Seth Frey writes:
There are mostly useless of bits of cognitive psychology that I’ve always loved. For example, a lot of categorization research about life on the edge of what objects are what. How flat can a bowl be before it’s a plate? How narrow can a mug be before it’s a cup? How big can a cup be before it’s a bowl? Can it have a handle and not be a cup? When does too much handle make it a spoon? These are questions that can be used to create little microcosms for the study of things like culture, learning, expectations, and all kinds of complexities around the kinds of traits we’re surprisingly sensitive to.
This reminds me of the old question of whether cereal with milk is a soup. The answer seems evidently no, but it’s not immediately clear why. If you eat cereal with milk for lunch instead of breakfast, does it become a soup? If not, I’d say the problem is not with the bits of cereal—some soups do have bits of bread or rice or something else floating in them—but with the milk. Milk is just too simple to be a soup base. For that matter, if you drank a bowl of tomato sauce, you’d be drinking tomato sauce, not tomato soup. So I think that for something to be “soup,” it requires some preparation or alteration of its non-soup components.
Surprisingly, often the dictionary can help with these sorts of challenges of understanding. Looking up “soup,” we get the definition, “a liquid food made by boiling or simmering meat, fish, or vegetables with various added ingredients.” That doesn’t sound completely right, either—for example, gazpacho is considered a soup and it doesn’t involve any boiling—but there is the “added ingredients” thing. Yes, cereal is added to the bowl of milk (or vice-versa), but the cereal doesn’t quite feel like an “ingredient.”
The dictionary gives this definition of “soup up”: “to improve the capacity for speed or increase the efficiency of (a motor or engine) by increasing the richness of the fuel mixture or the efficiency of the fuel, or by adjusting the engine.” This seems relevant, as it captures the idea of soup being rich in some way, being more than a sum of its parts. Cereal with milk is just the sum of its parts; it’s not “souped up.”
The point of this discussion is not to obsess over definitions of soup but rather to reflect that this sort of discussion can, paradoxically, be helpful in developing our understanding.
I say “paradoxically” here because it would be natural to think that a discussion such as “Is cereal a soup?” is pointless because definitions are arbitrary, you can define “soup” to be whatever you want it to be, and in any case no definition is completely precise—it’s a coastline-of-Britain sort of thing—and so this would all seem to be as pointless as an argument over whether the nickname for Stephanie should be spelled Steph or Stef.
But, no, I argue that the above example demonstrates that something can be learned from such a discussion, that it is not a sterile debate going in circles. I’d argue the debate is more of a helix than a circle, in that it moves forward while turning. To put it another way, the discussion (I wouldn’t call it a debate) on whether cereal is a soup is an entry point toward thinking harder about soup. I wouldn’t say that thinking hard about soup is the most important goal in life, but, yeah, I do feel that my understanding is slightly deeper now. I’ve filled in a couple of nodes in my general network of understanding.
How does this relate to statistics?
All our models are wrong but they can still be useful. Often the value of a model is in its refutation. Or, to say it more carefully: by being specific with our modeling, we can learn from the ways in which it does not accurately predict the world.
This sort of thing happens a lot in science, that the way to understand an idea is to make specific assumptions and push through their implications, looking at, in Frey’s words, “life on the edge of what objects are what.” And this is one reason I like an assumption-rich approach to statistics.
Set up a full joint distribution and use it to make (Bayesian) inferences, not because you “believe” your model but because you don’t believe it—because you don’t believe any model.
I agree with this, and think that in a lot of papers it would be great to read about the models that did not fit, and reasons for that in detail, instead of just the final results the authors were satisfied with (and maybe some robustness checks).
But unfortunately in a lot of fields it is discouraged to include these things in the final paper, at least in detail. And, of course, technical difficulties often make it difficult to fit a lot of models, and discussing them with simulated data fits etc can be a lot of work.
Incidentally, when children learn a language they love to generalize to abstract concepts from a few examples; eg most 2 year olds would have absolutely no problem calling milk with cereal a soup, or “make a soup” from orange juice, mayo, and potato chips at the dinner table and then cheerfully proceed to eat it. A lot of these terms are just cultural conventions.
Knowing what was tried and didn’t work is important.
Tamas: I think you’re absolutely right, but even Andrew and Aki didn’t want me to include the approaches that didn’t work in our Pathfinder paper with Lu Zhang. We also dropped the motivation from applying the intermediate value theorem to an optimization trajectory as it applies to heads, tails, and bodies of distributions. We also dropped discussion of typical sets and the related notion of wanting to sample points whose log density is near the expected log density (aka negative entropy). I had a series of visualizations in the original paper working through how this all applies to multivariate normal distributions. Instead, they said we needed to kick out the scaffolding because anything we say can be held against us by referees. They’re right, of course, which is why I think journal publications can be counterproductive. The only place you’ll find the motivation and the two approaches we tried that didn’t work is in my talks.
As a kid I would eat “ice cream soup” once it started melting. And it was clearly different than eating ice cream.
“the discussion (I wouldn’t call it a debate)”
Good example
“Surprising[ly] often the dictionary can help with these sorts of challenges of understanding.” I’m surprised that this is surprising: Isn’t a dictionary constructed by people thinking carefully about how specific words reflect our understanding of concepts? What else is a dictionary for?
The U.S. Customs Service often has to answer questions like this to classify imported goods, which questions then may end up in court. Two of my “favorites” from court cases (i.e., I’m soooo glad I don’t have to decide these things) were whether a replacement rear seat for a Ford Explorer is a passenger seat or a cargo seat, and whether an 11-piece Santa Clause costume is a festive costume or a costume for a religious holiday.
