Topic: How do subtle language changes influence decisions when judging guilt or innocence?
Statistical Concepts Covered: In this applet, you’ll expand on your knowledge of main effects and interactions (see the activity on Learning), and relationships to evaluate potential biases in interpreting graphs and issues with experimental designs.
Introduction:
The linguistic relativity hypothesis covered in your text presents the idea that “language shapes the nature of thought.” This hypothesis was applied to how we process color and the concept of time, but what about other topics? How might language impact our perception of blame and how much someone should be punished?
The data in this applet comes from research conducted by Fausey and Boroditsky (2009) exploring the impact of language on guilty verdicts, blame ratings, and liability estimates. More specifically, the researchers were primarily interested in testing agentive and nonagentive languaging. Agentive language refers to when a subject is responsible for an action, such as “he ripped”, “she killed”, and “he burned it”. This is in contrast to nonagentive language, which refers to when there is no subject responsible for the action, such as “it ripped”, “it died”, and “it burned”.
How might the wording of criminal trials or accident reports impact a person’s perception of whether the individual involved was guilty or not, how much blame they are responsible for, and the amount the person should be held liable? Let’s explore these questions further using the data from this study.
1) Before conducting their studies, Fausey and Boroditsky (2009) analyzed archival data of around 200,000 criminal trials in London. They evaluated whether particular key phrases were related to a guilty verdict. They were primarily interested in whether the use of agentive language, such as “killed it”, “broke it”, and “burned it”, resulted in more guilty verdicts than nonagentive language, such as “it died”, “it broke”, and “it burned”. In looking at the relationship between these two variables, let’s assume we are using agency as the independent or predictor variable, and percent of guilty verdicts as the dependent or outcome variable. What pattern can we observe in these three scenarios for agentive versus nonagentive language? (Cycle through the three “% Guilty” graphs.)
A. |
B. |
C. |
D. |
2) Based on question 1 and what you know about relationships and causality, can we state that having agentive language caused jurors to deliver a guilty verdict more often than nonagentive language? In other words, does agency impact verdict?
A. |
B. |
C. |
D. |
3) In the previous two questions we evaluated the relationship between agentive language and guilty verdicts when they were used as the independent and dependent variable, respectively. How does our interpretation differ if we use guilty verdicts (guilty or not guilty) as our independent variable and percentage of cases that use agentive language as the dependent variable? (Cycle through the three “% Agency” graphs.)
A. |
B. |
C. |
D. |
4) Based on the previous two questions, can we now state that guilt is really what is affecting the decision about which words to use, not the other way around? In other words, does verdict impact agency?
A. |
B. |
C. |
D. |
5) After conducting their initial analysis of criminal trials, Fausey and Borodistsky (2010) conducted a series of studies to further test the impact of agentive language on blame and liability. In the first study, they gave participants an account of property damage resulting from an accidental restaurant fire. Half of the participants were exposed to the account using agentive language, such as “she had ignited” and “she overturned”. The other half of the participants were exposed to the account using nonagentive language, such as “the napkin had ignited” and “table overturned”. All participants were asked to rate how much they blamed the person involved in the accident on a scale from 1 (not at all to blame) to 7 (completely to blame). Based on a graph of the data, did agentive language appear to impact ratings of blame? (Pick “Blame” for the group to display.)
A. |
B. |
C. |
D. |
6) Another variable the researchers measured in the study described in question 5 was liability. All participants were asked to provide a dollar amount for how much they believed the participant involved in the accident should be held liable. Did agentive language appear to impact liability estimates? (Pick “Liability” for the group to display.)
A. |
B. |
C. |
D. |
7) In the next study, the researchers decided to further evaluate the impact of language by having participates review the same scenarios, half reading the one with agentive language and half reading the one with nonagentive language. They were then presented with a blame rating from an independent panel. This imaginary panel either judged the person to be 1 (low), 4 (middle) or 7 (high) blame. A third of each of the two groups were exposed to a specific blame level, thus a participant was exposed to one type of language and one blame rating. The participants were then asked to estimate the financial liability of the person involved in the accident.
As we can see, there were two main effects (language and blame) and an interaction of these two effects on liability estimates. Remember what we covered in the Learning activity regarding main effects and interactions? The main effect is the impact of a single variable on what we are measuring, while the interaction is the combined impact of two main effects on what we are measuring. Let’s evaluate the main effect of language on liability estimates. Based on the graph, what can we say about the impact of being in the agentive or nonagentive language condition on liability? (Pick “Main effect of Language” for the group. Remember to evaluate only this main effect, ignoring any impact of blame. So, overall, how do the agentive estimates compare to the nonagentive estimates?)
A. |
B. |
C. |
D. |
8) Now, let’s evaluate the other main effect of blame on liability estimates. Based on the graph, what can we say about the impact of the imaginary panel’s low, middle, or high blame ratings that were presented to the participants? (Pick “Main effect of Blame” for the group. Remember to evaluate only this main effect, ignoring any impact of language. So, overall, how do the various levels of blame compare?)
A. |
B. |
C. |
D. |
9) In the previous two questions we evaluated the main effects, now let’s take a look at interpreting interactions. Using the Interaction graph, evaluate and determine whether there is a significant interaction between language and blame rating. Visualize the line connecting the three bars representing the agentive language conditions and the line connecting the three bars representing the nonagentive language conditions. Evaluate whether the lines are more parallel or perpendicular. How would you explain the interaction in relation to liability estimates?
A. |
B. |
C. |
D. |
10) How does our interpretation of the data in these two experimental studies, where participants were manipulated into certain conditions, differ from the analysis of the archived criminal trials? What about how the data was collected impacts the types of conclusions we can make?
A. |
B. |
C. |
D. |
Congratulations! You have completed this activity.