Effects of stimulus variability and adult aging on adaptation to
ur W
iver
006
ers
in e
pre
er t
In s
ure c
s. A
hou
d sp
at w
ung
tenc
bseq
in f
21/1
SS
Lomanto, 1984. These variations in speech rate influence,
tinguish speech cues that varied across two dimensions:
1among other things, listeners’ perception of phonemic
boundaries
Miller, Aibel, and Green, 1984; Miller and
Liberman, 1979, which play a key role in speech recogni-
tion. To effectively process speech, listeners must therefore
be able to adjust to changes in speech rate extremely rapidly.
When variations in speech rate are minimal, as is typically
the case in natural speech, listeners can accomplish this task
without apparent effort. However, when speech rates are ex-
tremely fast, adaptation becomes considerably more difficult.
Despite the challenge posed by rapid speech rates, lis-
teners are able to adapt to speech time compressed to rates
that are substantially faster than that to which they are ac-
customed. Studies consistently show that exposure to 10–20
sentences of time-compressed speech is sufficient to signifi-
cantly improve listeners’ comprehension of such materials

Dupoux and Green, 1997; Pallier et al., 1998; Sebastián-
Gallés et al., 2000. This perceptual learning is a robust phe-
nomenon that is largely preserved in healthy older adults
the frequency spectrum occurring during burst onset, and
2
the origin of the formant transitions. Different groups of lis-
teners were provided feedback that emphasized the fre-
quency or burst components of the stimuli. Listeners’ judg-
ment scores reflected this feedback, with decisions about
stimuli with conflicting burst and formant information fol-
lowing the dimension on which they were trained. In addi-
tion, listeners made less use of cues present in the dimension
on which they were not trained. These results support a
mechanism of perceptual learning of speech sounds that in-
volves a reallocation of attention during the learning process.
Given the above-mentioned view of perceptual learning,
it is reasonable to think that adaptation to time-compressed
speech occurs as listeners recalibrate phonemic boundaries to
accommodate the much more rapid speech rate. An outstand-
ing question is whether the redirection of attentional focus
necessary for this adjustment can be disrupted if listeners
hear speech at two speech rates that are far removed from
each other. If increasing attention to cues salient at one
speech rate comes at a cost to attention to cues at a second
rate, alternating between the two rates over the course of an
adaptation period should produce shifts in attention that
hinder learning.
a
Current affiliation: Interdepartmental Neuroscience Program, Yale Univer-
sity, New Haven, Connecticut 06520.
b
Electronic mail: wingfield@brandeis.edutime-compressed speech
Julie D. Golomb,
a

Jonathan E. Peelle, and Arth
Volen National Center for Complex Systems, Brandeis Un

Received 4 February 2006; revised 13 September 2
With as few as 10–20 sentences of exposure, listen
distorted compared to that which is encountered
examines the extent to which adaptation to time-com
the continuity of the exposure sentences, and wheth
listeners when they are equated for starting accuracy.
the degree of adaptation was assessed in four expos
to the same number of time-compressed sentence
presentation of the time-compressed sentences wit
alternated time-compressed speech and uncompresse
sentences. A fourth condition presented sentences th
uncompressed speech. For all conditions, neither yo
was influenced by disruptions to the exposure sen
showed reliable improvement across the first and su
learning mechanisms in speech perception that rema
? 2007 Acoustical Society of America. DOI: 10.11
PACS number
s: 43.71.Es, 43.71.Lz, 43.70.Dn M
I. INTRODUCTION
Speaking rates vary considerably in the course of every-
day listening. Although speech rates typically range from
140 to 180 words per minute
wpm, speakers reading from a
prepared script often exceed these rates
Stine et al., 1990.
Even a single talker will vary his or her speaking rate con-
siderably within a given conversation
Miller, Grosjean, andJ. Acoust. Soc. Am. 121 3
, March 2007 0001-4966/2007/1213ingfield
b

sity, Waltham, Massachusetts 02454
; accepted 1 January 2007
are able to adapt to speech that is highly
veryday conversation. The current study
ssed speech can be impeded by disrupting
his differs between young and older adult
eparate sessions conducted one week apart,
onditions, all of which involved exposure
continuous exposure condition involved
t interruption. Two alternation conditions
eech by single sentences or groups of four
ere separated by a period of silence but no
nor older adults’ overall level of learning
es. In addition, participants’ performance
uent sessions. These results support robust
unctional throughout the lifespan.
.2436635
Pages: 1701–1708

Peelle and Wingfield, 2005 in spite of other perceptual

Schneider, 1997 and cognitive
Wingfield and Stine-
Morrow, 2000 changes.
Perceptual learning is often characterized as the shifting
of attention toward cues that are relevant for a task and away
from those which are irrelevant
Francis et al., 2000; Gold-
stone, 1998; Nosofsky, 1986. For example, Francis et al.

