The syntax of Tzotzil auxiliaries and directionals:

the grammaticalization of "motion"

John B. Haviland
Reed College

The problem to be considered arises from the (Zinacantec) Tzotzil

encoding of simple motion, which leads to a syntactic puzzle. The syntactic
puzzle in turn leads back to the linguistic expression of motion and space, now
understood as results of a process of grammaticalization.
Consider first a group of English words like up, above, ascend, climb, lift,
and raise. There is clearly no etymological connection between them, although a
semantic analysis might well try to connect their meanings in some systematic
way. Perhaps, for example, the words would figure in a definitional circle: up
captures some notion of vertical trajectory; ascend means "go up"; raise means
"cause to ascend"; and so on. But there would be no MORPHOLOGICAL or
GRAMMATICAL reasons to make these interconnections, and each word would
deserve its own separate lexical entry.
Now consider, by contrast, the Tzotzil root muy. It gives rise to a number
of lexical forms, of which those of interest are exemplified in (1-6). i
(1)-(3) illustrate the root as a normal intransitive verb stem. Like all main
verbs in Tzotzil, it must be inflected for both aspect (ch- marking incompletive, l-
completive) and person. (i- is a 1st person absolutive prefix, and -otik is a 1st
person inclusive plural absolutive suffix.)
(1) CK:127
ch- i- muy ta ba k'atal toj
ICP-1A-ascend PREP top sideways pine
I would climb up the leaning pines.
(2) CK:127
ch- i- muy j- tuch' i tajchuche
ICP-1A-ascend 1E-cut ART mushroom_sp.
I climbed up to pick the lentinus mushrooms.
(3) SSS:91
l- i- muy- otik te yo`- e,
CP-1A-ascend-1PLINCL there where-CL
We went up there [on an observation tower].
In (2), the fully inflected intransitive verb chimuy 'I
ascend' has what appears to be a purposive adjunct, j-
tuch' i tajchuche '(in order that) I cut mushrooms' where
the transitive verb j-tuch' bears NO aspectual inflection
but only person marking (here a 1st person ergative
prefix with a presumed zero 3rd-person absolutive).
The absence of aspectual inflection is characteristic
of various purposive constructions in Tzotzil, and is
normally labelled "subjunctive." ii
(4) SSS:91
muy j- k'el-tik k'u s- muk'ul li mejiko-e
ascend(AUX) 1E-look-1PLINCL what 3E-size ART Mexico-CL
We went up to see how big Mexico City was.
(5) {constructed}
ch- muy ve`-ik- on ta jol na
 ICP-ascend(AUX) eat-SUBJ-1A PREP head house
I will go up to eat in the attic.
(2) contrasts rather sharply with (4) and (5), which
illustrate the use of muy as an AUXILIARY verb. Here
the auxiliary itself is NOT inflected for person,
though it bears aspectual affixes. (The bare form
illustrated in 4 is understood as completive; 5 shows
an incompletive prefix.) Following the auxiliary is
again a "subjunctive" verb, marked for person but not
for aspect. (5) shows that when this second verb
(which following Aissen 1993 I will call V2) is
intransitive, it bears an explicit subjunctive suffix -
ik. The auxiliary-V2 construction thus distributes the
morphology of the simple Tzotzil verb over two
different elements which are tightly bound together,
being separable only by a small set of clitics. iii The
auxiliary bears aspect, and V2 inflects for person.
The combination auxiliary-V2 has been called a "Motion-
cum-purpose" construction (Aissen 1984, 1987:16) to
reflect the characteristic gloss.
Finally, (6) illustrates a further DIRECTIONAL form derived from the same
root muy by suffixing -el.
(6) CK:398
ta la s-jipulan muyel yi` ta vinajel
ICP CL 3E-throw ascend(DIR) sand PREP heaven
He would keep tossing the sand up into the sky.
The directional closely follows a main verb as a kind of locative adverb. Here
neither the verb jip 'throw' nor the all-purpose preposition ta 'at, on, to, from, etc.'
conveys any information about the path involved. It is instead the directional
muyel that expresses the upward trajectory.
By contrast with the English words mentioned at the outset which were
SEMANTICALLY relatABLE but UNrelated at the level of morphology and syntax, it
is clearly desirable to connect these different uses of the Tzotzil root muy. How
must a lexical entry for this Tzotzil root be constructed so as to reflect the various
syntactic guises it can assume and also their underlying commonality of meaning?
Muy 'ascend' is one of a small set of Tzotzil roots which display these
different forms. All are ordinary intransitive verb roots, mostly verbs of
"motion." The entire set of roots which produce AUXILIARY verbs is shown in (7),
arranged into semantic classes that include deictically anchored motion, point-
oriented motion, enclosure- or region-oriented motion, and motion on a vertical
axis. iv There are also two roots in the set with an aspectual meaning. v
DIRECTIONALS are derived by suffixing -el to virtually the same verbal
roots that yield auxiliaries. (In [7] those roots which produce only auxiliaries are
shown in square brackets, and those which produce only directionals are shown in
parentheses.)
