Glossa 9977 Finholt

Download as pdf or txt
Download as pdf or txt
You are on page 1of 37

Finholt, Aron. 2023.

Copulas, Possession, and the Temporary-


Permanent Distinction in Mashi: Evidence for decompositional
HAVE. Glossa: a journal of general linguistics 8(1). pp. 1–37.
DOI: https://doi.org/10.16995/glossa.9977

Copulas, Possession, and the Temporary-Permanent


Distinction in Mashi: Evidence for decompositional
HAVE
Aron Finholt, University of Kansas, US, [email protected]

It is cross-linguistically common to find a distinction between “temporary” vs. “permanent”


states in copular verbs, e.g., Spanish estar and ser (see Deo et al. 2017). In Mashi [shr] (Bantu,
JD53; Maho 2009), a similar temporary-permanent distinction holds not only of the two copular
be-verbs in the language, -li (temporary state) and -ba (permanent state), but also of the two
possessive have-verbs, -dwiire and -jira; -dwiire indicates temporary possession, while -jira
instead indicates permanent possession. This parallelism provides a novel argument in favor
of a decompositional approach to possessive verbs (Freeze 1992; Kayne 1993; Harley 1998).
The analysis is further supported through patterns of allomorphy where have is realized in its
decomposed form, be + comitative, in Mashi. I analyze the patterns using a Spanning Insertion
analysis (Williams 2003; Svenonius 2012; 2016; Merchant 2015).

Glossa: a journal of general linguistics is a peer-reviewed open access journal published by the Open Library of
Humanities. © 2023 The Author(s). This is an open-access article distributed under the terms of the Creative Commons
Attribution 4.0 International License (CC-BY 4.0), which permits unrestricted use, distribution, and reproduction in any
medium, provided the original author and source are credited. See http://creativecommons.org/licenses/by/4.0/.
OPEN ACCESS
2

1 Introduction
From a typological perspective, it is relatively common to find languages that make semantic
distinctions using distinct copular verbs in non-verbal predication; there is a substantial literature
suggesting that copular verbs may reflect or encode the semantic properties of the predication
relation they help realize (Ramchand 1997; Greenberg 1998; Green 2000; Bochnak et al. 2011;
Deo et al. 2017, a.o.). For example, many languages are reported to distinguish between stage-
and individual-level prediction using different copular verbs. The Spanish copular be-verbs
estar and ser exhibit distributional and interpretational differences akin to the stage-individual
contrast (Carlson 1977; Diesing 1990; 1992; Kratzer 1995; Arche 2006 a.o.), with estar being
preferred with properties that hold of stages (1), and ser being preferred with properties that
hold of individuals (2).1

(1) El artista ✓está/#es presente/ausente/lejos


The artist estar/ser.pres.3sg present/absent/far away
‘The artist is present/absent/far away’ Spanish (Maienborn 2005: 173)

(2) Maria #está/✓es inteligente/altruista


Maria estar/ser.pres.3sg intelligent/altruistic
‘Maria is intelligent/altruistic’ Spanish (Deo et al. 2017: 8)

Another distinct line of research has investigated the distribution of copulas in ostensibly
possessive constructions (Freeze 1992; Kayne 1993, a.o.). The central observation in these works
is that some languages, like Russian (3), express possession via the combination of a copular
be-verb and a locative or adpositional element, rather than a distinct have-verb.

(3) U menja byla sestra


at 1sg.gen cop sister.nom
‘I had a sister’ Russian (Freeze 1992: 577)

In yet other languages, e.g., Portuguese (4)–(5), lexical have-verbs are sometimes found to be
interchangeable with be + locative/comitative ‘with’ constructions.

(4) O menino tem fome


the child has hunger
‘The child is hungry’

(5) O menino esta com fome


the child cop with hunger
‘The child is with hunger’ Portuguese (Freeze 1992: 587)

1
See Carlson (1977) and Kratzer (1995) for discussion of these terms.
3

Based on these observations, works such as Freeze (1992); Kayne (1993); Ritter & Rosen (1997);
Harley (1998); Beavers et al. (2009); Myler (2016; 2018), a.o., attempt to connect abstract be
and have compositionally, i.e., have is built from be.
In this paper, I bridge the gap between these two distinct lines of research — namely, the
semantic contribution of copulas, and the relationship between abstract be and have — through
an investigation of copular predication and possession in Mashi [shr] (Bantu, JD53; Maho 2009),
a Great Lakes Bantu language spoken in South Kivu, Democratic Republic of Congo.2 Similar
to Spanish, we find what appears to be a stage/individual distinction between the two copular
verbs in Mashi, -li and -ba; -li lexicalizes a stage-level description , while -ba lexicalizes an
individual-level description.
However, unlike other languages that make similar copular distinctions (e.g., Spanish),
the interpretive contrast shown to hold of the two be-verbs in Mashi extends to possession
and possessive predication as well. That is, just as -li and -ba lexicalize a distinction between
temporary states and permanent states, the two have-verbs in the language, -dwiire and -jira,
lexicalize a distinction between temporary possession (-dwiire), and permanent possession (-jira).
To my knowledge, this is the first time that such an interpretive overlap has been reported of
be- and have- verb pairs in a Bantu language.
Following previous work relating abstract be and have, I suggest that the interpretational
overlap between the copular be-verbs and possessive have-verbs in Mashi provides support for
a decompositional approach to have in which it is fundamentally composed from be. In this
paper, I will specifically model this as incorporation: have is composed of a verbal head vbe that
incorporates a relational element Phave that relates two DPs in a possessive relationship (Freeze
1992; Kayne 1993; Harley 1998, a.o.). In support of this analysis, I provide morphological
evidence to show that have has an underlyingly locative structure in Mashi. Specifically, I show
that in subjunctive and infinitival clauses, the lexical have-verbs in Mashi exist in complementary
distribution with the copular possessive construction be + comitative na (i.e., so-called with-
possessive Freeze 1992). Given that the comitative particle is commonly used in conjunction with
copular be to yield the possessive in Bantu (Gibson 2012; Halpert & Diercks 2014; Myler 2016),
I argue that na is an overt exponent of the relational element Phave presented in incorporation
accounts of have (Freeze 1992; Kayne 1993; Harley 1998).

2
Mashi (JD53) is spoken by over 650,000 people primarily in South Kivu (Lewis 2009), though sizeable Bashi com-
munities can also be found in neighboring East African countries and the United States. There is very little doc-
umentation on the language overall, and relatively few descriptions of the grammar can be found (see however
Polack-Bynon 1975). The data presented in this paper were collected via direct elicitation with a native speaker
consultant living in Kansas City. All judgements are assumed to reflect a single variety of Mashi, and therefore may
not be representative of all other varieties of the language. Elicitation involved a combination of direct translation,
well-formedness judgement tasks, and semantic judgement tasks — the latter being comprised of truth-value and feli-
city-judgment tasks, which provide a salient context against which an utterance may be judged (Matthewson 2004).
4

The goal of this paper is to present a novel decompositional approach to abstract have by
extending previous analyses of the stage-individual contrast (Carlson 1977; Diesing 1990; 1992;
Kratzer 1995; Arche 2006, a.o.), to have constructions. In doing so, this paper does not attempt
to explain the semantics of the stage-individual contrast, but rather show that whatever underlies
this contrast is involved in the derivation of both be and have in Mashi. As such, the major
takeaway from this paper is that the distribution and interpretation of the be/have forms in
Mashi can only be explained if have is composed from be.

2 Copular Predication
Though property concepts are often expressed as verbs in Bantu languages, non-verbal predicates
— in the form of property denoting nominals/adjectives — are also attested (Schneider-Zioga
2018; 2021; Gibson et al. 2019 a.o.). Generally speaking, Bantu languages exhibit significant
morphosyntactic variation in how non-verbal property concepts are predicated, i.e., how they
express non-verbal predication (Gibson et al. 2019). Though various morphosyntactic strategies
may be employed across the different sub-types of non-verbal predication (Higgins 1979;
Schneider-Zioga 2018; 2021; Gluckman et al. 2022), Bantu languages tend to express “pure”
predication via an overt copular be-verb (Gibson et al. 2019). In many languages, such as Digo
(E73), an invariant form of the copula — generally ni or nu — is used to establish a predication
relation between the predicate and its subject, regardless of subject noun class (6).3

(6) Mutu hiyu ni daktari


1.person 1.dem be 1.doctor
‘This person is a doctor’ Digo (Nicolle 2013: 298)

In contrast, other languages employ an inflectional copula (or copulas) that agree(s) with the
noun class of the subject.4 Such is the case in Mongo (C61); in (7) the copula -le obligatorily
expresses class 1 agreement with the subject, nkómbé ‘kite’.

(7) Nkómbé a-le mpúlu


1.kite 1sm-be 9.bird
‘The kite is a bird’ Mongo (Hulstaert 1965: 340)

3
Per Bantuist tradition, noun class identity is marked in the gloss via noun class number. Proper names are morpho-
logically unmarked, but behave otherwise as class 1 nouns since their referents belong in class 1. Subject noun class
agreement is expressed via a subject marker (sm). Agreement with class 1/2 nouns (e.g., human nouns) is subdivided
by person: 1sg/pl for 1st person, 2sg/pl for 2nd person, and 1sm/2sm for 3rd person. Glossing otherwise follows
the Leipzig Glossing Rules.
4
Bantu languages vary significantly in the number of copular forms they employ in non-verbal predication, with some
languages reported to use up to three distinct inflectional copulas, e.g., Kifuliiru (JD63; Finholt 2023). See Schneider-
Zioga (2018); Gibson et al. (2019) for further discussion.
5

Like Mongo, Mashi falls into the category of language that utilizes an inflectional copula to
establish a predication relation between a non-verbal property concept and its subject. In fact,
Mashi exhibits two distinct copular verbs that may be used in non-verbal predication: -li and -ba.5
As illustrated in (8), both -li and -ba may be used to predicate a non-verbal property concept, e.g.,
the adjective -bulee ‘blue’. Again, both copulas obligatorily exhibit subject noun class agreement
in such contexts.6

(8) O-ma-lunga ga-✓li/✓ba ga-bulee


aug-6-sky 6sm-be-li/be-ba 6agr-blue
‘The sky is blue’

While both copulas are generally available in examples like (8), they do not exist in free-variation.
At first glance, we find that the distribution and interpretation of the two Mashi copulas aligns
with the differences observed between the Spanish copulas estar and ser — which are famously
reported to encode a stage-individual-level contrast (Carlson 1977; Diesing 1990; Kratzer 1995;
Diesing 1992; Arche 2006 a.o.). In such descriptions of the Spanish copular system, ser is treated
as an individual-level copula in that it is used to predicate an inherent/permanent property of the
subject, while estar is treated as a stage-level copula, as it is instead used to predicate a transient/
temporary property of the subject. As such, the two copulas are reported to exhibit distributional
differences in non-verbal predication, with estar co-occuring with stage-level predicates, e.g.,
presente ‘present’ (9), and ser with individual-level predicates, e.g., inteligente ‘intelligent’ (10).

