Jump to ContentJump to Main Navigation
Language Change and Linguistic Theory, Volume IIMorphological, Syntactic, and Typological Change$

D. Gary Miller

Print publication date: 2010

Print ISBN-13: 9780199583430

Published to Oxford Scholarship Online: January 2011

DOI: 10.1093/acprof:oso/9780199583430.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (oxford.universitypressscholarship.com). (c) Copyright Oxford University Press, 2020. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 27 October 2020

The Mediopassive: Latin to Romance

The Mediopassive: Latin to Romance

Chapter:
(p.163) 7 The Mediopassive: Latin to Romance
Source:
Language Change and Linguistic Theory, Volume II
Author(s):

D. Gary Miller (Contributor Webpage)

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780199583430.003.0008

Abstract and Keywords

This chapter documents the changes from Latin to Romance in the coding of reflexive, anticausative, middle, and passive. The last three had the same morphological form in Early Latin, but until recently in Romance the first three have had the same syntactic expression. The Latin ‐r forms were triggered by altering the argument structure so that some argument could not be saturated in syntax. Reflexive forms replaced the ‐r forms in different structures at different times. The replacement began in the ergative verbs where ‘I sank myself’ had a bound anaphor in contrast to the lack of agentivity in the type ‘the ship sank itself’ , reanalyzed as an anticausative with reflexive merged in a projection for derived imperfectivity. Subsequently, the construction replaced the ‐r forms in certain other structures, and finally the middle and impersonal, but not the passive (within Latin, at least).

Keywords:   reflexive, anticausative, passive, Early Latin, Romance

7.1 Introduction

MEDIOPASSIVE is the term traditionally used to refer to formations that are employed in conjunction with passive syntax as well as the traditional middle uses, typically encompassing anticausative and operations that alter prototypical argument structure.

As shown in Chapter 6*, ergative, middle, passive, etc. are syntactically distinct and have a number of different properties. What they have in common is a reduction of the prototypical 1〈2〉 argument structure. That reduction, it will be argued here, is the main trigger for mediopassive ‐r morphology in Latin (cf. Miller 1993: ch. 10). This is true of structures in which some role is not available for saturation in syntax and of verb classes that are defective in terms of the 1〈2〉 prototype.

This chapter argues that the reflexive construction with began competing with the ‐r forms first in the ergative verbs. Since the ‐r forms represent syntactically distinct functions, the sē‐replacement of the ‐r forms, as one might expect, occurred at different times in different functions. Within the history of Latin, the middle was the latest structure to undergo formal renewal with . The passive was renewed by an auxiliated construction with esse ‘be’ plus past passive participle (PPP). Salience of the cue for this structure inhibited generalization of the construction to the passive until recently, within the Romance languages, where the replacement by continues to be carried out.

7.2 The Latin mediopassive

Mediopassive morphology in Latin has a number of consistent syntactic functions, but there is some inconsistency with its use by verb classes. Certain (p.164) kinds of unaccusatives take mediopassive (medpass) morphology while other subclasses strongly avoid it.

In Early Latin, there are residues of a contrast between active morphology for volition or control and non‐active in the absence thereof. But the contrasts in (1a) were early given up, leaving the language with two sets of forms to redistribute. With some verbs the ‐r forms prevailed (1a‐2), while other forms generalized active morphology (1a‐1), the result being a certain amount of morphological unpredictability in class (1a).

In class (1a‐1) there are traces of a prehistoric opposition between rūctat ‘produces a belch’ (voluntarily) and rūctātur ‘a belch occurs’ (involuntarily), but rūctātur disappears and rūctat ends up with both meanings. Likewise, lacrimat becomes the preferred form except in the Vulgate where lacrimātur prevails.

In subclass (1a‐2) the active forms are abandoned. In entry (d), the archaic contrast between tuet and tuētur is given up and CL tuētur has both sets of meanings. In (1a‐2f), vagat is mostly archaic and means ‘wanders, roams’ , while the medpass vagātur would seem to encode a lack of volition.

(1) Unaccusative (§6*.7)

  1. a) Volition/control

    1. 1) Active morphology prevails: body functions (residual in Old Latin)

      (a)

      rūctat

      ‘produces a belch’

      (: rūctātur

      ‘belches’ )

      (b)

      lacrimat

      ‘bewails’

      (: lacrimātur

      ‘weeps; sheds tears’ )

      (c)

      ēmungit

      ‘blows the nose’

      (: ēmungitur)

    2. 2) Mediopassive morphology prevails

      (a)

      (lāmentat :)

      lāmentātur

      ‘laments’

      (b)

      (luctat :)

      luctātur

      ‘struggles, strives’

      (c)

      (———)

      cōnscreātur

      ‘clears the throat’

      (d)

      Arch. tuet ‘watches, protects’ :

      tuētur

      ‘beholds, sees, views’

      (e)

      (proficīscit :)

      proficīscitur

      ‘sets out, departs; proceeds’

      (f)

      (vagat :)

      vagātur

      ‘wanders aimlessly’

  2. b) Affective unaccusatives: medpass morphology

    1)

    nāscitur

    ‘is born’

    2)

    moritur

    ‘dies’

    3)

    patitur

    ‘suffers, endures’

    4)

    collābitur

    ‘collapses’

    5)

    but dolet

    ‘feels pain; aches’ (dolētur rare)

    (p.165)

  3. c) Non‐affective unaccusatives: active morphology

    1)

    est

    ‘is’

    (*estur)

    2)

    ex(s)istit

    ‘exists’

    (*ex(s)istitur)

    3)

    cadit

    ‘falls’

    (*caditur)

    4)

    cubat

    ‘is reclining’

    (*cubātur)

    5)

    fluit

    ‘flows’

    (*fluitur)

    6)

    ruit

    ‘crashes’

    (*ruitur)

The affective unaccusatives (1b) involve an ‘affected’ 2‐role, e.g. being born directly affects the patient in a way that reclining, say, does not. This class avoided active morphology (since medpass could also indicate an affected argument) except in subclass (1b‐5). In general, statives in ‐ē‐ had no medpass (dolētur is a rare so‐called deponent), since state verbs can have no passive; cf. *iacētur ‘it is lain (by someone)’ .

Non‐affective unaccusatives (1c) took active morphology. This subclass can have no passive either; cf. *it was existed by us.

One productive use of non‐active morphology was to mark passive voice, as in (2). But even in this category exceptions are found. For instance, facit (2a‐1) has no passive. Its passive is supplied by fit, which is active in form. Since another meaning of fit is ‘becomes’ , it is also a non‐affective unaccusative (class 1c). The other major exception, in the opposite direction, is vāpulat, active in form but passive in function, as shown by (2b), common in early Roman comedy.

(2) Passive (productive): dēlet ‘destroys’ : dēlētur ‘is being destroyed’

  1. a) Exceptions

    1)

    facit

    ‘makes, does’

    :

    *facitur (fit)

    2) vāpulat ‘is (being) beaten’ (passive of verberat ‘beats’ Quintilian 9.3.7)

  2. b) Passive syntax for (2a‐2): ā mē vāpulat ‘is being beaten by me’

The passive was replaced in Late Latin by the auxiliated construction which originated as a stative, and was grammaticalized as the perfect and preterit passive. Latin PPPs are both stative and dynamic, e.g. laudāta est means both ‘she is praised’ (in a praised state) and ‘she has been/was praised’ (perfect/preterit) of laudātur ‘is (being) praised’ . Secondarily it functioned imperfectively, which prevailed in Vulgar Latin and Romance, replacing the ‐r forms in the present passive (viz. est laudāta ‘she is (being) praised’ ), the perfect/preterit passive being renewed by fuit laudāta ‘she has been (was) praised’ (cf. Kuryłowicz 1964: 56f.; Miller 2006: §1.8). This new passive construction became very productive in Late and Vulgar Latin.

