Version 2.1
[In this version, much or some of the materials have been deleted up to the end of section 3 of version 1.2 in
order to make the present organization of the book more transparent.]
Falsification, Corroboration and Repeatability in Generative
Grammar: an illustration by the so-called raising-to-object
construction in Japanese
Hajime Hoji
University of Southern California
http://www.gges.org/hoji/
1. Introduction
1.1. Goal of generative grammar
1.2. Goal of this book
1.3. The object of inquiry
1.4. Negative and positive predictions
1.5. Basic idea
1.6. The structure of an experiment
1.7. Some objections
1.8. Repeatability
1.9. Summary
2. So-called raising-to-object construction in Japanese
2.1. Various analyses
2.2. Summary
3. Falsification of any raising analysis
3.1. Experiment 4: Preliminary experiment (on PBC effects)
3.1.1. Exp 4: Prediction
3.1.2. Exp 4: Examples
3.1.3. Exp 4: Results
3.2. Experiment 5: Crucial experiments
3.2.1. Exp 5: Prediction
3.2.2. Exp 5: Examples
3.2.3. Exp 5: Results
3.3. Experiment 6: A follow-up experiment
3.3.1. Exp 6: Examples
3.3.2. Exp 6: Results
3.4. An attempt to save the raising analysis from being falsified by the PBC-related experiments
3.5. Summary
4. Falsification of any 'ECM' analysis
5. Corroboration for the Major Object hypothesis
5.1. Experiment 1: Preliminary experiment with XP-sika
5.1.1. Exp 1: Prediction
5.1.2. Exp 1: Examples
5.1.3. Exp 1: Results
5.2. Experiment 2: Crucial experiment with NP-o-sika
5.2.1. Exp 2: Prediction
5.2.2. Exp 2: Examples
5.2.3. Exp 2: Results
5.3. Experiment 3: Crucial experiment with Rokuna NP
5.3.1. Exp 3: Prediction
5.3.2. Exp 3: Examples
5.3.3. Exp 3: Results
5.4. A theoretical characterization of Mob and the Major Object construction.
5.5. Summary
6. An attempt to falsify the Major Object hypothesis
7. Some Details: the major subject and the major object
7.1. Yoon 2004 and 2005
7.1.1. Yoon's critique of the Major Object hypothesis
7.1.2. Problems for Yoon's (2005) proposal
7.2. Yoon 1989
8. The significance of preliminary experiments
8.1. The structure of an experiment (this has already been given above once)
8.2. Local disjointness effects
8.3. Inverse scope
9. Concluding remarks
10. Appendix: On the 'semantic' restriction on Mob
11. Taken out from the LSA handout
11.1. Goal
11.2. A few residual (or remaining) issues
11.2.1. On the parallelism between a major subject and Mob
11.2.2. On the (16) + [(14) or (17)] analysis
11.2.3. On the (neo)ECM analysis
11.3. Outline
1. Introduction
The content and the exposition of the introduction will be changed, based on how the book will turn out to be.
At the moment, the goal of the book is to suggest and illustrate a means to test a hypothesis in generative
grammar, addressing the questions of when a hypothesis should be considered falsified, corroborated or
neither, to be determined on the basis of the result of an experiment, whose design is to be specified by the
hypothesis in question, together with the auxiliary hypothesis or hypotheses. The discussion draws from
concrete empirical materials having to do with the so-called raising-to-object construction in Japanese (and
to a somewhat lesser degree in Korean). Those not familiar with Japanese (or Korean) would likely wonder
why they would have to read this book in order to get an idea about a general methodology in generative
grammar. I.e., why would a non-JP linguist want to listen to a methodological proposal that is crucially
based on empirical materials in a language that they are not familiar with? In a way, it would be much
more effective if I could illustrate my points on the basis of empirical materials in English. But we know
that I would in that case not be able to do what I can do with Japanese.
I think we are actually talking here about the strength and the weakness of what we do.
--To put it loosely for now,
--Strength: We pursue empirical generalizations in Japanese that are (significantly) more solid and rigorous
than most other works.
For this reason, our theoretical accounts of the hypothesized/claimed
generalizations often turn out to be on the right track or at least they usually do not turn out to be totally off
the mark. More importantly, we can rely on the expected/predicted correlations of judgments as a basis for
evaluating both the validity of a given generalization and our proposed account of it.
New Book1 Outline version 2.0
Hoji 2/26
--Weakness: Since our strength is closely connected with empirical materials in Japanese, only those who
have intuitions about the language can truly appreciate what we do especially when conflicting judgments
are reported by (numerous) other Japanese linguists in the field. Since what we propose is based on indepth empirical examination and since we try to adopt a generalization only on the basis of rigorous
empirical tests, we tend not to accept claims and generalizations about other languages unless they see some
indirect evidence for their validity in Japanese. This makes it more difficult than otherwise for us to
communicate with non-Japanese linguists or for non-JP linguists to appreciate what we do -- it is note
worthy that I seldom get questions about empirical issues about Japanese when I give a presentation before
an audience that consists mainly of non-JP linguists --; after all, they can only understand our empirical
claims through the lenses of their understanding of English. By contrast, the works by those linguists who
make frequent and regular reference to the 'known' generalizations in English (and other 'well-studied'
languages) are easier for the non-JP linguists to understand and hence cite.
1.1. Goal of generative grammar
If "the aim of science is, on the one hand, a comprehension, as complete as possible, of the connection
between the sense experiences in their totality, and, on the other hand, the accomplishment of this aim by the
use of a minimum of primary concepts and relations,"1 as Einstein (1936: 293) puts it, and if generative
grammar is that part of science whose aim consists of a comprehension of the connection between the sense
experiences as reflections of the language faculty, it follows that one of our tasks is to identify what the
relevant sense experiences are; cf. Chomsky 1955/1975: 37. Since our sense experiences, such as
introspective judgments about a given sentence in a given language, most likely reflect more than the
language faculty proper, such a task necessarily involves hypotheses about the nature of the relevant sense
experiences, in particular, hypotheses as to which aspects of our sense experiences under discussion are
reflections of our grammar, and in what theoretical terms they are to be expressed. At a particular stage of
theory construction, a given factor can be reasonably considered grammatical in nature only if it can be
expressed in terms of concepts postulated within the grammatical theory being developed.2 Every concept
and relation postulated in the theory, in turn, must be tightly related to the native speaker's linguistic
intuitions—often quite indirectly—as reflections of his/her grammar. It is in fact the tight connection
between (i) theoretical concepts and relations on the one hand and (ii) the speaker's linguistic intuitions on
the other that makes it possible to put forth definite predictions about the latter that are formulated in terms
of the former, thereby making the proposed theory/hypothesis falsifiable.
1.2. Goal of this book
In what follows, I would like to suggest and illustrate a means to test a hypothesis in generative
grammar, addressing the questions of when a hypothesis should be considered falsified, corroborated or
neither, to be determined on the basis of the result of an experiment, whose design is to be specified by the
hypothesis in question, together with the auxiliary hypothesis or hypotheses. This is in relation to the
problem in (1a). Once the method of evaluating a hypothesis is applied to two competing hypotheses, we
can arrive at a relative assessment of them, hence will be in a position to address the problems in (1b).
(1)
a.
b.
Problems with assessing a given hypothesis
Problems with comparing two competing hypotheses
--It is often not clear how a given proposal can be empirically tested.
1
Similarly, it is often not clear how to
The emphases are as in the original.
2
Chomsky (1955/1975: 61) thus states that "a field of investigation cannot be clearly delimited in advance
of the theory dealing with the subject matter; in fact, it is one of the functions of a theory to give such a
precise delimitation."
New Book1 Outline version 2.0
Hoji 3/26
arrive at a relative assessment of two (seemingly) competing hypotheses.
such as (2).
(2)
One thus encounters remarks
The merit of a hypothesis has to be determined in the context of the entire theory in which the
hypothesis is embedded; and hence it is not a simple matter to put the hypothesis to an empirical
test.
--What underlies remarks like (2) is xxx?
==>Provide the two illustrations by means of a dialogue; see the Mayfest handout and also the syllabus for
EALC 620-05f.
--This book is thus an attempt to deal with problems such as (1), and to the extent that successfully
answering or coping with the problems in (1) is a necessary condition for making what we do an empirical
science with progress in mind, this book is an attempt to pursue a way to make generative grammar such a
science or ensure that it is.
