Article Figures & Data
Figures
- Fig. 1.
Spectrograms of sample experimental stimuli.a, The sentence “The poster was advertising a concert to be held next week” in undistorted normal form. b, Segmented with 500 msec noise bursts alternating with 200 msec of undistorted speech. c, Vocoded speech shown with four frequency bands. d, Speech in noise with a background of speech-spectrum noise at a signal-to-noise ratio of −1 dB.
- Fig. 2.
Word-report scores from the pilot study and subject ratings during scanning for the 11 different kinds of stimuli. Each point represents the average (and SE) for 18 subjects in the pilot study (percentage of words reported correctly) or eight subjects in the fMRI study (fMRI ratings). Correlation between word report and ratings was r = 0.98;p < 0.001.
- Fig. 3.
Details of the scanning procedure. A sparse imaging technique was used (see Materials and Methods) in which a single stimulus item was presented in the silent periods between scans. A tone pip after each sentence cued the subject's intelligibility judgment (button press). Timings of sentence onset and offset and tone cue relative to scan onset were jittered across trials.
- Fig. 4.
a–c, Predicted BOLD signal for three contrasts: linear correlation between BOLD signal and intelligibility (a), a differential response to the three forms of distortion (b), and an elevated response to all forms of distorted speech (c), compared with clear speech and signal-correlated noise. Open triangles, Signal-correlated noise; circles, segmented speech;x symbols, vocoded speech; squares, speech in noise; filled triangles, normal speech.
- Fig. 5.
Areas showing a significant linear response to increasing intelligibility. Activations are shown superimposed on the mean EPI image across subjects and thresholded at p< 0.001, uncorrected for multiple comparisons. a, Sagittal sections depicting areas in which activation was observed for signal-correlated noise relative to rest (pink-blue scale) and areas in which activation correlates with report score (intelligibility response; yellow-red scale).b, Axial sections through Heschl's gyrus. Activation to sound is predominant on Heschl's gyrus, whereas correlation with intelligibility is observed posteriorly, inferiorly, and laterally to Heschl's gyrus. c, The intelligibility response pattern shown in a but with a mask to show voxels that are sensitive to the acoustic properties of the stimulus (dependent on distortion type). Because the identification of form-independent regions depends on the absence of any reliable difference between distortions, we cannot precisely localize the transition between form-dependent and form-independent regions and so present the mask ingraded color scale. The 95% contour of the mask (form dependence; p < 0.05) corresponds to the boundary between blue and green; the 99.9% contour (form dependence at p < 0.001) corresponds to the boundary between orange and red. Activation foci related to intelligibility and common to all forms of distortion (form independent) are shown in blue and listed in Table 1 (top). Arrow indicates the approximate location of the form-independent left anterior temporal lobe voxel from which data are plotted in f. d, Axial sections through Heschl's gyrus. Extending posteriorly and laterally from primary auditory cortex into belt and parabelt cortices, form-dependent (graded color) intelligibility responses give way to form-independent (blue) responses.e, Coronal section (y = −20) depicting a form-independent intelligibility response in left hippocampus. f, The graph shows the size of the response (percentage of signal change from the mean) for a form-independent voxel (→) in the left anterior temporal lobe (−58, −2, −24) against word-report score for the different listening conditions. Open triangle, Signal-correlated noise;circles, segmented speech; x symbols, vocoded speech; squares, speech in noise; filled triangle, normal speech. Error bars indicate SEM across subjects.
- Fig. 6.
BOLD signal in peak voxels showing a form-dependent response. Graphs show mean percentage signal change for each distortion type compared with the mean response to signal-correlated noise and normal speech. Error bars indicate SEM after between-subject variability has been removed, which is appropriate for repeated-measures comparisons (cf. Loftus and Masson, 1994). Braces show significance of pairedt tests comparing the three distortion types (*p < 0.05; **p < 0.01; ***p < 0.001). a, Form-dependent response in left superior temporal gyrus (−52, −28, 6).b, Right superior temporal gyrus (66, −16, 0).c, Left inferior frontal gyrus (−42, 20, −6).
- Fig. 7.
Areas showing an elevated response to distortion. Activations are shown superimposed on the mean EPI image across subjects and thresholded at p < 0.001, uncorrected for multiple comparisons. a, Sagittal sections depicting areas in which activation is significantly elevated for distorted speech relative to clear speech and signal-correlated noise and covaried for intelligibility (distortion-elevated response).b, The activation pattern shown in a but with a mask to show voxels that are sensitive to the acoustic properties of the stimulus (showing a distortion-elevated response that is form dependent). The mask is shown in graded color scale as in Figure 5. Areas exhibiting distortion-elevated activation that is form independent are shown in blueand listed in Table 1 (bottom). Arrow indicates the approximate location of the form-independent left inferior frontal voxel from which data are plotted in d. c, Axial sections through Heschl's gyrus showing distortion-elevated responses in temporal and frontal lobe regions. Form-dependent (graded color) and form-independent (blue) responses are observed in temporal and frontal lobe regions. d, Thegraph shows the response (percentage of signal change) of a voxel (→) in left inferior gyrus (−56, 16, −6) against word-report score. Error bars indicate SEM across subjects.
Tables
Region Coordinates Z-score x y z Form-independent correlates of activity with intelligibility R MTG 64 −8 −16 5.50* L MTG −60 −34 −2 5.20* L Hippocampus −22 −20 −12 4.02* L post MTG −62 −56 6 3.79* L angular gyrus −54 −60 22 3.71* L frontal operculum −58 16 −2 3.35* L SFS −10 56 30 3.26* L Precuneus −8 −50 30 3.26* Form-independent activity increases for distorted speech relative to normal speech and signal-correlated noise Vathal −12 −6 10 4.25* L frontal operculum −48 14 −6 4.22* L MTG −66 −20 −4 3.72* L posterior STP −52 −46 16 3.54* L orbitofrontal −32 52 −4 3.40 R intraparietal sulcus 40 −46 38 3.39 L MFG (premotor) −40 2 44 3.35 Cingulate gyrus −2 16 46 3.29 L orbitofrontal −36 42 −14 3.23 L anterior STG −58 −8 −6 3.20 L posterior STG −56 −54 8 3.15 R MFG (premotor) 46 8 32 3.14 L substantia nigra −6 −20 −16 3.12 We present coordinates of activation foci together with Z-scores and an estimate of location relative to gross anatomy for each contrast of interest. MFG, Middle frontal gyrus; MTG, middle temporal gyrus; SFS, superior frontal sulcus; STP, superior temporal planum; STG, superior temporal gyrus; Vathal, ventral anterior thalamic nucleus; R, right; L, left. All peak voxels exceeding p < 0.001 are reported. Voxels marked with an asterisk reach whole-brain FDR correction at p = 0.05. Note that, because the FDR correction is an adaptive procedure, statistical thresholds are at different values for the two contrasts.