(p.166) With the impersonal passive (3), indefinite subject interpretation results from existential closure of SpecνP[AG] (§§6*.10, 6*.12). Consequently, it can only be made with verbs that have a 1‐role to begin with, such as transitives and unergatives (3a). Also included here are verbs culturally classified as unergative, e.g. those in (3a‐2e).

(3) Impersonal passive

  1. a) Unergative verbs

    1. 1) Verbs of motion

      (a)

      venit

      ‘comes’

      :

      venītur

      ‘(some)one comes’

      (b)

      discēdit

      ‘departs’

      :

      discēditur

      ‘(some)one leaves’

      (c)

      it

      ‘goes’

      :

      ītur

      ‘one goes’

      (d)

      currit

      ‘runs’

      :

      curritur

      ‘there is running’

    2. 2) Other verbs

      (a)

      saltat

      ‘dances’

      :

      saltātur

      ‘there is dancing; one dances’

      (b)

      cantat

      ‘sings’

      :

      cantātur

      ‘there is singing; someone sings’

      (c)

      rīdet

      ‘laughs’

      :

      rīdētur

      ‘there is laughter; one laughs’

      (d)

      dormit

      ‘sleeps’

      :

      dormītur

      ‘someone sleeps’

      (e)

      vīvit ‘lives’ , manet ‘waits, stays’ , flet ‘cries’

  2. b) Unaccusative verbs (no impersonal passive; cf. §6*.8)

    1)

    ex(s)istit

    ‘exists’

    (*ex(s)istitur)

    2)

    cadit

    ‘falls’

    (*caditur)

    3)

    cubat

    ‘is reclining’

    (*cubātur)

    4)

    fluit

    ‘flows’

    (*fluitur)

    5)

    ruit

    ‘crashes’

    (*ruitur)

Naturally the unaccusative verbs (3b), which have no 1‐role to operate on, cannot make an impersonal passive. Crucially, these are logically possible (imagine *cubātur ‘there is reclining; someone reclines’ ), which means they are excluded by their formal properties.

Ergative verbs, with a causative : non‐causative alternation, productively require active morphology for the causative alternant and medpass for the anticausative/unaccusative alternant (4a). To these were assimilated the secondary causatives in (4b‐1, 2) beside the earlier uses of the same verbs in the medpass (cf. Stéfanini 1962: 181). Of those, laetat is rare but early (Livius Andronicus, Accius). The relation of albicātur (4b‐3) to albēscit ‘becomes (p.167) white’ is not clear, but it is likely in this instance that the medpass form is secondary (analogical to other such pairs).

(4) Ergative verbs

(causative : unaccusative / anticausative alternation §§6*.6, 6*.9)

  1. a) Externally caused change of state

    1)

    rumpit

    ‘breaks (sthg.)’

    :

    rumpitur

    ‘(sthg.) breaks’

    2)

    (re)clīnat

    ‘causes to lean’

    :

    (re)clīnātur

    ‘leans’

    3)

    volvit

    ‘rolls (sthg.)’

    :

    volvitur

    ‘(sthg.) rolls’

    4)

    minuit

    ‘diminishes (sthg.)’

    :

    minuitur

    ‘(sthg.) diminishes’

    5)

    (dis)solvit

    ‘melts (sthg.)’

    :

    (dis)solvitur

    ‘(sthg.) melts’

    6)

    mergit

    ‘sinks (sthg.)’

    :

    mergitur

    ‘(sthg.) sinks’

    7)

    movet

    ‘moves (sthg.)’

    :

    movētur

    ‘(sthg.) moves’

    8)

    vertit

    ‘turns (sthg.)’

    :

    vertitur

    ‘(sthg.) turns’

    (a)

    Unaccusative syntax:

    vertitur rota

    ‘the wheel turns’

    (b)

    Passive syntax:

    rota ā mē vertitur

    ‘the wheel is turned by me’

  2. b) Factitive/antifactitive

    1)

    laetat

    ‘gladdens’

    :

    laetātur

    ‘is glad’

    2)

    laetificat

    ‘makes glad’

    :

    laetificātur

    ‘rejoices’

    3)

    albicat

    ‘makes white’

    :

    albicātur

    ‘becomes white’

  3. c) Other verb types

    1)

    pāscit

    ‘pastures, feeds’

    :

    pāscitur

    ‘grazes, feasts’

    2)

    vehit

    ‘carries, conveys’

    :

    vehitur

    ‘rides; sails’

Examples of the syntax of unaccusative/anticausative alternants of the ergative verbs are given in (5). Each of those examples could in principle contrast with an active form that has the causative meaning, e.g. if immūtāre were substituted for immūtārier in (5a), the meaning would be ‘I cannot change (someone/something)’ . In (5b), the island is endowed with magical properties. In (5d), movērī is semantically parallel to the intransitive appropinquāre.

(5) Examples of anticausative syntax

  1. a)

    nōn

    possum

    immūtārier

    (Terence, Phormio 206)

    not

    can.1SG

    change.INF.MEDP

    ‘I can't change’

  2. b)

    ibi

    īnsula

    in

    aquā

    commovētur

    (Varro, De lingua latina 5.71)

    there island

    in

    water

    move.3SG.MEDP

    ‘there an island moves about in the water’

    (p.168)

  3. c)

    haec

    (Proserpina)

    ut

    serpēns

    modo

    in

    dexteram

    she

    (P.)

    as

    serpent

    now

    to

    right

    modo

    in

    sinisteram

    partem

    lātē

    movētur

    (Varro, LL 5.68)

    now

    to

    left

    side

    widely

    move.3SG.MEDP

    ‘(Proserpina) like a serpent moves widely now to the right now to the left’

  4. d)

    ubi

    vērō

    movērī

    et

    appropinquāre

    moenibus

    vīdērunt

    when

    but

    move.INF.MEDP

    and

    draw.near.INF

    wall.DAT.PL

    see.PF.3PL

    (Caesar, de Bello Gallico 2.31)

    ‘but when they saw that (the tower) was moving and approaching the walls’

  5. e)

    quod

    (sōl)

    movētur

    ā

    brumā

    ad

    solstitium

    (Varro, LL 6.8)

    that

    (sun)

    move.3SG.MEDP

    from

    bruma

    to

    solstitium

    ‘in that the sun moves from bruma (winter's day) to solstitium (solstice)’

  6. f)

    lūna

    dīcitur

    esse

    quae

    in

    caelō

    tribus

    viīs

    movētur

    moon

    is.said

    to.be

    which

    in

    sky

    three

    way.ABL.PL

    move.3SG.MEDP

    (Varro, LL 7.16)

    ‘(Diana) is said to be the moon which moves three ways in the sky’

  7. g)

    Itaque

    adulēscentēs

    mihi

    morī

    sīc

    videntur

    ut

    cum

    and.so

    youths

    to.me

    die.INF

    so

    seem.3PL

    as

    when

    aquae

    multitūdine

    flammae

    vīs

    opprimitur,

    water.GEN

    mass.ABL

    flame.GEN

    force

    squelch.3SG.MEDP

    senēs

    autem

    sīc

    ut

    cum

    suā

    sponte

    old.men

    but

    so

    as

    when

    sua

    sponte

    nūllā

    adhibitā

    cōnsumptus

    ignis

    extinguitur

    no

    applied

    force.ABL

    spent.PPP.M

    fire

    extinguish.3SG.MEDP

    (Cicero, De senectute 71)

    ‘and so, young men seem to me to die in such a way as when by a large mass of water a flame's potency is squelched, but old men (seem to me to so die) as when, all by itself, with no force added, a spent fire (just) goes out’ (cf. Claflin 1946: 196)

  8. h)

    Ita

    sēnsim

    sine

    sēnsū

    aetās

    senēscit

    so

    gradually

    without

    realization

    age/life

    grow.old.3SG

    nec

    subitō

    frangitur

    sed

    and.not

    suddenly

    break.3SG.MEDP

    but

    diūturnitāte

    exstinguitur

    long.time.ABL

    extinguish.3SG.MEDP

    (Cicero, De senectute 38)1

    ‘so gradually and imperceptibly life passes into old age; it does not suddenly break but over a long period of time (just) fades out’ (cf. Claflin 1946: 216f.)

