Individual differences in susceptibility to the ‘‘irrelevant
R
a
n
o
te
ly
rr
le
d
c
is
.5
~s
th
at
y
he
po
o
pa
iv
m
alAmerica. @S0001-4966~97!00110-0#
PACS numbers: 43.50.Qp, 43.72.Dv @GAD#
INTRODUCTION
In environmental noise research, the need to study indi-
vidual differences has always been more apparent than in
other areas of psychoacoustics. One obvious reason is that
much of the survey research concerned with noise evaluation
uses correlational statistics. Individual differences on a sub-
jective dimension ~e.g., annoyance! are correlated with other
subjective ~e.g., attitudinal! or objective measures ~such as
exposure levels! characterizing individuals or groups of re-
spondents. Appreciable individual variation and its reliable
measurement are crucial to this research approach.
Therefore, instruments for measuring individual differ-
ences in annoyance with various noise sources ~Job, 1988;
Taylor, 1984, for reviews!, in response criteria for reporting
distress ~Green and Fidell, 1991!, or in general noise sensi-
tivity ~Weinstein, 1978! have been developed. Nevertheless,
a recent review ~Staples, 1996! blamed a lack of understand-
ing of individual differences in reaction to noise for costly
policy failures in the implementation of noise abatement, or
traffic rerouting programs.
In contrast to the questionnaire-based noise evaluation
studies thus characterized, research into the objective effects
of noise on performance—focusing on the demonstration of
overall effects of experimental manipulations—has shown
little concern with individual differences ~Jones and Davies,
1984, for an earlier review!. There are occasional reports of
personality variables such as anxiety or extraversion interact-
ing with noise effects ~summarized in Smith and Jones,
1992!; analyses of the stability of individual differences,
however, turn out to be rather disconcerting. Smith et al.
~1981!, for example, in an experiment, which required sub-
jects to memorize lists of words both in the quiet and under
continuous white noise, found quiet-noise differences in re-
call scores, and in indices of higher-order cognitive process-
ing ~‘‘clustering’’! to produce correlations as low as r
50.05 between two sessions one week apart. That is, sub-
jects appearing particularly susceptible to noise in the first
session were not the same ones as those showing the largest
performance decrements in the second session, and the mag-
nitude of the correlation indicated almost nonexistent indi-
vidual stability of these noise effects.
This inconsistency of performance across sessions may
stem in part from the highly variable effects of continuous or
intermittent white noise. Generally, these earlier studies
show that white noise presented at high sound-pressure lev-
els may either improve, depress, or result in no change in
performance. Moreover, those factors that predict such out-
comes cannot be articulated with any degree of certainty.
Such inconsistency in mean effects suggests ~but is by no
a!
Electronic mail: wolfgang.ellermeier@psychologie.uni-regensburg.de
2191 2191J. Acoust. Soc. Am. 102 (4), October 1997 0001-4966/97/102(4)/2191/9/$10.00 1997 Acoustical Society of Americaspeech effect’’
Wolfgang Ellermeier
a)
and Karin Zimmer
Institut fu
¨
r Psychologie der Universita
¨
t Regensburg, 93040
~Received 20 November 1996; revised 4 April 1997;
Individual differences in objective effects of noise o
their distribution, temporal stability, and the precision
subjects had to memorize sequences of visually presen
auditory background conditions which were random
speech; ~2! pink noise; and ~3! silence. Individual ‘‘i
the difference in recall errors under speech and in si
range extending from slight facilitation to severe
experiment after four weeks, the individual differen
50.45. Internal consistency, a measure of the prec
measured in a single session, was moderate (a50
coefficients are severely attenuated by the use of
reliability of which is bound to be considerably lower
these are correlated. Given that the original error r
determined with reliabilities approaching 0.85, it ma
decrements due to noise can be reliably measured in t
measures of noise susceptibility collected to explore
variation exhibited only weak relationships with the
were quite inaccurate in assessing their individual im
a questionnaire-based measure of general noise sensit
variance in objectively measured performance decre
relationship was much stronger in female than in megensburg, Germany
ccepted 11 June 1997!
performance were analyzed with respect to
f measurement to be attained. Seventy-two
d digits while being exposed to one of three
mixed on a trial-by-trial basis: ~1! foreign
elevant speech effects,’’ operationalized by
nce, were normally distributed over a wide
isruption. When 25 subjects repeated the
es were replicated with a reliability of r
tt
ion with which individual effects can be
5). However, both retest, and consistency
ound-minus-silence! difference scores, the
an that of the original error scores whenever
es in a specific auditory condition can be
be concluded that individual performance
‘‘irrelevant speech’’ paradigm. Self-report
tential sources of the large inter-individual
bjectively measured noise effects: Subjects
irment in the three auditory conditions, and
ity only accounted for a small portion of the
ents, although in both cases the predictive
e subjects. 1997 Acoustical Society of

means definitive in suggesting! that reliability measures may I. GENERAL METHODbe poor in such settings.
