002X 3932’92 S5.00+0 00 0 19Y2 Prrpmnn Press I.td
DICHOTIC
LISTENING:
WHAT
DOES IT MEASURE?
LUTZ J~NCKE,* HELMUTH STEINMETZ~ and JENS V~LKMANN? *Department of Psychobiology and Psychocybernetics, Heinrich-Heine-University, Universitatsstral3e D-4000 Dusseldorf 1, F.R.G. and tDepartment of Neurology, Heinrich-Heine-University. Moorenstrage D-4000 Diisseldorf 1, F.R.G. (Receiced
2 Septemhw
1991; accepted
1, 5.
2 1 July 1992)
Abstract-Auditory lateralization was investigated in 26 right-handed and 26 left-handed, normal subjects using seven different dichotic listening tests in each proband (free recall of digit lists, free recall ofconsonant vowel (0’) svllables, four different CV syllable monitoring paradigms, and free recall of Morse codes). Reliabilttjes calculated with the formula of Spearman Brown were low for digit recall (0.29, corrected for test length: 0.50) but good for CV recall (0.83), CV monitoring (0.75~ 0.88). and Morse code recall (0.50, corrected for test length: 0.88). Nevertheless, intertest correlations were low, both for right- and left-handers (negative correlations ranging from -0.44 to -0.05, positive correlations ranging from 0.01 to 0.51). Only 38 77% of the right-handed and lefthanded subjects retained one direction of ear advantage across any combination of two tests. The data suggest that different dichotic tests reveal different results. This may be due to psychometrtc. procedural, or phonetic properties. We conclude that individual predictions of language dominance are not justified using the dichotic tests evaluated in the present study.
INTRODUCTION SINCE KIMURA’S original digit list paradigm [18, 191, several paradigms using other verbal stimuli have been proposed to increase the predictive validity of the dichotic method with respect to language lateralization (for overviews see Refs [2], [S], [7--91 and [36]). Among these tasks, the recall or monitoring of dichotic consonant-vowel (CV) syllables [30,35,38] has been consistently found to produce the most robust right-ear advantages (REAs) in right-handed individuals, an effect thought to be due to the superior temporal discrimination of acoustic signals by the dominant hemisphere [4, 71. In accordance with this assumption. several studies demonstrated REAs also for nonverbal acoustic stimuli with rapidly changing temporal cues [23, 391. Only a few investigations, however, directly assessed the predictive validity of dichotic paradigms with respect to language lateralization as determined by invasive methods [ll, 19, 37, 441. In contrast to the encouraging results of these studies, other authors were much more cautious on the validity issue, mainly due to poor dichotic retest or intertest comparabilities [3, 5,401. The present study addresses this problem. We report intertest correlations for seven dichotic paradigms, all known to produce strong REAs and thought to reflect language lateralization, but differing with respect to stimulus material (digits, CV syllables, or nonverbal stimuli), stimulus presentation (list or single-pair) or stimulus reproduction (free recall or target detection). The investigation thus attempts to further elucidate the question whether auditory lateralization is likely to be related to one
*Address for correspondence: Dr Heinrich-Heine-University Dusseldorf,
Lutz Jancke, Department of Psychobiology and Universitatsstrabe I. D-4000 Dusseldorf I, F.R.G. 941
Psychocybernetics.
942
L. J;(N~KE rl trl
specialized mechanism of the dominant hemisphere [4,38] indeed be expected to also underlie language dominance.
which, if this were the case, could
METHODS (a) Suhjrcrs We examined 52 healthy, paid volunteers without hearmg deficits in the speech frequency range (less than IO dB hearing loss on either ear tested by standard audiometry). Twenty-six of the 52 subjects were self-described sinistrals (mean age: 26.4 years; age range: 21 35 years) and 26 were self-described dcxtrals (mean age: 24.9 years age; age range: 2l--34 years). Most ofthem were university students. There were 13 males and I3 females in each handedness group.
Hand preference was assosed using the five-item questionnaire OIBKYIIFY [6,7]. Subjects were classified as either left-handed (Bryden scores ~0) or right-handed (Bryden scores >O) L6.71. This classilicatlon conformed with the subject’s self-description in all casts (XC a).
