The Intransparency of Parentheticalism

Abstract

It has frequently been observed that typical utterances of the form “I believe that P” are assertions of the embedded proposition P. Yet that the matrix clause “I believe that” should be semantically idle creates an interesting puzzle: linguistic orthodoxy holds that the utterance is an assertion about one’s doxastic state, not about the content of this state. In response to the puzzle, Tim Henning has recently proposed a new semantic theory, parentheticalism, according to which “S believes that P” expresses P from S’s point of view: the at-issue content in belief ascriptions is the embedded proposition P. The puzzle is then claimed to be resolved as follows: as speaker and doxastic subject are identical in the first-person case, “I believe that P” expresses P from the speaker’s own point of view. In this paper, I argue, first, that parentheticalism is highly doubtful and, second, that even if parentheticalism were true, it would be unable to resolve the transparency puzzle.

It has frequently been observed that typical self-ascriptions of belief of the form

(1) I believe that P

are “transparent” with respect to the doxastic matrix clause: they are not assertions about the speaker’s mental states, but weakened assertions of the embedded proposition P.¹ The doxastic clause – “I believe that” – is somewhat idle, its only role being that of diminishing assertoric force. Yet transparency provides, or seems to provide, a puzzle because, according to the orthodox view, an utterance of (1) is an assertion about a person’s mental state, not about the embedded proposition P. So the orthodox view on the asserted content appears to conflict with apparent linguistic data. A solution to this transparency puzzle either denies the apparent phenomenon, or it provides an alternative theory that explains why the speaker uses (1) to (weakly) assert the content P, and not the content I believe that P.²

Often, transparency is taken to reflect a deeper form of first-person asymmetry, indicating that there is something special about self-ascriptions of mental states.³ Recently, however, Tim Henning has deemed such a position as “fundamentally mistaken” (Henning, 2018, 25). He claims that the transparency puzzle can be straightforwardly resolved by reference to his general theory of parentheticalism. According to parentheticalism, with

(2) S believes that P,

a speaker Sʹ does not, primarily, make an assertion about S’s beliefs, but rather “claims that P from S’s point of view” (cf. Henning, 2018, 19). In belief ascriptions generally, the at-issue content, expressed by the speaker, is not S’s doxastic state but rather the embedded content P. The transparency of (1) is then supposed to follow from the additional fact that here the speaker Sʹ and the doxastic subject S are identical.

Henning’s theory is intriguing. It promises a unified treatment for first- and third-person ascriptions of belief states, and hence to preserve semantic continuity even in the face of transparency (Henning, 2018, 11). And as it does not refer to the peculiarities of the first-person pronoun occurring in (1), there is, according Henning’s theory, “nothing that is semantically special about the first-person case. It is just that here, the speaker happens to coincide with the subject to whom the claim is oriented” (Henning, 2018, 21; cf. 22 and 27). Closer examination, however, reveals substantial problems. In this response, I will argue that, even if I were to ignore my doubts about parentheticalism itself (see Sect. 2 for these doubts), parentheticalism could not resolve the transparency puzzle (Sect. 3). I begin, in Sect. 1, with a sketch of Henning’s theory.

A note on terminology. Henning uses the label “parenthetical use/reading” to cover the linguistic data of both the first- and the third-person cases (cf. Henning, 2018, 16), obviously presupposing what I dispute in Sect. 3: that the transparency of (1) and the peculiarities of (2) are phenomena of the same type. Moreover, by calling his specific proposal ‘parentheticalism,’ Henning suggests (misleadingly, in my view) that there is a conceptual link between the phenomena to be explained and the explaining theory. To clarify the situation, I will put the adjective “parenthetical” in scare quotes and employ the term “content shifting” to refer to the particular phenomenon associated with the ‘parenthetical’ use of (2) (see Sect. 1 below).

