| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
The Journal of Neurophysiology Vol. 87 No. 1 January 2002, pp. 122-139
Copyright ©2002 by the American Physiological Society
1Department of Neurology, Harvard Medical School and Massachusetts General Hospital, Boston 02114-2696; 2Research Laboratory of Electronics, Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge 02139; and 3Eaton-Peabody Laboratory of Auditory Physiology, Department of Otology and Laryngology, Harvard Medical School and Massachusetts Eye and Ear Infirmary, Boston, Massachusetts 02114
 |
ABSTRACT |
|---|
 TOP
 ABSTRACT
 INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Tramo, Mark Jude,
Gaurav D. Shah, and
Louis D. Braida.
Functional Role of Auditory Cortex in Frequency Processing and
Pitch Perception.
J. Neurophysiol. 87: 122-139, 2002.
Microelectrode studies in nonhuman
primates and other mammals have demonstrated that many neurons in
auditory cortex are excited by pure tone stimulation only when the
tone's frequency lies within a narrow range of the audible spectrum.
However, the effects of auditory cortex lesions in animals and humans
have been interpreted as evidence against the notion that neuronal
frequency selectivity is functionally relevant to frequency
discrimination. Here we report psychophysical and anatomical evidence
in favor of the hypothesis that fine-grained frequency resolution at
the perceptual level relies on neuronal frequency selectivity in
auditory cortex. An adaptive procedure was used to measure difference
thresholds for pure tone frequency discrimination in five humans with
focal brain lesions and eight normal controls. Only the patient with bilateral lesions of primary auditory cortex and surrounding areas showed markedly elevated frequency difference thresholds: Weber fractions for frequency direction discrimination
("higher"
"lower" pitch judgments) were about eightfold higher
than Weber fractions measured in patients with unilateral lesions of
auditory cortex, auditory midbrain, or dorsolateral frontal cortex;
Weber fractions for frequency change discrimination
("same""different" pitch judgments) were about seven times
higher. In contrast, pure-tone detection thresholds, difference
thresholds for pure tone duration discrimination centered at 500 ms,
difference thresholds for vibrotactile intensity discrimination, and
judgments of visual line orientation were within normal limits or only
mildly impaired following bilateral auditory cortex lesions. In light
of current knowledge about the physiology and anatomy of primate
auditory cortex and a review of previous lesion studies, we interpret
the present results as evidence that fine-grained frequency processing
at the perceptual level relies on the integrity of finely tuned neurons
in auditory cortex.
| |
INTRODUCTION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Many neurons in
mammalian auditory cortex respond selectively to pure tone frequencies
over a narrow range of the audible spectrum, and several anatomical
subdivisions of auditory cortex are organized topographically with
respect to frequency selectivity (for reviews, see Aitkin et al.
1984
; Brugge and Reale 1985; Clarey et
al. 1992; de Ribaupierre 1997; Kaas et
al. 1999; Phillips et al. 1991; Schreiner
1992). On theoretical grounds, it would seem reasonable to
propose that neuronal frequency selectivity at the physiological level
contributes to frequency discrimination at the perceptual level.
However, the functional effects of auditory cortex lesions in animals
and humans have been interpreted as evidence against this hypothesis.
In their authoritative review of a large number of lesion studies in
the Handbook of Sensory Physiology, Neff et al.
(1975) concluded, "Frequency discrimination may be learned or
relearned after bilateral lesions involving all or nearly all of
primary auditory cortex in animals such as the cat and monkey and in
human patients." Similar observations have fueled the claim that
"tonotopic organization at the cortical level is not necessary for
the perception of tones" (Guttman and Diamond 1963).
Still, few analyses of lesion effects have combined measurements of
frequency discrimination at psychophysical thresholds and precise
definition of pathoanatomical correlates. Moreover, a number of
methodological issues hamper straightforward interpretations of lesion
effects in cats, the principal source of animal data (for review, see
Elliot and Trahiotis 1972).
In the present paper, we provide psychophysical and anatomical evidence supporting the hypothesis that sharp frequency tuning in auditory cortex is functionally relevant to fine-grained frequency processing.
| |
METHODS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Subjects
CASE A1+.
MHS, a 35-yr-old mixed-handed man, formerly employed as a
machinist, suffered ischemic infarcts in the distribution of the right
middle cerebral artery in 1980 and the left middle cerebral artery in
1981. Details pertaining to the neurological history and examination,
audiological tests, neuropsychometric tests, brain imaging, and
electrophysiological studies have been previously published
(Mendez and Geehan 1988; Musiek et al.
1994; Tramo et al. 1990). We summarize the
auditory, pathoanatomical, and evoked potential findings here. We refer
to MHS as "case A1+" to designate distribution of his lesions, which involve A1 plus surrounding cortex bilaterally.
) fell below the
20th percentile, with performance on the Timbre, Tonal Memory, and Time
subtests below the 5th percentile. Magnetic resonance imaging
(MRI) revealed lesions involving the entire right transverse gyrus of
Heschl (TG), all or almost all of left TG, all or almost all of the
right superior temporal gyrus (STG, 98% by surface area estimation),
left posterior STG (20% of surface area), and other structures (Figs.
1-3).
Brain stem auditory-evoked potentials were normal, and middle latency
and long latency auditory-evoked potentials were abnormal in both
hemispheres. At the time of the present experiments, case
A1+ complained of difficulty understanding words that were spoken
or sung, especially over the telephone or in noisy environments. He
reported difficulty localizing sounds. He enjoyed listening to music
and was able to recognize familiar songs. Signs on neurological
examination included impaired speech perception, decreased speech
intonation, a spastic left hemiparesis, and left somesthetic loss.
Medications were limited to coumadin. He works at home raising two
young children with his wife, a grade school teacher.
|
|
|
PATIENT CONTROLS. Four young adults with neurological disease who did not have bilateral lesions of auditory cortex were recruited for paid participation. These patient controls provided a means of assessing nonspecific effects of brain lesions, such as difficulty sustaining concentration, on auditory task performance. We refer to these patient controls as "AC± cases" to designate variable, unilateral involvement of auditory cortex. Further details of each case are provided below.