(7) Tzotzil auxiliary and directional roots
Deictically anchored motion
ba(t) "go" (as directional "from time to time") k'ot "arrive there"
 tal "come" yul "arrive here"
Point-oriented motion
`ech' "pass by" (as directional, "away") (jelav "pass by")
sut "return" [`a(y) "go and return"]
kom "stay"
Enclosure or region oriented motion
`och "enter" lok' "exit"
Vertical axis motion
muy "ascend" yal "descend"
Aspectuals
[laj "finish"] lik "arise, start"
(vay "sleep")
All these roots directly produce intransitive verb stems, which in their
ordinary incarnations require a syntactic subject, cross-indexed by an absolutive
person affix. When the verb expresses motion, the subject in such cases (as in 1-3
above) is the entity in motion. However, in sentences like (4)-(6) there are no
person markers attached to the forms of the motion-verb root. The syntactic
puzzle posed by the use of these roots as auxiliaries and directionals is this: if the
underlying roots are verbs of motion, who or what is doing the moving? How do
we identify what Aissen (1993) calls the "Mover"?
Auxiliaries
On first examination the motion associated with an auxiliary verb patterns
as if the logical subject of the auxiliary motion verb were coreferential with the S
(intransitive subject) or A (transitive subject) argument of V2. Such a pattern is
parallel to an example like (2) where the motion verb--not an auxiliary--IS
inflected for person.
When V2 is intransitive, it has only a subject argument which seems
naturally to be associated with the auxiliary's motion (or lack of motion in the
case of kom 'stay'). Thus, in (5), the "I" who will eat in the attic (shown by a 1st
person ABSOLUTIVE suffix -on) is the "I" who will go up to the attic. (8) and (9)
are further examples with auxiliaries "stay" and "arrive (there)."
(8) Chichspn:78:
kom to `abtej-uk.
stay(AUX) still work-SUBJ(3A)
He stayed on to work.
(9) T130:329
naka xa la ch-k'ot cham-uk z-na.
just already QUOT ICP-arrive(AUX) die-SUBJ(3A) PREP+3E-house
She just reached home and died. (Laughlin 1977:318)
In these cases there seems to be both a causal and a temporal link between the
motion encoded in the auxiliary and the action of the main verb: a protagonist's
action follows (or is contemporary with) a preparatory (if not literally purposive)
motion by the same protagonist.
When V2 is transitive, again its subject (or agentive argument) appears to
be associated with the motion of the auxiliary. In (4) above, "we"--the people
cross-indexed by the 1st person plural inclusive ERGATIVE affixes on the verb
'look'--are the ones who ascend.
 Auxiliary constructions can also appear in the imperative, as in (10). The
auxiliary element bears a zero aspectual affix, and V2 shows normal morphology
for a transitive imperative.
(10) LOL6:120
`ech' `ak'-on ta sintalapa
pass(AUX) give-1A PREP Cintalapa
Pass by Cintalapa and drop me off (there).
The logical subject or A argument of V2 (in this case an unstated 2nd person) is
also understood to be the Mover.
Such facts seem to provide evidence for a
Nominative/Accusative style Subject relation in
Tzotzil, despite the ergative/absolutive agreement
morphology. This result would coincide with an
intuitive view of the purposive relation between the
motion encoded by an auxiliary, and a "goal" or
"purpose" described in V2. Here some willing agent
moves (and the motion is encoded in an auxiliary) in
order to bring about some intentional result (captured
by V2). Corresponding claims have been advanced about
some sister Mayan languages and their ancestors. vi
However, a more careful look at the behavior of
auxiliaries suggests that interpreting the motion
denoted by the auxiliary cannot be strictly a syntactic
matter. One sort of evidence is provided by examples
in which V2 is passive. The details of such
constructions are too complex to describe here, vii but
(11) will illustrate the problem. One passive form
that appears with auxiliaries has the suffix -el and
bears formally ergative prefixes that cross-index the
logical PATIENT.
(11) T84.TXT:226:
tal- em k- ik' -el z- na `a li rey
come(AUX)-STAT 1E-take-PASS PREP+3E-house PT ART king
They've come to take us to the King's house. (CK:208).
An example like (11) is difficult to gloss; it means something along the lines of
"there is coming for us to be taken to the King's house." Laughlin's translation
indicates that the Mover--the entity that "comes"--is understood as an indefinite
"they," which appears neither in the sentence nor, indeed, in the surrounding
discourse. The "we" who are to be taken have not "come" anywhere, since we
were already there. (The root tal denotes motion towards "here," the speaker's
deictic center.)