1 Parentheticalism and Transparency

Henning (2018, 17–21) observes that belief ascriptions of form (2) frequently display an interesting feature. In their ‘parenthetical’ use, the at-issue content, the proposition up for debate, is not the proposition that S believes that P, but the embedded proposition P itself: “In a typical utterance of [(2)], the at-issue content is that P. If I say: ‘In Paul’s opinion, there will be rain tonight,’ rejection or challenge (‘No way!’) must be heard as directed at a claim about the weather, not at a claim about Paul’s opinions” (Henning, 2018, 18). As the interlocutors’ reactions suggest, the at-issue content is not what is reported by the utterance but the embedded proposition P. Call this phenomenon (at-issue) content shift.

Henning proposes a semantic theory, parentheticalism, to explain the content shift of belief ascriptions such as (2): the speaker asserts that S believes that P, but this reported content – that S has a certain opinion – is only “backgrounded” (Henning, 2018, 18). The foreground is constituted by another semantic content of the same utterance: the embedded proposition P functions as the at-issue content, at stake for acceptance, rejection, or at least further consideration. Thus, according to Henning’s theory of parentheticalism, assertions of (2) in their ‘parenthetical’ uses have two semantic contents, S believes that P and P, which are, however, related: in virtue of the reported content (that S believes that P) the at-issue content (P) is expressed by the speaker.⁴

This brings us to a second point. Although sentence (2) has two contents in its ‘parenthetical’ use, there is only one commitment undertaken by the speaker: she asserts merely that S believes that P (Henning, 2018, 18, cf. 5), and is thus only committed to the backgrounded content. The speaker also expresses the at-issue content P, but typically remains neutral with regard to its truth: In “S believes there will be rain tonight,” the “utterance clearly does not commit [the speaker] to the truth of the content that there will be rain tonight, and [the speaker] cannot count as having asserted this content” (Henning, 2018, 19, cf. 20 and 41). Rather, the utterer of (2) expresses or claims the at-issue content P “from S’s point of view” and “on S’s behalf” (Henning, 2018, 19, 22).⁵ In this way, S is “held accountable for the truth of P” (Henning, 2018, 19). If P turns out false, not the speaker but S takes the blame. The at-issue content is oriented towards subject S.⁶

Let’s turn to transparency. Henning assumes the transparency of (1) and the content shift in third-person belief ascriptions to be of the very same type (because of which he employs the term “parenthetical use/reading” to refer to the linguistic data of both first- and third-person belief ascriptions⁷). The theory of parentheticalism, primarily designed to account for content shift in third-person belief ascriptions, is therefore deemed applicable to first-person instances of (2), as well. Henning reasons as follows: Parentheticalism, applied to (1), would yield the fact that “I believe that it will rain” concerns the weather. And as the speaker expresses the at-issue content from the perspective of the doxastic subject, it is the latter who is held accountable for the truth of the proposition that it will rain. Now, in (1), the speaker is identical to the doxastic subject. Hence she ‘inherits’ the subject’s commitments and does not remain neutral with regard to P: “When I say: [‘I believe that there will be rain tonight,’] listeners will regard me as committed to the truth of the content that there will be rain tonight” (Henning, 2018, 21). Speaker–subject identity turns the subject-orientation into a speaker-orientation: it turns an expression of P from the subject’s point of view into a claim that P made from the speaker’s point of view. The assertoric force in relation to the embedded content P is then “nothing more than simply a special case of the general feature of subject-orientation” (ibid.; italics in the original).

Henning’s solution to the transparency puzzle is likely to be one of the most simple and elegant on the market (at least if we bracket the complexities of parentheticalism itself): it does not rest on any contentious form of first-person asymmetry and avoids the intricacies of the de se.⁸ It relies only on the generally accepted fact that “I” refers to the speaker of the sentence. Nevertheless, I think that Henning’s proposal is fundamentally flawed in that it wrongly assumes transparency to be but a special case of content-shifting. I will thus argue that even if parentheticalism were the right explanation of content-shifting in (2), it could not solve the transparency puzzle. Before I turn to this argument, however, I will show that parentheticalism itself is highly problematic: as I will demonstrate by reference to a series of examples, it is implausible that parentheticalism provides a unified explanation of all cases of content-shifting.

