1
Domains within Words and their meanings: a case study.
Elena Anagnostopoulou & Yota Samioti
University of Crete
[email protected], [email protected]
To appear in: Alexiadou, A., H. Borer & F. Schäfer (Eds.) The syntax of Roots and the roots of
Syntax. Oxford: Oxford University Press.
1. Introduction1
The goal of this paper is to investigate the conditions under which idiosyncratic meaning
arises in word formation. More specifically, we investigate the hypothesis put forth in Arad
(2003, 2005) and Marantz (2001, 2007) that idiosyncratic meaning derives from locality
conditions on the interpretation of roots. This hypothesis is stated in (1):
(1) The Marantz / Arad Hypothesis
Roots are assigned an interpretation in the context of the first category assigning head/
phase head merged with them, which is then fixed throughout the derivation
We test (1) in the domain of Greek verbal adjectives and adjectival participles which provide
extensive evidence for attachment below vs. above higher heads (i.e. little v, Voice) of the
adjectivizing/participial suffixes –tos and –menos, respectively. Our investigation leads to the
following results:
(i) Affixes serving as verbalizers and the little v head contributing eventiveness cannot be
collapsed. Greek verbal adjectives are shown to be semantically and syntactically small (i.e.
they do not have event implications, a semantic property correlating with the fact that they do
not support adverbs and other modifiers), and yet they may contain verbalizers inside the
adjectivizing suffix -tos. This leads us to propose that there are two ways of interpreting the
Arad/Marantz hypothesis, depending on whether we take (morphological) verbalizers or
(abstract) eventivizing heads to qualify as the relevant phase heads in (1).
(ii) If we look at the interpretation of an unspecified root, as suggested by Arad (2003,
2005) on the basis of Multiple Contextualized Meaning (MCM) in Hebrew, then Greek
provides some support for the view that specified meaning is indeed determined at the point
of categorization. Even though many Greek roots have a rather fixed meaning, unlike
Hebrew, there are also roots whose meaning is unspecified, similarly to Hebrew.
Interestingly, the presence of a verbalizing head in verbal adjectives formed on the basis of
the latter type of roots serves as a context determining their interpretation (similarly to
prefixes with latinate roots in English and Ancient Greek roots in Greek), as expected under
Arad’s hypothesis. More specifically, we propose that roots fall into basic ontological types
1 Different versions of this paper have been presented at the 32
nd GLOW Colloquium (University of Nantes,
April 15-18 2009), the Roots Workshop (University of Stuttgart, June 2009) and the Workshop on Linguistic
Interfaces (University of Ulster, December 2-4 2010). We would like to thank the audiences for questions,
suggestions and discussions, especially Asaf Bachrach, Orin Percus, Peter Svenonius, Gertjan Postma, Danny
Fox, Alec Marantz, David Embick, Andrew Koontz Garboden, Hagit Borer, Lisa Travis, Christina Sevdali, Raffi
Folli, and Heidi Harley. Special thanks to Winfried Lechner, Artemis Alexiadou and Terje Lohndal for their
comments.
2
naming events, things and states following Harley (2005). Roots must be classifiable in
terms of this basic ontology. If they are not, a categorizing head serves to classify them. The
classification itself makes them acquire what is sensed as fixed “meaning”, which is then
retained throughout the derivation.
(iii) If we look at idiomatic interpretations, as suggested by Marantz on the basis of
idiomatic interpretations of adjectival passive participles vs. compositional interpretations of
verbal passive participles in English, then (1) is not supported by the Greek data. Verbal
adjectives involving affix attachment below v (whether v is taken to be a verbalizer or an
eventivizer) as well as participles showing affix attachment above v may both have idiomatic
meanings, and there is no necessary correlation between the meaning of verbs and the
meaning of participles which include a verbal head, contra Marantz (2001, 2007). On the
other hand, there is strong evidence from various directions that the head which delimits the
domain for idiomatic interpretations of adjectival participles and verbal adjectives in Greek is
Voice, the little v head introducing the agent, as was proposed in earlier work by Marantz
(1996, 1997). We conclude that the node that syntactically projects an agent serves as a
boundary for special meanings of idioms at the word level, similarly to what has been claimed
for idioms at the phrasal level.
The paper is organized as follows. In section 2, we provide the necessary background
on inner vs. outer affixation focusing on how this distinction has been related to the question
of idiosyncratic interpretation. In section 3, we present an outline of the properties of Greek –
menos participles and –tos verbal adjectives, investigating their architecture in connection to
their syntax, semantics and morphology. We present evidence that morphological verbalizers
should be dissociated from the abstract heads contributing eventiveness. This leads us to
conclude that the categorizing head v (vC), which verbalizes an a-categorial root, cannot be
collapsed with the eventivizing v head (vE), which functions as the BECOME/ RESULT/
FIENT operator identified in recent decomposition analyses. Finally, in section 4 we turn to
the question of domains for idioms. We argue that neither the verbalizing head vC nor the
eventivizing head vE serve as boundaries for special meanings; the relevant domain is defined
by Voice, the head projecting an agent.
2. Background: two domains for word formation, inner and outer affixation.
As is well known, there are two types of word formation, one forming words showing
irregularities, such as paradigmatic gaps, non-predictable meaning, irregular forms, and one
for morphologically productive, semantically transparent and morpho-phonologically
predictable forms. In many theories, this difference has been linked to the hypothesis that
there are two places for word formation: Lexicon vs. syntax (e.g. Wasow 1977, Dubinsky &
Simango 1996, Horvath & Siloni 2008), derivation vs. inflection (e.g. Anderson 1982;
Perlmutter 1988; Spencer 1991) level I vs. level II (e.g. Siegel 1979; Kiparsky 1982;
Mohanan 1986). Adjectival/stative vs. verbal/eventive passive participles in languages like
English have been widely assumed in the literature to present a paradigmatic case
exemplifying the double nature of word formation. Stative/ adjectival participles may show
special morphology, while eventive/ passive participles always show regular morphology:
(2) a. The shaven man
b. The man was shaved by John
Moreover, stative participle formation is associated with idiosyncrasy in meaning, unlike
verbal participle formation:
3
(3) a. The hung jury (#Someone hung the jury)
b. *The jury was being hung
In English and other languages the stative vs. eventive participle distinction correlates with a
split between derivation and inflection (Wasow 1977, Marantz 2001, 2007; Horvath & Siloni
2008). Stative participles are adjectives and passive ones are verbs, as shown by “very” vs.
“very much”-modification in (4) and (5). In English, “very” modifies exclusively adjectives
and “very much” exclusively verbs, as indicated by the contrasts between (4b) vs. (4c) and
(5b) vs. (5c), respectively. The fact that the stative participle respected is modified by “very”
in (4a) presents evidence that it qualifies as an adjective. On the other hand, passive respected
in (5a) is modified by very much qualifying as a verb.
(4) a. Your family was very respected Adjectival
b. John is very fond of your family
c. *John is very much fond of your family
(5) a. Your family was very much respected by the neighbors Verbal
b. John very much respects your family
c. *John very respects your family
Wasow (1977) argued that the properties distinguishing adjectival from verbal participles may
receive a principled explanation if the two participle types are formed in two different
components of the grammar (an approach widely adopted since then;2 see Horvath & Siloni
2008 for a recent analysis along these lines). On this view, adjectival passives are formed in
the lexicon, and they show idiosyncratic forms and meanings due to the fact that words in the
lexicon have special listed properties. On the other hand, verbal passives are formed in
syntax, the locus of regularity, productivity and compositionality, which in turn explains their
transparency in form and meaning.
For syntactic approaches to morphology such as Distributed Morphology (DM) or
Borer (2005), properties traditionally attributed to lexical listing (e.g. categorization,
argument structure) are reduced to mechanisms of the computational system. In such models,
there can be no well-defined distinction between lexical and syntactic word formation. This
raises the question of how the distinction between regular and irregular word formation can be
derived. In order to capture this distinction, Marantz (2001, 2007) and Arad (2003, 2005)
propose to reconstruct the ‘two places to build words” approach within the syntax, as ‘two
domains for word formation’ (see also Embick 2010). More specifically, they propose that
there are two domains for word formation delimited by a category-defining head, as shown in
(6). Attachment of x directly to the root, as in (6a), leads to irregular word formation, while
attachment above the category defining heads (little v, a, n), as in (6b), leads to regular word
formation:
(6) a. x b. x 3 3 Root x n, v, a 3 Root n, v, a
2 But cf. Strong Lexicalist approaches such as Levin & Rappaport’s (1986) who argue that all participles are
formed in the lexicon, adjectival participles are formed by verbal ones via a category changing rule, and the
properties distinguishing adjectival from verbal participles derive from their difference in category. See
Dubinsky & Simango (1996) for evidence against this view and Marantz (2007) for discussion of Wasow’s
analysis, as compared to Levin & Rappaport’s.
4
On this view, there is a split between “inner” and “outer” morphology, as defined in (7):
(7) “Inner morphology” attaches to roots or complex constituents below the first
little x (x={v,n,a}) node head (phase head) above the root. All morphology
above the first x node is “outer morphology” including all “category
changing” derivational morphology.” (Marantz 2007: 5; Marantz’s (3), (2))
The main properties associated with inner vs. outer affixation are summarized in Table 1:
Table 1
Inner Affixation Outer Affixation
Regularity Potential special form and
special meaning
Predictable form and
predictable meaning
Selection Attaches inside
morphology determining
lexical category
May attach outside
morphology determining
lexical category
According to these authors, phasehood provides the key towards understanding why there is
an inner vs. outer split in word-formation. Assuming that category defining heads are phase
heads in the sense of Chomsky (2000, 2001), they define cyclic domains leading to the
semantic interpretation and phonological spell-out of the chunk of syntactic structure which
includes the root or a more complex constituent plus the category determining head. Once
such phase heads are merged, the structure is shipped off to PF and LF for pronunciation/
interpretation and from that point on, spell-out and interpretation cannot be altered. Moreover,
further affixation cannot reach into properties of the Root. The empirical generalization this
proposal is based on will be called here “the (non-)compositionality generalization” in (8):
(8) The (non-)compositionality generalization
When affixes attach directly to the root, idiosyncratic meanings may arise.
