Colloquium and Events Schedule (2013-2014)
All colloquium talks are in B342 Wells Hall at 4:30pm unless otherwise noted.
Upcoming Talks and Events (2013-2014)
Our 2013-2014 colloquium series has ended! Check here in the fall for our 2014-2015 colloquium series.
Past Talks (2013-2014)
Gradient linguistic knowledge in bilingualism: From phonetics to grammar
Matt Goldrick (Northwestern University)
When: April 17
Bilinguals bring together two languages into one mind, allowing unique insight into the representation of linguistic knowledge. I'll present data from bilingual language switching experiments that suggest such interactions occur along a continuum--not just at the level of phonetic categories but also at the level of constraints on linguistic structure. This suggests that rather than forming discretely separated systems of knowledge, linguistic systems are represented within a continuous space of grammatical hypotheses. I'll outline a theoretical framework, Gradient Symbolic Computation, that allows us to formulate such hypotheses, and discuss its predictions for linguistic knowledge and language processing.
Bilinguals bring together two languages into one mind, allowing unique insight into the representation of linguistic knowledge. I'll present data from bilingual language switching experiments that suggest such interactions occur along a continuum--not just at the level of phonetic categories but also at the level of constraints on linguistic structure. This suggests that rather than forming discretely separated systems of knowledge, linguistic systems are represented within a continuous space of grammatical hypotheses. I'll outline a theoretical framework, Gradient Symbolic Computation, that allows us to formulate such hypotheses, and discuss its predictions for linguistic knowledge and language processing.
Using parsed corpora to study language history (GLEEFUL Keynote Address)
Anthony Kroch (University of Pennsylvania)
When: April 12, 2:30-4pm
Where: B342 Wells Hall
Over the past two decades, linguists with an interest in diachronic syntax have created syntactically annotated text databases, "parsed corpora," that have proved useful for the study of language change. These corpora allow us to study both the grammars of historical languages and the time course of syntactic changes. At present, we have parsed corpora of at least six languages for use in research: English, French, Icelandic, Irish, Portuguese, and Yiddish. In this talk, I will briefly discuss how such corpora are constructed and then give three examples, one from English, one from Yiddish and a third from French, revealing the sorts of information about grammar and change that can be extracted from them. The examples will show that combining grammatical analysis with the analysis of usage frequencies can provide new insights into both language history and linguistic theory.
Where: B342 Wells Hall
Over the past two decades, linguists with an interest in diachronic syntax have created syntactically annotated text databases, "parsed corpora," that have proved useful for the study of language change. These corpora allow us to study both the grammars of historical languages and the time course of syntactic changes. At present, we have parsed corpora of at least six languages for use in research: English, French, Icelandic, Irish, Portuguese, and Yiddish. In this talk, I will briefly discuss how such corpora are constructed and then give three examples, one from English, one from Yiddish and a third from French, revealing the sorts of information about grammar and change that can be extracted from them. The examples will show that combining grammatical analysis with the analysis of usage frequencies can provide new insights into both language history and linguistic theory.
Sources of information deployed during agreement attraction (MSULC Keynote Address)
Karl DeVries (University of California at Santa Cruz)
When: April 11, time TBA
Where: 310 Bessey Hall
Agreement attraction occurs when a sentence exhibiting failed subject-verb agreement is per- ceived as grammatical due to the presence of a second DP (the attractor) which matches the number features of the verb ((1) The key to the cabinets were on the table) (Bock & Miller 1991). Recent work suggests that agreement attraction is the result of interference during a content addressable memory search (Wagers et al. 2009), but it remains unclear what level of representa- tion agreement attraction is sensitive to. Crosslinguistic work on Dutch (Hartsuiker et al., 2001) and German (Hartsuiker et al., 2003) has shown that agreement attraction is insensitive to case distinctions that cannot be read off the morphology of the DP alone.
The current study examines whether agreement attraction is sensitive to non-local sources of number information by manipulating the disambiguation of a number ambiguous DP ((2) the dog that the sheep was/were following around are barking with glee). If agreement attraction is sensitive to a representation of the attractor that includes number disambiguation, we expect to see increased acceptance of sentences like (2) when sheep is disambiguated plural. I present the results of a speeded acceptability judgment task showing no such improvement. These results suggest that only number information derivable from the morphological form of the attractor is active during agreement attraction.
Where: 310 Bessey Hall
Agreement attraction occurs when a sentence exhibiting failed subject-verb agreement is per- ceived as grammatical due to the presence of a second DP (the attractor) which matches the number features of the verb ((1) The key to the cabinets were on the table) (Bock & Miller 1991). Recent work suggests that agreement attraction is the result of interference during a content addressable memory search (Wagers et al. 2009), but it remains unclear what level of representa- tion agreement attraction is sensitive to. Crosslinguistic work on Dutch (Hartsuiker et al., 2001) and German (Hartsuiker et al., 2003) has shown that agreement attraction is insensitive to case distinctions that cannot be read off the morphology of the DP alone.