All subjects were examined with the following dichotic tests: (i) digit list recall. (ii) CV recall, (iii) CV monitoring with target syllables;ta; or/da: (hit rate and reaction time). and (iv) Morse code recall. Tests (i) and (iv) were applied in one session, tests (ii) anJ (iii) In another session. Half of the subjects started with session I, the other with session 2. Interbals between sessions ranged from I to 5 days. Within each session. half of the subjects started with test (i) or (iv) and the other with test (ii) or (iii). (i) Dig/if list vt~c~/l. The test material consisted of 96 pairs of German monosyllabic digits (/ems.. ;rwei:. idrei,‘, ,‘vier.. ‘fiinf,. . ,/sechs,,,, ,‘acht,,‘. ,‘neun,.‘. , rehn . :elf,‘. ,‘zwolf:) spoken by a male person. These stimuli were band-pass filtered (upper corner frequency 10 kHz, 24 dB,,octave and lower corner frequency 80 Hz. 24 dB,‘octave), digitized (AjD conversion. 16 bit). and stored in computer memory (PDPI I computer). Each dlgit wa, edited and synchromred for intensity and onset with a speech editor program. The criterion for temporal alignment of digits was onset of articulatory rcleasc. After randomizatmn for digit order and exclusion of pairs of identical digits. the stimuli were D A converted (20 kHr, resolution I6 hit). Digits forming a pair were rccordcd on separate tracks of a magnetic recording tape usmg different channels of the D:A conbertcr. With this technique. competing digits uerc aligned with an accuracy of 50 /Isec. Thirty-three trials. each trial consisting of three digit pairs, were recorded on tape. Within each trial, &git pairs were presented with I-XX intcr%als (interval from onset ofpreceding digit to onset of following digit). Each trial was followed hq a 6-set pause. Dichotic pairs were presented using a stereo tape recorder (AIWA AD-F360). stereo headphones (Scnnhelser HD222). and at an intensity of about 70 dB (SPL). Suhjccts were instructed to pay attenllon to both ears. They were asked to write down the percaved stimuli immediately after each trial. lrrcspective of order or ear of presentation. After 16 trials, headphone orientation was reversed. Half of the subjects started with headphone orientation A (right-car channel right. left-car channel left). The other half started with rckerscd headphone orientation B. In both groups. the numbers of right-banders, leftbanders. men and women were counterbalanced. Prior to each experiment. 10 practice trials were carried out. (ii) C‘I’wc011. This test comprised paira of natural speech CV syllables (:ba:, :ga,, da., :p:~,‘, /ka,‘, and IIta; ) spoken by a trained phonetician (microphone .AKG 16OEl) which had heen band-pass liltercd (Rockland 432, upper corner frequency IO kHz 24 dB octave and lower corner frequency X0 Hz 24 dB:octave). Each syllabic WBSdigitized and stored in computer memory (l2-hit A;D converter. digitizing rate 20.229 Hz. ATARI MEGA ST4). The onset, duration. intensity and fundamental frequency of the stimuli were then edited and synchronized by means of a speech editor. The criterion for temporal alignment of the competing syllables was the onset of articulatory release. Each syllable was preceded by IO0 msec ofzero intenTit> before articulation started and followed by 50 msec ofrero intensity after vowel end. This manipulation served to remove clicks at the beginning and end upon D:A convcraion. The duration of syllables ranged from 310 to 360 msec, depending on voice onset times, with a +owel duration of 300 msec. Voice onset times (in mscc) for the stops were approximately b- IO. g=20. d = 15. p=40, k =60 and t = 50. The duration of the entire signals (syllables. leading and end interval with Lero Intensity) ranged from 450 to 500 msec. These edited Ggnals were stored on hard disk (MEGAFILE 30) for further use. For dichotic presentation, the stored sign& were D,‘A converted (I 2.bit. four-channel D’A converter, digitizing rate 20.229 Hz) by alternately converting a value of the first syllable and then the second syllable. Each syllahlc was conberted using a diflerent channel of the D.A con\crtcr. With this technique. competmg syllables were aligned with an accuracy of49.4 /(sec. After con>eraion the output of eachD A channel &as amplified (70 dB SPL as measured by a Bruel & Kjaer sound level meter) and presented via stereo headphones (Sennheiscr HD450). During the cupcriments. the subjects were seated in a sound-attenuated chamber. received instructions through a two-way intercom set and were instructed to pay attention to both cars. The presentation of dichotic stimuli and the recall procedure were controlled by the
DlCHOTIC
LlSTENlNG
943
computer. For presentation, the CV pair was loaded from hard disk into computer memory and D/A converted. After stimulus application, a diagram showing six squares containing the six syllables appeared on a monitor in front of the subjects for 5 sec. The subjects were instructed to “click” the syllable or syllable pair they had heard. Prior to each experiment, 30 practice trials were carried out. Test stimuli were randomized and arranged in eight blocks of 30 CV pairs in which each ofthe six syllables was paired with the other five (no syllable was paired with itself). Thus, the whole experiment consisted of 240 test trials, each trial consisting of one pair of syllables followed by the 5-set response interval. Headphone orientation was reversed after each block. Halfof the subjects started with headphone orientation A (right-ear channel right, left-ear channel left), whereas the other half of the subjects started with reversed headphone orientation B. In both groups, the numbers of right-handers, left-handers, men