I really like these discussions, not because they’re useless, but because they can tell us a lot about both cognition and language. For example, they completely shoot down simple theories of meaning that say that the meanings are “out there in the world for us to discover”. No they’re not. We make them up as cognitive agents and as Andrew says, we can put the boundaries wherever is convenient. That’s what Wittgenstein was getting at when he said the notion of chair was defined by family resemblance, not an exhaustive definition, as the logical positivists were trying to do.
If you think “soup” is hard, what about “red” (the noun)? The Oxford Learner’s Dictionary says “having the color of blood or fire”. Merriam-Webster, not to be outdone, says “a color whose hue resembles that of blood or of the ruby or is that of the long-wave extreme of the visible spectrum”. So they give up on trying to define the boundaries and just work by analogy. Collins says “Something that is red is the colour of blood or fire.” I’ll leave it to Andrew to define the boundary of plagiarism.
Raghu: Isn’t the Santa Clause a piece of a contract parents make with their kids? But it does bring up an issue I had in natural language processing when trying to extract “person” names from text. Is Santa Claus a person? Maybe a fictional person. They’re fictional persons. What about Bugs Bunny? He’s a fictional person, too, though he’s shaped like a rabbit.
If you like this kind of thing and want to combine it with statistics, you can do a lot worse than checking out cultural consensus theory. It’s a survey-based approach to discovering culturally defined concepts, which basically uses the same model as Dawid and Skene introduced in 1979 for deciding consensus among medical record texts (the earliest approach to modeling text crowdsourcing I know—Andrew and Jennifer helped me define a model for crowdsourcing back when I was working on natural language and it turned out to be equivalent). Of course, I don’t think there are well-defined boundaries, so their approach is only an approximation. They do introduce a mixture component in one of their models, but they don’t really account for the fact that there’s a bit of variance person to person on the boundaries.
“Raghu: Isn’t the Santa Clause a piece of a contract parents make with their kids?” — It was Kyle C, not me, who related the neat Santa Claus costume story, though perhaps I may be defined as Kyle C through family resemblance, not exhaustive definition. Not that I have ever met Kyle C, nor know what he looks like. I did, by the way, briefly wonder if one could make a costume made of large-print clauses from Christmas-related legal documents.
There’s already relevant contract law on this. https://www.youtube.com/watch?v=EpL565sEKCc
I submit that you have figured out why GPT-3 et. all are problematic.
Is it useful at all to present or accept something that is not a model as a model? But, I must admit I don’t know what a not-model is. In any case, here is my “soup” story to share.
At many Japanese restaurants in the U.S., miso “soup” is in the menu as a soup. But, thanks to Andrew’s discussion about soup, I realize miso “soup” (miso shiru in Japanese) is not a soup. It is more a broth. In Japan, it’s not usually in the menu and is given free of charge to the customers like green tea. (There sometimes are souped-up miso “soup” for charge, however.) Maybe, the original translation was due to ignorance or convenience. But, being called a soup, miso “soup” becomes a soup—being taken perhaps more legitimate or valuable than a broth at least in the menu. But, it’s a fake soup, pseudo soup, or whatever—just not a soup. Having complained about the translation or labeling, I can sympathize. Occasionally ridiculed as a raw fish eating savage in the 80’s, I can imagine few ate/drank miso “soup” even if it were offered free. I remember some of my American friends didn’t even eat rice or drink green tea, eating only shrimp tempura and chicken/ beef teriyaki with ice water or Coke. But, miso “soup” has survived (with sushi, sashimi, and other eccentric Japanese foods) in the States. In the end, I personally accept the great nutritional value of miso “soup” in my food life. I’m sure it is useful for Japanese restaurants to make some additional profit (as standard miso “soup” can be made very easily and cheaply even in the U.S. these days). And, some customers might actually enjoy and value it. It seems whether one believes miso “soup” is a soup or not does not matter.
“The answer seems evidently no, but it’s not immediately clear why. ”
Maybe it’s not immediately clear but there are many substantive differences between soup and cereal.
IMO these two are the essence:
-Grains are optional in soup, but seem to be the primary substance in cereal
-Cereal is often consumed with fruit in it, almost never with vegetables or meat; while soup is almost always with either vegetables or meat in it, rarely (never that I know of, but that might be wrong) with fruit.
Other differences between soup and cereal:
-For cereal there usually isn’t a recipe for how much – or even which – liquid to add, except for porridge
-Soup can be packaged dry or canned with the liquid and is commonly frozen
-Cereal is never packaged canned with any liquid and never frozen, it’s always preserved dry.
-Cereal becomes less appetizing and converges on disgusting the longer it is left in the milk
-Soup often becomes better with age, for days and possibly weeks; when properly preserved it doesn’t get worse
-Soup can be made with all foraged or hunted ingredients; cereal requires?? processed grains (cue new Kurlansky book: the history of cereal)
“Grains are optional in soup”, OK, but so what? Players with the last name of Curry are optional on basketball teams, but the Golden State Warriors are still a basketball team. You can’t argue that something is not a member of a class just because it has some features that are not required for membership!
Same thing applies to cereal often being consumed with fruit: just because people sometimes exercise an option with cereal that they don’t exercise with other soups does not prevent cereal from being a soup.
Your whole list is like that, in fact. You list things that are common in soup but uncommon in cereal, and vice versa, but none of these logically prevent cereal-in-milk from being a soup.
So from a conventional logic standpoint I don’t think your argument holds up at all, and I almost said so and then moved on. But if you reframe it somewhat then I think you’re onto something, which is that a lot of these definitional questions might be better thought of in a fuzzy logic way than in the sense of binary membership.
