analysis plan: prediction 2

[jessb0t]

Sass et al. (2012) found an interesting pattern in valence priming: positive words are highly effective in priming other positive words, but negative words don't really "prime" anything (that is, RT on lexical decision for target words is similar for both related (negative) and unrelated (positive) targets following a negative prime). More interesting yet, when they compared RTs for positive-prime>negative-target against negative-prime>positive-target, participants actually performed better in the latter. If positive words activate a larger semantic network whereas compensatory mechanisms prevent such extensive network activation following exposure to negative words, one would expect a greater likelihood of disfluency at a positive>negative switch than a negative>positive switch. Alternately, if shifting between valence contexts is akin to task-switching, the surprisal associated with the conflicting valence would be expected to impede performance in either direction, but particularly when hitting a valenced word contradictory to one's current mood state.

• A: positive activation and negative compensation: fewer disfluencies in negative>positive switches than reverse
• B: general surprisal: similar disfluency rates positive>negative and negative>positive
• C: mood-incongruent surprisal: enhanced disfluency rates when hitting switch that conflicts with current mood state

[jessb0t]

Planning

10/21/2021: Comment from JA:

key question: is switching between reading (aloud) positive and negative valence akin to task switching?
main goals:

• determine whether switching from one "tone" in a text to a word on the extreme other end of the valence spectrum requires more cognitive effort than reading within a relatively consistent valence framework
• (if more cognitive effort is, indeed, required) confirm whether a real-word task shows a similar pattern to the lexical decision literature, with positive-to-negative switches being more cognitively burdensome than negative-to-positive switches

Response from GB:

I think it makes sense to start with all passages being negative to positive or positive to negative. This would allow for testing Predictions 1 A/B if you compare reading speed/accuracy for the first half of the passages only, as a function of valence. This will also allow you to test something very similar to what you have written for Prediction 2, but not exactly what you have written. What you have written is to test whether valence in general influences switching. However, a design that lacks all positive or all negative passages will not allow you to test that. Instead, it will only allow you to test whether switch costs are greater for negative to positive or postive to negative. If you have a strong a priori reason to think there will be an asymmetry in the switch cost (and I think you do) then this is fine. But, if you think there is a good chance that the switch cost could potentially be equal, this design is not ideal. We discussed "back up" analyses to compare the switch period to an earlier period in the first passage, which will get at prediction 2 as you have written it, BUT, that is not the ideal design because then switches are confounded with location in the text. With all that said, I think this is all fine, as we can always run a follow up study that includes whole positive or whole negative to explore further.

[jessb0t]

Possible Confounds

**10/23/2021:**The following seem most likely to be potential confounds and should be taken into consideration during analyses:

• word frequency
• word length (both in syllables and letters)
• cumulative surprisal

[jessb0t]

Exclusions

• less than 75% accuracy on challenge questions
• answers "I learned my other language(s) first and English later." with age of English acquisition higher than 5 years