and women were counterbalanced. Previous experiments in our laboratory had shown a Pearson test-retest correlation of 0.81 for the laterality index (“e”) (see below) (15 healthy male right-handers: test-intervals: 3 4 weeks). (iii) CVmonitoring. The same stimuli and technical equipment as in the CV recall test were used. The stimuli were randomized and arranged in eight blocks, each containing 30 syllable pairs. resulting in a total of 240 trials. Each of the 30 possible pair-wise permutations of the six CV syllables occurred equally often. Because all possible CV pairs were equiprobable, one-third of the trials consisted of pairs containing the target. Fifty per cent of the targets were delivered to the right and 50% to the left ear. Thus, the a priori probability of target occurrence for each ear was 0.167. The interval between syllable pairs was 1.9 sec. Subjects were instructed to pay attention to both ears and press a button with his or her forefinger when the target syllable was heard. Reactions within 100~1000 msec after stimulus onset were accepted. Subjects underwent this procedure twice, once with the target :ta/ and once with the target ,‘da/. Headphone orientation and responding hand were counterbalanced for the whole experiment. Thus, 26 subjects started with headphone orientation A. I3 of whom were asked to respond to with the right hand (R) and another 13 of whom were to respond with the left hand (L). The other 26 subjects started with reversed headphone orientation (B) using the same order of responding hands. Respondmg hand and headphone orientation were changed after each block (ABABABAB with RLRLRLRL or LRLRLRLR: BABABABAB with RLRLRLRL or LRLRLRLR). In both groups, the numbers of right-handers. left-handers, men and women were counterbalanced. The frequency of target detection (hit rate), false alarms and the reaction times (RT) for each ear were stored. Previousexperiments in our laboratory had shown Pearson test-retest correlations of0.78 and 0.72 for the laterality indices (“e”) for ,/(a/ and ,‘da,‘hit rates. and correlations of 0.73 and 0.64 for the differences of right- and left-ear ,‘ta,’ and ,idai reaction times (15 healthy right-handed, malt subjects; test intervals: 34 weeks). All subjects were given 30 practice trials for each target syllable prior to the experiment. (iv) Morse code recr~I1. Seven “letters” represented by either dots or dashes were chosen from the international Morsecodealphabet (e. ,i . . ,s.. . ,h . . . . , t -, m --, o ---). A single stimulus ofa dichotic pair consisted of one or two Morse letters. A personal computer was used to control two electronic oscillators which produced the signals. The oscillators were set at frequencies of 1000 and 2000 Hz to allow distinction between the two channels. Stimuli were presented via stereo headphones (Sennheiser HD222) at an intensity level of 65 dB (SPL). All dichotic Morse pairs were matched in length. A dot was counted as one unit long, a dash as three units. a pause in a Morse letter as one unit. and a pause between subsequent Morse letters as three units. Unit length was 75 msec. During the whole test, 30 Morse pairs were presented dichotically. In 10 ofthese trials, dots were presented to the left ear, while dashes were presented to the right ear. Ten trials were reversed. and 10 trials employed mixed dot-dash sequences on both ears. The order of these three blocks of 10 trials was randomized across subjects. Each ear received identical numbers ofdots and dashes. None of the subjects had previous knowledge of Morse code. The probands proceeded at their own pace by pressing a button to summon a stimulus pair. They were instructed to direct attention to both ears and to report the perceived Morse sequences immediately after stimulus presentation in self-chosen order. (The “Morse” sequences were vocalized using two different lengths of the syllable /ta,‘.) The test was scored such that partial credit was given if one letter in a Morse “syllable” was correctly reported. For a “completely correct” response, one point was given. After live trials, headphone orientation was reversed. Half of the subjects started with headphone orientation A (right-ear channel right, left-ear channel left). The other subjects started with reversed headphone orientation B. In both groups, the numbers of right-handers. left-handers, men and women were counterbalanced. All subjects were given 10 practice trials prior to the experiment. Design of this paradigm was based on the assumption that ear advantages produced by lOOO- or 2000-Hz Morse code signals are likely to reflect hemispheric left-right differences in the processing of rapidly changing temporal cues. It is unlikely from the data of Gt?FFENand REYNOLDS [13] that lOOO- or 2000-Hz tones are subject to a bias due to tone height. (d) Daru anul~.sis (i) Accurac): score and hteraliry index “e”. Accuracy scores were determined for each test as percentages of correct responses. To assess ear asymmetry, the laterality index “e” according to HALWES 173 was used because this coefficient is independent of the subjects’ overall accuracy and has been shown to produce sensitive isolaterality curves 171. The following formulas applied: e= R-L/(R+ L) if overall accuracy was less than .50X, or e = R ~ L,‘(R, - L,) if overall accuracy was greater than 50% [R, L: total numbers of correct responses for targets presented to the right (R) or left (L) ear, R,, L,: total numbers oferrors for targets presented to the right (R,) or left (L,) ear]. It should be noted at this point that other indicators ofear asymmetry were also determined in the present
944
L.
