Foreword


StressTyp stands for Stress Typology. StressTyp is a database containing information about word-level stress patterns of (some of) the languages of the world. The database has been implemented in 4th Dimension, a database program for Apple Macintosh. At present 320 languages are represented in StressTyp.

Work on StressTyp began in 1992 as a pilot project of EUROTYP (1990-1994), a project on the typology of European languages, financed by the European Science Foundation (ESF). StressTyp is not intended to be limited to European languages, however, and work on StressTyp has continued after 1994.

We thank the ESF and the Faculty of Arts of Leiden University for financial support for a MacIntosh computer and printer. Through those who work on StressTyp additional financial support comes/came from the Netherlands Foundation of Scientific Research, NWO, the Royal Dutch Academy of Schiences (KNAW), the Holland Institute for Generative Linguistics (HIL), the Departments of General Linguistics and CNWS of Leiden University and the Department of General Linguistics of the Free University of Amsterdam.

A "stripped" version of the database can be found on the web at xxx. For those wishing to use the full version of the database a diskette containing the database (not the 4th dimension program) can be obtained by writing or e-mailing to the authors of this manual. A runtime version of the database (which does not allow editing its desing) is also available. For this runtime version you will need the 4th Dimension runtime programme, which we are allowed to distribute freely.


Harry van der Hulst & Rob Goedemans
(StressTyp)
Holland Institute of Generative Linguistics (HIL)
P.O Box 9515
2300 RA Leiden
The Netherlands

E-mail: Hulst@rullet.leidenuniv.nl
Goedemans@rullet.leidenuniv.nl
We would be grateful for receiving comments on the present form of this manual or the StressTyp project in general, and we welcome any form of collaboration in this area.

Linguistic results based on StressTyp should acknowledge the database.



Chapter 1: Contents of this Manual


This manual describes StressTyp, a database containing information about word-level stress patterns of (some of) the languages of the world. The database has been implemented in 4th Dimension, a database program for Windows 95/NT and Apple Macintosh and currently contains 320 languages.

This Chapter describes the content of this manual.

Chapter 2 provides information about the history and goal of StressTyp.

Chapter 3 specifies the sources that have been used and indicates the (current) limitations of StressTyp.

Chapter 4 gives a detailed description of the record structure. It explains the names of the fields and the values that may appear in each field. This chapter takes the coding system as the point of departure and relates it to the terminological usage of students of stress. It thus serves as a legend on the coding system that we have used.

Chapter 5 approaches the relation between coding system and linguistic terminology from the viewpoint of the latter. Taking linguistic terminology as the point of departure this chapter explains how various types of stress systems are coded in the database.

This manual further contains a number of Appendices. Appendix A (Data Entry Sheet) gives a printed out version of the record. Appendix B contains technical descriptions of the installation procedure, structure of StressTyp and residual problems. Appendix C contains the full questionnaire. Appendix F contains an alphabetic list of the languages that are currently represented in StressTyp.

NB: the original manual contains a Chapter 6, and Appenices D, E and G. In as much as these are applicable to the internet version, they will be added later. Chapter 6 will contain a manual for the application itself.

Chapter 2: History and Goal of the Project


2.1. History

Work on StressTyp began in 1992 as a pilot project of EUROTYP (1990-1994), a project on the typology of European languages, financed by the European Science Foundation (ESF).

EURO TYP consisted of 9 Theme Groups, each studying an aspect of European languages from a comparative and typological point of view. The topic of Theme Group 9 (coordinated by Harry van der Hulst) was Word Prosodic Systems. In the course of the EUROTYP project the question regarding storing primary and secondary language data received special attention and in 1991 it was decided to start two pilot projects, one of which is StressTyp. The idea was to develop an intelligent filing system for basically secondary data (i.e. rules, generalizations, patterns) on word prosodic systems.

The structure of the record was de veloped by Harry van der Hulst (HIL, Leiden), in collaboration with Aditi Lahiri (then at the Max Planck Institute, Nijmegen). The relevant equipment was made available by a grant from the EUROTYP project and further support of the Faculty of Letters of Leiden University. Kees van der Veer (Max Planck Institute, Nijmegen) implemented the record structure in 4th Dimension. After a period of using the database Rob Goedemans made a number of structural modifications. Some technical aspects of the database are described in Appendix B.

In october 1997, work on the Prosody in Indonesian Languages (PIL) project started. StressTyp was integrated in this project to ensure its continuation, and to provide an output platform for the information on the prosodies of Indonesian languages that will be gathered by the project's researchers. Since then StressTyp has been subject to a complete metamorphosis. The whole database has been ported from the old MacIntosh system to a new version of 4th Dimension (6.0.2) which runs under Windows 95/NT. Mac users should not worry, though. With this new version a Mac application of StressTyp can be generated in seconds. While solving some technical problems caused by the difference between MAC OS and Windows NT, Rob Goedemans also gave the database a facelift. The input/browse form looks more modern now, and has been enhanced with input filters, IPA character selection charts and other user-friendly input possibilities.

At this point, in november 1998, the internet version to which this manual belongs, is about to go on the web. The manual is being converted to HTML, and a runtime version of StressTyp is nearly ready.

2.2. Goal

The goal of StressTyp is to offer a quick entry to the primary and secondary literature on stress systems of the languages of the world. By primary literature we mean grammars and articles that provide descriptions of stress patterns, examples and the like. By secondary sources we refer to theoretical works on stress.

We plan to extend the content of the database by systematically trying to add information on language families or linguistic areas that are now underrepresented (see section 4.1).
The goal of the coding system has been to make it possible to search through the database for the occurrence of quite specific properties. With the search facilities of 4th Dimension, StressTyp can be instrumental in testing and developing hypotheses (after the current limitations of StressTyp have been dealt with; cf. section 3.2).

