The University of Queensland, Brisbane
Miriam Butt and Tracy Holloway King (Editors)
1998
CSLI Publications
http://www-csli.stanford.edu/publications/
The notion 'predicate' is important in any theory of syntax, and in the prototypical case a predicate is a single word. However, it has been clear for some time that there are cases where two words are syntactically distinct from each other, yet show the properties of a single predicate. Such cases have recently become known as complex predication (Alsina, Bresnan, and Sells 1997).
Causatives and permissives are perhaps the best-known cases of complex predication. As many authors have noticed, there are important similarities between syntactic causatives of the sort seen in Romance languages and the morphological causatives seen in languages like Japanese, Turkish, and Chichea.
In this paper, I will suggest that languages also vary between syntactic
and morphological expressions of directionality. In many languages, directionality
is indicated morphologically through affixation. Consider the following
forms from the Iroquoian language Oneida (Abbot 1981):
1) T-a-ha-hkwé:nvht-e?
cis-fac-3smA-descend-pft
'He came down.'
2) Y-a-ha-hkwé:nvht-e?
trans-fac-3smA-descend-pft
'He went down.'
In other languages, like Choctaw, a Muskogean language spoken in Mississippi
and Oklahoma, directionals are a small class of preverbal particles that
nevertheless show signs of functioning together with a following verb to
form a single complex predicate.
1.1 Preliminary data
Consider the directionals seen in the following Choctaw sentences:
3) Chokka' ila-h pit kanalli-tok.
house other-tns away move-pt
'They moved to a different house.'
4) Boswell bilika' yakni' habiina-t chokka' o-talaali-t
áyyaasha-tok-oo-sh,
Boswell near land receive-ss house on-lay-ss be:located-pt-part-ss
at pihlichi-t i-chokka' isht iya-ttook.
come& lead-ss III-house instr go-dpast
'... they had been alotted land near Boswell, where they had built a
house, so they came and led them to the house.'
5) Oklah Amazing Grace ot
taloow-aachi-h.
plur Amazing Grace go& sing-irr-tns
'They're gonna go sing Amazing Grace.' 11:10
In this paper I will look at the syntax and semantics of these particles
and explore how the two are to be connected to each other.
1.2 The meaning of the directionals, first attempt
The Choctaw directionals fall into two groups, which I will call single-event and dual-event directionals. Single event-directionals are used with verbs of motion to tell us about the orientation of that motion.
6) Single-event directionals:
pit 'motion away from (a reference point)'
iit 'motion towards (a reference point)'
[awiit 'motion towards (a reference point)'](1)
There are interesting questions about how this reference point is established, but for current purposes, we can say that the point of view is generally that of either the speaker or the subject of the sentence.(2)
Dual-event directionals tell us about the direction of movement prior to the start of another verb. They are quite close to English 'come and' and 'go and'.
7) Dual-event directionals
ot 'motion away from (a reference point)'
at 'motion towards (a reference point)'
These directional particles are never used alone, but always before some other verb.
Diachronically the directional come from reduced verbs. The final /t/ found in all of them is the same-subject switch-reference marker /-t/. The most likely origins are the following:
8) onah 'to arrive (there)' + /-t/ 'same-subject' > ot
alah 'to arrive (here)' + /-t/ 'same-subject' > at
pilah 'to throw/send' + /-t/ 'same-subject' > pit
?? + /-t/ 'same-subject' > iit
However, the particles are probably synchronically monomorphemic for
modern speakers.
2. The syntax of the directionals
2.1 Directionals are syntactically independent words
Choctaw is an SOV language. A directional particle always follows any overt object and precedes the verb:
9) Hattak-at tachi' at apa-tok.
man-nm corn come& eat-pt
'The men came and ate the corn.'
*Hattak-at at tachi' apa-tok.
Only one directional may used in a clause, even when the combination of a single-event and dual-event directionals might seem coherent, as we can see from the following example:
10) Chokka' ila-h (*ot) pit kanalli-tok.
house other-tns (go&) away move-pt
'They (*went and) moved to a different house.'
There is some disagreement about whether the directionals are actually separate words or whether they are prefixes.(3) Ulrich (1986) shows that the directionals share phonological properties with other morphemes that he labels clitics.
Despite their phonological relationship with the following verb, they
should be treated as having the syntactic status of independent words.
We can see this through their interaction with the word oklah.