(9) El artista ✓está/#es presente/ausente/lejos


The artist estar/ser.pres.3sg present/absent/far away
‘The artist is present/absent/far away’ Spanish (Maienborn 2005: 173)

(10) Maria #está/✓es inteligente/altruista


Maria estar/ser.pres.3sg intelligent/altruistic
‘Maria is intelligent/altruistic’ Spanish (Deo et al. 2017: 8)

Likewise, we find that the two copular be-verbs in Mashi exhibit distributional differences in
terms of the predicates they may co-occur with. When uttered out of the blue, -li is used to
describe time-bounded, stage-like properties, e.g., ‘sick’, (11), while -ba is instead used to describe
inherent properties, e.g., ‘tall’, (12).

5
The copular forms -li and -ba are reconstructed in proto-Bantu as the copulas *de and *bà, respectively (Meeussen
1967; Wald 1973). Cognates of both can be found throughout Bantu, although their distributions vary significantly
across languages (Gibson et al. 2019).
6
The present tense in Mashi is marked by the absence of tense morphology (i.e., T[+pres] is unpronounced at PF). In
other tenses, overt tense morphology surfaces between the subject marker and the verb root (see discussion of past
tense in section 3).
6

(11) Maria a-✓li/#ba mu-lwala


Mary 1sm-be-li/be-ba 1agr-sick
‘Mary is sick’

(12) Johne a-#li/✓ba mu-liri


John 1sm-be-li/be-ba 1agr-tall
‘John is tall’

Though the above examples point toward a lexicalized stage-individual contrast in both Spanish
and Mashi, the true nature of this distinction turns out to be much more complicated. In Spanish
for example, it is unclear whether the distributional differences between estar/ser (9)–(10) are
indicative of a true stage-individual contrast, as the two copulas do not fundamentally determine
which genre of predicate is available.7As presented in Deo et al. (2017), both Spanish estar
and ser may be used with canonically stage-level predicates and canonically individual-level
predicates, they simply yield a distinction in the particular context in which the predication
relation holds. Put simply, it is argued that the semantic properties of the predicate (alone) do
not dictate which copula will be employed, but rather the semantics of the copula dictate the
interpretation of the predicate.
Like what Deo et al. (2017) report with Spanish, we find that both -li and -ba in Mashi
may be used with canonically stage-level predicates or canonically individual-level predicates,
depending on the context. For example, while -ba is infelicitous with the stage-level predicate
‘sick’ in (12), it may be used if the property ‘sick’ holds indefinitely of the subject. In this sense,
the use of -ba does not yield an individual-level property persay per se, but rather a “permanent”
state; the property ‘sick’ holds permanently of Mary (13).

(13) Maria a-ba mu-lwala


Mary 1sm-beperm 1agr-sick
‘Mary is sick’ (Mary is chronically ill)

Similarly, while -li is dispreferred with individual-level descriptions uttered out of the blue (11),
it can be used with canonically individual-level predicates (e.g., ‘tall’) to yield a “temporary”

7
Going back as far as Ramsey (1894), estar and ser are reported to coincide with distinct interpretations of certain pre-
dicative adjectives that allow for either copula. In (i)–(ii), the predicate sucio ‘dirty’, does not restrict which copula
may be used, but rather the choice of copula restricts the interpretation of the predicate.

(i) El reporter está sucio


The reporter estar.pres.3sg dirty
‘The reporter is dirty (they need a shower)’

(ii) El reporter es sucio


The reporter ser.pres.3sg dirty
‘The reporter is dirty-minded/corrupt’ (Ramsey 1894: 313)
7

state reading. As such, we find that -li may be used with the predicate ‘tall’ in (14), but it
crucially yields a reading in which the denoted property holds only of a specific (limited) context
of evaluation (e.g., John is taller than Mary only at this moment). Such a reading is naturally
unavailable with -ba, which yields a permanent state (13).

(14) Context: John is normally shorter than Mary, but right now he’s wearing shoes that make him
taller than her.
Johne a-✓li/#ba mu-liri kulisha Maria
John 1sm-betemp/beperm 1agr-tall than Mary
‘John is taller than Mary’ (right now, but not usually)

Moreover, while both copulas were shown to be available in our original example (2, repeated
below), they crucially coincide with distinct readings. With the temporary state copula -li, the
speaker describes a property that temporarily holds of ‘sky’; the sky is blue right now, but it may
not be later. With the permanent state copula -ba, the speaker instead states a general fact about
— or a permanent property of — the sky; it is known that the sky is blue (even if it is dark and
stormy at the utterance time).

(2) a. O-ma-lunga ga-li ga-bulee


aug-6-sky 6sm-betemp 6agr-blue
‘The sky is blue’ (outside right now)
b. O-ma-lunga ga-ba ga-bulee
aug-6-sky 6sm-beperm 6agr-blue
‘The sky is blue’ (a general fact about the world)

Given the more flexible use of -li/-ba than what is generally described in Spanish, I use the terms
“permanent state” (-ba) and “temporary state” (-li), as opposed to “stage-level” and “individual-
level”, to describe the meanings of the two Mashi copulas.8

3 Possession and Possessive Predication


In this section, I will draw a direct line between the copulas -li/-ba and the two possessive have-
verbs in Mashi by showing that they share the same interpretive contrast.
Like with copular predication, Bantu languages vary with respect to the particular strategies
they use to encode both pure possession and possessive predication. Nonetheless, we generally

8
This terminology captures the fact that -ba does not correlate with “genericity”. Use of -ba in (i) yields a reading
where John lives in the lake, but not one where John routinely goes to the lake (generic reading).

(i) Johne a-ba omu-nyanja


Johne 1sm-be perm 18loc-9.lake
‘John is in the lake’
8

find that Bantu languages fall into two distinct categories in terms of how they express possession
(Gibson et al. 2019).
For many Bantu languages like Herero (R30), possessive have is expressed via a prefixal
subject agreement marker and the comitative particle na, meaning ‘and/with’ (15).

(15) U-na omu-tima omu-wa


1sm-com 3-heart 3agr-good
‘S/he has a good heart’ Herero (Möhlig & Kavari 2008: 214)

Other languages, e.g., Rangi (F33) in (16), instead employ an overt agreeing copula alongside the
comitative to express possession (Halpert & Diercks 2014; Myler 2016).

(16) A-ri na vi-jombulo vi-kuulu


1sm-be com 8-calves 8agr-big
‘S/he has big calves’ Rangi (Gibson 2012: 101)

Though these two strategies for encoding possession both employ the comitative particle —
which yields a comitative ‘accompanying’ meaning elsewhere — it is important to note that the
above examples encode genuine possession.9 That is, the subject in (15) is canonically interpreted
to possess a good heart herself, as opposed to simply being in the presence of a good heart that
is not her own (the comitative reading). While these two readings are often conflated in Bantu
languages that employ comitative na in the possessive, the crucial point here is that the genuine
possession reading is necessarily available.
Unlike languages like Herero and Rangi that make use of comitative na in possession, Mashi
exemplifies a much less common strategy for encoding possession in Bantu: it employs a lexical
have-verb. Moreover, what is particularly interesting about Mashi is the fact that it exhibits two
distinct have-verbs, -dwiire and -jira (17).10

9
This extends to possessive predication as well; (i) is interpreted as a predication relation, despite the presence of the
comitative particle.

(i) Ndi-ri na njala


1sg-be com 9.hunger
‘I am hungry’ (Lit., ‘I have hunger.’) Rangi (Gibson 2012: 92)
10
Following Halpert & Diercks (2014), I take the availability of pre-stem object marking to be a transitivity diagnostic
in Bantu languages. In contrast to what Halpert & Diercks (2014) show for the copular (be + comitative na) possess-
ives in Swahili (G41) and Zulu (S42), pre-stem pronominal object markers are available with both -dwiire and -jira in
Mashi, suggesting they are true transitive have-verbs.

(i) a. Johne a-ka-soma e-bi-tabu


John 1sm-hab-read aug-8-book
‘John reads books’

b. Johne a-ka-bi-soma
John 1sm-hab-2om-read
‘John reads them (books)’
9

(17) Johne a-✓dwiire/✓jira i-wazo li-nja


John 1sm-havetemp/haveperm 5-idea 5agr-good
‘John has a good idea’

Though both available in examples like (17), these two have-verbs are crucially not in free-
variation. In fact, they lexicalize the same temporary-permanent distinction observed of the
two be-verbs in section 2; -dwiire encodes temporary possession, while -jira encodes permanent
possession. Considering the relative infrequency of lexical have-verbs across Bantu languages
(Halpert & Diercks 2014; Myler 2016; Gibson et al. 2019), such an interpretive contrast is
unsurprisingly quite rare; this is to my knowledge the first time that such a system has been
reported in Bantu.
Let us first consider the interpretive overlap between the temporary state copula -li
and the temporary possessive -dwiire. Just as the copula -li describes a property that holds
exclusively of a limited context of evaluation, -dwiire expresses a time-bounded associative
relationship between two individuals; since the relationship between John and ‘a child’ in (18)
is not permanent (i.e., the child doesn’t live in his arms), only the temporary possessive -dwiire
is available.

(18) Johne a-✓dwiire/#jira oo-mw-ana omu-boko


John 1sm-havetemp/haveperm aug-1-child 18loc-5.arm
‘John has a child in his arms’

On the other hand, we find that just as the copula -ba encodes a permanent state, the have-
verb -jira expresses an inherent associative relationship between two individuals, viz. permanent

(ii) a. Johne a-✓dwiire/jira a-ba-ana


John 1sm-havetemp/have perm aug-2-child
‘John has children’

b. Johne a-ba-✓dwiire/✓jira
John 1sm-2om-havetemp/have perm
‘John has them (children)’

(iii) a. Oo-mutugari gu-✓dwiire/✓jira bi-gondo


aug-3.car 3sm-havetemp/have perm 8-tire
‘The car has wheels’

b. Oo-mutugari gu-bi-✓dwiire/✓jira
aug-3.car 3sm-8om-havetemp/have perm
‘The car has them (wheels)’

As pointed out by an anonymous reviewer, this raises questions about the status of the copular possessive construc-
tion (e.g., be + comitative na) in Mashi, which obligatorily appears in place of a have-verb in irrealis environments
(see section 4). This in turn gets at other issues related to transitivity and object marking in Bantu languages that
need further study. For this reason, I will leave any further discussion of these issues to future work.
10

possession. As such, only -jira is available in contexts where an associative relationship between
two individuals holds indefinitely, e.g., the inherent associative relationship between a father
and child (19).11

(19) Johne a-#dwiire/✓jira oo-mw-ana


John 1sm-havetemp/haveperm aug-1-child
‘John has a child (John is a father)’

While it may be tempting to treat the two have-verbs as encoding solely alienable/inalienable
possession based on the data above, I analyze (in)alienability, like the stage-individual contrast,
as a sub-case of the broader temporary-permanent distinction. Accordingly, both have-verbs may
describe alienable/inalienable possession, albeit with different interpretations. For example,
while -dwiire and -jira are both available in examples like (17), they yield distinct interpretations
of the possessive relationship; with -dwiire (20a), the relationship between John and ‘good idea’
is new (e.g., he just came up with it), while with -jira (20b), this relationship is long-term (e.g.,
John has had this good idea his whole life).

