Scope of generic and iterative verbs in Czech experimental evidence Mojmír Dočekal 21/10/2024 Heinrich Heine University Düsseldorf pluractionality and genericity in Czech a. chyt-nou-t rybu semelfactive 'to catch a fish once' b. chyt-a-t rybu iterative 'to catch fish more than once' c. chyt-áva-t rybu generic 'to catch fish regularly' this talk: experimental evidence for different scopes of genericity and iterativity Outline 1. Intro 2. Experiment 3. Theoretical implications joint work with Anna Woideovä Theoretical background • standard assumptions: the denotation of verbs is lexically pluralized (*): Krifka (l992),Kratzer (2007) • English arrive denotes the set of all singular and plural events • first approximation (non-formally stated in the traditional grammar: Kopečný (1962)) • semelfactivity: the verb denotes a single event • iterativity + genericity: the verb denotes a set of plural events • the debate in in 60s: derivational or inflectional morphology • agreement in the traditional grammars: one of the aktionsarts (next to ingresives, delimitatives, distributives, etc.) 4 Slavic languages • in Czech (and Slovak), both iterativity and genericity are productive, unlike in Russian, Polish, or other Slavic languages: Nubier (2017) • the traditional grammars don't distinguish between iterativity and genericity, though (umbrella term: iterativity) - Nubier (2017), Kosek (2014) • Filip (2017) (a.o.): genericity is a separate category clear differences: • iterative verbs: can be used in the present tense with the pure pluractional meaning • generic verbs: not Petr just now catchesIT.3SG/catchesGEN.3SG fishes 'Petr is catching fishes right now.' corpus evidence: pure imperfective (and also iterative) verb delat is the 120th most frequent verb in the Czech National Corpus, while the generic verb delavat is the 5147th most frequent verb (similarly for mluvit vs. mluvivat) Petr právě teď ryby. • diachronic evidence: in Diakorp v6 regex search for a generic version of verbs delavat 'to do regularly' and hubivat 'to kill regularly' yields 6 hits, the earliest from 1573 • unlike that the imperfective/iterative versions delat/hubit yields 576 hits starting in 1350 with around 50 hits from the beginning of 15th century and earlier • good evidence for: iterativity is more basic and less marked than genericity • genericity is more specific and less frequent Formal semantics of pluractionality: basic assumptions • operators vs. filters for event plurality: (3) a. operators: {a,6,c} —{a,b,c,a®b,b®c,a®c,a®b®c\ b. filters: {a,b,c,a@b,b@c,a@c,a@b@c\ —>► {a0b, b0 c, a 0 c, a 0 6 0 c} • operators (binominal/adverbial each, e.g.) sum together events: Zimmermann et al. (2002), Champollion (2016) • filters (pluractionals, distributive/dependent numerals) restrict the plural set of events: Lasersohn (2013), Cable (2014) Kuhn (2017), Kuhn and Aristodemo (2017), Kuhn (2019) 8 One distinction between operators and filters • operators can have wide or narrow scope with respect to plain indefinites: (4) Every techer examined one student. V > 3/3 > V • but filters allow only a narrow scope: (5) JEAN ONE WORD FORGET-rep. 3 > V 'Jean forgot one word repeatedly' Kuhn and Aristodemo (2017) • pre-experimental intuition: Czech generic verbs allow both narrow and wide scope w.r.t. indefinites • Czech iterative verbs allow only narrow scope w.r.t. indefinites • if true, generic verbs should be more similar to operators than filters • and iterative verbs should be more similar to filters than operators 10 • that would fit well with Filip's analysis of Czech generic verbs as dyadic operators (quasi-universal force with obligatory exceptions) • while Czech iterative verbs would be treated as filters 11 Research question (6) Do Czech generic verbs allow wide scope w.r.t. indefinites? a. Do they differ from Czech iterative verbs in this respect? • positive answer: empirical evidence for treating generic verbs operators and iterative verbs as filters • 1x3 design: 1 factor with 3 levels (7) a. Petr každý den chyt-nu-1 jednu rybu. Petr every day catch-SEM.3SG-PAST one fish 'Petr caught one fish