Bonus Article for Chapters 2, 9,10, and 11 of Research design explained

 

 

You may want to assign thefollowing article:

Landau, M.J., Solomon, S., Greenberg, J., Cohen, F., Pyszczynski, T., Arndt, J., Miller,C. H., Ogilvie, D. M., & Cook, A. (2004). Deliver us from evil: The effectsof mortality salience and reminders of 9/11 on support for President George W.Bush. Personality and Social Psychology Bulletin, 30, 1136-1150.

 

The authors use terror managementto explain President George W. Bush’s popularity. Once studentsunderstand this interesting article, there are over a 100 other terrormanagement articles they can read. The article is easy for your students toobtain (students who buy the book can get it by using the Infotrac®subscription that comes with Research design explained and students may also be able to get it from either http://psp.sagepub.com/cgi/reprint/30/9/1136.pdfor www.apa.org/divisions/div46/images/landau.pdf). Not only will students be able to obtain the article, they will be able to understand it Ņif you give students Table 1 (below).

There are many ways you can usethis article. If you want to assign a reading for Chapter 2 that emphasized thevalue of theory, you could assign the introduction. If you wanted to assign areading for Chapter 9, you could assign Study 1, a simple experiment that findsthat the mortality salience group is more supportive of President Bush than a controlgroup. If you want to assign a reading for Chapter 10, you could assign both Study 2 and Study 3. Study 2 is a three-group experiment that uses three subliminal primes (WTC, 911, and a control prime: 573) to see how these subliminal primes affect the accessiblity of death-related thoughts. Study 3 is a three-group (control [exam prime] group, mortality salience group, and aterrorism [9/11 prime] group) experiment that looks at the effect of those primes on support for President Bush. If you want to assign a reading forChapter 11, you could assign Study 4 (a 2 [prime] X 2 [candidate]) factorialexperiment, as well as Study 3 (because the researchers analyze some of Study 3's results as if it were a prime X political orientation hybrid factorial design).

 

 

 

Table 1

Helping Students Understand the Article

Section

 Tips, Comments, and Problem Areas

Title

Mortality salience: awareness that one may die; reminding people that they will die; mortality salience can be increased by asking  a person to think about one’s own death.

Abstract

Terror management theory (TMT): Terror management theory will be defined in depth in the introduction. However, if you want a short version, the following—a series of quotes and close paraphrases from Jeff Greenberg’s presentation at the session on Experimental Existential Psychology at APA in Hawaii, 2004—may satisfy you. “As animals, we are, because of evolution, oriented toward survival, but, as humans, we know we may die for reasons we can’t control, and we don’t know when we will die.” To prevent ourselves from being terrified about death, we accept  a belief system that has us continuing to be important even after death (e.g., becoming an angel, our memory living on, etc.). 

mortality: that one will die

subliminal: going into the mind without the person knowing it (in this case, because the messages were flashed on a computer screen for such a brief time that the participants were not consciously aware of having seen the messages).

 

Introduction

Opening quote

demagogues: political leaders who become powerful by appealing to people’s emotions and prejudices rather than to reason.

 

 

1st paragraph

posits: proposes; states; suggests

 

2nd paragraph

salience: awareness; attention paid to; “noticeability”

existential fear: fear of no longer existing; often accompanied by anxiety about being meaningless

mediate: come between; be the mechanism for how something happens

 

3rd paragraph (headed “TERROR MANAGEMENT THEORY”)

render: make

propensity: tendency

debilitating: draining; damaging

annihilation: complete destruction

last two sentences:  Rather than be paralyzed by fear, people decide that death is not the end by accepting some belief system that many around them hold (e.g., in the U.S., Christianity) that holds that if they live according to the code (e.g., they are a faithful Christian, Jew, or Muslim), (a) their life will meaningful and (b) they will live on and continue to be important after death—either in an afterlife or by being remembered forever.

 

4th paragraph

worldview: belief system that includes beliefs about why the world makes sense and what one’s role is in the world;  Christianity could be considered a worldview.

operationalized MS: manipulations of mortality salience

control inductions: instructions or stimuli presented to the control group. Good control inductions that are similar to the experimental manipulation improve the construct validity of the experimental manipulation. The authors argue that because the control inductions and  mortality manipulations both produced unpleasant feelings, the effect of the mortality salience manipulation is due to something other than the mortality salience manipulation making participants feel bad.

 

5th paragraph

Implicit: unspoken, unstated

Dual process: two processes, two mechanisms

Proximal: the immediate, initial reaction

Suppression: consciously decide to stop thinking about

Distal: the later reaction; in this case, the second phase

Death thought accessibility dissipates back to baseline level: thoughts about death are not any more present or do not come to mind any more readily than is normally the case.

