dowling 1978 - scale and contour

14
7/26/2019 Dowling 1978 - Scale and Contour http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 1/14 Psychological  Review 1978,  Vol. 85, No. 4, 341-354 Scale  and  Contour:  Two  Components  of a  Theory of  Memory  for  Melodies W.  Ja y  Bowling University  of  Texas  at  Dallas This article  develops a two-component  model  of how melodies are stored in long- and short-term  memory.  The first  component  is the  overlearned perceptual-motor schema of the musical scale. Evidence  is  presented  supporting the  lifetime sta- bility of scales and the  fact  that  they  seem  to  have  a  basically logarithmic  form cross-culturally.  The second component, melodic contour, is  shown  to  function independently of pitch interval  sequence in memory. A new experiment is re- ported, using  a  recognition  memory paradigm in  which tonal standard stimuli are confused with  same-contour  comparisons, whether  they  are  exact  transpositions or tonal answers,  but not  with  atonal comparison  stimuli.  This  result  is  con- trasted with  earlier  work  using  atonal  melodies  and  shows  the  interdependence of  the two components, scale and  contour. Remembering melodies  is a  basic  process  in the music  behavior  of  people  in all  cultures. This behavior  may  involve production,  as with  th e  singer  performing  a  song for an audi- ence  or the  participant  in a  significant social event trying  to  remember  th e  appropriate song.  Or it may involve recognition, as with th e  listener whose comprehension  of the  later developments in a piece depends on memory for  earlier parts.  In  this  article,  I  concentrate on  two  components  of  memory that  contribute to the  reproduction  and  recognition  of  melo- dies, namely, melodic contour and the musical scale.  I  maintain that actual melodies, heard Earlier  versions of  this article  were  presented at the Music Educators National  Conference,  Western Division,  San  Francisco,  1975,  and the  meeting  of New  American  Music, Tangents III, University of Wyoming,  1977. Support  for the  experiment reported here  was provided in  part  by a Biomedical Sciences Support Grant  to the  University  of  Texas  at  Dallas from  th e  National Institutes  of  Health. I  thank Dane Harwood, Darlene Smith,  James Bartlett, Drew  Harare,  Mark Goedecke, Cynthia Null,  Kelyn Roberts, and Klaus Wachsmann for help- ful  discussions and comments. And I especially wish to thank Mantle Hood for bringing the  data  on Indonesian tunings to my attention. Requests  for reprints  should  be  sent  to W. Jay Dowling,  Program  in  Psychology  and  Human  De- velopment, University of Texas at Dallas, Richardson, Texas 75080. or sung, are the product of two kinds of un- derlying  schemata. First, there  is the  melodic contour—the  pattern  of ups and  downs—that characterizes  a  particular melody. Second, there  is the  overlearned musical scale  to  which th e  contour  is  applied  and  that  underlies many different melodies.  It is as though  th e  scale constituted a ladder or  framework  on which the ups and  downs of the  contour  were  hung. Two examples epitomize the behavior this theory addresses. First,  if  people  in our  West- ern  European culture hear a melody  from some  non-Western culture using  a  non-West- ern  scale, their reproductions  of  that  melody will  use  their  own  Western scale, preserving the contour of the original melody. The scale functions  as a  classic example  of a  sensori- motor schema controlling perception and be- havior.  In  this example,  th e  non-Western melody  is  assimilated  to the  Western schema (Frances, 19S8,  p.  49).  Part  of the  education of  ethnomusicologists  is  directed toward  free- ing  them from  their native schemata  so  that they can hear accurately the  pitches  of non- Western music. Th e  second example involves  the use of melodic  contour  in the  structure  of  pieces  of music.  One way a  composer  can tie a  piece together  is to repeat  th e  same contour  at  dif- ferent  pitch levels, at  different  relative place- Copyright 1978  by the  American  Psychological Association, Inc.  0033-295X/78/8504-0341$00.75 341

Upload: jennmjo

Post on 02-Mar-2018

221 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 1/14

Psychological

  R e v i e w

1978,  V ol. 85, No. 4, 341-354

Scale  and  Contour:  Two  Components  o f a  Theory

of  Memory  fo r  Melodies

W . Ja y  Bowling

University  of

  Texas

  at  Dallas

This article

  develops a two-component

 model

 of how melodies are stored in long-

and short-term

 memory.

 The first component is the  overlearned perceptual-motor

schema

  of the musical scale.

  Evidence

  is

  presented

  supporting the

  l i fe t ime s ta-

bility of

  scales

  and the  f a c t

  that  they

  seem  to

 have

  a

 basically

  logarithmic  f o r m

cross-culturally.  The second component, melodic contour, is

  shown

  to

  funct ion

independently of

  pitch interval

  sequence in memory. A new experiment is re-

ported,  using  a recognition memory paradigm in  whi ch  tonal standard stimuli are

confused  with  s a me - c o n t o u r  comparisons, whether  they  are  exact  transpositions

or tonal

  answers,

  but not

  with

  atonal comparison

  stimuli.

  This

  result

  is

  con-

trasted  with  earlier  work  using  atonal  melodies  and

  shows

  the

  interdependence

o f  the two components, scale and

  contour.

Remembering melodies  is a basic process  in

the mus ic

  be hav io r

  o f

  people

  in a l l

  cul tures.

This  be hav io r

  m ay

  involve product ion,

  a s

with

  th e

  singer performing

  a

  song

 for an

  audi -

ence  or the  part ic ipant  in a  s ignif icant socia l

event  try ing

  to

  remember

  th e  appropriate

song.  Or i t may involve recognit ion, as with

th e

  l i s tener whose comprehension

  of the

  later

developm ents in a piece depends on mem o ry

f o r  earlier parts.  In  this  art ic le ,  I  concentrate

on  tw o  components o f  me mory

 that

 co nt ribute

to the

  reproduct ion

  and

  recognit ion

  o f

  melo-

dies , nam ely, melodic contour a nd the music al

scale.  I  maintain tha t ac tual melodies , hea rd

Earlier

  versions of  this

  article  were

  presented at

t he Mus i c Educa to r s Na t i ona l

  Conference ,

  Western

Divis ion,

  San  Franc i sco ,

  1975,

  and the

  meeting

  o f

N ew   Amer i c an  Music, Tangents III , Univers i ty of

W y o m i n g ,  1977. Suppo r t  for the  experiment reported

here  was provided in  part  by a Biomedical Sciences

Suppor t Grant  to the  University  o f  Texas  a t  Dallas

f rom  th e  Nat iona l Ins t i tutes  o f  Heal th .

I  thank Dane Harwood, Dar lene Smith ,  James

Bartlett , Drew  Hara re ,  Mark Goedecke , Cynthia

Nul l ,  Kelyn Rober ts , and Klaus Wachsmann for he lp-

ful  discuss ions and co mm ents. And I especial ly wish

to t ha nk Mant l e Hoo d fo r b ring ing t he  data  o n

Indonesian tunings

  t o my

  at tent ion.

Requests

  for reprints

  s hou ld

  be

  sent

  to W. Jay

Dowl ing ,

  P r o g r a m  in  P s y c h o l o g y  a nd  H u ma n  De-

velopment, Univers i ty of Texas at Dal las , Richardson,

Texas 75080.

or sung, are the product of two kinds of un-

derlying   schemata. Firs t , there

  is the

 melodic

c o n to u r — th e  pat tern

  of ups and

  downs—tha t

character izes   a  part icular melody. Second,

there  is the overlearned music al scale  to  which

th e co ntour

 is

 a pplied

  a nd

 that underlies many

different

  melodies.

  It is as

  though

  th e

  scale

constituted a ladder or

  framework

  on which

the ups and

  downs

  of the

  contour

  were

  hung .

Two examples epi tom ize the behavio r th is

theory addresses. First ,

  if

  people

  in our

  West-

ern

  European c ul ture hear a m elody

  from

some  non-Western culture using  a  non-West-

ern  scale , the ir reproduct ions  o f  that  melody

will

  use

  the ir

  o wn

  Western scale, preserving

the contour of the original melody. The scale

functions

  a s a

  classic example

  o f a

  sensori-

m o to r  schema control l ing percept ion

  and be-

hav ior .  In  this example,  th e  non-Western

melody   is  assimilated  to the  Western schema

(Frances ,  19S8,

  p.

  4 9 ) . Part

  of the

  education

of  ethnomusicologis t s  is  directed toward

  free-

ing

  them

  from

  the ir nat ive schemata

  s o

  that

they can hear accurate ly the  pitches  of non-

Western music .

Th e  second example involves  the use of

melodic

  contour

  in the

  structure

  o f

  pieces

  o f

music.  One way a  composer  can tie a  piece

together  is to

  repeat

  th e

  same contour

  a t

  dif-

ferent  pitch levels, at  different  relative place-

Copyright 1978  by the  Amer ican  Psychological Association, Inc.  0033-295X/78/8504-0341$00.75

341

Page 2: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 2/14

342

W. JAY

  D O W L I N G

 

r

  i f

  f

  i

4 5

  -I

t A

V

  '  ^

4-5

  -3

  4-5

57

 9 ̂ ifr  J

  ( I

  ?

 

P

 

\

  1 3

4 5

  -I  4 5

4-5

  -4  4-5

4 -5

  -I  4 - 6

4-5  -I  4 -7

 

r-

 

±

i

=N=

L

i

 mj

  2  -2-2

t>

2

 _^

J

 

o

-3

 

 1

 

Figure  J.  Sect ion  A  shows example s

  from

  Bee thoven 's P i ano Sona ta opus  14 no. 1, i l lu s t r a t ing

use of the

  s ame me lod i c co n tour w ith  different  in terval s izes . (In tervals between notes

  a re

  s h o w n

be low the  s t aves  in  semi tones . Excerp t s  a re  labeled with measure

  number s . )

  Sec t ion  B i s an

Amer ican  Indian example from  Kol insk i  (1970,  p.

  9 1 ) .

merits

  on the  sca le ,  o r  even  o n

  different

  scales.

Th e  repet it i on p rov ides un i ty wi tho u t beco m-

ing

  b o r i ng t h r ou g h b e i ng

  to o

  exac t . F i gure  1A

demonstra tes Beethoven ' s

  use of

  th i s device

in

  h is

  P i ano Sona ta opus

  14 no . 1 .

  Such

  a

device relies  on the  listener's

  ability

  to  recog-

n ize the melod i c con tour th r ough t r ansforma-

t ions

  o f

  p i t ch .

