Together at Last: How Tense and Aspect Interact in Simulation Semantics

1
Together at LastHow Tense and Aspect Interact
in Simulation Semantics
2
The Problem

• Aspectual theorists have proposed revealing
  models of type shifts, both explicit and
  implicit
• Explicit type-shifting
• Progressive Process ? State (e.g., She was
  singing)
• Perfect Event ? State (e.g., The Eagle has
  landed)
• Implicit type-shifting (coercion)
• State ? episode of stasis She was depressed
  twice.
• State ? inchoative event I liked her within a
  minute.
• These models rely on an ontology of idealized
  situation types (process, state, event, etc.).
• In this respect, they are incompatible with
  models of tense.

3
The Problem

• Unlike aspect, tense lacks an ontology.
• Tense is purely relational there are no
  tense-type entities corresponding to
  aspectual-type entities.
• Tense relates speech time to reference time (R).
• R is located relative to the encoded situation
  it constrains inference potential.
• Tense semantics are therefore based on a location
  metaphor R is speaker/hearer location.
• By contrast, aspectual semantics are based on an
  entity metaphor.
• But tense and aspect are much more similar than
  the traditional models imply.

4
Tense-Aspect Interfaces

• Tenses convey aspectual perspectives like
  attention to endpoints
• My ex-husband is/was Latvian.
• I took a cab over. The driver was/?is Latvian.
• Aspect determines direction of inclusion for R
• State Pat was in Cleveland in June. (state
  overflows R)
• Event Pat visited Cleveland in June. (event
  exhausted by R)
• Tense and aspect jointly encode overlap/sequence
• I told them I had (had) more time on the meter.
• Tenses select aspectual types, and as such are
  coercion triggers. a eu
• Margot sest retournée. Henri avait lair
  heureux.
• Marge turned around. Harry was (perf/imperf)
  happy.

5
Our Purpose

• We will propose a simulation-based model of tense
  that is isomorphic to models of phasal aspect
  proposed by Narayanan 1997 and Chang, Gildea
  Narayanan 2000.
• This model will capture tense-aspect interfaces,
  in particular the role of tense in triggering
  aspectual type-shifts.
• We will use this model to account for two
  traditionally mysterious kinds of semantic
  effects
• Habitual coercion She took the bus to work that
  summer.
• Present-tense coercion She takes the bus to
  work.
• We will model these type shifts by assuming
  x-schemas for tense that are combinable with
  aspectual x-schemas.
• Tense x-schemas will be known as sampling sets.

6
Coercion by Tense Habituals

• Habitual predications count as state
  predications
• Extensibility
• She smoked back then and I think she still does.
• She enlisted in 1985 and I think she still does.
• Present-tense reporting
• She smokes.
• Look! That woman smokes a cigar.
• However, via internal composition they are not
  states but iterated events (like pacing, bouncing
  a ball, etc.).
• No predication is born habitual, so in the
  absence of a dedicated habitual marker, habitual
  meaning must come from an implicit type shift.
• But what grammatical form performs this type
  shift?

7
Coercion by Tense Present

• The present tense is a state selector (Langacker
  1987)
• She is ill for three days.
• She dribbles the basketball.
• Present distorts Aktionsart projected by
  predicators
• Futures from perfective stempresent inflection,
  e.g., Latin amabo I will love (ltamav o)
• Habitual-event reports
• She visits her mother. (possible iff assigned a
  habitual reading)
• Progressive reports in, e.g., French
• Eh bien, à present, je me sens mieux. Le morale
  revient. Now I feel better. My morale is coming
  back. (Binet, Bidochon 8 42)
• Perfect reports in, e.g., French
• Ca fait dix minutes quelle nous parle de la
  moquette! Shes been telling us about the carpet
  for 10 minutes. (Binet, Bidochon 1017)

8
Simulation Semantics

• Inspired by biological control theory, Simulation
  Semantics models events as executing-, or
  x-schemas.
• An x-schema is a Petri net a weighted graph
  consisting of places (circles) and transitions
  (rectangles) connected by directed input and
  output arcs.
• A state is defined by the placement of a token (a
  black dot or number) in a particular place.
• The real-time execution semantics of Petri nets
  models the production and consumption of
  resources
• A transition is enabled when its input places are
  marked such that it can fire by movement of
  tokens from input to output.
• Arcs include resource, enable and inhibitory
  arcs.
• Actions have hierarchical structure, permitting
  embeddings.

