Abstract

Legate (2008) proposes that the postsyntactic component of the grammar plays an important role in morphological case phenomena. Therefore, it is predicted that postsyntactic processes like Linearization may affect morphological case assignment. I argue that this prediction is correct by demonstrating that prepositional case in Scottish Gaelic is assigned solely under linear adjacency with P. To account for this, I propose that Morphological Case Assignment is a postsyntactic process derivationally ordered after Linearization, as in Arregi and Nevins 2012. This accounts for several typologically rare properties, such as the presence of closest conjunct case.

Legate (2008:55) proposes that “abstract Case features are determined syntactically and realized in a postsyntactic morphological component . . . resulting in an imperfect relationship between syntax and morphology.” For Legate, case morphology is determined in two steps. The first restricts where in a syntactic structure a nominal may occur. The second calculates the surface shape of the nominal. Legate argues that the two processes track each other as faithfully as possible, but mismatches may result from opaque mappings from syntax to morphology.

In Legate’s framework of Distributed Morphology (Halle and Marantz 1993, 1994), syntactic structure contains only abstract feature bundles, without words in the familiar sense. Once the operations of syntax are complete, the structure is processed further in order to render abstract syntactic representations interpretable at PF.

These claims lead to the question addressed in this article: to what extent can postsyntactic processing determine the distribution of case morphology? To probe this question, I examine the distribution of case morphology in Scottish Gaelic. I show that prepositional case in Gaelic is sensitive to linear adjacency, not syntactic operations like Agree. If linear order is determined postsyntactically (e.g., Chomsky 1981 et seq.), then case morphology is predicted to be determined postsyntactically, contra proposals in Bobaljik 1995, Haeberli 2001, Preminger 2014, and Deal 2017.

In section 4, I propose that Morphological Case Assignment is a distinct postsyntactic process that is derivationally ordered after Linearization. The core proposal is in the order of operations in (1).

(1)

graphic

Because Morphological Case Assignment occurs late in this system, nominals may show sensitivity to the output of operations further upstream. This can include syntactic processes like Agree, as well as postsyntactic processes like Linearization, as argued here. In this way, (1) allows for the output of Linearization to determine case morphology while being compatible with earlier proposals, such as Baker’s (2015).

Section 1 begins with an examination of the morphological case inventory of Gaelic. This discussion is necessary because Gaelic marks case morphology through several complex morphophonological processes. Readers with only a passing interest in the peculiarities of Gaelic are invited to continue to section 2, which investigates linear adjacency effects on prepositional case assignment. Section 3 separates ϕ-agreement from linear adjacency, showing that prepositional case morphology tracks only linear relations. Section 4 lays out the theoretical architecture shown in (1) in more detail. Of particular interest are the obligatory closest conjunct case effects that prepositional case demonstrates. Section 5 concludes and provides several further comments on the scope of the proposal.

1 Scottish Gaelic Case Morphology

Scottish Gaelic is a Celtic language spoken by approximately 75,000 people in Scotland, primarily in the Hebridean archipelago off the country’s west coast. The majority of speakers live in the Outer Hebrides, although there are significant and stable communities in the Inner Hebrides, notably on the Isle of Skye. Given the insular nature of these communities, inter- and intraisland dialect variation can be great. In addition to the variation between communities, there are significant differences among generations and age levels within communities.

This article describes two broad varieties that I term conservative Hebridean Gaelic and innovative Hebridean Gaelic. Conservative Hebridean Gaelic is spoken by the older generations, ranging from about 60 on, in the Inner and Outer Hebrides, particularly Skye, Barra, North and South Uist, and Benbecula. It is also robustly attested in published works from the early to late 1900s. Innovative Hebridean Gaelic is the natural continuation of conservative dialects, spoken by the next younger generation. These speakers are now in their early 30s to late 50s and live in the same communities, especially Skye and Tiree in the Inner Hebrides, and Barra, North and South Uist, and Benbecula in the Outer Hebrides.

In this article, I rely primarily on published sources. Conservative and innovative Hebridean speakers were consulted during the summers of 2016 and 2017 for supplemental data. No significant dialect differences were found in the crucial respects, and the judgments of innovative speakers matched those of conservative speakers. The primary distinction between the two varieties for present purposes comes from the morphological case inventories they employ. Consider the typical definite DP case inventories in (2)–(3), respectively, for the masculine noun cat [khaht] ‘cat’ and the feminine noun caileag [khaljak] ‘girl’, each modified by beag [pφk] ‘little’. See Macaulay 1992, Gillies 2009, and Adger 2010a.1

(2)

graphic

(3)

graphic

(2)–(3) show that both varieties distinguish case and gender primarily through morphophonological processes. These include lenition, marked orthographically by an h after the first consonant, and final palatalization, marked by an i before the final consonant. Throughout, I mark lenited forms with an initial superscript +L and final palatalization with a final superscript +J.2

Let us step through the conservative inventory in (2). The masculine, unmarked nominal lacks both lenition and final palatalization, while in the prepositional case, both the head noun and the adjective are lenited, causing an cat beag [ǝn khaht pφk] ‘the little cat’ to surface as a’ chat bheag [ǝ xaht vφk].3 Likewise, the genitive is marked by both processes. As a result, an cat beag is pronounced as a’ chait bhig [ǝ xahtj vφkj].

Conservative feminine nouns are distinguished by a different distribution of these morphophonological processes. In the unmarked case, feminine nouns are marked by lenition both on the head noun and on adjectives. Therefore, caileag [khaljǝk] ‘girl’ and beag ‘little’ surface as a’ chaileag bheag [ǝ xaljak vφk]. In the prepositional form, both the nominal and the adjective are marked by final palatalization as well as lenition, yielding [ǝ xaljıkj vıkj]. Finally, the feminine genitive is the only form that is not marked by lenition. Instead, the definite article has a unique allomorph na [nǝ], and the noun and the adjective undergo mandatory final palatalization. In careful or overly formal pronunciation, -e [φ ~ ǝ] may be suffixed to both the noun and the adjective. In this case, the phrase is pronounced [nǝ khaljıkj(ǝ) pıkj(ǝ)].

In the innovative paradigm in (3), we see that some of these distinctions are lost. For instance, the prepositional case for both grammatical genders is marked only by lenition, the final palatalization in the feminine being lost. Likewise, final palatalization in the feminine genitive is generally lost, though the distinct feminine genitive definite article na is preserved.

Given that these categories are marked almost exclusively through intricate morphophonological processes, one may wonder if they are truly comparable to morphological case in other, more agglutinative languages. Let us review some evidence that supports the identification of these categories with morphological case.

First, the instances of Initial Consonant Mutation in (2)–(3) are radically different from the behavior of lenition elsewhere in the language. See Macaulay 1992, Gillies 2009, and Bosch 2010 for detailed discussions of Gaelic lenition. Usually, Initial Consonant Mutation behaves like an autosegment in Goldsmith’s (1976) sense: lenition is idiosyncratically associated with particular words and must dock onto the following word in a straightforward way.4 Consider the contrasts in (4)–(5).

(4)

  • a bràthair beag

    her brother little

    ‘her little brother’

  • *a +Lbhràthair beag

(5)

  • a +Lbhràthair beag

    his brother little

    ‘his little brother’

  • *a bràthair beag

  • *a bràthair +Lbheag

  • *a +Lbhràthair +Lbheag

In (4) we see that some words, like a [ǝ] ‘her’, never cause lenition. In comparison, in (4)–(5) we see that a [ǝ] ‘his’ and a [ǝ] ‘her’ are segmental minimal pairs distinguished solely in that only a ‘his’ triggers lenition. From this comparison, we see that the property of triggering lenition is unpredictably associated with particular lexical items and cannot be predicted morphophonologically.

In the (a)–(b) examples of (4)–(5), we see that lenition, where licensed, is obligatory. Now consider (5c–d). Here, we see that lenition associates only with the linearly closest word. In other words, lenition is a highly local process that may only involve a lexical lenition trigger and the following word.

These patterns differ from those of prepositional case in (2)–(3). Consider (6).

(6)

  • Thill Mamaidh [leis an ultach+Lbheag+Lphrìseil].

    returned Mommy with.DEF the bundle little precious

    ‘Mommy returned with the precious little bundle.’

    (CC, pg. 18)

  • Bha mi a’ bruidhinn [ris a’ L+chloinn+J L+mhòir+J].

    was I PROG speak.VN with.DEF the children big

    ‘I was speaking with the big children.’

    (Speaker: Griomsay Gaelic)

First, in (6a) we see that ultach bheag phrìseil ‘precious little bundle’ shows lenition through the entire phrase.5 This cannot be reduced to the pattern in (5) because in (5), no following adjectives can be targeted. Prepositional case assignment, as in (6a), is therefore different from regular lenition. This conclusion also holds for the lenition and final palatalization in (6b). Instead, these patterns are more consistent with gender and case concord, as I assume here.

Second, there are nouns that show suppletion for prepositional case, as illustrated in (7).6

(7)

  • bean ~ mnaoi

    [pφn] [mɾ

    graphic
    ]

    ‘woman ~ woman.PREP

  • bò ~ boin

    [poː] [ponj]

    ‘cow ~ cow.PREP

Mnaoi ‘woman’ and boin ‘cow’ do not occur elsewhere in the paradigms of these nouns and cannot be related via living phonological mechanisms to other forms. Therefore, these are suppletive forms.

