The analysis I propose specifies definiteness, interpretation (here: generic vs. existential), and focus for each nominal argument in the input. GEN adds prosodic phrase boundaries, stress and word order. (This is a matter of convenience; in keeping with Richness of the Base, any of these aspects could be specified - or left unspecified - in the input, subject to addition and/or change as governed by markedness and faithfulness constraints; that is, the present analysis is decidedly *not* taking a `translation'-point of view, according to which the OT-grammar models production or comprehension, but a `generation'-point of view, according to which the grammar as a whole generates the set of all well-formed sound-meaning pairs.)
I assume that prosodic phrasing is governed by (cf. Truckenbrodt 1995):
(1) Phonological Phrase Formation (pPF)
a set of constraints that conspire to build phonological phrases (pPs) of the form (argument-XP predicate) or (XP), but not, e.g. (XP YP) or (predicate)
(2) Intonational Phrase Headedness (iPH)
an alignment constraint that makes the rightmost phonological phrase within an intonational phrase (iP; more or less: the clause) the head of iP, i.e. maximally prominent
(3) Focus Prominence (FPro)
a constraint that keeps prosodic phrases without foci in them from being more prominent than prosodic phrases with foci in them
In addition, there are morphosyntactic constraints which influence linear order, such as - in the case of German - `subjects precede objects', `datives precede accusatives', `animates precede inanimates' etc., which I'll refer to as MS-constraints.
Non-canonical word order can arise if the structure that meets the MS-constraints and the phrasing constraints (1) and (2) optimally would yield an output that is at odds with (3); provided, that is, that the language in question has prosodic constraints outrank (or be tied with) MS-constraints. In (4), parentheses represent phonological phrases, capitals the main accent (=head of iP) (secondary accents, corresponding to the heads of pPs, are omitted):
(4) Who did you give the money to? ( -> DatO must be focus)
a.# (ich habe)(dem Kassierer)(das GELD gegeben) violates Foc.Prom.
b.# (ich habe)(dem KASSIERER)(das Geld gegeben) viol. iP-headedn.
c. (ich habe)(dem KASSIERER das Geld gegeben) viol. pPF
d. (ich habe)(das Geld)(dem KASSIERER gegeben) viol. Dat>Acc
`I gave the-ACC money to the-DAT teller.'
The ranking for German is as shown in (5), allowing the two grammatical options in (4c) and (4d), the former with a defective prosodic structure, the latter with morphosyntactically marked word order:
(5) FP, IPH >> 3PS >> pPF <<>> MS
(5) contains an additional constraint, 3PS (for `tripartite structure'), which concerns placement and interpretation of indefinites. I follow (some of) the authors mentioned above in assuming that an indefinite's interpretation depends on whether it winds up as part of the restrictive clause (RC -> generic) or as part of the nuclear scope (NS -> existential) of an invisible quantifier. The mapping from syntax to the semantics is governed by an iconicity constraint that seeks to `synchronize' background-focus structure and tripartite semantic structure:
(6) Iconicity Constraint for Tripartite Structure (3PS)
Left-align the NS with a phonological phrase containing a focus.
In a manner of speaking, 3PS places the RC-NS boundary between two pPs, the later one of which must contain a focus. As a consequence of (6), indefinites will have to precede the RC-NS boundary if specified as generic, but follow it if existential.
The most immediate consequence of (6) is that generics resists `integration' with their predicate, cf. (7):
(7) Wenn man in die USA einreisen will, muss man
a. (Vorstrafen)(ANgeben). violates pPF (V is its own pP)
b. (VORstrafen angeben). would violate 3PS, if NP is generic
`If one wants to enter the US, one has to report previous
convictions.'
(7a) can only be understood as generic (thus as a true statement), given that the violation of pPF can only be justified by the higher 3PS which wants there to be a pP-break following the indefinite. (7b) can only be understood as existential (thus false), because a generic indefinite would require a fragmented prosodic structure.
3PS is a farily high ranked constraint, but it will be violated whenever a generic is the only focus in a clause, given that in that case, no focus-containing pP can be found that follows the RC-NS boundary:
(8) Wenn man in die USA einreisen will, muss man ansteckende Krankheiten angeben. Darueberhinaus muss man
a. (VORstrafen angeben) violates 3PS
b.#(VORstrafen)(ANgeben) violates Focus Prominence (given that the V is not focused, yet maximally prominent)
`If one wants to enter the USA, one has to report contageous diseases. On top of that, one has to report previous convictions.'
The final and perhaps most striking effect of 3PS, however, is that it often favors generics to precede non-generics, (9), while it disfavors `scrambling' of non-focused existentials for prosodic reasons, (10), even where such scrambling is possible with definites, (11):
(9) Laut Gesetz muessen wir
a. (Ladendiebe)(der POLIZEI melden). violates Dat>Acc
b.#(der Polizei)(LADENdiebe melden). violates 3PS
`According to the law we have to deliver shoplifters to the police.'
(10)Wir verkaufen Huete in aller Herren Laender. Neulich habe ich
a. (einem BRASILIANER einen Hut verkauft) violates pPF
b.#(einen Hut)(einem BRASILIANER verkauft) violates 3PS
`We sell hats throughout the world. Recently I sold a Brasilian a hat.'
(11)Wenn ich schon mein Haus verkaufen muss, dann moechte ich wenigstens
a. (einem BRASILIANER mein Haus verkaufen) violates pPF
b. (mein Haus)(einem BRASILIANER verkaufen) violates dat>acc
`If I have to sell my house, i'd at least like to sell my house to a Brasilian.'
The analysis thus provides a comprehensive account of the data under consideration using a simple generalization, combined with ranked violable constraints.