FSI Category IV versus V is not just a numbering problem

Search for FSI language categories and Japanese will appear as Category IV on some pages and Category V on others. German may appear as Category II or III. The usual explanation is that the same hour bands were merely renumbered.

That explanation is too neat. The State Department's current National Foreign Affairs Training Center page and the legacy five-tier table differ in hours, grouping and stated endpoint. They should not be merged as though only the Roman numerals changed.

What the current official page says

The current page sets one target: an integrated ILR score of 3 in Speaking and Listening. It describes a typical week as 23 class hours plus 17 hours of self-study and publishes four categories:

The Category I range matters. Italian, Portuguese, Romanian, Dutch, Danish, Norwegian and Swedish are listed at 24 weeks, which derives to 552 class hours. French and Spanish are listed at 30 weeks, which derives to 690. “Category I” no longer means one interchangeable 600-hour value.

The final row also shows why this site stores the published class-hour figure rather than recomputing everything from weeks: 88 × 23 is 2,024, while the current page explicitly prints 2,200 class hours. Where the source supplies an hour count, the model preserves it.

Where the familiar five-tier table fits

The widely copied legacy table uses 600 / 900 / 1,100 / 2,200 and often labels the hardest group Category V. It is also commonly presented with a Speaking-3/Reading-3 endpoint.

That is not the same dataset as the current four-category table:

Only the last number is unchanged. The endpoint description also changed. A legacy Category V and a current Category IV may both contain Japanese at 2,200, but that one overlap does not make the two tables interchangeable.

Why this matters

A category number is not enough. “Category IV” can point to a current 2,200-hour group or a legacy table in which the same numeral meant something else.

An hour number is not enough either. A bare 600 may be a legacy baseline; it is not the current Spanish value, and it does not identify the target proficiency or whether self-study was counted.

The source version and endpoint are part of the number. Dropping them turns a reproducible schedule into folklore.

Earlier versions of this site made exactly that mistake: they treated the legacy five-tier scheme as the current four-tier scheme with shifted labels. The correction is not cosmetic. It changes three of the four baselines and removes the assumption that every language in Category I has the same hours.

What this site does now

The current model is keyed to the official page's named languages and current values:

• 552 or 690 within Category I, according to the language's listed weeks;
• 828 for Category II;
• 1,012 for Category III;
• 2,200 for Category IV.

It also records the model version, retrieval date, weekly workload and integrated Speaking-and-Listening endpoint. Legacy category labels remain useful for understanding older articles, but they are historical context, not alternate names for the current data.

When you meet an FSI category elsewhere, ask four questions: which source version, which named language, which hours, and which endpoint? If those are missing, the category cannot safely drive a learning-time estimate.

The current language table shows the per-language values, and the methodology separates official inputs from model choices.