• Abstract and Figures
• Public Full-text

March, 2010 Draft The Hypothetico-Deductive Method, the Language Faculty and the Non-Existence of Local Anaphors in Japanese Hajime Hoji University of Southern California

This paper explores how the hypothetico-deductive method can be applied to research concerned with the properties of the language faculty. The paper adopts Chomsky's (1993) conception of the Computational System (hypothesized to be at the center of the language faculty) and considers informant judgments to be a major source of evidence for or against hypotheses about the Computational System. Combined with two research heuristics (i) Maximize testability, and (ii) Maximize our chances of learning from errors, that leads us to a particular methodology. The paper examines, in accordance with the proposed method, the predictions made under the lexical hypotheses that otagai, zibun-zisin and kare-zisin in Japanese are a local anaphor. The experimental results show that the predictions are not borne out. If what underlies a local anaphor is closely related to "active functional categories" in the sense of Fukui 1986 and if, as suggested in Fukui 1986, the mental lexicon of speakers of Japanese lacks them altogether, this result is as expected. The failure to identify what qualifies as a local anaphor in Japanese, despite the concerted efforts by a substantial number of practitioners for nearly three decades, is therefore not puzzling but just as expected, given the hypothesis put forth in Fukui 1986. The paper also provides results of other experiments that indicate that the rigorous empirical standard advocated here is attainable, thereby providing us with hope that it is possible to rigorously apply the hypothetico-deductive method to research concerned with the properties of the language faculty.

Key Words hypothetico-deductive method, language faculty, Computational System, model of judgment making, confirmed schematic asymmetries, local anaphors, Japanese

1. Introduction: the general scientific method In the seventh lecture of his 1964 Messenger Lectures at Cornell University "Seeking New Laws," Richard Feynman states:1

In general, we look for a new law by the following process. First we guess it. Then we compute the consequences of the guess to see what would be implied if this law that we guessed is right. Then we compare the result of the computation to nature, with experiment or experience, compare it directly with observation, to see if it works. If it disagrees with experiment, it is wrong. In that simple statement is the key to science. It does not make any difference how beautiful your guess is. It does not make any difference how smart you are, who made the guess, or what his name is—if it disagrees with the experiment, it is wrong. That's all there is to it." (Feynman 1965/94: 150)

1 The passages below are taken from Feynman 1965/1994, which is a reproduction of his 1964 Messenger Lectures at Cornell University. The book was originally published in hardcover by BBC in 1965 and in paperback in 1967 by MIT Press. The page references are to the 1994 edition. The Feynman lectures can be viewed at http://research.microsoft.com/apps/tools/tuva/#.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 1/49 Feynman continues the above passage by adding the following "obvious remarks":2

It is true that one has to check a little to make sure that it is wrong, because whoever did the experiment may have reported incorrectly, or there may have been some feature in the experiment that was not noticed, some dirt or something; or the man who computed the consequences, even though it may have been the one who made the guesses, could have made some mistake in the analysis. These are obvious remarks, so when I say if it disagrees with experiment it is wrong, I mean after the experiment has been checked, the calculations have been checked, and the thing has been rubbed back and forth a few times to make sure that the consequences are logical consequences from the guess, and that in fact it disagrees with a very carefully checked experiment. (Feynman 1965/94: 150-1)

In this paper, I would like to explore how the above-mentioned general scientific method, which we can schematize as in (0), can be applied to research concerned with the properties of the language faculty.

(0) The general scientific method (i.e., the hypothetico-deductive method): Guess — Computing Consequences — Compare with Experiment

When we turn to the specific hypotheses below, we will be assuming, without discussion, that they are part of research concerned with the properties of the language faculty, and more in particular with those of the Computational System as it is hypothesized to be at the center of the language faculty. 3

2. Methodological preliminaries 2.1. The goal of generative grammar and the model of Computational System In what follows, I use generative grammar to refer to research concerned with the properties of the language faculty, and more in particular with those of the Computational System as it is hypothesized to be at the center of the language faculty and use the adjective generative accordingly.4 I also assume,

2 The "obvious remarks" should not be taken as reducing the significance of "the key to science" in the first quote. The point intended in the "obvious remarks" is not that we should not concern ourselves with empirical details and the testability of our hypotheses—that is given—; on the contrary, the point of the "obvious remarks" must be about the importance of empirical (as well as theoretical) rigor. The point seems to be either missed or misrepresented in Boeckx 2006, judging from the way Feynman's remarks are cited there. Similar remarks apply to the way Lakatos' work is cited in Boeckx 2006, as pointed out in Kuroda 2008: footnote 3. See also Newmeyer's 2008 review of Boeckx 2006 for much relevant discussion.

3 It goes without saying that a different research goal might be better served by a different methodology; and it is up to the researchers with a different goal in mind what criteria for testability they find the most effective in achieving their goal. One might raise issues as to whether such research would deserve to be called a science or should be regarded as so-called pseudo-science; but this paper is not concerned with such questions; cf. Popper 1959, 1963, Lakatos 1970/1978, 1973, and Feynman 1974 among many other works on the relevant topics. Despite sharing the goal as such, one might still wish to adopt a method different from (0). Those who wish to do so should specify how hypotheses pertaining to the Computational System can be put to rigorous test and how we can measure progress under such an alternative.

4 This is perhaps too narrow a characterization of generative grammar; see Culicover and Jackendoff 2005: chapter 1, for example, for remarks on a wide spectrum of research orientations and practices under the name of generative grammar. Generative grammar has been called "bio-linguistics" in recent years by some of the practitioners in the field. The inauguration in 2007 of the on-line journal BIOLINGUISTICS is perhaps a good example of the trend. One cannot, however, simply understand bio-linguistics as a "new name" for generative grammar because, after all, non-generative researchers had been calling their works as bio-linguistics, as can be seen by the title Bio-Linguistics: The Santa Barbara Lectures (2002) (John Benjamins) by T. Givon, for example, as pointed out in Newmeyer 2008. In this paper, we shall not be concerned with the terminological issues, and in what follows, I use generative

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 2/49 without discussion here, that a major source of evidence for or against our hypotheses concerning the Computational System is informant judgments, as explicitly stated by N. Chomsky in Third Texas Conference on Problems of Linguistic Analysis in English.5 Minimally, the language faculty must relate 'sounds' (and signs in a sign language) and 'meanings'. A fundamental hypothesis in generative grammar is the existence of the Computational System at the center of the language faculty. Since Chomsky 1993, it is understood in generative research that the Computational System is to be an algorithm whose input is a set of items taken from the mental Lexicon of speakers of a language and whose output is a pair of mental representations—one underlying 'sounds/signs' and the other 'meaning'. While what is in the mental Lexicon differs from language to language (and, to some extent, among speakers of a given language), a fundamental generative hypothesis is that the properties of the Computational System are universal. Following the common practice in the generative tradition since the mid 1970s, let us call the former a PF (representation) and the latter an LF (representation). The model of the Computational System (CS) can be schematized as in (0).

(0) The Model of the Computational System: Numeration  => CS => LF()  PF() Numeration : a set of items taken from the mental Lexicon LF(): an LF representation based on  PF(): a PF representation based on 

The PF and the LF representations in (0) are thus meant to be abstract representations that underlie a sequence of sounds/signs and its 'interpretation', respectively. Our hypotheses about the Computational System are meant to be about what underlies the language users' intuitions about the relation between "sounds/signs" and "meanings." The main goal of generative grammar can therefore be understood as demonstrating the existence of such an algorithm by discovering its properties. Construed in this way, it is not language as an external 'object' but the language faculty that constitutes the object of inquiry in generative grammar, as stated explicitly in Chomsky 1965: chapter 1. Given the reasonable assumption that informants' acceptability judgments can be affected by various non-grammatical factors, it is imperative, for the purpose of putting our hypotheses to rigorous test, that we have a reasonably reliable means to identify informant judgments as a likely reflection of properties of the Computational System. In what follows, we shall adopt a particular means to do so, a version of which we are led to, I maintain, once we adopt the basic assumptions noted above, along with the two research heuristics (i) Maximize grammar in the sense just noted.

5 Chomsky's remarks in Third Texas Conference on Problems of Linguistic Analysis in English May 9-12, 1958, published in 1962 by the University of Texas seem to point directly to what he had in mind at least around 1958, in my view more directly than what we find in his writings in the 1950s and 1960s and the subsequent years. One of many such remarks by Chomsky is reproduced here (p. 168).

Hill: Linguistic intuition is itself a system, almost a complete grammar. If it is good enough, why bother with any other grammar? Chomsky: Because I am interested in explaining intuition. If you cannot accept this as the purpose of linguistic study, I am lost. I would like to get a theory which will predict intuitions.

Obviously, informant judgments are not the only source of evidence for or against hypotheses about the Computational System. If one seeks evidence elsewhere, one must articulate how such 'evidence' is related to the hypothesized properties of the language faculty so as to ensure, and hopefully maximize, testability of the hypotheses. I take that to be a minimal methodological requirement for using evidence other than informant intuitions in empirical research concerned with the properties of the Computational System.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 3/49 testability and (ii) Maximize our chances of learning from errors.6

2.2. The model of judgment making As noted, the language faculty must relate 'sounds' (and signs in a sign language) and 'meanings'. By adopting the thesis that informant judgments are a primary source of evidence for or against hypotheses concerning the Computational System, we are committing ourselves to the view that informant judgments are, or at least can be, revealing about properties of the Computational System. Although it is not obvious how they may be so, it seems reasonable to assume that the Computational System is "made use of" when the informant judges a sentence presented to them. For, otherwise, it would not be clear how informant judgments could be revealing about properties of the Computational System. We can schematically express this as in (0).

(0) Embedding the Computational System in the model of judgment making:

(a, b)   => CS => LF()   PF()

a. : presented sentence b. (a, b): an intuition that two linguistic expressions a and b are related in a particular manner c. the informant judgment on the acceptability of  under (a, b)

The boxed part in (0) is the Computational System; see (0). The informant is presented sentence  and asked whether it is acceptable, or how acceptable it is, under a particular interpretation involving two linguistic expressions a and b, (a, b). As noted above, insofar as informant judgments are assumed to be revealing about properties of the Computational System, the Computational System must be part of the act of judgment making by the informant. It thus seems reasonable to hypothesize as follows: the informant must be coming up with a numeration when judging a sentence, and the numeration results in two output representations, and the informant must be making her or his judgment by making reference to the output representations. As a consequence of this reasoning, the following model of judgment making by informants presents itself.7, 8

6 Both heuristics are explicitly advocated by K. Popper; see Popper 1959 and especially Popper 1963 for a succinct summary of his methodological proposal. See section 5 for further discussion.

7 The model in (0), which is adapted from the proposal in a series of works by Ayumi Ueyama, including Ueyama 2010, can be understood as characterizing a specialized instance (or aspect) of the model of comprehension. It may be well to emphasize, as Ayumi Ueyama points out, that the act of judgment making, more often than not, requires that informants do something that is not involved in ordinary language use. As I hope will be made clear in the ensuing discussion, such idealization is necessary in extracting 'information' pertaining to the properties of the Computational System from informant judgments. It may be an interesting exercise to compare (0) with the model of comprehension discussed in Townsend and Bever 2001.

8 Hoji 2009: Appendix compares (0) with the model of judgment making suggested in Schütze 1996: 175.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 4/49 (0) The Model of Judgment Making by the Informant on the acceptability of sentence  with interpretation (a, b)9 (based on A. Ueyama's proposal):

Lexicon (a, b)   ≈≈>  Numeration ≈≈ ≈≈ => CS => =>  > Extractor10 >  LF() SR()    PF()        pf()

a. : presented sentence b. : numeration c. (a, b): the interpretation intended to be included in the 'meaning' of  involving expressions a and b d. LF(): the LF representation that obtains on the basis of  e. SR(): the information that obtains on the basis of LF() f. PF(): the PF representation that obtains on the basis of  g. pf(): the surface phonetic string that obtains on the basis of PF() h. the informant judgment on the acceptability of  under (a, b)

The "==>" in (0) indicates that a numeration is input to the Computational System (CS) and its output representations are LF and PF, and that SR and pf obtain based on LF and pf, respectively. What is intended by " ≈≈>," on the other hand, is not an input/output relation, as indicated in (0).

(0) a. Presented Sentence  ≈≈> Numeration Extractor: ... is part of the input to ... b. Numeration Extractor ≈≈> numeration : ... forms ... c. SR() ≈≈> Judgment : ... serves as a basis for ...

As discussed in some depth in Hoji 2009, the model of judgment making in (0) is a consequence of adopting the theses, shared by most practitioners in generative grammar, that the Computational System in (0) is at the center of the language faculty and that informant judgments are a primary source of evidence for or against our hypotheses pertaining to properties of the Computational System.

2.3. Informant judgments and fundamental asymmetry It seems reasonable to assume that the informant judgment  can be affected by difficulty in parsing and the unnaturalness of the interpretation of the entire sentence in question.11 That is to say, even if the informant (eventually) finds numeration  corresponding to the presented sentence  such that  results in pf() non-distinct from  and SR() compatible with the interpretation (a, b), that may not necessarily result in the informant reporting that  is (fully) acceptable under (a, b). On the other hand, if the

9 There is reason to believe that informant judgments would be qualitatively more difficult to handle if we dealt with simple (un)acceptability not involving (a, b) for the purpose of attributing the unacceptability in question to a property of the Computational System. This stems from the difficulty in distinguishing "acceptability" from "intelligibility." But I cannot elaborate on the issue here due to space considerations.

10 Numeration Extractor (in place of Parser in the original Ueyama model) is due to Yuki Takubo (p.c., December 2009).

11 This assumption, which is in accordance with our experience as researchers, can be shown to be supported by experimental results, as discussed in Hoji 2009.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 5/49 informant fails to find such , the informant's judgment should necessarily be "complete unacceptability" on  under (a, b). For it must have been deduced from the hypotheses in question that (a, b) arises only with such a numeration. This is the source of the fundamental asymmetry between a *Schema-based prediction and an okSchema-based prediction in terms of the significance of their failure (to be borne out); the asymmetry will be the most crucial conceptual basis of what will be presented in this paper; see below.

2.4. Empirical rigor, "facts," and Confirmed Schematic Asymmetries Before proceeding further, I would like turn to the following remarks by Feynman.12

The history of the thing, briefly, is this. The ancients first observed the way the planets seemed to move in the sky and concluded that they all, along with the earth, went around the sun. This discovery was later made independently by Copernicus, after people had forgotten that it had already been made. Now the next question that came up for study was: exactly how do they go around the sun, that is, with exactly what kind of motion? Do they go with the sun as the centre of a circle, or do they go in some other kind of curve? How fast do they move? And so on. This discovery took longer to make. The times after Copernicus were times in which there were great debates about whether the planets in fact went around the sun along with the earth, or whether the earth was at the centre of the universe and so on. Then a man named Tycho Brahe evolved a way of answering the question. He thought that it might perhaps be a good idea to look very very carefully and to record exactly where the planets appear in the sky, and then the alternative theories might be distinguished from one another. This is the key of modern science and it was the beginning of the true understanding of Nature—this idea to look at the thing, to record the details, and to hope that in the information thus obtained might lie a clue to one or another theoretical interpretation. So Tycho, a rich man who owned an island near Copenhagen, outfitted his island with great brass circles and special observing positions, and recorded night after night the position of the planets. It is only through such hard work that we can find out anything. When all these data were collected they came into the hands of Kepler, who then tried to analyse what kind motion the planets made around the sun. And he did this by a method of trial and error. At one state he thought he had it; he figured out that they went around the sun in circles with the sun off centre. Then Kepler noticed that one planet, I think it was Mars, was eight minutes of arc off, and he decided this was too big for Tycho Brahe to have made an error, and that this was not the right answer. So because of the precision of the experiments he was able to proceed to another trial and ultimately found out three things [i.e., Kepler's three laws of planetary motion, HH]." Feynman (1965/94; pp. 5-6))

Given that "[i]t is only through such hard work that we can find out anything," it is clear that we should bring the utmost rigor to our attempt to identify what the "facts" are. Without being able to identify what is likely a reflection of properties of the Computational System, neither could we specify the consequences of "our guess," nor could we compare them with the results of a "very carefully checked experiment" (because it would in that case not be possible to design such experiments, to begin with). See the Feynman remarks quoted at the outset of this paper. It is proposed in Hoji 2009 that what we can regard as a likely reflection of properties of the Computational System is a confirmed schematic asymmetry such that sentences conforming to one type of Schema are always judged to be completely unacceptable under a specified interpretation while those conforming to the other type of Schema are not necessarily judged to be completely unacceptable. The asymmetry follows from the considerations given in sections 2.2-2.4. In Hoji 2009, the former type of Schema is called a *Schema and sentences conforming to it are called *Examples; the latter type of Schema is called an okSchema and sentences conforming to it are called okExamples. We can thus characterize a *Schema-based prediction and an okSchema-based prediction as follow:

12 This is taken from the first lecture of his Messenger Lectures, "The Law of Gravitation: an example of physical law" reproduced in Feynman 1965/1994.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 6/49 (0) A *Schema-based prediction: The informant judgment on the presented sentence  under interpretation a, b) is always "completely unacceptable" for any *Example conforming to a *Schema.