1.3. The object of inquiry
--What is it that our hypothesis is about? What are the sense experiences as reflections of the language
faculty (or more accurately put, reflections of the formal aspects of the language faculty, if we assume that
there are formal as well as functional aspects to the language faculty)? How do we know what sense
experiences of ours count as such? How should we understand the relation between a theory/hypothesis
and what it is designed to describe and explain so as to be clear about how the theory/hypothesis can be
tested against the sense experiences under discussion?
--What requirements do we want to impose upon how a descriptive generalization is to be stated, in
accordance with our understanding of the above issues? What about requirements on a proposition in the
theory?
1.4. Negative and positive predictions
--Considerations about how a hypothesis can be tested and what significance can or cannot reasonably be
assigned to the result of an experiment lead us to conclude that it is negative predictions that we should be
focusing on rather than positive predictions.
--To be continued. (Explaining why that is.)
--For an experiment on the researcher-informant, a positive prediction can be tested meaningfully, and a
falsifying result of the experiment can be significant.
--We might also obtain the same effects in an experiment on a set of informants if the preliminary
experiments have been done correctly.
--Otherwise, it is not clear how significant a falsifying result of an experiment on a positive prediction might
be, unlike a falsifying result of an experiment on a negative prediction.
--It is also important that we go beyond obtaining confirmation for our negative prediction. ==>This leads
to the need t consider corroboration.
1.5. Basic idea
(3)
a.
b.
Eg*: an example in an experiment that is predicted to be impossible (under a specified
interpretation)
Egn : an example that forms a minimal pair with an Eg*n
New Book1 Outline version 2.0
Hoji 4/26
(4)
Falsification (the general idea)
A hypothesis is falsified iff Eg* is judged to be not as hopeless as predicted (under the specified
reading).
(5)
Corroboration, not in the sense of Popper (the general idea)
A hypothesis is corroborated iff Eg*n is judged unacceptable while Egn is judged to be
significantly more acceptable (under the specified reading).
1.6. The structure of an experiment
I suggest/propose that a given experiment be designed so as to make it maximally possible to test
particular singular hypothesis. Thus one prototypical experimental design would look like the following.
Hmain: the hypothesis being tested
Hsub: the hypothesis that has been independently corroborated or assumed to be valid
(6)
a.
b.
(7)
An experimental specification in accordance with Hmain and Hsub:
(i) Sentences of such and such form are unacceptable (under a specified interpretation).
(ii) Sentences of such and such form that are minimally different from those noted in (i) are not
necessarily unacceptable (under a specified interpretation).
--It may be the case that Hsub consists of more than one hypothesis.
--But if we want the result of an experiment to be directly relevant to a given hypothesis H, we must take it
to be Hmain and the other hypothesis or hypotheses to be Hsub, and the latter either must be independently
corroborated or assumed to be valid. Otherwise, the experimental result cannot be taken as (directly)
revealing as to the validity of the hypothesis in question. This relates to the issues having to do with
preliminary experiments and the repeatability issues.
--It must also be noted that an experiment can be conducted to test the validity of a generalization. To be
continued.
The importance of conducting necessary preliminary experiments before conducting a crucial experiment.

If one uses as Hsub a hypothesis that is shaky at best or is easily falsified (if not having already been
falsified)—which can easily result in the disconfirmation of a prediction—it is not even clear what
significance the result of such an experiment might have.
<==The point here perhaps need to be emphasized more than we might initially think is necessary. Make
reference to what we can learn from a falsifying result of an experiment on a given Hmain.
(8)
A specific format of an experiment
a. A set of example sentences are presented to informants, typically on a web page.
b. Informants are asked to judge each sentence by choosing one of the five radio buttons placed
under (i).3
(i)
Bad < ===== > Good
o o o o o
c. The five choices will be computed as in (ii), "2" corresponding to "Bad" and "+2" to "Good" but
the informants do not know what numeric values will be assigned to each of the five circles.
(ii)
2, 1, 0, +1, +2
3
No time limit is placed on the task. The informants can judge some of the examples on a CFJ during a
given visit to the web page, and can come back to the page to judge other examples later. They are also
allowed to change their judgments later.
New Book1 Outline version 2.0
Hoji 5/26

(9)
Important Clarification: An experiment can be conducted on just one informant who can be the
researcher him/herself.4
Falsification (a specific idea)
A hypothesis is falsified iff the average score for Eg* is greater than 1.0.
(10) Corroboration, not in the sense of Popper (a specific idea)
A hypothesis is corroborated iff the difference between the average score on Eg*n and that on Egn
is greater than 3.
1.7. Some objections
[What is given in (11) is based on Hornstein's comments at the LSA workshop.]
(11) a.
b.
c.
Why invent a new wheel? (A magnitude estimation method has been already developed, or
something like that.)
Ungrammaticality can be overridden (as has been shown in a number of experiments …);
Magnitude estimation is only a very first and very coarse way to test a hypothesis. Ultimate tests
will be offered by reading time tests, neuro xxx, etc. The reactions to grammatical factors are
extremely fast, as shown by psychological experiments, and it is not clear what the judgment tests
reveal (or something like that).
My responses at the moment:
To (11a) -- Need to find out what is out there, especially, if the method in question focuses on falsifiability.
To (11b) -- If a given sentence that is predicted by the theory to be unacceptable is judged to be not
unacceptable, fairly consistently, there must be something wrong with the theory.
To (11c) -- One might suggest that the real testing of our hypothesis is not done by means of acceptability
judgment tests, to begin with, but by something else such as what gets done in psycholinguistics field.
Notice that this move gives one the freedom to say anything he would like to say since he is not responsible
for empirical consequences. My view is that reading time tests, eye tracking, and other tests that are being
done in the field are only 'side evidence' and the main evidence comes from a carefully constructed sequence
of judgment tests.
1.8. Repeatability
This will perhaps be the most crucial part of the book…
"average"
--across occasions, across different realizations, across informants
1.9. Summary
2. So-called raising-to-object construction in Japanese
--Need to illustrate the general point(s) by some concrete empirical materials.
--Will use for that purpose the so-called raising-to-object construction in Japanese.
AU: 「この構文については、一見、starting assumption から大きく異なる分析が乱立していて、それぞれの
4
We can actually understand that one of the main purposes of conducting an experiment on a (sizable) set of
informants is for the researcher to obtain confirmation for the result of an experiment on him/herself.
New Book1 Outline version 2.0
Hoji 6/26
成果を比べることが難しいように見える。しかし、prediction という観点から整理しなおしてみると、一見
直接比較できないように思う分析でも、ちゃんと比べられるのだ、ということを示したい」みたいな説明が
あってもいいのではないかと思いました。
HH: I see your point. I will see what I can do.
(12)
John-wa Mary-o Itariazin da to omotteita.
John-TOP Mary-ACC Italian
be that thought
(Cf. John-wa Mary-ga
Itariazin da to omotteita.)
(13) a.
b.
John believed about Mary that she was an Italian.
John believed Mary to be an Italian.
2.1. Various analyses
(14) Raising analysis:
The o-marked NP5 in the structure in (12) (henceforth Mob) is 'base-generated' in 'its theta
position' in the embedded clause and gets raised to a position in the matrix clause.6
(15) ECM analysis:
Mob is 'base-generated' in 'its theta position' in the embedded clause and stays inside the
embedded clause.
(16) Major object analysis:
Mob is 'base-generated' in the matrix clause and is not part of the embedded CP at any stage of
derivation, and (12) corresponds more closely to (13a) than to (13b), in terms of the relevant
formal properties.7, 8, 9
(17) Movement-of-the-major-subject analysis:
Mob is 'base-generated' as the major subject in the embedded clause and gets raised to a position
in the matrix clause.10
(18) The combination of (14) and (15):
5
The use of NP instead of DP has no bearing in this presentation.
6
The proposal in Kuno 1976 is of this type.
7
Essentially the same analysis has been proposed for a similar construction in Madurese and other Javanic
languages in Davies 2001; see Davies and Dubinsky 2004: 243-251 and the references therein for more
details.
