Aspectual composition with motion verbs in Japanese: A scale-based account Yusuke Kubota March 8, 2014 Abstract In this paper, I propose an analysis of motion verbs in Japanese by building on and extend- ing recent scale-based approaches to verb meaning (Hay et al. 1999; Kennedy and Levin 2008; Beavers to appear; Pi~non 2008; Kennedy 2012, inter alia ). The proposed analysis captures the aspectual composition patterns exhibited by two types of motion verbs in Japanese uniformly in terms of the notion of scale composition. The analysis is compared with the previous proposals by Ueno (2007) and Iwamoto (2008), the former couched in a decomposition-based approach to lexical semantics, and the latter in Jackendo’s (1996) theory of ‘structure-preserving binding’. I argue that the present scale-based approach over- comes the diculties that these previous approaches face and enables a conceptually simple yet explicit modelling of aspectual composition. The paper also discusses implications on wider issues pertaining to lexicalization patterns and typological classication of motion verbs. Keywords: aspectual composition, scale-based semantics, scale composition, telicity, mo- tion verb
53
Embed
Aspectual composition with motion verbs in Japanese: A ...kubota.yusuke.fn/papers/kubota-motion.pdf · ‘walk’, hasiru ‘run’ and nagareru ‘ ow’ intuitively entail change
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Aspectual composition with motion verbs in Japanese:
A scale-based account
Yusuke Kubota
March 8, 2014
Abstract
In this paper, I propose an analysis of motion verbs in Japanese by building on and extend-
ing recent scale-based approaches to verb meaning (Hay et al. 1999; Kennedy and Levin
2008; Beavers to appear; Pinon 2008; Kennedy 2012, inter alia). The proposed analysis
captures the aspectual composition patterns exhibited by two types of motion verbs in
Japanese uniformly in terms of the notion of scale composition. The analysis is compared
with the previous proposals by Ueno (2007) and Iwamoto (2008), the former couched in
a decomposition-based approach to lexical semantics, and the latter in Jackendoff’s (1996)
theory of ‘structure-preserving binding’. I argue that the present scale-based approach over-
comes the difficulties that these previous approaches face and enables a conceptually simple
yet explicit modelling of aspectual composition. The paper also discusses implications on
wider issues pertaining to lexicalization patterns and typological classification of motion
The subscript straight(x)(init(t)) indicates that this measure function has a (derived)
minimum endpoint corresponding to the straightness of x at the beginning of t. The measure
function returns a differential (or comparative) degree that designates the relative degree
of straightness of x at the end of t against this minimum endpoint.9 Truth relative to
intervals is given by the following verbal pos(itive) morpheme, which is an operator that
takes a verbal measure function of type µ and returns a predicate of individuals relative to
temporal intervals of type 〈e, 〈i, t〉〉 (here and in what follows, I use the letterM for writing
variables over measure functions of type µ):
(16) [[posv]] = λMλxλt.M(x)(t) ≥ stnd(M)
The pos operator in (16) says that a sentence containing a verb that denotes a measure
function is true just in case the degree that the verbal measure function returns for the
interval at which it is evaluated satisfies the standard value of the scale associated with the
verbal measure function. Thus, the notion of standard plays a crucial role in defining the
truth conditions for sentence containing degree achievements.
The crucial assumption that enables a principled account of the ambiguity in sentences
like (13) (and the lack thereof for those in (14)) is that the scale for the measure function for
7K&L formulate their analysis in event semantics by employing an event variable instead of a temporal
interval variable. I adopt a simpler formulation using temporal interval variables since nothing in the ensuing
discussion crucially hinges on assuming a richer ontology that takes events as primitives.
8Here, d is the type of degrees and i is the type of temporal intervals.9This essentially amounts to the claim that the semantics of degree achievements contains the semantics
of comparatives at its core. Bobaljik (2012, chapter 5) discusses some cross-linguistic morphological facts
that provide motivation for this assumption.
14
verbs like straighten has both a maximum and a minimum endpoint. The maximum endpoint
is the one inherited from the underlying adjectival scale and the minimum endpoint is the
derived endpoint designating the initial degree. It is independently known that, in the
adjectival domain, for predicates with both a maximum and a minimum endpoint, both of
these endpoints are available as the standard for truth evaluation (Kennedy and McNally
2005; Kennedy 2007). Assuming the null hypothesis that the same principle for standard
setting is operative in the verbal domain as well, we have an explanation for the observed
ambiguity. If the standard is set to the minimum endpoint (corresponding to the initial
degree), we obtain a differential (or atelic) interpretation which is true just in case the final
degree is larger than the initial degree (i.e, if there was some positive change in degree in
the relevant interval). If, on the other hand, the standard is set to the maximum endpoint,
the sentence is true just in case the maximum endpoint is reached by the end of the relevant
interval. This results in a telic reading which, for (13a), says that the rod became completely
straight. With verbs like lengthen in (14), the scale associated with the measure function
has only a minimum endpoint designating the initial degree. Thus, it follows that such
sentences have only the atelic, differential interpretation.
3.2 Measure functions for motion verbs
I now extend the scale-based analysis of verb meaning to motion verbs. This is meant to
be part of a larger project of characterizing the meanings of verbs belonging to different
aspectual classes in terms of the notion of scale (see Beavers (2008b) for a similar attempt),
but since laying out the full (re-)classification of verbal aspect and justifying its empirical
and ontological underpinnings goes beyond the scope of the present paper, here I focus on
aspects that are relevant for the analysis of motion verbs, and leave the task of working out
a more complete theory for future work.
