John Benjamins Publishing Company
his is a contribution from Diachronica 31:3
© 2014. John Benjamins Publishing Company
his electronic ile may not be altered in any way.
he author(s) of this article is/are permitted to use this PDF ile to generate printed copies to be
used by way of ofprints, for their personal use only.
Permission is granted by the publishers to post this ile on a closed server which is accessible
only to members (students and faculty) of the author’s/s’ institute. It is not permitted to post
this PDF on the internet, or to share it on sites such as Mendeley, ResearchGate, Academia.edu.
Please see our rights policy on https://benjamins.com/#authors/rightspolicy
For any other use of this material prior written permission should be obtained from the
publishers or through the Copyright Clearance Center (for USA: www.copyright.com).
Please contact rights@benjamins.nl or consult our website: www.benjamins.com
Notes on the history of reciprocal
NP-strategies in Semitic languages
in a typological perspective*
Elitzur A. Bar-Asher Siegal
he Hebrew University of Jerusalem
Among Semitic reciprocal constructions, a division is seen between two types:
1) two-unit constructions, with two components, each illing a diferent argument position of the verb, and 2) one-unit constructions, with an anaphora that
co-refers with the subject (that must be plural) and occupies only the non-subject
position required by the verb. he goal of this paper is to explain how these constructions developed, speciically: 1) how did the various types of two-unit constructions evolve? and 2) could diachronic chains be identiied in order to explain
the development of the one-unit constructions from the two-unit constructions?
Previous work on question (1) focuses on the range of phrases that tend to develop into reciprocal markers. Such accounts, however, do not explain how these
constructions developed the speciic meanings they have. I argue that consideration of the semantics of these constructions is crucial for understanding their
evolution. Instead of ‘reciprocal constructions’ it is better to see them as denoting
‘unspeciied relations’. As for (2), various attempts have been made to explain such
processes focusing on Indo-European languages, which do not capture the Semitic
developments; therefore I propose an alternative hypothesis, according to which
the one-unit constructions result from a reanalysis of the two-unit constructions.
Keywords: reciprocal constructions, unspeciied pronouns, typology,
grammaticalization, reanalysis, agreement, Semitic languages
* I would like to thank Eitan Grossman and two anonymous readers for excellent and productive comments on earlier versions of this paper. Parts of this paper were presented at the seminar
of the Department of Hebrew Languages, Hebrew University of Jerusalem; Semitic Workshop,
Harvard University; the colloquium of the Department of Hebrew and Semitic Languages,
Bar-Ilan University; and at “Agreement from a diachronic perspective,” Philipps Universität
Marburg, Germany (October 4–5, 2012). I thank the audience in these forums for productive
comments. Finally I wish to thank my consultants Daniela Viale (Italian) and Moshe Bar-Asher
(Judeo-Arabic). his work is supported by the European Union grant IRG 030–2227.
Diachronica 31:3 (2014), 337–378. doi 10.1075/dia.31.3.02bar
issn 017–4225 / e-issn 15–714 © John Benjamins Publishing Company
338 Elitzur A. Bar-Asher Siegal
1.
Introduction
Reciprocal constructions are oten deined as a grammatical means for encoding symmetric relations (Lichtenberk 1985: 21, Kemmer 1993: 102, Nedjalkov
2007a: 6). A symmetric relation, in turn, is deined as a relation in which there are
two participants (A and B), and A stands to B in the same relation as B to A. he
current paper focuses on reciprocal constructions that have received various designations by scholars: nominal strategies (König & Kokutani 2006), pronominal
strategies (Nedjalkov 2007a: 12) and NP-strategies (Evans 2008). Constructions
are included here based on two criteria:
I. hey share the same range of uses (§2);
II. he encoding is non-verbal, i.e., verbs in the relevant constructions are transitive (unlike verbal encoding of reciprocity). hus, (1a), a reciprocal sentence,
which denotes a symmetric relation between its participants, has the same
predicate and argument structure as (1b):
(1) a. James and Beth love each other.
b. James loves Beth.
I use the term NP-strategies, since the relevant expressions, which constitute the
constructions, always ill the positions of the NP-arguments of the verb. Other
terms designate subtypes of such constructions.
he current study focuses on Semitic, with parallels from other families to
show broader typological implications. Ater presenting several types of constructions and classiications according to formal characteristics, this paper deals with
the most basic historical question: how do such constructions evolve?
he structure of the paper is as follows: ater describing the semantics of the
constructions under discussion (§2), I introduce various common types of reciprocal NP-strategies in Semitic, tracing their evolution from certain constructions
that also have a reciprocal reading (§3) and the development of others (§4). his
study also sheds light on the synchronic syntax of reciprocal NP-strategies, in
Semitic and non-Semitic languages (§5). §6 presents a diferent path of development, and §7 concludes.
2. he semantics of reciprocal NP-strategies — preliminary observations
I challenge and ultimately modify the common assumption in the typological literature that reciprocal constructions encode symmetric relations. As will become
clear this discussion will turn out to be instrumental for understanding the evolution of many of the relevant NP-strategies.
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 339
It has been repeatedly noted that cross-linguistically the same NP-expressions
that encode symmetric relations (e.g., English each other) express other relations
where strong reciprocity is impossible (Fiengo & Lasnik 1973, Dougherty 1974,
Lichtenberk 1985, Dalrymple et al. 1998, Williams 1991, Beck 2001, Haas 2010,
Evans et al. 2011). For example, the following sentence does not express a symmetric relation:
(2) hey were hiding behind each other.
Such examples are common among ancient languages as well:
(3) Ancient Greek:
οὐκ ἀθρό-ους
ἀναβιβάζ-ων,
ἀλλὰ κατὰ μέρ-η
NEG in.mass-M.PL.ACC advance.PTCP.M.SG.NOM but by division-N.PL.ACC
πυκν-οὺς
ἐπ᾽ ἀλλήλ-οις
crowded-ADJ.M.PL.ACC ater=RECP-M.DAT
“hey did not make the attack en masse, but by divisions in close order,
following each other.”
(Appian, Punic Wars, 18: 126)
(4) Akkadian:
2 kakkabān-i rab-ût-i… arki ah̬āmeš iṣarrū
two stars-gen big-pl-gen ater recp lash.dur.3.m.pl
“Two great shooting stars lash, one ater the other.”
(hompson Rep. 202 r. 4, Neo Assyrian)
(5) Jewish Babylonian Aramaic:
manḥī
a-hǝdāde
place.prs.pass.3m.pl on-recp
“hey are placed on top of each other.”
(B. Meṣi῾a 25a)
Considering (5), if X is on top of Y, then Y cannot be on top of X. he reciprocal
expression hǝdāde conveys that X is on top of Y, Y is on top of Z and so on, not a
symmetrical relation.
Dalrymple et al. (1998) and others have surveyed the logical relations expressed by so-called reciprocal pronouns in English. Taking the sentences in (6)
from Dalrymple et al. (1998), each has diferent truth conditions with regard to
the number of pairs that should exhibit the relation expressed by their predicates.
(6) a.
House of Commons etiquette requires legislators to address only the
speaker of the House and refer to each other indirectly.
b. “he captain,” said the pirates, staring at each other in surprise.
c. Five Boston pitchers sat alongside each other.
© 2014. John Benjamins Publishing Company
All rights reserved
340 Elitzur A. Bar-Asher Siegal
In (6a) there must be a symmetric relation between each possible pair; each of the
pirates in (6b) should stare at one of the other pirates (and perhaps also be stared
at), but not necessarily at every pirate; in (6c) all of the pitchers must be sitting in
a line, and each must be sitting either next to one other pitcher (on the two ends)
or next to two others (in the middle).
Moreover, sentences have diferent truth conditions in diferent contexts.
Compare (7a) in two diferent contexts (7b–c):
(7) a. hey will wake each other up.
b. I never put my twins in the same crib, because they will wake each other
up [i.e., it is suicient that only one of them will wake up the other].
c. hey made an agreement that they will wake each other up [understood
as in taking turns to sleep].
he truth conditions of (7b) do not include a symmetric relation, while those of
(7c) do.
In some languages the same expressions are used in reciprocal contexts and
in casuistic laws. For example, in Biblical Hebrew (§3.2.2), the two elements ’īšrē‘ēhû “man-his.fellow,” are used both in reciprocal contexts (8a) and in casuistic
laws (8b). In the latter only one person does something to the other:
(8) a.
way-yaḥăziqû
’īš bĕ-rō’š
rē‘-ēhû
and-hold.IPF.3.M.PL man in-head.of fellow-POSS.3.M.SG
“hen each man grabbed his opponent by the head.”
(2 Sam. 2:16)
b. wĕ-kī
yāzid
’īš ‘al rē‘-ēhû
and-when act.presumptuously.IPF.3.M.SG man on fellow-POSS.3.M.SG
lĕhorg-ô
bĕ-‘ormâ
kill.INF-ACC.3.M.SG in-cunning
“If someone acts presumptuously toward someone else, so as to kill him
with cunning…”
(Exod. 21:14)
While all the other logical relations that the NP-strategies cover are well attested,
only some of the Semitic constructions show up in casuistic laws. In §3.3 I argue
that, based on the origin of the NP-strategy, the use of this strategy in such laws
can be predicted.
I do not provide a complete semantic account for these strategies, but note
some crucial observations. his suggests the following characteristics for all of the
relationships described thus far:
(9) For a given set:
a. All members of the set should participate, as one of the arguments of the
relation denoted by the predicate with another member of the set.
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 341
b. It is crucially important for the interpretation of the sentences that it is
immaterial which member of the set takes which role in the relation; it is
only the number of applications of the relations between members of the
set that matters.
hese requirements it the following descriptive portrayal of the function of these
constructions, which we will designate ‘unspeciied constructions’ and the pronouns and anaphors used in these constructions ‘unspeciied pronouns/anaphoras’:
(10) Unspeciied constructions: expressions used in relations between two
(deined) sets (or more) without specifying which set occupies which
position.
hese relations are deined between ‘sets’ since the reciprocity can be a relation
between groups, not just individuals, as in the following Akkadian sentence:
(11) nišē
māt
Aššur māt
Karduniaš itti
people.of country.of Assyria country.of Babylonia with
ah̬āmeš ibballū
RECP mingle.3.M.PL.DUR
“he people of Assyria and Babylonia mingle with each other.”
(CT 34 39 ii 37, NA)
he formula in (12a) captures this deinition for the set A with two or more members and the relation R. For reasons that will become clear later I also provide in
(12b) the truth conditions when the set A has only two members:
(12) a. | A | ≥ 2 and ∀ x ∈ A ∃ y ∈ A (x ≠ y ∧ (Rxy ∨ Ryx))
b. | A | = 2 and ∃ x, y ∈ A (x ≠ y ∧ Rxy)
(12a) states that for a given set A, for each member of the set, it is true that it is
a member of a subset of two members of the set A, standing in the relation R.
(12a), however, does not account for how the speciic meaning of a sentence is
determined, an issue beyond our concern here. Still, such constructions only necessitate that each member of the set stands in a single relation to another member. Although this is necessarily true for all of the sentences with these pronouns,
these are not suicient conditions to capture only true sentences in many cases. As
example (7) demonstrates, the given context determines the speciic meaning of a
given sentence, and a full semantic account would require an explanation of how
it is speciied. hus (12) represents the basic meaning of the NP-strategies, further
speciied and strengthened in each context. Having this assumption, I will argue in
this paper, illuminates the origin of the reciprocal NP-strategies.
While typological discussions begin with prototypical symmetric relations
and examine which constructions denote them (Lichtenberk 1985, Kemmer 1993)
© 2014. John Benjamins Publishing Company
All rights reserved
342 Elitzur A. Bar-Asher Siegal
and consequently consider their usage in asymmetric relations as an “extended use
of a reciprocal marker” (Nedjalkov 2007a: 9), I argue for the opposite position: At
least for historical discussions, one should examine the entire range of functions
of the NP-strategies and see their evolution in this larger context. Accordingly,
semantically speaking, symmetric relations constitute only one subtype of the unspeciied relations.
Although I call these constructions ‘unspeciied constructions’ and the pronouns that appear in such constructions ‘unspeciied pronouns’, I still also use the
common terms ‘reciprocal constructions’ and ‘reciprocal pronouns’ interchangeably, as these constructions express prototypical symmetric relations too.
