'Online knowledge revision study 2'
(AsPredicted #110,454)
Download PDF

---

---

1) Have any data been collected for this study already?
No, no data have been collected for this study yet.

2) What's the main question being asked or hypothesis being tested in this study?
This study is closely related to "Online knowledge revision study 1" (see as.predicted file #100923) and aims to repeat this study with an improved material set.
The aim of both studies is to investigate how pictures potentially influence knowledge revision processes during reading.
The study "Online knowledge revision study 1" used 16 text passages introduced by O'Brien, Cook, & Guéraud (2010) and varied whether these text passages were presented without pictures or with pictures. The text passages existed in a consistent version and an outdated (also called qualified) version. In the consistent version, participants read information that was consistent with a subsequent target sentence. In the outdated version, participants first read information that was inconsistent with a subsequent target sentence, but that was revised before the target sentence was presented. This means that participants had to revise earlier information in the outdated version to perceive the target sentence as consistent. As dependent variable, reading time for the target sentence was measured. One weakness of the material set used in "Online knowledge revision study 1" was that the two different text versions (consistent vs. outdated) varied not only regarding the revision of the outdated information but also regarding the alignment between text and picture.
To eliminate this confounding variable, we constructed a new material set. Eight text passages are taken from Cook (2014) and adjusted for our purpose; eight text passages are constructed by ourselves. The main difference between the old and new material sets is that the pictures are combined with the same sentences in both text versions and that both text versions only differ regarding the revision of the outdated information
We hypothesize a significant interaction effect between the factors text version (consistent vs. outdated) and picture (text is presented without pictures vs. text is presented with pictures). In the consistent conditions, we expect no difference in the reading time for the target sentences as a function of the factor picture. In the outdated conditions, we expect faster reading times for the target sentences if the text is presented with a picture compared to if the text is presented without a picture. That means, for presentations without pictures, we expect to replicate the effect observed by O'Brien et al. (slower reading times of the target sentence in the outdated version). For presentations with pictures, we expect a smaller or no differences between both text versions.

3) Describe the key dependent variable(s) specifying how they will be measured.
Reading time for the target sentence in milliseconds.

4) How many and which conditions will participants be assigned to?
There are four within-subjects conditions resulting from crossing the factor text version (consistent vs. outdated) and the factor picture (text presentation without pictures vs. with pictures).
We will use four lists of texts containing one version of each text passage to counterbalance the different versions. Every list consists of 4 text passages with consistent information presented without pictures, 4 text passages with outdated information presented without pictures, 4 text passages with consistent information presented with pictures, and 4 text passages with outdated information presented with pictures. Within each list, the order of text passages is randomized. Participants are randomly assigned to one of these 4 lists.

5) Specify exactly which analyses you will conduct to examine the main question/hypothesis.
Reading time for the target sentences will be analyzed with a linear mixed effects model. Text passages (items) and participants will be included as random effects, text version (consistent vs. outdated ), and picture (without vs. with) will be included as fixed effects. If the interaction of text version * picture is significant, pairwise single comparisons will be conducted to test whether reading time differs for the factor picture within each level of text version. We will also test whether reading time differs for the factor text version within each level of picture.
Further, in case that the interaction is not significant, we will run additional analyses to test whether the single comparisons reach significance.
We will run additional analyses to test whether we were able to replicate the outdated-information effect without pictures reported by O'Brien et al. (2010).

6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations.
Exclusions of participants: Participants who do not agree to have their data processed will be deleted. Participants who do not speak English fluently or who report serious technical issues during participation will be excluded. Moreover, after every text passage, participants have to answer a yes-no comprehension question about text's content. After all 16 text passages, participants complete an old/new picture memory task. Participants with 6 or more wrong answers in the comprehension questions and/or 6 or more wrong answers in the picture memory task (equals ca. 40%) will be excluded.
Exclusion of target sentences' reading times: Per participant, target sentences' reading times below 500ms or above 7000ms will be excluded.

7) How many observations will be collected or what will determine sample size?
No need to justify decision, but be precise about exactly how the number will be determined.
We aim for 400 participants. Due to possible exclusion of participants, we will assess 450 participants.
We conducted a power analysis for the interaction effect of the factors text version and picture. Parameters were estimated based on data reported in O'Brien et al. (2010) and on data from the previous study (see as.predicted #100923).
With 400 participants, we can detect an interaction effect of -85ms with sufficient power (1-β >/= .8).
The parameters were set as follows:
Fixed effects (dummy coding: text version: consistent = 0, outdated = 1; Picture: not present = 0, present = 1): estimate intercept = 1400, estimate text version = 100, estimate picture = -40, estimate text version*picture= -85.
SD of random effects: SD participant = 600, SD text version = 150, SD residual = 600

8) Anything else you would like to pre-register?
(e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?)
Participants will be recruited via Prolific. Screening criteria will be previous study-participation (no participation in our related studies), language (first language = English, English as a fluent language), no language-related disorder, current country of residence (United States of America), and age (between 18 and 35).
Transformation of data: If the reading time is not normally distributed, the data will be log transformed. If this transformation is not adequate, another or no transformation will be chosen.
In line with the analysis procedure by O'Brien et al. (2010), we will conduct the same analyses with a different outlier exclusion. First, we will exclude target sentences' reading times below 750 ms and above 7500 ms per participant. Then, target sentences' reading times above/below 2.5 SD from the condition mean will be excluded.
As an exploratory analysis, we will examine the reading times for the sentence that follows the target sentence (i.e., spill-over sentence). For this analysis, we will exclude spill-over sentences' reading time below 500ms or above 7000ms per participant. Additionally, we will exclude spill-over sentences' reading times that follow after excluded target sentences. We will compute another linear mixed effects model with text passages (items) and participants as random effects, text version (consistent information vs. outdated information), and picture (without vs. with) as fixed effects. Furthermore, we will run the same analyses with the outlier exclusion procedure based on O'Brien et al. (2010); see above.
We will also analyze whether the error rate for the comprehension questions varies as a function of condition. If so, we may increase the allowed error rate.
Moreover, we will check whether the processing time for sentences varies as a function of picture presentation. Therefore, we will compare those sentences presented with a picture in the condition with pictures to the same sentences in the condition without pictures. The focus will be mainly on the second picture, as this picture illustrates the content relevant to the target sentence. We expect longer processing times for sentences provided with pictures compared to sentences without pictures.