RJC. RJC, a 36-yr-old right-handed man, formerly employed as a school administrator, suffered a small brain contusion secondary to closed head trauma in a 1987 motor vehicle accident. CT demonstrated a small hyperdensity in the dorsolateral midbrain in the region of the right inferior colliculus and rostral lateral lemniscus (Fig. 4). At the time of the present experiments, RJC complained of difficulty localizing sound, balancing stereo headphones and car speakers while listening to music, remembering details, sustaining concentration, and using his right arm because of pain dating back to traumatic injuries of the right clavicle and shoulder. Neurological examination was within normal limits except for decreased light touch sensation and tenderness over the supraclavicular region. Brain stem auditory-evoked potentials showed decreased wave V amplitude following left ear stimulation and a more moderately decreased amplitude following right ear stimulation. Electroencephalography (EEG) was normal. Medications included verapamil, which abolished posttraumatic migraines, and paroxetine, which successfully treated depression. RJC has a college degree and no formal music training; recently he has been trying to teach himself how to play guitar.
|
CASE CTS. CTS, a 27-yr-old left-handed woman, currently employed as an account manager, has a left frontotemporal arachnoid cyst, left congenital deafness, mild cerebellar tonsillar herniation, paroxysmal nocturnal dystonia, left frontal sharp waves and slowing on EEG, and a childhood history of learning difficulties. MRI in 1994 demonstrated absence of normal tissue involving the anterior half of STG, middle temporal, inferior temporal, middle frontal, and inferior frontal gyri, the temporal pole, and portions of the orbitofrontal gyri, precentral gyrus, amygdala, parahippocampal gyrus, hippocampus, and fusiform gyrus (Fig. 5). She came to neurological attention because of common migraine and musculoskeletal headaches that responded to naproxen treatment, treatment of her sleep disorder with carbamazepine, and modification of her diet. Neurological signs included sensorineural deafness of the left ear and moderate end-gaze horizontal nystagmus. CTS has a high school degree, never took music lessons, and enjoys dancing and listening to music during aerobic exercise.
|
CASE MAH. MAH, a 44-yr-old right-handed man, currently employed as a mail clerk, suffered closed-head trauma during a motor vehicle accident in 1975 that was complicated by cerebral contusions and complex partial seizures that typically begin with unformed auditory hallucinations. CT in 1990 demonstrated three areas of hypodensity involving the following structures: the posterior two-thirds of right STG, part of right TG, and adjacent portions of middle temporal, inferior temporal, supramarginal, angular, and postcentral gyri; part of the left inferior frontal gyrus and left precentral gyrus; and part of the left superior frontal gyrus (Fig. 6). EEG in 1993 showed right and left temporal lobe sharp waves and slowing. Excellent seizure control has been achieved over the past 2 years with carbamazepine and clorazepate. The neurological history is also notable for a mixed headache disorder (musculoskeletal headaches and common migraine), currently treated with naproxen; depression, currently treated with paroxetine; and past polysubstance abuse, including 5 yr of daily ethanol ingestion, now in remission for 12 yr. At the time of the present experiments, neurological signs included mild-to-moderate impairments in concentration, spatial orientation, verbal and visual recall, oculomotor control, gait, and distal peripheral nerve sensation. Laboratory data were remarkable for a mild anemia and folate deficiency. MAH has a high school degree, 3 years of college education, and no music training.
|
CASE YPT. YPT, a 28-yr-old right-handed woman, currently employed as bank clerk, suffered a spontaneous intracerebral hemorrhage in 1993. MRI demonstrated signal abnormalities involving posterior portions of the right middle and inferior frontal gyri (Fig. 7). At the time of the present experiments, YPT was being treated with carbamazepine for seizure prophylaxis, and neurological signs included a mild impairment in concentration and mild clumsiness of the left hand. YPT has a college degree and no music training.
|
NORMAL CONTROLS. Eight adults with no active symptoms of neurological or otological disease were recruited for paid participation through advertisements. The age range was 30-52 yr with a median of 40.5 yr. The normal controls had from 12 to 16 yr of education with a median of 15.5 yr. Four subjects were left-handed, four were right-handed. None of the subjects was formally trained in music or actively performing music.
Stimulus synthesis and presentation
Sinusoids were digitally synthesized using an Ariel DSP-96 board
and converted to analog signals at a sampling rate of 11 kHz per
channel using a linear-phase digital anti-alias filter with a bandwidth
of 4.96 kHz. Analog auditory signals were attenuated using Tucker-Davis
PA4 programmable attenuators. Vibrotactile stimuli were amplified using
a Crown amplifier model D-75. Sinusoidal tones (pure tones) were
presented via calibrated TDH-39P headphones with the subject seated in
a double-walled, soundproof booth. The method described by
Gescheider et al. (1990) for measurement of vibrotactile
intensity difference thresholds was adapted to deliver sinusoidal
vibrations at the thenar eminence via a custom-built 1.5 × 2.5 × 3 cm vibrator secured with an elastic bandage.
Adaptive procedure and experimental tasks
A two-interval, two-alternative forced-choice paradigm and an
adaptive procedure associated with a 79.4% probability of correct response (Levitt 1970) were used to measure difference
thresholds for judging whether: 1) the second of two pure
tone bursts was "lower" or "higher" in pitch than the first
(frequency direction discrimination); and 2) judging whether
the second of two pure tone bursts was the "same" as or
"different" from the first (frequency change discrimination). To
assess possible nonspecific effects of brain lesions (e.g., impaired
level of arousal, general attention, concentration, problem-solving) on
the performance of these two tasks, we also measured: 1)
difference thresholds for determining the direction of an intensity
change between two successive sinusoidal vibrations delivered to the
hand (vibrotactile intensity discrimination); and 2)
accuracy on a standardized match-to-sample test of visual line
orientation perception (Benton et al. 1983). To assess
whether the brain lesions affected auditory functions other than
frequency discrimination, we also measured: 1) difference
thresholds for determining the direction of a duration change between
two successive pure tones with a long center duration of 500 ms
(duration discrimination); and 2) intensity thresholds for
pure tone detection.
