Академический Документы
Профессиональный Документы
Культура Документы
net/publication/321658549
CITATIONS READS
0 124
7 authors, including:
Malcolm R Mcneil
University of Pittsburgh
130 PUBLICATIONS 2,544 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
New method for diagnosing acquired apraxia of speech - http://goo.gl/J429Ca View project
All content following this page was uploaded by Malcolm R Mcneil on 12 January 2018.
Clinical Focus
Purpose: The purpose of this study was to evaluate the Results: All participants with no brain injury and 72.7%
ability of persons with aphasia, with and without hearing of the participants with aphasia (24 out of 33) completed
loss, to complete a commonly used open-set word the NU-6. Furthermore, all participants who were unable
recognition test that requires a verbal response. Furthermore, to complete the NU-6 were able to complete the Picture
phonotactic probabilities and neighborhood densities of Identification Task. There were significant group differences
word recognition errors were assessed to explore potential on NU-6 performance. The 2 groups with normal hearing
underlying linguistic complexities that might differentially had significantly higher scores than the 2 groups with
influence performance among groups. hearing loss, but the 2 groups with normal hearing and
Method: Four groups of adult participants were tested: the 2 groups with hearing loss did not differ from one
participants with no brain injury with normal hearing, participants another, implying that their performance was largely
with no brain injury with hearing loss, participants with brain determined by hearing loss rather than by brain injury or
injury with aphasia and normal hearing, and participants with aphasia. The neighborhood density, but not phonotactic
brain injury with aphasia and hearing loss. The Northwestern probabilities, of the participants’ errors differed across
University Auditory Test No. 6 (NU-6; Tillman & Carhart, 1966) groups with and without aphasia.
was administered. Those participants who were unable to Conclusions: Because the vast majority of the participants
respond orally (repeating words as heard) were assessed with aphasia examined could be tested readily using an
with the Picture Identification Task (Wilson & Antablin, 1980), instrument such as the NU-6, clinicians should not be
permitting a picture-pointing response instead. Error patterns reticent to use this test if patients are able to repeat single
from the NU-6 were assessed to determine whether words, but routine use of alternative tests is encouraged for
phonotactic probability influenced performance. populations of people with brain injuries.
A
common impression among audiologists is that notion that PWA and other neurogenic communication
persons with aphasia (PWA) and other neuro- disorders are difficult to test does raise the concern about
genic communication disorders make up a limited appropriate service delivery for these patients and whether
portion of the clinical census but require extra time and they have reduced likelihood of referral for audiological
special adjustments in order to complete a standard audio- assessments and rehabilitation services, such as hearing
metric test battery (Palmer, Adams, Durrant, Bourgeois, & aids, auditory training, and counseling. This is a substan-
Rossi, 1998; Wilson, Shanks, & Flowler, 1981). However, tive issue because the epidemiological data on hearing loss
little has been published to support this impression. The and neurogenic communication disorders suggest that both
disorders become more common in early middle age, with
a
rates accelerating with increasing age (Agrawal, Platz, &
Geriatric Research, Education, and Clinical Center, VA Pittsburgh
Niparko, 2008; Chapman & Ulatowska, 1992). As the
Healthcare System, PA
b
Department of Communication Science and Disorders, University of
mean age of the population in the United States continues
Pittsburgh, PA to rise, so too will the occurrence of age-related comorbid
Correspondence to Sheila R. Pratt: spratt@pitt.edu
conditions.
Jillyn Roxberg is now at the University of Pittsburgh Medical
Center, PA.
Neurogenic Communication Disorders
Amanda J. Ortmann is now at Washington University, St. Louis, MO.
Editor-in-Chief: Sumitrajit Dhar It is estimated that 20% of the population with speech
Editor: Lauren Calandruccio and language impairment are over 65, and the majority of
Received April 17, 2017 these individuals have neurogenic disorders (Fein, 1983).
