jackendoff toward a formal theory tonal music

Upload: idschun

Post on 06-Jul-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    1/62

    Yale University Department of Music

    Toward a Formal Theory of Tonal MusicAuthor(s): Fred Lerdahl and Ray JackendoffSource: Journal of Music Theory, Vol. 21, No. 1 (Spring, 1977), pp. 111-171Published by: Duke University Press on behalf of the Yale University Department of MusicStable URL: http://www.jstor.org/stable/843480

    Accessed: 14/04/2010 10:38

    Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

    you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

    may use content in the JSTOR archive only for your personal, non-commercial use.

    Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

    http://www.jstor.org/action/showPublisher?publisherCode=duke.

    Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

    page of such transmission.

    JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of 

    content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

     Duke University Press and Yale University Department of Music are collaborating with JSTOR to digitize,

    preserve and extend access to Journal of Music Theory.

    http://www.jstor.org

    http://www.jstor.org/stable/843480?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=dukehttp://www.jstor.org/action/showPublisher?publisherCode=dukehttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/843480?origin=JSTOR-pdf

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    2/62

    TOWARD

    A FORMAL THEORY

    OF TONAL MUSIC

    Fred Lerdahl and

    Ray

    Jackendoff

    INTRODUCTORY

    REMARKS

    We take

    the

    goal

    of

    a

    theory

    of

    music to

    be a

    formal

    de-

    scription

    of

    the musical

    intuitions

    of

    an

    educated

    listener.

    By

    "musical intuitions" we mean

    the

    largely

    unconscious knowl-

    edge

    which a

    listener

    brings

    to

    music

    and

    which allows him to

    organize

    musical sounds

    into coherent

    patterns.

    By

    "educated

    listener"

    we

    mean

    not

    necessarily

    a trained

    musician

    but

    a

    listener who is

    aurally

    familiar with the musical idiom in

    question.

    Such a listener is able

    to

    identify

    a

    previously

    unknown

    piece

    as

    an

    example

    of

    the

    idiom,

    to

    recognize

    ele-

    ments of

    a

    piece

    as

    anomalous

    within the

    idiom,

    and

    gener-

    ally,

    to

    comprehend

    a

    piece

    within the

    idiom.

    The

    "educated

    listener" is an

    idealization.

    Rarely

    do two

    people

    hear the same

    piece

    in

    precisely

    the same

    way

    or

    with

    the same

    degree

    of

    richness.

    Nonetheless,

    while it is

    conceiv-

    able to hear a piece any way one wants to, there is normally

    substantial

    agreement

    on

    what

    are

    the

    most

    natural

    ways

    to

    hear a

    piece.

    Our

    theory

    is

    concerned

    not

    with

    particular

    in-

    stances of

    hearing,

    which

    are

    always

    subject

    to a

    degree

    of

    111

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    3/62

    variability,

    but with

    the

    idealized

    underlying

    competence

    which

    the

    educated

    listener

    brings

    to

    bear

    in

    understanding

    tonal

    pieces.

    A

    theory

    about

    a

    particular type

    of music

    is,

    ideally,

    a

    sub-

    set of a

    theory

    of all music. We are

    constructing

    our

    theory

    of

    tonal music with this

    larger

    perspective

    in

    mind.

    While

    many

    of our

    specific

    rules are

    applicable

    only

    to tonal

    music,

    the

    basic

    components

    of

    the

    theory

    are

    designed

    to

    accommodate

    music

    of

    different traditions

    and

    historical

    periods.

    A

    reader

    who

    is at all

    acquainted

    with

    contemporary

    lin-

    guistics

    will observe

    that the

    goals

    we

    have

    set for

    ourselves

    are

    in

    some

    ways parallel

    to the

    goals

    of transformational

    generative grammar, which seeks to describe the linguistic

    intuitions

    of

    a native

    speaker

    of

    a

    language

    and

    to

    discover

    those

    aspects

    of

    particular

    grammars

    which are

    common to

    all

    languages.

    Indeed,

    our

    way

    of

    thinking

    about

    music is

    pat-

    terned

    after the

    methodology

    of

    linguistics

    in

    that we

    demand

    strong

    motivation,

    formal

    rigor,

    and

    predictive

    power

    for

    every part

    of the

    theory.

    On

    the

    other

    hand,

    we do

    not

    ap-

    proach

    music

    with

    any preconceptions

    that the

    substance

    of

    our theory will look at all like linguistic theory, since language

    and

    music are

    on the face

    of it

    different

    manifestations

    of

    human

    cognitive

    capacity.

    Previous theories

    of tonal

    music have

    not met such

    de-

    mands

    of

    rigor

    and

    prediction.

    Even

    Schenker's

    theory,

    which

    can

    be

    construed

    as

    having

    much in common

    with

    the

    genera-

    tive

    approach

    to

    linguistics,

    is

    at

    bottom

    inexplicit.

    One

    of

    the

    virtues

    of a formal

    theory

    is not that it is

    necessarily

    more

    "true," but that, even where incorrect or inadequate, it clari-

    fies issues

    precisely.

    To elucidate

    in what

    sense our

    theory

    is

    modeled

    after lin-

    guistic

    theory,

    we

    must

    mention a

    common

    misconception

    about

    transformational

    grammar.

    It

    is often

    thought

    that a

    Chomskian

    generative

    grammar

    is an

    algorithm

    that

    grinds

    out

    grammatical

    sentences;

    this view

    suggests

    that

    a

    genera-

    tive music

    theory

    should be

    a

    device

    which

    composes pieces.

    There are two errors in this view. First, the sense of "gener-

    ate"

    in the term

    "generative

    grammar"

    s not that of an elec-

    trical

    generator

    which

    produces

    electricity,

    but

    rather the

    mathematical

    sense,

    in

    which

    it

    means

    to

    describe

    a

    (usually

    infinite)

    set

    by

    formal

    means.

    Second, Chomsky

    distinguishes

    112

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    4/62

    the

    "weak

    generative

    capacity"

    of

    a

    theory

    from

    its

    "strong

    generative

    capacity."'

    A

    grammar

    weakly

    generates

    a set

    of

    sentences

    and

    strongly

    generates

    a set

    of structural

    descrip-

    tions,

    where

    each structural

    description uniquely specifies

    a

    sentence,

    but not

    necessarily

    conversely.

    It is the notion of

    strong

    generation

    which

    is

    overwhelmingly

    of

    interest

    in

    linguistic

    theory.

    The

    same

    holds

    for music

    theory,

    since

    a

    theory

    which

    weakly

    generates "grammatical"

    tonal

    pieces

    would

    tell

    us

    nothing

    interesting

    about their

    structures.

    A

    strongly

    generative theory

    of

    tonal music

    would

    not

    merely

    give

    a

    description

    of what

    pieces

    are

    grammatical.

    Rather,

    it

    would

    have to

    specify

    each

    tonal

    piece

    together

    with its structural description, i.e., a specification of all the

    structure

    which the

    educated

    listener

    infers in

    his

    perception

    of

    the

    piece.

    If

    a

    given

    piece

    cannot

    be

    heard as

    tonal,

    the

    theory

    should be

    unable

    to

    give

    it a

    structural

    description;

    if

    a

    piece

    can

    convincingly

    be

    heard

    in

    several

    ways,

    the

    theory

    should

    give

    it a

    different

    structural

    description

    for

    each

    way

    of

    hearing

    it.

    We

    have

    found that a

    generative

    music

    theory

    must

    not

    only assign structural descriptions to pieces, but must differ-

    entiate

    the

    structural

    descriptions

    along

    a

    scale of

    coherence,

    weighting

    them as

    more or

    less

    "preferred"

    ways

    of

    hearing

    a

    piece.

    Thus,

    the

    theory

    is

    divided

    into

    two distinct

    parts:

    well-

    formedness

    conditions,

    which

    specify

    possible

    structural

    de-

    scriptions;

    and

    preference

    rules,

    which

    designate,

    out

    of

    the

    possible

    structural

    descriptions,

    those

    that

    correspond

    to

    the

    educated

    listener's

    hearing

    of

    any

    particular

    piece.

    There are various criteria for determiningwhich structural

    descriptions

    of

    a

    piece

    are

    "preferred."

    Among

    these,

    of

    course,

    are our

    own

    intuitions.

    Furthermore,

    beyond

    all

    seemingly

    self-evident

    intuitions,

    a

    "preferred"

    structural

    description

    will

    tend

    to

    relate

    otherwise

    disparate

    elements in

    a

    satisfying

    way

    and

    to

    reveal

    surprising analytic

    insights.