(0) An okSchema-based prediction: The informant judgment on the presented sentence  under interpretation a, b) is not necessarily "completely unacceptable" for okExamples conforming to an okSchema.

There are two crucial points intended by schematic asymmetry. One is that the contrast of significance is not between examples but it is between Schemata. The other is that the contrast must be such that a *Schema-based prediction (see (0)) has survived a rigorous disconfirmation attempt and is accompanied by the confirmation of the corresponding okSchema-based predictions. The formulation of a *Schema-based prediction in (0) is thus "definitive," so to speak. For an okSchema-based prediction, on the other hand, there is a continuum of formulations from one extreme in (0) to the other in (0), with (0) falling between the two extremes.

(0) An okSchema-based prediction—extreme version 1: The informant judgment on the presented sentence  under interpretation a, b) is not "completely unacceptable" for some okExample conforming to an okSchema.

(0) An okSchema-based prediction—extreme version 2: The informant judgment on the presented sentence  under interpretation a, b) is "fully acceptable" for any okExample conforming to an okSchema.

This is a reflection of the fundamental asymmetry between a *Schema-based prediction and an okSchema- based prediction. Under the formulation of an okSchema-based prediction as in (0) or (0), and taking the formulation of a *Schema-based prediction in (0) as 'invariant', we can state the fundamental asymmetry as follows: okSchema-based predictions cannot be disconfirmed and they can only be confirmed; *Schema-based predictions, on the other hand, can be disconfirmed although they cannot be confirmed. The informant judgment that  is not completely unacceptable under (a, b) (even if not fully acceptable) would therefore disconfirm a *Schema-based prediction because that would mean, contrary to the prediction, that there is numeration  corresponding to  that would result in LF() (hence SR()) compatible with (a, b) and PF() non-distinct from . While the marginal acceptability would thus disconfirm a *Schema-based prediction, it would be compatible with, and hence would confirm, an okSchema-based prediction as formulated in (0) or (0). Given that the ultimate testability of our hypotheses lies in their being subject to disconfirmation, what makes our hypotheses testable is the *Schema-based predictions they give rise to. To put it differently, it is by making *Schema-based predictions that we can seek to establish a "fact" that needs to be explained in research that is concerned with the properties of the Computational System, a "fact" that serves as evidence for or against hypotheses about the Computational System. Let us say that a predicted schematic asymmetry gets confirmed, i.e., a confirmed schematic asymmetry obtains iff the informants' judgments on *Examples are consistently "completely unacceptable" and their judgments on the corresponding okExamples are not "completely unacceptable." By using the numerical values of "0" and "100" for "complete unacceptability" and "full acceptability," respectively, we can more accurately express what we intend as follows: we say that a confirmed schematic asymmetry obtains iff the "representative value" of the *Schema is "0"13 and that of the corresponding okSchemata is

13 In actual practice, we must allow some room for the possibility of "errors" committed by informants. We might therefore have to be "content" with something like "5 or less" as the "representative value" of the *Schema among the entire informants, on the scale of "0" (for complete unacceptability) to "100" (for full acceptability), for example.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 7/49 higher than "0."14 On the basis of the considerations given above, I would like to maintain that confirmed schematic asymmetries are like "minimal units of facts" for research concerned with the properties of the Computational System. The *Schema-based prediction in question must survive a rigorous test of disconfirmation while at the same time the corresponding okSchema-based predictions must be confirmed. Otherwise, the predicted schematic asymmetry does not get confirmed. If the predicted schematic asymmetry does not get confirmed, i.e., in the absence of a confirmed schematic asymmetry, the hypotheses that have resulted in the prediction of the schematic asymmetry should not be used in deducing further theoretical consequences or deriving further empirical predictions. That is the main content of the methodology advocated in Hoji 2009.15 As noted above, while the requirement on the *Schema-based prediction is quite strict, how strict the requirement on an okSchema-based prediction should be may depend on various factors. We surely cannot expect to be able to convince others if the "representative value" of the okSchemata is "10," "20," or "30," for example, on the scale of "0" (for complete unacceptability) to "100" (for full acceptability), even if that of the corresponding *Schema is "0". While it is bound to be a subjective matter to determine what the "representative value" of the okSchemata should be in order for a confirmed schematic asymmetry to obtain, the researchers themselves perhaps should aspire to the "standard" suggested in the formulation of an okSchema-based prediction in (0), leaving aside its actual feasibility in every experiment.16 Be that as it may, one may suggest that identifying confirmed schematic asymmetries is analogous to the rigorous observation and recording of the positions of planets done by Tycho Brahe; see the Feynman remarks quoted above at the beginning of section 2.4.17

2.5. The significance of experiment results Before turning to the discussion of empirical materials, I would like to make one last point in relation to the significance of experimental results. Suppose that we have designed and conducted an experiment

It must be understood, however, that, if some informants consistently find *Examples of a given *Schema more or less acceptable, that should be regarded as a serious challenge to the prediction in question even if the "representative value" of the *Schema among the entire informants is quite low.

14 The "representative value" of a Schema is based on the informant judgments on the Examples that conform to the Schema; see Ueyama 2010 for more details.

15 The proposed method is called the Evaluation of Predicted Schematic Asymmetry (EPSA) method in Hoji 2009.

16 Hoji 2009 provides a great deal more discussion on the relevant issues, making reference to concepts such as informant's resourcefulness, single-informant experiments and multiple-informant experiments; section 6.4 below briefly addresses some of the issues. One of the points made in Hoji 2009 is that we can address across-informant repeatability meaningfully only if we have obtained within-informant repeatability (especially for the researchers themselves) where repeatability is in terms of a confirmed schematic asymmetry.

17 One may object that identifying confirmed schematic asymmetries is more "theory-driven" than the observation of the motion of the planets because the construction of a *Schema and the corresponding okSchemata is based on hypotheses about properties of the Computational System, hypotheses (which are called bridging statements in Hoji 2009) about how a certain type of informant intuition arises only if a certain structural condition is met at LF, and hypotheses (which are called pf-LF correspondences in Hoji 2009)—presumably derived from more "basic" hypotheses—about what LF representation(s) a particular surface phonetic string can, cannot, or must correspond to, etc. While such an objection might be justified in relation to the analogy drawn in the text between confirmed schematic asymmetries and Tycho Brahe's observations, it no longer applies once we start considering data accumulated with the aid of observation devices. The accuracy of the observation of the planetary motions was enhanced (dramatically) by the introduction of various observation devices, including telescopes, and how to interpret what is "observed" by such devices is based on theories of various phenomena, including optics.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 8/49 to see if a predicted schematic asymmetry gets confirmed, i.e., whether the *Schema-based prediction does not get disconfirmed and, furthermore, the corresponding okSchema-based predictions get confirmed; see (0)-(0) and the discussion thereabout. Does that mean that we are justified to conclude that we now have a confirmed schematic asymmetry? Recall that a confirmed schematic asymmetry obtains iff the *Schema- based prediction has survived a rigorous attempt of disconfirmation and at the same time the corresponding okSchema-based predictions are confirmed. The fact that the result of a particular experiment is in harmony with the prediction therefore does not necessarily mean that we have obtained a confirmed schematic asymmetry. For it is possible that the experimental result thus obtained might be crucially due to the particular choice of the lexical items in the Examples conforming to the Schemata in question. What is claimed by a *Schema-based prediction is that the informant judges any *Example (conforming to a *Schema) to be completely unacceptable under the specified interpretation. While the researcher might have tried his or her best to construct the *Examples that are most natural and the easiest to parse for the intended interpretation—as he or she in fact should—it is still possible that the researcher did not have enough ingenuity to construct *Examples that are not completely unacceptable under the specified interpretation. Once the predicted experimental results have obtained in his or her own experiment(s), the researcher should therefore invite other researchers to construct *Examples (as well as okExamples) in accordance with the predicted schematic asymmetries and to conduct their own experiments. That is to say, having obtained the expected informant judgments in our own experiment(s) is merely a start in terms of our rigorous disconfirmation attempt. Other interested researchers are thus strongly encouraged to conduct experiments themselves on the basis of the predicted schematic asymmetries, and make various adjustments on the lexical items in the actual Examples conforming to the Schemata, doing the best they can to construct *Examples of the *Schema that are not completely unacceptable under the specified interpretation. The prediction is that the *Examples conforming to the *Schema are completely unacceptable under the specified interpretation no matter what efforts might be made to render the *Examples not completely unacceptable. If the *Schema-based prediction(s) did not get a value very close to zero in any of those experiment, such a result would constitute a serious challenge to our hypotheses; we must consider how such informant judgments arise. That should be our basic attitude if we are interested in discovering the properties of the Computational System of the language faculty in line with the general scientific method schematized in (0). If the *Schema-based prediction does not get disconfirmed in many such experiments, we will finally be in a position to declare, with some confidence, that the *Schema-based prediction has survived a rigorous disconfirmation attempt, and to the extent that the corresponding okSchema-based predictions get confirmed, we can say, again with some confidence, that we have indeed obtained a confirmed schematic asymmetry.18 As our research advances, confirmed schematic asymmetries may become more and more general and abstract, and we expect to be able to deduce their explanations (i.e., hypotheses that account for them) from more basic and fundamental principles, approaching something that may deserve to be called a truly explanatory theory of the language faculty. No matter how abstract our theory of the language faculty may become, however, its empirical consequences should remain expressible, ultimately, in terms of confirmed schematic asymmetries.

18 The research attitude advocated here is thus quite different from one that takes the presence of some contrast between some examples for some speakers in the predicted direction as evidence in support of the hypotheses that give rise to the prediction under discussion. As argued above, the mere fact that such a contrast obtains between some examples for some speakers does not mean much for research concerned with the properties of the Computational System in line with the general scientific method schematized in (0). I might add in passing that if a *Schema does not specify anything about prosody or intonation, the claim must be that *Examples conforming to the *Schema is totally unacceptable no matter what prosody/intonation might be used; cf. Miyagawa and Arikawa 2007: 652 (at the end of their section 3) for a remark that seems to be based on a rather different view.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 9/49 3. Hypotheses and their empirical consequences 3.1. Anaphors in English It has been observed at least since the mid 1960s, that the informants' judgments are in accord with a general pattern as illustrated in (0).

(0) a. John recommended himself. b. *John thought that Mary had recommended himself.

Attempts have been made to express the contrast as a reflection of the Computational System, resulting in a hypothesis about the Computational System that has the effect in (0) and a hypothesis about the mental Lexicon of speakers of English as in (0), as discussed in Chomsky 1981.

(0) A [+A] category must have an antecedent in its local domain.

(0) Himself is marked [+A] in the mental Lexicon of speakers of English.

By defining "local domain" so as to ensure that in (0) NP2 is, but NP1 is not, in the local domain of NP3, the contrast in (0) is accounted for.19

(0) NP1 Verb [that NP2 Verb NP3]

That is to say, if one puts forth or accepts a hypothesis that expression  is marked [+A], one can make a testable prediction—as long as one also accepts something like (0) and the definition of "local domain" that has the effect noted above. One of the clearest predictions is that sentences containing  are unacceptable if  is an embedded object and is interpreted as sharing the same value as the matrix subject. We can state the predicted schematic asymmetry as follows:

(0) a. okSchema NP20 V himself NP=himself

b. *Schema NP1 V that NP2 V himself NP1=himself

c. okSchema NP1 V that NP2 V him NP1=him

As suggested above, what is predicted is a schematic asymmetry; more specifically, the prediction is that there are no Examples conforming to (0b) that are judged not completely unacceptable while there are Examples conforming to (0a) and (0c) that are judged (more or less) acceptable under the interpretations indicated in (0a) and (0c). We are not going to address here how robust the informant judgments are on the predicted schematic asymmetry; we only note here that an informal survey conducted a few years ago suggests that they are fairly robust and they are in accordance with (0).21

19 The use of "NP" in place of "DP" here and elsewhere in the paper is immaterial to the main points of the paper.

20 The choice of NP over DP is inconsequential in this presentation.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 10/49 3.2. Lexical Hypotheses in Japanese

3.2.1. Hypotheses In much of the generative research over the past 20 years, Japanese expressions such as otagai, zibun-zisin, and kare-zisin have been assumed to be marked [+A] in the sense noted in section 3, and they are called local anaphors in Japanese. Many generative works dealing with Japanese have provided some paradigm or other in support of such hypotheses and other works have derived and discussed various empirical as well as theoretical consequences by assuming the validity of the lexical hypotheses under discussion.22 The claim that otagai, zibun-zisin, and kare-zisin are local anaphors can be stated as in (0); see (0).23

(0) Specifications in the mental Lexicon of speakers of Japanese: a. Otagai is marked [+A]. b. Zibun-zisin is marked [+A]. c. Kare-zisin is marked [+A].

The properties of the Computational System are assumed to be universal, with the possible exception having to do with the so-called head parameter. The hypothesis about the Computational System having the effect in (0), repeated here, is thus considered universal.

(0) A [+A] category must have an antecedent in its local domain.

A natural application to Japanese of the notion of "local domain" as understood in relation to (0) would lead us to accept that in (0) NP2 is, but NP1 is not, in the local domain of NP3.

(0) NP1-ga [NP2-ga NP3-{o/ni} to] Verb 'NP1 Verb that NP2 Verb NP3'

With the language-specific lexical hypotheses in (0), the universal hypothesis in (0), along with the articulation of "local domains" in Japanese just given, we make testable predictions. We turn to them in the following subsection.

3.2.2. Predictions

3.2.2.1. Otagai The schematic asymmetries as indicated in (0) and (0) are among the consequences of adopting (0), (0a), and the characterization of the "local domain" as noted above.

(0) a. okSchema NP-ga/wa [NP1-ga otagai-o/ni V-ru/ta {to/no ka}] V-ru/ta under the reciprocal reading of otagai with NP1 as its "antecedent"

b. *Schema NP1-ga/wa [NP-ga otagai-o/ni V-ru/ta {to/no ka}] V-ru/ta

21 See Appendix for a summary of the informal survey.

22 Such works are in fact numerous and they include Katada 1991, Nishigauchi 1992, Saito 1992, 2003, and Takita 2009.

23 I leave aside the issue as to whether each of (0) is derived from more basic statements; this applies to (0) as well.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 11/49 under the reciprocal reading of otagai with NP1 as its "antecedent"

c. okSchema NP1-ga/wa [NP-ga karera-o/ni V-ru/ta {to/no ka}] V-ru/ta under the coreference between karera and NP1

(0) a. okSchema [[otagai-o/ni V-ru/ta] NP1] under the reciprocal reading of otagai with NP1 as its "antecedent"

b. *Schema [[[NP-ga otagai-o/ni V-ru/ta {to/no ka}] V-ru/ta] NP1] under the reciprocal reading of otagai with NP1 as its "antecedent"

c. okSchema [[[NP-ga karera-o/ni V-ru/ta {to/no ka}] V-ru/ta] NP1] under the coreference between karera and NP1

On the basis of the Schemata in (0), we can construct the Examples in (0) and (0).