    (p.169)

  9. i)

    …quod

    item

    fit

    in

    altīs

    flūminibus

    magnōque

    which

    likewise

    happens

    in

    deep

    river.ABL.PL

    great.and

    marī

    cum

    frangitur

    aestus

    (Lucretius 6.143f.)

    sea.ABL

    when

    break.3SG.MEDP

    tide

    ‘which likewise happens in deep rivers and the great sea when the tide breaks’ (cf. Claflin 1946: 218)

  10. j)

    omnia

    mūtantur

    (Ovid, Metamorphoses 17.165)

    all.NOM.PL.NT

    change.3PL.MEDP

    ‘all (things) change’

  11. k)

    memoria

    minuitur

    (Cicero, De senectute 21)

    memory

    decline.3SG.MEDP

    ‘the memory declines’

  12. l)

    augēscunt

    aliae

    gentēs,

    aliae

    minuuntur

    (Lucretius 2.77)

    increase.3PL

    some

    nations

    others

    diminish.3PL.MEDP

    ‘some nations increase, others diminish’

  13. m)

    nūbēs…

    disparguntur

    et

    ita

    diffunditur

    in

    clouds

    scatter.3PL.MEDP

    and

    so

    pour.3SG.MEDP

    in

    terrās

    lands.ACC.PL

    (Vitruvius, De architectura 8.2.2)

    ‘the clouds break and so it [rain water] pours onto the fields’

It is clear from the examples in (5) that non‐active morphology was productively used for the unaccusative alternant of ergative verbs, i.e. for anticausatives.

Medpass morphology was also used for the middle, as in (6). In (6b), note the semantic parallel with stative manēre ‘to stay, last’ . Another crucial fact about (6b) is that the ergative verb (sē commūtent ‘(before) they turn’ ) can have the replacement but the middle servārī was not replaced by sē servāre until the fourth century (see below).

(6) Syntax of the middle (§6*.11)

  1. a)

    bona

    facile

    mūtantur

    in peius

    (Quintilian 1.1.5)

    good.NT.PL

    eas(il)y

    change.3PL.MEDP

    in worse

    ‘good things change easily for the worse’

  2. b)

    alia

    enim

    quae

    manēre

    nōn

    possunt

    some.NT.PL

    now

    which

    last.INF

    not

    can.3PL

    antequam

    commūtent

    before

    REFL

    change.3PL.SBJ

    ut

    celeriter

    promās

    ac

    so

    quickly

    take.out.2SG.SBJ

    and

    vendās,

    alia

    quae

    servārī

    sell.2SG.SBJ

    others

    which

    keep.INF.MEDP

    possunt

    ut

    tum

    vendās

    can.3PL

    C

    then

    sell.2SG.SBJ

    cum

    cāritās

    est

    when

    dearness

    is

    (Varro, De re rustica 1.69.1)

    ‘some (crops), which cannot last, before they spoil, as quickly as possible you should take out and sell; others, which can keep, you should sell when the price is high’

    (p.170)

  3. c)

    haec

    omnia

    (pōma)

    in

    locō

    āridō

    et

    frīgidō

    these

    all

    (fruits)

    in

    place

    dry

    and

    cold

    … servārī

    rēctē

    putant

    (Varro, De re rustica 1.59.1)

    keep.INF.MEDP

    rightly

    think.3PL

    ‘all these (fruits) they rightly think keep well in a dry and cool place’

The ‘inherent’ reflexives (a misleading but traditional label) in (7) consist of actions typically conceptualized as applicable to oneself, especially the personal care subclass (7c), whence the name of the category.

The rationale for the active : medpass alternation (7b, c) is the relation to the ergative verbs (7a). Note that (am)plectuntur (7b‐3) is reciprocal.

(7) Inherent reflexive

  1. a) Compare ergative: movet ‘causes to move’ : movētur ‘moves (oneself)’

  2. b) Core verbs

    1)

    nitidat

    ‘makes bright; bathes’ :

    nitidātur

    ‘takes a bath'

    2)

    dissolvit

    ‘pays’

    : dissolvitur

    ‘frees oneself from debt’

    3)

    plectit

    ‘plaits, (inter)weaves’

    : (am)plectuntur

    ‘they embrace’

    4)

    tenet

    ‘holds, keeps’

    : tenētur

    ‘holds (oneself) back’

    e.g. vix

    teneor

    quīn

    dīcam

    scarcely

    hold.1SG.MEDP

    that.not

    say.1SG.SBJ

    (Plautus, Casina 239)

    ‘I hardly keep from talking’ (i.e. ‘I can hardly refrain from talking’ )

    (p.171)

  3. c) Personal care subclass

    1)

    colit

    ‘cultivates’

    :

    colitur

    ‘gets dressed’

    2)

    cingit

    ‘surrounds, girds’

    :

    cingitur

    ‘puts on a belt’

    3)

    amicit

    ‘wraps about’

    :

    amicītur

    ‘puts on a veil’

    4)

    fingit

    ‘shapes; adorns’

    :

    fingitur

    ‘gets dressed’

    5)

    ōrnat

    ‘furnishes, equips’

    :

    ōrnātur

    ‘gets (oneself) ready’

    6)

    ung(u)it

    ‘(be)smears, anoints’

    :

    ung(u)itur

    ‘anoints oneself’

    7)

    pūrgat

    ‘cleanses’

    :

    pūrgātur

    ‘washes (oneself)’

    8) lavat ‘washes’ [trans. and intr.] / lavātur (Varro, LL 9.106f.):

    (a) lavāmur

    ‘we bathe’ (holistic)

    (b) lavāmus

    ‘we wash (parts)’

In (7c–8), active and medpass morphology are used to encode a semantic difference involving holistic and non‐holistic interpretation, on Varro's testimony. This seems to be isolated in Latin, and slightly idiosyncratic. The opposite occurs in Russian where medpass morphology encodes the partitive relation (Geniušienė 1987: 56f.).

The syntax of the inherent reflexives is illustrated in (8). In (8a), all of the forms ending in ‐(r)ī are medpass infinitives, not indicated for reasons of space.

(8) Syntax of the inherent reflexives

  1. a)

    numquam

    concessāvimus

    lavārī

    aut

    fricārī

    aut

    tergērī

    aut

    ōrnārī,

    never

    stop.PF.1PL

    wash

    or

    rub

    or

    polish

    or

    adorn

    polīrī,

    expolīrī,

    pingī,

    fingī

    (Plautus, Poenulus 219)

    smooth

    refine

    paint

    trim

    ‘we never stopped washing (ourselves) or rubbing ourselves or polishing ourselves or adorning ourselves, smoothing ourselves, refining ourselves, making ourselves up, trimming ourselves’

  2. b)

    reliquās

    armārī

    et

    sēsē

    subsequī

    iussit

    remainder

    arm.INF.MEDP

    and

    REFL

    follow.INF

    order.PF.3SG

    (Caesar, BG 4.32)

    ‘the rest he ordered to armor up and follow him directly’

The point of (8b) is that the troops are to put on their armor, not get armed in the passive by someone else.

The indirect reflexive (9) signals self‐involvement as a goal rather than a patient/theme, as often emphasized by use with an object of inalienable possession (examples in (10)). Nevertheless, the alternations are largely archaic or residual. This was a productive medpass category in Greek but not Latin.

