APT . . .

A phrase is given in all capitals. The solver must find another word or phrase that forms a base of the appropriate type with the given phrase and for which the phrase is an apt clue. For example: “FEAR DOTH THREATEN THEE” clues the consonantcy Friday the thirteenth. The “apt” modifier can apply to other base types also: for example, NOT HERE / no, there is an apt heteronym; Loki-5’s “HE’S A KING” is an apt letter bank (for Genghis Khan).


The solution: après moi le déluge.

Note that an anagram is just an apt transposal (the type from which apt flats were generalized). Only flat types relating pairs of words that are quite different make interesting apt puzzles; an apt deletion, for example, wouldn’t be much of a challenge. So in practice, only apt consonantcies and letter banks have appeared.

Apt flats were invented by Ai.

 BIGRAM . . .

Instead of single letters, bigrams (two-letter groups) are the basic units of these puzzles. An example is the bigram reversal ONE = se-ra-ph, TWO = ph-ra-se. A bigram deletion: ONE = impetuous, TWO = impetus.

For more information, look up the next word in the puzzle’s name.

 TRIGRAM . . .

Instead of single letters, trigrams (three-letter groups) are the basic unit of these puzzles. They work just as bigram puzzles do, but they use three-letter instead of two-letter groups.

For more information, look up the next word in the puzzle's name.


Sometimes a flat is solvable after reading just the first line or two. Flats we never finished reading present only that first line or two. These shortened flats must be solvable based on only the information given, which often doesn’t include cuewords. This means that the clueing presented in the shortened verse must be complete.

FWNFR should not be used just as a way to present bases that the composer didn’t feel inspired to versify. “Flats we never finished writing” are often uninteresting, and are likely to be underclued. FWNFR tend to work best with simpler bases, but the existence of FWNFR as a type should not discourage composers from writing full flats on those bases.

FWNFR were introduced by Xemu.

The ugly crook was on the lam . . .

The solution: hideous, hideout

As a Pueblo, my home . . .

The solution: adobe, abode


Overloaded flats are puzzles in which a cueword can stand for any of two or more solution words.

It’s time to start the bacchanal;
Everyone disrobes. I count
Eleven folks (an odd amount)
Who’ll cause the bed to rock, in all.
I’m feeling just a little shy,
So I check out the dinner spread.
It’s just some veggies on a bed
Of stir-fried noodles, which I try --
It stinks! But it’s the only food,
So I fill up a plate and wander
Back to the bedroom, where I ponder
What to do. It might be rude
To cut in on a busy pair,
And everyone seems occupied . . .
Back to the kitchen. I must confide,
I’m not enjoying this AFFAIR.
=Lunch Boy

The solution: lo mein, love-in


Puzzle variations in which sounds are the basic unit instead of letters. An example of a phonetic beheadment: ONE = quest, TWO = west; a phonetic charade: ONE = lox, TWO = myth, WHOLE = locksmith; a phonetic curtailment: ONE = cute, TWO = queue. In a phonetic curtailment, TWO can be longer than ONE; what counts is the number of sounds, not the number of letters. In fact, the greater the change in spelling, the more interesting the base.

A flat is phonetic if any part is phonetic, even if some parts are not. For example: a phonetic charade, ONE = weigh, TWO = ding, ALL = wading. No flat is labeled phonetic if it can also work as a letter-based puzzle. Mite/rite is both a first-letter change and a first-sound change, but it is called a first-letter change. Mite/right has to be a first-sound change.

The underlined letters, as pronounced in the following words, stand for single sounds: loud, chin, whale, joke, sing, coin, ship, thin, this, vision. These sounds are indivisible. On the other hand, these represent two sounds: few and curable. The y and oo sounds are separate.

Modern pronunciation varies widely; you’re likely to encounter phonetic flats that don’t work in your speech. They are still legal and valid as long as MW substantiates their pronunciation. These pronunciation variations are common enough to be acceptable without comment in flats:

w = hw. For most Americans, where and wear are homonyms.

For many Americans, T and D have the same sound between two vowels if the second vowel is unstressed; latter and ladder are homonyms.

ä = ŏ. For most Americans, bother rhymes with father. For a minority, cot is a homonym of caught.

Before a vowel sound, ar = er = ār. Many Americans pronounce Mary and merry the same, and a large minority pronounce marry the same as the other two. 11C is not consistent on this point; NI3 explains it. Phonetic flats based on this pronunciation are not tagged “NI3 pronunciation.”

r = schwa or nothing. For many r-droppers, card and cod are homonyms, as are manners and mannas.

schwa = short i, most often unstressed. For many, language and languid are a last-sound change.

For more information, see the next word in the puzzle’s name.

 PICTURE . . .

The solution words are clued by elements shown in an illustration rather than in verse. Note that the solution words need not be nouns; actions and qualities of things or actions depicted can also serve as clues to verbs, adjectives, or adverbs.


For more information, look up the next word in the puzzle’s name.


A variation of a puzzle in which you have to reverse the result to read the final solution.

If the puzzle involves a single operation on one word or phrase to produce another, you reverse the result of the operation to get the second part. Thus, ONE = petal, TWO = late is an example of a reversed beheadment (first you behead, then you reverse).

If the puzzle involves breaking a word or phrase into two or more parts, you assemble the parts first, and then reverse the result, to get the whole word or phrase. Thus, ONE = red, TWO = rum, WHOLE = murder is an example of a reversed charade (first you join the parts, then you reverse the result).

A reversed rebus is called a suber.

A reversed heteronym is called a mynoreteh.

For more information, look up the next word in the puzzle’s name.

 WELDED . . .

A word or phrase is divided up into pieces (not usually words) that form a base of the given type. The solution need not be (and usually isn’t) a dictionary entry, but the number of dictionary entries in the answer shouldn’t be greater than the number of pieces in the base. Example: the welded second-sound change tech stocks.

Step up and vote for your favorite newt!
(But do not stuff the ROOT BEER BOOT.)

The solution: axolotl ballot box (because axolotlb transposes to allotbox). Note that the solution consists of two dictionary entries even though it contains three words, because ballot box is an entry phrase. (Note that the cuewords ROOT BEER BOOT are also a welded transposal.)

Most welded flats divide the solution into two pieces, but occasionally a welded flat with more pieces is seen-either because the given type of flat always has more than two solution words (a word deletion, for example), or because the type of flat sometimes has more than two (a transposal).

The place where the pieces are joined should fall within a word and not between words of the solution phrase (hence the term “welded”).

Welded flats are an Italian puzzle type that Hot introduced to the NPL.

This page was last updated on Thursday, June 18, 2015. /webmaster