every day' perf b. Petr chyt-a-1 jednu rybu. Petr catch-ITER.3SG-PAST one fish 'Peter caught one fish (repeatedly)' plur c. Petr chyt-áva-1 jednu rybu. Petr catch-GEN.3SG-PAST one fish 'Peter caught one fish (regularly)' gen 13 acceptability judgment task with 3 conditions: in a context strongly favoring narrow scope of the indefinite: Day Fish Monday Salmon Tuesday Trout Wednesday Bass Thursday Catfish Friday Tuna Saturday Cod Sunday Mackerel 14 • expectations (if generics are not iteratives): Condition Rating perf good plur the worst gen worse 15 • experiment run online on L-rex: Starschenko and Wierzba (2024) • 118 participants (without any compensation) • 3 conditions: 9 items (1x3 Latin square) • 9 fillers (bad: wide scope out of islands, good: wide scope out of non-islands) - again acceptability judgment task with context • 1 to 7 Likert scale (1 = the worst, 7 = the best) • 20 participants excluded (criterium: more than 4 points mean difference between good and bad fillers) • 98 participants in the final analysis 16 Descriptive statistics gen perl plui imperl Condition Figure 1: Barpiotwith standard errors 17 Figure 2: Distribution of ratings Condition Mean Median SE gen 3.11 2 0.14 pert 6.52 7 0.08 piur/imperf 2.32 1 0.12 19 Inferential statistics • Bayesian hierarchical model (in rstanarm: Goodrich et al. (2023)) with one fixed effect (condition) and random intercepts for participants and items • The full random effects model didn't converge • the model was run with default priors and 4 chains with 2000 iterations each (default) • the condition plur/imperf was treated as a reference level • plur/imperf: part of the verbs were pure iterative, part ambiguous between imperfective and iterative (but statistically, the parts were not credibly different) 20 % in Parameter Median 95% CI ROPE ROPE BF (Intercept) 3.09 [2.78,3.42] [-0.26,0.26] 0% 8.39e+14 conditionperf 3.43 [3.14,3.70] [-0.26,0.26] 0% 4.71e+21 conditionplur/imperf -0.77 [-1.05,-0.50] [-0.26,0.26] 0% 365.86 22 Discussion • the differences between the baseline and the other two conditions are credibly different: 1. perfective semelfactives are 3.5 points better than generic verbs • Bayes factor in favor of the existence of the difference: 4.71e+21 (extreme evidence) 2. iteratives are 0.8 points worse than generic verbs • Bayes factor in favor of the existence of the difference: 366 (also extreme evidence) • in traditional terms, the probability of zero hypothesis (no difference between iteratives and generics) is extremely low 23 Theoretical implications • answers to the research questions: (8) Do Czech generic verbs allow wide scope w.r.t. indefinites? a. Do they differ from Czech iterative verbs in this respect? • Czech generic verbs allow wide scope w.r.t. indefinites • they differ from Czech iterative verbs: the latter allow only narrow scope w.r.t. indefinites 24 First steps to formalization • iteratives are filters that restrict the set of events to pluralities • they require distribution across time (not participants): (9) Petr kych-a-1. Petr sneeze-ITER.3SG-PAST 'Peter sneezed (repeatedly)' • and as filters they unable to pluralize/make sum of events —> the cardinality of the indefinite requires the denotation of the indefinite to be singular 25 in this respect (distribution over time), they seem to differ from Czech dependent numerals (also filters), which require distribution across participants: {#Jeden tým/Oba týmy} získal(y) one team.SG.NOMboth teams.PL.NOM get.PL.PAST po dvou bodech. after two points.PL.LOC 'One team/Both teams got two points each' • generic verbs are operators that sum together events • in this respect they are close to universal quantifiers • the sum of events allows the multiplicity of the indefinite's denotation • only the atomic sub-events retain their singular denotation 27 Pieces of formalization • neo-Davidsonian event semantics: Champollion (2015) • thematic roles as constituents with NPs: (11) [onefishTh]=\V\e[V(e)Atheme(e) G lfish}A\theme(e) i] • lexical denotation of verbs is inherently pluralized: Krifka (1992), Kratzer (2007) 28 Filters the filters