draft, December 2007

comments welcome: melfraz@email.unc.edu

Default Case in OT Syntax

Melissa Frazier
University of North Carolina, Chapel Hill

Abstract
Default case appears when a DP cannot get case through normal syntactic mechanisms (Schütze
2001). In an OT system, default case is best analyzed as an example of emergence of the
unmarked. When case cannot be assigned, the least marked case is used. A constraint system is
developed that includes markedness constraints for assigning structural case and faithfulness
constraints for assigning nonstructural case (in partial agreement with Woolford 2001a). Default
case is controlled by the ranking of constraints that penalize the appearance of certain cases – the
lowest ranked of such constraints correlates with the default case of the language. This is an
argument against Woolford’s (2001a) proposed universal ranking of these constraints. The
constraint system developed here is used to analyze bare pronoun answers to questions (e.g.
“Who is going?”, answer: “Me.”) in English (where default case is used) and German (where
default case is not used).

1. Introduction

Case systems of languages show substantial diversity.1 One way case systems differ

from one another is in the use of default case. Default case seems to appear when a DP cannot

get case through normal means. Such DPs have case, but this case is not necessarily related to

the semantics or syntax of the DP, i.e. the case does not denote a grammatical relation or a theta

role, etc. For example, in the sentence ‘Me, I like beans’, me has accusative case, which is the

default case of English (Schütze 2001). There is language variation in terms of which case is the

default of the language and in terms of which constructions utilize default case. This paper

investigates default case systems in search of some basic properties that constrain the cross-

linguistic variation.

Because a general goal of this paper is cross-linguistic analysis, I believe theoretical

advantages can be gained by using a system based on competition, namely Optimality Theory

1
Unless stated otherwise, case in this paper will refer to both abstract case and surface morphological case marking.
 2

(OT, Prince and Smolensky 1993, McCarthy 2002). When we use violable constraints, these

constraints can be ranked in various ways to derive various case systems. If a pattern cannot be

derived from any ranking of the given constraints, that pattern is predicted not to exist in any

language.

The specific goals of this paper are thus to introduce constraints relevant to default case,

use the constraints to explain some basic differences between German and English in terms of

default case, and to discuss the languages that are and aren’t predicted to exist with the given

constraint set. I will make use of the types of case proposed by Woolford (2006, structural,

lexical, and inherent). I will amend this classification of case by adding another type – default

case – which is not structural, lexical, or inherent. Default case is an example of the emergence

of the unmarked. Following Woolford (2001a), constraints exist that rule out certain cases, i.e.

*ACCUSATIVE says that accusative case is forbidden. My claim is that default case correlates

with which ever of these constraints is ranked the lowest in a given language. For example, if

*ACC is dominated by all other such constraints (e.g. *NOMINATIVE, *DATIVE, etc.), accusative is

the default case for that language. This claim goes against the proposed universal ranking of

*DAT » *ACC » *NOM (Woolford 2001a).

This paper will proceed as follows. I will first present my assumptions about OT syntax

in §2. I will then outline different types of case and how to model them in §3. This section will

present relevant data from German and English and relevant OT constraints. In §4 default case

in English is analyzed and a challenge presented by English infinitives is addressed. German is

discussed in §5. Implications for a factorial typology are in §6 and conclusions follow in §7.

2. Assumptions about OT Syntax

As McCarthy (2002) points out, there are many issues that are taken for granted in OT
 3

phonology but can be controversial in OT syntax. Of particular interest here are three questions

he poses (p 193): What is the input? What are the candidates? and What are the faithfulness

constraints like? McCarthy’s answer to the first two questions is “just about anything” with a

caveat: the input must be able to “interface sensibly” with GEN and with faithfulness constraints,

and the candidates must be able to “interface sensibly” with faithfulness and markedness

constraints. Answers to the third question vary, but of course the nature of faithfulness

constraints is dependent on the nature of the input and the output candidates. Thus, the answers

to these three questions are dependent on one another. In this section I will make my

assumptions about OT syntax explicit with respect to the input, the output, and GEN.

Regarding the input, I will follow Legendre (2001:20) who says that there is some

agreement among OT syntacticians that the input must (at least) contain “predicate-argument

structure, lexical items, information and illocutionary features, level of argument prominence, as

well as familiar functional features (tense, aspect, and so on)...” I believe this viewpoint is

summarized by Lyngfelt (2000), Hendriks and De Hoop (2001), and Wilson (2001) who say that

formal semantic structure is the input for OT syntax. It is important to note that this approach

does not utilize syntactic structure in the input.

The output does, of course, contain syntactic structure. Because such structure is not

present in the input, it must be supplied by GEN, which is the universal candidate generator.

There can be no language specific requirements on GEN, nor is GEN the “appropriate repository

for language universals” (McCarthy 2002: 194). There is thus a general desire to keep GEN as

simple as possible. For the purposes of this paper, I assume that GEN creates candidates that

conform to the X' schema. I also follow Woolford (2001a:511), who says that any candidate

violating the general principles of Case theory “will be eliminated from the candidate set before
 4

the violable constraints apply.” One effect of this stipulation is that GEN can only create

candidates wherein all DPs have case. In accord with “freedom of analysis” (McCarthy 2002:

8), GEN supplies an infinite set of candidates for any given input, which is then evaluated by the

constraint ranking (as illustrated in further detail below).2

If the input contains formal semantic structure and the output contains formal syntactic

structure, it is clear the faithfulness constraints must be able to refer to the semantic structure of

string 1 (the input) and compare it to the syntactic structure of string 2 (the output candidate).

This means that OT faithfulness constraints, in addition to demanding similarity between string 1

and string 2, may also have to specify exactly what it means to be similar. For example,

Woolford’s (2001a) FAITH-LEX constraints (see (3) below) say that an inherent case-marking

feature (in the semantic input) must be checked (in the syntactic output). Thus, the input and the

output are the “same” if a DP in the output appears with the surface case marking that correlates

with a feature present in the input. On the other hand, some properties, like lexical items, are

consistent in both semantic and syntactic form. For example, it might be desirable to have a

faithfulness constraint MAX-LEX, which says that any lexical item present in the input must also

be present in the output. Such a constraint does not need to refer to semantic or syntactic

structure, meaning that comparison for similarity between input and output is straightforward.

