dc.description.abstract |
For humans with adequately operating auditory systems the arrival of acoustic
waves to the ear and the subsequent displacement of the tympanic membrane gives
rise to the sensations and perceptions commonly known as hearing. One aspect of
hearing is loudness, to which the field of psychophysics attempts to ascribe a physical
correlate intrinsic to the incoming acoustic wave. For over 50 years psychophysicists
have attempted to explicitly define the just-noticeable difference (JND) for loudness:
the smallest amount, in physical units, that a stimulus needs to be increased to
produce a perceptual difference. A review of the literature suggests that the problem
in defining the JND is twofold. Firstly there is the measurement problem, with
specific weighting on two competing measures of the JND: the level difference (ΔL)
vs. the Weber fraction (ΔX/X). Deciding between the two measures has proved
difficult: within the natural operating range of the auditory system the two measures
are directly proportional to one another. Secondly there is the units problem,
concerned with the physical unit in which the measure should be expressed. The units
problem exists because it has yet to be determined which property of the acoustic
stimulus accounts for the perception of loudness. There are several candidates:
acoustic pressure (p); acoustic intensity (I); acoustic power (P), and acoustic energy
(E). The necessity of determining which, if any, is the proper measure of the stimulus
is forced by the phenomenon of negative masking, which is manifest when the
stimulus is expressed in terms of pressure. Selecting among the various quantities is
complicated by the fact that the four measures are linearly related and constitute a
direct transformation of one another. In the laboratory context existing techniques
can produce stimuli that severely challenge the auditory system and break down the
proportionality exhibited by the various candidate JNDs. These techniques however
require greater scrutiny and development. Buus and Florentine (1991) have proposed
that the JND measure (i.e., Δp/p, ΔI/I, ΔL) that realizes a linear relationship with the
detectability index, d’, is the correct measure. The proportionality of the measures
makes such comparison complicated, and threshold values need to be inflated by
manipulating stimulus parameters in order to exceed the range of proportionality.
This is accomplished using short duration (10-ms) sinusoids and low level masking
noise. The experimental program commenced with a series of negative masking experiments to ensure the phenomenon is still relevant to short duration stimuli.
Laming’s (1986) sensory analytical model, which assumes the auditory system
responds to pressure, provided an acceptable fit to these data. Experiments on the
pedestal effect, a related phenomenon, generated fixed-increment functions that could
be compared to the masking functions generated from the negative masking data. A
consistent relationship between the two was found when the stimuli were expressed in
units of pressure. Attention then turned to improving the studies undertaken by Buus
and Florentine (1991) to determine what the proper measure of the JND should be. A
series of experiments, differentiated by stimulus conditions, suggest a JND measure
expressed in pressure units had the most linear relationship with d’. The results argue
for pressure as the correct measure of the stimulus, and Δp/p to be the most qualified
representative of the JND.
KEYWORDS: psychophysics, sinusoid, pressure, intensity, difference threshold,
short-duration stimuli, negative masking, pedestal effect, psychometric function,
masking function, fixed-increment function, d’ |
en |