Summary of paragraph: When confronted with thoughts of your own death, your first line of defense works to push those thoughts out of your mind; the second line of defense works to reassure you that death is not that bad because you are an important person (e.g., someone who God loves).

 

6th paragraph (headed “THE POPULARITY OF GEORGE W. BUSH”)

tenuous: weak, vulnerable

 

 

8th paragraph (headed “TMT AND SUPPORT FOR PRESIDENT BUSH”)

 

allay: calm; relieve; reduce

cosmically significant cultural scheme: very meaningful plan; belief that society is engaged in an important  and noble mission

 

secular: areas outside of religion; not dealing with religion or churches

 

resolute invocation: determinedly calling upon a higher power to either ask for help for an action or to justify that action

 

frailty: physical weakness; the body being vulnerable to injury

 

finitude:  having an end; not lasting forever

 

indefinite perpetuation: lasting forever

 

imbues: gives; fills

 

transcendent: going beyond the limited physical universe

 

primed: reminded (sometimes without the person’s knowledge as in subliminal priming studies)

 

impinge: attack; intrude; try to limit or weaken

 

aversively: unpleasantly, painfully

 

cognitions: thoughts

 

“experimental case study”: “testing an idea experimentally but using a specific well-known stimulus; in this case, the idea that mortality salience increases the degree to which people like charismatic leaders was tested by experimentally manipulating mortality salience but looking at its effect on a specific instance (case) of a  charismatic leader: Bush” (M. J. Landau, personal communication, September 8, 2004).

 

Study 1

 

Method

Procedure

MATERIALS AND PROCEDURE

1st paragraph

 

Attributes: characteristics

 

Filler questionnaires: questionnaires that were not looked at by the researcher. Sometimes, if the manipulation involves filling out a questionnaire, the control conditions will get filler questionnaires to establish that the effect of a manipulation is not due to activities involved in filling out just any questionnaire (e.g., time passing). In Study 1, the filler questionnaires were designed to get participants to believe that the study was about personality traits and social issues.

 

Affective consequences: effects on mood; emotional effects

 

Last line of first paragraph of Materials and Procedure section

The participant is more likely to strongly support traditional beliefs when some time passes (there is a delay, a lag) between when the participant gets the mortality salience manipulation (is reminded that the participant will die) and when the participant performs the dependent measure task. 

 

Final paragraph of Method section

Reverse-scored:  In this case, if a participant responded “1,” that would be scored as a “5”; if a participant responded “2,” that would be scored as a “4”; if a participant responded “3,” that would be scored as a “3”; if a participant responded “4,” that would be scored as a “2”; and if a participant responded “5,” that would be scored as a “1.”

Results and Discussion

Support for the president

Internal reliability: internal consistent (seepages 103-104 of Research design explained) ; responses on one item (question) correlate with responses to other items (questions). If items are internally consistent, researchers may sum up responses to items to produce a single overall (composite) score for each participant.

 

a : Cronbach’s alpha, an index (that can range from 0 to 1) of the degree to which there is consistency between how participants answer one question with how participant answer other questions. A high alpha (above .80) indicates that how a participants answer one question is strongly related to how they answer other questions. If the a  for questions making up a scale is above .80, that scale would be considered internally consistent: people agreeing with one item on a subtest item tend to agree with other items on that same subtest (for more on internal consistency, see page 104 of Research design explained).

 

ANOVA (analysis of variance):  A common statistical technique used to determine whether  one or more predictors were significantly related to an outcome measure (for more, see Chapter 11).

 

2 X 2 ANOVA: an ANOVA (see above) that looks at the relationship of (a) two variables  (in this case, mortality salience and gender) that each have (b) two levels (in this case, control versus mortality salience and men versus women) as well as (c) the interaction between those two variables on (d) an outcome variable (in this case, support for the president). If  participants who have one level of the predictor (e.g., mortality salience) get scores on the outcome measure that are—on the average—reliably higher than participants who had a different level of the predictor (e.g., the control group), researchers would say that there was a main effect for that variable (e.g., “There was a mortality salience main effect.”). In Study 1, if women had scored reliably higher than men on the support of the president measure, the authors would have reported “significant gender main effect.” An interaction between gender and mortality salience would mean that the effect of mortality salience was different for men than for women. For example, if the mortality salience manipulation increased men’s support for George Bush but did not increase women’s support for George Bush, there would be a gender by mortality salience interaction. (For more about interactions, see Chapter 11.)  Because there was no main effect for gender and no gender X mortality salience interaction, the researchers only needed to compare two groups: the control group and the mortality salience group. To make that comparison, they could have used either an F test (which they did) or a t test.