3. Types of reciprocal NP-strategies in Semitic languages
Among the constructions found in Semitic, a signiicant division is seen between
two types:1
I.
Two-unit constructions: constructions with two components, each illing a
diferent argument position of the verb.
II. One-unit constructions: constructions with a one-unit expression, which
co-refers with plural subjects and occupies only the non-subject position as
required by the verb, i.e., the forms analyzed as anaphors in Government
and Binding, where anaphors in this framework are variables that have to be
bound in their governing category. (When referring to these expressions I will
use the term anaphora.)
Akkadian, for example, has both types. he two-unit construction consists of a
repetition of ah̬um “brother,” while the one-unit type contains variants of ah̬āmiš/
ah̬āiš. he former was predominant in the earlier dialects (13a), while the latter
developed only in Middle Babylonian and Middle Assyrian (13b) (Bar-Asher
Siegal 2011).
(13) a.
Old Akkadian:
urkatam ah̬-um
ana ah̬-im
lā inappuš
aterwards brother-NOM to brother-GEN NEG make.a.claim.DUR.3.SG
“Aterwards one will not make a claim against the other.” (TCL 19 63:45)
b. Late Babylonian:
ah̬āmeš ippalū
RECP pay.DUR.3.M.PL
“hey will compensate each other.”
(Dar 321:29)
1. See Haspelmath’s (2007) division between ‘bipartite anaphor’ and ‘single-part anaphor.’
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 343
Classical Arabic uses only the two-unit construction, consisting of the repetition
of ba‘ḍ “some” (14a) (with a possessive pronominal suix attached to the irst unit,
agreeing with the participants of the reciprocal relation). his construction also
appears in the modern standard language (14b), but in addition to two other constructions: a one-unit construction with the irst element only (14c), and an apparent two-unit construction, with only one pronoun marked for case and a second
caseless pronoun (14d).
(14) a.
Classical Arabic:
danā
ba‘ḍ-u-hum
min ba‘ḍ-in
approach.PST.3.M.SG some-NOM-POSS.3.M.PL from some-GEN.IND
“hey approached each other.”
(AS 161, Kremers 1997: 31)
b. Standard Arabic:
qāla
ba‘ḍ-u-hum
li ba‘ḍ-in
say.PST.3.M.SG some-NOM-POSS.3.M.PL to some-GEN.INDF
“hey said to each other…”
(Cantarino 1975: 137)
c. Standard Arabic:
muraddidīna
‘alā masāmi‘-i ba‘ḍ-i-him
ḥikāyāt-i
repeat.AP.M.PL.ACC on ear.PL-GEN some-GEN-POSS.3.M.PL story-Pl.GEN
l-’ayyām-i
wa-l-layāl-ī
DEF-day.PL-GEN and-DEF.night.PL-GEN
“Retelling [lit. repeating to the ears of] to one another stories of the days
and nights…”
(Cantarino 1975: 137)
d. Standard Arabic:
tu‘azzizāni
ba‘ḍ-a-humā
l- ba‘ḍ
strengthen.IMP.F.DU some-ACC-POSS.3.DU DEF-some
“hey strengthen each other.”
(Kremers 1997: 55)
Likewise, Syriac employs both a two-unit construction, consisting of the repetition of ḥad “one” (15a) and a one-unit construction with the form ḥǝdādē (15b):
(15) a.
Syriac:
mallel[u]
re‘w-ātā
ḥad ‘am ḥad
speak.PST.3.M.PL shepherd-PL one.M with one.M
“he shepherds spoke with each other.”
(Luke 2:15)
b. Syriac:
mšaḥlp-īn
rēḥāy-hon
men da-ḥdādē
diferent-M.PL smell.POSS.3.M.PL from of.RECP
“heir smells are diferent from each other.” (Life of Simon Stylites 382:8)
Historically, while the two-unit constructions in Semitic resulted from grammaticalization of NPs to become reciprocal constructions, the anaphors of the oneunit constructions developed from two-unit constructions and not directly from
© 2014. John Benjamins Publishing Company
All rights reserved
344 Elitzur A. Bar-Asher Siegal
nominal expressions. he shit from two-unit constructions to one-unit constructions is well attested cross-linguistically. Compare, for example one…another in
English to einander in German. My goal is to better understand the mechanism
behind these developments, aiming at answering these questions:
I. How did the various types of two-unit constructions evolve?
II. Could diachronic chains be ofered in order to explain the development of the
one-unit constructions from the two-unit constructions?
Question (I) has been treated in the literature, focusing on the range of the phrases
that tend to develop into reciprocal markers (Heine & Kuteva 2002: 92, Heine &
Miyashita 2008: 177–182, Nedjalkov 2007b: 155). Such accounts, however, do not
explain how these constructions developed the speciic meanings they have. I argue (§§3.1–3.2) that consideration of the semantics, as in §2, is crucial for understanding the evolution of the Semitic constructions. As for question (II), various
attempts have been made to explain such processes focusing on Indo-European
languages, especially for Germanic (§3.2.3.1, §4.1). hese proposals, however, do
not it the Semitic developments; therefore I propose an alternative hypothesis
(§4.3). I also ofer evidence of a parallel process in Italian. I begin with diferent
types of two-unit constructions found among the Semitic languages, accompanied by hypotheses for their developments as expressions of unspeciied relations.
Figure 1 provides the range of construction types, all but one (constructions with
universal quantiiers) found in Semitic. heir organization is relevant to the origin
of the use of these structures to express unspeciied relations.
NP-strategies
Two-unit constructions
Nominal
constructions
Pronominal
constructions
Quantificational
constructions
Constructions with
universal quantifier
One-unit constructions (anaphors)
Extension of
a pronominal
construction
Repetition
of partitives
Original
pronouns
Repetition of
the NP
Bleached
nouns
Figure 1. he range of construction types used to express unspeciied relations.
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 345
3.1 Two-unit constructions: A nominal construction, a pre-grammaticalized
structure
A well-known phenomenon in Indo-European and other families (Nedjalkov
2007b: 154, Plank 2008: 359, Haas 2010: 11, Evans 2008: 64) is a repetition of nouns
without a speciication of the particular referent of each of its tokens. his repetition may express reciprocal relations (among other relations), as in the following
Latin proverb:
(16) hom-o
hom-ini
lup-us
est
Man-NOM.SG man-DAT.SG wolf-NOM.SG be.PRS.3.SG
“Man is wolf to man.”
Nedjalkov (2007b) and Haas (2010) comment that this type of construction is
mainly used in generic contexts. However, as we shall see among the Semitic languages, and beyond that, this is a wider phenomenon, as such repetitions are used
also with referential expressions. For example, reciprocity indicated by a repetition
of nouns is found in Biblical Hebrew:
(17) wat-ta‘ărōk
iśrā’ēl û-pĕlišt-īm
ma‘ărākâ liqra’t ma‘ărākâ
and-lead.ipf.3.f.sg Israel and-Philistine-pl battle
toward battle
“Israel and the Philistines were drawing up their lines facing each other [lit.
line of battle against line of battle].”
(1 Sam. 17:21)
his structure is not a two-unit construction in the literal sense, as it does not contain any component unique to this construction. It is not even a grammaticalized
construction for expressing unspeciied relations but rather a compositional way
of expressing such relations. Since the function of the relevant NP-strategies is to
express relations between two sets without specifying which set occupies which
position (as discussed in §2), a repetition of a noun in contexts where each token
of the noun clearly has a diferent referent (otherwise a relexive pronoun should
appear) is therefore expected for this function (Plank 2008: 359). It must be noted,
however, that the truth conditions of (16) are stronger than what is formulated in
(12). However, with deinite descriptions of plural nouns the truth conditions are
the same, as in the following sentences. Consider example (18), which can be read
in several ways:
(18) he boys fed the boys and the girls fed the girls.
his sentence can be true even if in a given set of boys and girls only some of the
boys fed all the other boys and some of the girls fed all the other girls. In certain
circumstances it can mean, of course, that all the boys were fed, and that the feeding was done by the boys (and similarly about the girls), or that all of the boys and
© 2014. John Benjamins Publishing Company
All rights reserved
346 Elitzur A. Bar-Asher Siegal
girls were divided into pairs of the same gender and participated in a reciprocal
feeding. Interestingly, although the deinite article with plural expressions implies
maximality, in this case the maximality is in the entire clause. hus, for the sentence to be true, all boys should be either the feeders or should be fed. he truth
conditions of a repetition of deinite plural expressions are, accordingly, similar
to what has been formulated in (12). As noted above, (12a) states that for a given
set A, for each member of the set, it is true that it is a member of a subset of two
members of the set A, standing in the relation R. Accordingly, a repetition of a
noun indicates this semantic relation in the following way:
(19) [For a given set of individuals denoted by NP, every individual is part of a
pair in which] — NP R NP
As the following sections will demonstrate, assuming that unspeciied constructions consist of the elements “NP R NP” sheds light on the existing constructions in
the Semitic languages. As for the irst part of (19), which is in brackets, it is beyond
the scope of the current paper to demonstrate how it is achieved in a compositional
way. In the case of a set with only two members, however, the formula of NP R NP
directly represents (12b). As we shall see, it is possible that these constructions
grammaticalized for cases where the reciprocity holds between two sets only. Once
it grammaticalizes, the same construction is used with larger sets as well.
In light of the above, example (20) from Akkadian, which does not necessarily
express symmetric relations, is also an NP-strategy expressing unspeciied relations:
(20) Old Babylonian:
qaqqar-um eli qaqqar-um2 utelli
surface-NOM over surface-NOM high.DUR.3.SG
“How much higher is one level [of water in the water clock] than the other
level?”
(TMB 26 50:3)
Negative sentences have only reciprocal readings:
(21) Late Babylonian, Achaemenid:
awīl-u
eli awīl-i
mimma el-īšu
person-NOM to person-GEN thing NEG-have.PST.3.SG
“hey do not owe each other a thing.”
(MDP 24 328:8)
his is actually a logical deduction: even if in positive sentences each noun/pronoun picks only one unspeciied referent, in negative sentences, with a wide scope
2. Concerning the nominative case, see Bar-Asher Siegal (2011: 28 n. 24) and Evans (2008: 64)
for a similar phenomenon in Bangla.
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 347
negation (taken as a sentential negation), each of the nouns must pick both. (22)
illustrates this:
(22) a.
“It is not true that one of the two did R to the other one.”
his sentence is semantically equivalent to:
“he two persons did not do R to each other.”
b. Formally, the following three formulae are semantically equivalent for
the set A that contains only the two members a, b: A={a, b}:
i. |A| ≥ 2 and ∀x ∈ A ~∃y ∈ A (x ≠ y ∧ (Rxy ∨ Ryx))
ii. ~(Rab ∨ Rba)
iii. ~ Rab ∧ ~Rba
he connection between the reciprocal expressions and repetitions of two nominal
expressions is seen in other related constructions:3 the pairs of pronouns one another with the antecedent in the sentence (dogs) replace the repetition of the NP dog:
(23) a. hey entered the room, dog ater dog.
b. hese dogs entered the room one ater the other.
Considering (24a), from Old Babylonian, in light of the semantics of these expressions, it becomes insigniicant that is not a ‘real’ reciprocal sentence (for it is most
likely that only one king defeats the other). It is equally unimportant that (24b) from
Jewish Babylonian Aramaic does not necessarily describe mutual visits. Such sentences are still relevant for our current discussion, due to the way unspeciied nouns
express relations between diferent sets. In both cases a repetition of a nominal expression conveys these relations (as is also relected in the English translation).
(24) a.
Old Babylonian:
šarr-um šarr-am ina kakk-i
idâk=ma
king-NOM king-ACC in battle-GEN defeat.DUR.3.SG =and
“One king will defeat the other in battle.”
(YOS 10, 56ii37)
b. Jewish Babylonian Aramaic:
nǝše
lǝgabbe nǝše
šǝkīḥ-ī
d-āzl-ī
gabr-e
women to
women ind.PTCP.PASS-PL REL-go.PTCP-PL man-PL
lǝgabbe gabr-e lā šǝkīḥ-ī
d-āzl-ī
to
man-PL NEG ind.PTCP.PASS-PL REL-go.PTCP-PL
“Women visit each other frequently; men do not visit each other
frequently.”