2 Problems for Parentheticalism

To fix our intuitions, let me give another example of content-shifting along the lines of (2). Suppose we are contemplating the prospects of investing in the stock markets. We have already discussed economic reports, company assets, and market shares. Now A says, without apparent change of topic,

(2′) Barren Wuffett believes the stock markets will rise to a new high,

thereby referring to the famous star of Wall Street. We may assume that A still intends to contribute to the preceding discussion and that acceptance (“That’s what I have been saying all along”) and rejection (“No, the crash is nigh”) are directed at the embedded proposition. Thus (2′) is a case of a content shift by Henning’s own lights: The at-issue content is not the proposition that Barren Wuffett believes that …, but the embedded proposition that the stock markets will rise to a new high. Furthermore, A does not herself assert the at-issue proposition, but she orients it towards Barren Wuffett. If the stock markets do not develop positively, we might well blame Barren Wuffett, not, however, A.⁹ According to Henning’s theory, therefore, the content shift of (2′) is to be explained by reference to parentheticalism: A expresses that the stock markets will rise from Barren Wuffett’s point of view.

I will now introduce a series of content-shifting cases which cannot be explained by parentheticalism, or at least not in a straightforward way. As will emerge, content-shifting is a very general phenomenon. I will not, however, provide a general theory of content-shifting, and rest content with identifying some content-shifting cases by reference to Henning’s criteria, i.e., speaker intention and the reaction of the interlocutors.

Assume that A were not to refer to what Wuffett believes but to what he asserts. Instead of uttering (2′), A says,

(3) Barren Wuffett asserts that the stock markets will rise.

A intends to contribute to the debate and the object of discourse is the embedded content: assent and dissent may take the same form as in case (2′). And, surely, there is a sense in which Wuffett would be to blame if the stock markets were to drop. In the present context, A’s utterance of (3) exhibits the content shift no less than (2′) does. But since (3) is an ascription, not of belief, but of assertion, parentheticalism would have to be expanded from the psychological to the illocutionary case. Parentheticalism cannot be limited to ascriptions of mental states alone.

While (3) may be subject to a plausible and indeed natural extension of parentheticalism, other cases may prove more problematic. Suppose we are in the midst of a debate on the temperature in the water boiler, when you say,

(4) The thermometer shows that the temperature is 65° C.

Your remark surely is pertinent to the discussion and it also survives Henning’s test: affirmation (“Of course, the boiler has been set on 65° C”), rejection, and challenge (“No way, the boiler has not been functioning for days!”) are plausibly construed as being directed at the embedded proposition, the claim that the boiler has that temperature. Furthermore, if the temperature is not 65° C, it’s not you who is to blame: the thermometer must be repaired or henceforth ignored.¹⁰ Thus, by Henning’s own lights, your utterance of (4) displays the content shift phenomenon as much as (2) and (3) do. The at-issue content is the embedded proposition P, oriented towards the thermometer, not the speaker. Yet how to account for the content shift in this case? Parentheticalism would have to claim that you express that the water has 65° C “from the point of view” and “on behalf” of the thermometer. Even if such a view may, perhaps, be defensible, commitment to it surely is a prima facie weakness for parentheticalism.

Nor do problems arise only for content-shifting in non-doxastic ascriptions. Consider doxastic cases of content-shifting in higher-order belief ascriptions of the form.

(5) S believes that S′ believes that P.

Suppose not only the stock markets but Barren Wuffett himself is an enigma to us. It is common ground that Wuffett is an expert on the stock markets, but A does not know, at least not directly, what Wuffett believes. Fortunately, A’s friend B is generally well informed about the famous Wall Street figure. A might well say, again while we are contemplating investing our money,

(5′) B thinks that Barren Wuffett believes that the stock markets will rise.

Plausibly, A’s utterance about B is ultimately economy-directed too: assent (“Of course, I have been saying this all along”) or challenge (“No way, the crash is nigh”) are directed at the proposition that the markets will rise. The at-issue content then is the ultimately embedded proposition that the stock markets will rise, and this is oriented not towards the speaker or towards B but, if only indirectly, towards Barren Wuffett.

