Abstract

This article analyzes mismatches between syntactic and prosodic constituency in Irish and attempts to understand those mismatches in terms of recent proposals about the nature of the syntax-prosody interface. It argues in particular that such mismatches are best understood in terms of Selkirk’s (2011) Match Theory, working in concert with constraints concerned with rhythm and phonological balance. An apparently anomalous rightward movement that seems to target certain pronouns and shift them rightward is shown to be fundamentally a phonological process: a prosodic response to a prosodic dilemma. The article thereby adds to a growing body of evidence for the role of phonological factors in shaping constituent order.

1 Introduction

What are the mechanisms that shape word order in natural language? A traditional and still widely favored answer to that question is that syntax has exclusive responsibility in this domain; in some traditions of investigation, in fact, syntax just is the study of word order. More recently the possibility has emerged, though, that word order is determined postsyntactically, in the process of what Berwick and Chomsky (2011) call ‘‘externalization’’—the translation of the hierarchical and recursive representations characteristic of syntax and semantics into the kinds of serial representations that the sensorimotor systems can manipulate. Given that overall conception, it is natural that some aspects of constituent order should be shaped by demands particular to phonology, and in recent years there have been many studies arguing for the role of phonology in shaping word order.1 We contribute to these debates here by analyzing one aspect of word order in Irish. The phenomenon at the heart of our investigations—the variable placement of pronouns—seems straightforward at first, but it has stubbornly resisted successful analysis despite a 25-year history of investigation. Our goal is to push as far as possible toward an exclusively phonological treatment of the phenomenon, one that involves no reference to any term from syntactic theory. We then assess the viability of that understanding, in terms of descriptive coverage and theoretical integration. Many questions and puzzles will remain in the end, but our core claim is that this purely phonological treatment is the most successful analysis currently available, by both criteria.

One reason why this outcome, if correct, is interesting is that the phenomenon in question (a rightward displacement of certain kinds of pronouns) does not wear its phonological credentials on its sleeve. Its phonological aspects emerge only under fairly close scrutiny. If the case is typical (and we know of no reason to think it atypical), there are probably many similar phenomena awaiting discovery. And in broad terms, the research program that then unfolds is closely consistent with the conjecture of Berwick and Chomsky (2011) that a great deal of the variation found among languages is properly located in systems of externalization.

2 The Phenomenon

2.1 Initial Observations

The phenomenon at the heart of our investigation initially seems strange because it runs counter to certain well-established typological tendencies—the tendencies, in particular, for phonologically light elements (clitics and so on) to displace toward clause-initial position and for phonologically heavy elements to displace to clause-final position. Irish presents a case in which light pronouns displace to the right, sometimes all the way to clause-final position. Despite its typological oddity, pronoun postposing is characteristic of all the Gaelic languages and has been a stable feature of those languages for a thousand years or more. For Irish, the core observations can be made quickly. (1) illustrates the normal position of the object in a finite VSO clause.

(1)

  • Fuair sé nuachtán Meiriceánach óna dheartháir an lá cheana.

  • get.PAST he newspaper American from.his brother the-other-day

  • ‘He got an American newspaper from his brother the other day.’

When the object is a simple pronoun, however, an alternative order is available, in which the object pronoun appears farther to the right than an object really ought to—in clause-final position in (2).2

(2)

graphic

Such displacements often leave pronominal objects quite distant from the verbs that select them.3

(3)

graphic

However, they do not always leave them in absolute clause-final position.

(4)

graphic
graphic

Finally, displacement of the pronoun, though often preferred, is never required. In the examples of (5), for example, the object pronoun appears in the normal position for direct objects, immediately following the subject (we will give many other such examples as we proceed).

(5)

graphic

Putting all of this together, we can summarize the principal puzzle by way of the informal diagram in (6), where the arrows indicate three possible ‘‘trajectories’’ (in a pretheoretical sense) for the displaced pronoun.4

(6)

graphic

The initial challenge then is to understand (6). To understand (6), though, we must first understand certain additional conditions that govern postposing. We lay these out in the next section.

2.2 Additional Conditions

There is in Irish an important distinction between strong and weak forms of personal pronouns. Although not represented in any standard orthography, the difference is crucial, as it turns out, for understanding how pronouns are placed in larger structures. Some of the relevant forms are laid out in (7).5

(7)

graphic

Strong forms of the pronouns can have an accent6 and their vowel nuclei are long; weak forms are unaccented and their vowels are characteristically shortened and centralized. The chart in (7) illustrates (in the final column) fully reduced variants, but unstressed pronouns may have either reduced or unreduced vowels. We return to some of the phonetic details in footnote 40 when more of the relevant material is in place; for now, the crucial observation is that when a pronoun undergoes postposing, it always appears in its weak, unaccented form.

A second important restriction is that weak pronouns never postpose from subject position of a finite clause.7

(8)

graphic

Instead of (8) we find the VSO order of (9), in which the subject pronoun cliticizes to the finite verb.

(9)

graphic

A third important property of pronoun postposing is that it has no consequences for, or correlations with, information structure or discourse function (Bennett, Elfner, and McCloskey 2015; cf. Mulkern 2003, 2011). Many displacements (to the left and right alike) involve the core concepts of discourse and information structure—topic, focus, and the like (see, e.g., Horn 1986).

Pronoun postposing is not such a displacement, and has no detectable pragmatic or semantic effect. In addition to the arguments developed by Doyle (1998:45) for this conclusion, we add two observations of our own. The first suggests that information structure properties of the pronoun itself play no role in postposing. The second suggests that information structure properties of the larger context play no role either.

The first conclusion is suggested by the fact that expletive pronouns postpose freely and under the same conditions as all other pronouns. Each example in (10) has a small clause complement that in turn contains a complement CP. The presence of a complement CP within the small clause licenses an expletive pronoun é ‘it’ as its subject. This pronoun may remain in subject position (leftmost in the small clause); but in each example in (10) it has in fact displaced rightward. (Unlike finite subjects, small clause subjects may undergo pronoun postposing; we discuss this difference in later sections.)

(10)

graphic

Since elements that lack semantic content cannot have information structure content, placement of the pronoun in such cases cannot be linked with the information structure status of the pronoun itself.

The second conclusion is suggested by an interesting kind of natural experiment, one showing that even when we hold the semantic content and context of utterance constant across tokens, we still find variability in pronoun placement. The crucial observations involve certain formulaic announcements broadcast regularly on Raidió na Gaeltachta, a radio network that serves Irish-speaking communities. These are death notices, broadcast as part of the local news for each region. Each notice announces a death and then gives details about funeral arrangements. As part of the 9 a.m. bulletin on Thursday, 24 January 2013, for example, the following two announcements were made in sequence by the same presenter from the studio in Donegal:

(11)

graphic
graphic

Each begins with a future impersonal form of the verb meaning ‘bury’, followed by a nonsubject pronoun and a sequence of temporal and locative modifiers (which are freely ordered with respect to one another). The importance of these observations now lies in the fact that they approach the conditions of a clean natural experiment. The formulaic and repetitive character of the notices comes as close as we are likely to get in natural settings to fixing semantic content and discourse context across utterances. If postposing really depends on such contextual factors, then when they are held constant, we should see a constant outcome. But we do not. Postposing may or may not apply and when it does apply, the displaced pronoun may appear in a range of positions. In (11a), for example, the pronoun postposes across a locative PP and a temporal adverb; in (11b), on the other hand, the pronoun postposes only across a temporal adverb. The variation that is possible here is particularly evident in the larger dataset discussed and analyzed in Bennett, Elfner, and McCloskey 2015, which draws on a collection of 114 such announcements broadcast between 1999 and 2002. In 30 of these, the pronoun was not postposed. Of the 84 instances in which postposing did take place, 10 had the pronoun in absolute final position, and the remaining 74 had it in shifted but nonfinal position. Within this group of 74 ‘‘partial’’ postposings, the pronoun appeared in a range of different positions.

It would strain credulity to maintain that the differences in pronoun placement here reflect aspects of communicative intention or discourse context.8 One would have to hold, for instance, that there was some shift in the discourse context or in the communicative intentions of the speaker between the uttering of (11a) and the uttering of (11b) a few seconds later—a shift, moreover, that was relevant in some way to the positioning of the pronoun. None of this seems plausible. Rather, when listening to a sequence of hundreds of such productions, it is hard not to be struck by the intuition that placement of the pronoun has to do at its core with the rhythmic planning of the utterance. We develop exactly that intuition in what follows. For now, we can bring together our observations so far as follows:

  • (i)

    In Irish, nonsubject pronouns in their weak forms may displace rightward.

  • (ii)

    This displacement may leave the pronoun in absolute clause-final position or in a range of positions between the canonical object position and clause-final position.

  • (iii)

    The displacement has no discernible semantic or pragmatic effect or trigger.

How might we understand all of this?

3 Syntactic Movement

We are by no means the first to tackle the problem of pronoun postposing. The earliest generative treatments were syntactic and attempted in various ways to assimilate postposing to familiar syntactic processes. A rightward-movement analysis was developed in Chung and McCloskey 1987, an article in which the phenomenon of pronoun postposing played a central role. Duffield (1995:66–81) later developed an account that assimilates pronoun postposing to cliticization of the Romance or Germanic type. On this account, the pronoun first undergoes a standard leftward cliticization—movement to a high position in the inflectional layer. Surface order is then accounted for by appeal to remnant movement. Here we focus on the strengths and weaknesses of the rightward-movement analysis. Our goal is to show that no syntactic account is likely to be successful, and the observations and arguments that suggest that conclusion for rightward movement suggest exactly the same conclusion for analyses of the remnant movement type.9 We begin, though, by laying out what the syntactic movement account does well. Doing this will expose some important aspects of the phenomenon that we have not yet touched on and will also give us a set of measures against which we can later assess our own proposals.

3.1 Apparent Head Government Effects

Pronoun postposing exhibits a set of restrictions that are very reminiscent of the head government requirement of the Empty Category Principle (ECP). To see this, observe first that not all postposings take place from object position. In the right circumstances, one can postpose subjects of complement small clauses (on which, see Chung and McCloskey 1987, McCloskey 2014). (12) illustrates a small clause complement (with accusative subject) to the unaccusative verb tarla ‘happen’. Here, postposing is freely possible.

(12)

graphic
  • since happen.PAST married on.him with more and twenty years her

  • ‘since she happened to have been married to him for more than twenty years’

  • (DCA 186)

Small clauses with accusative subjects also appear freely in discourse isolation, as (13a–b) illustrate.

(13)

  • a.

    graphic

    be.PAST the archbishop PROG come him dressed resplendently

    ‘The archbishop was coming. He (was) dressed resplendently.’

    (SR 19)

  • b.

    graphic

In such cases, in contrast with the complement small clauses of (12), postposing is unavailable.

(14)

graphic

An additional restriction is that there is no postposing from the position following the marker of negation in small clauses (gan), whether that clause is a complement (15) or stands in discourse isolation (16).

(15)

graphic

(16)

graphic

This pattern seems to reflect a deeper generalization: there is apparently no postposing from subject position in small clause complements of functional heads (as opposed to complements of lexical heads like tarla ‘happen’). We can see this most clearly in a kind of absolutive construction with agus ‘and’ illustrated in (17).

(17)

graphic

What all this suggests (or at least suggested to Chung and McCloskey (1987)) is that postposing is allowed in the contexts schematized in (18).

(18)

graphic

That is: a weak pronoun may postpose only if it is the complement of a lexical head or the specifier of the complement of a lexical head. This pattern was an expected one in the theoretical context of the time. It was natural to maintain that pronoun postposing, being a routine syntactic movement involving right-adjunction to a containing category, should be subject to the ECP. The ECP requires that the origin site of movement be governed by a lexical (open-class) head, where we can understand government as in (19) (see, e.g., Chomsky and Lasnik 1993).

(19) A head H governs α iff α is (the specifier of ) the complement of H.

With these assumptions in place, we can understand the contrasts documented above: the marker of negation is not a lexical head, nor is the element agus. In the case of root small clauses, there is no candidate lexical governor at all for the trace of pronoun postposing. Put another way, this account assimilates pronoun postposing to other well-studied rightward movements such as heavy NP shift, which exhibit a similar array of sensitivities (as documented especially in Rizzi 1990:chap. 1).

4 Grounds for Skepticism

Despite these successes, there are reasons to be skeptical about the rightward-movement analysis. There are, to begin with, theoretical concerns. One of the goals of the Minimalist Program for syntax, which we take to be an impressively successful enterprise, is to eliminate appeal to the ECP and to the relation of ‘‘government’’ upon which it depends. In addition, the rightward-movement account faces substantial empirical difficulties. In sections 4.14.3, we lay out those challenges, with an eye to establishing later that the phonological account we develop does better at meeting them. If we can build an account of pronoun postposing that involves no appeal to the notion of government but can match or exceed the head government account in descriptive coverage, we contribute to the goal of eliminating technological bloat from syntactic theory while still meeting fundamental descriptive responsibilities.

4.1 Postposing in the Absence of a Lexical Governor

The principal strength of the account developed by Chung and McCloskey (1987) is the understanding that it makes available of the restrictions on postposing just summarized. But in this lies also one of its principal weaknesses. Problems become apparent when additional cases are considered—cases in which small clauses appear as complements to a different set of functional heads.

(20)

  • a.

    [seo [DP XP]]

  • b.

    Seo na saighdiúirí ag teacht.

    DEMON the soldiers PROG come

    ‘Here come the soldiers.’

The element seo is a demonstrative particle and is proximate in its interpretation (see McCloskey 2004 for some discussion). In the use illustrated in (20), it takes a small clause complement (DP in composition with a predicative XP) and the interpretation is presentative—a deictic gesture toward a scene or state of affairs described by the small clause. Although seo is clearly a closed-class element and so should not qualify as a head governor, postposing applies freely from the subject position of its complement.

(21)

graphic

A further particularity of the language is that it allows small clause wh-questions (McCloskey 2011b).

(22)

  • a.

    Cá fhad Mac Alastair marbh anois?

    what long McAllister dead now

    ‘How long has McAllister been dead now?’

    (TGC 103)

  • b.

    Cén aois é?

    what age him

    ‘How old is he?’

  • c.

    Cad fá laghad mo mheas orthu?

    what reason smallness my respect on.them

    ‘Why do I respect them so little?’

    (DO 41)

  • d.

    Cé mhéad scéal ar fad agat?

    what amount story in-all at.you

    ‘How many stories in all do you have?’

    (MSN 271)

Pronoun postposing from the subject position of such small clauses is free.

(23)

graphic

Whatever one assumes about the syntax here ( presumably a CP superstructure above the predicative core of the small clause), there is no plausible head governor in such cases to license a trace created by movement of the pronoun. Although there may well be silent structural projections between the small clause and the CP layer in such cases, there is no reason to believe that such null functional heads could ‘‘lexically govern’’ the origin site of postposing.10 If there really were a head government requirement on postposing, then, it should fail in contexts like (23a–d). Such worries are amplified by the observations in the next section, where we consider the range of positions in which the displaced pronoun may ultimately appear.

4.2 Implausible Landing Sites

Postposed pronouns often surface in positions that are, from a syntactic perspective, deeply implausible as landing sites for movement. For the most part, the relevant cases involve small clauses with pronominal subjects.11 In (24), for instance, we have small clause complements to perception verbs, in which a pronominal subject has been postponed. We have seen this phenomenon already; what is striking about cases like (24a–c) is that the pronoun postposes to an intermediate position rather than clause-final position. In fact, it appears following the verb of which it is the subject, but preceding its complement.

(24)

graphic

Interpreted as syntactic movement, this would apparently involve a lowering of the pronoun into the middle of the predicate of the small clause, as illustrated in (25), where the arrows indicate postulated head movements and the black triangle indicates the linear position of the subject pronoun.

(25)

graphic

Example (21d), repeated here as (26), makes the same point in a slightly different way.

(26)

  • graphic

  • DEMONPROG put down to.me REFLEX him on the way C be.COND he with.her

  • ‘Here he goes describing to me how he would be with her.’

  • (CLENS 15)

Here the small clause subject is postposed, as in (24), but this time it appears between two complements—another apparent lowering. (27) is of the same general form and also shows postposing of the small clause subject.

(27)

  • graphic

  • I.saw PROG look on.her him lasciviously

  • ‘I saw him looking at her lasciviously.’

  • (LG 314)

In this case, though, the subject pronoun appears to the left of a manner adverb that modifies the VP of the small clause complement. But that is in turn the VP of which the pronoun, in syntactic and semantic terms, is the subject. The relevant structure is (28), where the pronoun is once again shown in its syntactically expected position and the black triangle indicates its actual position in the pronounced string.