These suppletive forms occur only in prepositional phrases, consistent with the view that such forms bear a distinct prepositional case. This is demonstrated with boinCOW.PREP’ in (8).

(8)

  • Bha aca, ’s bha ’m bain’ [aig a’ +Lbhoin] cho math.

  • was cow at.them and was the milk at the cow so good

  • ‘They had a cow, and the cow’s milk was so good.’

  • (BSR, pg. 10)

In (8), we see that

graphic
surfaces as when not governed by a preposition, but as boin in prepositional object position. This is straightforwardly consistent with prepositional case.

Further evidence that the suppletive stems in (8) bear prepositional case is that adjectival modifiers to these stems show the same concord as with other feminine nouns in the same environment. In (9), we see that adjectival modifiers of these suppletive stems undergo both lenition and final palatalization, as do the modifiers of other feminine nouns in this environment. See (2)–(3).7

(9)

  • . . . gun doireadh tu fiar [dha ’n +Lbhoinn+Lghlais+J ].

    C give.COND you grass for the cow gray.PREP

    ‘. . . that you would give grass for the gray cow.’

    (TB, pg. 152)

  • Dh’fhàg iad [leis a’ +Lmhnaoi òig+J ] e.

    left they with the woman young.PREP him

    ‘They left him with the young woman.’

    (MM, pg. 74)

Analyzing these suppletive stems as allomorphs of prepositional case allows us to straightforwardly account for the distribution of suppletive stems by positing Vocabulary entries such as (10).

(10) Vocabulary entries for COW and WOMAN

  • graphic
    [poː]

  • graphic
    bean [pφn]

  • graphic
    with prepositional case ⇔ boin [ponj]

  • graphic
    with prepositional case ⇔ mnaoi [mɾ
    graphic
    ]

Treating these patterns as parallel to better-known instances of morphological case marking provides a straightforward account of these data. I will therefore assume this to be so in what follows.8

2 Linear Adjacency and Prepositional Case

As the name suggests, prepositional case is assigned by prepositions. To understand how this assignment works, let us examine the syntactic relationship between P and the DP to which it assigns prepositional case.

There is evidence that P enters into a ϕ-agreement relation with its DP complement.9 As in other modern Celtic languages, ϕ-agreement is visible on P only when DP is a pronoun, and in such cases, the pronoun is always null. See, among others, Anderson 1982 and Stump 1984 for Breton, Pranka 1983 and McCloskey and Hale 1984 for Irish, and Adger 2000 for Scottish Gaelic. (11) and (12) show these patterns in Gaelic.

(11)

  • agam pro

    at.1SGpro.1SG

    ‘at me’

  • *agam mi

    at.1SG me

  • *aig mi

    at me

    (Robinson 2008)

(12)

  • aig a’+Lghille

    at the boy.PREP

    ‘at the boy’

  • *aige a’ +Lghille

    at.3SG.MASC the boy.PREP

    (Robinson 2008)

ϕ-agreement on the preposition aig ‘at’ is visible only in (11a) through the suffix -am. In this example, an obligatorily null pronominal controls agreement on P, as demonstrated by the ungrammaticality of the overt pronoun in (11b–c).10 In (12b), we see that overt agreement on P with an overt DP is ungrammatical; only (12a), with no ϕ-agreement on P, is possible.

There is independent syntactic reason to posit a silent pronominal in cases like (11a). First, as McCloskey and Hale (1984) observe for Irish, pro can host a relative clause.

(13)

  • Dhaibh pro -san [a dh’earbadh asam ].

  • for.3PLpro.3PL -EMPH.3PLREL trust.COND from.1SG

  • ‘For them who trusted me.’

  • (AMC, dedication)

If we assume that relative clauses must attach to a nominal projection in the syntax, then these data require that a silent DP be present.11

Second, consider the patterns of contextual allomorphy exhibited by a set of morphemes referred to in the Celtic literature as emphatic particles.12 These morphemes show full allomorphy for pronominal ϕ-features, as illustrated in (14).

(14)

graphic

Work in Distributed Morphology proposes that contextual allomorphy of the sort in (14) is subject to locality restrictions (Embick 2010, Bobaljik 2012, 2015, Moskal 2015, Bobaljik and Harley 2017). I adopt the locality conditions proposed by Bobaljik (2012), shown in (15).

(15) Where α triggers allomorphy on β

(15) requires that the nodes implicated in contextual allomorphy not be separated by XP boundaries. As highlighted by Bobaljik and Harley (2017), a head-complement structure meets these conditions, assuming the relevant features are present on the maximal projection of the complement.

With this in mind, consider the syntax of emphatic particles. Like demonstratives, emphatic particles only compose with definite DPs (McCloskey 2004, Adger 2013).

(16)

  • *(na) dealbh -an ud

    the.PL picture -PL yon

    ‘those pictures’

  • iad sin

    they that

    ‘those (guys)’

  • Daibhidh ud

    David yon

    ‘that David’

    (Adger 2013:106–107)

(17)

  • *(am) fear -sa

    the man -EMPH

    ‘the man’

  • i -se

    she -EMPH

    ‘she.EMPH

  • Màiri -sa

    Màiri -EMPH

    ‘Màiri.EMPH

    (Speakers: South Uist)

(16a) and (17a) show that demonstratives and emphatic particles cannot modify indefinite DPs. (16b–c) and (17b–c) show that other kinds of definite DPs such as pronominals and proper names can be modified by demonstratives and emphatic particles. Demonstratives and emphatic particles are also in complementary distribution, as shown in (18).

(18)

  • am boireannach seo

    the woman this

    ‘this woman’

  • am boireannach -sa

    the woman -EMPH

    ‘the woman.EMPH

  • *am boireannach seo-sa

  • *am boireannach-sa seo

    (Speakers: Barra, South Uist)

I analyze the complementary distribution in (18c–d) as indicating that demonstratives and emphatic particles occupy the same syntactic position. McCloskey (2004) and Adger (2013) propose that demonstratives, and by extension emphatic particles, occupy a position within the nominal extended projection (Grimshaw 2000, 2005) labeled DemP. Dem selects a definite DP complement,13 which derives the patterns in (16)–(17). This is shown in (19) for (17b).

(19)

graphic

In (19), DemP takes a DP complement, which is headed by the third singular feminine pronoun i ‘she’. Per the standard technology of projection (more precisely, perfect projection; Grimshaw 2000:117), the features of D are also present on DP. This means that no maximal projection boundaries will separate DemP from the feature bundle [+FEM, −PART, −AUTH], satisfying the locality conditions on contextual allomorphy in (15).

Interestingly, the appropriate emphatic allomorph occurs with prepositions that bear ϕ-agreement.

(20)

  • Bha fear [againn pro-ne ] co-dhiù.

    was one at.1PLpro.1PL -EMPH.1PL anyway

    ‘Anyways, we had one.’

    (MM, pg. 44)

  • Ghabh e [orm pro-sa ] cheana.

    got he on.1SGpro.1SG -EMPH.1SG already

    ‘He already got one over on me.’

    (TB, pg. 146)

In (20a), the preposition aig ‘at’ is inflected for 1PL features. This yields the form againn ‘at.1PL’. Crucially, the emphatic particle exhibits allomorphy for 1PL features as well, producing the form -ne (see (14)). Similar data are seen in (20b).

To meet the locality conditions in (15), (20a) must have a pronominal whose local relation with Dem is the same as in (19). Consider the structure in (21).

(21)

graphic

The only structural difference between (21) and (19) is that P enters into a ϕ-agreement relation with D in (21) while no P is present in (19).

Therefore, P seems to be a ϕ-probe in Gaelic, as proposed by McCloskey and Hale (1984), Andrews (1990), Legate (1999), and Brennan (2009) for Irish. This Agree relation between P and its complement could well be significant in light of work that proposes that Agree relations can license case morphology (Chomsky 2000, 2001, Frampton and Gutmann 2000, Legate 2008, Baker and Vinokurova 2010, Baker 2015). As an initial hypothesis, let us suppose that Agree is responsible for the distribution of prepositional case. One way of implementing this proposal is with the case realization rule in (22), inspired by Baker (2015).

(22)

  • Prepositional case assignment (First pass)

  • DP is assigned prepositional case when DP stands in a ϕ-agreement relation with P.

Interestingly, the distribution of prepositional case is more intricate than (22) suggests: surface linear adjacency between P and DP is also necessary for prepositional case. This effect was first observed by Adger (2000), who noticed that certain dialects allow eadhon ‘even’ to intervene between P and DP. In (23)–(24), we see that if eadhon ‘even’ linearly intervenes between P and DP, prepositional case cannot occur.

(23)

  • ri eadhon am ban-rìgh cridheil

    with even the queen kind

    ‘with even the kind queen’

  • *ri eadhon a’ +Lbhan-rìgh +Lchridheil

    with even the queen.PREP kind.PREP

    (Adapted from Adger 2000)

(24)

  • eadhon ris a’ +Lbhan-rìgh +Lchridheil

    even with the queen.PREP kind.PREP

    ‘even with the kind queen’

  • *eadhon ris am ban-rìgh cridheil

    even with the queen kind

    (Speakers: Barra, South Uist)

Eadhon ‘even’ seems to be an adjunct, because it may modify a variety of categories, including DP, AP, CP, VP, and PP.14

(25)

  • Bha [DPeadhan [DP seòrsa de dh’fhiamh]] aig saighdearan . . .

    was even sort of awe at soldiers

    ‘The soldiers had even a sort of awe . . .’