  That

  such processes

  a re no t

confined

  to

  Western music

  is

  s hown

  by  Kol in-

ski's

  (1970) example

  from

  the

  Flathead

  In-

d ian s (s ee F i gu r e I B ) . A d am s

  ( 1 9 7 6 )

  h a s

provided

  a

  gu ide

  to the

  uses

  o f

  con tour con-

ceptual izat ions   in  e t hn om u s i c o l ogy .

In

  w h a t

  fol lows,  I

  will  di scuss

  (a)

  scales

of  pi tch

  a nd

  thei r character i s t i cs

  and (b)

melod i c con tour s

  a nd

  their independence

  in

m e m o r y .  (Independence  is

  here used

  in the

sense of  referr ing  to  cases  where one remem-

bers

 one

  th ing wi thou t r emember ing

 a

  related

th ing . )  I  will  also  present  a new  exper iment

that

  i l l u s t r a tes per fo rmance

  in

  m em or y

  fo r

melo dies co nsistent with  th e  present two-

component theory .

Mus i ca l

  Scales

Several aspects

  o f

  musica l sca les

  a s

  they

func t i on  in var ious cu l tures are of par t i cu lar

interest

  to the

  psycho l og i s t . F i r s t ,

  scales

  con -

sist  of d i screte s teps of p i tch , typ ica l ly f ive

o r  seven

  to the

  oc t ave .  This  l imi t a t i on wou ld

seem

  to

  match qu i te wel l

  th e

  l imi t a t i ons

  o f

people

  a s

  informat i on p rocesso r s . Second ,

musica l sca les   of  p i t ch a r e approx imate l y

l o g a r i t hm i c w i t h r e sp ec t  to  f requency ,  a  fa c t

t ha t bear s d i r ec t l y

  on the

  prob lem

  o f

  pi tch

sca l ing

  in

  psychophys i c s . Th i rd , mus i ca l s ca l es

ca n be s l id co nt inuously up and do wn in p i tch

wi thou t d i s to r t ing the i r r e l a t i ve in terva l s i zes .

They serve  a s  h i gh l y s t ab l e

  senso r imo to r

schemata wi th great f lex ib i l i ty of appl i cat ion.

Discrete Scale Steps

W i th   fe w  excep t i ons ,  th e  cul tures  of the

world  use

  d i sc re te changes

  o f

  p i t ch a l ong

  a

mus i ca l s ca l e

  in

  creat ing thei r melodies . Since

th e  s ing ing vo i ce  a nd  many ea r l y in s t ruments

a re capab le o f p roduc ing a con t inuous var i a -

t ion of p i tch , how i s i t

  that

  the mus i c o f t he

world

  moves

  by

  di screte s teps? Helmh ol tz

( 1 9 5 4 )

  ra i sed th i s quest ion ,

  a nd

  some

  of the

answers

  he

  suggested

  are

  appeal ing even

  to-

d ay .

  Th e

  u l t im ate b io l og i ca l  func t ions

  o f

  dis-

crete steps

  are as

  ob s cu r e

  a s the

  func t ions

  o f

melody or of music   i t se l f .  However , g iven that

Page 3: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 3/14

M E M O R Y

  FOR

  MELODIES

343

we

  have melody, discrete scale  steps

  can be

very useful.  Helmhol tz argued

  that

  the di-

vision of the

  pitch continuum into discrete

steps   is  desirable  in  order  t h a t  th e  degree  o f

melodic   and rhy thmic movement migh t be

immediately apparent

  to the

  l istener.

  If

  pitch

moved

  con t inuous ly

  in a

  melody,

  h e

  argued,

i t  would  be

  mu ch more  difficult

  to

  decide

  ex-

act ly   h o w  mu ch  th e  pitch  h a d  changed  o r

what rhythmic , tempora l uni t s

  h a d  elapsed.

In  fact ,

  such

  a

  decision could

  be

  made only

o n

  the basis of a fixed set of discrete

  refer-

ence points anyway.  That  is , the  l istener

would  have

  to

  learn some

  form  o f

  scale

  a s a

cogni t ive  framework

  t h r ou gh w h i ch

  to

  deal

with

  th e melo dies  h e  hears , cont inuous  o r  not.

Th e  modern cogni t ive psychologis t  ca n

eas i ly carry Helmhol tz ' s argument

  o ne  step

fur ther .  Grant ing  t h a t  we  need discrete scale

steps  to

  comprehend melodic movement ,

  h o w

many should there be? Mil ler ' s

  ( 1 9 5 6 )

  classic

ar t ic le suggested that human beings  a re  l im-

ited in their categorical judgment capabil i ty

to

  using

  7 ± 2

  categories

  on any one

  dimen-

s ion. Most cu l tures

  fall

  with in th i s  range, us-

ing f ive or seven dist inct scale step categories

wi th in

  th e

  octave.

  A s

  will

  be

  discussed below,

the system of categories repeats cycl ical ly

every octave.  Thus ,  in  Miller 's terms,  a  pitch

would  be  categor ized  in  terms  of two  dimen-

sions:

  one for

  pi tch  level within

  th e

  octave

(of ten  called

  chroma)

  and the

  o ther

  for oc-

tave level. Certain cultures  t h a t  would  seem

at f irst

  s i g h t

  to

  provide exceptions

  to

  this

pat tern turn

  out on

  closer inspect ion

  not to.

For example, the mus ic of India typical ly

uses many very narrow intervals  o f  which  22

a re   avai lab le with in

  a n

  oc tave. However,

melodies

  are basical ly constructed out of

scales having seven

  or so  foca l

  pitches,

  and

the other neighboring tones

  a re

  used

  a s

  auxil-

iary tones around these  focal  pitches for the

purposes  o f  ornamenta t i on .

A t  this point ,

  I

  need

  to

  in t roduce

  a

  con-

ceptual

  scheme  fo r

  talking

  about musica l

scales.  This  scheme  h a s  four  levels o f abstrac-

t ion

  from

  the pitches of actual melodies and

shares cer ta in

  features with  th e

  conceptual

schemes  o f Hood

  ( 1 9 7 1 )

  a nd  Deutsch

  ( 1 9 7 7 ) .

I  have discussed  it in  greater detai l  in an-

other paper

  (Bowl ing ,

  No t e

  1). The

  mos t

abstract level is  t h a t  of the  psychophys ica l

scale,

  which

  is the general

  rule

  system  by

which

  pi tch intervals

  a re

  related

  to  frequency

in tervals

  o f

  tones.

  I  will

  a r gu e  t h a t

  the ap-

propr ia te  form  for the  psych ophy s ica l sca le

is   approx imate ly l oga r i t hmic , owing t o t he

fundamenta l

  place  o f  octave judgments  in hu-

man audi tory process ing.  Th e  second level  is

that

  of tonal mater ia l , which inc ludes the set

o f

  pitch intervals in use by a part ic ula r cul-

ture

  o r

  wi th in

  a

  part icu lar genre.

  In

  Western

music ,  the tonal material  would  include the

set of

  semitone intervals

  of the

  equal tem-

pered chromat ic sca le

  but not

  anchored

  to any

part icu lar  frequency.

  Th e

  third level

  is

  t h a t

o f

  th e

  tuning

  sys tem, which cons is t s

  of a se-

lect ion of a subset of the available pitch in-

tervals  from  the tonal mater ia l tha t are used

in

  actua l melodies .

 In

  Western music ,

  th e

  tun-

ing

  sys tem might cons is t

  of the set of

  intervals

represented by the white notes of the

  p iano—

the set of

  intervals with

  th e

  ascending cyc le

(in

  semitones)

  of [2, 2, 1, 2, 2, 2, 1]

  repeat-

ing

  every octave.

  Th i s

  set of intervals can

become the

  bas is

  of any of the

  heptatonic

modes (that is , the modes with seven notes

per

  o c t a v e ) :  m ajo r , minor , or the m edieval

church modes.

  The set of

  intervals

  of the

black notes on the p iano [2 , 3 , 2 , 2 ,

  3 ]

  c a n

also

  funct ion  a s a

  tuning sys tem

  for the

  West-

ern  pentatonic modes.

Th e  most concrete level  is  mode.  In  going

from  tuning sys tem to mode, two th ings hap-

pen.

  Firs t ,

  an anchor for the

  frequencies

  is

established,

  a nd

  what were pitch intervals

  a t

th e  m o re abst ract levels  g et  transla ted into

the pitches of notes. Second, a tonal

  focus

  in

th e

  tuning sys tem

  is

  selected,

  and a

  tonal

h iera rchy

  is

  established

  on the

  tuning sys-

tem.  (The tonal h i erarc hy determines which

notes

  in the  mod e  a re  more impor t an t  a nd

dynamic tendencies of the notes as they

  func-

t i on

  in

  melodies .)  These

  tw o

  operat ions

  a re

usual ly

  combined in Western music under the

rubric

  o f

  selecting

  a  tonality, fo r

  example,

tak ing the intervals of the whi te-note tuning

system

  and

  select ing

  the key of

  C-major ,

  A -

minor ,  Eb-major ,

  o r

  C-dorian. Since

  mode

  in

m y

  sys tem over laps cons iderably with

  th e

term   scale  as used in the phrases

  musical

  scale

a nd  diatonic scale,  I

  will

  often  use scale in

Page 4: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 4/14

344

W .

  JA Y  DOWLING

w h a t  follows  as a synonym for mode . I want

to

  caut ion

  th e

  reader, however,

  that

  scale

  is

also used in ways that are not synonymous

with   my use of  mode,  fo r  example,  in the

phrase  chromatic scale,  which i s more near ly

synonymous

  with

  m y  term  tonal material.  M y

use of scale is   informa l ;  a s t r ic t ly

  formal

  sys-

tem  would  use  only mode .  I use the  term

tonal

  to

  refer

  to

  melodies us ing

  th e

  notes

  o f

well-known

  modes.  And I  shou ld caut ion  th e

reader

  t h a t w h a t

  I

  h a v e

  to s ay

  here

  is

  most ly

i r re levant  to the dispute among proponents

o f  tempered, Py thag orean, and other natu-

ra l tuning systems wel l discussed  by  Ward

( 1 9 7 0 ) .

Logarithmic Scale

Th e  musical sca les  o f  most cul tures  repeat

themselves cycl ica l ly every octave.

  Tones

  an

oc t ave apa r t  a re  treated  a s  equivalent  in some

sense and are

  typica l ly

  given the same name

(C, D, E; do, re , mi; or

  B a r a n g ,

  Sulu,

  Data

[ I ndones i an ] ) .  Tones  t ha t  a re  psycho log i -

cal ly   a nd  mus ica l ly  an  oc t ave  a p a r t  have  a

rat io between their fundamental frequencies

of  approx imate ly 2/1. (The approxima te ly

will  be discussed below.) Ward (1954) has

shown

  that Westerners

  a re

  quite precise

  in

lOOr

so

 

60

sett ing two pure tones (presented alternately)

to the

  subjec t ive in te rva l

  of an

  oc tave .