Revision received June 23, 2017
Accepted August 1, 2017 Disclosure: The authors have declared that no competing interests existed at the time
https://doi.org/10.1044/2017_AJA-17-0041 of publication.
Figure 1. Average audiometric threshold for each group. Error bars indicate ±1 SD. PWA-NH = persons with aphasia with normal hearing;
PWA-HL = persons with aphasia with hearing loss; NBI-NH = individuals with no brain injury with normal hearing; NBI-HL = individuals with
no brain injury with hearing loss.
Note. PICA = Porch Index of Communicative Ability; CRTT = Computerized Revised Token Test; NBI-NH =
individuals with no brain injury with normal hearing; NBI-HL = individuals with no brain injury with hearing loss;
PWA-NH = persons with aphasia with normal hearing; PWA-HL = persons with aphasia with hearing loss.
latency response testing was completed with click stimuli Test (Arvedson, McNeil, & West, 1985) was administered
presented at 90 dB nHL and the N1-P2 (slow vertex poten- to assess sentence-level auditory language processing. It
tial with 500 Hz tone bursts presented at 50 dB HL). These was administered multiple times across a range of intensity
potentials were evoked to assess the integrity of the auditory levels, but only their asymptotic performance was used in
pathway and primary auditory cortices. The stimuli and this analysis. Both tests of aphasia were completed in a
presentation levels were selected to ensure audibility across sound-attenuated booth by licensed and ASHA-certified
all participants. All electrical recordings were made from speech-language pathologists and audiologists with previous
an electrode at vertex referenced to the mastoids bilaterally training and experience with the tests. Demographic data
(two-channel recording for auditory brainstem response and performance on the PICA and the overall mean CRTT
and middle latency response; single-channel with tied mas- score are shown in Table 2.
toids for N1-P2). Waves V, Pa, N1, and P2 latencies and
amplitudes were measured. The results were compared with Word Recognition Testing
local age-corrected and hearing loss–corrected norms and The NU-6 recordings from the Department of Veterans
needed to be within 3 SDs of the mean for participant inclu- Affairs compact disc were used as stimuli and included a
sion. Only one participant was excluded from the study female speaker (Department of Veterans Affairs, 1998).
because of the electrophysiolocal testing, and he was later The 50-item word lists were subdivided into half-lists of
diagnosed with a brain tumor. This participant would have 25 words, which is a common clinical practice and has been
been in the NBI-NH group if included. validated for the NU-6 in previous studies (Dubno, Lee, Klein,
The PICA was administered to document the pres- Matthews, & Lam, 1995). For this study, each ear was tested
ence and profile of aphasia. The PICA includes four hier- separately with one of four half-lists (1A, 1B, 2A, 2B). The
archical ordered subtests that contribute to the verbal PICA recordings were played from a compact disc recorder/player
score. These subtests were examined to determine whether (Denon DN-T645) and routed through a clinical audiom-
phonological output and speech motor control, rather than eter (GSI 61) and presented via insert earphones (ER-3)
language, explained the ability to complete the NU-6. Sub- at 40 dB SL (re: three-frequency pure-tone average). If
test XII requires the simple repetition of the word form, performance was lower than expected (below the lower
which necessitates the engagement of the output phonology 95% confidence interval, as established by Dubno et al.,
and speech motor control systems. It does not require accu- 1995), the test was readministered at 50 dB SL using a dif-
rate visual recognition of the simple objects used in the test ferent half-list. No presentation level exceeded a participant’s
but does require sufficient auditory perceptual resolution to uncomfortable loudness level. If performance improved
encode the speech sounds from the acoustic stimulus. Further- and was within the 95% confidence intervals established by
more, semantic, lexical, and phonological recall of the word Dubno et al. (1995), the results obtained at the elevated
form is not required for correct performance. Therefore, Sub- level were used in the analyses; otherwise, the initial test
test XII (imitate/repeat the names of common objects) is the results were used. The participants who were slow to respond