Parentheticalism in the form Henning presents to us cannot account for this case. According to this theory, A expresses that Barren Wuffett believes the stock markets to rise from B’s point of view. The at-issue content, expressed from B’s point of view, is hence the proposition that Barren Wuffett believes that the stock markets will rise. Thus, parentheticalism does not permit the ultimately embedded proposition – that the stock markets will rise – to be an at-issue content of A’s utterance, let alone for it to function as the at-issue content. Parentheticalism hence provides the wrong results for higher-order belief ascriptions.

In response to this problem, parentheticalists might propose modifying the theory. Consider, for example, the variant of parentheticalism according to which in higher-order belief ascriptions the at-issue content is not the first embedded proposition (the proposition that S believes that P), but the ultimately embedded proposition (the proposition P). This theory would then account for the content shift of (5) and (2) equally. Nevertheless, it seems implausible – even if we ignore that any such change of parentheticalism would be an ad hoc modification, motivated only by the threat of an apparent counterexample. First, recall that, according to parentheticalism, the speaker undertakes no commitment with respect to the truth of the proposition she expresses from some other person’s point of view. Yet if the utterer of (5) does not commit to S′’s believing that P, it seems impossible for the speaker to express P from S′’s point of view. Parentheticalism does not explain that the at-issue content – that the stock markets will rise – is oriented towards Barren Wuffett. Second, if the first problem were resolved, the theory would have the implausible result that with (2′) and (5′) the speaker would do the same: express that the stock markets will rise from Barren Wuffett’s position. Third and finally, not every ‘parenthetical’ use of (5) is directed at the ultimately embedded proposition. In contexts in which we are not concerned with the stock markets but with what Wuffett will do next, A’s statement (5′) will be of interest because it provides evidence on Wuffett’s beliefs, and not because it provides evidence on the stock markets. The at-issue content is then the proposition that Barren Wuffett believes that the stock markets will rise. An adequate parentheticalist theory would have to make room for the possibility that there are candidate at-issue contents beyond the ultimately embedded propositions.

To accommodate this desideratum, one might propose a version of parentheticalism according to which the utterer of (5) expresses all embedded propositions from the point of view of the respective subjects. The speaker of (5) would then express the proposition that S′ believes that P from S’s point of view and the proposition P from S′’s point of view. Beside the reported content – that S believes that S′ believes that P – there would be two possible at-issue contents, that S′ believes that P and P itself. But this theory does not seem to be satisfactory either. First, the identification of the at-issue content amongst the different candidates could not be a matter of semantics and would hence be inexplicable by reference to parentheticalism alone. Second, this theory does not escape the already mentioned problem that, as the speaker does not commit to S′ having such a point of view, it is very implausible that she can express P from S′’s point of view. Some required orientation shifts would remain unexplained. Third, and most importantly, given that belief ascriptions can be iterated indefinitely, the mentioned strategy would require an implausible multiplication of semantic contents for a single utterance. For nth-order belief ascriptions, such a theory would have to postulate, besides the reported content, n potential semantic contents.

For those parentheticalists who are still not abashed, let me point out that the phenomenon of content-shifting is not even restricted to that-clauses: at-issue content need not even be embedded. Suppose believer C inquires about the beginning of mass and D answers without apparent change of topic:

(6) Mr. Pious is already on his way.

D here obviously makes a case for the claim that mass will begin soon and that C better get her coat, too. That the beginning of mass is the at-issue content is confirmed by Henning’s test: assent (“Oh, now I remember that mass starts early today”) or dissent (“But mass is always in the evening”) are directed at that proposition. Yet such cases clearly defy explanation by parentheticalism. When the at-issue content does not verbally shine up anywhere, there is not even the theoretical option for parentheticalism to kick in. Nor would we expect a semantic theory to be necessary. There is an obvious and straightforward pragmatic alternative: D’s observation promises to give valuable information on the primary topic of interest, viz., the beginning of mass. The at-issue content of the assertion is hence not an additional semantic content, but pragmatically fixed as that proposition for which D provides evidential support. This would also explain the fact that, in different contexts, the at-issue content may be shifted to different propositions: if, e.g., C’s question does not concern the beginning of mass but, say, of the Bishop’s visit, the at-issue content of (6) changes accordingly.