In StressTyp we have tried to encode properties of stress systems without making a commitment to a specific theoretical framework, although it is necessarily the case that a coding system implicitly embodies a theory of stress. We have tried, however, to encode the relevant properties without avoiding redundancy, allowing for duplication of information and fields for informal statements and comments.

Since StressTyp now falls under the umbrella of the Prosody in Indonesian Languages (PIL) project, we will devote extra attention to Indonesian substrate languages in the near future. Indeed, a large protion of the languages that are currently being added are spoken in the Indonesian isles. Since some of these languages have tonal systems, and since the PIL project is phonetic as well as phonological in nature, we plan to expand StressTyp with some extra fields to accomodate broader description, renaming it ProsTyp.

In the more distant future we aim at embedding StressTyp in a network of related databases that provide information on various aspects of stress research, such as an annotated bibliography of stress (StressBib, currently in progress), a syllable structure database (SyllTyp, currently in progress), addresses of linguists who do research on stress (StressRes) and so on. The present global indicator for this imaginary network is StressEx (Stress Expert System).

By making the database available to other researchers we hope to benefit from their knowledge (or personal databases in whatever form) and cooperation in adding more languages to the system, and improving the quality of information presently contained in StressTyp.

To promote the use of StressTyp we have published a collection of articles under the title "Stress Patterns of the World", of which some are based on StressTyp information. In addition, this volume describes the database architecture and presents some direct numerical results and examples of queries.

A second volume of "Stress Patterns of the World" is in preparation. It will contain selected data ordered by language family, and general introductions to stress phenomena in those families.

Chapter 3: StressTyp Sources and Limitations


3.1. Sources

Although the work on StressTyp started in 1992, StressTyp is still in its first phase due to the fact that we have had only limited means to develop it more rapidly.

The first phase involves extracting (often elementary) information about stress systems from (cross)linguistic studies. Both strictly typological and theoretical so urces have been consulted.

The collection of material was started by Aglaia Cornelisse. She did not enter material into the database but collected the data on a proto version of the Data Sheet in Appendix A. The first sources consisted of the 'usual typological studies' such as Hyman (1977), Greenberg and Kashube (1978), Hayes (1981/95), Lockwood (1982), Halle and Vergnaud (1987) and so on, i.e. studies that provide the empirical basis for much of the present-day theoretical discussion. Later, Simone Langeweg assisted in adding data from more sources. The collected data were transferred into the database format (that had meanwhile been implemented) by Rob Goedemans and Ellis Visch. They then started checking the records for consistency and correctness and also added new languages. Checking is done by consulting the original sources, and often additional theoretical or descriptive studies. In this phase Ruben van de Vijver provided valuable assistance.

We plan to further check the data that are now contained in StressTyp and to add new information, both of languages already included and of new languages.

We would be very grateful to receive references to books or articles that contain useful information on stress systems, especially of languages that are not yet contained in the database (cf. Appendix F).

It is inevitable that others have developed data bases on stress systems (perhaps with less detailed record structures) or that information on stress systems is part of less specialized databases. We would like to be made aware of such systems and, more importantly, of the availability of the information contained in them. We are interested in collaborating with others if such systems are still accessible.

3.2. Limitations

At this point we intend to store information in the database on as many languages as possible. This means that the languages included now do not form a well defined sample by any criterion. Some language families or linguistic areas are clearly overrepresented (e.g. Australian languages) or underrepresented (European languages).

The data that StressTyp contains are as trustworthy as the information we found in the sources. If that information is wrong, StressTyp has copied that wrong information. (Of course, whenever we had any reason to believe that the information was wrong we did not copy it.) We have tried to trace the information back to the original descriptive source wherever this was possible.

Specifying values in database fields necessitates interpreting sometimes very limited information. This means that the information in StressTyp is very often rather incomplete.

The information stored for each language ranges from very elementary statements (like "initial stress", all further fields unspecified) to fairly detailed specifications for a number of fields. The record allows information on morphological structure, but this is mostly unspecified.

Misinterpretations on our part are also possible. The coding system requires interpretation of the sources. Records are not always faithful to any particular source. Where we have consulted more than one source for one language an attempt has been made to reconcile the sources. In doing so we may have come up with a coding that does not correspond to an actually existing dialect or language variety. Another factor that may have attributed to inconsistencies is that various people have been involved in the coding.

Every record specifies the sources on which we have based the coding (cf. Chapter 4).

Despite its limitations, our own experience is that StressTyp can be helpful in developing and testing hypotheses by offering data and properties of different languages in an identical format.

We emphasize that StressTyp cannot be held responsible for providing incorrect, or incomplete information. We encourage those who use StressTyp in publications to check crucial information in the original descriptive sources or with native speakers. We welcome all corrections and additions both regarding specific languages and the overall organization.


Chapter 4: Description of the Record Fields


4.0 Introduction/general remarks

In this chapter, all the record fields of the stress database are described in detail. Example sheets of the actual pages in the computerized version of the database are given at the end of this manual in Appendix A. The underlined words in capitals refer to the field names on these pages. All descriptions of the record fields are accompanied by sample languages from the database to provide as much clarification to the user as possible.

StressTyp is based on a theoretical perspective that competes with other metrical theories. Its starting point is the separation of main stress and rhythm (cf. "Stress Patterns" chapter 1). The separation of information on main stress and rhythmical structure is not commonly assumed, but we must emphasize that it is possible to interpret the information without making a commitment to this particular theory. The parameters used in this database embody a particular view on stress, but they are presented as purely 'descriptive'.

