University of Arizona
LING 495A/ 595A – Linguistics Department Colloquium
Spring 2016
Chair:
TA:
Schedule:
Massimo Piattelli-Palmarini (massimo@email.arizona.edu)
Zechy Wong (zechy@email.arizona.edu)
Fridays, 3:00pm – 4:30pm in Communication 311
Program
Friday, January 15
Reconsidering canonical forms and the benefits of “clear” speech
Meghan Sumner (Department of Linguistics, Stanford University)
Labels are powerful. A label for an object or an idea induces its representation in our
minds to become more categorical and less idiosyncratic than it might be in reality (Clark,
1998). Labels are all around us. They are the names we give to objects, the stereotypes we
use to describe people and behaviors, and using them reinforces the abstract concept,
overriding the specifics of our experience (Lupyan, 2007). Some labels we use quite
frequently in phonetics, speech perception, and spoken word recognition are “clear speech”,
“reduced speech”, “noisy speech” and “canonical forms”. In this talk, I give a brief overview
of these labels, and how they have influenced our thought and our aims to understand
speech processing, as a field.
This overview leads to the central question of some current research that I will discuss:
Given that the bulk of the research in the field shows that clear speech is understood more
quickly and accurately than casual speech, and that “mismatches” between signal and
canonical forms wreak havoc on the perceptual system, how is it that we understand one
another so well when nearly all of what we hear and say falls under the “reduced speech”
umbrella?
To address this question, I present a variety of studies investigating both the recognition of
clear and reduced words, the processing of clear and reduced sentences, and memory for
clear and reduced words. The data paint a story that is intuitive, given what we say and
hear, but counterintuitive given our strong sense of the benefits of “talking clearly”. The
project as a whole suggests that we move away from the notion of canonical forms, and
replace terms like “reduced” and “noisy” with “informative” and “typical”.
Friday, January 22
Undergraduate Meet-and-Greet
Prof. Diane Ohala
Friday, January 29
A Theory of Quantifier Context Dependence
Kristen Greer (Department of Linguistics, UCLA)
Quantifiers appearing in the DP are widely acknowledged to have meanings that are at
least partially determined by the context in which they are uttered. Three phenomena in
particular have been cited and treated as cases of quantifier context dependence: (a) the
domain restriction, (b) the ambiguity of certain quantifiers between proportional, reverse,
focus-affected, and cardinal interpretations, and (c) expectation readings. These are
roughly illustrated in (1)-(3), respectively.
(1) Every student failed the assignment.
a. ≠ Every student in the universe failed the assignment.
b. = Every student in the class failed the assignment.
(2) Many boys came to the party.
a. = Many of the boys came to the party. (Proportional)
b. = Many of the people who came to the party were boys. (Reverse)
(3) a. Many lawyers attended the meeting this year.
b. Many doctors attended the meeting this year.
(Where (3a) ≠ (3b) even if [[doctors]]=[[lawyers]], presumably because an
expectation is exceeded in one case but not the other)
Existing analyses account for these phenomena using a variety of formal tools, alternately
appealing to multiple linguistic forms (ambiguity) or to variable positions in the semantic
structure that are resolved to contents within a single model (extensional vagueness) or
within multiple models (intensional vagueness).
I develop a unified approach to quantifier context dependence, showing that the
phenomena illustrated in (1)-(3) can all be handled as cases of extensional vagueness,
wherein context (construed as a set of sets in a single model) provides content for variables
in semantic structure. I argue that there are two such contextual variables, one
representing the domain of quantification and one the restriction on this domain, and I
develop a theory of the syntactico-semantic structure of quantifiers that incorporates these
variables. I then show how this structure predicts their context-dependent behaviors in a
fully general, extensional way. The analysis has striking implications for our understanding
of natural language quantification, suggesting that there are two fundamental operators
forming generalized quantifiers (GQs) in natural language DPs: one that forms supersets
over its restrictor, and one that forms subsets over its restrictor.
Friday, February 5
An Act Apart: Processing NAI Content
Lyn Frazier (Professor, Linguistics, University of Massachusetts Amherst)
Potts (2005) investigated parentheticals, appositives, expressives and honorifics and
argued that they form a semantic natural class in that these "Not at Issue" (NAI)
expressions do not contribute to the truth-conditions of the embedding utterance, they
convey speaker commitments, and their semantic interpretation does not interact with "At
Issue" (AI) content. To explain these properties, he offered a multi-dimensional
semantics. See Schlenker (2010) for an alternative semantic account without multidimensionality, and for pragmatic accounts, see Harris and Potts (2009) and Amaral et al.