    In

    addition,

    we are

    aware

    of

    relevant

    research

    in

    experimental

    psychology

    and are

    concerned

    that its

    findings

    be in

    agree-

    ment with our theoretical constructions. Criteriawithin the

    theory

    itself

    include the

    internal

    consistency

    of

    the

    rules

    and

    their

    generalization

    from

    particular

    instances

    to

    the

    entire

    body

    of

    classical

    tonal

    music. We

    utilize

    within

    the

    frame-

    work

    of

    the

    theory

    such

    psychologically

    primary

    notions

    as

    113

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    5/62

    parallelism,

    articulation,

    and

    stability. Finally,

    the nature

    of

    our structures must be

    psychologically plausible

    in

    that

    their

    complexity

    must result from the interaction

    of a

    fairly

    small

    number

    of

    processes,

    each

    of which

    taken in

    isolation

    is

    rela-

    tively simple.

    Although

    it

    would

    be

    possible

    within

    the

    theory

    to

    gener-

    ate new tonal

    pieces,

    we

    have chosen

    only

    to

    generate

    struc-

    tural

    descriptions

    for

    existing pieces.

    It

    would

    seem

    to be

    inherently

    less

    rewarding

    to

    specify

    normative but dull

    pieces

    than

    to

    develop

    structural

    descriptions

    for works

    of

    lasting

    interest.

    Moreover,

    in the

    former

    case,

    there

    would

    always

    be

    the

    danger,

    through

    excessive

    limitation of the

    possibilities

    in

    the interest of conceptual manageability, of oversimplifying

    and

    thereby establishing

    shallow

    or incorrect

    principles

    with

    respect

    to music in

    general.

    For

    the common

    conception

    of

    a

    transformational

    genera-

    tive

    grammar

    as

    merely

    a

    sentence-generating

    device

    is

    mis-

    taken

    in

    a further

    respect.

    Linguistic

    theory

    is

    not

    simply

    concerned

    with the

    analysis

    of

    a limited set

    of

    sentences;

    rather it considers

    itself a branch

    of

    psychology,

    concerned

    with making empirically verifiable claims about one complex

    aspect

    of

    human

    mental

    life,

    namely

    language.

    Similarly,

    by

    putting

    our

    emphasis

    on the

    musical intuitions

    of the

    edu-

    cated

    listener,

    and

    by

    taking

    as our

    sample

    a

    highly

    complex

    body

    of

    music,

    we are

    asserting

    that the

    analysis

    of

    pieces

    of

    music,

    though

    not

    without

    a

    great

    deal

    of intrinsic

    interest,

    is

    not

    an end

    in

    itself.

    Rather

    the

    goal

    is

    an

    understanding

    of

    the

    mental

    process

    of musical

    perception,

    a

    psychological

    phenomenon. From this viewpoint, our theory of music is not

    just

    an

    analytic system,

    but makes

    strong

    claims

    about

    the

    delimitation

    of

    possible

    theories

    of musical

    cognition.

    By

    regarding

    music

    only

    as

    apprehended

    structure,

    we

    are

    deliberately

    avoiding

    the difficult

    issue of musical

    meaning.

    Whatever

    music

    may

    "mean,"

    it

    is in no sense

    comparable

    to

    the semantic

    component

    in

    language;

    there

    are no

    substantive

    parallels

    to

    sense and

    reference

    in

    language,

    or to

    such seman-

    tic

    judgments

    as

    synonymy, analyticity,

    and entailment.

    It is

    in

    the domain

    of

    syntax

    that the

    linguistic approach

    has rele-

    vance

    to music

    theory.

    Yet even

    here

    there are

    no

    substantive

    parallels

    between

    musical

    structure

    and

    such

    grammatical

    categories

    in

    language

    as

    noun, verb,

    adjective,

    noun

    phrase,

    114

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    6/62

    verb

    phrase,

    and

    so

    forth. The

    concepts

    of

    musical structure

    must

    be

    developed

    in terms

    of

    music

    itself.

    We

    likewise

    avoid

    the

    issue

    of

    aesthetic

    value.

    Nevertheless,

    that we

    are

    dealing

    with

    works of art rather than

    "mental

    objects"

    from the everyday world, such as sentences, is to a

    degree

    problematic.

    Whereas

    a sentence

    normally

    has a

    defi-

    nite

    meaning,

    structural

    ambiguity

    in a

    work

    of

    art

    is

    com-

    mon.

    And

    whereas a

    sentence

    normally

    has a definite

    function,

    it is in

    the

    nature of a

    work of art that it is

    appreciated

    and

    contemplated

    from various

    points

    of

    view and with various

    purposes

    in

    mind,

    not

    only by

    different

    people

    but

    by

    the

    same

    person

    on

    different

    occasions.

    These

    differences,

    how-

    ever,do not mean that the understandingof a work of art can

    take

    any arbitrary

    form

    whatsoever; rather,

    they

    mean

    that,

    to an

    extent,

    multiple

    understandings

    are

    possible,

    desirable,

    and

    even

    inevitable.

    In

    constructing

    our

    music

    theory,

    we

    have

    accounted for this

    state

    of affairs

    by

    building

    a

    system

    of

    interactive

    components

    and

    by

    emphasizing

    the

    "pre-

    ferred"

    nature

    of the

    resulting

    structural

    descriptions.

    Under

    our

    conception

    of

    music

    theory,

    then,

    the under-

    standing of a piece of music by the idealized listener consists

    in

    his

    finding

    the

    maximally

    coherent

    structural

    description

    or

    descriptions

    which

    can

    be

    associated

    with

    the

    piece's

    se-

    quence

    of

    pitch-time

    events.

    Maximizing

    coherence involves the

    interaction of

    a

    number

    of

    different

    domains

    of

    analysis,

    each

    of

    which

    must

    be

    repre-

    sented

    in

    the

    structural

    description.

    There

    are four

    with which

    we will be

    concerned

    here,

    to

    be

    termed

    grouping

    analysis,

    metrical analysis, time-span reduction, and prolongational

    reduction. As an initial

    overview,

    we

    may

    say

    that the

    group-

    ing

    analysis

    assigns

    group

    boundaries

    to

    the

    music in a hier-

    archic fashion

    at

    every

    level of

    a

    piece.

    The

    metrical

    analysis

    assigns

    a

    hierarchy

    of

    strong

    and weak

    beats. The

    time-span

    reduction

    designates

    "structural

    beginnings"

    and

    "structural

    endings"

    of

    groups,

    and

    assigns

    to the

    pitches

    a

    hierarchy

    which

    relates

    them

    to

    the

    grouping

    and

    metrical

    structures.

    The

    prolongational

    reduction

    assigns to the pitches a hier-

    archy

    which

    expresses

    harmonic

    and melodic

    continuity

    and

    progression;

    it is

    the

    closest

    equivalent

    in

    our

    theory

    to

    Schenkerian

    analysis.

    There

    are

    some

    abstract

    properties

    common

    to

    these

    four

    115

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    7/62

    domains

    of

    analysis.

    First,

    each

    domain

    partitions

    a

    piece

    into

    discrete

    regions,

    organized

    hierarchically

    in

    such

    a

    way

    that

    one

    region

    may

    contain

    other

    regions,

    but

    may

    not

    partially

    overlap

    with

    other

    regions.

    For

    example, Figure

    1

    a

    represents

    a possible kind of organization. Figure lb represents an im-

    possible organization:

    at

    j,

    two

    regions

    overlap

    within

    level

    2;

    and

    at

    k,

    a

    boundary

    in level

    3

    overlaps

    a

    region

    in

    level

    2.

    Another

    property

    common to these

    domains

    is that the

    processes

    of

    organization

    are

    essentially

    the same at all hier-

    archic levels.

    A related

    point

    is

    that the

    processes

    of

    organiza-

    tion of these domains

    are

    recursive, i.e.,

    capable

    of indefinite

    elaboration

    by

    the same

    rules.

    Other

    aspects

    of musical struc-

    ture, however, are not hierarchic in nature. For the present

    we shall

    ignore

    these

    dimensions.2

    As mentioned

    above,

    the rules which

    assign

    structural de-

    scriptions

    are

    categorized

    as

    well-formedness

    rules,

    which

    assign

    possible

    structures,

    and

    preference

    rules,

    which

    select

    coherent structures from

    possible

    structures.

    In

    addition,

    transformational

    rules,

    which convert structures into other

    structures,

    are

    needed

    for

    special

    cases

    (such

    as

    elisions)

    not

    generatedby the well-formednessrules. Although transforma-

    tional

    rules have

    been

    central to

    linguistic theory, they

    play

    a

    peripheral

    role in

    our

    music

    theory,

    at

    least

    at its

    current

    stage

    of

    development.

    The

    following

    discussion

    of

    the

    organization

    of

    the

    theory

    will be

    informal.

    Emphasis

    will

    be

    placed

    on how the

    compo-

    nents

    work

    in

    principle.