(Yebam. 26a)
My hypothesis is that most NP-strategies grammaticalized from a repetition of two
nouns. Before turning to this larger idea, Plank (2008: 359) importantly warrants:
3. he semantic relations between the constructions are beyond our concern; cf. Beck (2007).
© 2014. John Benjamins Publishing Company
All rights reserved
348 Elitzur A. Bar-Asher Siegal
“A more general drawback of the strategy of identical NP repetition is that it does
not work when the participants in a reciprocal relation are diferently categorized:
from ‘earl(s) hated queen(s)’, it is asking for too much to infer that the same relation also obtained in reverse between the same referents.” his drawback, I would
argue, is a motivation behind the grammaticalization of some of the NP-strategies.
3.2 Two-unit constructions: Pronominal constructions
his category includes grammaticalized constructions of unspeciied pronouns,
consisting of two pronominal expressions that ill the two argument positions of
the predicate. Pronouns in this context are taken in the general sense of free forms
whose interpretation depends on another referential element, namely, the antecedent. he antecedent in the case of the unspeciied relations is the set participating in the relation described by the verb.
his broader category of pronominal constructions can be subdivided into
several other types according to formal distinctions, related to the origins of the
components of these constructions.
3.2.1 Repetition of the same NP twice
Similar to the previous category, the relation between the participants is marked
with a repetition of NPs. Unlike in those §3.1, these constructions contain repetitions of various types of pronominal expressions.
3.2.1.1 Repetition of pronouns
he repeated elements function otherwise as pronouns in the grammar, such as
demonstratives and indeinite pronouns (as deined by Haspelmath 1997). For
example, consider the use of indeinite pronouns in the Judeo-Arabic Moroccan
dialect of Tailalt (25), proximal demonstratives in Aramaic dialects (26) and in
Mishnaic Hebrew (27) and the cardinal number “one” in other dialects of Aramaic
(28) and various other languages (Nedjalkov & Geniušienė 2007: 426), which
functions in other contexts as an indeinite pronoun.
(25) Judeo-Arabic Moroccan dialect of Tailalt:
muḥmmǝd u-musa
‘ṭaw
si
l-si
kadu
Muhammad and-Moses give PST.3.M.PL someone to-someone git
“Muhammad and Moses gave each other a git.”
(26) Biblical Aramaic (Oicial Aramaic):
wĕ-’arkubb-āt-ēh
dā’
lĕ-dā’
nāqš-ān
and-knees-PL-POSS.3.M.SG DEM.F.SG to-DEM.F.SG strike.PTCP-F.PL
“And his knees were striking one another.”
(Dan. 5:6)
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 349
(27) Mishnaic Hebrew:
᾿en dān-īn
lō᾿ ze
᾿ēt ze
NEG judge.PTCP-M.PL NEG DEM.M.SG ACC DEM.M.SG
“hey should not judge each other”
(28) Galilean Aramaic (Western Late Aramaic):
᾿innūn pǝlīg-īn
ḥdā ‘al ḥdā
they be.at.variance.M.PL one.F on one.F
“hey are at variance with each other.”
(t. Sanh. 5:4)
(y. Ḥal. 3:2)
his type of construction relies on the previous construction of a repetition of
two nominal expressions (§3.1), as pronouns appear instead of repeating the expressions themselves. (29) provides a scheme of such constructions (for a detailed
syntactic analysis, see §4.3).
(29) NPA Verb NPA => NPA — Pronouni∈A Verb Pronounj∈A
Accordingly, the historical development merely pertains to ordinary pronouns
(demonstratives, indeinite pronouns, etc.), as it is again the expected syntactic
coniguration. As the various pronouns demonstrate, they are in most cases in the
singular, which suggests that these constructions grammaticalized, as noted earlier, for unspeciied relations for sets with only two members (as in 12b). Moreover,
the strategy of repetition of nouns has a drawback (§3.1), namely, such a strategy
cannot be used when the participants of the unspeciied relation are of diferent
categories. When using pronouns, this problem is solved, as they are read in a
distributive fashion.
With pronouns marked for number and gender it is possible to capture iner
relations in the coniguration of the number of the participants, such as the distinction between (30a) and (30b) from Mishnaic Hebrew:
(30) Mishnaic Hebrew:
a. haś-śôkēr
’et hā-’ûmmān-īn
wĕ-hīṭ‘û
DEF-hire.PTCP.M.SG ACC DEF-cratsman-M.PL and-deceive.pst.3.m.pl
ze
’et ze
DEM.M.SG ACC DEM.M.SG
“If one hires cratsmen and they deceived one another.” (B. Meṣiʕa 6:1)
b. šĕttê
ḥăbûr-ôt bi-zman šem-mi-qṣāt-ān
rô’-īm
two.F.PL group-PL in-time REL-from-few-POSS.3.F.PL see.PTCP-M.PL
’ēllû
’et ’ēllû
DEM.PL ACC DEM.PL
“If two separate parties… if some members of each party are able to see
some members of the other company…”
(Ber. 7:5)
© 2014. John Benjamins Publishing Company
All rights reserved
350 Elitzur A. Bar-Asher Siegal
While the antecedents of the pronouns in both sentences represent plural entities (“cratsmen” and “two parties,” respectively), only (30b) contains a plural demonstrative. he reason is that (30a) describes a reciprocal relationship between
individuals (two cratsmen), whereas (30b) describes the reciprocal relationship
between sets (two parties). his is therefore not morphological but semantic
agreement, since the target of the agreement is controlled by the actual number
of members within each set participating in the reciprocal relation (cf. Glinert
1989: 69 in the context of Modern Hebrew and Heine & Miyashita 2008: 169–170).
On this account of the origin of such constructions, I assume that pronominal
constructions, in which the two pronominal expressions are not the same (i.e.,
pairs such as one…another in English or exad-hašeni “one-the second” in Modern
Hebrew, as discussed in §5), developed in the same way. In fact the semantics of
these constructions (see (12)) contain the distinctness requirement (x≠y); the second element in these pairs (i.e., other, second etc.) expresses this alterity.
3.2.1.2 Repetition of semantically bleached nouns
he pre-grammaticalized construction (§3.1) is likely the source for constructions
consisting of a repetition of nouns which have bleached semantically. Note the
following examples:
(31) a.
Akkadian:
innašqū
ah̬-u
ah̬-i4
kiss.ING.PST.3.M.PL brother-NOM brother-GEN
“hey began to kiss each other.”
(En. El. III132, Neo Assyrian [Standard Babylonian])
b. Amharic:
ǝrs bä-ras-aččǝn
annǝṭṭala
head in-head-POSS.1.PL ight.NEG.JUSSIVE.1PL.REC
“Let us not ight with each other.”
(Leslau 2000: 27)
Considering (31a) it is reasonable to posit the following stages of grammaticalization of the pronouns:
I.
Originally there was no single word dedicated to expressing reciprocity, and
the word ah̬um “brother” was used only in contexts in which the literal meaning of “brother” was relevant.
II. he semantics of ah̬um later bleached, and it became a pronoun.
4. he genitive is unexpected. See Bar-Asher Siegal (2011: 24 n. 7) and Von Soden (1931: 186–
187 n.1).
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages
It is conceivable that this process began when ah̬um was used in contexts such as
the following:
(32) Standard Babylonian:
bīt-u
itti bīt-i
inakkir
ah̬-u
ah̬-a
house-NOM with house-GEN hostile.DUR.3.M.SG brother-NOM brother-ACC
idâk
kill.DUR.3.M.SG
“Family will turn hostile against family, brother will kill brother.”
(KAR 148:13)
his relatively late example shows a context in which the original meaning of
“brother” is still relevant, but a more general translation is also possible: “one will
kill the other.” At an early stage, ah̬um was most likely to have been used only for
people, retaining the gender distinction, as in the following example with ah̬āt
“sister”:
(33) ah̬āt ah̬āt-am ina puzr-i
awāti umma;
sister sister-ACC in secret-GEN word DSM
“You (F.PL) are saying secretly to each other…”
(Kraus AbB 1 135:22)
Once grammaticalized, ah̬um can refer to animals (34a) and inanimate objects
(34b):
(34) a.
Neo Assyrian:
[šumma er-û]
ah̬-u
ah̬-i
issū=ma
COND eagle.PL.NOM brother-NOM brother-GEN call.DUR.3.MP =and
“When eagles call each other…”
(CT 39, [Plate] 25, Sm1376:9).5
b. Old Babylonian:
ah̬-um
ah̬-am
idris
brother-NOM brother-ACC press.PST.3.SG
“[two gates…] one presses the other”
(YOS 10 24:7)
Another somewhat expected development is the lack of gender agreement. A clear
example with a feminine antecedent has yet to be found.6
5. his example is in Standard Babylonian, a literary dialect that imitates the classical period of
Old Babylonian.
6. Due to the rarity of discourse situations with solely feminine referents, it is hard to ind such
examples. here is, however, a possible relevant example from Middle Babylonian (Bar-Asher
Siegal 2011: 30).
© 2014. John Benjamins Publishing Company
All rights reserved
351
352 Elitzur A. Bar-Asher Siegal
3.2.2 Extension of a pronominal construction
Similar to what has been suggested in (19), (12a) can be expressed also in the following way:
(35) [For a given set of individuals denoted by NP, every individual is part of a
pair in which] — someone R someone
he semantics of these pronouns, as described earlier, is ‘de facto’ an extension
of propositions with one set of unspeciied referents, where various indeinite
pronominal expressions are used. In fact, once again, the formula “someone R
someone” better represents (12b), which is the case when the unspeciied relation
is held between two participants only. In light of this, a construction comprised
of two indeinite pronouns is expected, as in (25) in the Judeo-Arabic Moroccan
dialect of Tailalt.
Similarly, in Biblical Hebrew the irst component ’īš “man” is a regular indeinite pronoun:
(36) Biblical Hebrew:
’īš kî
yiddōr
neder la-Yahwe
man when vow.IPF.3.M.SG vow to-the.Lord
“When someone makes a vow to the Lord…”
(Num. 30:3)
he extension of the use of the indeinite pronoun is accompanied by the addition
of a second diferent correlative component (marking the distinctness requirement): either ’āḥīw “his brother” (as in (37)) or rē‘ēhû “his fellow,” both nouns with
a genitive suix pronoun referring to the irst component, i.e., ’īš “man.”
(37) wĕ-’īš
’āḥ-īw
lō’ yidḥāqû
and-man brother-POSS.3.M.SG NEG prod.IPF.3.M.PL
“hey do not jostle each other.”
(Joel 2:8)
he supplement of a correlative presumably began when ’āḥīw/rē‘ēhû, otherwise
nouns with full lexical content, were used in contexts such as:
(38) wĕ-hirgû
’īš ’et ’āḥ-īw
wĕ-’īš
’et
and-kill.IMP.2.M.PL man ACC brother-POSS.3.M.SG and-man ACC
rē‘-ēhû
wĕ-’īš
’et qĕrōb-ô
fellow-POSS.3.M.SG and-man ACC neighbor-POSS.3.M.SG
“Each of you should kill his brother, his friend and his neighbor.”
(Exod. 32:27)
(38) provides a context in which the original meaning of “brother” is relevant.
Since the list of “brother,” “friend” and “neighbor” seems to relect “everyone,” it
was probably perceived, at least in other contexts, as a general expression for such
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 353
a relation. Consequently, at irst only ’īš was used as an indeinite pronoun with
various other participants (“brother,” “fellow,” etc.), but with time ’āḥīw/rē‘ēhû also
became part of such expressions and started to function as indeinite pronouns.
At an early stage such expressions were likely used to designate only human
beings, as in Middle Aramaic.7 Biblical Hebrew, however, exhibits a further stage
in the grammaticalization process, as these expressions also refer to animals (39)
and even inanimate objects (40):
(39) way-yittēn
’īš bitr-ô
liqra’t rē‘-ēhû
and-give.IPF.3.M.SG man cut-POSS.3.M.SG towards fellow-POSS.3.M.SG
“And he arranged the halves (of a heifer, a goat and a ram, each three years
old, along with a dove and a young pigeon) opposite each other.”
