Skillp1 Specs
Overview. The experiment will have two phases. In Phase 1, subjects will see words
presented one a time, and make a speeded response based on whether the word is an
animal, a place name, or a first name (all words will be presented in lower-case). In
this phase, each word will be presented 5 times; some words will be followed by
pauses (after the subject responds, the word disappears and the next trial does not
begin for 2 seconds) whereas for other words, the next trial begins immediately. In
Phase 2, the subject will be given a surprise quiz, and asked to recall as many words
as possible from the first phase.
Goals. We are interested in the possibility that post-response pauses may have
different effects on different kinds of learning. The experiment will allow us to assess
learning in two ways: by looking at the speedup in responding to each subsequent
presentation of the same word within phase 1, and by looking at recall probability in phase 2.
General Approach.
Two lists of 15 words will be provided, each containing 5 animal names, 5 place names,
and 5 first names. Call these lists ListA and ListB. For now, make a dummy list with
items like Aplace3 and Banimal4 and Bfirstname2. Each would show follow by a
marker, P, A or N, indicating what category the word belongs to. Example:
Aplace3 P
Banimal4 A
Bfirstname2 N
At the beginning of the experiment, ask the experimenter for the condition number of the
subject. If cond=1, then assign ListA to be followed by pauses, and ListB not to be. If
cond=2, then assign ListB to be followed by pauses, and ListA not to be.
Make a big list that contains 5 copies of each of the two lists appended one to the next, so
that every word appears 5 times in biglist, and biglist contains 150 items. Now shuffle
biglist. This is the set of words to be presented in Phase 1.
Phase
1.
Begin by presenting a fixation point (plus in center of screen) for two seconds. Then present
each word one at a time in the center of the screen, in fairly large letters. After the subject
responds to that word, remove the word. If the word is in pause condition, insert a 2-second
pause before you put up the next word.
Data from Phase 1. After each trial, write the following columns:
1. Trial number
2. Word presented
3. Correct response
4. Identity of subject's response
5. Whether response was correct (3=4?)
6. Response time in milliseconds
7. Pause presented after word? (Y or N)
8. Presentation number for that item
The only tricky thing here is determining presentation number for that item. It will be easier to
compute at runtime, so here's how it should work. This code should be 1, 2, …, or 5. It
reflects whether this is the first, second, …, fifth occurrence of that word within the sequence.
Obviously you can only compute this after you do the overall randomization of biglist.
Phase
2.
In this phase, the subject is asked to try to type in as many of the words as they can from Phase 1.
In the raw data file, store all of the letter strings that they type in.
Instructions
for Phase 1.
In this phase, we will present words one at a time on the screen. Each word is an animal name,
a place name, or a first name. There are 3 response keys, labeled ANIM for animal, PL for place,
and FIR for first name. Your task is simply to press the appropriate key. Please try to respond as
quickly and accurately as possible. When you have responded, the computer will proceed to the
next trial, sometimes after a brief pause.
Instructions
for Phase 2.
We are interested to see how many words you can recall from the first phase of the experiment. At
first you may find it hard to think of more than a few, but please take a little extra time to try to
remember as many as you can. Just type the words in any order you want to--we are just interested
in seeing how many total words you can recall.
Skillp1 Analysis Specs
In addition to the experiment running program, we will need a special-purpose analysis program.
This will read in the raw data files from all subjects, once they are available, and produce 4 different
summary datafiles (one file total, including data from all the subjects), which will be suitable to be
analyzed by ANOVA. As you know, ANOVA wants one row per datapoint, and labels that include
subject #. These summary data files are as follows:
1. Phase1_RT_Summary
This file will have 10 rows per subject corresponding to one row per condition in the 5X2 matrix
[Presentation# X Pause],
Sub# Pres# Pause? Mean-RT
So the 10 rows for subject 5 might appear like this
5 1 No 585
5 1 Yes 614
5 2 No 644
etc.
To compute Phase1_RT_Summary, you need to go over the Raw Data file, and basically average
over all the words within a condition, excluding any RT that falls outside of the RT boundaries. Don't
average in any RTs in excess of 5000 msec or below 200 msec.
2. Phase1_Exclusion_Totals
This is just like Phase1_RT_Summary, except that instead of RT it shows the total number of trials
excluded for that subject and condition in computing Phase1_RT_Summary. So the possible values
in this file are 0,1,2,…,15, since 15 is the largest possible number of words that could be excluded).
3. Phase1_Error_Summary
This is just like Phase1_RT_Summary, except that instead of RT it shows the total proportion of
errors for that subject and condition in computing Phase1_RT_Summary. Since there were 15 words
per condition per subject, this is the number of errors divided by 15, so the numbers in this file can only
take the values 0, 0.06, 0.13, 0.2, …, 1.0.
4. Phase2_Recall_Summary. This is based on the data from Phase 2. There are three columns
Subject# Pause? Proportion of words recalled
Now, of course Pause? refers to whether there was a pause in Phase1. Here is how you need to
compute this. You need to go over the raw data file events from the first phase, and find out which
15 words were presented with a pause after them. Then you look at the raw data record from phase
2, and see what proportion of these words were successfully recalled. So the possible values here
will be the digital versions of 0, 1/15, 2/15, …., 14/15, and 15/15.
Now, you might save time by having the experiment running program spit out a list of which words are
in which condition, and then the analysis program could look at that list in order to compute the
Phase2_Recall_Summary. If you want, that can go in the raw data file, as long as it won't get confused
with the record of what words were typed.
Maybe more elegant would be to have the analysis program look at the subject counterbalancing
condition number, and look at the word files for ListA vs ListB, and use that to infer which words
were presented with pauses and without pauses, rather than going back over the raw datafile to
determine which items were presented with pauses and without. This solution makes me slightly
nervous, though, because it doesn't incorporate any inherent check on what really happened in Phase1.