2.2 The position of oklah
Oklah is both a noun meaning 'people'and an element that indicates the plurality of a animate subject. In texts, the most natural position for oklah is after a direct object and immediately before the verb.
11) Hitokoosh chokfi' oklah falaama-tok.
and:then rabbit plur meet-pt
'And then they met a rabbit.' (T3:3)
Oklah may also appear before the direct object:
12) Oklah Amazing Grace ot taloow-aachi-h.
plur Amazing Grace go& sing-irr-tns
'They're gonna go sing Amazing Grace.' 11:10
13) Hattak-at oklah tachi' apa-tok.
man-nm plur corn eat-pt
'The men ate all the corn.'
However, it may not appear before the subject:
14) *Oklah hattak-at tachi' apa-tok.
plur man-nm corn eat-pt
(The men ate the corn.)
Oklah may also appear between the directional particle and the verb. All three of the following sentences are acceptable.
15) Hattak-at tachi' at oklah apa-tok.
man-nm corn come& plur eat-pt
'The men came and ate corn.' 10:202
OK Hattak-at oklah tachi' at apa-tok.
OK Hattak-at tachi' oklah at apa-tok.
Given the mobility of oklah, we should
treat it as a syntactically separable word. But since oklah may
follow the directional, the directional must also be a separate word.
2.3 A proposal
We can account for this distribution if we assume the following syntactic structure:
and the following distributional statement for oklah:
16) Distribution of oklah
Oklah must be left-adjoined to some projection
of the verb.
The single-event directionals pit, iit,
and awiit are used when the following verb includes a motion component.
But what, exactly, is a motion component?
Speaking more formally, the single-event directionals
are appropriate with verbs that contain the predicate GO in their lexical
decomposition. These fall into several semantic classes.
a.) verbs of simple physical motion
b.) verbs of transfer
c.) verbs of perception
d.) verbs of "directed emotion"
e.) verbs of speech and thought
f.) verbs of comparison
The verbs thus include cases where there is real
motion in the world and cases where the movement is abstract or metaphorical.
For reasons of space, this paper will only discuss the use of directionals
with a few of these classes. See Broadwell (1996) for a fuller discussion
of directionals and abstract motion.
3.2 Verbs of simple physical motion
The following verbs describe physical motion in
the real world, and all are appropriate with the single-event directionals.
kachih 'to send; to sell'
atohnoh 'to send, order'
tilhiilih 'to send (pl. obj.)'
oyyah 'to go up, climb'
tanablih 'to cross over'
ilhkolih 'to go (pl.)'
kanallih 'to move'
iyyakayyah 'to follow'
pilah 'to throw, send'
itokaahah 'to throw in the fire'
kochchah 'to go out'
ashaachih 'to gather'
okaachih 'to throw in the water'
lhayah 'to throw away'
abachakaalih 'to lift the head'
17) Chokka' ila-h pit kanalli-tok.
house other-tns away move-pt
'They moved to a different house.'
18) Iit pila-h.
toward throw/send-tns
'He threw it (toward me).'
19) Pit pila-h.
away throw/send-tns
'He threw it (away from himself/me).'
I assume the framework and notational conventions
of Conceptual Semantics (Jackendoff 1983, 1990). I believe this is most
compatible with a view of syntax like that of Lexical-Functional Grammar
(Bresnan et al. 1982), but it is probably possible to implement the semantic
suggestions here in other syntactic frameworks as well.
Let us assume the following sorts of lexical entries
for the items under discussion:



3.3.2 Unification and single-event directionals
ET shows a 'transparent event', using
the terminology of Butt, Isoda, and Sells (1990), Butt (1993, 1997), and
Alsina (1993). The idea is that directionals like pit and iit
are light verbs which are unable to denote events on their own, but combine
their lexical information with that of a following verb to form a complex
predicate. The unification of the lexical entries for pit and pilah
will yield the following result:(4)
We can state the rule for the unification of
the single-event directionals as follows:
20) Event Fusion--Directionals
Unify the lexical conceptual structure of a single-event
directional with the
[EVENT GO..] substructure of another
event.
After unification, THERE functions as a selectional
restriction on the goal of the complex predicate. Consider the following
example:
21) John-at aaípa'-ma towa'
pit pila-tok.
John-nm table-d:ac ball away throw-pt
'John threw the ball under the table.'
The presence of pit in this sentence requires
that the table be located far from John. The unified LCS for pit pilah
shows this by indicating that whatever the goal of the sentence is, it
is THERE.