(Gen. 15:9–10)
(40) ḥămēš ha-yĕrī‘-ōt
tihĕyenā
ḥōbĕr-ōt
’iššâ
’el
ive.F.PL DEF-curtain-F.PL be.IPF.3.F.PL join.PTCP-F.PL woman to
’ăḥōt-āh
sister-POSS.3.F.SG
“he ive curtains should join each other.”
(Exod. 26:3)
3.2.3 Quantiicational constructions
his category includes constructions with at least one quantiier as one of the elements constituting the construction.
3.2.3.1 Constructions with a universal quantiier
Previous studies, such as Plank (2008) and Haas (2010), focus mostly on constructions with a universal quantiier, such as each in English, as one of the elements. It
is not diicult to propose that sentence (41b) derives from (41a):
(41) a. Each one of them saw the other.
b. hey saw each other.
Only in the early stages of English was a reanalysis of each as part of the anaphor
possible, when each as a quantiier could be loated and appear immediately before
other. his is illustrated in the following example, cited by Haas (2010: 70):
(42) And there vppon they cast eche to other their gloves…
(Helsinki Corpus, ME IV [1420–1500])
7. he Aramaic translator of the Pentateuch, Onkelos, regularly translated Biblical expressions
containing ’īš “man” and ’āḥīw “his brother” with the corresponding Aramaic words: gǝbar and
᾿ā᾿ḥohī. he translation deviates from the Hebrew, and uses the cardinal number “one,” when
the participants of the relations are inanimate objects (for example, in the translation for sentence (40)). A similar phenomenon is found in Mishnaic Hebrew (Bar-Asher Siegal 2012: §4.3).
© 2014. John Benjamins Publishing Company
All rights reserved
354 Elitzur A. Bar-Asher Siegal
While (41a) and (41b) are likely related historically, at the synchronic semantic level they are diferent. Sentence (41a), with its universal quantiier, is not an unspeciied construction, as deined in (12). hus, the similarity between the sentences
holds properly only in sentences with two sets of participants in the reciprocal
relation.8 With a larger number of participants, while the each other construction
as in (41b) allows for ‘weak distributivity,’ constructions such as those in (41a) do
not allow for such a reading (Williams 1991). Compare (43a) and (43b):
(43) a. he children were kissing each other.
b. Each child was kissing the other.
While (43b) requires strong distributivity, that every child was a kisser of each and
every other child, (43a) could be true even with weak distributivity, as it describes
events in which some of the children were only the receivers of kisses, but not necessarily kissed by all of the other children. hus, the grammaticalization of a oneunit construction such as each other from constructions with a universal quantiier
involves a semantic shit.
Such a construction appears to be unattested in Semitic; the discussion of their
origin, however, raises the question whether the discussed diachronic changes involve a semantic shit.
3.2.3.2 Constructions with partitives
Although not a quantiier in the strict sense, the closest strategy found in the
Semitic languages is the repetition of a partitive in Arabic (see (14)). his construction should probably be understood similarly to constructions with a repetition of pronouns (§3.2.1.1). Once again this development can be understood in
light of nominal constructions (§3.1):
(44) NPA Verb NPA => NPA — some-of-NPA Verb some-of-NPA
Similar to the explanation for the constructions with the repetition of indeinite
pronouns, (44) is a natural reading of (12a) and in fact this is the only construction
that indicates that the set whose members participate in the unspeciied relations
may contain more than two members. Accordingly, a reciprocal reading can be
attained because, logically speaking, the quantiier some is used also to imply the
exhaustion of the set (i.e., when it is more natural to use the quantiier all).
8. Dougherty (1974) and Heim et al. (1991: 70), who argue for a synchronic relation between
the constructions in (41), admit that this relation is demonstrated properly in sentences with
only two sets of participants.
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 355
3.3 Intermediate summary and a further observation
here are two possible sources for the two-unit constructions:
(45) a.
hey evolved from the repetition of two nominal expressions. Instead
of a repetition of the nouns themselves, pronouns appear (this category
includes repetitions of partitives, original pronouns and bleached
nouns).
b. Or, they are a further development of a pre-existing indeinite pronoun,
either by repeating it (Moroccan Arabic, in (25)), or with the addition
of a correlative (Biblical Hebrew, in (37)). hus, schematically, there are
two constructions:
I. someone Verb someone
II. NPA — Pronouni∈A Verb Pronounj∈A
here is a clear diference: while (II) requires an antecedent, (I) does not.
Consequently, the expectation is that in contexts without an antecedent, (II) will
be unavailable and (I) will be available.9 his diference is crucial to the viability of
these constructions with casuistic laws. Laws by their nature are impersonal and
therefore appear without antecedents, as they state possible relations between two
unspeciied members of the legal community. Accordingly, only constructions of
type (I) are expected to be available for expressing such laws. hat this expectation is indeed met can be seen by comparing Biblical and Mishnaic Hebrew. As
illustrated in (8), Biblical Hebrew has type (II), and the pronominal construction
is used both in reciprocal constructions and in casuistic laws. In Mishnaic Hebrew,
where type (I) is the regular construction for expressing unspeciied relations, repetition of the demonstratives (27, 30a–b) is not used in casuistic laws. An alternative construction is used for this in Mishnaic Hebrew (Bar-Asher Siegal 2012).
4. he diachronic development from a two-unit to a one-unit
construction
4.1 Previous proposals
Since it is oten possible to trace the phonological derivation of the one-unit construction from the two-unit construction in some languages, while an explanation
for the opposing direction is not as readily available, I propose this as the direction
9. In constructions (II) and (I), when there is an antecedent, it restricts the domain for which
the indeinite pronouns are relevant.
© 2014. John Benjamins Publishing Company
All rights reserved
356 Elitzur A. Bar-Asher Siegal
of development and look for motivations. As noted, this is the process found in the
history of Akkadian, Arabic and Aramaic (see (13)–(15) above).
he only explanation I am aware of is Visser’s proposal (1963: 445), reformulated by Haas (2010: 83–86), namely that the one-unit formula is a reduced clause
functioning as an aterthought:
(46) a. he knights hugged; one hugged another.
b. he knights hugged1; one e1 another.
c. he knights hugged one another.
While this is plausible, it is likely only for a language like English, in which verbal
reciprocity is unmarked morphologically. In other languages, the form of the verb
in both clauses is not the same, and therefore the omission of the second verb is
less likely. Moreover, as discussed in Siloni (2002, 2012) and Bar-Asher (2009), an
equivalent of the third stage (46c) is ungrammatical in many documented languages (based on the results of Nedjalkov & Geniušienė’s 2007 cross-linguistic
questionnaire) since they are monovalent, as illustrated in (47b) with an example
from Modern Hebrew:
(47) a.
ha-᾿abir-im hitnašqu
exad nišeq
et ha-šeni
DEF-knight-PL kiss.RECP.PST.3.PL one.M kiss.PST.3.M ACC DEF-second.M
he knights hugged; one hugged another
b. *ha-᾿abir-im hitnašqu
exad et ha-šeni
DEF-knight-PL kiss.RECP.PST.3.PL one.M ACC DEF-second.M
Intended reading: “he knights hugged each other.”
hus, an alternative proposal is needed for families, like Semitic, that encode reciprocity verbally. I begin by noting the diferent syntactic behaviors of the one-unit
and two-unit constructions in Akkadian, similar to all other languages that have
both types of constructions.
4.2 he distribution of one-unit and two-unit constructions
While the Akkadian one-unit anaphora ah̬āmiš appears in the same clause as its
antecedent, the two-unit constructions, with the repetition of ah̬um, are used only
when the antecedents are not part of the clause. hus, the subjects of clauses with
ah̬āmiš are the parties participating in the reciprocal relations. hey appear either
separately (and coordinated) (48a), or as a group whose members participate in
such relations (48b):
(48) a.
šumma surd-û
u ārib-u
itti ah̬āmiš ṣalt-a
COND raven-NOM and falcon-NOM with RECP ight-ACC
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 357
īpušū=ma
do.PST.3.M.PL=and
“If a falcon and a raven ight, and…”
b. māt-āt-i
ana ah̬eiš iqabbûni
country-F.PL-OBL to RECP say.DUR.3.M.PL
“he countries say to each other…”
(CT39 30:35)
(Craig ARBT 1 26:8 NA)
With the two-unit constructions, the nominal representations of the participants
in the reciprocal relations are not the subject of the clause. In most cases, the irst
element between the two pronominal expressions is the subject in the nominative
(ah̬um), while the second element ills the slot of the other argument in the sentence and appears in the appropriate case, as in (49–50):
(49) ah̬-um
eli
ah̬-im
mimma ul īšu
brother-NOM toward brother-GEN something NEG have.PST.3.SG
“No one has a claim upon the other.”
(PBS 8/1 81:17, Old Babylonian)
(50) atta u nakir-ka
taṣṣabbatāma
ah̬-um
you and enemy-2MSG.POSS get.into.ight.DUR.2.M.SG brother-NOM
ah̬-am
ušamqat
brother-ACC destroy.DUR.3.SG
“You and your enemy will get into a ight, and one will destroy the other.”
(YOS 10 50:8, Old Babylonian)
Although antecedents are not the subject in two-unit constructions, they can still
appear as the topic:
(51) atta u nakrī-ka
ah̬-um
ina pāni ah̬-im
you and enemy-POSS.2.M.SG brother-nom from brother-gen
udappar
withdraw.DUR.3SG
“You and your enemy will withdraw from each other.”
(YOS 10 47:81, Old Babylonian)
he fact that “you and your enemy” in (51) functions as the topic and not as the
subject is indicated by the form of the verb (3.m.sg rather than 2.pl), which agrees
with the reciprocal pronoun. his can be demonstrated by contrasting (52) with
(53). With a verbal encoding of the reciprocity, the verb in (53) is 1.pl, unlike the
singular verbal adjective in (52):
(52) ištu pānānumma anāku u kâti awīl-um ana awīl-im paqid
since formerly
I
and you man-NOM to man-GEN VADJ.trust
“For some time now our relationship had been such that one trusted the
other [lit. I and you trusted man to man].” (TCL 17 31:8 f, Old Babylonian)
© 2014. John Benjamins Publishing Company
All rights reserved
358 Elitzur A. Bar-Asher Siegal
(53) inūma anāku u kâti ina GN nuštāt-û
when I
and you in GN see.RECP.PST.1.PL-SBJV
“When you and I saw each other in GN…” (PBS 7 108:10, Old Babylonian)
While the components of the two-unit constructions, ah̬um-ah̬am, ill the argument positions of the verb (subject, object and oblique), the one-unit anaphora,
ah̬āmiš, occupies all the arguments selected by the verb but never the subject position.
In conclusion, the two types of constructions difer not only in the number
of components they comprise, but also in the way each construction expresses
unspeciied relations with respect to the following characteristics:
(54) a. In two-unit constructions:
i. Each pronominal expression ills a diferent argument selected by
the predicate (subject, object, etc.).
ii. Each of the arguments selected by the predicate is illed with a
pronominal expression.
iii. he participants of the relation themselves, if mentioned, are not
part of the grammatical relations and therefore may only appear as
the topic of the sentence.
b. In one-unit constructions:
i. he anaphora co-occurs with the sets that participate in the
reciprocal relations.
ii. he participants in the relation hold the subject position, and
the anaphora occupies the position of the other argument of the
predicate.
hus, an analysis arguing that one-unit constructions derive historically from the
two-unit constructions should explain not only the merging of the two forms into
one, but also the shit in grammatical relations between the various components
of the sentence.
In the previous Akkadian examples of constructions with two independent
pronominal elements, ah̬um occupies the subject position and the verb, expectedly, is singular. Rarely, however, there are plural verbs:
(55) a.
ah̬-um
ah̬-am
lā ibaqqarū
brother-NOM brother-ACC NEG raise.a.claim.DUR 3.M.PL
“None should raise claims against the other.”
(YOS 8 99:19f, Old Babylonian)
b. ah̬um
ah̬am
ina
mê
lā
brother-NOM brother-ACC concerning water.OBL NEG
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 359
udarrasū
treat.opressively.DUR.3.M.PL
“One should not treat the other oppressively on account of the water.”
(TCL 7 23:29, Old Babylonian)
Such Akkadian sentences evoke the standard construction of Biblical Hebrew,
with a plural verb despite the fact that the NPs in all syntactic positions are morphologically singular:
(56) way-yaḥăziqû
’īš bĕ-rō’š
rē‘-ēhû
and-hold.IPF.3.M.PL man in-head.of fellow-POSS.3.M.SG
“Each man grabbed his opponent by the head.”