 

ps > .2: not statistically significant; little evidence that the difference between the groups is due to anything beyond chance; even if there is no relationship between the predictor and the outcome variable, there is more than a 20% chance of getting  the chances of getting differences that are as big or bigger than the observed differences.

that are at least as big as was observed.

 

F (1, 93) = 112.48, p < .001: The researchers did an F test (the test associated with ANOVA [ANOVAis described above]). The “1” in parentheses represents the degrees of freedom for the predictor. From this “1,” we know that there were two levels of the predictor (because the formula for the degrees of freedom = levels of treatment  -1). The “93” represents the degrees of freedom in the error term. The “112.48” is the value of “F.” This is a big F value (if the predictor  [mortality salience] was not related to the outcome variable [support for the President], we would expect an F around 1. An F of 4 would have been statistically significant. Our F value is much greater than 4.). This F is statistically significant, as indicated by the “p < .001.” If the researchers had done a t test instead of an F test, instead of reporting “F (1, 93) = 112.48, p < .001,” the researchers might have written  t (95) = 10.61, p < .001.

 

M: mean

 

SE: standard error

 

Control prime condition: in the control prime condition, participants wrote  about what emotions television aroused in them and about what happened to them when they watch ed television.

 

Scale’s midpoint:  “3” was the midpoint of the researchers’ 5-point scale; in this case, a “3” would be neutral, indicating neither agreement nor disagreement.

 

h2 : is a measure of effect size; can range from 0-1;  like r2;  .55 would be considered a large effect size; (for more on h2, see pages 165-166 of Research design explained).

 

Affect.

 

MANOVA: multivariate analysis of variance (see page 212). A MANOVA is like an ANOVA except that whereas an ANOVA typically only involves one dependent measure, a MANOVA involves more than one dependent measure. Thus, if you have several different measures that you do not want to combine into a single measure, you could either do several ANOVAs (one for each measure) or you could do one MANOVA. To oversimplify, the relationship between ANOVA and MANOVA is similar to the relationship between ANOVA and t tests. If you only have one comparison to make, t tests are fine. However, if you have more than one comparison to make—for example, you have a multiple-group experiment—most researchers would discourage you from doing multiple t tests (to see why they would frown on doing multiple t tests, see pages 307-308 of Research design explained). Instead of starting out doing multiple t tests, many researchers believe you should manage your risk of Type 1 errors by first doing a general ANOVA.  Then, if that ANOVA is statistically significant, many researchers would advise following up that ANOVA with t tests. Similarly, if you have one dependent measure, ANOVAs are fine. However, if you have more than one measure, some researchers believe you should do a MANOVA first. Then, if that MANOVA is significant, they would urge you to ahead and do the ANOVAs. (Note that not all researchers believe in this strategy. To understand why, see Huberty, C. J., & Morris, J. D. (1989.) Multivariate analysis versus multiple univariate analyses. Psychological Bulletin, 105, 302-308.

 

PANAS-X: a self-report mood scale; participants make 60 ratings—on a 1 (very slightly or not at all) to 5 (extremely) scale—about “the extent to which” they have felt a certain way  (e.g., sad) “during the past few weeks”; “PANAS” is an abbreviation for “positive and negative affect schedule”; positive affect would deal with positive emotions (feelings of being in a good mood) such as cheerfulness, confidence, peacefulness, and alertness; negative affect would deal with negative emotions (feelings of being in a  bad mood) such as fear, sadness, anger, and guilt.

 

… subscales and aggregate positive ….: The 60-item PANAS-X can yield 13 different scores. There are 11 separate subscales, one for each of 11 different, specific emotions (e.g., a fear subscale, a guilt subscale). In addition, the scale provides two overall, general  scores: one for positive emotions, one for negative emotions.

 

Engender: affect, create; change; cause

 

ANCOVA: An ANOVA that, in addition to using the manipulation as a predictor, also uses participants’ scores on one or more measures as  predictor(s). Sometimes, ANCOVA is used as a more sophisticated form of the blocked design (the block design, used to improve power by controlling for differences between participants,  is described on pages 364-365 of Research design explained). This time, however, Landau et al. are using ANCOVA to show that the mortality salience manipulation did not have its effect by changing participants’ moods. Specifically, the researchers showed that even when participants’ mood states are factored out (by using participants’ emotion scores as predictors), the manipulation still had an effect. Put another way, “ANCOVA can be seen as a form of ‘what if’ analysis, asking what would happen if all cases scored equally on the covariates, so that the effect of the factors over and beyond the covariates can be isolated” (Garson, 2004).