When affixes attach outside category defining heads, the result is a meaning
predictable from the meaning of the stem.
Arad (2003, 2005) presents extensive evidence from Hebrew denominal verbs supporting (8)
and the “inner” vs. “outer” architecture.
3. Greek –menos participles and –tos verbal adjectives: outer and inner architecture
Greek has a rich set of adjectives/ participles based on verbs, ending in -tos and –menos.3 As
has been argued for in Anagnostopoulou (2003) and Alexiadou & Anagnostopoulou (A&A
2008), –menos participles denote states resulting from prior events and are formed by outer
cycle attachment (above vP or above VoiceP). On the other hand, -tos verbal adjectives lack
event implications, a property taken by Anagnostopoulou (2003) and A&A (2008) to provide
evidence for inner-cycle attachment. If this picture is correct, then -menos and –tos forms
present an ideal empirical domain for testing the validity of the Marantz/Arad hypothesis (1),
and the non-compositionality generalization in (8). Assuming (1) and (8), –menos participles
3 Νakas 1978, Setatos 1984, Lascaratou 1991, Lascaratou & Philippaki-Warburton 1984, Anastasiadi-Simeonidi
1994, Markantonatou et al. 1996, Kordoni 2002, Αnagnostopoulou 2003, Alexiadou & Anagnostopoulou 2008,
Samioti 2008; in progress.
5
are expected to have a meaning predictable from the meaning of the corresponding verbs,
while –tos forms are expected to be highly idiosyncratic.
However, as will be shown in this section, a closer investigation of the morphology of
–tos forms in connection to their syntax and semantics reveals that this simple picture needs
to be revised. It turns out that the architecture of –tos verbal adjectives is considerably more
complex than described by Anagnostopoulou (2003) and A&A (2008), for two main reasons:
First, it is actually not easy to decide on the basis of morphology whether –tos verbal
adjectives involve inner or outer cycle attachment: sometimes –tos attaches directly to the
root and sometimes outside a verbalizer, with absolutely no difference in the interpretation
and syntax of the two cases. Importantly, –tos verbal adjectives uniformly lack event
implications and the concomitant syntactic differences associated with them, regardless of
whether –tos attaches directly to the root or to the root+verbalizer complex.
Second, -tos attachment directly to the root provides some evidence for Arad’s
hypothesis based on Multiple Contextualized Meaning (MCM). If the meaning of the root is
completely unspecified the presence of a verbalizer fixes it first, and then –tos attaches to the
Root+verbalizer complex. But idiomaticity is a different issue, and the domain for non-
compositional, unpredictable interpretation of verbal adjectives/ participles turns out to be
bigger than the domain defined by the first categorizing or eventivizing head: idiomatic
meanings may arise at any point in the extended vP domain, provided that the properties
associated with an agent are absent.
The discussion is organized as follows. In section 3.1, we summarize from the
previous literature the main evidence in favor of postulating a more complex architecture for
–menos participles than for –tos forms. In section 3.2, we concentrate on the properties of –tos
verbal adjectives which lead to a revision of the Anagnostopoulou (2003)/ A&A (2008)
picture in two respects: (a) There are classes of –tos verbal adjectives which express negation
and possibility/ability (the latter implicating an agent). These involve attachment above
vP/VoiceP, similarly to –menos participles. (b) A class of –tos verbal adjectives involves
attachment below the little v head contributing eventiveness but not necessarily below
verbalizers. It thus follows that morphological verbalizers cannot be equated with abstract
eventivizing (little v) heads. In section 3.3, we address the generalizations governing (i) the
distribution of –menos vs. –tos in participles/ verbal adjectives and (ii) the presence vs.
absence of verbalizers in Greek verbal adjectives. We argue that the morphological piece –
menos is always associated with the Perfect of Result (Kratzer 1994, 2001), i.e. it is present
with participles expressing that the subject is in the target or result state of a prior event. On
the other hand, –tos is present in a heterogeneous set of cases all having in common that they
uniformly lack event implications. We furthermore propose that verbalizers in Greek verbal
adjectives are required on roots that do not fall under the basic ontological type “event”,
following Harley’s (2005) root-typology. Having revised and refined the Anagnostopoulou
(2003)/A&A (2008) picture of the –menos vs. –tos distinction, we finally proceed to the
question of idiomaticity. In section 4, we test (1)/(8) by comparing the interpretations of the
several types of –tos verbal adjectives to those of -menos participles and of the corresponding
verbs. We conclude that agentivity and not eventiveness is the key property blocking
idiomatic readings from participles, in accordance with Marantz (1996, 1997) and contra
Marantz (2001, 2007).
3.1. The simple A 2003/ A& A 2008 picture: -menos is an outer- and –tos an inner-affix
Next to adjectives, Greek has two adjectival constructions: the participle in –menos and what
traditional grammars call the “verbal adjective” in –tos:
6
(9) anig-men-os anix-t-os “opened”/ “open”
open-men-masc.sg.nom open-t-masc.sg.nom
They have a similar function to adjectives, i.e. they appear in attributive and predicative
positions:4
(10) a. to anih-t-o parathiro
the-neut.sg.nom open-t-neut.sg.nom window
‘the open window’
b. to anig-men-o parathiro
the-neut.sg.nom open-men-neut.sg.nom window
‘the opened window’
(11) a. to parathiro ine anih-t-o
the window is open-t-neut.sg.nom
‘the window is open’
b. to parathiro ine anig-men-o
the window is open-men-neut.sg.nom
‘the window is opened’
Participles and verbal adjectives show a number of semantic and syntactic differences
discussed in Markantonatou et al. (1996), Georgala (2001), Kordoni (2002), Anagnostopoulou
(2003), A&A (2008), among others. The most fundamental one is that –menos participles
denote a state resulting from a prior event, while –tos verbal adjectives lack event
implications. They denote what has been called “a characteristic state” by Markantonatou et
al. (1996). Consider the examples in (12):
(12) a. I patates ine tiganis-men-es
The potatoes are fry-men-fem.pl.nom
‘The potatoes are fried’
b. I patates ine tigani-t-es
The potatoes are fry-t-fem.pl.nom
‘The potatoes are fried’
(12a) conveys the meaning that the potatoes are fried as a result of a frying event: they have
been fried e.g. a minute ago and are now ready to be eaten. On the other hand, (12b) simply
expresses the fact that the potatoes are cooked in a particular way (“characteristic state”
interpretation): they are fried (rather than e.g. cooked). A context bringing out this difference
between the two adjectival forms is (13), based on Embick (2004: 357). As pointed out by
Embick, the complement of verbs of creation cannot be a state resulting from a prior event
because this would be a contradiction. And as can be seen in (13), –tos forms expressing
characteristic states are licit in such a context, while –menos participles are not, due to their
eventive subcomponent:
(13) a. I porta chtistike anix-t-i/ *anig-men-i
The door-Nom built-NAct,3sg open/ opened
‘The door was built open/*opened’
b. Magirepsa to kotopoulo vras-t-o/ *vras-men-o
Cooked-I the chicken boiled/ boiled
4 Passives are verbal in Greek and therefore there is never an ambiguity between verbal and adjectival passives
in Greek, unlike English.