(p.172)

(9) Indirect reflexive

a)

cibat

‘feeds’

:

cibātur

‘takes food (for oneself)’

b)

pāscit

‘feeds, pastures’

:

pāscitur

‘feeds (oneself) on; grazes’

c)

licet

‘is/offers for sale’

:

licētur

‘bids for; makes an offer for (for oneself)’

d)

pignerat

‘gives as pledge’

:

pignerātur

‘takes as pledge; appropriates for oneself’

e)

(ulcīscit

‘avenges’ [rare] :)

ulcīscitur

‘takes vengeance for oneself’

Syntactic examples appear in (10). In (10a), the basic verb is ēmungī ‘blow the nose’ ; cf. (1a‐1c). It is here used with ‘eyes’ as its inalienable object, hence the implication of one's own eyes. The complementizer ut ‘as; that’ is a modal particle, as in (6b) above.

(10) Syntax of the indirect reflexive

  1. a)

    ut

    oculōs

    ēmungāre

    (Plautus, Casina 391)

    C

    eyes.ACC.PL

    nose.blow.2SG.MEDP.SBJ

    ‘may you blow your eyes out!’

  2. b)

    cōpulantur

    dexterās

    (Plautus, Aulularia 116)

    join.together.3PL.MEDP

    right(hands)

    ‘they shake hands (with me)’

On (10b), Stéfanini (1962: 179) objects to Wackernagel's reciprocal interpretation. The context (Aulularia 113ff.) shows that it means rather ‘they join their right hands (to mine); they shake hands (with me)’ . Stéfanini therefore claims it is semantically but not syntactically reciprocal in that ‘joining hands’ is an inherently reciprocal act. This also misses the contextual point. Medpass cōpulantur is opposed to active cōpulant ‘they join (someone else's) hands (to mine)’ and means ‘they join their own hands (to mine)’ .

The denominal unergatives in (11) constitute a very large class with over a hundred verbs in the Roman Republic alone (Flobert 1975: 65–85 etc.). These verbs productively make use of medpass morphology and are traditionally called ‘deponents’ , which is misleading and essentially claims there is no reason for their behavior.

(11) Denominal verbs (productive)

a)

parasītātur

‘plays the parasite (parasītus); sponges’

b)

philosophātur

‘plays the philosopher (philosophus), philosophizes’

c)

frūmentātur

‘gets grain (frūmentum)’

d)

piscātur

‘catches fish (piscis); fishes’

e)

fūrātur

‘is a thief (fūr); steals’

f)

arbitrātur

‘is judge (arbiter); judges; estimates; considers’

(p.173)

Denominal unergative verbs are created from nouns by conversion which, on some accounts, is interpreted as moving a noun into an empty ν or abstract DO, as in the English example in (12); cf. Lat. piscārī ‘to fish’ (11d) and see §6*.7.

(12) Derivation of the verb fish

The Mediopassive: Latin to Romance

Assuming the empty verb is prototypical and has two thematic roles 1〈2〉, removing the 〈2〉 role by incorporating it with the abstract verb head alters the argument structure (for saturation in syntax) in such a way as to trigger medpass morphology.

The triggers of mediopassive morphology are summarized in (13).

(13) Review/Overview of Latin medpass triggers

  1. a)

    Passive

    1*

    〈2〉

    (implicit argument remains for the discharged agent)

    rota

    ā

    vertitur

    wheel

    by

    me

    turn.3SG.MEDP

    ‘the wheel is turned by me’

  2. b) Ergative 1〈2〉 : 〈2〉   (anticausative has no 1‐role §6*.9)

    vertit ‘turns (x)’  : vertitur ‘x turns’

    vertitur rota ‘the wheel turns’

  3. c) Middle 1arb*〈Gen, 2〉 (genericized event, arbitrarized 1‐role)

    hic liber facile legitur (*legit) [example fabricated after (6) above]

    ‘this book reads easily’

  4. d) Inher Refl [1i+2i]〈〉  (internal argument not saturatable in syntax)

    pūrgat ‘cleanses (x)’ : pūrgātur ‘washes, gets washed’

  5. e) Impersonal 1*〈(2)〉 (applies to unergatives with a 1‐role to discharge)

    1. 1) unerg.: saltat ‘x dances’ : saltātur (ā nōbīs) ‘there is dancing (by us)’

    2. 2) unacc.: cubat ‘x reclines’ : *cubātur ‘there is reclining; one reclines’

  6. f) Denominal unergative 1〈2*〉 (incorporation of object N saturates the 2‐role)

    (p.174)

In all of these, mediopassive morphology signals the unavailability of some role for saturation in syntax. In the case of (13f), lexicalization of transferred meanings allows new transitivity, e.g. arbitrārī (11f) as ‘observe, consider’ takes accusative objects.

7.3 Replacement of ‐r forms by

Throughout the history of Latin the mediopassive ‐r forms were being replaced by reflexive at different times in different categories (see Flobert 1975; Kemmer 1988; Miller 1993: ch. 10; Cennamo 1993, 1999, etc.). The development is summarized here.

7.3.1 Ergative verbs

Ergative verbs (those with a causative : anticausative/unaccusative alternation) had formations competing with ‐r forms from the earliest Latin onwards. In Early Latin many of the examples in (14) were already equivalent.

(14) Ergative verbs

  1. a) scinditur / sē scindit ‘splits’

  2. b) rumpitur / sē rumpit ‘bursts’

  3. c) vertitur / sē vertit ‘turns’

  4. d) mergitur / sē mergit ‘sinks’

The original contrast mergit ‘sinks (something)’ : mergitur ‘(something) sinks’ was continued by mergit: sē mergit (14d).

In a construction such as that in (15a) the animate subject is a willful agent (cf. Haspelmath 1990: 44f.), and the reflexive is a true anaphor (1i〈2i〉). In (15b) by sharp contrast, the inanimate subject can have no agentivity or volition, and only the anticausative reading obtains (barring personification and the like).

(15) Animacy and volitional complementarity

  1. a)

    (ego)

    (sub)mersī

    [+volition]

    I

    me

    submerge.PF.1SG

    ‘I submerged myself’

  2. b)

    nāvis

    (sub)mersit

    [–volition]

    ship

    REFL

    submerge.PF.3SG

    ‘the ship sank’

    (p.175)

The reanalysis that allowed examples like (15b) occurred via statements like the civilization destroyed itself, in which civilization is elliptical for the people in it, who of course have volition. More to the point, in this instance, ‘ship’ was a frequent Roman metaphor for the ‘ship of state’ meaning the government. Metaphorically, then, the ship of state could in fact sink itself. In terms of language acquisition, a child hearing ‘the ship sank itself’ might not realize that it referred to the ship of state metaphor, whence the reanalysis of (15b) as the unaccusative counterpart to the true reflexive: 1i〈2i〉 → 〈2〉. What (15a) and (15b) have in common is that the subject and patient are one and the same. The volitional complementarity permits the type (sub)mergit (etc.) to formally replace (sub)mergitur in its ergative function ‘sinks’ . The change is loss of νP[CAUS/AG] with dependent case on generated in a projection for derived imperfectivity (§§6*.9, 6*.11).

There is another factor. Stephens (2006) shows that it is necessary to distinguish (i) virtual reflexives (type this problem practically solves ITSELF), (ii) lexical reflexives (Rania's enthusiasm revealed itself), and (iii) patient‐causer reflexives (the light turned itself on). Type (i) involves events caused by an external (unnamed) agent, (ii) involves non‐causative/spontaneous events, and (iii) are events caused by the patient‐subject. Lexical reflexives usually have abstract noun subjects and “often convey a spontaneous coming into existence or appearance on a scene” (Stephens 2006: 279, w. lit). The fact that the reflexive cannot be stressed is additional evidence it is not an argument.

The most frequent of the derived unaccusatives with reflexive syntax in Early Latin is the so‐called lexical type (an unfortunate label) an idea suggested itself. Also common is the patient‐causer type the glass shattered itself.2 In abstract noun reflexives, such as an opportunity like this presents itself once in a lifetime, itself is not a reflexive argument (as noted above) but a syntactic (rather than morphological) marker of derived imperfectivity (§6*.11, end). Early Latin examples equivalent to this appear in (16).

