Key Words dichotic listening - digits - learning disability - minimal detectable change - test–retest
reliability
INTRODUCTION
Dichotic listening tests are among the most common central behavioral tests used to
assess the functioning of the cerebral hemispheres, the interhemispheric transmission
of information, the central auditory nervous system maturation, and central auditory
processing disorders in adults and children ([Keith and Anderson, 2007 ]). Various materials are applicable in dichotic speech tests, including nonsense
consonant-vowel (CV) syllables, digits, monosyllabic words, and sentences ([Noffsinger et al, 1994 ]; [Jerger and Martin, 2006 ]; [Obrzut and Mahoney, 2011 ]).
Digits represent suitable speech materials for assessment of dichotic listening in
children with a wide range of linguistic ability. However, older children who are
tested with a double dichotic digits test may show a ceiling effect ([Strouse and Wilson, 1999a ]; [Strouse et al, 2000 ]; [Moncrieff and Musiek, 2002 ]; [Neijenhuis et al, 2003 ]). This led to the development of the American randomized dichotic digits test ([Department of Veterans Affairs, 1998 ]). The randomized dichotic digits test (RDDT) induces an uncertainty so that the
listener does not know in advance the number of digit-pairs in the incoming dichotic
item. This increases the difficulty of the test and reduces the performance of the
both ears, more pronouncedly the performance of the left ear ([Strouse and Wilson, 1999a ]). According to [Strouse and Wilson (1999b) ], the two-pair component of the RDDT is challenging enough to be a sensitive indicator
of binaural integration ([Strouse and Wilson, 1999b ]; [Moncrieff et al, 2016 ]). Currently, RDDT normative data are available for adults ([Strouse and Wilson, 1999b ]) and children ([Moncrieff and Wilson, 2009 ]; [Moncrieff, 2011 ]).
The RDDT has recently been adapted for use with the Persian language by [Mahdavi et al (2015) ] in two lists ([Mahdavi et al, 2015 ]). Both the American and Persian RDDTs (PRDDTs) use nine monosyllabic digits 1–10
(except for disyllabic digit 7 in the American version and digit 4 in the PRDDT).
In the Persian version of this test, similar to the original American test ([Department of Veterans Affairs, 1998 ]), there are 500-msec intervals between the digits and 4- to 8-sec intervals between
the items so that 8 sec is given to the test participants to repeat the digits after
the three-pair items, 6 sec after the two-pair items, and 4 sec after the one-pair
items. Supplemental Audio S1 is list 1 of the PRDDT that is available with the online
version of this article. Similar to its American counterpart, each list in the PRDDT
contains 54 items that are equally divided into the one-, two-, and three-pair items.
The RDDT has a unique structure with one-, two- and three-pair digits that are randomly
distributed in the list. Each ear gets 108 scores when receiving the full raw score
in each list so that 50% of this total score is obtained from the three-pair digits,
33% from the two-pair digits, and 17% from the one-pair digits. Therefore, the weight
of the three-pair digits in the total score of the RDDT is higher than that of the
two- and one-pair digits. Supplemental Appendix S1 is the score sheet of the PRDDT
that is available with the online version of this article.
A majority of normal right-handed participants show a right-ear advantage (REA) in
dichotic listening; this means that the number of correct answers for the right ear
as the dominant ear for language is higher than those for the nondominant left ear
([Moncrieff, 2011 ]). According to the anatomical model of Kimura, observation of REA in right-handed
individuals has its origin in the contralateral auditory pathway, which is a strong
neural route for conduction of speech signals to the language brain area located within
the left hemisphere. Auditory information received by the right ear will follow a
shorter path to the language-dominant hemisphere. However, auditory information received
by the left ear has to cross the corpus callosum, which may impose a lag relative
to the information of the right ear. Therefore, the speech signal to the right ear
is processed stronger and faster within the left hemisphere and suppresses the ipsilateral
route of the speech signal to the left ear ascending to the left hemisphere ([Moncrieff and Black, 2008 ]; [Musiek and Weihing, 2011 ]). It seems that working memory plays a role in the size of REA. Experiments with
dichotic digits have shown that when the number of digit pairs increases from one
pair to three pairs, REA increases due to inefficiency of the left ear in dichotic
listening ([Wilson and Jaffe, 1996 ]; [Strouse and Wilson, 1999a ]; [Moncrieff and Wilson, 2009 ]). [Penner et al (2009) ] highlighted the initial work of [Kimura (1961 ]; [1967 ]) and showed that the magnitude of REA increases as the number of letters in a pair
increases from 3 to 5 ([Kimura, 1967 ]; [1961 ]; [Penner et al, 2009 ]). REA of the PRDDT in a group of 18- to 25-yr-old adults did not differ from REA
of the three-pair dichotic digits ([Aghazadeh et al, 2015 ]; [Mahdavi et al, 2015 ]). However, calculating interaural asymmetry in dichotic listening using the traditional
method of REA has a disadvantage; individuals indicating left-ear advantage (LEA)
yield negative values for REA that lessen the average interaural asymmetry of a population.
