Difficulty: Intermediate | Prerequisites: Study One notes (same paper); familiarity with factorial (two-way) ANOVA and interaction effects.
Study Two extends Study One by asking whether the age of a rude social media user changes how people perceive them. It drops the Polite Agree condition and adds a second independent variable (Corey's age: young vs old), creating a 2 × 2 factorial design analysed with two-way ANOVA. This is a step up from the one-way ANOVA in Study One and introduces the concepts of main effects and interaction effects. If you have not yet reviewed Study One from this paper, do that first, since Study Two reuses the same stimulus materials and dependent variables.
Study Two tested whether a rude Facebook user being young or old would change how participants perceived them. It did not, for the most part. Rudeness of tone (Rude Disagree vs Polite Disagree) remained the dominant factor driving impressions of rudeness and anger. Age had a small, significant effect on anger ratings but no significant main effect on perceived rudeness, contradicting the researchers' hypothesis that younger users would be judged more harshly.
Factorial design (2 × 2)
An experimental design with two independent variables, each having two levels. Here: Condition (Rude Disagree, Polite Disagree) × Age (Young, Old) = four cells. Think of it as a grid where each participant falls into one of four boxes.
Main effect
The independent influence of one IV on the DV, averaging across levels of the other IV. In simple terms, does rudeness alone matter? Does age alone matter? Each of those questions is a main effect.
Interaction effect
The combined influence of two IVs that goes beyond their individual main effects. An interaction means the effect of one IV depends on the level of the other. Think of it as: "Does age change the impact of rudeness?" If so, that is an interaction.
Two-way ANOVA (factorial ANOVA)
The statistical test used when you have two IVs and one continuous DV. It produces an F-statistic for each main effect and for the interaction. In simple terms, it is one test that answers three questions at once: does IV1 matter, does IV2 matter, and do they combine in a non-additive way?
Phi coefficient
An effect-size measure for 2 × 2 chi-square tables (the equivalent of Cramer's V when both variables have exactly two levels). In Study Two, Phi = .023, indicating a very small (near-zero) effect for the manipulation check on age recall.
Adolescence and identity formation
The developmental period (roughly late teens to early twenties) when individuals establish their self-concept and are heavily influenced by peer feedback (Koutamanis et al., 2015). Relevant here because the researchers hypothesised younger users would be perceived as ruder, drawing on stereotypes about adolescent impulsivity.
Online disinhibition
The tendency to behave less cautiously online due to perceived anonymity. Combined with heavy social media use, it can amplify rude or aggressive behaviour (Lowry et al., 2016). Think of it as the "keyboard warrior" effect.
Ingroup / outgroup dynamics
The tendency for people to seek out like-minded others (ingroup) and react negatively to those with opposing views (outgroup). On social media, this drives defensive reactions to "fake news" accusations (Kim & Kim, 2019; Schumann et al., 2012).
Study One showed that rudeness of tone drives negative impressions, but the PD condition's ambiguous position raised a new question: does the perceived age of the rude person change how they are judged?
Adolescents and young adults are the heaviest social media users (Jordan, 2010) and have integrated these platforms into daily social life
Adolescence is a period of identity formation, where self-concept develops through peer interactions and self-esteem evaluation (Koutamanis et al., 2015)
Adolescents with poor peer relationships may engage in problematic internet use, including rude or aggressive online behaviour
Online anonymity combined with heavy use leads to disinhibited communication (Lowry et al., 2016)
Social media users cluster into ingroups and will defend their group's beliefs, sometimes seeking out opposing content specifically to argue against it (Schumann et al., 2012)
Adolescents are more likely to overshare online and leave themselves open to public judgement (Lin, 2013)
Age perception prediction: Participants would be more likely to associate rudeness with a younger user, because adolescence is stereotyped as a time of rebellion, immaturity, and disrespect
No age difference expected in the PD or PA conditions
Underlying assumption: Older individuals are seen as more mature and better able to handle criticism
Design: 2 × 2 between-subjects factorial. IV1: Condition (Rude Disagree, Polite Disagree). IV2: Age (Young < 25, Old ≥ 25)
Conditions: Rude Disagree Young (RDY), Polite Disagree Young (PDY), Rude Disagree Old (RDO), Polite Disagree Old (PDO)
Participants: 581 total (40.3% male, 59.7% female; mean age 26.62, SD = 9.588; 57.5% Hispanic, 24.8% African American, 14.3% Caucasian, 2.4% Asian American, 1.0% Native American)
Procedure: Same Facebook stimulus as Study One, but only the two Disagree conditions were used. Corey's age on the profile was manipulated to be either young or old. Six-part survey administered via Qualtrics
Key DVs: Same as Study One: "Corey seems like a rude person" and "Reading Corey's last post made me feel angry" (both 1 to 6 Likert scale)
Additional manipulation check: Could participants remember whether Corey was young or old?
