Latin verbs appear to both obey and disobey the Mirror Generalization (
A great deal of crosslinguistic evidence (see especially
(1) | |
Morphological derivations must directly reflect syntactic derivations (and vice versa). |
The concept in (1) is standardly interpreted as making available the following structural diagnostic:
(2) | If a root and some functional heads in its extended projection ( |
This diagnostic, which makes it possible to infer the composition of portions of clause structure from that of complex words, has been put to use very fruitfully indeed in contemporary work on morphosyntax.
Although (1) is often referred to as the Mirror Principle, it is now typically held—correctly, in our view—that it is not in fact a principle, but a generalization to be derived from more fundamental principles (
Because the MG has played such a prominent role in enabling us to understand syntax/morphology interactions, any apparent counterexamples to it need to be scrutinized to determine whether the MG can survive them—and, if so, what their etiology is. To that end, this paper will investigate certain apparent counterexamples to the MG in Latin, to determine what they reveal about the MG and the relation between syntactic and morphological structure more broadly.
The paper proceeds as follows. Section 2 presents the main empirical puzzle: Latin verbs seem to both obey and disobey the MG, but in different inflectional forms. Section 3 lays out the theoretical desiderata for a principled analysis of this situation. It then develops an analysis (the
Section 4 shows that the analysis, though developed for finite forms, extends straightforwardly to nonfinite forms. In particular, the view that Latin makes use of
Suppose, following fairly standard assumptions, that the extended projection of V is, in part, the following (argument phrases omitted; cf.
(3)
In Latin, the heads in the extended projection of V shown in (3) do seem to be packaged into a single morphophonological word. Therefore, the MG, in conjunction with the clause structure in (3), leads us to expect that the Latin verb word should display the following morpheme order:
(4) | Root > |
(> = ‘precedes’) |
If this were so, Latin verbs would transparently obey the MG. But in fact, they seem to both obey and disobey the MG, albeit in different parts of their paradigm, as will be shown in Sections 2.1 and 2.2. This is peculiar, and naturally gives rise to the following question:
(5) | Given the crosslinguistic robustness of the MG, how can it be that Latin verbs seem to both obey and disobey it (in different inflectional forms)? |
An analysis that answers this question will be developed in Section 3. But first, we will give concrete examples of Latin verb forms that appear to obey the MG (henceforth
In the pluperfect, the perfective Aspect morpheme (-
(6)
a.
laud-ā-
praise-
‘I had praised’
b.
laud-ā-
praise-
‘you had praised’
c.
laud-ā-
praise-
‘he/she/it had praised’
d.
laud-ā-
praise-
‘we had praised’
e.
laud-ā-
praise-
‘you guys had praised’
f.
laud-ā-
praise-
‘they had praised’
Likewise, in the future perfect, the same perfective Aspect morpheme (-
(7)
a.
laud-ā-
praise-
‘I’ll have praised’
b.
laud-ā-
praise-
‘you’ll have praised’
c.
laud-ā-
praise-
‘he/she/it will have praised’
d.
laud-ā-
praise-
‘we’ll have praised’
e.
laud-ā-
praise-
‘you guys will have praised’
f.
laud-ā-
praise-
‘they’ll have praised’
The forms in (6) and (7) follow the MG-compliant template in (8):
(8) | Root > Theme Vowel > |
In many passive forms, on the other hand, the exponents of Mood, Tense, and sometimes Aspect occur closer to the root than does the exponent of passive Voice, in seeming contravention of the MG. (The unexpectedly external position of the passive Voice morpheme is discussed in
As an example, consider the present passive subjunctive forms of
(9)
a.
laud-
praise-
‘(that) I may be praised’
b.
laud-
praise-
‘(that) you may be praised’
c.
laud-
praise-
‘(that) he/she/it may be praised’
d.
laud-
praise-
‘(that) we may be praised’
e.
laud-
praise-
‘(that) you may be praised’
f.
laud-
praise-
‘(that) they may be praised’
Nor is anti-mirroring in passive forms restricted to the paradigm in (9). The exponent of passive Voice also occurs farther from the root than do the following morphemes, among others (see the Appendix for the full paradigm):
(10)
a.
-
laud-ā-
praise-
‘he/she/it will be praised’
b.
-
laud-ā-
praise-
‘he/she/it was being praised’
c.
-
laud-ā-
praise-
‘(that) he/she/it might be praised’
All the forms examined in this subsection follow the template in (11), which appears to violate the MG. (The notation
(11) | Root (> Theme Vowel) > |
The question, then—as mentioned—is how it can be that some verb forms seem to violate the MG (and in a language in which other verb forms obey it, no less).