[Moncrieff (2011) ] recommends that ear advantage be calculated by subtracting the score of the nondominant
ear from the dominant ear to compensate for the bias produced by the participants
with LEA.
There are studies aimed at test–retest reliability of dichotic listening in children
([Mukari et al, 2006 ]), adults ([Hugdahl and Hammar, 1997 ]; [Strouse and Wilson,1999b ]), and patients ([Koomar and Cermak, 1981 ]; [Strouse and Hall, 1995 ]). As part of a study by [Mukari et al (2006) ], a 4-week test–retest reliability of Malay one- and two-paired dichotic digits was
evaluated in normal children aged 6–11 yr in the free recall condition. The right-
and left-ear scores in the retest session differed significantly from those in the
test session and the average improvement of the scores in the retest session was 2.28
and 5.16% for the one-paired dichotic digits and 3.06 and 6.58% for the two-paired
dichotic digits. In that study, ear advantage of the one-paired digits was decreased
significantly in the retest session ([Mukari et al, 2006 ]). [Strouse and Hall (1995) ] measured test–retest reliability of two-paired dichotic digits developed by [Musiek (1983) ] in 10 adults with Alzheimer’s disease. The control group consisted of ten normal
adults. Pearson’s correlation coefficients between the test and retest session in
the Alzheimer’s group were 0.82 and 0.97 for the right- and left-ear scores, respectively.
The results of the Alzheimer’s group were comparable to the control group (r = 0.79 and r = 0.85 for the right- and the left-ear scores, respectively), although variability
of scores of the Alzheimer’s group in the first session was higher than the control
group. In that study, changes of the scores in retest session did not exceed 10% in
any participant except for three cases of the Alzheimer’s group with 11–13% score
changes ([Strouse and Hall, 1995 ]). [Strouse and Wilson (1999b) ] reported intertrial reliability of the American randomized dichotic digits test
in adults aged 20–79 yr. The left-ear score of the 70- to 79-yr-old age group improved
from 64.5% in the first trial to 70.2% in the second trial for three-pair digits ([Strouse and Wilson, 1999b ]). [Koomar and Cermak (1981) ] investigated a 1-week test–retest reliability of CV and three-pair digits materials
in normal and learning-disabled (LD) children between the ages of 7 and 10 yr. Pearson’s
correlation and consistency of REA between test and retest sessions were considered
as indices of test–retest reliability. The results showed that both groups showed
an REA and that there were no significant differences in the size of ear advantage
between the two groups of children on both dichotic CVs and digits. The normal children
were more reliable than those in the LD group and both groups of the children showed
higher test–retest reliability in dichotic CVs than for the dichotic digits ([Koomar and Cermak, 1981 ]).
Relative reliability is defined as the degree to which individuals maintain their
position in a sample with repeated measurements and usually is assessed by a form
of correlation coefficient such as Pearson’s correlation and intraclass correlation
(ICC) coefficients. Absolute reliability is defined as the degree to which repeated
measurements vary for studied individuals ([Baumgartner, 1989 ]). In contrast to relative reliability that evaluates association of two variables,
absolute reliability is concerned with their proximity ([Bruton et al, 2000 ]). Indices of absolute reliability are expressed either in the actual units of measurement
or as a proportion of the measured values (dimensionless ratio). Absolute reliability
is estimated by indices such as standard error of measurement (SEM), method error,
coefficient of variation, Bland–Altman limits of agreement, and minimal detectable
change (MDC) ([Atkinson and Nevill, 1998 ]). [Beckerman et al (2001) ] introduced the ‘‘smallest real difference’’ or MDC and defined it as the limit for
the smallest change between two measurements, which indicates a real (clinical) change
for a single individual following intervention ([Beckerman et al, 2001 ]).
Emerging documents suggest that dichotic listening training interventions such as
Dichotic Interaural Intensity Difference and Auditory Rehabilitation for Interaural
Asymmetry can ameliorate dichotic listening deficits ([Musiek et al, 2004 ]; [2008 ]; [Moncrieff and Wertz, 2008 ]). We need indices for detecting the real change in the ear scores after a period
of dichotic listening training. Thus, typical trial-to-trial variation of the ear
scores is clinically important when measuring the outcome of a dichotic listening
intervention.