It is unavoidable that the translation of properties of certain systems (as described in our sources) into the format of the database is not always entirely straightforward, because they have special or sometimes even conflicting features. Often, there are different possible ways of storing properties of systems. In those cases, choices have been made to fit the description into the format of the database. In chapter 5, we clarify our choices. It exemplifies all types of stress systems identified in the theoretical literature with example languages. Cross-references to corresponding sections in chapter 5 and this chapter are provided.

Some general remarks concerning the notation used in the database have to be made. In a large subset of record fields, simple abbreviations are used to express information, such as Y(es) or N(o). In some cases, the interpretation of an abbreviation is dependent on the contents of a specific record field. For example, L can stand for left, expressing for instance at which edge footing starts, or for last in which case a specific syllable or position in the word is singled out. This practice is not unusual in metrical phonology either. The StressTyp user must be aware of these ambiguities.

Since this chapter follows the structure of StressTyp rather closely, some information on databases might be in order. A database contains records (this chapter basically describes the contents of a StressTyp record). These are packages of information relating to one item. In our case the records are related to languages. One record contains all the information we were able to find for a certain language. The information in a record is organised into fields. A field contains one piece of information or a set of related pieces. In our database, for instance, we have fields for language name, affiliation, examples, stress type etc. (see below).

In general, all abbreviations (indicating a field value or parameter setting) are exemplified in this chapter by a reference to one or more example language(s). Sometimes, no examples of an abbreviation occur in the database, which we indicate by "no example in StressTyp". This could be merely a coincidence or a systematic gap. No claims are made with respect to the relative markedness of a parameter setting.

Finally, if a field is completely irrelevant to a particular stress system, this is indicated by a hyphen. If information is not available from the theoretical or descriptive literature, this may be indicated by a blank or question mark. In general, blanks indicate that no effort has been made to uncover the information elsewhere, for reasons of time. A question mark indicates that the relevance of the record field is uncertain or that there is doubt about the interpretation of the information in the source. Sometimes a field does contain information but it is still accompanied by a question mark. In those cases, the information may be correct, but conclusive evidence is not available.

4.1 Language information

LANGUAGE
The name of the l anguage or dialect in the naming-conventions of Voegelin & Voegelin (1977). If the language is referred to by other names in the literature, these are added to prevent double records in the database

DIALECT OF
The name of the mother language of which the language specified in the record field language is a dialect

PATH
The family tree, root first (following Voegelin & Voegelin 1977)

AREA
All geographical a reas in which the language, or dialect of the language in question, is spoken

Example language:

Language: Bidyara-Gungabula
Dialect of: Mandandanji

Path Australian - Pama-Nyungan - Pama-Maric - Mari
Area South Queensland (Australia)

4.2 STRESS TYPE

4.2.1. List of items/abbreviations and connectives

The record field STRESS TYPE indicates the main stress type of the language by means of a code, identifying the position(s) of main stress. It can either be a simple abbreviation from a list of items, or a combination of abbreviations and one (or more) connective(s).

It is general practice to describe so called unbounded systems by statements like "last heavy, or first". "Last heavy" could identify any syllable in the word, even the first one if this is the only heavy syllable present. The first part of this statement, therefore, singles out only specific syllables, i.e. heavy ones, and identifies which one of these syllables is promoted. It identifies a specific syllable in a domain, rather than a specific position. The second part of the statement, however, identifies a specific position, in this case the first syllable in a domain.

The same practice is used in this database. An abbreviation not followed by a connective identifies a specific position in a word; when followed by a connective its interpretation slightly changes and identifies a specific syllable within a domain. The following combinations occur (x and y are variables representing a set of values that are actually used in the coding):

x = position x
x/ = x if heavy
x; = x and
% x = shift to x
x- = x is special case
(x) = additional information x to clarify and extend a preceding code

Thus, if it is followed by a connective, x is never the only stress position . Moreover, a code can contain more than one connective. The following complex codes can be identified in StressTyp (to be exemplified later):

x/y = x if heavy, otherwise y
x;y = x and y both occur in free variation, but x is dominant
x/y;z = x if heavy, otherwise y; in free variation with z, but x/ y is dominant
x%y = x if heavy, otherwise shift to y
x%y/z = x if heavy, otherwise shift to y if y is heavy, otherwise z
x-y/z = x is special case; if there is no x in domain, y if heavy, otherwise z

In general, the first abbreviation in a code identifies the special case, the final one the default case.

The following items can occur in a code. For convenience, we have indicated the specific interpretation of the codes when followed by the connective /, which is the connective that is used most often in StressTyp.


UNBOUNDED SYSTEMS

L = last L/ = last heavy
F = first F/ = first heavy


BOUNDED SYSTEMS

I = initial I/ = initial heavy
S = second S/ = second heavy
T = third T/ = third heavy
U = ultimate (final) U/ = ultimate heavy
P = penultimate P/ = penultimate heavy
A = antepenultimate A/ = antepenultimate heavy

MISCELLANEOUS

NMS = no main stress, i.e. all stresses are perceived as equally prominent
CN T = main stress depends on the counting of the syllables in a word
LEX = main stress is lexical
IRR = main stress is irregular
Pitch = main stress interacts with pitch-accent assignment
Tone = main stress interacts with tone-assignment

The code is deduced from information stored in record fields in the formal analysis of section 4.3.2. below. For many substitutions of x and y by the values above there are no actual examples in the database. An interesting question is whether this is merely a coincidence or a systematic gap.

4.2.2. Guidelines and exemplification of codes

4.2.2.1. Single abbreviations

Single abbreviations are used to identify quantity-insensitive systems. In systems where main stress is always found on a fixed syllable with reference to one of the edges of words, irrespective of the form of that syllable. Mostly such systems are described as bounded, but we include here systems that have been described, or could be described, as unbounded (see also chapter 5, section 5.1.2. for more details). The symbol identifies a syllable in a specific position in the stress domain.

