Strange Pronouns*Paul M. Postal |
(1) | a. | Michelle1 may believe __ and Amanda2 certainly does believe __ [that she is a genius]. |
b. | Michelle1 may believe __ and Amanda2 certainly does believe __ [that her mother is a genius]. |
Expressions (1a, b) have one rather banal reading, on which the highlighted pronoun is linked to neither of the subscripted DPs but rather is understood as denoting some individual distinct from those designated by Michelle and Amanda. I shall have nothing further to say about this reading.
But (1a, b) have two other readings, one involving so called de se interpretation and the other not.1 And, amazingly, on each of those subtly distinct readings, the pronouns she/her are understood as in some sense jointly anteceded by the subscripted DPs. That is, on these readings, (1a) is equivalent to the two sloppy readings of (2), although not of course to its strict reading.2
(2) | Michelle1 may believe that she1 is a genius and Amanda2 certainly does. |
Let us refer to cases of the relevant type as multiply anteceded (singular) pronoun (MASP) structures.
Since there are two (or more) different understood antecedents for the relevant MASP readings, one might naturally expect the highlighted pronouns in (1a, b) to show up as plurals, as in (3):
(3) | a. | Michelle1 may believe and Amanda2 certainly does believe [that they1,2 are geniuses]. |
b. | Michelle1 may believe and Amanda2 certainly does believe [that their1,2 mothers are geniuses]. |
But not only are (1a, b) grammatical on the relevant MASP readings, (3a, b) do not have those readings. Rather, for me, (3a, b) clearly associate with the denotation of each antecedent DP beliefs about multiple individuals, indicated as in (4):
(4) | a. | Michelle1 may believe that they1,2 are geniuses and Amanda2 certainly does believe that they1,2 are geniuses. |
b. | Michelle1 may believe that their1,2 mothers are geniuses and Amanda2 certainly does believe that their1,2 mothers are geniuses. |
Various questions then arise as to how to conceptualize MASP sentences like (1). Conceiving of anaphoric relations in terms of indices, as is of course now common, one might think of providing MASP structures with some kind of joint indices along the lines of the fusion indices of Fiengo and May (1994: 39). These authors, for example, provide the representation (5b) for an example like (5a) on the reading where they relates to the pair {John, Mary}:
(5) | a. | John told Mary that they should leave. |
b. | John1 told Mary2 that they1+2 should leave. |
However, Fiengo and May's discussion clearly contemplates such indices only for plural structures, as in (5a), where the proposition that was 'told' is one whose subject denotes multiple individuals. MASP structures contrast not only in that the surface pronouns they involve are singular, but in the fact that the logical structures they represent do not involve structures denoting multiple individuals but rather multiples of structures involving single individuals, as in (1).3
This notion of 'multiples' of structures might lead one back to notions of the 1960s and 1970s specifically, so called conjunction reduction (CR). This could permit assigning to sentences like (1a) abstract structures like (6):
(6) | Michelle1 may believe that she1 is a genius and Amanda2 certainly believes that she2 is a genius. |
CR would then inter alia have to bring about the combination of the two underlying complements into the single complement found in (1), with the special property that the combination yields a singular pronoun even when two distinct underlying pronouns are present. One might then claim precisely that MASP structures exist only if the pronoun occurs in a set of structures which involve CR. This could account for the contrast between e.g. (1) and (5).
In any event, it is obviously necessary to control the distribution of the MASP property, notably absent in e.g. cases like (7):
(7) | a. | * | Michelle1 told
Amanda2 that she was a genius. ('Michelle1 told Amanda2 that she1 was a genius and that she2 was a genius.') |
b. | * | Michelle1 and
Amanda2 respectively assumed and proved that she
was a genius. ('Michelle1 assumed that she1 and Amanda2 proved that she2 was a genius.') |
That is, (7a, b) lack any reading in which the pronoun links jointly to the two anecedents. The badness of (7a) might fall under a generalization that MASP cases exist only when there are multiple underlying containing constituents. But intuitively this generalization does not hold for (7b), which is, I think, nonetheless as bad as (7a).