The current study examines whether agreement attraction is sensitive to non-local sources of number information by manipulating the disambiguation of a number ambiguous DP ((2) the dog that the sheep was/were following around are barking with glee). If agreement attraction is sensitive to a representation of the attractor that includes number disambiguation, we expect to see increased acceptance of sentences like (2) when sheep is disambiguated plural. I present the results of a speeded acceptability judgment task showing no such improvement. These results suggest that only number information derivable from the morphological form of the attractor is active during agreement attraction.
Digital Humanities, Linguistics, and the Library
Thomas Padilla (Michigan State University)
When: April 10
Where: B243
Thomas Padilla, the new Digital Humanities, Linguistics, and Philosophy Librarian will introduce himself and discuss areas of potentially fruitful interaction between Digital Humanities and Linguistics. Thomas will also discuss resources that are available at the library to support Linguistics research. Throughout the discussion Thomas hopes to and engage faculty and students in a dialog dedicated to finding out how the library can best support Linguistics at Michigan State University moving forward.
Where: B243
Thomas Padilla, the new Digital Humanities, Linguistics, and Philosophy Librarian will introduce himself and discuss areas of potentially fruitful interaction between Digital Humanities and Linguistics. Thomas will also discuss resources that are available at the library to support Linguistics research. Throughout the discussion Thomas hopes to and engage faculty and students in a dialog dedicated to finding out how the library can best support Linguistics at Michigan State University moving forward.
Unlocking hyper-raising: lessons from syntactic optionality
Claire Halpert (University of Minnesota, Twin Cities)
When: April 3
In this talk, I argue that the Durban dialect of Zulu (Bantu) exhibits three-way syntactic optionality with subject-raising predicates. In particular, certain matrix predicates allow for optional subject raising out of an agreeing, finite embedded clause. When raising occurs, agreement between the raised subject and the matrix predicate, normally obligatory, is also optional. The presence of this type of stable syntactic optionality in a grammar poses certain challenges for syntactic theory. If two distinct grammatical constructions share an underlying argument structure and semantic interpretation, any theory that appeals exclusively to obligatory operations, constraints, or well-formedness/economy conditions to account for grammaticality patterns faces a challenge accommodating this optionality. With the Zulu optional raising constructions, optionality is not the only problem. An additional central puzzle is the question of why a subject may raise out of a finite embedded CP ("hyper-raising," Ura, 1994) in the first place. This type of movement seems to be unavailable in many languages, a fact that has been attributed to a general universal constraint on derivations (e.g. PIC, Chomsky, 2000).
I propose that it is the optionality itself (and in particular the optionality in agreement within the matrix clause) that is the key to understanding the availability of hyper-raising in Zulu. I use evidence from another optional agreement process to demonstrate that CPs are goals for agreement in Zulu. I then argue that hyper-raising is the result of the matrix predicate agreeing first with the embedded CP – obviating PIC effects along the lines of Rackowski and Richards (2005) – and subsequently with the embedded subject. This proposal for Zulu has implications for the crosslinguistic typology of raising and, more generally, for our understanding of movement out of syntactic phases. I further argue that though DP movement is constrained by information structure elsewhere in Zulu, these distinctions (predictably) do not arise with subject-raising predicates. It is the lack of a semantic difference, combined with the existence of two distinct routes to the satisfaction of an inviolable syntactic condition on agreement, that produces optionality in the hyper-raising construction. This type of approach to optional hyper-raising in Zulu departs from earlier "optional feature" proposals (e.g. Harford Perez, 1985; Ura, 1994; Zeller, 2006).
In this talk, I argue that the Durban dialect of Zulu (Bantu) exhibits three-way syntactic optionality with subject-raising predicates. In particular, certain matrix predicates allow for optional subject raising out of an agreeing, finite embedded clause. When raising occurs, agreement between the raised subject and the matrix predicate, normally obligatory, is also optional. The presence of this type of stable syntactic optionality in a grammar poses certain challenges for syntactic theory. If two distinct grammatical constructions share an underlying argument structure and semantic interpretation, any theory that appeals exclusively to obligatory operations, constraints, or well-formedness/economy conditions to account for grammaticality patterns faces a challenge accommodating this optionality. With the Zulu optional raising constructions, optionality is not the only problem. An additional central puzzle is the question of why a subject may raise out of a finite embedded CP ("hyper-raising," Ura, 1994) in the first place. This type of movement seems to be unavailable in many languages, a fact that has been attributed to a general universal constraint on derivations (e.g. PIC, Chomsky, 2000).
I propose that it is the optionality itself (and in particular the optionality in agreement within the matrix clause) that is the key to understanding the availability of hyper-raising in Zulu. I use evidence from another optional agreement process to demonstrate that CPs are goals for agreement in Zulu. I then argue that hyper-raising is the result of the matrix predicate agreeing first with the embedded CP – obviating PIC effects along the lines of Rackowski and Richards (2005) – and subsequently with the embedded subject. This proposal for Zulu has implications for the crosslinguistic typology of raising and, more generally, for our understanding of movement out of syntactic phases. I further argue that though DP movement is constrained by information structure elsewhere in Zulu, these distinctions (predictably) do not arise with subject-raising predicates. It is the lack of a semantic difference, combined with the existence of two distinct routes to the satisfaction of an inviolable syntactic condition on agreement, that produces optionality in the hyper-raising construction. This type of approach to optional hyper-raising in Zulu departs from earlier "optional feature" proposals (e.g. Harford Perez, 1985; Ura, 1994; Zeller, 2006).
