Memory systems of humans and machines - Comparability and transfer possibilities

2008-02 till 2012-10
Research Areas: 

MEHME investigates human long-term memory, specifically the priming system. Our projects use visual stimuli, mainly natural photographs and drawings. We aim to differentiate the amount of information necessary to find priming in healthy humans. First results show that high informative stimuli like photographs are faster and better processed, resulting in a speeded response even after a very brief presentation. A tentative explanation is that semantic memory facilitates the processing of stimuli which are close to our visually rich environment.

Methods and Research Questions: 

Our visual system has evolved to deal with a rich environment. Unconscious perception of information, priming, influences our decisions without our awareness of it. What are the single components of visual priming and how many features do humans need in order to show priming on a behavioral and neuroimaing level?

The focus of MEHME is to develop methods to transfer what we know about the workings of human long-term memory to artifical systems. Priming is an early developing long-term memory system in humans. It facilitates processing of information from one presentation to the next as well as across repetitive presentation without consciously learning it. The main research question of the project is how much information do humans need in order to detect priming. We use different approaches to answer this question. By measuring behavioral responses, like reaction times and changes in accuracy, we are able to investigate a larger group of participants and verifying our stimuli and paradigms. Additionally, we use neuroimaing methods, like functional magnetic resonance imaing (fMRI) and eye-tracking, to detect specific brain regions related to processing different amounts of information and inferring attention and the specific information participants focus on by analyzing their gaze pattern.

In this project we work with different sets of visual stimuli, mostly photos from real-life objects, animals, and scenes. From a specific set of stimuli we also created different conditions. Our stimuli material comprises of a set of 100 colored natural photos (animals and objects) which are also equally distributed regarding taller or smaller than a table. From this set natural drawings were made as well as simply line-drawings.
This stimuli material gives us the oportunity to differentiate between the amount of information presented to participants in a given task. Currently, we are collecting data for the different conditions and establish a paradigm that can be used within an fMRI. The fMRI study will be undertaken this fall. In the eye-tracking study the different natural scenarios are used and presented in different blocks in order to investigate differences in gazing between new, old-repeated, and similar stimuli.


The main results of this project should shed light on early visual processing of information, mostly outside of our conscious knowing. Our first results suggest that priming in healthy participants often involves semantic knowledge that facilitates even during a first very brief presentation of a stimulus its evaluation and subsequent response. Humans seem to be extremely good when it comes to process their surrounding and they seem to be able to filter out the most valueable information for them mostly without any effort. Technical systems still have problems with this kind of efficient filtering processes of their direct environment, even when it comes to only one sensory channel. The last couple dekades research on visual priming focused mainly on artificial stimuli and how distinct priming reacts to changes. Our results so far suggest that even so priming is a early developed memory system and works most of the time outside our direct awareness during encoding or retrieval it still is closely interconnected to semantic memory.