    To be more technical, what usually happens in ANCOVA is a three-step process. First, the covariate(s) are used as predictors in a regression equation to predict scores on the measure. Second, residuals are calculated by taking the score that, based on that participant’s score on the covariate, would be expected to get and subtracting the participant’s actual score from that predicted score. Third, an ANOVA is done using those residuals as scores. The idea is that if the ANOVA still finds a significant effect for the predictor, the predictor’s effect is not completely due to the covariate.

 

To get a better understanding on ANCOVA, consider two hypothetical cases. In the first case, the covariate is a very good predictor of the dependent measure; in the second, the covariate is not correlated with the predictor.

 

 

Case 1

Condition

Score on Covariate

Predicted score on DV

Actual score on DV

Residual (Predicted score  - Actual score)

Low

1

1

1

0

Low

1

1

1

0

Medium

3

3

3

0

Medium

3

3

3

0

High

5

5

5

0

High

5

5

5

0

 

Remember that, in ANCOVA, an ANOVA is done on the residuals.  In Case 1 (above), all the residuals are zero. Thus, the ANCOVA will not find an effect for condition.

 

Case 2

Condition

Score on Covariate

Predicted score on DV

Actual score on DV

Residual (Predicted score  - Actual score)

Low

5

3

1

2

Low

1

3

1

2

Medium

2

3

3

0

Medium

4

3

3

0

High

1

3

5

-2

High

5

3

5

-2

 

In Case 2 (above), it seems like there might be an effect for condition. Note that there is no evidence that the effect of condition is due to the covariate.

 

 

9/11 and support for Bush

 

touchstone: in this case, the authors might really mean “cornerstone,” “centerpiece,”  or “foundation.”

 

Veneration: admiration; feelings of great respect; worship

 

Affinities:  similarities

 

Study 2

1st paragraph

 

Supraliminal: not subliminal; at least at the level of conscious awareness.

 

Summary of 1st paragraph: Mortality salience manipulations have two effects: (a) they make people produce more death-related thoughts and then (b) people reduce the anxiety caused by these death-related thoughts by becoming more committed to a reassuring belief system (e.g., “We are the greatest country on earth”). An important point is that the manipulation is effective when the death-related thoughts are outside of conscious awareness. Note that the two basic effective types of mortality salience manipulations that increase participants’ commitment to comforting worldviews  rely on making sure that participants are not consciously focusing on death:

  1. making participants consciously aware that they will die and then letting time pass and distracting participants so that many of the effects of the mortality salience on the participants’ conscious thoughts have worn off.
  2. presenting the mortality manipulation subliminally (beneath the level of conscious awareness) by manipulations such as showing death-related words on a computer screen so fast that participants are not aware of what the words are.

 

 

 

Method

PARTICIPANTS AND DESIGN

 

911 was to remind participants of September 11.

WTC was to remind participants of World Trade Center.

573 was the control condition.

 

accessibility of death-related words: how easily (quickly, readily) words like “coffin,” “murder,” and “buried” came to mind.

 

MATERIALS

 

Participants thought they were deciding whether two words were similar or not. As you can see from the table below, the real purpose of this task was to expose participants to a subliminal message (“911,” “WTC,” or “573”). Participants did this ask 10 times so they were exposed to the critical prime (“911,” “WTC,” or “573”) 10 times.

 

 

911 condition

WTC condition

573  condition

First word (participant is aware of word)

Word  presented for .356 seconds

Word  presented for .356 seconds

Word  presented for .356 seconds

Second “word” (participant is not aware of this “word” and it differs depending on condition)

“911” presented for .0285 seconds

“WTC” presented for .0285 seconds

“573” presented for .0285 seconds

Second word (participant is aware of word)

Word presented for .356 seconds

Word  presented for .356 seconds

Word  presented for .356 seconds

 

Forward mask: a stimulus that overpowers (covers up) the weaker stimulus coming after it

 

Backward mask: a stimulus that overpowers  (covers up) the weaker stimulus that came before it.