8
The proposal is hinted at in Saito 1983, and explicitly argued for in Hong 1990 and Hoji 1991, and further
defended in Takano 2003. According to Hoji 1991, Kitagawa (1985) also suggests or argues for the
possibility of the Major Object analysis, along with the ECM-type analysis.
9
[Refer to the talk at the LSA workshop on the 'quasi ECM' in Greek.]
10
(James) Yoon 2004 argues for (17) (though he also admits the option of the Major Object analysis for
Mob that denotes a human, as of February 2005, at the Kyoto Workshop).
New Book1 Outline version 2.0
Hoji 7/26
The option in (14) and the one in (15) are both allowed.11
--logically possible combinations of the options (cf. the supplementary handout at the LSA workshop)
--some (not extensive) remarks on what has to be made explicit minimally re. the formal properties under
each analysis, w.r.t. crucial aspects such as (i) argument structure and its realization, (ii) theta assignment,
and (iii) case/Case, (iv) movement, in order to yield an experimental specification under each analysis.
2.2. Summary
3. Falsification of any raising analysis
The goal of this section:
To show that the raising analysis gets falsified when it is combined with the PBC hypothesis.
(19) A typical configuration of 'scrambling' examples that exhibit PBC violation effects (cf. the works by
S.-I. Harada and M. Saito and the subsequent works.)
(i) [CP ... NP-NOM ... t2... Verb Infl C]1 ... NP-DAT2 ... NP-TOP/-NOM ... t1 … Verb
(20)
PBC Hypothesis
<==A theoretical characterization of the PBC or something like that is needed.
(21)
Hypothesis on 'scrambling' in Japanese that has the consequence of (22).
(22)
A surface string corresponding to (19) necessarily contains a trace in the position of t2 in (21).
(20) is Hmain and (21) is Hsub; so (21) must be corroborated independently, or will just have to be assumed to
be valid.
--The Deep OS option with NP-DAT2 being the DL is not possible due to an independently hypothesized
property of 'long-distance OS'.
--The Surface OS option with NP-DAT2 being the DL would result in PBC violation.
--Suppose we go with (20) as Hmain and (21) as Hsub. We will then have to have a hypothesis of the general
phrase structure of JP as Haux.

Part of an experiment conducted by Y. Tsuboi will be introduced below, as Experiments 4 and 5.
3.1. Experiment 4: Preliminary experiment (on PBC effects)
3.1.1. Exp 4: Prediction
Prediction: The sentence corresponding to the configuration in (19) is unacceptable.
--It is not a priori obvious whether an empty category, assuming that it is represented, is a trace or something
else. So, the relevant issue must be addressed and the point has to be demonstrated.
11
The proposal in Hiraiwa 2002 is of the type in (18), claiming that Mob always moves from 'its theta
position' to a/the Spec of the embedded CP, and what is optional is the subsequent movement of Mob out of
the embedded CP.
New Book1 Outline version 2.0
Hoji 8/26
3.1.2. Exp 4: Examples12
(23)
keisatu-wa [John-ga
tyuugoku-ni nigeta to] danteisita.
police-TOP [John-NOM China-to
escaped that] determined
'The police determined that John had escaped to China.'
(24) a.
tyuugoku-ni keisatu-wa [John-ga nigeta to] danteisita.
China-to
police-TOP [John-NOM escaped that] determined
b.
[John-ga tyuugoku-ni nigeta to] keisatu-wa danteisita.
[ John-NOM China-to
escaped that] police-TOP determined
c.
*[John-ga nigeta to] tyuugoku-ni keisatu-wa danteisita.
[John-NOM escaped that] China-to
police-TOP determined
(25)
Yamada sensei-ga [John-ga Toyota-ni syuusyokusita to] omoikondeita (koto)
Yamada professor-NOM [John-NOM Toyota-at got:job
that] believed
(that)
'(that) Prof. Yamada believed that John had gotten a job at Toyota '
(26) a.
Toyota -ni Yamada sensei-ga
[John-ga syuusyokusita to] omoikondeita (koto)
Toyota-at Yamada professor-NOM [John-NOM got:job
that] believed
(that)
b.
[John-ga Toyota -ni
[John-NOM Toyota-at
syuusyokusita to] Yamada sensei-ga
omoikondeita (koto)
got:job
that] Yamada professor -NOM believed
(that)
c.
*[John-ga syuusyokusita to] Toyota-ni Yamada sensei-ga
omoikondeita (koto)
[John-NOM got:job
that] Toyota-at Yamada professor-NOM believed
(that)
3.1.3. Exp 4: Results
(27)
Serial No.
1
2
Eg4
failed
3
Eg4
cleared
4
Eg*4
cleared
9
10
Eg12
failed
11
Eg12
cleared
12
Eg*12
cleared
18
18
18
18
18
18
18
18
0
0.96
0.68
1.46
0.23
0.95
The # of
informants
Standard
Deviation
Average
0
1.41
+2
+0.67
+2
1.56
+1.56
+0.5
+1.94
1.61
Example No.
(23)
(24a)
(24b)
(24c)
(25)
(26a)
(26b)
(26c)
The scores on (24a) and (26a) indicate that so-called long-distance scrambling is not accepted fully by every
informant.
If we focus on the informants who accept (24a) and (26a) fully, and consider their judgments on (24c) and
(26c), we obtain the following result.13
12
The examples in (24) and (26) have the same 'basic meaning' as (23) and (25), respectively.
13
The 'adjustments' under discussion can be summarized as in (i).
New Book1 Outline version 2.0
Hoji 9/26
(28) Results focusing on the informants who have given "+2" to (24a) and (26a):
Serial No.
1
2
Eg4
cleared
3
Eg4
cleared
4
Eg*4
cleared
9
10
Eg12
cleared
11
Eg12
cleared
12
Eg*12
cleared
18
7
18
7
18
6
18
6
The # of
informants
Standard
Deviation
Average
0
0
0
1.05
0.68
0
0.23
1.12
+2
+2
+2
1.57
+1.56
+2
+1.94
1.5
Example No.
(23)
(24a)
(24b)
(24c)
(25)
(26a)
(26b)
(26c)
3.2. Experiment 5: Crucial experiments
3.2.1. Exp 5: Prediction
Prediction under the raising analysis:
A raising analysis of RtoO in Japanese predicts the examples of the form (29) to be unacceptable due to the
PBC, violated by the unbound trace as postulated in the sentence-initial CP in (29).
[CP ... t2 ...Verb Infl C]1 ... NP-ACC2 ... NP-TOP/-NOM ... t1 … Verb
(29)
3.2.2. Exp 5: Examples
(30)
keisatu-wa John-o tyuugoku-ni nigeta to danteisita.
police-TOP John-ACC China-to
escaped that determined
(31) a.
b.
John-o
[tyuugoku-ni nigeta to]
John-ACC [China-to
escaped that]
c.
[tyuugoku-ni nigeta to] John-o keisatu-wa danteisita.
[China-to escaped that] John-ACC police-TOP determined
(32)
keisatu-wa danteisita.
police-TOP determined
Yamada sensei-ga
John-o Toyota-ni syuusyokusita to omoikondeita (koto)
Yamada professor-NOM John-ACC Toyota-at got:job
that believed
(that)
(33) a.
(i)
'The police determined about John that he had escaped to China.'
John-o keisatu-wa [tyuugoku-ni nigeta to] danteisita.
John-ACC police-TOP [China-to
escaped that] determined
'(that) Prof. Yamada believed of John that he got a job at Toyota'
John-o Yamada sensei-ga [Toyota-ni syuusyokusita to] omoikondeita (koto)
John-ACC Yamada professor-NOM [Toyota-at got:job
that] believed
(that)
b.
John -o
[Toyota-ni syuusyokusita to] Yamada sensei-ga
omoikondeita (koto)
John-ACC [Toyota-at got:job
that] Yamada professor-NOM believed
(that)
c.
[Toyota-ni syuusyokusita to] John-o Yamada sensei-ga
omoikondeita (koto)
[Toyota-at got:job
that] John-ACC Yamada professor-NOM believed
(that)
a.
b.
If [the answer of (24a)] < 2, then [the answer of (24a)] and [the answer of (24c)] will be excluded
from the average.
If [the answer of (26a)] < 2, then [the answer of (26a)] and [the answer of (26c)] will be excluded
from the average.
