Instructions in contemplative practice frequently involve observing your own mental states. Watch the thought arise. Notice the feeling. Step back and see what's happening.
There's a difficulty embedded in this, and traditions have engaged with it seriously.
The regress
If you observe a thought, what is doing the observing?
If the answer is another mental event, then that event could also be observed, by something else, and so on. The regress is straightforward and has no obvious stopping point.
If the answer is a permanent observing self, standing outside the flow of experience, then that raises its own questions — what its relationship to the observed content is, and whether it can be found by looking.
The practical version of this difficulty appears immediately in practice. Instructed to observe the mind, people notice they're observing, then notice they're noticing, and the layers multiply uncomfortably.
The Buddhist analysis
The traditions that engage with this most directly generally deny the permanent observer.
The doctrine of not-self holds that no unchanging entity can be found among the components of experience. Investigation is directed at looking for the observer, and the traditional claim is that looking carefully produces the discovery that no such thing is found.
What's left is described as a process rather than a thing — awareness arising with its objects, moment by moment, with no separate witness.
Which resolves the regress by denying the premise. There isn't an observer separate from the observing.
Whether that resolves it philosophically or relocates the difficulty is a genuine question that has occupied commentators for a very long time.
The Advaita position
A different resolution from a different tradition.
Here the argument runs that the witness cannot be an object of observation precisely because it is what does the observing — a subject that can never appear among its own objects.
The analogy sometimes used is of an eye that sees everything except itself.
The practice associated with this involves progressively identifying what can be observed — body, sensations, thoughts, the sense of being a particular person — and recognising that each can be observed and is therefore not the observer.
What remains after everything observable is set aside is the position from which the enquiry was conducted.
That's a substantially different conclusion from the Buddhist one, and the disagreement between the traditions on this point is longstanding and philosophically serious.
The Western versions
Similar difficulties appear in European philosophy.
Hume, examining his own experience for a self, reported finding only particular perceptions and never a self having them — an observation with obvious parallels.
Kant argued that a unified consciousness must accompany all experience but that this is a formal condition rather than an object that could be encountered.
Contemporary philosophy of mind continues to work on related problems, including whether introspection is reliable and what it actually accesses.
Research on introspective accuracy has produced findings suggesting people are frequently mistaken about the causes of their own mental states, which complicates any practice that depends on observing them accurately.
Why this matters practically
Not merely an abstract puzzle.
Instructions to observe your thoughts can produce a specific unhelpful state: an anxious monitoring in which you're watching yourself watching, which is uncomfortable and unproductive.
Practitioners frequently report this in early practice, and the traditions have responses.
The general instruction is not to add layers but to relax the attempt to stand apart. Rather than positioning yourself as an observer separate from experience, allowing the observing to happen without a separate observer being constructed.
That's easier to say than to do, and it's frequently reported as a shift that occurs rather than one that can be performed deliberately.
The practical instruction
What seems to help, from teaching traditions.
Don't try to locate the observer. The instruction to notice thoughts is not an instruction to identify who is noticing.
Notice the noticing only once, if at all. Further layers are the mind doing what it does rather than progress.
Attend to the object rather than to the attending. Focus on the breath, not on the fact that you're focusing.
And treat the difficulty as information rather than as failure. The awkwardness of trying to observe yourself is precisely what the traditions were pointing at, and encountering it is the enquiry working rather than going wrong.
The honest position
This is an unresolved problem that serious people have worked on for millennia without agreement.
The traditions offer different answers, the answers are incompatible, and each is defended by people who have investigated far more thoroughly than most of us will.
Which suggests holding the question open rather than adopting a position, and treating the practical instructions as what they are — techniques for working with attention that don't require settling the metaphysics first.
The therapeutic parallel
Worth noting because the same move appears in a completely different context. Cognitive therapy involves observing one's own thoughts and recognising them as mental events rather than as facts about the world.
That is structurally the same operation, arrived at independently, for different reasons, without any of the metaphysical apparatus.
The therapeutic version does not raise the regress problem, because it does not need to. It treats the observing as a practical capacity rather than as evidence about the nature of the self, and stops there.
Which is a reasonable position and it suggests the philosophical difficulty is separable from the practical technique. The technique works whether or not anybody resolves who is doing the observing.