Hardware One source of error came from the fact that we were measuring the conductance of the force resistor. The FlexComp sampler used to collect the signal did not have a setting for potential. The signals acquired are therefore the inverse of what a resistance signal would look like. The use of a voltage isolator as an interface between the FlexComp and the resistor sensor, might provide a cleaner signal.
The FlexComp software was programmed to sample the EKG at 1000Hz and the sentic mouse at 31Hz but we discovered upon analysis that it was actually sampling everything at 31Hz.
The construction of the sentic mouse was plagued with technical problems. The result is bulky, odd-looking, and uncomfortable. Subjects often asked how to work the mouse before the trials began. Although the experimenter instructed them to use it as a natural mouse, clearly they were aware of the presence of the sensor and that effected the way they used the mouse. Future work to make this device more robust and to hide the sensor within the button on the mouse would benefit the collected data greatly.
Design After briefly reviewing the data collected from the first ten subjects, it was decided that a signal marker was needed for the sentic signal. The lack of a timing apparatus and the freedom of the user to move the mouse made the sentic click impossible to find in most trials. For the future, either an auditory tone as is used in Dr. Clynes work, or better yet the embedding of a switch in the button of the sentic mouse should be implemented. The switch will close a circuit sending an accurate timing impulse to the FlexComp. At the time of data analysis, this impulse signal can be used to extract the relevant mouse activity and sentic patterns.
The same random order of slides was presented to each subject. This was due to time restrictions. In future incarnations of this experiment, the order of slides should be shuffled for each subject to compensate for priming effects. For example in our study, the picture of the sky happened to fall after an especially gruesome face. The overall ranking of the sky was inflated due to the negative priming of the mutilation slide.
The slider bars used in the experiment always reset to zero. Thus if the subject failed to respond, they would tend to leave the measure at the zero mark. The impact of this is a dampening of the self-reported data as compared to the Lang database. Another contributor to the dampening effect is that the Lang pictorial scale is a nine-point scale, whereas the slider bars only measured seven points.