At stated errors and fixed pulse plateaus, 90% recovery time had no upper bound in three ECoG and two fMRI records under a positive-mixture model

We built a tool to bound visual recovery responses for any positive mixture of exponentials. Separate public human studies supplied ECoG broadband response areas and fMRI condition estimates. A fixed rule used five gaps to bound a withheld 533 ms response. The allowed ranges included all five reserved ECoG patients and all eight fMRI experiment records from six adults. The ranges were broad. Mean compatibility margins were 0.3188 and 0.2465 to 0.3128 single-pulse RMS units. Person 95% interval limits were [0.0102,0.6273] and the union [0.0113,0.5757]. The error allowance was each grid fit error plus 0.05; it was not a confidence level. An exact construction shows that a free plateau can leave full recovery time without an upper bound at any positive error slack. A later exploratory test fixed plateaus from isolated pulses. At stated errors, three ECoG records and two fMRI records still allowed arbitrarily late 90% recovery. Ten records had a conditional 533 ms upper bound. One negative measured plateau left the positive percentage undefined. All fifteen positive-plateau records needed more than 0.05 RMS fit error. Thus no shared neural mechanism was confirmed. The contribution is a continuum enclosure method, explicit time-limit certificates, and a reproducible evaluation in both modalities. Clinical sampling, reused selection, response zero, finite electrical tails, vascular effects and separate cohorts limit interpretation. Code, compact permitted data, source terms and pins, all failures, figures and repeat instructions are supplied.

visual recovery · ECoG · fMRI · positive mixtures · partial identification · extrapolation · human neuroscience