Human Perception &
Assessing adults' abilities to categorize and discriminate speech sounds can shed light on the mechanisms of speech perception. Listeners in these studies typically hear synthesized or natural speech sounds or complex non-speech sounds over headphones and respond to them using a handheld response box. Generally, these experiments last less than an hour, but learning experiments may take multiple sessions over many days. In other experiments, subjects play video games or train in learning the sounds of a new language.
Eyetracking can reveal auditory processing as it unfolds across time. We use the Tobii (
www.tobii.com) eye tracking system for studies with adult and infant participants to investigate the fine-grain time course of spoken language processing and auditory learning.
Another important tool is computational modeling. With the wealth of fine-grain data provided by the animal learning experiments, it is even possible to yoke performance of computational models with data from individual animals as they learn from their experience with speech. Computational models also inform our investigations of human speech perception, for example the effects of prior knowledge of a native language on speech categorization.
Magnetoencephalography (MEG) is a non-invasive neurophysiological method that measures the magnetic fields generated by neural activity of the brain. Magnetic fields are found whenever there is charge flow (current), whether in a wire or neuronal tissue. The magnetic field passes unaffected through brain tissues and the skull, so it can be recorded outside the head by the MEG system. The spatial distributions of the magnetic fields localize the sources of the activity within the brain, and the locations of the sources may be superimposed on anatomical images, such as MRI, to provide information about both the structure and function of the brain.
MEG offers a completely non-invasive, direct measure of brain function and very high temporal resolution. Events with time scales of milliseconds can be resolved with MEG, making it ideal for studies of auditory cognition and speech processing.
For more information, see the Center for Image-Guided Neurosurgery’s website (
or the MEG wiki (
Electroencephalography (EEG) is a non-invasive neurophysiological measure of the electrical activity of the brain. EEG is measured by recording from electrodes placed on the scalp.
We use a Neuroscan Synamps2 system with 64 electrode Neuroscan QuikCaps. Data is collected via Neuroscan Scan software and experiments are run using E-prime.
For more information about EEG at Carnegie Mellon, see the EEG Laboratory website ( http://memory.psy.cmu.edu/eeg_index.html ).
For more general EEG information, see the EEG wiki (
For helpful lab information, see the guide for CMU EEG researchers (
Perception & Learning Paradigms
Using animals in speech perception studies allows us to examine questions that are impossible, unethical, or extremely difficult to answer with human listeners. These experiments take advantage of the fact that animals' experience with speech can be easily manipulated experimentally. By observing the influence that this experience has on speech perception, we can learn about aspects of the development of speech perception and language acquisition.
The species that we've used in this work include the Chinchilla, the Japanese Quail, and the European Starling. At CMU, we use the gerbil as a model of speech perception. For more information about the use of animals in speech research,