(20) a. Johne a-dwiire i-wazo li-nja


John 1sm-havetemp 5-idea 5agr-good
‘John has a good idea’ (a new idea)
b. Johne a-jira i-wazo li-nja
John 1sm-haveperm 5-idea 5agr-good
‘John has a good idea’ (he’s profited on it for a long time)

Similarly in (21)–(22), the two have-verbs are distinguishable only in that -jira yields a
“permanent” possession reading, and -dwiire a “temporary” possession reading. In both cases, the
possession relationship between the car and the wheels is inalienable.

(21) Context: You’re naming all the things you know are true about cars.
Oo-mutugari gu-jira bi-gondo bi-nne
aug-3.car 3sm-haveperm 8-tire 8agr-four
‘The car has four wheels’

(22) Context: You see a car driving down the road with only three wheels. Shocked that it is
driving with a missing wheel, you point and exclaim:
Oo-mutugari gu-dwiire bi-gondo bi-sherho
aug-3.car 3sm-havetemp 8-tire 8agr-three
‘The car has three wheels’

11
Note that -dwiire can be used in (19) if the possessive relationship is transient, e.g., if John is watching over the child
temporarily.
11

While the use of the permanent possessive -jira in (21) follows from the general expectation that
cars have four wheels (i.e., a permanent associative relationship), the context in (22) presents a
context dependent associative relationship; since a car having only three wheels is exceptional
and holds only of this context, only temporary -dwiire is available.
In the previous sections, I have demonstrated that both the be-verbs and the have-verbs
in Mashi lexicalize the same temporary-permanent distinction. That is, just as the copula -li
encodes a temporary state and the copula -ba encodes a permanent state, the have-verb -dwiire
encodes temporary possession, and -jira permanent possession. To conclude this section, I would
like to discuss one additional observation that nicely illustrates this interpretive overlap: when
two predicates that convey the same (general) property exhibit different syntactic requirements,
the choice of be/have-verb coincides with the interpretational restrictions of the property they
both denote.
Consider the adjectival predicate ndaa ‘pregnant’, and the nominal predicate izimi ‘pregnancy’.
Though these two predicates differ in category, they attribute the same general property,
‘pregnant’, to their subject. Nonetheless, the two crucially differ in which predicational strategy
they employ; ndaa requires copular predication (23), while izimi requires possessive predication
(24). Moreover, since the property ‘pregnant’ holds only of a limited context of evaluation,
both predicates require the “temporary” copula/possessive; ndaa obligatorily occurs with the
temporary state copula -li, while izimi obligatorily occurs with the temporary possessive -dwiire.

(23) Maria a-✓li/#ba ndaa


Mary 1sm-betemp/beperm 1agr.pregnant
‘Mary is pregnant’

(24) Maria a-✓dwiire/#jira izimi


Mary 1sm-havetemp/haveperm 9.pregnancy
‘Mary is pregnant’ (Lit., ‘Mary has pregnancy.’)

Given that the two predicates ndaa and izimi describe the same general property, and necessarily
require the “temporary” copula/possessive, the only difference between the two is predicational
strategy. As such, the data in (23)–(24) again suggest a significant interpretive overlap between
the temporary state be-verb -li, and the temporary possessive -dwiire. While -li and -dwiire may
differ in terms of the morphosyntactic environment in which they may be realized (i.e., they
vary according to the predicational requirements of the predicate), they are interpretively quite
similar, since they may both be used to describe a single property like ‘pregnant’.

4 The link between be and have


To this point, I have shown that the be/have-verbs in Mashi exhibit a significant overlap in
interpretation. While this interpretive overlap alone is perhaps sufficient to make the claim that
12

(abstract) be and have are compositionally related in Mashi, there is also significant distributional
evidence to suggest that these forms are quite closely linked. Though the be-verbs (-li/-ba) and
have-verbs, (-dwiire/-jira) generally lexicalize a temporary-permanent distinction, their realization
is sensitive to Mood and Tense. Specifically, we find that the temporary-permanent distinction
is morphologically neutralized for both be and have in two specific contexts: the past tense,
and the subjunctive/infinitival (irrealis) moods. In each of these neutralizing contexts, be and
have exhibit similar invariant surface forms, e.g., the prototypical “temporary” forms -li/-dwiire
obligatorily surface in the past tense, while -ba (along with the comitative particle na in the case
of have) surface in the subjunctive/infinitival moods (see Table 1).

Environment Surface Form


Mood Tense Temporary Permanent
be +indic +pres -li -ba
+indic +past -li -li
+sbjv – -ba -ba
+inf – -ba -ba
have +indic +pres -dwiire -jira
+indic +past -dwiire -dwiire
+sbjv – -ba (+na) -ba (+na)
+inf – -ba (+na) -ba (+na)

Table 1: Distribution of be/have allomorphs.

As we have seen, Mashi encodes a temporary-permanent distinction in its be- and have-verb
pairs. This contrast is salient in the present indicative, where both be-verbs and both have-verbs
are available. In the past tense however, be and have obligatorily surface as -li and -dwiire —
their “temporary” forms in the present indicative. Crucially, these forms are invariable; they
surface regardless of the interpretation of the predication relation. This can be seen illustrated
for be below; -li ([-ri]) co-occurs with both temporary (25) and permanent interpretations (26)
in the past tense.12

12
That be surfaces as -ri in the presence of the past tense marker -li is taken to be a purely phonological process. This
reflects a common observation in Bantu languages that [l] and [r] exist in an allophonic relationship (Maddieson &
Sands 2019). This is true in related Kinyarwanda and Kinyamulenge (JD61; author notes), which exhibit a similar
tense-conditioned neutralization pattern: though two copular forms are attested in the present tense, only -ri (in free
variation with [-li]) appears in the past tense. As such, I argue that treating -ri as a phonological variant of copular
-li — as opposed to a distinct vocabulary item — better fits the broader typological picture. See however footnote 26
for discussion of an alternative approach. Note also that past tense -li is not a copula synchronically (c.f. Swahili),
though it may have a diachronic source as a copula.
13

(25) Context: Mary had a cold yesterday, but now she’s feeling better.
Maria a-li-ri mu-lwala
Mary 1sm-past-betemp 1agr-sick
‘Mary was sick’

(26) Context: Mary has had a chronic illness her whole life, but she’s been miraculously cured.
Maria a-li-ri mu-lwala
Mary 1sm-past-betemp 1agr-sick
‘Mary was sick’

Likewise, have may only surface as -dwiire in the past tense, again regardless of the interpretation
of the possession relation; the permanent have-verb form -jira is crucially unavailable (27).

(27) Johne a-li-✓dwiire/*jira o-mu-lali oku a-ja e Kinshasa


John 1sm-past-havetemp/haveperm aug-3-plan comp 1sm-go loc Kinshasa
‘John had a plan to go to Kinshasa’

Much like the past tense, we find that the subjunctive mood serves as a neutralizing environment
for both be and have. In this environment, be is obligatorily realized as the permanent state
copula -ba; the temporary state copula -li is unavailable (28).13

(28) O-shinganine o-*l(e)/✓be mu-liri o-twa a-ma-lehe


2sg-must.sbjv 2sg-betemp.sbjv/beperm.sbjv 1agr-tall 2sg-pick aug-6-fruit
oku-gu-murhi
15loc-3dem-3-tree
‘You must be tall to pick fruit from this tree’

Likewise, have is also limited to one possible realization in the presence of the subjunctive mood.
However, this form is distinct from the two have-verb forms we find in the present indicative;
both -dwiire and -jira are crucially unavailable.14 Instead, have is obligatorily realized as the
combination of the permanent state copula -ba and the comitative particle na, meaning ‘and/
with’ (29)–(30).

(29) Bi-kuhunire o-be na oo-mutugari


8sm-require 2sg-beperm.sbjv com aug-3.car
‘It is required that you have a car’

13
Like many other Bantu languages, Mashi generally expresses mood contrasts via a change in the final vowel (FV)
of the verb. The subjunctive mood is marked via the subjunctive final vowel, -e. This competes with the underlying
vowel in the verb root, yielding a surface form of -be (from -ba) for be.
14
Given that subjunctive mood is generally marked via the final vowel -e, there should be no morphological differ-
ence between the indicative and subjunctive forms of -dwiire, however, -dwiire is shown to be incompatible with the
subjunctive environment in (30). Correspondingly, the expected subjunctive-marked form of -jira, i.e., -jire, is also
unavailable in this environment, suggesting that the subjunctive imposes a general restriction on overt have-verbs.
14

(30) *Bi-kuhunire o-dwiire/jire oo-mutugari


8sm-require 2sg-havetemp.sbjv/haveperm.sbjv aug-3.car
‘It is required that you have a car’

Importantly, be (28) and have (29) both realize the permanent state copula -ba in the subjunctive
mood. The only difference between the two in this context is the presence of the comitative
particle na; have is realized as be + comitative.
We find a similar pattern to the subjunctive in the infinitival mood: both be and have realize
the permanent state be-verb form -ba — namely its infinitival form okuba (31), with have being
realized as the combination of -ba and comitative na (32).