(2 Sam. 2:16)
hus, the reciprocal pronouns in Biblical Hebrew (and many other languages) are
diferent from both standard ah̬um-ah̬am and ah̬āmiš. While at irst the various
combinations seem to be instances of a two-unit construction, verbal agreement is
(almost) always plural and thus the irst pronoun does not seem to hold the subject
position.
Such hybrid constructions are known from other languages. Furthermore,
with this type of construction in languages with morphological case, both the
topic and the irst element of the reciprocal pronouns are in the nominative, as is
the case in Icelandic:
(57) Þeir
elska
hvor
annan
they.NOM love.3.PL.IND one.NOM other.ACC
“hey love each other.”
(Everaert 1999: 69)
I propose irst a historical account for the three types of constructions in the following order:
(58) two-unit construction > hybrid construction > one-unit construction.
In principle, as will become clear, since the inal stage is the result of a reanalysis,
the two-unit constructions can already be reanalyzed and turned into one-unit
anaphors without an intermediate stage.
4.3 he diachronic development of the NP-strategies
I suggest the following structures for the three stages of development of the NPstrategies:
{NP1, NP2…NPn.NOM} verb.SG reciprocal-pronoun1.NOM.SG reciprocal-pronoun2.ACC.SG
Topic
Subject
Object
Figure 2. Stage I: Two-unit constructions.
© 2014. John Benjamins Publishing Company
All rights reserved
360 Elitzur A. Bar-Asher Siegal
he topic is in brackets as it is optional. When it does appear, though, it is in the
nominative, which may be attributed either to the fact that in Semitic languages
with morphologically overt cases, the case for the topic (which is never the subject) is also nominative; or alternatively, it could also be attributed to the fact that
it has the same referent as the subject of the sentence. his explanation is valid
beyond Semitic and may account for similar occurrences cross-linguistically.10
In the second stage, the only formal change is that the verb is plural:
{NP1, NP2…NPn.NOM} verb.PL reciprocal-pronoun1.NOM.SG reciprocal-pronoun2.ACC.SG
Topic
Subject
Object
Figure 3. Stage II: Hybrid constructions.
While one might assume that the plural agreement relects a reanalysis of the
nominative topic as the subject, it seems to have come about for a diferent reason. his could illustrate the well-known phenomenon of semantic rather than
morphological agreement (Corbett 2006: 155–160); in the case of reciprocals, this
confusion is due to the fact that part of the deinition of a reciprocal relation is that
more than one set occupies the subject position. As shown in (55), this phenomenon is already evident in Akkadian. Similarly, in Biblical Hebrew the verb is in
the plural, even in the case of a repetition of the same noun phrase in two diferent
positions in the sentence, where there is clearly no other antecedent (§3.1).
(59) ki gibbôr bĕ-gibbôr kāšālû
as warrior in-warrior stumble.PRF.3.M.PL
“One warrior will stumble over another.”
(Jer. 46:12)
It is impossible to explain the plural verb in (59) as a reanalysis of the topic as the
subject of the clause. Hence, the motivation behind the plural form of the verb
must be seen as semantic and not as a relection of a diferent syntactic relation in
the clause. his hybrid construction is likewise found in Standard Arabic in sentences with two-unit constructions, and we note a similar development in Arabic.
Compare (60) with (61). In both, the speakers participate in the reciprocal relations. In Modern Standard Arabic the verb is in the 1.PL:
(60) linusā‘ida
ba‘ḍ-u-na
ba‘ḍ-an
assist.SBJV.1.C.PL some-NOM-POSS.1.C.PL some-ACC.INDF
“Let us assist each other.”
(ar-ar.facebook.com/tohelpeachother)
10. Everaert (1990–1: 298–300) considers this use of the nominative in let dislocation structures to be evidence for the hypothesis that in Icelandic (and in other languages), the nominative case is “lack of case,” since this is a position of non-case-assignment. his matter deserves
independent examination.
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 361
In the classical period, in the Qur’an, the verb agrees with the ba‘ḍ- and is 3.m.sg:
(61) rabb-anā
stamta‘a
ba‘ḍ-u-na
bi-ba‘ḍ-in
lord-POSS.1.PL 3.M.SG.PST.make.proit some-NOM.1.PL in-some-GEN.INDF
“Our Lord! We made proit from each other.”
(6: 128)
Example (61) from the classical period exempliies stage I, with ba‘ḍ- as the grammatical subject. In example (60), from Modern Standard Arabic, the speakers, in
a reciprocal relation, are the grammatical subject (stage III, below), or at least in
control of semantic agreement (stage II).
In all the above examples, the verb either proceeds or follows both elements of
the construction. he Jewish Neo-Aramaic Dialect of Challa furnishes an example
with a plural verb form standing in between the two elements:
(62) xa lu
mšaboḥe ’ǝl-xé
one COP.3.PL praise.PRF to-one
“One is praising the other.”
(Fassberg 2010: 48)
Since the two elements of the construction are separated, one can assume that one
part is the subject and the other the object; thus this is another example of stage II
and accordingly another instance of semantic agreement.
A similar development can be traced in the history of Italian.11 he 13th–14thcentury Italian expression l’uno l’altro occurs as a two-unit construction, with a
verb in the singular form (thus, l’uno is the subject and l’altro the object). he two
elements can be separated, or they may occur together in post-verbal position.
(63) a.
Quando lo amico
ama
la sua
amica
per
when DEF friend.M.SG love.PRS.3SG DEF POSS.3.M.SG friend.F.SG for
dilettazione, e quella ama
lui per utilità,
non
pleasure
and DEM.F.SG love.PRS.3.SG him for convenience NEG
ama
l’uno
l’altro
per diritto bene
love.PRS.3SG DEF=one-M.SG DEF=other-M.SG for right good
“When the friend loves his girlfriend for pleasure and she loves him out
of convenience, the one does not love the other for the right reason.”
(Tesoro volg (XIII))
b. perche queste due cose
seguita
l’una
because DEM.F.PL two thing.FEM.PL follow.prs.3.sg DEF=one.f.sg
11. he data (and the translations) are taken from Vezzosi (2010). Vezzosi follows Plank (2008)
in assuming that un in Italian is similar to the English distributor each, a similarity that was
never shown.
© 2014. John Benjamins Publishing Company
All rights reserved
362 Elitzur A. Bar-Asher Siegal
l’altra
igualmente.
DEF=other.f.sg equally
“Because these two things follow each other equally…”
(Andrea Cappellano (XIV))
here are only a few examples in texts from the 13th century of two-unit constructions demonstrating signs characteristic of stage II, as l’uno l’altro occurs with a
plural verb; this, though, only occurs if the set participating in the relation is in
the topic position, as seen in example (64). Such a pattern, however, becomes very
frequent in 14th-century texts.
(64) La prima ragione si
è,
che le cose
della natura
DEF irst reason REFL be.3.SG REL DEF thing.F.PL of.DEF nature
generano
l’ una
l’ altra …
generate.PRS.3PL DEF one.FEM.SG DEF other-F.SG
“he irst reason is that the things of nature each generates the other.”
(Egidio Romano (volg., 1288))
Finally, one encounters a shit to a construction with the following changes: 1) the
verb is regularly found in a plural form; 2) at the end of the 14th century, the elided
form for the irst element gradually began to appear: l’un l’altro; 3) with complements, the preposition governs only the second element:
(65) a.
perché facciamo
l’un
l’altro
tapini …
because make.PRS.1PL DEF=one.M.SG DEF=other.M.SG miserable.M.PL
“Because we make each other miserable…”
(Bioardo Lib. 1 can. 2.17 (‘400))
b. il veder la miseria l’un
dell’altro
DET see.INF DEF misery DEF=one-M.SG of.DEF=other.M.SG
e l’aversi
compassione l’un
all’altro
and DEF=have.INF.REFL pity
DEF=one.M.SG to.DEF=other.M.SG
“Seeing each other’s misery and pitying each other…”
(Firenzuola Ragionamenti Giorn. 1 nov. 1.4 (500))
his last stage is similar to contemporary Italian, discussed in §5.
he last stage is a typical example of a topic reanalyzed as a subject.
Consequently, the two separate pronouns are conceived of as one unit, namely, the
one-unit reciprocal anaphora:
{NP1, NP2…NPn.nom}
Subject
verb.PL
reciprocal-pronoun1.reciprocal-pronoun2.acc.du/pl
Object
Figure 4. Stage III: One unit-construction (the reanalysis).
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 363
Once there is another subject in the sentence, a ‘reciprocal-pronoun1’ can no longer be the subject and so it must be analyzed, like ‘reciprocal-pronoun2’, as part of
an expression that produces the reciprocal meaning. As a result, both of the original pronominal expressions ill the same syntactic position. It is more accurate to
describe it as follows, indicating that at this stage there is only one element illing
the object position:
{NP1, NP2…NPn.nom}
Subject
verb.PL
recp.acc.pl
Object
Figure 5. Stage III: One unit-construction with an anaphora.
Clearly there are various changes that are taking place in the inal stage:
(66) a. A shit from two-unit to one-unit (univerbation)
b. A change of number for the pronominal expressions (singular => plural)
c. A change of cases (loss of the nominative element)
In previous analyses (Plank 2008, Haas 2010) the univerbation seems to be led by
the formal aspect (and not by the meaning or the syntax), and the reanalysis of the
two units as one is merely due to their syntactic proximity. In contrast, I argue that
once the topic is reanalyzed as the subject of the clause, this reanalysis is a necessary result, since the two units hold the same syntactic position. Accordingly,
while the previous analyses hold that univerbation is the only characteristic that
marks a construction as stage III, our current line of reasoning maintains that
each characteristic of this stage (66a–c) serves as a sign for a construction to be
considered as stage III. Even one indication is enough; it is expected that not all
will be formally expressed at once. I now show the formal expressions of each of
the characteristics of stage III, and how our analysis changes the perception of data
from several languages.
4.3.1 Univerbation
A shit to stage III anticipates an increasing frequency of use of these constructions, and such conditions lead to phonological reduction (Bybee 2003: 615–617,
inter alios). Indeed cross-linguistically one encounters phonological fusions by
which two-unit constructions become one-unit constructions. As both forms occupy one syntactic position, they are primed for such a fusion; such a shit its the
phenomenon of univerbation. Andersen (1987) subdivides this process as follows:
morphological univerbation (loss of morpheme boundaries), prosodic univerbation (stress shit) and segmental levelling (phonological reduction).
Among the Semitic languages, morphological univerbation, which also
includes prosodic and segmental changes as well, is well attested. In the Old
© 2014. John Benjamins Publishing Company
All rights reserved
364 Elitzur A. Bar-Asher Siegal
Babylonian texts from Susa, for example, the one-unit anaphora is ah̬mah̬am/im
or ah̬māmam/im. he former is clearly a fusion of the older two-unit construction
(a repetition of ah̬um, 13a). he elision of the second /h̬/ in the second form seems
to be a result of haplology: ah̬mah̬am > ah̬mah̬ a̬ m > ah̬māmam, as seems to be the
case also with the Syriac form (15b) *ḥadḥad > ḥadḥ ad+ē > ḥǝdādē (with the addition of the plural marker; see §4.3.2). In both instances, the loss of the consonant
results in a lengthening of the next vowel.
In addition to univerbation, Semitic languages exhibit two other ways in
which two-unit constructions become one-unit constructions.
1) Deletion — elision of one of the elements of the two-unit construction:
At stage III, two forms ill one syntactic position. Consequently an elision of one of
these forms, as in the Standard Arabic example (14c), is easily explained12:
(67) ba‘ḍ -u-humā li ba‘ḍ-an > li- ba‘ḍ-him
2) Frozen forms:
Such formulae seem to have two separate elements, but synchronically do not
function as two-unit constructions. In Amharic the pronominal expression consists of the repetition of the element ras “head” (31b) in the following formula:
(68) ǝrs bä +ras + pronominal (plural) suix agreeing with the subject
he element bä was originally the preposition bä “in”; however, this function has
been lost, and it appears in all constructions, regardless of their semantics.