A simpler non-passive case appears in (12):
(12) T109:116
ba nox ech'-uk ak'ubal yo`-e
go(AUX) only pass-SUBJ(3A) night there-CL.
Just go spend the night there. (CK:82)
Here the auxiliary bears no overt aspect marker, and the overall construction must
be interpreted as imperative. The preceding context makes clear that the
ADDRESSEE is meant to be the one to go [to a graveyard]. The V2 involves an
idiomatic expression ech' ak'ubal 'lit., the night passes.' A person who, in
English, would "spend a night" can appear in such an idiom only as an oblique
argument introduced by the relational noun -u`un.
(13) {constructed}
vokol i`-0- `ech' k- u`un ak'ubal
difficult CP-3A-pass 1E-ABIL night
I had a hard night. (Lit., I managed with difficulty for the night to pass.)
There appears to be no argument of V2 in (12) which can provide a logical
subject to the auxiliary ba 'go.'
A similar problem arises with example (14).
(14) MONOL:3:
ch-ba lok'-uk `akta noxtok
ICP-go(AUX) exit-SUBJ(3A) document too
(We) will go to have a document issued also.
The free translation here obscures the problem, for the original Tzotzil sentence
has no "we." The sentence
(15) {constructed}
ch-0-lok' akta
ICP-3A-exit document
A document will come out (i.e., be issued).
shows an expression used to describe the issuing of a signed, written declaration
which often serves to settle civil grievances in Zinacantán. The sentence is
intransitive, akta being the grammatical subject of the intransitive verb lok' 'exit.'
However, it is clearly not possible for such a document first to go (to the
townhall, in this case), in order then to be created and issued. It is the
complainants who do the going, with the issuing of the akta the proposed result.
This is made clear by the wider conversational context in (16): a man describes
his continuing disputes with a relative, and how they ultimately went to the
magistrate for settlement.
(16) MONOL:
(b) chibatikótik ta jteklum che`e,
We'll go to Zinacantán Center.
(c) ch-ba j-k'opon-tikótik preserente
We'll go to talk to the (municipal) president.
(d) ch-ba lok'-uk `akta noxtok
(We) will go to have a document issued also.
The sense of the auxiliary paired with lok' in (16d) derives in part from the
auxiliary ba occurring in the previous sentence (16c), which has a transitive V2 (-
k'opon 'talk to'). The DISPUTANTS will go to have the akta issued. The fact
remains, however, that there is no argument available in (14) (= 16d) which can
correspond to the "logical subject" of the auxiliary verb 'go.'
(17) is a last example. As I was about to leave Chiapas, a Zinacantec
promised to repay a loan when I returned:
(17) PV0492
ja` to ch-(y)ul k- ak'- b- ot
! still ICP-arrive_here 1E-give-BEN-2A
Just then I will give it to you arriving here.
If the Mover involved in the auxiliary yul "arrive here [i.e., in this case, to the
Chiapas village]" must co-refer with the ergative argument of the V2, (17) would
have meant that this Zinacantec, not I, would be leaving the village to return
sometime in the future--the exact reverse of the actual situation. It is almost as
though the speaker is transposing himself (Haviland 1991b) into my shoes, to say
in effect: "you have only to arrive here for me to repay you." But now the
understood Mover corresponds to the object of V2.
Aissen (1992, 1993) uses such facts to motivate a syntactic account of
Tzotzil auxiliaries in which the auxiliary is the functional head of a VP, selecting
for a subjunctively marker complement clause, but in which the auxiliary has no
nominal arguments, in particular no subject. Thus in a sentence like (2) muy
functions as a normal intransitive verb with a cross-indexed subject, whereas in
(4)-(5) muy is an auxiliary which cannot have its own subject, imparting instead
only its idealized spatial trajectory to the resulting purposeful action. The
auxiliary thus represents a GRAMMATICALIZATION of the parent verbal root: it
loses its ability to have a nominal argument, and it may, as I will argue briefly at
the end of the paper, undergo both semantic generalization and formal erosion.
The Mover in an auxiliary construction is the entity that proposes, or is in
a position to bring about the result expressed in V2. The auxiliary conjures a
scene, in interaction with the wider discursive context, and motion must be
assigned to arguments (explicit or implicit) by inference rather than by syntax.
What pragmatic parameters operate to allow the combination of understood
motion, coded in an auxiliary, with a purposive result coded as V2 requires
further investigation.
Directionals
The same verb roots from which Tzotzil auxiliaries derive provide the
source for the separate class of directionals, illustrated in (6) above. Unlike the
auxiliary, which comes at the beginning of the verb complex and inflects for
aspect, directionals immediately follow the main verb and are free of inflection,
verbal or otherwise. Indeed, multiple directionals can be combined in a single
Tzotzil sentence.