(0) a. okExample Mary-wa [John to Bill-ga otagai-ni toohyoosi-ta to] omoikonde-i-ta 'Mary thought that John and Bill had voted for each other.'

b. *Example John to Bill-wa [Mary-ga otagai-ni toohyoosi-ta to] omoikonde-i-ta 'John and Bill thought that Mary had voted for each other.'

c. okExample John to Bill-wa [Mary-ga karera-ni toohyoosi-ta to] omoikonde-i-ta 'John and Bill thought that Mary had voted for them.'

(0) a. okExample Sensei-wa [John to Bill-ga naze otagai-o suisensi-ta no ka] mattaku wakara-nakat-ta 'The teacher had no idea why John and Bill had recommended each other."

b. *Example John to Bill-wa [sensei-ga naze otagai-o suisensi-ta no ka] mattaku wakara-nakat-ta 'John and Bill had no idea why the teacher had recommended each other."

c. okExample John to Bill-wa [sensei-ga naze karera-o suisensi-ta no ka] mattaku wakara-nakat-ta 'John and Bill had no idea why the teacher had recommended them."

On the basis of the Schemata in (0), we can construct the Examples in (0) and (0).

(0) a. okExample [[ec sensyuu-no senkyo-de otagai-ni toohyoosi-ta] John to Bill]-wa Susan-ga dare-ni toohyoosi-ta ka sit-te odoroi-ta. 'John and Bill, who had voted for each other at the election last week, were surprised to learn who Susan had voted for.'

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 12/49 b. *Example [[ ec [[Susan-ga sensyuu-no senkyo-de otagai-ni toohyoosi-ta] to] omoikonde-i-ta] John to Bill]-wa Susan-ga dare-ni toohyoosi-ta ka sit-te odoroi-ta. 'John and Bill, who thought that Susan had voted for each other at the election last week, were surprised to learn who Susan had voted for.'

c. okExample [[ ec [[Susan-ga sensyuu-no senkyo-de karera-ni toohyoosi-ta] to] omoikonde-i-ta] John to Bill]-wa Susan-ga dare-ni toohyoosi-ta ka sit-te odoroi-ta. 'John and Bill, who thought that Susan had voted for them for the election last week, were surprised to learn who Susan had voted for.'

(0) a. okExample [[ec kondo-no yakusyoku-ni otagai-o suisensi-ta] John to Bill]-wa iroirona hito-ni meeru-o okut-te riyuu-o setumeisi-te-i-ru rasii. 'I hear that John and Bill, who had recommended each other for the new post, are emailing various people to explain why.'

b. *Example [[ec [Mike-ga kondo-no yakusyoku-ni naze otagai-o suisensi-ka] siritagat-te-i-ta] John to Bill]- wa iroirona hito-ni meeru-o okut-te riyuu-o sirabe-te-i-ru rasii. 'I hear that John and Bill, who wanted to know why Mike had recommended each other for the new post, are emailing various people to find out why.'

c. okExample [[ec [Mike-ga kondo-no yakusyoku-ni naze karera-o suisensi-ka] siritagat-te-i-ta] John to Bill]- wa iroirona hito-ni meeru-o okut-te riyuu-o sirabe-te-i-ru rasii. 'I hear that John and Bill, who wanted to know why Mike had recommended them for the new post, are emailing various people to find out why.'

3.2.2.2. Zibun-zisin Similarly, the schematic asymmetries as indicated in (0) and (0) are among the consequences of adopting (0), (0b), and the characterization of the "local domain" as noted above.

(0) a. okSchema NP-ga/wa [NP1-ga zibun-zizin-o/ni V-ru/ta {to/no ka}] V-ru/ta NP1 = zibun-zisin

b. *Schema NP1-ga/wa [NP-ga zibun-zizin-o/ni V-ru/ta {to/no ka}] V-ru/ta NP1 = zibun-zisin

c. okSchema NP1-ga/wa [NP-ga {kare/kanozyo}-o/ni V-ru/ta {to/no ka}] V-ru/ta NP1 = kare/kanozyo

(0) a. okSchema [[[α-ga zibun-zisin-o/ni V-T to] V-T] NP]-wa ...  = zibun-zisin

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 13/49 b. *Schema [[[NP-ga zibunzisin-o/ni V-T to] V-T] α]-wa ...  = zibun-zisin

c. okSchema [[[NP-ga {kare/kanozyo}-o/ni V-T to] V-T] α]-wa ...  = kare/kanozyo

On the basis of the Schemata in (0) and (0), we can construct the Examples in (0)-(0).

(0) a. okExample John-wa [Mary-ga zibun-zisin-ni toohyoosi-ta to] omoikonde-i-ta 'John thought that Mary had voted for herself.'

b. *Example John-wa [Mary-ga zibun-zisin-ni toohyoosi-ta to] omoikonde-i-ta 'John thought that Mary had voted for himself.'

c. okExample John-wa [Mary-ga kare-ni toohyoosi-ta to] omoikonde-i-ta 'John thought that Mary had voted for him.'

(0) a. okExample John-wa [Mary-ga zibun-zisin-o suisensi-ta to] bakari omotte-i-ta 'John firmly believed that Mary had recommended herself.'

b. *Example John-wa [Mary-ga zibun-zisin-o suisensi-ta to] bakari omotte-i-ta 'John firmly believed that Mary had recommended himself.'

c. okExample John-wa [Mary-ga kare-o suisensi-ta to] bakari omotte-i-ta 'John firmly believed that Mary had recommended him.'

(0) a. okExample Ziro-wa [Hanako-ga zibun-zisin-o hihansi-ta to] it-ta 'Ziro said that Hanako had criticized herself.'

b. *Example Ziro-wa [Hanako-ga zibun-zisin-o hihansi-ta to] it-ta 'Ziro said that Hanako had criticized himself.'

c. okExample Ziro-wa [Hanako-ga kare-o hihansi-ta to] it-ta 'John said that Hanako had criticized him.'

(0) a. okExample [[John-ga zibun-zisin-ni toohyoosu-ru to]-wa omotte-mo-mi-nakat-ta] Yoko]-wa John-ga dare- ni toohyoosi-ta ka sit-ta toki totemo odoroi-ta. 'Yoko, who did not even think that John might vote for himself, was very surprised when she

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 14/49 learned who John had voted for.'

b. *Example [[John-ga zibun-zisin-ni toohyoosu-ru to]-wa omotte-mo-mi-nakat-ta] Yoko]-wa John-ga dare- ni toohyoosi-ta ka sit-ta toki totemo odoroi-ta. 'Yoko, who did not even think that John might vote for herself, was very surprised when she learned who John had voted for.'

c. okExample [[John-ga kanozyo-ni toohyoosu-ru to]-wa omotte-mo-mi-nakat-ta] Yoko]-wa John-ga dare-ni toohyoosi-ta ka sit-ta toki totemo odoroi-ta. 'Yoko, who did not even think that John might vote for her, was very surprised when she learned who John had voted for.'

(0) a. okExample [[Bill-ga zibun-zisin-o suisensu-ru to] omikondei-ta] John]-wa Bill-ga dare-o suisensi-ta ka sit- ta toki totemo odoroi-ta. 'John, who firmly believed that Bill would recommend himself, was very surprised when he learned who Bill had recommended.'

b. *Example [ ec [Bill-ga zibun-zisin-o suisensu-ru to] omikondei-ta] John]-wa Bill-ga dare-o suisensi-ta ka sit-ta toki totemo odoroi-ta. 'John, who firmly believed that Bill would recommend himself, was very surprised when he learned who Bill had recommended.'

c. okExample [ ec [Bill-ga kare-o suisensu-ru to] omikondei-ta] John]-wa Bill-ga dare-o suisensi-ta ka sit-ta toki totemo odoroi-ta. 'John, who firmly believed that Bill would recommend him, was very surprised when he learned who Bill had recommended.'

(0) a. okExample [[Ziro-ga zibun-zisin-o hihansi-te-i-ru to] sit-ta] Hanako]-wa tyotto odoroi-ta. 'Hanako, who learned that Ziro was criticizing himself, was a little surprised.'

b. *Example [ ec [Ziro-ga zibun-zisin-o hihansi-te-i-ru to] sit-ta] Hanako]-wa tyotto odoroi-ta. 'Hanako, who learned that Ziro was criticizing herself, was a little surprised.'

c. okExample [ ec [Ziro-ga kanozyo-o hihansi-te-i-ru to] sit-ta] Hanako]-wa tyotto odoroi-ta. 'Hanako, who learned that Ziro was criticizing her, was a little surprised.'

3.2.2.3. Kare-zisin With (0c), we obtain the schematic asymmetries as indicated in (0) and (0) and we can construct Examples such as those in (0)-(0).

(0) a. okSchema NP-ga/wa [NP1-ga kare-zizin-o/ni V-ru/ta {to/no ka}] V-ru/ta

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 15/49 NP1 = kare-zisin

b. *Schema NP1-ga/wa [NP-ga kare-zizin-o/ni V-ru/ta {to/no ka}] V-ru/ta NP1 = kare-zisin

c. okSchema NP1-ga/wa [NP-ga {kare/kanozyo}-o/ni V-ru/ta {to/no ka}] V-ru/ta NP1 = kare/kanozyo

(0) a. okSchema [[[α-ga kare-zisin-o/ni V-T to] V-T] NP]-wa ...  = kare-zisin

b. *Schema [[[NP-ga kare-zisin-o/ni V-T to] V-T] α]-wa ...  = kare-zisin

c. okSchema [[[NP-ga kare-o/ni V-T to] V-T] α]-wa ...  = kare

(0) a. okExample Mary-wa [John-ga kare-zisin-ni toohyoosi-ta to] omoikonde-i-ta 'Mary thought that John had voted for himself.'

b. *Example John-wa [Mary-ga kare-zisin-ni toohyoosi-ta to] omoikonde-i-ta 'John thought that Mary had voted for himself.'

c. okExample John-wa [Mary-ga kare-ni toohyoosi-ta to] omoikonde-i-ta 'John thought that Mary had voted for him.'

(0) a. okExample Mary-wa [John-ga kare-zisin-o suisensi-ta to] bakari omotte-i-ta 'Mary firmly believed that John had recommended himself.'

b. *Example John-wa [Mary-ga kare-zisin-o suisensi-ta to] bakari omotte-i-ta 'John firmly believed that Mary had recommended himself.'

c. okExample John-wa [Mary-ga kare-o suisensi-ta to] bakari omotte-i-ta 'John firmly believed that Mary had recommended him.'

(0) a. okExample Hanako-wa [Ziro-ga kare-zisin-o hihansii-ta to] it-ta 'Hanako said that Ziro had criticized himself.'

b. *Example

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 16/49 Ziro-wa [Hanako-ga kare-zisin-o hihansii-ta to] it-ta 'Ziro said that Hanako had criticized himself.'

c. okExample Ziro-wa [Hanako-ga kare-o hihansii-ta to] it-ta 'Ziro said that Hanako had criticized him.'

(0) a. okExample [[John-ga kare-zisin-ni toohyoosu-ru to]-wa omotte-mo-mi-nakat-ta] Yoko]-wa John-ga dare-ni toohyoosi-ta ka sit-ta toki totemo odoroi-ta. 'Yoko, who did not even think that John might vote for himself, was very surprised when she learned who John had voted for.'

b. *Example [[Yoko-ga kare-zisin-ni toohyoosu-ru to]-wa omotte-mo-mi-nakat-ta] John]-wa Yoko-ga dare- ni toohyoosi-ta ka sit-ta toki totemo odoroi-ta. 'John, who did not even think that Yoko might vote for himself, was very surprised when he learned who Yoko had voted for.'

c. okExample [[Yoko-ga kare-ni toohyoosu-ru to]-wa omotte-mo-mi-nakat-ta] John]-wa Yoko-ga dare-ni toohyoosi-ta ka sit-ta toki totemo odoroi-ta. 'John, who did not even think that Yoko might vote for him, was very surprised when he learned who Yoko had voted for.'

(0) a. okExample [[John-ga kare-zisin-o suisensu-ru to]-wa omotte-mo minakat-ta] Yoko]-wa John-ga dare-o suisensi-ta ka sit-ta toki totemo odoroi-ta. 'Yoko, who did not even think that John would recommend himself, was very surprised when she learned who John had recommended.'

b. *Example [ ec [Yoko-ga kare-zisin-o suisensu-ru to] omikondei-ta] John]-wa Yoko-ga dare-o suisensi-ta ka sit-ta toki totemo odoroi-ta. 'John, who firmly believed that Yoko would recommend himself, was very surprised when he learned who Yoko had recommended.'

c. okExample [ ec [Yoko-ga kare-o suisensu-ru to] omikondei-ta] John]-wa Yoko-ga dare-o suisensi-ta ka sit- ta toki totemo odoroi-ta. 'John, who firmly believed that Yoko would recommend him, was very surprised when he learned who Yoko had recommended.'

(0) a. okExample [[Ziro-ga kare-zisin-o hihansi-te-i-ru to] sit-ta] Hanako]-wa tyotto odoroi-ta. 'Hanako, who learned that Ziro was criticizing himself, was a little surprised.'

b. *Example [ ec [Hanako-ga kare-zisin-o hihansi-te-i-ru to] sit-ta] Ziro]-wa tyotto odoroi-ta. 'Ziro, who learned that Hanako was criticizing himself, was a little surprised.'

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 17/49 c. okExample [ ec [Hanako-ga kare-o hihansi-te-i-ru to] sit-ta] Ziro]-wa tyotto odoroi-ta. 'Ziro, who learned that Hanako was criticizing him, was a little surprised.'

4. Experiments on "local anaphors" in Japanese24 4.1. The general design of experiments In our on-line experiments, the Examples are presented to the informants, including the specification of their intended interpretation. The Examples, the instructions, and the specification of the intended interpretations are all in Japanese, and the Examples presented to the informants do not contain ec, unlike the examples given in section 4.2. Bracketing is supplied when we thought that it would help the informants parse the sentence easily (typically indicating the sentence boundaries). The specifications of the intended interpretations are as in (0), for example, once they are translated into English.

(0) a. under the interpretation that "John voted for Bill and Bill voted for John" b. under the interpretation that karera 'them' and John to Bill 'John and Bill' refer to the same individuals c. under the interpretation that kare-zisin and Ziro refer to the same person

In the experiment on the predicted schematic asymmetries in (0) and (0), for example, the 12 Examples in (0)-(0) are presented to informants in a random fashion, (i) one at a time or (ii) three at a time (three Examples as a set of an okExample, a *Example, and another okExample, e.g., those in (0)), depending upon the test type chosen by each informant. Depending upon the test type of their choice, they either (i) choose "No" (for "not acceptable no matter what") or "Yes" (for "(more or less) acceptable") or (ii) indicate how acceptable they find each example by clicking one of the five radio buttons as in (0).

