Memory & Cognition
2000, 28 (6), 1029-1040
Listening to, and comprehending, natural speech re-
quires effective organization of the input as it is being
heard both in terms of linguistic constituents and higher
order elements at the discourse level. Although discourse
processing theories differ in detail, limitations in mem-
ory or processing resources are generally assumed to
limit the effectiveness of language processing, whether at
interpretive or postinterpretive stages (Caplan & Waters,
1999; Carpenter, Miyaki, & Just, 1994; Daneman & Mer-
ikle, 1996; Kintsch, 1994; van Dijk & Kintsch, 1983; Wa-
ters & Caplan, 1996). In using the term resources, we refer
to a pool of mental energy that is presumed to underlie the
successful execution of cognitive tasks. As such, tasks will
be less effectively conducted when their resource demands
are high or when the processing system has reduced re-
sources (Craik & Byrd, 1982; Kahneman, 1973; Salt-
house, 1991; Stine, 1995). One way memory or resource
constraints on language processing have been studied is
through experimental comparisons between healthy
young and older adults because of the common finding
that older adults are at a disadvantage relative to young
adults on tests of working memory or attentional resource
capacity (Craik & Byrd, 1982; Salthouse & Babcock,
1991; Wingfield, Stine, Lahar, & Aberdeen, 1988).
Although some theorists have focused on age-related
reductions in resource capacity, others have focused on
possible age differences in the effective allocation of
these resources. An example of this latter view is the self-
initiation hypothesis of Craik and Jennings (1992), which
suggests that older adults are less able to successfully di-
rect attentional resources to where they are needed, es-
pecially when performing a task that is particularly dif-
ficult. Applied to tasks such as discourse processing, this
view would imply that “automatic” aspects of discourse
processing should be performed similarly by young and
older adults, whereas less automatic aspects of the dis-
course processing task, such as elaborative inferencing
and the integration of information, would not be as con-
sistently well performed by older adults (Stine, Loveless,
& Soederberg, 1996). Consistent with this hypothesis, a
number of reading time studies have shown that older
adults, unlike young adults, are less likely to slow their
reading at sentence boundaries, an important area for inte-
gration during comprehension (Stine, 1990; Stine, Cheung,
& Henderson, 1995). That is, older adults are less likely
than young adults to allocate attentional resources where
text demands are greatest. Given that linguistic knowl-
edge is well preserved in normal aging (Light, 1990), age
differences in patterns of resource allocation in discourse
processing might be a major contributor to the signifi-
cant age-related declines in recall of written and spoken
discourse (Hartley, 1993; Wingfield & Stine, 1992).
One way to better understand the nature of older adults’
discourse processing is to examine the extent to which
age-related encoding failures relate to impaired memory
performance for discourse. Reading time studies using
the moving window or related techniques (Aaronson &
Ferres, 1984; Aaronson & Scarborough, 1976; Just, Car-
penter, & Woolley, 1982; Magliano, Graesser, Eymard,
Haberlandt, & Gholson, 1993) have shown that older
adults spend less time than young adults reading impor-
tant regions of text, although they are very similar to young
1029 Copyright 2000 Psychonomic Society, Inc.
We acknowledge support from NIH Grant R37 AG04517 from the
National Institute on Aging and support from the W.M. Keck Founda-
tion. D.T. also acknowledges support from NIH Training Grant T32
AG00204 to Brandeis University. We thank Lisa Connor, Fergus Craik,
Lisa Miller, Elizabeth Stine-Morrow, and Deborah Little for comments
on earlier versions of this manuscript. Correspondence should be ad-
dressed to A. Wingfield, Volen National Center for Complex Systems
(MS 013), Brandeis University, Waltham, MA 02454-9110 (e-mail:
wingfield@brandeis.edu).
Resource allocation during spoken
discourse processing: Effects of age and passage
difficulty as revealed by self-paced listening
DEBRA TITONE, KRISTEN J. PRENTICE, and ARTHUR WINGFIELD
Brandeis University, Waltham, Massachusetts
The allocation of processing resources during spoken discourse comprehension was studied in a
manner analogous to self-paced reading using the auditory moving window technique (Ferreira, Hen-
derson, Anes, Weeks, & McFarlane, 1996). Young and older participants listened to spoken passages in
a self-paced segment-by-segment fashion. In Experiment 1, we examined the influence of speech rate
and passage complexity on discourse encoding and recall performance. In Experiment 2, we examined
the influence of speech rate and presentation mode (self-paced vs. full-passage presentation) on recall
performance. Results suggest that diminished memory performance in the older adult group relative
to the young adult group is attributable to age-related differences in how resources were allocated dur-
ing the initial encoding of the spoken discourse.