An optimal, maximally principled solution to the mirroring/anti-mirroring puzzle just presented should satisfy the following theoretical desiderata:
(12) | a. | |
b. | ||
c. |
This challenge can be met. The present section will develop an analysis of mirroring and anti-mirroring in Latin that satisfies all three desiderata in (12).
On the analysis developed below, mirroring and anti-mirroring verb forms have the same derivation—which involves one step of XP-movement (
Consider the following mirroring form:
(13)
laud-ā-
praise-
‘he/she/it had praised’
On our analysis, (13) has the following derivation:
(14)
Five aspects of this derivation deserve comment.
First, the Voice head in finite clauses bears two
Secondly, Asp undergoes head movement to T (and is linearized to the left of T, just as the V-
Third,
Fourth, as suggested to us by Nico Baier, T is the locus of both tense and mood features—i.e., T(ense) and Mood are not projected as separate heads in Latin. This view has two advantages over an alternative on which T and Mood are separate heads in Latin. First, it makes it easy to understand why there are no clear cases of tense and mood being exponed separately in this language (see the Appendix). Secondly, if T and Mood were separate in Latin, Mood would presumably be higher, and we would therefore have to stipulate that leftward adverbials standardly analyzed as TP-adjuncts in English were MoodP-adjuncts in Latin (since, if they were TP-adjuncts, they would interrupt the verb word, which in fact is impossible). This stipulation is avoided on the T-only view adopted here: on this view, TP-adjuncts in English correspond to TP-adjuncts in Latin, and the latter do not interrupt the verb word.
Fifth, the external argument (EA) originates in [Spec,
On this analysis, then, the derivation of every synthetic finite verb form in Latin involves one step of XP-movement (
As illustrated in (14), the XP/X0 Analysis accounts straightforwardly for clearly mirroring verb forms. In the form derived in (14), for example—
Consider now the following anti-mirroring form:
(15)
laud-
praise-
‘(that) he/she/it may be praised’
On the XP/X0 Analysis, this anti-mirroring form has exactly the same syntactic derivation as the mirroring form examined in Section 3.1:
(16)
In (16), as in (14), Voice probes to value its
The anti-mirroring observed in
On the XP/X0 Analysis, then, all synthetic finite verbs in Latin, whether they appear to obey the MG or not, share a single, simple derivation, involving one step of head movement (Asp to T) and one step of phrasal movement (
On this analysis, the MG is in fact valid for Latin—contrary to initial appearances—but it must crucially be understood as a generalization about structures formed by operations on heads (e.g., head movement).
On the XP/X0 Analysis, the Latin verb word (e.g.,
If the XP/X0 Analysis is on the right track for Latin, then, it strongly supports an
See the Appendix for the full paradigm of all the (synthetic) finite, nonfinite, and deverbal forms of the sample verb
Before we extend the XP/X0 Analysis to nonfinite forms, it will be worth noting how it interacts with existing hypotheses about the locality domain for allomorph selection, given the allomorphy patterns observed within Latin verbs. The basic observation to be accounted for is that, within the verb word, a head can often trigger allomorphy on another head in a way that seems nonlocal in terms of hierarchical structure, given the structures proposed here.
For example, consider the verbs
(17)
a.
mān-
stay-
‘he/she/it had stayed’
b.
mon-
warn-
‘he/she/it had warned’
c.
see-
‘he/she/it had seen’
In each of (17a–c), a particular V idiosyncratically triggers the realization of AspPFV in a different way: as -
The XP/X0 Analysis is therefore incompatible with the hypothesis that, for two elements
(18) | Let |
|
a. | For all |
|
b. | For any |
Ostrove argues that, in Irish, a single exponent can sometimes realize the sequence of terminals Mood–√
Ostrove, then, invokes stretches primarily as targets for Vocabulary Insertion. However, his definition of stretch is modeled on Merchant’s (
(19) | |
Allomorphy {is conditioned only/can be conditioned} by an adjacent stretch. |
(Which version is more adequate is an open question.) If (19) is on the right track, the allomorphy patterns in (17) become unsurprising: in all three forms, the form of AspPFV is conditioned by the linearly adjacent stretch V–
Naturally, the proposals in (18)–(19) deserve much more detailed testing, and they are discussed further in Section 4.3. What is important here is that, as a reviewer notes, the XP/X0 Analysis is compatible with (a minimal extension of) one independently motivated existing approach to Vocabulary Insertion: Ostrove’s (
Section 3 developed an analysis of the derivation of Latin finite verb forms that relies crucially on
This result in itself furnishes a strong argument that Latin makes use of
This section argues that the answer is yes—and specifically, that a phenomenon of exactly this type emerges when we turn our attention from the finite to the nonfinite verb forms. The discussion that follows will focus on two nonfinite forms in particular: the Perfect Passive Participle and the Future Active Participle (cf.