I assume that manner of motion verbs and directional motion verbs are lexically associ-
ated with scales with different scale structures, and that this derives their distinct (lexical)
aspectual properties. More specifically, the former denote measure functions with open-
ended scales whereas the latter denote measure functions with maximally-closed scales.
15
The maximum endpoint of a directional motion verb is provided by the goal argument that
it subcategorizes for. Thus, noboru ‘ascend’ is of type 〈e, µ〉 lexically and denotes a verbal
measure function of type µ by combining with a goal argument:
(17) [[noboru]] = λyλxλt.ascend(y)(x)(t)
By combining with some goal argument (say, t, denoting the top of the mountain), (17)
returns a measure function (of type µ) ascend(t) which measures the progress of an upward-
directed motion toward the goal designated by t.
Manner of motion verbs, by contrast, denote measure functions associated with open
scales, reflecting their lexical aspectual property of being (underlyingly) activity predicates.
Thus, I assume that the manner of motion verb aruku ‘walk’ denotes the measure function
in (18), which measures how much of the physical motion constituting walking has taken
place during t.
(18) [[aruku]] = λxλt.walk(x)(t)
As in K&L’s analysis of degree achievements, truth relative to intervals is given by
converting the measure function via the verbal pos morpheme defined in (16) above. With
manner of motion verbs (which have the same type of open-ended scales as other activity
verbs), I assume that the standard is set to a context-dependent degree that designates
the threshold for determining whether the named activity has been ongoing throughout
the relevant interval. With a directional motion verb, the standard is simply set to the
endpoint of the scale, which corresponds to the goal. From this, it follows that the former
has the subinterval property characteristic of atelic predicates (down to a certain level of
granularity) whereas the latter does not. Thus, the basic aspectual distinction between the
two types of verbs follows from the different scale structures of the underlying scales.
The difference between manner of motion verbs and other more prototypical activity
verbs lies in their entailments. That is, unlike prototypical activity verbs, manner of motion
verbs denote activities that necessarily entail change of location. I assume that this entail-
ment invokes a measure function conversion operation that takes an underlying activity-type
16
measure function for manner of motion verbs and derives a new (also open-ended) measure
function that measures the change of location that results from the named activity. This
measure function conversion operation (of type 〈µ, µ〉) can be formalized as a lexical rule
that takes the original lexical entry of manner of motion verbs like (18) and returns a derived
entry like (19):
(19) path ([[aruku]]) = λxλt.path(walk)(x)(t)
Importantly, this measure function conversion operation is strictly based on uncancellable
entailments that arise (but are distinct) from the lexically-encoded meaning of the verb.
Thus, it should not be confused with more general types of ‘rules of construal’ including
coercion operations of various sorts whose interpretation often crucially depend on general
pragmatic knowledge as opposed to context-independent aspects of meaning. That is, deriv-
ing a change of location measure function out of the measure function of a genuine activity
verb like odoru ‘dance’ is not an option, even though there is nothing incoherent about the
notion of moving along a path in a dancing manner.
One might ask at this point whether there is any such thing as uncancellable entailment
that is distinct from lexically encoded meaning. That is, if some implication associated
with the sentence is uncancellable, doesn’t that mean that that implication is there because
it is part of the lexically-encoded, context-independent component of the meaning of some
word in the sentence? I would like to remind the reader that this is not necessarily so.
As pointed out in Kubota (2012), if sentence S expresses some proposition P as its main
assertion and additionally has a presupposition Q, and, if, furthermore, P and Q together
entail some proposition R, then R is just as uncancellable as P from a felicitous utterance
of S, but it does not necessarily have to be part of the lexically encoded meaning of (the
components of) S. Rather, in this case, R has the status of what one might call a derived
entailment, that is, an entailment that arises from an interaction between the narrower
sense of entailment (corresponding to the lexically encoded meaning) and the presupposition.
The change of location meaning of manner of motion verbs has exactly the status of such a
derived entailment. The verb expresses as its lexical meaning only the manner of motion.
17
But it is shared knowledge of competent speakers of Japanese that if the event named by
the verb takes place, it inevitably leads to an (at least relative) change of location. Thus,
an uncancellable entailment of change of location arises from the lexically encoded meaning
of the verb (together with its presupposition), but the change of location entailment itself
is not part of the lexically encoded meaning.
The assumption that a measure function conversion operation is available for manner of
motion verbs based on their derived entailments is both conceptually plausible and empiri-
cally supported.10 Conceptually, this operation is essentially a homomorphic mapping of a
measure function from one domain to another; there is a one-to-one correspondence between
the activity carried out over the course of some temporal interval and the change of location
that results from it. It is well attested in the literature that such homomorphic mappings
play significant roles in characterizing the semantics of natural language expressions, es-
pecially in the domain of aspectual composition (Krifka 1989; Nakanishi 2007). Note also
that the measure function conversion operation via the path operator obeys the principle of
monotonicity (Koontz-Garboden 2007) that morphological processes are generally known to
obey. The path operator takes some measure function as an input and produces a derived
measure function by merely relying on the (derived) lexical entailment of the input, without
destructively modifying its internal structure. Thus, the process is strictly monotonic. See
section 4.2 for more on the monotonicity principle, where this property becomes relevant in
relation to a comparison with a previous proposal by Ueno (2007).
Empirically, there are at least two pieces of evidence for the availability of this measure
function conversion operation for manner of motion verbs. The first comes from the inter-
pretation of the compound verb V-sugiru ‘V excessively/over-V’ (Yumoto 1997; Nakanishi
2007), which expresses excessive meanings.