The pragmatic account seems applicable also to those cases in which the at-issue content is shifted to the embedded proposition. Consider (4): Your remark has the function of providing evidence for the claim that the temperature in the water boiler is 65° C. That’s why you refer to the thermometer in the first place. (Compare: “What time is it?” – “The clock shows 3.”) Thus, your observation of the thermometer reading is of relevance, because (or in as far as) it promises to give valuable information on the primary topic of interest, viz., the water temperature. And the orientation is determined by the fact that the function of showing the temperature is ascribed to the thermometer. No reference to semantic peculiarities is necessary or plausible.

From here it is but a very short step to a pragmatic account of cases (2′) and (3). A’s reference to Wuffett plausibly has the same function as your remark on the thermometer. We are interested in Barren Wuffett’s beliefs and assertions because they promise to hold valuable information on the topic under discussion. Thus our cases can plausibly be given a pragmatic treatment as well: the content shift of both (2′) and (3) derives from discourse context (the discussion concerns the stock markets, and more information is welcome), common knowledge (everyone believes, and believes that the others believe, Barren Wuffett to be an expert in the field), and Grice’s cooperative principle, which demands the speech act to be a meaningful contribution to the discourse (Grice, 1989). The at-issue content is that proposition on which A provides relevant information or, in the present case, justification.

Needless to say, pragmatics might also provide the means to account for the vagaries of content-shifting in iterated belief ascriptions such as (5′). B’s belief provides direct evidence on Wuffett’s belief and indirect evidence for a positive economic development. If, as in the original case, the stock market is the topic at stake in A’s statement, it directs us at the ultimately embedded proposition. And this at-issue content, that the stock markets will rise, is oriented, however indirectly, towards Barren Wuffett because B thinks that Wuffett has this belief. The pragmatic explanation would even be sensitive to finer distinctions: that there is a higher-order belief involved ties in with the fact that the evidential support for the ultimately embedded proposition will be weaker than in the original, first-order case of form (2): in (5′), there is the additional risk that B is wrong in her belief-attribution. If, on the other hand, not the stock markets are at issue but Barren Wuffett’s doxastic state (we want to know what he will do next, and hence are interested in his opinion on the economy), pragmatics may explain why (5′) will be understood as concerning neither B nor the stock markets but Wuffett himself. A revisionary semantic theory would then be unnecessary.¹¹

To conclude. Of all the different cases of content-shifting, parentheticalism in its present form is at best able to handle first-order ascriptions of belief or assertion as in (2′) and (3). The applicability of parentheticalism to (4) is highly doubtful. Higher-order belief ascriptions of the form (5) would require major revisions in the theory, and even then success is not guaranteed. The shifting of content to unembedded propositions (as in (6)), finally, escapes the grip of parentheticalism altogether. Parentheticalism can plausibly cope with only a few of the content-shifting cases. Moreover, since it is a revisionary semantic theory, it comes at great theoretical cost. And in view of the fact that Gricean pragmatics promises a unified treatment of all content-shifting cases without additional theoretical baggage, it seems fair to say that parentheticalism with respect to (2), while not ultimately disproven, appears to be an unattractive option.

I will now claim that, even if we took that option, this would not help us in explaining the transparency of (1).