Two previous studies showed that the PRDDT has enough difficulty to detect REA in
young adults and has adequate test–retest relative reliability ([Aghazadeh et al, 2015 ]; [Mahdavi et al, 2015 ]). However, currently, there is no information about the test–retest reliability
of the PRDDT in children. The studies on test–retest reliability of dichotic listening
have been performed on normal children and adults using indices of the relative reliability
([Bakker et al, 1978 ]; [Strouse and Hall, 1995 ]; [Hugdahl and Hammar, 1997 ]; [Mukari et al, 2006 ]). The abovementioned studies have used Pearson’s correlation coefficients as a relative
indicator of test–retest reliability and average difference between the test and retest
sessions to control systematic changes due to factors such as the learning effect.
Learning disability is associated with dichotic listening deficit ([Weihing and Musiek, 2013 ]). Currently, there is no study on the absolute reliability of dichotic listening
results in LD children as a target population for dichotic listening training. Therefore,
this study was conducted with the purpose of comparing the results of the PRDDT and
the relative and absolute indices of test–retest reliability between TA and LD children.
MATERIALS AND METHODS
This study was performed using a test–retest design in which the participants were
tested with the PRDDT within a 7- to 12-day period.
Participants
Fifteen LD children (eight males and seven females) with an age range of 7–12 yr were
recruited from a study performed by[ Esmaili et al (2016) ] on specific learning disability in Iran. The study psychologist (A. P.) confirmed
the diagnosis of learning disability in the children based on the Wechsler intelligence
scale for children, teacher-created tests based on school textbooks ([Esmaili et al, 2016 ]), and the definition of learning disability provided in Diagnostic and Statistical
Manual of Mental Disorders-IV ([American Psychiatric Association, 2000 ]). Eleven children of the LD group were right handed and four of them were left handed.
The control group consisted of 15 right-handed typical achieving (TA) children with
an age range of 8–11 yr selected from three elementary schools whose first semester
reports showed typical educational achievement. Hearing thresholds of both groups
of the children were within normal limits (≤15 dB HL in frequencies 500–4000 Hz) and
interaural hearing threshold asymmetry was ≤10 dB. The parents of the children signed
a written consent form and the children were paid for participating in our study.
Procedures
A calibrated laptop (Dell Inspiron 6400, Dublin, Ireland) attached to headphones (Philips
SHM 6500/10) was used to administer the PRDDT. Before starting the test, we explained
the objective of the test and the participants were instructed how to perform the
test. Further, the participants got familiar with the test procedure through seven
practice items including one-, two-, or three-pair digits before starting the test
session. List 1 of the PRDDT was performed on the children at 70 dB HL under free
recall conditions in a very quiet room. The retest session was held within 7–12 days
with the average of 9.3 days after the test session.
Statistics
REA was calculated through the traditional method in which the left-ear score was
subtracted from the right-ear score with two directions, positive and negative. Dominant
ear advantage (DEA) was calculated by subtracting the nondominant ear score from the
dominant ear score. The normality assumption of the data was checked by the Kolmogorov–Smirnov
statistical test using the statistical software SPSS 21.0 (IBM SPSS Inc., Chicago,
IL). The average of the ear scores was compared between the right and left ears using
a paired t test. An independent t test was used for comparing the average of the REA and DEA ear scores between TA
and LD children. Fisher’s exact test was used for comparing the direction consistency
of REA between TA and LD children. A 0.05 significance level was considered for all
statistical tests.
Relative test–retest reliability of the ear scores and REA were estimated using Pearson’s
correlation coefficient and intraclass correlation coefficient model (ICC2,1 ) using a single measure two-way mixed effects model with both consistency and absolute
agreement definitions. A 95% confidence interval (CI) was constructed around ICC2,1 point estimation for the ear scores and REA. Interpretation of ICCs was based on
the known Fleiss’ classification of ICC as follows: <0.4: poor; 0.4–0.75: fair to
good; and >0.75: excellent ([Fleiss, 1986 ]).
SEM and SEM%—as absolute test–retest reliability indices—were calculated using the
raw ear scores. The square root of within-subjects mean squares was used for calculating
SEM Downham et al, 2005 ]).
The MDC was obtained by multiplying SEM by √2 and by 1.96, (MDC = SEM × √2 × 1.96).