(20) a. odori-sugiru (dance-sugiru): ‘dance too much’ (i.e. too much physical activity
10Moreover, the underlying notion of derived entailment receives independent justification elsewhere: as
discussed in Kubota (2013), this notion is also useful in characterizing the meanings of ‘manner of death’
verbs, another class of verbs that have been discussed in the recent literature as (putative) counterexamples
to the manner/result complementarity (cf. Beavers and Koontz-Garboden 2012; Husband to appear).
18
involved)
b. nobori-sugiru (ascend-sugiru): ‘ascend too much’ (i.e. ascend to a too high po-
sition)
c. aruki-sugiru (walk-sugiru): ‘walk too much’, ambiguous between:
• too much physical activity involved
• too much distance moved
Setting aside the interpretation which entails excessiveness of the number of distinct events
(which is available regardless of the verb type), the three types of verbs contrast with one
another in the interpretation of this compound verb. As shown in (20), with activity verbs,
the excessiveness expressed by sugiru pertains to how much the named activity has been
carried out. By contrast, with the directional motion verb noboru ‘ascend’, the excessiveness
is interpreted in terms of the distance of movement. On the present scale-based approach,
this contrast can be naturally captured by assuming that the verb sugiru is a kind of ver-
bal measure function intensifier which targets the degree argument of the measure function
denoted by the verb.11 Now, crucially, with manner of motion verbs, sugiru exhibits an
ambiguity between the ‘excessive activity’ interpretation and the ‘excessive change of loca-
tion’ interpretation. This ambiguity follows straightforwardly by assuming that V-sugiru
originally denotes an activity-type measure function which can be mapped to a change of
location measure function via the path operator.
A similar pattern is observed with the degree question with dore-kurai ‘how much’.
(21) a. Dore-kurai odot-ta? ‘How much did you dance?’
(asking about the length of time involved)
b. Dore-kurai nobot-ta? ‘How much did you go up?
(asking about the length of time / the length of physical path involved)
c. Dore-kurai arui-ta? ‘How much did you walk?’
(asking about the length of time / the length of physical path involved)
11See Nakanishi (2007) for further support for an analysis along these lines.
19
In addition to the duration interpretation available for all three sentence, this degree ques-
tion can be interpreted as asking for the physical distance moved with directional motion
verbs. Manner of motion verbs pattern like directional motion verbs in allowing for this
interpretation as well. This ambiguity, again, suggests that a derived change of location
measure function is available for them.
3.3 Scale composition
The analysis of aspectual composition I formulate below builds crucially on the notion of
‘scale composition’, or, more precisely, measure function composition. To ensure that
composite measure functions serve the purposes they are intended to, I posit both a sub-
stantive and a formal condition on the operation of measure function composition. The
substantive condition embodies the key intuition behind the notion of measure function
composition that two measure functions can be composed into one only if they simultane-
ously keeps track of the progress of a single event from two different perspectives. I call this
condition the commensurability condition on measure function composition.
(22) Commensurability condition on measure function composition:
For two measure functionsM1 andM2, the composite measure functionM1+M2 is
defined only if for any t and x that are in the domains ofM1 andM2,M1(x)(t) and
M2(x)(t) measure changes that pertain to different aspects of a single more general
change characterizable in terms of some common property (such as length, weight,
temperature, etc.).
In addition to this substantive condition, there is a formal condition called the structure-
preserving condition, which guarantees that the composite measure function is nothing
more (or less) than an ‘amalgamate’ of the original two measure functions.
(23) Structure-preserving condition on measure function composition:
For any two measure functionsM1 andM2, the composite measure functionM1+M2
satisfies the following properties:
a. Two projection functions π1 and π2 keep track of the correspondence between
20
the degree returned by the composite measure function M1+M2 on the derived
composite scale and the degrees returned by M1 and M2 on the original scales:
∀x∀t.π1([M1+M2](x)(t)) =M1(x)(t)
∀x∀t.π2([M1+M2](x)(t)) =M2(x)(t)
b. Endpoints of the composite scale are inherited from the original scales. Thus, if
the scale forM1 has a minimum endpoint MIN(M1), then the scale forM1+M2
also has a minimum endpoint, corresponding to MIN(M1):
π1(MIN(M1+M2)) = MIN(M1)
Analogous conditions hold for the maximum endpoint of the scale for M1 and
the minimum and maximum endpoints of the scale for M2.
c. If M1 has a minimum (maximum) endpoint, then, for any two degrees d1 and
d2 on M1 +M2, the relative distance from the endpoint for the two degrees
is isomorphic to the relative distance from the endpoint for the corresponding
degrees on the original scale M1:12
d1 −MIN(M1+M2)
d2 −MIN(M1+M2)=π1(d1)−MIN(M1)
π1(d2)−MIN(M1)
Analogous conditions hold for the maximum endpoint of the scale for M1 and
the minimum and maximum endpoints of the scale for M2.
(23a) pertains to the recoverability of the original measure function from the composite
measure function. That is, for any given pair of individual and temporal interval given
12This presupposes that degrees are modelled by numerical values and that the (dense and total) order
which establishes the ordering among degrees on a scale is the ‘greater than’ relation between numerical
values. For concreteness, I assume that closed scales are modelled by an interval (i.e. a continuous set) of
real numbers with the smallest or the largest numbers (or both, in the case of totally-closed scales) and that
open scales are modelled by an interval of numbers without either the smallest or the largest number.
