10 A Worked Example
This chapter walks through the complete process — from choosing an anchor theory to handling a reviewer request and writing a cover letter — using a fictional but realistic study in event cognition. The study is invented, but the problems it illustrates are real.
10.1 The study
Imagine you have conducted the following study:
Participants watched short video clips of everyday activities — making coffee, unpacking groceries, setting a table. At unpredictable moments the video paused and participants rated their surprise on a 7-point scale. Pause points were either at event boundaries (moments of natural transition between sub-actions) or within events (mid-action). The dependent variable was surprise rating. The hypothesis was that surprise ratings would be higher at boundaries than within events, because a boundary is the point at which the current event model stops fitting the incoming activity and prediction error rises.
This is a clean, focused study. It has one manipulation, one dependent variable, and one theoretical prediction. It should produce a clean, focused theoretical framing. It often does not — and here is why.
10.2 The first attempt: what goes wrong
Here is how a typical first draft of the theoretical framing might look:
Event cognition has been studied from multiple theoretical perspectives. Event Segmentation Theory (Zacks et al., 2007) proposes that people continuously segment the stream of activity into discrete events, with boundaries occurring at points of maximal change. Situation Models (Zwaan & Radvansky, 1998) describe the mental representations that readers and observers construct during narrative comprehension, including spatial, temporal, and causal dimensions. Narrative Comprehension accounts (Graesser et al., 1994) emphasise the role of goal structures and causal chains in organising event understanding. More recently, predictive processing frameworks (Clark, 2013) have proposed that perception involves continuous prediction and error-correction at multiple levels of abstraction. Each of these frameworks offers important insights into how people understand and remember everyday events.
In the present study, we examined surprise ratings at event boundaries and within events.
This paragraph mentions four theories in five sentences and commits to none of them. The reader finishes it knowing that event cognition is complicated and that something was measured. They do not know what the study is for.
Notice also the final sentence — “we examined surprise ratings at event boundaries and within events” — which describes the method without stating the prediction. The theoretical framing has done no work.
10.3 Applying the framework
Before rewriting, apply the tools from the previous chapters in order.
Step 1: The anchor exercise
This study tests the prediction of Event Segmentation Theory that surprise ratings will be higher at event boundaries than within events, as operationalized by a 7-point surprise rating scale at paused video clips, because a boundary is the moment at which the current event model ceases to fit the incoming activity and prediction error rises, making boundary moments feel more surprising than mid-event pauses.
This sentence is specific. It names a mechanism (a transient rise in prediction error at the point where the event model fails), connects it to a specific outcome (higher surprise ratings), and could only describe this study.
Step 2: Running the two filters
Situation Models:
- F1: Does it make a different prediction from EST for this design? No — both frameworks predict elevated surprise at boundaries, EST because perceptual prediction error rises when the event model stops fitting, Situation Models because the situational dimensions being tracked change at the same moment. They differ in mechanism but not in direction. → Citation only.
Narrative Comprehension accounts:
- F1: Does it make a different prediction for this design? No — these accounts address goal and causal structures in narrative, not perceptual boundary detection in video. → Citation only.
Predictive processing:
- F1: Does it make a different prediction? In this design, no — both EST and predictive processing predict elevated prediction error at boundaries, producing the same directional prediction for surprise ratings. → Citation only, with one sentence acknowledging the connection.
The result: Event Segmentation Theory is the anchor. The other three frameworks get at most one sentence between them.
Step 3: The landscape sentence
Other theories in the field that address event cognition include Situation Models, Narrative Comprehension accounts, and predictive processing frameworks. They appear only briefly in this paper because none of them generates a differential prediction for surprise ratings as a function of boundary location in this design — they fail F1.
This sentence is not for the paper. It is the written record of a principled decision, worked out before drafting begins. A reviewer who later asks why these theories were not discussed will get a precise answer, because the answer was already worked out here.
