Slide 1: Computer-Aided Psychological Experimentation
Slide 2: I set out to answer the question of why it makes sense to use computers for 
psychological experimentation. I wanted to find out in what ways they have been used and how they 
have improved upon existing experiments. Then, after researching, I wanted to design my own 
program that could be used to study human cognitive processes.
Slide 3: So why incorporate computers into psychology? It has long been known that computers 
excel when it comes to data collection, compilation and analysis. Long mathematical equations can 
be computed instantly. Software can perform operations on data and allow experimenters to 
restrict that data by enforcing specific value ranges their variables must meet in order to be 
accepted. With computers, data can be saved and restored easily to provide a more efficient 
method of experimental approach and study. Because computers enhance data analysis and in general 
ease workloads, there is no reason why computers can not do the same for psychological 
experiments. The computer can serve as a medium between the researcher and the subject. Also, 
computers can modify previous experiments by adding new dimensions to them, such as accurate 
timing and excellent randomization, and by allowing incorporation of other technologies such as 
audio and video equipment.
Slide 4: One advantage that computers have over other tools is that they can be used to simulate 
specific environments. At the University of Amsterdam and the Delft University of Technology, 
researchers are using virtual reality to study phobias like the fear of heights. The photo on the 
left is an actual picture of a stairwell, whereas the picture on the right is a virtual reality 
simulation. Thus, mental conditions that are triggered from certain stimuli can be examined 
through this simulation. The researchers are also concerned with virtual reality exposure therapy 
and the psychological aspects of virtual reality through experimentation. At this time, I would 
like to examine several psychological experiments that have been previously performed that could 
be improved upon by the use of computers.
Slide 5: In 1935, J. Ridley Stroop conducted an experiment that tested the difficulty of naming 
the ink color of a color word. His hypothesis for the Stroop Effect was that there would be a 
faster response time for congruent words, such as the word .blue. written in blue ink, compared 
to neutral words, like the word .house. written in yellow ink, compared to incongruent words, 
such as .red. written in orange ink. The independent variables for the experiment were the type 
of word: congruent, incongruent or neutral, and the number of stimuli per trial. The dependent 
variable was the response to the color of each word and its correctness.
Slide 6: The Stroop Effect demonstrates that we can read words more quickly and automatically 
then we can name colors. When we are required to name the color of the written word confliction 
occurs between our automatic and controlled processes. It then becomes difficult to name the 
color of the ink. Rather, we want to name the word itself, because reading is a more innate 
process than color-recognition and, thus, interference occurs.
Slide 7: Introducing computers to the Stroop experiment allows for additional results to be 
obtained, thereby expanding the dimensions of the original experiment. The experiment could 
require a verbal response, an updated version of what Stroop did when he conducted the experiment 
originally, but with a microphone interconnected with a computer rather than the participant 
directly relaying their response to the experimenter. The responses themselves could then be 
collected and played back via the computer so that experimenters could analyze the tone of voice 
of the participant throughout the trials to measure their level of apprehension in responding to 
the stimuli. That is, when the participant says .RED. confidently versus when he says .red. 
apprehensively. Would the level of apprehension begin to lower throughout the trials as the 
participant practices this mental struggle and gains better control over the interference? A 
computer application designed to measure changes in tone could be used to analyze this data. The 
experiment could also be done with a standard keyboard or with buttons labeled by color. With 
computers, another dependent variable could be introduced: the amount of time, measured in 
milliseconds, could be recorded between each response to analyze the time elapsed while 
attempting to process the information and give the correct response. This is significant because 
it allows researchers a way to document the struggle between automatic and controlled processes 
in a way much more precise than without computer assistance.
Slide 8: This next experiment examined differentiation between tones and was conducted by 
University of California San Diego psychologist and Professor Diana Deutsch. (1969). She 
hypothesized that as the overall number of tones increase, successful identification of the first 
and last tones decrease. Also, .when the test and probe tones differ in pitch, and neither of 
these pitches is included in the intervening sequence, errors are only 6 per cent. However, 
errors increase to 24 per cent with insertion in the third serial position of the intervening 
sequence of a tone at the same pitch as the probe tone (condition 8)..  