Thus, faithfulness constraints can only compare features that are present in the input to

features that are present in the output. Because input and output do not contain the same type of

structure, faithfulness constraints in OT syntax may have to be explicit about what they are

looking at in both strings (input and output candidate). This differs from faithfulness constraints

2
This means that the candidates created by GEN are created without reference to the input. According to freedom of
analysis, GEN actually creates the same infinite set of candidates for any input in any language. Most of these
candidates will, of course, be harmonically bounded such that they would not win no matter what the constraint
ranking (see example (1) below).
 5

in OT phonology, where input and output have the same type of structure (though not always – it

is generally assumed that certain prosodic information (like syllable boundaries) is not present in

the input though it is in the output). OT phonology faithfulness constraints usually say that an

output should be identical to the input in such and such a way. OT syntax faithfulness

constraints do the same thing, but often they must specify what it means for the output to be

identical to the input due to the different types of structure present in input and output.

I will illustrate these claims with a simple example. Consider the input 1s-AGENT go-
3
PRES.PROG. This input contains the lexical entry go and a reference to a first person singular

pronoun. Each of these entries is also supplied with some semantic information, denoted by

small capitals (which will be the notation used throughout this paper). This extra information

tells us that the first person singular pronoun should bear the theta-role of agent and that go

should be inflected for progressive aspect and present tense. There is no other information

expressed in the input.

The job of GEN is to create an infinite set of candidates with syntactic structure. Some

potential candidates are shown below.

(1) a. I am going b. am going I c. I go d. Ben had arrived

The winning candidate for this input in English would, of course, be (a). Various constraints

would rule out candidates (b-d). Candidate (b) violates constraints on word order in English.

3
Though I make the claim that the input contains formal semantic structure, for ease of exposition, I will use the
notation presented in “1s-AGENT go-PRES.PROG”, where lexical items are in lower case italics and all other
information (theta marking, tense, etc.) is in small capital italics.
 6

Candidate (c) is unfaithful to the input in that the verb does not agree with the tense and aspect

specifications. Candidate (d) is unfaithful to both the lexical and semantic specifications of the

input. It should be noted the candidate (b) could be a winner in another language with a different

constraint ranking (assuming different but semantically equivalent lexical items) and candidate

(c) could be the winner in a language with no tense or aspect morphology. Candidate (d) would

be harmonically bounded in any language, which means it would never be the optimal candidate

for this input no matter what the constraint ranking of the language is. This candidate was

included to illustrate the point that GEN is blind to the input, i.e. it does not only generate

relevant candidates. It is up to the constraint ranking, and hence the EVAL function to decide

which candidate wins. The “irrelevant” candidates, like (d), will be harmonically bounded and

will never win.

Another important point illustrated by (1) is the difference between faithfulness and

markedness constraints. Candidate (b) is fully faithful to the input; it does not deviate

semantically or lexically. It loses because of high-ranking markedness constraints that control

word order in English. Markedness constraints evaluate only the output candidate and, like GEN,

are blind to the input. In fact, a candidate like (d) would actually fare better than (b) in terms of

markedness because it utilizes correct English word order. On the other hand, candidates (c-d)

are ruled out by faithfulness constraints. Faithfulness constraints compare the output candidate

to the input and demand similarity.

3. Case in OT Syntax

It is clear that case is a source of language variation. Languages are often classified as to

whether they are nominative/accusative or ergative/absolutive, but neither of these is a perfectly

homogenous group. There is a general trend in the literature to handle case by making language-
 7

specific claims about the assignment of case. For example, with respect to default case, Hwang

(1997:86) says, “Default case is checked and licensed by T with the feature [-tense]... nominative

case is checked and licensed by T with the feature [+tense],” which is only applicable to English

and any other similar languages (as recognized by Hwang). On the other hand, Roehrs

(2002:224) says, “Nominative case is the default option in PF when a DP has not had a case

licensed,” which is not applicable to English.

In an OT system, the claims made above by Hwang and Roehrs could be encoded into

violable constraints, and because these constraints would be freely rerankable in any language,

there is no problem with the fact that not all languages seem to be affected by such constraints.

Furthermore, candidates that satisfy any of the relevant constraints would compete in all

languages, with the winner chosen by the constraint ranking. This approach will shed some light

on the theoretical nature of case assignment that would not be possible in a system that only

focused on the language-specific nature of case without taking into account cross-linguistic

generalizations. In an OT analysis, it is impossible to ignore cross-linguistic generalizations.

3.1 Default Case

In many languages, we find a pattern of one particular case appearing in constructions

where typical conditions for case assignment are not met. For example, in “Mad Magazine

expressions” in English like ‘What? Him eat carrots? Never!’ we see accusative case with him,

even though this pronoun is acting as the subject and as such would normally receive nominative

case (cf. he eats carrots). The same pattern can be found in English in a variety of constructions

(see Schütze 2001 for a summary). What all of the constructions have in common is that

accusative case is used no matter what case is normally assigned to such structural positions or

theta-roles. This evidence suggests that accusative is the default case of English.
 8

According to Schütze (2001:1-2), “The default case forms of a language are those that are

used to spell out nominal expressions (e.g. DPs) that are not associated with any case feature

assigned or otherwise determined by syntactic mechanisms.” He also notes, “We find a set of

elements behaving alike, where the members of that set have nothing in common other than the

fact that they do not belong to any natural class.” Default case is thus diagnosed as one case that

consistently covers an unnatural class of theta roles or structural positions. This contrasts with

what I will call assigned case, which is any case normally assigned to a DP that expresses

particular theta roles or appears in certain structural positions. In he eats carrots, he has assigned

case (nominative). In him eat carrots?, him has default case (accusative).