(28)

graphic

Construed as a rightward syntactic movement, this is again a lowering into the VP. Putting this together with the observations of (24), we conclude that pronoun postposing (in these instances) is a rightward lowering that sometimes places the lowered element among the complements of V and sometimes places it at the right edge of VP.

Of course, for all such cases, one might appeal to rightward extraposition of some XP around an already postposed pronoun; but there is little independent reason to believe in the required extrapositions, and for the final case (the manner adverbial go drúisiúil ‘lasciviously’ in (27)) the solution is particularly implausible, since there is no evidence that we know of that such adverbs are liable to extraposition.

Cases like (29) are more challenging still from a syntactic perspective. Here, the pronoun (the subject of the small clause complement to the predicate cuma ‘no matter’) again undergoes postposing.

(29)

  • graphic

  • COP.PRES no.matter PRED summer it or PRED winter

  • ‘It doesn’t matter whether it’s summer or winter.’

  • (U 20)

We can assume a syntactic structure along the lines of (30).12

(30)

graphic

In (29), the postposed pronoun appears inside the predicate of the small clause, another apparent lowering. Worse, though, is the fact that the pronoun appears in the middle of a disjunction— following the first disjunct but preceding the second. We know of no believable syntactic treatment of such cases. In particular, we know of no syntactic movement that can break a disjoined phrase into its component parts.13

Let us observe finally that pronoun postposing cannot always be construed in syntactic terms as a lowering. Consider, for instance, basic cases like (31a–b).

(31)

graphic

In such cases, the object pronoun appears to the right of a sequence of adverbial phrases that modify the VP in which the pronoun originates. If pronoun postposing is a syntactic movement, such displacements must be construed as raisings rather than lowerings (VP adjuncts being syntactically higher than the ‘‘core’’ VP that contains the base position of the object pronoun). What these observations jointly suggest is that if pronoun postposing is a syntactic movement, it is an exotic and ill-behaved one, blind to the hierarchical relations that are at the heart of syntactic computation.

We cannot say that it would be impossible to develop a syntactic account of these observations. For example, in a framework committed to leftward-only syntactic movement and no post-syntactic fixing of precedence relations (Duffield 1995, Kayne 2000), various kinds of remnant movement might be able to derive the unusual constituencies that postposing is sensitive to. The challenge would be to make such an analysis something more than an exercise in self-confirmation. In particular, given section 2.2 above and the more detailed treatment of the same issues in Bennett, Elfner, and McCloskey 2015, the movements that would be needed cannot be driven by factors or features grounded in the needs of discourse or information structure. Such movements would also need to be tightly constrained to avoid generating illicit cases of postposing (compare, e.g., (17b) and (23d)). We will not take up the challenge of constructing or assessing such an alternative here.

4.3 Prosodic Correlations

The observations of sections 4.1 and 4.2 take on added force when we observe that the rightward-movement analysis provides no obvious way of understanding the web of correlations with prosodic factors that is a central aspect of the postposing puzzle. We have already noted one aspect of that pattern, namely, that only weak (i.e., prosodically dependent) pronouns may undergo postposing. But the correlations go further. A particularly striking point concerns cases in which postposing may not apply. Three such examples are gathered in (32).

(32)

graphic

(32) illustrates three positions from which pronoun postposing is impossible: subject position of a nonfinite clause (32a), subject position of a small clause complement to a functional head (32b), and subject position of a root small clause (32c). A crucial property of such cases is that, in the absence of postposing, the unshifted pronoun may not be weak; instead, it must be pronounced in its strong form—accented and with a long vowel. This is true even when, as in (32c), the pronoun is expletive and questions of focus, emphasis, and so on are necessarily irrelevant (see Lucas 1979:94, §358(i) and Ó Baoill 1996:90 for this observation with respect to agus ‘and’). These phonological correlates form a central element of the general pattern, and they need to be integrated smoothly into the ultimate account of pronoun postposing.

5 Interim Conclusions

We can summarize our conclusions so far as follows. The syntactic movement analysis is successful in various ways, but it is also incomplete and troubling from the perspective of the syntactician. It is incomplete because it leaves unexplained those cases in which postposing unexpectedly succeeds; it is also incomplete in that it does not, in any of its current forms, provide a way of understanding a central aspect of the phenomenon—how it interacts with prosodic factors. It is troubling in that the range of positions in which the postposed pronoun may appear is very much at odds with reasonable expectations about syntactic constituency and about how constituency interacts with movement. It also has as its core a piece of theoretical machinery (the head government clause of the ECP) whose legitimacy is dubious. It makes sense, then, to explore alternatives. And since prosodic factors seem to be at the heart of the phenomenon, it makes sense to place those factors at the heart of the analysis. We do that here, building an analysis that pushes as far as possible toward a purely prosodic understanding of pronoun postposing. We then evaluate how successful that push has been.

Of course, in establishing a correlation between prosodic factors and positioning of the pronoun, we do not thereby establish that a prosodic account is superior to a syntactic account. Such correlations might emerge from the interaction between syntactic and other factors. The question to ask is what analysis yields a deeper and better-integrated understanding of the observations. How could we make that assessment? A successful analysis, we think, should have the following properties. At the empirical level it should

  • (i)

    deal well with the basic cases,

  • (ii)

    integrate the prosodic correlates of pronoun placement naturally,

  • (iii)

    let us understand why there are no pragmatic or discourse effects,

  • (iv)

    let us understand the optionality of postposing,

  • (v)

    provide an understanding of the range of positions in which displaced pronouns may appear,

  • (vi)

    incorporate whatever is right about the head government condition, and

  • (vii)

    gracefully incorporate those cases that pose difficulties for it.

    At the theoretical level, it should accomplish at least the following:

  • (viii)

    It should be well-integrated with a reasonable theory of how prosodic structure is built (in Irish and in general). In particular, the theory of prosody appealed to should have solid independent grounding rather than being tailored to the needs of the problem at hand.

  • (ix)

    It should be well-integrated with a reasonable theory of how constituent order is determined (in Irish and in general).

With these criteria in mind, we move on to our own proposals.

6 Prosody and Syntax

The intuition that drives our analysis is that pronoun postposing emerges from an interplay between principles of rhythmic structuring and the phonological properties of pronouns. In working this intuition out, our analysis situates pronoun placement squarely among the mechanisms that build prosodic structures. We thus develop a line of analysis that goes back to work by Adger (1997, 2007a) and has been explored by several others—Doyle (1998), McCloskey (1999), and Elfner (2011a, 2012). Our goal, though, is to go farther than previous proposals in meeting the desiderata laid out in the previous section. We begin by outlining the theoretical assumptions that we bring to these tasks.

6.1 Theoretical Background

The background that we assume (concerning prosodic structure and its relation to syntax) is relatively conventional and is informed by a great deal of work done over the past three decades. We take it as given that there is hierarchical structure in phonological representations, and furthermore that the relationship between syntactic constituency and prosodic constituency is not arbitrary. We also assume, as is conventional, that the correspondence between syntactic and prosodic representations is partial and imperfect.14 Understanding the syntax-prosody relationship, then, is a matter of understanding what the mechanisms are that guarantee nonarbitrary correspondence, but also what the mechanisms are that render the correspondence partial and opaque.

To begin, we take from recent work by Ito and Mester (especially 2012, 2013) the view that there are just three prosodic constituent types above the level of the word (we will have little occasion here to deal with prosodic organization below the level of the word). In order of inclusiveness, they are as follows:

(33) The hierarchy of prosodic constituents

  • a.

    The prosodic word (ω)

  • b.

    The phonological phrase (Φ)

  • c.

    The intonational phrase (ι)

From Selkirk 2009, 2011, we adopt the core mapping principles in (34), which assume the theory of bare phrase structure (Chomsky 1995a and subsequent work) as their syntactic backend.

(34) Core mapping principles

  • a.

    MATCHWORD

    Prosodic words correspond to the heads from which phrases are projected in the syntax (heads that will often have a complex internal structure determined by head movement).

  • b.

    MATCHPHRASE

    Phonological phrases correspond to maximal projections in the syntax.

  • c.

    MATCHCLAUSE

    Intonational phrases correspond to those clausal projections that have the potential to express illocutionary force (assertoric or interrogative force, for instance).15

The mapping constraints of (34), unadorned, predict a closer relation between syntactic and prosodic constituency than we in fact find—in Irish in particular. The sources of disparity are both general and language-particular. Here, we review the general mechanisms behind such ‘‘imperfect’’ mappings. The language-particular effects will be discussed in the following sections, with the phenomena that motivate them. We begin by being more precise about how we understand MATCH PHRASE, adopting from Elfner 2012:28 the formulation in (35).

(35)

  • MATCHPHRASE

  • Given a maximal projection XP in a syntactic representation S, where XP dominates all and only the set of terminal elements {a, b, c, . . . , n}, there must be in the phonological representation P corresponding to S a Φ-phrase that includes all and only the phonological exponents of a, b, c, . . . , n.

For simple cases, (35) has the expected and familiar consequences, giving rise to syntax-prosody correspondences like those in (36), where n′ indicates the phonological exponent of the terminal element n.

(36)

graphic

But now consider cases in which one of the terminal elements has no phonological exponent—if, for example, a of (36) were a trace of head movement. The predicted outcome in that case is (37).

(37)

graphic

The formulation of MATCH PHRASE in (35) licenses a ‘‘flattening’’ of syntactic structure in the sense that in the prosodic representation of (37b) a single Φ-phrase does double duty, as a correspondent of XP that dominates the phonological exponents of b and c, and as a correspondent of YP that does the same. Two maximal projections in syntax ‘‘collapse,’’ to speak metaphorically, into a single Φ-phrase in phonology. We will shortly examine cases in which this abstract scheme does vital empirical work.16

We also adopt two general conventions for interpreting the constraints in (34), both of which eliminate nonbranching structure from prosodic representations.

In its present form, the system of constraints in (34) underdetermines the outcome of syntaxprosody mappings in certain circumstances. The first such circumstance involves maximal clauses. A CP that carries illocutionary force should be mapped to an ι-phrase by the principle MATCH CLAUSE. But CPs are also maximal projections, and so should be mapped to a Φ-phrase by MATCH PHRASE. A similar issue arises for syntactic elements that are simultaneously maximal and minimal (in the system of bare phrase structure, which we assume here). Pronouns (members of the category D that take no complement) are in this category. They are minimal because they contain no proper subparts, maximal because they pass on no label to containing expressions. Given the mapping principles of (34), then, a pronoun should map to a prosodic word (being a minimal category D), but should also map to a Φ-phrase (being a maximal category DP). The same ambiguity holds for PPs that consist only of a prepositional head P, such as air ‘on him’.

We assume that such equivocations are not tolerated. Specifically, the principles that link lexical items with prosodic words and certain clause types with ι-phrases take priority over MATCH PHRASE. As a consequence, pronouns will preferably map to prosodic words (as will, for example, PPs that consist only of a prepositional head), and root CPs will preferably map to ι-phrases, not to Φ-phrases.17

This leaves open the possibility that a single syntactic constituent (say, a CP) could correspond to two independent but nested prosodic constituents (say, a Φ-phrase immediately dominated by an ι-phrase). Indeed, the mapping principles in (34), left to their own devices, will give rise to many cases in which a Φ-phrase consists exclusively of a phonological word. Consider the Irish DP bean, for instance, which means ‘a woman’ and which includes, on most understandings, a null indefinite determiner. The syntactic representation for such a phrase will be either (38a) (assuming X-bar theory) or (38b) (assuming bare phrase structure).

(38)

graphic

In (38b), the noun bean is simultaneously maximal and minimal in syntactic terms. We might then expect the prosodic representation in (39a).

(39)

graphic

We suggest that such nonbranching structures are not tolerated in prosodic representations (in Irish at least) and that in all such cases only the prosodic word is projected (39b), not the higher Φ-phrase (39a). Notice that this is another ‘‘flattening’’ of prosodic structure by comparison with syntactic representations.18

More important for the particulars of our account, though, will be the (conventional) claim that the syntax-prosody mapping is also rendered opaque because prosodic phrasings are optimized to meet certain purely phonological desiderata that have no parallel in syntax (Nespor and Vogel 1986, Inkelas and Zec 1995, Selkirk 2000, 2011). Such ‘‘optimizing distortions’’ may involve both hierarchical organization and linear order, and they will be crucial for our proposal about pronoun postposing. We will consider them in the following section as we introduce the Irish data that motivate them.

There are a number of ways in which our proposals might be integrated into a larger theoretical framework. Here we assume that the unordered hierarchical representations provided by syntax are subject to a once-off optimization—one in which matching constraints, constraints governing hierarchical aspects of prosodic structure, and purely phonological constraints conspire and compete simultaneously and in parallel to determine optimal outputs (as in Optimality Theory; Prince and Smolensky 1993, Selkirk 2011). It is an important property of that overall framework that for a given syntactic input there need not be a unique prosodic outcome. Given the variability in phrasing evident in the Irish data (which we outline in the next section), this is a welcome consequence. For our own core concerns, what will be crucial at almost every point is that a certain phrasing pattern be possible, not that it be the only phrasing allowed.

6.2 Prosodic Structure in Irish: A First Sketch

Given this background, consider now what prosodic structure we would expect for a simple finite clause in Irish. We assume the syntactic analysis schematized in (40).

(40)

graphic

In (40), the ‘‘inflected verb’’ is a fusion of at least four syntactic atoms: a verbal stem, a light verb v, a specification of tense, and a specification of polarity (Σ). The subject begins life inside the vP, in which it is thematically licensed, but raises to a position immediately below the expression of polarity. In this way, VSO order is derived (Chung and McCloskey 1987, McCloskey 1991, 1996b, 2011b,c).

What do we now expect as the prosodic structure of a VSO clause? Consider (41), which is just (40) with phonologically null elements (e.g., traces) removed. The boxed element at the left represents the complex inflected verb formed by successive applications of head movement through the extended projection of V.19

(41)

graphic

What should emerge from the syntax of (41) is the prosody of (42).20

(42)

graphic

In (42), the inflected verb forms a prosodic word w and phrases by itself; there are then Φ-phrase s corresponding to the subject (Φ3 ), to the complement (Φ4 ), and also to the constituent that includes the sequence of subject and complement (Φ2 ).

This is in fact an attested pattern, as confirmed both by controlled production studies and by observation of naturally occurring data. Elfner (2011b, 2012, 2013, 2015) identifies two pitch accents that function as boundary markers in Conamara dialects: one rising (L-H) and one falling (H-L). In (43), for example, there are L-H accents on the first stressed syllable of the verb and the first stressed syllable of the subject DP. There is also an H-L accent on the final stressed syllable of the subject and of the object.

(43)

  • Díolfaidh leabharlannaí dathuíl blathanna áille.

  • sell.FUT librarian attractive flowers beautiful

  • ‘An attractive librarian will sell beautiful flowers.’

This distributional pattern can be understood in terms of the prosodic structure in (44) ( predicted by the matching principles) and the mechanisms in (45).

(44)

graphic

(44) incorporates an extra annotation, distinguishing those Φ-phrase s that are minimal (dominating no other Φ-phrase ) and those that are not. The specific hypothesis developed and defended in Elfner 2012 is that in (45).

(45)

  • a.

    L-H accents associate with the stressed syllable of the leftmost prosodic word of all nonminimal Φ-phrases.

  • b.

    H-L accents associate with the stressed syllable of the final prosodic word of all Φ-phrases.

This pair of hypotheses yields an understanding of the distribution of accents in transitive clauses such as (43). The inflected verb hosts an L-H accent because it is at the left edge of a nonminimal (in fact, maximal) Φ-phrase ; the first word of the subject DP similarly hosts an L-H accent, being initial in the nonminimal Φ-phrase that includes both the subject and the object; the final word of the subject DP hosts an H-L accent because it is at the right edge of a Φ-phrase ; and the final word of the object DP hosts an H-L accent because it too is at the right edge of a Φ-phrase (three such phrases, in fact). For further arguments and exemplification, including extensions to much more complex syntactic structures than (43), see Elfner 2012:esp. chaps. 2, 3.

