Journal of Speech and Hearing Research, Volume 39, 177-190, February 1996

Lexical Learning by Children With

Specific Language Impairment:
Effects of Linguistic Input Presented
at Varying Speaking Rates

Susan Ellis Weismer

Linda J. Hesketh filir"""p""""s""ss "0MMsa I
Walsman Center on Mental
Retardation and Human Development
and
Department of
Communicative Disorders The purpose of this study was to investigate the impact of speaking rate variations in the
University of Wisconsin-Madison linguistic input provided to children during a novel word learning task. Thirty-two school-age
children participated in this investigation, including 16 children with specific language impair-
ment (SLI) and 16 normal language (NL) controls matched on mental age (MA). The younger half
of the NL group also served as a vocabulary level comparison for the older half of the children
with SLI. No significant rate effects were found for comprehension of novel words, with all
children performing at relatively high levels of accuracy. The group with SLI demonstrated the
same recognition accuracy pattern as MA matched controls for target labels versus phoneti-
cally similar/dissimilar foils only for words trained at slow rate. Rate effects were most
pronounced for items with the highest difficulty level, namely production of novel words.
Children with SLI produced significantly fewer words that had been presented at fast rate
during training than NL children matched on mental age or vocabulary level. Individual
differences and production error patterns on fast rate items were examined. The finding that
variations in speaking rate had a disproportionate impact upon word learning for children with
SLI was interpreted within a framework of limited processing capacity.

KEY WORDS: specific language impairment, lexical learning, speaking rate, limited
processing capacity, linguistic input

There is considerable evidence that even though children with specific language
impairment (SLI) do not have generalized cognitive deficits, they demonstrate certain
perceptual and cognitive limitations including temporal processing and memory
difficulties (see reviews by Bishop, 1992; Ellis Weismer, 1993; Johnston, 1992;
Leonard, 1987). The pattern of difficulties exhibited across nonlinguistic as well as
linguistic tasks has prompted a number of investigators to posit that these children
have limited processing capacities (Bishop, 1992; Ellis Weismer, 1994; Gathercole &
Baddeley, 1990, 1993; Johnston, 1994; Johnston & Smith, 1989; Kirchner & Klatzky,
1985; Lahey & Bloom, 1994; Leonard, McGregor, & Allen, 1992; Montgomery, 1993).
Although there is reason to suspect that children with SLI have certain processing
constraints, very little is known about how the manner in which linguistic signals are
presented affects their ability to process and acquire language.
Prosodic characteristics of linguistic input are easily manipulated and have the
potential to influence various levels of linguistic processing, presumably having an
impact on auditory perception/comprehension of language, and thereby language
use (cf. Ellis Weismer, 1992). In the current study, the particular prosodic modifica-
tion that was investigated involved variations in speaking rate. Examination of
manipulations of the linguistic input may provide insight into factors that facilitate or

D 1996, American Speech-Language-Hearing Assoclation 177 0022-4685/96/3901-0177

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 178 Journal of Speech and Hearing Research 39 177-190 February 1996

deter initial language learning, as well as shedding light on with these children, for both clinical and academic reasons,
processing mechanisms underlying SLI. involves vocabulary learning. It is well established that
Various studies have examined the effect of manipulating children with SLI exhibit late onset of words and often have
speaking rate on linguistic processing by individuals with restricted vocabularies at the point when they enter school
normal language and those with different types of language (cf. Leonard, 1988). There has been considerable interest in
disorders. Fast speaking rates have been shown to have an examining factors related to lexical learning difficulties in
adverse effect on auditory comprehension and on produc- children with SLI (Dollaghan, 1987; Rice, Buhr, & Nemeth,
tion. Older adult listeners (with normal language abilities) 1990; Rice, Buhr, & Oetting, 1992; Rice, Oetting, Marquis,
exhibit significantly reduced auditory comprehension of pas- Bode, & Pae, 1994). Rice et al. (1992) examined the effect of
sages presented at fast speaking rates compared with their pause insertion in a quick incidental learning (QUIL) task.
performance on passages presented at normal rates They reported that this particular prosodic modification of
(Schmitt, 1983; Schmitt & Carroll, 1985; Schmitt & Moore, the linguistic input had no effect on the initial comprehension
1989). It has been shown that children with auditory percep- of unfamiliar words by kindergarten children with SLI. Con-
tual language problems have more difficulty perceiving trary to these findings, a recent study by Ellis Weismer and
speech presented at a fast rate than do their age-matched Hesketh (1993) found that speaking rate variations signifi-
peers (Manning, Johnston, & Beasley, 1977). Further, chil- cantly affected comprehension and production abilities of
dren with learning disabilities evidence less accuracy than kindergartners with SLI on a novel word learning task. Inthat
age-matched controls at repeating sentences that were study, novel word learning was assessed under linguistic
presented at fast speaking rates (McNutt & Chi-Yen Li, input conditions involving variations in speaking rate, em-
1980). In a related line of research, numerous investigations phatic versus neutral stress, or use of supplemental visual
by Tallal and colleagues have demonstrated temporal pro- cues (i.e., gestures). It was hypothesized that certain of
cessing constraints for rapidly presented verbal and nonver- these input manipulations could serve to reduce the de-
bal stimuli in children with SLI (Fellbaum, Miller, Tallal, & mands of linguistic processing in various ways and thus
Curtiss, 1994; Johnston, Stark, Mellits, & Tallal, 1981; Tallal mitigate processing constraints of children with SLI. Sixteen
& Piercy, 1973a, 1973b; Tallal, Stark, & Mellits, 1985). kindergarten children participated in that preliminary study,
Findings from these studies indicate that the amount of time 8 children with SLI and 8 with normal language (NL). Results
children with SLI require for sensory information processing for the Rate Condition indicated that both groups correctly
is substantially greater than that required by children with produced significantly more novel words trained at slow rate
normal language abilities and that significant differences than fast rate and significantly more words trained at normal
between children with SLI and NL on tasks of temporal rate than fast rate. Children also demonstrated significantly
integration disappear when stimuli are presented more better understanding of novel words trained at slow rate
slowly. than at fast rate. Although the group x rate interaction did
Based on the findings from speaking rate manipulation not reach statistical significance inthat limited sample study,
studies, there is evidence to suggest that slower rates of there was a clear trend for the group with SLI to exhibit more
speech can promote auditory comprehension. Several stud- pronounced effects in response to rate variations than the
ies have reported that slower rates of speech led to im- NL group in both comprehension and production. The
proved comprehension of spoken paragraphs for adults with purpose of the present study is to further investigate the
aphasia (Lasky, Weidner, & Johnson, 1976; Pashek & Brook- influence of speaking rate manipulations upon language
shire, 1982; Weidner & Lasky, 1976), though findings from learning in a larger sample of children with SLI, using a
other investigations have called into question the positive modification of the original novel word learning task that
effects of rate reduction for this population (Blumstein, Katz, provides an assessment of recognition abilities in terms of
Goodglass, Shrier, & Dworetsky, 1985). Investigations of accuracy and reaction time measures, as well as providing
speech rate reduction with children have demonstrated measures of comprehension and production accuracy.
beneficial effects of slower speaking rate for children with The particular limited capacity framework that was em-
learning disabilities (McCroskey & Thompson, 1973) as well ployed in the current investigation was Just and Carpenter's
as for children with acquired aphasia (Campbell & McNeil, (1992) Capacity Theory of comprehension. The main thesis
1985). Using a dual processing paradigm, Campbell and of this theory is that cognitive capacity constrains compre-
McNeil (1985) found that slowing the rate of the primary hension and does so more for some individuals than others.
sentence improved comprehension of a secondary sen- Capacity is defined in terms of the maximum amount of
tence that had been presented at normal rate. This result activation available in working memory to support storage
was interpreted within a limited capacity model (Kahneman, and processing. According to the Capacity theory, when
1973), such that it was hypothesized that slowing the rate of task demands exceed the available resources, both storage
the primary sentence allowed more cognitive resources to and computational functions are degraded. Just and Car-
be allocated to the processing of the secondary sentence, penter's theory is based on Kahneman's (1973) limited
thereby increasing comprehension levels. capacity theory of attention and also draws on Baddeley's
We were interested in extending the findings from re- (1986) notions of working memory, in that there is a trade-off
search focused on the impact of variations in speaking rate between information maintenance and computation. There
upon language processing, to explore how this type of are, however, important differences between Baddeley's
modification of the linguistic input might affect language notions of working memory and the view underlying the
learning in children with SLI. One area of particular concern Capacity theory. Baddeley proposes that working memory is