Manipulation check (chi-square on age recall):
χ²(1) = 0.296, p > .05 (not significant)
Phi = .023 (very low effect)
Participants did not differ in their ability to recall Corey's age across conditions. This means the age manipulation did not produce a strong impression
DV 1: "Corey seems like a rude person" (two-way ANOVA):
Main effect of Condition (RD vs PD): F(1, 577) = 109.611, p < .05 (significant)
RD: M = 4.23, SD = 1.492
PD: M = 3.01, SD = 1.377
Main effect of Age (Y vs O): F(1, 577) = 2.147, p = .143 (not significant)
Young: M = 3.70, SD = 1.519
Old: M = 3.49, SD = 1.618
Interaction effect (Condition × Age): F(1, 577) = 4.233, p = .040 (significant)
PDY: M = 3.18, SD = 1.445
PDO: M = 2.75, SD = 1.227
RDY: M = 4.20, SD = 1.417
RDO: M = 4.28, SD = 1.615
The interaction suggests that within the Polite Disagree condition, younger Corey was rated slightly ruder than older Corey, but within the Rude Disagree condition, age made little difference
DV 2: "Corey's post made me feel angry" (two-way ANOVA):
Main effect of Condition (RD vs PD): F(1, 577) = 33.475, p < .05 (significant)
RD: M = 3.29, SD = 1.508
PD: M = 2.59, SD = 1.370
Main effect of Age (Y vs O): F(1, 577) = 5.035, p = .025 (significant)
Young: M = 3.05, SD = 1.487
Old: M = 2.76, SD = 1.456
Interaction effect (Condition × Age): F(1, 577) = 0.457, p = .499 (not significant)
PDY: M = 2.73, SD = 1.435
PDO: M = 2.37, SD = 1.239
RDY: M = 3.36, SD = 1.475
RDO: M = 3.17, SD = 1.561
Both rudeness and age independently increased anger, but the two did not combine in a special way
The hypothesis that age would substantially change perceptions of rudeness was largely unsupported
Tone of response (rude vs polite) remained the dominant driver of both perceived rudeness and anger, consistent with Study One
Age had a small but significant main effect on anger only: participants felt slightly angrier reading a young person's post than an older person's, regardless of tone
The significant interaction on the rudeness DV is noteworthy: within the polite condition, younger Corey was seen as ruder than older Corey, while in the rude condition, both ages were rated similarly high on rudeness. This suggests age expectations may matter more when behaviour is ambiguous
The researchers had expected younger users to be perceived as more immature and therefore ruder, but the data did not support this for overt rudeness. Once someone is clearly rude, age becomes irrelevant to the judgement
Test | Purpose in This Study | Key Output |
|---|---|---|
Chi-square (χ²) | Manipulation check: did age recall differ across conditions? | χ²(1) = 0.296, p = .586 (not significant) |
Phi | Effect size for the 2 × 2 chi-square | Phi = .023 (near zero) |
Two-way ANOVA | Test main effects of Condition and Age, plus their interaction, on each DV | See results above for all F-values |
R² (adjusted) | Proportion of variance explained by the model | "Rude person" DV: R² = .163 (adj. .159); "Angry" DV: R² = .064 (adj. .059) |
The finding that age plays a minimal role in how rudeness is perceived online has implications for content moderation. Platforms cannot assume that rude behaviour from younger users will be judged differently by the public. It also matters for educational interventions: teaching digital civility is relevant across age groups, not just for teenagers.
