PSC100Y

Introduction to Cognitive Psychology

• Two main kinds of statistics:!

• Descriptive: describe a set of data!
• Such as calculating the proportion of people in
the picture above wearing red shirts.!
• Inferential: When we take a small subset of a
Research Methods 1-1! group of people (a “sample”) and use them to
draw inferences about the broader group!
Populations and Samples! !

Sample!

Population!

• If we spent enough time, we could figure out the • Our goal with inferential statistics is to be able to draw
proportion of people in the whole crowd who were valid conclusions about a large or even infinite population
wearing red shirts.! even though we’ve only measured from a small subset of
that population (a sample of the population).!
• That would again be a descriptive statistic, because it
describes the people in the entire crowd.! • However, these conclusions are only probabalistic.!

• Fish! • Snow! • Fish or Soap?!

• Candy! • Home! • Rug or Candy?!
• Guitar! • Sky!
• Cake or Guitar?!
• Window! • Heart!
• Finger! • Radio! • Window or Saw?!
• Car! • Watch! • Toothbrush or Finger?!
• Printer! • Book!
• Car or Water?!
• Grass! • Towel!
• Printer or Salt?!
Goal: measure how well people can remember a set of 100 • Movie or Grass?!
words.!
Memory is tested by showing them pairs of words, one from
Subjects are given 10 minutes to look at the words.! the list and one not from the list, and having subjects circle
the word that had been on the list.!
They come back to the lab a week later, and we test their
memory.!

© S. J. Luck

All rights reserved
1

 Percent  Correct  on  
Test  
Subject  S1   68  
Subject  S2   51  
Subject  S3   60  
Subject  S4   70  
Subject  S5   58  
Subject  S6   67  
Subject  S7   60  
Subject  S8   64  
Memory experiment with a sample of 8 subjects who are Sample  Mean   62.3  
randomly selected from the 200 students in a cognitive
psychology class.!
Our goal is to estimate the average memory ability of the Goal: estimate the population mean!
whole class, but we don’t have time to do this with all 200,
so we just take a sample of 8 students from the class.! The sample mean is an estimate of the population mean!
- Number of subjects (N)!
They come to the lab, view the 100 words, and then come - Variability among subjects (standard deviation)!
back a week later for the memory test.!

Variability:!
Percent  Correct  on  
Sample Size:! Percent  Correct  on   • In this sample, all subjects got
Test   Test  
• The letter N denotes the between 51 and 70 percent
Subject  S1   68   Subject  S1   68  
Subject  S2   51   number of subjects in a Subject  S2   51   correct on the memory test.!
Subject  S3   60   sample.! Subject  S3   60  
Subject  S4   70   Subject  S4   70   • Given that the sample mean was
Subject  S5   58   • In our imaginary experiment, Subject  S5   58   62.3% correct and all 8 subjects
Subject  S6   67   N is 8.! Subject  S6   67  
scored between 51 and 70
Subject  S7   60   Subject  S7   60  
Subject  S8   64   • All else being equal, the Subject  S8   64   percent correct, is it very likely
Sample  Mean   62.3   sample mean becomes a Sample  Mean   62.3   that the actual population mean
N   8  
better and better estimate
N   8   is 90% correct?!
of the population mean as N • In other words, if we took the time to test all 200
increases.! subjects, is it very likely that the mean of all 200
students would be 90% correct?!
• It can be time-consuming and expensive to test a large
sample, so most experiments in cognitive psychology use • If the average across the whole class is 90% correct, it’s
sample sizes of between 8 and 50 subjects.! pretty unlikely that, just by chance, we would select 8
people who all scored 70% or less.!

Percent  Correct  on   Percent  Correct  

PSC100Y
Subject  S1  
Test  
68   Subject  S1  
on  Test  
59   Introduction to Cognitive Psychology
Subject  S2   51   Subject  S2   58  
Subject  S3   60   Subject  S3   62  
Subject  S4   70   Subject  S4   60  
Subject  S5   58   Subject  S5   57  
Subject  S6   67   Subject  S6   60  
Subject  S7   60   Subject  S7   63  
Subject  S8   64   Subject  S8   60  
Sample  Mean   62.3   Sample  Mean   59.9  
N   8   N   8  
• In this example, the sample mean is 59.9% correct, and all the values
are between 58 and 63 percent correct.!
• If we got a set of values like this, it would be pretty unlikely that the
population mean was 68% correct.!
• The less variability there is, the more accurately we can use the Research Methods 1-2!
sample mean as an estimate of the population mean.!
Measures of Variability!

