FoodTongue: Interview with Alan Huang

Alan Huang is a three- (soon to be four-) time alum of the Canada/USA Mathcamp, where the language FoodTongue started about five or six years ago. Alan learned FoodTongue just prior to his first Mathcamp, and he helped create the most recent dictionary, which is more or less all written in FoodTongue.

Brit: Hi Alan! In a sentence or two, what is FoodTongue?

Alan: Depending on the speaker, it's either a bold experiment in language design or a creative specimen of Mathcamp randomness (or an unnaturally long-lived bad idea, for certain non-speakers).

Brit: Can you tell us a bit more about the entries in the lexicon of FoodTongue? Is there a purely conventional connection between the symbols of the language and the entities referred to? Or is there a system or pattern which you follow when new words are introduced? Or is there an even closer connection between the symbols of the language and the entities referred to than that?

Alan: There's no general rule by which words are related to their meanings. Speakers are free to create any new words they need to express ideas, and the methods by which foods have been chosen include sound in English or other languages (lobster), properties of the food (popsicle), specific references (pie), related words (ketchup/mustard), themes (names of games), and randomness (Ice Cream).

Many words are thus easier to remember than if meanings were attached to random food names. Occasionally conflicts arise, but as FoodTongue evolves, ambiguous, hard-to-remember, and uncommon words tend to be replaced. The system also works well without a central authority -- the dictionary may be the closest FoodTongue has to one, but it's designed as a wiki so that it can respond to changes in the language, as opposed to instigating them.

Brit: What is the syntax like? Is the syntax that of English? Or does it have its own syntax? The phonology obviously must be that of English.

Alan: First of all, there's no inflection in FoodTongue, since there aren't many endings to add on to English nouns. (The obvious choice is to pluralize, but that has not been used yet. There has been discussion on grouping related foods; for example, if grape is a verb, raisin could be the future tense. But that was deemed infeasible.) So FoodTongue usually follows normal English word order (subject - verb - object, preposition - noun, etc.), and everything else (such as tense, number, and person) is done with separate words. However, grammar is still quite free-form as long as it's understandable, and speakers may have different styles based on preference or other languages they know. For example, a sentence that literally means "want have" would be understood as "I want to have it". A special case is music and poetry (of which there do exist FoodTongue examples). Since FoodTongue words tend to have more syllables compared to English translations, words are often omitted or moved to fit the meter.

Brit: Can you ascribe propositional attitudes to others in FoodTongue? You mentioned "want"-reports above. What about other attitude ascriptions? For example, can you say "I know that p", "I believe that p", "I hope that p", "I love that p", etc? More generally: is there a way of expressing tense and modality in the language? Can you talk about what once were the case but is no longer? Can you talk about what will be the case? Can you talk about what is merely physically, metaphysically or logically possible? Can you talk about what is logically impossible? If so, how do you do that?

Alan: All such constructions are usually just implemented with subordinate clauses, e.g. "I want that you be here". There are words for verbs like "think", "know", "love" and helpers like "should", "can", "must", which can be used with such clauses (or independently). Essentially everything else that's possible (and I don't know if some modals are in concise ways) is done explicitly, with words like "maybe" and "in the past" / "now" / "in the future". For example, "it PAST be, but NOW not" means "it was once the case, but no longer", and "not can be that..." means "it is impossible that".

Brit: Do you have ways of expressing indicative conditionals (e.g., "if John is home, then I will pay him a visit"), subjunctive conditionals (e.g. "if I had been a Mathcamper, I would have known FoodTongue"), conditional probabilities (e.g. "the probability that it will rain given how clowdy it is is 0.8"), etc?

Alan: There is no distinction between indicative and subjunctive conditionals; the difference is determined through knowledge and context. Probabilities are expressed by modifying "maybe", e.g. "if he NOW not be here, he big-maybe small-FUTURE be" would be understood as "if he is not here now, he will probably be here soon", but whether he is here now (and whether the speaker knows) is unstated.

Brit: Could you also say something about indexicals ("I", "now", "here"), demonstratives ("this", "that"), quantification ("there is an ...", "for all x") and anaphora (e.g., "A man entered the door. He took off his hat") Do you have a word referring directly to the speaker or the time of speech? Do you have words that refer only when accompanied by a demonstration (e.g. by an index finger or a gaze). Can you quantify over things as in "there are at least 30 people who are fluent speakers of the language"? Are there anaphors that refer to an entity already referred to, as in "The TV is in the living room. But it is broken".