What is encouraging about these results is that the phrasing they imply is consistent with earlier work on prosodic phrasing in a different dialect (Donegal Irish) by Bennett (2008), a study that relied on the distribution of pauses rather than on the distribution of accents. The phrasing in (44), for instance, is also signaled by the presence of sometimes quite long pauses separating the finite verb from the subject, as in the examples of (46), with their associated sound files (‘‘||’’ indicates a pause).21

(46)

graphic

The prosodic structure in (42)/(44) closely mirrors the corresponding syntax. But this is not the only possible outcome for a VSO clause. Also very common are phrasings in which the verb and the subject together form a prosodic constituent, to the exclusion of other material (an example is given in (52) below). Why should such departures from the ‘‘ideal’’ of full syntax-prosody isomorphism be common? They are common, we maintain, because they reflect the activity in the language of constraints governing eurhythmy.

A large body of research demonstrates that optimal prosodic structures conform to (47).22

(47)

  • BINARITY

  • Optimal prosodic constituents are binary-branching.

It follows from (47) that all of the following phrase types satisfy binarity requirements equally well:

  • a Φ-phrase having two Φ-phrase s as immediate constituents,

  • a Φ-phrase having two prosodic words as immediate constituents,

  • a Φ-phrase having a prosodic word and a Φ-phrase as immediate constituents.

However, we also take from Revithiadou (2004), Revithiadou and Spyropoulus (2009), and others the idea that optimal prosodic constituents are balanced, being roughly equal in length and having as subconstituents phrases of the same type. Myrberg (2010, 2013), for example, proposes the constraint in (48).

(48)

  • EQUALSISTERS

  • Sister nodes in prosodic structure should be instantiations of the same prosodic category.

In combination with BINARITY, the EQUAL SISTERS constraint suggests a ranking for the three prosodic structures in (49).

(49)

graphic

(49a) satisfies neither BINARITY nor EQUAL SISTERS. (49b) satisfies BINARITY, but not EQUAL SISTERS. (49c) satisfies both constraints. All other things being equal, then, (49c) should be favored over (49b), which should in turn be favored over (49a).

Both of these constraints have behind them a long history of thinking about the components of eurhythmy in natural language (among many others, see Nespor and Vogel 1986, Gussenhoven 1991, Ghini 1993, Tilsen 2011, to appear, and references there). Recent work has even suggested a functional basis for the structural parallelisms enforced by EQUAL SISTERS (Krivokapić 2007). Revithiadou (2004) and Revithiadou and Spyropoulus (2009) have further argued that the influence of phonological balancing constraints can lead to quite radical disparities between syntactic and prosodic constituency (see also Nespor and Vogel 1986, Selkirk 2011:sec. 3.1). This will be important in what follows.

Structures such as (42)/(44), which fully satisfy MATCH PHRASE, will always encompass a violation of EQUAL SISTERS at the level of the topmost Φ-phrase . Given the syntax in (41) and the mapping principles in (34), this much is inevitable; the finite verb (which must correspond to a prosodic word) will always have as its sister a maximal projection (which must correspond to a Φ-phrase ), making a violation of EQUAL SISTERS inescapable. It is this fact, we argue, that makes departures from the ideal of (42) so frequent. In considering the alternatives that in fact arise, we can begin with cases like (50), in which the subject consists of a single prosodic word.

(50)

  • Cheannaigh múinteoirí málaí bána.

  • buy.PAST teachers bags white

  • ‘Teachers bought white bags.’

For such cases (discussed in detail in Elfner 2012:chap. 4), the mapping principles of (34) and (35) will yield the structure in (51).

(51)

graphic

Such a structure involves two violations of EQUAL SISTERS (with respect to Φ1 and with respect to Φ2 ). What actually emerges in such a case is (52), in which the subject is phrased with the finite verb, and in which both violations of EQUAL SISTERS are thereby eliminated. (For detailed discussion and evidence, see Elfner 2012:chap. 4.)

(52)

graphic

The phrasing in (52) is intuitively accurate; more importantly, it predicts, in combination with (45), the empirically correct distribution of pitch accents in (50) (which we have marked in (52)). The same pattern can clearly be discerned in the naturally occurring examples of (53), with their associated sound files.

(53)

graphic

Notice in particular the often dramatically long pauses that follow the initial Φ-phrase in such examples. The distribution of accents in the first clause of (53b) is also consistent with the analysis developed in Elfner 2012 and discussed in section 6.2 above: L-H on the initial verb, H-L on the noun mílte.23

The rebracketing in (52) thus emerges from a negotiation between purely prosodic pressures (EQUAL SISTERS) and the mapping principles that govern syntax-prosody correspondences (MATCH PHRASE). In the terms of Optimality Theory ( Prince and Smolensky 1993), EQUAL SISTERS (ES) outranks MATCH PHRASE (MP). The prioritization of a strictly phonological demand thus derives a pattern of prosodic constituency that only partially resembles the syntactic representation.24

(54)

graphic

When, however, rephrasings such as those to be heard in (53) are less beneficial (when the subject is, for instance, syntactically and prosodically substantial, as in (46b)), there is a greater chance that the phrasing predicted by the mapping principles alone will actually emerge. See Elfner 2012 for a related model that engages more directly with the observed frequencies of each variant.

This is the foundation upon which we will build an understanding of postposing. Viewed in this light, postposing will emerge as another eurhythmic effect, reflecting the mechanisms we have described here, in interaction with another, arguably related constraint that we introduce in the next section. As we develop this argument, we will be guided by the methodology that brought us this far—relying in part on careful instrumental investigation of the kind found in Elfner 2012, in part on theoretical deduction (in particular, on the principle that similar cases should be treated in similar ways), and in part on impressionistic observation of naturally occurring data. We take it that each of these aspects of the investigation should inform and refine each other aspect.

7 Postposing as Prosodic Repair

7.1 The Core Proposal

It is an unavoidable consequence of the proposals laid out so far that in a finite clause, the direct object will always be placed, by the default linearization routines, at the left edge of at least one Φ-phrase : the one corresponding to the verbal projection that immediately contains it.25 This conclusion holds under all of the analyses of objecthood in Irish that we know of: if the object is a complement of V, if (as in Chomsky 2008) it raises to the specifier of V, or if (as in Bobaljik and Carnie 1996) it undergoes full object shift, moving to the specifier position of v (as is clearly the case in nonfinite clauses). In the first two variants, the relevant Φ-phrase will correspond to VP; in the third, it will correspond to vP. Our working assumption here will be that the third approach is correct, and that will matter at certain points; our core proposal, however, can be understood in the context of any of these theories.

The observation of the previous paragraph is an important one, because the left edge of a Φ-phrase is a position of prosodic strength, inhospitable to weak elements. If the direct object is realized as a pronoun in its weak form, then the result will be a flawed phonological object—one in which a dependent and accentless element occupies a position reserved for the prosodically strong. Our core proposal is that pronoun postposing is a repair for that prosodic imperfection,26 and furthermore that it is just one among a range of available repairs.

Let us be more specific. The phonological constraint that we take to be at the heart of these interactions is called STRONGSTART by Selkirk (2011). This is a constraint that penalizes elements at the left edge of a phrase, if they are relatively weak. There are various ways in which the constraint might be formalized, but we will work here with the version in (55).27

(55)

  • STRONGSTART

  • Prosodic constituents above the level of the word should not have at their left edge an immediate subconstituent that is prosodically dependent. For our purposes here, a ‘‘prosodically dependent’’ constituent is any prosodic unit smaller than the word.

We take it that STRONG START is one of a family of constraints that conspire to reward prosodic phrasings that are evenly balanced (note its similarity to Myrberg’s (2010) EQUAL SISTERS discussed above).28 There could well be separate (or gradiently violated) constraints that differ based on the number of levels of separation between the left and right sisters (structures in which the left sister is two steps lower than the right being less favored than structures in which there is just a one-step difference between the sisters). Our larger goals can be met in the context of a number of different understandings of STRONG START and its place in the firmament of prosodic constraints, but we will work here with the specific formulation in (55).

As we have illustrated, objects in VSO clauses will always be exposed at the left edge of a Φ-phrase—the Φ-phrase corresponding to vP, in fact—on the assumptions we are currently working with. It follows in turn that if the object is a weak pronoun (by definition less than a prosodic word), then a violation of STRONG START will result. If, however, such a pronoun were instead to appear at the right edge of the containing Φ-phrase , as in (56b), STRONG START would not be violated.

(56) Pronoun postposing (initial)

graphic

(56) assumes that postposed pronouns are right-adjoined at the level of the Φ-phrase . The core of our proposal could, we think, be maintained with different assumptions, but we assume adjunction here for several reasons. First, Elfner (2012:224) presents evidence based on the distribution of pitch accents that prosodic adjunction is the right interpretation of the positioning of the pronoun. Second, we take from recent work by Ito and Mester (e.g., 2006, 2009b, 2012) the idea that prosodic adjunction has a particularly central role in constructing optimal phonological representations.

We adopt an understanding of adjunction that is widespread in research on prosodic phonology, though sometimes only implicitly. In our usage, any structure of the form in (57) counts as an instance of adjunction, with K ranging over levels of the prosodic hierarchy (the linear order of K and K – n is irrelevant; see also Ito and Mester 2006, 2009a, 2012, Myrberg 2010, Elfner 2012:134–146, Sabbagh 2013).

(57)

graphic

One consequence of this interpretation is that adjunction will be ubiquitous in prosodic representations by comparison with syntax. A syntactic structure like (58a) will have the prosodic representation in (58b).

(58)

graphic

In the syntactic representation (58a), a. is not adjoined; in its prosodic counterpart (58b), however, α′ is right-adjoined to a complex two-segment Φ-phrase , the crucial syntactic distinction between XP and YP being lost, so to speak, in the translation from syntax to prosody. It will become clearer as we proceed why these commitments matter and what role they play in our analysis. For now, we just need to be clear about how this interpretation interacts with other terms we will be using. We assume (59).29

(59)

  • a.

    A category C dominates α iff α is contained within all of the segments of C.

  • b.

    A category C includes α iff α is contained within at least one segment of C.

In (58b), then, α′ is included in Φ, but not dominated by it. This structure is exactly mirrored in the schematic outline of postposing, (56b). There, the displaced pronoun is still included in Φ, despite having shifted away from its ‘‘base’’ position lower in the structure. We emphasize the distinction between inclusion and domination here because our understanding of MATCH PHRASE (35), and its interaction with postposing, hinges on this very point. We expand on these matters in sections 7.2 and 8.3.

With this much as background, we can illustrate the basics of our proposal. Consider (60), for which we expect the prosodic structure in (61), which presupposes the rebracketing of verb and subject discussed in the previous section.

(60)

graphic
  • bring.PAST my mother as-far-as house the.GEN school.GEN him

  • ‘My mother brought him as far as the school.’

(61)

graphic

This, in a nutshell, is pronoun postposing. It is the system’s response to the possibility of prosodic structures that are flawed in incorporating a violation of STRONG START. But there can be more than one way to repair structures that are rhythmically flawed. And in fact for contemporary varieties of Irish, there seem to be at least three distinct ways to resolve potential violations of STRONG START.

(62)

  • Option A

    Postpose the pronoun so that it appears at the right edge rather than at the left edge of a Φ-phrase —(56).

  • Option B

    Leave the pronoun in its syntactically expected position, but cliticize it to a preceding word or phrase, thereby removing it from the left edge of the Φ-phrase and avoiding a violation of STRONG START.

  • Option C

    Parse the pronoun as a prosodic word, in which case it is accented, no violation of STRONG START is incurred, and no repair is motivated.

Given the options of (62), what we have called the ‘‘optionality’’ of pronoun postposing (an inaccurate term, as it now turns out) in fact reflects the availability of three alternative repairs, only one of which involves displacement of the offending pronoun from its expected position. (63) is possible beside (60).

(63)

  • graphic

  • bring.PAST my mother him as-far-as house the.GEN school.GEN

  • ‘My mother brought him as far as the school.’

But the single orthographic form in (63) masks two distinct pronunciations, depending on whether the pronoun is accented (Option C) or in its weak form, enclitic on the preceding prosodic constituent (Option B).

(64)

graphic

The same options can be seen in the spontaneously produced examples of (65) and (66) with their associated sound files: Option B is at play in (65); Option C in (66).

(65)

graphic

(66)

graphic

As the data show, then, the three options are freely available in principle.30

7.2 The Constraints and Their Ranking

In implementing the analysis, we will assume an additional constraint, which we will name NO SHIFT, whose effect is to require that precedence relations in phonological representations should be isomorphic to the linear ordering of terminal elements determined from the syntactic representation.

(67)

  • NOSHIFT

  • If a terminal element α is linearly ordered before a terminal element β in the syntactic representation of an expression E, then the phonological exponent of α should precede the phonological exponent of β in the phonological representation of E.

The interpretation of this constraint is intended to be neutral with respect to the various ways in which linear order might be ‘‘read off’’ syntactic structure (see Kayne 1994, Frampton 2004, Fox and Pesetsky 2005, López 2009, and Elfner 2012 for more specific implementations). What matters for our purposes is that NO SHIFT penalizes pronoun postposing, and any other ‘‘phonological’’ displacements that result in orderings that are not faithful to those determined by syntactic linearization routines.

We also assume that the choice between strong and weak forms of pronouns is a free lexical choice and that different selections therefore give rise to distinct inputs (see Selkirk 1996:203–206, Anderson 2005, and Zec 2005 on English auxiliaries). As a consequence, if the strong form of a pronominal object is selected, no violation of STRONG START can be triggered and no repair will be required or possible. This much is more a decision of convenience than a decision of principle. While it would surely be possible to regulate this variation with grammatical devices (e.g., Mascaró’s (2007) PRIORITY constraint), nothing at present forces us down that more elaborate path (though some relevant issues will come up in section 8.5).

We will illustrate the workings of the system by way of example (60), repeated here as (68).

(68)

graphic

If a strong form of the pronoun is selected, we have the tableau in (69), in which, as throughout, we use the notation {ω D} to indicate a pronoun in its strong form.31

(69)

graphic

In the absence of a violation of STRONG START, the effect of NO SHIFT is to ensure a transparent mapping from linear order in syntax to precedence in prosodic representations. However, if the weak form of the object pronoun is selected, there are more possibilities to consider, as illustrated in (70).32

(70)

graphic

The high ranking of STRONG START in Irish (above MATCH PHRASE and NO SHIFT) eliminates the option in which the offending structure is not repaired. The ultimate outcome must therefore involve either Option A ( postposing) or Option B (leftward enclisis). These options are similar in that both represent ‘‘solutions’’ to the problem of STRONG START. However, they also differ crucially: Option A ( postposing) violates NO SHIFT but satisfies MATCH PHRASE, while Option B (leftward enclisis) violates MATCH PHRASE but satisfies NO SHIFT. This is probably clear for Option B, but some clarification is perhaps in order for Option A. We have assumed the definition in (35), repeated here as (71).

(71)

  • MATCHPHRASE

  • Given a maximal projection XP in a syntactic representation S, where XP dominates all and only the set of terminal elements {a, b, c, . . . , n}, there must be in the phonological representation P corresponding to S a Φ-phrase that includes all and only the phonological exponents of a, b, c, . . . , n.

Given (71), and given that we have assumed that postposing involves adjunction, it follows that both structures of (56) (our schematic representation of postposing) satisfy (71) equally well. Structure (56b) involves a single complex (two-segment) category Φ. Given the understanding of prosodic adjunction laid out at (59b), that complex category ‘‘includes’’ exactly the same elements as are included in structure (56a). Calculation of optimal satisfaction of that constraint, then, cannot proceed differently in the two cases; postposing, at least in such cases, is neutral with respect to MATCH PHRASE. That in turn leaves us in a position to understand why there are two ways of repairing structures that might otherwise violate STRONG START. If we assume that the relative ranking of MATCH PHRASE and NO SHIFT is variable (Anttila 2002; cf. Coetzee and Pater 2011), we expect two outcomes: the ranking MATCH PHRASE >> NO SHIFT results in pronoun postposing, while NO SHIFT >> MATCH PHRASE results in leftward enclisis.

Postposing will also in the general case be neutral with respect to the BINARITY constraint defined in (47). Postposing is adjunction, and adjunction by definition creates binary-branching structures. Provided that the material left behind by postposing (the contents of the lower Φ-phrase in (56b)) includes at least two prosodic constituents, structural binarity will be unaffected by displacement of the pronoun.33 (See footnote 41 for elaboration of this point.)