(18) t9007a1:805:
`al-a-ka`-ike `ich'-ik muyel tal `un
ART-2E-horse-PL take-(IMP)PL ascend(DIR) come(DIR) CL
As for your horses, bring them up.
In (19) the combination of a main verb of motion (the transitive -otes
'insert' derived from `och 'enter') with the motion directional tal 'coming' and the
aspectual batel 'now and then' requires an idiomatic English translation that mixes
motion and deictic elements in a way strikingly different from the Tzotzil.
(19) t9007a1:791:
ta x- otes- ik onox tal batel
ICP ASP+3E-put_in-PL anyway come(DIR) go(DIR)
They bring [their horses] here to put in [the corral] from time to time.
There is an apparent layering of the different categories, with the deictic
directionals tal(el) and `ech'el following all other directionals except the
aspectuals. The ideal ordering seems to be as in (20):
(20)
ENCL. VERT. DEIC. ASP.
ochel muyel ech'el likel
lok'el yalel tal batel
viii
(jelavel) (komel) (k'otel) (vayel)
(yulel)
 If a directional particle can be construed as a reduced reflex of its parent
motion verb, we can ask, as we did with auxiliaries, how to construe its putative
logical subject? In (6) it is the sand which ascends towards heaven. (21)-(24)
show further directionals in action.
(21) t9007a1:254
ta ka` ch-0-muy `ech'el li y-ajval
PREP horse ICP-3A-go*up go(DIR) ART 3E-owner
It is on horseback that the owners will go up away from here.
(22) t9006a1:241
i-0-lok' talel
CP-3A-exit come(DIR)
It came out towards here.
(23) LOL1:807:
ta j-k'an sutel noxtok ti j-pasajel `une
ICP 1E-want return(DIR) again ART 1E-fare CL
I (would) ask for my fare-money back again.
(24) LOL6:309:
ta j-nak' komel
ICP 1E-hide stay(DIR)
I would hide it away [equipment left at a distant market].
In these cases there is an apparent association between the directional and
the Absolutive argument of the main verb: either the S(ubject) of an intransitive
main verb (the ascending or exiting entities in (21) and (22)), or the O(bject) of a
transitive main verb (the fare returned in (23) or the hidden equipment left behind
in (24)). Even with passive main verbs, the syntactic S--the logical patient--is
clearly understood to be following the trajectory of the motion encoded in the
directional.
(25) T154:198
net'- e yalel ta `olon
press-PASS(3A) descend(DIR) PREP below
[H]e was pursued down (Laughlin 1977:117).
Once again, such a pattern has been claimed to obtain
for analogous constructions in sister languages. In
classical Quiché, for example, according to Dürr 1990:4
"[i]n transitive sentences [directionals] indicate the
direction of the patient in accordance with the general
ergative pattern of the language." ix An alternative
formulation would associate the motion of the
directional with that argument of the main verb
occupying some sort of "Theme" semantic role: an entity
construed as moving or changing.
However, this putative syntactic characterization also fails in Tzotzil.
Consider first directionals with derived antipassives. From most transitive verb
stems in Tzotzil it is possible to derive an intransitive stem by suffixing -van.
Typically, if the transitive stem fits in the frame "X does Y to Z" the suffixed -van
form fits in a corresponding intransitive antipassive frame "X does Y-ing
(typically to people)." Hence, from the transitive -toj 'pay for' one derives the
intransitive tojvan- 'pay for someone (e.g., pay a salary or a brideprice).' (26b)
shows what happens when intransitive verb stems in -van co-occur with
directionals.
(26) T145:198
a. lok' to la tal ti y-ajnile,
 (CP+3A)exit CL CL come(DIR) ART 3E-wife
b. ja` xa la tojvan lok'el tal
! CL CL (CP+3A)pay+van exit(DIR) come(DIR)
c. ti y-ajnil bankilal `une
ART 3E-wife brother CL
(a.) [The other one] got a wife. (c) The wife of the older brother (b) paid [the bride price] (Laughlin 1977:
195-6). Literally (jbh): (a) [the other one's] bride came out to here. (c) The wife of the older brother (b) paid
[for the new bride's] coming out.
The story concerns a man who managed to get a bride only through the
good offices of his older brother's wife. In clause (a) we are told that he got the
bride (she "exited" "coming"); in lines (b) and (c), which form a single clause, we
learn that the sister-in-law "paid for [the new bride]," also with directionals
"exiting" and "coming." Here, the Mover who "exits" her old house and "comes"
to her new husband's compound is the underlying, but syntactically unexpressed
(or lexically incorporated), logical patient of the intransitive main verb tojvan 'pay
[brideprice] for someone.' But this bride corresponds to no argument in the
sentence.