(0) Bad < ===== > Good o o o o o

(0) 0, 25, 50, 75, 100

What the informant has indicated is converted to numerical values as in (0), i.e., the worst score is converted to "0" and the best score to "100." Likewise, the "Yes" or the "No" answer in the "Yes-or-No" test gets converted to "0" or "100," respectively although the informants are not informed how their judgments get converted to numerical values. The informants are allowed to return to the experiment website and report their judgments in the same experiment again, and in fact as many times as they wish; they may repeat the same "test type" as before or different "test types" (as to "Yes-or-No" or "Five-ranking" and also as to "one at a time," "three at a time" (or "all in one sheet" in some cases)). In the event that one informant has reported his/her judgment on the same experiment more than once, regardless of the "test type," the average score on a given example by that informant is used when calculating the average score on that example by the entire informants for the experiment. The results we have obtained so far indicate that the choice of the "test type" does not make a significant difference at all. See, for example, the chart in (0), which shows the results of "different test types" of the experiment on the 12 Examples given in (0)-(0). 29 informants have responded and some have reported their judgment more than once.25

24 I should like to acknowledge that the program for the basic design of our on-line experiments has been created by Ayumi Ueyama.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 18/49 (0) A summary of the experimental results based on the different test types (as of March 11, 2010): EPSA [5]-#1 Schema A Schema B Schema C Five-ranking (in pairs) 99 values 95 99 values 57 99 values 83 Yes-or-No (one each) 63 values 95 61 values 55 63 values 85 Yes-or-No (in pairs) 64 values 98 62 values 59 64 values 76 Five-ranking (one each) 32 values 89 31 values 49 32 values 82

"Schema A" covers the okSchemata in (0a) and (0a), "Schema B" the *Schemata in (0b) and (0b), and "Schema C" the okSchemata in (0c) and (0c). As can be seen from (0), there really is not much difference among the different "test types."

4.2. The results of the experiments

4.2.1. Otagai Provided in (0) is a summary (as of January 30, 2010) of the results of the experiment on the predicted schematic asymmetry in (0) and (0).

25 17 of the 29 informants have provided their judgments more than once, and 8 of them have provided their judgments on the same example as many as 4 times.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 19/49 (0) A summary of the results of an experiment as of March 11, 2010:26 Hypothesis: Otagai is a local anaphor. Otagai is in the embedded object position. okNP-ga/wa [NP1-ga otagai-o/ni V-ru/ta {to/no ka}] V- Schema 1 A 58 values 97 ru/ta (under the reciprocal reading of otagai taking NP1 as its "antecedent") Schema group 1 *NP1-ga/wa [NP-ga otagai-o/ni V-ru/ta {to/no ka}] V- Schema 1 B 57 values 64 ru/ta (under the reciprocal reading of otagai taking NP1 as its "antecedent") okNP1-ga/wa [NP-ga karera-o/ni V-ru/ta {to/no ka}] V- Schema 1 C 58 values 87 ru/ta (with the coreference between karera and NP1) Otagai is in the embedded object position. The intended antecedent is the relative head. ok[[otagai-o/ni V-ru/ta] NP1] (under the reciprocal reading Schema 2 A 54 values 95 of otagai taking NP1 as its "antecedent")

Schema group 2 *[[[NP-ga otagai-o/ni V-ru/ta {to/no ka}] V-ru/ta] NP1] Schema 2 B 54 values 57 (under the reciprocal reading of otagai taking NP1 as its "antecedent.") ok[[[NP-ga karera-o/ni V-ru/ta {to/no ka}] V-ru/ta] NP1] Schema 2 C 54 values 73 (with the coreference between karera and NP1) 29 participants, 769 answers

"Schema group 1" is for (0) and "Schema group 2" is for (0). "Schema 1 A" covers the okExamples in (0a) and (0a), "Schema 1 B" the *Examples in (0b) and (0b), and "Schema 1 C" the okExamples in (0c) and (0c). Likewise, "Schema 2 A" covers the okExamples in (0a) and (0a), "Schema 1 B" the *Examples in (0b) and (0b), and "Schema 1 C" the okExamples in (0c) and (0c). "769 answers" means that there have been 769 occurrences of a reported judgment. As noted, some informants have judged the same example more than once; but in such cases the values in (0) are based on the average score on a given example by the same informant. The values of "Schema 1 B" and "Schema 2 B" should be close to "0" according to the predicted schematic asymmetries in (0) and (0). The informant judgments as indicated in (0) thus clearly disconfirm the *Schema-based predictions based on the lexical hypothesis in (0a).

4.2.2. Zibun-zisin A summary of the results of the experiment on the predicted schematic asymmetries in (0) and (0) is given in (0).

26 Corresponding to each Schema in (0), there are two Examples; see (0)-(0). As noted above, if one informant has reported his/her judgment on the same Example more than once, regardless of the "test type," the informant's average judgment on the Example is used when calculating the average score on that example by the entire informants for the experiment. If every informant had provided his/her judgment on at least one of the Examples for each Schema, the value for each Schema in (0) would be 58 (=29x2), as in the case of two Schemata in (0). Because some informants did not report their judgments on some Examples, some of the values in (0) are less than 58.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 20/49 (0) A summary of the results of an experiment as of March 11, 2010:27 Hypothesis: Zibun-zisin is a local anaphor. the topic construction okNP-ga/wa [α-ga zibun-zisin-o/ni V-T to] V-T (under the Schema 1 A 51 values 96 interpretation that zibun-zisin and α are understood to refer to the same individual)

Schema group 1 *α-ga/wa [NP-ga zibun-zisin-o/ni V-T to] V-T (under the Schema 1 B 51 values 64 interpretation that zibun-zisin and α are understood to refer to the same individual) okα-ga/wa [NP-ga kare-o/ni V-T to] V-T (under the Schema 1 C 51 values 80 interpretation that kare and α are understood to refer to the same individual) the relative clause construction ok[[[α-ga zibun-zisin-o/ni V-T to] V-T] NP]-wa ... (under the Schema 2 A 48 values 99 interpretation that zibun-zisin and α are understood to refer to the same individual)

Schema group 2 *[[[NP-ga zibun-zisin-o/ni V-T to] V-T] α]-wa ... (under the Schema 2 B 48 values 52 interpretation that zibun-zisin and α are understood to refer to the same individual) ok[[[NP-ga kare-o/ni V-T to] V-T] α]-wa ... (under the Schema 2 C 38 values 67 interpretation that kare and α are understood to refer to the same individual) 17 participants 538 answers

As in the case of the *Schema-based prediction in (0a) about otagai, the *Schema-based prediction in (0b) about zibun-zisin is also clearly disconfirmed.

4.2.3. Kare-zisin We have not obtained enough informant judgments on the examples in (0)-(0).

27 There are three Examples for each Schema in (0); see (0)-(0). "48 values" in (0) means that every informant has reported his/her judgment on at least one Example on each Schema.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 21/49 (0) A summary of the results of an experiment as of March 7, 2010:28 Hypothesis: Kare-zisin is a local anaphor. in the topic construction okNP-ga/wa [α-ga kare-zisin-o/ni V-T to] V-T (under the Schema 1 A 48 values 92 interpretation that kare-zisin and α are understood to refer to the same individual)

Schema group 1 *α-ga/wa [NP-ga kare-zisin-o/ni V-T to] V-T (under the Schema 1 B 48 values 58 interpretation that kare-zisin and α are understood to refer to the same individual) okα-ga/wa [NP-ga kare-o/ni V-T to] V-T (under the Schema 1 C 48 values 89 interpretation that kare and α are understood to refer to the same individual) in the relative clause construction ok[[[α-ga kare-zisin-o/ni V-T to] V-T] NP]-wa ... (under the Schema 2 A 48 values 93 interpretation that kare-zisin and α are understood to refer to the same individual)

Schema group 2 *[[[NP-ga kare-zisin-o/ni V-T to] V-T] α]-wa ... (under the Schema 2 B 48 values 59 interpretation that kare-zisin and α are understood to refer to the same individual) ok[[[NP-ga kare-o/ni V-T to] V-T] α]-wa ... (under the Schema 2 C 48 values 73 interpretation that kare and α are understood to refer to the same individual) 16 participants 459 answers

The *Schema-based prediction in (0c) about kare-zisin is thus also clearly disconfirmed. Under a general experimental design that was not as systematic as the one that is discussed above, we conduced experiments a few years ago and checked informant judgments on examples that conform to (0b) in experiments where those examples were given as "control." In one of those experiments, such examples conforming to (0b) were judged to be unacceptable consistently only by a few informants out of over 60 informants. In another such experiment, all of the 11 informants accepted examples conforming to (0b).

4.3. Other experiments We have also conducted experiments in which we checked the informant judgments on the *Schema- based prediction that otagai must be c-commanded by "its antecedent." In (0) I provide a summary of the results of one such experiment.

28 For each Schema in (0), there are three Examples; see (0)-(0). "48 values" in (0) means that every informant has reported his/her judgment on at least one Example on each Schema.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 22/49 (0) A summary of the results of an experiment as of March 11, 2010: Hypothesis: Otagai is a local anaphor. In Schema B, otagai is not c-commanded by its "antecedent." okNP1-ga [ogatai-no N]-ni/o V (under the reciprocal Schema 1 A 48 values 95 reading of otagai with NP1 as its "antecedent") Schema group 1 *[otagai-no N]-ga NP1-ni/o V (under the reciprocal Schema 1 B 48 values 72 reading of otagai with NP1 as its "antecedent") ok[karera-no N]-ga NP1-ni/o V (with the coreference Schema 1 C 48 values 66 between karera and NP1) 16 participants 251 answers

As indicated, the *Schema-based prediction is clearly disconfirmed.29 We have also conducted an experiment on examples with the 'respective' (instead of 'reciprocal') interpretation of otagai; the Schemas in those experiments are the same as those in (0) and (0), except that the specified intended interpretation is "under the respective reading of otagai with NP1 as its "antecedent"," as indicated in (0) for the sentence in (0b), for example, which I repeat here.

(0) Under the interpretation that John thought that Mary had voted for John and Bill thought that Mary had voted for Bill.

(0) b. *Example John to Bill-wa [Mary-ga otagai-ni toohyoosi-ta to] omoikonde-i-ta 'John and Bill thought that Mary had voted for each other.'

The results of those experiments, as indicated in (0) and (0) show that the informant judgments do not seem to be affected much at all by whether or not the "antecedent" of otagai is in the local domain of otagai (see (0)), or by whether or not the "antecedent" of otagai c-commands otagai (see (0)).

29 Hoji 2006b contains examples conforming to such a *Schema and other *Schemata in relation to otagai and reports that such *Schema-based predictions are also clearly disconfirmed.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 23/49 (0) A summary of the results of an experiment as of March 11, 2010: Hypothesis: Otagai is a local anaphor. Otagai is in the embedded object position. okNP-ga/wa [NP1-ga otagai-o/ni V-ru/ta {to/no ka}] V- Schema 1 A 29 values 47 ru/ta (under the respective reading of otagai taking NP1 as its "antecedent" ) Schema group 1 *NP1-ga/wa [NP-ga otagai-o/ni V-ru/ta {to/no ka}] V- Schema 1 B 30 values 51 ru/ta (under the respective reading of otagai taking NP1 as its "antecedent" ) okNP1-ga/wa [NP-ga karera-o/ni V-ru/ta {to/no ka}] V- Schema 1 C 29 values 97 ru/ta (with the coreference between karera and NP1 ) Otagai is in the embedded object position. The intended antecedent is the relative head. ok[[otagai-o/ni V-ru/ta] NP1] (under the respective reading Schema 2 A 30 values 47 of otagai taking NP1 as its "antecedent" )

Schema group 2 *[[[NP-ga otagai-o/ni V-ru/ta {to/no ka}] V-ru/ta] NP1] Schema 2 B 30 values 49 (under the respective reading of otagai taking NP1 as its "antecedent" ) ok[[[NP-ga karera-o/ni V-ru/ta {to/no ka}] V-ru/ta] NP1] Schema 2 C 30 values 86 (with the coreference between karera and NP1 ) 15 participants 349 answers

(0) A summary of the results of an experiment as of March 11, 2010: Hypothesis: Otagai is a local anaphor. In Schema B, otagai is not c-commanded by its "antecedent." okNP1-ga [ogatai-no N]-ni/o V (under the "respective Schema 1 A 33 values 57 reading" of otagai taking NP1 as its "antecedent" ) Schema group 1 *[otagai-no N]-ga NP1-ni/o V (under the "respective Schema 1 B 33 values 52 reading" of otagai taking NP1 as its "antecedent" ) ok[karera-no N]-ga NP1-ni/o V (with the coreference Schema 1 C 33 values 79 between karera and NP1 ) 12 participants 191 answers

In yet another experiment, informants were asked to report their judgments on examples like (0), among other examples, where there is no "antecedent" for otagai, at least in any obvious way. Two such examples are given in (0).

(0) a. Haru-no atatakana kaze-ga otagai-o totemo siawase-na kimoti-ni sita. Spring-GEN warm wind-NOM otagai-ACC very happy feeling-DAT made 'The warm spring wind made otagai (=them) feel very happy.'

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 24/49 b. Otagai-ga manzoku nara, boku-wa monku-o iwanai tumori da. otagai-NOM satisfied if I-TOP complaint-ACC say:not plan copula 'If otagai (=both of them) are satisfied, I will not raise issues.'

The value on Schema 1B in (0) is the average "score" on examples like (0), on the scale of 0-100.

(0) A summary of the results of an experiment as of March 11, 2010: Hypothesis: Otagai is a local anaphor.

Schema 1 A 43 values 95 ok Schema group 1 Schema 1 B 45 values 93 * Schema 1 C 46 values 100 ok 16 participants 212 answers

If otagai were like English each other, (0a) and (0b) would correspond to The warm spring wind made each other feel very happy and If each other is satisfied, I will not raise issues, respectively, and the acceptability of examples like (0a) and (0b) would be quite puzzling.

4.4. The absence of local anaphors in Japanese and Fukui's (1986) thesis We have seen that the *Schema-based predictions under the lexical hypotheses in (0) are clearly disconfirmed. It may be possible that someone can in the future come up with a way to modify and hence save a version of the lexical hypotheses in (0). Several attempts are in fact discussed in Hoji 2009 and it is concluded there that such attempts either end up being content-reducing (or degenerating) problemshift in the terms of Lakatos 1979/1978—resulting only in the elimination of the *Schema-based prediction without introducing a new *Schema-based prediction—or simply fail to save the hypotheses under discussion. I leave the challenge of saving those hypotheses in a theoretically progressive way to those who wish to make use of them in their theoretical discussion. While it is not possible to empirically demonstrate the non-existence of elements in Japanese that are marked [+A]—for it is not possible to empirically demonstrate the non-existence of anything—their non- existence in Japanese is an immediate consequence if we adopt the thesis put forth in Fukui 1986. Fukui (1986) proposes that the mental Lexicon of speakers of Japanese does not contain what is responsible for making functional categories "active." Given the assumption that what most crucially underlies a local anaphor is an "active functional category"—cf. Lebeaux 1983 and Chomsky 1986: 175f—it follows that Japanese does not have local anaphors. Given this, the results of the Experiments reported above are just as expected. That is to say, the fact that the researchers have so far failed to identify what qualifies as a local anaphor in Japanese despite the concerted efforts by a substantial number of practitioners for nearly three decades, is not puzzling, after all.30

30 As H.-D. Ahn (p.c., 12/12/2009) suggests, one might pursue the possibility that the hypotheses in (0) are valid but that otagai, zibun-zisin and kare-zisin always occur in a structural position in which they have a covert antecedent in its local domain. Such a move would save (0) from refutation (and one might even claim that it would also make it possible to maintain the thesis that Japanese shares a "universal property" of having [+A] elements), but it would result in the elimination of the *Schema-based predictions. Hence that would be like a content-reducing (or degenerating) problemshift in the sense of Lakatos 1970/1978. Furthermore, if we accepted the view that what formally underlies a local anaphor is something like an "active functional category," it would be puzzling that there

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 25/49 5. On the use of 'invalidated' hypotheses As pointed out explicitly in Duhem 1909/1982, predictions are derived or deduced from two or more hypotheses combined. Some hypotheses are, but others are not, (intended to be) testable. In the terms of Lakatos 1970/1978, the former are auxiliary hypotheses and the latter constitute the hard core of a research program. If a testable hypothesis is shown to be invalid in what is considered to be a "simple case," it seems reasonable not to use it for what we consider to be a "more complex case."31 From the perspective of research concerned with the properties of the Computational System, lexical (hence language-specific) hypotheses such as (0), repeated here, would be significant only insofar as they contribute to a discovery of properties of the Computational System by interaction with universal hypotheses about the Computational System.

