On the division of labor between roots and functional structure
Artemis Alexiadou (Universität Stuttgart) and Terje Lohndal (NTNU Trondheim)
Introduction: In frameworks such as Distributed Morphology (DM), and exoskeletal
approaches in general (e.g. Borer 2005, 2013) it is assumed that the lexicon consists
of roots and that these roots are categorized by functional elements. For instance, in
DM, in the presence of little v, a root is interpreted as a verb. An important question
raised by such approaches is whether roots have independent meaning of their own,
that is, do roots have substantial meaning independent of their syntactic
configuration? This paper proposes a typology of languages based on the division of
labor between little v and roots, and discusses the question of how roots are merged in
morphosyntactic representations.
Background: Arad (2003, 2005) shows that some languages have more roots than
others. She compares English to Hebrew. In Hebrew, the number of roots is highly
limited. These morphemes have very little meaning outside of a template that assigns
interpretation to them. English is different: in this language, roots seem to have some
substantial meaning which is rather independent of the syntactic configuration in
which they occur (Harley 2005, Alexiadou, Anagnostopoulou and Schäfer 2006, and
Levin and Rappaport 2008). Furthermore, unlike Hebrew, English does not have a
rich functional vocabulary that is responsible for fixing the interpretation of roots.
Thus we have two languages on each side of the scale: Hebrew with little root
independent meaning and several functional morphemes, and English with root
independent meaning and few functional morphemes. In this paper we compare
English and Hebrew, to the behavior of roots and affixes in earlier English, and to
Greek, both of which provide evidence for the existence of mixed languages where
some roots have substantial meaning and others do not.
Causativization in English: van Gelderen (2011) documents a series of verbal
valency changes in the history of the English language. Among others, she shows that
in the Middle English period, a new causativizer is introduced, namely –en. A few
examples are provided in (1) from van Gelderen (2011: 125).
(1)
awaken, brighten, fasten, frighten, ripen, sadden, strengthen, tighten
van Gelderen analyzes –en as an instance (a ‘flavor’, cf. Folli and Harley 2007) of
little v though with a clear causative semantics. She also points out that there are three
other causative suffixes: -ize, -(i)fy, -ate (e.g., advertize, abdicate, beautify). These are
not fully productive and were loaned from Greek and Latin. van Gelderen argues that
in Modern English, -en and zero derivations derive a causative from an unaccusative.
Pace van Gelderen, we argue that -en realizes a Cause head (for reasons outlined in
Alexiadou et al. 2006), and that the other ‘causative’ suffixes are pure verbalizers,
realizing little v. The main argument for this is that these suffixes are not productive
and do not provide real causative semantics. The history of causativization in English
demonstrates that little v can be morphologically realized even in English, and that
English at earlier stages looked a bit more like Hebrew.
Greek: Anagnostopoulou and Samioti (in press) argue that in Greek there are roots
with a substantial meaning, (2), and roots where the meaning depends on the syntactic
environment, see (3). In the latter case, the role of verbalizer affixes is to classify the
roots in terms of ontological categorization.
(2) a.
axn-os
‘steamN’ axn-iz-o ‘steamV’
b.
vid-a
‘screwN’ vid-on-o ‘screwV’
(3)
a.
kt-iz-o ‘build-1sg’
b.
sk-iz-o
‘tear-1sg’
They also argue that verbalizers are different from the head contributing eventiveness.
As can be seen in (4), in Greek stative participles in -tos include such verbalizing
elements. However, they lack eventive implications, they denote characteristic states.
For instance, they do not license manner modification (5), neither agent PPs or
instruments:
(4)
a.
axn-is-tos
‘steaming hot’
axn-iz-o
‘steam’
b.
vid-o-tos
‘screwed’
vid-on-o
‘screw’
(5)
a.
*To fagito
ine kala/ prosektika
magir-ef-t-o
The food
is well/ carefully
cooked
Idiomatic expressions: Lately, little v has been related to interpretation, and in
particular Arad (2003, 2005) and Marantz (2001, 2007, 2012) have argued that the
first categorizer determines the interpretation of a root. Anagnostopoulou & Samioti
(in press) show that Greek complicates this picture and does not conform to the
Arad/Marantz view (see also Harley 2011). They argue that the relevant head that
determines idiomaticity is Voice (cf. Marantz 1997, 2011). However, if Arad/Marantz
were right, it would obscure the claim that English is different from Hebrew.
Analysis: We argue that it is possible to devise a typology of roots and functional
morphemes where languages are sorted on a scale from Hebrew to English.
(6)
A scale from ‘empty’ roots to ‘contentful’ roots
Hebrew
>
Greek
>
Old English >
English
It follows that in Hebrew, functional morphemes and especially verbalizers are crucial
in determining the interpretation of a root. In contemporary English, this is not the
case, and the interpretation of the root is to a greater extent determined by the
meaning of the root itself. In cases of idiomatic dependencies, we argue that it is not
little v but Voice that is the crucial head. This suggests that little v plays different
roles in different languages: in Hebrew it is crucial for determining interpretation of
roots, whereas in English it only categorizes roots. In English, and partly in Greek, it
is Voice that is the more important head for determining (idiomatic) interpretation.
We argue that little v is always a verbalizer in all languages. It can also take
on additional roles, such as in Hebrew where it is strongly linked to determining the
interpretation of roots. These roles have to be acquired – Universal Grammar only
specifies that little v is always a categorizer. This means that little v cannot be the
locus of accusative Case, but that this has to be relegated to Voice or Aspect heads.
Thus little v does not play a role in argument structure, contrary to work by Chomsky
(1995, 2001, 2008). But where do little v and a root merge in the structure?
Locus of Root-Merger: Three views have been pursued with respect to the question
how roots are merged in morphosyntactic representations. On one view, roots are
always merged as complements of v (e.g. de Belder & Craenenbroek 2011). On a
second view, roots are merged as v modifiers (Marantz 2013). Finally, a third view
holds that some roots are merged as v's modifier while others as v's complement
(Embick 2004, 2010). The behavior of roots in transitivity alternations (Levin &
Rappaport Hovav 2008, Alexiadou & Anagnostopoulou 2012) provides evidence for
view (iii). In particular, as Alexiadou & Anagnostopoulou (2012) have shown, the
morphological make-up of Greek verbs of detachment strongly suggests that many
verbs participating in the locative alternation are manner verbs and non-alternating
ones are result verbs. Alternating verbs typically involve manner/instrument roots and
a verbalizer (7a), and following Embick (2004), they can take small clause
complements introducing the result. Non-alternating verbs are e.g. based on a state
root, plus a verbalizer (7b) and cannot take such complements.
(7)
a. sider-on-o lit. iron-v-1sg ‘iron’ b. atho-on-o lit. innocent-v-1sg ‘acquit’
If all roots had but one-way to merge this would be unexpected. This is also
consistent with Embick’s (2010) theory of allomorphy, though see Marantz (2013).