New Book1 Outline version 2.0
Hoji 10/26
3.2.3. Exp 5: Results
(34)
Serial No.
5
6
7
13
Eg8
failed
8
Eg*8
failed
Eg8
failed
14
15
Eg16
failed
Eg16
failed
16
Eg*16
failed
# of
informants
18
18
18
18
18
18
18
18
Standard
Deviation
1.25
1.15
1.46
1.33
1.67
1.13
1.24
0.91
Average
+1
+1.11
+0.56
+0.28
+0.56
+1.06
+0.89
0.06
Example No.
(30)
(31a)
(31b)
(31c)
(32)
(33a)
(33b)
(33c)
Only +2
14
# of
informants
+2
+0.8
+2
+0.0
10
10
9
9
'Failure' on Eg*8 and Eg*16 ==> Falsification
--I think it will be effective if I address the following question.
(35)
What would be wrong if we tried to focus on the informants who gave "+2" on Eg*8 and Eg*16,
and check their scores on Eg* and Eg* in the above experiment.
My answer (at this point) is as follows.
Consider the two ways of 'adjusting the scores' given below.
(36)
Focusing on the informants who have given "+2" (or "+1") to an Egn and consider their scores on
Eg*n and Egn, to determine whether the hypothesis is falsified or corroborated, with respect to
Eg*n and Egn.
(37)
Focusing on the informants who have given "2" (or "1") to an Eg*n and consider their scores on
Eg*n and Egn, to determine whether the hypothesis is falsified or corroborated, Eg*n and Egn.
What do (36) and (37) achieve? Given that the acceptability of Egs depends upon not only grammatical
factors but non-grammatical factors (and what is being tested in regard to an Eg, in some cases, might in fact
be how good a given informant is in imagining a certain pragmatic context.15 Therefore, choosing a subset
of the entire informants as in (36) can be understood as an attempt to reduce the noise of non-grammatical
factors. By focusing on those informants' scores on Eg*n and Egn, we can hope to obtain judgments of the
informants who can 'control' the non-grammatical factors. It is therefore significant if those informants in
question still find Eg* unacceptable since they are, presumably, good at controlling the non-grammatical
factors.
Choosing a subset of the entire informants as in (37), and 'ignoring' the informants who have given "0,"
"+1" or "+2" would have a significantly different meaning.
14
The figures on this and the next rows show the scores on Eg*s when we only counted the scores by the
informants who gave (29a) and (31a) "+2."
15
I will (try to) return to other issues related here, such as the 'patience' of each informant to go through a
few or more possible arrays (and hence the structural descriptions), corresponding to the surface string being
asked.
New Book1 Outline version 2.0
Hoji 11/26
--An 'adjustment' like this would have significance only if we assumed that the speaker judgment on Eg* can
be affected by non-grammatical factors, i.e., only if we assumed that Eg* can be felt to be not clearly
unacceptable due to non-grammatical factors.
--But if we made such an assumption, it would no longer be possible to falsify a hypothesis, at least in the
way that has been suggested above, and if we are interested in how to falsify a hypothesis in generative
grammar, we will have to come up with an alternative means to falsify a hypothesis.
(To be continued.)
--If we make a (37)-type adjustment on CFJ-32, we still do not get anything that resembles corroboration for
the raising analysis. <==Should provide the data/figures.
3.3. Experiment 6: A follow-up experiment
--Make reference to Yoon's 'non-human' restriction on Mob.
3.3.1. Exp 6: Examples
(38) a.
b.
keisatu-ga hannin-ga azito-ni
mukatteiru to danteisita
police-NOM suspect-NOM hideout-to is:going
that determined
'The police has determined that the culprit is on his way to the hideout'.
azito-ni
keisatu-ga hannin-ga mukatteiru to danteisita
hideout-to police-NOM suspect-NOM is:going
that determined
c.
hannin-ga azito-ni
mukatteiru to keisatu-ga danteisita
suspect-NOM hideout-to is:going
that police-NOM determined
d.
*hannin-ga mukatteiru to
azito-ni
keisatu-ga danteisita
suspect-NOM s:going
that hideout-to ipolice-NOM determined
(39) a.
keisatu-ga
police-NOM
ko-no kutu-o
yukuefumeininatta dansei-no
this-GEN shoes-ACC is:mossing
man-GEN
mono-dearu to
belongings-is that
danteisita koto
determined (fact)
b.
'(the fact that) the police has determined that this shoe belongs to the man who is missing'.
ko-no kutu-o
yukuefumeininatta dansei-no mono-dearu to keisatu-ga
this- GEN shoes-ACC is:mossing
man-GEN belongings-is that police-NOM
danteisita koto
determined (fact)
c.
ko-no kutu-o keisatu-ga
this- GEN shoes-ACC police-NOM
yukuefumeininatta dansei-no
is:mossing
man-GEN
mono-dearu to
belongings-is that
danteisita koto
determined (fact)
d.
yukuefumeininatta
is:mossing
dansei-no
man-GEN
mono-dearu to
ko-no kutu-o
keisatu-ga
belongings-is that this- GEN shoes-ACC police-NOM
danteisita koto
determined (fact)
(40) a.
daremo-ga
manga-o
syoaku-no
kongen-dearu to
minasiteita
everyone-NOM comic:book-ACC variety:of:vice-GEN source-is
that used:to:regard
New Book1 Outline version 2.0
Hoji 12/26
b.
'Everyone used to regard comic books to be responsible for a variety of vices'.
syoaku-no
kongen-dearu to manga-o
daremo-ga
minasiteita
variety:of:vice-GEN source-is
that comic:book-ACC everyone-NOM used:to:regard
3.3.2. Exp 6: Results
(41) Results
Serial No.
1
2
3
4
PBC1
Eg4
cleared
Eg4
failed
Eg4
cleared
Eg*4
cleared
PBC2
5
6
7
8
9
10
Eg8
failed
Eg8
failed
Eg8
failed
Eg*8
failed
Eg10
failed
Eg*10
failed
The # of informants
27
27
26
27
27
27
27
27
27
27
Standard Deviation
0.19
1.39
0.00
0.88
0.00
0.19
0.61
1.12
0.00
1.42
Average
+1.96
+0.81
+2.00
1.52
+2.00
+1.96
+1.81
+0.81
+2.00
+0.44
Example No.
(38a)
(38b)
(38c)
(38d)
(39a)
(39b)
(39c)
(39d)
Serial No.
11
12
13
14
15
16
17
18
19
20
PBC2
Eg12
failed
Eg*12
failed
Eg14
failed
Eg*14
failed
Eg16
failed
Eg*16
failed
Eg18
failed
Eg*18
failed
Eg20
failed
Eg*20
failed
The # of informants
Standard Deviation
27
0.00
27
1.31
27
0.19
27
1.17
27
0.19
27
1.13
27
0.00
27
1.34
27
0.00
27
1.45
Average
+2.00
+0.67
+1.96
+0.89
+1.96
+0.52
+2.00
+0.78
+2.00
+0.22
(40a)
(40b)
PBC1
Example No.
3.4. An attempt to save the raising analysis from being falsified by the PBC-related experiments
--One may try to save the raising analysis from being falsified by maintaining that the movement involved in
RtoO is an instance of A-movement and that A-movement, as argued (or suggested?) in Lasnik xxx, does not
leave a trace, hence not being subject to the PBC. This would indeed save the raising analysis from PBCrelated falsification; but such a move also would result in eliminating the only empirical argument for the
raising analysis.
3.5. Summary
Disconfirmation and falsification:
These results thus clearly disconfirm the negative prediction made by the raising analysis, hence falsifies the
hypothesis in question, and in turn provide support for the Major Object hypothesis, which does not make
such a negative prediction. As we have seen earlier, the negative predictions made under the Major Object
analysis, by contrast, have been confirmed, and the hypothesis corroborated (in the sense introduced above)).
4. Falsification of any 'ECM' analysis
--See the LSA supplementary handout.