least difficult of the verbal PICA subtests for most PWA. or appeared to struggle with verbal responses were allowed
In contrast, Subtest I requires sentence formulation after extra time by pausing the recording between words. The
accurate visual recognition of the test objects. Because all duration and number of pauses were not recorded. The par-
levels of linguistic formulation (semantics, syntax, lexical, ticipants were instructed to repeat the words that they heard
and phonologic) and recall from long-term memory are as clearly as possible, and the experimenter recorded errors
required for successful completion of this subtest, it is with broad phonetic transcription. Word recognition was
usually the most difficult of the verbal PICA subtests. The assessed with stimuli from the PIT (Wilson & Antablin,
task requirements for the other two verbal subtests are more 1980, 1982) if a participant was unable to complete the NU-6
intermediate in difficulty for typical PWA. as evidenced by an inability to repeat the test stimuli after
A 55-item version of the CRTT (McNeil et al., 2015) reinstruction and face-to-face practice with the examiner,
that mirrored the noncomputerized 55-item Revised Token or if the participant communicated that they were unable to
Subject Group Age Sex HFPTA (dB HL) Overall PICA Auditory PICA Verbal PICA CRTT PIT (% correct)
Note. NU-6 = Northwestern University Auditory Test No. 6; HFPTA = high-frequency pure-tone average for thresholds obtained at 1000,
2000, and 4000 Hz; PICA = Porch Index of Communicative Ability; CRTT = Computerized Revised Token Test; PIT = Picture Identification
Task; PWA-NH = persons with aphasia with normal hearing; PWA-HL = persons with aphasia with hearing loss; M = male; F = female.
the PWA who were able to complete the NU-6 had signifi- difference between observed and predicted group membership
cantly higher overall PICA scores than those who were as classified by the verbal PICA score, Hosmer-Lemeshow
unable to complete the test, Pillai’s trace = .260, F(3, 29) = χ2(9) = 12.250, p = .200, and the overall classification rate
3.396, p = .031, ηp2 = .260. They also had significantly was good with a receiver operating characteristic area of
higher verbal (oral language production) PICA scores than 0.801. A cutoff verbal PICA score of 11.64 (i.e., cutoff
those who were not able to complete the test (see Table 4), probability value of .740) could be used for classification
F(1, 31) = 9.519, p = .004, ηp2 = .235. However, there was no with minimal false negative and false positive rate using
difference between these two groups on the auditory PICA Youden’s index. Seventy percent of the PWA were correctly
or overall mean CRTT score, F(1, 31) = 2.122, p = .155, classified: 75% for those able to complete the NU-6 and
ηp2 = .064 and F(1, 31) = 0.002, p = .967, ηp2 < .001, 55.6% for those unable to complete the NU-6.
respectively. A separate model selection analysis on the ability to
The results of an all possible subsets model selection complete the NU-6, as predicted by PICA subtests, generated
analysis suggested the verbal PICA score was the best four comparable predictive models according to the Mallows’s
predictor of the ability to complete the NU-6 among Cp, AIC, SBC and Cox-Snell residuals. Subtests XII (imitate/
hearing status, age, verbal PICA, auditory PICA, and repeat the names of common objects) and IV (name common
overall mean CRTT based on the Mallows’s Cp, Akaike objects) significantly predicted the ability to complete
Information Criterion (AIC ), Schwarz Bayesian Criterion the NU-6 (Model 1), χ2(2) = 10.173, p = .006, Negelkerke
(SBC ), and Cox-Snell residuals, all diagnostic measures of R2 = .384. The ability to complete the NU-6 also was sig-
model quality and fit. The verbal PICA was significantly nificantly predicted by Subtests XII and IX (complete sen-
better than the other two test measures, χ2(1) = 7.362, tences by saying the names of common objects; Model 2),
p = .007, Negelkerke R2 = .290. There was no significant χ2(2) = 10.101, p = .006, Negelkerke R2 = .382, and by
Figure 2. Mean Porch Index of Communicative Ability (PICA) and overall mean Computerized Revised Token Test (CRTT) score for the
participants with aphasia who were able to perform the Northwestern University Auditory Test No. 6 (NU-6; solid black bars) and those
who were unable to perform the Picture Identification Task (PIT; gray bars). Error bars indicate ±1 SD. PWA NU-6 = persons with aphasia
who were able to complete the NU-6; PWA PIT = persons with aphasia who were only able to complete the PIT.