    (FSS, pg. 46)

  • Bhithinn [APeadhon [AP toilichte]] nuair a dhùisginn.

    be.COND.1SG even happy when REL awake.COND.1SG

    ‘I would be even happy when I woke up.’

    (FSS, pg. 42)

  • Bha cuid [CPeadhon [CP a fhuair am bàs ]].

    was some even REL got their death

    ‘There were some who even died.’

    (FSS, pg. 42)

  • [CPEadhon [CP nuair a bhiodh tu nad chadal]] . . .

    even when REL be.COND you in.your sleep

    ‘Even when you’d be asleep . . .’

    (AMC, pg. 11)

  • Dh’fheuch e ri [VPeadhon [VP am ban-rìgh a+L mharbhadh]].

    tried he with even the queen v kill.VN

    ‘He tried even to kill the queen.’

    (Adger 2000)

  • [PPEadhon [PP ann an 1944]] bha iad leis a’ +Lchianalas.

    even in 1944 was they with the sorrow.PREP

    ‘Even in 1944 they were still with sorrow.’

    (H, pg. v)

This freedom of attachment indicates that eadhon ‘even’ is not selected in the extended projection of any particular category. This is consistent with its characterization as an adjunct. Likewise, eadhon ‘even’ does not pattern like a DP; eadhon is not licensed as the complement of P by itself, in out-of-the-blue contexts or otherwise.

(26)

  • *Bha mi a’ bruidhinn ri eadhon.

  • was I PROG speak.VN with even

  • Intended: ‘I was speaking with even.’

  • (Speakers: Everywhere)

Therefore, the structure of the entire PP in (23a) is as shown in (27).

(27)

graphic

If P is a ϕ-probe as discussed above, it will search its c-command domain for valued ϕ- features. The most local goal that P encounters is DP. As an adjunct, eadhon ‘even’ is in no position to disrupt this relation.15 Furthermore, since eadhon ‘even’ is not a DP, there is no reason to posit that it has ϕ-features. Therefore, it is not predicted to act as a defective goal in the sense of Chomsky 2000 and Roberts 2010.

All of this means that P should successfully enter into an Agree relation with DP regardless of eadhon ‘even’. Therefore, the environment for the rule of prepositional case assignment in (22) is met even in (27). Why, then, is prepositional case morphology not assigned in (27)? I will consider two hypotheses. The first, in (28a), posits that for DP to be assigned prepositional case, it must both stand in a ϕ-agreement relation with P and be linearly adjacent to P. The second, in (28b), posits that only linear adjacency is required.

(28) A DP is assigned prepositional case when:

  • Hypothesis A

    DP stands in a ϕ-agreement relation with P and P⌢DP.

  • Hypothesis B

    P⌢DP.

    P linearly precedes DP.

To distinguish these hypotheses, we need an environment where P and DP are linearly adjacent, but do not stand in a ϕ-agreement relation. If prepositional case occurs without an Agree relation with P, then we have support for Hypothesis B as opposed to Hypothesis A.

3 Prepositional Case in Nonfinite Clauses

Nonfinite clauses provide the right environment to test the two hypotheses in (28). Consider the transitive nonfinite clause in (29).

(29)

  • Tha an gille ag iarraidh [PRO an cù a+L fhaicinn].

  • Is the boy PROG want.VNPRO the dog v see.VN

  • ‘The boy wants to see the dog.’

  • (Speakers: South Uist, Barra, Eriskay)

In (29), the object an cù ‘the dog’ precedes a nonfinite verbal form referred to as the verbal noun,16 and a morpheme (a+L) occurs between the object and the verbal noun. Following Carnie (1995), Adger (1996, 1997, 2000), and Robinson (2008), I identify a+L as v.17 v enters into a ϕ-agreement relation with the object, licensing movement to Spec,vP. Throughout this discussion, I abstract away from any higher functional structure, such as nonfinite TP. The derivation adopted here is shown in (30).

(30)

graphic

In (30), v enters into a ϕ-agreement relation with the direct object. This Agree relation licenses movement of the object to Spec,vP.

Evidence for ϕ-agreement again comes from the behavior of pronominal objects. As with the objects of prepositions, overt pronominal objects are forbidden in nonfinite clauses.

(31)

  • *Tha iad ag iarraidh [mi a+L ghlacadh].

    Is they PROG want.VN me v catch.VN

    Intended: ‘They want to catch me.’

  • *Tha an tidsear air [sinn a+L chronachadh ].

    is the teacher on us v reprimand.VN

    Intended: ‘The teacher has reprimanded us.’

    (Speakers: Everywhere)

Instead, v inflects for the ϕ-features of the pronominal object, as shown in (32). The full inflection paradigm for v is provided in (33).

(32)

  • Tha iad ag iarraidh mo+L ghlacadh pro.

    is they PROG want.VN v.1SG catch.VNpro.1SG

    ‘They want to catch me.’

  • Tha an tidsear air ar cronachadh pro.

    is the teacher on v.1PL reprimand.VNpro.1PL

    ‘The teacher has reprimanded us.’

    (Speakers: Everywhere)

(33)

graphic

In (32), we see that one of the conjugated forms in (33) replaces a+L in v. As with the objects of prepositions, only a silent pro is permitted with ϕ-agreement.

Evidence that pro occurs in object position18 in (32) comes from the same patterns of contextual allomorphy exhibited by emphatic particles illustrated in section 2.

(34)

graphic

In (34), we see that the correct allomorph of the emphatic particle occurs after the verbal noun in complement position. To satisfy the locality conditions on contextual allomorphy from (15), the emphatic particle must not be separated from the relevant ϕ-feature bundles by any XP boundaries (Bobaljik 2012). If the features of the pronoun were present only on v, then we would not have any account of the pattern in (34). But if a silent pronoun occupies direct object position, then we do have a potential analysis, shown in (35).

(35)

graphic

In (35), the ϕ-features of pro are copied onto v as a result of Agree, and the presence of pro allows Dem to be in the right local relation for contextual allomorphy.

With this, consider (36), in which a nonfinite clause is immediately preceded by P.19 In particular, observe the prepositional case on the apparent nonfinite object.20

(36)

  • Tha an cù air a’ +Lghille+Lbhàn a+L imlich.

    is the dog on the boy.PREP blond.PREP v lick.VN

    ‘The dog has licked the blond boy.’

  • Tha mi a’ feuchainn ris a’ +Lchaillich+J +Lbhig+J a+L fhaicinn.

    is I PROG try.VN with the old.woman.PREP little.PREP v see.VN

    ‘I am trying to see the little old woman.’

    (Speakers: Stafainn, Barra)

The sentences in (36) have the same verbal noun morphology and the same apparent movement of the object of the verbal noun as (29), but in (36) the direct object is assigned prepositional case. Clearly, prepositional case here correlates with the presence of a preposition, air ‘on’ in (36a) and ris ‘with’ in (36b).

There are two conceivable analyses of structures like (36a–b) in which P is immediately followed by an apparent nonfinite clause. First, P could take DP as a complement to the exclusion of the rest of the nonfinite clause, as in (37). Second, P could take the entire nonfinite clause as a complement, as in (38).

(37) [PP P DP] [vP v V]

(38) [PP P [VoiceP DP v V]]

These two possibilities make different claims about whether P and DP form a constituent that excludes vP. Adger (2010b) observes that constituents in Scottish Gaelic may be clefted. For instance, the grammaticality of (39b) indicates that P and DP in (39a), unsurprisingly, form a constituent.

(39)

  • Bha mi a’ bruidhinn [ris a’ +Lchaillich+J ].

    was I PROG speak.VN with the old.woman.PREP

    ‘I was speaking to the old woman.’

  • ’S ann [ris a’ +Lchaillich+J ] a bha mi a’ bruidhinn ____.

    CLEFT with the old.woman.PREPREL was I PROG speak.VN

    ‘It’s to the old woman that I was speaking.’

    (Speakers: Stafainn, Barra)

Therefore, if P and DP in examples like (36a–b) form a constituent, they should be able to be clefted too. Interestingly, P and DP in examples like (36) cannot be clefted. This is shown in (40b).

(40)

  • Bha mi a’ feuchainn ris a’ +Lchloinn+J a+L chuideachadh.

    was I PROG try.VN with the children.PREP v help.VN

    ‘I was trying to help the children.’

  • *’S ann [ris a’ +Lchloinn ] a bha mi a’ feuchainn ____ (a+L)

    CLEFT with the children.PREPREL was I PROG try.VNV

    c(h)uideachadh.

    help.VN

    Intended: ‘It’s the children that I was trying to help.’

    (Speakers: Stafainn, Barra, South Uist, Harris)

Sentences like (40b) are highly ungrammatical, and indeed unparsable to speakers. I take this ungrammaticality to indicate that P and DP do not form a constituent in (40a), in contrast with (39a). Therefore, (37) is not a possible structure for this sort of sentence. This leaves the structure in (38), developed in (41).

(41)

graphic

Crucially, observe that (41) and (30) employ the same structure within the nonfinite clause, the only difference being that in (41) the nonfinite clause is selected by P. This raises a question: if the direct objects of the nonfinite clauses in (41) and (30) exhibit the same syntactic behavior, why does the object in (41) receive prepositional case, while the object in (30) cannot? (42) provides a minimal pair.