Such examples suggest that the motion implied by directionals is also a
matter of inference and not of syntax, that directionals like auxiliaries have no
nominal arguments. The motion or trajectory is encoded as an adverbial
augmentation of the scene described in the overall clause. Such an analysis seems
especially pressing when we turn to examples in which the verb modified by
directionals contains no hint of motion in its ordinary meaning. In such cases, the
presence of directionals INJECTS motion into the action portrayed, and the
resulting inferred trajectory must be attributed to elements which find no overt
linguistic expression.
With verbs of speaking and perception, directionals can impose an
orientation, both deictic and non-deictic, on a scene, with no suggestion of a
literal moving argument.
(27) anvask4:96:
ja` taj lo`il i-y-a`i tal ta s-na
! DEM gossip CP-3E-hear come(DIR) PREP 3E-house
That's the gossip that he heard at his house.
In (27) the deictic tal 'coming' transposes the perspective adopted by the speaker
onto that of the receiver of words: the gossip came HERE to him. In (28) the
directional adds an upward trajectory to the looking.
(28) T152:62:
xi ta j-k'el muyel `une
thus ICP 1E-look ascend(DIR) CL
I was looking up (Laughlin 1977: 202).
With other sorts of main verb more inferential ingenuity is required to
construe the meaning of the directional. (29) exhibits the directional komel
'staying' whose non-motion must accrue to the object of the main verb 'burn' (here
a cornfield).
(29) T162:196
j-chik' xa komel
1E-burn CL stay(DIR)
I've burned it (Laughlin 1977: 390).
This line is spoken by a lazy husband after he has gone home after burning off
fields in preparation for planting. The directional komel implicates the fact that
the man himself is no longer in the fields (they "stayed behind," after he
departed), and that he has finished burning them. The overall effect of the
directional is thus partly aspectual.
Further varieties of motion are implied in (30)-(34), all of which have the
directional tal 'coming.'
(30) T61:234:
t'om tal volkan
(CP+3A)explode DIR volcano
The volcano erupted [and then its ashes] came.
(31) LOL1:875:
ta j-ch'ak tal j-pasajeb `une
ICP 1E-divide DIR 1E-fare CL
I will separate out (from the rest of the money) my bus fare [before bringing my money back home].
(32) BARIL:175:
l-i-laj tal `un
CP-1A-finish DIR CL
I was injured [on my way back here].
(33) t9007a1:988:
mi ja` lek ch-a-ve`-ik tal che`e?
Q ! good ICP-2A-eat-PL DIR PT
Should you then eat [before you come here]?
(34) T86:352:
mi`n a-toj tal a-ka?
Q 2E-pay DIR 2E-horse
"Did you pay for your mule [before bringing it back here]? (Laughlin 1977: 284).
One common element in all these inferred trajectories is that the action of the
main verb takes place either BEFORE or SIMULTANEOUSLY with the motion implied
in the directional. (For example, in (33) the question suggests that the addressee
should eat first and come here afterwards.) Apart from such a temporal sequence,
though, there is rather little consistent patterning. The directional depends on the
overall scene evoked by the verb and the rest of the context of situation to supply
a Mover.
Directionals can also accompany stative adjectival predicates. De León
(1991) reports the frequent use of egocentrically anchored directionals combined
with positional adjectives in an experimental interactive game in which
Zinacantecs offered verbal descriptions of photographs. (35) is an example from
de León's transcripts.
(35) tape2a1:26
cha` kot ich j-moj cholol ech'el
two NC chile one-blow lined_up(+3A) pass(DIR)
(These are) two chiles lined up [pointing away from me].
Similar examples abound in conversational and textual materials.
(36) LOL3:921
ja` tey tzakal `ech'el x-chi`uk
! there joined pass(DIR) 3E-with
It [went away] attached to him.
(37) T75:53
chukul komel ta te`el alampre
tied_up(+3A) stay(DIR) PREP post wire
[He had] left [the cow] tied to the fence post (Laughlin 1977: 228).
Directionals also accompany non-verbal expressions of location. These
may be formed with "relational nouns" (which denote parts or regions of their
grammatical possessors).
(38) t9006a1:46:
y-ak'ol tal Nachij
3E-above come(DIR) place_name
 (It is) above Nachij on this side (i.e., above coming).
More surprising, perhaps, is the use of the directionals tal and `ech'el to
modify ordinary locative phrases of the form PREPOSITION+NP.
(39) LOL2:222:
le` ta `ach'eltik `ech'el s-na-e
there PREP muddy-expanse DIR 3E-house-CL
His house is over there on the far side of the muddy place.
Here a whole scene is broken into deictic, locational, and trajectory elements by
the Tzotzil syntax. There is first a distal demonstrative le` 'there' (which also
serves as the predicate that can bear absolutive suffixes, in this case a zero third
person). It is complemented by a prepositional phrase, whose preposition ta is
semantically empty, 'at/near/in a muddy place.' The directional `ech'el in turn
suggests the image of, say, walking from the muddy place in a direction away
from "here." The same scene must be rather more statically re-packaged in the
English translation.