The example I usually use is “is a hot dog on a bun a sandwich”? I thought about making this point in the discussion of “what is a woman” https://statmodeling.stat.columbia.edu/2021/07/18/what-is-a-woman/ . I agree with Andrew that this kind of discussion can lead to some insights…but the insights are related to how we define things and why. We all know what a hot dog on a bun is, and we’re not going to learn anything more about it in discussing whether it’s a sandwich. This is not to say that such a discussion is unnecessary, indeed as others have pointed out there may be reasons (such as legal reasons) that we need a definitive yes/no answer. But that answer is going to driven by the way we choose to describe the world, not by the world itself. Whether the courts rule that a hot dog on a bun is a sandwich, or rule that it is not, a hot dog on a bun is what it is and the words we use to describe it do not change it at all.
Which brings me to where I agree with the relevance of your list without agreeing that it logically demonstrates that cereal-in-milk is not soup. I think of these categorical definitions in a fuzzy logic sense rather than a conventional logic sense. As you have noted there are many parameters that characterize foods that have liquid components (including soups). Chicken noodle soup is very soupy in just about every way there is. As Hiroaki points out above, miso soup (as sold in Japanese restaurants in the US) is less soupy in some ways: it’s more like a broth — a component of soup — than a complete soup. Same for tomato soup: as Andrew suggests, if you warm up some tomato juice you don’t necessarily get tomato soup. But I would say you don’t necessarily _not_ get tomato soup, either. In all of these things I think a strict yes/no definition might be imposed on them if required for legal reasons or something, but for our internal representations it is better to have a fuzzy class membership in which chicken noodle soup is very soupy, miso soup and a simple tomato soup are less soupy, cereal in milk is very much less soupy (but is still very soupy compared to dry cereal, or compared to a bowl of milk).
Phil says:
‘ “Grains are optional in soup”, OK, but so what? ‘
So what? Grains are 95% or more all of most cereal by weight, that’s so what!! Grains, in fact, even before processing, are often called “cereal”. That’s so what!! :) You can go to the store and find lots of soup with only a few percent grain by weight. You like the chicken noodle example – OK, what’s the grain content by weight of Campbell’s chicken noodle? 10%? Almost all “cereal” is 90% or more grain by weight, the rest being sugar and/or dried fruit. You think that distinction is irrelevant? I can’t imagine why. It’s universal.
“…without agreeing that it logically demonstrates that cereal-in-milk is not soup. ”
Why not?
You want the definition of something to be restricted to a narrow set of rules about its physical constitution. Why is that the only property you allow? Things are the way they are for practical reasons: how they’re made, shipped, stored and used. Sometimes technology overcomes barriers and things of different function converge: some day you may be able to buy your frosted flakes in a can with milk or in the freezer section. But not right now, so that’s a pretty significant distinction between soup and cereal.
Phil, imagine a bowl of tomato juice with Cheetos in it. Soup or cereal? :) You can properly classify it with the guidelines in my comment above. :0
this is so hilarious. reminds me of teaching undergrad labs.
If no non-cereal soups were mostly liquid and grains, you’d have a point. But that’s not the case, so you don’t.
For example, barley soup is mostly liquid and barley. There are also plenty of noodle soups in which wheat is a main ingredient.
Some soups have grains. Grains are, as you say, optional in soups. Therefore just because something has grains, that does not rule out it being soup.
Here, try this: not all people have blue eyes. Tom has blue eyes. Therefore Tom is not a person. See how that isn’t correct?
That’s the same logic you’re trying to use when you say: not all soups contain grain. Cereal-in-milk has grain. Therefore cereal-in-milk is not a suit.
You’re right, this is hilarious and is like teaching undergrads. We agree on that, at least!
We need to disambiguate a little. There is “cereal” and then there is “breakfast cereal.” “Cereal” is just processed grass (germ, bran, and one other component)–there’s something like a dozen+ cereals that have been harvested for thousands of years (corn, wheat, barley, oats, etc). “Breakfast cereal” is cereal that has been further processed into a form that doesn’t resemble the original cereal form. “Breakfast cereal, the meal” (for lack of a better phrase) is breakfast cereal served in a liquid vehicle, typically milk or milk substitute.
Barley soup would not be considered a breakfast cereal nor a breakfast cereal meal. The processing the grain goes through and the liquid vehicle the grain is served in matters. Barley soup is served in a savory broth–often beef-based–with vegetables and, typically, braised chunks of beef. The barley is minimally processed past the processing that made it into a cereal. Because of these different factors, in combination, barley soup is not a breakfast cereal.
As an extreme example: beer is typically made of cereal, but is not a breakfast cereal. It consists of processed malted barley (and water, hops and yeast). Would we call beer cereal, breakfast cereal, or a breakfast cereal meal? Or would we call beer soup? Or neither? I think it is neither any cereal nor any soup.
Grape Nuts *is* considered a breakfast cereal. It is made of malted barley flour, whole grain wheat flour, and yeast. Hey, there are some overlapping ingredients with beer! But it’s a breakfast cereal because of the additional processing the cereal ingredients go through.
As a final example, take a look at the differences between chocolate chip cookies and rice krispy treats and the corresponding breakfast cereals: Cookie Crunch and Rice Krispy Treat Cereal. What makes the former desserts and the latter breakfast cereals? One one hand, Cookie Crunch breakfast cereal consists of finely ground corn, corn meal, corn flour, and corn syrup–none of these ingredients are usually used in a chocolate chip cookie, but most of the ingredients were, at one point, cereals (corn). The breakfast cereal is cookie-like, but there are some distinct differences; they are very dry and have a long shelf life, among other differences; there is much less fat and cholesterol in the breakfast cereal, presumably because the breakfast cereal was not made with egg and butter. They certainly aren’t dried cookies; the processing matters.
Rice Krispy Treat cereal consists of rice, sugar, and some vitamins/minerals. Interestingly, in this case–a cereal (Rice Krispy) became a dessert (Rice Krispy Treat) which was then reformulated back into a sugary breakfast cereal (Rice Krispy Treat cereal). The difference between the final cereal and the dessert is that the cereal is not gooey, buttery, and mashamallowy; it’s instead quite dry and brittle, presumably for a longer shelf life and better properties in milk. Again, there is much less fat in the breakfast cereal than the dessert. Processing matters.