2000 used category-level feedback to train listeners to dis-© 2007 Acoustical Society of America 1701
/1701/8/$23.00

An alternate line of research has examined stimulus
variability as a means of increasing the overall learning and
A second goal of this experiment was to examine the
effect of shifts in speech rate on perceptual learning in oldergeneralizability of phonetic training
e.g., Barcroft and Som-
mers, 2005; Lively et al., 1993. Logan et al.
1991 trained
Japanese listeners on identification of /r/ and /l/, a discrimi-
nation that previous training paradigms had little success in
teaching. The authors effected a significant improvement in
identification scores by employing training words that varied
in placement and context of the consonant sounds
word-
initial and final positions, singleton and cluster environ-
ments, and in intervocalic positions, spoken by five talkers.
Follow-up studies using a single talker for training found that
listeners improved in the /r/-/l/ distinction for that talker, but
that this learning did not generalize to novel talkers
Lively
et al., 1993. These results indicate that, at least in some
contexts, variability in training stimuli may actually be ben-
eficial to listeners’ perceptual learning. This presumably oc-
curs because, rather than merely forming associations with a
single idiosyncratic token of a phoneme, with sufficient vari-
ability listeners are able to define a perceptual space that
includes both a prototype and several variants
e.g., Kuhl,
1991.
Dupoux and Green
1997 asked whether adaptation to
time-compressed speech might be affected by interrupting a
series of time-compressed sentences with presentation of
sentences heard at a different speech rate. The authors pre-
sented listeners with 10 sentences time-compressed to 38%
of their original duration followed by 5 uncompressed sen-
tences, and a final 5 sentences again compressed to 38% of
their original duration. Perceptual learning was assessed by
asking listeners, following each sentence, to report as many
words as possible from that sentence. Listeners’ recall accu-
racy improved over the course of the time-compressed sen-
tences, but showed a moderate drop for the two time-
compressed sentences immediately following the
uncompressed speech. However, listeners’ accuracy quickly
regained
and surpassed the level achieved at the end of the
initial 10-sentence set of compressed sentences. Dupoux and
Green
1997 concluded that there was a short-term, local
adjustment to the uncompressed speech that caused the drop
in performance, but that perceptual learning persisted
through a longer-term component, as evidenced by the rapid
recovery.
The initial drop in accuracy following a change in
speech rate reported by Dupoux and Green
1997 suggests
that listeners’ attentional focus may have indeed shifted to-
ward speech cues salient at the slower speech rate, hindering
their performance on the time-compressed sentences. It is
possible that, had there been more interruptions between the
exposure sentences, a summation of the small performance
drops associated with each interruption would have resulted
in a significant decline in overall learning. This would be
consistent with attentional processes that tune in to the cur-
rent or local speech rate at the expense of other speech rates.
Alternatively, it may be the case that such narrow local at-
tention is not necessary to adapt to speech at a new rate. The
primary goal of the current study was to investigate whether
frequent shifts in speech rate might have a significant impact
on perceptual learning.1702 J. Acoust. Soc. Am., Vol. 121, No. 3, March 2007adults. There are many reasons one would expect age to
adversely affect adaptation to rapid speech. First, older adults
typically have poorer hearing acuity than do young adults

Morrell et al., 1996, with this decline in auditory acuity
contributing to poorer performance on a variety of language
tasks
Sommers, 1997; Wingfield et al., 2006. Accompany-
ing these changes in peripheral hearing are age-related de-
clines in auditory temporal processing ability. In gap detec-
tion tasks, older adults typically require longer gaps between
sounds than do young adults to correctly discriminate one
from two tones
Schneider et al., 1994; Strouse et al., 1998;
Schneider and Hamstra, 1999. Fitzgibbons and Gordon-
Salant
1995 investigated listeners’ temporal resolution abil-
ity by comparing discrimination thresholds for changes in
the duration of tones embedded in sequences to those for the
same tones presented in isolation. They found that older
adults’ discrimination ability was significantly worse for
tones embedded in sequences compared to that of the young
adults. These findings have been upheld in studies using
modified natural speech sounds
Gordon-Salant et al., 2006.
Above and beyond these age-related changes in sensory
processing, older adults exhibit declines in several cognitive
domains important for language processing. These include
declines in attentional resources
McDowd and Shaw, 2000,
slower speed of information processing
Salthouse, 1996,a
diminished working memory capacity
Zacks et al., 2000,
and difficulty inhibiting irrelevant or distracting information

Tun et al., 2002. The result is that older adults perform
more poorly than their younger counterparts on a wide vari-
ety of language tasks, including exhibiting a greater diffi-
culty processing time-compressed speech
Gordon-Salant
and Fitzgibbons, 1993; Wingfield et al., 1999, 2003.
In addition to auditory and linguistic processing, adap-
tation to rapid speech involves learning new acoustic repre-
sentations of stored phonemic categories, and older adults
exhibit declines on a number of tasks involving learning and
memory. These include word generation, paired associates,
cued recall, and free recall
Craik, 1977; Craik et al., 1987;
Kausler, 1994. There is also evidence from the episodic
memory literature that supports age-related weakening of as-
sociations between items
Kahana et al., 2002; Wingfield and
Kahana, 2002. Although the literature on age-related
changes in perceptual learning is limited, the available evi-
dence would lead one to expect age-related deficits on this
front as well. For example, older adults show significant im-
pairments in prism adaptation
Fernandez-Ruiz et al., 2000
and semantic category visual search tasks
Gilbert and
Rogers, 1996; Rogers et al., 1994.
To the degree that adaptation to time-compressed speech
depends on the above-listed processes, such adaptation
should be adversely affected in adult aging. Contrary to this
expectation, it has recently been demonstrated that under
ideal training conditions—that is, an uninterrupted series of
training sentences in an otherwise quiet environment—older
adults initially adapt to time-compressed speech in a manner
comparable to young adults
Peelle and Wingfield, 2005.
However, this does not rule out age-related declines in per-Golomb et al.: Adaptation to time-compressed speech