10.4 The revised framing
Here is what the theoretical framing looks like after applying the framework:
Everyday activities have a natural structure. Actions unfold in sequences — making coffee involves boiling water, then adding grounds, then pouring — and people put the boundary in much the same place (Zacks & Tversky, 2001). These perceived transitions are event boundaries, and they play a central role in Event Segmentation Theory (EST; Zacks et al., 2007).
EST proposes that people maintain an ongoing model of the current event — a working representation of what is happening and what is likely to happen next. While the activity continues as expected, that model predicts well. At a transition, it no longer fits the incoming input, prediction error rises, and it is this transient rise that signals a boundary and triggers an update to a new event model. The theory therefore predicts that boundary moments should feel more surprising than within-event moments, where the current model is still fitting and prediction error remains low.
The present study tests this prediction directly. Participants watched video clips of everyday activities and rated their surprise at pause points located either at event boundaries or within events. If EST is correct, surprise ratings should be higher at boundaries than within events. Other accounts of event cognition — including Situation Models, Narrative Comprehension frameworks, and predictive processing accounts — address complementary aspects of event understanding but do not generate differential predictions for surprise ratings as a function of boundary location in this design.
Three paragraphs. One job each. The first establishes the phenomenon and orients the reader. The second introduces the anchor theory and its mechanism. The third states the specific prediction and disposes of the other frameworks in one sentence.
10.5 The reviews arrive
The paper is submitted and reviews come back. After waiting 24 hours, you sort the comments:
Must do:
- Reviewer 1 asks for clearer reporting of effect sizes → correct and straightforward, add to results section
- Reviewer 2 notes a missing reference for the boundary detection paradigm → correct, add the citation
Negotiate:
- Reviewer 1 suggests the discussion overstates the implications for predictive processing → legitimate concern, revise the discussion to be more precise about scope
Resist:
- Reviewer 2 requests a substantive discussion of predictive processing frameworks throughout the introduction → does not change the theoretical argument; one sentence already acknowledges it, and predictive processing fails F1 for this design
10.6 The reviewer response
Reviewer 2, Comment 3:
The authors fail to engage with predictive processing frameworks, which are highly relevant to the phenomena under investigation. A thorough discussion of Clark (2013) and related work is needed.
Response:
We thank the reviewer for raising predictive processing frameworks. We agree that predictive processing offers an important and compatible account of boundary-related phenomena. However, in the present design, predictive processing and Event Segmentation Theory generate identical directional predictions for surprise ratings at boundaries versus within-event pause points — both frameworks predict higher surprise at boundaries, driven by a transient rise in prediction error. Because our data cannot adjudicate between these accounts, a substantive discussion of predictive processing would imply a contrast our study is not designed to test. We have added a sentence in the introduction acknowledging the relationship between EST and predictive processing frameworks (p. 4, lines 67–69).
10.7 The cover letter
Dear [Editor],
We are pleased to resubmit our manuscript “Event boundaries increase surprise: Evidence from a video-based rating task” for consideration in [Journal]. The paper tests the prediction of Event Segmentation Theory that surprise ratings will be higher at event boundaries than within events, using a video-based paradigm with everyday activities.
We thank the reviewers for their careful reading. The reviews raised three main concerns: the reporting of effect sizes, a missing reference, and the treatment of predictive processing frameworks. We have addressed the first two fully. On the third, we have added a sentence acknowledging the relationship between EST and predictive processing, but have not added a substantive discussion — our reasons are explained in detail in the response document.
The major changes are as follows: effect sizes have been added throughout the results section; the Zacks & Tversky (2001) reference has been added on p. 2; the discussion has been revised to more precisely scope the implications for predictive processing accounts; and a sentence has been added on p. 4 acknowledging the relationship between EST and predictive processing frameworks.
Sincerely, [Authors]
10.8 What this example shows
The writer of the first attempt knew all four theories. What the revised version adds is a decision — a commitment to one theoretical account and a willingness to let that commitment do the work.
The reviewer response and cover letter follow directly from that commitment. Both restate a decision that was made, and written down, before any drafting began.