3.1.1 Data from English and German

Default case appears in a wide variety of contexts in English. This is not so in all

languages. English uses its default accusative case when pronouns are used as answers to

questions, as shown in table 1, while German does not. In table 1, an example is provided where

the pronoun represents the subject of an intransitive (S), the subject of a transitive (A), and the

object of a transitive (O, following Dixon 1979). We see that English uses accusative case for

each of these positions and that German has “case-matching” following a nominative/accusative

case system. Thus, we see that languages may or may not utilize default case for bare pronoun

answers. Given that these pronouns are the only overt element of the utterance, it is almost

surprising that default case is not used in these situations in German. One may wonder how the

German pronouns are getting case. The fact that there is language variation in case even in

simple one word responses will have important implications for the theory presented here (see

§4-5).
 9

Table 1: Bare pronoun (DP) answers to questions4

all bare pronoun answers use a 1st person singular pronoun

English German
Who is going? Wer geht?
who-NOM go-3S
S Me-ACC/*I-NOM. Ich-NOM/*Mich-ACC.

Who hit John? Wer hat Hans geschlagen?

who-NOM has Hans hit-INF
A Me/*I. Ich/*Mich.

Who did John hit? Wen hat Hans geschlagen?

who-ACC has Hans hit-INF
O Me/*I. Mich/*Ich.
3.1.2 Default Case in OT Syntax

Schütze (2001) says that default case results when case cannot be assigned by ordinary

“syntactic mechanisms”. In OT, the ordinary mechanisms are of course constraints, and more

importantly, the constraint ranking. Assigned case results when constraints effectively allocate

case based on structural positions (structural case), theta roles (lexical case), or inherent lexical

specifications (inherent case, see Woolford 2006 for an explanation of these different types of

cases, and §3.2 below). Default case results when these constraints fail to determine a winning

case. This suggests that default case is an example of the emergence of the unmarked and that

this type of case is thus determined by markedness constraints that rule out certain cases (i.e

*DATIVE, Woolford 2001a).

The data presented in table 1 is thus somewhat unexpected given Woolford’s (2001a)

proposed universal markedness hierarchy with respect to case. Using the fixed ranking of OT

constraints (which are defined in more detail below in §3.2), *DATIVE » *ACCUSATIVE »

*NOMINATIVE, nominative is the least marked case and should thus appear as default case.5 In

4
Thanks to a native speaker, Rahul Bhatnagar, for help with the German data.
5
In fact, Woolford (2001a:footnote 10) proposes that *NOM does not even exist. If the hierarchy really were
universal, this would be the correct approach because no language would ever penalize nominative case. The
 10

this paper I argue against the existence of a universal markedness hierarchy with respect to case,

based on the evidence presented in table 1. Woolford (2001a) admits in a footnote that

accusative default case in English is problematic for this constraint hierarchy and thus assumes

that other constraints control the use of an accusative default in English. I don’t think these other

undefined constraints are necessary or that a fixed universal hierarchy with respect to *CASE

constraints is necessary.6 I propose that in any given language the default case will correlate

with the lowest-ranked *CASE constraint, i.e. if *NOM is ranked lower than all other *CASE

constraints, nominative is the default case of that language.7

In this way, I believe that Schütze’s claims fit nicely with an OT framework. When a

constraint that is responsible for assigning case (via ordinary syntactic mechanisms) is dominated

by the *CASE constraints, case cannot be assigned by that constraint. In the winning candidate,

any DP occurring in the relevant position will appear with the default – the least marked – case.

In the next section I turn my attention to these other constraints that assign case.

3.2 Types of Case and Relevant Constraints

According to standard case theory, there are two types of case, commonly called

analysis presented here make crucial use of *NOM and so I believe it is a member of CON (the universal constraint
set).
6
The problem cannot be remedied by appealing to the impoverished case system of English. Schütze (2001) also
cites Norwegian, Danish, and Irish as having accusative default case, while German, Spanish, and a handful of other
languages have nominative default case. I have concluded that Yucatec Maya (an ergative Mayan language) has
dative default case (based on the fact that dative is always used with bare pronoun answers and further discussion
with a native speaker, Santiago Domínguez). It is clear that there is language variation with regard to default case.
7
It should be noted that Woolford (2001b) develops the universal case markedness constraint hierarchy to explain
Burzio’s Generalization (Burzio 1986:178). She looks at data from Hindi, German, Icelandic, and Faroese, shows
that the universal hierarchy makes the correct predictions with respect to case in these languages, and argues that
nominative objects appear when the subject has another case because nominative is the least marked case. While I
argue against the universality of the hierarchy I do not argue against this ranking of the *CASE constraints for these
four languages. Instead, I propose that this is the ranking of these constraints for Hindi, German, Icelandic, and
Faroese because all of these languages utilize nominative as the default case (Schütze 2001 (German and Icelandic),
Wunderlich 2000 (Hindi), Schütze and Wexler 1996 (Faroese)). There is thus not an immediate conflict between
her analysis and my rejection of the universal hierarchy. The prediction is made that if a language upholds Burzio’s
Generalization, this language will have a nominative default case. If a language is considered an exception to
Burzio’s Generalization, this language likely utilizes a default case other than nominative.
 11

structural and non-structural. Woolford (2006) argues that there is a further division in non-

structural case, shown below in (2). We thus have three different types of case: structural case is

assigned based on structural position (e.g. the specifier of TP receives nominative case), lexical

case is assigned idiosyncratically by particular lexical entries (e.g. a certain verb requires a dative

subject), and inherent case is assigned with respect to theta-roles (e.g. the agent receives

nominative case).

(2) types of case (Woolford 2006):

Case

Structural Non-Structural

Lexical Inherent

It seems that default case is different from each of these three types of case. Because

those DPs who receive it form an unnatural class, it cannot be assigned based on syntactic

structure or theta-roles. Furthermore, it is too consistent to be the unpredictable lexical case. We

are thus dealing with a fourth type of case and a third type of subdivision in the case schema.

Given an OT framework, it is worthwhile to look for explanations as to the differences between

these types of case in the constraints that assign them.