Dependent indefinites and their post-suppositions
Robert Henderson (Wayne State University)
When: March 20
This talk presents an analysis of a new scope puzzle that arises through the interaction of two lesser-studied constructions, dependent indefinites and verbal pluractionality. The result is a novel account of dependent indefinites that correctly predicts their grammaticality with pluractionals by recognizing two ways of establishing the covariation they require: (i) true distributive quantifiers, and (ii) pluractional operators that structure thematic dependencies. The core insight is that both routes, while compositionally different, lead to similar output structures in Dynamic Plural Logic (DPlL) (van den Berg 1996) or is close analogs (Brasoveanu 2008; Nouwen 2003), which is what dependent indefinites constrain. The analysis not only permits a better understanding of dependent indefinites in Kaqchikel, an endangered and understudied Mayan language of highland Guatemala, but it clarifies their place in a crosslinguistic typology of similar expressions (e.g., Balusu 2006; Choe 1987; Farkas 1997a, 2002; Yanovich 2005). Along the way we produce the first description and analysis of these phenomena in Kaqchikel.
This talk presents an analysis of a new scope puzzle that arises through the interaction of two lesser-studied constructions, dependent indefinites and verbal pluractionality. The result is a novel account of dependent indefinites that correctly predicts their grammaticality with pluractionals by recognizing two ways of establishing the covariation they require: (i) true distributive quantifiers, and (ii) pluractional operators that structure thematic dependencies. The core insight is that both routes, while compositionally different, lead to similar output structures in Dynamic Plural Logic (DPlL) (van den Berg 1996) or is close analogs (Brasoveanu 2008; Nouwen 2003), which is what dependent indefinites constrain. The analysis not only permits a better understanding of dependent indefinites in Kaqchikel, an endangered and understudied Mayan language of highland Guatemala, but it clarifies their place in a crosslinguistic typology of similar expressions (e.g., Balusu 2006; Choe 1987; Farkas 1997a, 2002; Yanovich 2005). Along the way we produce the first description and analysis of these phenomena in Kaqchikel.
Gradience, allophony, and chain shifts
Aaron Dinkin (University of Toronto)
When: February 20
Where: B243 Wells Hall
Labov (2010) notes the lack of “allophonic chain shifting”—change in an environmentally-conditioned allophone of one phoneme does not cause an equivalently conditioned allophone of a neighboring phoneme to shift into the allophonically abandoned space. This can be simply explained in terms of a modular feedforward architecture of phonology, under which the entities subject to chain shifting are not the most abstract underlying representations (phonemes) but the intermediate phonological representations that are the output of discretely-acting symbolic phonological rules and the input to gradient phonetic-implementation processes. I demonstrate this by examining the role of gradiency in the Northern Cities Shift and the Southern Shift.
Almost all North American dialects of English have /æ/ higher and fronter before nasals than elsewhere, but in some there are two phonologically discrete allophones and in some /æ/ is a single continuous entity. In my Upstate New York data (Dinkin 2009), the Inland North fringe and the Hudson Valley show similar incidence of Northern Cities Shift features except that the Hudson Valley has little raising of non-prenasal /æ/. In the Hudson Valley, discrete /æ/ allophones are the norm, and so the prenasal and non-prenasal /æ/ are phonetically independent of each other, while in the Inland North mostly has continuous /æ/ distributions. This difference in the phonological structure of /æ/ accounts for the difference in the regions’ openness to the Northern Cities Shift.
In the Southern Shift, the monophthongization of /ay/, which is disfavored in pre-voiceless environments, is said to trigger the lowering of /ey/. The feedforward account of chain shifting presented here predicts that the difference between pre-voiced and pre-voiceless /ay/ is gradient, not two discrete allophones. I test that hypothesis in the Inland South data from the Atlas of North American English (Labov et al. 2006). In the Atlas, the older speakers tend to show a gradient distribution of /ay/ from more to less monophthongal as predicted, supporting the hypothesis. Younger speakers, for whom monophthongization has gone to completion, occasionally show a few isolated and phonologically distinct diphthongal tokens, apparently as corrections toward the standard.
Where: B243 Wells Hall
Labov (2010) notes the lack of “allophonic chain shifting”—change in an environmentally-conditioned allophone of one phoneme does not cause an equivalently conditioned allophone of a neighboring phoneme to shift into the allophonically abandoned space. This can be simply explained in terms of a modular feedforward architecture of phonology, under which the entities subject to chain shifting are not the most abstract underlying representations (phonemes) but the intermediate phonological representations that are the output of discretely-acting symbolic phonological rules and the input to gradient phonetic-implementation processes. I demonstrate this by examining the role of gradiency in the Northern Cities Shift and the Southern Shift.