(p.176)

(16) Non‐argument reflexives with abstract noun subject (Early Latin)

  1. a)

    an

    nescībās

    quam

    eius

    modī

    hominī

    or

    not.know.2SG.IMPF

    how

    this

    kind.GEN

    person.DAT

    rārō

    tempus

    daret

    (Plautus, Bacchides 676)

    rarely

    time

    REFL

    give.3SG.IMPF.SBJ

    ‘Didn't you know how rarely an opportunity of this sort presents itself to someone?’

  2. b)

    omnibu(s)

    nōbīs

    ut

    rēs

    dant

    sēsē

    all.DAT.PL

    we.DAT.PL

    as

    things

    give.3PL

    REFL

    ita

    magnī

    atque

    humilēs

    sumus

    so

    great.NOM.PL

    and

    low.NOM.PL

    be.1PL

    ‘all of us, as things present themselves to us, are mighty or humble’

    (Terence, Hecyra 380)

Examples of this type provide the bridge from the ergatives in (17) to other classes of anticausatives.

As a result of the reflexives with abstract noun subject and the reanalysis in the ergative verbs, became equivalent to ‐r in that category. The alternation between movēbitur and sē moveat in (17a), a classic patient‐causer reflexive, underscores the equivalence of the two (cf. Stéfanini 1962: 203f.). For the ‐r form corresponding to (17b), see (5m) above.

(17) Examples of in the anticausative construction

  1. a)

    (columella)

    movēbitur

    eximitō

    dēnuō

    if

    (upright.pivot)

    move.FUT.MEDP.3SG

    take.out.FUT.IMPV

    anew

    eōdem

    modō

    facitō

    moveat

    same

    way.ABL

    do.FUT.IMPV

    lest

    REFL

    move.3SG.SBJ

    (Cato, De agricultura 20.1)

    ‘if (the upright pivot [of an oil‐mill]) moves, take it out; do it anew, in the same manner, so it won't move’ (cf. Stéfanini 1962: 186)

  2. b)

    medicāmentum

    diffūdit

    in

    vēnās

    drug

    REFL

    diffuse.PF.3SG

    in

    vein.ACC.PL

    (Quintus Curtius 3.6.16)

    ‘the drug diffused into the blood stream’

  3. c)

    semel

    amōris

    pōculum

    accēpit

    merī

    if

    once

    love.GEN

    cup

    take.PF.3SG

    undiluted.GEN

    eaque

    intrā

    pectus

    penetrāvit

    pōtiō

    that.and

    inside

    breast

    REFL

    penetrate.PF.3SG

    potion

    extemplō

    et

    ipsus

    periit

    et

    rēs

    et

    fidēs

    immediately

    and

    himself

    perish.PF.3SG

    and

    thing(s)

    and

    trust(Pl., Truculentus 43ff.)

    ‘as soon as he has taken the cup of undiluted love and that potion has penetrated deep inside, immediately he himself is done for, and his fortune and credit’

    (p.177)

Recall also antequam sē commūtent ‘before they turn (i.e. spoil)’ in (6b). Roman authors played with the ambiguity of ‐r and .3

The anticausatives with are continued in Late Latin and Romance, though with varying degrees of productivity in different Romance languages; cf. (18).

(18) Later survivors of anticausative [Late Latin, Old French]

  1. a)

    ūstulant

    focō

    in stomachō

    (Anthimius 3.6 [c6])

    burn.3PL

    REFL

    fire.ABL

    in stomach.ABL

    ‘they burn with the fire in their stomach’ (cf. Cennamo 1999: 128)

  2. b)

    elle colpes

    non

    auret

    por o

    no.s

    coist

    she faults

    not

    have.PLUPF.3SG

    for that

    not.REFL

    burn.PST.3SG

    (Eulalia 20 [c10])

    ‘she had no faults; because of that, she did not burn’

Example (18b) is the earliest occurrence of the anticausative in Old French, often cited, incorrectly, as a passive. The point is, even though they threw her in the fire, she would not burn.

(p.178) 7.3.2 Other unaccusatives

For certain other of the unaccusatives, e.g. type (1a‐2), the ‐r forms were renewed by in later Latin and Romance. Consider the continuation of lāmentātur ‘laments’ in (19a) and dēlectātur ‘takes delight in’ in (19b).

(19) Emotives with

  1. a)

    et

    tunc

    lāmentābunt

    omnēs

    tribūs

    terrae

    and

    then

    lament.FUT.3PL

    REFL

    all.NOM.PL

    tribes.NOM.PL

    earth.GEN

    (Sacrae Scripturae Matthew 24.30 h)

    ‘and then all the peoples of the earth will lament’ (cf. Cennamo 1999: 128) [the Vulgate has intransitive plangent ‘will mourn’ ]

  2. b)

    intereā

    cum

    Mūsīs

    nōs

    dēlectābimus

    meanwhile

    with

    Muses

    REFL

    delight.FUT.1PL

    (Cicero, Letters to Atticus 2.4.2)4

    ‘meanwhile I'll find delight with the Muses’

As to continuity, both verbs in French have a transitive active use and an experiencer meaning with se: lamenter ‘lament’ : se lamenter ‘bewail’ , délecter ‘delight’ : se délecter ‘take delight; relish’ . For Old French, compare the alternations of crembre ‘fear’ in (20).

(20) Emotive alternations of crembre ‘fear’ [Old French]

  1. a)

    soürs

    est

    Charles,

    que

    nulhome

    ne

    crient

    (Roland 549 [1080])

    secure

    is

    Ch.

    that

    no‐man

    not

    fear.3SG

    ‘Charlemagne is secure, (so) that he fears no one’

  2. b)

    jo

    me

    crendreie

    que

    vos

    vos

    meslissiez

    (Roland 257)

    I

    REFL

    fear.COND.1SG

    that

    you

    REFL

    mix.IMPF.SBJ.2PL

    ‘I'd be afraid that you'd get into a fight’

(p.179) 7.3.3 Inherent reflexives

In the inherent reflexives, originally ‐r forms and did not mean the same thing, but they assimilated early to the same meaning, as in (21).

(21) pūrgat ‘cleanses’ : pūrgātur ‘washes (oneself)’ ≠ sē pūrgat ‘exculpates oneself’ (→ ‘washes (oneself)’ )

In Early Latin, (22a) and the ‐r and formations of many other verbs were used interchangeably. The verbs in (22) were originally an extension (by lexical feature) of the ergative class in that the actives are transitive/causative and bears dependent accusative case (§6*.11, end).

(22) Inherent reflexives

a)

ōrnātur / sē ōrnat

‘gets (oneself) ready’

b)

tenētur / sē tenet

‘restrains oneself; holds back’

c)

levātur / sē levat

‘gets up; lightens; gets better’

d)

reclīnātur / sē reclīnat

‘leans back, reclines’

The syntax of (22a) is illustrated in (23); with ōrnā tē cf. ōrnārī in (8a).

(23) Syntax of (22a)

age

nunciam

ōrnā

tē,

Epidice,

come

now

adorn

REFL

E.

et

palliolum

in

collum

conice

(Plautus, Epidicus 194)

and

mantle.DIM

on

neck

throw.IMPV

‘come now, Epidicus, get (yourself) ready, and put a small mantle on your neck’

This body‐care subclass was continued largely unaltered in Romance; cf. (24).

(24) Body‐care verbs [Modern French]

  1. a)

    je

    raserai

    Jean

    I

    shave.FUT.1SG

    J.