7
‘I cooked the chicken boiled’
c. Eftiaksa tis patates tigani-t-es/ *tiganis-men-es
Made-I the potatoes fried/fried
‘I made the potatoes fried’
Crucially, the difference between the two forms with respect to event implications is
associated with a number of syntactic differences. –menos participles can be modified by
manner adverbs (14a) and can license by-phrases and instrument PPs (15a), while –tos verbal
adjectives cannot (14b, 15b):
(14) a. Afto to vivlio ine kala gra-men-o
This the book is well written
‘This book is well-written’
b. *Afto to kimeno ine kala grap-t-o
This the text is well written
(15) a. O tixos ine xtis-men-os me mistri/ apo ton ergati
The wall is built with trowel/ by the worker
‘The wall is built with a trowel/ by the worker’
b. *O tixos ine xtis-t-os me mistri/ apo ton ergati
The wall is built with trowel/ by the worker
Anagnostopoulou (2003), following Kratzer (2001), furthermore points out that -menos
participles can denote target and resultant states (Parsons 1990: 234-235). The former
describe states that are in principle reversible; the latter introduce states that hold forever after
the event that brings them about. Target state participles in (16) are compatible with the
adverbial akoma 'still', while resultant state participles in (17) are incompatible with it:
(16) Ta pedhia ine akoma krimena
The children are still hidden
(17) *To theorima ine akoma apodedigmeno
The theorem is still proven
Target and resultant state –menos participles are not only semantically but also syntactically
distinct. Target state –menos participles cannot license agent and instrument PPs and agentive
adverbials. As shown by (18), by-phrases and instrument phrases are incompatible with
akoma ‘still’:
(18) a. Ta lastixa ine (*akoma) fuskomena apo tin Maria
The tires are (still) inflated by the Mary
‘The tires are still inflated by Mary’
b. Ta lastixa ine (*akoma) fuskomena me tin tromba
The tires are (still) inflated with the pump
‘The tires are still inflated with the pump’
Moreover, note that there are two types of manner adverbials, i.e. (a) manner adverbs that are
result-oriented in that they modify the visible result of an event such as ‘sloppily’, ‘well’ and
(b) manner adverbs that modify the initiator of the action such as ‘carefully’, ‘deliberately’
8
(agent-oriented). As shown below, agent-oriented modifiers are incompatible with akoma
(19), while adverbs denoting the visible result (result-oriented) are compatible with it (20):
(19) To thisavrofilakio itan (*akoma) prosektika anigmeno
The safe was (still) cautiously opened
‘The safe was still cautiously opened’
(20) Ta malia mu ine (akoma) atsala xtenismena
The hair my is still sloppily combed
‘My hair is still sloppily combed’
On the basis of these considerations, A&A (2008) proposed that Greek has three types of
participles with three different structures, depending on the height of attachment of the
participle morpheme. According to A&A (2008), a layer Asp (=stativizer) attaches to the root
in –tos participles, to vP in –menos target state participles and to VoiceP in –menos resultant
state participles:
(21) [AspP Asp X] (where X= Root, vP or VP)
-tos verbal adjectives lack event implications (they are licensed after verbs of creation in (13),
they do not permit result-oriented modification) and agentivity (no agent-oriented
modification, no by-phrases and instruments). A&A (2008) take this to mean that they involve
root-attachment:
(22) ASP root attachment of Asp 3 ANIG ASP
OPEN
-t-
-menos target state participles which include the implication of an event (they are not licensed
after verbs of creation in (13), they license result oriented modifiers) but lack agentivity (no
agent-oriented modification, no by-phrases and instruments) involve v attachment (v is taken
by A&A 2008 to be the eventivizing head):
(23) Asp v attachment of Asp 3 v men 3 ANIG
open
Finally, -menos resultant state participles which include both implication of an event and
agentivity (agent-oriented modification, licensing of by-phrases and instruments) involve
Voice attachment (Voice is taken to introduce the external argument):
9
(24) Asp Voice attachment of Asp 3 Voice men 3 v 3 ANIG v
open
Important for present purposes is the proposal that –tos verbal adjectives instantiate ‘inner-
cycle’ attachment while –menos participles ‘outer-cycle’ attachment:
(25) root-cycle outer-cycle attachment
-menos functional head
-tos root
… root…
The distinction in (25) leads us to expect that –tos forms will show properties of inner
affixation and –menos forms of outer affixation from the perspective of hypotheses (1)/(8).
Prima facie evidence in favor of the “inner” vs. “outer” division in (25) comes from the
observation that there is a striking difference in the productivity of –menos as opposed to –tos
participles. While all verbs of the appropriate semantic type, i.e. all telic, many atelic and
even some (coerced?) statives, can form –menos participles, there are many gaps in the
formation of –tos participles. Some examples illustrating this are listed in (26):5
(26) a. vizag-menos *vizax-tos ‘nursed’
b. rimag-menos *rimax-tos ‘ruined’
c. doule-menos *doulef-tos ‘worked out’
d. louz-menos *lous-tos ‘shampoo bathed’
e. pirag-menos *pirax-tos ‘hurt, tampered
3.2. A more complex picture: different types of –tos participles
A closer look into the properties of –tos verbal adjectives reveals that the Anagnostopoulou
(2003)/ A&A (2008) classification of participles in (25) needs to be refined. As it turns out,
there are some clear cases of –tos involving outer cycle attachment and some more
controversial ones:
(I) A clear case of outer cycle attachment is discussed in Samioti (2008, in progress).
As pointed out by Markantonatou et al (1996), there is also a class of –tos participles denoting
ability/ possibility rather than characteristic state:
(27) a. Afti i dikaiologia ine pistef-t-i
This the excuse is believe-t-fem.sg.nom
‘This excuse can be believed/ is believable’
5
Note that out of 2722 verbs listed in the online Lexicon of Anastasiadi-Simeonidi
(http://www.komvos.edu.gr/dictionaries/dictOnLine/DictOnLineRev.htm) that have been checked so far, 688
verbs form both –menos and tos participles, 1866 form only –menos participles and 168 verbs only –tos
participles. These numbers need to be checked more carefully.
10
b. Afto to asteri ine ora-t-o
This the star is see-t-neut.sg.nom
‘This star can be seen/ is visible’
Other participles of this type are katortho-t-os ‘achievable’, bore-t-os ‘doable’, epitefk-t-os
‘doable’, antilip-t-os ‘perceivable’, aisthi-t-os ‘perceivable’, ap-t-os ‘touchable’, thea-t-os
‘visible’, ia-t-os ‘curable’, fori-t-os ‘transportable’, noi-t-os ‘conceivale, thinkable’, katanoi-t-
os ‘understandable’, anek-t-os ‘tolerable’, ipofer-t-os ‘tolerable’ etc. Samioti (2008, in
progress) argues that –tos in ability participles attaches above v. Evidence for this is drawn
from the fact that they license agent (28a) and instrument PPs (28b) and agent-oriented
adverbs of the type found in English middles (29):
(28) a. I istoria tou ine pistef-t-i apo olous.
The story his is believable by everyone
‘His story can be believed by everyone’
b. To vouno ine ora-t-o me kialia.
The mountain is visible with binoculars
‘The mountain is visible with binoculars’
(29) To mathima ine efkola katanoi-t-o.
the lesson is easily understandable
‘The lesson can be easily understood’
Samioti argues that ability –tos participles pattern syntactically with Greek middles as these
have been described by Lekakou (2005). According to Lekakou, the middle Voice in Greek is
a variant of the passive Voice, unlike English and other languages. An implicit external
argument is present and, therefore, by-phrases and instruments are licit (see Lekakou 2005 for
detailed argumentation). Adopting this analysis for ability/possibility participles, Samioti
proposes that –tos attaches to the (middle) VoiceP (see Samioti 2008, in progress, for details).
(II) A related, though less straightforward, case of high attachment is illustrated in
(30). Greek has a productive process yielding ability/possibility –tos participles which consist
of the verbal root and the adjectival prefix aksio- 'worth-':
(30) a. aksi-agapi-tos : worth loving
b. aksi-o-thavmas-tos: worth admiring
c. aksi-o-meleti-tos worth studying
d. aksi-o-katafroni-tos worth despising
Semantically, such formations clearly implicate an agent (a generic/ impersonal agent
corresponding to English “one”), pointing to the presence of Voice. Syntactically, they differ
from the participles in (27)-(29) in that they do not license adverbs and instruments, and
agentive PPs are marginally tolerated when they express universal quantification:
(31) a. To arthro ine aksiomeletito (*prosektika)
The article is worth-studying carefully
'The article is worth studying carefully'
b. O planitis ine aksiomeletitos (*me to tileskopio)
The planet is worth-studying with the telescope
'The planet is worth studying with the telescope'
11
c. To arthro ine aksiomeletito (?apo olus/ ?apo ton kathena
The article is worth-studying by all/ by the everyone/
?apo ton opjondipote)
by the anyone
'The article is worth being studied by all/ everyone/ by anyone'
Morphologically, these participles involve compounding, as evidenced by the fact that they
systematically employ the compound marker –o- indicated in (30) (see Ralli 2001, 2006 for
discussion of –o- in Greek).6 We tentatively suggest that they indeed contain Voice (perhaps
the type of “middle” Voice also contained in ability/possibility participles), while the
restrictions in (31) reduce to restrictions on compounding.7
(III) Negated participles instantiate a third case where, arguably and more
controversially, –tos participles present an instance of outer cycle attachment. Negated
participles obligatorily surface with –tos in Greek, as shown by the fact that the negated –
menos forms in (35) and ungrammatical. Note that, even in cases when there is no
corresponding non-negated –tos form available, such as (33) and (34), the presence of -tos
instead of –menos is triggered by negation:
(32) a. gra-men-os b. a-graf-t-os (grap-t-os)
written un-written
(33) a. pli-men-os b. a-pli-t-os (*pli-tos)
washed un-washed
(34) a. diavas-menos b. a-diavas-t-os (*diavas-tos)
read un-read
6 Some examples of compounds displaying this marker are the following:
(i) a. xart-o-peksia ‘card playing’
b. trapez-o-ipallilos ‘bank employee’
c. maxer-o-piruno ‘knife & fork’
d. Afstr-o-Ungaria ‘Austria-Hungary’ 7 The issue of compounding deserves further investigation. Samioti (in progress) compares the properties of
aksio-compounds to-tos participles involving compounding with the manner adverbs efkola- ‘easily’ and
dhiskola ‘with difficulty’. These are formations like efkol-o-diavas-tos lit. easy-read-tos “something that can be
easily read” and diskol-o-diavas-tos lit. difficult-read-tos “something hard to read’. Samioti points out the
following: (i) All three types of compounds disallow instrument PPs. (ii) In efkolo- and dhiskolo-compounds, the
adverbs efkola- ‘easily’ and dhiskola ‘with difficulty’ are manner adverbs, a fact which can potentially explain
why further manner adverbs cannot be licensed with them. Note, though, that this explanation does not
immediately carry over to aksio-compounds, because aksio- ‘worth’ is not a manner but rather an evaluative
adverb. (iii) Interestingly, aksio-compounds differ from efkolo- and dhiskolo- compounds w.r.t. the type of
agentive PP they license. While aksio-compounds marginally license PPs introduced by the preposition apo
‘from’, which generally introduces agents in Greek (see e.g. Alexiadou & Anagnostopoulou 2004, Alexiadou,
Anagnostopoulou & Schäfer 2006 for discussion and references), efkolo- and dhiskolo- compounds license
‘agentive’ PPs introduced by the preposition gia ‘for’, as in (i) (we would rather call these PPs ‘relevance’ PPs,
as they express the person for which something is easy or difficult to do):
(i) Afto to fagito ine efkol-o-magirefto gia oles tis nikokires
This the food is easily-cook-to for all the housewives
‘This food is easy to cook for all housewives’
A similar fact has been noted for some (though not all) middles, namely the existence of well-formed examples
like No Latin text translated easily for Bill (compare though to *These books don’t sell for the average
shopkeeper); see among others Stroik (1992, 1995, 1999) and Ackema & Schoorlemmer (1995) for discussion.