Almost all North American dialects of English have /æ/ higher and fronter before nasals than elsewhere, but in some there are two phonologically discrete allophones and in some /æ/ is a single continuous entity. In my Upstate New York data (Dinkin 2009), the Inland North fringe and the Hudson Valley show similar incidence of Northern Cities Shift features except that the Hudson Valley has little raising of non-prenasal /æ/. In the Hudson Valley, discrete /æ/ allophones are the norm, and so the prenasal and non-prenasal /æ/ are phonetically independent of each other, while in the Inland North mostly has continuous /æ/ distributions. This difference in the phonological structure of /æ/ accounts for the difference in the regions’ openness to the Northern Cities Shift.
In the Southern Shift, the monophthongization of /ay/, which is disfavored in pre-voiceless environments, is said to trigger the lowering of /ey/. The feedforward account of chain shifting presented here predicts that the difference between pre-voiced and pre-voiceless /ay/ is gradient, not two discrete allophones. I test that hypothesis in the Inland South data from the Atlas of North American English (Labov et al. 2006). In the Atlas, the older speakers tend to show a gradient distribution of /ay/ from more to less monophthongal as predicted, supporting the hypothesis. Younger speakers, for whom monophthongization has gone to completion, occasionally show a few isolated and phonologically distinct diphthongal tokens, apparently as corrections toward the standard.
Bare nouns in Malagasy: scope and aspectual interactions
Ileana Paul (University of Western Ontario)
When: February 6
In this talk, I discuss certain properties of bare nouns in Malagasy and explore ways to connect these properties. First, I show that nouns in Malagasy have general number and that bare noun arguments are interpreted as indefinite. Next, I provide data that illustrate that bare nouns can take either wide or narrow scope. This scopal behavior is somewhat surprising given what we know about bare nouns in other languages. Finally, I consider how bare nouns are interpreted in telic contexts. Again, the result are unexpected: bare nouns retain their number neutral interpretation, even in telic contexts. This runs counter to what has been observed for other number neutral languages. I argue that bare nouns are in fact headed by a null determiner, but I show that this does not explain the aspectual data. I conclude by exploring ways in which to fully integrate all of the properties of bare nouns.
In this talk, I discuss certain properties of bare nouns in Malagasy and explore ways to connect these properties. First, I show that nouns in Malagasy have general number and that bare noun arguments are interpreted as indefinite. Next, I provide data that illustrate that bare nouns can take either wide or narrow scope. This scopal behavior is somewhat surprising given what we know about bare nouns in other languages. Finally, I consider how bare nouns are interpreted in telic contexts. Again, the result are unexpected: bare nouns retain their number neutral interpretation, even in telic contexts. This runs counter to what has been observed for other number neutral languages. I argue that bare nouns are in fact headed by a null determiner, but I show that this does not explain the aspectual data. I conclude by exploring ways in which to fully integrate all of the properties of bare nouns.
The dative in English: Variable syntax in sociolinguistic perspective
Sali Tagliamonte (University of Toronto)
When: January 30
Recent research shows that corpora provide a good model of probabilistic syntactic choices (e.g. Bresnan, Cueni, Hikitina & Baayen, 2007; Bresnan & Ford, 2010; Bresnan & Hay, 2008). Variation between double object (1a) vs. prepositional dative constructions (with to), (1b), is a showcase feature.
The choice between these constructions has been found to have highly regular statistical patterns such that inanimate, indefinite, nominal and longer arguments, e.g. some all-dressed pizza, as in (1a), tend to occur in the final complement position, adhering to a suite of tendencies described as “harmonic alignment”. Yet research has demonstrated a contrast in the ranking of these effects across varieties of English. Australian English has more prepositional datives (1b) and tolerates longer arguments in the theme (the possessed element) than American English (Bresnan & Ford 2010) and New Zealand English is more sensitive to animacy (Bresnan & Hay, 2008). Yet studies of spoken conversation reveal that long, inanimate, non-pronominal themes are rare while pronouns and short themes prevail, e.g. give it to me/give me it. What happens to the strong tendencies of harmonic alignment in these contrasting conditions?
Further, the dative alternation has been diverging in varieties of English for several centuries (e.g. Bresnan & Hay, 2008), possibly due to pressure for shorter texts (e.g. Hinrichs, 2007); however shifting language styles and colloquialization may also be at play (e.g. Leech & Smith, 1994) all suggesting external (socio-cultural) influences. Yet so far there have been no analyses of socially stratified corpora where each speaker’s age, sex, education and social class are known and can be tested.
This paper seeks to redress this gap in the literature by providing a comparative cross-variety analysis of vernacular Canadian (CdnE) and British English (BrE). The data comprise over 3200 tokens, 120 verbs and 567 individuals between 9-107. In order to provide a viable statistical model of the dozen or more linguistic predictors (e.g. Bresnan et al., 2007), social factors, individuals, and lexical items, linear mixed effect modeling is employed (Team, 2007). The results reveal that the prepositional dative is relatively rare in spoken data and the difference in frequency between the two varieties (BrE 12%; Cda 20%) is statistically significant (p = >.001). The linguistic predictors diverge as well: e.g. in BrE theme animacy is a stronger effect than recipient animacy, whereas in CdnE (as in New Zealand) recipient animacy highly favours the prepositional construction e.g. to me. Despite a clear trend in the Canadian data towards increasing use of the double object construction, standard sociolinguistic indicators (sex, education) are modest at best. These findings corroborate the earlier studies in demonstrating that speakers of different varieties show subtle variation in the effects of probabilistic syntax. In corroborating the finding that different social contexts (namely Canada vs. Britain) significantly impact syntax, they also add support for a usage-based model of grammar.