    ‘I'll shave Jean’

  2. b) Jean se (*Ø) rase

    ‘Jean shaves’

7.4 Verbs of motion and change of state

The body‐care subclass (22a) and the directional/change of state verbs (22c/d) took an altered course in Romance. In Late/Vulgar Latin, verbs of motion tended to split along aspectual lines (e.g. telic / atelic), as in Egeria's (p.180) Peregrinatio ad Loca Sancta ‘Pilgrimage to the Holy Lands’ [c4e] (Väänänen 1987),5 where one finds the verbs in (25), cited for simplicity in the third‐person singular.

(25) Motion/change of state (Late Latin: Egeria)

  1. a) sē acclīnat ‘reclines’

  2. b) sē repōnit ‘reposes’

  3. c) sē levat ‘lifts oneself; arises’

  4. d) sē colligit (multitūdō) ‘(the crowd) collects itself; gathers’

  5. e) () plicat [winds (oneself)] ‘goes’

  6. f) sē tendit ‘directs oneself; goes’

  7. g) sē recipit ‘takes oneself back; returns’

  8. h) sē vādit ‘goes off’ (telic)

The syntax of several of these is illustrated in (26).

(26)

recipit

episcopus

et

vādent

take.back.3SG

REFL

bishop

and

go.FUT.3PL

REFL

ūnusquisque ad

(h)ospitium

suum

ut

each‐and‐every‐one.NOM.SG to

lodge

one's.own

C

resūmant

(Egeria 25.7 Prinz)

REFL

renew.3PL.SBJ

‘the bishop retires and each and every one will go off to their own lodge to recuperate’

Most of these began as transitive reflexive verbs, but by the time of Egeria, even some old intransitives have assimilated to this class; cf. (25h), which was simply intransitive vādit ‘goes’ in Classical Latin. Most late authors use the dative sibi with this verb (Orbán 1974: 243f.). Löfstedt (1911: 140) speculates that recipit sē influenced Egeria's construction, but Cennamo (1999: 132) attributes it to the growing equivalence of and sibi. The use with intransitive verbs is clear evidence that sibi/sē is not an argument but just a marker of telicity in some cases. This is due to the collapse of dative sibi, which because of its inherent goal orientation came to mark telicity, and accusative , which with anticausatives was merged in a projection for derived imperfectivity.

The spread of the Late Latin split between verbs with and sibi was apparently by lexical diffusion, given the large amount of confusion among (p.181) the verb classes that is documented in Cennamo (1999). Cennamo concludes (p. 137) that the core distribution was that sibi occurred with telic unaccusatives while occurred with atelic unergatives:

In particular, sibi seems to proceed from intransitive change of state/inherently directed change of location verbs to non‐directed change of location verbs and then states….

Se…spreads from anticausatives denoting change of state to states (mainly mental process, though it can also occur with location state verbs) and then to emotive speech acts and other activity verbs….

(Cennamo 1999: 140)

The reason, then, that vādit ‘goes’ was used with sibi is that it originally designated an inherently directed change of location, as in vāde tibi ‘off with you’ (Cennamo 1999: 123ff.). As sibi yielded to sē, vāde sē was reanalyzed regionally as telic ‘go off’ .6

7.5 Affective unaccusatives

Dative sibi was also frequent with the affective unaccusatives (class (1b) above), as in (27a), where its force was to convey a meaning of ‘sua sponte’ (or the like), continued to some extent in Romance with se forms (27b).

(27) Dative sibi with affective unaccusatives

  1. a)

    genus

    vulneris

    quī

    sibi

    nāscantur

    type

    wound.GEN

    that

    REFL

    be.born.3PL.SBJ

    (Mulomedicina Chironis 52 [c4e/5b])

    ‘a type of wound that arise [sic] (on their own)’ (Cennamo 1999: 122f.)

    (p.182)
  2. b) Semantic reflex of sibi transferred to se [Spanish]

    1. 1) Juan muriò (asesinado)

      ‘John died [of not necessarily internal causes] (was murdered)’

    2. 2) Juan se muriò  [accidental death; cause must be internal]

      ‘John died (on his own)’

    3. 3) Juan se muriò (*asesinado)

      ‘John died [on his own] (*he was murdered)’ (Cennamo 1999: 140)

Semantically, (27b‐1) is a simple (affective) unaccusative. In (27b‐2/3) se is connected with terminal aspect. Unergative verbs, by contrast, are largely inchoative, e.g. Sp. se dormió ‘fell asleep’ vs dormió ‘slept’ .

There was considerable encroachment of on sibi over the centuries until the latter finally disappeared in c8/9. Since the contrast is nowhere preserved formally in Romance, it will not be further illustrated here.

7.6 Victory of and telicity

Verbs with (originally sibi) continue to alternate with ‐r forms as late as c8 (see the list in Flobert 1975: 264f.). A number of verbs with and without se are continued in Old French (Hatcher 1942: 101–26; Pearce 1990: 191ff.) but not simply as options; the se forms are telic, as illustrated in (28).

(28) Change of state and motion verbs and telicity [Old French]

  1. a) lever ‘lift; raise’ :  sei lever ‘rise up; get up’

    1. 1)

      lievet sa main, fait sa b[en]eïçun

      (Roland 2194)

      ‘lifts his hand, makes his benediction’

    2. 2)

      granz

      est

      li

      calz

      si

      se

      levet

      la

      puldre

      (Roland 3633)

      great

      is

      the

      heat

      and

      REFL

      lift.3SG

      the

      dust

      ‘great is the heat, and the dust rises up’

  2. b) gesir ‘lie’ (intr./atelic) : se gesir ‘lie down’ (telic)

    1. 1)

      sur l'

      erbe

      verte

      veit

      gesir

      sun

      nevuld

      (Roland 2876)

      on the

      grass

      green

      see.3SG

      lie.INF

      his

      nephew

      ‘on the green grass he sees his nephew lie/lying’

    2. 2)

      li

      quens

      Rollant

      se

      jut

      desuz

      un

      pin

      (Roland 2375)

      the

      count

      R.

      REFL

      lie.PF.3SG

      beneath

      a

      pine

      ‘the count Roland lay down beneath a pine’

In (28a), an original transitive has a telic (originally reflexive) alternant with se, and in (28b), an original intransitive verb occurs with and without se to (p.183) mark the telic : atelic contrast. In summary, due to the collapse of dative sibi which came to mark telicity and accusative , intransitive verbs developed telic and atelic variants with and without respectively, whence telic spread to ergative verbs, first in the anticausative, where it was originally merged in a projection for derived imperfectivity (§6*.11, end).

7.7 Indirect reflexive

The indirect reflexive, as noted above, was not a productive category in Latin. However, in Late/Vulgar Latin, possibly under Greek influence, the category was revived by means of the dative reflexive, as illustrated in (29a).

(29) Indirect reflexive

  1. a)

    gustāvimus

    nōbīs

    loc[um]

    (Egeria 4.8 Prinz)

    taste.PF.1PL

    we.DAT

    place.ACC[?]

    ‘we took ourselves a snack there’7

  2. b)

    il

    se

    lave

    les

    mains

    [French]

    he

    REFL

    wash.3SG

    the

    hands

    ‘he washes his (own) hands’

  3. c)

    *ille

    sibi

    lavat

    manūs

    [putative Vulgar Latin]

    he

    to.himself

    washes

    hands

    ‘he washes his (own) hands’

  4. d)

    je

    crien

    qu'

    il

    se

    brit

    la

    cuisse

    I

    fear

    that

    he

    REFL

    break

    the

    leg

    (Chrétien de Troyes, Lancelot 1622K/1628P)

    ‘I fear that he [my horse] would break his leg’

    [MnFr.… qu'il ne se brise la cuisse (trans. Poirion 1994)]

There are no clear Latin examples of the French type (29b), which at any rate cannot be a continuation of a putative Latin structure (29c), because the indirect reflexive does not occur in Old French prior to 1150 (see Miller 2001b, w. lit). The first author to use rampant indirect reflexives is Chrétien de Troyes [c.1170–81]; cf. (29d).