(2007.)
Based on research conducted with Brian Dillon and Chuck Clifton, I will argue that NAI
expressions are complete but dependent speech acts, and presumably as a result, they may
be represented in a separate memory store from AI content. Evidence derives from
differential effects of lengthening NAI vs AI content (Dillon et al.,
2014). Additional evidence comes from distinct interpretations for comparable material
when it is expressed as AI vs NAI content, and evidence that expressives exhibit a similar
interpretation (the same range of interpretations and sensitivity to the same interpretive
principles) when the expressive stands as an utterance by itself, where it must be analyzed
as a speech act, and when it appears as an attributive adjective (Frazier et al., 2014). Online
evidence shows an interaction of a wh-dependency in the embedding sentence with the
status (AI vs. NAI) of an embedded structure containing a wh-dependency (Dillon et al.,
submitted), as expected if content in the same memory store may interact, but content in
different stores may not.
The resulting view is one where prosody, syntax, semantics, pragmatics and general
cognitive principles all play a role in explaining the properties and processing of NAI
content.
References
Amaral, Patricia, Craige Roberts & E. Allyn Smith. (2007). Review of The Logic of
Conventional Implicatures by Chris Potts. Linguistics and Philosophy 30(6): 707–
749.
Dillon, B., Clifton, Jr., C., & Frazier, L. (2014). Pushed aside: Parentheticals, Memory &
Processing. Language, Cognition and Neuroscience 29 (4), 483-498.
Dillon, B., Clifton, Jr., C., Sloggett, S., and Frazier, L. (Submitted) Not all relative clauses
interfere with filler-gap processing equally: Appositive relative clauses and the
organization of linguistic working memory.
Frazier, L., Dillon, B., & Clifton, Jr., C. (2014). A note on interpreting damn expressives:
transferring the blame. Language and
Cognition 29, 1-14.
Harris, J., & Potts, C. (2009). Perspective-shifting with appositives and expressives.
Linguistics and Philosophy 32, 523-552.
Potts, C. (2005).The logic of conventional implicatures. Oxford: Oxford University Press.
Schlenker, P. (2010). Supplements within a unidimensional semantics II: Epistemic status
and projection. Proceedings of Northeastern Linguistic Society 2009, GLSA.
Schlenker, P. (2010). Supplements within a unidimensional semantics II: Epistemic status
and projection. Proceedings of Northeastern Linguistic Society 2009, GLSA.
Friday, February 12
Minimalism: A view after 20 years
Norbert Hornstein (Professor, Linguistics, University of Maryland / College Park)
This talk will try to assess how far we have come in realizing the goals of the Minimalist
Program. Of course, doing this requires outlining what these goals were, and here I will
provide a somewhat idiosyncratic (though, of course, deeply faithful rendition) of what the
program was. In my opinion, the project had two parts:
1. A Reductive/unificatory aspect in which the GB like modularity of FL was shown to
be merely apparent, i.e. that the seven or so distinct grammatical modules with their
own distinctive operations and locality domains were just surface aspects of the
same operations and principles.
2. An analytical aspect which reduces all the unified dependencies to a very small
number (1?) of linguistically novel cognitive operations.
I argue that we have come a pretty long way in seeing what (1) entails and that the
evidence that it is correct is non-negligible. Thus, I think that there is reason to believe that
phrase building, movement, binding, control, agreement, case, etc. are all just aspects of the
same basic combinatoric machinery. There are problems here, but the outlines of a
plausible theory are visible.
I then investigate whether Merge is the secret sauce that lies behind this unification. I
suggest (as I did in the 2009 book) that we should decompose Merge into two more general
operations, one of which corresponds roughly to labeling and the other to set union, and
that this is what lies behind the kinds of chains we in fact see. I show that were this right
then the kinds of restrictions on Merge that we find would follow. In particular, were Union
the basic combination operation, we would expect phrase markers to be sets subject to
Extension, Inclusiveness and to contain copies.