Recent research shows that corpora provide a good model of probabilistic syntactic choices (e.g. Bresnan, Cueni, Hikitina & Baayen, 2007; Bresnan & Ford, 2010; Bresnan & Hay, 2008). Variation between double object (1a) vs. prepositional dative constructions (with to), (1b), is a showcase feature.
- a. Give me some all-dressed pizza. b. Give some all-dressed pizza to me.
The choice between these constructions has been found to have highly regular statistical patterns such that inanimate, indefinite, nominal and longer arguments, e.g. some all-dressed pizza, as in (1a), tend to occur in the final complement position, adhering to a suite of tendencies described as “harmonic alignment”. Yet research has demonstrated a contrast in the ranking of these effects across varieties of English. Australian English has more prepositional datives (1b) and tolerates longer arguments in the theme (the possessed element) than American English (Bresnan & Ford 2010) and New Zealand English is more sensitive to animacy (Bresnan & Hay, 2008). Yet studies of spoken conversation reveal that long, inanimate, non-pronominal themes are rare while pronouns and short themes prevail, e.g. give it to me/give me it. What happens to the strong tendencies of harmonic alignment in these contrasting conditions?
Further, the dative alternation has been diverging in varieties of English for several centuries (e.g. Bresnan & Hay, 2008), possibly due to pressure for shorter texts (e.g. Hinrichs, 2007); however shifting language styles and colloquialization may also be at play (e.g. Leech & Smith, 1994) all suggesting external (socio-cultural) influences. Yet so far there have been no analyses of socially stratified corpora where each speaker’s age, sex, education and social class are known and can be tested.
This paper seeks to redress this gap in the literature by providing a comparative cross-variety analysis of vernacular Canadian (CdnE) and British English (BrE). The data comprise over 3200 tokens, 120 verbs and 567 individuals between 9-107. In order to provide a viable statistical model of the dozen or more linguistic predictors (e.g. Bresnan et al., 2007), social factors, individuals, and lexical items, linear mixed effect modeling is employed (Team, 2007). The results reveal that the prepositional dative is relatively rare in spoken data and the difference in frequency between the two varieties (BrE 12%; Cda 20%) is statistically significant (p = >.001). The linguistic predictors diverge as well: e.g. in BrE theme animacy is a stronger effect than recipient animacy, whereas in CdnE (as in New Zealand) recipient animacy highly favours the prepositional construction e.g. to me. Despite a clear trend in the Canadian data towards increasing use of the double object construction, standard sociolinguistic indicators (sex, education) are modest at best. These findings corroborate the earlier studies in demonstrating that speakers of different varieties show subtle variation in the effects of probabilistic syntax. In corroborating the finding that different social contexts (namely Canada vs. Britain) significantly impact syntax, they also add support for a usage-based model of grammar.
Adding Causativity: (Anti-)Causatives Across Languages and Levels of Representation
Lisa Levinson (Oakland)
When: December 5th
The representation of causativity in verbs participating in the cross-linguistic "lexical" causative alternation, in which verbs apparently alternate between intransitive (“The ice melted”) and transitive (“The sun melted the ice”) realizations, has been an ongoing source of debate. Dowty (1979) proposed that the transitive variant contains an additional CAUSE relation (see also Lakoff 1965, Pesetsky 1995, Piñón 2001, Pylkkänen 2002, Hartl 2003, Rappaport Hovav and Levin in press). On this view, only the transitive is truly "causative." However, others have argued that the intransitive also encodes causativity, and that the difference reduces to the arguments introduced (Chierchia 1989/2004, Levin and Rappaport Hovav 1995, Kallulli 2006, Alexiadou et al 2006, Schäfer 2008, Koontz-Garboden 2009). In this talk, I will argue that the truth conditional semantics is as proposed by Dowty, whereby the transitive form is more semantically complex with respect to causativity. I will present supporting evidence from both intuitions about entailments and experimental data. These data from two behavioral studies show increased reading times for transitives in the alternation in comparison with phonologically-identical intransitives, independent of the difference in the number of arguments. I will further argue, however, that in terms of the linguistic representations which derive the truth conditions, the two variants are in another sense equally complex. This will be linked to the observation that the morphology associated with the alternation does not reflect the additional causative complexity proposed by Dowty, nor a reliable indication of any directional derivation between the transitive and intransitive variants.