MDC% was computed by dividing MDC by the mean of the test and retest scores and multiplying
the result by 100 Downham et al, 2005 ]).
RESULTS
There was no significant difference in the age means; the TA and LD children had age
means (±standard deviations [SDs]) of 9.2 ± 1.5 and 8.3 ± 1.3 yr old, respectively.
The average hearing thresholds of right and left ears in the frequencies of 500–4000
Hz did not show significantly difference between TA and LD children and between the
ears of each group of the children (p ranging from 0.069 to 0.615 for between-group and 0.063 to 0.564 for within-group
interaural hearing threshold asymmetry). [Table 1 ] indicates demographics and the individual ear scores in percent correct and REA
in percent and we report the results in the text as mean (in %) ± SD.
Table 1
Characteristics and Individual Data of Children with Means and Standard Deviations
of Ear Scores and REA in the Test and Retest Sessions (in %)
TA Children
LD Children
Participant
Sex/Handedness/Age (yr)
Test
Retest
Participant
Sex/Handedness/Age (yr)
Test
Retest
RE
LE
REA
RE
LE
REA
RE
LE
REA
RE
LE
REA
1
M/R/8
98.15
89.81
8.34
98.15
92.59
5.56
1
F/R/7
75.00
32.41
42.59
72.22
48.15
24.07
2
M/R/12
94.44
67.59
26.85
96.30
70.37
25.93
2
M/R/8
82.41
62.04
20.37
82.41
70.37
12.04
3
M/R/7
84.26
75.00
9.26
90.74
77.78
12.96
3
F/R/7
52.78
54.63
−1.85
53.70
61.11
−7.41
4
M/R/11
87.96
62.96
25.00
92.59
73.15
19.44
4
F/R/7
71.30
28.70
42.60
66.67
47.22
19.45
5
M/R/8
92.59
81.48
11.11
95.37
97.22
−1.85
5
F/L/8
61.11
94.44
−33.33
62.04
93.52
−31.48
6
M/R/8
91.67
58.33
33.34
94.44
65.74
28.70
6
M/R/9
80.56
62.96
17.60
86.11
43.52
42.59
7
M/R/9
91.67
80.56
11.11
93.52
87.96
5.56
7
F/R/7
64.81
29.63
35.18
70.37
42.59
27.78
8
M/R/8
90.74
62.96
27.78
92.59
65.74
26.85
8
M/R/8
71.30
41.67
29.63
75.00
54.63
20.37
9
M/R/11
97.22
72.22
25.00
96.30
75.00
21.30
9
M/R/9
78.70
61.11
17.59
73.15
70.37
2.78
10
F/R/8
91.67
79.63
12.04
96.30
82.41
13.89
10
F/R/11
69.44
40.74
28.70
71.30
46.30
25.00
11
F/R/9
92.59
83.33
9.26
94.44
98.15
−3.71
11
F/L/9
80.56
62.96
17.60
82.41
67.59
14.82
12
F/R/8
90.74
83.33
7.41
90.74
86.11
4.63
12
M/R/8
40.74
23.15
17.59
53.70
45.37
8.33
13
F/R/11
94.44
84.26
10.18
96.30
74.07
22.23
13
M/L/7
59.26
53.70
5.56
71.30
59.26
12.04
14
F/R/10
95.37
85.19
10.18
97.22
87.96
9.26
14
M/L/9
84.26
55.56
28.70
87.04
53.70
33.34
15
F/R/10
93.52
83.33
10.19
95.37
90.74
4.63
15
M/R/11
77.78
62.04
15.74
91.67
70.37
21.30
Mean (SD)
92.47 (3.5)
76.67 (9.6)
15.80 (8.9)
94.70 (2.3)
81.67 (10.8)
13.02 (10.6)
Mean (SD)
70.00 (12.3)
51.05 (18.6)
18.95 (19.0)
73.27 (11.4)
58.27 (14.2)
15.00 (17.7)
[Figure 1 ] is a scatter plot with a 45° line that illustrates agreement between the test and
retest results of the ear scores and REAs for the studied children in percent. The
more closely the scatter data points cluster around the 45° line, the greater the
agreement is between the test and retest results. A 45° line illustrates perfect agreement.
Figure 1 A bivariate plot showing test results (in %) on horizontal and retest results (in
%) on vertical axis. The 45° diagonal line represents perfect agreement between the
test and retest results.