Finally, gazpacho is not breakfast cereal. It is not made of cereal and it has a savory, vegetable base. Breakfast cereal the meal has a milk, milk substitute, or water base (for “hot” breakfast cereal).
Unfortunately, the only way to get around inane internet arguments is to start writing definitions as a lawyer would; it’s a way to formalize into words these distinctions we make that should otherwise be common sense.
1. You haven’t done that. “Doesn’t resemble its original form” is ambiguous, and you added an extra criteria of certain liquid vehicles being acceptable and some not without defining it.
2. I don’t believe you believe your own definitions. By your definition, rolled oat oatmeal is not breakfast cereal, nor is granola in milk, honey bunches of oats, or any of these:
https://www.quakeroats.com/products/cold-cereals/granola
3.) I have no reason to accept your definitions over mine. This kind of precision can be useful for standardization in a task driven context. And this kind of discussion can clarify what people mean when they say things, or can help people clarify to themselves what they mean. But the argument is unresolvable because words are made up and imbued with meaning by subjective life experience, so I don’t care about it.
The problem arises when one person comes up with a definition that works for them and confidently declares that “this is the correct definition”. It’s much worse if said definition misclassifies a lot of things even according to their own opinions. And worse still when they challenge people to find counterexamples, then become indignant when said counterexamples are found.
“Some soups have grains. Grains are, as you say, optional in soups. Therefore just because something has grains, that does not rule out it being soup.”
“Here, try this: not all people have blue eyes. Tom has blue eyes. Therefore Tom is not a person. See how that isn’t correct?”
Phil. Your characteristic sharpness is eluding you. I’m shocked at your simple misunderstandings.
The fact that grains are optional in soups is a distinct **property of soups**. The fact that grains are *not* optional in cereal is a distinct property of cereal. It does not rule out grains in soups. Blue eyes are optional in humans. Therefore they may still have blue eyes.
“1. You haven’t done that. “Doesn’t resemble its original form” is ambiguous, and you added an extra criteria of certain liquid vehicles being acceptable and some not without defining it.”
I agree–I had not done that, and never claimed I did. I am proposing that such a definition could be written if one took the time. I was trying to disambiguate what was being discussed: cereal, breakfast cereal, or breakfast cereal as a meal. Regarding liquid vehicles: milk, milk substitute, or water as inclusion criteria to be considered a breakfast cereal meal. Broth (chicken, beef, or vegetable) or puree (fruit or vegetable) as exclusion criteria to be considered as a breakfast cereal meal.
“2. I don’t believe you believe your own definitions. By your definition, rolled oat oatmeal is not breakfast cereal, nor is granola in milk, honey bunches of oats, or any of these:
https://www.quakeroats.com/products/cold-cereals/granola”
I do believe my own definitions. Re-read them. Rolled oat oatmeal is a cereal, but not a breakfast cereal. Rolled oat oatmeal, heat-processed for 10-15 minutes in hot water or milk, is a breakfast cereal and meal. This is essentially what they do to create instant/quick-cooking oats (pre-cooked, dried, and rolled). Granola is treated with dry heat (baked). et cetera
“3.) I have no reason to accept your definitions over mine. This kind of precision can be useful for standardization in a task driven context. And this kind of discussion can clarify what people mean when they say things, or can help people clarify to themselves what they mean. But the argument is unresolvable because words are made up and imbued with meaning by subjective life experience, so I don’t care about it.”
That’s fine. Even if this isn’t the case, you’re free to order a bowl of cereal at a restaurant and expect to get minestrone, or order a bowl of soup at a restaurant and get a bowl of Cheerios.
“The problem arises when one person comes up with a definition that works for them and confidently declares that “this is the correct definition”. It’s much worse if said definition misclassifies a lot of things even according to their own opinions. And worse still when they challenge people to find counterexamples, then become indignant when said counterexamples are found.”
My annoyed friend–why don’t you spend 15 minutes reading about cereals, breakfast cereals, and the like. These aren’t just my *definitions that work for me.* This is pretty standard stuff.
See: https://en.wikipedia.org/wiki/Cereal
versus: https://en.wikipedia.org/wiki/Breakfast_cereal
Also, I haven’t challenged anyone to find counterexamples, nor claimed my set of definitions were correct. I think that was my second post in this entire thread! I’m proposing that a set of definitions could be continuously refined until counter-examples are essentially eliminated, but such a set of definitions would likely be lengthy and complex (sort of like a legal document), because the criteria we use to differentiate something as simple as cereal from soup (or pizza from non-pizza, or hot dog from not hot dog) can be quite complex, because our inherent classification systems are pretty complicated things with all sorts of inclusion/exclusion rules based on sensory information and past experience, among other things.
chipmunk,
Grains are not optional in barley soup, yet barley soup is soup.
Replace “barley” with “cereal” and there you go.
WTF man.
Chipmunk: you have again drifted away from discussing what was actually written. Phil is NOT saying that grains being in soup contradicts your rule. He IS saying that your rules are inadequate to even resolve the ambiguity you responded to.
You said:
And here’s the problem with that which Phil was pointing out; imagine a bowl of honey nut cheerios in milk. Let’s attempt to classify it with these two guidelines you provided:
1. Grains are optional in soup, but required in cereal
2. Cereal is served with fruit, while soup is served with meat and/or vegetables.
Well:
1. Grains are in the cheerios in milk, but that doesn’t imply that it’s not soup because grains can be in soup optionally.
2. There is no fruit, meat, or vegetables, so rule 2 does not resolve anything either.
So, as Phil has pointed out, your rules fail to classify even the example you responded to.