(31) Bi-ri bi-nja oku-ba mu-liri


8sm-betemp 8agr-good inf-beperm 1agr-tall
‘It is good to be tall’

(32) Bi-kuhunire oku-ba na oo-mutugari


8sm-require inf-beperm com aug-3.car
‘It is required to have a car’

The important point here is that, just as in the past tense, the surface forms of be and have are
invariable in the subjunctive and infinitival moods, as they do not morphologically encode the
temporary-permanent distinction.
While both Tense and Mood condition the realization of be/have, it is worth noting the
distinct effect of Mood on the realization of have. As illustrated above, neither of the lexical
have-verbs (-dwiire/-jira) are available in the subjunctive/infinitival moods. Instead, possessive
have is obligatorily realized as the copular possessive construction, be + comitative na.
From a Bantu perspective (see section 3), it is not surprising to find such a copula-based
possessive construction in Mashi, as this sort of possessive strategy is commonly attested in
Bantu languages (Halpert & Diercks 2014; Myler 2016; Gibson 2012; Gibson et al. 2019).
Although this copular possessive construction is exceptionally limited to subjunctive and non-
finite clauses in Mashi, the fact remains that it exists in a complementary relationship with
the have-verbs -dwiire/-jira; when lexical have is unavailable, be surfaces (with the comitative
particle na) in its place.
Summarizing the above, we find that in the past tense, be and have are realized as their
“temporary” forms -li/-dwiire. In irrealis moods, be forms are realized as the “permanent”
form -ba, and have forms are realized as permanent copula -ba with the comitative particle.
When considered alongside the general interpretational overlap between the be/have-verbs, the
parallel effects of Mood and Tense on the realization of be and have provide a strong case for a
15

decompositional approach to possessive have (Freeze 1992; Kayne 1993; Ritter & Rosen 1997;
Harley 1998; Beavers et al. 2009; Myler 2016; 2018 a.o.). The alternation between the lexical
have-verbs and the copular possessive construction, be + comitative, provides particularly strong
support for a decompositional approach in which have is built via the incorporation of a locative
element by be, as I will discuss in subsection 4.1. Importantly, an incorporation-based approach
to decompositional have provides a straightforward analysis of the interpretational overlap
between the Mashi be/have-verbs; if have is built from be, then the temporary-permanent
distinction shown to hold of the have-verbs falls out of the fact that the be-verbs lexicalize the
same distinction.

4.1 Decomposing have


Following previous decompositionalist analyses of possessive have (Freeze 1992; Kayne 1993;
Ritter & Rosen 1997; Harley 1998; Beavers et al. 2009; Myler 2016; 2018 a.o.), I argue that the
overlap between the Mashi be/have-verbs is best accounted for under an approach where have
is directly composed from be.
In this section, I will introduce two previous decompositional analyses of possessive have
— one that posits incorporation (Freeze 1992; Kayne 1993; Harley 1998), and another that
posits contextual allomorphy (Myler 2016; 2018) — and discuss how each could fit the Mashi
data. Ultimately I will demonstrate that both of these analyses fail to capture the fact that the
lexical have-verbs, -dwiire/-jira, are strictly unavailable in the presence of irrealis mood (e.g., in
subjunctive and infinitival clauses). In light of this sensitivity to Mood, I will instead present a
decompositional analysis of possessive have that defines the domains for Vocabulary Insertion
and allomorphic conditioning according to spans — or contiguous strings — of terminal nodes
in a single extended projection (Williams 2003; Svenonius 2012; 2016; Merchant 2015; Hanink
2018). By defining locality conditions on allomorphy according to adjacent spans of terminal
nodes rather than adjacent heads, the analysis accounts for the fact that stem selection is sensitive
to morphosyntactic information in various places, which renders it unable to be otherwise
derived cyclically.

4.1.1 The P-Incorporation Account


Given the crosslinguistic tendency for languages to either encode possession via an overt lexical
have-verb, or the combination of copular be + locative, accounts like Freeze (1992), Kayne
(1993), and Harley (1998) suggest that have is underlyingly composed of two verbal heads: vbe
and a functional element P (or Phave) that serves to relate two DPs in a possessive relationship.
The core idea of these accounts is that have is realized through the incorporation of a lower
functional element Phave by a stative verbal head vbe (33).
16

(33) vP

v be PP

DPPossessor P

Phave DPPossessee

When one considers that Bantu languages often express have via the combination of
copular be and an overt prepositional element, usually the comitative particle na meaning
‘and/with’ (see section 3 for discussion), it makes sense to treat the parallelism between
the copular be-verbs and the possessive have-verbs in Mashi as being fundamentally built
into the syntax. That is, the overlap between copular -li/-ba and possessive -dwiire/-jira
makes sense if the two have-verbs are realized as be+Phave, and only differ in which be-verb
incorporates Phave: if beperm (-ba) incorporates Phave, it realizes the permanent possessive -jira
(42), whereas if betemp (-li) incorporates Phave, it instead realizes the temporary possessive
-dwiire (34b).

(34) a. [ v-ba Phave ] → /jira/


b. [ v-li Phave ] → /dwiire/

Note that, under this genre of analysis, there are two possible ways to derive the temporary-
permanent distinction. On one analysis, there are two semantically distinct vbe heads that may
enter into the derivation: v-li encodes a temporary state, while v-ba encodes a permanent state.
The second approach instead treats vbe as a meaningless copula, and attributes the temporary-
permanent (or stage-individual) contrast to the existence of two distinct Pred heads (Myler 2016;
2018).15 Under this approach, the distinction between betemp/beperm depends on which Pred head
moves/incorporates into vbe; the combination vbe + Predstage yields the temporary state copula
-li, whereas vbe + Predindiv yields the permanent state copula -ba. Deriving the two have-verbs
remains largely the same under this approach, with the sole difference being that a now complex
head vbe+Pred incorporates Phave (as opposed to just vbe). For the sake of consistency, I will

15
Following Wood (2015) and Myler (2016; 2018), I assume that vbe carries no semantic meaning, but instead serves a
purely syntactic function. As such, I take vbe to denote a type-neutral identity function (Myler 2016; 2018).

(i) Copula vbe


vbe ⟺ λx.x

This meaningless head serves to integrate non-verbal predicates into the verbal domain, since PredP — being non-
verbal —- cannot directly merge with clausal functional projections (see Myler 2016; 2018).
17

assume that the latter approach — which posits a syntactically represented Predstage/indiv head
below vbe — applies to both copular and possessive constructions in Mashi.16

(35) vP

v be PredP

DP PredP

Pred XP

Copular Predication

(36) vP

v be PredP

Pred PP

DP P

Phave DP

Possessive Predication

The idea here is that the overlap between the be- and have-verbs in Mashi is the direct result of
there being two derived forms of be. Depending on which be-verb — whether the complex head
vbe+Predstage or vbe+Predindiv — incorporates Phave, a different have-verb surfaces. As such, the
P-incorporation account directly explains the overlap that we see between the temporary state
copula -li and temporary possessive -dwiire with the near synonyms ndaa ‘pregnant’, and izimi
‘pregnancy’, in section 4. Crucially, the verbal head vbe and the predicative head Predstage are
involved in both constructions. With ndaa ‘pregnant’ — which requires copular predication —
the combination of vbe and Predstage surfaces as the temporary state copula -li (37). However,
since the nominal predicate izimi ‘pregnancy’, requires possessive predication, it exists within
a possessive relationship with the subject (e.g., Mary) — a relationship which is mediated by
the relational element Phave. Under the incorporation account, Phave is then incorporated by the
complex head vbe+Predstage, yielding the temporary possessive -dwiire (38).

16
This approach also provides a neat explanation of the stage-individual collapse observed in certain environments (see
section 5).
18

(37) vP

v be PredP

DP PredP
Maria

Predstage AdjP
ndaa
Maria ali ndaa ‘Mary is pregnant’

(38) vP

v be PredP

Predstage PP

DP P
Maria

Phave DP
izimi
Maria adwiire izimi ‘Mary is pregnant’

This same process also holds of the relationship between the permanent state copula -ba and
the permanent possessive -jira. If we instead found near synonyms that require the permanent
be/have-verbs, their structural representations would be identical to those in (37)–(38), with
the sole change being the presence of a Predstage head. In both cases, the relationship between
the be/have-verbs is compositional; the syntactic material that yields be is also present in the
composition of have.
Now, while the incorporation approach does well to account for the interpretational overlap
between the be/have-verbs, it crucially fails to capture the allomorphic effects of Tense (25) and
Mood (28)–(29). As I will show, the P-incorporation account has no way of accounting for the
fact that the surface realization of vbe — as either a be or have-verb — is simultaneously sensitive
to which Pred head is merged, the features associated with the Mood/Tense head(s), and also
whether Phave is present. To illustrate, let us walk through a derivation of be and have.
19

In the present indicative, either vbe and Pred (be) or vbe, Pred, and Phave (have) are realized
as one complex head. This is accomplished by having Pred move to vbe to create a complex
head, vbe+Pred, that then incorporates Phave in the case of have, or by having Pred incorporate
Phave, if present, before then moving to vbe. In either case, if Phave is present, it incorporates
with Predstage or Predindiv. However, as discussed above, the possible surface realizations of be/
have are restricted when Mood or Tense is modulated. In the presence of irrealis mood for
example, incorporation crucially does not occur; vbe is obligatorily realized as the permanent
state copula -ba (alongside the comitative na in the case of have), regardless of interpretation
(i.e., no matter which Pred head is merged). Similarly, the presence of past tense restricts
the realization of be/have to their temporary forms, -li and -dwiire, again regardless of
interpretation.
Herein lies the problem faced by the P-incorporation approach; if we assume cyclical
derivation, we can’t explain what does — and does not — incorporate, nor how those heads may
be realized. If have is built cyclically, vbe first encounters Pred, which determines the temporary-
permanent interpretation, and Phave, which distinguishes be from have, before interacting with
Mood/Tense later in the derivation. This is problematic however, since the features associated
with Mood/Tense may fundamentally determine the realization of vbe, regardless of what is
present lower in the structure (e.g., Pred/Phave).
At this point, I would like to head off two alternative analyses. The first has to do with
the hierarchical order of Mood and Tense: perhaps the incorporation account doesn’t work
because Mood is non-adjacent to vbe (e.g., the two are mediated by Tense). Note however that the
observed issues with incorporation would persist even if we reorder Mood and Tense, as we still
need some way to block incorporation countercyclically. The second alternative attributes the
blocking effects of Mood to linearization: maybe the subjunctive FV -e linearizes to right of the
verb, and blocks post-syntactic incorporation of na. This genre of analysis doesn’t work because
the infinitival form of the be-verb, okuba, doesn’t display this kind of morphological intervention,
despite exhibiting the same blocking effects as the subjunctive.