This is the core of our proposal. Besides its fundamental simplicity, its single most important property is that it provides a reason for the existence of pronoun postposing. Viewed in this way, postposing is not an isolated quirk of the grammar of Irish; rather, it is one cog in a larger machine that shapes the rhythmic structure of expressions. Prosodic factors are now not extraneous or added on—they are the heart of the matter, and it is then inevitable that the positioning of pronouns would exhibit a rich set of interactions with prosodic factors. It is also expected, rather than surprising, that postposing would be insensitive to pragmatic and discourse factors, since (on this view) the positioning of weak pronouns is shaped only by a drive for rhythmic balance of a certain kind.

These are important virtues, but we need now to assess how the proposal fares when we venture beyond the simplest cases and face the empirical challenges considered earlier in the article. We need to ask in particular how well the proposal deals with those cases that pose difficulties for purely syntactic accounts of postposing. That is the work of the next section.

8 Extensions and Challenges

8.1 A Simplification and an Extension

We begin by considering a facet of the postposing puzzle that we have not yet discussed. To start, notice that the informal presentation of the mechanism of postposing given in (56) is in fact illegitimate, if we are serious about developing a truly phonological account of the facts. (56) implies that postposing applies to light elements at the left edge of a Φ-phrase only if they correspond to pronouns (are members of the syntactic category D). But prosodic categories are ‘‘homogeneous’’ in the sense that Φ-phrase s that correspond, say, to PPs and those that correspond to VPs are indistinguishable in their behavior and properties. If that is the case, prosodic elements that correspond to D should be indistinguishable from similar elements that correspond to, say, the syntactic category P. And of course STRONG START as defined in (55) makes no reference to pronouns, to objects, or to any term from syntactic theory. Its effects (ill-formedness or associated repairs) should then be felt whenever an inappropriately light element appears at the left edge of a Φ-phrase , no matter what syntactic category that element corresponds to. All this being so, (56) would be better understood in the simpler and more general form of (72).

(72)

graphic

In (72), we have a truly prosodic, and truly nonsyntactic, account of postposing. Interestingly, there is evidence that the simplification in (72) also represents an empirical advance. Our discussion of pronoun postposing has so far been typical of theoretical treatments in focusing exclusively on the reordering of pronouns. But all descriptions of the phenomenon make clear that an exactly analogous reordering applies to certain kinds of prepositional phrases (see, e.g., Stenson 1981: 42–47, Ó Siadhail 1989:207–210, esp. 207). Ó Siadhail (1989:208) cites the doublet in (73), for example.

(73)

graphic

To these we can add the examples in (74) from our own observation.34

(74)

graphic

The elements that postpose in cases like these are PPs that consist solely of a prepositional head inflected for the person, number, and gender features of its (silent) object.35 They are mostly monosyllabic and they are all unaccented (like simple pronouns, these PPs have distinct weak and strong realizations). Cases such as (73) and (74), then, also have inappropriately light elements at the left edge of the Φ-phrase corresponding to VP, threatening a violation of STRONG START. It is unsurprising, then, that postposing should be available here as well.

Questions now arise, of course, about other kinds of prosodically light elements: functional elements such as C, D, and T. These are syntactic heads and therefore occur in the initial position of their maximal projections. Their phonological exponents—which are typically unaccented and weak—will then appear at the left edge of Φ-phrase s, in apparent violation of STRONG START. So why do these elements never postpose? This is an important question, and we return to it in section 8.6, when more of the necessary background has been put in place. For now, we move on to other aspects of postposing.

8.2 Subjects of Finite Clauses

The correct analysis of pronoun postposing must guarantee that subject pronouns in VSO clauses, even when weak, never postpose. (75), repeated from (8), is impossible.

(75)

  • *Chuir mo lámh ’mo phóca mé.

  • put.PAST my hand in.my pocket I

  • ‘I put my hand in my pocket.’

Simple application of MATCH PHRASE would derive a nested prosodic structure like (76) (cf. (51) and tableau (54)).

(76) (Φ chuir (Φ mé (Φ mo lámh ’mo phóca)))

Default prosodification may therefore place a subject pronoun in <-initial position—exactly the same configuration responsible for triggering the postposing of object pronouns (cf. tableau (70)). The puzzle, then, is why subject pronouns never postpose.

From the earliest discussions of the phenomenon, the intuition has been that postposing is preempted in such cases by the requirement that weak subject pronouns incorporate into the preceding verbal complex. We believe that this general approach is correct. The challenge, however, has always been to go beyond the level of intuition in spelling it out. And syntactic movement analyses are ill-placed to do that, since both leftward and rightward movements routinely target the subject position of finite clauses, as shown for wh-movement (in a cleft) in (77).

(77)

  • Is mé a tá __ tuirseach.

  • COP.PRES me C be.PRES tired

  • ‘It’s me that’s tired.’

In this observation, we have another important contrast between the conditions that govern pronoun postposing and those that govern syntactic movement. We also now have two questions to answer. The first is why syntactic movement of a subject is not blocked by whatever operation is responsible for incorporation of the subject (call it Subject Pronoun Incorporation); the second is why pronoun postposing is preempted by Subject Pronoun Incorporation.

The answer to the first question is clear. Subject Pronoun Incorporation is a postsyntactic phenomenon (Chung and McCloskey 1987:226–228, Doherty 1996:23–25, Ackema and Neeleman 2003). Doherty (1996:23) in fact argues that pronouns are incorporated into the verbal complex by way of a morphological operation, one that results in the creation of a complex morphological word. As evidence for this conclusion, he cites a range of phenomena that treat the incorporated pronoun in exactly the same way as person- and number-marking suffixes on the verb.36 If this is right, Subject Pronoun Incorporation will be invisible and irrelevant as far as syntactic operations are concerned. Following movement of the pronoun in, for instance, (77), linearization and trace elimination (if they are distinct) will apply and will yield as input to the morphology a representation in which there is no pronoun to incorporate—in which case, nothing more need be said. Syntactic operations cannot anticipate what might happen in the morphology and will never enter into competition with morphological operations.

But of course the heart of our argument here is that pronoun postposing is postsyntactic. If postposing were a syntactic movement, the logic of the previous paragraph should apply with equal and similar force to wh-movement and to postposing; we would then expect, contrary to fact, that weak pronominal subjects should postpose, just as they undergo wh-movement. However, if Doherty (1996) is right that Subject Pronoun Incorporation is a morphological operation, and we are right that pronoun postposing is part of prosodic structure-building, the observations fall into place. Assuming that morphological operations (including vocabulary insertion) create the input to prosodic structure-building, a weak pronominal subject will necessarily be the rightmost element within the verbal complex and will never trigger a violation of STRONG START (cf. Henderson 2012 on the timing of morphological insertion and prosodic parsing). Consequently, no repair will be warranted. Given this constellation of assumptions, the contrast between (8)/(75) and (77) is inevitable.37

8.3 Partial Postposing

Any suitable analysis must also provide an understanding of partial postposing. As we have already shown (see (4d) and (11)) and as is especially clear from the large dataset summarized in appendix B of Bennett, Elfner, and McCloskey 2015, variability of positioning is a central part of the pronoun-postposing puzzle. Consider (78).

(78)

  • graphic

  • show.PAST they it to my mother week ago in Derry

  • ‘They showed it to my mother in Derry a week ago.’

(78) is well-formed under conditions that should by now be familiar: the object pronoun can cliticize to the subject DP, or it can be realized as a full prosodic word in situ. But of course pronoun postposing is also an option, and the postposed pronoun may appear after any of the major postverbal constituents.

(79)

graphic

Following elimination of silent elements, the vP of (78) will have the syntactic structure shown in (80).

(80)

graphic

In determining what prosodic structure will be associated with (80), the crucial principle is MATCH PHRASE as formulated in (35), repeated once again in (81).

(81)

  • MATCHPHRASE

  • Given a maximal projection XP in a syntactic representation S, where XP dominates all and only the set of terminal elements {a, b, c, . . . , n}, there must be in the phonological representation P corresponding to S a Φ-phrase that includes all and only the phonological exponents of a, b, c, . . . , n.

  • Setting aside the various PPs, there is just one maximal projection in (80): vP1 . Neither vP2 nor vP3 is maximal (since each is immediately contained within a phrase with which it shares a label).38 That means that MATCH PHRASE will impose a relatively weak requirement on the prosodic realization of structures such as (80). It will be satisfied as long as there is a Φ-phrase that includes all and only the ( phonological exponents of the) terminal elements of vP1 . There are of course many ways in which that requirement can be met, and it follows in turn that those constraints that are concerned only with rhythmic balance (BINARITY and EQUAL SISTERS especially) will play a decisive role in determining which phrasings actually emerge. We believe that this is a correct outcome since complex vPs such as (78) can in fact be phrased in more than one way. For (78), we will expect at least the possibilities in (82).

(82)

graphic

What is most important for present purposes, though, is that we now understand the various possibilities exhibited in (79). A weak pronoun may adjoin to any of Φ3 , Φ2 , or Φ1 of (82a), avoiding a violation of STRONG START while still satisfying MATCH PHRASE and doing just as well with respect to BINARITY as the variant without postposing. This is one place where prosodic recursion plays a pivotal role in the analysis: flexibility in pronoun positioning reflects an interaction between MATCH PHRASE, as defined in (81), and the recursive nesting of Φ-phrases.

We now understand, then, the range of possible landing sites.39 But we must also understand why pronouns may not be shifted to positions within the various postverbal constituents (XP, YP, or ZP of (6)). From the starting point in (83), the variants in (84) cannot emerge.

(83)

  • graphic

  • bury.FUT.IMPERS him in graveyard Cruit Tuesday after mass the

  • mheán lae i dteach pobail Cheann Caslach.

  • midday in church Kincasslagh

  • ‘He will be buried in Cruit graveyard on Tuesday after midday mass in the church in Kincasslagh.’

(84)

graphic

But these contrasts too fall out from the basics of our proposal since postposing ‘‘into’’ a syntactic constituent like PP will always incur a gratuitous violation of MATCH PHRASE, which will require for each postverbal constituent in cases like (84) that there be a corresponding Φ-phrase that includes all and only its terminal elements. The presence of the weak pronoun within the Φ-phrase corresponding to PP will therefore force a violation of MATCH PHRASE. Since there will be no such violation in the examples of (79), they will always emerge as optimal by comparison, as shown in (85).

(85)

graphic

In this way, we derive the pattern of (6), repeated here as (86).

(86)

graphic

This was one of the principal goals we set for ourselves at the beginning of the article.40

8.4 Small Clauses

Many of the most challenging puzzles we have encountered center on the application of pronoun postposing to the subjects of small clauses. Here we argue that our proposals deal straightforwardly with core cases of this type and that they also extend gracefully to those cases that seem most troublesome for syntactic approaches. To begin, consider again (24a) and (27), repeated here in (87) and (88).

(87)

graphic

(88)

graphic

We have already shown (section 4.2) that such cases are troublesome for syntactic analyses of postposing. Far from being troublesome, however, they are expected given our prosodic proposals. As before, the syntactic starting point we assume is something like (89).

(89)

graphic

The syntax of (89) exposes a quirk of the system of mapping constraints in (34) that we have not yet addressed. Those mechanisms associate lexical items, maximal projections, and certain kinds of clauses with prosodic constituents (ω’s, Φ-phrase s, and ι-phrases, respectively). However, they provide no instruction about how to treat other syntactic constituents—those, in particular, that are neither maximal nor minimal. Consider (90), with specifier α, complement β, and a head H that crucially has phonological content.

(90)

graphic

The mapping principles ensure that HP will correspond to a Φ-phrase , that the head H will correspond to a prosodic word w, and that the specifier and complement will correspond to Φ-phrases (unless they happen to be syntactically minimal—for example, pronouns or other heads). But as things stand, the intermediate constituent, unlabeled in (90), which includes only the head H and its complement, will not be mapped to any prosodic constituent. But the prosodic word corresponding to H must be integrated somewhere, and given (35) it must be integrated in such a way that it is within the Φ-phrase corresponding to HP but outside the Φ-phrase corresponding to α and β. This will lead to the prosodic representation in (91).

(91)

graphic

But (91) violates the crucial BINARITY constraint (section 6.2). We expect, then, that structures such as (90)/(91), in which H has phonological content, will be prosodically unstable, at least in those languages in which the demands of binarity take precedence over the matching principles of (34). In such languages, of which Irish is certainly one, structures like (91) will always require a repair.

In the case of (89), if the mapping constraints of (34) were to apply without adjustment, the prosodic structure in (92) would result.

(92)

graphic

In (92), the complex word ag troid ‘fighting’ (lit. ‘PROG fight’) is orphaned in the higher Φ-phrase. But (92), of course, is not what emerges as the actual prosodic structure, given the fatal violation of BINARITY that it incurs (and also EQUAL SISTERS, if the PP is correctly mapped to a Φ-phrase ). One of the structures that can emerge to resolve this dilemma is (93), in which both BINARITY and EQUAL SISTERS are respected. The string mé ag troid constitutes a Φ-phrase in but corresponds to no syntactic constituent—a failure of isomorphism rooted again in the requirements of eurhythmy.

(93)

graphic

In (93), we show the pronoun in its strong form as a full prosodic word and so no further adjustment is required. If, however, the weak form of the pronoun were to be chosen (a syllable rather than a word), then a violation of STRONG START would result, since the pronoun is initial in Φ. A repair is thus required and the result is (87b), with the prosodic structure shown in (94), determined by the constraint ranking shown in (95).41

(94)

graphic

(95)

graphic

In the apparently contrasting example of (88), the pronoun follows the complement but precedes a VP-adverb. Such a case will have the syntactically expected order of (96) and the syntax in (97a). For exactly the same reasons as in the previous case, we now expect the prosodic structure in (97b).42

(96)

  • Chonac [é ag féachaint uirthi go drúisiúil].

  • I.saw him PROG look on.her lasciviously

  • ‘I saw him looking at her lasciviously.’

(97)

graphic

Given (97) with its STRONG START violation (the leftmost constituent of Φ3 is a syllable rather than the required word), the familiar repairs are available: right-adjunction to Φ1 or Φ2 , yielding the two legal outputs of (98).43

(98)

graphic

Once again, MATCH PHRASE is satisfied in both variants of (98) because the single maximal verbal projection of (97a) has a prosodic counterpart (Φ1 of (97b)) that includes all and only its terminals. These seemingly difficult cases, then, fall into place.

Another piece that falls into place without elaboration is the striking type of example in (99), which was presented in section 4.2 as being mysterious if postposing is syntactic.

(99)

  • graphic

  • COP.PRES no.matter PRED summer it or PRED winter

  • ‘It doesn’t matter whether it’s summer or winter.’

  • (U 20)

In such cases, the pronoun subject of a small clause postposes to a position apparently within the disjoined predicate of the small clause. We again assume the syntax in (100).

(100)

graphic

Very few of the particulars of (100) are important for present concerns. No matter how those details are filled in, the mapping principles of (34)/(35), acting in concert with BINARITY, would yield the prosodic representation in (101).

(101)

graphic

The representation in (101) simply carries through consistently our earlier assumptions about matching. The pronoun é, being a syntactic head, corresponds to a prosodic word; the intermediate projection consisting of the disjunction particle and its complement (the predicate ’na gheimhreadh) cannot be linked with any prosodic constituent; to meet the requirements of BINARITY, however, the disjunction particle phrases with its complement, the second disjunct. Both predicates, though phrasal in the syntax, consist only of a single prosodic word with a single accent (samhradh ‘summer’ and geimhreadh ‘winter’), each with an adjoined proclitic—the predicative particle ’na (/nә/). Therefore, they phrase only as prosodic words, consistent with our earlier commitments. This much is unremarkable.

What is however striking about (101) is that it is identical, in all respects, to the structures assigned by the mapping constraints to finite VSO clauses (see the discussion in section 6.2 around (53)). One of the examples from that discussion, (53a), is repeated here.

(102)

  • Sciob an cat an t-eireaball den luch.

  • cut.PAST the cat the tail off.the mouse

  • ‘The cat cut the tail off the mouse.’

The isomorphism between the prosodic structure associated with (102) and that associated with (100) reflects a core property of the mapping theory we rely on throughout—namely, its blindness to syntactic category distinctions. Despite their profound syntactic differences, both the finite verb in the VSO structure and the initial pronoun in the small clause structure map to a prosodic word whose sister is a Φ-phrase , one that in turn has a prosodic word and a Φ-phrase as its immediate constituents. That Φ-phrase in turn consists of two prosodic words. This structure (illustrated for (102) in (103a)) should be compared with (101). It may be worth stressing that the calculations that yield these isomorphic structures are identical in every detail for the two (syntactically very different) cases. Importantly, though, as we have already shown, the structure that actually emerges for cases like (102) is the one in (103b), as can be heard in the sound file associated with (53a).