I see at the bottom you wrote this:
You see your problem here is that you have a parallel imaginary discussion going on in your head about things not written. Nobody has disagreed with what you have written there in your bottom comment; I 100% agree with it. NOBODY is arguing that soup and cereal are “INDISTINGUISHABLE.” What people are disagreeing with are the things you actually wrote, not what you apparently meant.
As you go through these discussions, you seem to forget what they were even about to begin with. You say things like this:
But do you even remember saying this?
So I have to reiterate my advice; print these threads out, and underline the relevant parts before you respond.
Chipmunk –
You’re just dressing up opinions as facts. And there’s an underlying logical problem where you try to take nuance and jam it into false categorical (binary, as Andrew l discusses downstairs) distinctions.
As just one problematic example from your list:
> Soup can be packaged dry or canned with the liquid and is commonly frozen.
So that would mean that soup didn’t exist in the 1700s?
You’re just basically looking for differences, and trying to reverse engineer from them dichotomous distinctions.
You might as well say “Cereal isn’t soup because I like soup and I don’t like cereal.”
Another from the list, although most of them are problematic.
https://tinyurl.com/bddfkaxh
“You’re just dressing up opinions as facts.”
That’s patently absurd! :) Go to a grocery store and find a can that says “soup” that has fruit in it! Freeze your cereal with milk in it and eat it a month later – or better yet find some in the grocery store. Do you like it? :)))
There are practical reasons that things are **what** they are. Think about it: why do spoons have long handles and cup have short handles? :) In biology this is called “functional morphology”.
Many soups make use of tomatoes, peppers, limes, lemons. I would be shocked if there wasn’t a soup with any of these ingredients at your local whole foods or whatever.
Many Chinese soups make use of jujube
Pumpkin soup, butternut squash soup and the like are common in Europe and America. You can probably find a squash soup at your local grocery store
Vietnamese canh chua makes use of pineapples
Hungarian apple soup
Trader Joe’s used to sell this
https://freezermealfrenzy.com/2019/02/trader-joes-steelcut-oatmeal-review/
Chinese barley porridge is very often sold canned. Don’t know if that counts as cereal
Please, I beg of you: never stop commenting here
“make use of tomatoes, peppers, limes, lemons\.”
OK, but usually not as main ingredients, and please note: “in culinary terms, tomato is regarded as a vegetable (wikepedia)” So you missed that one
“Pumpkin soup, butternut squash soup and the like are common in Europe and America. You can probably find a squash soup at your local grocery store”
OK, we’ll make an exception for squash on both sides. Do you have squash in your cereal? I’m surprised you didn’t point that out! :)
“Hungarian apple soup…Chinese barley porridge is very often sold canned.”
Hey, that’s great. But not common.
“Trader Joe’s used to sell this”
Note: “used to”. :)))
Thanks, you’re proving my point about the forest. :)
Haha, fruit is also not the main ingredient in cereal! You said: “served with.”
You said: “never.” It was sold, so you are wrong. Now, because it wasn’t a successful product, you think it doesn’t count? A cereal you don’t like doesn’t count as cereal for semantic purposes? Besides, this is just what I knew about off the top of my head; people still sell canned porridge.
Uh, pumpkin flavored cereal is quite common actually, and oatmeal made in a squash soup base is a thing.
First of all, you challenged someone to find counterexamples; I’m just doing what you said to do.
Second, the topic of discussion is that definitions, like models, are fuzzy at the boundaries and never quite right. You cut in and claim there are are essential distinctions you can draw, but those distinctions also turn out to be fuzzy at the boundaries and not quite right and require exceptions for squash soup. So it kind of seems like the original post is right and your “essential distinctions” are not essential.
You seem to have trouble focusing, both in tracking the general topic of discussion and remembering the words you actually wrote that people are responding to. Before the 2nd grade statewide reading comprehension exam, Ms Rodriguez recommended that we go back and underline relevant phrases while answering questions. I didn’t do it, but it seems relevant. Maybe you could print these blog posts out and highlight the relevant parts when writing your comments?
Somebody: we need to distinguish between culinary vs botanical fruit. This is a very tired argument that Kenji Lopez Alt (among others) has tried to clear up. Pumpkin, squash, and tomatoes are not considered fruits from a culinary sense. They are botanically fruit (vegetables in culinary parlance). Read “Food uses” https://en.wikipedia.org/wiki/Fruit
I think Chipmunk makes a good point, and I think Chipmunk is really referring to culinary fruit here.
Also, pumpkin and squash soup are typically made with a vegetable stock–there are various ways to make this stock but it often includes: onion, celery, carrot, garlic, and some type of bouquet of herbs with a base of water.
The other soups you mention such as canh chua include a lot of vegetables and savory ingredients. These are all clearly not cereals.
Polenta is neither or a soup nor a cereal. It is much thicker than either and is also savory–usually cooked with onions/shallots and chicken or vegetable broth and served with butter. However, the cornmeal that polenta is made from can be used to make a breakfast cereal that is sometimes eaten in Italy. It’s just made boiling the polenta in water or some milk, and served much thinner than polenta (porridge-like consistency) and sometimes topped with a little sugar and cinnamon or a pat of butter.
> You cut in and claim there are are essential distinctions you can draw, but those distinctions also turn out to be fuzzy at the boundaries and not quite right and require exceptions
The difference between soup and cereal with milk is that you heat soup.
Except heated cereal with milk. And except cold soup.
Unanon –
Sssms to me.yYou’re kinda doing the same thing as Chipmunk.
When I use a word,’ Humpty Dumpty said in rather a scornful tone, ‘it means just what I choose it to mean–neither more nor less.