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 Ellis Weismer & Hesketh: Effects of Linguistic Input 179

composed of two distinct components, a central executive testing. Middle ear screening (per ASHA, 1990 guidelines)
and another component consisting of modality-specific and bilateral acuity screening (500, 1000, 2000, and 4000 Hz
storage buffers including the articulatory loop that is impli- at 20 dB HL per ASHA, 1985 guidelines) were conducted in
cated in language processing and retention. Rather than an audiometric sound booth. Visual screenings were per-
invoking modality-specific buffers, Just and Carpenter con- formed using the Snellen E chart. Children who served as NL
ceptualize working memory for language as a global set of controls had no history of speech, language, or hearing
processes and resources that perform language compre- problems, or any other exceptional educational needs ac-
hension. Although Capacity theory was developed as a cording to school records and background information
reading comprehension model, it has been extended to provided by parents. All children were monolingual, native
investigations of children with reading disabilities, and to English speakers. Based on parent report of their own
auditory language processing in adults with aphasia (cf. educational levels and occupations, families in both groups
Carpenter, Miyake, & Just, 1994). Similarly, it can be em- ranged from working class to upper-middle class using the
ployed to evaluate claims regarding capacity limitations in criteria employed by Hoff-Ginsberg (1991).
children with SLI and the impact that modifications of the All children with SLI had previously been diagnosed by a
linguistic signal have on their language processing and certified speech-language clinician and were receiving lan-
acquisition. guage services in their schools. Five of the 16 children in the
The research questions addressed by this study were: group with SLI were also demonstrating written language/
1. Do prosodic manipulations of the linguistic input involv- learning difficulties based on parent and school reports. A
ing variations in speaking rate affect lexical learning by battery of assessment measures was administered at the
children with SLI? same time as the experimental tasks for the purposes of
2. Do these linguistic input modifications result in different confirming diagnoses of language impairment and compar-
patterns of performance for children with SLI than for ing the performances of the children in the two groups. As
those with NL development? mentioned previously, nonverbal cognitive skills were as-
We hypothesized that presentation of linguistic stimuli at sessed using the Columbia and receptive vocabulary abili-
fast speaking rates may exceed children's capacity limita- ties were measured by the PPVT-R. Syntactic comprehen-
tions and lead to reduced novel word learning whereas the sion skills were assessed via the Grammatical Morpheme
additional processing time afforded by slow speaking rates (Section II)and Elaborated Sentences (Section III) subtests
may facilitate linguistic computation and storage, resulting in of the Test of Auditory Comprehension of Language-Re-
better learning. Additionally, we predicted that these input vised (Carrow-Woolfolk, 1985). Auditory comprehension
manipulations would have the most impact upon language was further evaluated by the Token Test for Children
learning by children with SLI given their presumed process- (DiSimoni, 1978). Twelve-minute narrative language samples
ing limitations. were analyzed using Systematic Analysis of Language Tran-
scripts (SALT) (Miller & Chapman, 1992) and findings were
compared to SALT Reference Data Base norms collected on
Method Madison area children (Leadholm, Miller, Contrucci, Britting-
ham, & Odell, 1992). It should also be noted that phonolog-
Participants ical skills of the children with SLI were assessed from the
spontaneous language samples. Based on this evaluation it
A total of 32 children participated inthis study, 16 children was determined that they did not demonstrate articulation
with SLI and 16 children with NL. Children with SLI had a errors on the sounds in the experimental target words or
mean chronological age of 7 years, 2 months (with a phonological processes such as final consonant deletion
standard deviation of 12 months). The group with NL also that would interfere with their production of these particular
had a mean CA of 7 years, 2 months (with a standard words.
deviation of 13 months). The group with SLI was comprised The results of the cognitive and language testing are
of 6 girls and 10 boys, whereas the group with NL included summarized in Table 1 in terms of group means and
3 girls and 13 boys. Children with SLI were compared to standard deviations (SD) for standard scores (or raw scores
same age children with NL who had equivalent mental ages where appropriate) for both the children with SLI and those
(MAs) on a nonverbal cognitive measure, the Columbia with NL. Although the mean score on the Columbia was
Mental Maturity Scale (Burgemeister, Blum, & Lorge, 1972). somewhat higher for the NL group than for the SLI group in
A subset of the children involving the oldest half of the group the MA comparison, a t-test revealed that this difference was
with SLI and the youngest half of the NL group was also not statistically significant and the overall distribution of
compared on the basis of a vocabulary match, using raw scores was reasonably well matched. As expected, the
scores from the Peabody Picture Vocabulary Test-Revised mean scores on all the language measures were significantly
(PPVT-R), Form M (Dunn & Dunn, 1981). lower for the group with SLI than for the NL controls
All children demonstrated normal-range cognitive abilities matched on MA (p < .05). Half of the children with SLI
(i.e., obtained Age Deviation Scores of at least 85 on the evidenced both receptive and expressive delays; the other
Columbia). None of the children evidenced physical, motor, half demonstrated difficulties in only one area (either com-
or emotional handicaps according to teacher and parent prehension or production). The vocabulary match involved
report and school records. Children in both groups dis- comparing older children with SLI to younger NL children
played normal hearing and vision as assessed at the time of who exhibited equivalent raw scores on the PPVT-R. The

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 180 Journalof Speech and Hearing Research 39 177-190 February 1996