© S. J. Luck

All rights reserved
2

 Percent  Correct  on   Difference  from   Absolute  Value  of   Percent  Correct  on   Difference  from   Absolute  Value  of  
Test   Sample  Mean   Difference   Test   Sample  Mean   Difference  
Subject  S1   68   5.8   5.8   Subject  S1   68   5.8   5.8  
Subject  S2   51   -­‐11.3   11.3   Subject  S2   51   -­‐11.3   11.3  
Subject  S3   60   -­‐2.3   2.3   Subject  S3   60   -­‐2.3   2.3  
Subject  S4   70   7.8   7.8   Subject  S4   70   7.8   7.8  
Subject  S5   58   -­‐4.3   4.3   Subject  S5   58   -­‐4.3   4.3  
Subject  S6   67   4.8   4.8   Subject  S6   67   4.8   4.8  
Subject  S7   60   -­‐2.3   2.3   Subject  S7   60   -­‐2.3   2.3  
Subject  S8   64   1.8   1.8   Subject  S8   64   1.8   1.8  
Sample  Mean   62.3   Sample  Mean   62.3  
N   8   N   8  

• The standard way to quantify variability is to look at how • For example, the score for subject 1, 68, is 5.8 higher than the
sample mean of 62.3.!
different each individual is from the average across the
individuals.! • The score for subject 2, 51, is 11.3 lower than the sample mean.!
• The average of all these differences is zero, because some are
• In our example experiment, we can do this by computing positive and some are negative, so we can’t just average together
the difference between each subject’s test score and the these differences to get a measure of the variability across subjects.!
sample mean.!

Percent  Correct  on   Difference  from   Absolute  Value  of   Percent  Correct  on   Difference  from   Absolute  Value  of  
Test   Sample  Mean   Difference   Test   Sample  Mean   Difference  
Subject  S1   68   5.8   5.8   Subject  S1   68   5.8   5.8  
Subject  S2   51   -­‐11.3   11.3   Subject  S2   51   -­‐11.3   11.3  
Subject  S3   60   -­‐2.3   2.3   Subject  S3   60   -­‐2.3   2.3  
Subject  S4   70   7.8   7.8   Subject  S4   70   7.8   7.8  
Subject  S5   58   -­‐4.3   4.3   Subject  S5   58   -­‐4.3   4.3  
Subject  S6   67   4.8   4.8   Subject  S6   67   4.8   4.8  
Subject  S7   60   -­‐2.3   2.3   Subject  S7   60   -­‐2.3   2.3  
Subject  S8   64   1.8   1.8   Subject  S8   64   1.8   1.8  
Sample  Mean   62.3   Sample  Mean   62.3  
N   8   Average  DeviaIon  =  5.0   N   8   Average  DeviaIon  =  5.0  
• Instead, we could just take the absolute value of each difference, Std  Dev   6.3  
making the negative numbers into positive numbers.!
• Instead of the average deviation, statisticians use something called
• We could then average together these difference scores.! the standard deviation.!
• This would give us what’s called the average deviation, which in this • It’s like the average deviation, but computed slightly differently (don’t
case is 5.! worry about the actual formula). !
• This means that, on average, the single-subject scores are 5 away
from the sample mean.!

Percent  Correct  on   Difference  from   Absolute  Value  of   Estimating the Population Mean!
Test   Sample  Mean   Difference  
Subject  S1   68   5.8   5.8   • In our imaginary memory experiment, we might want
Subject  S2   51   -­‐11.3   11.3   to know if subjects remembered anything at all a week
Subject  S3   60   -­‐2.3   2.3  
Subject  S4   70   7.8   7.8  
after they learned the word list.!
Subject  S5   58   -­‐4.3   4.3  
• If they didn’t remember any of the words, and they
Subject  S6   67   4.8   4.8  
Subject  S7   60   -­‐2.3   2.3   had just guessed for each word pair, they should have
Subject  S8   64   1.8   1.8   gotten a score of about 50% correct.!
Sample  Mean   62.3  
N   8   Average  DeviaIon  =  5.0   • Some people might be lucky guessers and get above
Std  Dev   6.3   50%, and others might be unlikely and get less than
50%.!
• The smaller the standard deviation, the more accurately
we can use the sample mean as an estimate of the • But we’d expect that the population mean would be
population mean.! 50% if none of the subjects remembered any of the
words from the original list, and they were just
• To have a really good estimate of the population mean, guessing.!
we need both a small standard deviation and a large N.!

© S. J. Luck

All rights reserved
3

 Percent  Correct  on   Difference  from   Absolute  Value  of  
Test   Sample  Mean   Difference  
Subject  S1   68   5.8   5.8  
Subject  S2   51   -­‐11.3   11.3  
Subject  S3   60   -­‐2.3   2.3  
Subject  S4   70   7.8   7.8  
Subject  S5   58   -­‐4.3   4.3  
Subject  S6   67   4.8   4.8  
Subject  S7   60   -­‐2.3   2.3  
Subject  S8   64   1.8   1.8  
Sample  Mean   62.3  
N   8  
Std  Dev   6.3  
Chance   50   • But it seems possible that, if the average across all 200
students was 50%, just by chance we might get 8
• In other words, we want to know if the population mean is subjects with an average score of 62.3%, even if they
greater than would be expected by chance (50%).! were guessing.!
• Our sample mean was 62.3% correct, which is quite a bit • And if we randomly chose 8 different people, we might
higher than 50%, but it seems possible that we might get get an average of something like 41% for those 8.!
8 subjects with an average score of 62.3%, even if they
were guessing.! • So we need a way !

One-Sample t Test!
PSC100Y
Introduction to Cognitive Psychology • Fish or Soap?!
• Rug or Candy?!
• Cake or Guitar?!
• Window or Saw?!
• Toothbrush or Finger?!
• Car or Water?!
• Printer or Salt?!
Research Methods 1-3! • Movie or Grass?!
One Sample t-Test Overview!