Woolford (2001a) proposes that lexical case is assigned via a faithfulness constraint,

shown in (3). If this is true, it would be ideal for inherent case, the other type of non-structural

case, to also be assigned via a faithfulness constraint, developed in (4).

(3) FAITH-LEX: A lexically specified inherent case marking feature must be checked.
(Woolford 2001a)

(4) FAITH-θ: A DP bearing the theta-role α has case β.

It should be noted that the above constraints really stand in for constraint families. For

example, one constraint in the F-θ family might be F-AGENT, as defined in (5). This constraint
 12

may be ranked differently from other constraints in the F-θ family. Thus a language may

prioritize which theta-roles are realized in which way.

(5) example F-θ constraint:

F-AGENTÆNOM: A DP bearing the theta-role of agent has nominative case.

While non-structural case is assigned by faithfulness constraints, I propose that structural

case is assigned by markedness constraints. Again, the constraint defined in (6) stands for a

constraint family. One constraint in this family, for example, may correlate with Hwang’s

(1997) statement quoted at the beginning of §3: if a DP α is the specifier of a TP that is [+tense],

then α must have nominative case. The STRUCCASE constraints require no assumptions about the

input. It does not matter what role a DP plays in the semantics of the input: a STRUCCASE

constraint looks for DPs in the specified position, and violation-marks are assigned if a DP is in

the appropriate syntactic position and does not bear the specified case.

(6) STRUCCASE: If a DP α is in the appropriate syntactic position to receive structural case β,

then α must have case β.

(7) example STRUCCASE constraint:

SC-SPECTPÆNOM: a DP in the spec of TP has nominative case

Thus far I have claimed that non-structural case is accounted for by faithfulness and

structural by markedness. There is no other constraint type that can account for default case, but

this is expected because the very nature of default case is that it is not assigned, i.e. we should

not expect a constraint that specifically says, for example, “such and such DPs receive default

case, which is accusative”. A particular phenomenon associated with OT is parallel to this idea

of default case not being directly controlled by markedness or faithfulness. In OT, the

appearance of a common form that is otherwise inexplicable is often connected with “the

emergence of the unmarked”. When higher ranked constraints cannot pick a winning candidate,

the candidate that wins is the one that is the least marked. I propose the default case of a
 13

language is the one that is the least marked, as determined by the ranking of multiple *CASE

constraints, defined in (8). Specifically, when F-LEX, F-θ, STRUCCASE, and *CASE are ranked in

such a way that the former three cannot select a winning candidate, the decision must be made by

the *CASE constraints. The winner will thus be the one that violates the lowest-ranked *CASE,

and so the unmarked case will emerge.

(8) *CASE: case α is forbidden. (Woolford 2001a)

(9) example *CASE constraint:

*NOM: nominative case is forbidden

Given these constraints, we can reevaluate the schema in (2) to include default case and

to define the different case types by what kind of constraint accounts for them:

(10) different types of case as controlled by different types of constraints (in partial agreement
with Woolford 2001a, 2006):
case

default: assigned
indirect markedness
structural: non-structural:
direct markedness faithfulness

4. Analysis of English

Now that we have constraints for assigning case and for determining default case, we can

proceed to an analysis of the English and German data presented in table 1 in §3.1.1. We need to

account for the fact that bare pronoun answers receive assigned case in German and default case

in English and for the fact that this default case is accusative (and not nominative). The analysis

of English presented in this section will illustrate some of the basic claims of this paper. We will

see how structural case is assigned by markedness constraints and how default case emerges

when the *CASE constraints outrank other relevant markedness constraints. I will also extend the

analysis to account for the absence of DP subjects in infinitive clauses (and their presence in
 14

Portuguese).

4.1 Bare Pronoun Answers

In this section, we will look at bare pronoun answers in English. For the time being, I

will assume that such answers result from a simple input specifying person and number of

pronoun and its theta-role (an assumption that will be adequate for now, but will change in §5,

when evidence against it is presented). We see in tableau (11) how an input specifying a first

person singular agent maps onto the accusative pronoun me. The winning candidate violates

both the F-θ constraint F- AGENT and *ACC, meaning both of these constraints must be

dominated by *NOM, as shown in (12). Because there is no syntactic structure, SC-SPECTP is

vacuously satisfied.

(11) English, bare pronoun answer (to the question Who is going?)
/1s-AGENT/ SC-SPECTP *NOM F-AGENT *ACC
I. *
) Me. * *

(12) *NOM » F-AGENT, *ACC

This ranking is compatible with the bare pronoun answer me when it correlates with the

object of a transitive verb. In this case, the output would not incur a violation of any F-θ

constraint (i.e. F-THEME), only of *ACC.

We must also be able to account for the appearance of assigned case in English. It is, of

course, possible to answer the question “Who is going?” with a full sentence, such as “I am

going.” I will assume, for now, that the answers “me” and “I am going” result from different

inputs, such that the later input contains the verb. In tableau (13), we see the winning full

sentence answer with the subject in nominative case. For space considerations, the structure of

these candidates is not shown, but it is the same structure given above for (1a). The losing

candidate fatally violates the STRUCCASE constraint SC-SPECTPÆNOM, which derives the
 15

ranking in (14).

(13) English, full sentence answer (to the question Who is going?)
/1s-AGENT go-PRES.PROG/ SC-SPECTP *NOM F-AGENT *ACC
) I am going. *
Me am going. * * *

(14) SC-SPECTP » *NOM » F- AGENT, *ACC

We now have a ranking that accounts for assigned case in English full sentence answers

and default case in English bare pronoun answers. It is worthwhile to reflect on what this

ranking tells us about English, given Schütze’s (2001) definition of default case as employed

here. In full sentence answers, subjects have structural case as assigned by SC-SPECTP, which is

a normal “syntactic mechanism”. This means subjects in full sentence answers have assigned

case due to the fact that SC-SPECTP dominates the *CASE constraints. To be exact, default case

cannot surface (i.e. the unmarked case does not emerge) in such syntactic positions because it is

more important for subjects in such sentences to have nominative case. In bare pronoun

answers, default case is utilized because SC-SPECTP cannot assign case (i.e. it cannot pick a

winner among the possible candidates) and F-AGENT is outranked by a *CASE constraint, namely

*NOM. Because normal syntactic mechanisms cannot assign case, the unmarked case is deemed

optimal by the grammar.