12
(35) a. *a-diavas-menos b. *a-pli-men-os
un-read un-washed
At first sight, negated participles behave as if they involve attachment to the root. They do not
imply a prior event, and this correlates with the lack of the key syntactic properties associated
with a vP/VoiceP architecture (see Anagnostopoulou 2003 following Kratzer 1994, and A&A
2008 who adopt a root-attachment analysis based on the facts in (36)-(38)). More specifically,
they resist manner modification, as illustrated in (36b), and they do not license by-phrases and
instruments, as shown in (37b) and (38b), respectively. In all respects they differ from their
non-negated –menos counterparts in (36a), (37a) and (38a):
(36) a. I Maria ine prosektika xtenis-men-i manner adverbs
The Maria is carefully combed
'Mary is carefully combed'
b. *I Maria ine aprosekta a-xtensi-t-i
The Maria is sloppily/not carefully un-combed
'*Mary is sloppily uncombed'
(37) a. To vivlio ine gra-men-o apo tin Maria by-phrases
The book is written by the Mary
'The book is written by Mary'
b. *To vivlio ine a-graf-t-o apo tin Maria
The book is unwritten by the Mary
'*The book is unwritten by Mary'
(38) a. To vivlio ine gra-men-o me grafomixani instruments
The book is written with typewritter
'The book is written with a typewritter'
b. *To vivlio ine a-graf-t-o me grafomixani
The book is unwritten with typewritter
'*The book is unwritten with a typewritter'
From the point of view of their productivity, morphology and interpretation, however, they
behave as if they involve outer cycle attachment (cf. Embick 2004: 359, fn 6 for English
negated un-participles). First, negated –tos participles are fully productive, similarly to –
menos partciples and unlike –tos ones. Second, in cases of morphologically conditioned stem
allomorphy, negated –tos participles do not show the same allomorphy as the corresponding
non-negated –tos form:
(39) a. a-graf-tos b. grap-tos c. *a-grap-tos
un-written written un-written
Third, they clearly form the negation of –menos participles, in cases where there is no
corresponding –tos form (see (33) and (34) above). And fourth, when all three forms exist,
negated participles negate –menos and not –tos participles. For example, a-graf-t-os in (32)
above means ‘something that has not been yet written’ and not ‘something that is not in a
written form’ (see Setatos 1984; Anastasiadi-Simeonidi 1994). But if negated participles
involve the negation of the –menos forms, then this leads to the postulation of an “outer-
cycle” architecture for them (contra Anagnostopoulou 2003 and A&A 2008). Under such an
analysis, the reason why -tos and not -menos surfaces on negated participles might be a
13
morpho-phonological one, related to stress: the suffix -ménos requires stress while the
negation prefix a- triggers obligatory recession of stress, a conflict resolved by the insertion
of -tos.
(IV) The final case we would like to discuss is the most interesting one from the
perspective of hypotheses (1)/(8) discussed in this paper. Note first that Greek productively
employs verbalizing suffixes which have been analyzed as root verbalizers by Alexiadou
(2001, 2009; see Giannakidou & Merchant 1999; Ralli 2001 for discussion of these elements):
(40) Root- verbalizing elements
Greek: -iz, - on-, -en/an, -ev,- -az, -a Alexiadou (2001, 2009):
(41) a. aspr-iz-o, kathar-iz-o b. pag-on-o ler-on-o
whiten cleaned freeze dirty
c. diaplat-en-o, arost-en-o d. sten-ev-o, berd-ev-o
widen, become sick tighten , confuse
e. diav-az-o, mir-az-o f. pul-a-o xal-a-o
read split, share sell destroy
As expected by the “outer-analysis” of –menos participles in (25), verbalizers (or their
allomorphs as in the (c), (f) examples) are systematically present in –menos participles,
providing morphological evidence for outer cycle attachment:
(42) a. aspr-iz-menos, kathar-iz-menos
whitened, cleaned
b. pag-o-menos, ler-o-menos
frozen, dirtied
c. diaplat-i-menos, arost-i-menos
widened, sickened
d. sten-e-menos, berd-e-menos
tightened, confused
e. diav-az-menos, mir-az-menos
read split, shared
f. pul-i-menos xal-az-menos
sold destroyed
Root verbalizers are often disallowed in –tos participles, a fact which was taken by A&A
(2008) as evidence for root attachment on –tos in (25):8
(43) a. *aspr-is-tos, *kathar-is-tos
whitened, cleaned
b. *pag-o-tos, *ler-o-tos
frozen, dirtied
c. *diaplat-i-tos, *arost-i-tos
widened, sickened
d. *sten-ef-tos, *berd-ef-tos
tightened, confused
8 For some of these examples (the ones based on adjectival roots) there is an alternative well-formed formation
based on the root + the adjectival ending:
*kathar-is-tos (cleaned) vs. kathar-os (clean), *aspr-is-tos (whitened) vs. aspr-os (white), *sten-ef-tos
(tightened) vs. sten-os (tight). A&A 2008 analyse this as an instance of blocking. See below for discussion.
14
e. *diav-as-tos, *mir-as-tos
read split, shared
But, quite unexpectedly, many Greek characteristic state –tos participles that do include such
verbalizing elements:
(44) a. axn-is-tos ‘steaming hot’ axn-iz-o ‘steam’
b. koudoun-is-tos ‘ringing’ koudoun-iz-o ‘ring (a bell)’
c. vathoul-o-tos ‘hollow’ vathoul-on-o ‘hollow out’
d. vid-o-tos ‘screwed’ vid-on-o ‘screw’
e. if-an-tos ‘woven’ if-en-o ‘weave’
f. magir-ef-tos ‘cooked’ magir-ev-o ‘cook’
g. fit-ef-tos ‘planted’ fit-ev-o ‘plant’
h. angali-as-tos ‘embraced’ angali-az-o ‘embrase’
i. evodi-as-tos ‘fragrant’ evodi-az-o ‘be fragrant’
Despite the presence of verbalizers, the –tos participles in (44) do not have event implications
(they denote characteristic states), and they do not license manner modification, agent PPs or
instruments:
(45) a. *To fagito ine kala/ prosektika magir-ef-t-o
The food is well/ carefully cooked
b. *To fagito ine magir-ef-t-o apo tin Maria
The food is cooked by the Mary
c. *Ta fita ine fit-ef-t-a me diaforetika ergalia
The plants are planted with different instruments
The characteristic state participles of the type illustrated in (44) are important to our
discussion, as they show that the abstract little v heads described in the decomposition
literature, i.e. the semi-functional heads introducing eventive interpretations (and licensing
result-oriented manner modification) must be dissociated from morphological verbalizers.
This poses an immediate problem concerning the proper understanding of (1) and (8): the
question is raised what counts as a phase head in –menos and –tos participles. The verbalizers
in (40) which are also present in –tos verbal adjectives like (44) or the abstract eventive little
v heads licensing modifiers and PPs, which are consistently absent from characteristic state –
tos verbal adjectives, even when they include a morphological verbalizer?
In the next sections, we will argue that morphological verbalizers indeed serve as
contexts for meaning assignment to unspecified roots, in accordance with (1)/(8), as these
generalizations have been interpreted by Arad (2003, 2005) in her work on Hebrew. But this
is a separate issue from the question of idiomatic interpretations, where the domain for
unpredictable meaning of participles is considerably larger, and neither verbalizers nor
eventivizers have a special status, but rather Voice.
3.3. Revisions and Generalizations
The discussion in the previous sections has led us to revise the typology of Greek verbal
adjectives in two respects:
a) First, –tos does not always attach below little v. There is one clear case where -tos attaches
above little v (ability –tos, Samioti 2008; in progress) and two more cases where -tos possibly
15
attaches above little v (aksio-compounding, negated –tos participles). It is an open question at
this point whether the different types of –tos verbal adjectives involve a single morphological
piece -tos inserted in stative environments lacking event implications, with the three different
interpretations resulting from the specific environments of insertion (a case of polysemy) or
whether Greek has three different -tos, i.e. ability -tos, negation -tos, characteristic state –tos
(a case of homophony). It is interesting though that Greek is not unique in showing the same
morphology for stativized expressions and for expressions denoting ability/possibility, i.e. the
counterparts of –able adjectives in English. Chichewa (Dubinsky & Simango 1996: 759, ex.
(17)) and Malagasy (Travis 2005) are similar in this respect.
b) Second, the morphological decomposition of –tos characteristic state verbal adjectives
leads to the identification of verbalizing heads in a number of cases, as was shown by the
examples (44). Crucially, as shown by (45), these morphological verbalizers do not have the
syntactico-semantic properties of the abstract little v head identified and discussed in the
decomposition literature (Embick 2004; Alexiadou, Anagnostopoulou & Schäfer 2006;
Marantz 2001, 2007; Harley 2012 and related literature), i.e. they do not contribute
eventiveness nor do they license modifiers and argumental PPs. Alexiadou (2009) reaches
exactly the same conclusion for result nominals in Greek, which may contain a verbalizer
which however does not contribute eventiveness and does not license arguments.
In view of these revisions, what are then the generalizations governing (i) the
distribution of –menos vs. –tos in participles/ verbal adjectives and (ii) the presence vs.
absence of verbalizers in Greek verbal adjectives?
