Thread
PessoaLabFTW 1 - 
2. Truth. 
3. Here's how we've been thinking about it, as part of a related in-prep project. Static rectangular models are the norm (top-right). 
4. Everyday experience and prominent models teach us that temporal dynamics are a central feature of fear and anxiety. "Fear"/acute-threat is supposed to be a rapid, phasic response to certain and imminent danger, whereas "anxiety"/potential threat is STB a sustained response to
5. uncertain or distal threat. Static models of the hemodynamic signal cannot reveal this complexity or determine how it's embodied in the human brain. Mechanistic studies in animals have yielded tantalizing clues, but the relevance of these discoveries to the complexity of the
human brain and human emotion remains unclear. In short, there's a considerable gulf between theory and contemporary fmri analytic approaches
Backfill: Per Hur...& Shack '20, & seen in these figs, phasic surges & other signals are psychol & biol plausible. Conventional fmri analytic approaches (boxcars) can reveal identical stats for wildly diff't signals. e.g. in left fig, sustained & phasic waves both yield T=7.75 
Back to PessoaLab: Addressing this gap requires a more dynamic modeling framework, let's see what my neighbors down the hall have come up with...
Best of both worlds: A priori ROI's to provide unbiased effect size estimates & exploratory whole-brain voxelwise analyses. Hypoth testing encompassed subcortical regions that are intensively scrutinized in rodents & frontocort regions that have dubious homologies in mice 
large healthy uni sample - 
[[ Taking dog to daycare. BRB :) ]]
PessoaLab vs ShackLab shocks - 
pre-processing similar to what we do and, like us, PessoaLab has moved to 4-mm smoothing kernel for voxelwise and unsmoothed for roi analyses 
Meat of the matter - 2 x 2 design with "offset-locked" modeling of BOLD signal 
This is uncertain likelihood instrumental threat (p=33%), with min. temporal uncertainty...so bear in mind that there is *no* "certain threat" (aka n.*P.*u) ... threat is always uncertain in likelihood, varies in intensity (lo/hi), & is under pseduo-operant control 
back to the fmri approach - we started masking out WM/CSF recently as well ... 
offset-locked splines (akin to FIRs) ... adjacent reinforcer delivery epoch w/ convolved mini-block ... like us, they report (very nice) VIF's ... indicating acceptable co-linearity in the 1st level model 
minor comment - the better reference for Nacewicz amygdala roi's is Tillman...& Shackman HBM '18 
[ BRB - now have to chauffeur kid to orthodontist appt ]
Nice touch! 
3 x 1.25 second TR average locked to end of threat/reward anticipation epoch 
Very conservative thresholding 
There will some (minimal?) contamination from onset effects ~12 seconds earlier). Bigger issue might be sustained effects. The mean of the offset epoch will ride on top of sustained effects. 
We've been tackling this using 2 approaches: theory driven decomposition of Onset, Sustained, and Phasic/Offset effects ("osp decomposition") and 
...and sequential (no overlapping) convolved mini blocks 
Onward Results!
This is ever so slightly misleading. Data are modeled & displayed at every TR (1.25 s), but only the last 3.75 s (3 TR's) are subject to inferential testing 
I'm assuming the figures show betas from the models that control for the convolved mini block outcome regressors?
Per prior work by PessoaLab, BST does what you would expect, but CeM does not respond to high threat (v low threat). BST shows valence and arousal effects, but V >> A
minor comment - MCC ROI seems to be missing the cingulate sulcus and paracingulate gyrus? figshare.com/articles/journal_contribution/shackman_dACCBlogPost_121515_pdf/2026014 
vmPFC not so into threat 
MCC likes threat (valence fx) 
Superduper cool new data, but I quibble with the interpretation, which muddles the descriptive vs inferential stats ... "ramping" was not tested; a "phasic surge" or "uptick" at the end of uncertain threat anticipation was tested (very rigorously!) 
Note - It's not just you! The in-prep project we're working is internally called "The Ramps Project" but we realized that neither of our approaches really "nails" parametric ramping across the anticipation, at least in our focal tests .... 
Sum: Extremely thoughtful and rigorous work from a friend and scientific inspire-er. We push each other to go further. There are other aspects of the work that are notable and interesting--esp the too-often-neglected Reward vs. Threat comparisons--but I ran out of time. Lab mtg!
Fruitful 75 min lab discussion. Single biggest issue was a concern with the task and interpretation of the results. Paper is framed to imply that game anticipation period (above turtle in fig) is the threat/reward anticipation epoch 
But! A blank screen waiting period is interposed b/w the end of the game and the onset of the Shock/$$. Effectively, the blank screen is a CS. 33% of trials it's CS+ (p=1.0 shock); 66% of trials it's CS- (p=0.0 shock). Analyses focused on last 3.75 s of game (above turtle) are
not the same, and cannot be interpreted as equivalent to certain-threat/acute-threat/"fear" in the usual (e.g. Mike Davis/RDoC) sense.
