TU Berlin

Modelling of Cognitive ProcessesIngo Fründ

Page Content

to Navigation

Ingo Fründ


Room FR 6003
Tel.: 030 - 314 78962
Email: ingo . fruend @ tu-berlin . de

Sekretariat FR 6-4
Franklinstr. 28/29
10587 Berlin

My position with the TU Berlin ends. You can find a recent version of my website at ingofruend.net

Research Topic

The retinal images from our environment are constantly changing. At the same time, our perceptual world is surprisingly stable. To achieve perceptual stability, our perceptual system needs to adapt to the changes in the world continuously. I am interested in the dynamical recalibration of perceptual processes, that are required to allow for perceptual stability. Processing might change from stimulus to stimulus over the time course of an experiment. I am interested in characterizing conditions for such adaptive processing and their impact on psychophysical performance.

For the analysis of psychophysical tasks, we typically make a number of assumptions about the responses of an observer. We assume that each stimulus is processed independently of all other stimuli. In addition, we assume that each response is independent of all other responses. However, these assumtions are not always justified. Typical models for psychophysical decision processes neglect these interdependencies between different stimuli and different responses. A consequence of this is underestimation of the variance in psychophysical data resulting in possible rejection of otherwise correct models. It can also lead to artificially small confidence intervals for parameters of these models. In this case, differences may be diagnosed where in fact no difference is present. We analyzed a large number of simulated psychophysical experiments and found that these errors can be very strong. We therefore developed diagnostics tools to detect violations of the above assumptions. We also derived a simpmle correction formulat that estimates the error in the data and corrects the inference accordingly. Psignifit is a toolbox for testing hypotheses about one of the simplest descriptive models in psychophysics — the psychometric function.
Together with Valentin Haenel, I reimplemented psignifit. The new version is available from http://psignifit.sourceforge.net. The new version of psignifit incorporates the developed tools.

The psignifit implementation merges all these interdependencies between sucessive stimuli and responses into a global distribution model. In addition to that, I try to characterize the interdependencies in psychophysical experiments more precisely. Natural vision develops over time. What we see at one moment in time will be very similar to what we see in the next moment. This heuristic helps a lot in interpreting the inherently ambiguous images we encounter in natural vision. In psychophysical experiments, this heuristic might be exactly what creates intedependencies in the way we process sucessive stimuli and respond to them. These interdependencies typically evolve over the time course of an experiment. I search for regularities in these serial dependencies. Knowing the rules that govern the evolution of serial dependencies provides insight into decision processes that are involved in psychophysical tasks. If we knew the rules that govern the evolution of serial dependencies it might be possible to design experiments that minimize these dependencies. This might provide a means of studying static and dynamic mechanisms of perceptual processing in isolation.

Serial dependencies are important already on the level of single trials. Typical experiments in psychophysics are performed as two-alternatives-forced-choice trials: An observer sees two stimuli. He knows that one and only one of these two stimuli contained a target feature. The observer's task is to indicate which of these stimuli contained the target. Typically the two stimuli are presented in temporal sucession. This type of task offers several theoretical and practical advantages. However, serial dependencies between the first and the second stimulus might counteract these advantages. I try to incorporate reaction times into the analysis of two-alternatives-forced-choice tasks. By using this additional dependent variable, I hope to be able to resolve dependencies in processing between the first and the second stimulus.

Curriculum Vitae

Department of Biological Psychology, Universtity of Magdeburg
Department of Biological Psychology, University of Magdeburg
Phd under supervision of Christoph Herrmann, thesis title "Speed in early vision"
University of Bremen
Studies of Mathematics
University of Bremen
Studies of Psychology


Quick Access

Schnellnavigation zur Seite über Nummerneingabe