4.1.2 Contextual Allomorphy Account


An alternative approach to the parallelism between the be/have-verbs in Mashi would be to take
have as a contextually conditioned allomorph of vbe (Myler 2016; 2018). Under this analysis, we
would assume the be/have-verb forms in Mashi to be four allomorphs of vbe that are conditioned
by the presence of lower Pred head as well as the presence of a transitive Voice head. Much like
the P-incorporation approach discussed previously, the contrast between the copular be-verb
allomorphs, -li/-ba, coincides with which Pred head is merged; Predstage yields the temporary
state copula -li (39a), while Predindiv yields the permanent state copula -ba (39b).
20

(39) a. [ vbe ] → /li/ / ____ Predstage


b. [ vbe ] → /ba/ / ___ Predindiv

The derivation of have is slightly more complicated. Myler (2016; 2018) presents have as
an allomorph of vbe that appears in the context of a transitive Voice head. To capture the fact
that Mashi exhibits two have-verbs that distinguish between temporary/permanent possession,
however, we need to posit that Mashi have is sensitive to both the presence of a transitive Voice
head and a lower Pred head. As such, the temporary have-verb is an allomorph of vbe that appears
in the context of a transitive Voice head and Predstage (40a), whereas the permanent have-verb
is an allomorph of vbe that appears in the context of a transitive Voice head and Predindiv (40b).

(40) a. [ vbe ] → /dwiire/ / VoiceD ____ Predstage


b. [ vbe ] → /jira/ / VoiceD ____ Predindiv

This analysis therefore accounts for the fact that the temporary-permanent distinction applies to
both the be- and have-verbs via the combination of a lower allomorphic trigger, e.g., Pred, which
serves to capture the temporary-permanent distinction, and a higher allomorphic trigger, e.g.,
the transitive Voice head, which yields have from either vbe+Predstage or vbe+Predindiv. Under
the assumption that this sort of contextual allomorphy can be sensitive to both higher and lower
triggers simultaneously (e.g., if allomorphy can “look both ways”; see Merchant 2015), and that
the allomorph of vbe that surfaces under transitive Voice retains/projects the lower Predstage/indiv
head, then such an account should be able to capture the Mashi data, as it would crucially yield
the same four way contrast.
That being said, there are two significant problems faced by the contextual allomorphy
account. The first is an issue of locality; given the observed effects of Mood/Tense, both heads
must be sufficiently local to condition allomorphy of vbe. To illustrate, let us consider the effect
of Mood on the realization of have. In the indicative mood, we find a temporary-permanent
distinction in the two have-verbs, -dwiire/-jira. This contrast is however neutralized in the irrealis
mood(s), as have obligatorily surfaces as the copular possessive construction, be + comitative
(e.g., -ba + na), regardless of interpretation. Assuming that Mood is above Voice, we can derive
the three observed have forms with the Vocabulary Insertion rules in (41).

(41) a. [ vbe ] → /dwiire/ / Mood[+indic] VoiceD ____ Predstage


b. [ vbe ] → /jira/ / Mood[+indic] VoiceD ___ Predindiv
c. [ vbe ] → /ba / / Mood[-indic] VoiceD ___ Predstage/indiv

The issue here is that, given general locality conditions on contextual allomorphy (Embick 2012),
it is surprising that the features associated with the Mood head fundamentally determine the
realization of vbe, as Mood is crucially non-adjacent to vbe (41c). Put simply, it is unexpected that
21

the surface realization of vbe may be conditioned by a non-adjacent head, Mood, as allomorphic
triggers are generally taken to be strictly local (Embick 2012; Merchant 2015).
One way around this adjacency issue is to define insertion and locality relative to spans of
consecutive heads, rather than individual terminal nodes (Merchant 2015; Myler 2018). The
idea here is that, like Merchant (2015) demonstrates for Greek stem allomorphy, sensitivity to a
non-adjacent head (e.g., Mood) renders it impossible to predict the surface realization of Mashi
be/have based solely on elements immediately adjacent to vbe. If however, we take insertion to
target a span of heads, e.g., [Mood-Voice-vbe-Pred], then said non-adjacent heads, e.g., Mood,
would be close enough to condition allomorphy of vbe, thus resolving any locality issues.
The second problem faced by the contextual allomorphy account concerns the Mood-
conditioned alternation between lexical have and the copular possessive construction, be +
comitative na. The issue here is that, if we take possessive have to be an allomorph of vbe
that appears in the environment of a transitive Voice head, we have no explanation for why
have should sometimes surface as a lexical be-verb, and sometimes as be + comitative na. By
defining the distinction between be/have according to Voice — as opposed to the presence/
absence of an adpositional element like Phave (Freeze 1992; Kayne 1993; Harley 1998) — it
becomes much more difficult to account for the complementary distribution of lexical have and
be + comitative, as these would crucially correspond to distinct structures.17 In contrast, this
alternation is quite amenable to the aforementioned P-incorporation approach, which assumes
that have is underlying be + Phave (e.g., -ba + comitative na).
In this section, I have shown that a portmanteau analysis where heads are collected onto
a single complex head, e.g., P-incorporation, cannot account for the surface patterns unless
it sometimes allows for collection/incorporation to be countercylically blocked. Likewise, a
contextual allomorphy account cannot derive the surface patterns unless it posits non-locally
conditioned allomorphy and also explains why have may be realized as be + comitative. Given
these observations and the fact that the derivation of Mashi have is simultaneously sensitive to
four distinct elements — namely Mood, Tense, Voice, and Pred — I conclude that the present
data are incompatible with standard approaches to P-incorporation and contextual allomorphy,
and instead propose an analysis that bridges the gap between them. In the following section, I

17
To account for the fact that Bantu comitative na may be used to express possessive have, Myler (2016) proposes two
allosemes of na: one with a comitative semantics (like psym in Levinson 2011), and one that is expletive.

(i) Allosemy of na
a. na ⟺ λP s,t.λye.λe s.P(e)∧companion(e,y) / ____s,t
b. na ⟺ λx.x

While this accounts for the interpretive flexibility of the Bantu comitative, it does not resolve the issue that the cop-
ular possessive construction, be + comitative, alternates with the lexical have-verbs in Mashi. If have is not built
from be + Phave, we shouldn’t expect these forms to be in complementary distribution.
22

lay out an analysis that takes the Mashi be/have forms to be allomorphs of vbe, and have to be
derived from be+Phave. Moreover, following accounts like Williams (2003); Svenonius (2012;
2016); Merchant (2015); Hanink (2018), I suggest that the Mashi data conform to an analysis
that takes the domain for Vocabulary Insertion to be not a single terminal node, but rather a span
of consecutive nodes in a maximal projection (Merchant 2015).

4.2 The case for span-conditioned allomorphy


As I have shown, the Mashi be/have data is incompatible with approaches that assume strict
locality conditions on allomorphic conditioning, and take lexical insertion to target individual
terminal nodes. Instead, similar to Myler (2018), I assume that locality may be defined according
to contiguous spans of heads. In light of previous work on cases of non-locally conditioned
allomorphy (Williams 2003; Svenonius 2012; 2016; Merchant 2015; Hanink 2018), I specifically
adopt the Spanning Insertion Hypothesis presented in Merchant (2015) and suggest that the
Mashi be/have data is best accounted for as a case of span-conditioned allomorphy. In this
section I show that it is possible to derive the correct be/have forms by defining lexical insertion
domains and locality conditions according to contiguous spans of terminal nodes, rather than
single terminal nodes. This analysis ensures that lexical insertion has access to/can be conditioned
by the four (sometimes non-adjacent) heads constraining the realization of vbe: namely Mood,
Tense, Pred, and Phave.
Though the suggestion that lexical insertion may target a string of consecutive nodes has
its roots in earlier works (Williams 2003; Svenonius 2012), I will treat the formal approach
laid out in Merchant (2015) as the baseline for the present analysis. In brief, Merchant (2015)
shows that certain transitive verbs in Greek show simultaneous sensitivity to both Voice and
Aspect, despite Aspect being syntactically non-local. Specifically, he shows that verbs like troo
‘eat’, exhibit distinct forms that vary according to the features associated with Voice — namely
[+/–active] — and also the features associated with Aspect — namely [+/–perfective]. To
illustrate, I have replicated the VI rules for troo ‘eat’, first presented in Merchant (2015) below
(42). As Merchant points out, these VI rules quite clearly violate standard locality conditions on
contextual allomorphy.18

18
Though many previous accounts have proposed strict adjacency requirements on allomorphic conditioning (Allen
1978; Siegel 1978; Embick 2010; 2012; Bobaljik 2012), Merchant (2015) specifically frames this as a violation of
the “concatenation” condition (Condition A2) proposed in Embick (2010; 2012). This condition broadly states that
the features of one morpheme may only condition the realization of another morpheme if the two morphemes are
adjacent.

(i) Condition A2
Contextual allomorphy requires concatenation (linear adjacency)
23


(42) a. eat → fáɣ(o) / ____ Voice[+act] Aspect[+perf]

b. eat → fáɣo / ___ Voice[–act] Aspect[+perf]

c. eat → tro(ɣ)

The idea here is that it is impossible to predict which stem form of eat will appear without
having access to both the adjacent head Voice and also the higher head Aspect; without access
to both heads, there is no way to differentiate between fáɣ(o), fáɣo, and tro(ɣ) — as the latter
appears in all imperfective contexts. Note however that the conditioning elements in (42), namely

Voice and Aspect, are discretely exponed. As such, the realization of eat is conditioned by
the heads Voice and Aspect, despite not being jointly realized with them. This observation is at
the heart of the distinction between insertion and conditioning spans (Merchant 2015). Canonical
insertion spans represent the consecutive series of nodes that are jointly targeted by Vocabulary
Insertion. However, spans are not uniquely a mechanism for insertion. In fact, spans of heads —
specifically the features associated with those heads — may serve as local allomorphic triggers
for an adjacent span (Merchant 2015; Hanink 2018, a.o.). So-called conditioning spans are
characterized as a consecutive series of nodes whose features may condition the realization of an
adjacent insertion span.19 Crucially, these two types of spans can be distinct, but they need not
be; insertion spans and conditioning spans may either target the same series of nodes, or they
may target distinct — but potentially overlapping — series of nodes. Given this distinction, Voice
and Aspect are argued to form a conditioning span in (42), as they condition the possible surface

form of the adjacent insertion span which realizes eat .
Although the relevant features conditioning stem-alternation in Greek may be distinct from
the features conditioning the realization of vbe in Mashi, the general observation is the same.
To capture the three possible realizations of have — temporary possessive -dwiire, permanent
possessive -jira, and the copular possessive -ba + na — Vocabulary Insertion need necessarily
have access to the four distinct heads that condition the realization of vbe. In light of the fact that
many Bantu languages encode have via the copular possessive construction be + comitative
na, I will assume, following the P-incorporation literature, that the relational element Phave is
involved in the derivation of have in Mashi. Nonetheless, like Myler (2016; 2018), I will treat
the be/have forms in Mashi as allomorphs of a single vbe head, with the caveat that this head is
realized as part of a span.
To capture the interpretive overlap between the be/have forms while also accounting for
the effects of Mood and Tense, I define two distinct spans. Both the structural contrast between

19
Under the Spanning Insertion Analysis (Merchant 2015), the features of one span may condition the realization of
another span so long as they are directly adjacent. As such, the features associated with the span [A1-A2-A3] may
condition the realization of another span [B1-B2-B3] so long as there are no heads that intervene between them.
24

be/have and the interpretive temporary-permanent distinction arise from the single insertion
span [vbe-Pred-Phave], with Pred determining the interpretation, and the presence/absence of
Phave distinguishing be and have. To capture the effects of Mood and Tense, I also define a
higher conditioning span, [Mood-T]. This span serves to restrict the possible surface forms of
the adjacent insertion span that realizes vbe. Importantly, the proposed multi-span analysis also
ensures that both spans, [Mood-Tense] and [vbe-Pred-Phave], may be discretely exponed. As I will
discuss below, this turns out to be a crucial aspect of the analysis, as we find that overt exponents
of [Mood-Tense] can co-occur with be/have forms. The basic idea of the multi-span analysis is
laid out in (43).20

(43) The Multi-span Analysis


Mood − Tense v − Pred − (Phave )
    be  
condi t ioning inser t ion

To illustrate how this works, consider first the effect that Tense has on the realization of be and
have. In the present tense we need to distinguish temporary/permanent forms, a distinction
which is neutralized in the past. This can be achieved with the following VIs (44a)-(44d). These
VI rules assume that Mood and Tense form a conditioning span, and are not implicated in the
insertion span that realizes vbe (indicated by frame boxes).