Starting from the first issue, the affix –menos signifies the Perfect of Result, i.e. it is
present whenever the participle denotes a target or resultant state resulting from a prior event
(Kratzer 1994; 2001). For –tos it is unclear whether it presents a case of polysemy or
homophony. Under the first option, there is a single morphological piece –tos inserted
whenever there is no prior event implied, maybe as an elsewhere form in the absence of a
Perfect of Result operator. Alternatively, there are three different –tos morphemes, namely
ability -tos, characteristic state –tos, and the negated Perfect of Result -tos inserted for
morphophonological reasons. We conclude that verbal adjectives with –tos involve either
inner- or outer-cycle attachment, contra A&& (2008), namely:
(i) Ability/ possibility –tos forms clearly show outer-cycle attachment, and they contain a
(middle) VoiceP.
(ii) –tos forms involving ‘aksi-o-’ compounding (e.g. aksi-o-meletitos ‘worth-studying’)
possibly also contain a VoiceP and hence are formed by outer-cycle attachment.
(iii) When a target/ resultant state participle is negated there are semantic and
morphological reasons for positing –tos attachment above vP or VoiceP, though the syntactic
evidence is less clear.
(iv) Characteristic state verbal adjectives do not have event implications and lack all
syntactic properties associated with a little v head providing evidence for inner-cycle
attachment. They pose, however, a puzzle. While –tos attaches directly to the root in a
number of cases, it attaches to the Root+verbalizer in another set of cases. This leads to the
second question that needs to be addressed, namely what is the explanation for the presence
vs. absence of morphological verbalizers in –tos characteristic state participles. In answering
this question, we would like to pursue hypothesis (46):
(46) Selection hypothesis:
-tos selects expressions naming events
16
With Harley (2005) we assume that roots fall into basic ontological types naming events,
things and states. We can then formulate the following generalizations concerning the
absence or presence of verbalizers inside –tos:
Generalization I: -tos directly attaches to roots which can be characterized as
Rootevent in terms of Harley’s ontology, as shown in (47). These forms involve roots with a
specified meaning (they express particular types of events) yielding verbs (as in the first
column of 47) by combining with the verbal inflectional ending either directly (as in
47b,d,e,f,g) or via the formative –n- (in 47a,c).9
(47) a. ftiax-n-o ‘make’ ftiax-tos ‘made’
b. lin-o ‘loosen’ li-tos ‘loose’
c. per-n-o ‘take’ par-tos ‘taken’
d. klev-o ‘steal’ klef-tos ‘stolen’
e. din-o ‘give’ do-tos ‘given’
f. plek-o ‘knit’ plek-tos ‘knitten’
g. klin-o ‘close’ klis-tos ‘closed’
In these cases, the selection requirement of –tos stated in (46) is satisfied because –tos
directly combines with roots naming events (and see fn. 6 on the question of whether root or
stem is the appropriate notion).
Generalization II: -tos does not combine with Rootthing. It combines with Rootthing +
verbalizer10
. In a number of cases illustrated in (48) roots have a specified meaning naming a
thing and can be combined directly with nominal inflection forming a noun (first column in
48). –tos cannot directly combine with such roots. As shown by the third column in (48), they
first become verbal by combining with a verbalizer (cf. the second column in (48) where such
roots are turned into verbs by a verbalizer plus the verbal inflection), and then –tos attaches to
the root + verbalizer complex:
(48) a. afr-os ‘foamN’ afr-iz-o ‘foamV’ afr-is-tos ‘foaming’
b. axn-os ‘steamN’ axn-iz-o ‘steamV’ axn-is-tos ‘steaming hot’
c. vid-a ‘screwN’ vid-on-o ‘screwV’ vid-o-tos ‘screwed’
d. koudoun-i ‘ringN’ koudoun-iz-o ‘ringV’ koudoun-is-tos ‘ringing’
e. koumb-i ‘buttonN’ koumb-on-o ‘buttonV’ koumb-o-tos ‘buttoned’
f. skep-i ‘roof’ skep-az-o ‘cover’ skep-as-tos ‘covered’
Note that in some cases, such as (49), the denominal verb formed by the Rootthing + verbalizer
is deviant (second column in 49, i.e. such expressions cannot name events and therefore they
cannot be verbs), while the corresponding –tos participle is perfect:
(49) kamban-a ‘bellN’ ??kamban-iz-o ‘bellV’ kamban-is-tos ‘sounding like a bell’
We propose that the presence of a verbalizer in formations of this kind is necessary to satisfy
the selection requirements of –tos in (46).
9 There are a number of further observations one can make here. First, the absence of –n- in the verbal adjectives
as in (47a) could be viewed as evidence that –tos directly attaches to the root and not to the verb, if it can be
ensured that the reason for the absence of –n- is not morpho-phonological. Second, in (47c, e, g) –tos attaches to
the perfective stem (marked by stem allomorphy in 47c, e and by the presence of –s- in 47g), a fact that could, in
principle, receive either a semantic or a phonological explanation. The question raised by such examples is to
what extent it is legitimate to talk in these cases about –tos attachment to roots rather than stems. We have to
leave these questions open for now. 10
See Alexiadou, 2009, for discussion of such formations.
17
An issue that needs to be discussed at this point concerns the criteria by which the
verbal and adjectival forms in the second and third columns of (48), (49) above are indeed
root derived (based on Rootthing), as opposed to being derived from zero related nouns (e.g.
based on Root plus a little n with a zero exponent). A way to decide this is to apply Kiparsky's
(1982) criteria. Kiparsky argues that root-based formations do not entail the existence of the
corresponding nouns while noun-derived ones do entail the existence of the corresponding
nouns (see Arad 2003, 2005 for extensive discussion of Hebrew based on Kiparsky). For
many of the forms in (48), (49) it can indeed be demonstrated that they are root derived. For
example, vidono and vidotos in (48c) do not entail the existence of a screw but rather they
refer to the type of movement required for putting two pieces together. In a similar manner,
koumbono and koumbotos in (48e) do not entail the existence of a button (they refer to closing
something by making use of a particular device), koudounizo and koudounistos (48d)
/kambanistos (49) do not entail the existence of a ring/bell (they mean sounding like a
ring/bell), skepazo and skepastos (48f) do not entail a roof (they mean cover and being
covered) and axnizo and axnistos (48b) do not entail steam (they mean hot and cosy like
anything that is steaming hot).11
Generalization III: As shown in (50), -tos does not combine with Rootstate + verbalizer
because there is an adjective blocking the –tos form (as pointed out by A&A 2008).
(50) aspr-iz-o 'whiten' aspr-iz-men-os aspr-os/i/o 'white' *is-tos
mavr-iz-o 'blacken' mavr-iz-men-os mavr-os/i/o 'black' *is-tos
kitrin-iz-o 'yellow' kitrin-iz-men-os kitrin-os/i/o 'yellow' *is-tos
prasin-iz-o 'green' prasin-iz-men-os prasin-os/i/o 'green' *is-tos
kathar-iz-o 'clean' kathar-iz-men-os kathar-os/i/o 'clean' *is-tos
megal-on-o 'grow' megal-o-men-os megalos 'big' *o-tos
In principle, however, the formation Rootstate + verbalizer + tos is possible. And indeed, such
forms do exist, but they have specialized uses. For example, the –tos form in (51) is used only
for food:
(51) kokin-iz-o 'redden' kokin-os/i/o 'red' kokin-is-tos ‘with a red sauce’
Overall, we have checked many –tos participles containing the verbalizers –iz-, -az-, -ar-, -
on-, -ev-, and it turns out that the majority of them are as in (48)/(49), i.e. they are based on
Rootthing+ verbalizer. These have corresponding nouns of the form Rootthing+ nominal
inflection. A few –tos participles with a verbalizer are based on Rootstate (adjective or adverb/
preposition), as in (51) and (52) below:
(52) a. stogil-ev-tos round-ef-tos ‘round/ rounded’
b. thab-o-tos misty-o-tos ‘misty/ blurred’
c. xor-is-tos without-is-tos ‘separate’
d. antam-o-tos together-o-tos ‘together’
There is a final set of cases, to be discussed immediately, which leads to a particular
interpretation of the hypothesis (1)/(8).
11
Of course, one would have to check all such formations available in the Greek lexicon in order to draw firm
conclusions concerning this issue. For example, by Kiparsky’s criteria the verbal and adjectival forms in (48a)
seem to qualify as being based on a zero-derived noun.
18
Generalization IV: verbalizers turn undefined roots into Constructevent and then –tos attaches
to the Rootundefined + verbalizer. A group of –tos participles is based on a Root with no clear
meaning (call it Rootundefined ) which combines with a verbalizer. This type of root has
unspecified meaning in the sense that (a) one couldn’t assign an exact meaning to it and (b)
there is no corresponding noun or adjective or verb of the form Rootundefined + Inflection. Two
different subcases fall under this category.