1030 TITONE, PRENTICE, AND WINGFIELD
adults in many other ways (Hartley, 1986, 1993; Hartley,
Stojack, Mushaney, Annon, & Lee, 1994). In one study
that exemplifies this approach, Stine et al. (1996) visu-
ally presented expository passages to a group of young
and older adults for immediate recall. Each passage was
presented one sector at a time and a participant-initiated
keypress triggered presentation of each new sector. This
procedure allowed for reading time to be measured for
each sector of text, thus providing a measure of how
young and older adults allocated attentional resources to
different regions of the texts during encoding.
Stine et al. (1996) found that both age groups showed
a sensitivity in their reading times to different levels of
the texts (e.g., word, sentence, and discourse levels). Older
adults’ reading times were shorter than those of the
young, however, specifically at clause boundaries. Stine
et al. suggested that this difference in patterns of reading
time probably resulted in a less stable memory trace of
text information for older adults because clause bound-
aries are important computational points for integrating
information. They also argued that the similarities in
reading behavior between the groups could be seen as re-
flecting a lack of compensation by the older adults for
age-related memory or resource constraints. They con-
cluded that successful performance requires that older
adults modify their behavior to overcome these age-related
limitations.
The Unique Problem of
Spoken Discourse Comprehension
Although resource allocation during the reading of
discourse has received considerable attention, the results
from this work may not be entirely generalizable to the
domain of listening. Comprehenders of spoken discourse
are faced with an inherently transient and noisy signal,
and their control over the flow of information during
spoken conversation is bound to be much coarser than a
reader’s control of how the eye moves across written text.
Even in word-by-word or sector-by-sector reading time
experiments in which comprehenders cannot backtrack
to read previous portions of a text, it is still the case that
readers can take as long as they need to fully encode a
particular word or segment before initiating the next seg-
ment. Listeners, by contrast, must hold the transient acous-
tic forms in memory until they are processed. Such dif-
ferences might thus result in a more memory-demanding
task environment for listeners than for readers.
One difficulty researchers have faced in addressing the
question of spoken discourse processing is finding an
adequate experimental method with which to study im-
mediate encoding of spoken discourse (i.e., “listening”
behavior). Wingfield and Butterworth (1984) employed
a technique called spontaneous segmentation to study
the encoding of spoken discourse. In this paradigm, lis-
teners control the flow of spoken input by interrupting
recorded speech for intermediate recall at points of their
choosing. This technique shows that listeners will reli-
ably interrupt the prose input at commonly defined syn-
tactic boundaries (Wingfield & Butterworth, 1984). Addi-
tionally, Wingfield and Lindfield (1995) showed that
for both normal and rapid speech, young and older adults
typically interrupt speech for recall at similar points
(clause and sentence boundaries), although older adults
recall less of their selected segments than do young adults,
especially with faster speech rates. These results suggest
that the older adults’ maintenance of an encoding strategy
that was appropriate for young adults does not serve them
well for ultimately recalling information from spoken
passages.
Although this technique suggests strong similarities
between young and older adults in spoken language en-
coding, there might be differences in encoding patterns
that the spontaneous segmentation paradigm does not
capture. Young and older participants, for example, might
be differentially sensitive to text characteristics other than
syntactic boundary structure. Such text characteristics
might range from word-level variables, such as word fre-
quency, to the overall conceptual structure at the discourse
level.
A useful technique for answering this question is the
so-called auditory moving window (AMW) technique
developed by Ferreira and colleagues (Ferreira & Anes,
1994; Ferreira, Anes, & Horine, 1996; Ferreira, Hender-
son, Anes, Weeks, & McFarlane, 1996). Analogous to self-
paced reading, the method has been successfully used to
study how listeners encode spoken text. In this tech-
nique, participants are given direct control of the tempo-
ral flow of speech as it is being heard. Participants listen
to spoken language that is presegmented at units (e.g.,
word or multiword length) chosen by the experimenter.
The presentation of each segment is initiated by a lis-
tener-controlled keypress. Of interest is the duration of
listeners’ pauses between the end of one segment and their
initiation of the subsequent segment. Ferreira and col-
leagues have used this technique to examine listening be-
havior for word-by-word presentations of sentences and
have found that, similar to word-by-word reading tech-
niques, AMW is sensitive to lexical frequency effects
and syntactic garden path effects (Ferreira, Henderson,
et al., 1996).
A potential concern with the AMW technique is that
it presents spoken language in a way that listeners do not
normally encounter (i.e., in word or multiword segments).
Using this technique, however, Ferreira, Henderson, et al.
(1996) have shown positive effects for the influence of
prosodic information. This suggests that, although pause
information might be reduced, presenting spoken lan-
guage in segments does not appreciably disrupt the percep-
tion of intonation, vowel lengthenings, and stress patterns
that accompany natural speech (Ferreira, Henderson,
et al., 1996).
The Present Study
Given the work on adult age differences in discourse
processing during reading, we were specifically interested
in testing the extent to which resource limitations in