Mfcttrf

There are words from which you can remove a "chunk", leaving a new word. Like this:

 WISHBONE

 WI SHBO NE

 WI SHBO NE 

WI <poof!> NE

WI NE

 WI NE

 WINE

There are also words that work the other way, for which inserting a "chunk" produces a new word. For example, you can insert the chunk AUTIFI into the word BEER to make BEAUTIFIER.

A "chunk" is a string of consecutive letters. It must consist of at least two letters (no single-letter chunks). It does not need to be a valid English word.

Now, what if I told you there are words from which you can remove a chunk, then insert a different chunk with different letters, and get the original word again?

What the heck am I talking about?!

There are actually thousands of such examples. I'm just looking for a general description of the pattern that creates this strange phenomenon.

(Too easy? Too hard? Try the counterpart subtraction paradox.)

I assume that

"with different letters" means only that the sequence of letters isn't the same, rather than that the (multi)set of letters isn't, because otherwise the thing seems to be genuinely impossible unless there's some sort of lateral-thinking nonsense going on.

In that case

D(ESP)ERATE can lose ESP to make DERATE and then gain SPE to make DE(SPE)RATE again. Or R(ESIGN)ED can lose ESIGN to make RED and then gain SIGNE to make RE(SIGNE)D again.

The general picture here is

that you have words ABCBD and ABD where A,B,C,D are arbitrary strings of letters. The easiest cases (as above) have B a single letter, but I bet there are some where B is longer. At any rate, you're then removing BC and inserting CB or vice versa.

[EDITED to add:]

Yes, B can certainly be longer. For instance, BANYANS can lose ANY and gain YAN or vice versa. Or consider HONEYMOONED; you can lose ONEYMO making HONED and then gain YMOONE to get HONEYMOONED again.

Are you talking about...

words that have the pattern "...X(...X)...", where X is any letter and () marks the chunk to remove? For example, you can take NIN out of FANNING to get FANG, then add NNI into the middle of it to getFANNING` again.

One suggestion

From

DECENT

You can remove the chunk

EC

and get

DENT

But if now you had the different chunk

CE

You can come back to the first word,

DECENT

Here is one

HANDSOME, from which you can remove ANDS to get HOME

and

add jumbled SAND to HOME to get HANDSOME again.

Few more are:

M(ORE)OVER, B(RACEL)ET

I'm pretty sure it's easy to understand what the OP is talking about if you've tried to solve this question before. :)

Solution:

If the word is [Left][Old chunk][Right], the new chunk must be added from somewhere else, splitting the left or right part into two. Let's say it consists of five parts, as in [I][II][Old chunk][III][IV] (I or IV can be empty):

If [I][II][Old chunk][III][IV] = [I][New chunk][II][III][IV],
[II][Old chunk] = [New chunk][II] ([Old chunk] is made up of [V][II], [New chunk] of [II][V])

If [I][II][Old chunk][III][IV] = [I][II][III][New chunk][IV],
[Old chunk][III] = [III][New chunk] ([Old chunk] is made up of [III][V], [New chunk] of [V][III])