3 The Failure of Parentheticalist Transparency

Transparency is the phenomenon that self-ascriptions of doxastic states of the form (1) are hesitant or weakened assertions of P, not about one’s doxastic states. Henning assumes that transparency is but a special case of content-shifting and hence also explained by parentheticalism: he claims that with an utterance of (1), like with an utterance of, say, (2′), the speaker expresses P from the doxastic subject’s point of view. Since, in (1), speaker and doxastic subject happen to be identical, the speaker expresses P from what turns out to be her own point of view and therefore, according to Henning, ‘inherits’ the doxastic subject’s P-related commitments. This explains why the speaker in fact asserts that P: the transparency of (1) is a consequence of parentheticalism plus the happenstance that speaker and subject are identical. As the theory does not refer to the first-person pronoun at all, let alone to its semantic or epistemic peculiarities, Henning concludes that transparency does not reflect a deeper form of first-person asymmetry; it is “not due to semantic or syntactic features specific to [(1)]” (Henning, 2018, 22).

To examine the success of parentheticalism with respect to transparency, we will now ignore our worries from Sect. 2 and assume that parentheticalism is the correct explanation of the content-shifting in first-order belief ascriptions: by uttering (2), the speaker indeed claims that P from S’s point of view. Our aim is to show that, even if parentheticalism is correct, it cannot account for the transparency of (1).

First doubts arise once we recall the nature of the task. According to the transparency thesis, an utterance of (1) is an assertion of P, and not also an assertion about the speaker, however backgrounded. Urmson, from whom Henning takes his major inspiration, is very clear about this when he says that his parentheticals are “not psychological descriptions” (Urmson, 1952, 482). And Wittgenstein says: “So it looks as if the assertion ‘I believe’ were not the assertion of what is supposed in the hypothesis ‘I believe’!” (Wittgenstein, 1953, part ii x, 190). It appears therefore that the transparency phenomenon to be explained comprises a positive claim – that (1) is an assertion of P – and a negative claim – that (1) is not an assertion of the proposition that she believes that P. Henning’s reconstruction on the basis of parentheticalism, however, fails to yield the negative claim of transparency. According to a parentheticalist construal, a person uttering (1) ends up asserting two propositions: she asserts the at-issue content P, and, in addition, she asserts the proposition that she believes that P, however backgrounded.

One way to get around this problem would be to claim that Urmson’s and Wittgenstein’s ‘negative’ claims about transparency are not to be taken literally; that their particular phrasings are mere hyperbole resulting from overemphasizing what is special in first-person belief ascriptions, namely that the embedded proposition P is the object of primary assertion. And upon my insistence that such an inclusivist construal of transparency would contradict not only the wording but also the spirit of both Urmson and Wittgenstein,¹² Henning might resort to the position that Urmson and Wittgenstein simply do not get the transparency phenomenon right in this respect; that we should liberate ourselves from their particular, exclusivist views, and thus permit a construal in terms of the unified theory of parentheticalism. Upon which I would retort that this conception would be revisionary not only with respect to the explaining theory, but also with respect to the phenomenon to be explained. And so on.

Yet, instead of pursuing this debate, let me address my main worry. In my view, parentheticalism cannot even account for the positive claim of transparency. Recall that, according to Henning’s suggestion, transparency results from parentheticalism and the additional fact that subject and speaker happen to be identical. No more is required. In particular, the fact that the speaker–subject identity in (1) is the result of the semantics of the first-person pronoun is but incidental to the case: “There is nothing that is semantically special about the first-person case. It is just that here, the speaker happens to coincide with the subject to whom the claim is oriented” (Henning, 2018, 21; cf. 22 and 27). Yet if nothing hinges on the presence of the first-person pronoun, any ‘parenthetical’ statement of the form (2) should, according to Henning, turn out to be transparent – provided only that speaker and subject happen to coincide. A little reflection shows that this cannot be correct.

Suppose, for example, that, in the context-shifting scenario we have assumed above, A happens to be Barren Wuffett. Given a parentheticalist analysis of (2′), A expresses the embedded proposition from Wuffett’s point of view. And as A is in fact Barren Wuffett, A expresses the embedded content from what turns out to be his own point of view. Henning would be forced to claim that (2′) is a transparent utterance: that A asserts that the stock markets will rise to a new high. Yet this is the wrong result, which becomes particularly obvious if we assume A to be ignorant of his own identity.¹³ Suppose, therefore, that A suffers from dissociative amnesia, is temporarily in a fugue state, and does not know that he is in fact Barren Wuffett. And suppose that A has earlier made the claim that the Dow Jones will climb to new record heights. Henning’s theory postulates that, in uttering (2′), A would repeat his earlier claim, and do so with reduced assertive force. But this is absurd. In uttering (2′), A does not make a claim about the stock markets, let alone a weakened claim. Recall that, in the situation we imagined, A’s utterance of (2′) has the function of supporting his earlier claim by citing an expert who is of the same opinion.