Similarly, the formula in Arabic, as expressed in (14d), seems to be an intermediate, frozen stage in the shit from two-unit to one-unit constructions, since
only the irst element has inlection that matches the syntax of the sentence. A
similar phenomenon occurs in Kannada (Bhat 1978: 44–45). On my analysis, frozen expressions in languages where fusion or elision did not occur, although they
do appear as the historical stage with two separated elements, synchronically do
not relect such grammatical relations. We explore this in §5.
It seems to be signiicant whether the two elements of the two-unit construction are separable or inseparable (cf. Haspelmath 2007: 2113). Earlier, the fact
that the two components in the Neo-Aramaic dialect of Challa are separable (62)
served as an indication of being at stage II; and in §5, the inseparability of the
two components in Hebrew would indicate the shit to stage III. he problem of
treating inseparability as a criterion lies in the nature of ancient languages. Usually
there is not enough data to determine whether the two components were indeed
12. Bar-Asher Siegal (2011) proposes an example of a similar deletion in Akkadian.
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 365
inseparable or simply appeared next to each other in all attested examples; we explore this in §5 for contemporary languages.
4.3.2 Insertion of agreement marker
According to our analysis the shit to a one-unit construction involves a shit in
the grammatical number of the pronominal expression. his trait is occasionally
relected by the development of an agreement feature, and it will be demonstrated
through the relevant expressions in Mehri. Occasionally Mehri expresses unspeciied relations with a repetition of ṭayt “one”:
(69) yeṭḳawḳ
ṣǝwayr ṭayt ð̣ar ṭayt
throw.IPFV.3.M.SG stone.PL one upon one
“hey throw stones at one another.”
(Johnstone 16:2)
he common one-unit anaphora ṭāṭīday- likely resulted from a fusion of the repetition of ṭayt. Importantly for the current discussion, this one-unit anaphora
agrees in number (plural or dual) with its antecedent (Rubin 2010: 50–51):
(70) a.
’āmǝrō
hǝ-ṭāṭīday-hi
say.PRF.3.M.DU to-RECP-DU
“hey (two) said to each other…”
b. tōli fǝhēmǝm
ṭāṭīday-hǝm
then understand.PRF.3.M.PL RECP-3.M.PL
“hen they understood each other.”
(Johnstone 4:17)
(Johnstone 59:14)
hus, once the one-unit anaphora grammaticalized, it became similar to other
pronouns and has the same nominal declension. hus, it resulted in a dependency
between the anaphora and the subject, hence the subject-pronoun agreement.
Similarly, the ending -ē in the one-unit anaphors ḥǝdādē (Syriac) and hǝdāde
(Jewish Babylonian Aramaic) in the Late Eastern Aramaic dialects, originating
from a repetition of ḥad (stage I), is probably the result of an additional agreement
feature. It is either a vestige of a dual form (-ay > -ē), or the usual Late Eastern
Aramaic plural marker -ē:
(71) ḥad ḥad > *ḥadḥad > *ḥadḥad > *ḥǝdād > [ḥǝdād+ē =>] ḥǝdādē
Bar-Asher Siegal (2011: 31) proposes a similar explanation for the Akkadian oneunit anaphora. he deletion of one of two components of the two-unit construction and an addition of a plural marker is known from Finnish, too (König &
Kokutani 2006: 281).
© 2014. John Benjamins Publishing Company
All rights reserved
366 Elitzur A. Bar-Asher Siegal
4.3.3 Change of case
As noted, Standard Arabic has a construction with two elements and only the irst
has case marking (14d). he second is syntactically frozen and always caseless. In
Russian, the irst is frozen, as the nominative form drug appears even when it is
unexpected:
(72) On
ne otliča-et
zolot-o
i med’
he.NOM NEG distinguish.IPFV-3.SG.PRES gold.N-SG.ACC and copper.F.SG.ACC
drug
ot drug-a
friend.NOM from friend-GEN
“He does not distinguish gold and copper from one another.”
(Knjazev 2007: 688)
In Kirghiz (Nedjalkov 2007b: 156) the reciprocal pronoun generally marks case on
the second element; sometimes, though, the case marking is on the irst element.
Occasionally the personal-possessive marker is added to both elements and the
postpositions are usually inserted between the elements.
(73) a.
b.
c.
d.
biri biri-Ø-n (Ø = 3.pl, -n = acc)
biri-Ø-n biri
biri-biri-biz-di (-biz- = 1.pl, -di = acc)
biri-biz-di biri-biz
Icelandic provides another interesting phenomenon, where there is only a shit
of case. In addition to the two-unit construction (in 57) with each pronominal
expression in a diferent case, there is another variant (hráinsson 1979: 129 n. 34)
with both elements in the accusative:
(74) a.
Þeir
elska
hvor
annan
they.NOM love.3.PL.IND one.NOM other.ACC
“hey love each other.”
b. Þeir
elska
hvorn annan
they.NOM love.3.PL.IND one.ACC other.ACC
“hey love each other.”
(Everaert 1990–1: 283 ex. 14)
hus, in all these constructions both elements share the same syntactic position,
and it is relected merely with the formal expression of the grammatical relations.
Finally, a priori fusion or deletion that creates the one-unit anaphora is more
likely if ‘reciprocal-pronoun1’ and ‘reciprocal-pronoun2’ are juxtaposed, and is
more plausible in verb-inal (SOV) languages and verb-initial (VSO) languages.
Among the Semitic languages with one-unit anaphors, (non-literary) Akkadian
is a strict verb-inal language (with SOV as the unmarked order) and Arabic is
a VSO language. Syriac also has VSO tendencies, despite having free word order
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 367
(Nöldeke 2001: 258–259). Mehri, though, has free word order. However, the default word order at the time of the emergence of a one-unit anaphora is not always
known; however, further cross-linguistic typological studies will be able to determine the correlation between word order, the shit to stage III and the type of
phonological developments involved.
he signiicance of this goes beyond its historical scope. As shall become clear,
stage I and stage III relect two types of constructions, with two distinct syntactic
structures. Previous studies of the NP-strategies have focused on the second type
(stage III), which resulted in various statements about reciprocal constructions —
statements which if true can only be true about this particular type of reciprocal
construction. hese are statements such as 1) reciprocal pronouns are never the
subject of the sentence (Everaert 1990–1, Nedjalkov 2007b: 154, Brame 1977: 387–
390 and Koster 1987 beyond the framework of Government and Binding) and 2)
reciprocal pronouns are grammatically plural (Fiengo & Lasnik 1973, Kamp &
Reyle 1993). hese observations motivated the analysis of such pronominal expressions in Government and Binding as anaphors (Chomsky 1981). Like relexive
pronouns, so must these reciprocals receive their reference from the same clause
(and with a particular syntactic coniguration). Unlike relexives, however, reciprocal pronouns supposedly have plural antecedents (Fiengo & Lasnik 1973). Even
if these observations are correct (and this has been disputed; see Everaert 1999,
Haas 2010: 18–19), our study shows their relevance for expressions in the one-unit
constructions only (as Beletti 1982 notes for Italian). In contrast, two-unit constructions, within this framework, function as pronouns and not as anaphors (cf.
König & Kokutani 2006). hus, there is no necessary relation between the semantics of reciprocal constructions and a speciic syntactic coniguration.
5. he Modern Hebrew and Italian constructions as frozen formulae
My key conclusion is that the shit to stage III is not merely at the formal level (univerbation); it is, rather, irst and foremost a shit in terms of the grammatical relations between the components of the constructions. his section illustrates how a
better understanding of the various aspects of the grammatical changes assists in
recognizing frozen forms. Focusing on constructions from Modern Hebrew and
Italian, I argue that despite appearing to contain two-unit pronouns (consisting of
the elements exad “one” hašeni “the second” and l’un “the one” l’altro “the other,”
respectively), structurally speaking, these are in fact one-unit anaphors. he argument for this claim relies on the fact that these constructions exhibit other characteristics of stage III.
© 2014. John Benjamins Publishing Company
All rights reserved
368 Elitzur A. Bar-Asher Siegal
Halevy (2011) has noted that in Hebrew the two elements are in one stress
unit, which suggests incipient grammaticalization. She claims, however, that since
there is no fusion, and since diferent elements can be inserted between the two
components of the construction, the grammaticalization process therefore has not
yet been completed (she, however, never deines what exactly is the relevant process of grammaticalization.)
However, in these constructions, the verb must be plural. his indicates that
Modern Hebrew is at least at stage II, if not stage III, in the development of the
NP-strategies. It is now possible to recognize a shit to this stage based on other
indications.
Languages may simultaneously have constructions of diferent types, as demonstrated by the following pairs:
(75) Modern Hebrew:
a. šoš ve-lea (ha)-axat ohevet
et ha-šniya
Šoš and-Lea def-one.f love.prs.f.sg acc def-second.f
b. šoš ve-lea ohavot
(a)xat et ha-šniya
Šoš and-Lea love.PRS.F.PL one.F ACC DEF-second.F
“Šoš and Lea love each other.”
(76) Italian:
a. (A proposito di quei ragazzi) l’uno
non condivide
As for
DEM.PL guy.PL DEF=one NEG share.PRS.3.SG
le idee dell’altro
DEF idea.PL of=DEF=other
b. Quei ragazzi non condividono le idee l’uno
dell’altro
DEM.PL guy.PL NEG share.PRS.3.PL DEF idea.PL DEF=one of=DEF=other
“he guys do not share each other’s ideas.”
An obvious diference is that the verbs in the (a) sentences are singular, while
those in the (b) sentences are plural. I now highlight a structural diference between the constructions. Consider the syntactic diferences between the various
pronouns in Hebrew:
(77) a.
odadnui
et ha-ylad-imj lelamed et aṣmamj
encourage.PST.1.PL ACC DEF-child-PL teach.INF ACC REFL.3.M.PL
matematika
math
“We encouraged the boys to teach themselves math.”
b. odadnui
et ha-ylad-imj lelamed otanui matematika
encourage.PST.1.PL ACC DEF-child-PL teach.INF ACC.1.PL math
“We encouraged the boys to teach us math.”
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 369
c.
*odadnui
et ha-ylad-imj lelamed et aṣmenui
encourage.PST.1.PL ACC DEF-child-PL teach.INF ACC REFL.1.PL
matematika
math
Intended meaning: “We encouraged the boys to teach us math.”
As shown in these sentences, only the object of the main clause controls the object
of the ininitive clause, requiring that it be a relexive pronoun. his pronoun agrees
in gender and number with its antecedent. Similarly, the following four sentences
all have the same intended meaning, as illustrated by the English translation:
(78) a.
(ha-)exad oded
et ha-šeni
lelamed
(DEF-)one.M encourage.PST.3.M.SG ACC DEF-second.M teach.INF
et aṣmo
matematika
ACC REFL.3.M.SG math
b. *(ha-)exad oded
et ha-šeni
lelamed
(DEF-)one.M encourage.PST.3.M.SG ACC DEF-second.M teach.INF
et aṣmam
matematika
ACC REFL.3.M.PL math
c. hem odedu
(e)xad et ha-šeni
lelamed et
they encourage.PST.3.PL one.M ACC DEF-second.M teach.INF ACC
aṣmam
matematika
REFL.3.M.PL math
d. *hem odedu
(e)xad et ha-šeni
lelamed
they encourage.PST.3.PL one.M ACC DEF-second.M teach.INF
et aṣmo
matematika
ACC REFL.3.M.SG math
“hey encouraged each other to teach themselves math.”
While the agreement in (78a) and the ungrammaticality of (78b) are expected, that
(78c) is grammatical and (78d) is not is surprising, considering that the formal
grammatical object of this sentence is hašeni “the second,” which is morphologically singular, as can also be seen in sentence (78a), and overtly it is the same form
in (78a) and (78c). he object of the main clause in (78c) is plural and the plurality of the relexive pronoun cannot be explained as a case of semantic agreement,
since if this were the case, (78b) should also be grammatical, as semantically the
sentences are equivalent.
Similar to (78a, c), in Italian each construction shows diferent agreement with
respect to the number of the relexive pronoun:
(79) a.