(i) A number of undefined roots represent sounds or movements or shapes (they are often
formed by reduplication). A verbalizer must necessarily attach to them before they become
verbs (surfacing with verbal inflectional endings), and then they enter further derivation
(becoming adjectives, as in (53), or nouns):
(53) a. kakar-is-tos ‘cackling’
b. tsitsir-is-tos ‘sizzling / frizzling’
c. trekl-is-tos ‘staggering’
d. tourtour-is-tos ‘shivering / shuddering’
e. gourl-o-tos ‘goggling’
f. koxl-as-tos ‘bubbling’
g. xarxal-ef-tos ‘rummaging’
h. paspat-ef-tos ‘fiddling’
(ii) There is a residue of roots which seem completely and totally undefined before a
verbalizer attaches to them, making them of type “event”. –tos suffixation follows:
(54) a. kt-is-tos ‘built’
b. sk-is-tos ‘slit’
c. str-o-tos ‘smooth, regular’
d. lig-is-tos ‘bent’
e. sik-o-tos ‘raised’
The two types of roots described above, especially the ones in (54), seem to us to be very
close to what Arad (2003, 2005) describes for Hebrew. One couldn’t exactly tell what they
mean. According to Arad (2003, 2005), there are two types of languages: In Hebrew-type
languages (also Georgian, Russian) roots may be assigned numerous interpretations in
different morphophonological environments (MCM). This correlates with the fact that (i)
roots are semantically underspecified and (ii) the inventory of roots in the languages in
question is small. On the other hand, in English-type languages most roots are assigned
meaning in one environment only (exception to this are Latinate bound roots like √fer, √cieve
etc.; see fn 12 below). This correlates with the fact that (i) they are semantically specified and
(ii) the inventory of roots is large. Greek seems to be an English-type language. A large
number of Greek roots are semantically specified.12
However, there are exceptions to this in
both English (the latinate roots mentioned above; see also fn. 12) and Greek (see Alexiadou &
Anagnostopoulou 2011, to appear for discussion of the Greek counterparts of words based on
proto-indoeuropean roots, where prefixes drawn from the Ancient Greek prepositional
inventory fix the meaning of unspecified roots). The examples in (53), (54) constitute a
further class of exceptions. In the cases listed in (53) and (54) lack of clear/ stable meaning
correlates with the inability to classify the roots according to Harley's ontology. Adding a
verbalizer to these roots on the one hand makes them classifiable in terms of Harley’s basic
12
As in English, exception to this are roots with a proto-Indo-European origin like √fer. These form multiple
verbs with very different interpretations depending on the prefix they combine with: dia-fer-o: differ, pro-fer-o:
pronounce, ana-fer-o: report, pros-fer-o: offer, ek-fer-o: formulate.
19
ontology and at the same time it fixes their meaning, which is then retained throughout the
derivation. We believe that this double function of verbalizers in the case at hand provides a
key towards understanding why categorizers serve as contexts for meaning assignment to
unspecified roots. Roots must be classifiable in terms of a basic ontology naming events,
things and states. If they are not, as in (53), (54) or in the cases of the Hebrew type MCM
Arad discusses, a categorizing head serves to classify them. The classification itself makes
them acquire what is sensed as fixed “meaning”, which is then retained throughout the
derivation. We conclude that the Greek facts in (53) and (54) support an interpretation of
(1)/(8) along the lines of Arad (2003, 2005), where the crucial head is the categorizing head
(in the cases we discuss vC) which makes roots classifiable in terms of Harley’s (2005) basic
ontology.
3.4. Summary: How many participles?
We have arrived at the following picture of Greek adjectival participles and verbal adjectives
w.r.t. the question of outer vs. inner cycle architecture:
(55) Greek Participles and Verbal Adjectives
1) Two types of -menos participles:
a) target state -menos participles outer attachment
b) resultant state -menos participles outer attachment
2) Four types of -tos participles:
a) ability -tos participles, outer attachment
b) aksio- compounding -tos participles outer attachment (possibly)
c) negation -tos participles outer attachment (possibly)
d) characteristic state -tos participles : direct attachment to Rootevent
attachment to Rootthing+ verbalizer
attachment to Rootstate+ verbalizer
attachment to Rootundefined+ verbalizer
(56) root-cycle = tos + Rootevent outer-cycle attachment= ability-tos
-tos VoiceP
-tos root
Voice … root…
Does not count as outer-cycle = tos + [Rootthing/state/undefined + verbalizer]
3 -tos 3
v root
There is one question raised by these participles which will only be mentioned briefly and
will not be answered here concerning the polysemy vs. homophony issue mentioned earlier in
20
the paper. In principle, this question can be asked for both –menos participles and –tos
participles. For –menos participles the question is the following. Are there two different –
menos suffixes one yielding target states and one yielding resultant states which have
different selection requirements and different syntax (target state –menos necessarily
attaching below Voice) or is there one –menos yielding resultant states or target states
depending on the context of insertion? In view of the close similarity in meaning between
resultant and target state participles, we are inclined to think that the second option is more
plausible: there is only one –menos suffix yielding states resulting from events; the resultant
vs. target state difference must derive from the context, e.g. the idiosyncratic properties of the
root, the size of the complement (if the complement is VoiceP then the state is necessarily a
resultant state), plus some further properties of the structure. For example, the presence of
adverbs has been noted by Kratzer to provide the component necessary for a target state to
verbal roots otherwise yielding resultant states ((57) is Kratzer’s example (21)):
(57) a. * Meine Haare waren immer noch geschnitten.
My hairs were still cut
‘My hair was still cut’
b. Meine Haare waren immer noch schlampig geschnitten
My hairs were still sloppily cut
‘My hair was still cut sloppily’
For –tos the issue is more involved since it is not clear that –tos has a uniform semantics in
characteristic state participles, negated participles and ability/possibility participles.
Moreover, the cross-linguistic evidence we reviewed does not provide straightforward support
in favor of the one over the other alternative hypothesis. On the one hand, English
distinguishes –ed (in negated and simple state adjectival participles) from –able (for
ability/possibility adjectives) supporting the homophony hypothesis. On the other hand,
Chichewa and Malagasy use the same element for both types of adjectives, like Greek,
providing crosslinguistic support in favor of the polysemy alternative (the unifying property
of all three kinds being the absence of a prior event).
Turning to the issue of non-compositionality discussed in this paper, characteristic state –
tos verbal adjectives are particularly relevant to the hypothesis (1)/(8) for two reasons:
a) First, they present evidence that the presence of a verbalizer does not, in itself,
contribute eventiveness and the syntactic properties associated with it.
b) Second, they present evidence that the presence of a verbalizer fixes the meaning of
unspecified roots which cannot be classified according to Harley's typology.13
We are now in a position to turn to the question of idiomaticity. We do so in the next section,
and we refer the reader to Anagnostopoulou & Samioti (in progress) for a more complete
discussion.
4. A bigger domain for Idioms
4.1. Neither vC nor vE are not boundaries for idioms
We assume a working definition of idioms along the lines of Svenonius (2005: 1): “....listed
syntactic structures which […] have unpredictable meanings in the way words do, but consist
of more than one “piece”. In multi-word idioms, “piece” is the phonological word. In single-
13
The cases where aspectual prefixes determine the meaning of verbal (i.e. event denoting) roots, as in English
with words based on Latinate roots, in Russian (Arad 2003: 775, 2005) and in Greek (mentioned in footnote 12)
should also be taken into account and might lead to a partial modification of this statement.
21
word idioms “piece” is the head (Roots, affixes). Here we are looking at single word
idioms. In view of the dissociation between verbalizing heads (categorizing heads, vC) and
little v (the head contributing eventiveness, licensing arguments etc., vE) introduced in the
previous discussion, the Marantz/ Arad hypothesis as a hypothesis about idioms can, in
principle, be tested in two ways:
a) Taking literally the proposal that once a root is categorized it is assigned a range of
meanings fixed for the rest of the derivation, what should be tested is whether the presence of
a verbalizer in –tos participles yields such an effect. In other words, are there significant
differences between –tos verbal adjectives with a verbalizer and –tos adjectives without a
verbalizer w.r.t. idiomaticity, in comparison to the corresponding verbs?
b) Assuming, alternatively, that it is the presence of eventive v which defines a cyclic
domain for interpretation (phase), it should be tested whether the presence of an eventive
little v fixes meaning in a way that affixes attached above it always lead to predictable
interpretations (outer affixation). In other words, are there significant differences between –
tos and –menos forms and the corresponding verbs w.r.t. idiomaticity?
The answer to the first question is negative. The internal composition of –tos verbal
adjectives denoting characteristic states does not correlate in any way with (non-)idiomaticity.
Characteristic state verbal adjectives of both types (simplex without a verbalizer or complex
with a verbalizer) can have idiomatic readings lacking from the corresponding verbs. This is
illustrated in (58) and (59) below:
Characteristic state -tos verbal adjectives showing direct attachment of –tos to Rootevent:
Verb Participle Idiomatic interpretation
of participle only
(58) a. sfing-o sfix-tos
tighten tight ‘careful with money’
b. ftin-o ftis-tos
spit lit. spitted ‘spitting image’
c. klin-o klis-tos
close lit. closed ‘introverted’
Characteristic state -tos verbal adjectives showing attachment of –tos to Root+ verbalizer:
Verb Participle Idiomatic interpretation
of participle only
(59) a. kol-a-o kol-i-tos ‘close friend’
glue-1sg lit. glued
b. xtip-a-o xtip-i-tos ‘striking’
bang, hit, whip lit. whipped
c. xon-ev-o xon-ef-tos ‘inside the wall’
digest no lit. meaning
d. karf-on-o karf-o-tos ‘very fast/ direct’
nail no lit. meaning
The answer to the second question is again negative. -menos participles may have
idiosyncratic meanings, just like the –tos verbal adjectives in (58) and (59):
(60) striv-o stri-menos geros strif-to tsigaro
twist 'crotchety old man’ lit. twisted (rolled) cigarette
22
Strikingly, the –menos participle in example (60) only has the idiomatic reading when
modifying a [+human] noun, while the verb and the –tos participle can only have the literal
meaning. But this is the reverse of what is expected if (1)/(8) are understood as applying to
idioms and if the relevant phase head is taken to be vE (the abstract eventivizing head). What
would be expected under such an interpretation of (1)/(8) is that the characteristic state –tos
participle has the idiomatic reading and the –menos participle the compositional meaning. It
is furthermore expected that the idiomatic reading of the –menos participle depends on the
idiomatic interpretation of the corresponding verb, i.e. exactly the opposite of what we see in
(60). As we discuss in detail in Anagnostopoulou & Samioti (in progress), there is extensive
evidence that there is no necessary correlation between the meaning of verbs and the meaning
of the corresponding –menos participles. We distinguish between the following three cases:
i. Idiomatic verb Non-idiomatic participle
kathar-iz-o kathar-iz-menos
lit. ‘clean’ only lit. ‘cleaned’
idiom. ‘kill’
(61) a. ton katharisan xtes to vrady
him cleaned-3sg yesterday the evening
‘They killed him yesterday evening’
b. ?aftos o anthropos ine katharismenos
this the man is cleaned
‘This man is cleaned’
ii. Non-idiomatic verb Idiomatic participle
trav-a-o trav-ig-menos
only lit. ‘pull’14
lit. pulled
idiom ‘far fetched’
Verb
(62)
a. O Janis travik-s-e tin porta
The Janis pulled the door
‘Janis pulled the door’
b. *O Janis travik-s-e to epixirima
The Janis pulled the argument
‘*Janis pulled the argument’
Participle
(63) a. ?I porta ine trav-ig-meni
The door is pulled
‘Janis is pulled’ lit interpretation
14
This verb can have an idiomatic reading as a phrasal idiom (i) either when it combines with the prefix para
(meaning exaggeration) or when combined with clitic doubled object to skini ‘the rope’:
(i) a. O Janis to paratravikse
The Janis it para-pulled
‘John went too far’
b. O Janis to travikse to skini
The Janis it pulled the rope
‘John went too far’
Crucially, for present purposes, one does not ‘pull the argument’ or ‘pull the story’; compare (62b) to (63b).
