ElasticTruth

 

BoM Linguistic Analysis 
The words they use when they are defending the record

It is an analysis of two documents the Bureau of Meteorology published in October 2026 — a website statement and a letter to a parliamentarian — and of what the wording is doing. 

You do not need to know anything about temperature records to follow it. You only need to read the two documents, which are public, and notice the choices.

Letter to Andrew Wilkie is here
BoM statement Oct 22 is here

Every quotation below is from one of those two documents. Nothing is paraphrased.

1. The passive voice that hides who did it

The statement says:

"We have conducted analysis of all underlying data examined in the report, which confirmed a small number of duplicated historical daily temperatures."

Read that again. The Bureau is the subject of the first clause: we have conducted analysis. Then the subject disappears. The duplicated temperatures "were confirmed." Who confirmed them? The sentence does not say. They confirmed themselves.

Compare the next sentence in the same statement:

"We have corrected these values in this year's update."

There, the Bureau is the subject and the verb is active. We have corrected. The diligence is described with an actor. The errors are described without one.

That is not accidental. It is how institutions write when they want to describe their own diligence and someone else's fault. The correction is something they did. The duplication is something that happened.

2. The denominator

The statement says:

"These account for fewer than 200 daily values from a dataset containing more than 7 million values."

Two hundred out of seven million is a very small number. That is the point of the sentence. The reader is meant to think: well, that's nothing.

But the question is not what proportion of the dataset is affected. The question is whether the pattern is consistent with isolated mistakes or with a systematic process. A proportion cannot answer a question about pattern. A dataset can contain 200 errors in 7 million values and still have a problem, if those 200 errors cluster in a way that random mistakes do not.

The sentence is written as though the proportion settles the question. It does not settle it. It avoids it.

3. The word "misunderstanding"

The statement says:

"The claims in the report are based on several misunderstandings."

This is the most effective sentence in the document, and the one worth the most attention. It does not say the report is wrong. It says the person who wrote it did not understand.

That is a claim about the author, not about the data. And it is made without a single example. If the report is based on misunderstandings, the statement could name one. 

It could say: on page 14, the author claims X, but the correct position is Y. It does not. It uses the word and moves on.

A claim that someone has misunderstood, made without specifying what they misunderstood, is not an argument. It is a way of not having one.

4. The hedge

The statement says the records were digitised using punch-card systems "which may have contributed to isolated transcription errors."

"May have" is doing a lot of work. It is not a claim that punch-cards caused the errors. It is a suggestion that they might have, which is enough to give the reader a cause without the Bureau having to commit to one.

The same statement says the Bureau "is confirming if it was the cause." Which means, as of the date of the statement, it did not know.

So the sentence reads: the errors may have been caused by punch-cards, and we are checking whether they were. That is two hedges in a row. A reader skimming the statement will remember "punch-cards" as the explanation. The document does not say that.

5. The closure

The statement says:

"We have corrected these values in this year's update, completed on 19 October 2026."

This is the sentence that ends the conversation. It moves the reader from is there a problem? to the problem is resolved. The date is specific, the verb is completed, and there is nothing left to ask.

The files tell a different story. The duplicates were still present in the public record on 26 October. They were still present on 28 October. They were removed some time between 28 and 29 October.

The closure was performed before the work was done. The sentence says "completed on 19 October." The file says otherwise.

And then there is the name

The website statement and the letter to the parliamentarian describe the same report in two different ways.

The letter, signed by the Bureau's CEO, says:

"I would like to thank Mr Berger for the considerable effort involved in identifying and documenting the anomalies and patterns described in the report."

Mr Berger. By name. A person who did work.

The website statement, issued a few weeks later, says:

"The Bureau of Meteorology is aware of a report produced by a member of the public…"

A member of the public. No name. No work. A category.

The same report. The same author. Two descriptions. The first acknowledges that a named person did the work. The second makes the work anonymous. 

That is not a coincidence of drafting. It is what happens when an institution decides that the person who found the problem should not be named.

What the wording is for

None of these techniques is unusual. Every large institution uses them. The passive voice, the denominator, the word "misunderstanding," the hedge, the closure, the shift from a name to a category — these are the standard tools of institutional defence, and they are used because they work.

They work because most readers do not check. They read the statement, remember the reassuring parts, and move on. The techniques are designed for readers who do not open the files.

This post is for readers who do. Every quotation above is in the two documents. Every claim about the files is checkable. The wording matters, and the wording is on the record.