  This

is   a ll

  t h a t

  is

  required

 to

 produce

 a

  logar i thmic

psychophys ica l sca le

  o f

 p i tch :

  Tones  a t

  equal

dis tances a long

  th e

  subjective  pitch  scale

(namely ,  an  oc tave apar t ) shou ld  be  related

by   equa l 'frequency  rat ios .

Complicat ions ar ise because the subject ive

octave usual ly represents

  a  frequency  rat io

j u s t  s l igh t ly grea te r than

  2/1.

 Ward (1954 ;

replicated

  by

  Burns, 1974, us ing non-Western

subjects

  and by

  othe rs [Sundberg

  &

  Lind-

qvist , 1973] us ing Westerners)  found  that

t h rough

  th e

  midrange

  of up to

  frequencies

of

  abou t

  900 Hz, the

  frequency rat io

  o f a

subject ive, pure-tone octave  wa s abou t

  2.02/1 .

Th i s

  ratio st i l l leads

  to an

  approximate ly

l oga r i t hmic

  scale.

  Ward's

  results

  a re

  plot ted

in the

  cu rve

  o f

  Figure

  2 ,  which

  shows

  th e

tuning

  of successive oc taves  as cumulat ive de-

viat ions

  from  the 2/1  frequency

  ratio

  in

logar i thmic uni t s (hundredths  o f a  semitone,

o r

  cents) . (A semitone represents a

  frequency

rat io  o f

  2

1/12

/1).

  Octaves having  a 2/1  fre-

quency rat io would  be  represented  in  Figure

2

  as a  horizontal l ine . Stra ight segments  o f

the curve with posit ive slope represent loga-

r i t hmic

  pi tch scales with

  a

  greater than

  2 /1

frequency

  rat io

  to the

  octave.

3

40

2 0

o

-10

25

j 5

IOOO

2

F R E Q U E N C Y

  OF

 UPPER NO TE   Hz)

Figure  2.

  Cum ula t ive tuning devia t ions o f oc taves in cents (h undredths of a semi tone) as a

funct ion  of the

  fundamenta l f requency

  of the

  upper note

  in the

  octave

  fo r

  Burmese ha rp

 (Xs;

Wil l i amson

  &

  Will iam so n, 1968)

  and for

  Indones ian gamelans us ing  sUndro  (circles)

  and  pilog

(tr iangles) tuning systems, (The Indonesian  data are based  on  H o o d  [1966; filled  s ymbo l s ;  = 2]

a nd   Sur jodiningra t , Sudar jana ,  & Susanto [1969; open sym bols ;  J V s =l l and 10 for  sUndro  a nd

ptlog,  respect ive ly ] . The  solid

  line

  is  based  o n Ward's  [1954]  data  fo r  pure- tone  octave  judgments . )

Page 5: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 5/14

M E M O R Y   FOR MELODIES

345

Figure

  2

  also shows some measurements

  o f

non-Western instrument tunings: a se t

  from

a   Burmese h arp (Wil l iamson  &  Wil l iamson,

1968)

  and  several

 sets

  from

  Indonesian

  game-

l ans (Hood, 1966;  Surjodiningra t ,  Sudarjana,

&  Susanto,  1969) .  Gamelan  refers

  to

  both

  th e

orchestra  and the set of  instruments , con-

sist ing

  chief ly

  o f  gongs  and  mar imbas .  Th e

correspondence  of the  tunings  and  Ward's

curve  is very good, especial ly considering the

sorts  o f  systemat ic deviat ions within  the oc-

tave  in gamelan tunings discussed by Hood

(1966 ) . One reaso n  t h a t  th e  fundamental f re-

quencies of  gamelan tunings  a re  free  to ap-

prox imate  th e  tunings o f  s ubject ive octaves  is

that there is not the same

  a t t empt

  a s

  found

in  th e

  West

  to

  make fundamentals

 o f

  h i g h e r

tones match upper part ia ls (overtones)

  o f

lower

  tones.  One  reason  fo r  this  is  tha t acous-

t ica l ly gongs have very i rregular ser ies   o f

upper part ials.

Western pianos

  a re

  tuned

  to

  octaves  that

a re

  s l ight ly greater than  2 /1  ratios,  but the

deviation  is  smal ler than  that  fo r  subjective

pure-tone  oc taves (Mart in  &  Wa rd, 1961).

The reason for the deviation is that piano

strings are no t ideal, frict io nless , vibra ting

bodies

  but

  have

  a

  certa in amount

  o f

  stiffness.

This

 stiffness

  is  most pronounced  at the  upper

and

  lower extremes

  of the

  keyboard, where

th e  rat io  of the  diameter  to the  length  of the

string

  is

  re lat ive ly large .

  The

  stiffness

  of the

strings leads

  th e

  frequencies

 o f

  upper part ia ls

to be

  progress ive ly sharper than t rue har-

monics (which  stand

  in

  integer  ratios

  to the

fundamenta l ) .

  Hence, tuning upper  funda-

mentals  to  partials  o f  lower notes will lead  to

stretched octaves,

  but

  they

  are not so

  severely

stretched as those of the human ear. Our

Western emphasis  o n  consonance  o f  simul-

taneous

  tones leads

  us to  prefer

  this

  piano

tuning over one in which melodic succession

would  be  g iven precedence. (See  P l o m p  &

Levelt,  1965,  for a  discussion  of the  contribu-

tion

  o f

  coincidence

  o f

  upper harmonics

  to the

consonance  o f  complex tones.)  These  con-

siderations also

  suggest

  that

  for the

  same rea-

sons,  a  cappella choirs  and  string

  quartets

will

  tend

  to use a 2 /1

  octave, especial ly dur-

ing   slow passages.

The precision of judgment of

  Ward's

(1954)  subjects  is a  very good reason  fo r  pre-

ferring  a

  quas i - l oga r i t hmic

  scale  to  other can-

didates

  for the

  psychophys ica l sca le re l a t ing

pitch   to  frequency  (Ward,  1970), Other rea-

sons inclu de the above cross-cul tural data and

th e

  fact

  t ha t  one o f  Ward's

  (19S4)

  subjec t s

who  had absolute pi tch produced the same

scale  by  note label ing  a s  with  th e  oc tave

j udgment  metho d. A noth er s t rand  o f evidence

is

  provided by Nul l

  ( 1 9 7 4 ) ,

  who obta ined a

l oga r i thmic

  scale using

  a

  s l ight ly

  modif ied

magni tude  es t imat ion method.  One of the

cleverest

  corroborat ions

  of the log

  scale

  is

provided

  by an

  experiment

  o f

  At tneave

  a nd

Olson

  ( 1 9 7 1 ) .  They observed that musical ly

untra ined  subjec t s

 h a d

  difficulty

  in

 p roduc ing

transposi t ions  o f  arbitrari ly selected pitch  in-

tervals. Then, they

  h i t

  upon

  th e

  idea

  that

cer t a in

  interval sequences might

  be

  over-

learned

  even  by the  unini t ia ted,  fo r  example,

th e  Na ti ona l B roadca s t ing Company (NB C )

chimes. They

  found

  t h a t  t he NBC ch imes had

always been presented at the same pitch level

— o n  th e  notes G-E-C  in the  middle register,

an  ac ronym  for the  General Electric Corpora-

t ion.

 They

  asked subjects t o  produce  th e  pat-

tern

  at var ious pi tch levels above and below

the or iginal . The scal ing quest ion was, Would

transposit ions  fol low  a log scale , or a power

funct ion,

  o r

  something e lse? Subjects over-

whelmingly

  responded with t ransposi t ions

a long  a  logar i thmic scale . Thus, g iven a  mean-

ingful

  task, untra ined subjects use a loga-

r i t h m i c  scale  fo r  p i t ch .

Attneave

  and

  Olson's  subjects  demonstrate

ho w

  stable the tonal sca le system is through-

o ut

  life

  once

  it is

  learned.

  So do

  Frances's

(1958)  subjects,  wh o

  found

  it  easier  to  notice

changes

  o n

  rehearing tonal melodies  than

atonal. Tonal scales constitute one of the most

durable families of perceptual-motor  schemata

that have been observed

  in

  psycho logy , rank-

ing   perhaps only

 a f ter

  th e  schemata  o f  natu-

ral language in their stabil i ty and resistance

to   change  in adul t  life.  In  fact ,  th e  same kinds

of  categorica l percept ion phenomena  found

in the

  phonetics

  o f a

  language seem

  to

  hold

f o r  musical sca le pi tch judg ments (Burns,

1974) .  It  would probably surprise most

American psych o log i s ts  h o w early  in  l ife  these

perceptual-motor schemata   a re  acquired.  Im -

Page 6: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 6/14

346

W. JAY  D O W LING

A  =

+4

  -2 +3 +2

B  Jl»f

+4

 -2 +3 +2

• T   f M

  3-1

  3 2

  3 -3

  4-2  -2

-1 +3+3-1

Dow l l ng  8  Fuj l tani

Standard

Target

Tonal Answer

Atonal Contour

R a n d o m

H

I

+2 +

  - 2 + 3

  1  -2 3

S t a n d a r d

T a r g e t

  o n t o u r

  a n d o m

+ 1+2-3  -3

Figure

  3.  Sections A-E are  examples o f  s t imul i

  f ro m

th e  present experiment.  (A i s the  sample  tonal stan-

dard  s t imu lus ;

  B i s the

  target s t imulus ,

  th e

  exact

t ranspos i t ion of the  s t andard;  C i s the  tonal  a n-

swer  lure,  whi ch  remains  in the key of the  s tandard

but at a

  different

  pi tch level ;  D is the  atonal lure

with   th e  s ame  c o n t o u r  as the  s t andard;  and E is the

r andomly  different  lure.) Sections

  F-I  a re

  examples

o f

  atonal s t imul i

  from

  Experiment

  1 o f

  Dowling

a nd  Fuj i tani , 1971.  (F is the  sample s tandard s t imu-

lus ; G i s the  target s t imulus ,  th e  exact t ranspos i t ion

of  th e  s t a nda r d ;  H i s t he  s a m e - c o n t o u r l u r e ;  and I

i s

  th e  r andomly different  lu re . A ll  staves  a re  notated

with  a  treble  clef  unders tood. Interval s  in semi tones

a re

 sh own under each  stave.)

berty   ( 1 9 69 )  found  that  8 year  olds  a re  able

to

  not ice sh i f ts

 o f

  sca les wi th in

 a

  melody  f rom

o ne   key to

  ano the r

  a

  semitone

  or two

  h igher .