The mean RE score was significantly better than that of the LE score in both groups
of children [TA children test session: t
(14) = 6.9, p = 0.000, retest session: t
(14) = 4.8, p = 0.000; LD children test session: t
(14) = 3.9, p < 0.005, retest session: t
(14) = 3.27, p < 0.01]. Therefore, the PRDDT revealed a clear average REA in both groups of children.
[Table 1 ] contains individual data, mean, and SD of the ear scores and REA recorded in the
test and retest sessions for the groups. Statistical analysis detected a significantly
poorer performance in LD children versus TA children in the both test and retest sessions
for RE scores [the test session: t
(28) = 6.8, p < 0.001, d = 2.48; the retest session: t
(28) = 7.15, p = 0. 0.000, d = 2.6] and LE scores [the test session: t
(28) = 4.74, p < 0.001, d = 1.73; the retest session: t
(28) = 5.08, p < 0.001, d = 1.85].
In TA children, average REA and DEA (in %) for the test session was the same (15.8
± 8.9) because none of the TA children produced left-ear dominancy and the average
REA in the retest session (13.02 ± 10.6) did not differ significantly from the average
DEA of the retest session (13.77 ± 9.5) [t
(14) = −13.82, p = 0.189]. In LD children, the average REA of the test session (18.95 ± 19.0) and
retest session (15.00 ± 17.7) was lower than the corresponding average DEA of the
test session (23.64 ± 12.11) and retest session (20.19 ± 10.9), although the observed
differences were not statistically significant [test session: t
(14) = −1.05, p = 0.308; retest session: t
(14) = –2.22, p = 0.242] so we presented only REA in [Table 1 ].
Between-group comparison of the score changes (in %) showed that the mean of the retest
change of RE score (2.22 ± 1.9) and LE score (5.00 ± 6.1) in the TA children did not
differ significantly from the mean retest change of RE scores (3.27 ± 5.9) and LE
scores (7.22 ± 9.9) in the LD children, [t
(28) = 0.650, p = 0.521 for RE scores; t
(28) = 0.739, p = 0.466 for LE scores], respectively. The mean REA of LD children in the test and
retest sessions did not differ significantly from the mean of the TA children in the
test and retest sessions, respectively [t
(28) = −0.581, p = 0.566 for test REA; t
(28) = 0.370, p = 0.714 for retest REA].
Comparison of average DEA in LD children in test and retest sessions (23.64 ± 12.1
and 20.18 ± 10.9, respectively) did not show statistically significant difference
from the corresponding average DEA in TA children (15.80 ± 8.9 and 13.77 ± 9.5, respectively)
[t
(28) = −2.02, p = 0.053 for the test DEA; t
(28) = −1.71, p = 0.097 for the retest DEA]. Although average DEA in LD children was not statistically
larger than average DEA in TA children, given the small number of participants and
the medium effect size of this difference in the test (Cohen’s d = 0.74) and retest sessions (Cohen’s d = 0.63), the observed difference is clinically important. Since LEA is more prevalent
in LD children ([Keith, 2007 ]), it is logical that we expect DEA to produce a larger interaural asymmetry than
REA in this group of children.
As presented in [Table 1 ], the average of the both ear scores in TA children enhanced significantly in the
retest session (p < 0.001 for RE, p < 0.01 for LE scores). In LD children, the RE score did not show a significant improvement,
however, their average LE score significantly increased in the retest session (p < 0.05).
The average of the test REA of both groups of the children remained unchanged statistically
in the retest session ([Table 1 ]). Thirteen of the TA children (86.7%) maintained the REA direction in the retest
session and REA changed to LEA in two (13.3%) of the TA children. All of the LD children
(100%) consistently indicated the same REA direction on the retest session. This difference
was not statistically significant (p value calculated by one-sided Fisher’s exact test = 0.241).
[Table 2 ] contains ICC coefficients of consistency and agreement and Pearson’s correlation
coefficient between test and retest sessions for right and left scores and REA. The
ICC coefficient of consistency of ear scores and REA was categorized as excellent
(ICC = 0.78–0.87) in both groups of children. The ICC coefficient of agreement of
the ear scores and REAs was categorized as fair to excellent in TA children and excellent
in LD children. Similarly, as Pearson’s correlation coefficient demonstrates, the
test results were very strongly (r > 0.80) and positively correlated to retest results in both groups of the children.
RE of the TA children obtained the lowest ICC coefficient of consistency and agreement
([Table 2 ]).