L’uno
ha incoraggiato
l’altro
ad insegnare a se stesso
DEF=one encourage.PST.3.SG DEF=other to teach.INF to REFL.3.SG
© 2014. John Benjamins Publishing Company
All rights reserved
370 Elitzur A. Bar-Asher Siegal
la matematica
DEF math
b. Si sono incoraggiati
l’un
l’altro
ad insegnare
they REFL encourage.PST.3.PL DEF=one DEF=other to teach.INF
a se stessi la matematica
to REFL.3.PL DEF math
“hey encouraged each other to teach themselves math.”
Relating this to the diachronic discussion in §4.3, we see that sentences (78c, 79b)
clearly relect stage III. In §4.3.2, I provided morphological evidence that the
grammatical number of the one-unit anaphora is overtly plural, with agreement
between the pronoun and its plural subject antecedent. Accordingly, the fact that
the relexive pronouns controlled by the reciprocal pronoun in Hebrew and Italian
are plural indicates that these languages, despite their appearances, already have
a one-unit construction at stage III. (he two-unit constructions, (75a) and (76a),
are, in fact, marked for high register in both languages.)
Several more observations can be made. (Real two-unit constructions will be
referred to as the a-sentences and frozen constructions as the b-sentences.) First,
another formal diference between the two constructions is that in the regular pronunciation of the b-sentence, the irst element exad “one.m” and axat “one.f” have
elided forms without the initial vowel: xad and xat respectively, a unique pronunciation for pronominal uses of the cardinal number “one.” Similarly, the Italian
pronominal expression uno elided and is expressed in this construction as un.
According to the current analysis, the b-sentences represent stage III, and as such,
it is only expected that this development would demonstrate the characteristics
of univerbation (§4.3.1). Having in mind Halevy’s (2011) observation of prosodic
univerbation (stress shit), these unique pronunciations of the cardinal number
“one” (e.g., exad as xad and l’uno as l’un) exhibit a further manifestation of univerbation at the segmental level (phonological reduction).
Second, in Hebrew in the a-sentences that illustrate stage I, the irst element,
exad, usually appears with a deinite article. In the b-sentences, however, a deinite
article before the irst element sounds like an attempt to speak in a higher register,
while the article before the second is mandatory (regardless of the semantics).
Moreover, changes in case (accusative, or prepositions) appear only before the second element, hašeni:
(80) a.
hem dibru
exad al
ha-šeni
they talk.PST.M.PL one.M about DEF-second.M
b. * hem dibru
ha-šeni
al (ha-)exad/ al
they talk.PST.M.PL DEF-second.M on (DEF-)one.M on
(ha)exad
ha-šeni
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 371
(DEF-)one.M DEF-second.M
“hey talked about each other.”
Le mie
amiche parleranno
l’un
dell’altro
DEF POSS.1.M.SG friend.PL speak.FUT.3.PL DEF-one of=DEF=other
b. * Le mie
amiche parleranno
l’altro dell’un
DEF POSS.1.M.SG friend.PL speak.FUT.3.PL DEF-one of=DEF=other
“My friends will speak about each other.”
(81) a.
hird, the two elements in these constructions form a constituent and are inseparable. In Hebrew, the irst element exad appears either before the case marker (et
in the accusative) or a preposition. hus, while the order of the arguments of a
verb with three arguments, such as natan “he gave,” is usually free (82a), the two
elements of the pronominal expression always come together (cf. 82b and 82c; for
Italian, see Belletti 1982: 104).
(82) a.
hem natnu
sefer la-more/hem
natnu
la-more
sefer
they give.PST.PL book to.DEF-teacher/they give.PST.PL to.DEF-teacher book
“hey gave a book to the teacher.”
b. hem natnu
exad la-šeni
sefer
they give.PST.PL one.M to.DEF-second.M book
“hey gave each other a book.”
c. *hem natnu
exad sefer la-šeni
they give.PST.PL one.M book to.DEF-second.M
Intended reading:.“hey gave each other a book.” (his sentence is
grammatical in the non-reciprocal reading: “one of them gave another.”)
Only in one case can another NP stand between the two components: if within a
prepositional phrase there is another NP standing in a possessive relationship to
the pronominal expression, then exad may appear right before the preposition, as
in (83):
(83) šney ha-gis-im
yad’u
exad al ma’as-av
both DEF-brother.in.law-PL know.PST.PL one.M on deed-PL.POSS.3.M.SG
šel ha-šeni
of DEF-second.M
“he two brothers-in-law knew about the afairs of each other.”
(http://www.haaretz.co.il/news/law/1.1659732)
Even in this case, however, it sounds as if (83) belongs to a higher register (as the
use of the pronominal possessive suix indicates as well), and therefore it is probably the adnominal possessive construction of the a-type. he standard construction would be as in (84) (for Italian, see Belletti 1982).
© 2014. John Benjamins Publishing Company
All rights reserved
372 Elitzur A. Bar-Asher Siegal
(84) ani ve-yadid šeli
ṣiyarnu male male al ha-yaday-im
I and-friend POSS.1.SG draw.1.PL much much on DEF-hand-PL
exad šel ha-šeni
one.M of DEF-second.M
“A friend of mine and I drew a lot on each other’s hands.”
(http://www.stips.co.il/singleask.asp?stipid=480452)
hese observations lead us to the following preliminary conclusions concerning
these constructions in Hebrew and Italian:
(85) i.
he pronominal expression is plural (as indicated by the verbal
agreement and the control of agreement with relexive pronouns).
ii. he two parts of the pronominal expression form a constituent.
iii. hey show formal indications of univerbation (as indicated by the
change of stress and by the elision in the irst element: exad > xad,
uno > un).
iv. Only the second elements hašeni, l’altro are overtly associated with the
grammatical relations of the sentence (as indicated by the location of
cases and prepositions).
One can thus view the two elements, exad … hašeni/l’un … l’altro, as discontinuous sequences of one constituent (cf. Belletti 1982), whose cores are hašeni and
l’altro (they receive case / the deinite article / prepositions), with exad/l’un joining
and losing their independence even at the phonological level. From a historical
point of view, both languages experienced a similar, relatively covert development
of a shit from a two-unit to a one-unit construction.
6. From one-unit to two-unit constructions
he above discussions follow the development of two-unit to one-unit constructions in various Semitic languages. his raises the question of whether there are
similar developments in the opposite direction, from one-unit to two-unit constructions. A relevant example can be found in the idiolects of some speakers of
the Jewish Neo-Aramaic of Arbel. While speakers of this dialect use dixle — the
one-unit anaphora found in many of the Neo Aramaic dialects in the area — for
some it is only the second component in a two-unit construction, where xa “one”
is the irst element:
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 373
(86) mxélu
xá l-d̀xle
beat.PST.3.PL one to-RECP
“hey beat one another.”
(Khan 1999: 223)13
How did this construction emerge? Without speciic historical data, there are
two possible explanations. On the one hand, the two-unit construction may have
evolved from the addition of the irst element xa to dixle, probably under the inluence of Modern Hebrew, or this construction could be seen as a relection of a
very old phenomenon. For this hypothesis consider the following details, which
are important for the reconstruction of this two-unit construction: 1) the one-unit
anaphora in the NENA [=North Eastern Neo-Aramaic] dialects is historically related to the Syriac anaphora ḥǝdādē (15b), which in turn derives from *ḥadḥadē,
a fusion of a repetition of ḥad “one” + an agreement marker; 2) the form xa is the
cardinal number “one” in the NENA dialects. (he inal /d/ of the Late Aramaic
form of the cardinal number ḥ ad “one” apocopated.) herefore, the two-unit construction consisting of xa and dixle may have resulted from a reanalysis of the
components of the one-unit anaphor as a two-unit construction:
(87) * ḥadḥadē > *xadxadē> *xadxalē > *xa dxalē > *xa dǝxǝlē > xa dixle
he last developments are the expected phonological changes in classical Aramaic:
1) shortening of unstressed vowels and 2) addition of a short vowel in a sequence
of two schwas. Regardless of whether the xa dixle pair is the result of an addition
of xa, or a reanalysis of *xadixle, it nonetheless represents the unusual shit from a
one-unit to a two-unit construction.
7.
Conclusions
he goal of this paper was to present various answers to the broad historical question of how reciprocal constructions evolve. In order to answer this question it was
crucial to distinguish between the one-unit and two-unit constructions, and then
to demonstrate that the one-unit and the two-unit constructions are diferent in
terms of their syntactic structure (§4.3).
13. Khan (1999) interviewed various informants and found a clear distinction between informant L and the others. L had a diferent background from the other informants (Khan, p.c.).
He lived in Jerusalem and spoke Neo-Aramaic and Arabic, whereas others came from rural
villages and spoke Neo-Aramaic and Kurdish. L was more inluenced by Modern Hebrew (in
one example, mentioned by Khan (example 85 above), xá-l-d̀xle was syntactically a one-unit
construction for L).
© 2014. John Benjamins Publishing Company
All rights reserved
374 Elitzur A. Bar-Asher Siegal
I argued that there are two sources for the two-unit reciprocal constructions in
Semitic: 1) a pre-grammaticalized nominal construction, and 2) an extension of a
pronominal construction (§3). In addition it was demonstrated that the grammaticalizations of such constructions are better understood when considering them as
denoting ‘unspeciied relations’ (deined in §2).
As for the one-unit anaphors, there are two diferent processes for their developments: 1) in Indo-European languages there is a reanalysis of constructions
with universal quantiiers (§3.2.3.1), and 2) in Semitic and in other languages as
well this construction developed from a reanalysis of two-unit constructions denoting unspeciied relations (§4.3). Although the result of both processes is the
same, in (1) there is also a semantic shit (§3.2.3). he fact that diferent constructions develop into a similar strategy to express the unspeciied relations naturally
leads to the conclusion that the one-unit construction is a semantically independent strategy to express this function, regardless of its origin. he semantics of the
one-unit construction, therefore, must be understood in its own right.
Furthermore, another conclusion from the data presented in this paper is that
the one-unit anaphors, at least in Semitic, always developed from previous twounit constructions. It would be interesting in further studies to examine whether
in other languages there are anaphors that grammaticalize directly from nominal
expressions without the intermediate stage of a two-unit construction.
Despite this direction of development, in §6 it became clear that while the
change from two-unit to one-unit occurs more oten, the opposite direction occurs as well. Considering the aforementioned need to explain the syntax and the
semantics of the one-unit and the two-unit constructions separately, it is not surprising that the development can go in both directions, as each strategy has its own
independent compositional way to express unspeciied relations.
he possible non-Semitic parallels that were mentioned throughout the paper
open the door to broader historical and typological work on these constructions.
References
Andersen, Henning. 1987. From auxiliary to desinence. In Martin Harris & Paolo Ramat (eds.),
Historical development of auxiliaries, 21–52. Berlin: Mouton de Gruyter.
Bar-Asher, Elitzur. 2009. A theory of argument realization and its applications to features of the
Semitic languages. Cambridge, MA: Harvard University dissertation.
Bar-Asher Siegal, Elitzur A. 2011. Notes on reciprocal constructions in Akkadian in light of
typological and historical considerations. Semitica et Classica 4. 23–42.
Bar-Asher Siegal, Elitzur A. 2012. Diachronic syntactic studies in the Hebrew pronominal reciprocal constructions. In Cynthia Miller & Ziony Zevit (eds.), Diachrony in Biblical Hebrew,
209–242. Winona Lake, IN: Eisenbrauns.
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 375
Beck, Sigrid. 2001. Reciprocals are deinites. Natural Language Semantics 9. 69–138. DOI:
10.1023/A:1012203407127
Beck, Sigrid. 2007. Pluractional adverbials. Journal of Semantics 24. 215–254. DOI: 10.1093/jos/
fm003
Belletti, Foris A. 1982. On the anaphoric status of the reciprocal constructions in Italian. he
Linguistic Review 2. 101–138. DOI: 10.1515/tlir.1982.2.2.101
Bhat, Shankara D. N. 1978. Pronominalization. Pune, India: Deccan College.
Brame, Michael K. 1977. Alternatives to the tensed S and speciied subject conditions. Linguistics
and Philosophy 1. 381–411. DOI: 10.1007/BF00353455
Bybee, Joan. 2003. Mechanisms of change in grammaticalization: he role of frequency. In Brian
D. Joseph & Richard D. Janda (eds.), Handbook of historical linguistics, 602–623. Oxford:
Blackwell. DOI: 10.1002/9780470756393.ch19
Cantarino, Vicente. 1975. Syntax of Modern Arabic prose, vol. 3. Bloomington: Indiana University
Press.