 

Fixation point: where the participant’s eyes are focused on

 

 

Results  and discussion

Checks on awareness of subliminal stimuli: Preliminary

c2 (10, N = 46) =11.42, p > .32. Each participant’s response to the six-option multiple-choice question  questions was categorized as “a,” “b,”
“c,” “d,” “e,” or “f.” In other words, rather than participants varying in the quantity (intensity) of their responses, participants varied in the
quality (type) of their responses With such categorical (qualitative) data, researchers would not use a t test or ANOVA. Instead, researchers would use a test suitable for qualitative data: the chi square (c2 ) test (for more on the chi-square test, see pages 211, 528-529, and 537-539 of Research design explained). The “10” indicates the degrees of freedom. (They had 10 degrees of freedom because [a] degrees of freedom = [rows –1] * [columns –1]; [b] they had a 3-row [the three conditions: 911, WTC, or 537] , 6-column [option chosen on the test] table; and [c) [3-1) * [6-1] = 2 * 5 = 10). The N = 46 means they had 46 participants. The 11.42 is the value of the Chi-square test and that value is, as you can see by looking at Table E-2 on page 539, not statistically significant at the .05 level. To be more specific, even if the manipulation had no effect on participant’s responses to the multiple-choice test, you could get differences at least this big 32% of the time.

 

Accessibility of death-related thoughts

One-way ANOVA: an analysis of variance that used one independent variable (one predictor).

 

2 versus 1 contrast: Comparing two groups against another in Study 2, seeing whether the two experimental conditions (911 prime and WTC prime) differ from the control (537 prime) condition.

 

Linear trend: that the change from group to group would follow a straight line. For example, a linear trend might have involved predicting that the 537 prime group would have the fewest death-related thoughts, WTC prime group would have more death-related thoughts, and that the 911 prime group would have the most death-related thoughts—and the that the difference between the 537 group and the WTC group would be roughly the same as the difference between the WTC group and the 911 group. However, note two important points. First, the statistical test for such a hypothesis would probably only detect whether the 911 group was significantly different from the 537 group. Second, and more importantly,

the researchers did not have hypothesize a linear trend.

 

Orthogonal contrast: comparisons (contrasts) that are independent  (orthogonal) of each other. With a set of orthogonal contrasts, finding a statistically significant effect for one contrast has no implications for whether any of the other contrasts will be statistically significant.

 

Study 3

Both the getting people to think about their own death manipulation and the getting people to think about 9/11 manipulation make people more likely to have death-related thoughts. The next step is to see whether both manipulations also increase support of President Bush.

 

Aversive: unpleasant, provoking strong, negative feelings

 

Nonaversive: not unpleasant

 

Per se: by itself; due specifically to its own unique qualities

 

Negative affect: unpleasant feelings; bad mood

Method

MATERIALS AND PROCEDURE.

 

Filler demographic questions: questions that asked about participants’ backgrounds (age, ethnicity, etc.) but were not of interest to the researchers.

If you have trouble understanding any part of this Method section, refer back to our notes on the Method section for Experiment 1.

Results

Support for the president.

Bush composite: liking of Bush that was measured by summing participants’ responses to the 3, five-point rating scale questions about Bush and dividing by 3. Scores could range from 1 (did not like Bush at all) to 5 (very strong supporter of Bush). To learn more about the scale, see the description of Experiment 1 on page 1140.

 

A 2 versus 1 contrast”: Comparing two groups against a third group; in Study 3, this meant comparing the two experimental conditions (mortality prime and terrorism prime) against the control (exam prime) condition. This could be done by taking the average of the two experimental groups and seeing whether that average was significantly different from the control group mean. To be more specific, the researchers probably (a) multiplied the mean of the mortality prime group by 1/2, (b) multiplied the mean of the terrorism prime group by 1/2, (c) added (summed) the two  results together, (d) subtracted  the control group mean from the sum obtained in step c, and then (e) determined whether  the difference obtained in step d was significantly different from zero.

Note that the MS group’s mean was not statistically significantly bigger than the terrorism group’s mean.

h2 : is a measure of effect size; can range from 0-1;  like r2;  (see pages 165-166 of Research design explained).

 

Note that the researchers went back and did an analysis on each of the three Bush questions (instead of using the measure based on combining all three questions). The results for each individual question were the same as the results for the measure that combined all three questions.

 

Affect.

MANOVA: multivariate analysis of variance (see page 212) When you have several (multiple) scores for each participant, doing one MANOVA is an alternative to doing multiple ANOVAs. To oversimplify, the relationship between ANOVA and MANOVA is similar to the relationship between ANOVA and t tests. As you may recall, if you only have one comparison to make, t tests are fine. However, if you have more than one comparison to make—for example, you have a multiple-group experiment—most researchers would discourage you from doing multiple t tests (to see why they would frown on doing multiple t tests, see pages 307-308 of Research design explained). Instead of starting out doing multiple t tests, many researchers believe you should manage your risk of Type 1 errors by first doing a general ANOVA.  Then, if that ANOVA is statistically significant, many researchers would advise following up that ANOVA with t tests. Similarly, if you have one dependent measure, ANOVAs are fine. However, if you have more than one measure, some researchers believe you should do a MANOVA first. Then, if that MANOVA is significant, they would urge you to ahead and do the ANOVAs.