TABLE 1. Summary of participant characteristics for the groups measures. Eight of the 16 children with SLI scored 2 SDs
with specific language impairment (SLI) and normal language below the mean on one or more language measure. Six
(NL) for the mental age comparison (MA) and the vocabulary
level comparison (VOCAB). Group means are reported for each children in the group with SLI performed below 1.5 SDs
variable, with standard deviations noted in parentheses. below the mean on at least one language measure and the
scores for the remaining 2 children fell more than 1 SD below
VOCAB the mean on two or more language measures. All of the
MA comparison comparison children with language disorders were judged to have a
SLI NL SLI NL specific language impairment such that there was a discrep-
(n = 16) (n = 16) (n = 8) (n = 8) ancy between their language skills and cognitive abilities
(e.g., the 2 children who met the minimum criterion of 1 SD
Age (months) 86 86 96 75 below the mean across several language measures dis-
(12) (13) (8) (7) played Age Deviation Scores of 114 and 116 on the
Columbia (ADSa) 105 110 102 113 Columbia).
(9) (9) (8) (12)
PPVT-Rb 97 113 95 114
(10) (11) (9) (13)
Raw score 87 86 Procedure
(8) (13)
Token (V)c 493 502 494 502 Children were seen individually for a total of three ses-
(6) (3) (7) (2) sions, each lasting approximately 1 hour. All testing was
Raw score 13 16 completed in the Language Processes Laboratory in the
(4) (2) Waisman Center. The first session was devoted to the
d
TACL (11) 41 53 41 57 administration of the majority of the assessment measures
(7) (11) (8) (13) (i.e., the Columbia, PPVT-R, language sample, hearing
Raw score 31 31
(6) (6) screening). During the second and third sessions the exper-
e
TACL (111) 43 57 41 60 imental task was administered, along with an auditory de-
(8) (7) (8) (6) tection measure, vision screening, and the remainder of the
Raw score 28 31 language assessment tests (TACL-R, Token). The experi-
(7) (4) mental task involved two training sessions-one in which
MLU - SALT' 5.3 8.1 6.1 8.1 three consonant-vowel-consonant (CVC) novel words were
(1.4) (1.8) (1.2) (2.4) taught and one in which three consonant-vowel-consonant-
aADS refers to Age Deviation Scores, which are standard score consonant (CVCC) targets were taught. The order of presen-
equivalents from the Columbia Mental Maturity Scale that have a tation of the experimental conditions (CVC vs. CVCC tar-
mean of 100 and standard deviation of 16.
bStandard score equivalents from the Peabody Picture Vocabulary gets) was counterbalanced across participants.
Test-Revised based on a mean of 100 and standard deviation of 15.
cStandard scores for Part V of the Token Test for Children, based on
a mean of 500 and standard deviation of 5. Experimental Task
dStandard scores from the Grammatical Morphemes Section II of
the Test of Auditory Comprehension of Language-Revised that have The basic paradigm was similar to that used in our prior
a mean of 50 and standard deviation of 10. investigation of novel word learning (Ellis Weismer & Hes-
eStandard scores from the Elaborated Sentences Section III of the
Test of Auditory Comprehension of Language-Revised that have a keth, 1993), with the addition of recognition items. Children
mean of 50 and standard deviation of 10. were shown a toy figure identified as an outerspace creature
fMean length of utterance (MLU) inmorphemes based on 12-minute named Sam and instructed that they were going to learn
narrative language samples analyzed using Systematic Analysis of some "funny-sounding" words in Sam's outerspace lan-
Language Transcripts (SALT). According to the SALT Reference guage. Modeling procedures were used to teach labels for
Data Base, the mean MLU for 7-year-olds' narrative samples is 7.32
with a standard deviation of 1.06 and the mean MLU for 9-year-olds' unusual objects, such that children heard a given target
narrative samples is 8.80 with a standard deviation of 1.64. word repeatedly being paired with a certain object. These
objects were ones that were judged to have low codability
vocabulary matched subgroups also demonstrated similar based on adults' inability to name them (as determined by
performance on other measures of language comprehen- responses from 20 adult judges who also served as listeners
sion (i.e., there were no significant differences between their for perceptual judgments of taped stimuli).
mean raw scores on these tests); however, it should be The novel target words were identified as the names of
noted that they differed in expressive language functioning. Sam's toys (i.e., object labels). The CVC targets consisted of
That is, the children with SLI evidenced significantly lower /pon/, /ted/, and /kub/, and the CVCC targets were /bimp/,
MLUs than the vocabulary-matched controls (p < .05). /tAnd/, and /gsnt/. Thus, the novel words all involved single-
None of the children selected for the NL group attained syllable forms comprised of early-developing sounds (stops
scores that fell below 1 SD below the mean for their CA on and the nasals /n/ and /m/). These target words were
any of the assessment measures. To be included in the presented in 5-syllable sentence frames so children heard,
group with SLI, children had to score more than 1.5 SDs for example, "Sam is by the /kub/." Each lexical item was
below the mean for their CA on at least one of the language consistently presented at a certain speaking rate for a given
measures administered at the time of testing or more than 1 child and this variable was counterbalanced across partici-
SD below the mean on two or more of the language pants. For instance, /pon/ was consistently presented at

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 Ellis Weismer & Hesketh: Effects of Linguistic Input 181

slow rate for one child, /ted/ at normal rate, and /kub/ at fast cally dissimilar (PD) novel word foils. The phonetically similar
rate; whereas, for another child /pon/ was always presented items varied from the target labels only in the final consonant
at normal rate, /ted/ at fast rate, and /kub/ at slow rate. (CVC) or final consonant blend (CVCC), that is, the PS item
Stimuli consisted of tape-recorded, natural speech that had for /kub/ was /kum/ and the PS item for /gent/ was /gEmp/.
been digitized (described below). The stimulus sentences Phonetically dissimilar items had no sounds incommon with
were presented at approximately 2.8 syll/sec for slow rate, target labels so they were maximally different, that is, the PD
4.4 syll/sec for normal rate, and 5.9 syll/sec for fast rate. item for /g£nt/ was /visk/. The PS and PD foils for the CVC
These particular target rates were chosen because they and CVCC target words are listed in Table 2. On the
correspond with values reported in the literature (cf. McNutt Recognition probes, CSpeech (Milenkovic, 1992) was used
& Chi-Yen Li, 1980; Schmitt & Moore, 1989), were used to present digitized stimulus utterances and to record re-
successfully in our prior investigation of speaking rate vari- sponse accuracy and reaction time. Responses were made
ations (Ellis Weismer & Hesketh, 1993), and were perceptu- via a button press on a mouse, using the index finger of the
ally discriminable (see Stimuli Validation section). preferred hand that was maintained in a uniform resting
Instructions for the comprehension and production por- position. Within the CSpeech program the resolution of
tions of the experimental task were as follows: measured values is 1 msec and resolution accuracy is 5
msec. Reported RTs on the Recognition probes reflect the
We're going to play a game. This is an outerspace man. His interval between the offset of the stimulus utterance and the
name is Sam. We're going to learn some funny-sounding words
in his outerspace language. These are words for the names of button press that terminated the timer.
Sam's toys. First, a person on the tape will tell you the names of On the Production and Comprehension probes children
the toys. Sometimes she will talk very fast [spoken at fast rate], were given feedback about the correctness of their re-
sometimes she will talk very slowly [spoken at slow rate], and sponses. Feedback consisted of comments such as "Good,
sometimes she will talk like we normally do [spoken at normal that's right" or "Yes, you got it" when the child responded
rate]. You listen and try to remember the words so when it's your
turn you can say the words or put Sam by the right toy. appropriately, or "No, that wasn't it" or "Good try, but that
wasn't quite right" for incorrect responses. The correct
Note that children were alerted about the variable that would object label was not mentioned in the feedback to ensure
be manipulated (i.e., speaking rate variations) through both that variation was not introduced in the number of expo-
verbal instructions and demonstration. The motivation for sures of the novel word each child received. Additionally,
providing these advance cues was based on pilot testing, children were given a penny for each correct response, so
which indicated that children often interrupted the task to both the verbal comments and tangible reinforcers served
question or comment on the varying speech rates when this as feedback concerning the correctness of the response. No
variable was not explicitly mentioned prior to the task. feedback was provided on the Recognition probes given the
The experimental task consisted of four types of trials: timed nature of the task.
Exposure trials, Production probes, Comprehension probes, The experimental task was divided into two sets of trial
and Recognition probes. During the Exposure trials, children blocks (see Appendix A). In Trial Block 1, a single exposure
were shown an array of three unfamiliar objects. The toy of each target word was followed by production and com-
figure of Sam was placed next to the target object and prehension probes. Then another exposure of each target
children heard the tape-recorded training stimulus sentence, was provided, followed by production, comprehension, and
"Sam is by the /pon/." Production probes, which immedi- recognition probes. This first block of trials was designed to
ately followed the exposure trials, involved a live-voice assess the initial phase of learning and was designated as
prompt, "What's this called?," as the experimenter pointed the Fast Mapping portion of the task. In the remainder of the
to each of the objects. Comprehension probes followed the task, target words were presented three times each (rather
production probes. On these probes an additional object than once) during each set of Exposure trials. The sequence
that served as a foil was added to the unfamiliar target of trials (exposure, production, comprehension, recognition)
objects. Children heard tape-recorded utterances instruct- was repeated four times within Block 2. Block 1 plus Block
ing them where to place Sam, e.g. "Put him by the /kub/." 2 constitute the Total Task, and represent the outcome of
On the Recognition probes, both accuracy and manual more extended training. The focus of the present report is on
RTs were recorded. Children heard sentences of the form
TABLE 2. Phonetically similar (PS) and phonetically dissimilar
"This is the " and were instructed to press a two- (PD) foils for the consonant-vowel-consonant (CVC) and con-
button mouse YES/NO. Instructions for the Recognition sonant-vowel-consonant-consonant (CVCC) target labels on
portion of the experimental task were as follows: the Recognition probes.