Difference thresholds measured adaptively are expressed as Weber fractions and reported as percentages. Weber fractions were calculated by averaging the six turnaround points in each run and averaging across runs.
FREQUENCY DIRECTION DISCRIMINATION.
Difference thresholds for frequency direction discrimination were
measured at center frequencies (Fc's) of 250, 500, 1,000, and 2,000 Hz. Our method was similar to that used by Moore (1973) to define frequency difference thresholds in his highly practiced subject TC. The trial design is schematized in Fig.
8 and the results of a typical run are
illustrated in Fig. 9. Each tone was
presented binaurally at 40 dB above the detection threshold at each ear
[i.e., 40 dB sensation level (SL)] with a starting phase of 0° and
a duration of 500 ms (plateau = 480 ms, rise time = 10 ms,
fall time = 10 ms). A 200-ms silent interval separated the two
tones. Subjects pressed the space bar on the computer keyboard to begin
each trial. Subjects were instructed to press a key labeled "L" if
the pitch of the second tone was lower than that of the first tone or
an adjacent, rightward key labeled "H" if the pitch of the second
tone was higher than that of the first tone. The order of "lower"
and "higher" trials was randomized. There was no constraint on
response time; subjects were encouraged to respond as soon as they knew
the answer and to guess if they were uncertain. If the subject pressed
an invalid key, the trial was discarded. The time between trials was
controlled by the subject and was typically one to a few seconds.
Following three successive correct trials, the frequency difference
(
F) was halved; after each incorrect trial,
F was doubled. The F associated with a reversal in the sign of the frequency change (i.e., an increase or
decrease in F) was defined as a turnaround point. A run
ended when the 6th turnaround point was reached. The mean of the six turnaround points was defined as the difference threshold for that run.
Each subject practiced with feedback for at least three to five runs
or, in a few cases, until performance worsened over any two successive
runs. After a few minutes rest period, three to five test runs were
completed without feedback. The initial value of F on
test runs was set one to two steps above the largest F on
incorrect practice trials so that the initial direction of
F change was downward. The subject's frequency
difference threshold was defined as the mean frequency difference
threshold of the test runs.
|
|
FREQUENCY CHANGE DISCRIMINATION. Difference thresholds for frequency change discrimination were measured at a center frequency of 1,000 Hz. Otherwise, the stimulus parameters, trial design, adaptive procedure, and definition of difference threshold were the same as in the preceding text. Subjects were instructed to press a key labeled "S" if the second tone sounded the same as the first tone or an adjacent, rightward key labeled "D" if the second tone sounded different from the first tone. The order of "same" and "different" trials was randomized.
DURATION DISCRIMINATION. Difference thresholds for duration discrimination were measured at a frequency of 1,000 Hz and a center duration of 500 ms. Stimulus parameters, trial design, adaptive procedure, and definition of the difference threshold were otherwise the same as the preceding text. Subjects were instructed to press a key labeled "S" if the second tone sounded shorter than the first tone or an adjacent, rightward key labeled "L" if the second tone sounded longer than the first tone. The order of "shorter" and "longer" trials was randomized.
VIBROACTILE INTENSITY (VTI) DISCRIMINATION. Difference thresholds for VTI discrimination were measured separately for each hand at a vibration frequency of 250 Hz and at a center vibration intensity of 20 dB SL. The vibrator was secured at the thenar eminence with an elastic bandage, and the arm was rested in the supinated position on a foam rubber pad with the subject seated. The trial design, adaptive procedure, and calculation of the difference threshold were the same as those used for frequency direction discrimination, frequency change discrimination, and duration discrimination. Because case A1+ has a left hemiplegia and thus could not respond manually with the unstimulated hand, all subjects responded by saying "weaker" or "stronger;" the examiner then entered the response on the computer keyboard. The order of "weaker" and "stronger" trials was randomized. Patient control MAH had a peripheral neuropathy, presumably caused by folate deficiency and chronic ethanol exposure in the past, so he was excluded from this control task.
VISUAL LINE ORIENTATION MATCH-TO-SAMPLE TASK.
The Judgment of Line Orientation test (Benton et al.
1983) was administered and scored in standard fashion. In this
task, subjects match the orientations of two lines to the two
corresponding orientations embedded in an array of eleven lines equally
spaced from 0 to 180 degrees.
PURE-TONE AND VIBROTACTILE DETECTION THRESHOLDS. Pure-tone and vibrotactile detection thresholds were measured for each ear and hand, respectively, at each test frequency using a one-interval down-up staircase procedure. Subjects were instructed to indicate when the tone was heard or the vibration felt. The examiner silently entered the response on a computer keyboard. All stimuli were 500 ms in duration with 10 ms rise and fall times. Initial intensity was at least 10 dB above thresholds determined in pilot measurements. The intensity was decreased randomly by 1-5 dB at pseudorandom intervals ranging from 1 to a few seconds. The descending threshold was defined as the first intensity at which no response was elicited. The intensity was then decreased by 10 dB, and the intensity stepped up by 1-5 dB. The ascending threshold was defined as the first intensity at which a positive response was elicited. Pure tone detection threshold was defined as the mean of the descending and ascending thresholds.
| |
RESULTS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Frequency direction discrimination
The scatter plot in Fig. 10 shows
F thresholds for frequency direction discrimination. Data
for each run and all subjects at all four center frequencies are
illustrated. There is a clear separation between case A1+
and the other 12 subjects.
