What is distinctive about consistency findings is this: no sentence is wrong when taken alone. Each is correct in itself; the flaw lives in the difference between them. That is why this type is seen only by a scan that looks at the text as a whole.
Definition
It is the same element occurring in different forms within the text. The sentences are correct one by one. The flaw is in the relationship.
Difficulty of seeing it
The finding is not noticed by looking at a single sentence. One has to look at the whole. The method is different.
Terminology inconsistency
The same concept is referred to with different renderings. The reader takes them for two separate things. It is the most common subtype.
Spelling inconsistency
The same name is written in different forms. This breaks matching across documents. The finding is recorded.
Abbreviation inconsistency
Use written out in one place and abbreviated in another produces confusion. The rule is defined in the guide. Its application is checked.
Format inconsistency
Date and number formats may vary within the text. The rule is set once. A scan is run.
Style inconsistency
A difference of tone between sections unsettles the reader. It is common in multi-person jobs. The guide is applied.
Inconsistency of address
A change in the form of address to the reader is a conspicuous flaw. The preference is set at the outset. A check is made.
Consistency within a document
Consistency within a single document is the basic expectation. Measurement starts here. The scope is clear.
Consistency across documents
The different documents of the same client must be consistent too. The scope of measurement widens. The memory is used.
Consistency across periods
The same renderings are used over the years. Even if the team changes, the language does not. The archive does its work.
Risk in multi-person jobs
As the number of people grows, the risk of inconsistency rises. A shared list becomes mandatory. Measures are taken.
Review in split jobs
After the parts are merged, an additional scan is run. The joins are the risky points. The step is added.
Automatic scan
The places where the same source expression was translated differently are found with a tool. The method is the most effective. The scan is mandatory.
Reverse scan
Cases where different source expressions were translated the same are scanned too. Both directions are looked at together. The scope widens.
False positives
Differences made because the context required them are not errors. The findings are filtered. The assessment belongs to a human.
Severity
Severity varies with the document type. It is high in a contract, low in a narrative text. The criterion is adapted.
Severity in contracts
Inconsistency in defined terms makes the provision disputable. Severity is kept at the top. The rule is strict.
Effect of using a memory
A shared memory raises consistency naturally. A memory that is not shared produces deviation. The infrastructure is set up.
Why it is measured separately
This category measures the quality not of individual translators but of the job as a whole and of its preparation; high inconsistency mostly points to a gap in the terminology list or to weak coordination. The root cause is sought there. The view is on the process.
Limits of measurement
If the source text itself is inconsistent, establishing consistency in the target is an intervention. The decision is put to the client. The finding is assessed accordingly.