The woman on the tape isgoing to tell you the names of the toys. PS foils PD foils
If she says the right name for the toy, press the button with the
happy face. If she says the wrong name for the toy, press the CVC targets
button with the sad face. I want you to press the button as fast /kub/ /kum/ /wis/
as you can, but try to be right. Let's practice. /pon/ /pob/ /fik/
/ted/ /ten/ /ZAg/
Pretraining was conducted prior to the Recognition probes CVCC targets
using real objects to ensure that the children understood the /gent/ /gemp/ /visk/
task. On the Recognition probes, target labels (L) were /blmp/ /bint/ /fAntf/
/tAnd/ /tAmp/ /oesk/
presented, as well as phonetically similar (PS) and phoneti-

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 182 Journal of Speech and Hearing Research 39 177-190 February 1996

the Total Task. Combining across the two training sessions this investigation served as listeners. A paired-discrimination
(one for CVC targets and one for CVCC targets), the Total task format was employed such that two sentences were
Task involved 84 exposure trials, 60 production probes, 60 presented on each trial. Listeners heard a pair of sentences
comprehension probes, and 240 recognition probes per and were required to judge whether the first or second
child. In the current report, Total Task data are reported. A sentence was produced at a faster rate than the other or if
separate report of the Fast Mapping data from this study both sentences were spoken at the same rate. Rate varia-
and from another investigation of linguistic input modifica- tions in utterances selected for inclusion on the stimulus
tion are being reported elsewhere (Ellis Weismer & Hesketh, tapes were judged with 99% accuracy.
1995).
Auditory Detection Measure
Stimuli In addition to the experimental task, an Auditory Detection
Measure was administered to all children to assess manual
The stimuli for this study consisted of tape-recorded,
reaction times (RTs) to nonlinguistic stimuli and provide a
natural speech utterances produced by an adult female
baseline against which to assess RTs on the Recognition
speaker. Rate changes were accomplished via natural
probes. In this task, children were instructed to press a
speaking rate adjustments since the sampling techniques
button as quickly as possible in response to a white noise
used in electronically altered (time-compressed or time-
stimulus that followed an alerting tone by randomly varying
expanded) speech potentially involve intelligibility concerns
intervals. The alerting tone consisted of a 1000 Hz pure tone,
similar to those of synthesized speech (Greene, Manous, &
300 msec in duration (generated by an HP 204D Oscillator).
Pisoni, 1984; Logan, Pisoni, & Greene, 1985), especially for
The white noise imperative stimulus was also 300 msec in
children with language disorders (Massey, 1988). Stimulus
duration (generated by a Grason Stadler 455-B Noise Gen-
utterances were recorded on a Harman/Kardon CD491
erator). The interstimulus interval (ISI) between the alerting
cassette deck in an acoustically isolated booth. A head-
tone and the noise stimulus signal varied between 500 and
mounted microphone (Shure SM10A) was used so that
1300 msec in 200 msec steps (500, 700, 900, 1100, and
consistent mouth-to-microphone distance could be main-
1300 msec). The time between the response and the start of
tained. Approximately 10 productions of each stimulus sen-
the next trial (alerting tone) was 2000 msec. The CSpeech
tence were initially recorded. Following acoustic measure-
program was used to present sequences and to store
ments and perceptual judgments (described below), the
responses. The Auditory Detection Measure consisted of
best exemplar of each sentence was selected for inclusion
five practice trials, 30 pre-test trials (presented before the
on the actual experimental tapes. Computerized speech
experimental task), and 30 post-test trials (presented at the
analysis programs (VOCAL, a program developed at the
end of the experimental task). Detection responses were
Waisman Center and CSpeech, Milenkovic, 1992) were used
made by pressing a button on a mouse. A uniform resting
to digitize utterances and arrange them in the desired
hand position was maintained, with children using the index
sequences for presentation during the experimental task.
finger of their preferred hand to respond. RTs on the
Taped stimuli for the exposure trials and comprehension
Auditory Detection Task were measured from the onset of
probes on the experimental task were presented in the
the noise signal to the button press terminating the timer.
sound field via a Harman/Kardon CD491 cassette tape deck
with Polk Audiospeakers. Recognition stimulus utterances
were presented directly from the computer via headphones Scoring and Interrater Agreement
(Sony MDR-7502).
Responses were scored as correct on the Production
probes if all sounds were produced correctly in the appro-
Validation of Stimuli priate sequence. Each child's responses were initially re-
corded on-line and were audiotaped for later analysis by a
Stimuli were selected on the basis of a combination of second judge. Both judges have graduate degrees in Com-
acoustic and perceptual criteria. The VOCAL and CSpeech municative Disorders and have extensive experience with
programs were used to perform acoustic analyses of the phonetic transcription. Production data from 8 randomly
stimulus utterances. Articulatory rate was determined by selected children across the two groups were rescored by a
measuring total utterance durations and converting these to second transcriber. This constituted approximately 15% of
syllables/second. Total utterance durations were measured the total production data (291/1920 production responses).
from the onset of frication noise for sentences beginning Interrater agreement was 93.8% based on a phoneme-by-
with a fricative ("Sam . . . ") or from the stop burst phoneme comparison (874/932 phonemes). On the compre-
("Put..."), to the last glottal pulse of the sentence-final hension probes, responses were scored as correct if the
word. A tolerance range of .2 syllables/second was al- child placed Sam by the appropriate object. A second judge
lowed inthe target rates of 2.8, 4.4, and 5.9 syllables/second viewed videotapes of six randomly selected sessions (out of
for slow, normal, and fast rate utterances, which still resulted 64) and independently recorded responses to the compre-
in nonoverlapping speaking rates. Perceptual judgments hension probes. Interrater agreement for the comprehension
were obtained for utterances that met the target rates based data was 100% (168/168). Recognition responses and re-
on acoustic measurements. Twenty adults who had normal sponses to the Auditory Detection task were recorded by the
hearing and language and were naive to the hypothesis of computer.

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 Ellis Weismer & Hesketh: Effects of Linguistic Input 183

Procedural Validity TABLE 3. Group means (and standard deviations) for the chil-
dren with specific language impairment (SLI) and normal lan-
Six of the 64 experimental sessions were randomly se- guage (NL) for the mental age comparison (MA) and the vocab-
lected to be video/audiotaped for analysis of procedural ulary level comparison (VOCAB). Mean accuracy scores are
reported for production and comprehension. For recognition,
validity. Three measures were analyzed, including: (a) the both accuracy and manual reaction times (sec) are reported.
examiner's pairing of novel objects with novel words during
the exposure trials; (b)appropriate feedback from the exam- VOCAB
iner for correct/incorrect responses on production trials; and MA comparison comparison
(c) appropriate feedback for responses on comprehension SLI NL SLI NL
trials. Results of these analyses indicated that the examiner (n=16) (n=16) (n=8) (n=8)
correctly paired the novel objects and words during expo-
sure trials with 99.6% accuracy (242/243), and provided Production
correct feedback for responses on both comprehension and (Total = 60)
Mean 12.1 25.7 17.9 22.2
production trials with 100% accuracy (180/180). (SD) (10.6) (13.0) (12.0) (12.5)
Comprehension
(Total = 60)
Results Mean 41.9 49.8 47.6 47.5
(SD) (12.5) (8.8) (7.0) (10.1)
A mixed-model design was employed in which the be- Recognition accuracy
tween-subjects factor was Group (SLI vs. MA or older SLI vs. (Total = 120)
Mean 91.5 94.3 101.3 82.3
VOCAB) and the within-subjects factors were Rate (slow, (SD) (23.1) (25.7) (6.8) (31.5)
normal, fast) and Syllable Shape (CVC, CVCC). Two sepa- Recognition reaction time
rate analyses were conducted to assess the influence of (in sec.)
Group, such that all of the children with SLI were compared Mean 1.25 1.02 1.17 1.17
to the total group of MA controls in one analysis and a (SD) (.22) (.48) (.24) (.63)
subset of the experimental and control groups (i.e., older SLI
and younger controls matched on VOCAB) were compared
in a different analysis. Separate MANOVA repeated mea- older SLI and younger vocabulary matched groups. No main
sures analyses were conducted for production, comprehen- or interaction effects for Rate were observed for the com-
sion and recognition data, such that the dependent variables prehension data, across either comparison.
represented measurements of the same variable across time
(cf. SPSS Reference Guide, 1990, p.392). For the Recogni-
tion accuracy data, these analyses were conducted using Recognition Data
both the original percentage data (percentages were neces-
On the Recognition items, both accuracy and reaction
sary since there were unequal numbers of the various item
time were examined. For Recognition Accuracy for the MA
types) and arcsine transformations of the percentage data.
matched groups, a significant three-way interaction was
Similarly, MANOVA analyses were completed on both the
found involving Group x Item Type x Rate [F(4, 108) = 3.31,
original RT data from the Recognition probes and the log
p < .05], as illustrated in Figure 1. Post hoc tests (Tukeys)
transforms of these data. Results reported for Recognition
indicated that the group with SLI demonstrated the typical
are based on the statistical analyses of the transformed
pattern of item difficulty exhibited by the NL group only at
data. For post hoc comparisons, Tukey HSD tests were
Slow rate; that is, a pattern in which accuracy was signifi-
used. To analyze error patterns in the production data,
cantly higher for recognition of PD items than PS (p < .05)
nonparametric statistics (Wilcoxon-Mann-Whitney tests)
and significantly higher for the target labels than PS items
were used due to the small number of responses in each
(p < .05). At normal rate, the recognition accuracy for the
category per child (see Siegel & Castellan, 1988, pp. 128-
SLI group did not differ significantly across item types and at
137).
Fast rate only the PD items were recognized with signifi-
Group means and standard deviations for the Total Task
cantly more accuracy than the PS items (p < .05). For the
(untransformed) data on the Production, Comprehension,
Vocabulary level comparison, there was a significant Item
and Recognition probes are reported in Table 3.
Type effect for Recognition Accuracy [F(2, 24) = 17.1, p <
.05] in which the ordering from most to least accurate was
Comprehension Data PD > L > PS, with each of the three comparisons being
statistically significant (p < .05).
Results of the MANOVA revealed a significant Group Analyses of Recognition Reaction Time for correct re-
effect for Comprehension such that the MA matched con- sponses revealed a significant Item Type x Syllable Shape
trols performed better than the group with SLI [F(1, 30) = x Rate interaction for the MA comparison [F(4, 68) = 4.74,
4.26, p < .05], though both groups performed at relatively p < .05), as illustrated in Figure 2. For the phonetically more
high levels of accuracy (70% for the group with SLI and 83% complex CVCC targets, reaction time to PS items was
for the NL group). Chance level of responding on the significantly longer than to PD items across all speaking
comprehension items was 25%. There was no significant rates (p < .05), mirroring the recognition accuracy data with
difference between the comprehension performance of the respect to the relative difficulty of these two item types. For