|
Figure 11 shows Weber fractions
averaged across runs for each subject class. Table
1 summarizes the difference thresholds for case A1+, AC± cases, and normal subjects on
each of the four psychophysical tasks that employed an adaptive
procedure (frequency direction discrimination, frequency change
discrimination, duration discrimination, and VTI discrimination). In
case A1+, the mean Weber fraction of the five runs at each
Fc ranged from 12.0 to 15.8%, with a mean average across Fc's of
13.7% (a little over 2 semitones). In AC± cases, the mean
Weber fraction ranged from 0.94 to 2.3% [mean ± SE = 1.7 ± 0.38%]. In normal subjects, the Weber fraction ranged
from 0.47 to 1.0% (0.72 ± 0.09%). The Weber fractions of our
untrained normal controls were only slightly larger than those
previously reported in highly-practiced normals and similarly showed a
decrease in Weber fractions with increasing Fc's in this frequency
range (Moore 1973).
|
|
We carried out a two-way ANOVA to assess the effects of subject class
[case A1+ vs. AC± cases vs. normal
controls] and Fc (250 vs. 500 vs. 1,000 vs. 2,000 Hz) on Weber
fractions (the dependent variable). The random factor was runs. As
expected from inspection of Figs. 10 and 11, there was a highly
significant effect of subject class [F(2,212) = 434.01; P < 0.0001]. There was no significant effect
of Fc on Weber fractions, indicating that, on average, F
thresholds were proportional to Fc [F(3,212) = 2.74;
P = 0.044]. There was an interaction between subject
class and Fc, indicating that the pattern of performance across Fc's
varied across different subject classes [F(6,212) = 4.83; P = 0.0001].
Given the effect of subject class and the interaction between subject class and Fc, we carried out separate ANOVAs on each subject class using Fc as the factor, runs as the random factor, and Weber fractions as the dependent variable. There was no effect of Fc in case A1+, indicating no significant variation in Weber fractions across the four Fc's tested [F(3,16) = 0.40, P = 0.76]. However, there was a significant effect of Fc in both AC± cases [F(3,56) = 7.77, P = 0.0002] and normal controls [F(3,89) = 17.42, P < 0.0001]. Thus in Fig. 10, the progressive decline in mean Weber fractions seen with increasing Fc was significant for these two subject classes.
To assess the effect of interindividual variation on Weber fractions within the AC± cases and normal controls, we carried out separate ANOVAs on each population using subject as a factor and runs as the random factor. For AC± cases, there was a significant effect of subject [F(3,56) = 3.38, P = 0.024], indicating that Weber fractions varied from patient to patient. For normal controls, there was a highly significant effect of subject [F(7,89) = 21.13, P < 0.0001], indicating that Weber fractions varied from individual to individual. Apparently, this within-class variability did not compromise detection of across-class variability on the initial two-way ANOVA.
In view of the significant effect of subject class, we also compared Weber fractions between subject groups. Because the number of runs for case A1+ and AC± cases was small, and because Weber fractions were not evenly distributed along an interval or ratio scale, we used a nonparametric test, the Mann-Whitney U test, to compare Weber fractions between subject classes.
At each Fc, the Weber fraction for case A1+ was
significantly higher than the mean Weber fraction for AC±
cases [Fc = 250 Hz, U = 100, P = 0.0008; Fc = 500 Hz, U = 85, P = 0.001; Fc = 1,000 Hz, U = 100, P = 0.0008; Fc = 2,000 Hz,
U = 75, P = 0.0012]. The Weber
fraction for case A1+ was also significantly higher than the
mean Weber fraction for normal controls [Fc = 250 Hz, U = 160, P = 0.0004; Fc = 500 Hz,
U = 140, P = 0.0005; Fc = 1,000 Hz, U = 165, P = 0.0004; Fc = 2,000 Hz, U = 140, P = 0.0005]. AC± cases had significantly higher Weber fractions than
normal controls for all four Fc's at the P 
0.05 level of significance and for one Fc at the P 0.01 level [Fc = 250 Hz, U = 441, P = 0.0234; Fc = 500 Hz, U = 329, P = 0.0118; Fc = 1,000 Hz, U = 463, P = 0.0146; Fc = 2,000 Hz, U = 316, P = 0.0072].
For all Fc's combined, there were highly significant differences between the mean Weber fraction in case A1+ and the mean Weber fraction in AC± cases [U = 433, P < 0.0001]. Likewise, there were highly significant differences between the mean Weber fraction in case A1+ and the mean Weber fraction in normal controls [U = 2,420, P < 0.0001]. The mean Weber fraction in AC± cases was also significantly higher than that in normal controls [U = 2,512, P < 0.0001].
Frequency change discrimination
In case A1+, the Weber fraction at Fc = 1,000 Hz was 6.9% (a little over 1 semitone). In AC± cases, the Weber fraction range = 0.61-1.4% and the mean ± SE = 1.0 ± 0.2%. In normal controls, the Weber fraction range = 0.31-1.5% and the mean ± SE = 1.1 ± 0.2%.
An ANOVA was carried out using subject class as the factor, runs as the random factor, and Weber fractions as the dependent variable. There was a highly significant effect of subject class [F(2,44) = 55.42; P < 0.0001]. Case A1+ had significantly higher Weber fractions than AC± cases [U = 40, P = 0.008]. Case A1+ also had higher Weber fractions than normal controls [U = 90, P = 0.005]. The Weber fractions of AC± cases and normal controls did not differ significantly [U = 213.5, P = 0.93]. We combined Weber fraction data for the 12 participants without bilateral auditory cortex lesions and found a significant difference between their mean Weber fraction (1.05 ± 0.15%) and the Weber fraction in case A1+ [U = 132, P = 0.0041].
To analyze the variability within the AC± and normal control populations, ANOVAs were run separately on each with individual subjects as the factor, runs as the random factor, and Weber fractions as the dependent variable. For AC± cases, there was an effect of subject [F(3,10) = 4.65; P = 0.028], indicating significant variability among the four patients. For the normal controls, there was also an effect of subject [F(7,22) = 4.05; P = 0.005].