There is a generalization that can be made at this point about default and assigned case in

an OT grammar. If a constraint of the F-θ, F-LEX, or STRUCCASE families dominates all *CASE

constraints, this constraint can influence assigned case (in contexts where the constraint is

applicable). Let’s call such constraints case-assigners in a given grammar. If a DP occurs in a

particular context such that no case-assigners are applicable, that DP will have default case (as

determined by the ranking of the *CASE constraints). In this manner, the grammar of a language

determines in what constructions DPs receive default case and in what constructions they receive
 16

assigned case. The more case-assigners the grammar has, the less often default case will appear,

as will be the case with German (see §5).

4.2 Infinitives

There is a perhaps undesirable implication that results from the previous analysis. If

default case is a viable option for any DP that does not receive case from the case-assigners, then

every DP can get case – either assigned or default case. This means that lack of case cannot

explain why DPs are prohibited in certain positions. Consider infinitive clauses, as shown in

(15). Such clauses never have overt subjects in English. A standard way to explain the absence

of subjects in this position is to say that a nonfinite TP cannot license case, and since a DP

cannot appear without case, a DP cannot appear in this position. This explanation will not work

with the system I have proposed because the constraint ranking devised so far predicts an

accusative subject to be grammatical. More generally, default case is always an option for a DP,

meaning there no reason any DP cannot get (default) case.

(15) English infinitives:

I tried PRO to leave.
*I tried me to leave.

Before working on a solution to this predicament, I would like to point out that the

problem is present in any analysis that utilizes default case and is not just a consequence of an

OT analysis. If the claim is made that default case results when normal syntactic mechanisms

for assigning case fail, then default case should be possible with the subject of an infinitive, just

as it’s possible with the subject of the nonfinite verb eat in Him eat carrots? The only way

around this problem for a non-OT analysis is to list every environment in which default case can

occur. Not only is such an approach unnecessarily stipulative, it is also against the nature of

default case.
 17

In an OT system, the solution lies in the constraints. It will be necessary for constraints

to directly penalize the appearance of subjects in infinitive clauses. Thus, it will be the case that

these DPs are not ruled out for not having case but are ruled out for other reasons. Given that

there is language variation in this matter, it is logical to expect this work to be done by violable

constraints. As shown below, Portuguese allows overt subjects in infinitive clauses, and even

uses agreement morphology on the verb. It should be noted that, because Portuguese is a pro-

drop language, a subjectless infinitive is also grammatical. However, because infinitives do use

agreement morphology and can appear with subjects, the claim is that the null subject of an

infinitive is little pro (just like the null subject of a matrix clause) and not big PRO (Garcia

2005).

(16) Portuguese infinitives (Garcia 2005:53):

Será difícil [eles/*PRO aprovarem a proposta ]
will be difficult [they/*PRO approve.infinitive+Agr the proposal]
"It will be difficult they to-approve-Agr the proposal (...for them to approve...)"

An analysis must be able to handle languages like English that completely disallow

subjects with infinitives and languages like Portuguese that do (optionally) allow them. It will

also be desirable for the analysis to account for the fact that a Portuguese-like language is

exceedingly rare. One way to do this would be for there to be multiple constraints penalizing the

appearance of overt subjects with infinitives. If there are many constraints disallowing them, it

becomes less likely for a language to have a constraint ranking with all such constraints ranked

low so that subjects could surface with infinitives as they do in Portuguese. Of course, we do not

want to just name a bunch of constraints that do the same job. If multiple constraints disallow

these subjects, these constraints should also be doing other work.

Due to the fact that some languages do not require overt subjects, it seems reasonable that

there is a constraint that penalizes overt subjects in any context, as defined below:
 18

(17) *OVERTSUBJ: no phonologically overt subjects are allowed

In addition to the general constraint *OVERTSUBJ, there should be contextual *OVSUB

constraints that only penalize overt subjects in certain environments, such as nonfinite and/or

non-matrix clauses. It is also likely that the properties of the null subject in a nonfinite clause

make it undesirable as an overt subject. Chomsky and Lasnik (1995) claim that PRO is

[+anaphor, +pronominal] (the only possible subject with these feature values). Reflexives and

reciprocals are also [+anaph], while pronouns are [+pro]. Hence, r-expressions are the only

possible subjects that are [-anaph, -pro]. It seems then, that languages might want to prohibit

either [+anaph] or [+pro] overt subjects but are unlikely to want to prohibit either [-anaph] or [-

pro] overt subjects. If there are two different *OVSUB constraints, one for [+anaph] and one for

[+pro], both would penalize an overt subject in PRO’s position.

(18) contextual *OVERTSUBJ constraints:

*OVSUB-[+anaphor]
*OVSUB-[+pronominal]
*OVSUB-embedded TP
*OVSUB-nonfinite TP

The constraints in (18) all penalize an overt subject in a nonfinite clause. However, they

all also penalize other types of overt subjects (except *OVSUB-nonfinite TP). Thus, they are

each responsible for ruling out a different set of overt subjects, and it is conceivable that these

different sets correlate with use of null subjects in different languages. The potential effect of

these constraints is illustrated in tableau (19). Here we see an input for the English sentence “I

tried to leave”. This input has two subjects, but only one is overt in the winning candidate. The

subject of the infinitive is deleted, which incurs a violation of MAX-LEX (as defined in §2), but

this is preferable to retaining an overt subject in an illegal position. The subject of the matrix

tensed clause is not deleted because there is nothing motivating deletion, and gratuitous
 19

violations of MAX-LEX are not allowed.