They  found sh i f t s

  o f

  mode

 that

  left

  th e

  under-

ly ing pa t tern

  o f

  interva l s unchanged

  (as in the

tonal answer stimuli of the experiment re-

por ted below) much more  difficult.  Zenatt i

( 1 9 6 9 )  found  that

  8  year  olds  c a n  recognize

three-note tona l melodies much better than

atonal.  Five year olds  find  tona l  a nd  atonal

melodies equally

  difficult

  because

  they

  have

no t

  y et

  thorough ly interna l ized

  th e  scale.

Eleven year olds

 find

 both  types equal ly  easy.

But with

  four -

  and six-note melodies, even

adults find the atonal  m o r e  difficult,  in agree-

ment with Frances

  ( 1 9 5 8 ) .

In

  th i s di scuss ion

  of

  musical scales ,

  I

  have

not tried

  to

  inco rpo ra te  Shepard ' s  (196 4)

hel ica l theory   o f  pi tch  scales.  I  wish  to  note,

however ,  that  I  bel ieve  th e  present theory

com p a t i b l e  with h i s .  It  only requires  that  th e

hel ix be sprung somewhat to a l low for Ward's

enlarged subjective octaves.

  I

  believe Shep-

ard's  demonstra t ion

  of the

  audi tory barber

po le could

  be

  repl icated using Ward's

  quasi-

l oga r i thmic  scale.  Th e  issue  of the

  acceptance

of

  th e  hel ica l model  a s a  pitch scale  h a s n o t

been settled, however.  Fo r  example , Deutsch

( 1 9 7 2 )  h a s  demonstra ted  that  in the  recog-

ni t ion of  fami l i a r

  tunes , chroma ( the qua l i ty

of

  C -ness, D-ness , or E-ness; do-ness, re-ness,

or mi-ness)  does  no t

  suffice

  as a cue t o  tune

ident i ty .

M e m o r y   fo r  Melodi c Co ntour

Fo r  everyone except  th e  small percentage

of  the popula t ion having abso lute pi tch , wel l -

known

  melodies must

  be

  s tored

  a s

  sequences

of

  pitch intervals between successive notes

(Deuts ch , 1969) .

  In

  this section,

  I

  have

  as-

sembled evidence  that  memory for the con-

t ou r

  (the ups and downs of the melodic

interva l s )  c a n  funct ion separa te ly  f rom  mem-

ory fo r

  exact interval s izes.

 That

  is , the

  con-

t ou r

  i s an abs trac t ion  from  the ac tua l melody

that  can be remembered independently of

pi tches or interva l s izes .

 This

  is true for mel-

odies in both shor t- term and long- term

m e m o r y .

Retrieval from Short-term Mem ory

The c ontour s o f

  brief,

  novel  atonal  mel-

odies

  can be

  retr ieved

  f rom

  shor t- term mem-

o ry   even when

  th e

  sequence

  o f

  exac t  intervals

canno t .

  This

  po in t

  is  illustrated  by an ex-

per iment

  by

  Dowl ing

  a nd

  Fuj i tani

  ( 1 9 7 1 ,  Ex-

per iment 1) . They  used  a  shor t- term recogni-

t ion memory paradigm   in  wh ich  th e  standard

Page 7: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 7/14

MEMORY

  FO R  MELODIES

347

s t imulus   o n

  each trial

  was a

  randomly gen-

erated

  five-note

  melody having smal l p i tch

intervals between successive notes.  Th e  com-

parison  stimulus followed  th e  standard  after

a

  brief

  pause. Three types of comparison mel-

odies  were  the same as the standard in both

contour  and  pitch intervals,  or had the  same

contour

  but  different

  intervals ,

  o r

  were dif-

ferent  in

  bo th con tour

  a nd

  intervals ( i .e. ,

were

  novel random sequences). These are i l-

lus t ra ted

  in

  Figure

  3,

  Sections

  F-I.

  Different

groups

  o f

  subjects

  had the  task  o f

  distinguish-

ing   amon g  th e

  following

  types  o f  comparison

melody: exactly the same versus random , ex-

actly  th e  same  versus  same contour,  and  same

contour  versus random. Half the subjects were

given

  comparisons start ing

  on the

  same pitch

as the  s tandard;  th e  o ther ha l f  h a d  com-

parisons wh os e start ing note was transposed

to

  another pitch level.

  Th e results

 showed

  that

when

  th e

  compar i son s ta r t s

  on the

  same pitch

level  a s the  standard, exact-same comparisons

a re  easily  distinguished  from  either random

comparisons

  o r

  same-contour ,

  different-inter-

va l  comparisons .  In  other words,  it was  easy

to reject  a ny  comparison s t imulus

  t h a t

  di d

no t  conta in exact ly  th e  same pitches  as the

standard. When  th e  comparison melodies were

transposed, ho wever, exact-same targets   a nd

sam e-contour co mparisons were easi ly dis-

tinguished  from  random  ones

  but

  almost

  im -

possible  to  tell  apart.  Listeners responded  o n

th e

  basis

  of the

  presence

  o r

  absence

  of the

contour

  and

  were unable

  to

  recognize

  th e

sameness  o f  intervals when these were added

to   th e  contour .

Thus,

  when

  th e  listener  is  t ry ing  to  retrieve

an

  exact

  set of

  intervals  from  memory ,

  th e

best

 he can do

  under

  the

 above conditions

 is

retrieve  th e  con tour .  It is cr it ical to  this  re -

sult  t h a t

  th e

  target melodies

  be

  atonal ,

  that

is , not

  using

  th e  pitches  o f a

  musical scale

famil ia r

  to the subjects. Frances  (1958 )  found

a l tera t ions

  in

  tonal melodies easier

  to

  detect

than a l tera t ions

  in

  atonal melodies.

  A

  ma j o r

theme  in the  present article  is  t h a t  there  a re

two

 components

 a t

  work

  in the

  normal

  process

o f

  melody recogni t ion: contour  and  scale.

Bowling

  and Fujitani  (1971 ,  Experiment  1)

explored  th e  extreme case where  th e  role  o f

th e  scale  h ad  been  all but  eliminated.  They

found  contour recogni t ion

  to be

  completely

dominant  over pitch interval recognit ion.

However,

  o ne

  would

  no t

  expect

  th e

  same

  re -

sult

  with  tonal  melodies.  The

  overlearned

scale

  framework

  shou ld make recogn i t i on

 o f

th e  difference  between a tonal target melody

and  an

  atonal

  lure having  the  same  contour

much

  easier. (These st imuli are i l lustrated in

Figures  3B and  3D.)

Dis t inguish ing

  between  a  tonal melody  a nd

a  tonal lure having  th e  same contour  as the

first  melody  but  starting  o n a

  different

  note

of  th e  same modal sca le should  be  very d i f -

ficult . The

  rela t ionship

  'between

  this  l a s t  pair

of

  melodies  is the  same  a s

 that

 o f a

  fugue

  sub-

ject  and its

  tonal answer.

  Such  a  pair  is i l-

lustrated

  by

  Figures

  3A and 3C.

  There

  a re

two   ways

  to

  th ink

  of the

  tonal  answer

  in

terms

  of the  conceptual scheme o f  tuning sys-

tem   and

  mode.

1.  We might th ink of the  tonal  answer as

consisting   of the  same  set of  diatonic inter-

vals

  translated

  into a new mo de. The

  interval

pa t te rn

  of the

  tuning sys tem

  (i n

  Figure

  3, the

whi te  notes of the piano) remains f ixed at the

same pitch level, while

  th e

  tonality

  of the

melody   in the sense of its starting pitch level

is

  moved  to a new place  in the  tuning  system.

T hus ,

  both

  th e

  s tandard

  and the

  tonal answer

in  Figure 3 (A and C ) h ave the d ia tonic in-

tervals

  [ +

 2 ,

  -1, + 2,  +1] ,  but the first is

in C-major  and the second is in A-minor.

2.  The tonal answer mig ht be tho ugh t of

a s

  remaining  in the  same mode  a s the  stan-

dard, which   is exact ly  the way a  tonal answer

is

  treated

  in a  fugue.  In  that  case,  th e

 pitches

of

  the mode remain fixed while the

  starting

pitch of the melody is

  shifted

  to a

  different

degree

  of the  scale.

  Both melodies

  in

  Figure

3 (A and C) would remain in C-major, and

th e

  difference

  between them would be that  th e

standard begins

  on the first

  degree

  and the

tonal answer on the sixth degree of the modal

scale.

I

  prefer

  th e  second  o f  these ch aracteriza-

tions

  fo r

  describing

  th e

  cognitive processing

involved

  in the

  experiment

  presented

  below.

This

  is

  because

  th e

  t ime interval between

  th e

presentat ion

  of the

  standard

  and the

  tonal

  an-

swer  is  short .  Th e  standard  is  coded  a s  being

in  a  part icu lar major mode. When  th e  listener

Page 8: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 8/14

348

W. JAY  B O W LING

hears the tonal answer immediately   following

th e

  standard, nothing

  in

  that comparison pat-

tern  demands that he change mode. There-

fore,

  he hears the tonal answer in the same

mode

  a s the

  standard. Diatonic interval pat-

tern (contour in a restricted sense) and mode

c an  function  a s  features  of the  standard.

Tonal answers would then  be  difficult  to  dis-

t inguish

  from  sta ndard st imuli because they

sha re these two features . Exact t ransposi t ions

force

  th e  listener  to  change mode  in  midtr ia l .

Exact t ransposi t ions share

  th e

  feature

  o f

hav ing   th e  same intervals  a s t h e  standards

when  measured in semitones ( that i s , a t the

level  of tonal

  ma te r i a l ) .

  The experiment thus

brings  these sets  o f  feature  s imilar i t ies into

confl ict  wi th each o the r .

Lest too persuasive an argument here make

th e  results  of the  experiment seem  a  foregone

conclusion, let me introduce a plausible the-

o ry   that makes  a

  different

  predict ion

  from

th e  two-component

  scale-contour

  theory.  It

could

  be, according to th is theory, that con-

tours are not stored in memory independently

of  interval sizes. Dowling  and  Fuj i t an i  g o t

the ir resul t because they used unnatural

atonal  melodies.  Atona l  intervals

  a re

  difficult

to

  remember, both because l i s teners have

  no t

had much prac t i ce wi th them and  because

they

  do not

  occu r

  a s  part  of an

  overlearned

scale schema. However,

  if we

  replicate Dowl-

ing   a nd  Fuj i tani us ing tonal materia ls ,  we

will  get a very  different  result . The interval

sequences  o f  tonal standard melodies will  be

easily learned because they   fall

  o n a

  well-

known

  scale.

