The Scientist

Manuscript ID: WB-SCI-2026-1002
Title: Invariant Butter-Down Terminal Orientation of Falling Toast: A Retrospective Account by an Investigator Who Has Recently Dropped Some
Reviewer: 2
Recommendation: Reject

Summary of the submission

The author asserts that buttered toast, released from any height, in any kitchen, on any occasion, comes to rest butter side down with probability one. The word "always" raises a breakfast grievance to a law of nature, a status previously held only by thermodynamics.

General assessment

The question has troubled the species for as long as it has had both bread and floors. Nothing in this manuscript shortens that period.

Major concerns

1. Anecdote. The evidence is an uncontrolled, unblinded, retrospective case series of the author's own breakfasts. The author is the experimenter, the outcome assessor and the person who has to clean the rug, so independence is not merely compromised. It has never been present.
2. "Always" asserts p = 1. One butter-up landing falsifies it, and the author does not report having looked for one. Please give the observed proportion with a 95% confidence interval. The reviewer cannot compute one, because n is not stated, and suspects the author cannot either.
3. Recall and publication bias. A butter-up landing is a non-event: the toast is picked up, eaten and forgotten. A butter-down landing on carpet is retold for years. The author's memory is a register of carpet incidents, and was compiled by a primate visual system uncalibrated since the Pleistocene that notices the floor only when it is sticky.
4. Confounding by initial conditions. Final orientation depends on release height, angular velocity at departure and how far the slice overhangs the edge before it tips. None of the author's tables, plates or wrists is characterised. Yeh et al. (BMJ, 2018) found no effect of an intervention when the drop was about 0.6 m, and the author has not reported the drop height at all.
5. Correlation is not causation. Butter co-occurs with breakfast, haste and a table, and any of these may be the true exposure. The obvious discriminating arm, toast buttered on both faces, predicts levitation and has not been run.
6. The p-value. A p-value is not the probability that the hypothesis is true (Wasserstein & Lazar, 2016). The reviewer nonetheless requests one.
7. Replication. Dropping more toast does not count: same investigator, same kitchen, same grudge. Of 100 published psychology findings, 36% replicated, against 97% significant in the originals (Open Science Collaboration, 2015). The author's domestic findings are unlikely to fare better.

Minor concerns

• "Lands" conflates first contact with resting orientation. Bounces were not recorded.
• "Toast" is not specified: bread type, degree of browning, butter mass and how evenly it was spread.
• There is no CONSORT-style flow diagram showing how slices were allocated to the floor, or how many escaped allocation by staying on the plate.

Requested additional analyses

1. A preregistered drop trial in which the outcome assessor is blinded to buttering, with butter replaced in the control arm by a placebo spread of identical mass and appearance.
2. High-speed footage recording angular velocity at departure, across a graded series of table heights, each tested in at least three independent kitchens.
3. A meta-analysis of every complaint the author has made about the universe, with a funnel plot to test for asymmetry.

Recommendation

Reject. Resubmission is encouraged once a multi-site, double-blind trial is complete. That will take about nine years, by which time the author will have switched to cereal, which would at least remove the exposure.

Confidential comments to the editor

The reviewer has not dropped toast since 1994.

Manuscript ID: WB-SCI-2026-1002
Title: Systematic Allocation of a Single Shopper to the Least Efficient Checkout Queue: A Retrospective Account by the Shopper Concerned
Reviewer: 2
Recommendation: Reject

Summary of the submission

The author asserts that, across an unspecified number of supermarket visits, they have been assigned the slowest available queue on every occasion. The universal quantifier "always" is offered without qualification. The implication is a mechanism, possibly cosmic, acting on one shopper.

General assessment

The manuscript addresses a question that affects most of the population. Unfortunately it reports data from only one member of it.

Major concerns

1. Anecdote is not data. This is an uncontrolled, retrospective, self-reported single-case series. The author was the subject, the investigator, the outcome assessor and the aggrieved party, and the last role has dominated the analysis.

2. Base rate. With k open tills and no information about them, the probability that the author's queue is the fastest is 1/k. With four tills the author should expect not to be fastest on 75% of visits. "Not the fastest" has been recorded as "the slowest", which is a different outcome. The author has rediscovered arithmetic and reported it as persecution.

3. Size-biased sampling. Long queues contain more people. A randomly chosen shopper is therefore more likely than a randomly chosen queue to be in a long one. Most of the species can make this complaint with equal sincerity, and on the arithmetic most of them would be right.

4. Sample size and confidence intervals. No n is reported. The proportion of visits on which the author's queue was slowest should be reported with a 95% confidence interval. "Always" is a point estimate of 1.00 with no interval, and a single counterexample would falsify it. The author has presumably had that counterexample at some point and does not remember it.

5. Recall and publication bias. A queue that moves normally produces no memory. A queue in which the customer ahead pays in coins and then queries a coupon produces an anecdote that gets retold at dinner. The literature, such as it is, consists entirely of the second kind.

6. Outcome measurement. The author's own queue was watched for the whole wait. The other queues were glanced at when the author was most annoyed. Measurement was neither blinded nor applied equally to each arm.

7. Correlation is not causation. Even if the association held, the author's choice of queue is not random. The author may systematically pick the queue that looks shortest, which is the queue everyone else picks next. The author's queue selection is a confounder, and so is the author.

Minor concerns

1. "Slowest" has no operational definition. It could mean wait time, throughput or the order of departure relative to a man the author had been watching.
2. The supermarket, the time of day and the number of tills are not reported. Yeh et al. (BMJ, 2018) show what happens when context is left out.
3. No CONSORT-style flow diagram shows how the author was allocated to a queue.
4. "I" is an underspecified population.

Requested additional analyses

1. A randomised trial in which a die chooses the author's queue, with a stopwatch on every till, over at least 200 visits.
2. A matched control shopper, blinded to the hypothesis, entering a neighbouring queue at the same moment.
3. A preregistered prediction, written before the author joins the queue, of which queue will be slowest, to measure the author's calibration.

Recommendation

Reject. The author is encouraged to resubmit with properly randomised, timed and preregistered data from several supermarkets. Collecting it will take around nine years, by which point self-checkout will have removed the queues and settled the question.

Confidential comments to the editor

I timed my own queue while reading this, and it was faster.