(42)

graphic

In (42a–b), the nonfinite clause is selected by the higher V, and prepositional case cannot be assigned to the embedded object. (42c–d) show that when the nonfinite clause is selected by P, prepositional case is assigned to the embedded object.

Recall the hypotheses in (28) for prepositional case licensing, repeated in (43).

(43) A DP is assigned prepositional case when:

  • Hypothesis A

    DP stands in a ϕ-agreement relation with P and P⌢DP.

  • Hypothesis B

    P⌢DP.

If no Agree relation is found in contexts like (42c), then we must pursue Hypothesis B. If evidence for Agree is found in these contexts, then Hypothesis A is viable.

In section 2, I proposed that pronominal ϕ-features on prepositions are licensed through an Agree relation with pro. Therefore, if P is in an Agree relation with the direct object in a nonfinite complement clause like (42c), P should be able to bear pronominal ϕ-features. As it turns out, this expectation is not borne out.

In (44a), we see that a pronominal object in a nonfinite clause selected by a P behaves identically to pronominal objects in other nonfinite clauses, examined in (31)–(35). v enters into a ϕ-agreement relation with pro. This Agree relation licenses pronominal ϕ-features on v, yielding the conjugated form mo.

(44)

  • Bha an t- arm a’ feuchainn ri mo+L chuideachadh pro.

    was the army v try.VN with V.1SG help.VNpro.1SG

    ‘The army was trying to help me.’

  • *Bha an t- arm a’ feuchainn rium c(h)uideachadh.

    was the army v try.VN with.1SG help.VN

    Intended: ‘The army was trying to help me.’

  • *Bha an t- arm a’ feuchainn rium mo+L chuideachadh pro.

    was the army v try.VN with.1SG v.1SG help.VNpro.1SG

    Intended: ‘The army was trying to help me.’

    (Speakers: Everywhere)

Now let us examine the form of the selecting preposition ri ‘with’. In the grammatical (44a), ri is not inflected for the ϕ-features of the pronominal object of the nonfinite clause it selects. If it were, that would yield rium ‘with.1SG’. Furthermore, (44b–c) demonstrate that such a conjugated preposition is ungrammatical, either with ϕ-agreement on v or without it.

These data might leave ground for skepticism. As we saw in (35), the syntax of pro in nonfinite clauses is apparently different from that of overt DPs because pro remains in situ while overt DPs move to Spec,vP. This diverging syntax could be the source of the ungrammaticality of (44b–c), rather than any failure of Agree. To see the confound, consider templates of the relevant nonfinite clauses.

(45) Skeletal templates of nonfinite clauses

  • With overt DP object

    P [VoiceP DP[+PREP]v[−ϕ] V]

  • b. With pro object

    P [VoiceP v[] V pro]

(45) demonstrates the dissimilarity in the two structures under consideration. In (45a), an overt DP object in the nonfinite clause moves to Spec,vP, and v does not show overt ϕ-agreement; see (30). This contrasts with a pro object that does not move to Spec,vP but does trigger overt ϕ-agreement on v; see (31)–(35).

Ideally, we should consider what happens when a pronominal object in a nonfinite clause moves to Spec,vP without triggering ϕ-agreement on v. A pronominal object is needed to provide the best chance of observing ϕ-agreement, while movement to Spec,vP and the lack of agreement on v would control for the confound in (44).

Gaelic provides the opportunity to investigate such a context. In addition to the pattern schematized in (45b), there is a second pattern that occurs only when a pronominal object inside a nonfinite clause is modified either by the reflexive marker fhèin or by an emphatic particle. This second pattern is illustrated in (46). The two patterns in (34) and (46) are largely in free variation, although (46) is strongly preferred in innovative varieties, while (34) has a conservative, perhaps literary flavor.

(46)

  • Am bheil thu airson [mi -se a+L fhaicinn]?

    Q is.DEP you for me -EMPH.1SG v see.VN

    ‘Are you for seeing me ~ Do you want to see me?’

    (Gillies 2009:297)

  • *Am bheil thu airson [mi -se mo+L fhaicinn]?

    Q is.DEP you for me -EMPH v.1SG see.VN

    Intended: ‘Are you for seeing me ~ Do you want to see me?’

In (46), the pronominal object of the nonfinite clause moves to Spec,vP, as evidenced by its placement to the left of v. Additionally, no overt ϕ-agreement occurs on v.

Observe that (46a) demonstrates the same properties as the target structure in (45a): the object moves to Spec,vP, and v does not exhibit ϕ-agreement. This parallelism allows us to formulate a more targeted hypothesis. If Agree between P and DP in examples like (46b), schematized in (45a), were to license prepositional case on DP, then we would expect to see ϕ-agreement on P when P selects a nonfinite clause of the sort in (46). Again, this expectation is not borne out. This is shown in (47), where we see that a P selecting a nonfinite clause of the sort in (46) may not show ϕ-agreement.

(47)

  • Tha Calum air [iad -san a+L phògadh].

    is Calum on them -EMPH.3PLV kiss.VN

    ‘Calum has kissed them.’

    (Adger 2010b:312)

  • *Tha Calum orra (iad) -san (a+L) p(h)ògadh.

    is Calum on.3PL (them) -EMPH.3PLV kiss.VN

    Intended: ‘Calum has kissed them.’

    (Speakers: Everywhere)

In (47a), when P selects a nonfinite clause of the sort in (46), P is not inflected for the ϕ-features of the pronominal object. If ϕ-features on P are licensed on P through Agree, then the ungrammaticality of (47b) indicates that P and the object of a nonfinite clause do not stand in an Agree relation.

These patterns reveal that no ϕ-agreement relation is established in the relevant contexts. Rather, the available evidence indicates the opposite. This is a reasonable result because, as Adger (2000:86) points out, “there seems to be no grammatical relationship between the preposition and the DP: the DP [is licensed in the nonfinite clause] and is complement of the embedded predicate.” In other words, P is not necessary to license DP in this configuration, as nonfinite clauses may license an object without a selecting preposition, as in (42a).

These data demonstrate that prepositional case morphology occurs in the absence of an Agree relation with P. Therefore, Hypothesis A in (43a) must be rejected. This leaves us with Hypothesis B, yielding our final morphological case assignment rule for prepositional case in (48).

(48)

  • Prepositional case assignment

  • DP is assigned prepositional case morphology when P⌢DP.

In other words, we have a mismatch between nominal licensing and case morphology. In nonfinite clauses, objects are licensed regardless of whether the nonfinite clause is selected by P or some other head. This suggests that a selecting preposition is not necessary to license an object DP in a nonfinite clause, but its presence immediately to the left of DP does license prepositional case. And indeed, if linear adjacency between P and DP is disrupted, prepositional case may not occur.

(49)

graphic

In (49), we see a nonfinite clause selected by P in which the adjunct eadhon ‘even’ linearly intervenes between P and DP. Unlike in the examples considered so far, prepositional case may not be assigned here, apparently because linear adjacency between P and DP is disrupted. Therefore, it seems that linear adjacency between P and DP, rather than Agree, provides the licensing context for prepositional case in Gaelic.

With this understanding of prepositional case assignment in Gaelic, let us consider its broader implications for the theory of morphological case.

4 Case Morphology and Linear Adjacency

Baker (2015:77–78) proposes that “[l]anguages . . . vary in what case assignment rules they include, as well as in their case realization rules.” If we adopt Baker’s proposal, we are left with a question: what are the possible forms of case realization rules? Baker (2015) identifies two kinds of case realization rules, although, in the spirit of his work, I take these only as starting points.

The first are Agree case realization rules (e.g., Chomsky 2000, Frampton and Gutmann 2000). (50) shows a basic Agree case realization rule.

(50)

  • Agree case realization rule

  • DP → [DP: M-case.X] / DP enters into an Agree relation with a node of category Y.

The prepositional case assignment rule motivated in section 3 is not reducible to an Agree case realization rule. This is because no Agree relation is required for prepositional case assignment.

Baker (2015) identifies a second kind of rule, building on a tradition that includes Marantz 1991, Bittner and Hale 1996, and other works. These are termed dependent case realization rules. Consider the dependent ergative case assignment rule in (51).

(51) If there are two distinct argumental NPs in the same phase such that NP1 c-commands NP2, then value the case feature of NP1 as ergative unless NP2 has already been marked for case. (Baker 2014:343)

Dependent case realization rules require that two conditions be met. First, there must be two distinct nominals in the same phase. Second, one nominal must c-command the other. A schematized dependent case realization rule is given in (52).

(52)

  • Dependent case realization rule

  • DP1 → [DP1 : M-case.X] / DP1 c-commands DP2 in the same phase and DP1 and DP2 are both arguments.

The rule in (48) cannot be reduced to a dependent case realization rule because these rules are fundamentally about two DPs within a certain domain competing for case morphology in order to be distinguished (Marantz 1991). The rule in (48) cannot be about case competition because this rule only references one DP.

Therefore, it is fitting to consider a third type of case realization rule for prepositional case in Scottish Gaelic. As linear adjacency is the sole necessary licensing context, I term this third kind of rule an adjacency case realization rule.

(53)

  • Adjacency case realization rule

  • DP → [DP: M-case.X] / DP is linearly adjacent to Y.

Adjacency case realization rules require only direct concatenation between a node Y and a DP. In this way, (53) captures the licensing contexts for prepositional case more successfully than a case realization rule framed in terms of Agree or dependent case.