In both the auxiliary and directional cases, then, inference rather than
syntax assigns putative logical arguments to the underlying (or perhaps "fictive"
[Talmy 1985]) motion expressed. This suggests that the auxiliary and directional
uses of a root like muy are no longer fully verbal: the argument structure available
to muy as a motion verb has been eroded away from muy the auxiliary (although it
is still enough of a verb to carry aspect) and still more from muyel the directional
which now involves only schematic motion, direction, or orientation.
AUX + DIR
Despite differences in detail, auxiliaries and directionals are formed from
the same motion roots. How do the two slots combine?
First, we have seen that there is an association between the motion of the
auxiliary and an understood (though possibly unexpressed) purposive agent in the
overall action. Similarly, we have seen that the directional specifies a trajectory
that is construed as simultaneous with or subsequent to the action of the main
verb. The ordering of auxiliary and directional also suggests iconically the
temporal sequencing of the two different phases of motion: the motion of the
auxiliary comes first, that of the directional afterwards. Thus, apparently
redundant combinations of semantically equivalent auxiliary and directional can
be understood to have non-redundant force.
(40) LOL3:150:
ja` ch-ba s-man `ech'el le` `une
! ICP-go(AUX) 3E-buy pass(DIR) that CL
He'll go to buy that (and take it away).
In (40) ba 'go (AUX)' is paired with `ech'el 'going (DIR).' The sequence implied
is: "he goes to buy [and afterwards take away] that thing."
The combination of auxiliary and directional is thus perfectly designed to
encode a ROUND-TRIP.
(41) LOL4:143:
ch-ba j-sa` tal k-ikatz ta jobele
ICP-go(AUX) 1E-search*for come(DIR) 1E-cargo PREP San_Cristóbal
I'll go to find my goods in San Cristóbal (and bring them back here).
Moreover, since two sorts of motion or trajectory can be grammatically
incorporated into every verb phrase by using both auxiliary and directional
positions, the main verbs themselves can be lexically specific, and the overall
topology of an action can still be economically encoded. The lexicalized
character of the main verb is also, thereby, distributed over the entire scene--the
action and both its preparatory and resulting paths. Thus consider the distribution
of the action of "grabbing" in (42).
(42) LOL1:348:
pinchi `alvanil ba s-tzak-on `ochel
damned mason go(AUX) 3E-grab-1A enter(DIR)
The damned mason just went and grabbed me [and hauled me] inside.

The grammaticalization of motion

Heine (1992:40-44) proposes a Motion schema--"X moves to/from Y"--as
one starting point for a grammaticalization chain in which a verb evolves into a
tense/aspect marker (TAM). An auxiliary, on Heine's account, is simply "a
linguistic item covering some range of uses along" (1992:93) such a chain. The
grammaticalization process comprises a sequence of diagnostic linguistic shifts
that include desemanticization, decategorialization, cliticization, and erosion (pp.
68ff.). For each shift, Heine posits an "Overlap model" that involves a three step
progression from a source stage to a target stage, with an intermediate stage where
constructions are ambiguous between the source and target patterns (pp. 66ff.).
Recent work by Zavala (1992) and Craig (1992) explores the
grammaticalization of auxiliary and directional elements in a Mayan comparative
context in considerably more detail than can be accommodated here. Tzotzil
auxiliaries and directionals are clearly consistent with Heine's scheme, falling
near the start of the chain when compared with, say, Akatek (Zavala 1992) or
Sakapultek (du Bois 1992) auxiliaries or Mam post-verbal directionals (England
1976). How do we understand the synchronic grammatical categories that result
from such a putative grammaticalization process?
1. Erosion
The verbal roots that give rise to AUX and DIR elements
show incipient phonological reduction of a
characteristic kind. Thus bat 'go' frequently appears
as AUX ba, and `ay 'go and return' yields AUX `a. x
Similarly, the suffixed DIR talel 'coming' is often
shortened to tal; moreover, where Zinacantec Tzotzil has
directionals `ech'el 'passing, away' and batel 'going, from
time to time,' nearby Chenalho Tzotzil uses `el and bel. xi
2. Decategorialization and cliticization
AUX and DIR elements are formally unlike verbs despite their verbal
provenance. The results of decategorialization have been the primary focus here:
for example, the Mover is not a syntactically governed argument of AUX and
DIR elements. AUX and DIR also have restricted clitic-like positions of
occurrence around a main verb (or other predicate).
The same roots that produce AUX and DIR elements also give rise to
normal intransitive verbs. Synchronic lexical rules must thus specify that from a
root like muy one can produce both a lexical verb and an auxiliary, differing
minimally as lexical vs. functional heads of VPs, along lines suggested by Aissen
(1993). The DIR arises from still more severe decategorialization, since it
functions as an adverbial enclitic. How exactly to represent such lexical relations
and the grammatical categories expressed by a single underlying root is a problem
still to be resolved.