Able 12.72 2.04 11.01 2.48 12.83 2.54 12.71 2.57 14.32 1.32
Unable 9.58 3.79 8.14 2.82 9.84 4.27 9.51 4.13 10.81 4.22
Note. NU-6 = Northwestern University Auditory Test No. 6; PICA = Porch Index of Communicative Ability.
Subtest XII and Subtest I (describe the function of common Youden’s index. With this cutoff, 82% of the participants
objects; Model 3), χ2(2) = 9.665, p = .008, Negelkerke R2 = were correctly classified (87.5% were able to complete the
.368. In a fourth model, Subtest XII alone was a significant NU-6; 66.7% were unable to complete the NU-6).
predictor of the ability to complete the NU-6, χ2(1) = The nine PWA who completed the PIT as an alterna-
9.635, p = .002, Negelkerke R2 = .367. Given Model 4, each tive speech recognition test produced scores ranging from
unit increase in Subtest XII score increased the likelihood 84% to 100% correct. The participants in the PWA-NH
of being able to complete the NU-6 by 66%, B = 0.505, exp group had higher scores (M = 93.33, SD = 5.47) than those
(B) = 1.657, χ2(1) = 4.655, p = .003. Model comparisons indi- in the PWA-HL group (M = 85.33, SD = 5.47), but statisti-
cated that Subtest XII was the major contributor because cal comparisons were not performed due to the small sample
none of the other three subtests significantly predicted the size. Because the equivalence between the NU-6 score and
ability to complete the NU-6 after adjusting for the Sub- the PIT score has not been documented in a quantitative
test XII performance. However, Subtest XII alone was not way, no comparison was conducted between the two tests.
adequate for classification, because there was a significant
difference between observed and predicted group member- Performance on the NU-6
ship, Hosmer-Lemeshow χ2(6) = 12.78, p = .047. The combi-
nation of Subtest XII and Subtest I (Model 3) classified The symmetry of the pure-tone thresholds between
the best among the competing models, Hosmer-Lemeshow ears was confirmed. There were no significant differences
χ2(6) = 7.04, p = .633, and the overall classification rate on pure-tone average (average of thresholds for 500, 1000,
was good with a receiver operating characteristic area of 2000, and 4000 Hz) between ears for any of the groups:
0.838 (see Figure 3). A cutoff probability value of .744 was NBI-NH, t(29) = −0.356, p = .725; NBI-HL, t(41) = 0.779,
used to minimize both false negative and positive rate using p = .440; PWA-NH groups, t(24) = 0.706, p = .487; or
PWA-HL, t(7) = 1.330, p = .225. Despite symmetrical
pure-tone averages, some participants demonstrated minor
Figure 3. The receiver operating characteristic curves for the verbal ear differences on the NU-6. Thirty-one participants had
Porch Index of Communicative Ability (PICA) score and two verbal
PICA subtest scores (Subtests I and XII) predicting the ability to identical NU-6 scores for both ears. Of the remaining par-
complete the Northwestern University Auditory Test No. 6. AUC = ticipants, performance differences between ears ranged
area under the curve. from 4% to 16%, which corresponded to one to four test
items in error. The magnitude of the ear differences on the
NU-6 was significant for groups (see Figure 4), F(3, 101) =
7.857, p = .0009, ηp2 = .189, and found for all four groups:
NBI-NH, F(1, 101) = 20.712, p = .0009; NBI-HL, F(1, 101) =
43.362, p = .0009; PWA-HL, F(1, 101) = 3.660, p = .008;
and PWA-NH, F(1, 101) = 4.879, p = .0009. Although rela-
tively small, subsequent analyses of the NU-6 scores were
based on the performance of the better ear because of the
significant ear differences.