9
Basic X-Schema Distinctions

10
X-Schemas and Aktionsart
11
A Schema Controller
iterate
Ready
Done
Start
Process
Finish
interrupt
resume
Cancel
Suspend

• The controller sends signals to the embedded
  schema.
• It transitions based on signals from the embedded
  schema.
• It captures higher level coordination of actions.

12
Phasal Aspects Map to Controller
Iterative (repeat)
Inceptive (start, begin)
Iterate
Ready
Done
Start
Process
Finish
interrupt
resume
Cancel
Suspend
Completive (finish, end)
Resumptive(resume)
13
Phases of Climb
Iterate
Ready
Done
Start
Process
Finish
resume
Suspend
interrupt
Cancel
BINDINGS
Energy Ready Standing
On top
14
About to Climb (Prospective)
Iterate
Ready
Done
Start
Process
Finish
resume
Suspend
interrupt
Cancel
BINDINGS
Energy Ready Standing
On top
15
Be- Climbing (Progressive)
Iterate
Ready
Done
Start
Process
Finish
resume
Suspend
interrupt
Cancel
BINDINGS
Energy Ready Standing
On top
16
Have- Climbed (Perfect)
Energy Ready Standing
On top
17
Tense and Reference Time

• Tense is defined by the location of reference
  time relative to speech time in the traditional
  (Reichenbach-based) view
• Past tense E,R lt S
• Present tense E,R,S
• In Simulation Semantics, R is a state of the
  controller schema.
• The location of R relative to S is captured by a
  time-depth representation called a sampling set
• Present single state
• Past two states (onset and offset)
• The final or unique state is the state of
  knowledge at S.

18
Aspect and Reference Time

• In Simulation Semantics, direction of inclusion
  is represented by the presence or absence of a
  transition between pre- and post-states in the
  sampling set.
• In the case of the present tense, there is only
  one state in the sampling set.
• Only states can be verified on the basis of a
  single temporal sample.
• Therefore, present-tense reports are state
  reports.

19
Aspect and Reference Time

• The sampling set for the past tense may include a
  transition between the two states.
• The presence or absence of this transition
  captures the difference between perfective and
  imperfective past
• Perfective Pat was in Cleveland yesterday for
  the trade show. (yesterday includes offset
  transition)
• Imperfective Pat was in Cleveland yesterday, and
  in fact has been there all month. (yesterday does
  not include an offset transition)
• The perfective and imperfective schemas are not
  alternate event types, nor are they alternate
  construals of a single situation they are sets
  of sets of places within the controller schema.

20
Aspectually Sensitive Past Tenses
Perfective
Imperfective
21
Aspectually Sensitive Past Tenses
Perfective
Imperfective
22
Type-Shifting Revisited

• In Simulation Semantics, a stative type shift is
  selection of a state of the controller.
• The combination of controller and base x-schema
  can (a) augment, (b) embed or (c) select a part
  of the base x-schema
• Augmentation State ? inchoative state, e.g.,
  Suddenly, I knew the answer.
• Embedding Event ? iterated event ? habitual
  state, e.g., She runs.
• Selection Inchoative ? state, e.g., Your soup is
  cooled.
• Additional examples present-tense coercion,
  progressive type-shifting, continuative perfect
  type-shifting.

23
Habituality

• Habitual situations, e.g., Jan ran, count as
  events, just like iterated actions, e.g., Jan
  paced back and forth.
• Habituals are therefore not intrinsically
  stative they are events embedded in an iterate
  schema

iterate
24
Stative Coercion

• Habituals become stative solely by virtue of
  being embedded in the stative process schema.
• This schema can be selected by the present tense.

present
ongoing
25
Conclusion

• In Simulation Semantics, tense and aspect are
  represented in a uniform manner.
• Rather than Aktionsart plus a time line and
  perspective, we have combination of simulations.
• Combinations model iteration, the selection of
  subparts and augmentation of schemas.
• Using these mechanisms, we account for
  tense-based coercions, e.g., stative coercion via
  present tense.
• How do we build ECG representations for tense
  inflections that abstract over the various
  morphological instantiations of tense
  (affixation, suppletion, etc.)?
• Can we assume that English, like Romance, has
  both state- and event-selecting past tenses?