The Future Active Participle (e.g.,
Two examples follow of the Perfect Passive Participle in a larger context. (Example (20) is adapted from
(20)
Livy,
…mūrus
wall.
et
and
porta
gate.
dē
from
caelō
sky.
touch-
era-nt.
be.
‘…the wall and the gate had been struck by lightning.’ (‘…the wall and the gate were having-been-touched from the sky.’)
(21)
Anna
Anna
praise-
es-t.
be-3
[C]
‘Anna {was/has been} praised.’ (‘Anna is having-been-praised.’)
On our analysis, the derivation of (21) is, in part, the following. (On the syntax of the auxiliary, see fn. 19. On the apparent
(22)
On this analysis, the participial morpheme -
The derivation in (22) yields the correct morpheme order, but no argument has yet been given that [
The Future Active Participle is shown in context in (23) (adapted from
(23)
Sallust,
Bellum
war.
write-
sum
I.am
quod
which.
populus
people.
Rōmānus
Roman.
cum
with
Iugurthā
Jugurtha.
…
gess-i-t.
wage-
‘I propose to write (lit. ‘I am going to write’) of the war which the Roman people waged with […] Jugurtha.’
(24)
Mārcus
Mark.
Annam
Anna.
praise-
es-t.
be-3
[C]
‘Mark is going to praise Anna.’
As mentioned, the Future Active Participle is built from the same structural core as the Perfect Passive Participle. Whichever allomorphs of the root and the participial morpheme appear in the Perfect Passive Participle, the same ones appear in the Future Active Participle (with a few exceptions—
Further scrutiny, however, reveals that the Future Active Participle has properties that seem rather contradictory. These have to do with the suffix -
On the one hand, -
On the other hand, that conclusion seems backwards.
(25)
Cicero,
opīniō
belief.
come-
bonī
good.
‘a belief in good to come’ (‘…good that is going to come’)
(26)
Cicero,
in
in
[…]
wish.for-
grātulātiōne
congratulation.
‘in the midst of […] the most gratifying (lit. ‘wished-for’) congratulations’
The paradox, then, is that the Future Active Participle’s categorial identity seems to be determined by the participial morpheme -
Just as the Mood/Voice anti-mirroring in certain finite forms (which is really two puzzles: the unexpectedly internal position of Mood and the unexpectedly external position of Voice) was argued in Section 3.2 to come about through
(27)
This analysis solves the paradox laid out above. On the analysis in (27), participial -
Let us recapitulate. Section 4.1 proposed that, in Perfect Passive Participles such as
The hypothesis that Latin makes use of
Before we consider the syntactic predictions of the analysis, a few words are in order about the apparent
There are at least two possible accounts of this that are compatible with the XP/X0 Analysis. One possibility, mentioned in fn. 17, is that Voice is simply absent from these participles. (If so, it must be possible for a head in the middle of a functional sequence to sometimes not be merged in, as argued by
Alternatively, this may be a matter of Vocabulary Insertion. In particular, it may be that Voice does successfully probe for
(28) | a. | Voice ⇔ ∅/ |
b. | Voice ⇔ ∅/AspPROSP ____ |
As long as insertion of (28-a) precedes the Fission that VoicePASS sometimes undergoes in other contexts (Section 3), Fission, and hence
The choice between the two accounts just laid out is left open here.
One final remark is in order. Consider the following infinitives:
(29)
a.
laud-ā-r-
praise-
‘to praise’
b.
laud-ā-r-
praise-
‘to be praised’
Here, as in the participles, Voice does not overtly expone
(30) | a. | VoiceACT ⇔ - |
b. | VoicePASS ⇔ - |
Here too, as long as insertion of (30-b) precedes the Fission that VoicePASS sometimes undergoes, Fission, and hence
Naturally, this discussion raises questions that merit more detailed investigation. What is important here, however, is that although Voice in Latin was argued in Section 3 to bear
The previous sections developed an analysis of Latin verbal morphology that is highly syntactic. On this analysis, Latin verb forms are built largely by narrow-syntactic operations—including, notably, XP-movement.