are interpreted as cardinality requirements: Kuhn and Aristodemo (2017), Kuhn (2019) the filter distrbuting over times (after Kuhn and Aristodemo (2017)): (12) [iterative] = XVXe[V(e) A 3e\ e" < e[r(e/) ^ r(e/f) input: verb denotation V output: the set of V-ing events with at least two events with distinct times 29 Petr Ag (vt, vt) caught vt) iter (vt, vt) one fish Th (vt, vt) (13) 3e[*caught(e) A 3e',e" < e[...] ... \theme(e)\ = 1] formalization of the conditions iter/imperf from Experiment the event is checked for a plurality (at least two events with distinct times), but the cardinality of the theme clashes with the context, forcing the dependency of the indefinite on days/events Operators • operators are interpreted as summing together events (again after Kuhn and Aristodemo (2017)) • Kuhn & Aristodemo's definition shortened and adapted: • quantification over times: as temporal location theta-role (14) [each TempLoc]=Al/Ae[3i?[e = 0 E]Vx[atom(x) —>► 3!e'[e' G E A V{e') A TempLoc{e') = x}} A Ve'[...]] • input: verb denotation V • output: the set of events (sum) e = 0 E\ for each atomic time there is a subevent e' of V-ing and vice versa 31 (vt, t) Each Day (vt, vt) Peter Ag (vt, vt) catch vt) OneFish Th (vt, vt) (15) 3e[3£[e = 0 E A \/x[atom(x) ->• 3\e'[\theme(e') 1-]]]] formalization of the conditions perf from Experiment explains the narrow scope of the indefinite: the sum of events has subevents in which each has the cardinality 1, but their sum is plural compatible with the context • the verb should be singularized Generic operators let's assume some dyadic generic operator (Filip's universal force with exceptions) in the template of Kuhn & Aristodemo's each operator: (16) [Gen]=3e[3£[e = 0 E A Gen similar to each operator but with a more nuanced quantification force 33 (vt) (vt, vt) (17) 3e[3E[e = 0£AGen[...][...][ • the formalization of the condition gen from Experiment • in restrictor, the subjects had to fill in the atoms (days of the week) pragmatically • that can be one of the reasons why the generic verbs were rated lower than the perfective verbs 34 Interim summary • the different scopal properties of generic and iterative verbs (w.r.t. indefinites) can be captured by the different formal semantics of the generic and iterative morphology • in a nutshell: generic verbs are operators (summing together events), while iterative verbs are filters (restricting the set of events) • overt operators (quantifiers over time) were accepted best in the context (condition perf) • gen was accepted worse (possible reason: pragmatical filling in of the restrictor) • iter was accepted the worst: the cardinality of the theme clashed with the context 35 Further implications Compatibility with the universal quantifiers • both generic verbs and iteratives are compatible with the universal quantifiers: (18) Petr každý den Petr every day {chyt-a-1 / chyt-áva-1} ryby. catch-ITER.3SG-PAST/catch-GEN.3SG-PAST fishes 'Peter caught fish every day (repeatedly/regularly)' • the iter operators are in the scope of the universal quantifier, but they can be interpreted vacuously • the cardinality check of the filter seems to be vacuous in this case 36 Kuhn & Aristodemo's solution: • scopable plurality (for pluractionals "-alt" in sign language): Kuhn and Aristodemo (2017) • the scope of gen in Czech would also seem to require access to the whole TP (not only V) • if true, then the difference between gen and iter is not configurational but interpretational 37 Prediction 1 • the interaction of generic verbs with quantifiers over time should be different from the interaction of iteratives with quantifiers over time • intuitions point in that direction but need further testing (19) V těch letech vždycky ráno, maminka líbala in those years always morning, mother kissed-Iter své dvě děti. her two children 'In those years, every morning, the mother kissed her two children (repeatedly)' (20) V těch letech vždycky ráno, maminka líbávala in those years always morning, mother kissed-Gen své dvě děti. her two children 'In those years, every morning, the mother kissed her two children (regularly each more than once)' Prediction 2 • sum operation: cumulative readings of generic verbs (analogous to the cumulative readings of universal quantifiers) • if in the scope of another plurality expression: (21) vs. (22) from Haslinger and Schmitt (2018) • if it does exist, it can complicate the previous prediction (21) Every girl in this town fed (the) two dogs. a. SCENARIO: Ada fed Carl and Dean. Bea fed Carl and Dean, true b. SCENARIO: Ada fed Carl. Bea fed Dean. false (22) The two girls in this town fed (the) two dogs. a. SCENARIO: Ada fed Carl and Dean. Bea fed Carl and Dean, true b. SCENARIO: Ada fed Carl. Bea fed Dean. true 39 first attempt to test the prediction: PLUR > ITER vs. ITER > PLUR (23) a. Petr a Karel skák-a-li (každý Petr and Karel jump-ITER.3PL-PAST (every jednou), once) 'Peterand Karel jumped (each once).' b. Maminka líb-a-la Marušku a mother kiss-ITER.3SG-PAST Maruška and Honzíka (každého jednou). Honzík (each once) 'Mother kissed Maruška and Honzík (each once).' • complicated by the extremely weak truth conditions of the iteratives 40 Prediction 3 • homogeneity: according to Kriz (2017), Russian dependent numerals are interpreted as homogenous (unlike Hungarian dependent numerals): (24) The girls danced. a. true iff all the girls danced. b. false iff none did. c. undef. iff some, but not all did. 41 Mal'ciki vypili po butylke. boys drank PO bottle The boys each drank a bottle' if true (intuitions are not clear to me), iterative verbs should be more similar to Czech dependent numerals (homogenous) than generic verbs Open questions • exact nature of the iterativity: distribution only over times? • relation of the interpretation to the morphosyntax: • the scope of gen should be semantically higher than iterative • because of restrictor and scope • but scope of the iterative is expected to be high according to Kuhn and Aristodemo (2017) 43 Thank you for your attention! 44 References i Cable, Seth. 2014. "Distributive Numerals and Distance Distributivity in Tlingit (and Beyond)." language, 562-606. Champollion, Lucas. 2015. "The Interaction of Compositional Semantics and Event Semantics." Linguistics and Philosophy 38: 31-66. ---. 2016. "Overt Distributivity in Algebraic Event Semantics." Semantics and Pragmatics 9: 16-11. Filip, Hana. 2017. "Genericity and Habituality." Unpublished manuscript. Düsseldorf: Heinrich Heine University Düsseldorf. Goodrich, Ben, Jonah Gabry, Imad Ali, and Sam Brilleman. 2023. "Rstanarm: Bayesian Applied Regression Modeling via Stan." https://mc-stan.org/rstanarm/. 45 References i i Haslinger, Nina, and Viola Schmitt. 2018. "Scope-Related Cumulativity Asymmetries and Cumulative Composition." In Semantics and Linguistic Theory, 197-216. Kopečný, František. 1962. Slovesný Vid v Češtině. Praha: Nakladatelství Československé akademie věd. Kosek, Pavel. 2014. Historická Mluvnice češtiny. Masarykova univerzita Brno. Kratzer, Angelika. 2007. "On the Plurality of Verbs." Event Structures in Linguistic Form and Interpretation 269 (300): 2. Krifka, Manfred. 1992. "Some Remarks on Polarity Items." Semantic Universals and Universal Semantics, 150-89. 46 References iii Kriz, Manuel. 2017. "In Soviet Russia, Alcohol Is Dependent on You." Festschrift for Martin Prinzhorn. Wiener Linguistische Gazette (WLG) 82: 173-80. Kuhn, Jeremy. 2017. "Dependent Indefinites: The View from Sign Language." Journal of Semantics 34 (3): 407-46. ---. 2019. "Pluractionality and Distributive Numerals." Language and Linguistics Compass 13 (2): e12309. Kuhn, Jeremy, and Valentina Aristodemo. 2017. "Pluractionality, Iconicity, and Scope in French Sign Language." Semantics and Pragmatics 10 (6): 1-49. Lasersohn, Peter. 2013. Plurality, Conjunction and Events. Vol. 55. Springer Science & Business Media. Nubler, Norbert. 2017. "Iterativnost." In. 47 References iv Starschenko, Alexej, and Marta Wierzba. 2024. "L-Rex Linguistic rating experiments [software], version 1.0.3." https://github.com/2e2a/i-rex/. Zimmermann, Malte et al. 2002. Boys Buying Two Sausages Each: On the Syntax and Semantics of Distance-Distributivity. Vol. 62. LOT Utrecht. 48