[BLANK_AUDIO] So let's look at how context effects alter answers to attitude questions. Context, in general, will influence the response, when it affects what enters into consideration when comprehending a question and forming judgements and evaluations. So again, think of the response process model that we talked about early in the class. And context effect can enter at any of these stages and alter ultimately the response. It's important that you think of context effects in a very wide way. So there many aspects to the survey that can form a context. It can be the subject of the questionnaire, the interviewer, the interview setting, instructions that you give in the questionnaire or through an interviewer, pictures that accommodate a question, wording of a question, response options, or question order itself. We'll have a few first examples, but later on also when we talk about layer and modes, we will come back to these context effects. Let's look at the first one, context effects comprehension. There are assimilation effects. That means the respondent includes prior questions into the interpretation in the understanding of a current item. There's an example here from a paper here from Tourangeau and Rasinski from 1988. "Do you favor or oppose passage of the Monetary Control Bill?" 27.5% reported in favor when this question was following a question on inflation, whereas 12.5% reported favor when following a neutral question. The Monetary Control Bill is related to anti-inflation, so this context has changed the amount of favorable answers to this particular question. A second way in which context effect can enter comprehension is the contrast effect. So that means the respondent excludes prior questions in interpreting and understanding the current item. We'll show you an example later where this context effect happens when we talk about something very specific versus something that is very general and the content of that question is related to happiness and marriage. I'll come back to that, but remember this context effect when we get to that example. Both of these have implications for questionnaire designers. They do mean, just like we said initially, that you should avoid ambiguous and vague terms. And you can't be specific enough about the attitude object that you want to talk about. At the retrieval stage the context prompts the accessibility or retrieval of certain consideration. That can happen both in assimilation and in contrast, where in contrast, it leads to excluding the retrieval of certain considerations. On the adjustment stage, again this has to do with the inclusion/ exclusion model Schwarz and Bless talk about this in their 1992 paper, it depends a little bit on constructing targets and standards. So, the retrieval of information and the application of this information has to do with the accessibility in memory. Some information is chronically accessible, others is temporarily accessible and context dependent. So, let me put some meat to this statement here. The representation of the target means that the information activated by context is included in representation of the target, that would mean an assimilation effect. The information can also be activated by context and is excluded from representation of the target, which is subtracted and creates this context effect. If the information is activated by context, is included in representation of the standard, then comparing based contrast effects will appear. So assimilation, substraction-based contrast effects, and comparing based contrast effects. Let me give you an example, it's rather technical. If the context is trustworthiness of Richard Nixon, meaning a question ahead or some information given about Richard Nixon, which often has a low rating, because of the scandal back then, and the target of the question is trustworthiness of politicians, then a low rating of politicians will appear due to the assimilation effect because Nixon is incorporated into the representation of politicians. Whereas the target is trustworthiness of a different politician, that is seen favourably, you'll get a high rating of that politician, in this example of Newt Gingrich, because of the contrast effect, because Nixon is incorporated into their representation of the standard which then lowers the bar. So here is an example for this effect, a question about politicians in general and about a specific politician and on the y axis a scale for trustworthiness that was used in this paper by Bless and Schwartz in 2010. If the scandal was brought to mind, those are the black bars, you see see much higher ratings of specific politicians then when the scandal was not brought to mind. But the difference is particularly strong between asking about politicians in general and asking about a specific politician. Context effects can also enter into reporting and mapping. Again, assimilation effects means that you get consistent responses to questions that are grouped together whereas contrast effects will lead to the opposite. So, if Nixon is mapped as the anchor for "least trustworthy", then you have these anchoring effects that change the rating or the scale points that you will pick for later questions. So, let's summarize these context effects for attitude questions. They can enter into the comprehension stage, because the content of preceding questions may influence respondents' interpretation of later questions, at the retrieval stage affecting accessibility, at the judgement stage in affecting information that's used to answer the proceeding questions is now more accessible in memory and used to construct either the representation of a target or the representation of a standard, and then reporting issues as in consistent reporting of questions that are grouped together and anchoring of rating scales. This has impact on questionnaire designers. First of all, there's a threat to the interpretation of our data and to the generalizability of the findings, both substantive and methodological. Think for example in trend analyses. So you field the same question time after time after time in the General Social Survey or some other omnibus type survey where various questions are collected, often by different sponsors, you might have problems maintaining the context over time. But even if you can manage to have the same context time after time after time, it could be that the meaning of the question changes over time. Also there might be an interaction with demographic characteristics, so the subgroups in your population. And, you know, although we have talked in these examples mostly about context in the sense of prior or preceding questions, keep in mind this can also be an interviewer, a setting, or a historic context that creates these effects. That's why it's important that you understand or go, anytime you're construct a questionnaire, that you go through the stages of answer formation, comprehension, retrieval, judgement, and response formation. And think about, "Is there something in this set up, in the interview in this context that might affect a particular stage in this?" So what to do? You can borrow questions when possible, assuming those questions worked in other places and not too much time has passed, and the like, you have control under the context and the like. Regardless there's lots of pre-testing needs to uncover context effects, data checking during the field, maybe behavioral coding, lots of analysis to check for context effect afterwards quantitatively. If you can, I mean ideally, you want to build in some experiments, that's often a good idea, some, as they call it A-B testing or split ballot experiments. But then also you can build in variables that can mediate the impact of context effects. And, you know, those can be variables related to cognitive capacity or cognitive load, variables related to motivation, any type of context distractions, interruptions. Be cognizant of those as you design the questionnaire and the instrument as a whole. In the next segment, I will come back to this example of specific versus general evaluations that we touched on earlier, when we talked about attitude effects.
