The Effect of Pitch Interval Magnitude on the Discrimination of Melodic Contours

1
The Effect of Pitch Interval Magnitude on the
Discrimination of Melodic Contours

• Tim Byron
• Kate Stevens
• MARCS Auditory Laboratories,
• University of Western Sydney.

2
Melodic Contour

• Shape of a melody.
• Why do we care?
• Jusczyk Krumhansl (1993) infants use it.
• Kolinski (1970) non-Western cultures and
  contour.
• Patel, Peretz, Tramo, Lebreque (2003) link
  between melodic contour and prosodic contour.

3
The Cognition of Melodic Contour

• Dowling (1978)
• Musically trained and untrained participants
  listening to pairs of short, unfamiliar melodies,
  in changed keys.
• Discrimination task
• Tonal Scale Melodic Contour.
• Ups and downs.
• Edworthy (1985) better in short unfamiliar
  melodies.

4
Pitch Interval Magnitude (PIM)

• Dowling (1978) did not test for this.
• Twinkle Twinkle Little Star
• SAME, LEAP UP, SAME, STEP UP, SAME, STEP DOWN.
• SAME, UP, SAME, UP, SAME, DOWN.

5
Evidence for Pitch Interval Magnitude

• Kim, Chai, Garcia Vercoe (2000) PIM most
  efficient.
• t Hart (1981) in speech, changes in F0 are
  only discriminable by listeners if changes of
  more than 3 semitones (i.e., changing from a step
  to a leap or vice versa).
• Stevens Latimer (1992) participants can
  distinguish melodies with identical contour on
  the basis of PIM.
• Stevens Latimer (1992) issues?

6
A Model of Memory for Short and Long Unfamiliar
Melodies
7
Hypotheses

• PIM can be used in discrimination.
• PIM discrimination is better in short than long
  melodies.
• Increased accuracy if change is to both PIM and
  contour in both short and long melodies.
• No difference in PIM based on direction.
• Musically trained participants more accurate than
  untrained participants at PIM, but they dont do
  PIM and contour differently.

8
Method

• 18 musically trained and 19 musically untrained
• Discrimination task with 128 trials.
• Trial pair of melodies in different keys. Same
  or different?
• 16 melodies half are 8-note melodies and half
  are 16-note melodies.
• No feedback.
• Controlled for serial position, metric position,
  implied harmony and note length of changed notes
  between different melody types.

9
Method Altered Melodies

• There were 5 different kinds of comparison
  melody
• 1. Target melodies.
• 2. Contour change melodies.
• 3. Step-to-leap melodies.
• 4. Leap-to-step melodies.
• 5. Contour interval magnitude change melodies.

10
Results PIM vs Same
Step-to-leap gt same p lt .001 Leap-to-step gt
same p lt .001
11
Results Short vs Long x PIM
Step-to-leap short gt long p n.s.
Leap-to-step short gt long p n.s.
12
Results ContourMagnitude vs Contour vs
Magnitude
In short melodies CM gt Contour p lt .04 CM gt
Step-to-Leap p lt .001 CM gt Leap-to-Step p lt
.001
In long melodies CM gt Contour p lt .01 CM gt
Step-to-Leap p n.s. CM gt Leap-to-Step p lt
.001
13
Results Step-to-Leap vs Leap-to-Step
Step-to-leap gt leap-to-step p lt .001
14
Results Musical Training
Step-to-leap musicians gt nonmusicians p lt
.001 Leap-to-step musicians gt nonmusicians p lt
.001
15
Discussion

• In support of model
• Participants can use PIM to discriminate between
  melodies.
• Musically trained participants more accurate than
  musically untrained participants at PIM (and
  dont appear to be doing PIM any differently to
  contour)
• Contour PIM change better than either just a
  contour change or just a pitch interval magnitude
  change in short melodies.

16
Discussion

• Against predictions
• No difference in salience between pitch interval
  magnitude changes in long and short melodies.
• Step-to-leap melodies are significantly easier to
  discriminate than leap-to-step melodies.
• No significant difference between long
  step-to-leap melodies and long contour
  magnitude melodies
• How to explain these results?
• Schellenberg (1997) melodic expectancy.

17
Further Research.

• PIM in unexpected melodies
• Role of temporal information in relative pitch
  memory for melodies.