3. Conceptual generalization.
Although as full lexical verbs, the roots in question
are often extended to a variety of metaphorical or non-
motion senses, xii as auxiliaries their meanings are
highly schematized, limited to the sorts of
trajectories and paths (Talmy 1985) expressed by closed
class grammatical elements in other languages (for
example, English verbal particles and prepositions).
The sense of literal motion may be generalized onto non-spatial domains
in characteristic and familiar ways. For example, the auxiliary ba(t) 'go' in the
incompletive aspect can have a clear future meaning, as in (43).
(43) CHID:27:
j-tak ta k'anele, yu`un ch-ba tal-uk
1E-send PREP wanting because ICP-go(AUX) come-SUBJ(+3A)
(However much [liquor]) I send for, it's going to come.
Given the mutually inconsistent meanings of bat 'go' and tal 'come' as MOTION
verbs the only apparently possible reading of the auxiliary construction ch-ba tal-
uk is, as in the idiomatic English translation, "it's going to (i.e., will) come." Bat
thus stands at a rather standard point along the path from motion verb to
tense/aspect marker.
As an extension of the Tzotzil aspect system,
auxiliaries exceptionally combine with predicates which
without auxiliaries could NOT bear aspect directly at
all. (44) shows that surface adjectives which derive
from "positional" roots (see Haviland 1992) can combine
with auxiliary verbs. xiii Uniquely, such normally
stative predicates can in an auxiliary construction
bear aspect.
(44)T83:32
ch-tal chotl-ik-on ta y-olon `one
ICP-come(AUX) sitting-SUBJ-1A PREP 3E-underneath avocado
I'm coming to sit under the avocado tree. (CK:197)
Such a construction is possible with positional adjectives denoting normal human
positions (e.g., va`al 'standing', puch'ul 'lying prone', even nakal 'at home,
residing') when the subject argument can be construed as a volitional agent.
With DIR elements, there are also semantic shifts and reanalyses. For
example, the DIR batel, which should mean 'away, going' (and DOES mean
exactly this in neighboring Tzotzil dialects) in Zinacantán has come to have more
of an aspectual flavor: 'from time to time.' In its place `ech'el 'passing' has been
recruited to fill the deictic "away from here" slot in the directional paradigm (as in
(21), (36), (39), (40)).
Nonetheless, such shifts do not simply represent a
dogged progression towards a TAM target, bleached of
motion entirely. The highly schematized trajectories
or paths encoded in AUX and DIR elements seem to be a
target in themselves. Thus the gap left by
reinterpreting `ech'el is in turn refilled by inventing
the otherwise idiosyncratic DIR jelavel 'passing.' xiv
Here Tzotzil and, indeed, languages across the Mayan family represent a
minor challenge or addition to Heine's view of auxiliaries as necessarily involving
a progression from Verb to TAM. Mayan languages seem to use AUX
constructions to encode not only tense and aspect, but also path and trajectory.
They use auxiliaries and directionals to build space directly, as it were, into
grammar.

i
Except where noted, examples are drawn from recorded
conversations or from Tzotzil texts published in
Laughlin 1977 (abbreviated as CK), or Laughlin 1980
(abbreviated SSS).
ii
See Haviland (1981), pp. 224-227, Aissen (1987), pp.
15-18, 214-229, and Aissen (1993), sect. 2.1, for
Tzotzil "subjunctive" form and usage. Aissen (1993)
bases much of her argument about the syntax of
auxiliary constructions on parallels with a causative
construction in which a fully inflected form of ak'
'cause' combines with a dependent clause whose verb
exhibits the same subjunctive morphology. Also
involving subjunctive form are desiderative
constructions with chak 'want' or -o`on-uk 'would like.'
iii
In recent work, Judith Aissen (1992, 1993) proposes a
syntactic account of why the morphology is distributed
in exactly the way it is, though that will not be a
direct concern here.
iv
The classification is presented in Haviland 1991a.
v
The root lik can mean both 'arise' and 'start.'
vi
For example, Dürr (1990) states that in classical
Quiché motion auxiliaries when combined with a
transitive main verb "indicate the movement of the
agent, contrary to the general ergative pattern" of the
language.
vii
More detailed argument about passive V2s is given in
Haviland 1991a:16-19, and especially Aissen 1984, 1993.
viii
The positions of jelavel and other parenthesized
directionals in this table are provisional.
ix
Similarly, Craig (1979:37) argues that cardinal
directionals in Jakaltek "express the movement of the
actor/subject" with intransitive main verbs but with
transitive verbs "refer to the movement of the
object/patient." Jakaltek deictic directionals
'toward/away' are calculated from the perspective of
the actor/subject. In languages with more highly
schematized directionals, no obvious pattern of
association is evident. See, e.g., England 1976 on
Mam.
x
Indeed, in the neighboring Tzotzil dialect of
Chamula, the apparent AUX la(j) 'finish' now functions as
the completive prefix in the normal verb paradigm.
xi
Aissen (1993) points out that the CV shape of the
reduced auxiliaries is quasi-canonical for functional
elements in the language, so that the apparent
phonological erosion has a kind of phonotactic target.
xii
Consider, for example, the extended meanings of `ech'
'pass' and lok' 'exit' as main verbs in (12) and (14).