It is therefore crucial to determine what predictions the XP/X0 Analysis makes about other aspects of Latin syntax, and whether these are correct. That is the objective of this section. Section 5.1 lays out and confirms the prediction that the verb word should move as an XP, skipping intervening heads, and not as a head. Section 5.2 lays out and confirms the prediction that the putative
It is possible that (some versions of) an alternative analysis—particularly an analysis on which the Latin verb word is a complex head, henceforth
On the XP/X0 Analysis, the Latin verb word is a sequence of morphemes—most of them phonologically dependent—that are suspended across a large syntactic expanse.
As mentioned in Section 3.3, if there are any overt arguments or adjuncts within
(31)
Cicero,
cum
when
[TP [
graviter
intensely
[
fīliī
son.
mortem
death.
maer-ē]]1-re-t
grieve-
______1 ]
‘when he was intensely grieving over the death of his son’
In (31), the verb word
In any of these situations, the string referred to here as the verb word
(32)
Livy,
…cum
since
cēterī2
others.
centūriās3
centuries.
[ΣP
not
[TP
[
exple]1-sse-nt
fill-
______1 ]].
‘…since the others did not win the required number of centuries.’
On the XP/X0 Analysis, then, when the Latin verb word is a constituent, it is a maximal projection (an XP—specifically, a TP); it is never a complex head. This leads to the following prediction:
(33) | |
The verb word should be able to move as an XP, skipping intervening heads—i.e., movement of the verb word should not be subject to the Head Movement Constraint (HMC, |
Sections 5.1.1–5.1.2 will show that the finite verb can move to the left of (thereby skipping) the left-peripheral heads
The complementizers
(34) | [ForceP Force [TopP Top [FocP Foc [FinP ut/cum/sīFin … ]]]] |
Significantly, the verb word can move to the left of these complementizers, as in (35)–(36) with
(35)
Tibullus 2.6.42
…nōn
not
ego
I
sum
am
tantī,
that.much.
[
cry-
so.that
illa
that.
semel ].
once
‘…I am not worth a single tear from her.’ (‘…I am not worth that much that she may cry once.’)
(36)
Ovid,
Spect-ā-tu-m
watch-
veni-unt,
come-3
veni-unt
come-3
[
[
watch-
so.that
ipsae
self.
] …
]
‘They come to watch, they come to be watched themselves…’ (‘…they come so that they may themselves be watched…’)
On the XP/X0 Analysis, whenever
(37)
(For convenience, (37) shows -
(38)
Because the verb word
As mentioned, the finite verb can also move to the left of the Fin head
(39)
Lucretius,
Oscit-a-t
yawn-
extemplō,
immediately
[
touch-
when
līmina
thresholds.
villae
house.
] …
‘He yawns as soon as he has reached the threshold of the house…’
On the XP/X0 Analysis, when the verb
In (35)–(39), Fin is preceded by only one clausemate element: the verb word. On the XP/X0 Analysis, this is because, in those examples, all the overt elements within TP except the verb word scramble out of TP before the (remnant) TP moves to the left of Fin. If this is correct, then, since scrambling is (descriptively speaking) optional in Latin, there should be sentences in which some nonverbal material does not scramble out of TP, and TP moves to the left of Fin, carrying this material along in addition to the verb word. This expectation is borne out by sentences such as (40) (adapted from
(40)
Vergil,
Ō
oh
[TP [
mihi
me.
praeteritōs
past.
______2
refer]3-a-t
bring.back-
______3 ]4
[FinP
if
[ScrP
Iuppiter1
Jove.
annōs2
years.
______4]] …
‘If only Jove brought back to me the years that have gone by…’
In (40),
A defender of a Long Head Movement (LHM) analysis of (35)–(39) (cf. fn. 24) might argue that, in (40),
Finally, consider (41)–(42) (adapted from
(41)
Horace,
[TP [
aulaea
canopy(
ru]1-a-nt
fall.down-
______1 ]2
if
‘the canopy falling’ (‘if the canopy should fall’)
(42)
Vergil,
[TP [
īnsān-ī-r-e
be.crazy-
lib-e]1-t
be.pleasing-
______1 ]2
since
[ScrP
tibi3
you.
______2 ]
‘since you like to be crazy’ (‘since being crazy is pleasing to you’)
In both examples, the additional material carried along (
Recapitulating, our claim that movement of the verb word alone past Fin ((35)–(39)) is in fact TP-movement is lent plausibility by (40)–(42), in which the phrasal nature of the movement past Fin is particularly clear.
