Skip to main content

Identifying Oscillatory Signatures of Predictive Coding in Hierarchical Auditory Networks with MEG

Periodic Reporting for period 1 - PredAMEG (Identifying Oscillatory Signatures of Predictive Coding in Hierarchical Auditory Networks with MEG)

Reporting period: 2017-01-01 to 2018-12-31

Elucidating the mechanisms involved in the communication of predictive signals across brain regions is currently one of the fundamental questions in cognitive neuroscience. This problem is not only relevant for understanding normal brain functioning but also potentially important for identifying core deficits underlying neurological and psychiatric disorders such as schizophrenia.
The PredAMEG project aimed to address these questions through the investigation of spectral signatures underlying sensory predictions and prediction-error generation across the auditory hierarchy in healthy volunteers and in schizophrenia patients. Specifically, we aimed to disentangle the neural mechanisms involved in inter-areal brain communication during normal brain functioning and assess whether bottom-up or top-down auditory predictive signaling is disrupted in schizophrenia.
"We developed an auditory paradigm that allows the assessment of the neural correlates of sensory predictions and prediction-error generation by presenting trains of repetitive sound sequences with different lengths, thus allowing to manipulate the predictability of acoustic inputs. Sounds in the first position within the trains were ""unpredictable"", while subsequent repetitions (up to nine presentation) led to more ""predictable"" events until a new sound sequence emerged hence restarting the predictability manipulation. Three different sound sequences were employed, consisting of four 50 ms sine waves followed by a fifth sound (inter-stimulus interval = 200 ms) that could be either a local change in frequency, a frequency repetition or a sound omission. In sum, the paradigm consisted of three sound patterns (change, repetition and omission) that could be presented under three predictability conditions (""unpredictable"": 1st presentation; ""low-predictable"": 2nd-3rd presentations; and ""high-predictable"": 4th-9th presentations). Participants were scanned using magnetoencephalography (MEG), asked to ignore the sounds and focus on a silenced film in order to drive their attention away from sound stimulation and prevent expectation effects. In the pilot phase of PredAMEG, task parameters were refined to obtain an adequate signal and finally applied to a sample of 23 healthy volunteers who underwent a MEG recording and a subsequent structural magnetic resonance imaging. MEG data was analyzed following state-of-the-art source reconstruction techniques that allowed disentangling the brain regions involved in the processing of repetition effects. We employed functional connectivity analyses to assess information flow in a hierarchical auditory-frontal network. A Granger causality approach was implemented to assess directed communication between target brain regions in distinct oscillatory bands.
Results obtained so far in healthy participants indicate that: 1) auditory networks use a multiplexing system for communication across brain regions. That is, bottom-up signaling from auditory to high-order prefrontal regions is conveyed by neural oscillations in the theta (~ 4-8 Hz) and gamma (> 40 Hz) bands. Top-down communication, on the other hand, is preferentially driven by alpha/beta oscillations (9-40 Hz). 2) Such an asymmetric communication pattern is tightly linked with the signaling of predictive signals. Prediction error signals, associated with the processing of unexpected events, elicit a dominant bottom-up communication between sensory to prefrontal regions in theta- and gamma-bands. Conversely, more repeated or predictable sounds sequences elicit stronger the top-down alpha/beta signaling from prefrontal to auditory regions. The outcomes of phase 1 have been disseminated in one international conference and several national conferences and workshops in the United Kingdom. A manuscript with the outcomes of phase 1 is under preparation and foreseeably published in 2019.
Phase 2 of the PredAMEG project is currently ongoing. Phase 2 involves the recording of a similar sample (n = 25) of schizophrenia patients and healthy controls. Identical analyses as in phase 1 will be carried out, in addition to the statistical assessment of group differences. Given the advancement made during phase 1, the outcomes from phase 2 are expected to be published during 2019.
The results from phase 1 suppose the empirical validation of theoretical assumptions at the core of the “predictive coding” framework, which translate into a strengthening of the currently accepted scientific paradigm in neuroscience.
Furthermore, results from the PredAMEG project advance in the understanding of the mechanisms underlying novelty and deviance detection, which have relevant implications in the research of ageing, language perception and clinical conditions such as psychosis, coma and vegetative state or dyslexia.
Finally, at completion of phase 2, we expect to observe a distinct pattern of neural communication in schizophrenia patients characterized by impaired bottom-up and top-down connectivity, and decreased predictability effects, suggesting a disturbance in predictive coding. Theses findings will provide critical insights into the pathophysiology of schizophrenia and will be of relevance to translational research by clarifying the oscillatory mechanisms involved in aberrant auditory processing.