For those participants who completed the NU-6, a
multiple regression analysis was conducted to determine
whether age and verbal PICA, auditory PICA, overall
PICA, and overall mean CRTT score significantly predicted
NU-6 performance (in rau) and should be considered as
covariates in the analysis. The regression was significant,
F(5, 90) = 4.552, p = .001, adjusted R2 = .157, and the
results showed a significant negative prediction of NU-6
rau score by age, B = −0.751, F(1, 90) = 21.520, p = .0009,
ηp2 = .193. However, no significant prediction of the NU-6
performance was found for the other factors, including
verbal PICA scores, B = 5.090, F(1, 90) = 2.519, p = .116,
ηp2 = .027 (see Figures 5 and 6). Hence, age was used as a were all other assumptions for the following analyses. After
covariate in the following group comparisons. removing the variance due to differences in age, a significant
A one-way between-subjects analysis of covariance was group effect was observed, F(3, 91) = 16.772, p = .0009,
performed on the NU-6 rau scores as a function of group ηp2 = .356. As expected, the NBI-NH group (M = 106.326,
after adjusting for age. The homogeneity of regression assump- SE = 2.569) had significantly higher scores than the NBI-HL
tion was met, F(3, 88) = 0.937, p = .426, ηp2 = .031, as (M = 84.100, SE = 2.180), F(1, 91) = 42.617, p = .0009,
Figure 5. The relationship between age and the Northwestern University Auditory Test No. 6 (NU-6) score (in rau).
Regression analysis indicated a significant correlation at p < .05. PWA-NH = persons with aphasia with
normal hearing; PWA-HL = persons with aphasia with hearing loss; NBI-NH = individuals with no brain injury
with normal hearing; NBI-HL = individuals with no brain injury with hearing loss.
ηp2 = .319, and PWA-HL (M = 77.712, SE = 6.529), n = 4), including combined information from both ears
F(1, 91) = 16.196, p = .0009, ηp2 = .151. Similarly, the tested separately, because no significant differences were
PWA-NH (M = 101.282, SE = 3.284) had significantly observed between ears on any of the phonotactic and neigh-
higher scores than the NBI-HL group, F(1, 91) = 18.382, borhood measures: average phoneme probability, F(1, 64) =
p = .0009, ηp2 = .168, and PWA-HL group, F(1, 91) = 9.869, 2.511, p = .118, ηp2 = .036; average biphoneme probability,
p = .002, ηp2 = .098. However, the two groups with normal
hearing did not differ from each other, F(1, 91) = 1.508,
p = .223, ηp2 = .016, nor did the groups with hearing loss, Figure 7. The Northwestern University Auditory Test No. 6 (NU-6)
F(1, 91) = 0.887, p = .349, ηp2 = .010. Word recognition word recognition performance by group after adjusting for age. The
performance for each group is illustrated in Figure 7. results are expressed in rau. Covariates appearing in the model
were evaluated at age = 63.89. Error bars indicate ±1 SE. PWA-NH =
persons with aphasia with normal hearing; PWA-HL = persons with
aphasia with hearing loss; NBI-NH = individuals with no brain injury
Phonotactic Probabilities and Neighborhood with normal hearing; NBI-HL = individuals with no brain injury with
Density of Error Productions hearing loss.