See Haviland 1991a:43-55 for further semantic details.
xiii
Such cases in which the complement selected by the
auxiliary is not (formally) a verb at all represent a
problem for Aissen's (1993) analysis of auxiliaries as
functional heads in which, following Abney 1987, she
takes it as a defining characteristic of functional
heads that they select unique complement types.
Positional predicate adjectives cannot ordinarily bear
aspect, although they do have a number of verb-like
properties (see Haviland 1992). Note that AUX is NOT
possible with fully verbal stems derived from such
roots.
xiv
This form is unique in both phonological shape and
lexical provenance, since it is the only directional
derived from a disyllabic stem, and indeed, from a
DERIVED stem formed by suffixing an intransitivizer -av to
the root jel 'change, exchange.'
References
Abney, Stephen. 1987. The English noun phrase in its sentential aspect. Unpublished Ph.D. dissertation.
MIT, Cambridge, Ma.
Aissen, Judith L. 1984. "Control and command in Tzotzil purpose clauses." Proceedings of the Tenth
Annual Meeting of the Berkeley Linguistics Society, 559-571.
Aissen, Judith L. 1987. Tzotzil Clause Structure. Dordrecht: Reidel.
Aissen, Judith L. 1992. Tzotzil auxiliaries of motion. Paper presented at the 91st annual Meetings of the
American Anthropological Association, San Francisco, Dec. 4, 1992.
Aissen, Judith L. 1993. Tzotzil auxiliaries and related constructions. Ms.
Craig, Colette. 1979. Jacaltec, field work in Guatemala. Languages and Their Speakers, ed. by Timothy
Shopen, 3-57. Cambridge, Ma.: Winthrop.
Craig, Colette. 1992. Grammaticalization of space in Mayan: Jakaltek directionals in family perspective.
Ms.
Dürr, Michael. 1990. Reference to space in colonial Quiché. Función, 1989-90, Universidad de
Guadalajara, Jalisco, México.
de León, Lourdes. 1991. "Space games in Tzotzil: creating a context for spatial reference. " Working paper
no. 4, Cognitive Anthropology Research Group, MPI for Psycholinguistics, Nijmegen, The
Netherlands.
du Bois, John W. 1992. Topographical strategies for topic management in Sakapulteko Maya. Paper
presented to the Annual Meetings of the American Anthropological Association, Dec. 1992, San
Francisco.
England, Nora C. 1976. Mam directionals and verb semantics. Mayan Linguistics I, ed. by Marlys
McClaren, 202-211. UCLA, American Indian Studies Center.
Haviland, John B. 1981. Sk'op sotz'leb: el tzotzil de San Lorenzo Zinacantán. Mexico City: Universidad
Nacional Autónoma de México.
Haviland, John B. 1991a. The grammaticalization of motion (and time) in Tzotzil. Working paper #2,
Cognitive Anthropology Research Group at the Max Planck Institute for Psycholinguistics.
Haviland, John B. 1991b. Projections, transpositions, and relativity . Paper prepared for Wenner-Gren
conference on Rethinking Linguistic relativity, Ocho Rios, May 3-11, 1991. Working Paper:
Cognitive Anthropology Research Group, Nijmegen.
Haviland, John B. 1992. Seated and settled: Tzotzil verbs of the body. Zeit. Phonetik, Sprachwissenschaft
und Kommunikationsforschung 45.543-561.
Heine, B. 1992. Auxiliaries. Ms.
Laughlin, Robert M. 1977. Of cabbages and kings, tales from Zinacantán. Washington, D.C.: Smithsonian.
Laughlin, Robert M. 1980. Of shoes and ships and sealing wax, sundries from Zinacantán. Washington,
D.C.: Smithsonian.
Talmy, Leonard. 1985. Lexicalization patterns: semantic structure in lexical forms. Language typology and
syntactic description, Vol 3, ed. by Timothy Shopen, 57-149. London: Cambridge University Press
Zavala Maldonado, Roberto. 1992. Mayan directionals: patterns of grammaticalization. Paper presented at
the 91st annual Meetings of the American Anthropological Association, San Francisco, Dec. 4,
1992 .