  Thus ,

  changes in those intervals

will

  be easily noticed whether the change is

to an

  atonal melody

  or to a

  tonal melody

  in

another mode.  Th e  listeners should  a t  least

do   better  than  chance

  in

  discr iminat ing

  ex-

a c t  transposit ions  from  tonal answers.  (T o

give this theory

  i ts

  due,

  I

  should note  t h a t

pilo t work with a few professio nal music ians

convinces me

  t h a t

  they can  perform  in the

wa y

  just descr ibed).

Th e  two-component theory,  on the  other

hand, claims  t h a t  even with

  tonal

  melodies,

contour

  (in the

  sense

  of ups and

  downs  mea-

sured  in  diatonic intervals)  a nd  interval sizes

(measured

  in

  semitones

  at the

  level

  o f  tonal

materia l)  a re  stored independently,  th e  latter

s imply a s a

  mode label .

  Tonal i ty  c an

  func-

t ion

  as a cue to

  dist inguish

  a

  tonal melody

from  a n  atonal one.  But  changes  in  interval

size  that leave tonal i ty intact  will

  be

  difficult

to  notice.

Experiment

 o n

  Tonal Melodies

Thi s   experiment

  is

  based

  on the

  paradigm

o f  Dowl ing  a nd  Fuj i tani (197 1, Experiment

1 ) .  It copies the condit ions of

  t h a t

  experi-

ment  in  which compar i son me lodie s were

transposed (i.e. , began on a   different  pi tch

from  th e  standard melodies)  and in  which  th e

subject 's task  was to  reco gnize only exact

t ranspos i t ions of the  standards.  Th e  most  im -

por t an t  difference  between

  the two

  experi-

ments i s  that  in Dowling and Fuj i tani , a l l

melodies  were atonal ; whi le  in the  present

experiment,  al l but one  type  o f  compar i son

melody were tonal . Other  differences

  in

  method

derived

  from  differences  in the  avai lable

equipment,  subject populations, and the de-

s i rab i l i ty   o f

  equalizing expected interval size

between  tonal and atonal melodies used in

th e  present experiment. Since  th e  most com-

parable condit ions in the  tw o

  experiments—

those

  o f  dist inguishing targets  from  randomly

different

  comp a r i s on s— gav e  rough l y compara -

ble results

  ( 8 1 %

  a nd

  8 4%

  vs .  89 ),  the ef-

fects  o f  changes  in  these other variables  a re

apparent ly negl ig ible  fo r  purposes  o f  estab-

l i sh ing  the qual i ta t ive resul ts toward which

this  study is directed.

Using   tonal melodies

  in

  this paradigm

makes possible  th e  in t roduct ion  of one  more

type of comparison melody: the tonal answer.

In  this type,  th e  compar i son beg ins  o n a d i f -

ferent  pitch  from

  th e

  standard

  but

  stays

with in  th e same diato nic scale.  T h u s ,  th e  pitch

intervals between tones  will  generally  be

changed,

  but the

  melody will st i l l

  be

  tonal .

Figure 3 shows examples of th is and the other

types of comparison st imuli used in the pres-

ent

  experiment  and  t ha t  o f  Dowling  a nd

Fuji tani .

Method

Subjects.

  Twenty-o ne students

  a t the

  Univers i ty

o f  Texas

  a t

  Dal la s served

  in

  four  separa te

  g r o u p

Page 9: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 9/14

M E M O R Y  F O R M E L O D I E S

349

sess ions , mos t

  o f

  t h e m

  receiv ing

  ext ra credi t

  in

upper level  psychology  courses

  fo r

  their  participa-

t ion . Subject s were d ivided into g roups on the bas i s

o f  a  pos texper iment ques t ionnaire . Subject s  in the

exper ienced  g r o u p  had 2 o r  m o r e y e a r s  o f  m u s i c a l

t r a in ing ( inc lud ing s t udy ing a n in s t rument o r vo i ce

or p lay ing a n ins t rum ent in an ensemb le but ex-

c lud ing t ak ing mus i c co ur se s o r  s ing ing  in  c h o i r ) .

In the  inexper ienced g roup, subject s  h a d  less  t h a n

2

  y ea r s t r a in ing . M eans

  for the two

  g ro ups were

5.0

  and .14 years t ra i n ing , re spect ive ly . The mea n

age o f subject s was 30.4 years and was comparable

in

  t h e t wo g ro ups .

  There

  were three males and seven

fema les

  in the

  exper ienced g roup

  a nd

  seven males

a nd  f o u r  females  in the  inexper ienced g roup  (a  d i f -

ference  t h a t  p r o b a b l y  reflects

  a

  tendency

  to

  g ive

females

  mus ic le s sons  in our

  c u l t u r e ) .

  P e r f o r m a n c e

o f

  ma le s and

  fema les

  in t h e t wo g ro ups was r o ug h ly

co m parab le , w i t h a s l i g h t supe r io r i t y o f expe rienced

males.

Procedure.  Subjec t s were ins t ruc ted  that  t h i s

  wa s

a n  expe r imen t  in  m e m o r y  fo r  melodies .  O n  e ach  trial,

t h ey h e a rd a pa i r o f  br ief  m elodies , and th e ir  t a sk

was t o s ay wh e t h e r t h e t wo were t h e s ame o r d i f -

ferent.

  Fo r  th is purpose , t hey responded us ing  a

f o u r - c a t e g o r y

  sca le with responses

  o f  sure

  same,

same, different,

and sure  different . There were

48

  t r ia l s in the exper iment and they wro te the ir re -

sponses on a shee t o f paper provided. The exper i -

menter expla ined tha t a l l compar ison s t imul i , even

those th a t were the sam e, would s t a r t on a

  different

note

  f rom  th e

  s t a n d a r d . W h a t

  wa s

  i m p o r t a n t

  in de-

t e rmining sameness  w a s  t h a t  no t  o n ly  t he ups and

downs

  bu t

  a l s o

  th e

  dis t ances among

  th e

  notes

  be

t he same.

  Th e

  exper imenter

  referred  t o t he

  poss i -

bi l i t y o f s ing ing  Happy  B i r t h da y s t a r t ing o n d i f -

ferent

  notes  a s a n

  illustration

  of how the

  same  mel-

o dy co u ld o ccur a t  different  pi t ch leve ls as long as

th e

  d i s t ance s amo ng

  th e

  notes remained

  th e

  s ame .

Th e  exper imenter presented  o ne  example  o f  e ach  o f

th e

  types

  o f s t imulus pa i r s .  Af t e r  responding to ques-

t ions ,  th e  exper imenter t hen presented  the 48  tr i a l s .

Stimuli.  St imul i were play ed  o n a  fre sh ly  tuned

Ste inway piano

  a nd

  recorded

  o n

  t ape .

  They

  were

presented  to  subject s over loudspeakers  a t  c o m f o r t a -

ble

  l i s tening leve ls v ia h ig h -q ua l i t y reproduct ion

equipment . The t iming o f s t imul i was cont ro l led by

a Da vis t imer produc ing c l icks every .67 sec , wh i ch

were bare ly audible on the t ape . In a l l s t imul i , t he

ra te

  of presenta t ion o f t he quar te r-no te va lues  ( J  )

of  F igure

  3 was 3 tones per sec . Thus , each s t imulus

was 2 sec in dura t io n. There wa s a 2-sec blank in-

te rva l be tween s t andard

  a nd

  co mpar i s o n s t imul i

  o n

each   tr i a l  and a  5-sec response interval fo l lowing

th e

  co mpar i s o n s t imulus .

  A

  warn ing , co ns i s t ing

  o f

th e

  exper imenter ' s vo ice announcing

  th e

  t r i a l num-

ber, preceded

  th e

  onse t

  o f

  each t r ia l

  by 2

  sec.

A ll  s t andard s t imul i began  o n  midd le  C  ( f u nda -

menta l

  f requency  of 262

  Hz) . Th e re

  are 16

  poss ib le

co n t o ur s

  for f ive-note

  sequences , provid ing unisons

are exc luded. The contours o f t he  standard  s t i m u l i

on each o f t he 48  tr i a l s  of the experiment were se-

lec ted by us ing th ree success ive random permutat ions

o f  th e

  o rde r

  of the 16

  c o n t o u r s .

  T h u s ,

  e a c h c o n t o u r

appeared  as a  standard

  exact ly

  three  times in the

exper imen t .

  Th e  interva l s izes be tween no tes  o f t he

t ona l s t imul i were chosen with the

  fo l lowing

  p r o b a -

bi l i t ies  o f

  d i a t o n i c s c a l e s t eps :

  P (±\

  s tep)

  =.67;

P

  (±2

  s teps)

  =

  .33.

There were

  f o u r

  t y pe s

  o f

  co mpar i s o n me lo dy . Each

o f  t h e se  f o u r  types  o c c u r r ed  equal ly  of ten

  in the

th r ee

  b lo cks

  of 16

  t r i a l s ( r ando mly de t e rm ined) .

  A ll

co m par i s o n me lo d ie s beg an o n e i t h e r t h e E abo ve

or the A

  be low middle

  C

  ( r ando mly de t e rmined) .

Th e se t r an spo s i t i o n s we re ch o sen a s be ing mo de ra t e ly

d i s t an t  f r om

  C

  b o t h

  in

  pi t ch leve l

  (+4 and  — 3

semitones ) and in  sh a red p i t ch e s  ( t h r e e  a nd  f o u r ,

r e s pec t ive l y ) .  (F rance s ,  1958,  h a s s h o w n

  r e mo t e

t r an spo s i t i o n s

  in the

  la t te r sense harder

  to

  r e c o g -

nize  t h a n

  near

t ranspos i t ions . )   Target  co mpar i s o n

s t imu l i

  (see F igure

  3B )

  were exact t ranspos i t ions

  to

the s t anda rds to t he key o f E o r A. They re t a ined

exact ly t he in terva l s izes o f t he  s t anda rds . Tona l

an s wer

  co mpar i s o n s t imul i ( s ee

  F igure

  3C) were lures

t h a t

  s t ar ted

  on E or A but  that

  rema ined

  in C, the

s a m e  key a s t he  s t anda rd ,  and had the  s ame co n-

t o u r

  a nd

  d ia t o n i c in t e rva l s . Th us ,

  th e

  interva l s izes

in  semitones d id no t remain the same as in the s t an-

dard , wh i le t he tona l sca le remained the same.