(103)

graphic

In our introductory discussion, we attributed the emergence of (103b) to the fact that it eliminates the violations of EQUAL SISTERS incurred by (103a), while satisfying BINARITY just as well.

What is more important at present, though, is that (103a) and (101) are indistinguishable. The logic of internal consistency will therefore demand that we treat (99) in exactly the same way as these VSO cases, with the consequence that we must expect (104) as a possible phrasing for a syntactic structure like (99).

(104)

graphic

The rebracketing seen in (104), like that in (103b), satisfies both BINARITY and EQUAL SISTERS. The predicate ’na shamhradh consists of a single prosodic word with one accent, as does the predicate ’na gheimhreadh. Disjunctive has both a prosodically strong accented form /no:/ and a prosodically weak form /nә/; in (104), the former is deployed. We have then a sequence of two Φ-phrase s with parallel internal constituencies, and the structure overall is rhythmically balanced. This phrasing is evident to the ear, and it emerges directly when the pronoun is in its strong form, with the word order in (105).

(105)

  • is cuma é ’na shamhradh nó ’na gheimhreadh

  • COP.PRES no.matter it PRED summer or PRED winter

  • ‘It doesn’t matter whether it’s summer or winter.’

If, however, the weak form of the pronoun were to be used instead, the familiar violation of STRONG START would result and the further readjustment in (106) would be motivated.

(106)

graphic

There is another possibility. As in our earlier discussion of ‘‘partial postposing,’’ the pronoun might instead attach to the higher Φ-phrase , yielding the equally possible (107), with the prosodic structure in (108).

(107)

  • is cuma ’na shamhradh nó ’na gheimhreadh é

  • COP.PRES no.matter PRED summer or PRED winter it

  • ‘It doesn’t matter whether it’s summer or winter.’

(108)

graphic

These possibilities emerge as necessary and expected in the context of our proposals, then. No elaboration is necessary in order to handle them, and in fact if the crucial rebracketing in (104) were excluded, and the postposing in (29)/(99) not predicted, we could justifiably be accused of internal inconsistency. We take this result to be important, since this example type poses such profound difficulties for syntactic accounts of pronoun postposing.

8.5 Root Small Clauses

Consider a final small clause type. As we have already illustrated (in section 3), Irish permits root small clauses with assertoric force. But such clauses differ sharply from complement small clauses in that their subjects may not postpose.

(109)

  • a.

    É gléasta go niamhrach.

    him dressed resplendently

    ‘He was dressed resplendently.’

    (SR 19)

  • b.

    *Gléasta go niamhrach é.

This fact was attributed by Chung and McCloskey (1987) to the absence of a licensing head governor for the trace of postposing. How might we understand the failure in (109b) in prosodic terms? The first observation to be made is that STRONG START is clearly at play in such cases. Pronouns at the left edge of root small clauses must appear in their strong forms; weak forms are absolutely excluded. We take this to be one of the signature effects of the activity of STRONG START. The analytical puzzle then is double-edged: to understand why the constraint is operative in the contexts of (109), and then to understand why the only available repair is Option A of (62)—strengthening in place. Option B (enclisis in place) is impossible for obvious reasons (the absence of a host). But why should postposing be impossible?

The commitments that we have taken on entail a prosodic structure like (110b) for (108), given the conventional syntax in (110a).

(110)

graphic

The entire small clause corresponds to an L-phrase because it carries assertoric force (see the second clause of (34) and the accompanying discussion) and the pronoun is phrased as a prosodic word—an option that is always available. However, if the pronoun were to appear in its weak form, we would have a violation of STRONG START as defined in (55). This much lets us understand why weak pronouns may not appear in the subject position of such clauses and why strong pronouns must instead be deployed.

What, then, explains the ban on postposing? Consider the options, given (110b). Note first that the pronoun may not adjoin to the Φ-phrase corresponding to the predicate XP, nor to any position within that Φ-phrase . Either of these deformations would convert the binary-branching ι-phrase in (110b) into a nonbranching constituent that dominates only a single Φ-phrase . Such structures run afoul of BINARITY, which, as we have noted repeatedly, is highly valued in Irish. In this way, we can understand the ill-formedness of ‘‘lowering’’ an utterance-initial pronoun into a subordinate Φ-phrase in structures like (110b).

The only available option therefore would be to adjoin the pronoun at the root—to the ι-phrase. But that too, we suggest, is impossible. Specifically, we hypothesize that right-adjunction of prosodically weak elements to the ι-phrase is forbidden (in Irish) and that that is what calls off the possibility of postposing in a structure like (110). This seems plausible in language-internal terms, and it may reflect a more general pattern. There is an intriguing parallel here to certain effects in Bosnian-Croatian-Serbian discussed by Werle (2009:364–370) (see also Harizanov to appear). Werle discusses an effect in those varieties that he calls the ‘‘utterance-final effect.’’ He shows that clitics of a certain class that have a relatively free distribution may not in general appear at the right edge of the utterance, arguing that the option is tolerated only if there is no alternative way of realizing the relevant structure. For Irish, there will always be such an alternative—namely, Option C of (62), in which a strong form of the pronoun is used, as shown in (109).44 See Bennett, Elfner, and McCloskey 2015 for more discussion of interactions between ιP phrasing and pronoun postposing.

This completes our discussion of the major types of small clause and their interaction with postposing. The principal task that remains is to assess how we can understand the facts that seem to suggest a role for the ECP in determining pronoun position. That is the work of the next section.

8.6 Head Government Revisited

The head government requirement of the ECP was at the core of the syntactic analysis of postposing developed by Chung and McCloskey (1987). In sections 3.1 and 4.1, we reviewed the observations that motivated that analysis and also some reasons to be skeptical of it. In particular, we pointed to cases like (22a) and (23d), repeated in (111), in which postposing is possible despite the absence of a lexical governor.

(111)

graphic

Such observations add force to the methodological impulse to eliminate proper government from our theoretical and descriptive arsenals; but of course the facts that seemed to motivate the head government requirement remain. In this section, we argue that those facts are better understood in the context of our prosodic proposal than in the context of the head government account. Further, we argue that the prosodic account lets us understand why the earlier proposal yielded such a good approximation of the facts.

The head government clause of the ECP draws a crucial distinction between lexical (open-class) heads and functional (closed-class) heads. The former are licensers of movement; the latter are not. But this distinction has a prosodic correlate. Functional heads tend to be prosodically dependent (Truckenbrodt 1999, among many others), while lexical heads tend to be prosodically independent. The existence of such a correlation suggests the possibility of rethinking the relevant patterns in phonological terms. But the correlation is in turn only approximate—not all functional heads are prosodically dependent. The existence of such ‘‘corner cases’’ should provide a way of distinguishing empirically between analyses based on proper government and analyses based on prosodic considerations.

We begin by showing that the troublesome case in (111b) is expected given the approach now on the table. The prosodic structure linked with an example such as (111b) will be (112).

(112)

graphic

In (112a), we represent the accusative subject pronoun in its strong variant, as a prosodic word, which means that there is no violation of STRONG START and therefore no repair. Of course, if the weak version of the pronoun is selected, STRONG START will be violated and the prosodic structure that will emerge is (112b), an entirely appropriate outcome.45 What was a problem of undergeneration for the head government proposal now falls into place quite naturally.

What of the cases, though, in which the head government requirement correctly rules out impossible postposings? Setting aside root small clauses (which we have already dealt with in section 8.5), such cases are of two kinds: negated small clauses (113) and small clause complements to functional rather than lexical heads (114).

(113)

  • a.

    Ba mhinic gan é sa bhaile.

    COP.PAST often NEG him in.the home

    ‘He was often not at home.’

  • b.

    *Ba mhinic gan __ sa bhaile é.

(114)

  • a.

    Agus [é i mBaile Átha Cliath] . . .

    and him in Dublin

    ‘While he was in Dublin . . .’

  • b.

    *Agus [ __ i mBaile Átha Cliath é] . . .

The minimal syntactic structure we could assume for such cases is shown in (115a) and (115b), respectively.46

(115)

graphic

Unsurprisingly, both the marker of negation gan and the subordinator agus (being function words) are prosodically dependent. The negative marker is [gən]; agus is realized variously as [əs], [gəs], or [əgəs]. Unadjusted prosodic structures corresponding to ΣP of (115a) and (115b) then will be as in (116a) and (116b), respectively.

(116)

graphic

Such structures raise two questions: (a) Why does the presence of the function word to the left of the pronoun call off the possibility of postposing? (b) There is a STRONG START violation in Φ1 of (116a–b), which is induced by the presence of the prosodically dependent element at its left edge. Why does this potential violation not itself trigger postposing of the offending element, the negative marker [gən]? This latter question is the one we postponed from section 8.1.

The two questions are intimately linked. The key property of these structures is that the function words at their left edges are proclitics that are dependent on the prosodic constituent to their right. We take it that this dependence should be modeled formally by way of left-adjunction of the dependent element to the immediately following prosodic constituent. Which constituent, though, does the dependent element adjoin to: the first word of Φ2 , or Φ2 itself? This kind of question is notoriously hard to resolve (for relevant discussion, see Selkirk 1996, Hall 1999, Ito and Mester 2009a), but for one class of cases at least that has been closely studied, the answer seems reasonably clear. McCloskey (1996a) shows that complementizers in Irish are subject to postsyntactic lowering and that they left-adjoin to the inflected verb (not to a phrase) to form a ‘‘verbal complex’’ within which complex patterns of allomorphy are observed (see Oda 2012: chap. 6 and Acquaviva 2014 for recent, detailed analyses of these matters within the framework of Distributed Morphology). Let us assume that this result is general and that all prosodically dependent functional heads similarly adjoin to a prosodic word below them and to their right.47 If gan in (116a) and agus in (116b) adjoin to the subject pronoun, we expect that the pronoun will be required to appear in its strong variant, in order to serve as a host for the clitic.48 In that case, the initial elements of (116a) and (116b) will be as in (117).

(117)

graphic

This prediction is correct. In all of the cases under consideration here (e.g., (113) and (114)), subject pronouns must appear in their strong and accented forms (section 4.3). There will of course now be no violation of STRONG START with respect to Φ2 of (116), since the structures illustrated in (117) are the leftmost immediate constituents of Φ2 and are full (complex) prosodic words. In the absence of a violation, no repair is needed and postposing is unmotivated. In effect, adjunction of the prosodically dependent function words ‘‘protects’’ the pronoun from being at the left edge of the Φ-phrase and therefore guarantees, by forcing the pronoun into its strong form, that there will be no violation of STRONG START. The results we need are secured. Pronouns will never postpose from such positions, nor will the functional heads that left-adjoin to them.

Important questions remain, of course. The generalization we rely on here is that functional heads in Irish, when prosodically dependent, are always proclitic and are never enclitic. It is this property that forces the crucial left-adjunction shown in (117), ensuring that there will be no postposing of either element. Weak pronouns and weak PPs, by contrast, are neither full prosodic words nor proclitics (pronouns at least are clearly enclitic) and therefore the mechanisms we discussed earlier come into play for them: postposing or left-adjunction in situ. The relevant generalization about functional heads is true and correct—they are all proclitic—but of course one would like to know whether this is simply a freestanding pattern, an irreducible regularity in the data to which learners are sensitive, or whether it is perhaps a reflection of some broader pattern or deeper principle (such as a preference to phrase syntactic heads with their complements; see Anderson 2005). We have at present no answer to these questions, but we note that they arise no matter what one assumes about pronoun postposing.49 However these questions are resolved, the conclusion we stress here is that the observations that seemed to argue for a head government condition on postposing now fall within the range of understanding, without appeal to the ECP or to the mechanisms that the ECP depends on. The relevant observations emerge as reflections of an interplay between our core proposal and some well-grounded, independent aspects of prosodic organization in Irish.

This success takes on added significance when we consider a final, special case in which pronouns may postpose, even though there is again no lexical governor to license the postposing. Recall from section 4.1 that small clause complements to the demonstrative particle seo routinely host pronoun postposing from their subject positions. The structure itself is illustrated in (118); the possibility of postposing is demonstrated again in (119).

(118)

  • a.

    [seo [DP XP]]

  • b.

    Seo na saighdiúirí ag teacht.

    DEMON the soldiers PROG come

    ‘Here come the soldiers.’

(119)

  • graphic

  • in.the end DEMONPROG come him

  • ‘In the end, here he comes.’

  • (PNG 83)

Our earlier discussion highlighted cases like (119) as problematic for a syntactic analysis that includes head government as a central requirement. The difficulty is that seo is clearly just as much a closed-class element as negation or the coordinator agus and should therefore be incapable of licensing the postposing seen in (119). The interesting puzzle is why (119) should be different from the apparently similar (114), such that postposing is possible in the former but impossible in the latter. In fact, we know of no syntactic difference between these two cases that would allow an understanding of their different behaviors. However, the elements in question differ crucially in their prosodic characteristics. As noted above, agus is unaccented and prosodically dependent. The particle seo, on the other hand, is exceptional among function words in being accented and prosodically independent; it is the ‘‘corner case’’ we sought.

We know that seo ( phonemically /∫ɔ/ or /∫o/) is a full prosodic word because, unlike all of the unaccented elements we have dealt with so far, it can stand alone as the single accented element in an utterance.

(120)

  • Seo é.

  • DEMON it

  • ‘Here it is.’

(121)

  • a.

    Cén dóigh a ndearna tú é?

    what way C did you it

    ‘How did you do it?’

  • b.

    Mar seo.

    as this

    ‘Like this.’

In (120), the only element in the utterance apart from seo is a pronoun. This pronoun may appear in its weak form, in which case it is prosodically dependent on the demonstrative. In the response of (121b), the only other element is the unaccented preposition mar (/mər/), which is proclitic on the demonstrative. Finally, seo may appear in a coordinate structure, as complement, for instance, to the preposition idir ‘between’.

(122)

  • idir seo agus Doire

  • between DEMON and Derry

  • ‘between here and Derry’

This last position is one that is absolutely restricted to prosodically independent, accented elements. Weak forms of pronouns, for instance, are impossible in the context of (122). For cases such as those in (123), the pronouns must be pronounced in their full accented forms—[meː] and [eː]—rather than in their weak (unaccented) forms.

(123)

  • a.

    idir mé agus é

    between me and him

    ‘between me and him’

    (SD 324)

  • b.

    idir é agus Fionntrá

    between it and Ventry

    ‘between there and Ventry’

    (PNG 573)

But once we establish that seo is not prosodically dependent, we understand why postposing is possible in cases like (119). The schematic prosodic structure for (119) will be as in (124).

(124)

graphic

In (124), the subject pronoun appears in its strong variant and postposing is not warranted. If, however, the pronoun were to appear in its weak form, there would be a violation of STRONG START with respect to Φ2 and postposing would be an available and warranted repair. Because seo is prosodically independent, no adjunction applies in (124) to protect the pronoun from appearing at the left edge of the Φ-phrase and the STRONG START violation is inescapable. A contrast that is mysterious in syntactic terms emerges as inevitable in prosodic terms.

We take this last case to be especially revealing since the contrast between (119) and (114) can be viewed as a sort of well-designed natural experiment—one in which everything is held constant except for one factor (prosodic) and we observe the possibility or impossibility of postposing covarying with that (prosodic) factor. And it is exactly this that makes the case so difficult for any syntactic account. These observations also let us understand why the head government analysis is so seductive. It provides a very good approximation of the facts, precisely because functional heads are in the typical case prosodically dependent. It is only when the two properties (syntactic and prosodic) exceptionally come apart, as they do in the case of demonstrative seo, that we can catch a glimpse of the truth behind the confound.

9 Assessment and Conclusion

In section 5, we laid out a set of criteria by which success in our domain of investigation might be measured. Now that our proposals have been developed, it is time for the assessment.

Needless to say, many puzzles and mysteries remain. That said, however, the proposals we develop here go farther, in terms of empirical coverage, than any extant account of postposing that we are aware of, for any of the Gaelic languages. In addition, they let us understand why postposing comes trammeled up with prosodic connections and correlations but shows no sensitivity to pragmatic or discourse factors: it has to do only with rhythmic organization. The account also provides an understanding of the ‘‘optionality’’ of postposing and of the fact that postposed elements may appear in a range of positions (though always at the right edge of a Φ-phrase). They let us understand why postposing gives the appearance of being sensitive to head government, but they also gracefully integrate those cases that are incompatible with such a requirement. Also integrated is a set of observations that suggest that postposing is sensitive to a kind of constituency that is not the constituency found in syntactic representations. Rather, the relevant representational system is that of prosodic constituency. Moreover, the prosodic phrasings we make crucial appeal to are clearly real, regardless of how we understand the mechanisms that generate those structures. Perhaps most important is the fact that the mechanisms that do this empirical work are well-integrated with a reasonable overall view of phonological phrasing in Irish and in general. The core mechanism appealed to (prosodic adjunction) is well-established and unexotic, as are the various phonological constraints upon which our proposals rely.