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 184 Journal of Speech and Hearing Research 39 177-190 February 1996

RECOGNITION
MA MATCHED GROUPS
Item Type-Op
100 . B
-
-

* PD-NL
O PD-SLI
90 * L-NL
O O
o L-SLI
0
OJ 80 O A PS-NL
O

!
Q A
O a
O
O
A PS-SLI

70
I A
[]----
60

a0o __ I
--

BLOW NORMAL FAST

Rate
FIGURE 1. Mean percent Recognition Accuracy on the phonetically dissimilar foils (PD),
phonetically similar foils (PS), and target labels (L) for the group with specific language
impairment (SLI) and the MA matched group with normal language (NL) for the slow, normal,
and fast rate targets.

the less complex CVC targets, RT was not significantly A significant Group x Syllable Shape interaction was
different for the various item types except at Fast rate. For observed in Recognition Reaction Time for both the MA
CVC targets presented at Fast rate during training, RT to PS comparison [F(1, 17) = 6.63, p < .05] and the Vocabulary
items was significantly longer than reaction time to PD items level comparison [F(1, 8) = 5.36, p < .05]. For the MA
(p < .05), as was the case for the CVCC targets. No comparison, this interaction stemmed from the fact that the
significant rate effects in Recognition RT were found for the NL group responded significantly more quickly to CVC
Vocabulary comparison. targets than the group with SLI (p < .05). For the Vocabulary

RECOGNITION
MA MATCHED GROUPS
Item Type-Syll. Shape
2 -

* PD-CVC

o PD-CVCC
l
1.5 A la * L-CVC
A
A A a L-CVCC
Pi
W Id A Ps-Cvc
0 1 0 0 A PS-CVCC
0
o
Af
0.5

0 I B~~~~~~~
I !
SLOW NORMAL FAST

RATE
FIGURE 2. Mean Recognition reaction time (sec) for the MA matched groups on the consonant-
vowel-consonant (CVC) and consonant-vowel-consonant-consonant (CVCC) phonetically dis-
similar foils (PD), phonetically similar foils (PS), and target labels (L) for the slow, normal, and
fast rate targets.

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 Ellis Weismer & Hesketh: Efects of Linguistic Input 185

level comparison, the younger NL children responded sig-

nificantly faster to CVC items than CVCC (p < .05) and the PRODUCTION
MA MATCHED GROUPS
older children with SLI responded significantly faster than 100
the younger, NL children for the CVCC items (p < .05). This
was the only finding suggesting that the children with SLI
80
performed better than the younger vocabulary controls on
certain items. For the Vocabulary level comparison, a signif-
icant main effect for Item Type was also found [F(2, 16) = o
0 60
38.5, p < .05]. Post hoc analyses indicated that RTs
increased inthe following order, PD < L < PS, with all three 40
I0
comparisons being statistically significant (p < .05). These
results exactly parallel those for Item Type effects in Recog- 20
nition Accuracy; that is, the PD items resulted in the highest
number correct and the shortest RTs whereas the lowest 0
number correct and longest RTs were associated with PS Slow Normal Fast
items. RATE
A second set of analyses was conducted to evaluate LSLI EINL
Recognition Reaction Time on PD versus PS items, which FIGURE 3. Mean percent correct Production of novel target
both required a NO response, while omitting L items that words presented at slow, normal, and fast rate for the group
required a YES response. The basic pattern of findings was with specific language impairment (SLI) and the mental age
identical to the previously reported combined analyses for matched group with normal language (NL). (Error bars denote 1
both the MA comparison and the Vocabulary level compar- standard deviation above the mean.)
ison.
Results (based on t-test analyses) from the Auditory significantly fewer fast rate items correctly than the NL
Detection Measure administered before and after the exper- children matched on MA (p < .05) or vocabulary level (p <
imental task indicated that there were no statistically signif- .05). Further, the SLI group produced significantly fewer
icant differences in RTs across the groups or across pre- to items correctly that had been presented at Fast rate than at
post-task conditions. Group means and standard deviations Normal rate (p < .05). Results for the MA comparison are
for the Auditory Detection Measure are presented in Table 4. graphed in Figure 3 and those for the Vocabulary level
Since no significant group differences were found, actual comparison are presented in Figure 4.
scores (rather than adjusted scores) were used in the
analyses of recognition reaction time reported above.
Production Error Pattern Analysis

Production Data To further explore the Group x Rate interaction described

above, production error patterns were examined for the Fast
Analyses of production data indicated a significant main rate targets for the SLI and MA matched groups using
effect for Syllable Shape for the MA comparison [F(1, 30) =
8.42, p < .05] and the Vocabulary matched comparison [F(1,
14) = 11.61, p < .05], such that novel words consisting of a PRODUCTION
CVC syllable shape were produced correctly by all children VOCABULARY MATCHED GROUPS
significantly more often than those with a CVCC syllable 100
shape. A significant Group x Rate interaction was also
found for the MA comparison [F(2, 60) = 3.21, p < .05] and 80
the Vocabulary level comparison [F(2, 28) = 5.43, p < .05]. (-
Post hoc tests revealed that children with SLI produced a
0
0 60
0
TABLE 4. Means (and standard deviations) for pre- and post- U 40
a-
task reaction times (RT) on the Auditory Detection Measure for el
'U
the children with specific language impairment (SLI) and normal
language (NL) for both the mental age (MA) and vocabulary level
(VOCAB) comparison.

Pre-task
RT
Post-task
RT
20

0 FL Slow Normal Fast

RATE
MA comparison OSLI E]NL
SLI (n = 16) .635 (.300) .651 (.224)
NL (n = 16) .568 (.092) .599 (.149) FIGURE 4. Mean percent correct Production of novel target
VOCAB comparison words presented at slow, normal, and fast rate for the group
SLI (n = 8) .519 (.089) .573 (.073) with specific language impairment (SLI) and the vocabulary
NL (n = 8) .588 (.084) .657 (.153) matched group with normal language (NL). (Error bars denote 1
standard deviation above the mean.)