(44) a. [ vbe ] → jira / Mood[+indic] + T[+pres] Predindiv + Phave

b. [ vbe ] → ba / Mood[+indic] + T[+pres] Predindiv

c. [ vbe ] → dwiire / Mood[+indic] Pred + Phave

d. [ vbe ] → li / Mood[+indic] Pred

The above VI rules are crucially ordered. As such, the highest rules apply in the most specific
contexts, e.g., -jira surfaces exclusively in the present indicative when Predindiv and Phave are
present (44a), while -ba surfaces in the present indicative when only Predindiv is present (44b).
The rules in (44c) and (44d) are underspecified for both Tense and Pred to reflect the more varied
distribution of -dwiire/-li. That is, unlike -jira (44a) and -ba (44b), which are restricted to the
present indicative and obligatorily yield a permanent interpretation, -dwiire and -li may appear
in either the past or present indicative, and are only restricted to a temporary interpretation in

20
I acknowledge that the proposed analysis will ultimately need to explain how Aspect fits into the defined spans.
Given that the relevant features for the data in this paper are Mood and Tense, I will not attempt to account for
Aspect here.
25

the present tense; -dwiire and -li may co-occur with either Predstage or Predindiv in the past tense.21
By ordering (44a) and (44b) before (44c) and (44d), we ensure that -dwiire and -li appear in both
the present tense, where they must co-occur with Predstage and yield a temporary interpretation,
and also the past tense, where they are the only forms available.
Turning now to the effect of Mood, we need to add a default form of vbe that appears just in
case Mood is not indicative (45a), as well as a default form of Phave that applies in cases where
Phave is not a proper part of the insertion span (45b).22

(45) a. [ vbe ] → ba
b. [ Phave ] → na

The default vbe rule in (45a) functions as the elsewhere condition; if no other rules apply, realize
-ba. It is therefore ordered after (44d). This ensures that vbe is realized as -ba in the subjunctive
and infinitival moods, again regardless of what is present below it (e.g., Pred/Phave). Note that
this default form is crucially distinct from the homophonous form, -ba (44b), that strictly surfaces
in the present indicative with Predindiv. I will return to the discussion of this fifth, default form
of vbe in section 5.
Now, unlike the previous analyses presented in section 4, the spanning analysis presented
here offers a built-in solution for the fact that Mood/Tense can not only trigger allomorphy
of vbe, but also be discretely exponed. As we have seen, the features associated with Mood/
Tense can determine the realization of be/have; both past tense and irrealis mood serve to
(morphologically) neutralize the temporary-permanent contrast for be/have. Importantly, the
conditioning effects of Mood and Tense each correlate with overt exponenence. For example,
the conditioning effects of past tense correlate with an overt realization of the Tense head, e.g.,
past tense -li in 25). Similarly, the conditioning/blocking effects of the subjunctive and infinitival
moods correlate with an overt exponent of the Mood head, e.g., subjunctive FV -e and infinitival
prefix (o)ku- (29)-(30). Crucially however, overt exponents of Mood and Tense cannot co-occur

21
It is outside the scope of this paper to explain the collapse of the temporary-permanent contrast in the past tense. It is
an interesting question as to why the permanent forms of be/have are limited to the present tense. For the purposes
of this paper, I will set aside any further exploration of this topic.
22
It is theoretically possible that the default form of vbe realizes the span [vbe-Pred], as opposed to just [vbe]. I have
chosen the latter conceptualization for two reasons. First, given that irrealis mood yields the copular possessive, be
+ na, with have, it is not clear whether Pred is realized with vbe or is simply unpronounced. Second, if we attribute
the “blocking” effect of irrealis mood to a shortening of the insertion span targeting vbe — which would allow Phave
to be realized in isolation under the rule in (45b) — it is favorable to treat this shortened span as just [vbe], as this
would ensure that the blocking effect applies to both be and have, rather than just the latter.
26

with each other; we find no contexts in which be/have forms appear with both overt tense
morphology and overt mood morphology.23
Under the current analysis, the fact that Mood/Tense features trigger allomorphy of vbe, and
that Mood/Tense morphology may co-occur with an overt realization of vbe follows if the heads
Mood and Tense form a distinct conditioning span that is sometimes targeted for insertion. The
idea here is that the specific combination of features present in the span [Mood-T] conditions the
realization of the adjacent span [vbe-Pred-Phave], and, in some cases, receives its own exponent.
Assuming that a span of heads may trigger allomorphy of another adjacent span (Merchant 2015;
Hanink 2018), this analysis ensures that both Mood and Tense are sufficiently local to condition
the realization of vbe regardless of ordering, and that they are jointly realized (to the exclusion of
vbe). As such, this analysis resolves the aforementioned issues with both countercyclic blocking
(section 4.1.1) and locality (section 4.1.2). The following Table 2 illustrates the possible be/
have surface forms according to the features present in two proposed spans.

Conditioning Span Insertion Span Surface Realization


Mood Tense vbe Pred Phave [Mood-T] [vbe-Pred-Phave]
be +indic +pres – indiv × – -ba
+indic +pres – stage × – -li
+indic +past – stage/indiv × -li -li
+sbjv – – stage/indiv × -e -ba
+inf – – stage/indiv × (o)ku -ba
have +indic +pres – indiv ✓ – -jira
+indic +pres – stage ✓ – -dwiire
+indic +past – stage/indiv ✓ -li -dwiire
+sbjv – – stage/indiv ✓ -e -ba + na
+inf – – stage/indiv ✓ (o)ku -ba + na

Table 2: Surface forms of be/have under the spans analysis.

In the present indicative, Mashi encodes a temporary-permanent contrast in both its be-verbs
and have-verbs. With respect to the current spans analysis, we can characterize this according to
the features associated with the relevant spans; when the conditioning span [Mood-T] contains
the features [+indic, +pres], the temporary-permanent contrast is morphologically overt.
When Phave is absent, the Pred head that is merged determines which be-verb, -li or -ba, is realized

23
In this context, I assume overt mood morphology to be any transparent realization of Mood features. As such, I
assume that the feature [+indic] is generally null-marked in Mashi (as opposed to being realized by the final vowel
-a (Nurse 2008).
27

(46)-(47). Importantly, the conditioning span [Mood-T] is not targeted for insertion, as the
present indicative is null marked in Mashi.

(46) Mood

Mood T
+indic
T v
+pres
v be Pred

Predstage ...

-li

(47) Mood

Mood T
+indic
T v
+pres
v be Pred

Predindiv ...

-ba

When Phave is present, the Pred head instead determines which have-verb, -dwiire or -jira, is
realized (48)–(49). As above, the span [Mood-T] is not targeted for insertion.

(48) Mood

Mood T
+indic
T v
+pres
v be Pred

Predstage Phave

Phave ...

-dwiire
28

(49) Mood

Mood T
+indic
T v
+pres
v be Pred

Predindiv Phave

Phave ...

-jira
In contrast, when the conditioning span contains the features [+indic, +past], the temporary-
permanent contrast is morphologically neutralized. In this context, be obligatorily surfaces as the
temporary be-verb form -li ([-ri]) (50), and have obligatorily surfaces as the temporary have-verb
form -dwiire (51), both regardless of which Pred head is present in the lower insertion span. In
both cases, this neutralizing effect correlates with the overt realization of the span [Mood-T] as
the past tense marker -li.
(50) Mood

Mood T
+indic
T v
+past
v be Pred

Predstage ...

-li -ri

(51) Mood

Mood T
+indic
T v
+past
v be Pred

Predstage Phave

Phave ...

-li -dwiire
29

Finally, if the conditioning span [Mood-T] exhibits either the feature [+sbjv] or [+inf], vbe is realized
as the be-verb form -ba, regardless of what is below it (e.g., Pred/Phave). This results in the morphological
neutralization of the temporary-permanent contrast, and also the blocking of “incorporation”; in (52)-
(55), the insertion span is restricted to a single node, vbe, leaving Pred unpronounced, and Phave — if
present — to be realized as the default na (see 45b). Like past tense, the conditioning effects of irrealis
mood correlate with overt exponents of [Mood-T], e.g., infinitival (o)ku- and subjunctive -e.24

(52) Mood

Mood T
+sbjv
T v

v be Pred

Pred ...

-e -ba

(53) Mood

Mood T
+inf
T v

v be Pred

Pred ...

(o)ku- -ba

(54) Mood

Mood T
+sbjv
T v

v be Pred

Pred Phave

Phave ...

-e -ba na

24
For our purposes it is irrelevant that the linear order of vbe and the exponent of [Mood-T] differs in the subjunctive
(expressed by final vowel -e) and infinitive (expressed by prefix oku-). What matters here is that [Mood-T] conditions
the realization of vbe and may be discretely exponed.
30

(55) Mood

Mood T
+inf
T v

v be Pred

Pred Phave

Phave ...