The representation of causativity in verbs participating in the cross-linguistic "lexical" causative alternation, in which verbs apparently alternate between intransitive (“The ice melted”) and transitive (“The sun melted the ice”) realizations, has been an ongoing source of debate. Dowty (1979) proposed that the transitive variant contains an additional CAUSE relation (see also Lakoff 1965, Pesetsky 1995, Piñón 2001, Pylkkänen 2002, Hartl 2003, Rappaport Hovav and Levin in press). On this view, only the transitive is truly "causative." However, others have argued that the intransitive also encodes causativity, and that the difference reduces to the arguments introduced (Chierchia 1989/2004, Levin and Rappaport Hovav 1995, Kallulli 2006, Alexiadou et al 2006, Schäfer 2008, Koontz-Garboden 2009). In this talk, I will argue that the truth conditional semantics is as proposed by Dowty, whereby the transitive form is more semantically complex with respect to causativity. I will present supporting evidence from both intuitions about entailments and experimental data. These data from two behavioral studies show increased reading times for transitives in the alternation in comparison with phonologically-identical intransitives, independent of the difference in the number of arguments. I will further argue, however, that in terms of the linguistic representations which derive the truth conditions, the two variants are in another sense equally complex. This will be linked to the observation that the morphology associated with the alternation does not reflect the additional causative complexity proposed by Dowty, nor a reliable indication of any directional derivation between the transitive and intransitive variants.
Extreme Modality
Paul Portner (Georgetown)
When: November 14th
This talk makes a contribution to two issues in the semantics of modality:
(i) How can the gradability properties of modal operators be compositionally modeled?
(ii) Why are there two strength levels of modal necessity, and how do they differ?
I will show that strong necessity modals (e.g., must, crucial, certain) have the properties of extreme adjectives like huge (e.g., Paradis 2001, Rett 2008, Morzycki 2012), in contrast to weak necessity modals (should, important, likely); see (1)-(2).
(1a) Susan positively/flat-out/downright must/*should call her mother.
(1b) It is positively/flat-out/downright crucial/*important for Mary to call her mother.
(1c) It is positively/flat-out/downright certain/*likely that Mary will call her mother.
(2a) Susan very much *must/should call her mother.
(2b) It is very *crucial/important for Mary to call her mother.
(2c) It is very *certain/likely that Mary will call her mother.
Focusing on priority-type modals, I will sketch a scale-based analysis which integrates Morzycki's (2012) ideas about extreme adjectives with treatments of necessity modals in premise semantics (e.g., von Fintel & Iatridou 2008, Rubinstein 2012). The idea is to construct modal degrees by manipulating the premises used in interpreting the modal (the ordering source), with the distinction between weak necessity/non-extreme modals and strong necessity/extreme modals drawn in terms of whether the premises being manipulated are up for debate or not.
This talk makes a contribution to two issues in the semantics of modality:
(i) How can the gradability properties of modal operators be compositionally modeled?
(ii) Why are there two strength levels of modal necessity, and how do they differ?
I will show that strong necessity modals (e.g., must, crucial, certain) have the properties of extreme adjectives like huge (e.g., Paradis 2001, Rett 2008, Morzycki 2012), in contrast to weak necessity modals (should, important, likely); see (1)-(2).
(1a) Susan positively/flat-out/downright must/*should call her mother.
(1b) It is positively/flat-out/downright crucial/*important for Mary to call her mother.
(1c) It is positively/flat-out/downright certain/*likely that Mary will call her mother.
(2a) Susan very much *must/should call her mother.
(2b) It is very *crucial/important for Mary to call her mother.
(2c) It is very *certain/likely that Mary will call her mother.
Focusing on priority-type modals, I will sketch a scale-based analysis which integrates Morzycki's (2012) ideas about extreme adjectives with treatments of necessity modals in premise semantics (e.g., von Fintel & Iatridou 2008, Rubinstein 2012). The idea is to construct modal degrees by manipulating the premises used in interpreting the modal (the ordering source), with the distinction between weak necessity/non-extreme modals and strong necessity/extreme modals drawn in terms of whether the premises being manipulated are up for debate or not.
The Acquisition of Universal Quantifiers in Indonesian: A Preliminary Report
I Nyoman Aryawibawa (Universitas Udayana)
When: November 7th
Current study specifically concerns with the acquisition of universal quantifiers in Indonesian, e.g., masing-masing/setiap “each/every”, semua “all”. Previous studies, e.g., Vendler (1967), Ioup (1975), Brooks & Braine (1996), Brooks et al. (2001), showed that all and each/every have innate collective, e.g., All the boys are riding an elephant, and distributive, e.g., Each boy is riding an elephant, representations respectively. This study mainly aims to reinvestigate the innateness of the two meanings.
The study used a picture selection task. Five different sets of stimuli were employed. Three different groups of subjects participated: 20 younger children (4-to-6-year old), 30 older children (7-to-12-year old), and 30 adults (17-to-22-year old).
The results indicated that there was no evidence that the younger children were able to restrict the use of quantifiers to its domain, i.e., the noun it modifies. Furthermore, there was even no evidence that they could assign the distributive and collective meanings to masing-masing “each” and semua “all” respectively. In contrast, over 95% of the older children’s and adults’ use of the quantifiers was correct. Additionally, their assignment to the distributive meaning of masing-masing “each” was over 90% correct. For semua “all”, nevertheless, only about 50% of their responses represented the canonical meaning, i.e., collective.
Moreover, the results from quantifier spreading tests informed that half of the younger children preferred a symmetrical relation between masing-masing “each” and its domain, while the other half preferred an asymmetrical relation. Thus, their preference to the symmetrical response, as suggested by Philip (1995), is not observed here.