J~CKF
et ul.
study without revealing different results (e.g. simple left -right difference. left right difference relative to total score). In addition, differences between reaction times for correct right- or left-ear target detection were calculated in the monitoring tests. A negative reaction time difference indicated a REA, a positive difference a left-ear advantage (LEA). Laterality indices and reaction time differences were used for further analysis. (ii) Splil-half‘ correlations. To estimate test reliability, we calculated the “consistency” using the formula 01 Spearman-Brown (R,, = (2 * R, *)/( I + R, J. R,,: consistency coefficient, R, z: Pearson correlation between halves of each test 12271).Variables of the Spearman Brown calculations were determined as follows, taking counterbalancing of headphone orientation into account. For digit list recall, the data of the first eight trials of the first half of the test and of the first eight trials of the second half of the test were pooled. From these, a latcrality index was computed. A second laterality coeficient was computed from the pooled data of the other trials. For CV recall and CV monitoring. laterality indices were determined from the first and last four blocks. For Morse code recall. the data from the first five trials of each block were pooled. From these a laterality index was computed. A second laterahty coeficient was determined from the second five trials of each block. Because reliability depends heavily on test length (high reliability with long tests), the Spearman- Brown reliabilities had to be corrected for this effect. Thus, we calculated an estimate of reliability which corrected for overall test length by using the Spearman Brown relation (LR;, = (n’,‘n* R,,)/( 1 + (n’,‘n - I ) * R,,)], R;, : corrected reliability, R,,: original reliability, n: number of test stimuli. n’: number oftest stimuli for the longer test version). For digit list recall and Morse code recall we estimated the reliability for a test length of 240 stimuli pairs. that is, for the same number of stimuli as in CV recall and CV monitoring. (iii) Sturistical ana/~.sis. A two-way multivariate analysis of variance (MANOVA) with “handedness” and “sex” as factors was computed for the laterality coefficients. The accuracy scores were analysed with a three-way univariate analysis of variance (ANOVA), with “sex”, ” handedness” and “dichotic tebts” (repeated measurements factor) as factors. In addition. frequency data for all dichotic tests concerning the Incidence of ear advantage in connection with handedness and sex w-ere subjected to seven Hierarchical Log- Linear analyses [I 51. Furthermore, 21 pair-bypair cross-tabulation analyses were computed and analysed with the 2’ test for each handedness group. Finally. 21 Spearman- Rank-intertest correlations were performed for each handedness group. Because several tests were computed for one sample, type-l-error was adjusted according to HOLM [16]. A significance level of 10% and a twosided test problem was chosen. All statistical analysts were performed with the statistical software package SPSSPC-V2.I.
RESULTS We obtained the following reliabilities based on the Spearman-Brown formula (see Methods): digit list recall, 0.29 (RI,: 0.50); CV recall, 0.83; /ta/ hit rate, 0.85; /a/ reaction time, 0.88; /da/ hit rate, 0.75; /da/ reaction time, 0.78; Morse code recall, 0.50 (R;,: 0.88). These reliabilities were similar for right- and left-handers. As shown in Table 1, the REA incidences in the different dichotic tests (Table 1) corresponded roughly to those reported in other studies on normal individuals [7], ranging from 58 to 85% in the 26 right-handers and from 54 to 70% in the 26 left-handers of the present study. Due to sample size, there was no statistically significant effect of handedness on REA or LEA incidence in any test. Furthermore, there was no evidence for an influence of sex or sex x hand interaction. Hit rate and reaction time for detection of the syllable /da/ in the CV monitoring task were surprisingly low in right-handers (Table 1). Table 2 shows the group means and standard deviations of laterality scores obtained for the different tests. There was considerable variation of laterality scores in both hand groups. The coefficient of variation (S.D. * loo/mean) ranged from 133 to 780”/0 in right-handers. In left-handers. these coefficients were higher, ranging from 224 to 1542%. Table 2 demonstrates that right-handers showed a significant REA only for digit list recall. Four tests yielded a trend towards a REA in this group (CV recall, /ta/ hit rate, ita/ reaction time, Morse code recall). In left-handers, there was no significant REA in any of the dichotic tests. Two tests revealed a trend towards a REA in this hand group (ita/ and /da/ hit rates). The MANOVA performed for the laterality coefficients revealed no significant result [handedness: F(7,42)=1.51, P=O.191; sex: F(7,42)=2.39,P=O.O37; handxsex: F (7,42)= 1.10. P=O.379]. However, this MANOVA and subsequent ANOVAs revealed
DICHOTIC
945
LISTENIMi
Table 1. Percentages of right-handers (RH, n = 26) and left-handers (LH, n=26) showing a right-ear advantage (REA), a left-ear advantage (LEA) or no ear advantage (NEA) in seven different dichotic tests REA (%) RH LH Digit list recall CV recall CVmon /ta/ hit rate CVmon /ta/ RT r 1 CVmon /da,’ hit rate CVmon /da/ RT r-l Morse code recall
85 77 73 73 58 58 70
LEA or NEA (%) RH LH
61 61 61 61 54 70 54
15 23 27 27 42 42 30
CVmon: CV monitoring, RT r I: difference between reaction presented to the right (r) or left (I) ear.
Table 2. Group
39 39 39 39 46 30 46 times for targets
means and standard deviations (S.D.) of laterality scores (RH, n =26) and left-handers (LH, n = 26) RH
Digit list recall CV recall CVmon ita,’ hit rate CVmon ;ta/ RT r -1 CVmon /da,/ hit rate CVmon ,/da/ RT r- 1 Morse code recall
Mean
S.D.
t
0.21 0.04 0.09 -36.95 0.01 9.19 0.06
(0.28) (0.10) (0.13) (65.10) (0.14) (71.73) (0.15)
3.75 2.23 3.40 2.84 0.46 0.63 2.06
P
DICHOTIC
LISTENING:
WHAT
DOES IT MEASURE?