    To

    give

    a

    complete

    account

    would

    exceed

    our

    present

    purpose,

    which

    is to

    convey

    in

    general

    the

    goals, operations, and implications of the theory as a whole.

    We

    begin

    with

    a discussion

    of

    rhythmic

    structure in terms of

    grouping

    analysis

    and

    metrical

    analysis.

    Then we

    develop

    the

    two

    modes

    of

    pitch

    hierarchization,

    time-span

    reduction and

    prolongational

    reduction.

    Finally,

    we

    apply

    the

    completed

    system

    to

    a

    piece

    of some

    complexity.

    RHYTHMICSTRUCTURE

    When

    hearing

    a tonal

    piece,

    the listener

    naturally

    organizes

    the sound

    signals

    into

    units such

    as

    motives, themes,

    phrases,

    periods,

    theme-groups,

    sections,

    and the

    piece

    itself.

    Our

    generic

    term for these

    units

    is

    "group."

    The

    grouping

    analysis

    116

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    8/62

    Figure

    la

    lb

    3)

    ..

    .

    J

    4)

    k

    a

    Figure 2a

    2b

    : *

    /

    0,~?

    117

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    9/62

    picks

    out the

    groups

    and indicates them

    in

    the

    structural

    de-

    scription

    by

    slurs

    beneath

    the musical

    notation. The

    grouping

    well-formedness

    rules restrict the

    possible

    grouping

    structures

    to the

    kind

    of hierarchic

    organization

    discussed

    above,

    where

    Figure

    1a was

    possible

    and

    Figure

    lb was not

    possible.

    Under

    these

    conditions,

    both

    Example

    1

    a

    and

    Example

    Ib are

    possi-

    ble

    groupings.

    However,

    while

    Example

    la would

    appear

    to

    show

    the

    cor-

    rect

    grouping,

    Example

    lb

    is

    absurd.

    In

    order to

    select

    the

    actually

    heard

    grouping

    or

    groupings

    (such

    as

    Example

    la),

    as

    against

    all the

    merely

    possible

    groupings

    (such

    as

    Example

    lb),

    we

    develop

    the

    grouping preference

    rules. These

    provide

    the criteria for pickingout groups, and are classified according

    to

    principles

    of

    (a)

    articulation

    of

    boundaries, (b)

    parallelism

    in

    structure,

    and

    (c)

    symmetry.

    Group

    boundaries

    are articulated

    by

    such factors as dis-

    tance

    between

    attack

    points,

    rests,

    slurs written into the

    music,

    change

    in

    register, change

    in

    texture,

    change

    in

    dy-

    namics,

    and

    change

    in timbre.

    A

    further

    articulatory

    device

    is

    the

    harmonic

    cadence,

    which from

    the

    phrase

    level

    upward

    normally signifies the ending of groups;this will be discussed

    later.

    Parallelism

    in structure involves

    some

    kind

    of

    repetition

    or

    similarity

    in the

    music,

    such

    as

    a

    motive,

    a

    sequence,

    a

    section,

    and so

    forth. The

    similarity

    is

    particularly

    crucial at

    the

    begin-

    ning

    of

    groups;

    for even

    if

    they

    diverge

    later

    on,

    they

    are

    still

    perceived

    as

    parallel.

    In tonal

    music,

    parallelism

    is the

    major

    factor

    in all

    large-scale

    grouping.

    Related to parallelismis the principle of symmetry, which

    states that

    the

    ideal

    subdivision

    of

    any

    group

    is

    into

    equal

    parts.

    In

    Example

    la,

    all the

    groupings

    are

    assigned

    by

    the

    parallelism

    rule

    (reinforced

    by

    various

    articulatory

    criteria),

    except

    for the

    groupings

    marked

    s,

    which

    are

    due

    to

    the

    sym-

    metry

    rule;

    thus,

    each 4-measure

    group

    subdivides

    not

    only

    into

    1

    +

    1

    +

    2,

    but,

    at the

    next

    level,

    into 2

    +

    2.

    A

    grouping

    transformational

    rule is

    required

    to account

    for grouping overlaps, which as such do not meet the well-

    formedness

    conditions

    of hierarchic

    organization.

    The trans-

    formational

    rule

    relates

    well-formed

    underlying

    groupings

    to

    the

    musical

    surface,

    thereby preserving

    the

    sense that

    overlaps

    are variations

    on

    normal hierarchic

    grouping.

    Thus,

    in

    Exam-

    118

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    10/62

    Example

    la.

    Beethoven: Sonata

    Op.

    2,

    No.

    2,

    Scherzo

    ^^f

    $

    r f

    1

    ff

    r>TI

    p

    lb

    P

    A-

    $

    T

    ^-

    z

    C

    $%

    4^

    v.?

    t^

    m J j J

    ^

    v$^ .^^t^ v^rf^^' ^

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    11/62

    pie

    2,

    the

    events at

    u

    and

    v

    function as

    both

    endings

    and

    beginnings

    of

    groups;

    these

    overlaps

    are

    disentangled

    at

    under-

    lying

    levels.

    We turn

    now

    to metrical

    structure,

    which

    is

    independent

    of

    grouping

    structure but interactive with it. Given

    appropriate

    musical

    cues,

    the listener

    will

    instinctively

    infer a

    regular,

    hierarchic

    pattern

    of

    beats

    to

    which

    he

    relates

    the

    actual

    musical sounds.

    The

    metrical

    analysis

    assigns

    to a

    piece

    such a

    pattern

    of beats and

    indicates

    them

    in

    the structural

    descrip-

    tion

    by

    dots beneath

    the

    musical notation and above

    the

    grouping

    slurs,

    as in

    Example

    3.

    Each

    level

    of

    dots

    represents

    a

    marking-off

    of the

    music

    into equal time-spans. A dot at a particularlevel representsa

    judgment

    that

    that

    moment

    in the

    music is a

    beat

    at that

    level. If a

    beat

    at

    a

    particular

    level

    is felt

    to

    be

    "strong,"

    or

    "down,"

    it

    is a

    beat

    at

    the next

    larger

    level

    and

    receives an

    additional

    dot.

    (Thus,

    metrical

    structure

    does

    not exist

    with-

    out

    at least

    two

    levels of

    dots.)3

    The

    process

    is the same

    whether

    at the

    level of the

    smallest

    note value

    or

    at

    a

    hyper-

    measure

    level. The

    notated

    meter

    is

    usually

    an intermediate

    metrical level.

    Theoretically,

    the

    dots

    could

    be built

    up

    to

    the level of a

    whole

    piece.

    However,

    the

    perception

    of relative metrical

    stress fades

    over

    long

    timespans.

    In

    addition,

    at

    large

    levels,

    metrical

    structure

    is heard

    within

    the context

    of

    grouping

    structure,

    which

    is

    rarely regular

    at

    such

    levels;

    and

    without

    regularity

    there can be

    no

    metrical structure.

    Therefore,

    metrical

    structure

    is a

    comparatively

    local

    phenomenon.

    The metrical well-formedness rules assure the hierarchic

    condition

    that a beat

    at a

    particular

    level

    must also

    be

    a

    beat

    at all

    smaller

    levels. Characteristics

    of

    metrical

    well-formed-

    ness in

    classical

    tonal

    music

    include

    the

    equal

    spacing

    of beats

    and

    the

    provision

    that at

    each

    successive

    level the

    distance

    between

    beats

    must be

    either

    two

    or

    three times that

    of the

    immediately

    lower

    level. Musical

    styles

    of other cultures

    and

    historical

    periods

    often

    require

    more

    complicated

    rules

    of

    metrical well-formedness; the rhythmic complexities of tonal

    music

    arise

    from

    the

    interaction

    of metrical

    structure

    with

    grouping

    structure

    and

    pitch

    structure.

    It

    hardly

    needs

    emphasizing

    that bar

    lines and

    beams

    be-

    tween

    notes

    are notational

    devices

    and

    not

    part

    of

    the

    physi-

    120

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    12/62

    Example

    2.

    Mozart:

    Sonata

    K.

    279,

    I

    Example

    3.

    Bach:

    "O

    Haupt

    voll

    blut und

    Wunden"

    r.,N

    I i I

    I,

    "

    /t4

    J

    LJ

    i

    .

    LL

    t rbr

    r r

    r

    C r

    CJ

    I

    n I

    j j

    J

    u Jj

    *0

    It*

    0.

    .

    '

    I.

    '

    *

    r

    ?

    ?

    e

    *

    0

    0)..

    .

    .

    I.

    - -

    I,

    ..

    , T14 _

    - I

    ...

    dr

    I

    - I

    I

    II

    --

    I

    I

    I I I

    _I

    I

    I I

    I

    ?

    ? ?

    ? ?

    *

    ?

