1The Problem with Basic Alignments

Big Star, Little Star

1the problem with basic alignments

candidates can violate immediate ordering constraints (precedence/subsequence) in one of two ways

being on the wrong side (= side violation)

being too far from the host (= adjacency violation)

xPy
... yxz ... / *
... xzy ... / *

this is fine if the relevant element can win a competition as it will immediately precede or follow the host

the issue is: what happens to losers?

do they stay on the right side, but accept a non-adjacent position (=elbowing)?

do they swap sides and stay close to the host (=side switching)?

The data seem to support elbowing:

the old manthe determiner is elbowed to the left by the adjective

* old man the

give Mary the moneythe direct object is elbowed to the right by the indirect object

* the money give Mary

2a possible solution

given that there are two ways to violate a constraint and that only one of these ways is acceptable, suppose we claim that the two cases involve different types of violation:

adjacency violation  incurs a violation marked by * (little star)

side violation  incurs a penalty marked by  (big star)

a big star violation is worse than any number of little star violations:

xPy
xayb ... / *
xaby ,,, / **
xab ... y / ** ...
abyx ... / !

This predicts that elbowing is the universal strategy for dealing with losers

3Universal word order phenomena

of the 6 possible orders of the 3 basic elements of a sentence, all are attested

however, some are more frequent than the others

SOV = 45%

SVO = 42%96% = SO

VSO = 9%

VOS = 3%

OVS = 1%(some uncertainty on correctness of analysis)

OSV = <1%

the uncertainty of OVS as a basic word order follows from the fact that the data is not straightforward

Hixkaryana allows various word orders and the verb is marked to indicate which argument is subject and object.

there is a hierarchy 3rd > 1st > 2nd: if the subject is higher in this hierarchy than the object, then one morpheme is used and if it is lower then another morpheme is used

in cases where they are the same, word order distinguishes them

in this case the object is the argument in front of the verb and the subject is after

3rd V 3rd(e.g. man eat jaguar) = O V S = the jaguar eats the man

on the basis of this, it is argued that OVS is the basic word order and others are derived

it also needs to be considered that this statistical analysis forces some round pegs into square holes.

there are many languages which have questionable status with regard to this typology

Hungarian is not easily categorised as any of these

all 6 possibilities are quite frequent (but what has frequency got to do with it)

there is no syntactic reason to think that the subject is given any special treatment, as opposed to other arguments (i.e. perhaps Hungarian has no subject

Note: thematic role, Case, discourse properties are not the same as subject

not all subjects are agents

not all subject are nominative

not all subjects are topics

possible the theoretically most justified assumption is that Hungarian has a V arg basic word order, all others being derived from this

this follows from the fact that all pre-verbal positions are associated with a particular interpretation, but post verbal positions are free

A number of languages show even greater word order freedom than Hungarian

Warlbiri, an Australian language, is known to have almost totally free word order, with the exception that the auxiliary is in second position

4Possible account in AS of word order patterns

Is ‘subject’ basic?

a number of approaches argue that the notion subject is derived

the subject is just an argument that happens to occupy a certain position

hence different arguments can be subjects under different conditions

others argue that it is a basic notion of human grammars

that different arguments are associated with subject in different constructions is a matter of grammatical process

for present purposes we will assume that subject is assigned in the input and hence is subject to possible constraints

this might be justified on the observation that the subject is usually an argument, though it is an argument given special treatment

other arguments typically have a position different from the subject

in SVO languages, the only argument to precede the verb is the subject, all others follow

John saw MarySVO

He reacted to the newsSVPP

I gave the money to MarySVOPP

I said he leftSVIP

I promised him that I would goSVOIP

it seems to me that he is wiseSVPPIP

so whichever argument is assigned to the subject is given special treatment

However, it has to be accepted that the subject is also usually an argument and hence it will also be subject to whatever conditions apply to arguments

The analysis

sP/Fpsubject precedes/follows predicate

aP/Fp argument precede/follows predicate (i.e. head initial/head final distinction)

Languages where there is no apparent subject

we wouldn’t want to say that such languages have no subject in the input

inputs should be universal

so the lack of the subject must be grammatical

the obvious suggestion is that the feature is deleted

this can be achieved by the following ranking

sPpsFp > Faith(sub)

SVO

this is easy to achieve

sPpsFpi.e. subject first language

aFpaPpi.e. arguments in general follow the predicate

sPpaFpi.e. although the subject is an argument, it still comes first

Faith(sub) > sPp/sFpthe subject feature does not delete

5Problems

OVS is just as easy to produce (just opposite rankings) and so it is not easy to see why SVO is more frequent than OVS

SOV and VOS word orders cannot be achieved

this is because of harmonic binding

a candidate is harmonically bound if there is another candidate which does equally well/badly or better on all constraints

sPp / sFp / aPp / aFp
OSV /  / * / 
SOV / * /  / * / 
VSO /  /  / *
VOS /  / * /  / *

it doesn’t matter which order the constraints are in, OSV will always beat SVO and VSO will always beat VOS

therefore SOV and VOS can never be grammatical

VOS is not a very common language – but it is more common than VSO

SOV is the most common language type!!!

6Possible solutions

introduce a sP/Fo constraint

this might be the basis of an account for why S precedes O in most of the world’s languages

though it is not clear why sPo should be more frequently ranked above sFo

this necessitates that we assume that object is a universal feature, assigned in the input

but there is little reason to believe that objects are given any special treatment

in general objects are DP arguments that are not subjects