LUTZ J~NCKE,* HELMUTH STEINMETZ~ and JENS V~LKMANN? *Department of Psychobiology and Psychocybernetics, Heinrich-Heine-University, Universitatsstral3e D-4000 Dusseldorf 1, F.R.G. and tDepartment of Neurology, Heinrich-Heine-University. Moorenstrage D-4000 Diisseldorf 1, F.R.G. (Receiced
2 Septemhw
1991; accepted
1, 5.
2 1 July 1992)
Abstract-Auditory lateralization was investigated in 26 right-handed and 26 left-handed, normal subjects using seven different dichotic listening tests in each proband (free recall of digit lists, free recall ofconsonant vowel (0’) svllables, four different CV syllable monitoring paradigms, and free recall of Morse codes). Reliabilttjes calculated with the formula of Spearman Brown were low for digit recall (0.29, corrected for test length: 0.50) but good for CV recall (0.83), CV monitoring (0.75~ 0.88). and Morse code recall (0.50, corrected for test length: 0.88). Nevertheless, intertest correlations were low, both for right- and left-handers (negative correlations ranging from -0.44 to -0.05, positive correlations ranging from 0.01 to 0.51). Only 38 77% of the right-handed and lefthanded subjects retained one direction of ear advantage across any combination of two tests. The data suggest that different dichotic tests reveal different results. This may be due to psychometrtc. procedural, or phonetic properties. We conclude that individual predictions of language dominance are not justified using the dichotic tests evaluated in the present study.
INTRODUCTION SINCE KIMURA’S original digit list paradigm [18, 191, several paradigms using other verbal stimuli have been proposed to increase the predictive validity of the dichotic method with respect to language lateralization (for overviews see Refs [2], [S], [7--91 and [36]). Among these tasks, the recall or monitoring of dichotic consonant-vowel (CV) syllables [30,35,38] has been consistently found to produce the most robust right-ear advantages (REAs) in right-handed individuals, an effect thought to be due to the superior temporal discrimination of acoustic signals by the dominant hemisphere [4, 71. In accordance with this assumption. several studies demonstrated REAs also for nonverbal acoustic stimuli with rapidly changing temporal cues [23, 391. Only a few investigations, however, directly assessed the predictive validity of dichotic paradigms with respect to language lateralization as determined by invasive methods [ll, 19, 37, 441. In contrast to the encouraging results of these studies, other authors were much more cautious on the validity issue, mainly due to poor dichotic retest or intertest comparabilities [3, 5,401. The present study addresses this problem. We report intertest correlations for seven dichotic paradigms, all known to produce strong REAs and thought to reflect language lateralization, but differing with respect to stimulus material (digits, CV syllables, or nonverbal stimuli), stimulus presentation (list or single-pair) or stimulus reproduction (free recall or target detection). The investigation thus attempts to further elucidate the question whether auditory lateralization is likely to be related to one
*Address for correspondence: Dr Heinrich-Heine-University Dusseldorf,
Lutz Jancke, Department of Psychobiology and Universitatsstrabe I. D-4000 Dusseldorf I, F.R.G. 941
Psychocybernetics.
942
L. J;(N~KE rl trl
specialized mechanism of the dominant hemisphere [4,38] indeed be expected to also underlie language dominance.
which, if this were the case, could
METHODS (a) Suhjrcrs We examined 52 healthy, paid volunteers without hearmg deficits in the speech frequency range (less than IO dB hearing loss on either ear tested by standard audiometry). Twenty-six of the 52 subjects were self-described sinistrals (mean age: 26.4 years; age range: 21 35 years) and 26 were self-described dcxtrals (mean age: 24.9 years age; age range: 2l--34 years). Most ofthem were university students. There were 13 males and I3 females in each handedness group.
Hand preference was assosed using the five-item questionnaire OIBKYIIFY [6,7]. Subjects were classified as either left-handed (Bryden scores ~0) or right-handed (Bryden scores >O) L6.71. This classilicatlon conformed with the subject’s self-description in all casts (XC a).