To conclude, all the evidence suggests that children’s universal quantifier acquisition is delayed until they are approximately at age 7. The learning process, in the sense of Bowerman (2001), and Borer and Wexler’s Maturation Hypothesis (1987) might factor in the delay. In other words, the innate collective and distributive meanings of all and each/every are not confirmed in this study.
Current study specifically concerns with the acquisition of universal quantifiers in Indonesian, e.g., masing-masing/setiap “each/every”, semua “all”. Previous studies, e.g., Vendler (1967), Ioup (1975), Brooks & Braine (1996), Brooks et al. (2001), showed that all and each/every have innate collective, e.g., All the boys are riding an elephant, and distributive, e.g., Each boy is riding an elephant, representations respectively. This study mainly aims to reinvestigate the innateness of the two meanings.
The study used a picture selection task. Five different sets of stimuli were employed. Three different groups of subjects participated: 20 younger children (4-to-6-year old), 30 older children (7-to-12-year old), and 30 adults (17-to-22-year old).
The results indicated that there was no evidence that the younger children were able to restrict the use of quantifiers to its domain, i.e., the noun it modifies. Furthermore, there was even no evidence that they could assign the distributive and collective meanings to masing-masing “each” and semua “all” respectively. In contrast, over 95% of the older children’s and adults’ use of the quantifiers was correct. Additionally, their assignment to the distributive meaning of masing-masing “each” was over 90% correct. For semua “all”, nevertheless, only about 50% of their responses represented the canonical meaning, i.e., collective.
Moreover, the results from quantifier spreading tests informed that half of the younger children preferred a symmetrical relation between masing-masing “each” and its domain, while the other half preferred an asymmetrical relation. Thus, their preference to the symmetrical response, as suggested by Philip (1995), is not observed here.
To conclude, all the evidence suggests that children’s universal quantifier acquisition is delayed until they are approximately at age 7. The learning process, in the sense of Bowerman (2001), and Borer and Wexler’s Maturation Hypothesis (1987) might factor in the delay. In other words, the innate collective and distributive meanings of all and each/every are not confirmed in this study.
How kids do and don’t make mistakes: The nature of regressions in the acquisition of phonological grammars
Anne-Michelle Tessier (University of Alberta)
When: October 24th
Children’s acquisition of their L1 phonological grammar is typically assumed to progress from an initial universal state towards a language-specific one, as learners change their grammar hypotheses incrementally to better approximate the target language. One well-known challenge to this view comes from the many reports of ‘U-shaped’ phonological development, in which a child’s production temporarily regresses, moving further from the target rather than closer. This talk has two main claims to make about the nature of phonological regressions, and their consequences for phonological acquisition and theory-building.
The first half of the talk argues that U-shapes should not emerge directly from the phonological learner’s automatic workings, i.e. just by its methods of processing errors and building new grammar hypotheses (construed here as constraint re-ranking). Building on existing and novel analyses of longitudinal data (such as Inkelas and Rose, 2007; Menn,1971), I will argue that the kinds of phonology on which children regress show both lexical and child-specific quirks, which combine to suggest that they are derived outside the learning ‘core’ in ways I will define. Part of this argument comes from the empirical observation that regressions do not usually demonstrate trade-offs: in particular that as more complex syllables are acquired, other previously-learned structures do not regress (data from Smith, 2010, and re-analyzing data from Donahue, 1986).
Given the assumption that a phonological learning algorithm should not create regressions on its own, the second half of the talk demonstrates that an OT learner which builds grammars incrementally and restrictively from its accumulated errors can in fact easily regress to create trade-offs of the (nearly?) unattested type. From this unwanted result, I conclude that learners must choose one of two alternative paths: either a re-ranking algorithm that learns more selectively from its errors, or else a more derivational grammar like Harmonic Serialism (McCarthy, 2007), which turns out to avoid making regression-prone errors in the first place. This study ultimately sheds some new light on the possible pathways for child phonological development, and on the mechanisms by which that development might be achieved.
Children’s acquisition of their L1 phonological grammar is typically assumed to progress from an initial universal state towards a language-specific one, as learners change their grammar hypotheses incrementally to better approximate the target language. One well-known challenge to this view comes from the many reports of ‘U-shaped’ phonological development, in which a child’s production temporarily regresses, moving further from the target rather than closer. This talk has two main claims to make about the nature of phonological regressions, and their consequences for phonological acquisition and theory-building.
The first half of the talk argues that U-shapes should not emerge directly from the phonological learner’s automatic workings, i.e. just by its methods of processing errors and building new grammar hypotheses (construed here as constraint re-ranking). Building on existing and novel analyses of longitudinal data (such as Inkelas and Rose, 2007; Menn,1971), I will argue that the kinds of phonology on which children regress show both lexical and child-specific quirks, which combine to suggest that they are derived outside the learning ‘core’ in ways I will define. Part of this argument comes from the empirical observation that regressions do not usually demonstrate trade-offs: in particular that as more complex syllables are acquired, other previously-learned structures do not regress (data from Smith, 2010, and re-analyzing data from Donahue, 1986).