21
as arguments to the composite measure function M1+M2, the degree returned by the
composite measure function represents an abstract value on a ‘composite scale’ from which
one can uniquely retrieve the ‘corresponding’ degrees on the original scalesM1 andM2 by
applying the projection functions π1 nad π2 to it. Based on these projection functions, we
can then define the endpoints of the composite scale. In short, the scale for the composite
measure function has a maximum (minimum) endpoint if and only if the scale for one of
the original measure functions has a maximum (minimum) endpoint. The correspondence
between the endpoints of the composite scale and the original scales is mediated by the
projection functions π1 and π2. Note that, from the definition in (23b), it follows that if the
scales for the original measure functions M1 and M2 both have a maximum (minimum)
endpoint, the two endpoints are identified with each other and inherited to the composite
measure function M1+M2. (23c) imposes a further structural isomorphism between the
original and composite scales. The idea is that the composite scale not only inherits the
endpoints of the original scale(s) but also preserves the form of the scale relative to the
endpoint(s). This last condition becomes important in the analysis of degree modifiers,
which are expressions that target the structures of the measure functions that they modify.
3.4 Accounting for aspectual composition
I now account for the patterns of aspectual composition from section 2. Since this requires
one to make certain assumptions about the syntax-semantics interface explicit, I list below
in (i)–(iv) the relevant assumptions I make. My analysis does not depend on any specific
syntactic theory and no operation other than function application is needed. Thus, syntactic
node labels are largely irrelevant, and I omit them entirely in the trees given below.
(i) Verbs denote (verbal) measure functions of type µ (= 〈e, 〈i, d〉〉).
(ii) Path phrases are syntactically treated as adjuncts, and semantically denote verbal
measure function modifiers of type 〈µ, µ〉.13
13This assumption may receive some support from the fact that path phrases are unlike subcategorized
direct objects in not obeying the double-o constraint, as noted by Matsumoto (1996a) (citing Poser (1983)):
22
(iii) Degree modifiers are of type 〈µ, 〈e, 〈i, t〉〉〉. They have the same type as the verbal pos
operator.
(iv) Measure phrases denote degrees of type d and are arguments of the Meas operator of
type 〈d, 〈µ, 〈e, 〈i, t〉〉〉〉 (which is of the same type as the pos operator except that it
takes a degree as an additional argument).
The last two assumptions are compatible with the standard assumptions about the functions
of measure phrases and degree modifiers in the semantics of gradable adjectives (see, e.g.,
Kennedy and McNally (2005)).
I start with composition with path phrases. As noted above, I assume that a path phrase
denotes a measure function modifier of type 〈µ, µ〉 that takes a verbal measure function as
an argument and composes it with a measure function that it lexically contributes (which
measures the progress of motion along some physical path it names), returning the composite
measure function as an output. For example, the denotation for the path phrase sono
17Let d =M(k)(t) and M1 = move(the-trail). Then, from (23c)
d−MIN(M)
MAX(M)−MIN(M)=
π1(d)−MIN(M1)
MAX(M1)−MIN(M1)
Since d = MIN(M)+MAX(M)2
d−MIN(M)
MAX(M)−MIN(M)=
1
2
Thus,
π1(d)−MIN(M1)
MAX(M1)−MIN(M1)=
1
2
Thus,
M1(k)(t) = π1(d) =MAX(M1) + MIN(M1)
2
31
otherwise itidan(M)(x)(t) is undefined.
The modified measure function is identical to the original measure function (and hence does
not change its endpoint-oriented scale structure), but it imposes an additional condition on
the temporal interval t over which it is evaluated. That is, t has to consist of subintervals
t1, . . . , tn which form a contiguous non-overlapping chain such that at each of these subin-
tervals ti, the event encoded by M progresses by the unit corresponding to one step (of
the stairs). I omit the analysis of a full sentence since the relevant compositional details of
measure function modification is the same as in the above examples.
Finally, I sketch briefly how path-denoting adverbial clauses may be analyzed via the
notion of scale composition. The role that these adverbials play in the sentence is similar
to the role of prepositional phrases in English in examples like the following:
(43) a. The train ran through the tunnel.
b. John walked over the mountain.
c. John went across the boarder.
d. The ball rolled down the slope.
This is consistent with the observation often made in the literature that English has many
prepositions specifying the structures of paths whereas corresponding information tends to
be expressed by verbs in Japanese (see, e.g., Kageyama et al. (2011)).
Zwarts (2005, 2008) analyzes the compositional semantics of English prepositional phrases
in detail by formalizing the notion of ‘path’ in algebraic terms. Zwarts’s notion of ‘path’
is largely compatible with the notion of scale in the present approach. Though technical
implementations differ in some details, in both approaches, the most fundamental kinds of
information encoded in the abstract representations posited are the endpoint-oriented and
incremental structures of objects which (in the case of path-denoting expressions) correspond
to actual physical paths.
32
Thus, for the most part, Zwarts’s analysis of English prepositional phrases can be re-
cast in the present scale-based setup straightforwardly and this gives us a starting point for
analyzing the semantics of Japanese path-denoting adverbial clauses in a way analogous to
English prepositional phrases. I will only sketch a beginning of this project since incorpo-
rating the whole algebra of paths developed by Zwarts in the present scale-based system is
beyond the scope of this paper.
Of particular interest to us is that, as noted by Zwarts, such path-denoting prepositional
phrases often exhibit variable telicity effects; the facts are largely parallel in Japanese path-
denoting adverbial clauses. For example, the prepositions up and down are similar to degree
achievements in that, when combined with motion verbs, they express incremental change
of height. As may be expected, when the object of the preposition is definite, we see the
familiar variable telicity effect:
(44) a. The car went/ran up the hill in/for 30 minutes.
b. Kuruma-wacar-TOP
30-pun(-de)30-minute-in
saka-oslope-ACC
nobot-teascend-TE
hasit-ta.run-PAST
‘The car went up the slope in/for 30 minutes.’