To the extent that the proposals are viewed as successful, it may be worth asking what the ingredients of that success are. There are several. One is the commitment to a certain kind of recursion in prosodic structure, one that allows one phonological phrase to contain another; this is the ingredient that allows postposed pronouns a range of final resting places and therefore lets us understand what we have called here ‘‘partial postposing’’ (a core property of the phenomenon, as we have shown).

A second key ingredient is the idea that the need to create optimal prosodic constituents leads to the emergence of prosodic constituents like mé ag troid ‘me PROG fight’ in (87) or é ’na gheimhreadh in (29)/(99), which are bizarre from a syntactic perspective (as extensively discussed in Nespor and Vogel 1986). But it is exactly this ‘‘bizarre’’ constituency that postposing seems to be sensitive to.

A third important factor has been what we might call the ‘‘homogeneity’’ of prosodic constituents within a given category. A consequence of that commitment is that structures that are syntactically very different indeed from one another (e.g., finite VSO clauses and certain small clauses involving disjunction) end up being indistinguishable from a prosodic perspective and so support instances of postposing that are, again, very bizarre indeed when viewed in syntactic terms.

We set out to develop and assess a deliberately radical version of the prosodic approach, one that used no term at all from syntactic theory in its formulation, but only the primitives provided by prosodic theory (see (72)). It is interesting, we think, that that proposal goes as far as it does. There remain observations about postposing that are not easily understood either on a syntactic account or on the account we offer here: the fact, for example, that the preverbal subjects and objects of nonfinite clauses resist postposing (see Chung and McCloskey 1987: 228–234 and (32a) above) even though these are positions from which leftward and rightward syntactic movements are freely possible.

(125)

graphic

(126)

graphic

We also do not yet know how to integrate the observation that postposing seems to show across-the-board effects, as in (127).

(127)

graphic

It remains unclear to us whether these lacunae reveal a failure of understanding of the syntax (poorly understood at present) or whether they suggest that an analysis is needed that draws on systems of representation that are part syntactic, part phonological—a less radical version of the present proposal in a certain sense (see Chung 2003, Göksel, Kabak, and Revithiadou 2013). Time will hopefully tell.

Appendix A: Sources of Sound Samples

graphic

Appendix B: Sources of Examples

ACS: Ag Coimeád na Síochána, Páid Ó Súilleabháin

AGMTS: Ar Gach Maoilinn Tá Síocháin, Pádraig Ó Cíobháin

AO: Aistí Ó Chléire, Donnchadh Ó Drisceoil

BEAL: Béaloideas, Journal of the Irish Foklore Society

CC: Cruithneacht agus Ceannabháin, Tomás Bairéad

CLENS: Cín Lae Eibhlín Ní Shúilleabháin, ed. by Máiréad Ní Loingsigh

CM: An Chéad Mhám, Seán Bán Mac Meanman

CTP: Cuimhne an tSeanpháiste, Micheál Breatnach

DC: Don Cíochóta, trans. by An tAthair Peadar Ó Laoghaire

DCA: Dith-Chéille Almayer, Joseph Conrad, trans. by Seosamh Mac Grianna

DGD: Deoir Ghoirt an Deoraí, Colm Ó Ceallaigh

DO: Dialann Oilithrigh, Donchadh Ó Céileachair

FB: Feamain Bhealtaine, Máirtín Ó Direáin

FF: Fonn na Fola, Beairtle Ó Conaire

GAT: Gabhla An tOileán, Máirín Uí Fhearraigh

I: Ise, H. Rider Haggard, trans. by Niall Ó Domhnaill

LG: Le Gealaigh, Pádraig Ó Cíobháin

LNT: An Leacht Nár Tógadh, Séamas Ó Conghaile

M: Mise, Colm Ó Gaora

MBS: Mura mBuafam—Suathfam, Maidhc Dáinín Ó Sé

MSF: Mo Sgéal Féin, Peadar Ua Laoghaire

MSN: Micí Sheáin Néill: Scéalaí agus Scéalta, Cathal Póirtéir

NGTTS: Na Gabh Thar Tí Stiofáin, Máire Uí Fhlatharta

ONH: Ór na hAitinne, Tomás Bairéad

OTA: Ón tSeanam Anall, Scéalta Mhicí Bháin Uí Bheirn, ed. by Mícheál Mac Giolla Easbuic

PNG: Pobal na Gaeltachta, ed. by Gearóid Ó Tuathaigh, Liam Lillis Ó Laoire, and Seán Ua Súilleabháin

SAIL: Seanchas Amhlaoibh Í Luínse, ed. by Seán Ó Cróinín

SB: Séidean Bruithne, Joseph Conrad, trans. by Seosamh Mac Grianna

SD: Sionnach Ar Mo Dhuán, Breandán Ó hEithir

SJSJ: Seachrán Jeaic Sheáin Johnny, Mícheál Ó Conghaile

SOOT: Seanchas ón Oileán Tiar, Tomas Ó Criomhthain

SR: Sciuird chun na Rúise, Pádraig Ó Fiannachta

TGC: Thiar i nGleann Ceo, Tadhg Ó Rabhartaigh

U: Unaga, trans. by Eoghan Ó Neachtain

Notes

The research reported on here has its origins in the meetings of the Prosody Interest Group at the University of California, Santa Cruz. We are grateful to the group and especially to its conveners Junko Ito and Armin Mester, for their input and for providing the initial impetus for the project. We are also grateful to Lisa Selkirk for input and advice throughout. Discussions with Andrew Dowd, Vera Gribanova, and Boris Harizanov contributed a lot to our work. In Ireland, discussions with Ailbhe Ní Chasaide and her colleagues in the phonetics laboratory of Trinity College Dublin were extremely helpful, as were discussions with Lillis Ó Laoire (NUI Galway) and Brian Ó Curnáin (Dublin Institute for Advanced Studies). Important financial support was provided by the CrISP group (http://stanford.edu/~gribanov/CrISP.html) and by the the Social Sciences and Humanities Research Council of Canada. Most importantly we benefited from the work of two reviewers for Linguistic Inquiry, whose suggestions and questions improved the article immeasurably.

2 Here and below, we occasionally highlight the position of a postposed pronoun by placing it in a box; we also indicate the syntactically expected position of the pronoun by way of the symbol __.

3 Many of the examples used in this article have been taken from published sources. When this is the case, it is indicated by way of a tag consisting of an abbreviation of the title of the publication followed by the page number on which the example appears, or the date of broadcast in the case of material excerpted from radio broadcasts. The abbreviations are explained in appendix B.

4Stenson (1981:42–45) and Ó Siadhail (1989:207–210) provide clear overviews of the basic facts.

5 See, for instance, Lucas 1979:93, Ó Baoill 1996:88–89, and Ó Sé 2000:155–159 on the distinction between weak and strong forms of the pronouns in Munster and Ulster varieties.

6 The text here glosses over certain very important questions about phonetic realization. As a matter of exposition, we will continue to use the terms strong pronoun and accented pronoun as virtual synonyms. But strictly speaking it is inaccurate to say that ‘‘strong’’ forms of pronouns are always accented, and even more inaccurate to say that weak forms of pronouns never bear a pitch accent. Probably the most accurate statement is that strong pronouns can bear an independent pitch accent (a target tone, in more technical parlance). We say ‘‘independent’’ because there are circumstances in which weak pronouns may end up bearing an accent, but only in virtue of being at the right edge of a larger prosodic domain that happens to carry a final boundary tone for other reasons (see McCloskey 2011b for one such case). We might call these ‘‘dependent’’ pitch accents. None of this gainsays the fundamental point, recognized by all observers, that pronouns in the language come in strong and weak variants.

7 This observation holds for the modern language, but the facts were somewhat different in earlier periods. See Ahlqvist 1975/6, Breatnach 1994:269–270, and (for some brief discussion) Bennett, Elfner, and McCloskey 2015.

8 The only proposals that we know of that attribute such force to pronoun postposing are those in Mulkern 2003, 2011. We consider those proposals in detail in Bennett, Elfner, and McCloskey 2015.

9 For arguments against the remnant movement proposal, see Adger 1997:12–16, Doherty 1997, Doyle 1998, and McCloskey 1999.

10 Similar observations surely hold of what look like small clause relative modifiers of the type in (i).

  • (i)

    • a.

      Níor mhórán lá imthighthe iad.

      NEG.PAST many days gone them

      ‘It wasn’t many days that they had been gone.’

      (ONH 58)

  • b.

    Níor mhórán achair ag cainnt liom é.

    NEG.PAST much length-of-time.GENPROG speak with.me him

    ‘It wasn’t a long time that he had been talking to me.’

    (ONH 93)

11 The syntactic structures presented below are largely plausible, we believe (for discussion and defense of many of the assumptions made, see Chung and McCloskey 1987, McCloskey 2011c), but the details are rarely crucial for the points we want to establish.

12Chung and McCloskey (1987) in fact take the particle ’na to be a Pred head, which selects predicative nominals (NP rather than DP) and which agrees with the subject of the predication. On that view, SC of (30) is PredP and coordination is at the level of the immediate projection of Pred. None of these issues is central for our concerns here.

13 Technically, whether (29) would involve a violation of Ross’s (1967) Coordinate Structure Constraint depends on whether we understand that constraint as applying to lowering rules (or movement into, rather than out of, coordinate structures). To address that issue, we would first have to concede the legitimacy of syntactic lowerings. Apparent subject lowerings in Chamorro (Chung 1990) and Tagalog (Sabbagh 2005) raise similar issues.

14 There are important dissenters from this point of view—Wagner (2005, 2010) and Pak (2008), for instance—but indirect reference theories have represented the mainstream view since the mid-1980s. See, for instance, Selkirk 1984: esp. chap. 8 and Nespor and Vogel 1986.

15 We could syntacticize this mapping principle by assuming that illocutionary force is signaled by a designated functional head, perhaps one of the heads in the C layer, as in Rizzi 1997, and that it is the presence of this head that the mapping principles are sensitive to. Alternatively, we could assume a more direct linkage between pragmatic and prosodic principles.

16 (35) also licenses a structure involving nonbranching recursion, in which there are two distinct Φ-phrases corresponding to XP and YP, respectively. It is a question whether this additional option should be allowed. Elfner (2012) presents evidence from Irish that it should not. If the possibility of nonbranching recursion is to be excluded in principle, we should add a requirement that prosodic structures should be the most economical structures (in terms of number of nodes) that satisfy matching and other requirements (see Selkirk 1996). The effect of such a requirement would again be to ‘‘flatten’’ prosodic representations. See also Inkelas and Zec 1995, Uechi 1998, Truckenbrodt 1999, and Tokizaki 2006.

17 Whether this prioritizing emerges from a language-particular constraint ranking or is more deeply embedded in the mapping system is a question we cannot take up here.

18 We implement this principle below by assuming that the constraint BINARITY outranks MATCH PHRASE, thereby licensing departures from strict syntax-prosody isomorphisms when such isomorphisms would include nonbranching structure. If languages vary on this point, MATCH PHRASE and BINARITY must be constraints that can compete, as we assume here.

19 If trace realization is one component of linearization (as in Frampton 2004 or Fox and Pesetsky 2005), (41) emerges as the direct outcome of those procedures.

20 When the clause is a root clause, or otherwise carries illocutionary force, the topmost node of (42) will in fact be an ι-phrase. For expository purposes, we will ignore this topmost layer of structure except when it is directly relevant to our concerns.

21 Sound files for some of the examples used in this article are available at http://www.mitpressjournals.org/doi/suppl/10.1162/LING_a_00209. Their filenames are keyed to example numbers so that, for instance, the file corresponding to example (46a) is ‘‘46a.wav.’’ The sources from which the files were extracted are given in appendix A.

22 For relevant discussion and alternatives, see for instance Ito and Mester 1992, 2006, 2009a, Ghini 1993, Mester 1994, Inkelas and Zec 1995, and Selkirk 2000.

23 Things are less clear for the second clause of the coordination, where information structure effects having to do with parallelism add an extra layer of complexity, resulting for instance in deaccenting, and perhaps proclisis, of the repeated finite verb.

24 There is a simplification here, in that the ranking in (54) suggests that MATCH PHRASE can always be sacrificed to satisfy eurhythmic constraints like BINARITY and EQUAL SISTERS. Given the logic of constraint ranking in classical Optimality Theory, this would lead us to expect that (44), for example, should be ill-formed (because it violates EQUAL SISTERS but satisfies MATCH PHRASE).

As Elfner (2012) points out, this difference in well-formedness correlates with the number of constraint violations: (51), which violates EQUAL SISTERS twice, undergoes rebracketing as in (52); but (44), which violates EQUAL SISTERS only once, does not. What is needed, it seems, is a theory that is more tolerant of structures that violate a relatively high-ranked constraint. Such less-than-optimal structures would be recognized by the grammar but defined as flawed to some degree, perhaps as reflected in frequency of use. See Elfner 2012:esp. chap. 4 for an account of syntax-prosody interactions in such a framework—Harmonic Grammar (Legendre, Miyata, and Smolensky 1990), in which constraints are weighted rather than ranked. Space limitations prevent us from addressing these important questions more extensively here.

25 This is true except for the case in which the object is syntactically minimal (a pronoun or other one-word XP) and is the only surviving element of vP. In this circumstance, no Φ-phrase will be projected corresponding to vP.

26 Our proposals share an important intuition with Adger’s (1997) account: namely, that normal syntactic routines risk placing weak pronouns in positions incompatible with their inherent stresslessness and are therefore overridden. On Adger’s account, most objects must shift leftward to a VP-external position, a position to which nuclear stress is necessarily assigned. Weak pronouns do not participate in this otherwise obligatory movement for reasons of case licensing. They therefore remain in their thematic positions, low in the VP and in an unstressed position. Weak pronouns are then ordered with respect to other VP-internal elements, such as adverbs, on purely prosodic grounds. On that proposal, weak pronouns appear further to the right than other direct objects not because they have displaced rightward in the syntax, but because they have failed to undergo an otherwise required leftward movement and are thus eligible for prosodic scrambling within the VP. We will not attempt here a detailed comparison between Adger’s proposals and our own, in part at least because Scottish Gaelic is the main focus of Adger’s investigation and we do not know if the kinds of observations that support our conclusions for Irish extend to that language. It is distinctly possible (as Adger has suggested tentatively to us) that the facts are different in Scottish Gaelic in relevant ways and that the two varieties work differently, despite superficial commonalities.

27 See Elfner 2012:157 for a slightly different formulation, with a subtly different range of empirical consequences. Both of these versions in turn differ slightly from the formulation in Selkirk 2011. The formulation in (55) has as a consequence that a prosodic word appearing at the left edge of a prosodic constituent with a Φ-phrase as a right sister will not induce a violation. See our discussion of VSO structures at (42) for evidence that this is the correct interpretation. STRONG START is clearly active in the phonology of Irish; see Elfner 2012:chaps. 3–4 for extensive discussion of the role it plays in shaping other aspects of prosodic constituency in Conamara Irish. Harizanov (to appear) discusses a range of effects in Macedonian and Bulgarian that he attributes to STRONG START, in a formulation close to that in (55). Sabbagh (2013) discusses a counterconstraint, WEAK START, which seems to drive certain apparent lowerings in Tagalog.

28 One might take STRONG START to be a positionally sensitive (i.e., left-edge) version of EQUAL SISTERS.

29 For the syntactic analogue, see May 1985, Chomsky 1986, 1995b:chap. 4, 2001, Kracht 1998, Chametzky 2003, and McCloskey 2006, among many others. Our presentation here is informal, but see Kracht 1998 for a full formal treatment.

