Allowing the subject in this experiment more freedom than Dr. Clynes permitted will both introduce significant noise into the data and simulate more realistic conditions for the natural expression of emotions. The use of a mouse rather than a formal sentograph offers an interesting mix of strongly adhering to and vastly straying from the guidelines set forth by Dr. Clynes in his experiments (see Sentics Chapter 4). Constraints such as arm and finger position as well as handedness are maintained while others like back posture and rigidity of body movement are sacrificed. Additionally, since the goal of the experiment is simply to extract positive versus negative reactions, and not to try to replicate the emotion patterns discovered by Dr. Clynes, the training period will be eliminated in favor of an untrained naturally occurring expression of emotion.
The self-reported data collected and the SAM data collected through
the IAPS project are the benchmarks against which the sentic data were
processed. If the sensor data correlates with valence than success has
been achieved, we can reject the null hypothesis. In our setup we
found that our subject's responses had highly significant correlation
to those in the Lang study.
The female data demonstrated a Pearson correlation of 0.885
(p<.001) to the Lang results, and the men's 0.757 (p<.001). This
significance supports the validity of our experimental setup.
By taking the difference of the front sensor signal from the back a
measure of valence is derived. It is this difference that holds the
sentic information. A person reacting to a negative valence stimulus,
pushing away on the sensor, showed a positive bump in this new
signal. Similarly, positive valence resulted in a dip in the new
signal, as predicted by Dr. Clynes. The sentic data collected
correlated significantly within as well as between
subjects. It is
noteworthy to mention that simply taking the sum of the data streams
would yield intensity information.
The results from the Winton, Putnam, and Krauss experiment concerning
the relation between HR and valence were not verified here, both
because we did not have access to the exact EKG sensor positioning
used in that experiment, and because we could not replicate the signal
processing Winton, Putnam, and Krauss conducted. We did not get a
chance to compare our data collected by the HR monitor against either
the self-reported data or the sentic detector, due to technical and
time constraints. This is left open as an avenue for future
work.