Does Your Dog Actually Feel Guilty? What the Studies Show

A frequently cited study had owners scold dogs for getting into forbidden food, but the dogs were sometimes scolded when they had not actually misbehaved and sometimes not scolded when they had; the so-called guilty look, with lowered head, averted eyes and a tucked tail, appeared based on whether the owner scolded the dog, not based on whether the dog had actually done anything wrong. This suggests the look is a response to the owner's behaviour and tone in the moment rather than evidence of the dog recalling and feeling bad about its own earlier action.

This does not mean dogs have no inner emotional life, but it does challenge the specific claim that the guilty look represents the complex, self-reflective emotion of guilt as commonly understood in humans, which involves evaluating one's own past actions against a moral standard.

The behaviours people label as guilt, appeasement postures such as lowered body, avoided eye contact, ears back and a lowered tail, are well documented across canine body language generally as signals of submission or an attempt to defuse perceived tension, not specific to situations involving wrongdoing. A dog may show these signals simply in response to an owner's annoyed tone, posture or facial expression upon walking into a room, independent of whether the dog actually did anything.

This reframing matters practically: if the guilty look is really a response to the owner's immediate cues rather than memory of a past action, then scolding a dog after discovering a mess, hoping the guilty look confirms it 'knows what it did', is both based on a misunderstanding and unlikely to be effective training, since dogs do not reliably connect a delayed consequence to an earlier action.

Current understanding in canine cognition research generally supports dogs experiencing primary emotions such as fear, joy, anger and affection, which do not require complex self-reflection to exist. More complex, self-conscious emotions like guilt, shame or pride, which in humans require evaluating one's own actions against an internalised standard, are considerably harder to demonstrate in any non-human animal, and the evidence in dogs specifically remains inconclusive rather than settled in either direction.

This is an active area of ongoing research rather than a fully closed question, and some researchers argue dogs may have a simpler precursor capacity worth investigating further, while others maintain the appeasement-response explanation adequately accounts for all the observed behaviour without needing to invoke guilt.

Because the guilty look is most consistent with a reaction to the owner's present behaviour rather than evidence of memory-based guilt, scolding a dog for something discovered after the fact, such as a chewed shoe or a house-training accident found later, is unlikely to teach the intended lesson and may instead simply make the dog warier of its owner's moods around messes in general, without changing future behaviour.

A more effective approach focuses on preventing access to tempting but forbidden items through management, and on reward-based training that reinforces the behaviour the owner actually wants in the moment it happens, rather than relying on a delayed correction that the dog cannot meaningfully connect to its earlier action.

Understanding that the guilty look reflects a reaction to immediate human cues rather than retrospective self-judgement helps owners interpret their dog's behaviour more accurately and respond in ways that are both kinder and more effective at changing behaviour going forward. It also illustrates a broader pattern in animal behaviour research: behaviours that look emotionally complex to a human observer sometimes have a simpler, well-supported explanation that does not require attributing human-like cognitive processes.

This does not diminish the genuine bond and communication between dogs and their owners; it simply means that bond is better supported by understanding dogs on their own cognitive terms rather than projecting a specifically human emotional framework onto behaviours that may have a different origin.