Table 2
ICC Coefficients of Consistency and Agreement with 95% CI and Pearson’s r with Calculated SEM, SEM%, MDC, MDC%, and Reference Band of 95% MDC for Ear Scores
and REAs between Test and Retest Sessions in TA and LD Children
ICC Coefficient (95% CI)
Consistency***
Agreement***
Pearson’s r **
SEM
SEM%
MDC
MDC%
Reference Band (95% MDC)
TA children
Right ear
0.78 (0.47 to 0.92)
0.62 (−0.05 to 0.88)
0.86
1.46
1.44
4.03
3.99
−1.60 to 6.43
Left ear
0.82 (0.57 to 0.94)
0.74 (0.22 to 0.92)
0.83
4.68
5.47
12.93
15.13
−7.53 to 17.33
REA
0.80 (0.51 to 0.93)
0.78 (0.46 to 0.92)
0.81
4.34
27.87
11.99
77.03
−14.99 to 8.99
LD children
Right ear
0.87 (0.67 to 0.96)
0.85 (0.57 to 0.95)
0.88
4.55
5.88
12.57
16.25
−9.04 to 16.10
Left ear
0.82 (0.55 to 0.94)
0.76 (0.30 to 0.92)
0.85
7.56
12.81
20.89
35.39
−13.09 to 28.69
REA
0.80 (0.49 to 0.93)
0.79 (0.49 to 0.92)
0.81
8.31
45.34
22.97
125.29
−27.24 to 18.70
Notes : ***p < 0.001 for the ear scores and REA; **p < 0.01 for the ear scores and REA.
[Table 2 ] presents SEM and SEM% for the ear scores and REAs. In both groups of children, SEM
of the RE was lower than SEM of the left ear. The SEM value of REA in the LD children
was higher than the corresponding value in the TA children. Calculated SEM% in LD
children is 1.5–4 times higher than SEM% in TA children. Although ICCs and Pearson’s
correlation coefficient show a similar relative reliability of the ear scores and
REAs for both groups of children, absolute reliability of the ear scores and REAs
is considerably poorer in LD children.
Also presented in [Table 2 ] are MDC and MDC%. In both groups of children, MDC of the right ear was lower than
MDC of the left ear. The MDC value of REA in the LD children was higher than the corresponding
value in the TA children.
DISCUSSION
The main purpose of the current study was evaluating and comparing the test–retest
reliability of the PRDDT between TA and LD children using both indices of relative
and absolute reliability and determining MDC for it. The results of this study showed
that LD children performed more poorly in the PRDDT for both right and left ears than
TA children. Both groups of children produced higher RE than LE scores but there was
no significant difference between the average of REA in the LD and TA children. This
is due to poorer performance of both ears of the LD children for the PRDDT. The results
demonstrated that the PRDDT could detect a group difference between the TA and LD
children only based on the ear scores.
Four LD children have LEA and we expected average DEA to be larger in size than average
REA in these children but two methods (REA versus DEA) of calculating interaural asymmetry
of dichotic listening did not show statistically significant average difference in
LD children or TA children. However, [Moncrieff (2011) ] demonstrated that DEA detected a larger average interaural asymmetry than REA in
normal children. The discrepancy might be due to a difference in scoring technique,
which is based on two-pair components of the RDDT in the American format and overall
scores of all items in the Persian RDDT. More research with similar inclusion criteria,
sample size, and scoring is needed for comparing the American and Persian RDDT.
We did not find any study using total-score RDDT with learning disability for comparison.
Most of the studies on learning disability with digit materials have used double dichotic
digits ([Moncrieff and Musiek, 2002 ]; [Moncrieff and Black, 2008 ]; [Pinheiro et al, 2010 ]; [Ghannoum et al, 2014 ]) that are easier than RDDT. However, our results are consistent with many studies
summarized by [Moncrieff et al (2016) ] including out-of-date studies ([Thomson, 1976 ]; [Keefe and Swinney, 1979 ]; [Pelham, 1979 ]; [Aylward, 1984 ]) on children with a reading disability—a specific learning disability—and a recent
study ([Pinheiro et al, 2010 ]) on LD children.
Reliability Indices
According to [Downham et al (2005) ], the reliability cannot be analyzed based only on relative indicators and an analysis
of measurement errors must be performed as a complement ([Downham et al, 2005 ]). Since each of the reliability indices has their own advantages and disadvantages,
researchers emphasize considering both absolute and relative reliability coefficients
([Lexell and Downham, 2005 ]).
Mean Difference
The mean difference between test and retest sessions demonstrated that the ear scores
were susceptible to a learning effect. In the TA children the scores of both ears
and in the LD children the scores of the left ear improved in the retest session.