A t o n a l

  s ame-co n t o ur co mpar i s o n s t imul i ( s ee F ig -

ure  3D )

  were lures t ha t re t a ined

  th e

  c o n t o u r

  o f t h e

s t a n d a r d

  but

  tha t used interva ls randomly se lec ted

wi t h o u t r e g a rd  fo r  tona l sca le s .  Th e  pro bab i l i t i e s  o f

in te rva l s izes o f t he a tona l lures were   P  (±1 semi-

t o n e )

  =  .17,

 P (±2

  semi t o ne s ) =.33, and

 P

  (±3

  semi-

t on es )  — .50. (These probabi l i t ie s were c hosen  s o

t h a t t h e expec t ed in t e rva l

  size

  of a t one sequence

w o u l d  be 2 .33 semitones . Th is i s co mpa rable to t he

expected interval s ize

  o f

  2 . 2 7 s e m it o ne s

  for the

  tonal

sequences.  The

  latter

  figure was

  calculated

  as a

weigh ted average o f t he d ia tonic in te rva l   s izes ,  with-

o ut  reg a rd  fo r  s t a r t ing p i t ch .  Th e  f a c t  that  a l l  stan-

dard  s t imul i began  o n C ,  imm edia te ly next  to a

1-semitone

  in t e rva l , wo u ld l o wer t h i s e s t ima t e s l i g h t l y

in the

  present case . )

  Th e

  rando m  different  s t imul i

(see  F ig ure  3E )  were lu res h av ing  different  c o n -

t o u r s

  f r om

  t h e s t anda rds ( r ando m ly s e le c ted) a nd

different  dia tonic sca le in te rva ls se lec ted jus t as with

th e  s t anda rds .

Data

  analysis.  Th e fo ur - c a t eg o ry co nf idence judg -

ments were used

  to

  genera te individua l memory

  o p-

era t ing charac te r i s t ics (MOC)

  fo r

  e ach sub je c t

  fo r

each o f t he th ree types o f s t imulus compar isons . Hi t

ra tes  were g iven  by  responses  t o t he  t arge t s t imul i .

These

  h i t

  ra tes were plo t ted separa te ly ag a ins t t h ree

sets

  o f

  fa lse  a l a rm r a t e s g iven

  by

  responses

  to

  t o na l

answer ,  atonal

  contour,

  and random  lures  to

  give

t h ree areas under

  the M OC fo r

  e ach sub je c t .  Areas

under the  MOC can be in terpre ted as an es t imate o f

w h a t

  th e

  pro po r t i o n co r re c t wo u ld

  be in the

  ca se

w h e r e  c h a n c e p e r f o r m a n c e  is .50  (Swets , 1973) .

Area s   under  the M OC were eva luated us ing  an un-

weig h t ed-means ana ly s i s  o f  v a r i a n c e  fo r  unequal ce l l

s izes due to the unequal numbers o f exper ienced

and inexper ienced sub ject s .

Page 10: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 10/14

350

W.

  JA Y

  BOWLING

Table

 1

Areas  Under

  th e

  Mem ory Operating

Characteristic  of the

  Experiment

  Compared

with Similar  Conditions  of Dowling  and

Fujitani's

  (1971}

  Experiment 1

G r o u p

Target

  Target

vs. vs.  Target

t o n a l

  a t ona l v s .

answer

  c o n tou r r a ndom

Experienced

Inexperienced

Dowling   a nd

Fuj i t an i

.48

.49

.79

.59

.53

.84

.81

.89

Results

Table

  1  shows  th e  mean areas under  th e

M O C

  for the two

  g roups

  and the

  three s t imulus

compar i sons . The main  effect  of s t imulus

compar i son type  was  s ignif icant ,

 F(2,

  38 )  =

77.35,  p <

  .001. Distinguishing between tar-

gets

  a nd

  tona l answer lures

  wa s

  very  difficult,

with c hance perform anc e in bo th g roups . Dis-

t inguish ing  targets

  from

  atonal  same-contour

lures

  wa s

  somewhat eas ier ,

  a nd

  d i s t ingu i sh ing

targets  from  random lures was eas ies t of a l l .

The Exper ience X St imulus

 Type

  in te r a c t i on

was  s ignif icant ,  F(2,  38) =  8.01,  p <  .001,

ma in ly

  reflecting

  a  difference  o f  abi l i ty  in

dis t inguish ing ta rgets   from  atonal  same-con-

t ou r  lures .  Th e  m a i n  effect  o f  experience  ap-

proached  s ignif icance

  at the .05

  level.

  This

modest

  effect

  i s consonant wi th the modest

corre la t ions   found  by D owl ing and Fuj i t an i

between

  per fo rmance on the i r  task  and ex-

perience.

Genera l Discuss ion  o f

  Short - term

  Recogni t ion

of

  Melodies

Th e  present results i l lustrate  th e  impor tance

of

  scale as

  well

  as contour in shor t- term rec-

ogni t ion memory for melodies .  This  po int i s

b r o ug h t

  out by

  com par ison wi th

  th e

  results

of

  D owling and Fuj i tani (see Table  1) . Both

studies

  found

  that

  the two

  melodies

  a re

  rela-

t ive ly easy   to  dis t inguish  in

  that

  they have

different  contours,  with performance  in the

.80s.  Dowl ing a nd  Fuj i tani ' s  subjects  found  it

difficult  to  dis t inguish between  tw o

  atonal

melodies wi th the same contour but  different

interval

  sizes,

  performing  at  around  the

chance level of .50. In the present experiment,

subjec ts  found

  it

  easier

  to

  re jec t

  a n  atonal

compar i son me lody when  it was  preceded  by

a

  tona l s t anda rd me lody .

 This

  wa s

  especially

true  o f  experienced subjects  wh o  presumably

h a v e  a f i rmly  interna l ized modal sys tem.

What subjec ts  in the  present experiment

found

  extremely  difficult,  performing  a t

chance ,

  was

  dis t inguish ing between exac t

transpos i t ions

  o f

  comparison melodies

  to new

tona l keys and

  sh i f t s

  o f the contour a l ong the

same dia tonic sca le  as the  standard. Both ex-

perienced

  a nd

  inexperienced  subjects

  h a d

t rouble

  wi th th i s

  task.  Phenomeno log i ca l l y ,

th e

  compar i son s t imul i

  in

  Pairs

  A-B and

A-C o f F igure 3 sound

  natural,

while  th e

compar i son

  in

  P a i r

  A-D

  sounds  strange.

This  resul t i l lus tra tes

  th e

  separateness

  o f

th e  func t ions  o f

  c o n to u r

  a nd

  mode.

  Th e

  func-

t ion of mode is not to fix a set of intervals

in  semitones

  a s

  belong ing

  to a

  melody.

  If it

were,  tona l answers would not be confused

with exac t t ranspos i t ions .  Th e  mode  is  simply

a   f r amework  o n  which  th e  contour  m a y b e

h u n g .

  Fo r

  brief

  melodies heard only once,

  th e

po in t  on the  modal sca le where  th e  melody

beg ins

  i s no t

  taken into account. What sub-

jects seem

  to

  a c c o un t

  for is

  that  both

  th e

mode  and the  contour  are the  s ame  in the

tw o   melodies.

This  resul t should  not be  taken  to  mean

that

  diatonic scale intervals  a re  somehow psy-

cho log ica l ly equa l .

  Th e

  concept

  o f

  subjec t ive

equa l i ty   applies  best

  to the

  level

  of the

  psy-

chophys i c a l

  scale.

  Th e

  aes thet ic purpose

  o f

using   differently  sized  pitch  intervals

  in mo-

da l  scales would  be  lost  if al l the  intervals

in the mode were subjec t ive ly the same. Dif-

ferent

  intervals  a re  used  because  they pro-

vide melodic interest. Melodies  a re  translated

into

  different

  modes because  that  i s more in-

teresting than rei terating them   in the  same

mode.

  M us i c i ans  find  tuning sys tems

  that

  use

only subjectively equal intervals dul l because

they deny

  a

  ma jo r s ource

  o f

  melodic

  variety.

This  is the  principal cr i t ic ism  of the  whole-

tone  scale with which

  Debussy  and

  others

experimented around

  th e

  turn

  of the

 century.

Moreover ,

  Frances

  (1958, Exper iment

  2 )

found  evidence  that  makes

  th e

  subjective

Page 11: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 11/14

M E M O R Y  FO R  M E L O DI E S

351

equa l i t y

  o f  modal intervals  difficult  to  believe

in.

  Frances

  mistuned certain notes

  o n a

  piano

and then placed these notes in various melodic

contexts. He

  found

  t h a t

  when the mis tuning

was in the

  d i rec t i on

  of the

  dynamic tendency

o f

  the tone in i t s modal context , the a l tera-

t ion was much less l ikely to be not iced than

when the

  mi s t u n in g

  was in the

  oppos i te

  di-

rec t ion.  Th i s  means  t h a t  the subject ive inter-

va l s ize changes with modal context . Such

changes

  would

  be  consonant

  with

  th e

  theory

t h a t  dia tonic intervals tend

  to be

  subject ively

equal provided that the changes due to con-

text make large intervals

  (in the

  sense

  of the

psychophys r c a l

  sca le and tonal mater ia l

levels)

  smal ler and smal l in tervals larger .

However ,

  Frances  go t t he  oppos i te resul t .  In

al l of his cases, i t was small intervals   t h a t

became

  sm a l le r.

Th e  confus ion  o f

  target melodies

 a nd

  tona l

answers does

  no t

  occur wi th

  well-known

  mel-

odies s tored in long-term memory. At tneave

a nd  Olson

  ( 1 9 7 1 )

  have shown that with  fa -

mi l i a r  melodies, exact interval s izes

  a re

  pre-

c ise ly remembered. In

  fac t ,

  famil iar melodies

can be

  used

  a s

  mnemonic s

  fo r

  ret r ieving sca le

intervals ,

  a s

  when

  a

  m usic s tudent uses

  th e

song Over

  There to

  remember descending

m ajo r

  s ixths

  ( —  9

  semitones)

  or the

  song

  There's

  a Place for Us for asc ending mino r

sevenths ( + 10 sem ito nes). But even for fa-

mi l i a r  melodies , the contour and mode can

funct ion

  independently, as when  Frere

Jacques

o r

  Twinkle, Twinkle

a re

  sung

  in

a mi nor key o r when their co ntours are recog-

nized in

  spite

  o f

  distorted interval s izes

(Bowling   & Hol lombe,

  1 9 7 7 ) .