I: main stress is on the first syllable
cf. Maranunggu

S: main stress is on the second syllable
cf. Araucanian

T: main stress is on the third syllable
cf. no example in StressTyp

U: main stress is on the ultimate or final syllable
cf. Weri

P: main stress is on the penultimate syllable
cf. Warao

A: main stress is on the antepenultimate syllable
cf. Macedonian

4.2.2.2. Usage of the connective "/"

The slash is used to identify systems in which the position of main stress is influenced by weight, i.e. for quantity-sensitive systems (see also section 4.3.2. stress weight and 4.4. heavy for stress). Their code contains two symbols x and y, separated by the connective slash: x / y. The first symbol 'x' identifies the position of main stress if the stress domain contains heavy syllables. The second symbol 'y' identifies the default case, i.e. only light syllables are present in the stress domain.
Systems that are identified with more than two symbols are special and other connectives are used. They will be exemplified below.

QUANTITY-SENSITIVE UNBOUNDED systems are divided into the four (familiar) types using L and/or F and the connective /. The stress domain comprises the entire word (cf. chapter 5, section 5.1.2.):

L/L: main stress on last heavy syllable in domain;
if no heavy syllables (i.e. only light syllables in domain) on last syllable
cf. Golin

F/Fmain stress on first heavy syllable in domain;
if no heavy syllables (i.e. only light syllables in domain) on first syllable
cf. Juwalarai

L/Fmain stress on last heavy syllable in domain;
if no heavy syllables (i.e. only light syllables in domain) on first syllable
cf. Chuvash
F/Lmain stress on first heavy syllable in domain;
if no heavy syllables (i.e. only light syllables in domain) on last syllable
cf. Goroa

In QUANTITY-SENSITIVE BOUNDED systems (see also chapter 5, section 5.2.), the stress domain is limited to two syllables and located at or nearby one of the two edges of a word, which is called a window. Four situations can occur, where square brackets indicate the edges of the stress domain, on either the left or the right side of the word:

(1) a. [h h] c. [h l]
b. [l l] d. [l h]

The code that StressTyp gives specifies only what happens in case (1a) and (1b). The symbol 'x' before the slash identifies the position of main stress when both syllables are heavy, i.e. (1a). The symbol 'y' after the slash identifies the position of main stress when both syllables are light, i.e. (1b).

When the two syllables in the stress domain are unequal in weight, cf. cases (1c) and (1d), it is always the heavy syllable which bears main stress. This information is considered as redundant; therefore, (1c) and (1d) are left out of consideration in determining the code.

In the code, we use the following terminology:

(2) # [I S] [P U] #
| | | |
first lastfirst last

It is our intention to eventually replace all abbreviations by 'last' and 'first' (following more closely, van der Hulst, 1996). As a consequence of this decision, domains, in which the syllables are unequal in weight will be covered implicitly by the code too. Consider as an illustration an unbounded system like (i) and a bounded system like (ii):

(i) a.     * b.  * c.    * d. *
[l l h l h l] [h l l l] [l h l l] [l l l l]

(ii) a.   * b.   * c.     * d.   *
#[h h] #[h l] #[l h] #[l l]

Both systems can be described as 'first/first'-systems: if there is more than one heavy syllable, the first heavy one catches the accent; if there are no heavy syllables the first syllable catches the accent. However, if there is only one heavy syllable that syllable receives the accent, independently of its position in the word. Since there is only one heavy syllable it can be captured by the statement: first heavy in domain. With respect to bounded systems, the information at which edge of the word the two-syllable domain is located is not expressed in the stress type code any more. However, since this information is stored elsewhere too, i.e. in record field stress domain, this is not seen as a loss.
Four main stress codes will be mentioned here that have a stress domain that is aligned to the left edge of the word. The stress type contains a choice from the symbols I (initial) and S (second), identifying 'first' and 'last' in the domain and the connective /. Schematized:

(3) # [I S]
| |
first last


I/I main stress is on first heavy syllable in domain;
if no heavy syllables, stress is on the first syllable
   *    *
  #[h h] #[l l] cf. Southeastern Tepehuan

I/S main stress is on first heavy syllable in domain;
if no heavy syllables, stress is on the last syllable
   *      *
  #[h h] #[l l] cf. Hopi

S/I main stress is on last heavy syllable in domain;
if no heavy syllables, stress is on the first syllable
     *    *
  #[h h] #[l l] cf. Capanahua

S/S main stress is on last heavy syllable in domain;
if no heavy syllables, stress is on the last syllable
     *      *
  #[h h] #[l l] cf. Archi

The following stress codes can be identified when the stress domain is aligned to the right edge of the word. The stress code contains a choice from the symbols P and U, identifying 'first' and 'last' in domain, and the connective /. Schematized:

(4) # [P U]
| |
first last

U/U main stress is on last heavy syllable in domain;
if no heavy syllables, stress is on the last syllable
   *    *
  [h h]# [l l]# cf. Yapese

U/P main stress is on last heavy syllable in domain;
if no heavy syllables, stress is on the first syllable
   *  *
  [h h]# [l l]# cf. Tol

P/U main stress is on first heavy syllable in domain;
if no heavy syllables, stress is on the last syllable
 *    *
  [h h]# [l l]# cf. Aklan

P/P main stress is on first heavy syllable in domain;
if no heavy syllables, stress is on the first syllable
 *  *
  [h h]# [l l]# cf. Awadhi

More complex are systems in which extrametricality interferes. The two-syllable stress domain is shifted one syllable from the edge. These systems are identified by the use of the symbols T and A in their code. Schematized:

(5) [S T] [A P]σ #
| | | |
first lastfirst last

As an example, consider:
P/A stress domain comprises the antepenultimate
and penultimate syllable
main stress is on last heavy syllable in domain;
if no heavy syllables, stress is on the first syllable
    *  *
  [h h]σ# [l l]σ# cf. Latin


4.2.2.3. Usage of the connective ";"

The semicolon is used in those cases where there is a certain degree of unpredictable free variation in the position of main stress, although restricted to some specific positions. Therefore, it is used typically for bounded systems. In general, there is one dominant and regular pattern, but there are exceptions, found at (a) specific position(s). Therefore, main stress is not to be considered irregular. For example:

P;A main stress is penultimate in most cases; in a few cases, stress occurs on the antepenultimate syllable. These antepenultimate cases have to be specified lexically.
cf. Tacana

See also, section 4.2.2.6. for the contrast with the use of IRR.