LD children usually show a LE deficit in dichotic listening but some also demonstrate
a RE deficit. Both deficits are targeted by auditory rehabilitation for interaural
asymmetry ([Moncrieff and Wertz, 2008 ]). Because dichotic listening training programs seek to diminish this weakness, information
about the learning effect of RDDT is clinically important. The LD children without
any intervention increased their average LE scores significantly, by 7.22%. A similar
improvement (7%) was reported by [Weihing and Musiek (2013) ] for the weaker ear of a control group of children when tested pre- and post-dichotic
interaural intensity difference training by double dichotic digits test ([Weihing and Musiek, 2013 ]). This index of reliability does not provide information regarding individual differences
and it is better to complement it with other reliability indices ([Weir, 2005 ]; [Zaki et al, 2013 ]).
Pearson’s Correlation Coefficient
In both groups of children, Pearson’s r showed a very strong positive relationship between the results of test and retest
sessions. Pearson’s r for the ear scores obtained in the current study was in line with the study by [Strouse and Hall (1995) ] and consistent with the upper range of correlations (0.6–0.8) reported in the previous
studies using CV materials ([Hugdahl and Hammar, 1997 ]; [Gadea et al, 2000 ]). Pearson’s r shows how test and retest results vary together without any information about the
agreement of the results ([Bruton et al, 2000 ]). A strong correlation between the two sets of data does not guarantee that the
two sets are highly repeatable, as Pearson’s r does not see systematic or fixed errors. Thus, this index should be used in conjunction
with other indicators for assessment of reliability ([Weir, 2005 ]; [Zaki et al, 2013 ]).
ICC: Consistency and Agreement
In the current study, both consistency and agreement of dichotic listening performance
were calculated using the ICC method. The ICC model used in the current study is similar
to Pearson’s r . As shown in [Table 2 ], the ICCs with the consistency definition and Pearson’s r coefficients are equivalent. Indices of agreement between the two measurements such
as dichotic listening scores before and after dichotic listening training could be
used for specifying expected improvement after intervention ([de Vet et al, 2006 ]). The results demonstrated that absolute agreement between test and retest scores
was lower than the consistency of the scores. This discrepancy originates from the
fact that ICC agreement is defined as the extent to which identical measurements are
obtained in test and retest sessions and considers both systematic and random errors
while ICC consistency does not take into account systematic differences (learning
or practice effect) in the scores between the test and retest sessions ([Downham et al, 2005 ]; [de Vet et al, 2006 ]). As [Figure 1 ] displays, the RE scores of the TA children have the lowest dispersion and are very
close to the perfect agreement line. However, its ICC coefficient unexpectedly is
the lowest in comparison to other variables. This is due to the dependency of ICC
to heterogeneity of studied participants in performance measured. If between-subjects
variability is low but real trial-to-trial consistency of a measured performance is
high, it may result in a small ICC ([Weir, 2005 ]). This relatively misleading result of ICC has occurred for the RE scores of the
TA children because TA children are less heterogeneous for the RE score than the LE
score, REA, and the results of LD children ([Figure 1 ]).
SEM and SEM%
Absolute reliability indicators such as SEM provide information about the degree by
which the repeated measurements vary for individuals. Since SEM estimates random variation
of the performance across repeated measures, it is an important indicator for test
sensitivity in detecting a change of performance over time. If SEM of the performance
in a test–retest study is small, detection of intervention-related changes is easier
([Downham et al, 2005 ]; [Lexell and Downham, 2005 ]). In a repeated measures design, it is not known how much of the variability originates
from the changes in mean and how much from typical variation of measured performance.
This limitation of SEM can be overcome by calculating SEM%, which expresses measurement
variability as a coefficient of variation ([Lexell and Downham, 2005 ]).
The previous studies on reliability of dichotic listening have not addressed within-subject
variation of the ear scores or REAs. SEM indicates the precision of the ear scores
and a CI can be constructed based on its value to identify true ear scores ([Downham et al, 2005 ]). SEM% specifies how much of the observed changes in the ear scores and REAs can
be attributed to typical variation. This type of reliability may be clinically useful
when changes of the ear scores are detected after a period of dichotic listening training.
As presented in [Table 1 ], the average REAs in both groups of the TA and LD children did not show statistically
significant differences between test and retest sessions. This may be interpreted
as high reliability of REAs. However, comparing the averages between the two sets
of measurements did not provide any information on individual differences ([Bruton et al, 2000 ]). On the other hand, SEM% of REAs is considerably greater than that of the ear scores
([Table 2 ]). Typical variation of REA of the PRDDT in our sample of the LD children is 9.3
(45.34% × 20.47 row scores). High within-subject variation of REA in repeated measures
may limit its adequacy as an indicator of posttraining improvement in LD children.