This proposal leaves us with three relations that are hypothesized to license case morphology: Agree, the dependent case relation (which holds between two DPs in the same phase when one DP c-commands the other), and linear adjacency. This diversity of licensing contexts may seem disconcerting, and one wonders what constraints, if any, might limit the potential licensing environments for case morphology. To explore this question, let us examine the broader theory in which this proposal is couched: Distributed Morphology (DM; Halle and Marantz 1993, 1994). Much work in DM proposes that the syntax-morphology interface consists of a series of derivationally ordered functions (see Arregi and Nevins 2012 and Bobaljik 2012 for implementations). The output of these functions adjusts the (morpho)syntactic structure so that it is interpretable by the phonology proper. In this way, “morphology” is distributed over several operations, some syntactic and some postsyntactic. This model of the postsyntax is shown in (54).

(54)

graphic

We focus on the left branch of (54). After the syntax, the derivation undergoes the initial processing referred to as the syntax-morphology interface (presumably in phases; Chomsky 2000, 2001, 2008). This interface is hypothesized to consist of several ordered operations, including Linearization (Marantz 1984, 1988, Bonet 1991, Arregi and Nevins 2012) and nominal concord (Norris 2014). Once the morphosyntactic structure has been adjusted, phonological content is inserted. Let us adopt the position that Vocabulary Insertion targets terminal nodes (Halle and Marantz 1993).21 The output of Vocabulary Insertion is the input to the narrow phonology, and ultimately to PF.

The primary claim of this article is that Morphological Case Assignment must have access to the output of Linearization.22 Assuming that Linearization is an operation in the postsyntax, it follows that morphological case assignment rules are also postsyntactic operations that can be ordered.23 In this way, at the point in the derivation when case morphology is assigned, the system has access to the output of the syntax, modulated by some degree of postsyntactic processing.

How much postsyntactic processing must occur before Morphological Case Assignment is ultimately an empirical question, as it requires close examination of the licensing contexts in various morphological case systems crosslinguistically and an understanding of the feeding relations at play. Gaelic provides evidence for the several orderings of postsyntactic operations, although crosslinguistic work will likely reveal more. For instance, that adjectival modifiers show concord for prepositional case in Gaelic, as we have seen throughout, suggests that concord occurs postsyntactically as well, ordered after both Linearization and Morphological Case Assignment (see Norris 2014).24 The proposed order of operations for Linearization, Morphological Case Assignment, concord, and Vocabulary Insertion is shown in (55).

(55)

graphic

(55) provides a nonexhaustive list of operations in the syntax-morphology interface, and captures the feeding relationship between Linearization and Morphological Case Assignment that is needed to account for prepositional case assignment in Gaelic. In the context of such a view of the syntaxmorphology interface, we can rephrase the question raised at the beginning of this section: what information is available when morphological case is determined?

(55) provides a possible answer: only information available in the output of operations that precede Morphological Case Assignment should play a role in determining case morphology. Assuming (55), certain types of case realization rules are predicted to be impossible.

(56) Impossible morphological case assignment rules

  • Concord case realization rule

    DP → [DP: M-case.X] / A modifier of DP is marked with a feature F through concord.

    M-case X is licensed iff DP is modified by a concord-bearing adjective.

  • Phonological case assignment rule

    DP → [DP: M-case.X] / A segment of DP in position X is specified for phonological feature F.

    M-case X is licensed iff DP contains a fricative segment.

The rules in (56) should be impossible. Ultimately, whether such rules exist is an empirical question, but I know of no proposed rules of either type.

With this larger framework in place, let us consider other ways in which the effects of linear adjacency could be tested in Gaelic prepositional case assignment. First, consider another case of linear adjacency effects in morphology: Closest Conjunct Agreement (CCA). Consider the Standard Arabic example in (57).

(57)

In (57), the subject is a coordinate structure consisting of a feminine name and a masculine name. The verb qaraʔat ‘read.3.FEM.SG’ agrees only with the closest conjunct, the feminine name ʕaliyaa, rather than with the entire coordinate structure, which presumably bears plural features.

Given that CCA is widely attested, one might wonder if Gaelic prepositional case also shows closest conjunct effects. It is not clear what the expectations should be, as Weisser (2018) has claimed that “closest conjunct case” does not exist. On this basis, he contends that morphological case assignment must be a syntactic process because “case assignment is not subject to linearity effects in the same way that agreement is.” Weisser’s proposal, which is incompatible with the present account, leads to an expectation: if prepositional case in Gaelic is assigned on the basis of linear adjacency, then prepositional case should exhibit closest conjunct case effects.

Gaelic provides the opportunity to test this hypothesis, as the language allows coordination of DP below P. Strikingly, closest conjunct case is obligatory.

(58)

graphic
graphic

(59)

graphic

First consider (58). Here we see two DPs, an gille beag ‘the little boy’ and an cù fliuch ‘the wet dog’, that are coordinated below the P ris ‘with’. In (58a), we see that only the leftmost conjunct may be inflected with prepositional case. Marking the rightmost conjunct with prepositional case, or not marking either conjunct with prepositional case, is ungrammatical (58b–c).

This pattern is obligatory and robust in Scottish Gaelic, the same pattern being found with speakers from Tiree, Barra, North Uist, Harris, West Lewis (Uig), and Northeast Lewis (Back). In fact, speakers are aware of this aspect of their language. The speaker who provided (59a) finished by saying, “A’ chlann, uh-huh, and that’s the way it would be for me . . . if you don’t have the ris immediately next to it.”25

Therefore, Gaelic provides a counterexample to Weisser’s (2018) hypothesis that closest conjunct case does not exist. At the same time, these data have a natural account in the theory proposed here. In (58)–(59), both DPs are syntactically licensed as objects of P, but only the leftmost conjunct is linearly adjacent to P. In this way, only this leftmost conjunct will meet the licensing requirements of the prepositional case morphology assignment rule in (48). A system in which case morphology is directly connected to DP-licensing will be unable to account for this pattern without further stipulation.

5 Conclusion and Looking Forward

In this article, I have argued that case morphology must be calculated in the postsyntax. In sections 2 and 3, I showed that linear adjacency is necessary to license prepositional case morphology. Next, I accepted the well-established premise that linear order is calculated postsyntactically (Chomsky 1981, 1993, 2000, 2001, 2008, Marantz 1984, 1988, Bonet 1991, Kayne 1994, Bobaljik 1995, Harley and Noyer 1999, Embick and Noyer 2001, 2007, Fox and Pesetsky 2005, Embick 2010, Berwick and Chomsky 2011, 2016, Arregi and Nevins 2012).26 If the syntax precedes Linearization, and Linearization feeds the calculation of case morphology, it follows that the calculation of morphological case must be determined postsyntactically.

To account for this, I proposed that after Linearization in the postsyntax, DPs are subjected to case realization rules. Because Morphological Case Assignment occurs late in the derivation, we can model sensitivity to various types of information determined upstream. In this way, we can understand why linear relations, adjacency in particular, determine the distribution of prepositional case.

A final comment is in order about the scope of this proposal. By placing Morphological Case Assignment in the postsyntax, I envision a system that expands the possible licensing environments for morphological case. This proposal is intended to be compatible with other systems of morphological case assignment, such as those presented by Legate (2008) and Baker (2015). Indeed, I take the expanded range of possible licensing environments argued for here as a welcome confirmation of these authors’ general approach of calculating case morphology in the postsyntax.

Notes

1 Plural nouns are not discussed throughout because the case and gender inventories in the plural in all contemporary varieties are greatly diminished.

2 Throughout, I will only present definite nominal phrases. This is because indefinites are not always assigned prepositional case. This contrast is illustrated between (i) with the masculine noun baile ‘town’ and (ii) with the feminine nouns clach ‘stone’ and bean ‘woman’.

    • Thàinig mi chun an t-saoghail [ann am baile beag àlainn ].

      came I to the world.GEN in town little beautiful

      ‘I came into this world in a small, beautiful town.’

    • *[ann am +Lbhaile+Lbheag àlainn ]

      in town.PREP little.PREP beautiful

    • anns a h-uile taigh [anns a’ +Lbhaile+Lbheag againn fhìn ].

      in each house in.DEF the town little at.us REFL.1

      ‘in each house in our own little town’

      (AMC, pg. 9)

    • Chuir e a làmh air cloich+J+Lmhòir+J.

      put.PAST he his hand on stone.PREP big.PREP

      ‘He put his hand on a big stone.’

      (FSS, pg. 66)

    • . . . ri mnaoi+Lshaoghalta a rachte ga+L phòsadh . . .

      with woman.PREP worldly.FEMREL go.AUTO v.3MASC marry.VN

      ‘ . . . to a worldly woman that he was sent to marry . . . ’

      (SD, pg. 4)

    In (iia), both the head noun clach [khlɣax] ‘stone’ and the adjective mòr [mɔːɾ] undergo final palatalization, becoming cloich [khlɣɣjç] and [mɔːɾj], respectively. As noted in the text, this final palatalization is a distinctive characteristic of feminine DPs in prepositional case. Mnaoi, the suppletive stem of bean ‘wife, woman’ illustrated in (iib), occurs only in prepositional case contexts (see the main text).

    While this gender difference is interesting, definiteness is not precise enough to explain why prepositional case is blocked in (ib). Consider the behavior of proper names: masculine proper names do not receive prepositional case.