 

PANAS-X: a self-report mood scale; consists of 60 (1-5) scales; “PANAS” is an abbreviation for “positive and negative affect schedule”; positive affect would deal with positive emotions (feelings of being in a good mood) such as cheerfulness, confidence, peacefulness, and alertness; negative affect would deal with negative emotions (feelings of being in a bad mood) such as fear, sadness, anger, and guilt.

See comments on study 1

 

Engender affect: change mood; cause people to have certain emotions.

 

 

Political orientation.

 

1st paragraph, 1st part

The manipulation did not seem to have an effect on how participants rated themselves on the 1 (very conservative) to 9 (very liberal) scale. On that scale,  “5” would be politically neutral. The participants average score was  5.8, suggesting that the participants were slightly on the liberal side of neutral (To be sure that the participants were significantly above the midpoint, we would need the results of a statistical test showing that the average  score for the participants was significantly different from 5.)

 

1st  paragraph, 2nd part

The researchers used multiple regression (see pages 212 and 530-535 of Research design explained) to see whether  political orientation affected the results. They could have used ANOVA, but multiple regression was a more powerful technique because regression allowed the researchers to use participants’ actual scores (1-9) on the political orientation scale rather than making the researchers group participants into “conservatives”  (participants with scores 1, 2, 3, who might be coded as “1’s”), “moderates” (participants with scores 4, 5, and 6, who might be coded as “2’s”), and “liberals” (participants with scores 7, 8, and 9, who might be coded as “3’s”) and then doing analyses based on the code that was assigned to each participant.  To understand why categorizing participants into a few groups—a necessary step in ANOVA—is not as powerful as using a correlation/regression technique that uses participants’ actual score, read page 172 (especially table 6-7) of Research design explained.

 

As you can see from Figure 2, the significant main effect for political orientation is due to conservatives liking Bush more than liberals (the conservative line is above the liberal line), and the significant main effect for condition seems to be due to the mortality and terrorism groups liking Bush more than the exam group. (Normally, we would want to have post hoc contrasts or planned comparisons to establish which groups differed from each other. However,  making these comparisons was not vital because the authors had already done a very similar analysis (the ANOVA reported in the “Support for the President” section) that had established that, in terms of support for Bush, the mortality and terrorism groups differed from the exam group, but did not differ from each other).

 

 

2nd paragraph

The authors had expected that the mortality salience (MS) manipulation would have the same effect (boosting support for Bush in the mortality and terrorism conditions) on both conservatives and liberals. If the MS had produced the same effect on conservatives and liberals, the two lines (the conservative and liberal lines) in Figure 2 would be parallel. As you can see, they are not. The lines are parallel from the Exam to the Mortality conditions, but the lines are not parallel from the mortality condition to the terrorism condition. This tendency for the manipulation to have a different effect for conservatives than for liberals is called an interaction and it was statistically significant (p = .05).

       Figure 2 is based on roughly 1/3 of the participants. The two-thirds that were closest to the mean (5.8) were not included. Thus, you could view  the graph as comparing participants who scored in the top 1/6th for conservatism versus participants who scored in the top 1/6th for liberalism.

 

b: Like Pearson r (the typical correlation coefficient), b gives you an index of the relationship of the predictor variable with the outcome variable. Indeed, if the regression equation has only one predictor, b will be the same as r. One way of interpreting b is to say that for every increase of one standard deviation in the predictor variable, the outcome variable will change by b number of standard deviations. In this study, a b of -.48 means that an increase of one standard deviation in liberalism would be accompanied by a .48 standard deviation decrease in liking of Bush. Thus, you could look at b is as the slope of the regression line. A negative sign means that line is sloping down—and the bigger the b, the steeper the slope.

        Although the b for political orientation in the 9/11 salience condition seems much smaller (almost 0) than the b’s for the other conditions, the researchers needed to do a statistical test to see which groups differed from each other. They found out that, indeed, political orientation was not as important in the 9/11 salience condition as political orientation was in the other conditions. That is, participants in the 9/11 salience condition liked Bush regardless of their political orientation. Put another way, the Political Orientation X Exam vs. 9/11 Salience interaction was significant because the terrorism prime had a stronger effect on liberals than on conservatives.

 

3rd (final) paragraph

 

 

Spurious: not real; an apparent relationship between variables  that is due to chance or to some other factor.

