Consider the following sentences, with the judgments in accordance with what has been given in the literature.

(1)　a.　John loves himself.
　　b.*John thinks Mary loves himself.
(2)　a.　John recommended himself for that position.
　　b.*John thought that Mary had recommended himself.
(3)　a.　Mary, who had firmly believed that Chomsky would recommend her, was shocked to death when she found out that Chomsky recommended Bill instead. (where her=Mary)
　　b.*Mary, who had firmly believed that Chomsky would recommend herself, was shocked to death when she found out that Chomsky recommended Bill instead.

The judgments on (1)-(3) by 9 native speakers of English who I have recently consulted with are as follows, in terms of the average scores computed as in the handout posted in [21051] (on the scale of -2 to +2).

(4)　The judgments on (1)-(3):
　　a.　(1a): +2
　　b.　(1b): -2
　　c.　(2a): +1.78
　　d.　(2b): -1.56
　　e.　(3a): +2
　　f.　(3b): -1.44

By the 'criteria' stipulated in the handout posted in [21051], the hypothesis that a reflexive in English that appears in an argument position is a local anaphor has been corroborated, at least insofar as the limited date gathered re. (1)-(3). The conclusion is consistent with the judgments that have been reported and widely accepted in the literature.

It has been assumed in many works on Japanese syntax that zibunzisin is a local anaphor, hence on a par with a reflexive in English (in terms of their formal properties). When we check the native speakers' reactions to the Japanese examples corresponding to (1b), (2b), and (3b), however, a very different result obtains. The judgments by 23 native speakers of Japanese on the Japanese analogues of (1b), (2b), and (3b), are as follows.

(5)　The judgments on the Japanese analogues of (1b), (2b), and (3b): (SD=standard deviation)
　　a.　On the JP analogue of (1b): +0.35 (SD: 1.31)
　　b.　On the JP analogue of (2b): -0.13 (SD: 1.39)
　　c.　On the JP analogue of (3b): +0.62 (SD: 1.53)

The hypothesis that zibunzisin (that appears in an argument position) is a local anaphor has thus been falsified, fairly blatantly, by the 'criteria' alluded to above. What is noteworthy is the fact that the speaker judgments on the Japanese analogues of (1b), (2b), and (3b) vary considerably while those on English (1b), (2b), and (3b) are fairly uniform. The figures next to SD in (5) is the standard deviation. I should also note that only two speakers among the 23 speakers judged all of the Japanese analogues of (1b), (2b), and (3b) as "-2," in sharp contrast with the result that six or more speakers (among the nine) judged all of English (1b), (2b), and (3b) as "-2."

The numbers of informants are fairly small, especially for the English sentences. But the readers of this posting are encouraged to check the judgments of their friends about the sentences given above and also the sentences in (1'b), (2'b), and (3'b), which correspond to (1b), (2b), and (3b), respectively, and see for themselves how strikingly different the native speakers' reactions to these examples are.

(1'b)　ジョンはメリーが自分自身に惚れていると思い込んでいた。（ジョン=自分自身)
　　John-wa Mary-ga zibun-zisin-ni horete iru to omoikonde ita. (John = zibun-zisin)
(2'b)　ジョンはメリーが自分自身を推薦したとばかり思っていた。（ジョン=自分自身)
　　John-wa Mary-ga zibun-zisin-o suisensita to bakari omotteita. (John = zibun-zisin)
(3'b)　チョムスキーが自分自身を推薦すると思い込んでいたジョンは、チョムスキーがビルを推薦したと知って愕然とした。(自分自身=ジョン)
　　Chomsky-ga zibunzisin-o suisensuru to omoikondeita John-wa Chomsky-ga Bill-o suisensita to sitte gakuzen to sita. (John = zibun-zisin)

As noted in the handout posted in [21051], when a given example is claimed to be unacceptable for a grammatical reason, we should not expect any lexical/pragmatic adjustment to improve the status of such an example in any significant way, which is basically what we seem to observe in the case of examples like (1b), (2b), and (3b) in English, assuming that the pragmatics for (1b), (2b), and especially the one for (3b), favor the intended 'long-distance' reading. In the case of the Japanese analogues of (1b), (2b), and (3b), it is perhaps the case that the lexical and pragmatic factors do significantly affect the acceptability judgments. The fact that many speakers find the Japanese analogues of (1b), (2b), and (3b) not so bad is totally unexpected under the hypothesis that zibunzisin is a local anaphor; 11 out of 23 speakers gave "+2" to (3'b), presumably finding them to be fully acceptable. In case one might wonder, the non-linguist informants' judgments and the linguist informants judgments on (1'b), (2'b), and (3'b) are not very different from each other.

At this point, a proponent of the hypothesis that zibunzisin is a local anaphor might suggest that when the relevant examples are judged acceptable, zibunzisin is taken to be 'emphatic' or something other than a regular local anaphor. In order to defend such a thesis, one would have to show under what linguistic or pragmatic contexts, zibunzisin 'of this type' can be excluded so that we can identify an environment in which a local anaphor zibunzisin can reliably be used as such. Until such a demonstration is done, there does not seem to be much merit in using zibunzisin in our syntactic experiment that addresses formal properties of the language in question. As noted in my Mayfest abstract in [21482], we can also expect a reaction that makes reference to the existence of a 'contrast', which might go as follows. The average score on (3'b) is +0.62; but an example in which zibunzisin has its antecedent in its local domain must be close to +2. Hence there is a significant contrast between the two and that is what we are interested in capturing. As also briefly noted in my Mayfest abstract, one would maintain such a position only if one were not concerned with making one's hypothesis falsifiable, as far as I can tell. What makes the contrast significant is the status of the unacceptable example, and its unacceptability is claimed to be due to grammar (or some formal property of the language in question). Therefore, when the average response to such an example is +0.62, i.e., when the negative prediction made under the hypothesis in question is clearly disconfirmed, the contrast itself does not have much significance at least for a research that aims at discovering formal properties of the language in question and ultimately of UG.

One might think that what is presented above constitutes an exceptional case where it goes so directly against what has been assumed widely in the literature, even if the results reported above get replicated in other instances of essentially the same experiments. It, however, seems to be the case that many other hypotheses that have been widely accepted in the field get falsified or fail to be corroborated once we conduct an experiment on them, 'with the attitude of falsification'. For example the claims/hypotheses on (i) the locality of so-called numeral floating quantifiers (see the WCCFL paper posted in [18703]) and (ii) the Proper Binding Condition effects in so-called raising-to-object construction in Japanese (see the handout posted in [21051]) are among them. If we use the kind of paradigms that have been given in the literature, even hypotheses that have to do with (iii) subjacency effects with respect to scrambling out of a complex NP and (iv) 'long-distance scrambling' not exhibiting so-called A-properties seem to get falsified. The claims/hypotheses noted above are among those that have been very widely accepted. Except for (i), there has not been much controversy, at least in the published works, as to the validity of the claims/hypotheses in question. I.e., they are not like some other empirical claims for which apologetic remarks such as "the relevant judgments are very subtle" are given. It is therefore not difficult to imagine what would happen to the empirical claims that are accompanied by such an apologetic remark once we conduct an experiment on them.