    ?

    e

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    13/62

    cal

    signal,

    as

    pitches,

    durations,

    dynamics,

    and timbre

    are.

    From

    the

    physical .signal

    the

    metrical

    preference

    rules

    assign

    to

    a

    piece

    the

    actually

    heard

    metrical

    structure

    (or structures)

    instead

    of the

    myriad

    well-formed

    but

    inappropriate

    metrical

    structuresapplicable to it. For example, they select the metri-

    cal

    structure

    in

    Example

    3 rather

    than one in a

    triple

    meter,

    or

    one

    which

    places

    the

    strongest

    downbeat

    of the

    passage

    on

    the

    opening

    event.

    The metrical

    preference

    rules

    can be classified

    according

    to

    principles

    of

    (a)

    cues

    for

    strong

    beats, (b)

    parallelism

    with

    grouping

    structure,

    and

    (c)

    regularity

    of

    pattern.

    The

    cues in

    the music

    for

    relatively strong

    beats

    include

    such

    factors

    as

    attack, accent, change of dynamic, registerof pitch, harmonic

    change,

    and

    suspensions.

    Added

    to

    these is the listener's

    tendency

    to ascribe

    parallel

    metrical

    structures to

    parallel

    grouping

    structures

    (especially

    rhythmic patterns

    which

    form

    groups).

    If there

    is

    any

    regularity

    to

    these

    various

    cues,

    the

    listener

    extrapolates

    an

    entire

    metrical

    hierarchy,

    which

    he

    will renounce

    only

    in

    the

    face of

    strongly

    contradictory

    cues.

    Syncopation

    takes

    place

    when

    there are

    strong

    contradictory

    cues which yet are not strong enough to overridethe inferred

    pattern.

    A metrical

    transformational

    rule is needed

    for metrical

    overlaps,

    in

    which

    a shift

    in the metrical

    structure occurs

    in

    such a

    way

    that

    the

    same

    moment

    in a

    piece

    serves

    a

    double

    metrical

    function.

    The

    transformational

    rule

    deletes

    one

    set

    of

    dots

    in favor of

    the other

    at

    the

    musical

    surface.

    In

    Figure

    2a,

    the weaker

    metrical

    function

    is

    deleted;

    in

    Figure

    2b,

    the

    stronger is deleted. (It may be helpful to think of the largest

    level of

    dots

    as

    representing

    the measure

    or half-measure

    level.)

    Although

    it

    is conceivable

    for metrical

    overlaps

    to

    arise

    independently,

    they

    generally

    happen

    as

    a

    consequence

    of

    grouping

    overlaps

    which

    take

    place

    on

    relatively

    weak

    metri-

    cal stresses.

    On the

    other

    hand,

    if a

    grouping

    overlap

    takes

    place

    on a

    relatively

    strong

    metrical

    stress

    (as

    in

    Example

    2),

    no metrical

    overlap

    results.

    Using

    familiar

    terminology,

    we

    call a

    combined

    grouping

    and

    metrical

    overlap

    an elision.

    We

    further

    distinguish

    between

    elisions

    in which

    the weaker

    metrical

    function

    has been

    deleted

    (as

    in

    Figure

    2a),

    and

    elisions

    in

    which

    the

    stronger

    metrical

    function

    has

    been

    122

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    14/62

    deleted

    (as

    in

    Figure

    2b).

    Example

    4 is

    typical

    of

    the

    former

    kind;

    the

    effect at

    s is one of

    jarring

    reorientation.

    In

    the

    latter

    kind,

    the effect is rather

    one of

    retrospective

    awareness

    that

    a shift

    has

    taken

    place;

    there is an

    example

    of this

    in

    the

    Schumann

    song

    analyzed

    later in this

    paper

    (pp.

    149ff).

    Some

    general

    reflections

    are in order

    concerning

    the

    organ-

    ization of the

    grouping

    and

    metrical

    components.

    It seems

    to

    us essential that the two not

    be confused

    inadvertently.

    It

    is

    tempting,

    for

    example, perhaps

    on the

    basis of a

    misleading

    spatial analogy,

    to

    attribute

    duration to the

    concept

    of

    metri-

    cal

    stress,

    so that first

    a

    beat,

    then a

    measure or

    motivic

    grouping,

    and

    finally

    a

    phrase

    receives

    "strong"

    or

    "weak"

    markingsas the analysis proceeds to higher levels. In ourview,

    however,

    the notion of "beat"

    or

    "metrical stress"

    can

    only

    be understood

    as a

    point

    in

    time,

    without

    duration;4

    hence

    our use

    of dots to

    signify

    metrical structure.

    Beats are

    correct-

    ly

    analogous

    to

    equidistant

    geometric points

    rather

    than

    to

    the lines

    drawn between

    them;

    while

    rhythms

    and

    groups

    have

    duration, then,

    beats do not. The

    metrical

    component

    assigns

    metrical stress

    not to

    groups

    but

    to beats

    within

    groups.

    Of course the

    listener senses

    that a

    group

    has a

    certain

    weight

    if

    within it

    there is a

    strong

    beat. But this

    does

    not

    mean

    that the

    group

    as

    a

    whole

    receives

    metrical

    weight;

    for

    the weak beats

    are all

    equally

    weak.

    In

    Example

    5a,

    for in-

    stance,

    the

    E-flats are

    metrically

    equal,

    a

    fact

    obscured

    in

    Example

    5b,

    in

    which

    metrical

    properties

    appear

    to be in-

    cluded within a

    grouping

    notation.5

    Just as groups as such do not receive metrical stress, metri-

    cal

    structure as

    such

    does not

    possess any

    inherent

    grouping.

    Whether a

    weak beat is

    heard

    as an

    afterbeat or

    as an

    upbeat

    is

    entirely

    a

    matter of the

    grouping

    associated with

    it.

    Similar-

    ly,

    at

    a

    larger

    level of

    grouping, say

    the

    paradigmatic

    ante-

    cedent-consequent

    pattern

    in

    Figure

    3,

    there is

    nothing

    in

    the

    metrical

    structure

    which

    prevents

    the half

    cadence

    in m. 4

    from

    resolving

    as a full

    cadence to

    the

    tonic in

    m.

    5.

    It

    does

    not so resolve only because of the intervening group bound-

    ary.6

    Figure

    3

    poses

    a

    question

    of a

    different

    sort.

    A

    phrase

    can

    be

    roughly

    characterized as the

    lowest

    level of

    grouping

    which

    has

    a

    structural

    beginning,

    a

    middle,

    and a

    structural

    ending

    123

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    15/62

    Example

    4.

    Haydn: Symphony

    no.

    104,

    I,

    mm.

    17ff.

    i$f

    r

    f

    r

    0

    1

    8

    i

    j^

    .

    a

    r

    4^

    r

    Y)r

    r J

    8

    -

    .

    .

    .

    ?

    Example

    6.

    Beethoven:

    "Hammerklavier"

    Sonata

    op.

    106,

    I

    j

    f

    a

    -

    t

    '

    t

    *

    *

    ;

    'r

    'f

    r

    r

    rI

    """=~~''(t

    i

    cedent)

    ~

    _ ~

    ntec

    de nt

    124

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    16/62

    : **

    :

    :

    r'

    '

    I

    ?

    ':

    , '

    :

    .

    .

    .

    I

    j[

    fr'f

    TS

    '

    rvm

    f=

    u=ro

    .

    -

    -

    L'

    @

    *

    \

    'F

    rr

    rJ

    JJ

    j

    'r

    r

    ir

    J

    J

    (extended

    consequenT)

    125

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    17/62

    Example 5a. Mozart: Symphony no. 40, I

    ibb

    t

    n ,

    n

    n

    ?

    .

    ?

    .

    ?

    .

    . .

    ?

    ? ? ?

    ?

    ? ?

    ?

    ?

    Sb

    ;9

    kr

    f

    ,

    r'

    i

    v

    i

    1,

    /

    ,Example

    7. Bach

    Example

    7.

    Bach:

    "0

    Haupt"

    126

    *

    .I

    I

    ._

    ....

    ,

    -,

    -

    I

    '..,

    ?

    ,

    ,,

    .

    ? ,

    ?

    ? ,

    ?

    1

    1 t

    1

    1 ?

    J

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    18/62

    Figure

    3

    I

    w%%V I-

    measure:

    1

    2

    3

    4

    5

    6

    7

    8

    Figure 4a

    (a'I

    pitch

    events:

    W z

    4b

    (b)

    (C)

    w

    y

    z

    4c

    w

    x

    y

    z

    127

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    19/62

    (a cadence).

    Thus,

    a

    phrase

    normally

    has a kind of structural

    gravity

    at or

    near its

    group

    boundaries.