The semicolon can also be preceded by more specific information, eg:

P/U;P main stress is penultimate if heavy, otherwise main stress is ultimate. Exceptions to this dominant pattern are always found at the penultimate syllable. Therefore, it is not an irregular system.
cf. Aklan

4.2.2.4. Usage of the connective "-"

The hyphen occurs in the code for systems in which there is a not a simple contrast in syllable-quantity or syllable-prominence, but syllable types must be represented on a scale. (e.g. many languages make a three-way contrast between light, heavy and superheavy syllables). As a consequence of having more than two weight distinctions, main stress can be found on more than two different positions. The special case (i.e. 's uperheavy') is indicated in front of the stress code, followed by the hyphen. After the hyphen, the code contains two symbols and the connective / which covers the predictable variation of stress within a two-syllable domain (cf. chapter 5, section 5.3.2). For example:

U-P/A main stress is on the final syllable if it is 'superheavy'; if such a syllable is absent, stress is on the last heavy syllable in domain, i.e. the penultimate syllable; if no heavy syllables in domain, main stress is on the first (=antepenultimate) syllable.
cf. Damascene Arabic
L-P/A main stress is on the last 'superheavy' syllable in the entire word; if such a syllable is absent, main stress is on the last heavy syllable in a two-syllable domain located one syllable away from the right edge of the word; if no heavy syllables are present in the two-syllable domain stress is on the first syllable in that domain.
cf. Klamath
 

4.2.2.5. Usage of the connective "%"

The percentage mark is used to identify what we call 'broken-window' systems (see also chapter 5, section 5.3.1.). These are quantity-sensitive systems that show a shift outside the two-syllable window of main stress in case both syllables inside the window are light. The syllable position to which main stress has been moved is identified and preceded by the percentage mark:

P%A/P    main stress is on first heavy (=penultimate) syllable in (right edge) domain; otherwise, i.e. if (l l), there is a shift to the antepenultimate 
                syllable, under the condition that this syllable is heavy; otherwise, i.e. the antepenultimate syllable is not heavy, main stress is on first
               
syllable in domain (=penultimate).

 *           *       *<--

[h h]#   l [l l]#   h [l l]#               cf. Maithili

 

U%A   main stress is on last heavy syllable in domain; otherwise, i.e. if two light syllables, main stress is shifted to the antepenultimate position, irrespective of the make-up of this syllable

   *      *<--      *<--

            [h h]#   l [l l]#   h [l l]#              cf. Hindi

4.2.2.6. Other categories

4.2.2.6.1. NMS

NMS   there is no distinction between primary and non-primary stresses, i.e. they are all perceived as equally prominent according to a descriptive source.

cf. Central Siberian Yupik

X (NMS)

there is no distinction between primary and non-primary stresses, but X (i.e. a specific position is always stressed.

cf. Gidabal I (NMS): stress occurs on the first syllable, regardless of weight and on

every syllable which surfaces with a long vowel

 

If NMS is used, without any further specification, the record fields on main stress contain hyphens, indicating that these fields are irrelevant. The relevant information is stored in the record fields on rhythm.

However, in X(NMS) cases, the record fields of the formal analysis are filled as if there were a distinction between main stress and rhythm. We suspect that in most systems requiring NMS a distinction between main stress and secondary stress can nevertheless be made. The fact that a certain syllable is always stressed may point in this direction.

 

4.2.2.6.2. CNT

CNT is always used between parentheses and preceded by either L or F. It is used to identify what we call 'count systems' (cf. Van der Hulst 1992). In these systems the position of main stress depends on counting syllables in terms of feet, which, in most cases, leads to a difference in the position of main stress between odd-numbered words and even-numbered words. These systems are analyzed as unbounded systems because the whole word is relevant for the computation of main stress. Therefore, we use L or F. See also chapter 5, section 5.1.3.

L(CNT) last foot in word carries main stress, i.e. stress is found at the right edge of the word

cf. Cayuga

F(CNT) first foot in word carries main stress, i.e. stress is found at the left edge of the word

cf. MalakMalak

Notice, that last/first must not be interpreted as last/first assigned foot. It is always the last assigned foot which carries main stress. Last/first identifies at which edge main stress is located, as in the unbounded systems described in section 4.2.2.2.

 

4.2.2.6.3. LEX

LEX is used to identify systems that use lexical accent marking. This can mean that main stress is unpredictable and can be found on any syllable in the word, although each word has a predictable stress position once the location of the lexical accents is given. These positions are diacritically marked for each word. Most typically, additional information must be specified, LEX is added in those cases, sometimes between parentheses.

LEX    main stress is completely unpredictable and can be found anywhere in the word. Main stress has to be specified in the lexicon by means of a diacritic accent or marker.

cf. Spokane

F/F (LEX)

main stress is on the first syllable that has a diacritic marker; otherwise, i.e. if there are no diacritic accents, main stress is found on the first syllable

cf. Russian

 

L/LEX

main stress is on the last heavy syllable; if there are no heavy syllables, i.e. only light syllables, main stress is unpredictable and can be found anywhere in the word. In those cases, stress has to be specified in the lexicon by means of a diacritic accent.

cf. Mountain Cheremis

 

4.2.2.6.4. IRR

IRR is used if main stress varies unpredictably in a bisyllabic domain in a word. They are bounded. Often, there is one (or more) identifiable pattern(s). Exceptions, however, cannot be captured by specifying a specific position. Therefore, the use of IRR is additional information, often added between parentheses or preceded by the connective semicolon (indicating free variation, see section 4.2.2.3).