All subjects were examined with the following dichotic tests: (i) digit list recall. (ii) CV recall, (iii) CV monitoring with target syllables;ta; or/da: (hit rate and reaction time). and (iv) Morse code recall. Tests (i) and (iv) were applied in one session, tests (ii) anJ (iii) In another session. Half of the subjects started with session I, the other with session 2. Interbals between sessions ranged from I to 5 days. Within each session. half of the subjects started with test (i) or (iv) and the other with test (ii) or (iii). (i) Dig/if list vt~c~/l. The test material consisted of 96 pairs of German monosyllabic digits (/ems.. ;rwei:. idrei,‘, ,‘vier.. ‘fiinf,. . ,/sechs,,,, ,‘acht,,‘. ,‘neun,.‘. , rehn . :elf,‘. ,‘zwolf:) spoken by a male person. These stimuli were band-pass filtered (upper corner frequency 10 kHz, 24 dB,,octave and lower corner frequency 80 Hz. 24 dB,‘octave), digitized (AjD conversion. 16 bit). and stored in computer memory (PDPI I computer). Each dlgit wa, edited and synchromred for intensity and onset with a speech editor program. The criterion for temporal alignment of digits was onset of articulatory rcleasc. After randomizatmn for digit order and exclusion of pairs of identical digits. the stimuli were D A converted (20 kHr, resolution I6 hit). Digits forming a pair were rccordcd on separate tracks of a magnetic recording tape usmg different channels of the D:A conbertcr. With this technique. competing digits uerc aligned with an accuracy of 50 /Isec. Thirty-three trials. each trial consisting of three digit pairs, were recorded on tape. Within each trial, &git pairs were presented with I-XX intcr%als (interval from onset ofpreceding digit to onset of following digit). Each trial was followed hq a 6-set pause. Dichotic pairs were presented using a stereo tape recorder (AIWA AD-F360). stereo headphones (Scnnhelser HD222). and at an intensity of about 70 dB (SPL). Suhjccts were instructed to pay attenllon to both ears. They were asked to write down the percaved stimuli immediately after each trial. lrrcspective of order or ear of presentation. After 16 trials, headphone orientation was reversed. Half of the subjects started with headphone orientation A (right-car channel right. left-car channel left). The other half started with rckerscd headphone orientation B. In both groups. the numbers of right-banders, leftbanders. men and women were counterbalanced. Prior to each experiment. 10 practice trials were carried out. (ii) C‘I’wc011. This test comprised paira of natural speech CV syllables (:ba:, :ga,, da., :p:~,‘, /ka,‘, and IIta; ) spoken by a trained phonetician (microphone .AKG 16OEl) which had heen band-pass liltercd (Rockland 432, upper corner frequency IO kHz 24 dB octave and lower corner frequency X0 Hz 24 dB:octave). Each syllabic WBSdigitized and stored in computer memory (l2-hit A;D converter. digitizing rate 20.229 Hz. ATARI MEGA ST4). The onset, duration. intensity and fundamental frequency of the stimuli were then edited and synchronized by means of a speech editor. The criterion for temporal alignment of the competing syllables was the onset of articulatory release. Each syllable was preceded by IO0 msec ofzero intenTit> before articulation started and followed by 50 msec ofrero intensity after vowel end. This manipulation served to remove clicks at the beginning and end upon D:A convcraion. The duration of syllables ranged from 310 to 360 msec, depending on voice onset times, with a +owel duration of 300 msec. Voice onset times (in mscc) for the stops were approximately b- IO. g=20. d = 15. p=40, k =60 and t = 50. The duration of the entire signals (syllables. leading and end interval with Lero Intensity) ranged from 450 to 500 msec. These edited Ggnals were stored on hard disk (MEGAFILE 30) for further use. For dichotic presentation, the stored sign& were D,‘A converted (I 2.bit. four-channel D’A converter, digitizing rate 20.229 Hz) by alternately converting a value of the first syllable and then the second syllable. Each syllahlc was conberted using a diflerent channel of the D.A con\crtcr. With this technique. competmg syllables were aligned with an accuracy of49.4 /(sec. After con>eraion the output of eachD A channel &as amplified (70 dB SPL as measured by a Bruel & Kjaer sound level meter) and presented via stereo headphones (Sennheiscr HD450). During the cupcriments. the subjects were seated in a sound-attenuated chamber. received instructions through a two-way intercom set and were instructed to pay attention to both cars. The presentation of dichotic stimuli and the recall procedure were controlled by the
DlCHOTIC
LlSTENlNG
943
computer. For presentation, the CV pair was loaded from hard disk into computer memory and D/A converted. After stimulus application, a diagram showing six squares containing the six syllables appeared on a monitor in front of the subjects for 5 sec. The subjects were instructed to “click” the syllable or syllable pair they had heard. Prior to each experiment, 30 practice trials were carried out. Test stimuli were randomized and arranged in eight blocks of 30 CV pairs in which each ofthe six syllables was paired with the other five (no syllable was paired with itself). Thus, the whole experiment consisted of 240 test trials, each trial consisting of one pair of syllables followed by the 5-set response interval. Headphone orientation was reversed after each block. Halfof the subjects started with headphone orientation A (right-ear channel right, left-ear channel left), whereas the other half of the subjects started with reversed headphone orientation B. In both groups, the numbers of right-handers, left-handers, men