Zwarts (2005) also notes that prepositions such as over, across, and through, while normally
entailing that the relevant path has been totally covered, can also induce atelic, partitive
readings in certain contexts. At least some of the corresponding Japanese path denoting
adverbial clauses seem to be similarly ambiguous.18
(45) a. John walked over the hill for/in one hour.
b. John-waJohn-TOP
3-pun(-de)3-minute-in
hasi-obridge-ACC
watat-tecross-TE
arui-ta.walk-PAST
‘John walked over the bridge for/in 3 minutes.’
The patterns of aspectual composition found here can be analyzed by means of scale
composition, in a way essentially parallel to the other cases discussed above. For example,
18Wataru ‘go across’ and tooru ‘go through’ seem to exhibit this ambiguity more readily whereas yokogiru
‘cross’ seems to be unambiguously telic. I currently do not have an explanation for this variability.
33
for (45b), we can analyze the path denoting adverbial clause as a measure function modifier
in (46) (with y designating the goal of upward movement existentially bound at the top S
level, as in the previous examples), which combines with the measure function denoted by
the verb to yield the composite measure function in (47).19
(46) λMλxλt[M+move(the-slope)+ascend(y)]
(47) λxλt[path(run)+move(the-slope)+ascend(y)]
The measure function from saka-o ‘the slope’ is bounded on both ends, and this scale
structure is inherited to the whole composite measure function denoted by the sentence.
The variable telicity effect in (45b) is then predicted in the same way as in other examples
above. A similar account goes for the other examples in (44) and (45).
To summarize, the present scale-based approach captures the telicity of motion verbs by
the structures of the scales underlying verbal measure functions. The structures of lexically-
encoded measure functions can be modified by expressions that combine with the verbs, and
this accounts for the various effects on telicity in aspectual composition. The notion of scale
composition proposed in this paper provides a conceptually natural and formally explicit
way of capturing these patterns. In the rest of this section, I discuss how this approach
extends to other cases of aspectual composition, by sketching an analysis of incremental
theme verbs.20
3.5 Other applications of scale composition
The notion of scale composition proposed above can be extended to other types of aspectual
composition. As already noted, English directional prepositional phrases seem to constitute
an empirical domain which naturally lends itself to this approach. Here, I sketch an extension
19In addition, the conditions on measure function composition in (23) need to be generalized to cases
involving more than two measure functions. I omit the details, but this extension is straightforward.20Japanese verbs corresponding to degree achievements such as atatamaru ‘warm’ have different aspectual
properties than English degree achievements (Kageyama 1996; Sugioka 2009). But as noted in Kubota
(2010), they exhibit variable telicity in construction with the auxiliaries iku and kuru. Such facts can also
be accounted for naturally in the present scale-based approach. See Kubota (2010) for details.
34
to another major area of aspectual composition, namely, the treatment of incremental theme
verbs.21
For accomplishment verbs that take incremental themes such as taberu ‘eat’, tateru
‘build’ and yomu ‘read’, I assume that the verb itself denotes a measure function with
an open-ended scale that measures how much of the process (ultimately leading to the
culmination of the event) the subject of the sentence has engaged in during the relevant
interval. Thus, lexically, these verbs have the same type of meanings as activity verbs. The
endpoint is not lexically encoded in the verb but comes from the incremental theme.
For the incremental theme, I assume that, corresponding to the move function for path
phrases, there is a function called affect that applies to the denotation of an incremental
theme object and converts it to a measure function that keeps track of the progress of an
event involving some agent that incrementally affects it, based on the material part-whole
structure of the object. This affect function is essentially a transitive version of Kennedy’s
(2012) intransitive partof function. Thus, just like path phrases, the incremental theme
denotes a measure function modifier of type 〈µ, µ〉:22
Then the meaning of (49) is calculated as in (50).
(49) John-gaJohn-NOM
sonothat
hon-obook-ACC
yon-daread-PAST
21See Rothstein (2004, 2012) for a closely related approach to analyzing the internal structure of accom-
plishments by establishing a link between the ‘process’ component and the ‘incremental change’ component
of the meanings of accomplishment verbs via a homomorphic mapping. Although Rothstein’s analysis is not
couched in scale-based terms, the gist of her proposal is largely compatible with the analysis I outline below.22One might think that this analysis is inadequate since it does not explicitly identify the incremental theme
object as the theme of the ‘process’ part of the meaning of the verb contributed by the measure functionM
in (48). I take this to follow from the commensurability condition on scale composition formulated in (22).
That is, in order for the two measure functions to compose into a single composite measure function, they
have to each pertain to different aspects of some common incremental change. The only way in which this
condition can be satisfied with the two measure functions read and affect(the-book) in (50) is for the book
in question to be identified as the object of the reading process measured by the first measure function.
(54) Lexical Conceptual Structure for manner of motion verbs (e.g. hasiru ‘run’)MOVE ([Thing α ])
AFF ([Thing ]α, [ ] )[FOR
[GO ([Thing α ], [Path ] )
]]
Here, MOVE and AFF are primitive predicates. MOVE represents physical motion that
is not necessarily related to change of location and AFF stands for a two place ‘affect’
relation that holds between an Actor and a Patient. In short, for both simple activity verbs
and manner of motion verbs, the subject of the sentence is characterized as an entity that
performs some physical activity that (potentially) affects some other entity. These common
components in the lexical representations are supposed to capture the commonality between
the two types of verbs.
