How could we know if an animal has an experience?

It will take us a long time to have strong evidence about what many other species experience. In the meantime, if there is a possibility that they can have negative and positive experiences, we should treat them as if they do. Perhaps later, with strong methods and a lot of caution, we can start to provide evidence that certain animals probably do not experience.
Here, I attempt to explain some difficulties in learning about the experiences of other animals and what interesting new methods are being developed to learn more about their experiences.

Animal behavior and physiology should be interpreted with a lot of caution.

For animal advocacy, it would be ideal to know exactly which animals can have negative and positive experiences in order to not spend resources on those that cannot experience them. Of course, this knowledge sounds utopian. Experiences cannot be measured as simply as something like temperature, and even in humans, we have to rely on what they say about their experience, which is affected by motivation, forgetfulness, and emotional unawareness. These issues are described in Elizabeth Irvine’s book, Consciousness as a Scientific Concept: A Philosophy of Science Perspective (2012).

Still, if experiences have effects in animals, they might not be impenetrable to the scientific method. Unfortunately, the assumptions that are made about the effects of experience muddle what was already a complicated topic. Another work of Irvine’s, Developing valid behavioral indicators of pain (2020), convinced me me that many experiments which are cited as evidence of what an animal feels are far from getting at the presence of an experience. For instance, some researchers try to provide evidence of an animal’s ability to experience by looking for motivational tradeoffs. These studies try to identify if an animal makes any decisions that take into account many criteria. This search is driven by a theory of the function of positive and negative experiences; they might help you decide which choice is best by allowing the creature to have a common currency of preference for all of its possible actions. At this point, experiments like this suffer from a major problem: we’re not really sure that the theory this search is based on is true; animals might be able to make complicated decisions without having an experience.

If we’re not making these types of assumptions, it’ll take a while to provide strong evidence of an animal’s ability to experience, as I will discuss in the next section. In the meantime, if there is a 50% chance that an animal can have a negative experience, and there is a easy way to (maybe) help it, we should consider helping it now and not wait around to be 90% confident. Accordingly, I think a more imminent focus should be payed to which organisms may possibly have positive and negative experiences, which might be explored by thinking about minimum criteria needed to have an experience. (I don’t have much to say on this topic because I have not read much about it at all.) Even for advocacy, I am not sure that strong evidence of animals having negative experiences is needed to change minds or behaviors. (In one study from 2022, 62% of people in the U.S. agreed that “animals used for food have approximately the same ability to feel pain and discomfort as humans.” ) The practical implications of these new methods may be limited for now, but exploring them could help us narrow down our list of concerns. More speculatively, perhaps the investigation of experience using these methods will somehow help us answer questions of what animals like and how much.

The effects of experience can be investigated with scientific tools, which is being done in cool new ways.

One way of investigating the impacts of an experience is to present different groups with experienced stimuli and non-experienced stimuli and look for differences in their behaviors. Typically, in order to do this, researchers take advantage of the delay that occurs between raw input and experience; your eyes get data about what’s in front of you, and there is a delay before that data is simulated as things you see. Before sight is simulated, you can still be affected by the information your eyes are taking in, similarly to how your hand pulls away from a hot surface before feeling pain. Among other techniques, one way to “mask” visual stimuli (making it not experienced) is to flash an image for a very short duration.

Finding behavioral differences after manipulating whether a stimulus is experienced has had a long and controversy-ridden journey. Beyond the problems of trusting what people say they have experienced, it is difficult to confirm that behavioral differences are the result of the manipulation of experience, and not simultaneous manipulation of other non-experienced processes. (For instance, what if the short duration made it hard for me to recognize what the image was, but I actually did experience it?)

Still, these methods, when used with care, can help us understand the behaviors that are facilitated by experiences in humans. After finding these in humans, we can see if an animal shows those behaviors and if they depend on the duration of the stimulus.

This is what Shay Ben-Haim et al. did in their 2021 study “Disentangling perceptual awareness from nonconscious processing in rhesus monkeys (Macaca mulatta).” In this paper, researchers had participants choose between 2 treasure chests on a computer screen, with the goal to choose the one with treasure. Before choosing, a star flashed in front of the treasure chest without treasure. For participants in the masked stimulus group, they flashed for 33 ms, and for others, they flashed for 250 ms. From earlier studies, it seemed that when information was presented and not experienced, it was hard for people not to use it, presumably because they couldn’t consciously decide not to. Similarly here, they hypothesized that if participants were unaware of the stars, they would have a hard time choosing the correct treasure chest. They expected a very different pattern to arise for participants who were aware of the stars; they thought these participants would quickly learn that they needed to select the treasure chest opposite to the star.

They found what they expected; when the participants were aware of the stars, they learned the pattern and chose more correct answers over time. When these stars flashed subliminally (below conscious perception) and participants were unaware of them, participants did not only fail to learn the pattern, but they also responded slower than controls who were subliminally flashed stars in front of both treasure chests, suggesting that their performance was impaired by the non-experienced stimuli. To phrase this last detail differently, one non-experience group was given no helpful information (stars flashing in front of both sides), and they were able to pick their answers faster than the non-experience group that was given the stars that flashed opposite to the correct choice.