If this is right, my conclusion is that Minimalism is alive and well and making progress. I
think that this might come as a surprising conclusion to many.
Friday, February 19
Beyond Decomposition: How experiments might help shape morphological theory
Alec Marantz (Silver Professor of Linguistics and Psychology Departments of Linguistics and
Psychology, NYU)
Over the last decade, neuro and psycholinguistic experimentation has accumulated
supporting the hypothesis that words are decomposed down to their roots in
comprehension, during both visual and auditory presentation, lending credence to
linguistic theories such as Distributed Morphology, which insist on such decomposition in
the analysis of word structure. However, this work has also dissolved the putative
distinction central to Pinker’s Words and Rules framework between the “memorized” and
the “constructed” – usage frequencies are relevant to the processing of all words and
phrases, no matter how transparent or regular. So, for example, the transition probability
between a stem and a suffix of a morphologically complex word modulates (obligatory)
decomposition independent of regularity. The lack of correlation between memorization
and regularity allows us to recast Pinker’s Words and Rules approach to integrating
linguistics with cognitive neuroscience as an “Atoms and Rules” approach, emphasizing the
distinction between the ontology of linguistic pieces (morphemes) and the generalizations
about their order and arrangement. I will discuss how some recent finding from NYU’s
Neuroscience of Language Lab might feed back into the development of morphological
theory, given the Atoms and Rules approach and the observation that there is no escape
from frequencies even for the most regular of rules.
Friday, February 26
Inferential Organization in Grammar Systems: Word Structure. Paradigm
Organization and Learnability
Farrell Ackerman (Professor, Linguistics; Director, Human Development Program, UC San
Diego); work in collaboration with Rob Malouf (SDSU)
Speakers of languages with complex morphology and multiple inflection classes confront
a large learning task whose solution raises fundamental questions about the structure of
words, and the organization of morphological systems. This task receives a general
formulation as the Paradigm Cell Filling Problem (PCFP) in Ackerman et al. (2009):
PARADIGM CELL FILLING PROBLEM: Given exposure to an inflected wordform
of a novel lexeme, what licenses reliable inferences about the other
wordforms in its inflectional family?
The essential challenge, as formulated in the PCFP, is not new, and proposed answers to
it have a similar profile (Paul 1891, Hockett 1967, Paunonen 1976, Bybee 1985,
Anttila 1989, Wurzel 1989, see also Fertig 2013): analogical inferences from (incomplete
sets of) forms belonging to known inflectional patterns permit reasonable guesses
concerning likely candidates for unknown forms. Descriptive observations about the
implicational organization of morphological systems have been reconceptualized and
quantified in renascent word‐based approaches to morphological analysis (see the
detailed overview and description in Blevins 2016), where two interdependent
explanatory dimensions of part/whole relations are developed: the internal structure of
words interpreted in terms of discriminability among related words and the external
relations among words as reflected in paradigm organization, where words are parts of
(complex) paradigms which are interpretable as (adaptive discriminative) systems of
patterns. These two dimensions are evident in the Nilo-Saharan language Fur (Waag
2010), where the combinatorics of affixes and stem variation reflected in segment length,
tonal melodies and metathesis distinguish related words and provide available inferences
for predicting the forms of unencountered words. This is exemplified in the simple
patterns for the 1 st person singular completive versus the 3rd person singular completive
for the verb `to speak’ in (1):
1a.
ʔ-ɪ́rsɪ́ŋɔ
1SG-spoke
‘I spoke’
1b.
rɪ̀sɪ̀ŋɔ̀
spoke
‘s/he spoke’
The tonal melodies (1a) and (1b) exhibit opposite values for their person contrasts: all
highs on 1st singular versus all lows on 3rd singular. Additionally, 1st singular is associated
with a prefix i.e., ʔ, while its stem represents a methathetic variant of the 3rd singular stem
form, i.e., ɪ́rX versus rɪ̀X. To know the form for 1sg completive is also to know the form of
the 3sg completive, and vice versa. The mutual inferential relations in (1) are trivial, but
they become more complex when the whole system of Fur morphosyntactic properties and
encodings is considered.