To consider how task requirements influenced F
thresholds, we compared Weber fractions at Fc = 1,000 Hz for
frequency change discrimination ("same""different" judgments)
and frequency direction discrimination ("lower""higher"
judgments; Fig. 12). In case
A1+, the mean Weber fraction for frequency direction
discrimination was larger than the mean Weber fraction for frequency
change discrimination by a factor of 1.8. In contrast, for the eight
normal controls, the mean Weber fraction for frequency direction
discrimination was smaller than that for frequency change
discrimination by a factor of 0.7. The ratio in case A1+ was
thus many SEs above the mean for normal controls (Table
2) [Student's t(7) = 12.4, P < 0.001]. There was also a significant
difference between case A1+ and AC± cases with
respect to the ratio of Weber fractions for frequency direction:change
discrimination [t(3) = 28, P < 0.001]. The latter ratio for AC± cases did not differ
significantly from that for normal subjects [t(10) = 1.24, P = 0.30]. This pattern of results indicates a
disproportionate elevation in Weber fractions for frequency direction
discrimination compared with frequency change discrimination in
case A1+. The combined ratio in AC± cases and
normal subjects (0.88) was significantly lower than that in case
A1+ [t(11) = 3.25, P < 0.0077].
|
|
Duration discrimination
In case A1+, the Weber fraction for duration discrimination was 34.4%. In AC± cases, the Weber fraction range = 11.8-32.5% (21.9 ± 5.0%). In normal controls, the Weber fraction range = 11.7-28.06% (19.3 ± 2.9%).
An ANOVA with subject class as the factor, runs as the random factor, and Weber fractions as the dependent variable revealed a significant effect of subject class [F(2,48) = 5.55; P = 0.0068]. Separate ANOVAs on the AC± and normal control populations revealed significant variability within both subject groups [respectively, F(3,14) = 10.21; P = 0.0008; F(7,22) = 3.02; P = 0.022].
In case A1+, the Weber fraction was significantly higher than that of AC± cases and normal controls (respectively, U = 48, P = 0.008; U = 85, P = 0.0048). There was no significant difference between the Weber fractions of AC± cases and normal controls (U = 349; P = 0.093). The Weber fraction in case A1+ was significantly higher than the mean Weber fraction of AC± cases and normal controls combined (combined mean = 20.2 ± 2.0; U = 133, P = 0.015).
This pattern of results is qualitatively similar to that observed for
F thresholds and indicates an elevation in difference thresholds for duration direction discrimination around a long center
duration of 500 ms in case A1+.
Vibrotactile intensity (VTI) discrimination
In case A1+, the Weber fraction in the hand ipsilateral to the lesion of sensorimotor cortex was 37.1%. In AC± cases (excluding MAH), the Weber fraction range in the dominant hand = 8.60-32.3% (22.5 ± 8.8%). In normal controls, the Weber fraction range in the dominant hand = 7.30-14.8% (10.0 ± 0.91%).
An ANOVA with subject class as the factor, runs as the random factor, and Weber fraction as the dependent variable revealed a highly significant effect of subject class [F(2,36) = 46.86; P < 0.0001]. A separate ANOVA on the AC± cases revealed significant variability from patient to patient [F(2,4) = 99.87; P = 0.0006]. Within the normal control group, there was no significant intersubject variability [F(7,20) = 2.03; P = 0.102].
The Weber fraction in case A1+ was significantly higher than
that in AC± cases (U = 30, P = 0.0062). The Weber fraction in case A1+
was significantly higher than that of normal controls (U = 140, P = 0.0012). There was no
significant difference in Weber fractions between AC± cases
and normal controls (U = 111, P = 0.223). The Weber fraction in case A1+ was significantly
higher than the mean Weber fraction in AC± cases and normal
controls combined [combined mean = 13.7 ± 2.86%;
U = 185, P = 0.0004]. This pattern of
results is qualitatively similar to those observed for F
thresholds and duration difference thresholds and indicates an
elevation in VTI difference thresholds in case A1+.
Visual line orientation match to sample
Case A1+ correctly answered 29 of the 30 items on the
Judgment of Line Orientation test. This performance is classified as "superior" according to scoring standards based on the distribution of corrected scores in Benton's normative sample (Benton et al. 1983). The mean scores for AC± cases (22.2) and
normal controls (22.8) were in the "average" range.
Performance ratios
For each adaptive task, a performance ratio was computed as WFA1+ divided by WFother, where WFA1+ is the Weber fraction for case A1+ and WFother the mean Weber fraction for AC± cases and normal controls combined. The results are illustrated in Fig. 13. Performance ratios for frequency direction discrimination (15.4) and frequency change discrimination (6.6) were far higher than the performance ratios for the duration direction discrimination (1.71), and VTI direction discrimination (2.59).
|
To further compare the results among auditory tasks for case A1+ and the other subjects, we computed performance ratios using the VTI difference threshold as a reference, since it was obtained in a different sensory modality using the same procedures (Table 2). The ratio of frequency direction discrimination to VTI discrimination was significantly higher in case A1+ than in AC± cases [t(2)]= 17.0, P < 0.01], normal controls [t(7) = 18.7, P < 0.001], and AC± cases and normals combined [t(10) = 25.7, P < 0.001]. A similar pattern of results was observed for the ratio of frequency change discrimination to VTI discrimination: A1+ versus AC±, t(2) = 6.50, P < 0.05; A1+ versus normal controls, t(7) = 4.00, P < 0.01; A1+ versus AC± and normal controls combined, t(10) = 4.89, P < 0.001. In contrast, the ratio of duration discrimination to VTI discrimination was not significantly different in case A1+ and AC± cases [t(2) = 0.89, P > 0.20]. Moreover, the ratio of duration discrimination to VTI discrimination was significantly lower in case A1+ than in normal controls [t(7) = 4.28, P < 0.01] and in AC± cases and normal controls combined [t(10) = 3.78, P < 0.01].