Consider as an example Sangil, in which the dominant pattern can be described by P/P. There are exceptions to this pattern in both ways, i.e. a final light or heavy syllable can receive main stress, or, if both syllables are unequal in weight, the light syllable can catch main stress instead of the heavy one. Its code will be P/P;IRR. This code expresses the fact that there are exceptions which cannot be specified more precisely.

Now consider Dutch. Part of the Dutch stress system is that words ending in three light syllables have either stress on the antepenultimate syllable or on the penultimate syllable, the latter being dominant: pįgina / pyjįma. Stress on the final syllable is truly exceptional. Although the stress system of Dutch is more complicated, this part of the system could be captured by the code: P;A (IRR). The use of the semicolon says that there is variation between P and A, with P as dominant pattern. The addition of IRR implies that the second position is regarded as exceptional.

We are aware of the somewhat vague use of IRR in general, although its interpretation in specific cases will be clear.

 

4.2.2.6.5. Pitch and tone

Pitch and Tone are used incidentally and sometimes between parentheses to indicate that there is an interaction between pitch or tone assignment and metrical structure. 'Pitch' or 'tone' are added if these terms are used as such in the record field quote which is based upon a theoretical or descriptive source. Specifics about interaction between pitch/tone and metrical structure assignment are provided in the record fields stress repair and remarks.

stress type: pitch                 stress type: L/- (CNT: tone)

cf. Chocktaw                            cf. Seneca

 

4.2.3 Stress type information continued

SOURCE QUOTATION 

A quotation that describes the stress pattern of the language in words (not necessarily literally quoted from the literature) with reference (author, year and page numbers).

EXAMPLES

Examples that illustrate the various stress patterns in the language. Examples are glossed and represented with IPA phonetic symbols in the LaserIPA TM version (see Appendix E). Sounds for which this IPA set gives no symbol may be represented by any standard typographic symbol not otherwise used in the description.

DESCRIPTIVE SOURCE 

Complete reference to authors that have done field work on the language. If a reference is preceded by an asterisk, the source has been consulted by us.

THEORETICAL SOURCE 

Reference to authors that have analyzed the stress pattern in a certain framework. Only author and year are given. Complete references are provided in a library file (StressBib). If no analysis exists but a name is given anyhow, address and E-mail address of the person who is responsible for the information of the language entry can be found in the address file (StressAdr). The information provided by this person is translated by us into the database framework. If no names are given at all, StressTyp is responsible for incorporating the information on the basis of a descriptive source.

4.3 Formal analysis

4.3.1 General remarks

In this database we have made a distinction between main stress and rhythm.

4.3.2 Main stress

The (main) stress domain contains exactly two syllables, irrespective of their weight in case of a bounded system, and it comprises the entire word in case of an unbounded system. If extrametricality occurs, the stress domain is shifted one syllable inside the word.

There are seven record fields for information on main stress. On the basis of these seven fields on main stress, the stress type code of the language is determined (see section 4.2. above).

STRESS DOMAIN

L          domain is bounded and located at the left (L) edge of the word

cf. Djirbal, Capanahua

R         domain is bounded and located at the right (R) edge of the word

cf. Polish, Yapese

U         domain is unbounded (U), i.e. the entire word (modulo extrametricality)

cf. Chuvash, Khalka Mongolian

-           domain is irrelevant, i.e. there is no distinction between prima­ry/non-pri­mary

stresses

cf. Central Siberian Yupik (but see also Gidabal: I (NMS), cf. section 4.2.2.6.1)

STRESS EXTRAMETRICALITY

L          extrametricality occurs on the left (L) edge of the word

cf. Iliaura/Alyawara, Western Aranda

R         extrametricality occurs on the right (R) edge of the word

cf. Macedonian, Chamorro

N         extrametricality is not (N) involved

                        cf. Yapese, Capanahua

 

STRESS EXTRAMETRICAL UNIT

seg      segment is extrametrical

cf. Alyawara (vowel)

mora    mora is extrametrical

cf. Hindi

syll       syllable is extrametrical

cf. Chamorro, Macedonian

foot      foot is extrametrical

cf. Unami

 

STRESS WEIGHT

Y          the position of main stress is determined on the basis of syllable-quantity, syllable-prominence, rhythmically strong positions or diacritically marked syllables, i.e. weight is Yes (Y)

cf. Capanahua (syllable weight): quantity-sensitive system

cf. Pirahć (syllable-prominence): prominence system

cf. MalakMalak (rhythmically strong positions): count system

cf. Russian (diacritically marked syllables): lexical accent system

N         the position of main stress is determined without reference to syllable-quantity,

syllable-prominence, or rhythmically strong positions, i.e. main stress

assignment is quantity-insensitive: weight is No (N)      

cf. Garawa, Cavinena

Specific information about what counts as heavy in a system is stored in the record field heavy for stress.

 

STRESS HEAVY SYLLABLE

L          stress is located on the leftmost (L) heavy syllable in domain

cf. Southeastern Tepehuan, Aklan (bounded)

cf. Khalka Mongolian, Komi-Permyak (unbounded)

R         stress is located on the rightmost (R) heavy syllable in domain  

cf. Capanahua, Jicaque (bounded)

cf. Chuvash, Golin (unbounded)

-           system is quantity-insensitive

cf. Araucanian, Garawa

In a bounded quantity-sensitive system this field specifies the position of stress in a domain with two heavy syllables, i.e. [h h]. If the two syllables in the stress domain are unequal in weight, i.e. [h l] or [l h], it is always the heavy syllable which bears stress.