One-Sample t Test!
Percent  Correct  on   Difference  from   Absolute  Value  of  
Test   Sample  Mean   Difference  
Null Hypothesis (H0): Population mean = 50%! Subject  S1   68   5.8   5.8  
Subject  S2   51   -­‐11.3   11.3  
Alternative Hypothesis (H1): Population mean ≠ 50%! Subject  S3   60   -­‐2.3   2.3  
Subject  S4   70   7.8   7.8  
Use sample data to compute t value and p value! Subject  S5   58   -­‐4.3   4.3  
Subject  S6   67   4.8   4.8  
Reject null hypothesis (accept alternative) if p < .05! Subject  S7   60   -­‐2.3   2.3  
Subject  S8   64   1.8   1.8  
Sample  Mean   62.3  
N   8  
Std  Dev   6.3  
Chance   50  
t   5.54  
p   0.001  

• Because the p value is less than .05, we would reject the

null hypothesis and conclude that the population mean
was not 50%!

© S. J. Luck

All rights reserved
4

 One-Sample t Test! One-Sample t Test!

Null Hypothesis (H0): Population mean = 50%! Null Hypothesis (H0): Population mean = 50%!

Alternative Hypothesis (H1): Population mean ≠ 50%! Alternative Hypothesis (H1): Population mean ≠ 50%!
Use sample data to compute t value and p value! Use sample data to compute t value and p value!

Reject null hypothesis (accept alternative) if p < .05! Reject null hypothesis (accept alternative) if p < .05!

Population mean = 50%! Population mean ≠ 50%! Population mean = 50%! Population mean ≠ 50%!
(H0 is true)! (H1 is true)! (H0 is true)! (H1 is true)!
! !
Sample mean is not TRUTH!! Type II error! Sample mean is not TRUTH!! Type II error!
significantly different (True negative)! (False negative)! significantly different (True negative)! (False negative)!
from 50%! from 50%!
Sample mean is Type I error! TRUTH!! Sample mean is Type I error! TRUTH!!
significantly different (False positive)! (True positive)! significantly different (False positive)! (True positive)!
from 50%! ! from 50%! !

One-Sample t Test!
True Negative!
Null Hypothesis (H0): Population mean = 50%!
• True Negative: When the null hypothesis is true and we
accepted it.! Alternative Hypothesis (H1): Population mean ≠ 50%!

• When the null hypothesis is true, we will get a true Use sample data to compute t value and p value!
negative in 95% of experiments.!
Reject null hypothesis (accept alternative) if p < .05!

Population mean = 50%! Population mean ≠ 50%!

(H0 is true)! (H1 is true)!
!
Sample mean is not TRUTH!! Type II error!
significantly different (True negative)! (False negative)!
from 50%!
Sample mean is Type I error! TRUTH!!
significantly different (False positive)! (True positive)!
from 50%! !

One-Sample t Test!
False Positive!
Null Hypothesis (H0): Population mean = 50%!
• In 5% of experiments in which the null hypothesis is true,
just by chance we may get a sample with unusually high Alternative Hypothesis (H1): Population mean ≠ 50%!
or unusually low scores, leading to a large t value and a p
value that’s < .05.! Use sample data to compute t value and p value!

Reject null hypothesis (accept alternative) if p < .05!

• We then reject the null hypothesis and accept the
alternative hypothesis, even though in fact the null Population mean = 50%! Population mean ≠ 50%!
hypothesis is true (false positive).! (H0 is true)! (H1 is true)!
!
• In other words, even though the population mean is at Sample mean is not TRUTH!! Type II error!
significantly different (True negative)! (False negative)!
50%, our sample mean is high enough that we incorrectly from 50%!
conclude that the population mean is > 50%! Sample mean is Type I error! TRUTH!!
significantly different (False positive)! (True positive)!
from 50%! !

© S. J. Luck

All rights reserved
5

 One-Sample t Test!
True Positive!
Null Hypothesis (H0): Population mean = 50%!
• True positive: The null hypothesis is false (the alternative
Alternative Hypothesis (H1): Population mean ≠ 50%! hypothesis is true), and we conclude that null hypothesis
is false.!
Use sample data to compute t value and p value!
• In other words, the population mean was greater than
Reject null hypothesis (accept alternative) if p < .05!
50%, and we concluded that it was greater than 50%
Population mean = 50%! Population mean ≠ 50%!
because our p value was less than .05.!
(H0 is true)! (H1 is true)!
!
Sample mean is not TRUTH!! Type II error!
significantly different (True negative)! (False negative)!
from 50%!
Sample mean is Type I error! TRUTH!!
significantly different (False positive)! (True positive)!
from 50%! !

False Negative! Not Quite Significant!

