Would other PCs eventually notice two rogues talking in Thieves' Cant?

Question

I have two PCs in the party that know Thieves' Cant and they constantly talk to each other and pass secrete messages to each other via Thieves' Cant. Another player thought that because Thieves' Cant is being used so often he should have a chance to recognize that something more is being said, he has not suggested he would know the hidden message but that perhaps notice that a message or something was being added to the normal conversation.

The PHB states:

... a secret mix of dialect, jargon and code allows you to hide messages in seemingly normal conversation. Only another creature that knows thieves' cant understands such messages. It takes four times longer to convey such a message than it does to speak the same idea plainly.

His reasoning is that mixing in the jargon, dialect, etc would make the conversations seem odd and allow him to start figuring out they are not normal conversations after a while. He also stated that Thieves' Cant can only be understood by the listener if they were clued in that there was a secret message but he did not provide any grounds for it.

Is he right on these two points or is Thieves' Cant done in such a way that anyone knowing it will pickup the message with out prompt and anyone listening in over and over on multiple conversations would never pick it up?

Answer 1

No, because the point of Thieves' Cant is that observers can't tell it's being used. The conversations being lengthy isn't a giveaway either, since that just means that the conversations the messages hide in are normal length and the Cant message is ¼ the length of the carrier message. For longer Cant messages, just talk more about mundane things like your sister's guild training or that horse you were admiring in the last town.

---

But even if we assume (or it's been house-ruled) that a listener can tell when Cant is and isn't being used, just not what the message is…

Still no, not if they're using Thieves' Cant properly. To use Thieves' Cant properly, you have to use it all the time with your fellow rogues, just as part of normal casual conversations with them, so that even if someone twigs that the conversations are “not normal”, you don't give away why you're using Cant by presenting an obvious pattern. If the only time a pair of rogues use Thieves' Cant is when they're up to no good, even the most Lawful Stupid paladin can figure out that “unusual conversations = I should start interfering!”

When it's used all the time between the two rogues, even someone who has learned to recognise the use of Cant won't be able to tell that they're up to mischief. It's definitely not grounds for making intuition checks or anything to figure out that something is up. When every conversation between the rogues is “not normal”, then it stops being suspicious.

Part of the point of Thieves' Cant is that knowing that there's a secret meaning is not the same thing as knowing that there's an important secret meaning. Two rogues bantering using Cant could as easily be talking about inconsequential things — ribbing each other, cracking jokes, discussing last night's hangover — as they could be talking about real plans. When all the talk between two rogues is laced with impenetrable jargon and slang, nobody can tell when to be concerned and when it's innocent.

The same principle is observed in modern cryptography — if you only encrypt sensitive messages, then anyone monitoring you knows when you're discussing sensitive topics and can probably figure out the message based on context, like time, other activity, and receiver.

Thieves who speak in Thieves' Cant only when they're discussing doing something improper aren't very good at being thieves.

Answer 2

Eventually, yes, particularly if you as the DM rule that.

If you want to explain why, you can make it as simple as an active Insight or Perception Check, or a Passive Insight or Perception check (or use an Intelligence based skill like Investigation if that makes more sense to you to use Intelligence to address language issues). Caveat: the eventually piece must play a part in that exposure over time is required to begin to get a clue.

Two RL examples where I've seen this happen when someone senses that the conversation is in the same language but they aren't getting it:

1. Cockney Slang. If you listen to rhyming Cockney Slang, you will eventually get the idea that something other than what is being said is being said, as some sentences end up sounding ludicrous. Until you have a given piece of slang explained to you, it still won't make sense, but you'll notice there is something off about the communication.

2. Military Jargon. I spent over two decades in the military, and developed a habit of using Military slang and jargon in my everyday speech. Now and again when having a discussion I'd confuse whomever I was speaking to, or someone in a group who wasn't in the 'in' group I was. They often commented or noticed that they didn't understand what I just said. My wife got used to this and often got me to explain in normal conversational speech what I was talking about. She even used a hand gesture that was part of our slang -- the 'cut' gesture, hand parallel to the ground, fingers pointing toward throat, wave hand back and forth with wrist -- to signal to me that I needed to cut the slang and talk like a normal person.

Another part of Thieves' Cant used to be hand gestures, and other body language signals. You could equate some of this to RL and how some people flash gang signs. If you aren't clued in, you may not grasp what just happened. Over time, though, you begin to pick up on that cue even if you don't understand its meaning.