(p.184) 7.8 Middle

One of the last categories to undergo formal renewal with was the middle, in which cannot possibly be a reflexive argument. It is at most a marker of derived imperfectivity (§6*.11, end). Classical Latin used only the ‐r forms, as in (6) above. The earliest recorded instance of a middle with is in c2/3, proving that the middle is syntactically different from the other structures with ‐r or . See (30).

(30) Middle [c2/3+]

  1. a)

    facile sēsē regit

    (Tertullian, Mantle 5.3)

    easily REFL rule.3SG

    ‘it controls easily’

  2. b)

    mēla…

    tōtō

    annō

    servāre

    possunt

    (Palladius, Agr. 3.25.18)

    apples

    whole

    year

    keep

    REFL

    can.3PL

    ‘apples can keep a whole year’

The second example (30b) is directly comparable to Varro's servārī possunt ‘can keep’ in (6b) above. There is no way (30b) can be passive (pace Cennamo 1993: 73).8

The middle is continued in the modern Romance languages with a reflex of , as in the French and Italian examples in (31).

(31) Romance middle (cf. Miller 1993: 232, w. lit)

  1. a) French

    1. 1)

      le

      livre

      se

      lit

      facilement

      the

      book

      REFL

      read.3SG

      easily

      ‘the book reads easily’

    2. 2)

      ce

      papier

      se

      lave

      this

      paper

      REFL

      wash.3SG

      ‘this paper is washable’

  2. b) Italian

    i

    bambini

    si

    lavano

    volentieri

    the

    kids

    REFL

    wash.3SG

    gladly

    ‘the kids wash gladly’

(p.185) 7.9 Passive

The final and latest category to be renewed by was the passive. Since the Latin ‐r passive was replaced by an auxiliated formation with esse ‘to be’ , the recent productivity of the auxiliated passive inhibited early generalization of to passive structures. That is, the microcue for a very different kind of passive formation was very robust. Although se/si passives in Romance continue to evolve and have not yet ousted the auxiliated alternative, no ‐r forms subsist in any of the above categories.

Latin never had a genuine unambiguous passive of the type in (32a) with an agent phrase. Some alleged passives have been adduced, e.g. (32b, c).

(32) Putative/alleged passives [Vulgar/Late Latin]

  1. a)

    *littera

    scrībit

    ā(per)

    scrīptōre(m)

    [constructed example]

    letter

    REFL

    write.3SG

    by

    writer

    ‘a letter is written by a writer’

  2. b)

    dōnec

    vulnus

    limpidet

    until

    REFL

    wound

    clean(se).3SG.SBJ

    (Mulomedicina Chironis 86 [c4e/5b])

    ‘until the wound cleanses itself’ (patient‐causer reflexive? anticausative?)

  3. c)

    et

    fecerunt

    se

    ducentos

    quinquaginta

    solidos

    in lucro

    and

    made.3PL

    REFL

    200

    fifty

    denarii

    in profit

    (San Millán 232.18 [c11])

    ‘and a profit of 250 denarius was made’ (Cennamo 1993: 74)

    [more likely: ‘they made themselves 250 denarii in profit’ ]

The type in (32a) with agent phrase is totally out of the question for any period of Latin (cf. Wistrand 1941; Stéfanini 1962: 200). Many Romance languages still cannot have an agent phrase with a se/si‐passive, unlikely because the use of the se‐construction implies that “the referent is universally unknown” (Afonso 2008: 211). As the microcues for true se/si‐passives become more robust, the situation is changing. The closest of the constructions to a passive that Latin ever had is found in examples like (32b), fraught with ambiguity, despite the certainty of Cennamo (1993: 73) who declares it a ‘bipartite passive’ . Nor is Cennamo's passive translation of (32c), from early Spanish Chancery documents, any more compelling. Cennamo has not demonstrated that the case and agreement relations are to be totally ignored and that the sentence cannot just mean ‘they made themselves two‐hundred and fifty denarii in profit’ , construing se in the sense of older sibi ‘for themselves’ . After all, long before the time of this document, dative sibi had been replaced by se.

(p.186) 7.10 Impersonal passive

The impersonal construction (§6*.12) with transitives and unergatives (type (3a) above) is well represented with se/si in early Romance texts; cf. (33).

(33) Impersonal se passive [Old French]

or

se

cante

(Aucassin et Nicolette, header for verse passages [c13])

now

REFL

sing.3SG

‘now (it) is sung; now one sings’

In Latin the same phrase would have been expressed by cantātur ((3a‐2b) above).

The impersonal construction remained productive elsewhere in Romance (34a, b), where it can be analyzed as T checking a zero‐person feature (Mendikoetxea 2008: 304f.), but in French was early replaced by on ‘one’ (34c).

(34) Impersonal passive ‘one eats well here’ [Modern Romance]

a)

Italian

si mangia bene qui

b)

Spanish

se come bien aquí

c)

French

*(il) se mange bien icion mange bien ici

The French change occurred through reanalysis of the impersonal passive as an active with indefinite subject, exactly like the Portuguese reanalysis of vendem‐se from ‘are sold’ to vende‐se ‘(some)one sells / they sell’ (§6*.12). The difference is that in French the active impersonal was replaced by on, already frequent in Old French; cf. (35).

(35) Active with indefinite subject for impersonal passive [Old French]

siet

e. l

cheval

qu'

om

claimet

Veillantif

(Roland 2127 [1080])

sit.3SG

on.the

horse

that

one

call.3SG

V.

‘he (Roland) sits on the horse that one calls (is called) Veillantif’

Although the French change has been seen as a necessity because French is not a null‐subject language (cf. Posner 1996: 274; Mendikoetxea 2008: 317–20, w. lit), the reality is that the reanalysis with on preceded the parametric shift to overt subjects. See (28b‐1) above for an example with null subject from the same text, the Chanson de Roland.

7.11 Conclusion

The Latin ‐r forms were triggered by altering the argument structure in such a way that some argument could not be saturated in syntax (and/or was affected). The forms replaced the ‐r forms in different structures at different times. The replacement began in the ergative verbs where ‘I sank myself’ (p.187) involved a volitional agent but ‘the ship sank itself’ did not. This was bolstered by the type ‘an opportunity presented itself’ . In both of these, the reflexive was not an argument of the verb but a syntactic (as opposed to morphological) marker of the anticausative. The change was a reanalysis entailing contextual loss of νP[CAUS/AG] with dependent case on that was merged in a projection for derived imperfectivity (§§6*.9, 6*.11).

Subsequently, the construction replaced the ‐r forms in certain other unaccusatives (change in lexical features), then other structures, and finally the middle and indirect reflexive, but not the passive (within Latin, at least). In all of these, is initially merged in a projection for derived imperfectivity and bears dependent accusative case. Later, due to the coalescence of accusative and dative sibi which, because of its inherent goal orientation came to mark telicity, intransitive verbs developed telic and atelic alternants with and without respectively. From there, telic spread in parts of Romance to ergative verbs, first in the anticausative, where it was originally merged in a projection involving derived imperfectivity but was subsequently reanalyzed as merged in an aspectual projection involving telicity (§7*.4).

The earlier changes involved structures with no νP[CAUS/AG] projection and no 1‐role. The later changes reintroduced νP[AG] but only with features to syntactically discharge the 1‐role. These changes occurred in the following order: (i) the middle with a genericized [AG] feature, (ii) unergative impersonals with a [PASS] feature, and (iii) the passive itself.

The reason the passive was the most recent of the changes was that it underwent formal renewal by means of an auxiliated construction whose productivity initially hindered generalization of . The microcue for a different kind of passive structure was very robust. By the beginning of Romance, the impersonal passive with was well entrenched, and there are signs of a se/si‐ passive developing. Crucially, however, the new se/si‐ passive has not succeeded in ousting the auxiliated construction in any Romance language to date, proving that passive, ergative, middle, etc. are all syntactically distinct, as argued in Chapter 6*.

