BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Mountains of memory in a sea of uncertainty: Sampling the external
  world despite useful information in visual working memory - Julie de Falc
 o
DTSTART:20250213T140000Z
DTEND:20250213T150000Z
UID:TALK228565@talks.cam.ac.uk
CONTACT:Adam Triabhall
DESCRIPTION:This week we will discuss and debate a recent paper by Sahakia
 n and colleagues (2023).\n\nAbstract: “A large part of research on visua
 l working memory (VWM) has traditionally focused on estimating its maximum
  capacity. Yet\, humans rarely need to load up their VWM maximally during 
 natural behavior\, since visual information often remains accessible in th
 e external world. Recent work\, using paradigms that take into account the
  accessibility of information in the outside world\, has indeed shown that
  observers utilize only one or two items in VWM before sampling from the e
 xternal world again. One straightforward interpretation of this finding is
  that\, in daily behavior\, much fewer items are memorized than the typica
 lly reported capacity limits. Here\, we first investigate whether this low
 er reliance on VWM when information is externally accessible might instead
  reflect resampling before VWM is actually depleted. To this aim we devise
 d an online task\, in which participants copied a model (six items in a 4x
 4 grid\; always accessible) in an adjacent empty 4x4 grid. A key aspect of
  our paradigm is that we (unpredictably) interrupted participants just bef
 ore inspection of the model with a 2-alternative-forced-choice (2-AFC) que
 stion\, probing their VWM content. Critically\, we observed above-chance p
 erformance on probes appearing just before model inspection. This finding 
 shows that the external world was resampled\, despite VWM still containing
  relevant information. We then asked whether increasing the cost of sampli
 ng causes participants to load up more information in VWM or\, alternative
 ly\, to squeeze out more information from VWM (at the cost of making more 
 errors). To manipulate the cost of resampling\, we made it more difficult 
 (specifically\, more time-consuming) to access the model. We show that wit
 h increased cost of accessing the model (which lead to fewer\, but longer 
 model inspections)\, participants could place more items correctly immedia
 tely after sampling\, and they kept attempting to place items for longer a
 fter their first error. These findings demonstrate that participants both 
 encoded more information in VWM and made attempts to squeeze out more info
 rmation from VWM when sampling became more costly. We argue that human obs
 ervers constantly evaluate how certain they are of their VWM contents\, an
 d only use that VWM content of which their certainty exceeds a context-dep
 endent “action threshold”. This threshold\, in turn\, depends on the t
 rade-off between the cost of resampling and the benefits of making an acti
 on. We argue that considering the interplay between the available VWM cont
 ents and a context-dependent action threshold\, is key for reconciling the
  traditional VWM literature with VWM use in our day-to-day behavior” (Sa
 hakian et al.\, 2023).\n\nReference: Sahakian\, A.\, Gayet\, S.\, Paffen\,
  C. L. E.\, & Van der Stigchel\, S. (2023). Mountains of memory in a sea o
 f uncertainty: Sampling the external world despite useful information in v
 isual working memory. Cognition\, 234\, 105381–105381. https://doi.org/1
 0.1016/j.cognition.2023.105381\n
LOCATION:https://cam-ac-uk.zoom.us/j/92612577704?pwd=MUtqMjVQdXNmUTVIYjRkM
 G1NUW9GZz09
END:VEVENT
END:VCALENDAR