CHANGELOG

# Changes to experiment / stimuli

Several changes were made to the experiment and stimuli through piloting and
development. The data presented in the paper were all `ses-04`, as described
below.

Things that have been constant: stimuli are presented 300 msec on, 200
msec off, with a stream of digits at fixation to use for a 1-back
distractor task. 10 blank "classes" (4 seconds in length) are randomly
interspersed on each run. 12 runs are gathered (or attempted to, first
two sessions ran out of time because of on-the-fly bug-fixing).

1. ses-pilot00 (git commit b88434d6af8cdc92fb741c99954fae05af02f651,
   Aug 23, 2017): negative spiral stimuli had `w_r < 0`, digits for
   the distractor task were shown every stimulus (500 msec). Only
   sub-wlsubj042 scanned using this protocol. Parameters `w_r` and
   `w_a` ran from `2^2.5` to `2^7.5`, with values every half-octave
   (rounded to the nearest integer), for each of the four main
   stimulus classes.
2. ses-pilot01 (git commit 2ab9d11c8ba077f53997dea5e525c53ef9c0dd64,
   Oct 9, 2017): negative spiral stimuli switched to have `w_a < 0`,
   digits are now shown every other stimulus (every
   second). sub-wlsubj001, sub-wlsubj042, and sub-wlsubj045 all
   scanned using this protocol.
3. ses-01, ses-02 (experiment git commit
   3f0920aee0f4f8f198c0f258b63482afbe47e3de, git commit
   aa661f8f0093a7e444fc60796a48d82006679596 for creation of local
   spatial frequency maps for all stimulus types): `w_r` and `w_a` now
   run from `2^2.5` to `2^7` in half-octave steps; this allows us to
   get an extra half-degree closer to the fovea. Also adds constant
   stimuli (spatial frequency constant across the whole image) to
   serve as a sanity check. The two sessions include one with the
   log-polar stimuli, one with the constant (these are referred to ask
   task-sfp and task-sfpconstant, respectively). The stimuli no longer
   have an alpha value and the calculation of their spatial frequency
   is correct (and is correct for the pilot stimuli as well).
4. ses-03 (git commit 69558708537c4d1a82617b05f6e39b4f2c8d7d9a): adds
   16 seconds of blank time at the beginning of each run (to improve
   estimation of baseline) and an extra 8 seconds at the end of each
   run. Still task-sfp (only log-polar stimuli).
5. ses-04 (git commit c7d6ea6543b368f3721ec836e7591f6c86baa438): same
   experimental design as before, but with new task,
   task-sfprescaled. We were concerned about the effect of the scanner
   projector's modulation transfer function, that we might be losing
   contrast at the higher frequencies (because of the projector's
   pointspread function, which effectively acts to blur the
   image). This could result in a lower response to those high spatial
   frequencies, simply from the reduced contrast (rather than the
   higher spatial frequency). In order to test this, we measured the
   projector's MTF (see the
   [spatial-calibration](https://github.com/WinawerLab/spatial-calibration/)
   Github repo for more details) and then constructed our stimuli so
   that their amplitude is rescaled in a spatial frequency-dependent
   manner; the amplitudes of lower spatial frequencies is reduced such
   that, when displayed by the scanner projector, both the high and
   low spatial should have (approximately) the same contrast. We then
   gather these new measurements and run the same analysis to compare
   the effect. We also add the stimuli for task-sfpconstantrescaled
   (same thing for the constant stimuli), but it's unclear if we will
   gather that data (would be ses-05).