    • ainmean a sgrìobh Fear Chanaigh sìos bho+L Dhòmhnall Bàn

      names REL wrote Fear Chanaigh down from Dòmhnall blond

      ‘names Fear Chanaigh wrote down from blond Dòmhnall’

      (MM, pg. 166)

    • *. . . bho+L [Dhòmhnall +LBhàn ]

      from Dòmhnall blond.PREP

    (iii) shows a masculine name, Dòmhnall Bàn ‘blond Dòmhnall’, in a PP. If this DP received prepositional case, then we would expect the adjective bàn ‘blond’ to be lenited under case concord with Dòmhnall.

    Therefore, semantic definiteness cannot be the proper characterization. Rather, it seems that masculine DPs must be marked with the definite article in order to receive prepositional case. Exactly how both the “definite article effect” and the gender effect can be captured must be left to future research.

3 The loss of the nasal on the definite article between an cat beag [ǝn khaht pφk] ‘the little cat (unmarked)’ and a’ chat bheag [ǝ xaht vφk] is a routine part of Gaelic phonology and not relevant for present purposes.

4 In the autosegmental approach adopted here, I assume that lenited forms are phonologically derived from unlenited forms via admittedly opaque phonological processes (see Wolf 2007). Alternatively, as a reviewer points out, these data could be captured by claiming that lenited forms are allomorphs triggered by the abstract +L diacritic through the general approach taken by Green (2006). Either approach could be adopted without affecting the conclusions of this article.

5 Of course, the word ultach ‘bundle’ cannot undergo Initial Consonant Mutation because it begins with a vowel. But this word can still be shown to bear prepositional case. The definite article for vowel-initial masculine nouns in the unmarked case is an t- [ǝn t], with the t- being lost in the prepositional case.

6Moskal (2015) proposes that nominal root suppletion for morphological case is impossible. She discusses the bean ~ mnaoi alternation, suggesting a phonological source. The phonological proposal seems to be dubious synchronically, and cannot account for the bo ~ boin alternation. Therefore, I consider these data a true counterexample to her hypothesis.

7Bhoinn is an orthographic variant of boin ~ bhoin.

8 While this sort of case morphology may seem foreign, other systems that mark morphological case through morphophonological processes are attested. These include Nias Selatan (Austronesian; Brown 2001), where ICM marks absolutive case, and Nuer (Nilo-Saharan; Trommer 2011), where consonant mutations can mark genitive case. See Merrill 2014.

9 The analysis of these synthetic endings as ϕ-agreement is not uncontroversial. See Hale 1987, 1990, Ackema and Neeleman 2003, Brennan 2009, and Diertani 2011 for a variety of other approaches. My goal is not to argue that ϕ-agreement is the best analysis of prepositional synthetic endings. Rather, my purpose is to investigate what the consequences of such an account would be with respect to theories of morphological case assignment.

10 The obligatory silence of the pronoun in these contexts is beyond the scope of this article, though see Pranka 1983, McCloskey and Hale 1984, Hale 1987, Doron 1988, Ackema and Neeleman 2003, and Jouitteau and Rezac 2006.

11 We can also be sure that the relative clause in (13) is not a free relative because free relatives are introduced by the separate complementizer na.

  • Leugh mi [na sgrìobh thu ].

    read I FREE.REL wrote you

    ‘I read whatever you wrote.’

    (Adger 2013:11)

As free relative na is not present in (13), there must be some nominal projection to host those relative clauses. pro provides such a host.

12 An identical pattern is found with the reflexive marker fhèin, which has the allomorph fhìn for first person ϕ-features.

13 I represent DemP as being head-final, though I assume that a structure with movement would be equivalent. See McCloskey 2004.

14Eadhan in (25a) is an orthographic variant of eadhon; both are pronounced [φɣǝn].

15 Of course, rightward adjuncts are in no position to disrupt Agree in this language, as demonstrated by the cases of prepositional-marked nominals with rightward AP adjuncts throughout. See in particular (2)–(3), (6), and (9). Therefore, we must conclude that the relationship that eadhon ‘even’ disrupts is linear, not structural. Many thanks to Jorge Hankamer for reminding me of this important piece of the puzzle.

16 See Carnie 2011 for a review of verbal nouns in Irish. The Gaelic data, as far as I can ascertain, are the same.

17 The morpheme a+L, glossed here as v, is elided in all dialects before vowel-initial verbal nouns, such as fhaicinn [aɪçkiɲj] ‘see’, in the traditional orthography and in natural speech. It is presented here for ease of exposition.

Additionally, the core of these data—namely, the OV order in nonfinite clauses—holds across almost all dialects. South Uist and Eriskay are specified here because of certain lexical choices in these data, such as gille ‘boy’. These lexical choices are irrelevant for present purposes.

18 Apparently, this silent pro does not undergo A-movement to Spec,vP either. Therefore, we see complementary distribution among ϕ-agreement, movement, and an overt pronominal. While these threads are tantalizing, unraveling them would take us too far afield for present purposes.

19 As a reviewer points out, the assumption that these morphemes are prepositions is not without controversy, particularly with regard to air in (36a). Given the aspectual interpretation of air in these contexts, it is widely assumed that air here is a homophonous aspect marker. See Reed 2011 and the citations therein. The primary motivation for claiming that this air is of a different category, apart from the semantics, seems to come from Lamb’s (2001) claim that it does not assign prepositional case. This conclusion is clearly not compatible with the data I gathered. Likewise, it is unclear whether this same sort of aspectual analysis would apply to ris ‘with’ in (36b). Furthermore, if this air is not of category P, the core analysis of prepositional case assignment does not seem to be affected. Therefore, for reasons of expository clarity, I will assume that these morphemes are prepositions.

20 Examples of this sort are difficult to find in corpora because prescriptive grammars and style guides proscribe prepositional case in this configuration (Gillies 2009). Of course, this prohibition proves its existence. At the same time, there seems to be some dialect variation. The data in (36) were replicated with speakers from Tiree, South Uist, Harris, and West Lewis (Uig). No prepositional case marking at all was found on Eriskay. Prepositional case marking was found with speakers from North Uist, but it did not occur in this configuration.

  • graphic

21 But see Caha 2009, Radkevich 2010, Svenonius 2012, Embick 2015, and Merchant 2015 for alternatives.

22 See Heath 2007 for a similar conclusion.

23 This is not a trivial claim. Several authors, among them Bobaljik (1995, 2008), Haeberli (2001), Preminger (2014), and Deal (2017), have suggested that Agree relations in some languages are only possible with DPs that bear a particular morphological case. If Agree relations are determined in the syntax, then morphological case must be in the syntax as well. Here, I follow the spirit of works by Polinsky and Potsdam (2001), McFadden (2004), Bhatt (2005), Baker (2012, 2015), and Polinsky (2016) and interpret these effects as syntactic in origin, rather than morphological.

I also assume that morphological case assignment rules operate at once, for lack of evidence to the contrary. We could likewise envision a system in which what I identify as a single operation actually consists of a family of operations that are each ordered separately. Demonstrating this would require bleeding relations in addition to the feeding relations which I argue that Gaelic shows. I am aware of no arguments of this shape at this time, although they may come from the interaction of morphological case and ϕ-agreement in Icelandic. See especially Sigurðsson 1993 and Bobaljik 2008.

24 Alternatively, an Agree-based approach such as Baker’s (2008) could be compatible with this analysis if concord is separated into two processes: establishment of Agree relations in the syntax and postsyntactic concord realization of the sort shown in (55). For reasons of theoretical parsimony, I prefer the postsyntactic analysis of concord in this system, as it requires only one operation of postsyntactic concord, rather than two.

25 The noun clann ‘children’ is feminine and morphosyntactically singular, idiosyncratically lacking a morphological plural. Despite this, clann is obligatorily interpreted as plural in all cases.

26 While this idea is firmly entrenched, it is not without challenge. See Bruening and Al Khalaf 2015.

Acknowledgments

My thanks first and foremost to the Gaelic speakers who made this article possible, particularly Margaret Anne Beggs, Màiri Liz Nic Kennedy, Coinneach Mac an Aoidh, Calum Mac Nèill, Niall Mac ’IllEathain, Sarah Fergueson, Sarah MacEachen, Veronica Beaton, Rev. Iain Urchardan, Margaret Fraser, Angela McKay, Eilidh Sgaimeall, Catrìona Nic Iain, Eilidh Nic Phaidin, and Murdina Stiùbhairt. Many thanks as well to my friends Dòmhnall Beggs, Màiri Nic ’IllEathain, Lauren Falconer, and Leah Jaques for facilitating much of this fieldwork. This work grew out of my qualifying exam at the University of California, Santa Cruz. My sincere thanks to Sandy Chung for chairing the qualifying exam, as well as to the rest of the committee: Jim McCloskey, Jorge Hankamer, and Ruth Kramer. My thanks in particular to Jim and Sandy for their help in preparing this article. My thanks as well to the two anonymous reviewers for their comments and careful consideration.

Abbreviations: AUTH = author, AUTO = autonomous verbal form, C = complementizer, COND = conditional, DEF = definite, DEP = dependent verbal allomorph, EMPH = emphatic, FEM = feminine, GEN = genitive, MASC = masculine, PART = participant, PL = plural, PREP = prepositional case, PRES = present, PROG = progressive, Q = interrogative, REFL = reflexive, REL = relativizer, SG = singular, VN = verbal noun.