--The most 'promising' way to save the 'ECM' analysis is to postulate a position at the edge of the embedded
clause and assume that an empty category heads the entire embedded clause that contains the maximal
projection headed by -to. The following two aspects of such an approach seem to be to be unsatisfactory, at
least given my current understanding of things. (i) It is not clear what the nature of such embedded clause
New Book1 Outline version 2.0
Hoji 13/26
and the empty category, or how it is independently motivated. (ii) The edge position in such an approach
would share the semantic-functional properties with 'about NP' in English (as in the 'proleptic' construction);
and such a state of affairs seems to me to be more peculiar or less natural than the sharing of the relevant
properties by 'about NP' in English and the major object as postulated in the Major Object analysis.
--The movement of non-constituents (assuming that X' cannot be moved, which follows from the assumption
(which I think is adopted in the Minimalist discourse) that has the effect that a grammatical operation can
operate only on a head and a maximal projection. <==Is this supposed to follow from the Phase
Impenetrability or something, I wonder.
--Cf. The Minimalist account of the RtoO in English, as presented in Radford?
idea to include it for a PR purpose. See Yasu1's summary document.
<==It might not be a bad
5. Corroboration for the Major Object hypothesis
Goal of this section:
To show that the Major Object analysis gets corroborated by a set of experiments which have to do with
negative-sensitive elements.
--Recall that corroboration, as understood here, obtains only if a given hypothesis makes a negative
prediction that is confirmed. It is important that we distinguish clearly between corroboration in this sense
and 'support' for a given hypothesis that only 'makes a prediction' about contrasts.
(42) a.
b.
Taro-wa manga-sika yoma-nai. / *yomu.
Taro-TOP comics-all:but read-Neg / *read
'Taro does not read any kind of book but comics.'
Taro-sika manga-o yoma-nai /* yomu (koto)
Taro-all:but comics-ACC read-Neg / *read
'Nobody but Taro reads comics.'
(43) a.
b.
Saikin
recently
rokuna-sakka-ga syoo-o {tora-nai / *toru}.
good-writer-NOM
award-ACC get-Neg / *get
'Recently, no good writers have received an award.'
Taro-wa itumo rokuna-koto-o {si-nai / *suru}.
Taro-TOP always good-thing-ACC
do-Neg / *do
'Taro always does damn things.'
(44) (Kataoka to appear: (4))
Rokuna-N must be c-commanded by Neg at LF.
(45) (Kataoka to appear: (23), slightly restated) (Cf. also Aoyagi and Ishii 1994.)
XP-sika must be in a mutual c-command relation with a projection of Neg at LF.
(46) a.
b.
... NPI ... [CP ... to] ... Neg
*.. NPI ... [CP ... V-Neg-I to] ...
New Book1 Outline version 2.0
Hoji 14/26
5.1. Experiment 1: Preliminary experiment with XP-sika16
--The Kataoka proposal is based on paradigms involving xx [and has been established independently of the
considerations in the next subsection. <==Really?]
(47) a.
b.
c.
d.
e.
BVA
Q scope
'scrambling'
resumption
Binding Principle B effects
--(44) and (45) will be used as an Hsub in the crucial experiment in the next subsection.
--In this subsection, we will
5.1.1. Exp 1: Prediction
(48) The predictions under (45):
XP-sika in the matrix
Neg in the
A
matrix
Neg in the
B (-2)
embedded
XP-sikain the embedded
D (-2)
C
5.1.2. Exp 1: Examples
(49) The predictions on actual examples:
XP-sika in the matrix
Neg in the
Eg: (50a), (51a)
matrix
Neg in the
Eg*: (50b), (51b)
embedded
(50) a.
XP-sikain the embedded
Eg*: (50d), (51d)
Eg: (50c), (51c)
Mary-ga Bill-ni-sika [John-ga NY-ni itta
to] iwanakatta (koto)
Mary-NOM Bill-DAT-all:but John-NOM NY-to go:past that say:not:past (fact)
'(the fact that) Mary did not say to anyone except Bill that John had gone to NY'
b.
*Mary-ga Bill-ni-sika
Mary-NOM Bill-DAT-all:but
[John-ga NY-ni ikanakatta to] itta
(koto)
John-NOM NY-to go:not:past that say:past (fact)
(Intended reading??) '(the fact that) Mary said to anyone except Bill that John had not gone to NY'
c.
Mary-ga Bill-ni
[John-ga NY-ni-sika ikanakatta to] itta
(koto)
Mary-NOM Bill-DAT John-NOM NY-to-all:but go:not:past that say:past (fact)
'(the fact that) Mary said to Bill that John had not gone to anywhere except NY'
d.
*Mary-ga Bill-ni
Mary-NOM Bill-DAT
[John-ga NY-ni-sika
John-NOM NY-to-all:but
itta to] iwanakatta (koto)
go:past that say:not:past (fact)
16
The materials in 3.1 and 3.2 are originally part of CFJ-40 in the supplementary handout by Hoji at
http://www.hmn.bun.kyoto-u.ac.jp/langlogic/.
New Book1 Outline version 2.0
Hoji 15/26
(Intended reading??) '(the fact that) Mary did not say to Bill that John had gone to anywhere
except NY'
(51) a.
Mary-ga Bill-ni-sika
Mary-NOM Bill-DAT-all:but
[John-ga NY-ni Susan-o turete itta to] iwanakatta (koto)
John-NOM NY-to Susan-ACC take go:past that say:not:past (fact)
'(the fact that) Mary did not say to anyone except Bill that John had taken Susan to NY'
b.
*Mary-ga Bill-ni-sika [John-ga NY-ni Susan-o turete ikanakatta to] itta
(koto)
Mary-NOM Bill-DAT-all:but John-NOM NY-to Susan-ACC take go:not:past that say:past (fact)
(Intended reading??) '(the fact that) Mary said to anyone except Bill that John had not taken Susan
to NY'
c.
Mary-ga Bill-ni
[John-ga NY-ni-sika Susan-o
turete ikanakatta
Mary-NOM Bill-DAT John-NOM NY-to -all:but Susan-ACC take go:not:past
to] itta (koto)
that say:past (fact)
'(the fact that) Mary said to Bill that John had not taken Susan to anywhere but NY'
d.
*Mary-ga Bill-ni [John-ga NY-ni-sika Susan-o turete itta
to] iwanakatta (koto)
Mary-NOM Bill-DAT John-NOM NY-to-all:but Susan-ACC take go:past that say:not:past (fact)
(Intended reading??) '(the fact that) Mary did not say to Bill that John had taken Susan to
anywhere but NY'
5.1.3. Exp 1: Results
(52)
1
Eg2/4
cleared
2
Eg*2
cleared
3
Eg2/4
cleared
4
Eg*4
cleared
9
Eg10/12
cleared
10
Eg*10
cleared
11
Eg10/12
cleared
12
Eg*12
cleared
28
28
28
28
28
28
28
28
0.26
1.32
0.37
1.15
0.35
0.93
0.26
0.9
Average
+1.93
1.36
+1.93
1.21
+1.86
1.64
+1.93
1.43
Example No.
(50a)
(50b)
(50c)
(50d)
(51a)
(51b)
(51c)
(51d)
Serial No.
The # of
informants
Standard
Deviation
clear and fail
Eg2/4 clear2/4 , Eg2/4 clear2 fail4, etc.
5.2. Experiment 2: Crucial experiment with NP-o-sika
5.2.1. Exp 2: Prediction
(53) The predictions under the Major Object hypothesis, together with (45):
NP-o-sika as Mob
Neg in the matrix
E
Neg in the embedded
F (2)
5.2.2. Exp 2: Examples
(54) The predictions on actual examples:
NP-o-sika as Mob
New Book1 Outline version 2.0
Hoji 16/26
Neg in the matrix
Neg in the embedded
Eg: (55)
Eg*: (56)
(55) (Corresponding to E in (53))
John-ga Bush-no
Iraku seisaku-o-sika
John-NOM Bush-GEN Iraq policy-ACC -all:but
(sore-ga)
it-NOM
kenkoku-no
seisin-ni
national:foundation-GEN spirit-DAT
hansiteiru to iwanakatta (koto)
is:against that say:not:past (fact)
'(the fact that) John did not say about anything else but Bush's Iraq policy that it is against the
spirit of the country's founders.'
(56) (Corresponding to F in (53))
*John-ga Bush-no Iraku seisaku-o-sika
John-NOM Bush-GEN Irap policy-acc-all:but
hansiteinai
is:against:not
(sore-ga)
it-NOM
kenkoku-no
seisin-ni
national:foundation-GEN spirit-DAT
to itta (koto)
that say:past that
(Intended reading) '(the fact that) John said about anything else but Bush's Iraq policy that it is not
against the spirit of the country's founders.'