F(1, 64) = 2.866, p = .095, ηp2 = .041; and neighborhood ηp2 = .217 (see Figure 9). The NBI-NH group (M = 3.034,
density, F(1, 64) = 1.718, p = .194, ηp2 = .025. SE = 1.241) had significantly larger positive deviations from
Because the PWA-HL group retained only four par- target than the PWA-NH group (M = −5.657, SE = 1.687),
ticipants, analyses were performed with and without the F(1, 64) = 17.226, p = .0009, ηp2 = .212. The NBI-HL group
PWA-HL group to document whether the small sample (M = 1.097, SE = 1.17) also had significantly larger posi-
size in this group biased the relationship among the other tive deviation than the PWA-NH group, F(1, 64) = 10.824,
three groups. The total number of errors was compared p = .002, ηp2 = .145. No other group comparisons were sig-
across groups using a one-way between-subjects ANOVA. nificant for neighborhood density. Average deviation from
The average deviation from target for the errors also was
assessed with one-way ANOVAs applied separately for pho-
neme probability, biphoneme probability, and neighbor- Figure 9. Comparison of the error production by the four groups in
hood density as a function of group (including the PWA-HL terms of neighborhood density deviation from the target. PWA-NH =
group). The assumption of normality was met for each group persons with aphasia with normal hearing; PWA-HL = persons with
aphasia with hearing loss; NBI-NH = individuals with no brain injury
on all variables, and all other assumptions were met. with normal hearing; NBI-HL = individuals with no brain injury with
The number of error productions was significantly hearing loss.
different among groups, F(3, 64) = 9.742, p = .0009, ηp2 =
.313. Post hoc comparisons with Bonferroni correction
showed that the NBI-HL group (M = 8.833, SE = 1.407)
produced significantly more errors than the NBI-NH
group (M = 2.687, SE = 1.492), F(1, 64) = 8.979, p = .004,
ηp2 = .123, and the PWA-NH group (M = 1.962, SE = 2.028),
F(1, 64) = 7.752, p = .007, ηp2 = .108, but significantly fewer
errors than the PWA-HL group (M = 20.750, SE = 3.656),
F(1, 64) = 9.256, p = .003, ηp2 = .126. The PWA-HL group
had significantly more errors than the NBI-NH group,
F(1, 64) = 20.927, p = .0009, ηp2 = .246, and the PWA-NH
group, F(1, 64) = 20.201, p = .0009, ηp2 = .240. There was
no difference in the number of errors between the NBI-NH
and PWA-NH groups, F(1, 64) = 0.083, p = .774, ηp2 = .001.
These results are consistent with the previous analyses of
NU-6 performance using percent correct (in rau) as the
dependent measure.
The average deviation from target for neighborhood
density was significant for group, F(3, 64) = 5.921, p = .001,
same way hearing loss affected the NBI participants. It Although the current study did not focus on the impact
was evident that hearing threshold was the dominant effect of hearing loss on aphasia, the descriptive data (Table 2)
and largely independent of aphasia. This pattern of results is showed that the overall PICA and overall mean CRTT score
consistent with early work by Jauhiainen and Nuutila (1977) of PWA with hearing loss were lower than PWA without
and more recently by Pratt et al. (2007), which showed that hearing loss—suggesting a potential negative influence of
the speech perception of stroke survivors was more adversely the auditory impairment. Systematic investigation of the
affected by hearing loss and reduced audibility than the effects of peripheral hearing loss on aphasia symptoms
presence of aphasia. and severity should be included in future research pursuits.
NBI-NH .045 .002 .047 .004 .0024 .00031 .0026 .00039 13.34 1.28 14.18 1.79
NBI-HL .046 .003 .046 .002 .0025 .00033 .0025 .00037 13.62 1.53 13.93 1.67
PWA-NH .045 .002 .047 .003 .0023 .00034 .0027 .00045 13.03 1.54 13.97 1.60
PWA-HL .046 .002 .046 .002 .0025 .00031 .0027 .00061 12.78 1.42 12.98 0.68
Note. NBI-NH = individuals with no brain injury with normal hearing; NBI-HL = individuals with no brain injury with hearing loss; PWA-NH =
persons with aphasia with normal hearing; PWA-HL = persons with aphasia with hearing loss.