That

  memory

  fo r

  exact interval s izes

  of fa-

mi l i a r

  m elodies is so goo d raises the quest ion

o f

  how such melodies are s tored.  They  cannot

be s tored s imply as contour p lus sca le , s ince

that would leave the skips unspecif ied.  Twin-

kle,  Twinkle

and the  Andante

theme

  from

Haydn's

  Surprise Symphony

  wo uld be stored

near ly iden t i ca l l y :  th e co n tou r  [ 0 , +,0,

 +,0,

— ±, 0,

  —

0,  — 0,  — ]  (where  0 =  unison,

+ = up, and —

 =

  dow n) beginning

  on the

tonic and in the major mode. What is needed

i s a way of

  spec ify ing

  skips along the diatonic

sca le . What  I

  will

  suggest  is a  three-valued

dimens ion of  quas i - l inguis t ic marking. Inter-

vals

  of one

  dia tonic s tep

  will  be

  u n mark ed ;

two-step intervals wi ll be ma rked

  s

for sk ip ;

and

  larger leaps wi l l be marked

  be,

where

x  gives  th e  n u m b e r  o f  steps.  In  th i s sy s tem,

  Twinkle, Twinkle becom es

  [0 ,

  +

]4

,

  0, + ,

0,

 -, -, 0, -, 0, -, 0, -],

  whi le  Haydn ' s

  Andante becomes

  [0, +

s

, 0, +

s

, 0, -„, +,

0,  — „ 0,

  —

 s

,  0,

  —

 8

] .  T h i s

  sys tem takes

  ad-

vantage  of the  f ac t  tha t re l a t ive ly nar row  in-

terva ls predominate  in the  mus ic  of the  world

(Bowling,  1968). Of the 26  intervals  in the

first  two

  phrases

  of the two

  songs jus t c i ted,

12  a re  un i sons  (0

  s t e p s ) ,

  7 are one  s tep (un-

m a r k e d ) ,

  6 are two s teps (ma rked s ) , and

1 i s

  four

  steps (marked

  14 ) .

In order to document th i s predominance

o f  sma l l d i a t on ic in terva l s more  ful ly

  and to

s h o w

  tha t us ing the present sys tem of mark-

ing   would  put

  cons iderably less load

  o n

  mem-

o ry

  t h an  would  remem bering l i tera l ly every

interval s ize,

  1

  counted

  a l l t he

  intervals

  in

the co l lect ion of 80 Appalachian songs by

Sha rp  a nd

  Rarpe le s

  ( 1 9 68 ) .  I  chose th i s co l -

lec t ion  because

  it was

  compi led

  by

  careful

s cho l a r s  from

  a n

  almost exclus ively voca l t ra-

di t ion

  tha t does no t u se ha rmoniz ing accom-

paniments .  Th e  col lect ion  is  jus t abou t  th e

r ight s ize to produce s tab le  da t a ,  and the

songs in i t

  were selected with o ut

  a ny

  obvious

biases regarding the phenomenon under in-

vest igat ion.  I first  categorized  the 80  songs

into those us ing heptatonic modes ( seven

notes

  to the

  o c t av e ;

  46

  songs )

  a nd

  those

  us -

ing

  pentatonic modes

  (five

  notes to the oc-

tave;

  34 songs) . For

  convenience,

  and be-

cause

  th e

  existence

  o f a

  system

  o f

  hexa ton ic

modes

  i s no t

  indisputable,

  I

  treated

  the 21

songs that used only six notes of the  scale  a s

heptatonic (an example of such a song is

  Twinkle, Twinkle ).

  This

  decision would

  if

any th ing b i a s

  th e

  interval count aga inst smal l

d ia tonic intervals , s ince

  a

  hexatonic  step

across

  th e  note

  omitted  from

  the hepta tonic

mode

  wa s

 counted

  a s a

  skip.

  (It is

  interest ing

t h a t  the note omit ted  from  the heptatonic

mode was

  a lways

  one of the

  pair

  o f

  notes

  in

th e  underly ing tuning sys tem

  that

  stands

  in

th e

  unique 6-semitone interval

  to

  each other

— B

  or F in the

  white-note tuning

  sys tem—

a nd

  t h a t

  th e

  6-semitone interval

  did not

  occur

in any of the heptatonic songs .) A subsequent

Page 12: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 12/14

35 2

W .

  JAY BOWLING

Table

  2

Percentages  of  Diatonic

 Intervals

  of

Different

  Sizes

  in the 80  Songs  of  Sharp

and

  Karpeles  (1968},  with  Unisons

Omitted

Mode

 type

Interval s ize

  in

diatonic  steps

2

  3 4 >4

Heptatonic

(46

  songs,

1,544 intervals)

Pentatonic

(34 songs,

1,334 intervals)

56

  30 8 3 2

81 15 2 1 1

check

  on the

  da ta  showed

  that

  separa t ing

  th e

hexatonic songs  from

  th e

  hep ta ton ic

  h a s a

negligible  effect  on the

  frequency

  dis t r ibu-

tion of interval sizes.

Th e

  songs were strophic.  That

  is ,

  m o r e

  o r

less the same melody was repeated for the

several verses.  Fo r  present purposes, each  in-

terval

  in one

 st rophe

  o f

  each song

 wa s

 co unted

once.  Where  th e  r h y t h m  differed  from  strophe

to   s t rophe (because  o f

  different

  words ) ,

  I

used the rhythm of the f irst strophe.

Unisons accounted  fo r  23 and 2 5% of

th e

  intervals

  in the

  heptatonic

  and pentatonic

modes, respect ively. Taking unisons

  and un-

marked diatonic  steps  together accounted for

66 and 86% of al l the

  hep ta ton ic

  a nd

pentatonic intervals . Omit t ing unisons

  from

cons iderat ion

  al lows us to see what percent-

ages of + and — intervals need to be marked

  s

o r

  1.

Table

  2

  shows these

  da t a .

  Fo r

both mode types ,

  th e

  unmarked intervals

  a c -

cou n t

  fo r

  over ha l f

  th e

  cases. Only  13%

  o f

hep ta ton ic  and 4% of the  pentatonic intervals

need  to be  coded  as leaps.

Moreover ,

  th e

  leaps

  a re

  used

  in

  generally

predictable ways in the songs. Of 202 leaps

o f  three or more steps  in the heptatonic songs ,

106  ( 5 2 % )  occur red  in one of the

  following

tw o   contexts :  (a )  pick-up notes  a t the

  start

o f

  a

  phra se

  o r

  leaps

  to the

  s t a r t

  of a new

phrase

  (a s

  between

  th e

  second

  a nd

  th i rd

phrases o f

  Twinkle,

 Twinkle ;  75 instances)

and (b)  passages  in  which  th e  same interval

a nd  rhy thmic pa t t e rn inc lud ing  th e  leap  wa s

repeated

  in the

  same song

  (as in the first and

penul t im a te phrases of Twinkle, Twinkle ;

31  ins t ances ) .  Th e  occur rence o f  leaps  is  also

redundant with mo de. A mo ng the pentatonic

modes, for  example,  th e  mode whose ascend-

ing intervals in semitones are [2, 3, 2, 2, 3]

had only

  1%

  leaps

  in its

  songs versus

  4%

for

  th e

  other modes.

  A nd  each

  mode

  has i t s

own

  pattern of more and less probable leaps.

Fo r  example,  no  leaps were observed between

th e  seventh  a nd

  four th

  degrees  of the  m a j o r

mode  ( the

  6-semitone

  in t e rv a l ) .

The scarc i ty of leaps and their redundant

usage when they do occur reduces cons idera-

bly the load on memory. Therefore, i t seems

plaus ib le  to  charac ter ize  th e  memory s t o rage

of

  the p i tch mater ia l of an actua l melody as

a co mbinat ion of mo de and co ntour , inc lud-

ing

  a

  spec i f i c a t i on

  of the

  start ing pitch level

in

  th e

  mod e

  and the

  mark ing

  o f

  skips

  and

leaps.  It is  impor t an t  to  note that th i s  ana l -

ysis

  is

  restricted

  to the

  pi tch mater ia l . With

rea l melodies , rhythm must p lay   a n  imp o r t an t

par t in memory . Ko l insk i  (196 9 )  in h is anal-

ysis

  o f  var iants  of the  song

  Barbara Allen

provides num ero us ins tances in which the

very same contour and mode, with a change

in  rhy t hm, b ecome  a

  different

  song ,  often  in

ano ther cu l tu re .  Fo r  example,  a  change  in

r h y t h m  turns

  one of the

  var iants into

  th e

sextet

  from  Lucia

  di

  Lammermoor.  In  fact ,

it is

  ar t i f i c i a l

  even to think of  rhy thmles s

melodies.

  In the

  state

  of our

  knowledge,  this

is  a  necessary ar t i f ic ia l i ty , s ince  th e  problem

becomes  enormo us ly complex when rhy thm

  is

added.

Thinking

Rather than jus t ho ld ing

  a

  melodic phrase

in  short - term memory, subjects in

  Bowl ing ' s

( 1 9 7 2 )

  s tudy

  o n

  melodic t ransformat ions

  h a d

to turn it upside down,  backwards ,  o r b o t h .

Recogn i t i on

  fol lows

  that order  o f  increas ing

difficulty.  Of  interes t here  i s the  fact  t h a t

subjects

  were able only to recognize the mel-

odic  contour when so t ransformed and not

th e  exact interva l s izes .

Long-term Memory

In addit ion to being preserved in short-

term memo ry, melodic co ntours a l so seem to

Page 13: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 13/14

M E M O R Y

  FOR

  MELODIES

353

be

  retr ievable  f rom  long- te rm mem ory inde-

pendently

  o f

  interval s izes. W h ite (1960) dis-

torted  familiar melodies such

  a s

  Yankee

Doo dle and On Top of Old  Smokey by

chang ing   the pi tch intervals between notes

by

  doubl ing them, adding

  a

  semitone,

  a nd

so on .

  Whi l e

  th e

  undistorted melodies were

recognized

  94 of the  t ime, melodies sub-

jected

  to

  these contour-prese rv ing t ransforma-

t ions  were

  recognized

 a t

  abou t

  the 80%

  level.

Other c o ntour-preserving t ransform at io ns such

a s

  sett ing

  a ll

  in te rva l s

 to 1

 semito ne produced

a greater decrement in performance ( to

around  th e  65

level).

  However, contour-

dest roying

  t ransformat ions produced even

g reater decrements. For example, rando mizi ng

th e

  intervals

  o r

  subt rac t ing

  2

  semitones  from

them

  even when this changed the ir s ign pro-

duced performance

 in the

 45-50 range.

  (I t

is

  difficult  because

  o f

  sampl ing prob lems

  to

s a y

  w h a t

 th e

  chance level

 in

  White 's task was,

though i t was undoubtedly in the  10-20%

range. The

  rhy t hmic pa t te rn s

  of the

  tunes

presented alone

  on one

  pi tch were recog nized

with 33%  accuracy . )

Bowling   and Fuj i tani

  (1971 ,

  Experiment

2)  refined  som ewhat Wh i te' s approach us ing

five  f ami l i a r  tunes whose  first two  phrases

could  be s ty l ized in to the same rhythmic pa t -

tern.