23
b. To epixirima ine trav-ig-meno
The argument is pulled
‘The argument is far fetched’ Only idiom. interpretation
iii. Idiomatic verb: one meaning Idiomatic participle: another
(64) xon-ev-o xon-e-menos
lit. ‘digest’ lit. ‘digested’
idiom. ‘like’ idiom. ‘understood’
Verb – Participle literal
(65) a. O Janis xonep-s-e to fagito
The Janis digested the food
‘Janis digested the food’
b. To fagito ine xone-meno
The food is digested
‘The food is digested’
Verb idiomatic – Participle idiomatic
(66) a. O Janis den xonevi ta mathimatika
The Janis not digests the math
‘Janis dislikes mathematics’
b. Ta mathimatika den ine xone-mena
The mathematics not are digested
‘Math is not understood
In conclusion, neither vC nor vE constitute boundaries for idiosyncratic/ un-predictable/ non-
compositional meanings of participles in Greek, contrary to what one might expect on the
basis of (1)/(8).
4.2. Agentive Voice is a boundary for idioms
While idiomatic interpretations are not blocked by vC or vE, agentive features systematically
destroy non-compositional interpretations of –tos and –menos forms. Specifically:
(a) –tos verbal adjectives denoting ability/ possibility which implicate an agent, never
have idiomatic readings. All participles belonging to this class have exactly the same
meanings as the corresponding verbs, and their interpretations are always predictable. Some
of these forms are listed below:
(67) katortho-t-os ‘achievable’, bore-t-os ‘able’ ‘doable’, epitefk-t-os ‘doable’, antilip-t-os
‘perceivable’, aisthi-t-os ‘perceivable’, ap-t-os ‘touchable’, thea-t-os ‘visible’, ia-t-os
‘curable’, fori-t-os ‘transportable’, noi-t-os ‘conceivale, thinkable’, katanoi-t-os
‘understandable’, anek-t-os ‘tolerable’, ipofer-t-os ‘tolerable’ etc.
(b) Whenever the adjectival prefix aksio- (‘worth’) combines with a idiomatic
participle yielding a modal ability/possibility interpretation which implicates an implicit
agent, the non-compositional meaning is lost, as shown in (68):
24
(68)
a. trav-ig-menos aksi-o-travix-tos
lit. pulled only lit: worth pulling
idiom ‘far fetched’
b. stri-menos aksi-o-strif-tos
lit. twisted only lit: worth twisting
idiom. ‘crotchety’
(c) Agentive adverbs like ‘deliberately’ and agent-oriented manner adverbs like
´carefully’ systematically block idiomatic interpretations (see Alexiadou, Anagnostopoulou &
Schäfer among many others for discussion of these properties in connection to agentive
Voice):
(69)
a. trav-ig-menos prosektika / skopima travigmenos
lit. pulled only lit.: carefully/ deliberately pulled
idiom ‘far fetched’
b. stri-menos prosektika/ skopima strimenos
lit. twisted only lit: carefully/ deliberately twisted
idiom. ‘crotchety’
(d) The same effect is trigerred by agentive and instrument PPs in (70):
(70) stri-menos jeros idiom. ‘crotchety man’
BUT
stri-menos apo kapion/ me kati only lit. ‘twisted by someone / with something’
The effects of agentivity illustrated above are consistent with Marantz’s (1996, 1997)
generalization that the syntactic head that projects an agent defines a locality domain for
special meanings:
(71)
3 boundary for domain of special meaning (Marantz’s 1997 (6))
agent 3 v
head projecting agent
Predictions: (Marantz’s 1997 (7)):
(72) a. No idioms with fixed agents
(root in agent position, context for special meaning)
b. No eventive-passive idioms, but possible non-eventive stative idioms)
c. No idioms with causative morpheme and lower agentive verb, but
possible idioms with causative and lower non-agentive verb)
In Marantz’s (1997) system, the difference between verbal and adjectival participles w.r.t.
idiomaticity, i.e. prediction (72b), is linked to agentivity; in Marantz (2001, 2007), on the
other hand, it is linked to the verbal/eventive vs. adjectival/stative nature of English
participles. While the two hypotheses are difficult to tell apart on the basis of English
participles which collapse agentivity with eventiveness and verb-hood, and lack of agentivity
25
with stativity and adjective-hood, the Greek data discussed in this paper show that agentivity
and not eventiveness or verb-hood is crucial. Throughout the paper and in this section, we
have only looked at adjectival/ stative forms with or without verbalizers, event implications
and agentivity. We saw that idiomatic meanings freely occur with –menos participles and -tos
verbal adjectives as long as the relevant morphological constructs do not implicate an agent.
But when agentivity features are present, compositional interpretations are enforced on both –
tos and –menos forms preserving the literal meanings of the corresponding verbs. Assuming,
furthermore, that agentivity is linked to the head Voice, this leads to the conclusion that Voice
is a boundary for special meanings within the word. In an approach based on phasehood, this,
in turn, entails that Voice and not v qualifies as a phase head (in line with Chomsky 2001;
contra Marantz 2007).
5. Concluding remarks
In this paper, we discussed the issue of (non-)compositionality of meaning on the basis of a
case study, namely Greek –menos participles and –tos verbal adjectives denoting
characteristic state, negation and ability/possibility. We investigated the Marantz/Arad
hypothesis in (1)
(1) The Marantz / Arad Hypothesis
Roots are assigned an interpretation in the context of the first category assigning
head/ phase head merged with them, which is then fixed throughout the derivation
and addressed two questions:
First, which head counts as a phase head for meaning assignment, the abstract
eventivizing head vE or the verbal categorizing head vC? We provided evidence in favor of
dissociating the two heads based on Greek characteristic state –tos verbal adjectives which
lack event implications but can include verbalizing heads. We offered some evidence that
verbalizing heads lacking event implications indeed serve as contexts for meaning assignment
to roots that have completely unspecified meaning and are therefore not classifiable in terms
of a basic typology “thing”, “event”, “state” (Harley 2005). This is in line with what Arad
(2003), (2005) proposes for Hebrew.
The second question concerns idiomaticity within words, what is the domain for
special meaning in single word idioms. We argued that this domain is larger than the first
phase head, whether we take the relevant head to be vE or vC. The boundary for special
meanings is Voice, i.e. the head that syntactically projects and agent, as proposed by Marantz
(1996, 1997) and contra Marantz (2001, 2007).
After we wrote the first draft of this paper, it came to our attention that Marantz (2011)
revises his 2001/2007 position with respect to idioms and adopts a view very close to the one
argued for in the present paper. Marantz (2011) distinguishes between two meaning domains:
(a) What he calls “the domain for contextual allosemy”. This is determined immediately, by
the first phase head, a contentful categorizing head, vE in our terms. Marantz argues that the
vC head we identify in our work on Greek participles (i.e. the verbalizer included in
characteristic state –tos participles) should be viewed as a semantically null head, akin to do
in English do-support, which, being semantically empty, does not block the (meaning
assignment) relationship between the phase head above it (taken to be little a, the adjective
forming head) and the root below it. (b) The domain for idioms. This must be separated from
contextual allosemy, and it has as a boundary the Voice head that introduces the external
argument, in accordance with Marantz (1996, 1997). We discuss Marantz (2011) in more
26
detail in Anagnostopoulou & Samioti (in progress), where we focus on the domains for
idioms.
References Ackema, P. & Schoorlemmer, M. (1995), Middles and nonmovement. Linguistic Inquiry 26:173–197.
Alexiadou, A. (2001), Functional structure in nominals: nominalisation and ergativity (Amsterdam:
John Benjamins,).
Alexiadou, A. 2009. On the role of syntactic locality in morphological processes: the case of Greek
nominals. A. Giannakidou & M. Rathert, QP, Nominalizations and the role of DP. Oxford
University Press.
Alexiadou, A & E. Anagnostopoulou. 2004 Voice Morphology in the Causative-Inchoative
Alternation. Evidence for a Non-Unified Structural Analysis of Unaccusatives. In: A. Alexiadou,
E. Anagnostopoulou and M. Everaert (Eds.) The Unaccusativity Puzzle,114-136. Oxford: OUP.
Alexiadou, A. & E. Anagnostopoulou. 2008. Structuring participles. Proceedings of WCCFL 26.