Phonotactic Probability and Neighborhood Density the groups with normal hearing than in groups with hearing
impairment for both phonotactic probability and neighbor-
Essentially no significant differences in phonotactic
hood density. However, the high levels of variability precluded
probability or neighborhood density were found in the cur-
definitive conclusions about how hearing loss, phonotactic
rent study. What was most notable about the results of
probability, and neighborhood density interact, especially
phonotactic probability and neighborhood density was the
in PWA. Nonetheless, the lack of significant differences for
large variability within groups. However, there were some
the error patterns among groups provides some support for
interesting patterns worthy of discussion. Although not signif-
the application of the NU-6 word lists with adults regard-
icant, the pattern of the results was consistent with the
less of aphasia.
assumption that the phonotactic probability and neighbor-
hood density properties of the errors were similar between
the two groups with normal hearing and between the two PIT
groups with hearing loss but diverged between groups with Because a majority of the participants with aphasia
normal hearing and with hearing loss. According to the were able to perform the NU-6, examination of the PIT
neighborhood activation model proposed by Luce and results relative to the NU-6 and the tests of aphasia was a
Pisoni (1998), spoken words are recognized in the context limitation. The PIT was administered to participants with
of phonologically similar words activated in memory, which verbal PICA scores that were generally low, indicating
then compete for recognition. The number of neighbors and that speech/language production was a notable indicator
the frequency of those neighbors influence lexical retrieval of whether a participant was able to perform the NU-6 or
in speech production as well. During the retrieval process, needed the PIT to assess word recognition ability. Another
phonologically similar words may block or compete with limitation was the small number of participants (n = 8) who
each other, resulting in slower and less accurate processing, had aphasia and had been diagnosed and treated for their
as they do in models of spoken word recognition (Vitevitch, hearing loss prior to their stroke. During our recruitment
2002). Hearing loss has a direct impact on the accuracy of efforts, it became evident that many clinicians were reluctant
word recognition because of missing or distorted informa- to manage the hearing loss of PWA and that for some
tion that likely interacts with phonotactic probability during participants and their families hearing loss was less of
speech perception tasks. Most errors occurred on high a priority than their aphasia. Another consideration was
phonotactic probability words within the NU-6 lists, which that a substantive number of the referrals with “normal
is consistent with the neighborhood activation model pre- hearing” had undiagnosed hearing loss.
dictions, but in some instances hearing loss might have Aphasia is associated with attention and working
altered the composition of the word and its relationship memory deficits that interfere with the ability to use lan-
to potential competitors. guage, with some arguing that these deficits are the actual
Theoretically, the average biphoneme probability source of aphasia (Hula & McNeil, 2008; McNeil et al.,
should reflect better a word’s average phoneme probability. 2004). As such, the added cognitive load and listening effort
Because a biphoneme is inherently a larger phonological observed with hearing loss and reduced audibility (McCoy
unit than a single phoneme, it likely provides a more accu- et al., 2005; Ng, Rudner, Lunner, Pedersen, & Rönnberg,
rate estimate of the probability of a meaningful word. In 2013; Tun, McCoy, & Wingfield, 2009) likely has an adverse
the current study, a slightly different pattern of average impact on language and cognitive performance in many
biphoneme probability from neighborhood density was PWA, which has consequences for accurate diagnosis,
observed, probably due to differences in the calculations prognosis, appropriate treatment, and quality of life. Thus,
and the different corpuses used in the calculations. The error it is important that PWA have access to prompt and appro-
words’ average biphoneme probabilities obtained from the priate audiological interventions. The impression of this
two PWA groups were marginally higher than the target. In study that PWA have inadequate hearing healthcare war-
addition, the difference between ears was slightly larger in rants further investigation.