(19) infinitives in English

1s-AGENT try-PAST 1s-AGENT leave-INF *OVSUB/context MAX-LEX *OVSUB
) I tried to leave * *
I tried me/I to leave *! **
tried to leave **!

(20) {*OVSUB/context} » MAX-LEX » *OVSUB

If any of the *OVSUB contextual constraints proposed above dominate MAX-LEX, the

correct candidate wins. It is not necessary for all of them to dominate the faithfulness constraint,

which is denoted by the curly braces around the contextual constraint in (20). Thus, in any

language where one or more of the contextual *OVSUB constraints dominates MAX-LEX, this

language will disallow subjects in infinitive clauses.

In Portuguese, there are two grammatical outputs – one with and one without an overt

subject in the nonfinite clause. Each output is the winning candidate for a different input, where

the inputs either do or do not specify the subject of the nonfinite clause. In tableau (21), where

the subject of the infinitive is specified, we see that MAX-LEX outranks all of the *OVSUB

constraints – both the general and the contextual constraints. In tableau (22), where the subject

of the infinitive is left unspecified, we see that the losing candidate is harmonically bounded.

There is no motivation for inserting a subject where it is not lexically provided.

(21) infinitives in Portuguese (subject of infinitive in the input)

1s-AGENT try-PAST 1s-AGENT leave-INF MAX-LEX *OVSUB/context *OVSUB
) I tried I to leave * **
I tried to leave *! *

(22) infinitives in Portuguese (subject of the infinitive not in the input)8

1s-AGENT try-PAST leave-INF DEP-LEX *OVSUB/context *OVSUB
) I tried to leave *
I tried I to leave *(!) **(!) *(!)

8
DEP-LEX penalizes insertion of lexical elements not present in the input.
 20

(23) MAX-LEX » *OVSUB/context, *OVSUB

The difference between English and Portuguese thus has to do with the ranking of MAX-

LEX and the *OVSUB constraints. It is crucial that, for English, only one of the proposed

contextual *OVSUB constraints must dominate MAX-LEX, while for Portuguese, MAX-LEX must

dominate all contextual OVSUB constraints. This system predicts a Portuguese-like language to

be rare and an English-like language to be common, which is indeed the case.

4.3 Local Summary

This section has shown how to account for both assigned and default case. Case-

assigners are those constraints (of the STRUCCASE, F-LEX, or F-θ families) that outrank all

*CASE constraints. Only those DPs that occur in positions where no case-assigners are

applicable will appear with default case.

This analysis makes the strong claim that any DP can get case. The fact that DPs cannot

appear in certain positions is thus not a consequence of the inability to get case – it is a

consequence of constraints that ban DPs in these positions. By using rankable violable

constraints that penalized the appearance of certain DPs, we can account for a language like

English where no DP subjects are allowed in infinitive clauses and a language like Portuguese,

where such subjects are allowed.

5. Analysis of German

The German case system is a definite contrast to that of English. In this section we will

look at how to account for case-matching bare pronoun answers and how to extend this analysis

to sluicing, where there is also case-matching. The German data will require altering some of the

assumptions made in the previous section. Specifically, I will show that all inputs in OT syntax

must be able to be mapped onto complete sentences. When a statement, such as a bare pronoun
 21

answers to a question, is not a complete sentence, this is because the input that created the

statement contains a feature demanding the deletion of extraneous material.

5.1 Bare Pronoun Answers

In German, when the answer to a question correlates with the subject, the pronoun is in

nominative case, but when the answer to a question correlates with the object, the pronoun is in

accusative case. Thus, it is not the *CASE constraints that account for case in these

constructions, but rather some case-assigner. It seems that the pronouns are displaying structural

case, as subjects are in nominative and objects are in accusative. If this is so, there must be some

syntactic structure to these bare pronoun answers so that they can get structural case. This would

be possible if the rest of the sentence was elided, such that the verb and any other elements are

not pronounced but their structure is left behind. In (24a-b) we see the structure that remains

after the deletion of the VP (and the subject for (b)).

(24) elided structures for bare pronoun answers

a. Ich (gehe). b. (Hans hat) mich (gelschlagen).

If such structure is present in the output, its corresponding semantic structure must be

present in the input. I propose that all inputs in OT syntax must contain enough information to

be interpreted as full sentences, or enough information to be given a truth value. The bare

pronoun answers and full sentence answer still result from different inputs and that difference is
 22

in features that specify certain information should be elided. Hence the bare pronoun answer

results from the same input as the full sentence answer except that the former contains a feature

(i.e. [+elide]) that demands the deletion of certain overt elements.9

If this is so, then the English bare pronoun answers must also result from fully specified

inputs with this [+elide] feature. The question then arises as to why the English pronouns are not

able to get structural case. I believe it would be appropriate to make use of contextual case-

assigning constraints that only work when lexical entries are not elided. As shown in (25) with

an example, such constraints would able be able to assign case in clauses where no elision has

occurred. It is thus the case that the positional case-assigning constraints are high-ranking in

English, while the general case-assigning constraints are low ranking and cannot assign case in

structures where elision has occurred, such as bare pronoun answers to questions (note that

default case occurs in all elliptical structures in English (Schütze 2001)).

(25) STRUCCASE-NOELISION: If an NP α is in the appropriate syntactic position (in a structure

with no elision) to receive structural case β, then α must have β case.

Given these new assumptions, the tableau from (11) is redone below, in order to illustrate

the importance of these contextual case-assigners. The candidates do not show structure, but

they have structure similar to that displayed in (24) where the structure of the full sentence exists

though all lexical elements except the answer have been deleted. This tableau derives the

ranking shown in (27), where we see that *NOM dominates the general SC-SPECTP constraint.

Of course, English does utilize structural case in non-elided contexts. This is possible because

the contextual SC-SPECTP-NOEL can be ranked separately from the general constraint. This

9
For now I will assume that [+elide] calls for the deletion of old information. What is old information is determined
by the discourse and putting the current analysis into the context of a discourse is beyond the scope of this paper. It
should also be noted that having [+elide] in the input should not guarantee deletion – whether or not deletion occurs
would be subject to the constraint ranking. A discussion of what constraints control deletion is also beyond the
scope of this paper.
 23

constraint should be high ranking in English and can thus assign structural case when no elision

has occurred.