Another test case is provided by the left-peripheral element =
Consider (43) (adapted from
(43)
Cicero,
against
Brutus.
mē
me.
dic-t-ūr-um
speak-
put-ā-s?
think-
‘Do you think I will contradict Brutus?’
The head
On the XP/X0 Analysis, the verb word, when it is a constituent, is an XP (a TP). Thus, the XP/X0 Analysis leads us to expect that the verb word (= TP) should perhaps be able to move to [Spec,
This expectation is met. Sentences in which
(44)
Cicero,
know-2
mē
me.
dē
about
rēbus
things.
mihi
me.
nōtissimīs
extremely.well.known.
dīc-e-r-e?
speak-
‘Do you know that I am speaking of matters with which I am fully familiar?’
On the XP/X0 Analysis,
Someone who was skeptical about the XP/X0 Analysis might claim that the verb word undergoes head movement to
In (44), the (infinitival) clausal complement of V scrambles out of (or is extraposed from) TP before TP moves to [Spec,
(45)
Cicero,
ship.
people.
Roman.
owe-
‘Whether they are obligated to provide a ship to the Roman people. […]’
(46) | [ |
Devine & Stephens (
On the XP/X0 Analysis, the verb word is often not a constituent. However, if there is no other overt material within TP, then the verb word is the reflex of a (phrasal) constituent—TP—and can therefore be expected to undergo XP-movement. All of Section 5.1 so far has argued that this expectation is met.
However, although the verb word sometimes corresponds to a phrasal constituent (TP), no operations can assemble its parts into a (complex) head on the XP/X0 Analysis—which therefore makes the following prediction:
(47) | |
The Latin verb word should be unable to move as a head. |
This prediction is testable. As foreshadowed in (32), and discussed further in Section 5.3, we take
Consider first (48) (adapted from
(48)
Caesar,
Interiōrēs
inlanders.
plērīque
most.
frūmenta
corn.
not
ser-u-nt…
grow-
‘Most of the inlanders do not grow corn…’
If the verb word can undergo head movement to
(49)
[*]Interiōrēs
inlanders.
plērīque
most.
frūmenta
corn.
grow-
(‘Most of the inlanders do not grow corn…’)
In our (admittedly nonnative) judgment, though, (49) and other such sentences are completely unacceptable; and, to our knowledge, such sentences are unattested. The verb word, then, apparently cannot undergo head movement to
On the XP/X0 Analysis, a central part of the derivation of Latin verb forms is
(50) | |
The putative |
The reason for this is straightforward. If, as the XP/X0 Analysis contends,
The expectation in (50) is met, as shown in (51) (adapted from
(51)
Cicero,
cum
when
[TP [
graviter
intensely
[
fīliī
son.
mortem
death.
maer-ē]]1-re-t
grieve-
______1 ]
‘when he was intensely grieving over the death of his son’
Likewise, in (52) (adapted from
(52)
Cicero,
…cūr
why
nōn
not
[TP [
item
likewise
ut
as
tum,
then,
dērēctō
directly
et
and
palam
openly
[
regiōnem
region.
illam
that.
pet-ī]]1-v-ēru-nt
seek-
______1 ]?
‘…why didn’t they go to that region just as [they did] then, directly and openly?’
Since one of the fundamental properties of adjunction is iterability (
(53)
Cicero,
…ut
that
[TP [
in
in
senātū
Senate.
[
nōn
not
semel
once
sed
but
saepe
often
multīs = que
many.
verbīs
words.
huius
this.
[
mihi
me.
salūtem
safety.
imperī
empire.
atque
and.also
orbis
orb.
terrārum
lands.
adiūdic-ā]]]1-ri-t
attribute-
______1 ].
‘…that he credited me in the Senate, not once but often and at length, with having saved the empire and the world.’
Our claim that the adjuncts in (51)–(53) are adjoined to
(54) | a. | But [ |
b. | But [ |
|
c. | But [ |
Summarizing, the XP/X0 Analysis leads us to expect that it should be possible for the core
Section 5.2 showed that, as predicted by the XP/X0 Analysis, it is possible for one or more
However, arguments can also precede
(55)
Cicero,
quod
because
mortem
death.
hominis
person.
necessārī
close.
grievously
fer-ō
bear-1
‘because I take the death of a close friend grievously’
Likewise, in (56) (adapted from
(56)
Caesar,
ut
that
cohortēs
cohorts.
ex
out.of
castrīs
camp.
quickly
ēdūc-e-re-t
lead.out-
‘that he should lead the cohorts out of the camp quickly’
The XP/X0 Analysis, then, requires that (57) hold:
(57) | In Latin, arguments generated within |
But the hypothesis in (57)
(58)
[ΣP
not
[TP [
ego
I.
illud
that.
parvī
small.
aestim]1-ō
estimate-1
______1 ]].
‘I don’t consider that a small thing.’