The lesson is threefold. First, while the structures represented by the ‐r forms and the constructions appear very similar, they confirm the syntactic evidence that they are in fact distinct because the replacements in different syntactic functions occurred at different times and places. The order of changes is explicable by an account that accords each of the similar structures its own features and distinguishes those that do and do not project νP (itself with different features), but remains entirely mysterious in cognitive radial category models (e.g. Kemmer 1988), which have no way to explain any specific sequence of changes.

(p.188) Secondly, the change of reflexive anaphors to mediopassive markers is frequent (see Geniušienė 1987; Miller 1993: chs. 810; Lidz 1996: ch. 3; Basilico, forthcoming), but asymmetrical. There is no attested change of a mediopassive marker to a reflexive anaphor. This is typical of grammaticalization (Ch. 3*) but totally unexpected on radial category and prototype accounts (cf. Haspelmath 1990: 54f.).

Finally, morphology is dependent on syntax.9 It is not unusual for syntactic expression to precede the creation of a morphological exponent. For instance, older English had a progressive that could appear with passive syntax (a house is building) long before the auxiliated passive (a house is being built) filled out the category with formal voice marking (Miller 2002a: 275ff.). Such lags in overt marking have often given the impression that a syntactic change was later than it was, illustrating the necessity of distinguishing the two. For other passive structures that antedate their morphological exponent, see §§9*.4.2.3, 10*.9.

The options are for each similar syntactic structure to have its own exponent (e.g. passive with an auxiliated construction), or to have a single exponent for all structures sharing some syntactic feature. In this instance, the unavailability of an argument for saturation in syntax was the trigger for ‐r morphology in Latin.10 At the same time, and the auxiliated construction gave Latin the opportunity to have different exponents for different syntactic functions. But the spread of from one structure to another has brought Romance back to essentially the situation that Latin had: one exponent for a variety of related syntactic structures. One difference is that se/si also expresses telicity, but the Latin ‐r forms did not. Another is that the change never affected structures in which the 2‐role was not available for saturation in syntax, e.g. the denominal unergatives. Restriction to structures without νP or a 1‐role and those with νP and a discharged 1‐role is explicable on a syntactic account but not a radial category or prototype account.

It is the tension between having one exponent for a variety of similar syntactic structures or giving each structure its own exponent that creates an endless cycle of changes. This is part of the competition among interacting processes and factors discussed throughout this work as the reason language change cannot be stopped.

(p.189) It seems likely that the similar changes in other Indo‐European languages are to be motivated the same way and are not a product of areal diffusion, except perhaps in Baltic, East Slavic, and North Germanic (see Ramat 2008: 153–9, w. lit). Ramat herself recognizes more than one ‘grammaticalization area’ , as she calls it. Another relevant consideration is that the similar changes in Scandinavian (Miller 1993: ch. 9; Harbert 2007: 322–9; cf. §6.13.1 above) were over a thousand years later than the changes that were already in progress in early Latin. Askedal (2009: 39) suggests that, since the passive use of Scandinavian ‐s(t) decreases from east to west, contact with Baltic and Slavic may have influenced that development. In any event, “[t]he morphological differences and the geographical distribution within Germanic preclude any assumption of Romanization” (Askedal 2009: 39).

Notes:

(1) Note the alliteration and assonance in the first line, as part of the poetry of prose, a fitting complement for Cicero's discussion of Cato on the ending of life.

(2) Virtual reflexives seem not to have played a role. Verbs of the cut class are frequent in virtual reflexives (Hale and Keyser 1987: 17; Stephens 2006: 291), e.g. the toughest carrots virtually slice THEMSELVES with this handy tool. In Early Latin, secāre, lacerāre, caedere (all ‘cut’ in some manner) do not occur with virtual . Therefore only the first two types and the ergative verbs are relevant to the Latin change.

(3) Cf. Cicero (Catilina 1.32):

quā rē

sēcēdant

improbī

sēcernant

ā

bonīs

therefore

depart.3PL.SBJ

malfactors

separate.3PL.SBJ

REFL

from

good

ūnum

in

locum

congregentur

one

in

place.ACC

congregate.3PL.MEDP.SBJ

mūrō

dēnique

quod

saepe

iam

dīxī

sēcernantur

ā

nōbīs

wall.ABL

finally

which

often

now

say.PF.1SG

separate.3PL.MEDP.SBJ

from/byus

‘let the traitors then depart, let them separate themselves from (us) loyal citizens,let them gather in one place, and finally, as I have often said now, let them separate/be separated from/by us by means of a wall’

Do both sēcernant sē and sēcernantur mean ‘separate (themselves)’ or is Cicero making a contrast, using sēcernant sē rather than equivalent sēcernantur, in order to distinguish it from passive sēcernantur? Cicero suggests that if the malfactors do not willingly separate themselves from ‘the good’ , that separation will be imposed on them, but he is clever enough not to assert that in any unambiguous way. The ambiguity could have been rectified had he so desired, but he seems to have deliberately created the ambiguity (ā nōbīs ‘from us’ ~ ‘by us’ ) to obviate any culpability on his part.

(4) Of course Cicero can also use the older ‐r form, as in (i), which Claflin (1946: 198f.) compares to (19b).

  1. ((i))

    dēlectāmur

    cum

    scrībimus

    (Cicero, De finibus 1.1.3)

    if

    take.delight.MEDP.1PL

    when

    write.1PL

    ‘if we find joy when we write’

(5) Egeria's stay in Jerusalem has been dated from Easter 381 to Easter 384 (see Wilkinson 1981: 337f.).

(6) How telicity is determined/valued is disputed. Borer (2005b) uses AspQ. Harley (2005) rejects a telicity‐checking functional projection and proposes that the position of the object and its inherent properties determine telicity. Thompson (2006: 214) argues that “Goal PPs…have a feature [bounded] and combine with the bounded Asp head and bounded verb to result in a telic reading” (cf. §6*.7). Perfectivizing particles and affixes are linked to high transitivity and telicity (Thompson 2006: 214; Basilico, forthcoming). Perfectivity is bounded Aspect, progressive unbounded. Since particles and affixes involve a functional projection, the FP account of telicity may be on the right track. However, the final result state also contributes to telicity, as in (i‐c), which Ramchand (2008: §2.1.2) contrasts with (i‐b).

  1. ((i))

    1. (a)) *John ate the bagel until 3:00.

    2. (b)) Mary dried the cocoa beans (for 12 hours/in 12 hours).

    3. (c)) Mary dried the cocoa beans bone dry (in only 12 hours/*for 12 hours).

It is unlikely that bone dry in (i‐c) raises to SpecAspP to check telicity, and there are indeed multiple sources of telicity (Basilico, forthcoming), but covert raising (at SEM) cannot be excluded. I will assume, then, that was reanalyzed as merged in an aspectual projection.

(7) The sentence has been variously interpreted. This is the usual interpretation today, given that in hortō ‘in the garden’ follows, which makes no sense under the interpretation ‘we enjoyed the place’ . The reading loco is then maintained and interpreted in the sense of Spanish luego ‘there’ . While there are problems for all interpretations, the point for our purposes is that the structure is an indirect reflexive expressed with the dative case.

(8) On ‘Proto‐Romance’ mēla in (30b) for CL māla ‘apples’ , see Svennung (1935: 116ff.).

(9) This does not mean that morphological change cannot spawn syntactic change, but it is indirect, via feature changes. Collapse of the case system and coalescence of dative sibi with accusative left the latter with a telic feature that caused (non‐argument) to be merged in an aspect projection.

(10) Similar arguments have been made for other languages. For instance, on the account by Lidz (1996), a certain argument in Lexical Conceptual Structure is not linked to Argument Structure.