5.2.3. Exp 2: Results
(57)
5
Eg7
failed
6
Eg8
failed
7
8
Eg*7
cleared
Eg*8
cleared
The # of
informants
28
28
28
28
Standard
Deviation
1.32
1.31
0.81
0.94
Average
+0.5
+0.29
1.64
1.43
Example No.
(55) with sore-ga
(55) w/o sore-ga
(56) with sore-ga
(56) w/o sore-ga
Serial No.



What makes a sentence unacceptable?
What makes a sentence acceptable (to a given speaker)?
Negative predictions and Eg* vs. Eg
(58) Results focusing on the informants who have given Egs +2 or +1:
Serial No.
The # of
informants
Standard
Deviation
Average
Example No.
5
Eg7
cleared
6
Eg8
cleared
7
8
Eg*7
cleared
Eg*8
cleared
18
14
18
14
0.49
0.49
0.94
0.59
+1.39
+1.43
1.67
1.71
(55) with sore-ga
(55) w/o sore-ga
(56) with sore-ga
(56) w/o sore-ga
New Book1 Outline version 2.0
Hoji 17/26
w/o and w/ sore-ga --- the actual examples presented to the informants
implications of examples with 'resumption'
5.3. Experiment 3: Crucial experiment with Rokuna NP17
5.3.1. Exp 3: Prediction
(59) The predictions under the Major Object hypothesis, together with (44):
Rokuna NP as embedded subject
E
Neg in the embedded
Rokuna NP as Mob
F (2)
5.3.2. Exp 3: Examples
(60) The predictions on actual examples:
Rokuna NP as embedded subject
Eg: (61a)
Neg in the embedded
(61) a.
b.
Rokuna NP as Mob
Eg*: (61b)
(Corresponding to E in (59))
Hanako-wa [rokuna otoko-ga kanozyo-no kurasu-ni inai to]
omotte ita.
Hanako-TOP good
man-NOM her-GEN class-in
be:not that thinking was
'(Roughly) Hanako had thought that no good man was in her class.'
(Corresponding to F in (59))
*Hanako-wa
rokuna otoko-o kanozyo-no kurasu-ni inai
to
Hanako-TOP good man-ACC her-GEN
class-in be:not that
'(Roughly)
omotte ita.
thinking was
Hanako had thought about no good man that he was in her class.'
5.3.3. Exp 3: Results
(62) Results
1
Eg2
cleared
2
Eg*2
cleared
The # of
informants
33
33
Standard
Deviation
0.54
0.48
+1.79
(61a)
1.79
Serial No.
Average
Example No.
(61b)
The results of Experiments 1, 2, and 3 all corroborate the Major Object hypothesis.
5.4. A theoretical characterization of Mob and the Major Object construction.
5.5. Summary
6. An attempt to falsify the Major Object hypothesis
--The indeterminate-agreement-based argument
17
The materials in this subsection are originally part of CFJ-32 (see footnote 16).
New Book1 Outline version 2.0
Hoji 18/26
--Hmain and Hsub
Hmain: the MOB hypothesis
Hsub: the c-command requirement on the indeterminate 'pronoun'.
--The falsification of the Hsub.
--An alternative account of the 'indeterminate agreement' phenomenon?
--The Kawazoe proposal/suggestion and its consequences
7. Some Details: the major subject and the major object
I would not be surprised if the content of this chapter will be incorporated in earlier chapters.
7.1. Yoon 2004 and 2005
--James Yoon's (2004 and 200) major-subject/raising account
7.1.1. Yoon's critique of the Major Object hypothesis
(63) Yoon's critique of the Major Object hypothesis18
a. The theta role and the case/Case on Mob
b. Mob can be made into the subject of a passive sentence without necessarily involving the
'affected' interpretation.
c. Case/case stacking
(64) HH's informal responses
a. (Re. (63a))
The same as "about NP" in (13a) in English, re. the theta role, I suspect. As to case/Case, I need
to make some general remarks on 'case marking' in JP, without much direct substantiation of the
view that I assume/adopt…
b. (Re. (63b))
Some remarks on passives in JP: No A-movement in Japanese.
*John thought Mary to be a genius (* for many speakers, I guess)
Mary was thought to be a genius
?????John thought of Mary to be a genius. (Any example like this that is acceptable?)
c. (Re. (63c))
Some 'cases' need not be related to a theta position for 'its licensing', to begin with, and case
stacking seems acceptable only with those 'cases'. <==I will perhaps have to include brief
discussion on the Subjacency effects in relation to this.
As to (64b) and (64c), we should consider whether the MOB hypothesis, as Hmain, would give rise to a
negative prediction and with what Hsub.
(65) a.
b.
c.
What would be Hsub, if the observations in (63b) and (63c) were to falsify the MOB hypothesis?
(For (63b))
The Hsub must have the effect that the Mob cannot be the subject of a passive sentence without
invoking the affected interpretation on it. How should it be formulated, and what could be Hsub
in an experiment to test it as Hmain? Unless such an Hsub, serving as Hmain, gets corroborated, or if
such an Hsub gets falsified, then the observation in (63b) does not falsify the MOB hypothesis.
(For (63c))
The Hsub must have the effect that the Mob cannot be the subject of a passive sentence without
18
Maybe, the "Major Object" is not a good term and Takano's (2003) "prolepsis" is a better term to use, it
seems.
New Book1 Outline version 2.0
Hoji 19/26
invoking the affected interpretation on it. How should it be formulated, and what could be Hsub
in an experiment to test it as Hmain? Unless such an Hsub, serving as Hmain, gets corroborated, or if
such an Hsub gets falsified, then the observation in (63c) does not falsify the MOB hypothesis.
7.1.2. Problems for Yoon's (2005) proposal
(66) a.
b.
c.
d.
What derives the movement?
The Mob option is also allowed under certain conditions ((i) when there is a raised NP, and (ii)
when the NP (as the Mob) denotes a human)
What would be the account of case stacking?
The absence of PBC effects
--The answer to (66a) is not clear. (Yoon stated at the Kyoto workshop that the movement is like
scrambling. At the LSA workshop, he stated that he did not have the answer to that.)
--How could (66b) be stated in theoretical terms, i.e., how could it be derived in the theory?
--Are the generalizations noted in (66b) valid, to begin with?
-7.2. Yoon 1989
--Yoon's (1989) major-subject/ECM account
8. The significance of preliminary experiments
The point here should be about 'repeatability' and what significance we should assign to the judgments of a
given informant or a given set of informants. The point is likely to be made in an earlier chapter, perhaps
before the chapters on the RTO-related specific discussion. But, for now, I am thinking that it would be
useful and effective to add the relevant discussion after some concrete empirical materials have been
presented.
It now seems, at least from my own perspective, that the main contribution of this book has to do with what I
can articulate in regard to the point that will be made in this chapter (as well as in an earlier part of the book
where the relevant discussion is provided.)
8.1. The structure of an experiment (this has already been given above once)
(67) a.
b.
(68)
Hmain: the hypothesis being tested
Hsub: the hypothesis that has been independently corroborated
an experimental specification in accordance with (67) (and other auxiliary assumptions), e.g.:
(i) Sentences of such and such form are unacceptable (under a specified interpretation).
(ii) Sentences of such and such form that are minimally different from those noted in (i) are not
necessarily unacceptable (under a specified interpretation).
The importance of conducting necessary preliminary experiments before conducting a crucial experiment.

If one uses as Hsub a hypothesis that is shaky at best or is easily falsified (if not having already been
falsified)—which can easily result in the disconfirmation of a prediction—it is not even clear what
significance the result of such an experiment might have.
8.2. Local disjointness effects
(69) a.
John believes that he is popular.
New Book1 Outline version 2.0
Hoji 20/26
b.
*John believes him to be popular.
8.3. Inverse scope
(70) a.
b.
c.
someone loves everyone
someone believes everyone to be popular
someone believes that everyone is popular
9. Concluding remarks
In addition to providing support for the Major Object analysis of the so-called RtoO in Japanese (and
arguably in Korean),19 I have suggested in this talk that it is necessary for us to bind ourselves by the criteria
of the sort alluded to above in regard to falsification and corroboration (if we want to be taken seriously by
linguists outside generative grammar, and perhaps more importantly by researchers in the neighboring
disciplines and beyond, in regard to the claim that we are engaged in an empirical science with progress in
mind).