But Ueno’s analysis also crucially distinguishes these two types of verbs by addition-
ally specifying the GO relation (embedded under the FOR predicate which is supposed to
represent a ‘purpose’ relation) for manner of motion verbs. As it stands, however, (54) is
basically a representation of an activity-type verb since the main predicate of the whole
lexical representation is MOVE (which, unlike what the name suggests, does not encode
change of location). In order to capture the commonality between manner of motion verbs
and directional motion verbs more explicitly (and thereby making it possible to formulating
a uniform analysis of their similar behavior reflected, for example, in their compatibility
with path phrases), Ueno introduces an operation called ‘Clause Inversion’, which is essen-
tially an operation that flips the superordinate/subordinate relation between the MOVE
(activity) and GO (change of location) predicates in the lexical representation of manner of
motion verbs. By applying Clause Inversion to (54), we obtain the output in (55):
(55)
GO ([Thing α ], [Path ] )
AFF ([Thing ]α, [ ] )
BY
[MOVE ([Thing α ])
]
41
(55) is a representation of a motion verb since the main predicate of the whole lexical rep-
resentation is now GO (which is a primitive representing directional motion in Jackendoff’s
(1990) framework).
Thus, Ueno’s proposal captures the dual status of manner of motion verbs by directly
representing both the manner meaning and the change of location meaning in the lexical
representation of these verbs. As noted already, this goes against the manner/result comple-
mentarity principle. Since manner/result complementarity is itself an empirical hypothesis
whose status is still controversial (see Beavers and Koontz-Garboden (2012); Husband (to
appear) for some recent discussion), violating it is not necessarily a problem. However, once
one admits lexical representations of the form of (54)/(55), one immediately faces the issue
of how to constrain the availability of this type of representation to a particular class of
verbs so that the otherwise robust generalization embodied in the manner/result comple-
mentarity is captured. As we see below, this is precisely the source of the theoretical and
empirical problems that Ueno’s analysis suffers from.
The theoretical problem for Ueno’s proposal is that, in addition to the manner/result
complementarity, it violates another otherwise robust generalization governing the organi-
zation of the lexicon. That is, word formation processes are generally known to obey the
principle of monotonicity, which dictates that a morphological operation (including zero
derivation) can only add information to its input. Koontz-Garboden (2007) formulates this
principle explicitly as follows:
(56) The principle of monotonic composition (Koontz-Garboden 2007)
Word meaning is constructed monotonically on the basis of event structure constants
and operators.
This principle can be thought of as a counterpart (within the domain of morphology) of
the principle of compositionality at the phrasal level. Like compositionality at the phrasal
level, it is a guiding principle whose validity is ultimately subject to empirical challenge,
but given that abiding to it leads to a much more principled organization of the lexicon,
apparent counterexamples should be scrutinized carefully. Ueno’s Clause Inversion operation
42
violates the monotonicity principle since it looks inside the complex internal structure of the
lexical representation of a manner of motion verb like (54) and destructively reverses the
superordinate/subordinate relationship between the two key primitive predicates in it.24
Introducing such a powerful operation in the grammar may be justified if it is the only
way of providing an adequate solution for the empirical problem at issue. However, it turns
out that Ueno’s analysis is empirically inadequate as well. The problem comes from the
patterns of directional auxiliaries from section 2. Specifically, once one admits complex
representations like (54)/(55) that simultaneously encode an activity-type meaning and
a (directional) motion meaning as possible lexical meanings of simplex verbs, there is in
principle no reason why such representations cannot be assigned as the meanings of more
complex, derived linguistic expressions such as *moete-iku (in the intended sense of ‘go
burning’ in (5c)). The representation in (55) seems exactly right for such a composite verb,
but then, it remains mysterious why this structure cannot be assigned as the representation
for such complex verbs.
Before concluding the discussion in this section, I would like to clarify one key difference
between Ueno’s proposal and the scale-based analysis that I have proposed above.25 In
a sense, my scale conversion operation is similar to Ueno’s Clause Inversion in that both
are operations for converting a ‘manner’-type representation to a ‘change-of-location’ type
representation for manner of motion verbs. One might then think that the present proposal
may run into similar problems as Ueno’s proposal. However, note crucially that, unlike
Ueno’s account, the ‘manner’-type meaning and the ‘result’-type meaning are never simul-
taneously present in the representation of the lexical meaning of a verb (formalized as a
measure function) in the present analysis. This way, we can maintain the manner/result
complementarity at the level of lexically-encoded meanings of verbs but at the same time
provide a principled explanation for the apparently problematic status of manner of motion
24Note that, in this respect, this operation is crucially different from the related operations of lexical
subordination (Levin and Rapoport 1988) and the GO-Adjunct rule (Jackendoff 1990) that Ueno argues are
inappropriate for characterizing the properties of manner of motion verbs in Japanese.
25I’d like to thank two JEAL reviewers for pushing me to clarify on this point.
43
verbs. Moreover, as already discussed in section 3, the operation of scale conversion is con-
strained to be available only for lexically triggered entailments. Thus, the issue of potential
overgeneration of the sort discussed above does not arise in the present proposal.