 

Study 4

Assuaging existential concerns: calming fears about whether one’s life is meaningful; relieving people’s worries that, after they die,  they will either not continue to exist or not be remembered.

 

per se:  by itself alone

 

Assess specificity to concerns about mortality”: The issue is whether  the mortality salience manipulation is just inducing unpleasantness and so its effects would be duplicated by any manipulation that made people unhappy. If that were the case, the manipulation is not having its effect by making people concerned about their own death but merely by making people upset. To address this concern, the researchers used a control topic that, like the mortality manipulation, was unpleasant and dealt with the body: pain. Thus, if the mortality manipulation was only having its effect by making people feel uncomfortable, the mortality group would not differ from the control group.

 

Method

MATERIALS AND PROCEDURE

 

Filler questionnaires: questionnaires that are not used for the responses participants will put on the sheets—those responses will not be analyzed—but for the affect the questionnaire will have on the participant. In this case, the filler questionnaires were to get participants to believe that the study was about personality traits and social issues.

 

MS: mortality salience; awareness that one may die; reminding people that they will die; mortality salience can be increased by asking a person to think about one’s own death.

 

PANAS-X: a self-report mood scale; participants make 60 ratings—on a 1 (very slightly or not at all) to 5 (extremely) scale—about “the extent to which” they have felt a certain way  (e.g., sad) “during the past few weeks”; “PANAS” is an abbreviation for “positive and negative affect schedule”; positive affect would deal with positive emotions (feelings of being in a good mood) such as cheerfulness, confidence, peacefulness, and alertness; negative affect would deal with negative emotions (feelings of being in a  bad mood) such as fear, sadness, anger, and guilt.

 

Results

Opinion survey.

 

(a = .94): Cronbach’s alpha, an index (that can range from 0 to 1) of the degree to which there is consistency between how participants answer one question with how participant answer other questions. A high alpha (above .80) indicates that how a participants answer one question is strongly related to how they answer other questions. If there is a high alpha among questions make up a scale, that scale is internally consistent: people agreeing with one item on a scale item (question) tend to agree with other items (questions) on that same subtest (for more on internal consistency, see page 104 of Research design explained).

 

Composite index scores: the scores computed by adding up responses on the four scales and dividing by 4. Thus, if a participant’s ratings on the four scales were 1, 2, 2, and 3, that participant’s composite index score would be 2 (1 + 2 + 2 + 3)/4.

 

h2 : is a measure of effect size; can range from 0-1;  like r2;  (see pages 165-166 of Research design explained).

 

Note that the main effects were based on comparing the average of two groups representing one level of a variable (e.g., the pain groups) with the average of two other groups representing another level of the variable (e.g., the mortality salience groups). Note that the interaction showed that the effect of mortality salience was different for the George Bush groups than the John Kerry groups. Finally, note that to help us interpret this interaction involving four groups, the researchers used t tests to let look at two groups at a time. The first t test compared the average score of the group of participants who rated George Bush in the pain condition to the average score of another group of participants who rated George Bush in the mortality condition. The second t test compared the average score of the group of participants who rated John Kerry after being primed about pain to the average score of the group of participants who rated John Kerry after being primed about their own death. The third t test compared the average score of the group of participants who rated George Bush in the mortality condition to the average score of the group of participants who rated John Kerry in the mortality condition.

 

midpoint: In this case, the midpoint of the 1-9 scale was 5, which was considered political neutral (neither conservative nor liberal).

 

All ps < .001: The researchers repeated the analyses that they did on the composite measure on the individual rating scales that made up the composite measure. For each rating scale, the pattern was the same as for the composite measure. Specifically, they found differences between conditions on each rating scale were big enough that such differences would be expected to occur less than once in 1000 times if the manipulation did not have an effect. 

 

 

Affect.

MANOVA: multivariate analysis of variance (see page 212). A MANOVA is like an ANOVA except that whereas an ANOVA typically only involves one dependent measure, a MANOVA involves more than one dependent measure. Thus, if you have several different measures that you do not want to combine into a single measure, you could either do several ANOVAs (one for each measure) or you could do one MANOVA. To oversimplify, the relationship between ANOVA and MANOVA is similar to the relationship between ANOVA and t tests. If you only have one comparison to make, t tests are fine. However, if you have more than one comparison to make—for example, you have a multiple-group experiment—most researchers would discourage you from doing multiple t tests (to see why they would frown on doing multiple t tests, see pages 307-308 of Research design explained). Instead of starting out doing multiple t tests, many researchers believe you should manage your risk of Type 1 errors by first doing a general ANOVA.  Then, if that ANOVA is statistically significant, many researchers would advise following up that ANOVA with t tests. Similarly, if you have one dependent measure, ANOVAs are fine. However, if you have more than one measure, some researchers believe you should do a MANOVA first. Then, if that MANOVA is significant, they would urge you to ahead and do the ANOVAs. (Note that not all researchers believe in this strategy. To understand why, see Huberty, C. J., & Morris, J. D. (1989.) Multivariate analysis versus multiple univariate analyses. Psschological Bulletin, 105, 302-308.