The emphasis placed on negative predictions makes it possible to obtain a clearer understanding of the
nature of (at least some aspects of) judgmental fluctuation and disagreement. Judgmental fluctuation on
Eg* is significant and serious; it could directly lead to the falsification of a hypothesis. Judgmental
fluctuation on Eg, on the other hand, is much less significant and serious. Although it could affect whether
our hypothesis gets corroborated, it would not bear on whether our hypothesis is falsified. Failure to
recognize this difference seems to me to have resulted in a (not uncommon, if not prevailing) attitude of not
being compelled to make one's hypothesis falsifiable, which is generally accompanied by the lack of
concerns for articulating what should count as a falsification of a hypothesis.
I also wish to suggest that recognizing this point helps us appreciate the real significance of a minimal
pair. The preceding discussion suggests that obtaining a contrast is not sufficient for a given hypothesis to
be considered plausible (let alone compelling). A contrast may obtain even when a hypothesis is falsified,
in the sense defined above, e.g., the score on Eg* can be "0" and that on its corresponding Eg can be "2." 20
If a specific example is an instance of Eg* in accordance with a hypothesis (Hmain), combined with an
independent hypothesis (Hsub) (along with some auxiliary assumptions), such a result must have been
deduced from the hypotheses in question, and there is no excuse for one not to seriously doubt H main if many
speakers accept such an example, even to varying degrees. It is hoped that proceeding in the manner
advocated above would make it possible to address the issues about repeatability in a much more concrete
and realistic manner than has been possible in the past, so far as I can tell.21
19
An outstanding issue that has yet to be addressed is how the so-called accusative marker -o is licensed,
assigned, etc. Under the Major Object analysis, it must be possible for -o to 'function' like a postposition,
so to speak, reminding us of the 'adverbial use' of -o discussed in Kuroda 1978=1992: chap. 6, 228. Also
relevant here is Y. Takubo's recent discussion of no koto (the handout downloadable at
http://www.hmn.bun.kyoto-u.ac.jp/langlogic/).
20
Maybe, I should make reference to the "PBC" effects in RtoO (for some speakers, to varying degrees) and
their absence in the "Object Control" cases.
21
Among the issues that need to be addressed is that of repeatability. Although repeatability across
informants appears to be emphasized in the preceding discussion, it is also (and perhaps more) important to
evaluate repeatability within a single informant in terms of (i) different example sentences of the same
formal properties in question and (ii) different occasions of testing the same examples. An articulation of
the relevant notions here would become crucial when we were faced with a situation where the researcher's
own judgments and the informants' judgments differ from each other substantially. I am assuming here that
the researcher must have conducted an experiment on him/herself and must have obtained corroboration for
New Book1 Outline version 2.0
Hoji 21/26
References
Aoyagi Hiroshi & Toru Ishii. 1994. On NPI licensing in Japanese. In Japanese Korean Linguistics 4, 295311.
Davies, W. D. (2001) Against raising in Madurese (and other Javanic languages)," Proceedings of the
Chicago Linguistic Society (CLS) 36, 57-69.
Davies, W. D. and Dubinsky, S. (2004) The Grammar of Raising and Control: A Course in Syntactic
Argumentation, Malden, MA: Blackwell Publishers.
Harada, Naomi. (2003) "Raising to Object is NOT an edge phenomenon," Paper presented at the January
2003 Annual Meeting of the Linguistics Society of America, unpublished manuscript, ATR International.
Hiraiwa, Ken. (2002) "Raising and indeterminate agreement," Second Draft (June 2002) (The document,
states "A revised version to appear in the Proceedings of WCCFL 21," but the paper is not included in the
proceedings.).
Hoji, Hajime. (1991). "Raising-to-object, ECM and the major object in Japanese," A talk given at Japanese
Syntax Workshop at University of Rochester.
Hoji, Hajime. (1995) "Demonstrative binding and Principle B," in NELS 25, 255-271. GLSA, University of
Massachusetts, Amherst.
Hoji, Hajime. (2005) "A Major Object Analysis of the so-called Raising-to-Object Construction in Japanese
(and Korean)," a talk presented at the Workshop on Japanese and Korean Linguistics, Kyoto Univeristy,
February 21-22, 2005. http://www.hmn.bun.kyoto-u.ac.jp/langlogic/
Hong, Kisun. (1990) "Subject-to-object raising in Korean," in Katarzina Dziwirek, Patrick Farrell, and Elias
Mejias-Bikandi (eds.), Grammatical relations: A cross-theoretical perspective, 215-225, Stanford: CSLI.
Kataoka, Kiyoko. (2004) Hiteibun-no Koozoo: Kakimaze-bun to Hiteekoohyoogen (Syntactic Structure of
Japanese Negative Sentences: Scrambling Construction and Negation-sensitive Elements). Doctoral
dissertation. Kyushu University, Fukuoka, Japan.
Kataoka, Kiyoko. (to appear) "'Neg-sensitive' Elements, Neg-c-command and Scrambling in Japanese," in
JK14.
Kitagawa, Yoshihisa. (1985) "Barriers to Government," NELS 16, 249-273.
Kuno, Susumu. (1976) "Subject Raising," in Shibatani, ed., Syntax and Semantics: Japanese Generative
Grammar, Academic Press.
Kuroda, S. -Y. (1978) "Case Marking, Canonical Sentence Patterns, and Counter Equi in Japanese," in J.
Hinds & I. Howards eds., Problems in Japanese Syntax and Semantics, Garland Press, New York, pp. 3051. Reprinted in Kuroda (1992) chapter 6, pp. 222-239.
Kuroda, S. -Y. (1992) Japanese Syntax and Semantics, Kluwer, Dordrecht.
Marantz, Alec. (1983) "Raising and Category Types in Japanese," in Y. Otsu, et al. eds., Studies in
Generative Grammar and Language Acquisition, ICU, Tokyo.
Saito, Mamoru. (1983) "Comments on the Papers on Generative Syntax," in Y. Otsu, et al. eds., Studies in
Generative Grammar and Language Acquisition, ICU, Tokyo.
Takano, Yuji. (2003) "Nominative objects in Japanese complex predicate constructions: a prolepsis
analysis," Natural Language and Linguistic Theory 21, 779-834.
Takubo, Yukinori (2005) "An Overt Marker for Individual Sublimation in Japanese," a talk presented at
the Workshop on Japanese and Korean Linguistics, Kyoto University, February 21-22, 2005.
Tomoda, E. K. (1976-77) "Raising and point of view: Evidence from Japanese," Papers in Japanese
Linguistics 5, 361-376
Yoon, James. (to appear) "Raising Specifiers. A Macroparametric Account of Subject-to-Object Raising in
Some Altaic Languages," The proceedings of the Formal Altaic Linguistics.
Yoon, James. (2004) "Raising and Prominence," a handout for the talk given at Language Education Institute,
Seoul National University, 7/26/04.
Yoon, J. E. (1989) "ECM and Multiple Subject Constructions in Korean," in S. Kuno, et. al. eds., Harvard
the hypothesis in question, before proceeding to conduct an experiment on a group of informants to obtain
(further) confirmation for his/her hypothesis.
New Book1 Outline version 2.0
Hoji 22/26
Studies in Korean Linguistics 3.
10. Appendix: On the 'semantic' restriction on Mob
Mob is claimed here to be on a par with about Mary in (71) in English, at least at an observational level,
along the lines of Hoji 1991, and Takano 2003.
(71)
John firmly believed about Mary [CP that she was an Italian].
If we want to give a formal content to this observational point, we might be able to make some additional
predictions. We might, for example, assume that due to some formal property of NP-o in the 'construction'
in question (which, by hypothesis, is the same as that of about NP in (71)) the grammar gives the following
instructions, so to speak, to the language user.22
(72) a.
b.
NP-o in the 'construction' in question, i.e., Mob, denotes some entity about which one can hold
some belief/assumption/judgment/etc. can be held (depending upon the predicate used).