These results indicate that within the auditory modality, difference
thresholds for frequency direction discrimination and frequency change
discrimination were elevated more than difference thresholds for
duration direction discrimination centered at 500 ms. In addition, they
emphasize that the F threshold elevations in case
A1+ cannot be attributed to differences related to nonauditory task requirements of the adaptive method.
Pure-tone detection thresholds
Figure 14 plots pure-tone
detection thresholds at each Fc for case A1+ and normal and
patient controls combined. "Reference" thresholds correspond to 25 dB hearing level (HL), that is, 25 dB greater than the zero reference
level for mean audible pressure recommended by the International
Standards Organization (1975). For case
A1+ as well as controls, detection thresholds were within normal
limits for both the left and right ears by clinical standards: the mean
of the thresholds at 500, 1,000, and 2,000 Hz was <25 dB HL
(AAO-ACO 1979; Yantis 1994). Still, left
ear detection thresholds for case A1+ were more than 2 SEs
higher than those of controls at all four frequencies tested, raising
the possibility of mild hearing loss contralateral to his complete or
near-complete right auditory cortex lesion.
|
| |
DISCUSSION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
To summarize the main results: bilateral auditory cortex lesions involving A1 were associated with elevations in frequency difference thresholds. For frequency direction discrimination (lower-higher pitch judgments), Weber fractions associated with a 79.4% probability of correct response averaged 14% at center frequencies between 250 and 2,000 Hz. These Weber fractions were about eightfold higher than those measured in neurological patients without bilateral lesions of auditory cortex, and they were almost 20-fold higher than Weber fractions measured in normal controls. For frequency change discrimination (same-different pitch judgments), Weber fractions averaged 7% at 1,000 Hz, about sevenfold higher than the Weber fractions observed in the other groups. The pattern of results on other auditory, tactile, and visual tasks indicates that the observed elevations in frequency difference thresholds cannot be attributed to nonspecific lesion effects on task performance or on hearing in general.
We interpret the present findings as evidence that fine-grained frequency processing at the perceptual level relies on the integrity of finely tuned neurons in A1 and surrounding areas. This interpretation is based on the anatomical localization of the lesions in case A1+ and current knowledge about anatomical-physiological correlates in humans and nonhuman primates.
Anatomical considerations
Deficits in frequency processing in case A1+ were associated with chronic, bilateral hemispheric infarcts involving all of the transverse gyrus of Heschl (TG) in the right hemisphere; all or almost all of TG in the left hemisphere; the posterior one-fifth of the left superior temporal gyrus (STG) and adjacent inferior parietal lobule; all or almost all of right STG; and portions of right frontal, parietal, medial temporal, and inferior temporal cortex. None of the other four neurological patients, including one with multiple cerebral contusions and right TG and STG damage (case MAH), had a single experimental run in which the frequency difference threshold was higher than the 23 thresholds measured in case A1+.
TG runs across the middle third of the superior surface of STG deep
within the lateral (Sylvian) fissure. Some individuals possess two or
more TGs in one or the other hemisphere (Campain and Minckler
1976; Musiek and Reeves 1990), and most have a
greater volume of white matter on the left than on the right
(Penhune et al. 1996). TG receives the geniculotemporal
radiation and contains the highest concentration of neurons and
myelinated fibers in all of superior temporal cortex (Brodmann
1909; Campbell 1905; Economo
1929; Galaburda and Sanides 1980;
Rademacher et al. 1993). These cytoarchitectural and
connectional features, combined with electrophysiological data on
minimum response latencies measured with intracortical electrodes
(Liegois-Chauvel et al. 1991, 1994, 1995), cortical
surface electrodes (Celesia 1976; Celesia and Puletti 1969; Celesia et al. 1968), subdural
strip electrodes (Lee et al. 1984), and scalp electrodes
(Scherg et al. 1989), indicate that all of A1 resides
within TG. However, TG and A1 are not co-extensive: minor portions of
TG (and all supernumerary TGs) contain non-koniocortical fields
(Rademacher et al. 1993).
Whether all of STG may be considered auditory association cortex and
whether auditory association cortex extends beyond the boundaries of
STG remain uncertain. Cortical depth and surface electrode recordings
of click-evoked potentials in humans suggest that all or almost all of
auditory cortex lies within the posterior two-thirds of superior
temporal cortex (Celesia 1976; Celesia and
Puletti 1969; Celesia et al. 1968; Howard
et al. 2000; Lee et al. 1984). Likewise, maximal
increases in glucose metabolism have been observed in posterior
superior temporal cortex during binaural stimulation with combined
speech and music (Mazziotta et al. 1982). Oxygen
utilization increases from the posterior extreme to almost the anterior
extreme of STG during binaural stimulation with speech and white noise
(Binder et al. 1994), and blood flow
increases in the middle temporal gyrus as well as STG during binaural
stimulation with speech and nonspeech sounds (Belin et al.
2000). In macaques, free-field stimulation with a potpourri of
species-specific vocalizations, human voices, and environmental sounds
increases glucose metabolism throughout superior temporal cortex, from
the posterior end of the lateral fissure anteriorly to the temporal
pole (Poremba et al. 1999).
It is likely that residual auditory functions in case A1+
are subserved by spared auditory association cortex residing in left
anterior STG and by subcortical structures. Although inputs to
the left auditory association cortex from damaged left A1 and damaged right auditory cortex were interrupted, auditory information may have reached the spared regions via ascending pathways from the
dorsal and medial nuclei of the medial geniculate complex, the
pulvinar, the posterior nucleus, the suprageniculate nucleus, and/or
the limitans nuclei. These parallel ascending inputs to auditory
association cortex are well-documented in macaques (Akert et al.
1959; Burton and Jones 1976; Chow
1950; Clark 1936; Clark and Northfield
1937; Hackett et al. 1998; Hashikawa et
al. 1992, 1995; Jones and Burton 1976;
Locke 1960; Mesulam and Pandya 1973; Molinari et al. 1995; Morel et al. 1993;
Pandya and Sanides 1973; Pandya et al.