Seems to me, essentially the argument goes like: “Cereal with milk is different than soup because cereal with milk is cereal with milk. With the exception of some cereal with milk which I say isn’t soup”
For me, generally I think of cereal with milk as being different than soup. And I think most other people do also. But when I’m asked to think about it, I realize that the boundaries are somewhat fuzzy, and the distinction is somewhat arbitrary (in the sense of being subjective). But I do still think there’s a broad difference of a sort that I’m not sure I can explain.
Maybe it’s more that there’s a collection of potential incision/exclusion criteria that can reach a critical mass at some stage that triggers my mental models to put an item more or less solidly in one category or the other. Perhaps thinking of them as overlapping sets (like in a Venn diagram) is a useful way to think about it.
What doesn’t work for me is some kind of didactic assertion that there’s a clear and objective distinction. Mostly that doesn’t work for me because it boils down to a subjective view on definitions (such as whether certain items are vegetables or fruits) that are often context-dependent.
If you really want to get into it, certain varieties of pumpkins and hard sweet squash (not summer squash) can be arguably either. They’re sweet, are typically paired with cream, sugar, and “warming” spices (by the flavor bible’s convention, meaning cinnamon, cloves, cardamom, nutmeg), and are served in dessert dishes you would never put “vegetables” in, like pie. At the same time, you wouldn’t eat these squashes raw (though you would eat a summer squash raw). On the other hand, few people eat plantains raw, and they’re usually served in savory dishes. I’m not interested in getting into it though. My point is exactly what’s written above
What I object to is coming in and disagreeing with the above and arguing that “there are essential distinctions”, throwing around words like “never” and “always” when they do not apply. I think the principles “a soup does not have fruit” and “a cereal is packaged dry” are great rules of thumb. I strongly disagree that they make a good point because lots of people have fine rules of thumb; what they claim is that they have “essential distinctions.” An essential distinction does not have exceptions; something that ”never happens” NEVER happens. And some of these exceptions are not even remotely uncommon if you think in a global context; once you go to the most populous country in the world they are no longer even good rules of thumb.
I guess this point is actually hard to follow since now two people have made this mistakes.
1. The claim was that cereal is not served with fruit, and it was challenged to find soup with fruit in it
2. Canh Chua a soup served with fruit
I am not arguing that Canh Chua is a cereal or that its primary ingredient is a fruit. Cereal’s primary ingredient is also not fruit. It is a common counterexample to a specific claim.
Again, grits is a cornmeal based porridge that is commonly made with broth and meat or seafood and eaten for breakfast (and other meals) in the southern United States. In China, various porridges made from barley or rice or sometimes even oatmeal, along with meat or seafood broths, are common breakfast foods. It is also not uncommon for these to be sold canned. There are also dessert soups made from no grain, but are made with a fruit and sugar based broth and served with a calcium sulfate coagulated tofu (“silken” in Japanese American parlance, “soft” in Chinese American). Jujubes are also a sweet fruit that are a core ingredient in savory herbal broth bases.
The point, again, is not that these are stupid unusable heuristics, but that they are subjective heuristics, not “essential distinctions,” and that Chipmunk said “never” about things that have happen and continue to happen. They are not even usable heuristics in a global context.
I recall enjoying some Campbell’s fruit soup when I was a kid. I don’t know if they still sell it.
A web search for “fruit soup recipes” got about 10,000 hits. These included https://www.allrecipes.com/recipes/175/soups-stews-and-chili/soup/fruit-soup/ which has 50 such recipes.
There’s a Wikipedia entry on fruit soup. See https://en.wikipedia.org/wiki/Fruit_soup.
Bob76
Joshua–yes, agree that definitions can be fuzzy. However, we often need to clearly make distinctions for practical purposes (USDA regulation, marketing regulations, expectations at restaurants, general communication with others). Given that, I think it’s useful to undertake an exercise to determine how many (and what type) of rules are needed to tease out breakfast cereals from soups, like a game of Guess Who?
An issue with the ongoing conversation, as I saw it, was people were using different versions of “cereal” in their discussion. Cereals have a pretty clear agricultural definition. There is also something that is colloquially known as “cereal” at the supermarket. Then there is another definition that involves consumption (“milk and cereal”).
In order to have a clear discussion about defining cereal and soup, would’t it help to have some common ground?
Unanon –
> In order to have a clear discussion about defining cereal and soup, would’t it help to have some common ground?
Yes. Grounding convos in shared definitions is something that often fails to take place in blog arguments. And yes, the definition of cereal is important to this convo, particularly since there are such contrasting definitions.
“So that would mean that soup didn’t exist in the 1700s?”
I bet in some parts of the world it was frozen and stored in cold cellars.
“You’re just basically looking for differences, and trying to reverse engineer from them dichotomous distinctions.”
No, I’m not. They are real distinctions. How was soup preserved in the 1700s? Probably only by freezing or maybe some kind of pickling or salt-based preservation?? I don’t know, but other than that it probably had to be kept continuously warm. Cereal, on the other hand, could be stored dry and preserved for long periods.
You and somebody found some soups with fruit! Hey, great. Who cares? Most of the other distinctions still stand.
>>“Hungarian apple soup…”
>Hey, that’s great. But not common.
You mean uncommon in your specific locality.
It is very common in Hungary. So are soups with apricot, peach, pear, cherry or plum.
Is polenta a soup or a cereal?
Is shrimp and grits a soup or a cereal?
Congee may be the most commonly eaten food on the planet.
Somebody–forgive me if I misunderstand, but I wonder if you’re getting into false dichotomies here. I don’t think restaurants or cookbooks would classify shrimp and grits as a soup or a cereal. Polenta also would not be classified as either. Do you consider shrimp and grits or polenta as a soup or cereal?
That said, I think polenta is a decent borderline case. It’s made of a true cereal (corn) that is processed (ground) with some parts removed. Polenta is usually made with the course ground cornmeal and is usually pretty viscous. It’s more of a “mush” than a breakfast cereal. For example, when polenta has cooled down, it is sometimes grilled.
