Episodic Working Memory
Visual events in the world have a rich episodic structure that must be
encoded into working memory. In a dynamic visual scene, such
as
when driving a vehicle, objects frequently emerge from occlusion or
disappear. Both the order and the duration of a sequence of
such
events carry important information. Even during search in a
static scene, the saccadic nature of fixational eye movements creates
brief patterns of input to the visual system, and it is important to encode these objects as distinct
representations.
Experiments involving rapidly presented objects using rapid serial
visual presentation (RSVP) provide valuable clues as to what tricks the
visual system uses to create temporally accurate
representations.
For example, if two instances of the same item are presented within a
sequence of items, subjects often see only a single instance of the
repeated item, a phenomenon known as repetition blindness (Kanwisher
1991, Mozer 1989).
A similar (though distinct and dissociable) error is found if subjects
are encoding two target items; subjects will frequently report only the
first of the two items. This deficit is known as the
attentional
blink (Raymond Shapiro & Arnell 1992).
The attentional blink (AB) is a particularly odd phenomenon because
only if the two targets are separated by about 200-400 milliseconds is
the second item blinked. If the two items are closer (~100
milliseconds), subjects frequently report them both! More
surprising is the catalog of findings which show how the attentional
blink can be attenuated or even switched off. The
blink is reduced if subjects are distracted (Olivers &
Nieuwenhuis 2006), perceive motion (Arend, Johnston, & Shapiro 2006), or is visually cued (Nieuwenstein, van del Lubbe, & Hooge 2005). The blink seems entirely absent if subjects encode a whole string
of consecutive items (Olivers, Van Der Stigchel & Hulleman 2007; Kawahara, Kumada, & Di Lollo 2006, Nieuwenstein & Potter 2006).
These results
make it
difficult to interpret the AB as the result of a processing
limitation. Rather, this attentional shutdown hints at a strategic
attentional mechanism designed to enhance the episodic separation between target
items that are presented separately. This point is elaborated in recent modelling work (STST; Bowman & Wyble 2007, eSTST; Wyble, Bowman & Nieuwenstein, In Press).
Another curious finding in visual perception, repetition blindness,
also informs this modelling effort. RB
occurs as the result of a circuit that avoids
encoding
spurious instances of a single item. In effect, the model is
designed to err on the side of caution; it avoids encoding one item as
two copies, at the expense of sometimes missing an actual
repetition.