Original Sources

TitleAuthorDialect (Island)
AMC Air mo Chuairt Ealasaid Chaimbeul Barra (Outer Hebrides) 
BSR Bha Siod ann Reimhid Lisa Storey Vatersay (Outer Hebrides) 
CC Còco is Crùbagan Floraidh NicDhmhaill Benbecula (Outer Hebrides) 
FSS Fo Sgáil a’ Swastika Dòmhnaill Iain MacDhòmhaill South Uist (Outer Hebrides) 
H Hiort: Far na laigh a’ ghrian Calum MacFearghuis South Uist (Outer Hebrides) 
MD Màiri Dhall agus Sgeulachdan Eile Donnchadh MacGillIosa Lewis (Outer Hebrides) 
MM Muinntir Mhiughalaigh Lisa Storey Mingulay & Vatersay (Outer Hebrides) 
SD Sgialachdan Dhunnachaidh Donnachadh ’Ic Dhunnachaidh South Uist (Outer Hebrides) 
TB Tales of Barra John ‘The Coddy’ MacPherson Barra (Outer Hebrides) 
TitleAuthorDialect (Island)
AMC Air mo Chuairt Ealasaid Chaimbeul Barra (Outer Hebrides) 
BSR Bha Siod ann Reimhid Lisa Storey Vatersay (Outer Hebrides) 
CC Còco is Crùbagan Floraidh NicDhmhaill Benbecula (Outer Hebrides) 
FSS Fo Sgáil a’ Swastika Dòmhnaill Iain MacDhòmhaill South Uist (Outer Hebrides) 
H Hiort: Far na laigh a’ ghrian Calum MacFearghuis South Uist (Outer Hebrides) 
MD Màiri Dhall agus Sgeulachdan Eile Donnchadh MacGillIosa Lewis (Outer Hebrides) 
MM Muinntir Mhiughalaigh Lisa Storey Mingulay & Vatersay (Outer Hebrides) 
SD Sgialachdan Dhunnachaidh Donnachadh ’Ic Dhunnachaidh South Uist (Outer Hebrides) 
TB Tales of Barra John ‘The Coddy’ MacPherson Barra (Outer Hebrides) 