(26) English, bare pronoun answer (to the question Who is going?)
/1s-AGENT go-PRES.PROG SC-SPECTP-
[+elide]/ NOEL *NOM SC-SPECTP F-AGENT *ACC
I. *
) Me. * * *

(27) * NOM » SC-SPECTP, F-AGENT, *ACC

Assigned case occurs in German in bare pronoun answers, and so these pronouns must be

able to get structural case even though the rest of the sentence has been elided. In the following

tableaux, we see how the bare pronoun answers for German are derived. Again, structure is not

shown in the tableau, but the structure for the candidates is the same as that shown in (24).

(28) German, bare pronoun answer (to the question Who is going?/Wer geht?)
/1s-AGENT gehen-PRES.PROG [+elide VP]/ SC-SPECTP *ACC *NOM
) Ich. *
Mich. * *

(29) {SC-SPECTP or *ACC} » *NOM

(30) German, bare pronoun answer (to the question Who did John hit?/Wen hat Hans
geschlagen?)10
/Hans schlagen-PAST 1s-THEME [+elide subj, V]/ SC-COMPVP *NOM *ACC
Ich. * *
) Mich. *

(31) {SC-COMPVP or *NOM} » *ACC

In (29) and (30), the curly braces denote optionality – one of the constraints within them

must dominate the other constraint. This means that no explicit rankings are derived in the

above tableaux, but they do show us the importance of the STRUCCASE constraints – due to the

transivity of dominance at least one of these STRUCCASE constraints must dominate both *CASE

constraints (*ACC and *NOM). In other words, structural case occurs with bare pronoun answers

10
The constraint SC-COMPVP demands that the complement of the VP have accusative case.
 24

in German because of the high-ranking general STRUCCASE constraints.

5.2 Sluicing

Further evidence for the German analysis presented so far comes from sentences where a

specific type of ellipsis has occurred. Sluicing happens when an IP is elided and all that remains

is the C (Merchant 2001). This is illustrated in (32), where we see the only remaining part of the

embedded clause is the head of the CP. In (32a) the word for ‘who’ is inflected for dative case.

This is because the verb schmeicheln ‘flatter’ triggers lexical dative case in its objects (i.e. the

case of the object is assigned by a F-LEX constraint). This lexical case is assigned even when the

verb is elided. In (32b) structural accusative case is assigned to ‘who’ as the object of the verb

loben ‘flatter’ (again, even though the verb and the rest of the IP is elided), which does not

trigger lexical case.

(32) sluicing in German (Merchant 2001:42-43)

a. Er will jemandem schmeicheln, aber sie wissen nicht, {wem /*wen}.
He wants to flatter someone-DAT, but they don’t know who-DAT/*who-ACC (he
wants to flatter)

b. Er will jemanden loben, aber sie wissen nicht, {*wem /wen}.

He wants to flatter someone-ACC, but they don’t know *who-DAT/who-ACC (he
wants to flatter)

The sluiced clauses show that indeed elided structure and lexical entries can control

structural and lexical case, respectively. Structural case with elision was illustrated in tableaux

(28, 30), but we have not yet seen lexical case. In the tableau below, we will consider a

simplified input. This is just the input for the relative clause without the infinitive construction,

such that the expected relative clause output would be wem er schmeichelt ‘who he flatters’. As

demonstrated in this tableau, the case-assigner for lexical case must dominate the case-assigner

for structural case in order for lexical case to surface. Because these are general constraints, it

does not matter that the lexical entries are elided. If the input contained loben ‘flatter’, a verb
 25

that does not have the [+dat] feature, an output with accusative ‘who’ is expected because there

is no conflict with F-LEX (i.e. there is no semantic feature for the output to faithful to) and so

structural case is assigned by SC-COMPVP.

(33) lexical case in German

/3s-AGENT schmeicheln[+dat] wer [+rel] [+elide]/ F-LEX SC-COMPVP
wen-ACC *
) wem-DAT *

(34) F-LEX » SC-COMPVP

5.3 Local Summary

The analysis of German has shown that there are indeed more case-assigners in German

than there are in English. Specifically, in the ranking that accounts for German, the *CASE

constraints are dominated by a variety of other constraints that control case. This ranking does

not allow default case to surface in contexts where it does surface in English. The prediction is

made that in any language, the constructions that use default case are determined by the ranking

of the *CASE constraints with respect to the case-assigners.

6. Factorial Typology

When new constraints are proposed in an OT framework, it is desirable to consider what

predictions are made by the existence of the constraints by looking at a factorial typology. Given

three types of potential case assigners, STRUCCASE, F-LEX, and F-θ, different grammars depend

on how these constraints are ranked in reference to the *CASE constraints. Specifically,

whichever of the potential case-assigners dominate the *CASE constraints determines where

assigned case appears in the language and where default case appears. In one way this

variability is advantageous because languages do differ from one another as to whether default

case or assigned case appears in a given construction. In another way, however, it is possible

that too much variability is predicted by the theory presented here.

 26

Imagine a constraint ranking where the *CASE constraints dominate all STRUCCASE, F-

LEX, and F-θ constraints. Such a language would only utilize one case, whichever is the least

marked, meaning all DPs would have default case. Whether or not such a language exists

depends on whether case refers to abstract case or to surface morphological marking. If the

latter, it is clear the Mandarin Chinese is such a language, where neither nouns nor pronouns

receive any sort of overt case-marking (Yip and Rimmington 1997:8,20), meaning some default

case is always signaled by a null marker. However, Chinese can still be said to have abstract

case, as diagnosed by movement, verb type, and noun meaning (Li 1971).