This point is strengthened if we consider definite descriptions. Assume that E is, unbeknownst to her, the richest person in town. Suppose further that she does not believe that the stock markets will rise. She believes, however, that the richest person in town believes that the stock markets will rise. In the appropriate context, E’s sincere utterance

(2′′′) The richest person in town believes the stock markets will rise

displays content-shifting no less than (2′) does. (E might utter this sentence in a conscientious attempt to provide all the available evidence for and against the at-issue proposition that the stock markets will rise, a proposition she considers to be false.) According to parentheticalism, E expresses that the stock markets will rise from the perspective of the richest person in town. And because E is in fact that person, it would follow that E asserts that the stock markets will rise. But this is surely wrong. As I have constructed the case, E does not even believe that the stock markets will rise. And even if I were to alter the case and assume that E has changed her mind and is now more optimistic about the economy, she does not assert that the stock markets will rise, however weakly, by saying (2′′′). After all, she does not know that she is the person she refers to when she provides evidence for her newly acquired belief. I conclude that Henning’s theory fails: transparency does not flow from parentheticalism and the simple fact that speaker and subject are identical.

It might be tempting to fix the problem by augmenting or replacing the mere identity requirement with the demand that the speaker and, perhaps, the audience (correctly) believe that she, the speaker, is identical to the subject. As this would require departing from the basic tenets of parentheticalism, I will not here discuss this suggestion. Two remarks may suffice. First, it is highly questionable whether this modification would do the job. Because of his enduring success in the past, self-confident Barren Wuffett (and the audience) might actually be convinced that his beliefs are excellent indicators of truth. Thus, even knowing who he is, Barren Wuffett might, in the given context, utter (2′) with the intention of providing evidence for the previous claim that the stock markets will rise. A construal according to which he simply repeats a weakened assertion about the stock markets would then be very implausible, which would show the revised theory to overgenerate as well. And second, the belief required would likely have to be a de se belief. Otherwise it would again be possible to construct counterexamples on the basis of the speaker’s failure to identify with the doxastic subject. But if an explanation of transparency requires reference to de se beliefs, it seems that transparency would have to be a first-person phenomenon after all. Henning would have to abandon the claim that transparency is not, essentially, de se.

4 Conclusion

I have argued that Henning’s parentheticalism is highly doubtful: it postulates a revisionary semantics but is of questionable value for the vast array of ascriptions of that-clauses exhibiting content-shifting, e.g., ascriptions of non-doxastic that-clauses and of higher-order belief clauses. And it cannot, in principle, account for cases in which the at-issue content is not an embedded clause at all. Since there is a conservative, pragmatic approach which may provide a unified treatment of the many cases of content-shifting, parentheticalism seems to be both ineffective and too costly. At the very least, the pragmatic approach must be explored first. Furthermore, I have argued that even if we were to ignore these doubts and buy into parentheticalism, Henning’s account of transparency cannot be correct: it is unable to account for the negative claim of transparency. Even worse, it fails to restrict the phenomenon of transparency to first-person ascriptions of belief. It is, of course, not excluded that a modified account of parentheticalism, involving further or different criteria, is able to handle the latter problem and to avoid the problematic instances of (2). But, first, given the difficulties of parentheticalism, this does not seem to be a very promising route to take. And second, it appears that all remotely plausible candidates must be such that transparency is effectively restricted to first-person ascriptions of mental states of form (1).

Henning does not show that transparency is not a first-person phenomenon. And it may well turn out that transparency is not a case of content-shifting at all.¹⁴