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 186 Journal of Speech and Hearing Research 39 177-190 February 996

nonparametric analyses (Wilcoxon-Mann-Whitney tests). The results of this investigation support the conclusion
Since production errors were subdivided into a number of that manipulations inthe speaking rate of the linguistic input
categories, this analysis was only completed on the largest influence lexical learning in children with SLI. Children in the
set of data involving the MA matched groups. There was no group with SLI produced significantly fewer novel words
significant difference in the number of attempted produc- correctly that had been modeled in utterances presented at
tions for the two groups (i.e., there were equivalent numbers fast rate than at normal rate, and significantly fewer fast rate
of "no responses"), indicating that both groups demon- words than NL peers. This finding indicates that the limita-
strated a basic understanding of the task and willingness to tions exhibited by children with SLI in processing rapidly
respond. However, there were a number of differences in presented information that have been extensively docu-
patterns of errors exhibited by the groups. The group with mented by Tallal and colleagues (cf. Tallal, 1988) also
SLI was significantly more likely than the NL group to translate into learning difficulties for these children.
convert the novel words trained at fast rate into real words Contrary to our expectations, presentation of novel words
(W= 735.5, p < .05). This effect could not be explained on at slow rate did not result in significantly better learning for
the basis of a misunderstanding of the task by the group the group with SLI. Unlike prior studies demonstrating
with SLI, since the groups were equivalent with respect to beneficial effects of slowed rate on auditory comprehension
their substitution of real words for novel words for targets for adults with aphasia (e.g., Pashek & Brookshire, 1982),
trained at Normal rate (W = 1110, p = .3323). Within-group our task involved target words embedded in sentences
analyses for the children with SLI indicated that real words rather than paragraphs. We might speculate that the distinc-
were substituted significantly more often for fast rate targets tion between linguistic input presented at normal versus
than for Normal rate targets (W = 855.5, p < .05). slow rates would become even more apparent due to
Children with SLI made significantly more changes in the cumulative effects of increased processing time and re-
syllable shape of the novel words trained at fast rate than the duced processing load as the task progresses beyond the
NL group, who rarely made such changes in their produc- individual sentence level to the level of connected discourse.
tions (W = 795.5, p < .05). These changes involved both In examining this issue of response to slow presentation
additions and omissions of sounds (e.g., /kub/ was incor- rate, individual analyses were revealing. When individual
rectly produced as /krub/ as well as /ku/). Thus, the failure to difference scores across the rates were computed, we found
maintain the basic CVC or CVCC syllable shape of the target that 6/16 (38%) of the children in the group with SLI
did not just appear to reflect a tendency to reduce the form evidenced benefits of slower speaking rates in Comprehen-
phonetically to increase the ease of production. sion whereas no NL children demonstrated consistent ben-
Mislabeling was another error pattern that distinguished efits for slower rate in terms of their comprehension scores.
the groups, though this type of error occurred less often than All but one of the children with SLI who appeared to perform
the previous two kinds of errors. The group with SLI was better at slow rate had evidenced language comprehension
significantly more likely than the NL group to confuse the deficits on at least one of our assessment measures. Differ-
label for one object with that for another novel object (W = ence scores across rates indicated that 8/16 (50%) of the
894.5, p < .05) (e.g., referring to the /pon/ as the /kub/). children with SLI demonstrated improved performance un-
An analysis of vowel and consonant errors was conducted der slow rate for production, as did a few of the younger NL
for the incorrect productions that did not involve the use of children. Six out of 8 of these children with SLI exhibited
real words or labels for other objects, that is, for instances in expressive language delays. It was also noteworthy that at
which the child's intended target appeared to be the novel the level of the individual child, there was a lack of corre-
word. Although the group with SLI made more errors of both spondence between comprehension and production perfor-
types than the NL group for targets trained at fast rate, only mance with respect to rate effects; that is, of the 8 children
the difference between the two groups' vowel errors was who demonstrated production benefits for slow rate, only 3
statistically significant (W = 1214.5, p < .05). also showed benefits for comprehension.
As far as the predictions of differential responsiveness of
Discussion the groups to these linguistic input modifications, some
group interaction effects were observed for recognition, but
Children with SLI were less proficient than MA matched the clearest pattern was found for production. Children with
peers at comprehending and producing novel words. Al- SLI produced significantly fewer fast rate target words than
though there were some group differences in Recognition, MA matched or Vocabulary matched peers. In our prelimi-
these items turned out to be quite easy for all children. No nary study, both the children with SLI and NL responded
significant group differences were observed in terms of more poorly to fast rate than other rates, but there was a
baseline levels of processing speed on the Auditory Detec- clear trend for the group with SLI to exhibit more pro-
tion Measure; that is, the groups did not differ in RTs to nounced effects in response to rate variations than the NL
nonlinguistic auditory stimuli. It is important to note, though, group in both comprehension and production. The lack of a
that the detection task did not entail memory or sequencing statistically significant group interaction effect in the prior
demands. With respect to the Vocabulary level comparison, study was most likely due to the relatively small sample size.
the performance of children with SLI was equivalent to that Consider now the pattern of results that was observed
of the younger controls with one exception: Children with with respect to comprehension versus production effects.
SLI demonstrated poorer production of words trained at fast Manipulations of speaking rate primarily affected production
speaking rates than the younger, NL controls. of novel words, rather than comprehension or recognition of

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 Ellis Weismer & Hesketh: Effects of Linguistic Input 187

the words. Examination of the overall levels of accuracy for than the NL controls. However, the group with SLI also was
comprehension, recognition, and production items, shows significantly more likely than NL children to confuse the label
that adjustments in the linguistic input have had the greatest for one object with that for another novel object. These types
effect on the most difficult items, namely production. Com- of mislabeling errors seem to suggest an additional problem
prehension accuracy levels ranged from 70%-83% correct with association processes involved in mapping the label to
(25% chance level) and recognition ranged from 76%-79% the appropriate referent or in lexical retrieval, rather than
correct (50% chance level), whereas production ranged simply a deficit in the phonological store of the word.
from 20%-43% correct. Consistent with our findings, we In terms of clinical implications, the findings indicate that
would expect on the basis of Capacity Theory that linguistic fast rates of presentation provide a degraded signal for
input modifications would have the greatest impact as children with SLI and should be avoided in instructional
cognitive demands of the task increased and as available situations. Individual analyses suggest that some children
resources were stretched to the limits. Although we found with language difficulties may find slowed speaking rate
rate effects in comprehension in our preliminary study, the beneficial, though this was not true for the group as a whole.
overall level of accuracy was lower than in the current study A good deal of individual variation would be expected in
since comprehension items were apparently more cogni- response to input modifications. In future studies we plan to
tively demanding for the younger, kindergarten children who assess the clinical utility of these input manipulations for
participated in that study. real-language targets with a limited number of children,
Rather than viewing this as an ease of task or cognitive using single-case treatment designs.
load difference, we might entertain the possibility that com- Prior to concluding, some general comments are war-
prehension and production are dissociable mechanisms. ranted pertaining to control group issues and interpretation
This claim has been made within the adult aphasia literature, caveats. In studies of language disorders, the basis for
as well as on the basis of evidence for individual differences comparison is critical to the interpretation of findings. Given
inearly patterns of development in language comprehension the nature of the questions posed by this investigation we
and production (Bates, 1993; Bates, Bretherton, & Snyder, opted to focus most heavily on an MA comparison, with a
1988). Although the focus of Just and Carpenter's (1992) secondary look at a vocabulary comparison. How should we
Capacity Theory is language comprehension, they have also interpret the fact that there were no significant differences
briefly addressed language production mechanisms, sug- between the group with SLI and younger, vocabulary-
gesting that capacity limitations similarly operate on speak- matched children, with the exception of production of fast
ing but that comprehension and production draw on sepa- rate targets? Does this mean that extant language level,
rate pools of working memory resources. Our prior findings vocabulary comprehension in this case, plays a central role
and those of others indicate that linguistic input modifica- in determining the extent to which children are able to learn
tions can operate on both comprehension and production; new words? We would argue against this interpretation on
nevertheless, the way in which they influence these lan- several counts. To begin with, one should be cautious about
guage mechanisms may be somewhat different due to interpreting the lack of statistically significant findings in
differing processing constraints. these relatively small subgroups. More importantly, vocab-
Next, we might consider the nature of the problem under- ulary level was not found to be significantly correlated with
lying the disproportionately poor production of fast rate performance on the novel word learning task in the current
targets by the children with SLI and question whether these study, even though the initial assumption that vocabulary
were the result of phonological working memory constraints knowledge should be relevant to this type of lexical task had
or more generalized capacity constraints. Gathercole and logical appeal. Rice and colleagues (Rice et al., 1992; Rice et
Baddeley (1990, 1993) have suggested that phonological al., 1994) also failed to find significant correlations between
memory deficits underlie SLI; however, van der Lely and PPVT scores and unfamiliar word learning in children with
Howard (1993) and Rice et al. (1994) report that their findings SLI. Interpretations of findings based on language matches
do not support this claim. Unlike Baddeley's (1986) theory of within investigations of language disorders have been ques-
working memory, Just and Carpenter (1992) do not posit tioned by Plante, Swisher, Kiernan, and Restrepo (1993).
modality-specific storage buffers akin to the articulatory Studies employing such designs usually equate children on
loop. Using Just and Carpenter's theory as a framework, we the basis of a single index of language (as we did in the
can speculate that capacity limitations may affect phonolog- present study) across a number of possible domains (syn-
ical representation or other aspects of lexical processing if tax, semantics, pragmatics) and processes (comprehension
activation is inadequate, but the source of the problem and production), so groups are not actually "language-
would not be seen as stemming from a specific phonological matched" in a more general sense. Further, when compar-
memory deficit tied to speech-based rehearsal mecha- ing younger, NL children to older children with language
nisms. In our investigation, there was some evidence of disorders, it is evident that they differ considerably in a
differences in the groups' abilities to recognize labels versus number of important ways, including world knowledge,
PS or PD foils, but, overall, these items were surprisingly conceptual abilities, memory span, and rehearsal strategies,
easy for all children. A production error analysis of fast rate that may affect learning and contribute to the findings as
target words suggested that capacity limitations affected much as the one aspect of language functioning in which
phonological representations, in that children with SLI made they are similar.
significantly more errors involving substitution of real words In conclusion, the results of this investigation indicate that
for novel words, changes in syllable shape, and vowel errors speaking rate variations influence the acquisition of novel