This raises the issue of whether MASP structures must involve something else or something else in addition. A possibility is that they need to involve extraction of the containing constituent, on the controversial assumption that, as argued in Postal (1998: Chapter 4), RNR structures are extraction structures. Notably, MASP interpretations seem possible in the much less controversial left extraction cases of (8):
(8) | a. | That she/her mother is a genius, Michelle may believe and Amanda certainly does believe. |
b. | It was her mother's being a genius that Michelle doubted and Amanda rejected. |
However, MASP structures are also possible in passives, passives with extraposition, object raising structures and object deletion structures:
(9) | a. | That she/her mother was a genius was believed by Michelle and proven by Amanda. |
b. | It was believed by Michelle and proven by Amanda that she/her mother was a genius. | |
c. | That she/her mother was a genius was impossible for Michelle to believe or for Amanda to prove. | |
d. | That she/her mother was a spy was too shocking for Michelle to believe or for Amanda to contemplate. |
And there is little hope of taking all these cases to involve extractions.
Thus not much in the way of successful generalizations about MASP cases has been reached. It might be, though, that MASP interpretations are possible only in a constituent X which relates to multiple gaps in the way that an across-the-board extractee relates to gaps. This could be argued to be true in the RNR and left extractions cases and also in those of (9). However, even this generalization is insufficient since good MASP RNR structures like (1) contrast with RNR cases like (10), which seem not to have good MASP readings:
(10) | Michelle1 may have believed and
certainly told Amanda2 [that she was a
genius]. ('Michelle1 may have believed that she1 was a genius and certainly told Amanda2 that she2 was a genius.') |
Here there would be across-the-board like gaps after believed and told Amanda, and yet a MASP interpretation seems impossible. Perhaps there is in addition some kind of parallelism requirement, since the roles of the two antecedents in the two conjuncts of (10) are different.
All the good MASP cases so far have involved ordinary pronouns. But I believe that, although a bit strained perhaps, there are good MASP structures manifesting reflexives:
(11) | a. | Elmer may have criticized and Otto certainly did criticize himself (and only himself). |
b. | It was himself that Elmer may have criticized and Otto certainly did criticize. | |
c. | Himself, Elmer may have criticized and Otto certainly did criticize. |
Further, it seems possible to have MASP-like cases of control:
(12) | a. | Helen may like and Jill surely does like dressing (herself) in pink. |
b. | Helen spoke to Jill about dressing
(herself) in pink. ('Helen spoke to Jill about Helen's dressing herself in pink and about Jill's dressing herself in pink.') | |
c. | Dressing (herself) in pink, Helen may like and Jill surely does like. |
It seems then that the principles governing MASP structures are largely unknown at present. Sentences involving such may then provide a useful focus for thinking about natural language syntax insofar as this is linked to pronominal antecedence, coordination, etc.
Fiengo, Robert and Robert May (1994) Indices and Identity, Cambridge, Massachusetts, The MIT Press.
Hankamer, Jorge (1971) Constraints on Deletion in Syntax, New Haven, Yale University Doctoral Dissertation.
Higginbotham, James (1992) 'Reference and Control' in Richard K. Larson, Sabine Iatridou, Uptal Lahiri, and James Higginbotham (eds.), Control and Grammar, Dordrecht, Holland, Kluwer.
Paul M. Postal (1998) Three Investigations of Extraction, Cambridge, Massachusetts, The MIT Press.
Tancredi, Christopher (1997) 'Pronouns and Perspective', in Hans Bennis, Pierre Pica, and Johan Rooryck (eds.), Atomism and Binding, Dordrecht, Holland, Foris Publications.
* I would be embarassed but not surprised to be reminded that Jorge had something to say about MASP structures in his fine thesis of 1971, unfortunately now rarely cited. I would check it out but my own copy of this has disappeared. Happy Birthday Jorge. [Back]
1 See e.g. Higgginbotham (1992) and Tancredi (1997) for discussion of de se readings. [Back]
2 For relevant discussion of these readings, see Fiengo and May (1994). [Back]
3 This is a simplification, as seen in:
(i) | [The two doctors]1 may believe and [the two lawyers]2 do believe [that they1,2 are geniuses]. |
This has a MASP interpretation, showing that what is at issue is not really singular forms and hence that the terminology is adopted here is misleading. What is at issue might be stated more abstractly as involving single instances of syntactic forms X linked to multiple antecedents, where X has multiple logical correspondents each linked separately to the logical correspondents of the multiple antecedents. That is why (5a) is not a MASP structure. [Back]