(o)ku- -ba na

To summarize this section, I presented an analysis of be/have allomorphy in Mashi that invokes
the Spanning Insertion Hypothesis as presented in Merchant (2015). Specifically, I argued that
the conditioning effects of Mood and Tense on the realization of be/have are best accounted
for by stipulating the presence of two distinct spans: a higher span [Mood-T], and lower span,
[vbe-Pred-Phave]. This analysis was shown to capture two distinct observations: that Mood/Tense
features serve to jointly condition the realization of vbe, and that overt Mood/Tense morphology
co-occurs with surface realizations of vbe. On the basis of this analysis I further defined an ordered
set of VI rules that capture the full surface pattern of vbe allomorphy.

5 Discussion
In this section, I would like to address two related consequences of the proposed VI rules (repeated
below in 56), namely the existence of two homophonous vbe allomorphs, -ba1 and -ba2, and the
‘default’ status of the temporary forms -li and -dwiire.

(56) a. [ vbe ] → jira / Mood[+indic] + T[+pres] Predindiv + Phave


b. [ vbe ] → ba / Mood[+indic] + T[+pres] Predindiv

c. [ vbe ] → dwiire / Mood[+indic] Pred + Phave

d. [ vbe ] → li / Mood[+indic] Pred


e. [ vbe ] → ba

f. [ Phave ] → na

Perhaps the most significant consequence of the proposed VI rules is that they result in two
homophonous allomorphs (-ba) of vbe — one that applies strictly in the present indicative (56b)
and one that that serves as the elsewhere form (56e). While undesirable, this turns out to be
a necessary stipulation in the analysis. The problem is that in the indicative, -ba must surface
exclusively in the presence of T[+pres] and Predindiv, but in the subjunctive/infinitive, -ba surfaces
31

regardless of what other material is present (e.g., Pred/Phave). By defining two distinct -ba
allomorphs, we ensure that -ba — but not -li — surfaces with Predindiv in the present indicative
(56b) and also that no other allomorph surfaces in the presence of Mood[-indic] (56e).25 Crucially,
these two environments cannot be captured by one synchronic rule; to keep the temporary-
permanent contrast between -li/-ba in the present indicative and ensure that only -ba appears in
Mood[-indic] contexts, we need to define two homophonous allomorphs, -ba1 and -ba2.26
That being said, we do find reasonable support for the existence of two homophonous vbe
allomorphs in Mashi. First, we find analysis internal support in the fact that vbe is sometimes
realized in isolation rather than as part of a larger span. Consider the “blocking” effects of the
subjunctive and infinitival moods; while -ba strictly realizes the span [vbe-Predindiv] in the present
indicative (56b), irrealis mood limits the size of the insertion span that realizes vbe, resulting in
vbe being realized as -ba (56e) and Phave — if present — being realized as na (56f). Given that
this blocking effect coincides with the morphological neutralization of the temporary-permanent
contrast (i.e., only -ba is available), it is not unreasonable to suspect that -ba is the default
realization of vbe in isolation; Pred could simply be unpronounced. As such, one -ba allomorph
realizes the span [vbe-Predindiv] (56b), while the other realizes the span [vbe] (56e).
Assuming Mashi does exhibit three distinct copular verb forms, viz. -li, -ba1 and -ba2, we
would expect to find other languages that make a similar tripartite distinction. This is indeed
what is observed; such a system is reported in closely related Kifuliiru (Bantu, JD63), which

25
The distribution of -ba is another motivating factor in attributing the temporary-permanent contrast to distinct Pred
heads (Myler 2016; 2018) rather than distinct vbe heads. If -li/-ba coincide with distinct vbe heads, it would be difficult
to explain why -ba appears in irrealis environments regardless of interpretation; if v-ba is semantically specified as an
individual-level copula, it should not appear in contexts where a stage-level (temporary) reading is available. If the
temporary-permanent contrast instead stems from vbe allomorphy, this distribution may be explained as the result of
there being a default form of vbe (56e).
26
An anonymous reviewer rightfully points out that it is possible to capture the allomorphy facts in Mashi with an
alternative analysis that does not posit two homophonous -ba allomorphs. In this alternative analysis, there is a
single -ba allomorph in (ic), and it is the elsewhere form. Additionally, the past tense form of the copula [-ri], which
I assume to be a phonological variant of -li (see footnote 12), is assumed to be a separate Vocabulary Item (ib). This
analysis offers a straightforward account of why the temporary-permanent contrast is neutralized in the past tense;
Tpast is included in the insertion span for -ri.

(i) a. [ vbe ] → li / T [pres] Predstage

b. [ vbe ] → ri / T [past] Pred

c. [ vbe ] → ba / Pred

d. [ vbe ] → dwiire / T [realis] Pred + Phave

e. [ vbe ] → jira / T [pres] Predindiv + Phave

f. [ Phave ] → na

Though this analysis nicely accounts for the Mashi data, I ultimately adopt the VI rules in (56) since they better
capture the parallelism between the be and have forms, which is one of the central observations in this paper. That
said, these analyses do not necessarily need to be distinguished here. Importantly, both analyses capture the core
theoretical proposal made in this paper: be and have are compositionally related.
32

exhibits three copular forms in predicative-environments, -li, -ba, and -tula (Finholt 2023). While
the interpretative distinctions between these copulas differ from those in Mashi (see Finholt 2023
for discussion), the fact that a closely related language exhibits a three-way copular system at all
provides significant typological evidence to support the analysis that there are two distinct -ba
allomorphs in Mashi. Just like Kifuliiru, Mashi has three copular be forms: -li, -ba1, and -ba2. The
fact that there are two -ba allomorphs in Mashi may simply be attributed to a case of accidental
homophony stemming from an unknown series of diachronic changes. I will leave any further
explanation of such diachronic processes to future work. What is important here is that there is
typological evidence to support the existence of three distinct copular be forms in Mashi.
The second consequence of the proposed VI rules in (44) follows directly from the existence
of two -ba allomorphs: if one -ba allomorph arises strictly in the present indicative (56b) and
another serves as the elsewhere form (56e), then the temporary forms -dwiire (56c) and -li
(56d) become de-facto default forms of have and be respectively, as they occur in much less
specific contexts (e.g., [+/–pres]). The exact reason why the temporary (“stage-level”) forms
behave as default forms for be/have is unclear, and I will not attempt to resolve this issue
here. That said, one possible analysis would be to assume that the temporary forms we see in
the present indicative — where the temporary-permanent contrast is morphologically overt —
realize unmarked forms of Pred. In this sense, the temporary-permanent distinction would be
characterized by a distinction between unmarked Pred and marked Predindiv. This would allow
for the temporary forms to surface in contexts where Pred is underspecified, e.g., in the presence
of past tense (56c)–(56d). Of course, such an analysis would not explain why the temporary
forms may appear with both temporary and permanent interpretations in the past tense, e.g.,
-li in (25)–(26). If the temporary forms realize unmarked Pred, we would expect them to be
unavailable with permanent interpretations, which arise from the Predindiv head. I will leave any
further discussion and/or account of such issues to future work.

6 Conclusion
In this paper, I present novel data on copular predication and possession in the Bantu language
Mashi (JD53; Maho 2009). The central observation is that Mashi exhibits an interpretive overlap
in its be-verbs and have-verbs; just as the two be-verbs, -li/-ba, lexicalize a distinction between
“temporary” and “permanent” states, the two have-verbs, -dwiire/-jira lexicalize a distinction
between “temporary” and “permanent” possession.27 From these descriptive observations, this

27
What this says about languages that make a stage-individual distinction in their be- but not have-verbs (e.g., Spanish)
is unclear. One possible interpretation of the analysis presented here is that such languages do extend the stage-in-
dividual contrast to have, but this contrast is morphologically covert. I will not comment any further on this issue.
What is important here is that the stage-individual (temporary-permanent) contrast must extend to have construc-
tions in at least some languages.
33

paper makes a theoretical point. Following a long line of previous research, I argue that the
interpretive overlap between the be/have-verbs provides novel evidence of a compositional
relationship between abstract be and have (Freeze 1992; Kayne 1993; Ritter & Rosen 1997;
Harley 1998; Beavers et al. 2009; Myler 2016; 2018 a.o.). I specifically show that Mashi be/
have allomorphy — which I characterize as vbe allomorphy — provides strong support for an
incorporation approach to decompositional have whereby have is derived via the incorporation
of a locative element Phave by a verbal head vbe (Freeze 1992; Kayne 1993; Ritter & Rosen 1997;
Harley 1998 a.o.). Under certain morphological conditions, e.g., in the presence of irrealis mood,
both lexical have-verbs in Mashi are unavailable; in this environment have is obligatorily realized
as the copular possessive (i.e., with-possessive Freeze 1992) construction be + comitative
na. Given that the combination of be and the comitative particle na is a common strategy for
encoding possession in possession in Bantu languages in general (Gibson 2012; Halpert & Diercks
2014; Myler 2016 a.o.), I treat na as an overt exponent of the locative element Phave described
in incorporation accounts (Freeze 1992; Kayne 1993; Ritter & Rosen 1997; Harley 1998 a.o.).
To account for the interpretive overlap between be/have and fully capture vbe allomorphy,
I invoke a spanning insertion analysis modeled after that presented in Merchant (2015). This
approach allows for a simpler analysis of complex allomorphic patterns by defining the target for
Vocabulary Insertion as a contiguous series, or span, of heads, and defining locality conditions
according to the adjacency of said spans. I propose two distinct spans of heads that are relevant
to the realization of vbe in Mashi. One span, namely [vbe-Pred-Phave], serves to directly realize vbe.
This span captures the general contrast between be/have, which is distinguished by the presence/
absence of Phave, and also the broader temporary-permanent distinction, which I attribute to a
distinction between two Pred heads, Predstage/indiv. The second span, [Mood-T] is not directly
involved in the composition of be/have, instead serving to condition its realization. By defining
two distinct spans, [Mood-T] and [vbe-Pred-Phave] we account for both the conditioning effects
of the features associated with Mood and Tense, and also the fact that Mood/Tense morphology
may co-occur with overt exponents of vbe.
In summary, this paper presents a novel decompositional approach to abstract have that
extends the stage-individual (or temporary-permanent) contrast (Carlson 1977; Kratzer 1995;
Diesing 1990; 1992; Arche 2006 a.o.) to have constructions, relates be and have via incorporation
(Freeze 1992; Kayne 1993; Ritter & Rosen 1997; Harley 1998, a.o.), and defines the domain for
insertion and allomorphic conditioning according to spans of heads (Merchant 2015).
34

Abbreviations
agr : agreement
aug : augment
com : comitative
comp : complementizer
cop : copula
dem : demonstrative
indic : indicative mood
inf : infinitival mood
hab : habitual
loc : locative
mod : modal
perm : permanent
poss : possessive pronoun
pres : present tense
sbjv : subjunctive mood
sg : singular
sm : subject marker
temp : temporary

Ethics and consent


The collection of human subject data was approved by the University of Kansas IRB Board. Data
was collected as part of two separate IRB-approved projects: 00148259 - Bantu complementizers,
and 00146220 - Documentation of African Languages in Kansas.