 

STRESS if no HEAVY SYLLABLE

tr         in two syllable-domain, stress is on lefthand syllable, i.e. trochaic: [* .]

cf. Garawa, Piro (QI-bounded)

cf. Capanahua, Jicaque (QS-bounded)         

ia         in two syllable-domain, stress is on righthand syllable, i.e. iambic:   [. *]

cf. Araucanian, Weri (QI-bounded)

cf. Hopi, Yapese (QS-bounded

L          in unbounded domain, stress is on first syllable, i.e. located at left (L) edge of word

cf. Khalka Mongolian, Chuvash

R         in unbounded domain, stress is on last syllable, i.e. located at right (R) edge of word

cf. Golin, Komi-Permyak

 

For a bounded quantity-insensitive system, this is the only field that provides information about the position of main stress. For a bounded quantity-sensitive system, this field specifies the position of stress in a domain with two light syllables, i.e. [l l]

 

STRESS REPAIR

Y         there is a shift outside the two-syllable stress window, if both syllables inside the window are light

cf. Manam, Maithili

N         no repair, stress falls always inside the two-syllable window

cf. Maranunggu

In chapter 5, section 5.3., we refer to systems in which there is a shift outside the two-syllable window, as 'broken-window' systems.  The stress code of this type of system contains the connective %. Information about the specific conditions under which (main) stress shift occurs, can be found in the record field Repair.

 

4.3.3 Rhythm

There are nine fields providing information about non‑primary stress, headed under RHYTHM.  The rhythm domain consists of all syllables of a word, i.e. including the syllables of the stress domain. Rhythmic feet are assigned in such a way that the syllable that is stressed in the stress domain is also stressed in the rhythm domain (see also chapter 5, section 5.2.1.). In other words, rhythm is sensitive to the presence of main stress, rather than the other way around, as is mostly assumed in the metrical literature.

 

RHYTHM

Y          there is a distinction between main stress and non-primary stress (Yes)

cf. Garawa, Maranunggu, Seri

N         only main stress is present, i.e. there are no non-primary stresses (No)

                        cf. Chamorro

RHYTHM DIRECTION

L          rhythm, i.e. non-primary stress assignment, starts at the left (L) edge of the word

cf. Maranunggu

R         rhythm, i.e. non-primary stress asignment, starts at the right (R) edge of the word

cf. Garawa, Warao

E         rhythm is bi-directional and echoes away from both edges (E) ("edge-in") of the word

cf. Auca

C         rhythm is bi-directional and echoes away from main stress in both directions towards the edges (C=Centrifugal)

cf. no example in StressTyp

RHYTHM EXTRAMETRICALITY

L         extrametricality occurs on the left (L) edge of the word

cf. Iliaura, Western Aranda

R         extrametricality occurs on the right (R) edge of the word

cf. Macedonian, Chamorro

N         extrametricality is not (N) involved

cf. Yapese, Capanahua

 

RHYTHM EXTRAMETRICAL UNIT

seg      segment is extrametrical

cf. Western Aranda

mora    mora is extrametrical

cf. Hindi

syll      syllable is extrametrical

cf. Palestinian Arabic

 

RHYTHM WEIGHT

Y          the position of non-primary stress is determined on the basis of syllable-quantity, i.e. weight is Yes (Y)

cf.  Palestinian Arabic, Aklan, Seri

N         the position of non-primary stress is determined without reference to syllable

-quantity, i.e. the system is quantity-insensitive, weight is No (N)      

cf. Weri, Capanahua

Specific information about what counts as heavy in the system is stored in the record field heavy for rhythm.

RHYTHM TYPE

tr         the general rhythmic pattern is trochaic, i.e. regular alternation of a stressed and an unstressed syllable, i.e. a sequence of (* .)

cf. Maranunggu, Garawa

ia         the general rhythmic pattern is iambic, i.e. regular alternation of an unstressed and a stressed syllable, i.e. a sequence of (. *)

cf. Aklan, Cayuga

bo        both (bo) "iambs" and "trochees" are found

cf. no example in StressTyp

-           no regular alternation of non-primary stresses

cf. Gidabal in which only long vowels are stressed


The regular alternating pattern can be disturbed by weight. In some quantity-sensitive cases, it is crucial that trochaic is interpreted as a moraic trochee, i.e. either a sequence of two light syllables ( l l) or one heavy syllable (h) (cf. Hayes 1995); this is specified in the record field remarks, cf. Cahuila, Cairene Arabic. In all other cases, the trochee may be taken to be unbalanced (i.e. just avoiding (lh) and (hh)).

RHYTHM REPAIR

Y          the rhythmic surface patterning deviates in some cases from the one specified by the fields Rhythm Type

            and/or Rhythm Ternary and must be repaired.

cf. Chugach

N         the rhythmic pattern does not need repair.

cf. Maranunggu

This field is specified as Yes, for instance in cases in which rhythm is both ternary and quantity-sensitive; in this case, a situation could arise where syllables are skipped or left over in parsing. In some cases these unparsed syllables are supplied with structure or incorporated. Specific information about the conditions of repair is stored in the record field Repair (see also chapter 5, section 5.4).