But the arguments can also precede
(59)
…cum
since
cēterī2
others.
centūriās3
centuries.
not
[TP [
exple]1-sse-nt
fill-
______1 ].
‘…since the others did not win the required number of centuries.’
Alternatively, some arguments can follow
(60)
Corbulō2
Corbulo.
… [ΣP
not
[TP [
eam
that.
speciem
show.
īnsignium
emblems.
et
and
armōrum
weapons.
praetul]1-i-t
display-
______1 ]] …
‘Corbulo did not make such a display of emblems and weaponry…’
This flexibility of argument positioning vis-à-vis
In permitting arguments to optionally move leftward out of
As alluded to above, a significant advantage of the XP/X0 Analysis is that it accounts for anti-mirroring forms such as passive subjunctives (e.g.,
This section considers two other analyses of anti-mirroring in Latin passive verbs (
Embick (
(61)
As Embick acknowledges, synthetic passive verb forms pose a problem for this analysis. In his assessment, “[t]he reason for this is that whereas the feature [pass], which is directly related to the passive forms, appears in the √-
Embick’s solution to this problem is worth quoting in detail: “if the only starting positions for the [pass] feature are below T-Agr, the structure must be readjusted to place [pass] in T-Agr [… T]wo mechanical operations are required: first, the [pass] feature must be separated from Asp through the process of
(62)
This analysis does manage to account for the fact that, although the passive Voice head (or the [pass] feature, in Embick’s terms) originates low in the syntactic structure, it surfaces to the right of Asp and T in the verbal complex. However, it accounts for this at a considerable cost: it is forced to posit two obligatory morphological operations (one instance of Fission and one of Morphological Merger). These operations are apparently posited exclusively to handle the anti-mirroring morpheme order in passive forms; owing to their lack of independent motivation, they must be considered stipulations.
It might be objected that the XP/X0 Analysis posits
Calabrese (
Calabrese proposes that the passive morpheme (what was analyzed in Section 3.2 as a VoicePASS head) is actually a reflexive-like clitic (“DR”) originating in [Spec,VoiceP] (in passives of transitives) or in direct object position (in the case of unaccusatives with passive/nonactive morphology—i.e., deponent unaccusatives). On his analysis, the verb word is assembled largely by head movement, which causes it to surface in Mood, but also partially by certain postsyntactic operations that are not directly relevant here.
On Calabrese’s analysis, DR undergoes LHM and adjoins to a high functional head F. It subsequently undergoes morphological merger with the complex Mood head (= the verb word). Finally, “[t]he characteristic property of Latin /-r/ [= DR —Authors] is a further merger operation that adjoins DR to AGR” (
(63)
Derivation of passive anti-mirroring according to Calabrese (2019)
It might be objected that this derivation is rather complex, since DR moves three times. However, the first two movements may be independently motivated, since they are modeled after analogous movements that, according to Calabrese, the Italian clitic
If DR underwent only the first two movements in (63), it would surface at the left edge of the verb word, whereas in fact it surfaces at the right edge. To solve this problem, as mentioned, Calabrese posits that it is an idiosyncratic property of DR that it adjoins to Agr. This analysis, then—like Embick’s—can derive the morpheme order in anti-mirroring passive forms only by making at least one stipulation specific to the passive morpheme.
Both Embick’s (
The XP/X0 Analysis, by contrast, accounts for this anti-mirroring order with no passive-specific stipulations whatsoever (Section 3). On the XP/X0 Analysis, the anti-mirroring order in these forms simply follows from their syntactic derivation, which is shared with all other synthetic finite verb forms in Latin, including those whose morpheme order is mirroring rather than anti-mirroring. This provides a significant argument for the XP/X0 Analysis and against the two alternatives just considered.
As we have seen, Latin verbs present a morpheme-ordering puzzle: they seem to both obey and disobey the Mirror Generalization (MG), albeit in different inflectional forms (
The result that has emerged is that, contrary to initial appearances, the MG is in fact valid for Latin—but it should be understood as a generalization about structures formed by operations on heads. Head movement, being extremely local, readily gives rise (in conjunction with other operations) to Latin verb forms displaying morpheme orders that straightforwardly obey the MG. Phrasal movement, by contrast, is subject to much less strict locality conditions than head movement (or at least the canonical head movement at play in the relevant Latin verb forms), and can consequently bring about apparent violations of the MG. This, we have argued, is exactly what happens in apparently anti-mirroring Latin verb forms.