and women were counterbalanced. Previous experiments in our laboratory had shown a Pearson test-retest correlation of 0.81 for the laterality index (“e”) (see below) (15 healthy male right-handers: test-intervals: 3 4 weeks). (iii) CVmonitoring. The same stimuli and technical equipment as in the CV recall test were used. The stimuli were randomized and arranged in eight blocks, each containing 30 syllable pairs. resulting in a total of 240 trials. Each of the 30 possible pair-wise permutations of the six CV syllables occurred equally often. Because all possible CV pairs were equiprobable, one-third of the trials consisted of pairs containing the target. Fifty per cent of the targets were delivered to the right and 50% to the left ear. Thus, the a priori probability of target occurrence for each ear was 0.167. The interval between syllable pairs was 1.9 sec. Subjects were instructed to pay attention to both ears and press a button with his or her forefinger when the target syllable was heard. Reactions within 100~1000 msec after stimulus onset were accepted. Subjects underwent this procedure twice, once with the target :ta/ and once with the target ,‘da/. Headphone orientation and responding hand were counterbalanced for the whole experiment. Thus, 26 subjects started with headphone orientation A. I3 of whom were asked to respond to with the right hand (R) and another 13 of whom were to respond with the left hand (L). The other 26 subjects started with reversed headphone orientation (B) using the same order of responding hands. Respondmg hand and headphone orientation were changed after each block (ABABABAB with RLRLRLRL or LRLRLRLR: BABABABAB with RLRLRLRL or LRLRLRLR). In both groups, the numbers of right-handers. left-handers, men and women were counterbalanced. The frequency of target detection (hit rate), false alarms and the reaction times (RT) for each ear were stored. Previousexperiments in our laboratory had shown Pearson test-retest correlations of0.78 and 0.72 for the laterality indices (“e”) for ,/(a/ and ,‘da,‘hit rates. and correlations of 0.73 and 0.64 for the differences of right- and left-ear ,‘ta,’ and ,idai reaction times (15 healthy right-handed, malt subjects; test intervals: 34 weeks). All subjects were given 30 practice trials for each target syllable prior to the experiment. (iv) Morse code recr~I1. Seven “letters” represented by either dots or dashes were chosen from the international Morsecodealphabet (e. ,i . . ,s.. . ,h . . . . , t -, m --, o ---). A single stimulus ofa dichotic pair consisted of one or two Morse letters. A personal computer was used to control two electronic oscillators which produced the signals. The oscillators were set at frequencies of 1000 and 2000 Hz to allow distinction between the two channels. Stimuli were presented via stereo headphones (Sennheiser HD222) at an intensity level of 65 dB (SPL). All dichotic Morse pairs were matched in length. A dot was counted as one unit long, a dash as three units. a pause in a Morse letter as one unit. and a pause between subsequent Morse letters as three units. Unit length was 75 msec. During the whole test, 30 Morse pairs were presented dichotically. In 10 ofthese trials, dots were presented to the left ear, while dashes were presented to the right ear. Ten trials were reversed. and 10 trials employed mixed dot-dash sequences on both ears. The order of these three blocks of 10 trials was randomized across subjects. Each ear received identical numbers ofdots and dashes. None of the subjects had previous knowledge of Morse code. The probands proceeded at their own pace by pressing a button to summon a stimulus pair. They were instructed to direct attention to both ears and to report the perceived Morse sequences immediately after stimulus presentation in self-chosen order. (The “Morse” sequences were vocalized using two different lengths of the syllable /ta,‘.) The test was scored such that partial credit was given if one letter in a Morse “syllable” was correctly reported. For a “completely correct” response, one point was given. After live trials, headphone orientation was reversed. Half of the subjects started with headphone orientation A (right-ear channel right, left-ear channel left). The other subjects started with reversed headphone orientation B. In both groups, the numbers of right-handers. left-handers, men and women were counterbalanced. All subjects were given 10 practice trials prior to the experiment. Design of this paradigm was based on the assumption that ear advantages produced by lOOO- or 2000-Hz Morse code signals are likely to reflect hemispheric left-right differences in the processing of rapidly changing temporal cues. It is unlikely from the data of Gt?FFENand REYNOLDS [13] that lOOO- or 2000-Hz tones are subject to a bias due to tone height. (d) Daru anul~.sis (i) Accurac): score and hteraliry index “e”. Accuracy scores were determined for each test as percentages of correct responses. To assess ear asymmetry, the laterality index “e” according to HALWES 173 was used because this coefficient is independent of the subjects’ overall accuracy and has been shown to produce sensitive isolaterality curves 171. The following formulas applied: e= R-L/(R+ L) if overall accuracy was less than .50X, or e = R ~ L,‘(R, - L,) if overall accuracy was greater than 50% [R, L: total numbers of correct responses for targets presented to the right (R) or left (L) ear, R,, L,: total numbers oferrors for targets presented to the right (R,) or left (L,) ear]. It should be noted at this point that other indicators ofear asymmetry were also determined in the present
944
L.
J~CKF
et ul.