Percent  Correct  on   Difference  from   Absolute  Value  of  
Test   Sample  Mean   Difference   • Imagine that we get results like in the previous slide from an
Subject  S1   45   -­‐11.0   11.0   experiment in which the population mean is truly greater than
Subject  S2   47   -­‐9.0   9.0   50%.!
Subject  S3   63   7.0   7.0  
Subject  S4   55   -­‐1.0   1.0   • The sample mean is 56%, which is a fair amount above chance,
Subject  S5   63   7.0   7.0   but the p value is .06.!
Subject  S6   52   -­‐4.0   4.0  
Subject  S7   58   2.0   2.0   • We must accept the null hypothesis, even though the
Subject  S8   65   9.0   9.0   population mean may actually be greater than 50%.!
Sample  Mean   56.0   6.3  
N   8   • However, we don’t ever really know the true population mean,
Std  Dev   7.6   and p > .05, so we can’t be sure that it’s greater than 50%.!
Chance   50  
t   2.24   • We suspect that it’s greater than 50%, but we just don’t have
p   0.06   enough evidence to make a strong claim that it’s greater than
50%.!
• False negative: The null hypothesis is false (the
alternative hypothesis is true), but p is > .05 and we • One cannot usually draw a strong conclusion when p > .05. We
(incorrectly) conclude that null hypothesis is true.! accept the null hypothesis, but we do not have strong
evidence that the null hypothesis is true.!

One-Sample t Test!
False Negative!
Null Hypothesis (H0): Population mean = 50%! • False negative: Even though the null hypothesis is false,
the p value is greater than .05, and you have to accept
Alternative Hypothesis (H1): Population mean ≠ 50%! the null hypothesis.!
Use sample data to compute t value and p value! !
• In other words, you’ve accepted the hypothesis that
Reject null hypothesis (accept alternative) if p < .05! subjects are at chance, but this is an incorrect
conclusion.!
Population mean = 50%! Population mean ≠ 50%!
(H0 is true)! (H1 is true)!
!
Sample mean is not TRUTH!! Type II error!
significantly different (True negative)! (False negative)!
from 50%!
Sample mean is Type I error! TRUTH!!
significantly different (False positive)! (True positive)!
from 50%! !

© S. J. Luck

All rights reserved
6

 p > .05!
p < .05! PSC100Y
Introduction to Cognitive Psychology

• This asymmetry between the null and alternative hypotheses reflects

the fact that scientists don’t want to claim that we’ve found an
effect unless we’re really sure.!

• So we use this p < .05 criterion, which means that we have false
positives only 5% of the time.!
Research Methods 1-4!
• But the flip side is that we get a fairly large number of false
negatives.! One Sample t-Test Equation!

One-Sample t Test!
One Sample t Test!

Percent  Correct  on  

• The first component of the equation is the difference
Test   between the sample mean and chance.!
Subject  S1   68   • The bigger this difference is, the bigger our t value will
Subject  S2   51   [Sample Mean – Chance]! be.!
Subject  S3   60   t =!
Subject  S4   70   [Std Dev ÷ sqrt(N)]! • In other words, if the sample mean is far from chance,
Subject  S5   58  
then the population mean is more likely to be different
Subject  S6   67  
Subject  S7   60   from chance, so this gives us a bigger t value.!
Subject  S8   64  
[62.3 – 50]! • That makes sense: If the sample mean was 90% and
Sample  Mean  
N  
62.3  
8  
t =! chance is 50%, we would have a lot of confidence that
Std  Dev   6.3  
[6.3 ÷ sqrt(8)]! the population mean was greater than chance.!
Chance   50   • But if the sample mean was only 55%, we wouldn’t have
t   5.54  
p   0.001   as much confidence that the population mean was greater
than chance.!

One-Sample t Test!
One Sample t Test!

Percent  Correct  on  

• The second component of the t equation takes
Test   into account the amount of variability among the
Subject  S1   68   test scores, which we quantify as the standard
Subject  S2   51   [Sample Mean – Chance]! deviation.!
Subject  S3   60   t =!
Subject  S4   70   [Std Dev ÷ sqrt(N)]! • The standard deviation appears in the denominator
Subject  S5   58   of our t equation, where it’s divided by the square
Subject  S6   67  
root of the number of subjects.!
Subject  S7   60  
Subject  S8   64  
[62.3 – 50]! • When we have a numerator and a denominator,
Sample  Mean  
N  
62.3  
8  
t =! making the denominator bigger makes the result
Std  Dev   6.3  
[6.3 ÷ sqrt(8)]! smaller.!
Chance   50   • So, if there is more variability among subjects, the
t   5.54   standard deviation will be bigger, and the t value
p   0.001  
will be smaller.!

© S. J. Luck

All rights reserved
7

 One-Sample t Test!
One Sample t Test!
• To see how this works, let’s look at a different
set of data values with the same mean but
• More variability among the individual test greater variability.!
scores means that we’re less certain about
Percent  Correct  on   Percent  Correct  on  
what the population mean is,! Test   Test  
• This means that we’re less certain that the Subject  S1   68   Subject  S1   78  
Subject  S2   51   Subject  S2   41  
population mean differs from chance.! Subject  S3   60   Subject  S3   40  
• So, more variability means a smaller t.! Subject  S4   70   Subject  S4   75  
Subject  S5   58   Subject  S5   48  
Subject  S6   67   Subject  S6   77  
Subject  S7   60   Subject  S7   55  
Subject  S8   64   Subject  S8   84  
Sample  Mean   62.3   Sample  Mean   62.3  
N   8   N   8  
Std  Dev   6.3   Std  Dev   18.1  
Chance   50   Chance   50  
t   5.54   t   1.91  
p   0.001   p   0.098  

One-Sample t Test!
One Sample t Test!