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 188 Journalof Speech and Heanng Research 39 177-190 February 1996

vocabulary by children with SLI. Presentation of novel words Carpenter, P., Miyake, A., & Just, M. (1994). Working memory
at fast speaking rates had a disproportionately adverse constraints in comprehension: Evidence from individual differ-
effect on the production of these words for children with SLI ences, aphasia, and aging. In M. A. Gernsbacher (Ed.), Handbook
of psycholinguistics (pp. 1075-1122). San Diego: Academic
compared to NL peers. As has been suggested by Rice and Press.
colleagues (1994), it seems likely that multiple, interacting Carrow-Woolfolk, E. (1985). Test for Auditory Comprehension of
factors may underlie the problems that children with SLI Language-Revised. Allen, TX: DLM Teaching Resources.
have in word learning. Findings from the present study are DiSimoni, F. (1978). Token Test for Children. Allen, TX: DLM
Teaching Resources.
consistent with the claim that processing capacity limita- Dollaghan, C. (1987). Fast mapping in normal and language-
tions, especially temporal processing constraints, appear to impaired children. Journal of Speech and Hearing Disorders, 52,
be at least one component of the difficulty that these 218-222.
children are experiencing. Like other researchers who have Dunn, L., & Dunn, L. (1981). Peabody Picture Vocabulary Test-
invoked various accounts of capacity limitation explanations Revised. Circle Pines, MN: American Guidance Service.
Ellis Weismer, S. (1992). The role of prosodic variables in language
of SLI (e.g., Bishop, 1992; Johnston, 1994), we acknowledge processing by children with specific language impairment. In R. S.
that its usefulness rests on the ability to make more specific Chapman (Ed.), Child talk: Processes in language acquisition and
predictions about the particular patterns of strengths and disorders (pp. 125-137). Chicago: Mosby Year Book.
weaknesses demonstrated by children with SLI and to tie Ellis Weismer, S. (1993). Perceptual and cognitive deficits in
children with specific language impairment: Implications for diag-
these more directly to observed profiles of linguistic abilities; nosis and intervention. In H. Grimm & H. Skowronek (Eds.),
we are working toward this goal in our ongoing research. Language acquisition problems and reading disorders: Aspects of
The implications of these results are that it may behoove us diagnosis and intervention (pp. 75-101). Berlin: de Gruyter.
to carefully consider how linguistic models are presented to Ellis Weismer, S. (1994). Factors influencing novel word learning
children and to seek more integrated frameworks of speech- and linguistic processing in children with specific language im-
pairment. Invited talk at the 15th Annual Symposium on Research
language-cognitive processing that can help us better un- in Child Language Disorders, Madison, WI.
derstand the task facing children with limitations in language Ellis Weismer, S., & Hesketh, L. (1993). The influence of prosodic
learning. and gestural cues on novel word acquisition by children with
specific language impairment. Journal of Speech and Hearing
Research, 36, 1013-1025.
Ellis Weismer, S., & Hesketh, L. (1995). Impact of linguistic input
Acknowledgments modifications on fast mapping by children with specific language
impairment. Manuscript in preparation.
We would like to thank Chris Everman Hollar and Cheryl Neylon Fellbaum, C. Miller, S., Tallal, P., &Curtiss, S. (1994). Evidence for
who have contributed to this project in various ways, including the relation between temporal processing deficit and specific
language sample and acoustic analyses, data entry, and reliability language impairment. Paper presented at the 15th Annual Sym-
scoring. Others involved at different phases of the project include posium on Research in Child Language Disorders, Madison, WI.
Pao-Hsiang Chi, Dawn Meyer, and Samantha Wolfson. Special Gathercole, S., & Baddeley, A. (1990). Phonological memory
thanks to the Madison, Middleton-Cross Plains, and Monona Grove deficits in language disordered children: Is there a causal con-
School Districts for their cooperation and to the speech-language nection? Journal of Memory and Language, 29, 336-360.
clinicians who helped recruit children, especially to Kathy Lyngaas. Gathercole, S., & Baddeley, A. (1993). Working memory and
Thanks also to the children and parents who participated in this language. Hove, UK: Lawrence Erlbaum Associates.
project. This project was funded by NIH NIDCD grant Greene, B., Manous, L., & Pisoni, D. (1984). Perceptual evaluation
#1R29DC01101 awarded to the first author. of DecTalk: A final report on version 1.8. Research on Speech
Perception (Progress Report No. 10). Bloomington, IN: Indiana
University.
Hoff-Ginsberg, E. (1991). Mother-child conversation in different
References social classes and communicative settings. Child Development,
62, 782-796.
ASHA. (1985, May). Guidelines for identification audiometry. Asha, Johnston, J. (1992). Cognitive abilities of language impaired chil-
27, 49-52. dren. In P. Fletcher & D. Hall (Eds.), Specific speech and language
ASHA. (1990). Guidelines for screening for hearing impairment and disorders in children (pp. 105-116). London: Whurr.
middle ear disorders. Asha, 32 (Suppl. 2), 17-24. Johnston, J. (1994). Cognitive abilities of children with language
Baddeley, A. (1986). Working memory. Oxford: Oxford University impairment. In R.V. Watkins & M. L. Rice (Eds.), Specific language
Press. impairment in children (pp. 107-121). Baltimore: Brookes Publish-
Bates, E. (1993). Comprehension and production in early language ing.
development. Monographs of the Society for Research in Child Johnston, J., & Smith, L. (1989). Dimensional thinking in language
Development, 58, 222-242. impaired children. Journal of Speech and Hearing Research, 32,
Bates, E., Bretherton, I., & Snyder, L. (1988). From first words to 33-38.
grammar: Individual differences and dissociable mechanisms. Johnston, R., Stark, R., Mellits, E., &Tallal, P. (1981). Neurological
Cambridge: Cambridge University Press. status of language-impaired and normal children. Annals of
Bishop, D. V. M. (1992). The underlying nature of specific language Neurology, 10, 159-163.
impairment. Journal of Child Psychiatry, 33, 3-66. Just, M., & Carpenter, P. (1992). A capacity theory of comprehen-
Blumstein, S., Katz, B., Goodglass, J., Shrier, R., & Dworetsky, sion: Individual differences in working memory. Psychological
B. (1985). The effects of slowed speech on auditory comprehen- Review, 99, 122-149.
sion in aphasia. Brain and Language, 24, 246-265. Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ:
Burgemeister, B., Blum, L., & Lorge, I. (1972). Columbia Mental Prentice-Hall.
Maturity Scale (3rd ed.). New York: Harcourt Brace Jovanovich. Kirchner, D., & Klatzky, R. (1985). Verbal rehearsal and memory in
Campbell, T., &McNeil, M. (1985). Effects of presentation rate and language-disordered children. Journal of Speech and Hearing
divided attention on auditory comprehension in children with an Research, 24, 446-453.
acquired language disorder. Joumal of Speech and Hearing Lahey, M., & Bloom, L. (1994). Variability and language learning
Research, 28, 513-520. disabilities. In G. Wallach & K. Butler (Eds.), Language learning