3.3.3 Dual-event directionals
I'll assume the following sort of lexical entries
for ot and at:
Unpacking the formalism, these lexical entries
say that:
a.) the event described (Ex) is incomplete
and must occur with a second event (Ey), and
b.) the time of Ex precedes that of
Ey, and
c.) the Theme of Ex is the Actor of
Ey.
The LCS for this entry is combined with that
of another verb via a different rule, Argument Fusion (Jackendoff 1990),
which inserts the LCS for the second event into the available argument
slot. For a sentence like (22), Argument Fusion yields a LCS like the following:
22) John-at towa' ot pila-h.
John-nm ball go& throw-tns
'John went (there) and threw the ball.'

4. Other verb classes
4.1 Verbs of transfer
The following verbs of transfer also appear with
single-event directionals, and thus must contain a motion component.
imah 'to give'
ipiitah 'to give (to several), to distribute'
Consider the following examples.
23) Iit am-a-h!
toward 1sIII-give-tns
'Give it to me!'
24) John-at Mary pit im-a-tok.
John-nm Mary away III-give-pt
'John gave it to Mary.'
I assume that the verb 'give' contains a motion
component in its lexical semantics, along the following lines.
Given this representation, the previously formulated
rules give us the right semantics.
4.2 Perception
All verbs of perception in Choctaw are compatible
with the single-event directionals.(5) From
the perspective of English, it is surprising that hearing and smelling
are using with the 'away' directional pit. This appears to reflect
a rather different folk theory of perception than that shown in English
sentences like The sound/odor came to me from across the room.
Consider the following examples:
25) Leslie-at Sandy pit pisa-tok.
Leslie-nm Sandy away see-pt
'Leslie saw Sandy.'
26) Mary-at ofi' pit haklo-tok.
Mary-nm dog away hear-pt
'Mary heard the dog.'
27) Pit chi-ashshowa-l-aana-h.
away 2sII-smell-1sI-pot-tns
'I can smell you.'
Let us assume that these have lexical entries
like the following:
The moving objects for hakloh and
ashshowah seem to be auditory and olfactory equivalents of the gaze.
See Broadwell (1996) for some discussion of the linguistic encoding of
perception and Whorfian considerations raised by these lexical entries.
4.3 Emotions
Emotions fall into two classes that we can call
"directed" and "non-directed" emotions. Directed emotions seem to be based
on a cultural metaphor that envisions certain kinds of thoughts as travelling
through abstract space to reach their objects.
Some directed emotions
inoktalhah 'to be jealous'
iholloh 'to love'
inokhakloh 'to be sad about, to
grieve over'
ayokpachih 'to like'
Some non-directed emotions
noklhakachah 'to be startled'
inokshoopah 'to be afraid of'
inokoowah 'to be angry at'
I will not attempt to formalize the lexical entries
for the directed emotions, but I assume that [EVENT GO ...]
constitutes some subportion of them, and that [EVENT GO ...]
is lacking in the non-directed emotions.
Many other classes of verbs are incompatible with
directionals, including all statives and many activities:
28) *Ofi-yat pit homma-h.
dog-nm away red-tns
(The dog is red.)
29) *?John-at pit taloowa-h.
John-nm away sing-tns
(John sang.)
These verbs presumably have no [EVENT
GO ...] constituent in their lexical decomposition.
They will fail to unify with the single-event
directionals. However, so long as a verb has an Actor, it is compatible
with a dual-event directional.
5. Putting it all together
Recent developments in LFG (Butt 1993, Alsina
1993) allow clauses with discontinuous heads. This approach has been successfully
applied to serial verbs, Romance causatives, and Urdu light verb constructions.
(This is approximately the class of verbs that have been treated as involving
abstract incorporation or "restructuring" in Principles and Parameters
approaches.)
The c(onstituent)-structure for a sentence with
a directional will be as follows, this time with the relevant functional
annotations:
In this
c-structure, both the verb and the particle serve as co-heads of the clause.
This c-structure is related to a f(unctional)-structure
like the following:
This has the effect of making the directionals
part of the sentence predicate along with the verb.
Finally, both these syntactic representations
are in correspondence with the conceptual structure of the sentence. Following
Butt (1993), argument structure can be viewed as a subpart of conceptual
structure.
6. Conclusion
The approach to directionals suggested here puts
most of the work of accounting for their distribution in the lexical entries
of the items involved and in two independently needed rules of semantic
composition - Argument Fusion and Event Fusion. The constituent structures
posited are relatively simple. This is a result of shifting the explanatory
burden from syntax proper to its connection to semantics, and the result
is a less syntacto-centric approach to explanation.
7. Bibliography
Abbot, Clifford. 1981. Here and
there in Oneida. International Journal of American Linguistics.
47:50-58.
Alsina, Alex. 1993. Predicate
composition: A theory of syntactic function alternations. Ph.D. thesis,
Stanford University.
Alsina, Alex; Joan Bresnan; and
Peter Sells, eds.1997. Complex Predicates. CSLI Lecture notes 64.
Stanford, CA.
Bresnan, Joan et a. The mental
representation of grammatical relations. Cambridge, MA: MIT Press.
Broadwell, George Aaron. 1990.
Speaker and SELF in Choctaw. International Journal of American Linguistics
57:411-425.
Broadwell, George A. 1996. Directional
particles and abstract motion in Choctaw. Proceedings of the 1996 Mid-America
Linguistics Conference. pp. 53-66. (Available at http://www.albany.edu/anthro/fac/broad.html)
Butt, Miriam. 1993. The structure
of complex predicates in Urdu. Ph.D. thesis, Stanford University.
Butt, Miriam. 1997. Complex predicates
in Urdu, in Alsina, Bresnan, and Sells.
Butt, Miriam; Michio Isoda; and
Peter Sells. 1990. Complex predicates in LFG. ms. Stanford University.
Byington, Cyrus. 1915. A dictionary
of the Choctaw language. BAEB 46.
Jackendoff, Ray. 1983. Semantics
and cognition. Cambridge, MA: MIT Press.
Jackendoff, Ray. 1990. Semantic
structures. Cambridge, MA: MIT Press.
Munro, Pamela and Catherine Willmond.
1994. Chickasaw: An analytical dictionary. Norman, OK: University
of Oklahoma
Nicklas, T. Dale. 1974. The
elements of Choctaw. Ph.D. thesis, University of Michigan.
Talmy, Leonard.1983. How language
structures space, In H.Pick and L. Acredolo, eds. Spatial Orientation:
Theory, research, and application. New York: Plenum
Ulrich, Charles. 1986. Choctaw
morphophonology. Ph.D. thesis, UCLA.
Abbreviations and orthography
Thanks to Miriam Butt, Wallace
Chafe, Donna Gerdts, Jack Martin, and Pamela Munro for helpful comments
on this analysis. Special thanks to Edith Gem and Henry Willis who provided
all the Choctaw examples not otherwise attributed. Symbols in the orthography
have their usual phonetic values, with the following exceptions: <sh>
and <ch> are approximately as in English, <lh> = a voiceless lateral fricative, and underlining represents nasalization.
The following abbreviations are
used: 3msA=third person singular masculine agent, ac=accusative, cis=cislocative,
comp=complementizer, con=constrastive, dpast=distant past, ds=different
subject, fac=factual mode, foc=focus, hn=hn-grade (iterative aspect), irr=irrealis,
l=l-grade (a stem form that appears before some suffixes), loc=locative,
n=n-grade (durative aspect), m=nominative, part=participle, pft=perfective,
pl=plural, prev=previous mention, pt=past, super=superessive, ss=same subject,
tns=tense, trans=translocative. There are three sets of Choctaw person-number
agreement markers, labelled I (approximately 'Actor'), II (approximately
'Patient'), and III (approximately 'Goal/source'). Person markers are glossed
as follows 1sI = 1st person singular, I agreement class; 2pI
= 2nd person plural, II agreement class, etc.
Notes
1. Awiit is a somewhat archaic variant of
iit, and modern speakers consider it essentially synonymous. It
occurs extensively in the Choctaw translation of the Bible.
2. See Broadwell (1990) for discussion of an analogous
problem of point of view in the interpretation of evidential particles.
3. In Chickasaw, for example, the comparable particles
are written as part of the same word as the following verb by Munro and
Willmond (1994)
4. Actually, in order to perform unification in
the technical sense, we need to recast these lexical-conceptual representations
as attribute-value matrices of the sort used in HPSG and LFG. This is a
relatively trivial point of notation.
5. It is, of course, perfectly possible to use these
verbs without a preceding directional. For many speakers, a directional
with a verb of perception emphasizes the distance of the perceived object.
However, other speakers report no semantic differences of this sort.