Choosing how a brain-computer interface acquires its signals is, at its core, a single decision about how much surgical risk a project can justify in exchange for neural data precision. Scalp EEG is non-invasive, cheap and safe, but its signals pass through skull and scalp, arriving noisy and spatially coarse. ECoG sits on the brain's surface under a craniotomy, capturing far cleaner signals at millimeter-scale resolution. The Utah Array penetrates cortical tissue to record individual neurons at roughly ten-micron resolution, powering the most demanding neuroprosthetics, but at the highest clinical risk. There is no best method—only the right point on that trade-off curve for a given project.

Understanding BCI Signal Acquisition: The Core Trade-offs

Researchers conventionally sort BCI signal acquisition methods by whether surgery is required, classifying them as non-invasive or invasive; a 2020 refinement by He and colleagues added a semi-invasive middle category, which is where ECoG typically lands. A review of BCI technologies frames the field's central tension plainly: each acquisition technique embeds trade-offs that must be matched to a specific application or user group rather than ranked in the abstract.

One useful way to think about the spectrum is the distinction a medical-engineering crossover review draws between "global" BCIs, which sample broadly distributed activity, and "local" BCIs, which detect precisely localized signals. Local systems capture more accurate signals, but because they depend on a limited number of neurons, they may not fully reflect complex cognitive processes. The same review notes that for domains like affective computing, implanting minute electrodes could actually impair the fidelity that biometric applications need. In other words, more precision is not automatically better—it must match what the application is trying to decode.