Single Trial Reconstruction of Acoustic Stimuli from Multielectrode Cortical Recordings of Awake Ferrets
Zusammenfassung der Projektergebnisse
Our perception of the world is not uniquely determined by the outside stimulus, but can depend on our expectations and the stimulus history. The goal of the present research project was to understand how the brain's representation of the world differs between different percepts of the same stimulus. We investigated this question in the auditory domain, using ambiguous stimuli and decoding methods applied to the neural population response to estimate the perceived stimulus. We presented a set of ambiguous stimuli, socalled Shepard pairs, to awake ferrets, while recording from their auditory cortex using chronic electrode arrays. A Shepard pair is a sequence of two Shepard tones. Each Shepard tone is a complex acoustic stimulus, composed of octave spaced pure-tones. Shepard tones separated by less than 6 semitones are perceived as clearly ascending or descending in pitch. If separated by 6 semitones, the Shepard pair becomes ambiguous - giving rise to the socalled tritone paradox – where different listeners have convincing but different percepts of the same stimulus. In our experimental paradigm, the Shepard pair was preceded by a sequence of Shepard tones - a socalled bias sequence - which has been shown in human experiments to reliably influence the perception of the Shepard pair. A bias sequence within a half-octave above the first Shepard tone leads to an ascending percept, and vice versa below. Prior to our study, it was assumed that the bias sequence should act to reduce the distance between the Shepard pair, thus leading to the perception of a step connecting the Shepard pair along the region where the bias sequence was presented. We applied a range of decoding methods to the neural population data and find that the opposite is true in the primary auditory cortex: the Shepard tones move apart in pitch. We verified this percept in a modified human psychophysical experiment, obtaining the same repulsive shifts as in the neural data. Using a combination of single-cell analysis and human magnetoencephalography (MEG), we demonstrate that the biasing sequence suppresses the neural representation locally and this suppression is predictive of the perceptual decision. We propose that the suppression creates a local imbalance in the population of pitch-change selective cells located at the edge of the bias sequence. This imbalance underlies our percepts of ascending and descending. Both model and neural data appear to be consistent with this hypothesis. In summary, we find that the percept of an ambiguous stimulus is shaped by the local suppression state of the auditory cortex. More generally, we propose that the suppression state could serve as a short-term memory store of the immediate stimulus history. Since the suppression was found to last for multiple seconds in human MEG data, we propose that the suppression state can be actively sustained, if required for a perceptual task. A necessary step for the research described above was to record from neural populations in behaving, potentially freely moving animals. As part of the research project, I therefore developed two enabling technologies. First, a recording system - MANTA - was developed, which allows state-of-the-art neural recordings with 100s of channels and provides ample flexibility to adapt to new paradigms/analyses. While this system was only intended as a local tool, it has turned into a versatile, feature-rich research tool in use in multiple labs, due to its low system costs and high degree of flexibility (a publication and online hosting of the system is forthcoming). Second, a positioning system for electrode arrays was developed, which increases the yield and eases the operation of a chronic implant. Similarly, originally only intended as a local tool, the EDDS ArrayDrive has become a tool in more widespread use, and will be manufactured by an external partner.
Projektbezogene Publikationen (Auswahl)
- (2012) Adaptive mechanisms underlying the perceptual bias of ambiguous stimuli in auditory cortex. ARO
Englitz et al.
- (2012) Investigating the Perceptual Mechanisms of Ambiguous Stimuli in Auditory Cortex. Auditory Cortex Meeting
Akram, Englitz, et al.
- (2012) Learning the Stimulus Space Geometry from Auditory Neural Population Responses. Auditory Cortex Meeting
Englitz & Shamma
- (2012) Putting the Tritone Paradox into Context: Insights from Neural Population Decoding & Human Psychophysics, in Basic Aspects of Hearing: Physiology and Perception , edited by B.C.J. Moore, R.D. Patterson, I. Winter, R.P. Carlyon, and H.E. Gockel (Springer, New York). International Symposium on Hearing 2012
Englitz et al.
- (2012) The Array Drive : Optimizing the Yield and Flexibility of Chronic, Multielectrode Array Recordings. SfN
Englitz et al.