Chomsky, Noam. 1981. Lectures on Government and Binding: he Pisa lectures. Dordrecht, he
Netherlands: Foris.
Corbett, Greville. 2006. Agreement. Cambridge: Cambridge University Press.
Dalrymple, Mary, Makoto Kanazawa, Yookyung Kim, Sam Mchombo & Stanley Peter. 1998.
Reciprocal expressions and the concept of reciprocity. Linguistics and Philosophy 21. 159–
220. DOI: 10.1023/A:1005330227480
Dougherty, Ray C. 1974. he syntax and semantics of each other constructions. Foundations of
Language 12. 1–47.
Evans, Nicholas. 2008. Reciprocal constructions: Towards a structural typology. In Ekkehard
König & Volker Gast (eds.), Reciprocals and relexives: heoretical and cross-linguistic explorations, 33–103. Berlin: Mouton de Gruyter.
Evans, Nicholas, Alice Gaby, Stephen C. Levinson & Asifa Majid. 2011. Introduction: Reciprocals
and semantic typology. In Nicolas Evans, Alice Gaby, Stephen C. Levinson & Asifa Majid
(eds.), Reciprocals and semantic typology (Typological Studies in Language 98), 1–28.
Amsterdam: John Benjamins. DOI: 10.1075/tsl.98.01intro
Everaert, Martin. 1990-1. Nominative anaphors in Icelandic: Morphology or syntax? In Werner
Abraham, Wim Kosmeijer & Eric Reuland (eds.), Issues in Germanic syntax (Trends in
Linguistics: Studies and Monographs 44), 277–305. Berlin: Mouton de Gruyter.
Everaert, Martin. 1999. Types of anaphoric expressions: Relexives and reciprocals. In Zygmunt
Frajzyngier & Traci S. Curl (eds.), Relexives: Forms and functions, 63–83. Amsterdam: John
Benjamins.
Fassberg, Steven E. 2010. he Jewish Neo-Aramaic dialect of Challa. Leiden: Brill.
Fiengo, Robert & Howard Lasnik. 1973. he logical structure of reciprocal sentences in English.
Foundations of Language 9. 447–468.
Glinert, Lewis. 1989. he grammar of Modern Hebrew. Cambridge: Cambridge University Press.
Haas, Florian. 2010. Reciprocity in English: Historical development and synchronic structure. New
York: Routledge.
Halevy, Rivka. 2011. Reciprocal constructions between syntax and lexicon. L’shonénu 73. 401–
422.
Haspelmath, Martin. 1997. Indeinite pronouns (Oxford Studies in Typology and Linguistic
heory). Oxford: Oxford University Press.
Haspelmath, Martin. 2007. Further remarks on reciprocal constructions. In Vladimir P.
Nedjalkov (ed.), Reciprocal constructions, 2087–2115. Amsterdam: John Benjamins.
© 2014. John Benjamins Publishing Company
All rights reserved
376 Elitzur A. Bar-Asher Siegal
Heim, Irene, Howard Lasnik & Robert May. 1991. Reciprocity and plurality. Linguistic Inquiry
22. 63–101.
Heine, Bernd & Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge: Cambridge
University Press. DOI: 10.1017/CBO9780511613463
Heine, Bernd & Hiroyuki Miyashita. 2008. he intersection between relexives and reciprocals:
A grammaticalization perspective. In Ekkehard König & Volker Gast (eds.), Reciprocals and
relexives: heoretical and cross-linguistic explorations, 169–223. Berlin: Mouton de Gruyter.
Kamp, Hans & Uwe Reyle. 1993. From discourse to logic: Introduction to model theoretic semantics of natural language, formal logic and discourse representation theory. Dordrecht, he
Netherlands: Kluwer Academic.
Kemmer, Suzanne. 1993. he middle voice: A typological and diachronic study. Amsterdam: John
Benjamins.
Khan, Geofrey. 1999. A grammar of Neo-Aramaic: he dialect of the Jews of Arbel. Leiden: Brill.
Knjazev, Jurij P. 2007. Lexical reciprocals as a means of expressing reciprocal relations. In
Vladimir P. Nedjalkov (ed.), Reciprocal constructions, vol. 1-5, 115–146. Amsterdam: John
Benjamins.
König, Ekkehard & Shigehiro Kokutani. 2006. Towards a typology of reciprocal constructions:
Focus on German and Japanese. Linguistics 44. 271–302. DOI: 10.1515/LING.2006.010
Koster, Jan. 1987. Domains and dynasties: he radical autonomy of syntax. Dordrecht, he
Netherlands: Foris.
Kremers, Joost. 1997. How Arabs speak to each other about themselves: A study of nafs and baʿḍ in
Modern Standard Arabic. Nijmegen: Katholieke Universiteit Nijmegen MA thesis.
Lichtenberk, Frantisek. 1985. Multiple uses of reciprocal constructions. Australian Journal of
Linguistics 5. 19–41. DOI: 10.1080/07268608508599334
Leslau, Wolf. 2000. Introductory grammar of Amharic. Wiesbaden: Otto Harassowitz.
Nedjalkov, Vladimir P. 2007a. Overview of the research: Deinitions of terms, framework,
and related issues. In Vladimir P. Nedjalkov (ed.), Reciprocal constructions, vol. 1, 3–114.
Amsterdam: John Benjamins.
Nedjalkov, Vladimir P. 2007b. Encoding of the reciprocal meaning. In Vladimir P. Nedjalkov
(ed.), Reciprocal constructions, vol. 1-5, 147–208. Amsterdam: John Benjamins.
Nedjalkov, Vladimir P. & Emma Geniušienė. 2007. Questionnaire on reciprocals. In Vladimir P.
Nedjalkov (ed.), Reciprocal constructions, vol. 1-5, 379–434. Amsterdam: John Benjamins.
Nöldeke, hodore. 2001. Compendious Syriac grammar. Winona Lake, IN: Eisenbrauns.
Plank, Frans. 2008. houghts on the origin, progress, and pronominal status of reciprocal forms
in Germanic, occasioned by those of Bavarian. In Ekkehard König & Volker Gast (eds.),
Reciprocals and relexives: heoretical and cross-linguistic explorations, 347–373. Berlin:
Mouton de Gruyter.
Rubin, Aaron D. 2010. he Mehri language of Oman. Boston: Brill.
Siloni, Tal. 2002. Active lexicon. heoretical Linguistics 28. 383–400.
Siloni, Tal. 2012. Reciprocal verbs and symmetry. Natural Language & Linguistic heory 30.
261–320. DOI: 10.1007/s11049-011-9144-2
hráinsson, Höskuldur. 1979. On complementation in Icelandic. New York: Garland.
Vezzosi, Letizia. 2010. Micro-processes of grammaticalization: he case of Italian l’un l’altro.
In Stathi Katerina, Elke Gehweiler & Ekkehard König (eds.), Grammaticalization: Current
views and issues (Studies in Language Companion Series 119), 343–372. Amsterdam: John
Benjamins. DOI: 10.1075/slcs.119.18vez
© 2014. John Benjamins Publishing Company
All rights reserved
Notes on the history of reciprocal NP-strategies in Semitic languages 377
Visser, Fredericus h. 1963. An historical syntax of the English language, Part 1: Syntactical units
with one verb. Leiden: Brill.
Von Soden, Wolfram. 1931. Der hymnisch-epische Dialekt des Akkadischen. Zeitschrit für
Assyriologie und vorderasiatische Archäologie 40. 163–227.
Williams, Edwin. 1991. Reciprocal scope. Linguistic Inquiry 22. 159–173.
Résumé
Parmi les constructions réciproques des langues sémitiques il existe une division centrale entre
deux types : 1) les constructions à deux composantes, chacune d’entre elles occupant une position argumentale diférente ; 2) les constructions à une composante, où celle-ci occupe une
position non-sujet régie par le verbe et partage les traits de nombre pluriel avec le sujet. L’objectif
de cet article est de comprendre le mécanisme du développement de ces constructions et de
répondre aux questions suivantes : 1) comment ont évolué les diférents sous-types de constructions à deux composantes ? 2) Peut-on expliquer le développement de la construction à une
composante à partir de celle à deux composantes à travers les changements diachroniques ? La
première question a déjà été abordée dans la littérature, visant la portée des stratégies conceptuelles qui ont tendance à se développer en marqueurs réciproques. Or, de telles descriptions
n’expliquent pas comment les constructions ont acquis leurs sens spéciiques. Dans la présente communication je propose qu’il est essentiel de prendre en compte la sémantique de ces
constructions ain de comprendre l’évolution des constructions sémitiques. Plus précisément,
je proposerai qu’à la place de ‘constructions réciproques’ il est plus approprié de les considérer comme des constructions à ‘relations non-spéciiées’ (la sémantique exacte sera déinie et
formalisée dans la communication). Quant à la deuxième question, de multiples tentatives ont
été faites auparavant pour expliquer de tels processus, se concentrant sur les langues indo-européennes. Or, ces propositions ne sont pas compatibles avec les langues sémitiques, et je proposerai donc une hypothèse alternative. Cette étude s’intéresse aux langues sémitiques, et comprend quelques démonstrations parallèles provenant d’autres familles de langues (notamment
les langues indo-européennes). Celles-ci permettent de voir l’intérêt et les limites d’une étude
typologique plus fouillée sur ce thème.
Zusammenfassung
Die semitische Sprachfamilie erlaubt für den Ausdruck von Reziprozität zwei grundsätzlich verschiedene Konstruktionstypen: 1) Zweiteilige Konstruktionen: Diese enthalten zwei
Komponenten, welche je unterschiedliche Positionen in der Argumentstruktur des Verbs
besetzen; 2) Einteilige Konstruktionen: Diese enthalten ein einziges Element, welches mit
Pluralsubjekten koreferiert und die vom Prädikat selektierte nicht-Subjekt Position besetzt.
Ziel dieser Arbeit ist es, anhand folgender Leitfragen die Entstehungsmechanismen dieser beiden Konstruktionen zu erforschen: 1) Wie entwickelten sich die verschiedenen Formen der
zweiteiligen Konstruktionen? 2) Kann eine diachronische Verschiebung die Entwicklung von
einteiliger hin zu zweiteiliger Konstruktion erklären? Auf die erste Frage wird in der Literatur
vor allem in Hinblick darauf eingegangen, welche Elemente sich zu Reziprozitätsmarkern
© 2014. John Benjamins Publishing Company
All rights reserved
378 Elitzur A. Bar-Asher Siegal
entwickeln. Beschreibungen dieser Art erklären jedoch nicht, wie sich die spezielle Bedeutung
der Konstruktionen entwickelt hat. Im vorliegenden Artikel wird die hese vertreten, dass es
für das Verständnis der Entstehung der semitischen Reziprok-Konstruktion wesentlich ist,
auch deren Semantik zu berücksichtigen. Insbesondere wird argumentiert, dass die fraglichen
Konstruktionen besser als ‘unspeziizierte Relationen’ anstatt als ‘reziproke Konstruktionen’
aufzufassen sind (eine genaue formale Semantik wird im Folgenden deiniert). Hinsichtlich
der zweiten Frage wurden bereits zahlreiche Versuche unternommen, die relevanten Prozesse
zu analysieren, wobei der Schwerpunkt hier auf den indogermanischen Sprachen liegt. Diese
Erklärungsversuche sind jedoch nicht auf das Semitische übertragbar, sodass in dieser Arbeit
ein alternativer Erklärungsversuch unternommen wird. Obgleich die vorliegende Studie ihren
Schwerpunkt auf die semitischen Sprachen legt, werden gelegentlich Parallelen zu anderen
Sprachfamilien — hauptsächlich der indogermanischen — aufgezeigt. Zweck dieser gelegentlichen parallelen Betrachtungsweise ist es, auf die Möglichkeiten und Aussichten einer umfassenderen typologischen Studie hinzuweisen.
Author’s address
Elitzur A. Bar-Asher Siegal
School of Language Science/he Department of the Hebrew Language
Language Logic Cognitive Center
he Hebrew University of Jerusalem
Mt. Scopus, Jerusalem 91095
ebas@mail.huji.ac.il
© 2014. John Benjamins Publishing Company
All rights reserved