(0) Specifications in the mental Lexicon of speakers of Japanese: a. Otagai is marked [+A]. b. Zibun-zisin is marked [+A]. c. Kare-zisin is marked [+A].

The investigation of such interaction, however, would have little significance if the lexical hypotheses under discussion have been shown to be invalid, as in the case of (0). We have seen that the hypotheses in (0) are invalid by examining the predictions made by (i) each of the lexical hypotheses in (0), (ii) the universal hypothesis in (0) and (iii) the hypotheses that ensure (0).

(0) A [+A] category must have an antecedent in its local domain.

(0) NP1 is not in the local domain of NP3 in the representation corresponding to (0).

(0) NP1-ga [NP2-ga NP3-{o/ni} to] Verb 'NP1 Verb that NP2 Verb NP3'

The validity of what is intended by (0) is widely assumed although there may be disagreements concerning exactly how to express it. Similarly, to the extent that the notion "local domain" is to have any empirical content in relation to (0) in Japanese, it seems clear that (0) must be valid. If either (0) or (0) were invalid, the disconfirmation the *Schema-based predictions discussed in section 4 would not be of much significance. We in fact would not have been able to come up with *Schema-based predictions (that can be tested meaningfully) based on (0), to begin with. It would be of little use to test the empirical predictions based on hypotheses that have been shown to be invalid. If we had obtained confirmed schematic asymmetries in accordance with the predictions made under (0), combined with (0) and (0), we would have acquired new tools, so to speak, for our further probe into properties of the Computational System. But we did not. Now that the *Schema-based predictions made under (0) have been disconfirmed, it follows from the above reasoning that it is premature and perhaps ill- advised to use the hypotheses in (0) to make further predictions on the basis of their interaction with other hypotheses. In the preceding discussion, we have focused on the Subject-Object-Verb order (SOV). Given that the Object-Subject-Verb order (OSV) is more "complex" than SOV, as most practitioners in does not seem to be any confirmed schematic asymmetry in support of the presence of an "active functional category" in Japanese. I am not aware of any empirical evidence in support of the existence of DPs in Japanese and of the EPP (feature) in Japanese that forms a confirmed schematic asymmetry in accordance with the method advocated here.

31 See Popper 1963: 965-966, for example, for general remarks of much relevance.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 26/49 generative grammar dealing with Japanese seem to assume, it thus follows that there would be little merit to considering predictions about OSV made under (0). The use of the hypotheses (0) in addressing issues pertaining to more "complex" cases of OSV—including so-called long-distance OSV or multiple OS constructions (i.e., so-called multiple scrambling constructions)—would be even more problematic. This is in the spirit of the heuristic "Maximize our chances of learning from errors." If a new *Schema-based prediction were made by assuming, contrary to fact that the hypotheses in (0) were valid, it would not clear what significance we could assign to the result of an experiment on the new "predicted" schematic asymmetry. If the *Schema-based prediction got disconfirmed, we could not attribute it to the newly- introduced hypothesis/ses because we independently know that the *Schema-based predictions under (0) have been disconfirmed; hence, the disconfirmation of the *Schema-based prediction would not tell us anything about the newly-introduced hypothesis/ses. If we obtained a confirmed schematic asymmetry, we would not be justified to consider the newly-introduced hypothesis/ses and the "original hypothesis" to be both valid because we already know that the latter is not.32 Being faced with the disconfirmation of the *Schema–based predictions under (0), one might make recourse to the notion that science progresses in the ocean of anomalies anyway (see Lakatos 1970: 48-52 and Feyerabend 1975: chapter 5, for example), and maintain that the methodological proposal made in the preceding discussion goes against what is practiced in mature sciences; cf. Boeckx 2006 for remarks that one might regard as endorsing such a view. I should like to note that anomalies in mature sciences, such as physics, are solid and in fact very precise observations that resist an account within a given research program, e.g., the orbit of Uranus (before the discovery of Neptune) and the anomalous precession of Mercury's orbit, to mention two of the most celebrated instances of anomalies within the Newtonian research program. What we have discussed above is the failure to obtain confirmed schematic asymmetry, which I maintain, on the basis of the reasoning advanced in section 2, is a "smallest unit of facts" for research concerned with the properties of the Computational System. The failure to obtain a confirmed schematic asymmetry in research concerned with the properties of the Computational System should not be likened to an anomaly in mature sciences. After all, we still seem to be at a stage where we are trying to obtain establish "smallest units of facts" for research concerned with the properties of the Computational System of the language faculty once we decide to approach the Computational System by the hypothetico-deductive method. We will face an anomaly only if we have obtained a sufficiently large number of confirmed schematic asymmetries. An anomaly arises when some confirmed schematic asymmetries appear to resist a coherent account with respect to the rest of the confirmed schematic asymmetries. Given that the properties of the Computational System are hypothesized to be universal, with the possible exception of the so-called head parameter, it follows that hypotheses about the Computational System should be testable with respect to any language unless there is a language-specific reason otherwise. How we can actually attain the testability of hypotheses about the Computational System in regard to a given language deserves serious discussion. The fundamental challenge is that, while hypotheses about the Computational System are necessarily quite abstract as it is meant to apply to any language, actual experiments to test the validity of those hypotheses must be quite concrete because they are concerned with specific predictions in a given language. A minimally satisfactory articulation and discussion of the relevant issues would have to include how, in general, predictions are to be made so as to maximize our chances of learning from the failure of our predictions to be borne out. We would need to articulate the model of prediction making, so to speak. That, however, is one of the many topics that cannot be dealt with in this paper due to space considerations. Be that as it may, it is clear from the

32 See Poincaré 1952: chap. 9, 151-152 for relevant remarks. The general point just made seems to be rather poorly understood in the field (at least in works dealing with Japanese syntax), judging from the continued use of various types of hypotheses that have been shown not to be backed up by a confirmed schematic asymmetry, not only in regard to "local anaphors" but also in regard to variable binding, quantifier scope, "floating quantifiers," and other empirical "domains." The illustration of this, however, will have to be made on a separate occasion due to space considerations.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 27/49 disconfirmation of the *Schema-based predictions illustrated above that it would be ill-advised to use otagai, zibun-zisin, or kare-zisin in an experiment in which we test the validity of hypotheses about the Computational System because we know that their use would only make our experiments unreliable.33

6. Confirmed schematic asymmetries in Japanese 6.1. So-NPs vs. a-NPs Recall that the mere presence of some contrast among some Examples (for some speakers) as predicted by the hypotheses in question does not constitute sufficient ground for a confirmed schematic asymmetry. Given the fundamental asymmetry between a *Schema-based prediction and an okSchema- based prediction, what needs to be demonstrated is that the former survives a rigorous disconfirmation and at the same time the latter gets confirmed. Thus, even if there were some speakers who detected a significant contrast among some relevant Examples in question, that in and by itself would not mean much; it must be demonstrated that the informants judge any *Example conforming to the *Schema completely unacceptable while at the same time judging okExamples of the corresponding okSchema significantly more acceptable. While one may not raise issue with the disconfirmation of the *Schema-based prediction in relation to (0), one might nonetheless wonder if that might, in general, be too high a standard for actual research because we perhaps cannot fully control various non-grammatical factors in our experiments. In response to such a possible objection, I should like to mention before closing the paper that it in fact seems possible to obtain a confirmed schematic asymmetry in accordance with the above-mentioned standard. Japanese has three non-interrogative demonstrative prefixes ko-, so-, a-, as exemplified in (0).34

(0) a. ko-no hito 'this person' b. so-no hito 'that person' c. a-no hito 'that person'

(0) a. ko-ko 'this place' b. so-ko 'that place'

33 As noted in section 4.4, it is possible that we have not yet been able to eliminate noise and that the hypotheses in (0) can be maintained with some modification. But that has to be demonstrated before we start using otagai, zibun- zisin, or kare-zisin in experiments that are clearly "more complex" than the cases involving SOV.

34 The standard literature on demonstratives in Japanese includes Matsushita 1930/1978, Sakuma 1936/1951/1983, Kuno 1973 and Kuroda 1979. Kinsui & Takubo 1992 provides an extensive literature survey; cf. also Takubo 1984, Takubo & Kinsui 1996, 1997 and Hoji et al. 2003. A more exhaustive paradigm of Japanese demonstratives is given in (i). (i) Ko-So-A-ko-re 'this thing' ko-tira 'this way' ko-tti 'this way' ko-ko 'this place' ko-itu 'this guy'so-re so-tira so-tti so-ko so-itua-re a-tira a-tti a-soko a-itu

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 28/49 c. a-soko35 'that place'

Let us refer to NPs such as (0) as ko/so/a-NPs. While both so-NPs and a-NPs can be anaphorically related with another NP within the sentence, as indicated in (0), it has been noted in works such as Hoji 1991 that so-NPs can but a-NPs cannot be anaphorically related to a non-singular-denoting NP.

(0) Toyota-ga {so-ko/a-soko}-no bureeki mondai-o kakusite ita. 'Toyota kept its break problems from the public.'

This is illustrated in (0).

(0) a. Kanarino kazu-no seizika-ga {so/*a}-itu/-no hisyo-o hihansita. considerable number-GEN politician-NOM that-guy-GEN secretary-O criticized 'Lit: A considerable number of politicians criticized that guy's secretary.'

b. Ni-wari-izyoo-no tihoozititai-ga {so/*a}-ko-no keisatu-o hihansita. 2-1/10-more-GEN local:government-NOM that-place-GEN police-ACC criticized 'Lit: 20% or more local governments criticized its police.' 'Each of 20% or more local governments criticized its police.

So-ko and so-itu are singular-denoting, as observed in Hoji 1995; see (0).

(0) a. Toyota1-ga Nissan2-ni [FBI-ga [{so-no ni-sya/*so-ko}]1+2-o tokubetutyoosa-no Toyota-NOM Nissan-DAT FBI-NOM {that-gen 2-compay/that-place}-ACC special:investigation- GEN taisyoo-ni siteiru to] tugeta target-DAT is:making that told

'Toyota1 told Nissan2 that the FBI was targeting the {two companies/it}1+2 for a special investigation.'

b. Aru ninensei1-ga aru itinensei2-ni [sensei-ga [so-itu-*(tati)]1+2-o hometeiru to] certain sophomore-NOM certain freshman-DAT professor-NOM that-guy-tati-ACC was:praising that tutaeta rasii. told it:seems

'It seems that some sophomore1 told some freshman2 that the professor was praising {them/him}1+2.'

Reasonable conclusions to draw on the basis of observations like (0) and (0) are as in (0) and (0).

(0) So-ko and so-itu are singular-denoting.

(0) So-ko and (to a lesser extent) soi-tu can be related with a non-singular-denoting phrase, unlike aso-ko and (to a lesser extent) ai-tu.

Leaving aside exactly how to express them in theoretical terms, (0) and (0) give rise to definitive testable

35 The so in a-soko is unrelated to the demonstrative prefix so-.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 29/49 predictions, i.e., predicted schematic asymmetries. Although (0) and (0) have never been challenged in the literature, it has not yet been demonstrated that they are indeed valid. In the following two subsections, I shall report results of experiments that confirm the validity of (0) and (0). The results also show that it is possible to obtain a confirmed schematic asymmetry in accordance with the methodological proposal suggested above.

6.2. Experiment (I): Split antecedence Based on (0), and the assumption that so-no 2-sya 'those/the two companies' and so-itu-tati 'them, that/the guy and others' are plural-denoting, we obtain predicted schematic asymmetries as indicated in (0) and (0).

(0) a. okSchema:

NP1-ga NP2-ni [NP-ga [α {so-itu-tati/so-no 2-sya}]-o V-ta] to V. (α=NP1+NP2; i.e., under the interpretation that the referent of α includes that of NP1 and that of NP2.)

b. *Schema:

*NP1-ga NP2-ni [NP-ga [α {so-itu/so-ko}]-o V-ta] to V. (α=NP1+NP2; under the interpretation that the referent of α includes that of NP1 and that of NP2.)

c. okSchema:

NP1-ga NP2-ni [NP-ga [α {so-itu/so-ko}]-o V-ta] to V. (α=NP2; i.e., under the interpretation that the referent of α is that of just one singular-denoting NP.)

(0) a. okSchema:

NP1-ga NP2-ni [α {so-itu-tati/so-no 2-sya}]-no N-ni tuite/kansite V. (α=NP1+NP2; i.e., under the interpretation that the referent of α includes that of NP1 and that of NP2.)

b. *Schema:

*NP1-ga NP-2-ni [α {so-itu/so-ko}]-no N-ni tuite/kansite V. (α=NP1+NP2; i.e., under the interpretation that the referent of α includes that of NP1 and that of NP2.)

c. okSchema:

NP1-ga NP-2-ni [α {so-itu/so-ko}]-no N-ni tuite/kansite V. (α=NP1; i.e., under the interpretation that the referent of α is that of just one singular-denoting NP.)

On the basis of the Schemata in (0), we can construct the Examples in (0) and (0). The examples in the experiment do not contain indices and the instructions given are like "under the interpretation that so- itu-tati 'those guys' refers to the sophomore and the freshman under discussion" (for (0a), for example), "so-itu 'that guy' refers to the sophomore the freshman under discussion" (for (0b), for example), so-itu refers to the freshman under discussion" (for (0c), for example).

(0) a. (Cf. (0b).) okExample

Aru ninensei1-ga aru itinensei2-ni [sensei-ga so-itu-tati1+2-o hometeiru to] certain sophomore-NOM certain freshman-DAT professor-NOM that-guy-tati-ACC was:praising

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 30/49 that tutaeta rasii. told it:seems

'It seems that some sophomore1 told some freshman2 that the professor was praising them1+2.'

b. (Cf. (0b)) *Example

Aru ninensei1-ga aru itinensei2-ni [sensei-ga so-itu1+2-o hometeiru to] certain sophomore-NOM certain freshman-DAT professor-NOM that-guy-ACC was:praising that tutaeta rasii. told it:seems 'It seems that some sophomore told some freshman that the professor was praising them.'

c. okExample Aru ninensei-ga aru itinensei-ni [sensei-ga so-itu-o hometeiru to] certain sophomore-NOM certain freshman-DAT professor-NOM that-guy-ACC was:praising that tutaeta rasii. told it:seems 'It seems that some sophomore told some freshman that the professor was praising him.'

(0) a. okExample

Aru itinensei1-ga aru ninensei2-ni so-itu-tati1+2-no kyoodoo kenkyuu-nituite soodansita certain freshman-NOM certain sophomore -DAT that-guy-tati-GEM joint research-about consulted

'Some freshman1 consulted with some sophomore2 about their1+2 joint research.'

b. *Example

Aru itinensei1-ga aru ninensei2-ni so-itu1+2-no kyoodoo kenkyuu-nituite soodansita certain freshman-NOM certain sophomore-DAT that-guy-tati-GEM joint research-about consulted

'Some freshman1 consulted with some sophomore2 about their1+2 joint research.'

c. okExample Aru itinensei-ga aru ninensei-ni so-itu-no kenkyuu-nituite soodansita certain freshman-NOM certain sophomore-DAT that-guy-GEM research-about consulted 'Some freshman consulted with some sophomore about the guy's research.'