Thus, the decision should be made as to whether or not the constraints cited and proposed

in this paper refer to abstract case or to surface case-marking. Either choice has undesirable

consequences. If all the constraints refer only to surface case-marking, the existence of a

language like Chinese is predicted and the factorial typology does indeed correlate with the set of

real languages. However, the system presented here would lose a significant amount of

explanatory power because abstract case is not accounted for at all. On the other hand, if the

constraints discussed here refer both to abstract and surface case, as was the original intention,

the factorial typology does not correlate with real languages. A language without any diagnostics

for abstract case is unprecedented to my knowledge. I argue against Woolford’s fixed hierarchy

because it is too restrictive, but in the process I have proposed a constraint set that is too

permissive.

At present, I do not have a solution to this dilemma, but there is a reason it is perhaps not

so problematic. While there is a significant emphasis on factorial typologies in any OT work, it

is also recognized that not every permutation of the constraints should compose a real grammar.

For example, one logically possible constraint ranking is one in which every markedness
 27

constraint dominates every faithfulness constraint (and vice versa). There is no known (or

expected) language, however, in which this is so. Such a language would map every

phonological input onto the least marked syllable or every semantic input onto the least marked

sentence. The reason such a language does not exist is not because theory predicts it to be

impossible but because it would fail as a communication system. With this in mind, I believe

that the “all default case” language should not exist for the same reasons. There is a possible

ranking that leads to all DPs having default case, but such a ranking would not be useful for

communication and so such a grammar would never arise. There thus seems to be a diachronic

explanation for why certain rankings do not correlate with actual grammars. If a diachronically

sensible explanation can be found, it should not be troubling when certain constraint rankings are

not attested.

7. Conclusions

This paper presents a theory of default case in OT syntax based in part on data from bare

pronoun answers to questions in English and German. I follow Woolford’s (2001a) claim that

non-structural case is controlled by faithfulness constraints, and I propose that structural case is

controlled by markedness constraints. Default case does not fit into either of these patterns, but

rather it is an example of emergence of the unmarked as determined by the ranking of a series of

*CASE constraints. This proposal goes against Woolford’s proposal that these *CASE constraints

exist in a universally fixed ranking.

The constraints proposed here were used to account for default case in English bare

pronoun answers and assigned case in German bare pronoun answers. When the analysis of

English was extend to infinitives, the need for both general and contextual *OVERTSUBJ was

demonstrated. Further support for the use of structural case in bare pronoun answers in German
 28

came from sluicing.

The difference between English and German is directly related to the ranking of case-

assigners relative to the *CASE constraints. More case-assigners dominate the *CASE constraints

in German. This led to the central theory proposed here: case-assigning constraints in a language

must dominate all *CASE constraints. If a given DP cannot receive case from the case-assigners,

it will have default case.

 29

Acknowledgements
I have received many helpful comments from Misha Becker, Randall Hendrick, and Jennifer
Smith. All mistakes remain my own.

References
Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge: MIT Press.
Chomsky, Noam and Howard Lasnik. 1995. In The Minimalist Program: 13-128.
Dixon, R. M. W. 1979. Ergativity. Language 55: 59-138.
Halle, Morris and Alec Marrantz. 1993. Distributed Morphology and the Pieces of Inflections. In The
View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. K. L. Hale and
S. J. Keyser, 111-167. Cambridge: MIT Working Papers in Linguistics.
Hendriks, Petra and Helen de Hoop. 2001. Optimality Theoretic Semantics. Linguistics and Philosophy
24:1-32.
Hwang, Kyu-Hong. 1997. Nominative and Default Case Checking in Minimalist Syntax. Doctoral
Dissertation. University of Washington.
Kang, Seung-Man. 1998. The Syntax and Morphology of Default Case. Doctoral dissertation. University
of Florida.
Legendre, Géraldine, Jane Grimshaw, and Sten Vikner (eds.) 2001. Optimality-Theoretic Syntax.
Cambridge MA: MIT Press.
Legendre, Géraldine. 2001. Introduction to Optimality Theory Syntax. In Optimality-Theoretic Syntax,
Legendre et al. (eds.), 1-27.
Li, Ying-Che. 1971. An Investigation of Case in Chinese Grammar. South Orange, NJ: Seton Hall
University Press.
Lyngfelt, Benjamin. 2000. OT Semantics and Control. Ms. Göteborg University. [ROA-411]
Merchant, Jason. 2001. The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis. Oxford:
Oxford University Press
McCarthy, John J. 2002. A Thematic Guide to Optimality Theory. Cambridge, UK: Cambridge University
Press.
Prince, Alan and Paul Smolensky. 1993. Optimality Theory: Constraint Interaction in Generative
Grammar. Ms, Rutgers University and University of Colorado, Boulder. Rutgers Center for
Cognitive Science technical report 2. [2002: updated version, ROA-537]
Roehrs, Dorian. 2002. Toward Eliminating Case as a Driving Force for Movement: “Licensing” of Case
and Default Case. In New Insights in Germanic Linguistics, III, Irmengard Rauch and Gerald F.
Carr, eds, 221-247. New York, NY: Peter Lang.
Schütze, Carson. 2001. On the Nature of Default Case. Symtax 4: 205-238.
Schütze, Carson and Kenneth Wexler. 1996. Subject Case Licensing and English Root Infinitives. In
Proceedings of the 20th Boston University Conference on Language Development: 670-681.
Wilson, Colin. 2001. Bidirectional Optimization and the Theory of Anaphora. In Optimality-Theoretic
Syntax, Legendre et al. (eds.), 465-508.
Woolford, Ellen. 2001a. Case Patterns. In Optimality-Theoretic Syntax, Legendre et al. (eds.), 509-543.
Woolford, Ellen. 2001b. Burzio’s Generalization, Markedness, and Constraints on Nominative Objects.
Ms. [ROA-457]
Woolford, Ellen. 2006. Lexical Case, Inherent Case, and Argument Structure. Linguistic Inquiry 37: 111-
130.
Wunderlich, Dieter. 2000. Optimal Case in Hindi. Ms: http://user.phil-fak.uniduesseldorf.de/~wdl/
Yip, Po-Ching and Don Rimmington. 1997. Chinese: An Essential Grammar. London: Routledge.