    If

    groups

    as such have

    no metrical

    stress,

    and

    if metrical

    structure

    is

    an

    extrapolation

    of

    multi-leveled, regularly spaced beats,

    how are

    these

    points

    of

    gravity

    accounted

    for in the

    structural

    description?

    In

    Figure

    3

    there

    is no conflict

    between

    metrical stress and

    structural

    weight

    in

    mm.

    1

    and

    5;

    but

    mm.

    4

    and

    8,

    where

    the cadences

    occur,

    are

    metrically

    relatively

    weak

    (depending

    on

    the

    example,

    the

    cadences

    might

    even

    happen

    in

    the second

    halves

    of these

    measures).

    One

    solution

    might

    be

    to move the

    second

    strong

    metrical

    stress of

    each

    phrase

    from

    the down-

    beats

    of the third

    and seventh measures

    to the

    points

    at which

    the cadences actually occur.7 However, there are two strong

    reasons

    against

    such a

    revision

    of the

    metrical

    component.

    First,

    it

    would

    entail

    surrendering

    one

    of the formal

    proper-

    ties

    common

    to

    all the

    domains

    under

    discussion,

    namely

    that

    the

    processes

    of

    organization

    are

    essentially

    the same at

    all

    hierarchic

    levels.

    Secondly,

    it

    would

    mean

    giving

    up

    the tradi-

    tional

    distinction

    between

    cadences

    which

    take

    place

    at

    weak

    metrical

    points

    and cadences

    which

    take

    place

    at

    strong

    metri-

    cal points. The latter are particularlyimportant for achieving

    large-scale

    arrival

    and-in

    conjunction

    with

    grouping

    overlaps

    -continuation.

    Thus,

    in

    Example

    68 the

    consequent

    phrase

    is

    extended

    so that

    the cadence

    at

    q

    arrives

    on a

    strong

    beat

    and

    overlaps

    with

    the

    succeeding phrase.

    According

    to the

    pro-

    posed

    revision,

    the

    cadence

    of the antecedent

    phrase

    at

    p

    also

    would

    have

    to

    be

    metrically strong,

    with

    the result that

    the

    metrical

    distinction

    between

    p

    and

    q

    would be

    lost.

    This is

    plainly not acceptable.

    The

    points

    of

    gravity

    in a

    phrase,

    then,

    are not to

    be

    inter-

    preted

    as

    metrical

    phenomena.

    Rather, they

    are

    hierarchically

    important

    elements

    produced

    by

    the

    interaction

    of

    pitch

    structure

    and

    grouping

    structure;

    and

    they

    stand with

    metrical

    structure

    in a

    contrapuntal

    relation,

    so to

    speak,

    which

    under-

    lies

    much

    of the

    rhythmic

    richness

    of

    tonal

    music. The

    metri-

    cal

    component

    therefore

    remains

    as

    originally

    set forth.

    As

    will be seen below, the proper distribution of structural

    weight

    in a

    phrase

    emerges

    in the

    analysis

    as

    part

    of

    the

    time-

    span

    reduction,

    which

    treats

    a

    phrase

    as an elaboration

    of its

    beginning

    and

    ending.

    128

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    20/62

    PITCH

    REDUCTION

    Both the

    time-span

    reduction and

    the

    prolongational

    re-

    duction

    assign

    hierarchic structures

    to all

    the

    pitch

    events in

    a

    piece.9

    To

    represent

    these two kinds of

    reduction,

    we have

    invented two somewhat different

    "tree"

    notations;

    these

    visu-

    ally

    resemble,

    but are

    substantially

    different

    from,

    the

    tree

    notation

    utilized

    in

    linguistics. Linguistic

    trees

    represent

    "is-

    a"

    relations: a noun

    phrase

    followed

    by

    a

    verb

    phrase

    is

    a

    sentence;

    a

    verb followed

    by

    a noun

    phrase

    is

    a verb

    phrase;

    and

    so

    forth. Our musical

    trees,

    however,

    do not involve

    gram-

    matical

    categories.

    The

    fundamental

    relationship

    which

    they

    express is that of a sequence of pitch events as being an elabo-

    ration

    of

    a

    single pitch

    event.

    The

    dominating

    event,

    that

    of

    which a

    sequence

    of

    events is an

    elaboration,

    is

    always

    one

    of

    the events in

    the

    sequence;

    the

    remaining,

    subordinate events

    in the

    sequence

    are

    heard

    as

    relatively

    ornamental.

    "Reduc-

    tion"-the

    process

    of

    recursively

    substituting single

    events for

    sequences

    of

    events-can be

    thought

    of

    as the

    inverse of

    elab-

    oration.

    In the following exposition we begin with the time-span

    component

    and

    then turn to the

    prolongational

    component.

    In

    both

    cases,

    after

    outlining

    the notations

    and the

    basic

    principles

    of

    reduction,

    we

    apply

    the

    components

    first

    to

    an

    abstract

    antecedent-consequent

    pattern,

    then

    to the

    opening

    eight

    measures of

    Mozart's

    Sonata,

    K.

    331.

    In

    the

    tree

    notation

    for the

    time-span

    reduction,

    a

    "right

    branch"

    (

    X),

    in

    which

    a

    line

    to the

    right

    attaches

    to a line

    to

    the left, denotes the subordination of an event to the preced-

    ing

    event

    within

    that

    region

    at that

    level;

    a "left

    branch"

    (

    X),

    in

    which

    a

    line

    to

    the

    left

    attaches

    to a

    line to the

    right,

    denotes the

    subordination of

    an

    event to the

    following

    event

    within

    that

    region

    at that level.

    The

    well-formedness

    condi-

    tions

    for these

    trees

    prohibit

    both the

    crossing

    of

    branches,

    as

    in

    Figure

    4a,

    and

    the

    assignment

    of

    more

    than one

    line of

    the

    tree

    to the

    same

    event,

    as in

    Figure

    4b.

    (The

    letters in

    paren-

    theses signify reductional levels.)

    The

    relevant

    notion of

    elaboration in

    the

    time-span

    reduc-

    tion is

    elaboration

    at

    successive levels

    within

    more or

    less

    equally

    spaced,

    discrete

    time-spans.

    Within

    each

    time-span,

    or

    region,

    a

    dominating

    event must

    be

    found;

    that

    is,

    all

    other

    129

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    21/62

    events in

    the

    region

    are elaborations

    of

    that event.

    Thus,

    at

    level

    (c)

    in the well-formed tree

    in

    Figure

    4c,

    x is in

    the

    region

    of

    w

    and

    is

    an elaboration

    of

    w;

    y

    is

    in

    the

    region

    of

    z

    and

    is

    an

    elaboration

    of

    z.

    At level

    (b),

    all

    four

    events are in the

    same

    region,

    and z

    (together

    with x and

    y)

    is an elaboration

    of

    w.

    Before

    the

    rules

    which

    establish

    the tree

    structure

    can

    be

    applied,

    it is

    necessary

    to select the

    regions

    of

    application

    at

    every

    hierarchic

    level

    within

    a

    piece.

    These

    regions

    for all but

    the most local

    levels of

    analysis

    consist of the

    groupings

    assigned

    by

    the

    grouping

    preference

    rules. Within the

    lowest

    grouping

    level,

    smaller

    regions

    are

    chosen in terms of levels of

    metrical structure. In these smallerregions, a given weak beat

    is bracketed with

    the

    preceding

    strong

    beat,

    unless the

    pre-

    ceding strong

    beat

    is

    separated

    from

    the weak

    beat

    by

    a

    group

    boundary;

    in

    this

    latter

    case,

    the

    weak

    beat

    is

    bracketed

    with

    the

    following strong

    beat.

    In musical

    terminology,

    this means

    that

    a weak

    beat

    is an afterbeat

    unless it is situated

    at the

    beginning

    of a

    group,

    in

    which

    case it

    is

    an

    upbeat.

    See

    Exam-

    ple

    7;

    the

    brackets

    indicate

    the

    sub-group regions

    of

    applica-

    tion. In a given tree, each level of branchingcorresponds to a

    region

    of

    application

    for

    the

    preference

    rules.

    Given these

    regions

    of

    application,

    the

    preference

    rules for

    the

    time-span

    reduction

    choose

    the

    syntactically

    most co-

    herent reduction

    (or

    reductions)

    from

    all the

    possible

    but

    mostly implausible

    reductions

    of a set of

    pitch

    events.

    Syn-

    tactic coherence

    in

    this domain

    can be

    thought

    of

    in terms

    of

    stability.

    These

    preference

    rules are

    classified

    as

    (a)

    those

    which ascertain the most stable structure (the tonic), and (b)

    those which

    establish

    the

    hierarchy

    of relative

    stability

    in rela-

    tion

    to the most

    stable

    structure.

