Frequency-dependent fine structure in the frequency-following response: The byproduct of multiple generators
The frequency-following response (FFR) is an auditory-evoked response recorded at the scalp that captures the spectrotemporal properties of tonal stimuli. Previous investigations report that the amplitude of the FFR fluctuates as a function of stimulus frequency, a phenomenon thought to reflect multiple neural generators phase-locking to the stimulus with different response latencies. When phase-locked responses are offset by different latencies, constructive and destructive phase interferences emerge in the volume-conducted signals, culminating in an attenuation or amplification of the scalp-recorded response in a frequency-specific manner. Borrowing from the literature on the audiogram and otoacoustic emissions (OAEs), we refer to this frequency-specific waxing and waning of the FFR amplitude as fine structure. While prior work on the human FFR was limited by small sets of stimulus frequencies, here, we provide the first systematic investigation of FFR fine structure using a broad stimulus set (90 + frequencies) that spanned the limits of human pitch perception. Consistent with predictions, the magnitude of the FFR response varied systematically as a function of stimulus frequency between 16.35 and 880 Hz. In our dataset, FFR high points (local maxima) emerged at ˜44, 87, 208, and 415 Hz with FFR valleys (local minima) emerging ˜62, 110, 311, and 448 Hz. To investigate whether these amplitude fluctuations are the result of multiple neural generators with distinct latencies, we created a theoretical model of the FFR that included six putative generators. Based on the extant literature on the sources of the FFR, our model adopted latencies characteristic of the cochlear microphonic (0 ms), cochlear nucleus (˜1.25 ms), superior olive (˜3.7 ms), and inferior colliculus (˜5 ms). In addition, we included two longer latency putative generators (˜13 ms, and ˜25 ms) reflective of the characteristic latencies of primary and non-primary auditory cortical structures. Our model revealed that the FFR fine structure observed between 16.35 and 880 Hz can be explained by the phase-interaction patterns created by six generators with relative latencies spaced between 0 and 25 ms. In addition, our model provides confirmatory evidence that both subcortical and cortical structures are activated by low-frequency (<100 Hz) tones, with the cortex being less sensitive to frequencies > 100 Hz. Collectively, these findings highlight (1) that the FFR is a composite response; (2) that the FFR at any given frequency can reflect activity from multiple generators; (3) that the fine-structure pattern between 16.35 and 880 Hz is the collective outcome of short- and long-latency generators; (4) that FFR fine structure is epiphenomenal in that it reflects how volume-conducted electrical potentials originating from different sources with different latencies interact at scalp locations, not how these different sources actually interact in the brain; and (5) that as a byproduct of these phase-interaction patterns low-amplitude responses will emerge at some frequencies, even when the underlying generators are fully functioning. We believe these findings call for a re-examination of how FFR amplitude is interpreted in both clinical and experimental contexts.