The way this experiment works is as follows: After the initial tone, the test tone, participants 
heard a set of four different tones, a pause, and then a final tone, the probe tone. Subjects 
were told they could ignore the four middle tones if they wished. They were then asked to 
identify if the probe tone was identical to the test tone. Here, A represents the pitch of the 
test tone, and of the probe tone when the two are identical. B represents the pitch of the probe 
tone when it differs from A. Tones were generated by a computer but were originally recorded on 
tape; it would seem more appropriate now to save them directly to a hard drive so that rewinding, 
fast-forwarding and tape maintenance could be eliminated altogether.
Deutsch recognized that computers can be of great assistance to experiments like these and stated 
.COMPUTER-GENERATED tonal sequences have several advantages for investigations of immediate 
memory: the stimulus parameters are simple and can be exactly controlled, and the items cannot be 
readily rehearsed.. The independent variables for this experiment were number of tones and pitch 
variance amongst those tones. The dependent variable was whether or not the participant could 
correctly identify the first and last tone as the same or different tone.
Slide 9: Gender recognition studies have been done previously without computers and simply by 
masking specific areas of the face with black tape, where the faces are on large note cards. The 
hypothesis for this study is that as more prominent facial features are masked, the greater the 
chance of an incorrect gender identification. The independent variable is the level of difficulty 
introduced by masking of specific facial features. The dependent variable is the participant.s 
ability to correctly identify the gender of the face.
Slide 10: Computers can aid this experiment by manipulating and masking facial features on the 
fly, and by quickly supplying hundreds of faces from a large database for use. As with other 
computer-aided experiments, the computer can be used to record the exact amount of time required 
to correctly (or incorrectly) identify the gender throughout the trial. An interesting experiment 
that could utilize this approach to gender recognition would be one that tested if a participant 
could recall names of people with prominent features when those features are masked. The faces of 
a dozen people the participant had just met could be digitally mapped and by blurring out key 
facial features and presenting the new faces on the screen, the experimenter could judge if other 
features allow the participant an alternate method of correctly naming the people; that is, if 
someone.s nose was prominent, would the participant be able to recall that person.s name by 
examining their chin or ears instead or do they require the more prominent feature for 
identification?
Slide 11: The previous experiments show various methods that psychologists employ to further 
understand the human mind and its processes. However, there is a deeper level process that I 
focused on and examined for experimentation, a process that we as humans utilize everyday, the 
process of reading. Reading is an automatic process, which means it requires very little 
attention to do. As we read and analyze the lexical structure of a sentence we begin to parse the 
elements of the sentence into groups of words and ultimately form a grammatical tree as shown 
here. The sentence .I saw her duck. can form two similar trees with different meanings depending 
on the context. In sentence (a), her is being used as a possessive determiner, and duck is a 
noun. In sentence (b), her is the noun phrase and duck is a verb. Two very different meanings 
from what appears to be the same sentence.
Slide 12: Understanding sentence processing allows us to examine what is known as the garden-path 
phenomenon. A garden-path sentence is one that tends to lead the reader astray, or down the 
.garden path.. The level of difficulty of a garden-path sentence varies by its design. In 
particular, when there are ambiguities involving whether a noun is the object of one clause or 
the subject of another, the sentence becomes more difficult to parse. In this example, the reader 
automatically pairs together what is believed to be the action the NP is performing, the verb 
.hunted..  However, as the sentence shows after careful analysis, the deer is not the object of 
the verb .hunted. but rather the subject of the verb .ran., while the verb .hunted. is the 
subordinate, intransitive verb. If the verb or verbs used in the sentence have a weak 
association, that is, .ate. and .soda., the easier the sentence will be to comprehend. 