 

Univariate main effects: statistically significant effects of a predictor on a single dependent measure

 

ANCOVA: An ANOVA that, in addition to using the manipulation as a predictor, uses participants’ scores on one or more measures as  predictor(s). Sometimes, ANCOVA is used as a more sophisticated form of the blocked design (the blocked design is described on pages 364-365 of Research design explained). In this study, however, the researchers are using ANCOVA to show that the mortality salience manipulation did not have its effect by changing participants’ moods. Specifically, the researchers showed that even when participants’ mood states are factored out (by using participants’ emotion scores as predictors), the manipulation still had an effect.

 

 

Political orientation.

 

Mean of 5.8: The scale went from 1 (very conservative) to 9 (very liberal). Thus, “5” would be politically neutral.

 

The researchers used multiple regression (see pages 212 and 530-535 of Research design explained) to see whether political orientation affected the results. The researchers could have used ANOVA, but multiple regression is a more powerful technique because regression allowed the researchers to use participants’ scores (1-9) on the political orientation scale rather than grouping participants into “conservatives”  (participants with scores 1, 2, 3), “moderates” (participants with scores 4, 5, and 6), and “liberals” (participants with scores 7, 8, and 9).  To understand why categorizing participants into a few groups—a necessary step in ANOVA—is less powerful than using a correlation/regression technique that uses participants’ actual scores, read page 172 (especially table 6-7) of Research design explained.

 

 

The “unsurprising Political Orientation X Leader interaction” is unsurprising because it indicates that conservatives rated Bush more highly than Kerry, whereas liberals rated Kerry more highly than Bush. 

 

The predicted prime X Leader interaction was the result of mortality salience prime increasing ratings for Bush but decreasing ratings for Kerry.

 


three-way interaction”:  An interaction between two predictors  is a two-way interaction. The two –way interaction means that we have to qualify statements we make about the two predictors involved in that interaction. Thus, because of the prime X leader interaction, we could not simply say that the mortality salience manipulation increases liking for a leader or that Bush was rated more highly than Kerry. A three-way interaction involves three predictors (in Study 4, the three predictors were prime, leader, and political orientation). If there had been a three-way  interaction , that prime X leader X political orientation interaction might have qualified the two-way (prime X leader ) interaction. Specifically, a significant three-way interaction might have meant that the prime X leader interaction occurred only for conservatives. However,  rather than obtaining a significant three-way interaction (indicated by a p value of less than .05), the p value was greater than .46. Thus, there was no evidence of a three-way interaction.

 

Discussion

 

General Discussion

Third paragraph

The failure to find an effect of the manipulations on political orientation could be due to the measure of political orientation being insensitive. (To learn more about sensitivity, see Chapter 5).

 

Sixth paragraph

 

Ingroup cohesion: liking of people of your own group; group members being attracted to and getting along with each other

 

Shared external threat: common enemy

 

Superordinate identities:  sense of being part of a general group; in this case, seeing oneself as being an American

 

Ingroup solidarity: a high degree of ingroup cohesion (ingroup cohesion is defined above)

 

Parsimonious: simplest

 

The authors suggest the following possibility:

External threats  à increased death-related thoughts  à ingroup favoritism. To use the language of Chapter 2, they are proposing the death-related thoughts may mediate the relationship between external threats and ingroup favoritism.

 

Political Implications

 

Salient: noticeable; obvious

 

Rational choice: logical decision

 

Irrational symbolic protection: people feeling that the things they believe in (themselves, as well as their ideals) have been shielded from harm.

 

Bode well: be a positive sign

 

Attenuate: weaken; reduce; decrease

 

Rife: full of; dominated by

 

Imminent: likely to happen soon

 

Rhetoric: speech or writing that persuades people but may lack meaningful substance

 

Psychological equanimity: calm; not overly fearful; not being stressed out; steady

 

 

 

 

 

 

 


To Chapter 2 Menu

To Chapter 9 Menu

To Chapter 10 Menu

To Chapter 11 Menu

Back to Bonus Article Menu

Back to Research Design Explained Home Page