On the basis of the Schemata in (0), we can construct the Examples in (0) and (0).

(0) a. (Cf. (0a).) okExample

Toyota1-ga Nissan2-ni [FBI-ga so-no ni-sya1+2-o tokubetutyoosa-no taisyoo-ni Toyota-NOM Nissan-DAT FBI-NOM that-gen 2-compay-ACC special:investigation-GEN target-DAT siteiru to] tugeta is:making that told

'Toyota1 told Nissan2 that the FBI was targeting the two companies1+2 for a special investigation.'

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 31/49 b. (Cf. (0a).) *Example

Toyota1-ga Nissan2-ni [FBI-ga so-ko1+2-o tokubetutyoosa-no taisyoo-ni Toyota-NOM Nissan-DAT FBI-NOM that-place -ACC special:investigation-GEN target-DAT siteiru to] tugeta is:making that told

'Toyota1 told Nissan2 that the FBI was targeting it1+2 for a special investigation.'

c. okExample Toyota-ga Nissan-ni [FBI-ga so-ko-o tokubetutyoosa-no taisyoo-ni Toyota-NOM Nissan-DAT FBI-NOM that-place -ACC special:investigation-GEN target-DAT siteiru to] tugeta is:making that told 'Toyota told Nissan that the FBI was targeting it for a special investigation.'

(0) a. okExample

Toyota1-ga Nissan2-ni so-no ni-sya1+2-no kaikosyasuu-nituite renrakusita Toyota-NOM Nissan-DAT that-gen 2-compay-ACC number:of:layoff-about contacted

'Toyota1 contacted Nissan2 about the number of layoffs that the two companies1+2 had made.'

b. *Example

Toyota1-ga Nissan2-ni soko1+2-no kaikosyasuu-nituite renrakusita Toyota-NOM Nissan-DAT that-place-ACC number:of:layoff-about contacted

'Toyota1 contacted Nissan2 about the number of layoffs that it+2 had made.'

c. okExample Toyota-ga Nissan-ni soko-no kaikosyasuu-nituite renrakusita Toyota-NOM Nissan-DAT that-place-ACC number:of:layoff-about contacted 'Toyota contacted Nissan about the number of layoffs that it had made.'

A summary of the result of the experiment is given in (0).

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 32/49 (0) Result of the experiment as of March 2, 2010: Hypothesis: So-itu and so-ko are necessarily singular-denoting while soitura, so-no #-sya are not. α is in the embedded clause while NP-ga and NP-ni are in the matrix clause.

ok NP1-ga NP2-ni [NP-ga [α {so-itu-tati/so-no 2-sya}]-o V- Schema 1 A 36 values 67 ta] to V. (α=NP1+NP2) (Under the interpretation that the referent of α includes that of NP1 and that of NP2.)

Schema group 1 *NP1-ga NP2-ni [NP-ga [α {so-itu/so-ko}]-o V-ta] to V. Schema 1 B 36 values 2 (α=NP1+NP2) (Under the interpretation that the referent of α includes that of NP1 and that of NP2.)

ok NP1-ga NP2-ni [NP-ga [α{so-itu/so-ko}]-o V-ta] to V. Schema 1 C 36 values 77 (α=NP2) (Under the interpretation that the referent of α is that of just one singular-denoting NP. ) There is no clause boundary between α and NP-ga/NP-ni.

ok NP1-ga NP2-ni [α{so-itu-tati/so-no 2-sya}]-no N-ni tuite/kansite V. (α=NP1+NP2) (Under the interpretation Schema 2 A 36 values 78 that the referent of α includes that of NP1 and that of NP2.) Schema group 2 *NP1-ga NP-2-ni [α{so-itu/so-ko}]-no N-ni tuite/kansite Schema 2 B 36 values 3 V. (α=NP1+NP2) (Under the interpretation that the referent of α includes that of NP1 and that of NP2.)

ok NP1-ga NP-2-ni [α{so-itu/so-ko}]-no N-ni tuite/kansite Schema 2 C 36 values 73 V. (α=NP1) (Under the interpretation that the referent of α is that of just one singular-denoting NP.) 18 participants 309 answers

"Schema group 1" is for (0) and "Schema group 2" is for (0). "Schema 1 A" covers the okExamples in (0a) and (0a), "Schema 1 B" the *Examples in (0b) and (0b), and "Schema 1 C" the okExamples in (0c) and (0c). Likewise, "Schema 2 A" covers the okExamples in (0a) and (0a), "Schema 1 B" the *Examples in (0b) and (0b), and "Schema 1 C" the okExamples in (0c) and (0c). "309 answers" means that there have been 309 occurrences of a reported judgment. As noted, some informants have judged the same example more than once; but in such cases the values in (0) are based on the average score on a given example by the same informant. The values of "Schema 1 B" and "Schema 2 B" should be close to "0" according to the predicted schematic asymmetries in (0) and (0) and those of "Schema 1 A," "Schema 2 A," "Schema 1 C" and "Schema 2 C" should be significantly higher than that. The informant judgments as indicated in (0) are thus very much in harmony with the predicted schematic asymmetries in (0) and (0), making us hopeful that confirmed schematic asymmetries can be established here.

6.3. Experiment (II): So-NPs vs. A-NPs We have seen strong evidence in support of (0), repeated here.

(0) So-ko and so-itu are singular-denoting.

Given (0), one might find it puzzling that so-ko and (to a lesser extent) soi-tu can be related with a non-

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 33/49 singular-denoting phrase, as illustrated in (0a); cf. (0), repeated here.

(0) So-ko and (to a lesser extent) soi-tu can be related with a non-singular-denoting phrase, unlike aso-ko and (to a lesser extent) ai-tu.

By considering the anaphoric relation indicated in (0a) as an instance of bound variable anaphora and by representing it as BVA(A, B) (BVA for "bound variable anaphora"), with its characterization as given in (0), we can state predicted schematic asymmetries under (0), as in (0) and (0).36

(0) An initial characterization of BVA(A, B): We intend BVA(A, B) to have the following properties: a. A is not singular-denoting; i.e., either there are two or more individuals or entities that are 'expressed' by A or there is no individual or entity expressed by A. b. B does not refer to a particular individual or entity. c. B is singular-denoting. d. B is understood to 'express the same thing' as A is understood to express; i.e., the value of B co- varies with that of A.

(0) a. okSchema: NP1-ga so-NP-no N-o V (Under BVA(NP1, so-NP))

b. *Schema: NP1-ga a-NP-no N-o V (Under BVA(NP1, a-NP))

c. okSchema: NP1-ga a-NP-no N-o V (Under the interpretation that a-NP is intended to refer to something/someone specific)

(0) a. okSchema: NP1-ga [ ... so-NP ... N]-ni V (Under BVA(NP1, so-NP))

b. *Schema: NP1-ga [ ... a-NP ... N]-ni V (Under BVA(NP1, a-NP))

c. okSchema: NP1-ga [ ... a-NP ... N]-ni V (Under the interpretation that a-NP is intended to refer to something/someone specific)

On the basis of the Schemata in (0) and (0), we can construct the Examples in (0)/(0) and (0)/(0), respectively. The informant have been asked to judge (0a), for example, under the interpretation that having criticized the local police applies to 20% or more local governments. The instructions for (0c), for example, includes "where a-soko refers to some place specific."

36 As suggested above, there must be a language-specific lexical hypothesis that yields the effects of (0). I do not make an attempt to articulate the hypothesis here. Ueyama 1998 puts forth a formal proposal that has precisely the effects of (0).

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 34/49 (0) a. okExample Ni-wari-izyoo-no tihoozititai-ga so-ko-no keisatu-o hihansita. 2-1/10-more-GEN local:government-NOM that-place-GEN police-ACC criticized 'Lit: 20% or more local governments criticized its police.' 'Each of 20% or more local governments criticized its police.

b. *Example Ni-wari-izyoo-no tihoozititai-ga a-soko-no keisatu-o hihansita. 2-1/10-more-GEN local:government-NOM that-place-GEN police-ACC criticized 'Lit: 20% or more local governments criticized its police.' 'Each of 20% or more local governments criticized its police.

c. okExample Ni-wari-izyoo-no tihoozititai-ga a-soko-no keisatu-o hihansita. 2-1/10-more-GEN local:government-NOM that-place-GEN police-ACC criticized '20% or more local governments criticized its (some specific place's) police.

(0) a. okExample Ni-wari-izyoo-no tihoozititai-ga [ so-ko-o hihansita zassikisya]-ni 2-1/10-more-GEN local:government-NOM that-place-ACC criticized magazine:reporter-DAT renraku-o totta. contact-acc made 'Lit: 20% or more local governments contacted the magazine reporter(s) who had criticized it.' 'Each of 20% or more local governments contacted the magazine reporter(s) who had criticized it.'

b. *Example Ni-wari-izyoo-no tihoozititai-ga [ a-soko-o hihansita zassikisya]-ni 2-1/10-more-GEN local:government-NOM that-place-ACC criticized magazine:reporter-DAT renraku-o totta. contact-acc made 'Lit: 20% or more local governments contacted the magazine reporter(s) who had criticized it.' 'Each of 20% or more local governments contacted the magazine reporter(s) who had criticized it.'

c. okExample Ni-wari-izyoo-no tihoozititai-ga [ a-soko-o hihansita zassikisya]-ni 2-1/10-more-GEN local:government-NOM that-place-ACC criticized magazine:reporter-DAT renraku-o totta. contact-acc made '20% or more local governments contacted the magazine reporter(s) who had criticized it (i.e., some specific institution).'

(0) a. (Cf. (0a).) okExample Kanarino kazu-no seizika-ga so-itu-no hisyo-o hihansita. considerable number-GEN politician-NOM that-guy-GEN secretary-O criticized

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 35/49 'Lit: A considerable number of politicians criticized that guy's secretary.'

b. (Cf. (0a).) *Example Kanarino kazu-no seizika-ga a-itu-no hisyo-o hihansita. considerable number-GEN politician-NOM that-guy-GEN secretary-O criticized 'Lit: A considerable number of politicians criticized that guy's secretary.'

c. okExample Kanarino kazu-no seizika-ga a-itu-no hisyo-o hihansita. considerable number-GEN politician-NOM that-guy-GEN secretary-O criticized 'A considerable number of politicians criticized that guy's (i.e., some specific person's) secretary.'

(0) a. okExample Kanarino kazu-no seizika-ga so-itu-o hihansita zassikisya]-ni considerable number-GEN politician-NOM that-guy-ACC criticized magazine:reporter-DAT renraku-o tottta. contact-ACC made 'Lit: A considerable number of politicians contacted the magazine reporter(s) who had criticized that guy.'

b. *Example Kanarino kazu-no seizika-ga a-itu-o hihansita zassikisya]-ni considerable number-GEN politician-NOM that-guy-ACC criticized magazine:reporter-DAT renraku-o tottta. contact-ACC made 'Lit: A considerable number of politicians contacted the magazine reporter(s) who had criticized that guy.'

c. okExample Kanarino kazu-no seizika-ga a-itu-o hihansita zassikisya]-ni considerable number-GEN politician-NOM that-guy-ACC criticized magazine:reporter-DAT renraku-o tottta. contact-ACC made 'A considerable number of politicians contacted the magazine reporter(s) who had criticized that guy (i.e., some specific guy).'

The result of the experiment (as of 2/28/2010) is summarized in (0).

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 36/49 (0) The result of the experiment as of March 11, 2010: Hypothesis: So-NPs can be B of BVA(A, B) but a-NPs cannot. So-NP or a-NP is in the position of "__" in the object "__ no N." okNP1-ga so-NP-no N-o V Schema 1 A 47 values 76 (Under BVA(NP1, so-NP)) *NP1-ga a-NP-no N-o V Schema group 1 Schema 1 B 48 values 5 (Under BVA(NP1, a-NP)) okNP1-ga a-NP-no N-o V Schema 1 C 48 values 96 (Under the interpretation that a-NP is intended to refer to something/someone specific) So-NP or a-NP is inside a relative clause contained in the object. okNP1-ga [ ... so-NP ... N]-ni V Schema 2 A 48 values 74 (Under BVA(NP1, so-NP)) *NP1-ga [ ... a-NP ... N]-ni V Schema group 2 Schema 2 B 46 values 3 (Under BVA(NP1, a-NP)) okNP1-ga [ ... a-NP ... N]-ni V Schema 2 C 47 values 94 (Under the interpretation that a-NP is intended to refer to something/someone specific) 25 participants 464 answers

"Schema group 1" is for (0) and "Schema group 2" is for (0). "Schema 1 A" covers the okExamples in (0a) and (0a), "Schema 1 B" the *Examples in (0b) and (0b), and "Schema 1 C" the okExamples in (0c) and (0c). Likewise, "Schema 2 A" covers the okExamples in (0a) and (0a), "Schema 1 B" the *Examples in (0b) and (0b), and "Schema 1 C" the okExamples in (0c) and (0c). "464 answers" means that there have been 464 occurrences of a reported judgment. As noted, some informants have judged the same example more than once; but in such cases the values in (0) are based on the average score on a given example by the same informant. As in the case of (0), the informant judgments summarized in (0) are thus very much in harmony with the predicted schematic asymmetries indicated in (0) and (0).

6.4. A rigorous standard for confirmed schematic asymmetries and informants' resourcefulness The experimental results reported above indicate that it is possible to obtain about 5 as the mean value on a *Schema, suggesting that we might want to aspire to the standard according to which a confirmed schematic asymmetry can be regarded as being established only if the mean value on a *Schema is about 5, with a sufficiently large number of informants. Given the claim made by a *Schema- based prediction, repeated below, one might object that the mean value on a *Schema that is higher than 0 should be regarded as disconfirmation of the *Schema-based prediction.

(0) A *Schema-based prediction: The informant judgment on the presented sentence  under interpretation a, b) is always "completely unacceptable" for any *Example conforming to a *Schema.

The objection is valid if the informants make no errors in reporting their judgments (such as by hitting the wrong button inadvertently) and if all the informants perfectly understand the instructions for a given experiment and the intended interpretation provided. It is perhaps possible to minimize such "errors" by

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 37/49 conducting preliminary experiments and screen out the informants whose judgments clearly seem to deviate from the large majority of the informants and not in any systematic way. But it is not clear that we can eliminate such errors completely; and hence, it seems to me to be reasonable to set our goal around 5, rather than 0, for the value on a *Schema. There is a significant difference between linguistically-naïve informants and linguistically-"trained" informants in regard to their judgments on okExamples. The former tend to judge the okExamples less acceptable than the latter. There is, however, not a significant difference in regard to their judgments on the *Examples in the experiments reported in this section. For example, of the 20 informants in the experiment reported in (0), 9 are linguistically-naïve informants and 11 are not; the mean values of the judgments reported by the former group on the (a), (b) and (c) Examples in (0)-(0) are 59, 13, 97, respectively while those reported by the latter group are 97, 0, 98, respectively. This is related to informants' resourcefulness and is as expected. The former group's mean values on (a) Examples also get affected significantly depending upon the choice of the "bindee" (e.g., so-ko vs. soi-tu as the "bindee") while such is not the case with the latter group; the former groups' mean score on the *Examples conforming to the *Schemata in (0a) and (0a) are 72 with so-ko and 47 with so-itu while the latter groups' mean scores are 99 with so-ko and 94 with so-itu. This is also as expected in light of the informant resourcefulness; see Ueyama 1998 and Hoji et al. 1999. Some informants judge *Examples completely unacceptable but also judge the corresponding okExamples fairly to completely unacceptable. It is therefore not clear that those informants' unacceptability judgments on the *Examples are due to the condition that we are concerned with. Clear unacceptability judgments on *Examples is therefore more revealing if they are accompanied by clear acceptability judgments on the corresponding okExamples. In this sense, the judgments by linguistically- trained informants can be more revealing than linguistically-naïve informants, contrary to the general claim fairly widely accepted in the field; see Schütze 1996.