However, the REA direction in TA and LD children showed a high intersession consistency.
Previous studies have reported both consistency and inconsistency of REA in some of
their participants. The more relevant study is [Strouse and Wilson (1999b) ], which reported intertrial reversal of REA direction in five participants (25%)
under 30 yr old for the American RDDT. Similarly, relative frequency of reversal of
REA direction was 30% in a study on one-pair dichotic digits performed by [Pizzamiglio et al (1974) ] and 29% in a study on CV syllables performed by [Blumstein et al (1975) ]. [Koomar and Cermak (1981) ] found lower test–retest reliability of REA for three-pair dichotic digits in LD
versus normal children with an age range of 7–10 yr. It seems that consistency of
REA direction is also dependent on size of REA so that individuals with small REA
are more likely to shift REA direction in the retest sessions ([Blumstein et al, 1975 ]). In the current study, all LD children maintained REA direction in the retest session
while this consistency was observed in 86.7% of TA children with no significant difference
between TA and LD children.
MDC and MDC%
MDC can be used to assess minimal changes of the ear scores required to be 95% confident
that the induced changes after dichotic listening training are true changes and not
measurement errors ([Downham et al, 2005 ]; [Lexell and Downham, 2005 ]). MDC and MDC% were higher in the LD children versus the TA children, especially
for REA ([Table 2 ]).
To ensure that the dichotic listening deficit of a child measured by RDDT has clinically
been changed after dichotic training, we need a “reference band” for RDDT. This range
extracted from test and retest sessions was computed and presented in [Table 2 ] and is entitled 95% MDC. If changes of ear scores after dichotic training are within
this range, it cannot be considered a clinically important change. Since dichotic
training is expected to increase ear scores, the improvement (posttreatment score
minus pretreatment score) should exceed the higher fence of this reference range.
MDC%, which is independent of the units of measurement, is more easily interpreted.
Suppose that we administer a dichotic training protocol on this group of LD children
and expect to improve the left-ear deficit compared to pretreatment ear scores with
a raw score of 55, SEM of 7.56, and SEM% of 12.81%. This means that the LE scores
have a typical variation of 7.04 (12.81% × 55). We are 95% confident that the true
raw score is within (55 ± [2 × 7.56]) or 39.8–70.1. If the posttraining score reaches
68, the audiologist cannot be sure that a true change has occurred. MDC%, presented
in [Table 2 ], implies that if we want to achieve a true change, the average of the LE raw score
of the LD children has to exceed 74.4 (35.39% × 55 + 55).
As denoted in [Table 2 ], in both groups of the children, MDC% of the ear scores is considerably lower than
that of REA. This is because of the fact that REA is a difference value and is affected
by typical variation of both the RE and LE scores. Therefore, considering indices
of measurement variability for dichotic listening scores may facilitate clinical decision-making
([Stratford, 2004 ]).
The studies performed on dichotic listening training have not reported MDC and compared
the outcome between experimental and control groups or used a before–after design
for demonstrating efficacy of dichotic listening training ([Katz et al, 1984 ]; [English et al, 2003 ]; [Moncrieff and Wertz, 2008 ]). This study introduces MDC, which may be applicable for identifying real alternations
of dichotic listening scores after dichotic listening interventions. However, according
to [Beckerman et al (2001) ] MDC is a clinometric indicator of the measuring tool and reflects the magnitude
of change that confidently did not result from typical variation of the measured performance,
whereas “clinically relevant change” is a change arbitrarily considered as important
change by clinicians ([Beckerman et al, 2001 ]).
CONCLUSION
LD children showed test–retest relative reliability as high as TA children in the
ear scores and in size and direction of REA measured by the PRDDT. However, the indices
of absolute reliability revealed that the ear scores and REA were less reliable within
the LD children versus the TA children. Establishment of a reference band of minimal
detectable change may be useful for clinical tracking of training-related improvements
in the ear scores.
Abbreviations
CI:
confidence interval
CV:
consonant-vowel
DEA:
dominant ear advantage
ICC:
intraclass correlation
LD:
learning disabled
LE:
left ear
LEA:
left-ear advantage
MDC:
minimal detectable change
PRDDT:
Persian randomized dichotic digits test
RDDT:
randomized dichotic digits test
RE:
right ear
REA:
right-ear advantage
SD:
standard deviation
SEM:
standard error of measurement
TA:
typical achieving