Our analysis of anti-mirroring in Latin lends plausibility to the notion that other instances of anti-mirroring crosslinguistically, including instances not yet discovered, also owe their existence to phrasal movement. However, the analysis does not entail that phrasal movement is the only possible culprit for anti-mirroring in any language. Whether a particular instance of anti-mirroring is due to syntactic XP-movement, to postsyntactic operations, or perhaps even to something else is an empirical question.
On the XP/X0 Analysis, every synthetic finite verb form in Latin, whether it appears to obey the MG or not, has exactly the same derivation
The analysis satisfies the theoretical desiderata in (12). It is
A perhaps surprising consequence of the analysis is that the Latin verb word, though a phonological word, is not always a syntactic constituent. (And when it is a constituent—because there is no overt nonverbal material within
The additional file for this article can be found as follows:
Appendix to ‘XP- and X0-movement in the Latin verb: Evidence from mirroring and anti-mirroring’. DOI:
See Bybee (
We will refer to the root of the verb (i.e., the head of the complement of
Assuming iterated head-raising with left-adjunction (cf.
This question is of interest not only because apparent counterexamples to the MG need to be scrutinized, for the reasons discussed above, but also because the development of Latin into Romance provides an excellent laboratory in which to learn how morphosyntactic change happens (
Embick (
For Latin verb paradigms, see Gildersleeve & Lodge (
The 2
For convenience, every subjunctive verb is translated here as if it were the main predicate of a rationale clause (e.g., ‘so that I may be praised’), but the interpretation of Latin subjunctive verbs is context-dependent (
Following Heck & Müller (
We assume for concreteness that Latin nominals are KPs, headed by a case morpheme K (cf.
The claim that the
A reviewer asks what drives this
Two final remarks are in order about
Secondly, (14) shows the theme vowel -
We say “exactly the same syntactic derivation” because there is one minor postsyntactic difference: when VoicePASS bears the features [Pers:3] and [Num:
This is true of the diagnostic version of the MG ((2)). Baker’s formulation ((1): “Morphological derivations must directly reflect syntactic derivations (and vice versa)”) is almost completely obeyed on the XP/X0 Analysis even if attention is not restricted to operations on heads. As a reviewer notes, together with head movement, “phrasal syntactic movement can produce a morphological derivation that directly reflects (in linearization) the syntactic computation, including both External/Internal Merge.” The only caveat is that the small amount of postsyntactic Fission in Latin (Section 3.1) does not appear to directly reflect anything in the syntactic derivation.
For further discussion of this hypothesis, see Section 5.1, particularly fn. 24.
There are a couple of oddities in the imperative forms.
This complement is shown in (22) as an AspP containing Asp and Voice heads, but little hinges on this; the complement could in principle be as small as VoiceP or
See fn. 21 for more on our analysis of -
This portion of the structure is not compatible with the Final-Over-Final Condition (
A question arises here as to how the auxiliary
(i)
[TP [
In (i), the highest Voice head, which is in the extended projection of
First, on the XP/X0 Analysis, when T is merged in,
Secondly, (i), contra Embick (
Naturally, (i) requires further elaboration and testing, and comparison with alternatives—which, for space reasons, we leave for future work.
Another positive consequence of this analysis is that the relation between Agr and
Embick (
(i)
[ModP -
(adapted from
However, Embick does not specify what the relevant “aspectual properties” are, and it is not clear that all structures containing -
Furthermore, even if deverbal structures with -
(ii)
[Adj0 [
(TV = theme vowel)
The theoretical issues raised by -
The parentheses in (28-a) are shorthand. AspPFV should be mentioned in this Vocabulary Item iff it is projected in Perfect Passive Participles (this was left open in fn. 17).
AspPFV and
However, if AspPFV is not projected in Perfect Passive Participles (and hence not mentioned in (28-a)), (28-a) will not require a revised definition of stretch.
A drawback of this analysis is that the Vocabulary Items in (28) and (30) are somewhat stipulative. As a reviewer notes, however, there may be an alternative analysis that overcomes this issue to some degree. On this analysis, the
The prediction in (33) is based on the traditional assumption that phrasal movement, but not head movement, can skip intervening heads. This assumption has been questioned, however. Harizanov & Gribanova (
Such derivations, in which the verb word moves to the left of Fin, are made use of only in poetry, not in prose (
(i)
a.
They come [
b.
They come [
c.