Finely ground cornmeal is sometimes eaten for breakfast in Italy. When it is served as a breakfast porridge, it is typically less viscous. It’s also served with butter and perhaps some (culinary) fresh/dried fruits, and sometimes cinnamon and/or sugar. At least that’s how I ate it growing up.
Classification depends on preparation method, textures, consistency, other ingredients, region, among other things. For example, despite sharing many similar main ingredients, Grape Nuts is a breakfast cereal and beer is not a breakfast cereal.
If I were serving someone shrimp and grits, I wouldn’t say it was cereal. When I say cereal, I mean the boxed stuff in the cereal aisle of the supermarket–I wouldn’t even describe oatmeal as “cereal”. But I was surprised to learn in middle school that many people will say “cereal” or especially “hot cereal”, then serve me oatmeal or grits or some other porridge. I’ve been in places where “grits” was the breakfast “hot cereal” of choice.
Now while shrimp and grits is a traditional breakfast food, I’ve never been told I was getting a hot cereal and received shrimp and grits. But since grits with butter and salt is “breakfast cereal” in some parts, it wouldn’t be a huge leap and I wouldn’t correct anyone if it happened.
Anyways, I’m not interested in the semantic debate. I’m just responding to specific claims and challenges made by other people. Chipmunk said that his distinctions cut to the essence of a distinction, but are not powerful enough to distinguish cereal in milk from soup, and said that things “never” happen when they happen all the time, and told someone to find a soup with fruit in it as if they would fail, when lots of soups have fruit in them (especially lemon/lime). I do think his distinctions are pretty much true in the part of the United States where I live, and I think cereal in milk is very distinguishable from soup; again, I am responding to specific claims from other people.
Somebody–thanks, that’s very reasonable. I think there’s ambiguity pertaining to how people apply the term “cereal.” I’m not an etymologist, nor a food historian, but I do know that cereal is an agricultural term that includes rice, corn, barley, oats, millet, rye, and maize–among other types of grass. I imagine when this stuff was processed and turned into something edible–maybe it started with oats, or ground corn (what’s simpler than mechanically grinding something up and boiling it in water?) –it was called “cereal” (after the more general, agricultural term) and continued to be called that as non-perishable industrialized food manufacturing became a thing with boxes of Corn Flakes, Wheaties and Cheerios.
> soup is … rarely … with fruit
There are fruit soups, alas.
https://www.allrecipes.com/recipes/175/soups-stews-and-chili/soup/fruit-soup/
And tomato is actually a fruit, botanically, so there’s that too.
Andrew wrote: “Milk is just too simple to be a soup base. For that matter, if you drank a bowl of tomato sauce, you’d be drinking tomato sauce, not tomato soup. So I think that for something to be “soup,” it requires some preparation or alteration of its non-soup components.”
But if you eat a bowl of cream-of-tomato soup, you are eating a mixture of tomato sauce and milk (or cream).
Martha, good point. Milk is also the liquid soup base for many chowders; it’s just thickened with a roux (mix of butter and flour). For seafood chowders, we usually add a little clam juice/fish stock to the mix too.
I know nothing at all about Japanese food or soup vs. cereal but recall the Maricopa County recount where election deniers were looking for traces of bamboo in the ballots in order to claim China was secretly altering the vote. Suppose the election deniers found instead, traces of chop suey or chow mein on the ballots, each of which is unknown in China. However, from a distorted perspective, chop suey or chow mein sort of has a soupcon, dash, bit, hint of resemblance to a Western notion of Chinese cuisine. Further, even if nothing at all is found, that in itself proves how clever the rigging of the election was.
I think of models as ways of quantifying and qualifying uncertainty.
And I think of this kind of discussion as being very similar.
Maybe asking whether cereal with milk is a soup is essentially like building a (mental) model of soup.
Another example where mapping alternatives with meaning equivalence versus alternatives with surface similarity is needed. The reason why it is not done explicitly is (quoting Andrew) that “people want certainty, but they cannot have it”. Give me a definition to understand conceptually is the same.
Regarding impasses. Just gave a talk with Chris Gotwalt, the Chief Data Scientist at JMP on the analytic journey in a specific project. We report on dead ends too, and people loved it https://community.jmp.com/t5/Discovery-Summit-Americas-2022/Different-goals-different-models-How-to-use-models-to-sharpen-up/ta-p/505854
It’s so good to see one of these without a discussion of sandwiches.
Sorry, I mentioned sandwiches in an earlier comment.
Thought about it, but refrained.
let’s not forget Zuppa Inglese – the archetypal “English Soup”
This is a comical thread.
While “somebody” has shown my original claims about fruit are incorrect, the can be modified to be correct. Overall, I’m more convinced now that it’s ridiculous to claim that soup and cereal with milk are “indistinguishable”. Yes, of course there are exceptions to *some* of the rules in probably any set of rules about the distinction between soup and cereal. But I have **NO DOUBT** that such a set of rules can be developed that would work >95% of the time.
The real world is messy, but suitable distinctions exist. Why the hell do you think we’ve been able to survive for so long? :) The lot of you should go read the descriptions for many similar wildflower species (like Indian Paintbrushes). They’re much harder to use effectively than the distinction between soup and cereal.
AFAIK nobody has claimed that cereal and soup are indistinguishable. That wouldn’t even make sense in the context of this discussion.
The question you’re referring to is whether cereal (in milk) can be reasonably called a kind of soup. Most people say No — you are clearly strongly in the No camp yourself, as am I — but your supposed proofs that the answer is No are not logically coherent, or least some of them are not. For instance, you keep on saying that cereal contains grains and that soup does not have to contain grains. Well, no, it doesn’t have to, but it can.