References

References
Ackema,
Peter
, and
Ad
Neeleman
.
2003
.
Context-sensitive Spell-Out
.
Natural Language and Linguistic Theory
21
:
681
735
.
Adger,
David
.
1996
. Aspect, agreement and measure phrases in Scottish Gaelic. In
The syntax of the Celtic languages: A comparative perspective
, ed. by
Robert
Borsley
and
Ian
Roberts
,
200
222
.
Cambridge
:
Cambridge University Press
.
Adger,
David
.
1997
.
VSO and weak pronouns in Celtic
.
Canadian Journal of Linguistics
42
:
9
31
.
Adger,
David
.
2000
. VSO clause structure and morphological feature checking. In
Syntactic categories
, ed. by
Robert
Borsley
,
79
100
.
San Diego, CA
:
Academic Press
.
Adger,
David
.
2010a
. Gaelic morphology. In
The Edinburgh companion to the Gaelic language
, ed. by
Moray
Watson
and
Michelle
Macleod
,
283
303
.
Edinburgh
:
Edinburgh University Press
.
Adger,
David
.
2010b
. Gaelic syntax. In
The Edinburgh companion to the Gaelic language
, ed. by
Moray
Watson
and
Michelle
Macleod
,
304
352
.
Edinburgh
:
Edinburgh University Press
.
Adger,
David
.
2013
.
A syntax of substance
.
Cambridge, MA
:
MIT Press
.
Anderson,
Stephen
.
1982
.
Where’s morphology?
Linguistic Inquiry
13
:
571
612
.
Andrews,
Avery
.
1990
.
Unification and morphological blocking
.
Natural Language and Linguistic Theory
8
:
507
557
.
Aoun,
Joseph
,
Elabbas
Benmamoun
, and
Dominique
Sportiche
.
1994
.
Agreement, word order, and conjunction in some varieties of Arabic
.
Linguistic Inquiry
25
:
195
220
.
Arregi,
Karlos
, and
Andrew
Nevins
.
2012
.
Morphotactics: Basque auxiliaries and the structure of Spellout
.
Dordrecht
:
Springer
.
Baker,
Mark
.
2008
. A macroparameter in a microparametric world. In
The limits of syntactic variation
, ed. by
Theresa
Biberauer
,
351
374
.
Amsterdam
:
John Benjamins
.
Baker,
Mark
.
2012
.
On the relationship of object agreement and accusative case: Evidence from Amharic
.
Linguistic Inquiry
43
:
255
274
.
Baker,
Mark
.
2014
.
On dependent ergative case (in Shipibo) and its derivation by phase
.
Linguistic Inquiry
45
:
341
379
.
Baker,
Mark
.
2015
.
Case: Its principles and parameters
.
Cambridge
:
Cambridge University Press
.
Baker,
Mark
, and
Nadya
Vinokurova
.
2010
.
Two modalities of case assignment: Case in Sakha
.
Natural Language and Linguistic Theory
28
:
593
642
.
Berwick,
Robert
, and
Noam
Chomsky
.
2011
. The biolinguistic program: The current state of its development. In
The biolinguistic enterprise
, ed. by
Anna Maria
Di Sciullo
and
Cedric
Boeckx
,
19
41
.
Oxford
:
Oxford University Press
.
Berwick,
Robert
, and
Noam
Chomsky
.
2016
.
Why only us?
Cambridge, MA
:
MIT Press
.
Bhatt,
Rajesh
.
2005
.
Long distance agreement in Hindi-Urdu
.
Natural Language and Linguistic Theory
23
:
757
807
.
Bittner,
Maria
, and
Ken
Hale
.
1996
.
Ergativity: Toward a theory of a heterogeneous class
.
Linguistic Inquiry
27
:
531
604
.
Bobaljik,
Jonathan David
.
1995
.
Morphosyntax: The syntax of verbal inflection
.
Doctoral dissertation, MIT, Cambridge, MA
.
Bobaljik,
Jonathan David
.
2008
. Where’s phi? Agreement as a post-syntactic operation. In
Phi theory: Phifeatures across interfaces and modules
, ed. by
Daniel
Harbour
,
David
Adger
, and
Susana
Be´jar
,
295
328
.
Oxford
:
Oxford University Press
.
Bobaljik,
Jonathan David
.
2012
.
Universals in comparative morphology: Suppletion, superlatives, and the structure of words
.
Cambridge, MA
:
MIT Press
.
Bobaljik,
Jonathan David
.
2015
.
Suppletion: Some theoretical implications
.
Annual Review of Linguistics
1
:
1
18
.
Bobaljik,
Jonathan David
, and
Heidi
Harley
.
2017
. Suppletion is local: Evidence from Hiaki. In
The structure of words at the interfaces
, ed. by
Heather
Newell
,
Màire
Noonan
,
Glyne
Piggott
, and
Lisa
Travis
,
141
159
.
Oxford
:
Oxford University Press
.
Bonet,
Eulàlia
.
1991
.
Morphology after syntax: Pronominal clitics in Romance
.
Doctoral dissertation, MIT, Cambridge, MA
.
Bosch,
Anna
.
2010
. Phonology in modern Gaelic. In
The Edinburgh companion to the Gaelic language
, ed. by
Moray
Watson
and
Michelle
Macleod
,
262
282
.
Edinburgh
:
Edinburgh University Press
.
Brennan,
Jonathan
.
2009
. Pronouns, inflection, and Irish prepositions. In
Papers in syntax
, ed. by
Patricia
Irwin
and
Violeta Vázquez Roja
Maldonado
,
1
32
.
NYU Working Papers in Linguistics 2
. http://as.nyu.edu/content/dam/nyu-as/linguistics/documents/nyuwpl/brennan_irish_prepositions_09_ny uwpl2.pdf.
Brown,
Lea
.
2001
.
A grammar of Nias Selatan
.
Doctoral dissertation, University of Sydney
.
Bruening,
Benjamin
, and
Eman
Al Khalaf
.
2015
.
Linear order in syntax: Selection in coordination
.
Talk presented at NELS 46, Concordia University, October
. http://udel.edu/~bruening/Downloads/SelectionInCoordination4.1.1.pdf.
Caha,
Pavel
.
2009
.
The nanosyntax of case
.
Doctoral dissertation, University of Tromsø
.
Carnie,
Andrew
.
1995
.
Non-verbal predication and head movement
.
Doctoral dissertation, MIT, Cambridge, MA
.
Carnie,
Andrew
.
2011
.
Mixed categories in Irish
.
Lingua
121
:
1207
1224
.
Chomsky,
Noam
.
1981
.
Lectures on government and binding
.
Dordrecht
:
Foris
.
Chomsky,
Noam
.
1993
. A minimalist program for linguistic theory. In
The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger
, ed. by
Kenneth
Hale
and
Samuel Jay
Keyser
,
1
52
.
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
.
Deal,
Amy Rose
.
2017
. Syntactic ergativity as case discrimination. In
WCCFL 34: Proceedings of the 34th West Coast Conference on Formal Linguistics
, ed. by
Aaron
Kaplan
,
Abby
Kaplan
,
Miranda K.
McCarvel
, and
Edward J.
Rubin
,
141
150
.
Somerville, MA
:
Cascadilla Proceedings Project
.
Diertani,
Chaya
.
2011
.
Morpheme boundaries and structural change: Affixes running amok
.
Doctoral dissertation, University of Pennsylvania, Philadelphia
.
Doron,
Edit
.
1988
. On the complementarity of subject and subject-verb agreement. In
Agreement in natural language: Approaches, theory, description
, ed. by
Michael
Barlow
and
Charles A.
Ferguson
,
201
218
.
Stanford, CA
:
CSLI Publications
.
Embick,
David
.
2010
.
Localism versus globalism in morphology and phonology
.
Cambridge, MA
:
MIT Press
.
Embick,
David
.
2015
. On the targets of phonological realization. In
The morphosyntax/phonology connection
, ed. by
Vera
Gribanova
and
Stephanie
Shih
,
255
283
.
Oxford
:
Oxford University Press
.
Embick,
David
, and
Rolf
Noyer
.
2001
.
Movement operations after syntax
.
Linguistic Inquiry
32
:
555
595
.
Embick,
David
, and
Rolf
Noyer
.
2007
. Distributed Morphology and the syntax/morphology interface. In
The Oxford handbook of linguistic interfaces
, ed. by
Gillian
Ramchand
and
Charles
Reiss
,
289
324
.
Oxford
:
Oxford University Press
.
Fox,
Danny
, and
David
Pesetsky
.
2005
.
Cyclic linearization of syntactic structure
.
Theoretical Linguistics
31
:
1
45
.
Frampton,
John
, and
Sam
Gutmann
.
2000
.
Argument is feature sharing
. http://mathserver.neu.edu/~ling/pdf/agrisfs.pdf.
Gillies,
William
.
2009
. Scottish Gaelic. In
The Celtic languages
, ed. by
Martin
Ball
and
Nicole
Mu¨ller
,
230
304
. 2nd ed.
New York
:
Routledge
.
Goldsmith,
John
.
1976
.
Autosegmental phonology
.
Doctoral dissertation, MIT, Cambridge, MA
.
Green,
Anthony
.
2006
.
The independence of phonology and morphology: The Celtic mutations
.
Lingua
116
:
1946
1985
.
Grimshaw,
Jane
.
2000
. Locality and extended projection. In
Lexical specification and insertion
, ed. by
Martin
Everaert
,
Peter
Coopmans
, and
Jane
Grimshaw
,
115
133
.
Amsterdam
:
John Benjamins
.
Grimshaw,
Jane
.
2005
.
Words and structure
.
Stanford, CA
:
CSLI Publications
.
Haeberli,
Eric
.
2001
.
Deriving syntactic effects of morphological case by eliminating abstract case
.
Lingua
111
:
279
313
.
Hale,
Ken
.
1987
.
Incorporation and the Irish synthetic verb forms
.
Ms., MIT, Cambridge, MA
. http://lingphil.mit.edu/papers/hale/papers/hale035.pdf.
Hale,
Ken
.
1990
. Some remarks on agreement and incorporation. In
Studies in generative grammar 1
,
117
144
.
Seoul
:
The Korean Generative Grammar Circle
.
Halle,
Morris
, and
Alec
Marantz
.
1993
. Distributed Morphology and the pieces of inflection. In
The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger
, ed. by
Kenneth
Hale
and
Samuel Jay
Keyser
,
111
176
.
Cambridge, MA
:
MIT Press
.
Halle,
Morris
, and
Alec
Marantz
.
1994
. Some key features of Distributed Morphology. In
Papers on phonology and morphology
, ed. by
Andrew
Carnie
and
Heidi
Harley
,
275
288
.
MIT Working Papers in Linguistics 21
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Harley,
Heidi
, and
Rolf
Noyer
.
1999
.
State-of-the-article: Distributed Morphology
.
Glot International
4
:
3
9
.
Heath,
Jeffrey
.
2007
.
Bidirectional case-marking and linear adjacency
.
Natural Language and Linguistic Theory
25
:
83
101
.
Jouitteau,
Mélanie
, and
Milan
Rezac
.
2006
.
Deriving the complementarity effect: Relativized minimality in Breton agreement
.
Lingua
116
:
1915
1945
.
Kayne,
Richard
.
1994
.
The antisymmetry of syntax
.
Cambridge, MA
:
MIT Press
.
Lamb,
William
.
2001
. Scottish Gaelic.
Munich
:
Lincom Europa
.
Legate,
Julie Anne
.
1999
. The morphosyntax of Irish agreement. In
Papers on morphology and syntax, cycle one
, ed. by
Karlos
Arregi
,
Benjamin
Bruening
,
Cornelia
Krause
, and
Vivian
Lin
,
219
240
.
MIT Working Papers in Linguistics 33
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Legate,
Julie Anne
.
2008
.
Morphological and abstract case
.
Linguistic Inquiry
39
:
55
101
.
Macaulay,
Donald
.
1992
. The Scottish Gaelic language. In
The Celtic languages
, ed. by
Donald
Macaulay
,
137
248
.
Cambridge
:
Cambridge University Press
.
Marantz,
Alec
.
1984
.
On the nature of grammatical relations
.
Cambridge, MA
:
MIT Press
.
Marantz,
Alec
.
1988
. Clitics, morphological merger, and the mapping to phonological structure. In
Theoretical morphology: Approaches in modern linguistics
, ed. by
Michael
Hammond
and
Máire
Noonan
,
253
271
.
Leiden
:
Brill
.
Marantz,
Alec
.
1991
. Case and licensing. In
Proceedings of ESCOL ’91
, ed. by
Germán
Westphal
,
Benjamin
Ao
, and
Hee-Rahk
Chae
,
234
253
.
Ithaca, NY
:
Cornell University, Cornell Linguistics Club
.
McCloskey,
Jim
.
2004
.
Mutation and innovation in Northern Irish
.
Ms., University of California, Santa Cruz
. https://people.ucsc.edu/~mcclosk/PDF/mutation.pdf.
McCloskey,
Jim
, and
Ken
Hale
.
1984
.
On the syntax of person-number inflection in Modern Irish
.
Natural Language and Linguistic Theory
1
:
487
533
.
McFadden,
Thomas
.
2004
.
The position of morphological case in the derivation: A study on the syntaxmorphology interface
.
Doctoral dissertation, University of Pennsylvania, Philadelphia
.
Merchant,
Jason
.
2015
.
How much context is enough? Two cases of span-conditioned stem allomorphy
.
Linguistic Inquiry
46
:
273
303
.
Merrill,
John
.
2014
.
A typological overview of consonant mutation
.
Unpublished PhD prospectus
. Available at http://linguistics.berkeley.edu/~merrill/files/Merrill_prospectus_consonant_mutation.pdf.
Moskal,
Beata
.
2015
.
Domains on the border: Between morphology and phonology
.
Doctoral dissertation, University of Connecticut, Storrs
.
Norris,
Mark
.
2014
.
A theory of nominal concord
.
Doctoral dissertation, University of California, Santa Cruz
.
Polinsky,
Maria
.
2016
.
Deconstructing ergativity: Two types of ergative languages and their features
.
New York
:
Oxford University Press
.
Polinsky,
Maria
, and
Eric
Potsdam
.
2001
.
Long-distance agreement and topic in Tsez
.
Natural Language and Linguistic Theory
19
:
583
646
.
Pranka,
Paula Marie
.
1983
.
Syntax and word formation
.
Doctoral dissertation, MIT, Cambridge, MA
.
Preminger,
Omer
.
2014
.
Agreement and its failures
.
Cambridge, MA
:
MIT Press
.
Radkevich,
Nina
.
2010
.
On location: The structure of case and adpositions
.
Doctoral dissertation, University of Connecticut, Storrs
.
Reed,
Sylvia
.
2011
. The semantics of Scottish Gaelic tense and aspect. In
Formal approaches to Celtic linguistics
, ed. by
Andrew
Carnie
,
255
282
.
Newcastle
:
Cambridge Scholars Publishing
.
Roberts,
Ian
.
2010
.
Agreement and head movement: Clitics, incorporation, and defective goals
.
Cambridge, MA
:
MIT Press
.
Robinson,
Christine
.
2008
.
Agreement in Scottish Gaelic: A Distributed Morphology analysis
.
Master’s thesis, University College Dublin
.
Sigurðsson,
Halldór Ármann
.
1993
.
Agreement as head visible feature government
.
Studia Linguistica
47
:
32
56
.
Stump,
Gregory
.
1984
.
Agreement vs. incorporation in Breton
.
Natural Language and Linguistic Theory
2
:
289
348
.
Svenonius,
Peter
.
2012
.
Spanning
.
Ms., CASTL, University of Tromsø
. ling.auf.net/lingbuzz/001501/curr ent.pdf.
Trommer,
Jochen
.
2011
.
Phonological aspects of Western Nilotic mutation morphology
.
Doctoral dissertation, Universität Leipzig
.
Weisser,
Philipp
.
2018
. Why there is no such thing as closest conjunct case. In
NELS 47
, ed. by
Andrew
Lamont
and
Katerina
Tetzloff
,
219
233
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
. http://philippweisser.de/Weisser-NELS 47.pdf.
Wolf,
Matthew
.
2007
. For an autosegmental theory of mutation. In
University of Massachusetts occasional papers in linguistics 32
, ed. by
Leah
Bateman
,
Michael
O’Keefe
,
Ehren
Reilly
, and
Adam
Werle
,
315
304
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.