1994; Poliak 1932; Rauschecker et al.
1997; Siqueira 1965; Walker
1937). In addition, the ventral nucleus of the medial geniculate complex, the main source of afferent input to A1, provides weak inputs to auditory association cortex surrounding A1.
Physiological considerations
At present, little is known about frequency processing at the
cellular level in human auditory cortex. However, a wealth of knowledge
is available in the experimental animal literature (for reviews, see
Clarey et al. 1992; de Ribaupierre
1997; Kaas et al. 1999). For phylogenetic,
anatomical, and functional reasons, knowledge about the single-unit
physiology of macaque auditory cortex may be especially relevant to
understanding frequency processing in humans.
Like humans, macaques house koniocortex within the middle third of the
superior surface of superior temporal cortex deep within the lateral
fissure (Bonin and Bailey 1947; Sanides
1972; Walker 1937). While it is not customary to
designate TG in the macaque, several authors have identified a small
elevation along the superior surface of STG that stains heavily for
Nissl, myelin, cytochrome oxidase, and parvalbumin (Bonin and
Bailey 1947; Jones et al. 1995; Merzenich
and Brugge 1973; Pandya and Sanides 1973;
Tramo 1998). Bonin and Bailey (1947)
remarked that the superior surface of STG in Macaca mulatta
"bears a very slight elevation, recognizable when one is acquainted
with Heschl's [transverse] gyrus in man, but easily overlooked
otherwise" (p. 17). In M. fuscata, Jones et al.
(1995) identified a similarly situated "annectant" gyrus as
the gross anatomical correlate of koniocortex and high-density parvalbumin staining. These observations indicate that macaques have a
rudimentary gross morphological homologue of human TG which houses
histologically-defined A1. For heuristic purposes, it might be
advantageous to refer to this structure as the "transverse gyrus"
in macaques as well as humans.
Microelectrode mapping studies of A1 and other "core area" fields
in anesthetized macaques have demonstrated frequency-selective neurons
that are topographically organized with respect to "best frequency"the pure tone frequency associated with maximal firing at
a given intensity (Kosaki et al. 1997; Merzenich
and Brugge 1973; Morel et al. 1993;
Pfingst and O'Connor 1981; Rauschecker et al.
1997). In addition, several fields in the "belt area" of auditory association cortex surrounding A1 contain neurons that are
frequency-selective and topographically organized. In general, neurons
in these belt fields are less sharply tuned than neurons in core fields
(Kosaki et al. 1997; Rauschecker 1995,
1997). In alert rhesus monkeys, 500-ms and 50-ms pure tones
presented at frequencies and intensities similar to those used in the
present experiment evoke responses in individual core area neurons over bandwidths ranging from approximately 0.02 octaves (about a quarter of
a semitone) to 5 octaves (60 semitones); the majority of cells respond
over 0.1 octaves (about a semitone) or more (Recanzone et al.
2000; Tramo 1998; Tramo et al.
1999). Macaques can discriminate 50-ms pure tones differing in
frequency by about 3% (half a semitone) at 500 Hz and 60 dB SPL
(Sinnott and Brown 1993). Thus there may exist
individual neurons with frequency selectivity sufficient to support
frequency discrimination at the difference thresholds observed in our
control subjects. Still, the overwhelming majority of core area neurons
in macaques respond over bandwidths greater than a semitone. The
results of microelectrode recordings in anesthetized owl monkeys who
had been trained on a frequency-change discrimination task indicate
that perceptual acuity correlates with the number of A1 neurons tuned
to frequencies used in the task (Recanzone et al. 1993).
In human auditory cortex, extracranial recordings of local field
potentials and functional imaging techniques have provided evidence of
frequency selectivity. Butler (1968) showed that the peak-to-peak amplitude of the N100 potential recorded at the vertex is
sensitive to the frequency difference between successive sinusoidal tones (monaural stimulation at 70 dB SL, duration = 600 ms,
ISI = 4.4 s, probability of occurrence standard tone:deviant
tone = 3:1). The larger the frequency difference between tones,
the larger the amplitude. Butler attributed the correlation between frequency difference and N100 amplitude to frequency-dependent refractoriness and differences in the amount of overlap between frequency-selective neuron populations. The effect of frequency difference on N100 amplitude is independent of the harmonic
relationship between the standard and deviant tones (Kussmaul et
al. 1992). The amplitude and latency of tone-evoked magnetic
potentials peaking around 100 ms poststimulus (N100m) vary
monotonically with the width of surrounding notched noise (Sams
and Salmelin 1994). N100 and N100m amplitude differences can be
elicited only at or above frequency difference thresholds for
same-different discriminations (Csépe et al. 1992;
Sams et al. 1985). These findings suggest that neural
mechanisms underlying the sensitivity of N100/N100m amplitude to
frequency differences are relevant to frequency resolution at the
perceptual level.
The location of frequency-selective neurons generating the N100 and
other potentials has been extensively investigated using dipole source
analyses of scalp-recorded electrical fields (Giard et al.
1994; Scherg and Cramon 1985; Scherg et
al. 1989) and magnetic fields (Gallen et al.
1993; Hari et al. 1984; Kuriki et al.
1995; Lavikainen et al. 1995; Nakasato et
al. 1995; Reite et al. 1994; Tiitinen et
al. 1993). To summarize the main results: most or all neural
generators of lateral scalp potentials peaking between 18 and 200 ms
after acoustic stimulation lie within TG or close by along the superior
surface of STG, and some potentials peaking between 150 and 300 ms may
originate within the lateral surface of STG, dorsolateral prefrontal
cortex, and parietal cortex. In general, the effects of focal lesions
on extracranial electrical field potentials (Aaltonen et al.
1993; Graham et al. 1980; Kileny et al.
1987; Knight et al. 1988; Kraus et al.
