iiinova — where your digital self keeps evolving

John Thorp @johnthorp.bsky.social · 09/12/2025

in sum, we generalize the RME to both schematic and episodic relations, also showing that individual differences mediate that inference process when the paradigm allows them to. if you're still with me, i really appreciate being a part of me procrastinating both grading and packing up my apartment

010

John Thorp @johnthorp.bsky.social · 09/12/2025

Cue soap box about fear generalization being adaptive, with even our dirty harrys detaching their memories from the particularities that might stop them from properly pattern-completing when a similar cue rides into town.

110

John Thorp @johnthorp.bsky.social · 09/12/2025

that's not the case for the high confidence recognition, but interestingly we do see that our non-generalizers have worse location memory for the items within the shocked scene, relative to the unshocked scene (bottom right).

100

John Thorp @johnthorp.bsky.social · 09/12/2025

but what about differences between the two scenes? remember we didn't see any differences between the shocked and unshocked scene in the main effects – maybe our cool hand lukes who best know the difference between them actually treat them differently?

100

John Thorp @johnthorp.bsky.social · 09/12/2025

First, on the top, we show that our nervous nellies who become afraid of beaches generally recreate our effect from experiment 1, down to mostly decreasing false alarms to congruent foils. bass drop.

100

John Thorp @johnthorp.bsky.social · 09/12/2025

ok so if something about the consent procedure where I walk you through that I'm going to shock you and it's not supposed to be painful but i'm going to make sure it's not fun – if that makes you nervous – you don't learn what causes that shock as well as you could. but does it affect memory?

100

John Thorp @johnthorp.bsky.social · 09/12/2025

I then make a basic generalization measure based on the shock expectancy ratings (top). this generalization was related to the skin conductance response to the unshocked scene (bottom left, sick) as well as the state anxiety they reported at the beginning of the experiment (bottom middle, whoa)

100

John Thorp @johnthorp.bsky.social · 09/12/2025

what if i told you that participants ran the full gamut of attributing the shock fully to the shocked beach, attributing the shock fully to every beach, with some lying somewhere in the middle. we get this from the shock expectancy ratings of subjects 1, 2, and 3 (respectively) in the top figure

100

John Thorp @johnthorp.bsky.social · 09/12/2025

i admit, in a typical fear conditioning experiment where everyone has the same experience and reaches the same conclusion, this is where I leave you. but because this is a "weaker situation" i want to show you some individual differences in the inference process www.sciencedirect.com/science/arti...

sciencedirect.com

What's wrong with fear conditioning?

Fear conditioning is one of the prime paradigms of behavioural neuroscience and a source of tremendous insight in the fundamentals of learning and mem…

100

John Thorp @johnthorp.bsky.social · 09/12/2025

"ok, so you made all beaches scary and it messed with the general schema, and then you made a recent, specific beach scary and it reactivated all recent, specific associates. can i go now"

100

John Thorp @johnthorp.bsky.social · 09/12/2025

but while confident recognition increased, we got a nice dissociation where participants had worse specific location memory for incongruent objects. if it walks like systems-level consolidation, and it talks like systems-level consolidation..

100

John Thorp @johnthorp.bsky.social · 09/12/2025

now on the memory test, we see a new pattern of results where high-confidence recognition is increased for both incongruent and congruent items. this effect is constant for the shocked and unshocked scene (which is unsurprising considering how bad participants were at the source memory)

100

John Thorp @johnthorp.bsky.social · 09/12/2025

To get closer to the latter, we split our beach schema into a CS+ (shocked) scene and CS- (unshocked) scene. Ostensibly, one should be learning that "myrtle beach" (shocked) sucks and "wrightsville beach" (unshocked) rules (or is at least better than myrtle)

100

John Thorp @johnthorp.bsky.social · 09/12/2025

Put colloquially, do you get home sunburned and dehydrated and infer "going to the beach" sucks, or do you get home with horsefly bites and an upset stomach from some bad seafood and infer that "myrtle beach" sucks (no hate). which things would you remember to stay away from, in each case?

100

John Thorp @johnthorp.bsky.social · 09/12/2025

time for our true favorite buzzword: latent inference. what if our conditioning procedure only taught our participants something about the general schema of the beach? And the focus on and restructuring of this schema deprioritized their recent episodic experiences at the beach?

100

John Thorp @johnthorp.bsky.social · 09/12/2025

weird.. plenty of evidence tells us humans can reactivate and generalize information between one-shot paired associates.. and if this was really about reactivation and upregulated systems-level consolidation, then decreased false alarms wouldn't really be my first guess..

100

John Thorp @johnthorp.bsky.social · 09/12/2025

to be specific, those congruent foils are objects that belong at the beach but that you didn't actually see on day 1 (so on day 2 I show you a beach pail and shovel and you correctly say you didn't see it)

100

John Thorp @johnthorp.bsky.social · 09/12/2025

now, two interesting things: 1) no effect on the incongruent items, so, seemingly, the way we structured conditioning didn't touch those one-shot, episodic encounters. and 2) the effect in the congruent items was mostly driven by a reduction in false alarms to the congruent foils

100

John Thorp @johnthorp.bsky.social · 09/12/2025

24 hours later on a surprise memory test for the objects, we generalized the RME such that participants showed improved recognition for the congruent items from the schema that went on to be shocked. critically, as for all retroactive effects, the objects were neutral at the time of encoding.

100

John Thorp @johnthorp.bsky.social · 09/12/2025

we set up the experiment as the cleanest combination of a typical RME and a schema congruent vs. incongruent study as we could. participants saw a scene of a beach come on screen, rated a bunch of objects as congruent or incongruent, and after 60 trials of this were conditioned to fear the beach.

100

John Thorp @johnthorp.bsky.social · 09/12/2025

is it a coincidence my first project in new york was called TGiF and consisted of collecting 120 pictures of objects you would see during a weekend beach or camping trip in north carolina? regardless, fight to make your cope into your work, kids.

100

John Thorp @johnthorp.bsky.social · 09/12/2025

now, how the memory literature often operationalizes these schemas (e.g., objects found while going to the beach) overlaps with how "thematic categories" are discussed in the category and conditioning literature, spawning my name for the project: Thematic Generalization in Fear

110

John Thorp @johnthorp.bsky.social · 09/12/2025

(a web that, of course, can be particularly sidelined in fear conditioning when directly generalizing rodent behavioral paradigms with well-specified biological mechanisms to how humans think, learn, and feel)

100

John Thorp @johnthorp.bsky.social · 09/12/2025

this was my first project upon coming to grad school in nyc from north carolina, and directly came out of giving myself the formidable – though gloriously concrete – task of reading everything @joeydunsmoor.bsky.social ever wrote.

111

John Thorp @johnthorp.bsky.social · 09/12/2025

new preprint! we generalize the retroactive memory enhancement from categories of objects (like animals, tools) to schemas of objects (things at the beach / camping), with a bit of a twist about whether you find all beaches threatening or just one specific beach 🤔 osf.io/preprints/ps...

osf.io

OSF

151