Nobody else has succeeded in making you realize this, so I don’t know why I’d have more success, but I’ll try. Most of your proofs assume the thing you are trying to prove: that “cereal in milk” is a different class from “soup”. You then describe the way these classes differ. The problem with this approach is that if you do _not_ assume that they are different then you don’t find anything in the definition of the “soup” class that prevents “cereal in milk” from being a member. Cereal in milk is served cold, but some soups are also cold. Cereal in milk has grains as an ingredient, as do some soups. Cereal in milk has milk as a base, as do some soups. And so on.
You’re wrong about wildflower species (or any other type of species) but, perhaps tellingly, that’s a poor analogy. A better analogy would be determining whether a flower is a plant is a wildflower versus some other kind of flowering plant. We aren’t trying to tell whether a given bowl of food is chicken noodle soup or whether it’s cheerios in milk.
“Cereal in milk is served cold, but some soups are also cold. Cereal in milk has grains as an ingredient, as do some soups. Cereal in milk has milk as a base, as do some soups. And so on.”
Phil, I think your example is too simple. I don’t think people classify cereal vs soup on such basic, univariate criteria. They classify based on specific combinations of yes/no to those questions in addition to many other criteria, and I think we do this stuff pretty intuitively, despite it being pretty difficult to write down.
e.g.
Couple ways to classify breakfast cereal as a meal eaten with culinary fruit vs no culinary fruit:
Breakfast Cereal: (AND is served cold AND grains AND milk base NO savory liquid base AND culinary fruit NO meat NO vegetable) OR (AND served cold AND grains AND milk base NO savory liquid base NO meat NO vegetable)….etc.
Some types of soups…a subset of the larger set of soups:
Soup: (AND hot AND grain AND milk base AND savory liquid base NO culinary fruit AND meat AND vegetable) OR (AND cold NO grain NO milk base AND savory liquid base AND culinary fruit NO meat AND vegetable) OR (AND hot NO grain NO milk base AND savory liquid base NO culinary fruit AND meat AND vegetable)….
There are many dimensions used in conjunction to classify this stuff: presence/absence of grain, presence/absence of broth (chicken, pork, beef, vegetable, or other), presence/absence of dairy or dairy-like products, temperature, presence/absence of vegetables, presence/absence of fruit, viscosity, grain size, texture, and seasonings. Geographic location/culture plays a part in classifying certain porridge such as congee.
I think there are combinations above that can adequately classify cereals vs soups pretty well–at least the “gotcha” cases people are trying to come up with in this message board can be handled pretty well.
I also think there are some simpler classifiers that get 95%+ things right, at least in the US:
Cereal: dairy base, served cold, with breakfast cereal (a processed form of cereal, which itself is a harvested form of grass)
Soup: savory/broth-based with meat and/or vegetables.
A word means whatever people think it means. If people think “soup” includes cereal in milk then it does. I agree it’s easy to define soup so that cereal in milk is not included. Also easy to define soup so as to include cereal in milk.
My personal definition of soup does not include cereal in milk but that’s not the point of Andrew’s post and honestly it seems pretty stupid to argue about. My only point in these comments (other than my first one) is to point out logical fallacies in arguments that purport to prove that cereal in milk is not soup. I agree it is not soup (in my book) but some of the arguments are fallacious.
Phil said: ” it seems pretty stupid to argue about. ”
Well, yes. But a point that could have been here is that people are really good at dealing with concepts like these in everyday life, yet when you try to code them up (e.g. in an AI demo or law) they turn into a horrible mess almost instantly.
So the interesting question is can you just ignore this problem, pray that no one will notice, and actually think that GPT-17 will be “intelligent” with no mechanism (or theory) for dealing with this issue?
David,
People manage to get by pretty well without a perfectly unambiguous definition of ‘soup.’ Perhaps AIs can do so as well. I’m not asserting that that is the case, but it’s not obvious to me that it’s not the case.
I can’t speak for others, but for me there’s a big family of words (and abstract concepts represented by those words) for which I don’t have strict definitions. How many grains of sand does it take to be a heap? What additional ingredients, and in what quantity, do you have to add to tomato juice before it is soup? How flat do a hot dog and its bun have to be before it’s a sandwich? And indeed the answers to these are context-sensitive. If I’m putting a menu together I might put “hot dog” in the sandwich section even though I don’t really think of it as a sandwich and I don’t think most other people do either. I’m fine with a fuzzy-logic view of the world in which a plant has some membership in the shrub category and some in the tree category. Every thing is what it is. Our word for the thing is not the thing.
But then we run into human-created questions in which we force things into categories. Who should be allowed to compete in the women’s division of a sporting event? Is a hot dog on a bun a sandwich for legal purposes? If a landscaper is contractually obligated to plant a tree and they plant a toyon, have they fulfilled their contract? Could an AI trained on today’s principles write a cogent argument in either direction on any of these? I dunno, I guess we could try it and see.
Au contraire, this thread would seem to indicate that humans think they are great at making subtle distinctions, but often fail in practice.
GPT is being trained to become a good used-car salesman, and has demonstrated that learning ability. It has been not been trained to function as a general intelligence, and probably has a small fraction of the capacity of a human cerebral cortex (in terms of the number of nodes in its neural networks).
Humans learn how things work by trial and error, and create stories, called theories, to help others learn those things. Neural networks with trial-and-error training is our current theory of how intelligence works and in my view it has been very successful.
When automobiles were first introduced horses were more versatile and a common saying was “get a horse”. Automobiles need roads and gas or charging stations, Neural networks need massive computer systems and a lot of focused training, to learn complex tasks. “Get a human” still works but already there are things which neural networks can do better (after a learning process, whose results are stored as parameters which can be transferred to another, untrained, network). Meanwhile, different network configurations and training regimens are being tried and the theory is being developed further.