The calculation of informativity concerning combinations patterned ingredients and
meanings associated with them attested in individual languages, such as Fur, has been
the main object of Information‐Theoretic measures in these new word-based formal
models. The L(ow) C(onditional) E(ntropy) C(onjecture) (Ackerman and Malouf 2013)
represents a cross-linguistic hypothesis concerning complex morphological systems:
morphological systems seem to display organization in terms of low conditional
entropies, reflecting high predictability between known words and their unknown
variants. In effect, the LCEC is a way of solving the PCFP, providing learners with cues to
facilitate good guesses about previously unencountered words. Given the huge
variability in the cross‐linguistic shapes of words and their patterns of relatedness, the
LCEC, by hypothesis, reflects a strategy by which language change is guided by
learnability considerations.
The learnability problem identified by descriptivists is put in new perspective by recent
research revealing that the inflectional stimuli that learners experience is highly skewed
and incomplete: following Zipfian distributions, small numbers of inflected words are
heard frequently providing partial paradigm information, while increasing the corpus
size does not provide exposure to the “missing” words from the complete paradigm, but
merely reinforces the distributions found in smaller samples. (Bonami and Beniamine
2015, Ramscar and Blevins 2015). Yang (2010) utilizes these results to argue against the
class of word and paradigm models. Following Bonami and Beniamine 2015, Ramscar
and Blevins 2015 and Blevins 2016, we argue, on the contrary, that these results suggest
the necessity for complex morphological systems to be organized in line with wordbased models that are centrally concerned with quantifying conditional entropy in
morphological organization: the learning paradox raised by Zipfian distributions of
stimuli points to the necessity of something like the LCEC and correlatively, for the type of
word‐based morphological model within which it operates.
Given this, we explore empirical data that confirm and/or challenge the LCEC,
suggesting that both help to refine the possible scope of the conjecture and the nature of
its extensions to data beyond those analyzed in Ackerman and Malouf 2006, Ackerman
et. al. 2009, Bonami and Luis 2013, Ackerman and Malouf 2013, Bonami 2014,
Bonami and Beniamine 2015, Sims 2015, Stump and Finkel 2015, Blevins 2016, among
others.
Friday, March 25
Bound subjects, phases and postponement of transfer
Howard Lasnik (Distinguished University Professor in the Department of Linguistics at the
University of Maryland / College Park)
Family of Questions (sometimes called pair-list) readings (here abbreviated FoQ) with WHs
and universal quantifiers most typically arise when the WH originates in the same clause as
the universal:
(1) Who did everyone see FoQ T
(2) Who do you think everyone saw FoQ T
May (1985) presented an important analysis of some such cases, but as pointed out by
Sloan (1991), it incorrectly predicts the possibility of such readings even when the
universal and the WH-trace are not clause-mates:
(3) Who does everyone expect [Mary to see t]? FoQ *
(4) Who does everyone think [Mary saw t]? FoQ *
May (1977) had already noted the absence of family of questions readings in examples like
(4). However, Sloan reported that May gave her examples with structures very similar to
those of (3) and (4) that do allow the reading:
(5) Who does everyonei expect [PROi to see t]? FoQ T
(6) Who does everyonei think [hei saw t]? FoQ T
It has long been known that clauses without overt subjects (especially infinitival clauses)
do not act like full clauses with respect to a wide variety of phenomena. Postal (1974)
introduced the notion ‘quasi-clause’ for some such cases and later Rizzi (1982) developed a
comprehensive theory of ‘restructuring’. There is now a vast literature on this topic. Some
version might cover (5). Only sporadically mentioned are situations where a bound
pronominal subject makes a complement clause similarly permeable. I will explore some
further constructions behaving in a similar way, especially WH-island exemption observed
by Ross (1967), as illustrated in (7), and consider a possible explanation in terms of phases,
based on joint work with Tom Grano.
(7) a. He told me about a book which I can't figure out [where PRO to obtain t]
b. Which books did he tell you [why he wanted to read t] *Mary
Bibliography
Grano, Thomas and Howard Lasnik. 2015. How to neutralize a finite clause boundary:
Phase theory and the grammar of bound pronouns. Ms. Indiana University and
University of Maryland.
May, Robert. 1977. The grammar of quantification. Doctoral dissertation, MIT, Cambridge,
Mass.
May, Robert. 1985. Logical Form: Its structure and derivation. Cambridge, Mass.: MIT Press.