RHYTHM ITERATIVE

 

Y            there is more than one non‑primary stress; rhythm is assigned iteratively (Yes)

cf. Maranunggu, Garawa

N            there is only one non-primary stress; rhythm is assigned non-iteratively

(No).

cf. Jabulajabula, Nengone

 

RHYTHM TERNARY

N         the rhythmic pattern is binary, not ternary (No)

cf. Aklan, Maranunggu, Capanahua

ia         the rhythmic pattern is ternary; the rhythmic foot comprises a complex head

of two syllables and a non-head of one syllable. the relation between the two

syllables within the complex head is iambic; the relation between complex head

and non-head is specified in the field Rhythm Type

cf.                                       Rhythm Type             Rhythm Ternary

Chugach                             ([. *]. )  trochaic           iambic    (=amphibrachus)

No ex. in StressTyp            (.[.  *])  iambic              iambic   (=anapest)

tr          the rhythmic pattern is ternary; the rhythmic foot comprises a complex head

of two syllables and a non-head of one syllable. the relation between the two

syllables within the complex head is trochaic; the relation between complex

head and non-head is specified in the field Rhythm Type

cf.                                              Rhythm Type            Rhythm Ternary

Cayuvava                        ( [* .] .)  trochaic                        trochaic (=dactylus)

Sentani                           (.[*  .] )  iambic              trochaic   (=amphibrachus)

 

                    For details, see chapter 5, section 5.4.

 

4.3.4 FORMAL ANALYSIS continued: degenerate feet and subminimal words

DEGENERATE FEET 

Y          there are incomplete feet (i.e. monosyllabic or monomoraic) to express

non-primary stresses at left-over syllables at the edge of a word

cf. Maranunggu

N        there are no incomplete feet (i.e. monosyllabic or monomoraic), i.e. there are

           no non-primary stresses at left-over syllables at the edge of a word

cf. Warao

Since in this stress database, stress and rhythm are separated, degenerate feet =Yes covers rhythmic properties only. This field does not provide any information about the occurrence of monosyllabic words. This is specified in a separate field Subminimal Words. The settings of these two fields are determined independently.

SUBMINIMAL WORDS

Y          there are subminimal words

cf. Araucanian, Maranunggu

 

N         there are no subminimal words

cf. Cayuga, Maithili

In quantity-sensitive systems, subminimal words are monomoraic, i.e. light syllables; in quantity‑insensitive systems subminimal words are monosyllables, irrespective of their content. Specific information about monosylla­bles, i.e. whether they are stressed or not, can be found somewhere else in the record, most likely in source quotation or remarks.

 

4.4 Basic syllable types 

TEMPLATE                          

Specification of the types of syllables that occur in the system, e.g. CV, CVC etc.

OBLIGATORY ONSETS    

Y            onsets are obligatory

N            onsets are not obligatory

BRANCHING ONSETS    

Y              branching onsets are allowed

N            branching onsets are not allowed

LONG VOWELS                  

Y            long vowels are present underlyingly

N            long vowels are not present underlyingly

CLOSED SYLLABLES

Y            there are closed syllables

N            there are no closed syllables

GEMINATES

Y            there are geminates

N             there are no geminates 

CODA RESTRICTIONS

Specification of coda restriction if any. If not: -

 

HEAVY FOR STRESS

Indication of what counts as heavy with respect to main stress in the

system: syllable or segment types (possibly in the form of a scale) (i.e. syllable

-quantity or syllable-prominence), diacritically marked syllables or rhythmically strong

positions (see also chapter 5, section 5.1.3. and 5.3)

HEAVY FOR RHYTHM

Indication of what counts as heavy with respect to non-primary stress(es) in the system: syllable or segment types that are heavy for rhythm, i.e. syllable- quantity

 

4.5 Repair strategies

REPAIR

This field specifies under which conditions Stress Repair and/or Rhythm Repair occur. With respect to so-called 'broken-window' systems, containing the connective % in their Stress Type and with Stress Repair set on Y, this fields specifies whether shift outside the two-syllable stress window, containing (l l), takes place always or only if the target syllable is heavy, see the use of the connective % (cf. section 4.2.2.5. and, chapter 5, section 5.3.1.).

 With respect to so-called 'ternary' systems, for which Rhythm Repair is set on Y, this field specifies under which circumstances unparsed syllables are grouped into a foot after all or under which circumstances reparsing occurs (see also chapter 5, section 5.4.).

             Finally, Repair can contain information about destressing of syllables.

 

4.6 Exceptional patterns

EXCEPTIONAL PATTERN

Description of the exceptional stress patterns that have to be specified in the lexicon.

EXAMPLE PATTERNS

Examples of exceptional stress patterns. Examples are glossed and are represented with IPA phonetic symbols in the LaserIPA TM version (see Appendix E). Sounds for which this IPA set gives no symbol may be represented by any standard typographic symbol not otherwise used in the description.

EXCEPTION SOURCE

            Reference of source of exceptions (author + year+ pagenumber(s)).

 

4.7 Processes

PROCESSES

Description of (phonological) processes that occur in the language (especially the ones related to stress).

EXAMPLE PROCESSES

Examples of the processes stated above. Examples are glossed and are represented with IPA phonetic symbols in the LaserIPA TM version (see Appendix E). Sounds for which this IPA set gives no symbol may be represented by any standard typographic symbol not otherwise used in the description.

 

4.8 Remarks

REMARKS

This field can contain any (additional) information relevant with respect to the system, especially if certain choices have been made to fit a more or less problematic system into the format of the database.

It contains information about the (un)stressedness of monosyllables if subminimal words occur in the language. 

It can specify the relation between metrical structure and tone or pitch assignment if the Stress Type contains an indication for Tone or Pitch.

And, finally, it states explicitly that all stresses are perceived as equally prominent if the Stress Type contains NMS between parentheses.

 

4.9 Morphology and stress

In these entries information about the role of morphology is stored, i.e. the behaviour of prefixes and suffixes, clitics, the stress patterns of compounds, etcetera. This part of the stress database is in a very preliminary state. Record fields are incorporated but only in a few cases information has been stored.