Percent  Correct  on  

• The standard deviation is divided by the
Test   square root of the number of subjects to form
Subject  S1   68   the denominator.!
Subject  S2   51   [Sample Mean – Chance]! • So, as the number of subjects goes up, the
Subject  S3   60   t =!
Subject  S4   70   [Std Dev ÷ sqrt(N)]! overall denominator goes down.!
Subject  S5   58  
• And when the denominator goes down, the t
Subject  S6   67  
Subject  S7   60   value goes up. So a larger N typically leads to
Subject  S8   64  
[62.3 – 50]! a larger t value.!
Sample  Mean  
N  
62.3  
8  
t =!
Std  Dev   6.3  
[6.3 ÷ sqrt(8)]!
Chance   50  
t   5.54  
p   0.001  

One-Sample t Test!
One Sample t Test - Summary!

Percent  Correct  on  

• The t equation has a numerator and a
Test   denominator.!
Subject  S1   68   • The numerator is the difference between the
Subject  S2   51   [Sample Mean – Chance]! sample mean and chance.!
Subject  S3   60   t =!
Subject  S4   70   [Std Dev ÷ sqrt(N)]! • The bigger this is, the more likely it is that
Subject  S5   58  
population mean is greater than chance.!
Subject  S6   67  
Subject  S7   60   • So a bigger difference gives us a bigger t value
Subject  S8   64   and a smaller p value.!
Sample  Mean   62.3  
N   8   • And a small p value is usually a good thing.!
Std  Dev   6.3  
Chance   50  
t   5.54  
p   0.001  

© S. J. Luck

All rights reserved
8

 One Sample t Test - Summary! PSC100Y
Introduction to Cognitive Psychology
• The denominator is the standard deviation
divided by the sqrt of the number of subjects.!
• If there is more variability, the stddev is
bigger, and the t value is smaller.!
• So, more variability means less confidence, a
smaller t value, and a larger p value.!
• The standard deviation gets divided by N, so a
larger N gives you a smaller denominator and a
larger t value.!
• This makes sense, because your sample mean
is a better estimate of the population mean
when you have more subjects.! Research Methods 1-5!
! Two Sample t-Test Overview!

Comparing 2 Conditions!
Condition 1: Rote Memorization!
Condition 2: Elaborative Encoding!
Rote Memorization Group! Elaborative Encoding Group!
• Fish! • Snow!
• Candy! • Home!
• Guitar! • Sky!
• Window! • Heart!
• In this imaginary experiment, we’re going to randomly
• Finger! • Radio! sample 8 subjects from our class of 200 students and
assign them to the rote memorization group!
• Car! • Watch! • And we’re going to randomly sample 8 different
• Printer! • Book! subjects for the elaborative encoding group.!

• Grass! • Towel!

Percent  Correct   Percent  Correct  

Goals of Inferential Statistics!
on  Test   on  Test   • Our goal is to test the hypothesis that the population
Elabora6ve   Rote   mean of the elaborative encoding group is greater than
Encoding     Memoriza6on     the population mean of the rote memorization group.!
Subject  E1   66   Subject  R1   73  
Subject  E2   72   Subject  R2   70   • We don’t have access to the population means, so we’ll
Subject  E3   95   Subject  R3   52   use the sample means to estimate the population means.!
Subject  E4   66   Subject  R4   61  
Subject  E5   85   Subject  R5   69   • We’ll combine the sample means with the standard
Subject  E6   83   Subject  R6   72   deviations and sample sizes to decide if the difference
Subject  E7   75   Subject  R7   50   between sample means is big enough to justify concluding
Subject  E8   61   Subject  R8   60  
that the population means are different.!
Sample  Mean   75.4   Sample  Mean   63.4  
N   8   N   8   • Remember: the goal of inferential statistics is to draw
Std  Dev   11.5   Std  Dev   9.0   conclusions about the populations on the basis of the
t  =  2.32  
 
samples.!
p  =  0.036  
• And usually we’re trying to determine whether the
population means are the same or different.!

© S. J. Luck

All rights reserved
9

 Summary of Results!
Percent  Correct   Percent  Correct  
• After a week, the elaborative encoding group scored an on  Test   on  Test  
average of 75.4% correct on the memory test.! Elabora6ve   Rote  
• The sample mean for the rote memorization group is Encoding     Memoriza6on    
Subject  E1   66   Subject  R1   77  
63.4% correct.!
Subject  E2   72   Subject  R2   73  
• The p value was < .05, so this is a statistically significant Subject  E3   95   Subject  R3   75  
difference between the two groups.! Subject  E4   66   Subject  R4   70  
Subject  E5   85   Subject  R5   70  
• This would allow us to conclude that the population mean Subject  E6   83   Subject  R6   76  
for the elaborative encoding condition is truly greater Subject  E7   75   Subject  R7   67  
than the population mean for the rote memorization Subject  E8   61   Subject  R8   63  
Sample  Mean   75.4   Sample  Mean   71.4  
condition.!
N   8   N   8  
• When we use a t test to conclude that two populations Std  Dev   11.5   Std  Dev   4.8  
differ, we will be right 95% of the time.! t  =  0.91  
 
p  =  0.380  
• A difference like this will occur by chance only 5% of the
time.!

