SupplementaryMaterial: Interpretable multi-timescalemodelsforpredictingfMRIresponses tocontinuousnaturalspeech