    The tonic-at

    any

    level,

    local

    or

    global-is

    selected with

    reference to

    the available

    pitch

    collection and cadential

    struc-

    ture at the

    appropriate

    level.

    The

    rules of

    relative

    stability

    or

    instability

    are,

    in

    broad

    musical

    terms,

    the

    principles

    of

    relative

    consonance

    or disso-

    nance. For example, a local consonance is more stable than a

    local

    dissonance;

    a triad

    in

    root

    position

    is

    more stable

    than

    its

    inversions;

    a chord

    is more

    stable

    if its melodic

    note

    is the

    same

    pitch-class

    as

    its

    root;

    the

    relative

    stability

    of

    two

    chords

    can be determined

    by

    the relative

    closeness

    to the local

    tonic

    130

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    22/62

    of

    their

    roots on the circle

    of

    fifths;

    conjunct

    linear

    connec-

    tions

    are more

    stable

    than

    disjunct ones;

    a

    pitch

    event

    is

    more

    stable

    if it is in a

    metrically

    stronger position;

    and

    so

    forth.

    The

    preference

    rules

    use all of these criteria

    in

    picking

    out

    which event in a

    given region

    is

    dominating.

    Thus

    (and

    let

    us assume here

    the

    "preferred"

    grouping

    and

    metrical

    structures)

    both

    Example

    8a and

    Example

    8b

    are

    well-formed

    time-span

    reductions.

    The

    preference

    rules

    select

    the

    reduction in

    Example

    8a as

    opposed

    to

    any

    other

    possible

    reduction,

    such as

    Example

    8b.

    In

    these

    examples,

    and

    in

    all

    succeeding time-span

    reduc-

    tions,

    a

    notation

    formally

    equivalent

    to the

    tree

    appears

    beneath the music proper: the relation of pitch structure to

    metrical structure

    is

    expressed by

    the

    placement

    of

    the

    syn-

    tactically

    most

    significant

    event

    within a

    bracketing

    on

    the

    strongest

    beat within

    that

    bracketing.'? Although

    the

    two

    notations

    express

    the

    same

    information,

    we

    retain

    both

    in

    the

    time-span

    reduction

    because,

    as

    will

    become clear

    in

    longer

    examples,

    they

    serve

    somewhat

    different

    purposes.

    The one

    below the

    music

    is useful in

    hearing any

    particular

    hierarchic

    level. The tree-which, though unfamiliar, is easily compre-

    hended

    with

    a

    little

    effort-gives

    a

    picture

    of all

    the

    levels in

    relation to

    one

    another;

    moreover,

    it

    is

    illuminating,

    in

    con-

    nection

    with a

    particular

    piece,

    when

    compared

    to the

    pro-

    longational

    reduction.

    At

    relatively

    local

    levels,

    the tree

    for the

    time-span

    reduc-

    tion

    correlates

    with

    the

    metrical

    analysis

    to

    produce

    the

    paradigmatic

    situations in

    Figure

    5. Both

    (h)

    and

    (i)

    pertain

    to afterbeats: in (h) the event on the afterbeat is the less

    stable

    event,

    such

    as

    a

    passing-

    or

    neighboring-tone

    or

    -chord;

    in

    (i)

    the

    event on

    the

    downbeat is

    the

    less

    stable

    event,

    such

    as

    a

    suspension

    or an

    appoggiatura-tone

    or

    -chord. Both

    (j)

    and

    (k)

    pertain

    to

    upbeats:

    in

    (j)

    the

    event

    on

    the

    upbeat

    is

    less

    stable than

    the event

    on its

    associated

    downbeat;

    in

    (k)

    the

    event

    on

    the

    downbeat

    is

    less

    stable than

    the

    event

    on

    its

    associated

    upbeat.

    In

    these

    ways,

    the

    relation

    of

    syntactically

    significant events to metrical structure is made clear. At large

    levels,

    where

    metrical

    structure

    is

    no

    longer

    hierarchically

    operative,

    the

    right

    and left

    branchings

    simply

    denote

    subor-

    dination

    within

    grouping

    structure.

    Instances

    of

    all

    four

    paradigmatic

    situations

    occur

    in

    Exam-

    131

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    23/62

    Example

    8a.

    Mozart:

    Sonata

    K.

    331,

    I

    8b

    (a)

    tli

    I

    :

    I

    I

    N

    -

    I

    K

    r

    Pr

    p

    r PF

    P

    I)

    r

    pr

    r prF

    (e)

    I

    I-

    -

    W

    L###9

    L1

    P

    r

    rr

    ^

    (b)

    #

    i

    1

    (a)

    #

    s--

    132

    (e)

    (

    :';rr r

    ;

    Lfr

    il(d)

    $

    #

    <

    I

    I

    IjI.

    ?(c)-i

    i

    i

    (b)

    1

    .

    I

    11

    (a)

    0#

    11

    J Ad

    J.

    J J -

    {8tPtI,

    _i m m

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    24/62

    Figure

    5

    (h)

    (i)

    (

    (

    A

    . X

    X

    *

    Figure6

    (9)

    level

    (h)

    =

    [b4

    [c

    [

    [?c

    level(g)

    =

    [b8]

    [Cs]

    measure: 1

    2

    3

    4

    5

    6

    7

    8

    *

    0

    * *

    .

    .

    4

    4

    8

    133

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    25/62

    pie

    9;

    h, i,

    /,

    and

    k in the music stand

    for

    Figure

    5

    (h), (i),

    (j),

    and

    (k),

    respectively.

    (The

    sub-phrase bracketing

    is

    the

    same

    as

    in

    Example

    7.)

    To

    carry

    the

    reductional

    process

    any

    further

    than we

    have

    in Example 9, it is necessary first to develop conceptions of

    the "structural

    beginning"

    and

    the "structural

    ending"

    of a

    group.

    By

    structural

    ending,

    we

    mean

    either

    a

    V-I

    progression

    when

    it

    appears

    at the end

    of a

    group,

    i.e.,

    the

    full

    cadence,

    or

    those

    variations

    on the

    full

    cadence

    known

    as the

    half

    cadence

    and

    the

    deceptive

    cadence.

    The cadence

    is

    designated

    as a

    syntactic

    unit,

    both

    elements

    of which

    are retained at

    the

    appropriate

    levels

    in the

    time-span

    reduction,

    with

    the

    first

    element subordinate to the last.l1

    By

    structural

    beginning

    we

    mean the

    most

    stable

    event

    early

    in

    a

    group

    in

    which

    there

    is a structural

    ending.

    There

    must

    normally

    be

    at least

    one

    intervening

    event

    (the

    "mid-

    dle")

    between

    this stable

    event and the cadence

    if the former

    is

    to be

    designated

    as

    a

    structural

    beginning.

    Thus the

    smallest

    grouping

    levels,

    such as those

    specifying

    motives,

    usually

    do

    not have structural

    beginnings

    and

    endings;

    all

    groups

    from

    the phraselevel on up do have them.

    The

    structural

    beginning

    and the cadence

    of a

    group

    are

    specially

    labeled

    in the

    reduction,

    with

    b

    standing

    for "struc-

    tural

    beginning"

    and

    c for "cadence."

    They

    dominate

    hier-

    archically

    all other

    events

    in a

    group.

    As

    a

    visual

    aid,

    we

    place

    a

    number,

    signifying

    the

    number

    of measures

    spanned,

    within

    each

    grouping

    slur beneath

    the

    music;

    the

    same number

    ap-

    pears

    as

    a

    subscript

    to

    the

    b

    and

    the c

    for each

    group.

    Thus

    in

    Figure 6 each b and each c receives the subscript "4" at level

    (h)

    in the

    reduction,

    since

    they

    begin

    and

    end

    four-measure

    groups.

    However,

    the

    first b and

    the last

    c are also

    the struc-

    tural

    beginning

    and

    the cadence

    for the

    entire

    eight-measure

    group;

    therefore,

    at

    this

    next

    region

    of

    application,

    they

    re-

    ceive the

    subscript

    "8"

    and

    are

    retained

    at level

    (g).

    In effect

    this

    labeling

    creates

    a

    double-layered pitch

    hier-

    archy

    between

    those

    events

    which

    are b's

    and c's and

    those

    which are not. At local levels, all other

    events

    in a

    group

    are

    subordinate

    to

    the

    group's

    structural

    beginning

    and

    ending.

    At

    more

    global

    levels,

    structural

    beginnings

    and

    endings

    are

    subordinate

    or

    dominating

    by

    virtue

    of

    the hierarchic

    struc-

    ture

    of

    the

    groups

    for which

    they

    function.

    At

    the end of

    the

    134

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    26/62

    Example

    9.