30 Other factors, of course, may intervene to favor one option over the other. Given our general approach, those factors should be prosodic. In that light, consider a small variation on (60).

graphic

If the object pronoun is weak, then of course postposing (Option A) is possible.

graphic

For the two in-situ options, we have (iii): the strong pronoun in (iiia), enclisis to the left in (iiib).

graphic

In this case, however, Option B is strongly dispreferred in comparison with the other two options, at least for the small number of speakers we have so far been able to consult. We can understand this contrast by observing that when the pronoun is incorporated into the first Φ-phrase , avoiding a violation of STRONG START, it leaves in its wake a violation of BINARITY in the second Φ-phrase (the preposition chun is a stressless proclitic /(ә)n/). In this respect (iiib) contrasts crucially with (60), whose second Φ-phrase contains two accented elements—fhad and scoile—and therefore two prosodic words. It is in full compliance with both BINARITY and EQUAL SISTERS. There is no syntactic difference at all between (64b) (optimal) and (iiib) (strongly dispreferred). There can therefore be no syntactic understanding of the sharp difference in acceptability that separates them (see also Carnie 2013).

31 The tableaux that follow presuppose the rebracketings of verb and subject described earlier.

32 Note that postposing must be triggered by STRONG START rather than EQUAL SISTERS: the output of postposing (70a) violates EQUAL SISTERS at least as much as the ungrammatical (70c), which leaves the weak pronoun in situ (this also suggests a ranking of STRONG START >> EQUAL SISTERS, which we omit here). See Elfner 2012 for a system in which STRONG START also conditions rebracketings like (52), which we attribute instead to the influence of EQUAL SISTERS (section 6.2).

33 There is surely more to be said about interactions between prosodic adjunction and the family of binarity constraints. See Ito and Mester 2006:sec. 4.1 for a perceptive overview.

34 For additional discussion and for arguments that these cases and pronoun postposing reflect the same phenomenon, see McCloskey 1999.

35 For the syntactic analysis of such items, see McCloskey and Hale 1984, McCloskey 2011a and references cited there, especially Brennan 2008.

36Doherty (1996:23) in fact argues that the operation in question is simply head movement, citing evidence that it is subject to the Coordinate Structure Constraint. This position is consistent with the claim that incorporation is postsyntactic if head movement is itself postsyntactic (Chomsky 2000, Boeckx and Stjepanović 2001, Harley 2004). The evidence is difficult to interpret, but what matters for now is the conclusion that, whatever mechanism is at play, the object that emerges is a single morphological word.

37 If Doherty (1996) is wrong about the status of Subject Pronoun Incorporation and it is in fact a prosodic incorporation, then more must be said. Specifically, (the constraint that drives) Subject Pronoun Incorporation would have to be prioritized over postposing. It is all the more crucial in this scenario that pronoun postposing be understood in prosodic terms, so that it can compete with postposing. A relevant observation in weighing these alternatives is that in earlier stages of Irish, pronoun postposing did in fact apply to subject pronouns (Ahlqvist 1975/6, Breatnach 1994:269–270, Bennett, Elfner, and McCloskey 2015). The ultimately correct account, then, must allow for this variation and must allow a reasonable understanding of what changed between the 11th century and the 20th century.

38 Or equivalently: the entire three-segment category vP is maximal.

39 One aspect of the phenomenon not touched on here is that in complex VPs, placement of the pronoun in absolute clause-final position is somewhat rare, though clearly well-formed (see also Doyle 1998:45). For detailed discussion of the facts and for a proposal about why this should be so, see Bennett, Elfner, and McCloskey 2015.

40 Although postposed pronouns are unaccented, their vowel is often surprisingly long. Given that on our account postposed pronouns always appear at the right edge of phonological phrases, we might understand this as a right-edge lengthening effect. Alternatively, the lengthening of domain-final pronouns might be a kind of parasitic lengthening triggered by phrase-level boundary tones (see, e.g., Silverman and Pierrehumbert 1990). Specifically, vowel lengthening may be required to provide sufficient vocalic material to realize phrasal melodies at the right edge.

41 Notice that (94) violates BINARITY to a greater extent than (93), given the nonbranching Φ-phrase that dominates ag troid. Such structures may be reparsed into a single, binary Φ-phrase that includes the pronoun, (Φ ag troid é) (see section 7, Elfner 2012:224, and footnote 16 above). Alternatively, violations of BINARITY may simply be tolerated when the only other option is to violate STRONG START. A structure that positions the pronoun to the right of the PP le ridirí will (correctly) satisfy MATCH PHRASE—for the reasons discussed in section 8.3.

42 The manner adverb might be better taken to adjoin to VP than to vP. However, the prosodic outcomes are identical if this is the case. In (97), we parse the PP uirthi as a prosodic word because it is, in syntactic terms, both minimal and maximal; see the discussion of (34).

43 Exactly as in the cases discussed in section 8.1, uirthi too can undergo postposing, which we assume is triggered by the alternative bracketing (Φ (Φ é ag féachaint) (Φ uirthi go drúisiúil)).

  • (i)

    • a.

      Chonac é ag féachaint go drúisiúil uirthi.

    • b.

      Chonac ag féachaint é go drúisiúil uirthi.

      For reasons that we would like to understand better, however, the outcomes in (ii) seem to be impossible.

  • (ii)

    • a.

      *Chonac ag féachaint é uirthi go drúisiúil.

    • b.

      *Chonac ag féachaint go drúisiúil é uirthi.

The sequence of two prosodically dependent elements (one a syllable, one a foot) at the edge of a Φ-phrase seems not to be tolerated. Perhaps this too reflects phrase-level eurhythmic pressures—specifically, an aversion to adjacent unaccented or weak elements (a kind of phrasal lapse avoidance). There is a good deal here that we do not yet understand, however.

44 The interpretation offered here may require us to rethink our position of convenience that strong and weak forms are different lexical items and therefore determine distinct inputs. The logic of the text discussion implies that strong forms may compete with weak forms in determining the outcome in (109).

45 The relative clause cases of footnote 10 should be amenable to the same analysis.

46 There could well be additional functional projections involved in these structures, but since the relevant heads will be null, they will make no difference to the prosodic calculations we are about to discuss.

47 One would of course want to find phonological processes whose domain is the minimal or maximal prosodic word and use such processes to probe questions of constituency. Green (2000) discusses much relevant material, but does not settle the matter. We conjecture, however, that lenition may well be a morphological process whose domain of application is exactly that of complex prosodic words, the adjoined function word being the lenition trigger, the host being the lenited element. Such a theory would successfully bring together the two core classes of cases: lenition of the second element of a compound, as in seanbhean ‘old woman’ and lenition of a lexical word by a function word such as a preposition, complementizer, or determiner, as in an bhean ‘the woman’. If that is correct, then the prosodic structure in (117) must be correct too. Fully exploring this conjecture and defending it, however, would take an article at least as long as the present one. We spare our current readers that pain.

48 This requirement follows from Selkirk’s (1996) HEADEDNESS clause: prosodic words must contain a stressable element. [gәn] being weak, the pronoun [eː] must appear in its strong, stressable form.

49 An alternative one might explore is that ordering statements that position heads may be prioritized over the various prosodic constraints we have been exploring here. If it were the case, for example, that a constraint HEAD INITIAL outranked STRONG START, which in turn outranked the constraints ordering specifiers and complements, the results we want would be guaranteed. As pointed out in the text discussion, the elements that do in fact postpose ( pronouns and single-word PPs) are never complement-taking heads.