The purpose of Ben-haim et al.’s study was not only to create this cool protocol; it was also to use a protocol like it on a handful of rhesus monkeys (Macaca mulatta) and see what happens. For the monkeys, the protocol was mostly the same except they were given juice when they chose the correct treasure chest. The monkeys showed the same pattern; when the stimulus was presented at longer durations, they got more and more correct over time. When it was presented at 33 ms, they did not, and responded more slowly than the two star control.

With only those experiments, it’s hard to be confident that a change in experience is necessarily what is causing the differences in behavior. For instance, might the difference be caused by how important the animal thinks the stimulus is? Maybe at short durations, they still experience it, but assume it’s irrelevant. To try to figure out if these differences were really due to differences in experience, the researchers attempted to make humans notice the short-duration stars. When humans were told that they were being presented with quick flashes of stars, participants, when asked, said that they were sometimes able to see them. Interestingly, after made aware of the stars, participants were able to learn the pattern. These participants wouldn’t always be able to see them, and on these occasions they would get more incorrect than occasions than when they said they could. Even though the stimulus length wasn’t varying, their behaviors varied in ways that corresponded with their awareness, which is consistent with the idea that their lack of awareness of short-duration stimuli drove the differences in behaviors. Because they manipulated awareness in a second way and found consistent results, it seems that these behavioral differences in humans are from differences in experience.

In a paper talking about this study, Crump and Birch (2021) point out that the authors were not able to do the second manipulation of awareness in monkeys, which means that there is less evidence that the behavioral differences are caused by differences in awareness for them. Crump and Birch (2021) describe that one potential way to get around this would be to find a threshold for awareness in stimuli duration where more behavioral differences occur, which would provide evidence that these behavioral differences are coming from differences in awareness. This threshold would imply a systematic difference in processing at certain durations of stimuli, which matches better to consciousness than other explanations.

(By the way, this is the only study that I know of that has used this type of protocol on animals, but I would expect more eventually.)

Perhaps these methods may be extended to help us cross certain things off of our list of concerns, but this would need to be done with a lot of caution.

Because I think that we should treat animals that have the possibility of being sentient as sentient, I would want these methods to focus on finding evidence of a lack of experience in animals, which would probably take many experiments for each species.

Let’s say we were working with flies on a similar protocol as the one described above, and they never showed the learning behavior at any duration of stimulus. (I have no idea if that is likely to be true based on current evidence.) I don’t think this would be much evidence at all that a fly wasn’t conscious because it seems pretty likely that the suite of behaviors facilitated by awareness in humans would not be exactly the same size as those in flies. For instance, when a human is aware of their emotions, they might be able to employ more self-control as a result of that awareness. It’s easy to imagine that this is not the case for all animals. Similarly, maybe a fly can’t learn to do certain things, but it still has an experience that allows it to do other stuff.

Another limitation is that these results may also mean that a fly doesn’t experience vision, which doesn’t mean it has no positive and negative experiences. In fact, it could still have positive and negative experiences as a result of non-experienced visual processing. To narrow into the effects of positive and negative experiences, we will need to develop new protocols to mask negative or positive experiences and protocols that show behavioral differences (both in humans). Perhaps something like a short duration of heat on the skin may work as a negative stimulus to be masked. We would also need to contrast behavioral differences that come out of positive and negative experiences from those that are the result of experiences in general.

Here is an example of the possibility I’m proposing. In two groups, we allow people to drink water. Each time they drink water, they are given experienced or non-experienced bursts of heat on their skin, depending on the group they are in. As a control, another group is allowed to drink water however they would like without any application of heat. We find that the non-experienced bursts participants pick up the water glass and put their lips to the glass, but do not drink anything from it. The group with experienced heat does not pick up the glass or drink the water, and the group without heat drinks the water. We cannot find a comparable behavioral effect with neutral stimuli, so we think the behaviors are related to it being a harmful stimulus. We find a similar pattern to the non-experienced humans in flies, where they will still do an action that gets them very close to the negatively conditioned stimulus, but will not touch it. In other insects, we see that above a certain threshold of duration, they will not approach the stimulus at all.

In this example, again, I would say we do not have much evidence that the fly doesn’t have negative experiences. Maybe heat is not a negative experience for it, or maybe this particular behavior is for some other reason unavailable to flies. We would need to try a variety of protocols on its behaviors and on what we are masking for the fly to learn more.

To learn as much as we can from these type of experiments, we can also map the results of these experiments onto an evolutionary tree, which could perhaps reveal when certain characteristics evolved.

When theories abound, ask yourself, how would this be confirmed or disconfirmed? I have enjoyed being humbled by how little I know during this project, and I hope you approach learning similarly.

Thank you to Melissa, Ren, and the Conner lab for your feedback. Thank you to Elizabeth Irvine for your response to my questions. Thank you to Jonathan Birch for your generous reading suggestions.