    Bach:

    "O

    Haupt"

    /

    \

    /

    \

    /

    o)(\

    Ab

    b

    (a)(a)

    (b)

    (b)

    C;

    c1

    I

    I

    I

    I

    F

    r

    r

    r

    r

    '

    '

    r

    , , ,,,

    r

    J

    J

    J

    J

    J

    J

    ((#

    i

    i

    X

    J-r

    j

    j

    I

    w

    f,

    r Pfrf

    I ij jI

    I

    (b)i

    (a)

    135

    ?

    ?

    ?

    ?

    *

    ? ?

    ?

    ?

    ? ?

    ?

    ?-

    ?

    *

    ?

    ?

    *

    :

    *J

    ?

    .

    ,P

    44

    j j j j

    -

    j

    .

    ( WI----wj 6 - p - p tp FF -1-

    I

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    27/62

    reductional

    process,

    there

    remains

    only

    the structural

    begin-

    ning

    and

    the

    structural

    ending

    for the

    piece

    as a whole.

    We

    are now

    in a

    position

    to

    investigate

    a

    complete

    time-

    span

    reduction.

    Example

    10

    gives

    the

    music12

    and

    Example

    11 represents its analysis. By this point, Example 11 should

    on the whole

    be

    self-explanatory.

    The

    following

    remarks are

    in the

    nature

    of

    annotations.

    1.

    The best

    way

    to "read"

    Example

    11

    is first to examine

    the

    grouping

    and

    metrical

    analyses,

    then

    to hear the

    levels in

    rhythm,

    as notated

    beneath

    the

    music,

    in

    their reductional

    order.

    If the

    analysis

    is

    correct,

    each level should

    sound

    like a

    natural simplification of the previous level. Any difficulties

    which

    the

    reader

    initially

    has

    in

    deciphering

    the

    tree

    can be

    cleared

    up

    by

    a

    step-by-step

    comparison

    with

    the notation

    underneath.

    2. Level

    (f)

    in

    Example

    11 has

    already

    been reduced

    from

    the actual

    music

    (Example

    10)

    to what

    is felt

    to

    be

    its

    small-

    est beat

    level.

    This

    is a

    convention

    which

    we

    always

    follow

    in

    constructing

    the

    time-span

    and

    prolongational

    reductions

    for

    a piece, partly because syntactically unimportant detail is

    thereby

    eliminated,

    and

    also because

    the

    two

    reductions

    do

    not

    significantly

    diverge

    beneath

    this level.

    3. It would

    be

    possible

    to

    assign

    further

    low-level

    grouping

    slurs within

    these

    eight

    measures.

    However,

    since

    the

    indica-

    tions

    for

    these

    groupings

    are somewhat

    conflicting,

    and

    since

    such

    groupings

    would

    not

    affect

    the

    analysis

    as

    a

    whole,

    we

    choose

    not to

    assign

    them.

    Therefore,

    the weak

    beats

    within

    the lowest groupingslurs are simply bracketed as afterbeats.

    4. The

    structural

    beginnings

    and

    cadences

    are labeled

    at

    each

    level,

    but

    they

    do

    not receive

    subscripts

    until the

    reduc-

    tional

    process

    has reached

    the

    grouping

    levels

    for which

    they

    function.

    5.

    The

    selection

    of which

    events

    are

    dominating

    is

    straight-

    forward

    except

    in mm.

    3 and

    7 at

    level

    (d).

    In m. 3 there

    is a

    conflict

    in

    the

    preference

    rules between

    the

    metrically strong-

    er position of the F-sharp-E-A

    chord and

    the

    more

    stable

    structure

    of

    the

    V6

    chord.

    The

    F-sharp-E-A

    chord

    is

    chosen

    for

    reasons

    having

    to do with

    the

    phrase

    as a whole:

    the

    regu-

    larity

    of harmonic

    rhythm

    is

    preserved,

    and

    a

    descending

    line

    in

    the bass

    from

    the

    tonic

    to

    the dominant

    is created.

    The

    136

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    28/62

    Example

    10. Mozart: Sonata

    K.

    331,

    I

    Andante

    grazioso.

    A I

    (ŝa rr%p-r

    f*-*

    a

    (-)

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    29/62

    K,

    ,

    I

    k I K .

    .

    I

    K I

    .

    f,

    P

    rr

    rr

    r

    r

    G

    rPt

    d

    r

    Gr

    r

    'r

    rr,

    r

    P

    ^r2i

    3

    ,r

    rr

    r

    r

    P

    r

    '

    r

    pr

    p

    pC

    r

    cJ

    4

    8

    (e) on>

    I

    I

    i

    ii i

    ,

    I

    i

    .

    i

    I

    i1

    [)

    g

    ~

    J

    2

    _

    2

    '

    tb 2 2 2 2t

    c

    J

    (C)

    *'Io

    |I,

    1,

    ;

    B

    114)

    (C

    1^^

    C4

    i

    4

    4

    4

    ca

    ,

    aYI

    8c

    c

    -

    ~~~~~~~(a)

    ,a>ll D

    I b 3

    r4

    C

    138

    Example

    11

    n.,

    A I

    k I

    K

    I

    I

    I

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    30/62

    same

    chord is

    chosen

    in m.

    7

    because

    of

    the

    parallelism

    with

    m.

    3.

    6. Note that if the

    cadences were

    not labeled and

    retained,

    in

    m.

    4 the

    I

    chord instead of the

    V

    chord would have

    been

    selected at level (d). The labeling also causes the retention at

    levels

    (d),

    (c),

    and

    (b)

    of

    both elements

    of

    the

    full

    cadence

    in

    m. 8.

    7. Levels

    (c)

    and

    (b)

    can

    be construed

    as

    Schenkerian

    "background"

    evels.

    Schenker's own

    analysis

    of

    this

    passage13

    gives

    the

    melodic structural

    weight

    to

    the

    E

    in

    the

    first

    mea-

    sure rather

    than to

    the

    opening

    C-sharp.

    In

    our

    theory,

    it

    would

    be

    possible

    to achieve his

    result

    by

    applying

    the

    "arpeg-

    giation rule," i.e., by regardingthe first measure as a broken

    chord

    and therefore

    compressing

    it

    into one

    event

    placed

    on

    the

    downbeat

    of

    the measure.

    However,

    such

    a

    decision

    pro-

    duces difficulties

    at

    intermediate

    structural levels

    (especially

    level

    (d)).

    8.

    In

    the

    large

    levels

    (b)

    and

    (a),

    the sense of

    metrical

    struc-

    ture

    has

    faded to the

    point

    that it is

    largely inoperative.

    The

    events

    at these

    levels in the

    notation

    beneath the

    music

    do

    not receive rhythmic values because there are no more dots in

    the metrical

    analysis

    to which

    such

    values

    could be

    related.

    It

    is at

    this

    point

    in

    the

    reductional

    process

    that

    pitch

    events

    can

    be

    thought

    of as

    rhythmically

    free-floating.

    9.

    It

    would

    perhaps

    be

    sufficient

    to

    stop

    the

    reductional

    process

    at level

    (b).

    Level

    (a)

    simply

    selects

    the

    most

    stable

    event

    from

    those

    available at level

    (b).

    As is

    often

    the

    case,

    the

    most

    stable event here

    is the

    last

    chord; thus,

    the

    rest of

    the piece is a left branch to it. This situation can be inter-

    preted

    to

    mean

    that,

    in a

    sense,

    all

    the

    events of

    a

    piece

    except

    the

    last

    constitute a

    delay

    of

    the

    moment of

    complete

    repose,

    which

    is

    the

    ending.

    If

    the

    point

    of

    maximal

    stability

    happened

    in

    the

    middle

    of

    a

    piece,

    there

    might

    be

    no

    reason

    for

    the

    piece

    to

    continue.

    10. We

    have

    considered these

    eight

    measures

    as

    if

    they

    were

    a

    complete

    piece.

    If

    the

    entire

    theme,

    which is

    18

    measures

    long,

    were

    analyzed,

    the

    opening event would eventually be

    labeled

    b18

    and thus

    would

    dominate all

    other

    events in

    the

    first

    eight

    measures.

    11.

    Observe

    how the

    geometry

    of

    the

    tree,

    while

    accurately

    conveying

    the

    hierarchy

    of

    pitch

    events,

    also

    mirrors

    visually

    139

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    31/62

    the

    partitioning

    of the

    piece

    into

    time-spans

    (groupings

    and

    bracketings).

    Moreover,

    the tree

    notation makes

    visually

    clear

    the interaction of the

    pitch

    structure

    with

    the

    metrical struc-

    ture at all

    pertinent

    levels.

    There

    are

    important

    hierarchic

    aspects

    to

    the

    pitch

    struc-

    ture

    in

    Example

    10 which do not

    emerge

    in the

    time-span

    reduction

    in

    Example

    11. To

    capture

    these

    aspects,

    we must

    develop

    the

    domain

    of

    analysis

    called

    the

    prolongational

    re-

    duction.