*They come [they may be watched
d.
**They come [may be watched
e.
*They come [they may be watched themselves
f.
*They come [they may themselves be watched
Crucially, we assume that (non-avant-garde) poetry in a language can differ syntactically from prose in that language only in ways in which (prose registers of) different languages can differ from one another—i.e., only in ways that the theory of syntax independently allows—clearly the most restrictive and hence most interesting hypothesis. That is, we adopt the Development Hypothesis (on which see
Zyman (
The
The LHM analyst might concede this point but respond that, on the analysis in (40), it is not clear where the TP moves or what effect this movement has on information structure, if any. The point is well taken, but since FinP is left-headed in Latin, the conclusion that (a constituent containing)
It might be objected that it is a weakness of the XP/X0 Analysis that its advocates must posit scrambling out of TP (prior to remnant TP movement) in (44), (35), and (39). This objection has little force, however, because scrambling is a prominent and amply attested feature of Latin syntax (see fn. 32 and Section 5.3), so it is hardly a liability for the XP/X0 Analysis that its defenders must posit scrambling in certain sentences. Crucially, too, there are sentences in which (on the XP/X0 Analysis) little or nothing scrambles out of TP before TP moves to the left periphery, so the phrasal nature of the movement can be seen clearly ((40)–(42), (45)).
The sentences containing
(i) Cicero, those. things. manūs hands. adfer-r-e lay.on- nōn not dubit-ā-stī…? hesitate- ‘Did you not hesitate to lay your hands on those things…?’
Because
Since the placement of
We do not claim that only the XP/X0 Analysis gives rise to this expectation about constituent order. Bailey (
It might be objected that, Latin word order being flexible, the constituent order in (51) reveals little about its derivation. But this would be conceptually and empirically misguided. Conceptually, “flexible word order” is an observation, not an explanation. Thus, the acceptability of (51) is not explained by saying that Latin has “flexible word order”; rather, the observation that Latin has “flexible word order” is to be explained by a principled analysis of the mechanisms giving rise to it, as is well appreciated (
(i)
subject > direct object > indirect object/oblique argument > adjunct > goal/source phrase > nonreferential direct object > verb
Questions arise here as to how particular constituent orders are derived on the XP/X0 Analysis. For example, in analytic passives (fn. 19), the order Perfect Passive Participle > internal argument > Anna laudāta
(ii)
[TP [
This hypothesis is independently motivated by sentences like (iii) (judgment from Samuel Zyman, p.c.):
(iii) [Alabada ______1 [praised. en on la the tele]2, TV Ana1 Anna no not fue ______2. was ‘Praised on TV, Anna wasn’t.’
In (iii), the internal argument
No adverbial can intervene between Voice and the rest of the verb word, suggesting that there is no left-adjunction to VoiceP in Latin, but only to
We take argument movement out of a moved
(i)
[Which building]2 did it seem that [on
In (i), both instances of
We argue, then, that the subject, like other arguments, can scramble out of the moved
The important question of what drives scrambling in Latin must, for space reasons, be left for future work.
As noted by Danckaert (
(i) Gaius, Praetor praetor. hērēdēs heirs. fac-e-r-e make- ⟨* ⟨*can-3 nōn not ⟨ ⟨can-3 ‘The praetor cannot appoint heirs.’
On Danckaert’s analysis, this is because the verb word is a complex head and thus cannot cross the intervening head
Relatedly, the XP/X0 Analysis makes a prediction about OV structures like (i) (adapted from
(i) Cicero, [TP [ PRO innumerābilem uncountable. pecūniam money. fac-e]1-r-e make- ______1 ] ‘to make an enormous amount of money’
Whenever the subject in a structure like (i) either is null (as in (i)) or scrambles out of TP, the object-plus-verb-word string should pass constituency tests, since, on the XP/X0 Analysis, it is the reflex of a constituent (TP). Danckaert (
Embick uses the abbreviated notation “√-
Embick exemplifies this structure with imperfect indicative passive forms like (i):
(i) laud-ā-bā-t-ur praise- ‘he/she/it was being praised’
—hence the choice of imperfective Asp, past T, and 3
To be fair, accounting for the anti-mirroring in passive forms is not Embick’s (
Embick (
Our (63) omits structure and operations that are not directly relevant here. For clarity, we have added a
Modulo certain minor postsyntactic idiosyncrasies (fn. 13).
Crucially, the
This material is based upon work supported by the National Science Foundation Graduate Research Fellowship Program under Grant No. DGE-1339067. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.
The authors have no competing interests to declare.