  They

  used three kinds  o f  distortion:

preservat ion  o f  con tour ,

  preservat ion

  o f

  con-

tour  plus relat ive interval size (i .e. , the

grea ter - than/ les s - than

  re lat ionships

  o f

  succes-

sive intervals were preserved), and preserva-

tion of the underlying harmonic structure of

th e

  melody whi le des t roy ing contour . The

contour-preserving

  distort ion produced a dec-

rement

  in

  performance  from

  99%

  (undis-

torted) dow n to 5 9% . Preserving th e re lative

interval  sizes only boosted  this  a

  little

  (to

66%). Des t roy ing contour led to a

  perform-

ance

  o f

  28%. (Chance level

  was

  probably

between

  20% and  3 0 % ) .

Melodic contours  o f  fami l i ar  tunes  can be

recognized even though   th e  distortion  of in-

terval sizes  is  very great . Deutsch

  ( 1 9 7 2 )

showed

  that dis tort ing a tune by adding or

subtra ct ing o ne or two

 oc taves  from  each pitch

made  it  very  difficult  to  recognize . But i f th e

octave distortion  is  carried  out  with a  preser-

va t ion  o f  con tou r , per fo rmance  is  remarkably

improved (Dowling

  & Ho l lombe,  1 9 7 7 ) .

One  final  consideration  is  that  long-term

memory

  has t he

  func t ion

  in

  musical ly non-

l i tera te

  societies

 o f

  preserving

 th e

  musical cul-

ture ,  inc luding

  th e

  melodies . Wh at

  we

 would

expect  from  the present theory is that var i-

ants

  o f a

 tu ne would sha re certa in s imilar i t ies ,

namely, those  that  ar i se  from  very s imilar

con tours  be ing hung  on the  under ly ing sca le

f ramework

  in

  var ious ways .  This  variat ion

m i g h t  occu r

  fo r

  three principal reasons: for-

get t ing

  of the

  o r i g ina l  intervals ,

  th e

  desire

  to

create  interesting innovations

  by

  manipulating

in terval  relat ionships (as sho wn in Figure

1) , and  changes o f  ins t ruments and  scale sys-

tems  that necess i tate t ransformat ions

  o f

  mel-

odies  (as in the

  case

  o f

  adapt ing

  a

  non-West-

ern melody   to a  Western  scale).  A l l o f  these

processes  a re  probab ly opera t ing ,  fo r  example,

in  th e

  var i a t ions

  of the

  tune  Barbara  Allen

studied

  by Seeger (1966) and Kol inski

( 1 9 6 9 ) .  Th e

  most s t r iking resul t

  o f

  Seeger's

s tudy   of 76  variants  o f  Barba ra Al len in

the U.S. Library of Congress i s   that  they

fall

  into famil ies

  based

  on the

  shar ing

  o f

close ly re lated contours .  These  contours

  fall

onto the ir sca les

  in

  var ious ways , produc ing

a

  profus ion

  o f

  s im i l a r

 but

  interest ingly var ied

melodies.

Reference

  Note

1.

  Do wl ing ,  W. J.  Musical scales  and  p s y c h o p h y s i c a l

sca le s :

  Their

  psy ch o lo g i c a l  real i ty. In T.  Rice  &

R .  Falck

  (Eds . ) ,  Cross-cultural approaches

  to

music:

  Essays in honor of

  Mieczyslaw

  Kolinski.

M anusc r ip t  in  prepa r a t i o n .

References

A d a m s ,  C . R .  M e lo d i c co n t o ur t y po lo g y .

  Ethno-

musicology,  1976 ,

 20 ,

  179-215.

A ttneave, F, , & Olson, R.

  K .  Pitch

  a s med ium: A

ne w

  appro ach  to  psy ch o ph y s i c a l s c a l ing .  American

Journal

  of

  Psychology,  1971,

  84 ,

  147-166.

Burn s ,  E . M.

  Octave adjus tment

  by

  non-Western

mus ic ians .  Journal  of the  Acoustical Society  of

America,  1974 , 56,  S25 . (Abs t r a c t )

Deut sch ,

  D .

  Music recogni t ion.  Psychological

  Re-

view,

 1969,

 76 ,

 300-307.

Deut s ch ,

  D .

  Octave genera l iza t ion

  a nd

  tune recogni-

tion.  Perception  •  Psychophysics,  1972 ,

  11 ,

  411-

412.

Page 14: Dowling 1978 - Scale and Contour

7/26/2019 Dowling 1978 - Scale and Contour

http://slidepdf.com/reader/full/dowling-1978-scale-and-contour 14/14

354

W . J A Y B O W L I N G

Deut sc h ,

  D .

  M e m o r y

  a nd

  a t t en t i on

  in

  m u s i c .

  In M.

C r i t ch le y

  & R . A .  Henson

  (Eds . ) ,

  Music and the

brain.  Lo ndon : He inemann , 1977 .

Dowl ing ,  W. J.  Rhythmic

  fission and the

  perceptual

organization  of  tone

  sequences.

  Unpub l i shed doc -

tora l di sser ta ti on , Ha rvard U niver s i ty , 1968.

Do wl ing ,

  W. J . Recogn i t i on o f me l od i c t r an s f o rma -

t ions: Invers ion , re t rograde, and re t ro g rade inver -

sion.  Perception  •  Psychophysics,  1 9 7 2 , 12,  417-

421 .

Dowl ing ,

  W .

  J.,

  &

 F u j i t a n i ,

  D. S.

  Con tou r , i n t e rva l ,

a nd

  p i tch recogn i t i on

  in

  m e m o r y

  fo r

  melod ies .

Journal  oj the  Acoustical Society  of  America,

1971,49, S24-S31.

Dowl ing , W.

  J.,

  &  Hol l ombe ,  A. W. The percept ion

o f  melodies distor ted by spl i t t ing into several oc-

taves:  Effects  o f

  increas ing proximi ty

  a nd

  melod ic

c on tou r .  Perception

  &

  Psychophysics,  1 9 7 7 ,

  21,

60-64.

Frances , R.  La . perception de la

 musique.

 P a r is :  Vrin ,

1958.

Helmhol tz , H. von

  On the

  sensations

  oj  tone.

  N ew

Yo rk : D over , 1954 .

Hood ,  M .  Slendro  a nd  pelog  redefined.  Selected  Re-

ports,

  U CL A  Institute  of

  Ethnomusicology,

  1966,

1(1),

  28-48.

H o o d ,  M .

  Th e

  ethnomusicologist.  N ew  Y o r k :  M c -

Graw-Hil l , 1971.

Imber ty , M.

  L'acquisition  des

  structures tonales

  chez

I'enjant.  P a r i s :

  Kl incks ieck ,

 1969.

Kol insk i ,  M .  Ba rba r a A l l en :  Tonal  versus melodic

s t ruc ture ,

  part II.

  Ethnomusicology,

  1969,

 13,

  1-73.

Kol insk i ,

  M .

  Review

  o f Ethnomusicoiogy

  oj the Flat-

head

  Indians  by  A l a n  P . M e r r i a m .

  Ethnomusicol-

ogy,  1970, 1 4,  77-99.

Mar t in , D.  W.,  & Wa rd , W. D. Sub jec t ive eva lua-

t i on of musica l sca le temperament in p i anos .  Jour-

nal

  of the  Acoustical Society  oj  America,  1961,  33,

582-585.

Mi l ler , G. A. The magic number seven , p lus or minus

two .  Psychological  Review,  1956,  63 ,  81-97.

N u l l ,  C . H .  Symmetry  in  judgments  oj  musical  pitch.

Unpubl i shed doc tora l

  dissertation,

  M i c h i g a n  State

Unive r s i t y ,

 1974.

P l o m p ,  R .,

  &

  Levelt ,

  W. J. M.

  Tonal  consonance

a nd   cr i t i ca l

  bandwid th .  Journal oj the Acoustical

Society oj  America,  1965, 38,  548-560.

Seeger,  C .  Vers ions  a nd  var i an t s  of the  tunes  o f

  Ba rbara Al len . Selected Reports, UC LA  Insti-

tute  oj

  Ethnomusicology,  1966, 1(1 ), 120-167.

S h a r p ,

  C .,

  & Karpe les , M.  Eighty English  folk  songs

from   th e  southern Appalachians.  Cambr idge , Mass . :

M IT  Press, 1968.

Shepa rd ,  R. N. Ci rcu la r i ty in judgments of re l a t ive

p i t c h .  Journal  of the  Acoustical Society  of  America,

1964,  36,  2346-2353.

Sundbe rg ,

  J. E. F., &  Lindqvist ,  J.  Musica l oc taves

a nd   p i t c h .  Journal oj the Acoustical Society oj

America,  1973 , 54 ,  9 2 2 - 9 2 9 .

Sur j o d in ing ra t , W. ,

  Suda r j ana ,  P .

  J.,

  &  Susan to ,  A .

Penjettdikan dalam   pengukuran  nada gamelan

gamelan  djawa

  terkemuka

  di Jogjakarta dan

Surakarta.  Jo g jak ar ta , Indones i a : Univer s it a s

Gadj ag   Mada , 1969 .

Swets ,  J . A . The  re l a t ive opera t ing charac ter i s t i c

in   p s y c h o l o g y .  Science,  1973, 182,

  990-1000.

Ward, W. D. Subjec t ive musica l p i t ch .

  Journal oj the

Acoustical Society oj America,  1954,  26 , 369-380.

W a r d ,  W. D.  Musica l percept ion .  In J. V. Tobias

( E d . ) ,

  Foundations  of  mo dern auditory theory

(V o l .

  1) . New

  Yo rk : Ac ademi c P r e ss , 1970 .

Whi te , B. W. Recogn i t i on of d i s tor ted melod ies .

American Journal oj Psychology,  1960, 73 ,  100-107.

W i l l i a m s o n ,

  M .  C .,  & Wi l l iamson , R. C. The co n-

s t r uc t i on

  a nd

  deco r a t i on

  of one

  Burmese h a rp .

  Se-

lected  Reports, UCLA  Instituteo f

  Ethnomusicol-

ogy,  1968, 1 ( 2 ) ,  45-76.

Zena t t i ,  A, Le

  deve loppement  genetique

  de la per-

cept ion musica le . Monographies  Francoises  de Psy-

chology,  1969,

 No. 17 .

Received September

  19,

 1977

 •