    In the

    parameter

    of

    rhythm

    in this

    domain,

    events

    come before or after other

    events,

    but

    they

    are not

    measured

    according

    to some metrical

    conception.

    The

    relevant

    notion

    of elaboration is elaboration by harmonic and melodic con-

    nection.

    There are two

    kinds of elaboration

    in the

    prolongational

    reduction:

    prolongation,

    in

    which

    a

    pitch

    event

    is

    elaborated

    into

    two

    or

    more

    copies

    of

    itself;'4

    and

    contrast,

    in

    which

    a

    different,

    relatively

    ornamental,

    pitch

    event is

    introduced.

    A

    prolongation

    is

    represented

    by

    two branches

    extending

    from

    a

    small circular

    node,

    as in

    Figure

    7a. Neither event takes hier-

    archic priority; rather,both events are thought of as an exten-

    sion over

    time of what

    at a

    deeper

    level is the same event.

    In

    contrast,

    hierarchic subordination

    is

    designated by

    right

    and

    left

    branching,

    as in the

    time-span

    reduction.

    However,

    where-

    as

    in the

    time-span

    reduction

    these

    branchings

    only

    indicate

    the

    subordination

    of

    one event to

    another,

    in

    the

    prolonga-

    tional reduction

    they

    receive

    a further

    interpretation.

    A

    right

    branch

    signifies progression

    in a

    piece,

    whether

    at

    a

    local

    level

    as in Figure 7b or at a large structural level. A left branch is

    utilized

    only

    at local

    levels and

    signifies

    delay

    in

    relation

    to

    the

    bass,

    as

    at level

    (c)

    in

    Figure

    7c.

    (Since

    metrical

    structure

    does

    not

    play

    a

    role in

    the

    prolongational

    reduction,

    Figure

    7c

    represents

    a

    suspension

    considered

    only

    with

    respect

    to its

    pitch

    structure.)

    The well-formedness

    conditions

    and

    the

    preference

    rules for

    the

    prolongational

    reduction

    are

    similar

    to

    those for

    the

    time-

    span reduction.

    The well-formedness

    conditions

    preclude

    the

    crossing

    of branches

    and

    the

    assignment

    of more

    than

    one

    branch to

    the

    same event.

    Given a

    sequence

    of

    pitch

    events,

    the

    preference

    rules

    select

    a

    hierarchy

    according

    to

    principles

    of

    140

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    32/62

    Figure

    7a

    (a)

    (IM

    --

    _

    -

    U

    *--

    (a)

    (b)

    (b)

    J>I^vwv-Y

    I

    -

    I

    ||

    [ba]

    [C4]

    [b,]

    [cs]

    measure: 1 2 3 4

    5

    6 7 8

    time-span

    reduction

    time-span

    reduction

    (a)

    >

    (a)

    (b)

    )

    (b)

    (b)

    >

    (c)

    >

    (c

    (C)

    )

    (C

    I- s- ^-y

    I-

    ^^

    --1[

    1 2 3 4 5 6 7 8

    prolongational

    reduction

    7b

    7c

    (a)

    (b

    O

    1

    (a)

    (b)

    I

    *?

    Figure

    8

    141

    i .

    1i=

    -I1dn

    .

    f

    Ii

    11

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    33/62

    stability largely

    resembling

    those

    for the

    time-span

    reduction.

    Meter is

    disregarded;

    prolongations

    are

    maximized.

    At

    any given

    level down

    to the

    phrase

    level,

    the

    sequence

    of

    events

    available

    to

    the

    preference

    rules

    for the

    prolonga-

    tional

    reduction

    is

    precisely

    the same

    sequence

    as at the

    equivalent

    level in the

    time-span

    reduction.

    In other

    words,

    the

    hierarchy

    of

    b's

    and

    c's as

    designated

    in the

    time-span

    reduction

    heavily

    determines

    the

    prolongational

    reduction.

    Again,

    we illustrate

    with

    a

    typical

    antecedent-consequent

    pattern.

    In the

    time-span

    reduction

    in

    Figure

    8,

    levels

    (a),

    (b),

    and

    (c)

    refer

    to events

    labeled

    as

    structural

    beginnings

    and

    endings;

    these events

    at

    equivalent

    levels become

    the material

    for the prolongational reduction in Figure 8.

    Even when

    precisely

    the

    same events are

    available

    at a

    given

    level

    for

    the two

    kinds

    of

    reduction,

    the reductions

    draw

    radically

    different

    connections

    among

    these events.

    For in-

    stance,

    in

    Figure

    8

    the

    V chord

    in the

    full cadence

    in m.

    8 is

    a

    left branch

    to the

    ensuing

    I

    chord

    in the

    time-span

    reduction,

    but

    is a

    right

    branch

    from

    a

    prolonged

    I

    chord

    in the

    prolon-

    gational

    reduction.

    In the

    former

    case,

    it

    is

    a

    left branch

    be-

    cause it is within the time span of the final tonic; in the latter

    case,

    it is a

    right

    branch

    because

    it

    progresses

    to

    the

    final

    tonic.

    (In

    both

    cases,

    it

    is subordinate

    to

    the

    tonic

    according

    to

    principles

    of

    stability.)

    Large-scale

    right

    branching

    in

    the

    prolongational

    reduction

    always

    indicates

    significant

    syntactic

    "progress"

    n a

    piece.

    Note, too,

    that

    at

    level

    (c)

    in

    Figure

    8

    the

    prolongational

    reduction

    brings

    out

    connections

    of harmonic

    identity

    not

    captured in the time-span reduction. The tree for the time-

    span

    reduction

    would

    look

    the

    same even

    if the

    b

    for

    the

    consequent

    phrase

    were

    an

    entirely

    different

    chord;

    in the

    prolongational

    reduction,

    however,

    such

    a

    change

    would

    pro-

    duce

    a

    right

    branch

    at

    level

    (c)

    instead

    of the

    prolongation

    represented

    there.

    On

    the

    other

    hand,

    the

    tree

    for

    the time-

    span

    reduction

    expresses

    grouping

    structure, something

    not

    conveyed

    in the

    prolongational

    reduction

    itself.

    Thus,

    for

    example, the tree for the prolongational reduction cannot

    differentiate

    between

    the

    full cadence

    in m.

    8,

    and

    the

    phrase-

    ending

    half cadence

    followed

    by

    the

    phrase-beginning

    tonic

    chord in

    mm. 4-5.

    142

  • 8/17/2019 Jackendoff Toward a Formal Theory Tonal Music

    34/62

    From the

    phrase

    level

    down

    to the smallest beat

    level-that

    is,

    for all events

    which

    are not

    b's and

    c's

    except

    for the

    most

    local details-the

    prolongational

    reduction

    is

    constructed

    without

    reference to

    the

    time-span

    reduction.

    As

    a

    result,

    within this

    region

    the two reductions

    usually

    differ not

    only

    in the connections

    made but

    in the actual

    sequences

    of

    events

    at

    corresponding

    levels.

    To facilitate

    reading

    the

    prolongational

    reduction,

    we

    give,

    beneath the music in a

    non-rhythmic

    notation,

    the

    actual

    sequence

    of events for each level from

    the

    phrase

    on

    down.

    (See

    Example

    12.)

    Unlike the

    tree, however,

    this

    secondary

    notation

    does

    not

    convey

    the

    types

    of

    elaboration within a

    sequence of events.15For each level from the phraseon up, it

    is

    sufficient to mark the events

    which,

    as

    b's

    and c's

    in

    the

    corresponding time-span

    reduction,

    have

    caused the

    equiva-

    lent

    sequence

    of events in the

    prolongational

    reduction;

    this

    marking

    is

    accomplished

    by

    placing

    the

    corresponding

    letters

    in

    parentheses

    at the

    appropriate places

    just

    below the

    music.

    Example

    12

    represents

    the

    prolongational

    reduction

    for

    the first

    eight

    measures of

    Mozart's

    K.

    331.

    Remarks:

    1. The most local level in the

    prolongational

    reduction

    (level

    (h))

    corresponds

    with

    the

    "lowest

    beat level"

    as

    deter-

    mined

    in

    the

    time-span

    reduction

    (Example

    11,

    level

    (f)).

    In

    placing

    this

    lower

    boundary

    on

    the

    prolongational reduction,

    we are

    in

    effect

    claiming

    that

    beyond

    this

    level

    local detail

    is

    not

    of

    prolongational

    significance.

    2. Level

    (a)

    represents

    the level

    of

    abstraction

    at

    which the

    A

    major

    root

    position

    triad is

    totally

    unelaborated. The

    high-

    est level of the prolongational reduction for any classical tonal

    piece always

    results in

    an

    undifferentia