5 Conclusion
In this concluding section, I would like to discuss some typological implications of the
present scale-based approach to verb meaning. The patterns of motion verbs have often been
discussed in the literature in relation to Talmy’s (1985; 2000) typological distinction between
verb-framed and satellite-framed languages (cf. Kageyama (1997); Matsumoto (1997); Ueno
(2007); Kitahara (2009) for some recent discussions on Japanese). In Talmy’s typology,
Japanese is classified as a verb-framed language, in which the notion of path is primarily
encoded in the verb rather than in a satellite (such as particles and pre-/postpositional
phrases). In particular, it does not allow for the pattern exhibited by run to the station
in English (a typical satellite-framed language), where the verb encodes the manner of
motion and the path is expressed by the satellite. The typological distinction is taken to
extend to other constructions as well, most notably the resultative construction: it has
been observed in the literature (Kageyama 1996; Washio 1997) that the so-called strong
resultatives (Washio 1997) in which an activity-type verb that does not entail any change of
state on the object combines with a result phrase (as in hammer the metal flat in English)
is impossible in Japanese.26 This is naturally expected given that change of location and
change of state are closely related notions. That is, here again, the satellite-framed pattern of
English (in which a purely manner-denoting verb combines with a ‘path’-denoting satellite)
26Note also in this connection that English allows for this type of meaning composition in motion con-
structions too. As discussed and analyzed in detail in Isono (2013), even purely activity-type verbs that do
not express any notion of change of location such as verbs of sound emission and light emission can combine
with path phrases:
(i) a. The cart rumbled down the street.
b. A shooting star flashed across the sky.
44
contrasts with the verb-framed pattern of Japanese (in which a verb that is not associated
with a ‘path’-oriented meaning fails to license resultatives).
The facts about resultatives are actually somewhat subtler than this. In particular,
as noted by Washio (1997), the set of verbs that license resultatives in Japanese is larger
than the set of change of state verbs. While activity verbs such as tataku ‘hammer’ and
hikizuru ‘drag’ do not license resultatives, in addition to (unequivocally) change of state
verbs such as kooru ‘freeze’ and kogeru ‘burn’ in (57), there is a set of verbs (many of which
are surface-contact verbs) that lexically do not entail any definite change of state but which
nevertheless license resultatives, including migaku ‘polish’, huku ‘wipe’ and haku ‘sweep’ in
(58):27,28
(57) a. Ike-gapond
katikati-nisolid
koot-ta.freeze-PAST
‘The pond froze solid.’
b. Zaimoku-gatimber-NOM
makkuro-niblack
koge-ta.burn
‘The timber burned black.’
(58) a. John-gaJohn-NOM
kutu-oshoe-ACC
pikapika-nishiny
migai-ta.polish-PAST
‘John polished the shoes shiny.’
b. John-gaJohn-NOM
yuka-ofloor-ACC
tiri hitotu nakuwithout a speck of dust
hai-ta.sweep-PAST
intended: ‘John swept the floor without a speck of dust.’
c. John-gaJohn-NOM
teeburu-otable-ACC
kireiniclean
hui-ta.wipe-PAST
intended: ‘John wiped the table cleanly.’
Washio (1997, 10) notes that although these latter verbs do not necessarily entail any change
27Washio (1997) provides evidence that these verbs are aspectually atelic (unlike genuine change of state
verbs, which are telic) and should be classified as activity verbs rather than accomplishments.28Kageyama (1996) claims that kireini in (58c) is a manner adverb rather than a result phrase, but his
argument is based on a dubious assessment of data where he takes other result phrases such as pikapika-ni
in (58a) to be incompatible with this class of verbs.
45
of state, they nonetheless ‘imply that if the state of the patients do change, then they would
change in certain fixed directions’, in other words, that they ‘have a “disposition” toward
certain states’. In what follows, I call this class of verbs ‘implicit change of state verbs’.29
The intermediate nature of implicit change of state verbs is somewhat reminiscent of the
status of manner of motion verbs. In both cases, the verb is lexically an activity predicate
which is aspectually atelic, but at the same time the notion of change (of location or state)
is implicitly present as a derived entailment (albeit as a modalized one in the case of implicit
change of state verbs).
The above observations suggest a somewhat different characterization of the parallel
between resultatives and motion constructions in Japanese than Talmy’s original binary
classification of language types. In particular, the fact that verbs exhibiting the inter-
mediate status and which are aspectually not prototypical change of location/state verbs
similarly license composition with path/result-oriented expressions (i.e. path phrases for
manner of motion verbs and result phrases for implicit change of state verbs) is intriguing.
Ideally, an analysis of resultatives in Japanese should capture this parallel in some theo-
retically principled way. Such an analysis in fact suggests itself quite naturally within the
present scale-based setup. Specifically, the close parallel between motion constructions and
resultatives can be captured by assuming that, just like motion constructions, resultatives
in Japanese involve the operation of scale composition as the key mechanism of meaning
composition. This indeed seems exactly the right analysis for change of state verbs like
kooru ‘freeze’ and kogeru ‘burn’. With change of state verbs, both the verb and the result
29The notion of implicit change here seems to be closely related to the notion of sublexical modality
discussed by Koenig and Davis (2001), in particular, their energetic modals, whose modal bases consist of
worlds ‘in which the actions or situations denoted by the verb achieve the goals that motivate those actions
or the inherent consequences which result from the occurrence of the situation’ (Koenig and Davis 2001, 89).
I take it that Washio’s notion of ‘disposition’ toward certain fixed states can be formalized more precisely
in modal semantics along the lines suggested by Koenig and Davis. Koenig and Davis invoke the notion
of sublexical modality to overcome certain difficulties of a purely entailment-based treatment of argument
linking. The fact that such closely related notions have independently been proposed in separate empirical
domains suggests that this is indeed a linguistically relevant concept.