Your Thief player is partially right in this regard:

Is he right on these two points or is Thieves' Cant done in such a way that anyone knowing it will pickup the message with out prompt and anyone listening in over and over on multiple conversations would never pick it

But he's also partially wrong. Over time, if a similar topic is being talked around, and a person is observant, they can often decode a message -- even if decoding is not 100% complete.

Recommendation for a Ruling:

As above, apply a Wisdom or Intelligence based ability checks (DC assigned by you since some messages are simple and some are complex) for a mechanical way to sort out a partial or total grasp of a message topic that has been noticed before. For an initial message where cues had not been observed before, it's far less likely for someone not in the 'in' group will pick up on the cues.

Answer 3

Without supernatural intervention (relevant Divination spells/successful appeals to Knowledge deities), I would rule "maybe," subject to a series of Hard/Very Hard Insight/Intelligence and Perception/Wisdom checks, over several tries for the same subject. (Friendly reminder, in the PHB, Medium DC = 15, Hard = 20, Very Hard = 25; Nearly Impossible = 30). Also, part of the Cant would be the cues to signal a secret message is being sent—and these would just be part of the normal conversation (consider encryption headers wrapped around the un-encrypted message, except in this case, the headers are also intelligible and seem to be the actual message).

Think of TC as being a language where non-speakers have, at best, superficial grasp of the phonetics/syntax, but know nothing of the underlying semantics. Thus, the first check would determine whether the non-Rogue PC has a reasonable shot of knowing about Thieves' Cant (TC's) existence and purpose—maybe a Medium/Hard History/Intelligence check? Alternatively, I would allow a Hard/Very Hard Insight/Intelligence check, to determine something is amiss. Ultimately, the PC can't be actively trying to learn about something he/she don't know exists/is going on.

To actually decipher the message, consider the following: you have just been dumped in a place where you don't speak the language—you can understand the sounds the speakers are making, and can replicate them, but you don't know what the sounds mean. Moreover, the speakers are actively but subtly trying to prevent you from learning the semantics: any attempt on your part to clarify/confirm your understanding of anything is met with deceit and misinformation (so, you would even be led to believe you have the right meaning when you don't). How quickly do you think you'll learn a word or phrase, let alone the language? Probably never. Put differently, we learn languages through heuristics, and part of that process is confirming correct understanding of a term: making the right connection between the symbol (sounds/gestures) and the object/concept (what is actually being talked about). You would need a proverbial Rosetta stone to test your understanding—you need some way of checking the message you think you got against what you think it means. More concretely, unless you hear the same phrases/gestures followed by the same sets of events, you won't be able to decipher meaning: language comprehension is a two-sided equation (symbols on one side, stuff being talked about on the other), if you only ever get one side, you'll never figure it out.

So, I would argue the PC would need to pass a second Very Hard Insight/Int check to have some "grasp" of message from the rest (a sort of placeholder—the PC has heard the same phrase repeatedly, so, it remembers it but doesn't know what it means), then a Perception/Wisdom check to catch the relevant event (and this could be trivial or impossible, depending on the circumstance), and another Int check to make the connection between the message and the event (variable difficulty depending on how much transpired between the message being discussed and the event happening, whether the PC witnessed the event, etc., but leaning towards Hard/Very Hard unless overriding circumstances). After this connection is made several times, then the PC has some understanding of that specific message. Restating from the previous paragraph: it should be nearly impossible to decipher a message without access to the corresponding event. Regardless of your intelligence, no string of sounds/gestures will reveal their meaning on their own.

Finally, I would rule that the Thieves would easily catch someone actively trying to decipher their message (an Easy Perception check); or if the PC is also trying to hide their surveillance, a Perception check by the Thieves against a Stealth check by the PC.

In sum:

To understand the message, the PC would have to pass:

1. A Medium/Hard History check, to make them aware of TC (and start looking for it) OR a Hard/Very Hard Insight check, to help them determine something's up.
2. A Very Hard Insight/Intelligence check to have some grasp of the message (the placeholder)
3. A Perception check (variable difficulty depending on circumstances) to become aware of the event happening.
4. A variable [baseline: Hard/Very Hard] Intelligence check to make the connection between the message and event. After making the right connection between the placeholder message from 2 to the event from 3 at least 2-3 times, then the PC can claim to understand the message.
5. Bonus: since the PC trying to decipher the message would need to actively look at the thieves (to catch all the subtleties of TC), the thieves could catch the PC either with an easy (passive?) Perception check or one against a Stealth check by the PC (if trying to be discreet).