(What appears to be) the CP complement of the verb in the 'construction' in question denotes a
property that can be attributed to some entity, reasonably and meaningfully.
Consider again the English examples in (13), repeated here, and their schematic representations in (73).
(13) a.
b.
(73) a.
b.
John believed about Mary that she was Italian.
John believed Mary to be Italian.
A believes about B that IP
A believes B to be VP
As noted, one of the hallmark properties of the construction in (73b) in English (and other languages that
have a structure formally corresponding to (73b)) is that B in (73b) can be a pleonastic element and/or an
idiom chunk, which clearly would not satisfy the condition imposed on Mob in (72a). A prediction is
therefore that if we place as Mob something that is like a pleonastic element, an idiom chunk, or any other
element that cannot satisfy the condition in (72a), the resulting sentence is unacceptable. Japanese does not
seem to have an (overt) pleonastic element, and it is not entirely clear how to identify idioms. But, to the
extent that we can identify such elements, the prediction does seem to be borne out.23
22
Tomoda (1976-77: 372) represents the meaning of NP-o in question as NP-ni tuite (regarding NP). The
intuitions recorded in (72) have been expressed in various ways in the literature.
(i) a.
(Hong 1990: 223)
"The object of the special class of (RTO) predicate should 'denote a specific thing, in the sense
that its reference is registered in the speaker's mind so that it can be identified by the speaker (Lee
1989:12)."
b.
The semantic relation between the ACC-marked NP and the following IP in RtoO is the same as that
between a major subject and an IP that follows it. (J.-E. Yoon 1989, James Yoon to appear,
2004)
23
Hoji 1991 makes an attempt to provide an argument of this sort on the basis of observations such as the
following, presumably taking the ga-marked NP in (v) below as 'part of a predicate'.
(i) (Hoji 1991: (50))
John believes it to have rained (while he was asleep).
(ii) (Hoji 1991: (51))
a.
Watasi-wa [CP ame-ga (neteiru aida-ni) hutta to] omotta.
I-TOP
rain-NOM while sleeping fell that thought
'I thought that it [had] rained (while I was asleep).'
New Book1 Outline version 2.0
Hoji 23/26
11. Taken out from the LSA handout
[The materials below are just for myself; I am keeping them just in case I may be able to use part of it later.]
11.1. Goal
Goal: To defend the Major Object analysis, by showing (74).
(74) a. Specific negative predictions made by this analysis are confirmed
b. A negative prediction made by a raising analysis re. Proper Binding Condition (PBC) effects gets
disconfirmed.
(74b) would falsify any analysis of the so-called raising-to-object (henceforth RtoO) in Japanese (and in
Korean as well, to the extent that the relevant considerations are applicable) in which the raising of Mob
from 'its theta position' to some higher position is assumed to take place obligatorily, and its trace is claimed
to subject to PBC, as is indeed the case of the majority of the proposals in the literature.
Question: Which of the analyses noted above does this apply?
11.2. A few residual (or remaining) issues
11.2.1. On the parallelism between a major subject and Mob
J.-E. Yoon (1989): A 'major-subject' analysis, combined with the ECM approach, without the raising part of
b.
(iii)
(iv)
(v)
(vi)
*Watasi-wa ame-o
(neteiru aida-ni) hutta to omotta.
I-TOP
rain-ACC while sleeping fell that thought
(Hoji 1991: (52), slightly adapted)
a.
*John believes of/about it that it (has) rained.
b.
John believes of/about Mary that she VP.
(Hoji 1991: (53))
a.
John-wa [CP henna nioi-ga
(itumo) suru to] omotteita.
John-TOP strange smell-NOM always do that thought
'John thought that it (always) smelled strange.'
b.
*John-wa
henna nioi-o
(itumo) suru to omotteita
John-TOP strange smell-ACC always do that thought
(Hoji 1991: (54))
a.
Ame-ga hutta.
'It rained.'
b.
Henna nioi-ga suru.
'It smells funny.'
c.
Oto-ga suru.
'There is a sound.'
Takano 2003: 822 provides the examples in (vi) in support of the same point.
(Takano 2003: (79))
a.
John-wa soko-made te-ga
mawar-anai
to itta.
John-TOP there-to hand-NOM get:around-not that said
'John said that he couldn't take good care of it.'
b.
John-wa te-ga
soko-made mawar-anai
to itta.
John-TOP hand-NOM there-to
get:around-not that said
c.
*John-wa te-o
soko-made mawar-anai to itta.
John-TOP hand-ACC there-to
get:around-not that said
New Book1 Outline version 2.0
Hoji 24/26
J. Yoon 2004.
But:
==> It would not be falsified by (74b).
Parallelism between Mob and a major subject is not complete, at least in Japanese.
(75) (=Hoji 1991: (43), (43b) slightly adapted)
a. IBM-wa [soko-no atarasii konpyuutaa-no himitu]i-o asita-no
kisyakaiken-de
IBM-TOP it-GEN new
computer-GEN secret-ACC tomorrow-GEN press:interview-at
[CP Hitati-ga (supai-o tukatte) {proi/sorei-o} nusunda to] happyoosuru tumorida
Hitachi-NOM spy-ACC by:using
it-ACC stole
that announce intend
b.
'IBM intends to announce about [the secret of their new computer]i at tomorrow's press interview
that Hitachi stole iti (by using spies).'
*?[{IBM/soko}-no atarasii konpyuutaa-no himitu]i-ga [CP Hitati-ga
(supai-o tukatte)
IBM / it-GEN
new computer-GEN secret-NOM
Hitachi-NOM spy-ACC by:using
{proi/sorei-o} nusunda/nusumidasita
it-ACC stole
'It is [the secret of their new computer]i that Hitachi stole (by using spies).'
Mob is closer to the 'aboutness' topic than to the major subject.
(76) (=Hoji 1991: (44), slightly adapted)
[{IBM/soko}-no atarasii konpyuutaa-no himitu]i-wa [CP Hitati-ga
IBM / it-GEN
new computer-GEN secret-TOP
Hitachi-NOM
(supai-o tukatte)
spy-ACC by:using
{proi/sorei-o} nusunda/nusumidasita
it-ACC stole
'As for [the secret of their new computer]i, Hitachi stole iti (by using spies).'
The major subject hypothesis perhaps cannot be maintained even without involving the raising of Mob,
without (an) additional stipulation(s).
11.2.2. On the (16) + [(14) or (17)] analysis
If the raising analysis ((14) or (17)) is combined with the Major Object analysis, i.e., if RtoO need not always
involve the raising of Mob, (74b) would not falsify such an analysis. Unless a restriction on either of the
two options is provided, however, such an analysis is not falsifiable. James Yoon (at the 2/2005 Kyoto
Workshop) suggests the Major Object option is available only if Mob denotes a human. At least in
Japanese, the negative prediction made by such a suggestion gets disconfirmed, as we will see in 4.3.
11.2.3. On the (neo)ECM analysis
The (neo)ECM analysis would maintain that the embedded position occupied by Mob behaves as if it were a
matrix element, due to the special property of such an edge position (and the general workings of the
grammar involving it). What would be peculiar is, as will be noted in the Appendix below, that one would
have to assume that such an embedded edge position has the property in (77).
(77)
It denotes some entity about which one can hold some belief/assumption/judgment/etc.
(depending upon the choice of the predicate).
(77) is shared by about NP in (13b), but not shared by the NP that is analyzed as undergoing raising in the
'raising constructions' in English (hence the possibility of an pleonastic element as such an NP). The
(neo)ECM analysis, without involving the raising of Mob, therefore could be maintained only at the expense
of a stipulating like:
New Book1 Outline version 2.0
Hoji 25/26
Stipulation needed for a (neo)ECM analysis:
The embedded edge position occupied by Mob has precisely the properties of Mob under the Major Object
analysis and about NP in (13b), not shared by the "ECM'ed" NP in English.
11.3. Outline
Section 2 suggests a concrete means of determining falsification and corroboration in generative grammar,
so as to be able to evaluate a given analysis/hypothesis.
Section 3 illustrates how the Major Object hypothesis is corroborated.
Section 4 illustrates how the raising analysis is falsified.
Section 5 contains brief remarks on the general structure of an experiment, and the significance of
preliminary experiments.
Section 6 is the concluding section.
New Book1 Outline version 2.0
Hoji 26/26