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 Ellis Weismer & Hesketh: Effects of Linguistic Input 189

disabilities in school-age children and adolescents (pp. 354-372). Rice, M., Buhr, J., & Nemeth, M. (1990). Fast mapping word
New York: Macmillan. learning abilities of language-delayed preschoolers. Journal of
Lasky, E., Weidner, W., & Johnson, J. (1976). Influences of Speech and Hearing Disorders, 55, 33-42.
linguistic complexity, rate of presentation, and interphrase pause Rice, M., Buhr, J., &Oetting, J. (1992). Specific-language-impaired
time on auditory-verbal comprehension of adult aphasic patients. children's quick incidental learning of words: The effect of a
Brain and Language, 3, 386-395. pause. Journal of Speech and Hearing Research, 35, 1040-1048.
Leadholm, B., Miller, J., Contrucci, V., Brittingham, K., & Odell, Rice, M., Oetting, J., Marquis, J., Bode, J., & Pae, S. (1994).
B. (1992). Language sample analysis: The Wisconsin guide. Mad- Frequency of input effects on word comprehension of children
ison, WI: Wisconsin Department of Public Instruction. with specific language impairment. Journal of Speech and Hear-
Leonard, L. (1987). Is specific language impairment a useful con- ing Research, 37, 106-122.
struct? In S. Rosenberg (Ed.), Advances in applied psycholinguis- Schmitt, J. (1983). The effects of time compression and time
tics (pp. 1-39). Cambridge, England: Cambridge University Press. expansion on passage comprehension by elderly listeners. Jour-
Leonard, L.(1988). Lexical development and processing in specific nal of Speech and Hearing Research, 26, 373-377.
language impairment. In R. Schiefelbusch & L. Lloyd (Eds.), Schmitt, J., &Carroll, M. (1985). Older listeners' ability to compre-
Language perspectives: Acquisition, retardation and intervention hend speaker-generated rate alteration of passages. Journal of
(2nd ed., pp. 69-87). Austin, TX: Pro-Ed. Speech and Hearing Research, 28, 309-312.
Leonard, L., McGregor, K., & Allen, G. (1992). Grammatical Schmitt, J., & Moore, J. (1989). Natural alteration of speaking rate:
morphology and speech perception in children with specific The effect on passage comprehension by listeners over 75 years
language impairment. Journal of Speech and Hearing Research, of age. Journal of Speech and Hearing Research, 32, 289-306.
35, 1076-1085. Siegel, S., &Castellan, N.J. (1988). Nonparametric statistics for the
Logan, J., Pisoni, D., & Greene, B. (1985). Measuring the segmen- behavioral sciences (2nd ed.). New York: McGraw-Hill.
tal intelligibility of synthetic speech. Research on Speech Percep- SPSS Reference Guide. (1990). Chicago: SPSS Inc.
tion, Progress Report No. 11. Bloomington, IN: Indiana University.
Tallal, P. (1988). Developmental language disorders. In J. Kavanagh
Manning, W., Johnston, K., & Beasley, D.(1977). The performance
of children with auditory perceptual disorders on a time-com- & T. Truss, Jr. (Eds.), Learning disability proceedings from the
pressed speech discrimination measure. Journal of Speech and national conference (pp. 181-272). Parkton: York Press.
Hearing Disorders, 42, 77-84. Tallal, P., & Piercy, M. (1973a). Defects of non-verbal auditory
Massey, H. (1988). Language-impaired children's comprehension perception in children with developmental aphasia. Nature, 241,
of synthesized speech. Language, Speech, and Hearing Services 468-469.
in Schools, 19, 401-409. Tallal, P., & Piercy, M. (1973b). Developmental aphasia: Impaired
McCroskey, R., & Thompson, N. (1973). Comprehension of rate rate of non-verbal processing as a function of sensory modality.
controlled speech by children with specific learning disabilities. Neuropsychologia, 11, 389-398.
Journal of Learning Disabilities, 6, 621-628. Tallal, P., Stark, R., &Mellits, D. (1985). Identification of language-
McNutt, J., & Chi-Yen Li, J. (1980). Repetition of time altered impaired children on the basis of rapid perception and production
sentences by normal and learning disabled children. Journal of skills. Brain and Language, 25, 314-322.
Learning Disabilities, 13, 25-29. van der Lely, H., & Howard, D. (1993). Children with specific
Milenkovic, P. (1992). CSpeech [computer program]. Madison, WI: language impairment: Linguistic impairment or short-term mem-
Paul Milenkovic. ory deficit? Journal of Speech and Hearing Research, 36, 1193-
Miller, J., & Chapman, R. (1992). SALT: Systematic Analysis of 1207.
Language Transcripts [computer program]. Madison, WI: Lan- Weidner, W., &Lasky, E. (1976). Interaction of rate and complexity
guage Analysis Laboratory, Waisman Center, University of Wis- of stimulus on the performance of adult aphasic subjects. Brain
consin-Madison. and Language, 3, 34-70.
Montgomery, J. (1993). Haptic recognition of children with specific
language impairment: Effects of response modality. Journal of
Speech and Hearing Research, 11, 273-290.
Pashek, G., & Brookshire, T. (1982). Effects of rate of speech and Received November 15, 1994
linguistic stress on auditory paragraph comprehension of aphasic Accepted May 15, 1995
individuals. Journal of Speech and Hearing Research, 25, 377-
383.
Plante, E., Swisher, L., Kiernan, B., & Restrepo, M. (1993). Contact author: Susan Ellis Weismer, PhD, University of Wiscon-
Language matches: Illuminating or confounding? Journal of sin-Madison, Waisman Center Room 473, 1500 Highland Avenue,
Speech and Hearing Research, 36, 772-776. Madison, WI 53705-2280. E-mail: SWEISMER@vms. macc.wisc.edu

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
 190 Journal of Speech and Heanng Research 39 177-190 February 1996

Appendix
Format of Trials

Trial Block 1 The A, B, C sequence is repeated a total of two times, Trial Block 2 The A, B, C sequence is repeated four times, with D
followed by D (i.e., ABC ABCD). occurring after the second and fourth presentation
(i.e. ABC ABCD ABC ABCD).
A. Exposure Trials: "Sam is by the _ " (Fast rate)
(3 items) "Sam is by the " (Slow rate) A. Exposure Trials - 9 items, 3 repetitions of each word
"Sam is by the _ " (Normal rate)
B. Production Probes - 6 items, each word probed twice
B. Production Probes:"What's this called?" (Live voice)
(Normal rate target) C. Comprehension Probes - 6 items, each word probed twice
(3 items) (Fast rate target) D. Recognition Probes - 24 items, each word probed four times
(Slow rate target) plus two probes of the phonetically
C. Comprehension Probes: Child is instructed to place Sam by similar and dissimilar word for each
the correct toy. target
(3 items) "Put him by the " (Slow rate target) Trial Block 1 = Fast Mapping
"Put him by the " (Normal rate target) Trial Block 1 + Trial Block 2 = Total Task
"Put him by the " (Fast rate target)
D. Recognition Probes: "This is the "
(Yes/No response; button press)
(12 items; Each target word is presented twice, plus one
word that is phonetically similar and one that is dissimilar to
the target)

Downloaded From: http://jslhr.pubs.asha.org/pdfaccess.ashx?url=/data/journals/jslhr/929263/ by a La Trobe Univ User on 05/09/2017

Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx