The psychoacoustic results were used to choose three signal-to-noise ratios for each listener to explore the effect of SNR on an fMRI correlate of target audibility. The first S/N (SNR₁) was selected so that the target signal was audible for both noise conditions and both interaural phase conditions. This S/N was set to 6 dB above the masked threshold for the S ₀ N ₀-UN configuration, i.e., the most effective masker condition. The second S/N (SNR₂) was selected so that the target signal in the S ₀ N ₀ configuration was not audible for the uncorrelated noise but audible for the comodulated noise (“halfway” between S ₀ N ₀-UN and S ₀ N ₀-CM thresholds). This S/N was picked to search for a correlate of the CMR effect alone, irrespective of binaural cues. The third S/N (SNR₃) was selected so that the target was inaudible in both S ₀ N ₀ configurations but audible for both S _π N ₀ configurations (“halfway” between S ₀ N ₀-CM and S _π N ₀-UN thresholds), to identify a correlate of the BMLD effect, largely independent of the effect of the modulation of the masker signal.

Thirteen different sound conditions in total were used during the fMRI recording for each of the 21 listeners. These included the conditions S ₀ N ₀-UN at SNR₁, SNR₂, and SNR₃; S ₀ N ₀-CM at SNR₁, SNR₂, and SNR₃; S _π N ₀-UN at SNR₁ and SNR₃; S _π N ₀-CM at SNR₁ and SNR₃; N ₀-UN and N ₀-CM alone with no target signal; and finally a condition with no signal presentation as a baseline control. The selected signal-to-noise ratios are shown as additional symbols in Fig. 48.2, to illustrate the relationship of the fMRI design relative to the results from the preceding detection experiments. As mentioned above, signal-to-noise ratios were chosen individually based on the results for the masked thresholds determined before. A subgroup of 16 listeners showed largely homogeneous results in their detection thresholds. The mean selected S/N for these listeners were SNR₁  =  0.1 (1.1) dB, SNR₂  =  −9.2 (1.1) dB, and SNR₃  =  −14.7 (1.8) dB. Five more listeners showed bigger deviations from this general trend, mainly with respect to the size of the CMR effect, which was very small for some of the participants. They were still included the fMRI study, with correspondingly different signal-to-noise ratios for the experimental conditions (not listed in detail here).

Each stimulus presentation was a sequence of twelve 400-ms noise bursts together with the respective target at 1 kHz and jittered by ±24 Hz from burst to burst. After each presentation of a stimulus block (i.e., during the respective scan acquisition), the participants indicated by button presses whether they heard the slightly varying tone in the masking noise. This task was introduced to maintain the participants’ attention. Stimuli were played via MR-compatible insert earphones (Sensimetrics S14) using a fixed masker level of 60 dB SPL. Throughout the full fMRI experiment, each condition was repeated 36 times, giving a total of 468 brain images. The order of conditions was fully randomized, and the complete session was divided into three runs with 156 scans each. Functional MRI data were acquired using a Siemens Sonata 1.5 T MRI system. For the functional data, 21 axial EPI slices (in-plane resolution 3 × 3 mm; thickness 5 mm, echo time TE  =  63 ms to maximize BOLD contrast) were acquired covering most of the cortex, including the whole of the temporal lobes and frontal regions. Sparse imaging with clustered volume acquisition was used (Hall et al. 1999). The total volume acquisition time TA was 2.7 s. On each trial, there was a 5-s stimulus interval followed by the 2.7-s scanning interval, making a total repetition time of TR  =  7.7 s. A T ₁-weighted high-resolution anatomical image (176 sagittal slices, TR  =  2.11 s, TE  =  4.38 ms) was also collected for each subject.