46
phrase denote changes of state. The composite measure function that results from scale
composition then measures the progress of change of state from two different perspectives
simultaneously.30
The fact that implicit change of state verbs license resultatives can then be accounted
for in terms of a measure function conversion operation in a way essentially parallel to the
treatment of manner of motion verbs. Here again, what triggers the measure function con-
version is the derived entailment of the verb. Specifically, due to the modalized entailment
of ‘change toward a certain fixed state’, the activity-type measure function originally speci-
fied as the meaning of the verb can be converted to a change of state measure function that
measures the progress of the resultant change. Note crucially that, by taking the modal
component into account, the derived entailment of change is an uncancellable entailment:
in Koenig and Davis’s (2001) terms, the consequent change inevitably takes place in all
worlds that are in the modal base of the ‘energetic’ modal lexically encoded in the verb.
The derived measure function is then composed with the measure function of the result
phrase to yield a composite measure function.
There are a number of details that need to be spelled out in order to work this out as
an adequate analysis of resultatives, including the issues of how to represent verb meanings
with modal components in a scale-based setup and at which step of meaning composition
an implicit change is turned into an actually entailed change (which seems to be necessary
in order to ensure that composing the derived verbal measure function with that of the
result phrase yields a coherent composite measure function). I have to leave these issues for
future work. The point here is that, if this line of analysis is generally on the right track,
it offers an attractively simple explanation for the parallel between motion constructions
and resultatives in Japanese.31 The perspective offered in such an approach is more refined
30See Uegaki (2014) for a recent scale-based analysis of resultatives in Japanese. While I believe that the
approach I have speculated on here shares the basic analytic intuition with Uegaki’s proposal, there are certain
technical assumptions that he makes which makes the comparison of the two not totally straightforward. A
careful investigation of the semantics of resultatives is left for future work.31It is worth noting in this connection that Kageyama (1996, 1997) seeks to explain the typological pattern
of Japanese in terms of available options for LCS composition, along lines similar to the present proposal
47
than the original Talmian classification, and is also in line with the recent rethinkings on
the Talmian typology (see, e.g., Beavers et al. (2010)) that attempt to account for the
typological variation observed across languages from more general principles pertaining to
the organization of the lexicon and the set of morpho-syntactic resources available in specific
languages. On the view that I have argued for here, the essentially verb-framed nature of
Japanese derives from the fact that verbs themselves must (at least implicitly) encode a
change-type measure function in their lexical meanings in order to license path phrases
or result phrases. This retains the key insight of the Talmian classification, but at the
same time, it naturally accommodates the fact that classes of verbs exhibiting intermediate
properties exist both in the domain of motion verbs and in the domain of verbs licensing
resultatives, a fact which arguably goes beyond the coarse binary classification of language
types entailed by the original Talmian typology.
References
Beavers, John. 2008a. On the nature of goal marking and event delimitation: Evidence from Japanese.
Journal of Linguistics 44:283–316.
Beavers, John. 2008b. Scalar complexity and the structure of events. In J. Dœlling and T. Heyde-
Zybatow, eds., Event Structures in Linguistic Form and Interpretation, 245–265. Berlin: de
Gruyter.
Beavers, John. to appear. Aspectual classes and scales of change. To appear in Linguistics.
Beavers, John and Andrew Koontz-Garboden. 2012. Manner and result in the roots of verbal mean-
ing. Linguistic Inquiry 43:331–369.
Beavers, John, Beth Levin, and Shiao Wei Tham. 2010. The typology of motion expressions revisited.
Journal of Linguistics 46:331–377.
in certain respects. In a nutshell, Kageyama’s proposal is that the verb-framed pattern of Japanese derives
from the fact that it lacks the operation of freely composing process-type representations of activity verbs
with result-denoting representations of the satellites to yield accomplishment-type meaning representations.
The underlying intuition that the verb itself has to encode the notion of change in order to combine with
path/result-denoting satellites is similar to the present proposal. However, in implementing this idea in
a decomposition-based setup, Kageyama requires that verbs that can appear in resultatives are uniformly
change of state verbs, an assumption that Washio (1997) convincingly demonstrates is problematic.
48
Bobaljik, Jonathan. 2012. Universals in Comparative Morphology: Suppletion, superlatives, and the
structure of words. Cambridge, Massachusetts: MIT Press.
Bochnak, M. Ryan. 2013. Two sources of scalaritywithin the verb phrase. In B. Arsenijevic,
B. Gehrke, and R. Marın, eds., Studies in the Composition and Decomposition of Event Pred-
icates, 99–123. Dordrecht: Springer.
Demonte, Violeta and Louise McNally, eds. 2012. Telicity, Change, and State: A Cross-Categorial
View of Event Structure. Oxford: Oxford University Press.
Hay, Jennifer, Christopher Kennedy, and Beth Levin. 1999. Scalar structure underlies telicity in
‘degree achievements’. In T. Matthews and D. Strolovitch, eds., Proceedings from Semantics and
Linguistic Theory IX , 127–144. Ithaca, New York: CLC Publications.
Henderson, Robert. 2013. Quantizing scalar change. In T. Snider, ed., Proceedings of SALT 23 ,
473–492. eLanguage.
Husband, E. Matthew. to appear. Rescuing manner/result complementarity from certain death. In
Proceedings from the Annual Meeting of the Chicago Linguistic Society , vol. 47, ??–??
Isono, Tatsuya. 2013. Polysemy and Compositionality . Tokyo: Hituzi Syobo.