Postal, Paul M. 1974. On raising: One rule of English grammar and its theoretical
implications. Cambridge, Mass.: MIT Press.
Rizzi, Luigi. 1982. Issues in Italian syntax. Dordrecht: Foris.
Ross, John Robert. 1967. Constraints on variables in syntax. Doctoral dissertation, MIT,
Cambridge, Mass. Published as Infinite syntax! Norwood, N.J.: Ablex (1986).
Sloan, Kelly. 1991. Quantifier-wh interaction. In MIT Working Paper in Linguistics 15, 219237.
Friday, April 1
Grammaticalization and Repair as a Resolution to Labeling Algorithm Failures
Robert LaBarge (Department of Linguistics and Applied Linguistics, Arizona State University)
This talk is an extension of work originally presented at ALC 8 and 9. In the first of these
talks, I argued that grammaticalization occurs as the result of labeling difficulties
experienced by the child acquirer. In the second, I argued that derivations attempt to label
and repair difficult structures in real-time. Here, I will combine the two arguments,
showing that repair and grammaticalization are essentially the same strategy, but only the
latter is available to the child acquirer. But why? This talk will focus on three phenomena:
the first is a possible head-head Merged structure in Chinese which shows evidence of
historic change from verb to modal. The second is a phrase-phrase Merged structure in
Macedonian which shows evidence of historic change from demonstrative to definite
article. In both cases, the question as to why new structures do not replace the old ones
entirely will be addressed. The third case involves Paul Postal-type raising-to-object ECM
constructions and asks why such constructions require a null subordinate C. In all cases, I
argue for an exoskeletal style scaffolding (a la Borer) that guides both grammaticalization
and repair, but in different ways.
Friday, April 15
Undergraduate Research Forum
Prof. Diane Ohala
Friday, April 22
Change in aspect and argument structure
Elly van Gelderen (Regents’ Professor, English Department, Arizona State University)
By sketching some of the changes that affect the argument structure throughout the history
of English, I shed light on the universality of the aspectual division in manner and result,
the major theta-roles that depend on this, and the special status of the Theme. For instance,
I show that unaccusatives are reanalyzed as causatives or copulas, due to the persistence of
the Theme, but not as unergatives or unergatives as unaccusatives. Object experiencers are
reanalyzed as subject experiencers but not the other way round. The reason for this is that
verbs hang on to their basic aspectual classification and their Themes, and that the
appearance of certain theta-roles is constrained by others.
Friday, April 29
Prosodic phonology of Nez Perce double reduplication
Kathryn Pruitt (English Department, Arizona State University)
This talk presents joint work with Amy Rose Deal (UC Berkeley) on the prosodic phonology
of fully-reduplicated and doubly-reduplicated adjectives in Nez Perce. Full reduplication is
associated with adjectives in Nez Perce, which may be derived from other categories as in
(1) or from bound roots as in (2) (Aoki 1994). (Other morphological means of marking or
deriving adjectives are available in Nez Perce but will not be discussed here.)
(1) qaʔán-qaʔan ‘respected’
(2) kaʔáw-kaʔaw ‘empty, unoccupied’
cf. qaʔán ‘to respect’ (V)
*kaʔáw
Further, plural marking on adjectives takes the form of a prefixed /Ci-/ reduplicant, where
C is the first consonant of the root as in (3) (Aoki 1963).
(3) kuhét → ki-kuhét
‘long (sg., pl.)’
This talk will focus on two phenomena that emerge when fully-reduplicated adjectival
forms are pluralized with /Ci-/ reduplication, creating doubly-reduplicated forms:
overapplication of syncope as in (4), which is analyzed as templatic backcopying (e.g.,
Gouskova 2007), and special treatment of monosyllabic stems according to their weight as
in (5) (Aoki 1994).
(4) Sg.
Pl.
qatat-qátat qi-qtat-qí-qtat
lokóy-lokoy li-lkóy-li-lkoy
(5) Sg.
Pl.
Light syllable base
‘solid, compact, hard’
‘slender’
kúc-kuc
ki-kúc-kuc
Heavy syllable base
qees-qées
qi-qees-qí-qees
maqs-máqs mi-maqs-mí-maqs
‘small’
(*ki-kúc-ki-kuc)
‘spotted, blurred’
‘yellow’
(*qi-qees-qées)
(*mi-maqs-máqs)