1982; Praamstra et al. 1993) and magnetic field
potentials (Makela 1993; Makela and Hari
1992) support the notion that neurons in TG and adjacent STG
are the major contributors to evoked potentials peaking after about 13 ms and before 200 ms poststimulus onset. In our case A1+,
middle- and long-latency auditory-evoked potentials are grossly abnormal (Musiek et al. 1994).
Evidence that neurons within human auditory cortex are topographically
organized with respect to frequency selectivity was first obtained
using magnetoencephalography (MEG). Some authors reported a low- to
high-frequency gradient of the N100m potential along the anterior to
posterior axis of posterior superior temporal cortex (Elberling
et al. 1982; Pantev et al. 1988). Others
reported a low- to high-frequency gradient along its lateral to medial axis (Huotilainen et al. 1995; Romani et al.
1982a,b; Yamamoto et al. 1992), consistent with
early claims based on lesion effects (Economo 1929). One
octave is represented by approximately 2-3 mm of cortical surface.
Analyses of scalp-recorded electrical potentials also suggest that
posterior superior temporal cortex contains a topographic organization
of frequency-selective neurons, with the low-frequency representation
anteriorly and slightly lateral to the high-frequency representation
(Scherg et al. 1989).
Functional imaging studies have provided converging evidence of
frequency selectivity and a low- to high-,
anterolateral-to-posteromedial frequency representation in and around
TG, albeit with considerable intersubject variability (Lauter et
al. 1985; Strainer et al. 1997; Wessinger
et al. 1997). Conspicuously missing from some human studies is
the multiplicity of topographically organized fields one would expect
on the basis of microelectrode mapping experiments in macaques. Over
the past several years, several MEG studies have reported evidence for
two frequency representations in posterior superior temporal cortex
(Cansino et al. 1994; Pantev et al. 1995;
Tiitinen et al. 1993; Verkindt et al.
1995). Recent functional MRI results obtained by
Talavage and colleagues (2000) via stimulation with
high- and low-pass filtered music and band-limited, amplitude-modulated
noise show four separate high-frequency regions and four separate
low-frequency regions. Among the most significant pieces of evidence
that human auditory cortex contains frequency-selective neurons is the
microelectrode study of an epilepsy patient by Howard and
colleagues (1996). Of 26 neurons encountered in three penetration sites spanning 8 mm of TG, 19 were judged to be sharply tuned.
In summary, on the basis of the gross topography of the lesions observed in case A1+, physiological-anatomical correlates in macaques, and electrophysiological data in humans, case A1+ likely had complete loss of frequency-selective neurons in right A1 and surrounding areas, complete or near-complete loss in left A1, and sparing of some frequency-selective neurons in left belt areas anterior and adjacent to A1. Thus the fact that his ability to resolve frequencies was coarsened but not abolished may be attributable to preservation of neurons in the belt areas of left STG.
Functional effects of auditory cortex lesions in primates
Perhaps the most influential experiments contributing to the
belief that primate auditory cortex plays no role in frequency discrimination were carried out by Evarts (1952a,b). Two
young M. mulatta learned an operant conditioning task that
required differential responses to 350 Hz ("go") and 3500 Hz
("no-go") tones. Correct responses were rewarded with food, and
incorrect responses were punished by electric shock. The intensity of
the tones was approximately 50 dB SL re: human detection thresholds. After they reached the response criterion of 80% correct, all but a
small portion of anterior superior temporal cortex was aspirated in the
left and right hemispheres. Part of the insula was also taken. The
extent of the lesions was confirmed postmortem, and near-complete
retrograde degeneration of the parvocellular portion of the medial
geniculate nucleus was observed. Ten days post-ablation, both monkeys
performed well in their initial training session and quickly
re-achieved the preoperative performance criterion. Two additional
macaques were trained an unspecified amount of time after surgeries
that had been carried out for a previous experiment. One macaque had
near-complete bilateral lesions of superior temporal cortex and one had
a near-complete lesion on the right and a partial lesion on the left
that spared a large part of the superior surface of posterior superior
temporal cortex. Both of these previously lesioned macaques learned the
task as quickly as the first two macaques had learned it before their surgeries. Evarts interpreted these findings as evidence that near-complete bilateral lesions of auditory cortex do not impair the
initial learning or post-lesion retention of frequency discrimination. He found the results "difficult to reconcile with the strict
tonotopic organization" found in primate auditory cortex and
speculated that small remnants of remaining auditory cortex were "of
great functional importance." Inasmuch as Evarts' task incorporated a frequency difference threshold corresponding to a Weber fraction of
164%, these experiments simply did not assess the role of auditory cortex in fine-grained frequency discrimination.
On the basis of animal lesion experiments on pure tone frequency
discrimination and his demonstration of impaired virtual pitch
perception following unilateral right anterior temporal lobectomy in
patients with intractable epilepsy, Zatorre (1988) argued that "simple frequency discrimination (i.e., with pure tones,
or with complex tones when the fundamental [frequency] is present) is
not permanently disrupted even by large bilateral lesions of auditory
cortex." However, patients who committed more than 25% errors on a
control task, in which complex tone fundamental frequencies (F0's)
were present, were excluded from the missing F0 task. In addition, the
F0 differences tested by the method of constant stimuli (Weber
fraction = 40%) were well above normal threshold (cf.
Patterson 1973). Using a dichotic-diotic match-to-sample recognition task, Sidtis and Volpe (1988) found impaired
complex pitch perception with F0 present in a population of right
hemisphere stroke patients; they did not use pure tone stimuli nor did
they report which gyri were lesioned. Robin et al.
(1990) subsequently demonstrated elevations in the F0
difference threshold averaged across five stroke patients with lesions
involving right TG and/or STG (mean Weber fraction approximately 14%);
five patients with similar left-sided lesions had normal thresholds
(mean Weber fraction, less
, AC± cases. 
, normal
subjects.
, case
A1+; 
, reference level corresponding to the upper limits of
normal by clinical standards (AAO-ACO 1979