to the identity of Roots, or to syntactic features. In the first part of the talk, I will provide an overview of what is at issue in each of these domains. The key questions to be considered center on how it might be possible to make empirical comparisons across approaches that make very different-looking starting assumptions, and hence have very different ideas about what might or might not be possible in principle in either of these areas. Against this background, I will then provide some specific comparisons as illustrations; time permitting these will include both 'standard' examples of morphological operations/representations (i.e. typical PF operations), as well as some new areas that I believe shed light on these questions (e.g. linear effects; interactions with ellipsis; and infixation/reduplication).