Reference

Ackema,
Peter
, and
Ad
Neeleman
.
2003
.
Context-sensitive Spell-Out
.
Natural Language and Linguistic Theory
21
:
681
735
.
Acquaviva,
Paolo
.
2014
.
The categories of Modern Irish verbal inflection
.
Journal of Linguistics
50
:
537
586
.
Adger,
David
.
1997
.
VSO order and weak pronouns in Goidelic Celtic
.
Canadian Journal of Linguistics
42
:
9
29
.
Adger,
David
.
2007a
.
Pronouns postpose at PF
.
Linguistic Inquiry
38
:
343
349
.
Adger,
David
.
2007b
.
Stress and phasal syntax
.
Linguistic Analysis
33
:
238
266
.
Agbayani,
Brian
, and
Chris
Golston
.
2010
.
Phonological movement in Classical Greek
.
Language
86
:
133
167
.
Agbayani,
Brian
,
Chris
Golston
, and
Dasha
Henderer
.
2010
.
Prosodic movement
. In
WCCFL 28
, ed. by
Mary Byram
Washburn
,
Katherine
McKinney-Bock
,
Erika
Varis
,
Ann
Sawyer
, and
Barbara
Tomaszewicz
,
231
239
.
Somerville, MA
:
Cascadilla Proceedings Project
.
Ahlqvist,
Anders
.
1975/6
.
On the position of pronouns in Irish
.
Éigse
16
:
171
176
.
Anderson,
Stephen R
.
2005
.
Aspects of the theory of clitics
.
Oxford
:
Oxford University Press
.
Anttila,
Arto
.
2002
.
Morphologically conditioned phonological alternations
.
Natural Language and Linguistic Theory
20
:
1
42
.
Anttila,
Arto
.
2008
.
Phonological constraints on constituent ordering
. In
WCCFL 26
, ed. by
Charles B.
Chang
and
Hannah
Haynie
,
51
59
.
Somerville, MA
:
Cascadilla Proceedings Project
.
Anttila,
Arto
,
Matthew
Adams
, and
Michael
Speriosu
.
2010
.
The role of prosody in the English dative alternation
.
Language and Cognitive Processes
25
:
946
981
.
Bennett,
Ryan
.
2008
.
Donegal Irish and the syntax/prosody interface
. .
Bennett,
Ryan
,
Emily
Elfner
, and
James
McCloskey
.
2015
.
Pronouns and prosody in Irish
. In
Proceedings of the XIV International Congress of Celtic Studies, Maynooth 2011
, ed. by
Liam
Breatnach
,
Ruairí Ó
hUiginn
,
Damian
McManus
, and
Katharine
Simms
,
19
74
.
Dublin
:
Dublin Institute for Advanced Studies, School of Celtic Studies
.
Berwick,
Robert C
., and
Noam
Chomsky
.
2011
.
The biolinguistic program: The current state of its development
. In
The biolinguistic enterprise: New perspectives on the evolution and nature of the human language faculty
, ed. by
Anna Maria Di
Sciullo
and
Cedric
Boeckx
,
19
41
.
Oxford
:
Oxford University Press
.
Bobaljik,
Jonathan David
, and
Andrew
Carnie
.
1996
.
A Minimalist approach to some problems of Irish word order
. In
The syntax of the Celtic languages: A comparative perspective
, ed. by
Robert
Borsley
and
Ian
Roberts
,
223
240
.
Cambridge
:
Cambridge University Press
.
Boeckx,
Cedric
, and
Sandra
Stjepanović
.
2001
.
Heading toward PF
.
Linguistic Inquiry
32
:
345
355
.
Breatnach,
Liam
.
1994
.
An Mheán-Ghaeilge
. In
Stair na Gaeilge
, ed. by
Kim
McCone
,
Damien
McManus
,
Cathal Ó
Háinle
,
Nicholas
Williams
, and
Liam
Breatnach
,
221
333
.
Coláiste Phádraig, Maigh Nuad
:
Roinn na Sean-Ghaeilge
.
Brennan,
Jonathan
.
2008
.
Prepositions in Modern Irish: Agreement and impoverishment
. In
WCCFL 26
, ed. by
Charles B.
Chang
and
Hannah
Haynie
,
105
113
.
Somerville, MA
:
Cascadilla Proceedings Project
.
Carnie,
Andrew
.
2013
.
Phonologically motivated pronoun post-posing in Scottish Gaelic
. In
Microvariation, minority languages, minimalism and meaning: Proceedings of the Irish Network in Formal Linguistics
, ed. by
Catrin
Rhys
,
Pavel
Iosad
, and
Alison
Henry
,
91
100
.
Newcastle upon Tyne
:
Cambridge Scholars Publishing
.
Chametzky,
Robert
.
2003
.
Phrase structure: From GB to Minimalism
.
Malden, MA
:
Blackwell
.
Chomsky,
Noam
.
1986
.
Barriers
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
1995a
.
Bare phrase structure
. In
Government and Binding Theory and the Minimalist Program
, ed. by
Gert
Webelhuth
,
383
439
.
Cambridge, MA
:
Blackwell
.
Chomsky,
Noam
.
1995b
.
The Minimalist Program
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2000
.
Minimalist inquiries: The framework
. In
Step by step: Essays on Minimalist syntax in honor of Howard Lasnik
, ed. by
Roger
Martin
,
David
Michaels
, and
Juan
Uriagereka
,
89
155
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2001
.
Derivation by phase
. In
Ken Hale: A life in language
, ed. by
Michael
Kenstowicz
,
1
52
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2008
.
On phases
. In
Foundational issues in linguistic theory: Essays in honor of Jean-Roger Vergnaud
, ed. by
Robert
Freidin
,
Carlos
Otero
, and
Maria Luisa
Zubizarreta
,
133
166
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
, and
Howard
Lasnik
.
1993
.
The theory of principles and parameters
. In
Syntax: An international handbook of contemporary research
, ed. by
Joachim
Jacobs
,
Arnim von
Stechow
,
Wolfgang
Sternefeld
, and
Theo
Vennemann
,
506
569
.
Berlin
:
Walter de Gruyter
. Also published in Noam Chomsky, The Minimalist Program, 13–127. Cambridge, MA: MIT Press (1995).
Chung,
Sandra
.
1990
.
VP’s and verb movement in Chamorro
.
Natural Language and Linguistic Theory
8
:
559
619
.
Chung,
Sandra
.
2003
.
The syntax and prosody of weak pronouns in Chamorro
.
Linguistic Inquiry
34
:
547
599
.
Chung,
Sandra
, and
James
McCloskey
.
1987
.
Government, barriers, and small clauses in Modern Irish
.
Linguistic Inquiry
18
:
173
237
.
Coetzee,
Andries
, and
Joe
Pater
.
2011
.
The place of variation in phonological theory
. In
The handbook of phonological theory
, ed. by
John A.
Goldsmith
,
Jason
Riggle
, and
Alan C. L.
Yu
,
401
434
.
2nd ed. Malden, MA
:
Wiley-Blackwell
.
Culicover,
Peter W
., and
Ray
Jackendoff
.
2005
.
Simpler syntax
.
Oxford
:
Oxford University Press
.
Doherty,
Cathal
.
1996
.
Clausal structure and the Modern Irish copula
.
Natural Language and Linguistic Theory
14
:
1
48
.
Doherty,
Cathal
.
1997
.
Particles and projections in Irish syntax: A review
.
Linguistic Analysis
26
:
227
251
.
Doyle,
Aidan
.
1998
.
Obligatory and optional movement: The evidence from Irish
. In
Projections and mapping: Studies in syntax
, ed. by
Piotr
Stalmaszczyk
,
43
56
.
Lublin
:
Wydawnictwo Folium
.
Duffield,
Nigel
.
1995
.
Particles and projections in Irish syntax
.
Dordrecht
:
Kluwer
.
Elfner,
Emily
.
2011a
.
The interaction of linearization and prosody: Evidence from pronoun postposing in Irish
. In
Formal approaches to Celtic linguistics
, ed. by
Andrew
Carnie
,
17
40
.
Newcastle upon Tyne
:
Cambridge Scholars Publishing
.
Elfner,
Emily
.
2011b
.
Recursive phonological phrases in Conamara Irish
.
In
WCCFL 28 online proceedings, ed. by Mary Byram Washburn, Sarah Ouwayda, Chuoying Ouyang, Bin Yin, Canan Ipek, Lisa Marston, and Aaron Walker
. .
Elfner,
Emily
.
2012
.
Syntax-prosody interactions in Irish
.
Doctoral dissertation, University of Massachusetts, Amherst
.
Elfner,
Emily
.
2013
.
Recursivity in prosodic phrasing: Evidence from Conamara Irish
. In
NELS 40
, ed. by
Seda
Kan
,
Claire
Moore-Cantwell
, and
Robert
Staubs
,
191
204
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Elfner,
Emily
.
2015
.
Recursion in prosodic phrasing: Evidence from Connemara Irish
.
Natural Language and Linguistic Theory
33
:
1169
1208
.
Fox,
Danny
, and
David
Pesetsky
.
2005
.
Cyclic linearization of syntactic structure
.
Theoretical Linguistics
31
:
1
45
.
Frampton,
John
.
2004
.
Copies, traces, occurrences, and all that evidence from Bulgarian multiple wh-phenomena
.
Ms., Northeastern University, Boston, MA. Available at http://www.math.neu.edu/ling/
.
Ghini,
Mirco
.
1993
.
Φ-formation in Italian: A new proposal
.
In
Toronto working papers in linguistics 12.2, ed. by Carrie Dyck, 41–79. Available at twpl.library.utoronto.ca/index.php/twpl/article/view/6343
.
Göbbel,
Edward
.
2007
.
Extraposition as PF movement
. In
Proceedings of WECOL 2006
, ed. by
Erin
Bainbridge
and
Brian
Agbayani
,
132
145
.
Fresno
:
California State University, Department of Linguistics
.
Göksel,
Aslı
,
Barış
Kabak
, and
Anthi
Revithiadou
.
2013
.
Prosodically constrained non-local doubling
.
The Linguistic Review
30
:
185
214
.
Green,
Anthony Dubach
.
2000
.
The prosodic representation of clitics in Irish
. In
Clitics in phonology, morphology, and syntax
, ed. by
Birgit
Gerlach
and
Janet
Grijzenhout
,
181
218
.
Amsterdam
:
John Benjamins
.
Gussenhoven,
Carlos
.
1991
.
The English Rhythm Rule as an accent deletion rule
.
Phonology
8
:
1
35
.
Gutiérrez-Bravo,
Rodrigo
.
2005
.
Subject inversion in Spanish relative clauses
. In
Romance languages and linguistic theory 2003
, ed. by
Twan
Geerts
,
Ivo van
Ginneken
, and
Haike
Jacobs
,
115
128
.
Amsterdam
:
John Benjamins
.
Hall,
T. Alan
.
1999
.
Phonotactics and the prosodic structure of German function words
. In
Studies on the phonological word
, ed. by
T. Alan
Hall
and
Ursula
Kleinhenz
,
99
131
.
Amsterdam
:
John Benjamins
.
Halpern,
Aaron
.
1995
.
Topics in the displacement and morphology of clitics
.
Doctoral dissertation, Stanford University, Stanford, CA
.
Harizanov,
Boris
.
To appear
.
The role of prosody in the linearization of clitics: Evidence from Bulgarian and Macedonian
. In
Formal Approaches to Slavic Linguistics 22
, ed.
by Ivona Kučerová
.
Ann Arbor
:
Michigan Slavic Publications
.
Harley,
Heidi
.
2004
.
Merge, conflation, and head movement
. In
NELS 34
, ed. by
Keir
Moulton
and
Matthew
Wolf
,
239
254
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Henderson,
Robert
.
2012
.
Morphological alternations at the intonational phrase edge: The case of K’ichee’
.
Natural Language and Linguistic Theory
30
:
741
789
.
Horn,
Laurence
.
1986
.
Presupposition, theme, and variations
. In
CLS 22: Papers from the Parasession on Pragmatics and Grammatical Theory
, ed. by
Carlos
Gussenhoven
and
Tomas
Riad
,
168
192
.
Chicago
:
University of Chicago, Chicago Linguistic Society
.
Inkelas,
Sharon
, and
Draga
Zec
.
1995
.
The syntax-phonology interface
. In
The handbook of phonological theory
, ed. by
John A.
Goldsmith
,
535
549
.
Cambridge, MA
:
Blackwell
.
Ito,
Junko
, and
Armin
Mester
.
1992
.
Weak layering and word binarity. LRC Working Paper 92-9. Santa Cruz: University of California, Linguistics Research Center
.
Reprinted in
A new century of phonology and phonological theory: A festschrift for Professor Shosuke Haraguchi on the occasion of his sixtieth birthday
, ed. by
Takeru
Honma
,
Masao
Okazaki
,
Toshiyuki
Tabata
, and
Shin-ichi
Tanaka
,
26
65
.
Tokyo
:
Kaitakusha
.
Ito,
Junko
, and
Armin
Mester
.
2006
.
Prosodic adjunction in Japanese compounds
. In
Formal Approaches to Japanese Linguistics (FAJL 4)
, ed. by
Yoichi
Miyamoto
and
Masao
Ochi
,
97
112
.
MIT Working Papers in Linguistics 55. Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Ito,
Junko
, and
Armin
Mester
.
2009a
.
The extended prosodic word
. In
Phonological domains: Universals and derivations
, ed. by
Baris
Kabak
and
Janet
Grijzenhout
,
135
194
.
Berlin
:
Mouton de Gruyter
.
Ito,
Junko
, and
Armin
Mester
.
2009b
.
The onset of the prosodic word: The final C syndrome, English r-sandhi, and the constraints on syllable and word onsets
. In
Phonological argumentation: Essays on evidence and motivation
, ed. by
Steve
Parker
,
227
260
.
London
:
Equinox
.
Ito,
Junko
, and
Armin
Mester
.
2012
.
Recursive prosodic phrasing in Japanese
. In
Prosody matters: Essays in honor of Elisabeth Selkirk
, ed. by
Toni
Borowsky
,
Shigeto
Kawahara
,
Takahito
Shinya
, and
Mariko
Sugahara
,
280
303
.
London
:
Equinox
.
Ito,
Junko
, and
Armin
Mester
.
2013
.
Prosodic subcategories in Japanese
.
Lingua
124
:
20
40
.
Kandybowicz,
Jason
.
2009
.
Embracing edges: Syntactic and phono-syntactic edge sensitivity in Nupe
.
Natural Language and Linguistic Theory
27
:
305
344
.
Kandybowicz,
Jason
.
2015
.
On prosodic vacuity and verbal resumption in Asante Twi
.
Linguistic Inquiry
46
:
243
272
.
Kayne,
Richard S
.
1994
.
The antisymmetry of syntax
.
Cambridge, MA
:
MIT Press
.
Kayne,
Richard S
.
2000
.
On the left edge in UG: Reply to McCloskey
.
Syntax
3
:
44
51
.
Kracht,
Marcus
.
1998
.
Adjunction structures and syntactic domains
. In
The mathematics of syntactic structure: Trees and their logics
, ed. by
Hans-Peter
Kolb
and
Uwe
Mönnich
,
259
299
.
Berlin
:
Mouton de Gruyter
.
Krivokapić,
Jelena
.
2007
.
The planning, production, and perception of prosodic structure
.
Doctoral dissertation, University of Southern California, Los Angeles
.
Legendre,
Géraldine
,
Yoshiro
Miyata
, and
Paul
Smolensky
.
1990
.
Harmonic Grammar: A formal multilevel connectionist theory of linguistic well-formedness: Theoretical foundations
. In
Proceedings of the Twelfth Annual Conference of the Cognitive Science Society
,
388
395
.
Hillsdale, NJ
:
Lawrence Erlbaum
.
López,
Luis
.
2009
.
Ranking the Linear Correspondence Axiom
.
Linguistic Inquiry
40
:
239
276
.
Lucas,
Leslie W
.
1979
.
Grammar of Ros Goill Irish, County Donegal
.
Belfast
:
Queen’s University Belfast, Institute of Irish Studies
.
Manetta,
Emily
.
2012
.
Reconsidering rightward scrambling: Postverbal constituents in Hindi-Urdu
.
Linguistic Inquiry
43
:
43
74
.
Mascaró,
Joan
.
2007
.
External allomorphy and lexical representation
.
Linguistic Inquiry
38
:
715
735
.
May,
Robert
.
1985
.
Logical Form: Its structure and derivation
.
Cambridge, MA
:
MIT Press
.
McCloskey,
James
.
1991
.
Clause structure, ellipsis and proper government in Irish
.
Lingua
85
:
259
302
.
McCloskey,
James
.
1996a
.
On the scope of verb raising in Irish
.
Natural Language and Linguistic Theory
14
:
47
104
.
McCloskey,
James
.
1996b
.
Subjects and subject-positions in Irish
. In
The syntax of the Celtic languages: A comparative perspective
, ed. by
Robert
Borsley
and
Ian
Roberts
,
241
283
.
Cambridge
:
Cambridge University Press
.
McCloskey,
James
.
1999
.
On the right edge in Irish
.
Syntax
2
:
189
209
.
McCloskey,
James
.
2004
.
Irish nominal syntax 1: Demonstratives
.
Ms., University of California, Santa Cruz. Available at http://ohlone.ucsc.edu/~jim/papers.html
.
McCloskey,
James
.
2006
.
Questions and questioning in a local English
. In
Crosslinguistic research in syntax and semantics: Negation, tense, and clausal architecture
, ed. by
Raffaella
Zanuttini
,
Héctor
Campos
,
Elena
Herburger
, and
Paul H.
Portner
,
87
126
.
Washington, DC
:
Georgetown University Press
.
McCloskey,
James
.
2011a
.
Inflection and silent arguments in Irish
.
Handout for a talk at CASTL, University of Tromsø, September 2011. Available at http://ohlone.ucsc.edu/~jim/papers.html
.
McCloskey,
James
.
2011b
.
Polarity, case, and the cartography of the inflectional layer in Irish
.
Paper presented at GIST 3, Cartographic Structures and Beyond, University of Ghent, 13 May 2011
.
McCloskey,
James
.
2011c
.
The shape of Irish clauses
. In
Formal approaches to Celtic linguistics
, ed. by
Andrew
Carnie
,
143
178
.
Newcastle upon Tyne
:
Cambridge Scholars Publishing
. .
McCloskey,
James
.
2014
.
Irish existentials in context
.
Syntax
17
:
343
384
.
McCloskey,
James
, and
Kenneth
Hale
.
1984
.
On the syntax of person number marking in Modern Irish
.
Natural Language and Linguistic Theory
1
:
487
533
.
Mester,
Armin
.
1994
.
The quantitative trochee in Latin
.
Natural Language and Linguistic Theory
12
:
1
61
.
Mulkern,
Ann
.
2003
.
Cognitive status, discourse salience, and information structure: Evidence from Irish and Oromo
.
Doctoral dissertation, University of Minnesota, Minneapolis and St. Paul
.
Mulkern,
Ann
.
2011
.
Left right behind: Irish pronoun postposing and information structure
. In
Formal approaches to Celtic linguistics
, ed. by
Andrew
Carnie
,
179
202
.
Newcastle upon Tyne
:
Cambridge Scholars Publishing
.
Myrberg,
Sara
.
2010
.
The intonational phonology of Stockholm Swedish
.
Stockholm
:
Acta Universitatis Stockholmiensis
.
Myrberg,
Sara
.
2013
.
Sisterhood in prosodic branching
.
Phonology
30
:
73
124
.
Nespor,
Marina
, and
Irene
Vogel
.
1986
.
Prosodic phonology
.
Dordrecht
:
Foris
.
Ó Baoill,
Dónall P
.
1996
.
An teanga bheo: Gaeilge Uladh
.
Dublin
:
Institiúid Teangeoláıochta É ireann
.
Oda,
Kenji
.
2012
.
Issues in the left periphery of Modern Irish
.
Doctoral dissertation, University of Toronto
.
Ó Sé,
Diarmuid
.
2000
.
Gaeilge Chorca Dhuibhne
.
Dublin
:
Institiúid Teangeoláıochta É ireann
.
Ó Siadhail,
Mícheál
.
1989
.
Modern Irish: Grammatical structure and dialectal variation
.
Cambridge
:
Cambridge University Press
.
Pak,
Marjorie
.
2008
.
The postsyntactic derivation and its phonological reflexes
.
Doctoral dissertation, University of Pennsylvania, Philadelphia
.
Prince,
Alan
, and
Paul
Smolensky
.
1993
.
Optimality Theory: Constraint satisfaction in generative grammar
.
Technical Report RuCCS TR-2, Center for Cognitive Science, Rutgers University, New Brunswick, NJ. Published, Malden, MA: Blackwell (2004)
.
Revithiadou,
Anthi
.
2004
.
Phonological vs interface grammars of P-phrasing
.
Ms., University of the Aegean, Rhodes
.
Revithiadou,
Anthi
, and
Vassilios
Spyropoulus
.
2009
.
A dynamic approach to the syntax-phonology interface
. In
InterPhases: Phase-theoretic investigations of linguistic interfaces
, ed. by
Kleanthes
Grohmann
,
202
233
.
Oxford
:
Oxford University Press
.
Richards,
Norvin
.
2010
.
Uttering trees
.
Cambridge, MA
:
MIT Press
.
Rizzi,
Luigi
.
1990
.
Relativized Minimality
.
Cambridge, MA
:
MIT Press
.
Rizzi,
Luigi
.
1997
.
The fine structure of the left periphery
. In
Elements of grammar
, ed. by
Liliane
Haegeman
,
281
337
.
Dordrecht
:
Kluwer
.
Ross,
John R
.
1967
.
Constraints on variables in syntax
.
Doctoral dissertation, MIT, Cambridge, MA. Published as Infinite syntax!, Norwood, NJ: Ablex (1986)
.
Sabbagh,
Joseph
.
2005
.
Non-verbal argument structure: Evidence from Tagalog
.
Doctoral dissertation, MIT, Cambridge, MA
.
Sabbagh,
Joseph
.
2013
.
Word order and prosodic structure constraints in Tagalog
.
Syntax
17
:
40
89
.
Selkirk,
Elisabeth
.
1984
.
Phonology and syntax: The relation between sound and structure
.
Cambridge, MA
:
MIT Press
.
Selkirk,
Elisabeth
.
1996
.
The prosodic structure of function words
. In
Signal to syntax
, ed. by
James L.
Morgan
and
Katherine
Demuth
,
187
213
.
Mahwah, NJ
:
Lawrence Erlbaum
.
Selkirk,
Elisabeth
.
2000
.
The interaction of constraints on prosodic phrasing
. In
Prosody: Theory and experiment—Studies presented to Gösta Bruce
, ed. by
Gösta
Bruce
and
Merle
Horne
,
231
261
.
Dordrecht
:
Kluwer
.
Selkirk,
Elisabeth
.
2009
.
On clause and intonational phrase in Japanese: The syntactic grounding of prosodic constituent structure
.
Gengo Kyenku
136
:
35
73
.
Selkirk,
Elisabeth
.
2011
.
The syntax-phonology interface
. In
The handbook of phonological theory
, ed. by
John A.
Goldsmith
,
Jason
Riggle
, and
Alan C. L.
Yu
,
435
484
. 2nd ed.
Malden, MA
:
Wiley-Blackwell
.
Silverman,
Kim
, and
Janet
Pierrehumbert
.
1990
.
The timing of pre-nuclear high accents in English
. In
Papers in laboratory phonology 1: Between the grammar and the physics of speech
, ed. by
John
Kingston
and
Mary
Beckman
,
72
106
.
Cambridge
:
Cambridge University Press
.
Stenson,
Nancy
.
1981
.
Studies in Irish syntax
.
Tübingen
:
Max Niemeyer Verlag
.
Tilsen,
Sam
.
2011
.
Metrical regularity facilitates speech planning and production
.
Laboratory Phonology
2
:
185
218
.
Tilsen,
Sam
.
To appear
.
Utterance preparation and stress clash: Planning prosodic alternations
. In
Speech production and perception: Planning and dynamics
, ed. by
Susanne
Fuchs
,
Pascal
Perrier
,
Melanie
Weirich
, and
Daniel
Pape
.
Frankfurt am Main
:
Peter Lang Verlag
.
Tokizaki,
Hisao
.
2006
.
Linearizing structure with silence: A Minimalist theory of syntax-phonology interface
.
Doctoral dissertation, University of Tsukuba
.
Truckenbrodt,
Hubert
.
1999
.
On the relation between syntactic phrases and phonological phrases
.
Linguistic Inquiry
30
:
219
255
.
Uechi,
Akihiko
.
1998
.
An interface approach to topic/focus structure
.
Doctoral dissertation, University of British Columbia
.
Vicente,
Luis
.
2005
.
Word order variation in Basque as prosodically-driven movement
.
Ms., Universität Potsdam
.
Wagner,
Michael
.
2005
.
Prosody and recursion
.
Doctoral dissertation, MIT, Cambridge, MA
.
Wagner,
Michael
.
2010
.
Prosody and recursion in coordinate structures and beyond
.
Natural Language and Linguistic Theory
28
:
183
237
.
Werle,
Adam
.
2009
.
Word, phrase, and clitic prosody in Bosnian, Serbian, and Croatian
.
Doctoral dissertation, University of Massachusetts, Amherst
.
Zec,
Draga
.
2005
.
Prosodic differences among function words
.
Phonology
22
:
77
112
.
Zec,
Draga
, and
Sharon
Inkelas
.
1990
.
Prosodically constrained syntax
. In
The phonology-syntax connection
, ed. by
Sharon
Inkelas
and
Draga
Zec
,
365
378
.
Stanford, CA, and Chicago
:
CSLI Publications and University of Chicago Press
.
Zubizarreta,
Maria Luisa
.
1998
.
Prosody, focus, and word order
.
Cambridge, MA
:
MIT Press
.

Supplementary data