Students often confuse a non-significant main effect with "no effect at all." The age main effect on the rudeness DV was p = .143. That means the data did not provide sufficient evidence to reject the null hypothesis at alpha = .05. It does not mean age has zero influence; the sample or the manipulation may simply have been insufficient to detect it
A significant interaction does not require significant main effects on the same variable. The rudeness DV showed no main effect of age but did show a significant interaction, which is entirely possible and interpretable
Students sometimes think a 2 × 2 ANOVA requires four separate t-tests. A factorial ANOVA handles all comparisons in one model, controlling for the other factor and testing the interaction simultaneously
The non-significant manipulation check for age (Phi = .023) is often misread as a failure of the study. It is a finding: participants did not strongly encode Corey's age, which helps explain why age had limited impact on the dependent variables
⚠️ Be able to read a two-way ANOVA summary table and identify which F-values correspond to main effects versus the interaction. The source labelled "ConditionIV" is the main effect of rudeness condition; "AgeIV" is the main effect of age; "ConditionIV * AgeIV" is the interaction.
⚠️ Know how to interpret a significant interaction alongside a non-significant main effect. The rudeness DV had a significant interaction (p = .040) but no main effect of age (p = .143). An exam question might ask you to explain what this pattern means.
⚠️ Understand why Study Two used only two conditions (RD and PD) rather than three. The PA condition was dropped to create a clean 2 × 2 factorial, which simplifies interpretation and increases statistical power per cell.
⚠️ The low R² values (.163 and .064) are worth noting. They mean the model explains only a modest proportion of the variance in participants' responses. Other unmeasured factors clearly influence impressions of rudeness and anger.
True or False: Study Two used a one-way ANOVA. (False, it used a two-way (factorial) ANOVA.)
Fill in the blank: The four conditions in Study Two were RDY, PDY, ______, and PDO. (RDO.)
True or False: Age had a significant main effect on the "rude person" measure. (False, p = .143.)
True or False: The interaction between Condition and Age was significant for the anger measure. (False, p = .499.)
Fill in the blank: The effect-size measure used for the 2 × 2 chi-square was ______. (Phi.)
Q: Why did the researchers drop the Polite Agree (PA) condition from Study Two?
A: To create a balanced 2 × 2 factorial design (Condition × Age). Including PA would have created an unbalanced design with three levels on one factor and two on the other, complicating the analysis and interpretation of interaction effects.
Q: The interaction effect on "Corey seems like a rude person" was significant (p = .040), but the main effect of age was not (p = .143). Explain what this pattern means.
A: It means that age alone did not significantly shift rudeness ratings when averaged across conditions. However, the effect of age depended on the condition: within the polite condition, younger Corey was rated ruder than older Corey, whereas within the rude condition, both ages were rated similarly. The combination of the two variables produced a pattern that neither variable produced alone.
Q: Participants reported feeling significantly angrier when Corey was young (M = 3.05) compared to old (M = 2.76). What is the practical significance of this finding, given the R² = .064?
A: The effect is statistically significant but practically small. The model explains only 6.4% of the variance in anger ratings, meaning 93.6% of the variation comes from other factors. The mean difference of 0.29 on a 6-point scale is modest. This illustrates why statistical significance and practical significance are distinct concepts.
Q: What does a Phi coefficient of .023 tell you about the manipulation check for age?
A: It tells you there was virtually no association between the condition participants were in and their ability to recall Corey's age. The age manipulation was either too subtle or not salient enough for participants to encode reliably.
Q: If you were designing Study Three to follow up on these findings, what would you change?
A: Possible improvements include making the age manipulation more salient (e.g., explicit age labels, age-typical profile photos), using a wider age gap, or measuring implicit age stereotypes to see whether those predict differential perceptions even when explicit ratings do not differ.
This connects to developmental psychology (age stereotypes, adolescent identity formation) and to the methods curriculum's coverage of factorial designs. The jump from a one-way ANOVA in Study One to a two-way ANOVA here is a natural progression that mirrors how PSY 3215 builds from simpler to more complex analyses. The interaction effect is a concept that carries into higher-level courses covering moderation, regression, and mixed designs.
Facebook rudeness, age and online perception, factorial ANOVA, two-way ANOVA, main effect, interaction effect, Phi coefficient, manipulation check, adolescent social media behaviour, online disinhibition, ingroup outgroup dynamics, identity formation, PSY 3215, FIU psychology, advanced research methods, between-subjects factorial design