Given the assumption that a phonological learning algorithm should not create regressions on its own, the second half of the talk demonstrates that an OT learner which builds grammars incrementally and restrictively from its accumulated errors can in fact easily regress to create trade-offs of the (nearly?) unattested type. From this unwanted result, I conclude that learners must choose one of two alternative paths: either a re-ranking algorithm that learns more selectively from its errors, or else a more derivational grammar like Harmonic Serialism (McCarthy, 2007), which turns out to avoid making regression-prone errors in the first place. This study ultimately sheds some new light on the possible pathways for child phonological development, and on the mechanisms by which that development might be achieved.
Aspectual Abstraction
Laura Wagner (Ohio State University)
When: September 26th
Children's acquisition of aspect has received a great deal of attention for children's conservative use of aspectual forms in production (that is,the fact that they fail to use many grammatical combinations that are available to them). This talk will discuss experimental evidence for children's ability to generalize their knowledge of aspect to novel instances, showing they command an abstract representation of the categories. Both lexical and grammatical aspect generalizations will be covered, with some discussion of the different developmental trajectories between them.
Children's acquisition of aspect has received a great deal of attention for children's conservative use of aspectual forms in production (that is,the fact that they fail to use many grammatical combinations that are available to them). This talk will discuss experimental evidence for children's ability to generalize their knowledge of aspect to novel instances, showing they command an abstract representation of the categories. Both lexical and grammatical aspect generalizations will be covered, with some discussion of the different developmental trajectories between them.
The acoustic dynamics of vowels and the sociophonetics of an American English variety spoken in
Michigan’s Upper Peninsula
Wil Rankinen (Indiana University / Michigan State University)
When: September 12th
Informed by the acoustic dynamic nature of vowel qualities and their formants’ trajectories, there is some debate over the ‘best’ means of capturing the acoustic characteristics of a vowel quality (e.g., one-, two- or multi-point measurements). Vowels are data rich acoustic entities. As such, vowel analysis requires quantitative procedures that can minimally extract relevant information from a vowel (e.g., time, F1, F2, F3, bandwidths) but also can be generalizable enough to accurately extract measurements for different qualities; traditionally, measurements are taken a fixed position or several fixed- positions in a vowel’s trajectory/duration for all vowel qualities.
The present paper introduces a procedure (e.g., based on the gestural model by Brown & Goldstein, 1990) based on our knowledge of monophthongs and diphthongs as targets (i.e., vowel trajectories toward a target or targets). This procedure objectively and efficiently extracts either one-point (i.e., for monophthongs) or two-point (i.e., for diphthongs) measurements as the target(s) for each individual vowel token at non-fixed positions for the duration of its trajectory.
Using this extraction procedure, the present paper examines the acoustic vowel data of two immigrant-heritage English speaking communities located in the Marquette County of Michigan’s Upper Peninsula. The data presented is based on an 88-speaker sample of a 130+ speaker corpus collected in 2007 and 2008. All 88-speakers for this paper are monolingual English speakers and native to the area of study. The speakers are stratified across the social factors consisting of heritage background, age, sex and socioeconomic status. This paper addresses two general sociophonetic characteristics found in the data:
1. The two heritage communities exhibit differences in the overall shape of their vowel system.
2. Canadian English vowel characteristics (i.e., Canadian shift and Raising) are prevalent in the individual vowels of younger speakers, particularly among the younger Italian-Americans.
Such observations illustrate how historical dynamics in a non-urban American English community are heavily dependent on grouping that arises from an emigrant heritage.
Informed by the acoustic dynamic nature of vowel qualities and their formants’ trajectories, there is some debate over the ‘best’ means of capturing the acoustic characteristics of a vowel quality (e.g., one-, two- or multi-point measurements). Vowels are data rich acoustic entities. As such, vowel analysis requires quantitative procedures that can minimally extract relevant information from a vowel (e.g., time, F1, F2, F3, bandwidths) but also can be generalizable enough to accurately extract measurements for different qualities; traditionally, measurements are taken a fixed position or several fixed- positions in a vowel’s trajectory/duration for all vowel qualities.
The present paper introduces a procedure (e.g., based on the gestural model by Brown & Goldstein, 1990) based on our knowledge of monophthongs and diphthongs as targets (i.e., vowel trajectories toward a target or targets). This procedure objectively and efficiently extracts either one-point (i.e., for monophthongs) or two-point (i.e., for diphthongs) measurements as the target(s) for each individual vowel token at non-fixed positions for the duration of its trajectory.
Using this extraction procedure, the present paper examines the acoustic vowel data of two immigrant-heritage English speaking communities located in the Marquette County of Michigan’s Upper Peninsula. The data presented is based on an 88-speaker sample of a 130+ speaker corpus collected in 2007 and 2008. All 88-speakers for this paper are monolingual English speakers and native to the area of study. The speakers are stratified across the social factors consisting of heritage background, age, sex and socioeconomic status. This paper addresses two general sociophonetic characteristics found in the data:
1. The two heritage communities exhibit differences in the overall shape of their vowel system.
2. Canadian English vowel characteristics (i.e., Canadian shift and Raising) are prevalent in the individual vowels of younger speakers, particularly among the younger Italian-Americans.
Such observations illustrate how historical dynamics in a non-urban American English community are heavily dependent on grouping that arises from an emigrant heritage.