Limitations of Human Visual Working Memory

The present empirical study investigates limitations of human
visual working memory (VWM). The experiments of the present work
involve the experimental paradigm of change detection using simple
geometrical objects in the form of rectangles of different colour,
length, and orientation. It can be shown, that a limited
performance in the temporary storage of visual information has
multiple sources. Limitations of VWM can be attributed to a limited
capacity or a limited duration, but also to limitations in
retrieval, which so far has found only little attention. Key
findings of the present study show, that a capacity limitation
cannot be described by a simple and generally valid size of the
store. It is in fact generally acknowledged that the capacity
limitation of VWM is object-based, which means that the capacity
can roughly be determined by the number of visual objects. However,
it could be shown that the complexity of the objects has an
influence on memory performance: Experimental evidence shows, e.g.,
that memory performance decreases, when an object is characterized
not only by one feature (e.g. colour), but by a number of features
(e.g. colour, orientation and size). The results are explained by
increased storage demands for the binding of the features. Other
key findings of the present study relate to the process of
retrieval of information from VWM. For example, an asymmetric
performance pattern could be observed: In a change detection task a
memory performance was observed that corresponds to a capacity of 3
- 4 objects. In contrast a drastic decrease of performance
corresponding to a capacity of only 1 object was observed, when the
task was to find a matching item among changed distractors. These
results lead to the idea of a change signal, by which the empirical
data can be explained. The change signal is elicited by a local
mismatch between the information stored in VSTM and perceptual
online information. The retrieval process is efficient, when the
change signal can be used in the memory task. However, retrieval is
extremely limited, when in the presence of multiple changes a less
efficient strategy has to be applied. In the course of the present
study, moreover, it became evident that there are various links
between VWM processes and visual attention. Visual attention is
probably crucial for feature binding in VWM. In retrieval the
change signal probably involves mechanisms of visual attention. The
present study could, therefore, contribute to a clarification of
the relation between VWM and visual attention