This experiment uses functional magnetic resonance imaging (fMRI) to examine brain activity in response to speech. Because the inherent loudness of the scanner noise may be a confounding factor, it has been incorporated into the design of a speech recognition in noise task. A goal of the experiment is to identify areas of brain activation associated with processing acoustic/phonetic information within words with and without sentence context. Structural and functional images are acquired for 15 coronal slices from normal-hearing subjects. Two listening conditions are included: passive and active listening. In both conditions, two types of sentences are presented (Kalikow et al., 1977; JASA 61, 1337-1351). Low-context sentences provide no contextual information so that the final word is not predictable from the other words in the sentence. High-context sentences include contextual information so that the same final word is highly predictable, for example:
High Context Low Context
Stir your coffee with a spoon. Bob could have known about the spoon.
The natives built a wooden hut. He should know about the hut.
High Context Low Context
Stir your coffee with a spoon. Bob could have known about the spoon.
The natives built a wooden hut. He should know about the hut.
Speech is presented to subjects' right ears through pneumatic non-metallic noise-reduction earphones during two six-minute runs. In the first six-minute run, "passive listening," subjects are instructed to concentrate on their breathing. In the second six-minute run, "active listening," subjects are instructed to listen to the sentences and silently determine the last word. Within each six-minute run, one-minute blocks of speech and noise are alternated with thirty seconds of noise only ("rest") as shown below:
The imaging modality used is Blood Oxygenation Level Dependent, or BOLD fMRI, which is based on the change in magnetic signal emitted by hemoglobin when it is carrying oxygen, revealing areas of the brain which are more active (or in high demand). With a scanner capable of recording very fast images (here, 5/sec), brain activity at rest and during a specific task may be compared. This information is being used in combination with MRI which provides information about the structure of the brain, enabling simultaneous studies of structure and function in individual subjects. The technology is magnetic rather than radioactive, which allows multiple scans at no risk to the subject.
For three representative subjects, images are shown of the areas activated during "active listening" to high-context sentences (in color, p<0.001) merged onto one of 15 serial coronal structural MRI scans. Significant activation is shown in both the left and right superior temporal gyri.
This work was supported by NIH and Picker International, Inc. Many helpful contributions from Jayne Ahlstrom and Donna Roberts are gratefully acknowledged.
