• discussion/manuscript almost done
• IME figure
  • A: one red, one black angle
  • B: add colors per monkey
• Fig. 5: say something like "drop 4 outliers"
  • try to get *** aligned across panels
• Fig. 6: put B transpose as bottom row
• Fig. 3: confirm firing rates are high because of population activity, not because we're using threshold crossings?
  • max(nanmean of spikes for all sessions
  • just put high number, and either mark it in figure caption or in methods
  • note in caption, explain in methods
  • also explain in main text?
    • mention in methods that this value increases with the number of channels contributing to that dimension
• find avg firing rate per channel per session for each monkey
  • report for each monkey, mean +/- sdev
  • mention this is in main text right when I mention we have threshold crossings
• Fig. 1D: remove the -10 and 10, keep the 0
• SSS: data is closest to the data, but significant decrease in variance, blah
  • just don't overstate
  • C: gray lines should be same thickness
  • C: maybe try filling the gray?
  • B: and then have the dots be filled or empty appropriately

washout data

• could be an effect of pert. mapping having low variance, explaining the huge jump

• red (pert vs. int) shows presence of washout

• blue (washout vs. int) shows memory trace

• steve sanity check: error right before washout (through Int. IME) should be almost exactly the error right after washout

• try to quantify this memory trace effect

  • need to decide whether we include trials prior to complete washout

• look at this for all sessions

• look at nelson sessions in 03/04 vs. 06/07

  • does 06/07 show more of this generalization/memory-trace thing?

• look at cursor error through IMEs from previous days

• look at OMPs for baseline for snapback during washout

  • find sessions with no washout effect, but also with learning