Different Results! • However, this doesn’t mean that elaborative encoding

isn’t a better strategy than rote memorization.!

• Imagine that we had gotten the results from the • The population mean might be quite a bit higher for the
previous slide.! elaborative encoding condition than for the rote
memorization condition, but due to random chance, our
• The sample mean is now only a little higher for samples might not have done a good job of reflecting the
the elaborative encoding group than for the rote population means.!
memorization group.!
• As a result, the t value is a little smaller, and the • In other words, this might be a false negative.!
p value is no longer < .05.!
• We know that the probability of a false positive is only
• We could no longer conclude that elaborative 5%, but we don’t usually know the probability of a false
encoding leads to better memory than rote negative.!
memorization.!
• False negatives are pretty common, so we just can’t draw
a strong conclusion from the lack of a significant effect.!

Dependent and Independent Variables! Independent Variable!

Independent variable: Learning strategy! • The independent variable is what the experimenter
Levels: Elaborative encoding vs. Rote memorization! manipulates.!
Percent  Correct   Percent  Correct  
on  Test   on  Test   • It’s called an independent variable because the
Elabora6ve   Rote   experimenter randomly assigns the subject to particular
Encoding     Memoriza6on     values of this variable, so it is independent of the
Subject  E1   66   Subject  R1   77   characteristics of the subjects.!
Subject  E2   72   Subject  R2   73  
Subject  E3   95   Subject  R3   75   • In our imaginary memory experiment, the independent
Subject  E4   66   Subject  R4   70   variable is the learning strategy that the experimenter
Subject  E5   85   Subject  R5   70   tells the subject to use.!
Subject  E6   83   Subject  R6   76  
Subject  E7   75   Subject  R7   67   • An independent variable has different levels. These are
Subject  E8   61   Subject  R8   63   the values of the independent variable.!
Sample  Mean   75.4   Sample  Mean   71.4  
N   8   N   8   • In our memory experiment, the independent variable of
Std  Dev   11.5   Std  Dev   4.8   learning strategy has two levels: Elaborative encoding and
rote memorization.!

© S. J. Luck

All rights reserved
10

 Dependent and Independent Variables!
Dependent Variables!
Dependent Variable!
Percent  Correct   Percent  Correct  
• The dependent variable is what you measure
on  Test   on  Test   from the subject.!
Elabora6ve   Rote  
• It’s called a dependent variable because it
Encoding     Memoriza6on    
Subject  E1   66   Subject  R1   77  
depends on which level of the independent
Subject  E2   72   Subject  R2   73   variable a given subject is in.!
Subject  E3   95   Subject  R3   75  
• In our memory experiment, the dependent
Subject  E4   66   Subject  R4   70  
Subject  E5   85   Subject  R5   70   variable is the percent correct on the memory
Subject  E6   83   Subject  R6   76   test.!
Subject  E7   75   Subject  R7   67  
Subject  E8   61   Subject  R8   63  
Sample  Mean   75.4   Sample  Mean   71.4  
N   8   N   8  
Std  Dev   11.5   Std  Dev   4.8  

t Test for Two Independent Samples!

PSC100Y
Null Hypothesis (H0): ! !Population mean for Condition 1 =
Introduction to Cognitive Psychology ! ! ! !Population mean for Condition 2!

Alternative Hypothesis (H1): !Population mean for Condition 1 ≠

! ! ! !Population mean for Condition 2!

Use sample data to compute t value and p value!

Reject null hypothesis (accept alternative) if p < .05!

Population means are Population means differ

equal to each other! from each other!
(H0 is true)! (H1 is true)!
Samples means are not TRUTH!! Type II error!
significantly different (True negative)! (False negative)!
from each other! Probability = ???!

Research Methods 1-6! Samples means are

significantly different
Type I error!
(False positive)!
TRUTH!!
(True positive)!
A Closer Look at Populations and Samples! from each other! Probability = 5%!

t Test for Two Independent Samples!

• In statistics, the term
“population” doesn’t actually
Condition 1: Rote Memorization! apply to a specific group of
Condition 2: Elaborative Encoding! individual people.!
• It applies to an infinite number
of hypothetical people.!
Null Hypothesis (H0): ! !Population mean for Condition 1 =
! ! ! !Population mean for Condition 2!
• The population here is an infinite number of hypothetical
Alternative Hypothesis (H1): !Population mean for Condition 1 ≠ people of the type who might sign up for this class.!
! ! ! !Population mean for Condition 2!
• When we randomly choose 8 people from this class of
200 to be in the rote memorization group, we think of
this as being approximately the same as choosing 8
people from an infinite population of people who might
take the class.!

© S. J. Luck

All rights reserved
11

 Rote Memorization Elaborative Encoding
PSC100Y Population! Population!
Introduction to Cognitive Psychology

Rote Memorization Elaborative Encoding

Sample! Sample!