Alexiadou, A & E. Anagnostopoulou. 2011. Decomposing the Greek verb. The view from the clear-
alternation. Paper presented at Approaches to the Lexicon (Roots II), The Hebrew University of
Jerusalem, June 13-16.
Alexiadou, A & E. Anagnostopoulou. To appear. Manner vs. Result Complementarity in Verbal
Alternations: A View from the Clear-Alternation. To appear in the Proceedings of NELS 42,
University of Toronto.
Alexiadou, A., E. Anagnostopoulou & F. Schäfer. 2006. The Properties of Anticausatives Cross-
linguistically. In M. Frascarelli (ed.), Phases of Interpretation. Mouton de Gruyter, 187-212.
Anagnostopoulou, E. 2003. Participles and Voice. In A. Alexiadou, M. Rathert & A. von Stechow
(eds.) Perfect Explorations. Mouton de Gruyter.
Αναστασιάδη-Συμεωνίδη, Α. 1994. Το τεμάχιο –τος στα ρηματικά επίθετα της Νεοελληνικής. Studies
in Greek linguistics 15: 473- 484.
Anderson, S. 1982. Where's Morphology? Linguistic Inquiry 13: 571-612.
Anderson, S. 1992. A-Morphous Morphology. Cambridge: Cambridge University Press.
Arad, M. 2003. Locality constraints on the interpretation of roots: The case of Hebrew denominal
verbs. Natural Language & Linguistic Theory 21: 737–778.
Arad, M. 2005. Roots and patterns: Hebrew morpho-syntax. Dordrecht: Springer.
Aronoff, M. 1976. Word formation in generative grammar. Linguistic Inquiry Monograph 1.
Cambridge, MA: MIT Press.
Baker, M. 1985. The Mirror Principle and Morphosyntactic Explanation. Linguistic Inquiry 36: 373-
416.
Baker, M. 1988. Incorporation. A Theory of Grammatical Function Changing. Chicago, Illinois:
University of Chicago Press.
Borer, H. 2003. Exo-skeletal vs. endo-skeletal explanations: syntactic projections and the lexicon. M.
Polinsky & J. Moore (eds.) Explanation in Linguistic Theory. Stanford: CSLI.
Borer, H. 2005. Structuring Sense. Oxford: OUP.
Chomsky, N. 1970. Remarks on Nominalization. Studies on Semantics in Generative Grammar, 11-
61. The Hague: Mouton.
Chomsky, N. 2000. Minimalist inquires: The framework. In: Martin, R./Michaels, D./uriagereka, J.
(ed.) Step by step. Essays on Minimalist Syntax in Honor of Howard Lasnik, 89-151. Cambridge
Mass: MIT Press.
Chomsky, N. 2001. Derivation by phase. Ken Hale: a life in language. Cambridge, MA: MIT Press.
Dubinsky, S. & Simango, R. 1996. Passive and stative in Chichewa: Evidence for Modular
Distinctions in Grammar. Language 72: 749-81.
Embick, D . 2004. On the structure of resultative participles in English, Linguistic Inquiry 35: 355-
392.
Embick, D. 2010. Localism vs. Globalism in Morphology and Phonology. Cambridge MA: MIT Press.
27
Georgala, E. 2001. The translational correspondence between the Modern Greek formations ending in –tos
and –menos and their equivalent forms in German. Ms. Institute for Natural Language Processing of the
University of Stuttgart.
Giannakidou, A. & J. Merchant. 1999. Why Giannis can’t scrub his plate clean: On the absence of
resultative secondary predication in Greek. In A. Mozer (ed.) Greek Linguistics’ 97: Proceedings of
the 3rd
International Conference on Greek Linguistics, 93-103. Athens: Ellinika Grammata.
Hale, K, & Keyser, J. 1993. On argument structure and the lexical expression of syntactic relations.
K. Hale & S. J. Keyser. (eds.) The View from Building 20: 53-109. Essays in Honor of Sylvain
Bromberger. Cambridge, MA: The MIT Press.
Hale, K. & Keyser, J. 1998. The basic elements of argument structure. In. H. Harley (ed.) MIT
Working Papers in Linguistics 32. Cambridge, MA: The MIT Press.
Halle, M. & Marantz, A. 1993. Distributed Morphology and the Pieces of Inflection. K. Hale & S. J.
Keyser (eds.) The View From Building 20, 111-176. Cambridge, Mass.: MIT Press.
Harley, H. 2005. How do verbs take their names? Denominal verbs, manner incorporation and the
ontology of roots in English. In N. Erteschik-Shir & T. Rapoport (eds). The syntax of Aspect.
Oxford University Press.
Harley, H. 2012. Lexical decomposition in modern generative grammar. In Handbook of
Compositionality, edited by Wolfram Hinzen, Markus Werning and Edouard Machery. Oxford:
OUP.
Horvath, J. & T. Siloni. 2008. Active Lexicon: Adjectival and verbal passives. 2008. In Sharon Armon-
Lotem, Gabi Danon and Susan Rothstein (eds.) Current Issues in Hebrew Linguistics. 105-134.
John Benjamins. Kiparsky, P. 1982. Lexical Phonology and Morphology. Seoul: Linguistics in the Morning Calm,
Hanshin.
Kordoni, V. 2002. Participle-adjective formation in Modern Greek. LFG Meeting, July 2002. Athens,
Greece.
Kratzer, A. 1994. The Event Argument and the Semantics of Voice. Ms., University of Massachusetts
at Amherst.
Kratzer, A. 2001. Building statives. Berkeley Linguistic Society, 26.
Lascaratou, Ch. 1991. How "adjectival" are adjectival passive participles in Modern Greek and
English? Glossologia 7-8: 87-97.
Lascaratou, Ch. & Philippaki-Warburton I. 1984. Lexical versus transformational passives in Modern
Greek. Glossologia 2-3: 99-109.
Lekakou, M. 2005. In the middle somewhat elevated. The semantics of middles and its crosslinguistic
realization. Ph.D dissertation, UCL.
Levin, B. & Rappaport, M. 1986. The formation of adjectival passives. Linguistic Inquiry 17: 623-661.
Marantz, A. 1996. “Cat” as a phrasal idiom. Consequences of late insertion in Distributed
Morphology. Ms. MIT.
Marantz, A. 1997. No escape from Syntax: Dont' try morphological analysis in the privacy of your
own lexicon. In: Dimitriadis, A, Siegel, L (eds) University of Pennsylvania Working Papers in
Linguistics, Philadelphia:University of Pennsylvania, pp 201-225.
Marantz, A. 2001. Words and Things. Ms., MIT & NYU.
Marantz, A. 2007. Phases and words. S. H. Choe et al, (eds.) Phases in the theory of grammar. Seoul:
Dong In Publisher.
Marantz, A. 2011. Locality Domains for Contextual Allosemy in Words. Handout. NYU.
Markantonatou, S., Kaliakostas, A., Bouboureka, V., Kordoni, V. & Stavrakaki, V. 1996. Μία
(λεξική) σημασιολογική περιγραφή των ρηματικών επιθέτων σε –τός. Studies in Greek Linguistics
17: 187-201.
Mohanan, K. 1986. The Theory of Lexical Phonology. Dordrecht: Reidel.
Νάκας, Θ. 1978. Για τη μετοχή και το ρηματικό επίθετο σε –τος όπως εμφανίζονται στην κοινή Νέα
Ελληνική και στις διαλέκτους. Β’ Συμπόσιο Γλωσσολογίας του Βορειοελλαδικού Χώρου (Ήπειρος-
Μακεδονία-Θράκη), 241-262.
Parsons, T. 1990. Events in the Semantics of English. A Study in Subatomic Semantics. Cambridge Mass:
MIT Press.
28
Perlmutter, D. 1988. The split Morphology Hypothesis: Evidence from Yiddish. Theoretical
Morphology, 79-101.
Pesetsky, D. 1995. Zero Syntax: Experiencers and Cascades. Cambridge Mass.: MIT Press.
Ράλλη, Α. 2001. Μορφολογία. Athens: Patakis.
Ράλλη, Α. 2006. Η Σύνθεση Λέξεων. Athens: Patakis.
Σετάτος, Μ. 1984. Παρατηρήσεις στα ρηματικά επίθετα σε –μένος και –τος της κοινής νεοελληνικής.
Studies in Greek Linguistics. Proceedings of the 5th Annual Meeting of the Department of
Linguistics of the Aristotle University of Thessaloniki, Thessaloniki.
Samioti, Y. 2008. The structure of –tos participles expressing ability/possibility. Studies in Greek
Linguistics 29.
Samioti, Y. In progress. Issues in the Lexicon – Syntax Interface and applications in Modern Greek as
a 2nd
Language. The case of Participles. Ph.D. dissertation University of Crete.
Siegel, D.1979. Topics in English Morphology. New York: Garland.
Spencer, A. 1991. Morphological Theory. Oxford: Blackwell.
Stroik, T. 1992. Middles and movement. Linguistic Inquiry 23: 127-137.
Stroik, T. 1995 On middle formation: A reply to Zribi-Hertz. Linguistic Inquiry 26: 165-171.
Stroik, T. 1999. Middles and reflexivity. Linguistic Inquiry 30: 119-131.
Svenonius, P. 2005. Extending the Extension Condition to Discontinuous Idioms. Linguistic Variation
Yearbook 5: 227-263.
Travis, L. 2005. Agents and Causes in Malagasy and Tagalog. In The Syntax of Aspect: Deriving
Thematic and Aspectual Interpretation, eds. Nomi Erteschik-Shir and Tova Rapoport, 174-189.
New York: Oxford University Press.
Wasow, T. 1977. Transformations and the Lexicon. P. W. Culicover, T.Wasow & J. Bresnan (eds.)
Formal Syntax, 327-360. New York: Academic Press.