6.5. Experiment (III): The structural condition for variable binding The condition on the availability of BVA addressed above is lexical in nature and hence it is language-specific, at least in terms of how the effects of the condition can be detected. What about the structural condition on BVA? The results of some experiments so far seem to indicate that it is possible to obtain a confirmed schematic asymmetry not only in regard to the lexical conditions addressed above but also in regard to the structural condition for BVA. For space reason, I will report only the results of just one of many such experiment and do not provide the actual examples used in the experiment. Given the structural condition of c-command on the availability of BVA and the hypothesis (or the hypotheses that have the effect) that the S(ubject) asymmetrically c-commands the O(bject) in the SOV order in Japanese, we have the predicted schematic asymmetry in (0).

(0) (Where the V is not an "ergative verb.") a. okSchema NP-ga [ ... so-NP ... ]-ni/o V BVA(NP, so-NP) b. *Schema [ ... so-NP ...]-ga NP-ni/o V BVA(NP, so-NP) c. okSchema [ ... so-NP ...]-ga NP-ni/o V (With so-NP "referring to" an individual/object that has been mentioned in the preceding discourse)

(0) is the typical BVA paradigm in Japanese in the SOV order, widely accepted as being valid, with the 'dependent term' being a so-NP. Given in (0) is a summary of the results (as of February 28, 2010) of an experiment with 55% izyoo-no NP '55% or more N' as the intended "binder" and so-ko 'it' as the intended

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 38/49 "bindee," i.e., with the intended BVA(55% izyoo-no NP, so-ko).

(0) Result of the experiment as of March 2, 2010: so-ko-no N is a matrix argument. ok55% izyoo-no NP-ga so-ko-no N-o/-ni V (under BVA(55% Schema 1 A 30 values 99 izyoo-no NP, so-ko)) Schema group 1 *so-ko-no N-ga [ N-ga 55% izyoo-no NP-ni V sita to] V Schema 1 B 30 values 3 (under BVA(55% izyoo-no NP, so-ko)) okso-ko-no N-ga [ N-ga 55% izyoo-no NP-ni V sita to] V Schema 1 C 29 values 94 (with so-ko referring to a specific institution) so-ko is an argument in a relative clause. ok55% izyoo-no NP-ga [ ... so-ko ...]-o/-ni V (under Schema 2 A 30 values 99 BVA(55% izyoo-no NP, so-ko)) Schema group 2 *[ ... so-ko ...]-ga 55% izyoo-no NP-o/-ni V (under Schema 2 B 30 values 10 BVA(55% izyoo-no NP, so-ko)) ok[ ... so-ko ...]-ga 55% izyoo-no NP-o/-ni V (with so-ko Schema 2 C 30 values 94 referring to a specific institution) The binder and the bindee are separated by a clause boundary. ok55% izyoo-no NP-ga [N-ga so-ko-no N-ni V sita to] V Schema 3 A 30 values 99 (under BVA(55% izyoo-no NP, so-ko)) Schema group 3 *so-ko-no N-ga 55% izyoo-no NP-o/-ni V (under BVA(55% Schema 3 B 30 values 8 izyoo-no NP, so-ko)) okso-ko-no N-ga 55% izyoo-no NP-o/-ni V (with so-ko Schema 3 C 30 values 97 referring to a specific institution) 16 participants 600 answers

Experiments with different intended "binders" such as kanari-no kazu-no N 'a considerable number of N' and NP-sae 'even NP' have gotten comparable results. I report here only the mean values on the Schemata in those experiments.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 39/49 (0) Result of the experiment with kanari-no kazu-no N 'a considerable number of N' as the intended "binder" as of March 2, 2010: so-ko-no N is a matrix argument. okkanari-no kazu-no N-ga so-ko-no N-o/-ni V (under Schema 1 A 43 values 97 BVA(kanari-no kazu-no N, so-ko)) Schema group 1 *so-ko-no N-ga [ N-ga kanari-no kazu-no N-ni V sita to] V Schema 1 B 43 values 11 (under BVA(kanari-no kazu-no N, so-ko)) okso-ko-no N-ga [ N-ga kanari-no kazu-no N-ni V sita to] V Schema 1 C 44 values 92 (with so-ko referring to a specific institution) so-ko is an argument in a relative clause. okkanari-no kazu-no N-ga [ ... so-ko ...]-o/-ni V (under Schema 2 A 44 values 96 BVA(kanari-no kazu-no N, so-ko)) Schema group 2 *[ ... so-ko ...]-ga kanari-no kazu-no N-o/-ni V (under Schema 2 B 43 values 28 BVA(kanari-no kazu-no N, so-ko)) ok[ ... so-ko ...]-ga kanari-no kazu-no N-o/-ni V (with so-ko Schema 2 C 44 values 92 referring to a specific institution) The binder and the bindee are separated by a clause boundary. okkanari-no kazu-no N-ga [N-ga so-ko-no N-ni V sita to] V Schema 3 A 44 values 94 (under BVA(kanari-no kazu-no N, so-ko)) Schema group 3 *so-ko-no N-ga kanari-no kazu-no N-o/-ni V (under Schema 3 B 44 values 5 BVA(kanari-no kazu-no N, so-ko)) okso-ko-no N-ga kanari-no kazu-no N-o/-ni V (with so-ko Schema 3 C 44 values 90 referring to a specific institution) 22 participants 858 answers

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 40/49 (0) Result of the experiment with NP-sae 'even NP' as the intended "binder" as of March 2, 2010: so-ko-no N is a matrix argument. okNP-sae-ga so-ko-no N-o/-ni V (under BVA(NP-sae, so- Schema 1 A 30 values 99 ko)) Schema group 1 *so-ko-no N-ga [ N-ga NP-sae-ni V sita to] V (under Schema 1 B 30 values 6 BVA(NP-sae, so-ko)) okso-ko-no N-ga [ N-ga NP-sae-ni V sita to] V (with so-ko Schema 1 C 29 values 95 referring to a specific institution) so-ko is an argument in a relative clause. okNP-sae-ga [ ... so-ko ...]-o/-ni V (under BVA(NP-sae, so- Schema 2 A 30 values 99 ko)) Schema group 2 *[ ... so-ko ...]-ga NP-sae-o/-ni V (under BVA(NP-sae, so- Schema 2 B 30 values 7 ko)) ok[ ... so-ko ...]-ga NP-sae-o/-ni V (with so-ko referring to a Schema 2 C 30 values 89 specific institution) The binder and the bindee are separated by a clause boundary. okNP-sae-ga [N-ga so-ko-no N-ni V sita to] V (under Schema 3 A 30 values 99 BVA(NP-sae, so-ko)) Schema group 3 *so-ko-no N-ga NP-sae-o/-ni V (under BVA(NP-sae, so- Schema 3 B 30 values 3 ko)) okso-ko-no N-ga NP-sae-o/-ni V (with so-ko referring to a Schema 3 C 30 values 99 specific institution) 14 participants 486 answers

The results of these experiments, although the number of the informants is still relatively small, suggest that it is perhaps possible to obtain the mean value of about 5 on a *Schema in regard to structural (hence universal) hypotheses as well.37 The experiments reported or alluded to in (0)-(0) deal with SOV. Given that we can be hopeful in obtaining a confirmed schematic asymmetry in SOV, we are in a position to prepare and conduct experiments on more complex cases, such as those involving OSV, using the "binder-bindee" pairs that have given fairly robust and as-predicted results. The results of such experiments addressing various other issues—including reconstruction effects, 'resumption', etc.—also seem to confirm that it is possible to keep to the suggested standard for obtaining a confirmed schematic asymmetry. And hence I should like to maintain that we aspire to attain such a rigorous standard if we intend our research to be concerned with a discovery of the properties of the Computational System of the language faculty.

37 Space considerations prevents me from discussing why the values on the *Schemata in (0)-(0), especially those on Schema 2-B in (0), are not as low as predicted. What modification of the hypotheses have been pursued and what follow-up experiments have been conducted in accordance with the modification will have to be addressed in a separate work; see Hoji 2003 for some relevant discussion.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 41/49 7. Concluding remarks This paper started out by considering the basic scientific method, what Feynman calls "the key to science," as schematized in (0), repeated here.

(0) The general scientific method (i.e., the hypothetic-deductive method): Guess — Computing Consequences — Compare with Experiment

After pointing out what informant judgments can reasonably be considered as a reflection of properties of the Computational System, I adopted, with slight modification, the model of judgment making suggested in Ueyama 2009, and proceeded to examine predictions made under the lexical hypotheses in (0), repeated here.

(0) Specifications in the mental Lexicon of speakers of Japanese: a. Otagai is marked [+A]. b. Zibun-zisin is marked [+A]. c. Kare-zisin is marked [+A].

The hypotheses in (0), combined with the universal hypothesis in (0), along with the articulation of "local domains" in Japanese noted in section 3.2.1, make definite and testable predictions. As we have observed in section 4, the *Schema-based predictions are clearly disconfirmed.38 I have also noted that this result is in fact as expected. If what underlies a local anaphor is closely related to "active functional categories" in the sense of Fukui 1986, and if the mental Lexicon of speakers of Japanese lacks them altogether, as suggested in Fukui 1986, the absence of local anaphors in Japanese is exactly what we expect.39 As stated in section 2.5, obtaining the expected informant judgments is merely a start in terms of our rigorous disconfirmation attempt. That is to say, other interested researchers are strongly encouraged to conduct experiments themselves on the basis of the predicted schematic asymmetry, making various adjustments on the lexical items in the actual Examples conforming to the Schemata, doing the best they can to construct *Examples of the *Schema that are not completely unacceptable under the specified interpretation. The prediction is that, despite such efforts, the *Examples of the *Schema will still be completely unacceptable under the specified interpretation. It is in light of the above that we must appreciate the significance of the experimental results reported in section 4, which clearly invalidate the lexical hypotheses in (0). I would be quite surprised if the *Schema-based predictions under the lexical hypotheses in (0) did not get disconfirmed in other "instantiations" of the same experimental design. What is crucial, however, is not whether we might actually obtain results that would be in harmony with the predicted schematic asymmetries in some experiment; what is crucial is that we should never obtain results that would disconfirm the *Schema- based predictions, which would result in the failure of obtaining a confirmed schematic asymmetry. That is to say, even if we obtained in some experiments results in harmony with the predicted schematic asymmetries, that would not be nearly as significant as there being a result of an experiment that disconfirms the *Schema-based predictions under the lexical hypotheses, for the reasons adduced in the preceding discussion. Since the inception of generative grammar in the 1950s, there has been concern about in what way our hypotheses with regards to the Computational System of the language faculty can be put to empirical

38 I have thus concluded that the hypotheses in (0) should not be used in deducing further theoretical consequences or deriving further empirical predictions if we wish to discover properties of the Computational System and if we wish to maximize our chances of learning something about the Computational System from our failed predictions.

39 Narita to appear contains interesting discussion pertaining to the Fukui thesis and related issues.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 42/49 test. Chomsky's remarks on the relation between acceptability and grammaticality, which are frequently found in his writings until about the mid 1970s, can be understood to be a manifestation of his own concern. A number of publications appeared in the late 1960s and in the 1970s addressing the issue (e.g., many of the articles included in Cohen 1974, Cohen and Wirth 1975, and Eckman 1977); see Schütze 1996 for an extensive literature review. Concerns seem to have resurfaced in recent years; cf. the 2005 special issue of Lingua on "Data in Theoretical Linguistics" and the 2007 special issue of Theoretical Linguistics volume on "Data in generative grammar." It is perhaps safe to state that until the introduction of the model of Computational System in Chomsky 1993, it was not possible to embed the model of Computational System within a model of judgment making in any straightforward way that would enhance (if not ensure) the testability of hypotheses about the Computational System in a compelling manner. Without a model of judgment making that embeds the model of Computational System, it would not be clear what significance could be assigned to judgments by informants; cf. notes Error: Reference source not found and Error: Reference source not found. I should like to conjecture that the lack of a model of judgment making such as the one adopted here has resulted in the failure of the field at large to come to terms with the fundamental asymmetry between *Schema-based predictions and okSchema-based predictions, which in turn has often resulted in a confusing state of affairs when being faced with judgmental disputes on a crucial paradigm and also in heavy reliance on "statistically significant contrasts" in "experimental works." (Recall that, according to the methodology proposed above, a "statistically significant contrast" in and by itself by no means guarantees that the informant judgments under discussion may be a reflection of properties of the Computational System unless the *Schema-based prediction has survived a rigorous disconfirmation attempt.) This paper adopts a model of judgment making that embeds Chomsky's (1993) model of the Computational System, and we obtain, as a consequence, the fundamental asymmetry between a *Schema-based prediction and an okSchema-based prediction, which in turn provides us with a conceptual basis for adopting a specific means to identify what may be a likely reflection of properties of the Computational System. I have suggested that research concerned with the properties of the Computational System must take confirmed schematic asymmetries as the "smallest units of facts" to account for and to use as evidence for or against our hypotheses about the Computational System. There are a number of merits to working with schematic asymmetries and building our proposals on confirmed schematic asymmetries. For example, technical details tend to make empirical consequences of various proposals opaque unless conscious efforts are made to articulate how the proposals under discussion can be put to empirical test. Crucial reliance on confirmed schematic asymmetries help us transcend the technical details and differences of frameworks, making it much more straightforward for us to determine, in terms of their empirical merit, which of the alternative proposals are to be valued more highly than the others. It also helps us understand how we can deal with, and/or proceed in, cross- linguistic research in a meaningful and effective way. Once we decide to evaluate an alleged generalization in a given language based on whether it is indeed 'backed up' by a confirmed schematic asymmetry, we can prevent a great deal of confusion from arising in dealing with other languages (as well as other 'phenomena' in the same language). Most importantly, it makes us hopeful that we might be able to make generative grammar an empirical science, i.e., a field where the general scientific method in (0) can be rigorously applied. If the ultimate testability of our hypotheses lies in their being subject to disconfirmation, what makes our hypotheses testable is the *Schema-based predictions they give rise to, as discussed in section 2.4. One of the two most crucial research heuristics adopted here is "Maximize testability."40 Proceeding with our research based on confirmed schematic asymmetries is a way to ensure testability—and it is related to the point made in Nakaya 1958: 17 that the intrinsic limitation of science is that it is a discipline where we extract phenomena that are reproducible.41 Working with confirmed schematic asymmetries, although it

40 The other is "Maximize our chances of learning from errors." See section 5.

41 Nakaya adds that the reproducible phenomena thus extracted must be analyzable statistically. This is a

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 43/49 ensures reproducibility, does not necessarily lead to maximizing testability. An effective way to try to maximize testability is to work with phenomena and pursue hypotheses that would most likely lead us to a number of *Schema-based predictions. To the extent that the properties of the Computational System are hypothesized to be universal, the empirical consequences of hypotheses that are backed up by a confirmed schematic asymmetry in one language should be testable with respect to any other language unless there is a reason otherwise, as noted earlier. "Reasons otherwise" may abound, however, starting with lexical differences among languages, which tend to make it difficult, if not overwhelming, to replicate the same confirmed schematic asymmetry among different languages. For this reason, the designs of experiments testing the same hypotheses about the Computational System are likely to appear quite different on the surface, depending upon what language is dealt with in a particular laboratory. A successful research concerned with the properties of the Computational System must therefore commit itself to systematic correspondences between schematic phonetic strings in a particular language and abstract representations that are stated in universal terms. Here we may see the merit of pursuing Reinhart's (1983: chapter 1) conjecture, restated here somewhat, that a semantic interpretation, i.e., a semantic intuition, involving dependency of an expression A on another expression B must be based on a c-command relation at LF (between what corresponds to A and what corresponds to B at LF) if the interpretation/intuition is based on, hence is revealing about, the Computational System or properties of the language faculty that are directly related to the Computational System. By focusing on phenomena and pursuing hypotheses in line with Reinhart's conjecture, we may be able to make our experiments free from much of the noise that might arise from the properties of the lexical items used in the actual examples and hence from factors that may be outside the Computational System.42 For example, we can check the availability of BVA(every such and such, it) in examples like (0).