Research Methods 1-7!

False Positives and negatives in Two Sample
Comparisons!

Subject! t Test for Two Independent Samples!

Pool!
Null Hypothesis (H0): ! !Population mean for Condition 1 =
! ! ! !Population mean for Condition 2!

Alternative Hypothesis (H1): !Population mean for Condition 1 ≠

! ! ! !Population mean for Condition 2!

Use sample data to compute t value and p value!

Reject null hypothesis (accept alternative) if p < .05!

Population means are Population means differ

equal to each other! from each other!
Rote Memorization Elaborative Encoding (H0 is true)! (H1 is true)!
Sample! Sample! Samples means are not TRUTH!! Type II error!
significantly different (True negative)! (False negative)!
from each other! Probability = ???!
Samples means are Type I error! TRUTH!!
significantly different (False positive)! (True positive)!
from each other! Probability = 5%!

t Test for Two Independent Samples! t Test for Two Independent Samples!
Null Hypothesis (H0): ! !Population mean for Condition 1 = Null Hypothesis (H0): ! !Population mean for Condition 1 =
! ! ! !Population mean for Condition 2! ! ! ! !Population mean for Condition 2!

Alternative Hypothesis (H1): !Population mean for Condition 1 ≠ Alternative Hypothesis (H1): !Population mean for Condition 1 ≠
! ! ! !Population mean for Condition 2! ! ! ! !Population mean for Condition 2!

Use sample data to compute t value and p value! Use sample data to compute t value and p value!

Reject null hypothesis (accept alternative) if p < .05! Reject null hypothesis (accept alternative) if p < .05!

Population means are Population means differ Population means are Population means differ
equal to each other! from each other! equal to each other! from each other!
(H0 is true)! (H1 is true)! (H0 is true)! (H1 is true)!
Samples means are not TRUTH!! Type II error! Samples means are not TRUTH!! Type II error!
significantly different (True negative)! (False negative)! significantly different (True negative)! (False negative)!
from each other! Probability = ???! from each other! Probability = ???!
Samples means are Type I error! TRUTH!! Samples means are Type I error! TRUTH!!
significantly different (False positive)! (True positive)! significantly different (False positive)! (True positive)!
from each other! Probability = 5%! from each other! Probability = 5%!

© S. J. Luck

All rights reserved
12

 t Test for Two Independent Samples! t Test for Two Independent Samples!
Null Hypothesis (H0): ! !Population mean for Condition 1 = Null Hypothesis (H0): ! !Population mean for Condition 1 =
! ! ! !Population mean for Condition 2! ! ! ! !Population mean for Condition 2!

Alternative Hypothesis (H1): !Population mean for Condition 1 ≠ Alternative Hypothesis (H1): !Population mean for Condition 1 ≠
! ! ! !Population mean for Condition 2! ! ! ! !Population mean for Condition 2!

Use sample data to compute t value and p value! Use sample data to compute t value and p value!

Reject null hypothesis (accept alternative) if p < .05! Reject null hypothesis (accept alternative) if p < .05!

Population means are Population means differ Population means are Population means differ
equal to each other! from each other! equal to each other! from each other!
(H0 is true)! (H1 is true)! (H0 is true)! (H1 is true)!
Samples means are not TRUTH!! Type II error! Samples means are not TRUTH!! Type II error!
significantly different (True negative)! (False negative)! significantly different (True negative)! (False negative)!
from each other! Probability = ???! from each other! Probability = ???!
Samples means are Type I error! TRUTH!! Samples means are Type I error! TRUTH!!
significantly different (False positive)! (True positive)! significantly different (False positive)! (True positive)!
from each other! Probability = 5%! from each other! Probability = 5%!

t Test for Two Independent Samples! t Test for Two Independent Samples!
Null Hypothesis (H0): ! !Population mean for Condition 1 = Null Hypothesis (H0): ! !Population mean for Condition 1 =
! ! ! !Population mean for Condition 2! ! ! ! !Population mean for Condition 2!

Alternative Hypothesis (H1): !Population mean for Condition 1 ≠ Alternative Hypothesis (H1): !Population mean for Condition 1 ≠
! ! ! !Population mean for Condition 2! ! ! ! !Population mean for Condition 2!

Use sample data to compute t value and p value! Use sample data to compute t value and p value!

Reject null hypothesis (accept alternative) if p < .05! Reject null hypothesis (accept alternative) if p < .05!

Population means are Population means differ Population means are Population means differ
equal to each other! from each other! equal to each other! from each other!
(H0 is true)! (H1 is true)! (H0 is true)! (H1 is true)!
Samples means are not TRUTH!! Type II error! Samples means are not TRUTH!! Type II error!
significantly different (True negative)! (False negative)! significantly different (True negative)! (False negative)!
from each other! Probability = ???! from each other! Probability = ???!
Samples means are Type I error! TRUTH!! Samples means are Type I error! TRUTH!!
significantly different (False positive)! (True positive)! significantly different (False positive)! (True positive)!
from each other! Probability = 5%! from each other! Probability = 5%!

t Test for Two Independent Samples!