Funding information
This work has been partially funded by NSF grant 2140000-908-1003445, PI: John Gluckman,
University of Kansas.

Acknowledgements
I would especially like to thank my advisor John Gluckman, as well as Philip Duncan, Andrew
McKenzie, and the other members of the Research in Field and Formal Linguistics seminar at the
University of Kansas for their invaluable feedback on this project. I would also like to thank the
anonymous Glossa reviewers and the audiences at NELS 52 at Rutgers University, and the 2022 LSA
35

Annual Meeting in Washington D.C. for their comments and questions. Considerable thanks as well
to Emily Hanink, Jens Hopperdietzel, and Andrew Koontz-Garboden for inviting me to present this
work to the Syn-Sem Lab at the University of Manchester. Finally, I would like to express my sincerest
thanks to Mitima Burhungane for his time and considerable effort in teaching me about Mashi.

Competing interests
The author has no competing interests to declare.

References
Allen, Margaret. 1978. Morphological investigations: Storrs, CT: University of Connecticut, Storrs
dissertation.
Arche, María J. 2006. Individuals in time: Tense, aspect and the individual/stage distinction.
Amsterdam & Philadelphia: John Benjamins Publishing. DOI: https://doi.org/10.1075/la.94
Beavers, John & Ponvert, Elias & Wechsler, Stephen. 2009. Possession of a controlled substantive:
Light ‘have’ and other verbs of possession. In Friedman, Tova & Ito, Satoshi (eds.), Proceedings of
Semantics and Linguistic Theory 18. 108–125. Amherst, MA: Amherst: University of Massachussetts.
DOI: https://doi.org/10.3765/salt.v18i0.2502
Bobaljik, Jonathan David. 2012. Universals in comparative morphology. Cambridge, MA: MIT
Press. DOI: https://doi.org/10.7551/mitpress/9069.001.0001
Bochnak, Ryan & Grinsell, Timothy & Yu, Alan. 2011. Copula agreement and the stagelevel/
individual-level distinction in Washo. In Stein, Mark J. (ed.), Workshop on Structure and
Constituency in Languages of the Americas (WSCLA). vol 16. 1–10.
Carlson, Greg. 1977. Reference to Kinds: University of Massachussetts, Amherst dissertation.
Deo, Ashwini & Sanchez-Alonso, Sara & Pinango, Maria. 2017. Alternative circumstances of
evaluation and the ser/estar distinction in Spanish. Ms. Yale University. http://ling.auf.net/
lingbuzz/003543.
Diesing, Molly. 1990. The syntactic roots of semantic partition: University of Massachusetts,
Amherst dissertation.
Diesing, Molly. 1992. Indefinites. MIT press.
Embick, David. 2010. Localism versus globalism in morphology and phonology. MIT Press. DOI:
https://doi.org/10.7551/mitpress/9780262014229.001.0001
Embick, David. 2012. Contextual conditions on stem alternations: Illustrations from the Spanish
conjugation. In Irene Franco, Sara Lusini & Saab, Andrés L. (eds.), Romance languages and
linguistic theory 2010: Selected paper from ‘Going Romance’ Leiden 2010, 21–40. Amsterdam: John
Benjamins. DOI: https://doi.org/10.1075/rllt.4.02emb
Finholt, Aron. 2023. Three-way copular contrast in Kifuliiru. 97th Annual Meeting of the
Linguistic Society of America, Denver, CO.
Freeze, Ray. 1992. Existentials and other locatives. Language 68. 553–396. DOI: https://doi.
org/10.2307/415794
36

Gibson, Hannah. 2012. Auxiliary placement in Rangi: a dynamic syntax perspective: SOAS, University
of London dissertation.
Gibson, Hannah & Guérois, Rozenn & Marten, Lutz. 2019. Variation in Bantu copula constructions.
In Fábregas, Antonio & Marín, Rafael (eds.), The grammar of copulas across languages, 213–242.
Oxford: Oxford University Press. DOI: https://doi.org/10.1093/oso/9780198829850.003.0011
Gluckman, John & Born, Stephanie & Finholt, Aron & Foster, Jack & Whittington, Lucy. 2022.
Non-verbal predication in Kihavu. 53rd Annual Conference on African Linguistics, University of
California, San Diego.
Green, Lisa. 2000. Aspectual be-type constructions and coercion in African American English.
Natural Language Semantics 8(1). 1–25. DOI: https://doi.org/10.1023/A:1008392112529
Greenberg, Yael. 1998. An overt syntactic marker for genericity in Hebrew. In Rothstein,
Susan (ed.), Events and grammar, 125143. Kluwer Academic Publishers. DOI: https://doi.
org/10.1007/978-94-011-3969-4_6
Halpert, Claire & Diercks, Michael. 2014. More on have and need. In Kandybowicz, Jason &
Major, Travis & Torrence, Harold & Duncan, Philip T. (eds.), African linguistics on the prairie:
selected papers from the 45th Annual Conference on African Linguistics, 289–305. Berlin: Language
Science Press. DOI: https://doi.org/10.5281/zenodo.1251742
Hanink, Emily A. 2018. Super light-headed relatives, missing prepositions, and spanconditioned
allomorphy in German. The Journal of Comparative Germanic Linguistics 21(2). 247–290. DOI:
https://doi.org/10.1007/s10828-018-9096-6
Harley, Heidi. 1998. You’re having me on: Aspects of have. In Guéron, J. & Zribi-Hertz, A. (eds.),
La grammaire de la possession, 195–226. Paris: Université Paris X – Nanterre. Higgins, Francis
Roger. 1979. The Pseudo-Cleft Construction in English. Routledge.
Hulstaert, Gustaaf. 1965. Grammaire du Lomongo. deuxième partie: Morphologie. Tervuren, BEL:
Musée royal de l’Afrique centrale.
Kayne, Richard S. 1993. Toward a modular theory of auxiliary selection. Studia Linguistica 47(1).
3–31. DOI: https://doi.org/10.1111/j.1467-9582.1993.tb00837.x
Kratzer, Angelika. 1995. Individual-level predicates. In Carlson, Greg & Pelletier, Francis Jeffry
(eds.), The generic book, 125–175. Stanford, CA: CSLI Publications.
Levinson, Lisa. 2011. Possessive WITH in Germanic: HAVE and the role of P. Syntax 14(4).
355–393. DOI: https://doi.org/10.1111/j.1467-9612.2011.00159.x
Lewis, M. Paul (ed.). 2009. Ethnologue: Languages of the world. Dallas, TX, USA: SIL International
sixteenth edn.
Maddieson, Ian & Sands, Bonny. 2019. The sounds of the Bantu languages. In The Bantu languages,
chap. 3. Routledge. DOI: https://doi.org/10.4324/9781315755946-3
Maho, Jouni. 2009. NUGL Online: the web version of the new updated Guthrie List, a referential
classification of the Bantu languages. Gothenburg: Department of Oriental and African Languages,
University of Gothenburg.
Maienborn, Claudia. 2005. A discourse-based account of Spanish ser/estar. Linguistics 43(1).
155–180. DOI: https://doi.org/10.1515/ling.2005.43.1.155
37

Matthewson, Lisa. 2004. On the methodology of semantic fieldwork. International Journal of


American Linguistics 70(4). 369–415. DOI: https://doi.org/10.1086/429207
Meeussen, Achille Emile. 1967. Bantu grammatical reconstructions. Africana linguistica 3(1).
79–121. DOI: https://doi.org/10.3406/aflin.1967.873
Merchant, Jason. 2015. How much context is enough? Two cases of span-conditioned stem
allomorphy. Linguistic Inquiry 46(2). 273–303. DOI: https://doi.org/10.1162/LING_a_00182
Möhlig, Wilhelm & Kavari, Jekura Uaurika. 2008. Reference grammar of Herero (Otjiherero). Köln:
Rüdiger Köppe Verlag.
Myler, Neil. 2016. Building and interpreting possession sentences. Cambridge, MA: MIT Press. DOI:
https://doi.org/10.7551/mitpress/9780262034913.001.0001
Myler, Neil. 2018. Complex copula systems as suppletive allomorphy. Glossa: a journal of general
linguistics 3(1). 1–43. DOI: https://doi.org/10.5334/gjgl.214
Nicolle, Steve. 2013. A grammar of Digo: a Bantu language of Kenya and Tanzania. Dallas, TX: SIL
International.
Nurse, Derek. 2008. Tense and Aspect in Bantu. Oxford University Press.
Polack-Bynon, Louise. 1975. A Shi grammar. Tervuren: Musée Royal de l’Afrique Centrale,
Annales – Sciences Humaines, No. 86.
Ramchand, Gillian Catriona. 1997. Aspect and predication: The semantics of argument structure.
Oxford University Press.
Ramsey, Marathon Montrose. 1894. A textbook of modern Spanish. New York: Spaulding.
Ritter, Elizabeth & Rosen, Sara Thomas. 1997. The Function of have. Lingua 101. 295–321. DOI:
https://doi.org/10.1016/S0024-3841(96)00024-1
Schneider-Zioga, Patricia. 2018. Non-verbal predication in Bantu. To appear in The Oxford Guide
to the Bantu Languages. http://ling.auf.net/lingbuzz/004350.
Schneider-Zioga, Patricia. 2021. A’-agreement is not only morphological: the case of focus
agreement. Move and Agree Forum, University of Britsih Columbia and McGill University.
Siegel, Dorothy. 1978. The Adjacency Constraint and the theory of morphology. In Stein, Mark
J. (ed.), Proceedings of NELS 8. 189–197. Amherst, MA: University of Massachusetts, Graduate
Linguistic Society. https://scholarworks.umass.edu/nels/vol8/iss1/18.
Svenonius, Peter. 2012. Spanning Ms., University of Tromsø. https://ling.auf.net/lingbuzz/001501.
Svenonius, Peter. 2016. Spans and words. In Siddiqi, Daniel & Harley, Heidi (eds.), Morphological
metatheory, 201–222. Amsterdam: John Benjamins. DOI: http://doi.org/10.1075/la.229.07sve.
Wald, Benji. 1973. Syntactic change in the Lake languages of Northeast Bantu. Studies in African
Linguistics, Vol 4, No 3 (1973).
Williams, Edwin. 2003. Representation theory. Cambridge, MA: MIT Press.
Wood, Jim. 2015. Icelandic morphosyntax and argument structure, vol. 90 (Studies in Natural
Language and Linguistic Theory). Dordrecht: Springer.

You might also like