(0) a. Every such and such did such and such to its such and such. b Its such and such did such and such to every such and such. c. Which one of its such and such did every such and such did such and such to?

The results of an experiment of this sort conducted in Japanese are promising; but elaboration has to wait for another occasion.

8. Appendix: Local anaphors in English and "local anaphors" in Japanese In section 4, I have made reference to local anaphors in English, but no mention was made there in regard to informant judgments on examples with anaphors in English. Although we have not done commonly-held view and is stated in Weinberg 1992:7 as "[w]hat a successful scientific explanation would have to accomplish" is "the qualitative understanding of phenomena." (The emphasis is as in the original.) One may speculate that heavy reliance on "statistically significant contrasts" in disciplines outside physical sciences, including linguistics, stems from adopting this focus on "the qualitative understanding of phenomena" but at the same time realizing the immense difficulty in obtaining data of a categorical nature that can be deduced from a proposed system; cf. Lakatos 1970: 176, note 1 (reproduced in Lakatos 1978: 88-89, note 4) and Loftus 1996: 169 for some relevant remarks. See the remarks above in this section. One of the main theses of the research reported here is that it is possible to deepen our understanding of the properties of the Computational System on the basis of data that are of a categorical nature in the sense of confirmed schematic asymmetries as discussed above. The demonstration of its validity, one may suggest, will have far-reaching implications for research on various aspects of the mind, beyond the language faculty.

42 As suggested above, what would be crucially needed for experiments of this sort are hypotheses that have the effects of ensuring specific correspondences between the relevant elements in a schematic phonetic string and their structural relation at LF. Much of syntactic research in individual languages in fact seems to be concerned with precisely this, leaving aside whether it is directly concerned with the properties of the Computational System.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 44/49 relevant experiments using the general experimental design introduced above, one of the earlier experiments we conducted contains examples with reflexives in English. This Appendix introduces the results of that experiment that pertain to reflexives in English, along with the results of earlier experiments on zibun-zisin in Japanese. Consider the Japanese *Examples in (0) and (0) and their English analogues in (0) and (0).

(0) *Examples a. John-wa Mary-ga zibun-zisin-ni horeteiru to omoikonde ita. John-TOP Mary-NOM zibun-zisin-DAT is:in:love that believed 'John believed that Mary liked himself.'

b. (=(0a)) John-wa Mary-ga zibun-zisin-o suisensita to bakari omotteita. John-TOP Mary-NOM zibun-zisin-ACC recommended that only thought 'John firmly believed that Mary recommended himself.'

(0) *Example Chomsky-ga zibun-zisin-o suisensuru to omoikonde ita John-wa, Chomsky-NOM zibun-zisin-ACC will:recommend that believed John-TOP Chomsky-ga Bill-o suisensita to sitte gakuzen-to-sita. Chomsky-NOM Bill-ACC recommended that know be:shocked 'John, who had firmly believed that Chomsky would recommend himself, was shocked to death when he found out that Chomsky had recommended Bill instead.'

(0) a. *John thinks Mary loves himself. b. *John thought that Mary had recommended himself.

(0) *Mary, who had firmly believed that Chomsky would recommend herself, was shocked to death when she found out that Chomsky recommended Bill instead.

In the experiments whose results are reported below, the informants were asked, mostly on-line, how acceptable they found each example on the survey sheet/page (under a specified interpretation). Each example was judged on the 'scale' given in (0), and the five choices were computed as in (0), with "0" corresponding to "Bad" and "100" to "Good" although the "survey sheet/page" did not indicate the numeric values assigned to the five radio buttons.

(0) Bad < ===== > Good o o o o o

(0) 0, 25, 50, 75, 100

The informant judgments on these examples are as in (0).

(0)

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 45/49 Number of informants Mean Score Standard Deviation who accepted it43, 44 (0a) 37 out of 68 62.13 36.00 (0b) 25 out of 67 57.72 37.44 (0) 41 out of 69 64.86 38.15 (0a) 0 out of 19 5.26 10.19 (0b) 1 out of 19 14.47 21.94 (0) 1 out of 19 17.11 23.02

The judgments on (0) and (0) by the non-linguist informants and those by the linguist informants are provided in (0) and (0), respectively.

(0) Non-Linguists only: Number of informants Mean Score Standard Deviation who accepted it (0a) 23 out of 43 65.12 35.00 (0b) 14 out of 42 60.47 37.05 (0) 28 out of 43 69.19 36.10

(0) Linguists only: Number of informants Mean Score Standard Deviation who accepted it (0a) 14 out of 25 57 37.09 (0b) 11 out of 25 53 37.63 (0) 13 out of 26 57.69 40.29

The informant judgments on the *Examples in English were fairly robust and close to what is predicted. The informant judgments on the *Examples in Japanese, on the other hand, were far from what is predicted. The experiments were conducted a few years ago (mostly in 2005 and 2006) and their design was not as rigorous as that of the experiments discussed in section 4 above. Nonetheless, the results of those experiment on zibun-zisin are very much in harmony with what is reported above in section 4 from the more recent experiments.

9. References Boeckx, Cedric. 2006. Linguistic Minimalism: Origins, Concepts, Methods, and Aims. Oxford University Press, New York. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Mass.: MIT Press. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1986. Knowledge of Language. New York: Praeger. Chomsky, Noam. 1993. "A Minimalist Program for Linguistic Theory," in K. Hale & S. J. Keyser (eds.), The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger. Cambridge, MA: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. MA: The MIT Press.

43 The score of "75" or "100" is taken in the context of this particular exposition as "an example being judged to be acceptable." This applies to the two other charts below.

44 The total numbers of the informants are not the same for (0a), (0b), and (0c) because not every informant provided their judgments on every example in a given experiment.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 46/49 Cohen, David. 1974. ed. Explaining linguistic phenomena. Washington, D.C.: Hemisphere Publishing Corporation. Cohen, David, and Jessica R. Wirth. 1975. eds. Testing Linguistic Hypotheses. Washington, D.C.: Hemisphere Publishing Corporation. Culicover, Peter and Ray Jackendoff. 2005. Simpler Syntax. Oxford: Oxford University Press. Duhem, Pierre. 1954/1982. The Aim and Structure of Physical Theory. Jules Vuillemin, Princeton University Press. (Translated by Philip P. Wiener from the second edition, published in 1914 by Marcel Rivière & Cie., Paris, under the title La Théorie Physique: Son Objet, Sa Structure. (The date of the original publication is 1906. The page references are to the 1991 Princeton Paperback.) Eckman, R. Fred. 1977. ed. Current themes in linguistics: bilingualism, experimental linguistics, and language typologies. Washington, DC: Hemisphere Publishing Corporation. Feyerabend, Paul. 1975. Against Method. New Left Books. (The third edition published by Verso in 1993.) Feynman, Richard. 1965/1994. The Character of Physical Law. New York: The Modern Library. Feynman, Richard. 1974. "Cargo Cult Science," 1974 Caltech commencement address, reproduced in Feynman 1997, pp. 338-346. (Available at: http://www.physics.brocku.ca/etc/cargo_cult_science.php; reproduced in Feynman 1997.) Feynman, Richard. 1997. Surely You're Joking, Mr. Feynman!. New York: Norton & Company. (The hardcover was originally published in 1985.) Fukui, Naoki. 1986. A Theory of Categories Projection and Its Applications, Doctoral dissertation, MIT. Hoji, Hajime. 1995. "Demonstrative Binding and Principle B," NELS 25: 255-271. Hoji, Hajime. 2003. "Falsifiability and Repeatability in Generative Grammar: A Case Study of Anaphora and Scope Dependency in Japanese," Lingua 113: 377-446. Hoji, Hajime. 2006a. "Assessing Competing Analyses: Two Hypotheses about 'Scrambling' in Japanese," in Ayumi Ueyama, ed., Theoretical and Empirical Studies of Reference and Anaphora—Toward the establishment of generative grammar as an empirical science, a report of the Grant-in-Aid for Scientific Research (B), Project No. 15320052, Supported by Japan Society for the Promotion of Science, Kyushu University, pp. 139-185. (Available at: http://www.gges.org/hoji/research/hp- papers.cgi.) Hoji, Hajime. 2006b. "Otagai," in Ayumi Ueyama, ed., Theoretical and Empirical Studies of Reference and Anaphora—Toward the establishment of generative grammar as an empirical science, a report of the Grant-in-Aid for Scientific Research (B), Project No. 15320052, Supported by Japan Society for the Promotion of Science, Kyushu University, pp. 126-138. (Available at: http://www.gges.org/hoji/research/hp-papers.cgi. The paper was circulated originally in 1997.) Hoji, Hajime. 2009. A Foundation of Generative Grammar as an Empirical Science. Book manuscript. University of Southern California, Los Angeles. Hoji, Hajime, Satoshi Kinsui, Yukinori Takubo, and Ayumi Ueyama. 1999. "Demonstratives, Bound Variables, and Reconstruction Effects," Proceedings of the Nanzan GLOW, The Second GLOW Meeting in Asia, September 19-22, 1999. 141-158. Hoji, Hajime, Satoshi Kinsui, Yukinori Takubo, and Ayumi Ueyama. 2003. "Demonstratives in Modern Japanese," In A. Li, and A. Simpson, eds., Functional Structure(s), Form and Interpretation: Perspectives from East Asian Languages, Routledge, London, England, pp.97-128. Katada, Fusa. 1991. "The LF Representation of Anaphors," Linguistic Inquiry 22: 287-313. Kinsui, Satoshi, and Yukinori Takubo. 1992. "Nihongo Sizisi Kenkyuusi kara/e (From/Toward a History of Research on Japanese Demonstratives)." In S. Kinsui & Y. Takubo, eds., Sizisi (Demonstratives), pp. 151-192, Hituzi Syobo, Tokyo. Kuno, Susumu. 1973. The Structure of the Japanese Language, The MIT Press, Cambridge. Kuroda, S.-Y. 1979. "(Ko), So, A nituite (On (ko-), so- and a-)," in Eigo to Nihongo to (English and Japanese), Kurosio Syuppan, Tokyo, pp.41-59.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 47/49 Kuroda, S.-Y. 2008. "Mathematics and Generative Grammar—"Beyond Explanatory Adequacy" and Mathematical Realism of Language: A Fable for Naoki Fukui," Sophia Linguistica 56: 1-36. Lakatos, Imre. 1970. "Falsification and Methodology of Scientific Research Programmes," in I. Lakatos and A. Musgrave (eds.), Criticism and the Growth of Knowledge, Cambridge University Press. pp. 91-195. (Reprinted as Lakatos 1978: chapter 1; the page references are to Lakatos 1978.) Lakatos, Imre. 1973. "Science and Pseudoscience," included in Lakatos 1978 as "Introduction: Science and Pseudoscience." (pp. 1-7) (The page references are to Lakatos 1978; the transcript can be obtained at http://www.lse.ac.uk/collections/lakatos//Default.htm.) Lakatos, Imre. 1978. The Methodology of Scientific Research Programmes: Philosophical Papers Volume 1, edited by John Worrall and Gregory Currie, Cambridge University Press. Lebeaux, David. 1983. "A Distributional Difference between Reciprocals and Reflexives," Linguistic Inquiry 14: 723-30. Loftus, R. Geoffrey. 1996. "Psychology will be a much better science when we change the way we analyze data," Current Directions in Psychological Science, 161-171. (Available at: http://faculty.washington.edu/gloftus/.) Matsushita, D., 1930/1977. Hyoozyun Nihongo hoo (The usage of the standard Japanese). Tokyo: Benseisya. (The original 1930 publication was from Tyuubunkan Shoten. The page reference in the text is to the 1977 edition.) Miyagawa, Shigeru. and Koji. Arikawa. 2007. "Locality in syntax and floating numeral quantifiers," Linguistic Inquiry 38: 645-670. Nakaya, Ukitarou 1958 Method in Science (Kagaku no houhou), Tokyo: Iwanami. Narita, Hiorki. to appear. "The Tension between Explanatory and Biological Adequacy. A Review of Naoki Fukui's (2006) Theoretical Comparative Syntax: Studies in Macroparameters," Lingua. Newmeyer, J. Fredrick. 2008. "A Review of Linguistic Minimalism: Origins, Concepts, Methods, and Aims. by Cedric Boeckx," Language 84: 387-395. Nishigauchi, Taisuke. 1992. "Syntax of Reciprocals in Japanese," Journal of East Asian Linguistics 1: 157-96. Poincaré, Henri. 1952. Science and Hypothesis. New York: Dover Publications. (The English translation of LA SCIENCE ET L'HYPOTHESE (1902)) Popper, K. 1959. The Logic of Scientific Discovery. London and New York: Routledge. (The English translation of Logic der Forschung (1934).) Popper, Karl. 1963. "Science: Problems, Aims, Responsibilities," Federation Proceedings (Baltimore), Federations of American Societies of Experimental Biology Vol. 22, Issue 4: 961-972. Popper, Karl. 1972. Objective Knowledge, Oxford University Press. Reinhart, Tanya. 1983. Anaphora and Semantic Interpretation: Chicago: University of Chicago Press. Saito, Mamoru. 1992. "Long Distance Scrambling in Japanese," Journal of East Asian Linguistics 1: 69- 118. Saito, Mamoru. 2003. "A Derivational Approach to the Interpretation of Scrambling Chains," Lingua 113: 481-518. Sakuma, K., 1936/1951/1983. Gendai Nihongo no Hyoogen to Gohoo (Expressions and Usages in Modern Japanese). (The second and the third editions were published in 1951 and in 1966, respectively. The latter was published in 1983 by Kurosio Publishers, Tokyo.) Schütze, Carson. 1996. The Empirical Base of Linguistics: Grammaticality Judgments and Linguistic Methodology, University of Chicago Press. Takita, Kensuke. 2009. "VP-scrambling, Linearization Preservation, and the Theories of Control," NELS 40 abstract. (http://web.mit.edu/nels40/program/abstracts/NELS40Takita.pdf) Takubo, Yukinori and Satoshi Kinsui. 1996. "Hukusuu no Sinteki-ryooiki niyoru Danwa-kanri (Discourse Management in terms of Multiple Mental Domains," Cognitive Studies: Bulletin of the Japanese Cognitive Science Society, Vol.3, No. 3, pp.59-74. Takubo, Yukinori & Satoshi Kinsui. 1997 "Discourse Management in terms of Mental Spaces," Journal of Pragmatics, Vol. 28, No.6. Elsevier Science, Amsterdam. pp.741-758.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 48/49 Townsend, D. J. & T. G. Bever. 2001. Sentence Comprehension: The Integration of Habits and Rules, MIT Press. Ueyama, Ayumi. 1998. Two Types of Dependency, Doctoral dissertation, University of Southern California, distributed by GSIL publications, University of Southern California, Los Angeles. Ueyama, Ayumi. 2010. "Model of Judgment Making and Hypotheses in Generative Grammar," in S. Iwasaki, H. Hoji, P. Clancy, and S.-O. Sohn, eds., Japanese/Korean Linguistics 17, CSLI. pp. 27- 47. (Available at: http://www.gges.org/hoji/research/hp-Ayumi.cgi.) Weinberg, Steven. 1992. Dreams of a Final Theory: The Scientist's Search for the Ultimate Laws of Nature. New York: Pantheon Books.

0b8ad8b2d4775bc7d97cbe761df2f4a9.doc 49/49

Load more

  49

Copy Link