While the studies thus characterized all employed broad-
band noise of relatively high level (>80 dB), more recent
research, pioneered by Colle and Welsh ~1976! as well as by
Salame
´
and Baddeley ~1982!, found highly replicable overall
noise effects ~no single instance of improvement has been
encountered, for example! when using temporally structured
sounds of moderate intensity, and a particular task requiring
recall ‘‘in the correct order.’’ The phenomenon referred to
has been termed the ‘‘irrelevant speech effect’’ ~ISE!, since
the presentation of auditory material ~typically speech!
which the subject is told to ignore, and which is of no sig-
nificance to the task performed, has sizable effects on the
serial recall of visually presented items such as letters or
digits ~for reviews see Jones and Morris, 1992; Jones, 1993;
Jones et al., 1996!. In recent years, the ISE paradigm has
become prototypical for studying moderate-level noise ef-
fects in a situation representative of modern office environ-
ments. The quickly growing number of studies on the effect
have either addressed the cognitive mechanisms involved
~e.g., Salame
´
and Baddeley, 1982; Buchner et al., 1996!,or
the properties of the auditory distractors producing maximal
interference ~e.g., Jones and Macken, 1995a; Jones et al.,
submitted; Ellermeier and Hellbru
¨
ck, in press!; individual
differences, however, have not been analyzed to our knowl-
edge, and even the presentation of standard errors seems to
be the exception rather than the rule ~see however,
LeCompte, 1994; Jones and Macken, 1995b!.
In our opinion, the need for laboratory studies of indi-
vidual differences in the susceptibility to noise expressed in a
recent review ~Staples, 1996! is best addressed by looking at
the paradigm characterized above, for which there is ample
and consistent evidence of performance disruption. Conse-
quently, the present study was designed to collect data on a
fairly large number of subjects (N572) under standard ‘‘ir-
relevant speech’’ conditions. More specifically, a foreign
language ~Japanese! was used to elicit the effect uncon-
founded by semantic content, and a ‘‘placebo control’’ ~pink
noise! was presented in addition to the quiet baseline, in
order to control for unspecific or expectation-based effects
due to the mere presence of an acoustical distractor.
The study was conducted with two goals in mind: The
primary goal was to assert the presence of individual differ-
ences in noise susceptibility in a controlled laboratory ex-
periment, and to show that these can be reliably measured,
applying established psychometric methods derived from
classical test theory ~Nunnally, 1978; Kline, 1993; Lienert
and Raatz, 1994!. According to this approach, errors in the
‘‘irrelevant speech’’ paradigm are treated much like errors in
an intelligence test, the precision and replicability of which
is to be determined. A secondary goal was to link behavioral
effects to differences in self-reported noise sensitivity in or-
der to explore ~a! if subjects are able to assess their suscep-
tibility to a given noise, and ~b! if the personal attribute of
‘‘noise sensitivity’’ ~Weinstein, 1978! may account for some
portion of the variance in error rates observed in the labora-
tory.2192 J. Acoust. Soc. Am., Vol. 102, No. 4, October 1997 WA. Subjects
Seventy-two students at the University of Regensburg
~median age 24, range 19–44; 31 male, 41 female! partici-
pated as subjects. A random subset of this sample consisting
of 25 subjects was asked and agreed to participate in a retest
session four weeks later. Hearing problems, knowledge of
Japanese, or prior experience in ‘‘irrelevant speech’’ experi-
ments were exclusion criteria for the experiment. All sub-
jects were naive both with respect to the literature on noise
effects, and to the specific hypotheses being investigated.
B. Apparatus and stimuli
1. Visual stimuli
The visual material to be memorized consisted of ran-
dom permutations of the digits 1 through 9, presented se-
quentially in the center of a colour monitor. The digits were
about 2 cm in height and appeared for 800 ms each, with
200-ms pauses between digits.
2. Auditory stimuli
The ‘‘irrelevant’’ auditory materials were recorded and
played with 8-bit resolution and an 22-kHz sampling rate
using a ‘‘Soundblaster-compatible’’ PC sound card. Two
types of auditory materials were used: ~1! Japanese speech,
and ~2! pink noise. The speech sample consisted of a 15-s
segment from a lecture given by a male speaker. The noise
sample was recorded from a Bruel & Kjaer ~type 1405! noise
generator. These single tokens of speech and noise were
shaped to have smooth onsets and offsets, and to yield
A-weighted, energy-equivalent sound-pressure levels, L
eq
,of
76 dB, as verified by measurements at the headphones using
a Cortex Electronic ~model MK 1! artificial head system.
Due to its continuous and broadband nature, the pink-noise
sample appeared louder: The mean computed loudness levels
were 44.7 sone for the noise, and 25 sone for the speech
sample. The auditory stimuli were presented diotically over
headphones ~Beyerdynamic DT 550! in a quiet, but not
sound-treated laboratory room @ambient A-weighted sound
level approximately 40 dB#.
C. Procedure
Each trial was initiated by a 2-s visual warning signal ~a
square of decreasing size cueing the subject to the point of
fixation!, after which the stream of 9 digits was displayed at
a rate of 1 per s. Following a 5-s retention interval, a 333
numerical array consisting of the numbers 1 through 9
prompted subjects to enter the correct serial order by sequen-
tially clicking the computer mouse on the digits displayed.
On sound trials, the acoustical background ~speech or
pink noise! was present both during the encoding and the
rehearsal phase, for a total of 14 s. No sound was presented
during the self-paced recall period. Subjects were told to
ignore the auditory input while quietly rehearsing the num-
ber sequence.
In order to be able to measure individual differences
unconfounded with practice effects, the auditory conditions2192. Ellermeier and K. Zimmer: Differences in noise susceptibility