Percent  Correct   Percent  Correct  
Null Hypothesis (H0): ! !Population mean for Condition 1 = on  Test   on  Test  
! ! ! !Population mean for Condition 2! Elabora6ve   Rote  
Encoding     Memoriza6on    
Alternative Hypothesis (H1): !Population mean for Condition 1 ≠
Subject  E1   66   Subject  R1   73  
! ! ! !Population mean for Condition 2!
Subject  E2   72   Subject  R2   80  
Use sample data to compute t value and p value! Subject  E3   95   Subject  R3   69  
Subject  E4   66   Subject  R4   61  
Reject null hypothesis (accept alternative) if p < .05! Subject  E5   85   Subject  R5   55  
Subject  E6   83   Subject  R6   76  
Population means are Population means differ Subject  E7   75   Subject  R7   62  
equal to each other! from each other! Subject  E8   61   Subject  R8   64  
(H0 is true)! (H1 is true)!
Sample  Mean   75.4   Sample  Mean   67.5  
Samples means are not TRUTH!! Type II error! N   8   N   8  
significantly different (True negative)! (False negative)!
Std  Dev   11.5   Std  Dev   8.5  
from each other! Probability = ???!
t  =  1.56  
 
Samples means are Type I error! TRUTH!! p  =  0.142  
significantly different (False positive)! (True positive)!
from each other! Probability = 5%!

© S. J. Luck

All rights reserved
13

 t Test for Two Independent Samples!
PSC100Y
Null Hypothesis (H0): ! !Population mean for Condition 1 =
! ! ! !Population mean for Condition 2! Introduction to Cognitive Psychology
Alternative Hypothesis (H1): !Population mean for Condition 1 ≠
! ! ! !Population mean for Condition 2!

Use sample data to compute t value and p value!

Reject null hypothesis (accept alternative) if p < .05!

Population means are Population means differ

equal to each other! from each other!
(H0 is true)! (H1 is true)!
Samples means are not TRUTH!! Type II error!
significantly different (True negative)! (False negative)!
from each other! Probability = ???!
Samples means are
significantly different
Type I error!
(False positive)!
TRUTH!!
(True positive)!
Research Methods 1-8!
from each other! Probability = 5%!
Two Sample t-Test Equation!

t Test for Two Independent Samples!

Two Sample t-Test – Numerator!
Elabora6ve   Percent  Correct   Rote   Percent  Correct  
Encoding     on  Test   Memoriza6on     on  Test  
Subject  E1   66   Subject  R1   73   • With a two-sample t test, we have two sample
Subject  E2   72   Subject  R2   70  
means, and the numerator is the difference
Subject  E3   95   Subject  R3   52  
Subject  E4   66   Subject  R4   61  
between the two sample means.!
Subject  E5   85   Subject  R5   69   • In the example from the previous slide, the
Subject  E6   83   Subject  R6   72   numerator would be 75.4 minus 63.4, which is
Subject  E7   75   Subject  R7   50   12.!
Subject  E8   61   Subject  R8   60  
Sample  Mean   75.4   Sample  Mean   63.4   • The bigger the difference between our sample
N   8   N   8   means is, the bigger our t value will be, and the
Std  Dev   11.5   Std  Dev   9.0   smaller our p value will be.!
t  =  2.32  
 
p  =  0.036   • That makes perfect sense, because a big
difference between sample means gives us more
[Sample Mean 1 – Sample Mean 2]! evidence that the population means are
t =! different.!
[Pooled Std Dev ÷ sqrt(N)]!

t Test for Two Independent Samples!

Two Sample t-Test - Denominator!
Elabora6ve   Percent  Correct   Rote   Percent  Correct  
Encoding     on  Test   Memoriza6on     on  Test   • The two standard deviations must be
Subject  E1   66   Subject  R1   73   combined to form a “pooled standard
Subject  E2   72   Subject  R2   70   deviation”.!
Subject  E3   95   Subject  R3   52  
Subject  E4   66   Subject  R4   61   • Other than pooling the two standard
Subject  E5   85   Subject  R5   69   deviations, this is just like the t equation that
Subject  E6   83   Subject  R6   72   you saw before when we had only a single
Subject  E7   75   Subject  R7   50   sample.!
Subject  E8   61   Subject  R8   60  
Sample  Mean   75.4   Sample  Mean   63.4   • And just like before, anything that makes the
N   8   N   8   standard deviations bigger will make our t
Std  Dev   11.5   Std  Dev   9.0   value smaller and make it less likely that the
t  =  2.32  
  effect will be statistically significant.!
p  =  0.036  

[Sample Mean 1 – Sample Mean 2]!

t =!
[Pooled Std Dev ÷ sqrt(N)]!

© S. J. Luck

All rights reserved
14

 Population!
• The goal of experiments in cognitive psychology is to
draw generalizable conclusions about people.!
• These conclusions don’t necessary apply to every
individual person in the population, but we’d like to think
that our experimental results apply to people on average.!
• But we have limited time and resources, so we can’t test
everybody. We can just test a sample of the population.!
• Inferential statistics allow us to draw conclusions from the
population on the basis of a sample of the population.!

© S. J. Luck

All rights reserved
15