study without revealing different results (e.g. simple left -right difference. left right difference relative to total score). In addition, differences between reaction times for correct right- or left-ear target detection were calculated in the monitoring tests. A negative reaction time difference indicated a REA, a positive difference a left-ear advantage (LEA). Laterality indices and reaction time differences were used for further analysis. (ii) Splil-half‘ correlations. To estimate test reliability, we calculated the “consistency” using the formula 01 Spearman-Brown (R,, = (2 * R, *)/( I + R, J. R,,: consistency coefficient, R, z: Pearson correlation between halves of each test 12271).Variables of the Spearman Brown calculations were determined as follows, taking counterbalancing of headphone orientation into account. For digit list recall, the data of the first eight trials of the first half of the test and of the first eight trials of the second half of the test were pooled. From these, a latcrality index was computed. A second laterality coeficient was computed from the pooled data of the other trials. For CV recall and CV monitoring. laterality indices were determined from the first and last four blocks. For Morse code recall. the data from the first five trials of each block were pooled. From these a laterality index was computed. A second laterahty coeficient was determined from the second five trials of each block. Because reliability depends heavily on test length (high reliability with long tests), the Spearman- Brown reliabilities had to be corrected for this effect. Thus, we calculated an estimate of reliability which corrected for overall test length by using the Spearman Brown relation (LR;, = (n’,‘n* R,,)/( 1 + (n’,‘n - I ) * R,,)], R;, : corrected reliability, R,,: original reliability, n: number of test stimuli. n’: number oftest stimuli for the longer test version). For digit list recall and Morse code recall we estimated the reliability for a test length of 240 stimuli pairs. that is, for the same number of stimuli as in CV recall and CV monitoring. (iii) Sturistical ana/~.sis. A two-way multivariate analysis of variance (MANOVA) with “handedness” and “sex” as factors was computed for the laterality coefficients. The accuracy scores were analysed with a three-way univariate analysis of variance (ANOVA), with “sex”, ” handedness” and “dichotic tebts” (repeated measurements factor) as factors. In addition. frequency data for all dichotic tests concerning the Incidence of ear advantage in connection with handedness and sex w-ere subjected to seven Hierarchical Log- Linear analyses [I 51. Furthermore, 21 pair-bypair cross-tabulation analyses were computed and analysed with the 2’ test for each handedness group. Finally. 21 Spearman- Rank-intertest correlations were performed for each handedness group. Because several tests were computed for one sample, type-l-error was adjusted according to HOLM [16]. A significance level of 10% and a twosided test problem was chosen. All statistical analysts were performed with the statistical software package SPSSPC-V2.I.
RESULTS We obtained the following reliabilities based on the Spearman-Brown formula (see Methods): digit list recall, 0.29 (RI,: 0.50); CV recall, 0.83; /ta/ hit rate, 0.85; /a/ reaction time, 0.88; /da/ hit rate, 0.75; /da/ reaction time, 0.78; Morse code recall, 0.50 (R;,: 0.88). These reliabilities were similar for right- and left-handers. As shown in Table 1, the REA incidences in the different dichotic tests (Table 1) corresponded roughly to those reported in other studies on normal individuals [7], ranging from 58 to 85% in the 26 right-handers and from 54 to 70% in the 26 left-handers of the present study. Due to sample size, there was no statistically significant effect of handedness on REA or LEA incidence in any test. Furthermore, there was no evidence for an influence of sex or sex x hand interaction. Hit rate and reaction time for detection of the syllable /da/ in the CV monitoring task were surprisingly low in right-handers (Table 1). Table 2 shows the group means and standard deviations of laterality scores obtained for the different tests. There was considerable variation of laterality scores in both hand groups. The coefficient of variation (S.D. * loo/mean) ranged from 133 to 780”/0 in right-handers. In left-handers. these coefficients were higher, ranging from 224 to 1542%. Table 2 demonstrates that right-handers showed a significant REA only for digit list recall. Four tests yielded a trend towards a REA in this group (CV recall, /ta/ hit rate, ita/ reaction time, Morse code recall). In left-handers, there was no significant REA in any of the dichotic tests. Two tests revealed a trend towards a REA in this hand group (ita/ and /da/ hit rates). The MANOVA performed for the laterality coefficients revealed no significant result [handedness: F(7,42)=1.51, P=O.191; sex: F(7,42)=2.39,P=O.O37; handxsex: F (7,42)= 1.10. P=O.379]. However, this MANOVA and subsequent ANOVAs revealed
DICHOTIC
945
LISTENIMi
Table 1. Percentages of right-handers (RH, n = 26) and left-handers (LH, n=26) showing a right-ear advantage (REA), a left-ear advantage (LEA) or no ear advantage (NEA) in seven different dichotic tests REA (%) RH LH Digit list recall CV recall CVmon /ta/ hit rate CVmon /ta/ RT r 1 CVmon /da,’ hit rate CVmon /da/ RT r-l Morse code recall
85 77 73 73 58 58 70
LEA or NEA (%) RH LH
61 61 61 61 54 70 54
15 23 27 27 42 42 30
CVmon: CV monitoring, RT r I: difference between reaction presented to the right (r) or left (I) ear.
Table 2. Group
39 39 39 39 46 30 46 times for targets
means and standard deviations (S.D.) of laterality scores (RH, n =26) and left-handers (LH, n = 26) RH
Digit list recall CV recall CVmon ita,’ hit rate CVmon ;ta/ RT r -1 CVmon /da,/ hit rate CVmon ,/da/ RT r- 1 Morse code recall
Mean
S.D.
t
0.21 0.04 0.09 -36.95 0.01 9.19 0.06
(0.28) (0.10) (0.13) (65.10) (0.14) (71.73) (0.15)
3.75 2.23 3.40 2.84 0.46 0.63 2.06
P