Slide 13: The left example depicts an incorrect grammatical tree. The reader assumes the verb 
.hunted. to be transitive, that is, to have an object, and this confuses the reader into 
associating that word with the object. However, when we come to the next verb, .ran,. we 
recognize that we have made a mistake. The grammatical tree on the right is the correct parsing 
of the sentence: .While John hunted, the deer ran into the woods..   So when we read why do we 
prefer one structure over another? Because we innately attempt to use what is known as .minimal 
attachment. while reading . that is, we prefer grammatical structures involving the least amount 
of operations, and, thus lesser complexity. We also utilize what is called Late Closure . which 
means that we would rather attach sentence constituents lower down on the parsing tree, rather 
than higher, thus, producing a cleaner result the first time through.
Slide 14: In an effort to further understand how we parse sentences, psychologists have performed 
experiments utilizing eye-tracking devices. The way that this works is that a camera and 
computer-based apparatus that records eye fixations and eye movements are attached to the head of 
the participant. An infrared light is then fixated on the pupil and tracks the pupil.s movements. 
The timing of these movements, that is, the location and duration of fixations, is recorded. 
Regressive eye movements are recorded whenever the participant.s eyes move back to a 
previously-read portion of text. The experimenters noted that most content words were fixated on, 
while less meaningful words, such as .of, the, and., were rarely fixated on.
Slide 15: The diagram here shows how the sentence is read in groups of words. When we encounter 
the word ran, our eyes go back to the group .the deer. and then immediately to .into the woods.. 
With eye-tracking equipment, it is easy to track eye movements of participants while reading 
garden-path sentences.
Slide 16: I wanted to simulate an experiment where eye-tracking equipment was involved. Without 
infrared eye-tracking equipment at my disposal I decided to create an alternate tool that would 
obtain similar results. Of course, without the actual equipment the experiment is less accurate 
but a much more affordable option. My application presents an alternate method of tracking eye 
movement by measuring the amount of time the participants spend on each word. The design of the 
experiment is as follows: Participants read 15 random sentences out of an available 35 (of which 
25 were garden-path and 10 were non-garden-path sentences). Timing began after they read the 
first word and then pressed the spacebar. Upon pressing the spacebar again the previous timer 
would stop and a new timer would start. The results of each sentence were then sent to an output 
file so I could analyze the data. My hypothesis was that non-garden-path sentences would take far 
less time than garden-path sentences, and that as the level of difficulty increased for the 
garden-path sentences (mainly sentences 15-25) the amount of time to read the sentence would 
increase. I also predicted that whenever a participant came across a garden-path sentence and up 
to the point where they had to mentally re-examine the sentence, the word that caused the 
re-examination would have a higher amount of time recorded than the rest of the words. I will now 
give you a brief tutorial of my program.
[Show program]
Slide 17: This plot depicts the average time spent per sentence, regardless of sentence size. The 
trickier sentences are shown by the peaks in the middle of the plot. The non-garden-path 
sentences, which I presumed to be easier, are found towards the end, where the points on the 
graph are lower.
Slide18: This graph shows the average time per word. The average sentence size in my program was 
10 words in length. As the graph shows, there were no sentences of length less than 7 and none 
greater than 15. Participants spent more time per word on shorter sentences, such as .The horse 
raced past the barn fell.. These garden-path sentences are more difficult to understand because 
of greater confusion.
Slide19: Here is a glimpse of some of the results I obtained from my experiment. What I 
discovered is that in some sentences, the word directly after the area of confusion also had a 
higher measure of time. This seems to indicate that the mental re-examination of the sentence has 
confused the reader and the reading process has not yet fully recovered.
Slide 20: In conclusion, computers hold a vital place in psychological experimentation. 
Previously performed experiments can be modified to obtain additional data. Software can collect, 
compile and analyze data with extreme accuracy and efficiency. As technology advances, the topic 
of psychological experimentation with computers will heat up. In what ways will virtual reality 
come to benefit these kinds of experiments? Will artificial intelligence play a role in future 
experimentation, allowing us to create programs that adapt in real-time to the response of the 
participant to maximize results? It is difficult to be certain, but the face of psychological 
experimentation will change with the advancement of human technology.