Alan: There are words for "I" / "you" / "he/she" (plurals are formed by prepending "all"), "in the past" / "now" / "in the future", and "here"/ "there". Demonstratives are handled by "here" / "there" or by the word for "one", which serves generically either to specify or to refer to nonhuman objects ("one book be good; one be big and" means "this book is good; it is big also"). Humans are referred to by the pronoun "he/she". All of these words can also be specified nonverbally (e.g. a gesture in person, or a picture in a blog entry).

There is one relative pronoun ("that") and one interrogative (which can take a noun, as in "what person"). Quantifiers are expressed with "be x that y" and "all x y". When referring to mentioned entities, "one" and "he/she" may be omitted if understood. For example, "should talk he/she; be here" could mean "you should talk to her; she is here".

The word for "and" has several other meanings, including "add", "also", "so", "then", and "more". Comparisons and such use "and" (in the sense of "more") and/or an adjective or adverb followed by the relative pronoun (acting as "than"). For example, "and that 30" means "at least 30"; "and big that Dan" or "big that Dan" means "bigger than Dan". It sounds confusing as I write it, but in practice it's quite natural.

One caveat -- numbers. Usually, all numbers greater than four are lumped into "many". Thus, "at least 30 people speak FoodTongue" would simply be translated "many person talk food-tongue". When relatively small numbers must be named, as in specifying Mathcamp 06 / 07 / 08, the numbers will be added together ("four and two", etc.). There is a base-five system that uses zero through four, but it's not really in common use.

Brit: How is the language used? Is it used only in writing on certain internet sites? Or is it also spoken when you meet other speakers of the language?

Alan: FoodTongue is probably most often spoken on the Internet. Speakers often make blog entries or have IM conversations containing (or entirely in) FoodTongue. However, it may also be heard whenever speakers meet in person -- likely math and science competitions, university events, and Mathcamp itself, where it's seldom difficult to find someone speaking or teaching FoodTongue.

Brit: What was the reason for inventing the language? Was it started as a kind of experiment? There is something very secretive about the language. What is that all about?

Alan: The language started at Mathcamp 2004. I was not there, so my source is the first FoodTongue dictionary, distributed in that year. One evening, at a gathering of campers, the idea came up of trying to communicate entirely with food words. Soon the campers formed a basic vocabulary, and began teaching others. At that first meeting it was decided, as noted in the dictionary, that FoodTongue should not when possible be explained in other languages, which has remained a central principle. So it's not that the language is being kept secret, but that questions asked about it often cannot be answered.

Brit: Why have a principle to the effect that FoodTongue should not when possible be explained in other languages? Was it because you wanted the language to develop naturally?

Alan: As I said, I was not present at FoodTongue's creation, but I suspect that because it grew out of a suggestion to communicate entirely with food words, there was a desire to maintain that ideal by not bringing in other modes of verbal communication. As Waffle [Eric Wofsey] said, "The goal of Food Tongue is not to be understood. The goal is to communicate within a restricted framework." Or perhaps there is a tradition at Mathcamp to confuse people by not explaining things (but also to willingly demonstrate for those who show interest). Whatever the case, those who have learned FoodTongue since its inception have also been willing to continue upholding the principle. I think that as it's often brought up as a matter of following the spirit of the language, there's an element similar to suspension of disbelief, where speakers agree to abide by the rules or risk breaking the experience. It should also be noted that there is no restriction on discussing FoodTongue in other languages, only on explaining individual words.

Brit: So how did you manage to acquire FoodTongue prior to your first Mathcamp? Did you have a mentor?

Alan: The best way to teach FoodTongue without violating its spirit is in person, because gestures and other nonverbal communication are allowed. I was taught it before Mathcamp 2006 by a friend who had gone in 2005, and he demonstrated many of the words for me (for example, pointing at his watch to show the word for time). I should note that the *new* dictionary is written in FoodTongue -- the first dictionary, which I mentioned before, has two-way translations between English and FoodTongue. Clearly, this violated its own principle, so it was decided (I believe) that the dictionary should only be used by fluent speakers, as a reference. (The stated goal of the dictionary was to reunite some dialects that had arisen, for which breaking the principle was useful.)

Brit: What do you use the language to talk about? All sorts of topics? Math? The language itself?

Alan: FoodTongue generally doesn't have the vocabulary needed to delve deep into any field (although agglutination and calquing may produce approximations), but anything it has words for (or that words can be created for) is fair game, from talking about school to playing a card game. If anything, FoodTongue is particularly suited to Mathcamp culture; some words (grapefruit, smarties) refer to concepts specific to Mathcamp, and a few of those have become the primary means of naming those concepts (meatloaf, starfruit), at least among my friends. In practice, conversations may switch into and out of FoodTongue regardless of the subject at hand (or for gratuitous Mathcamp references).

Brit: How many people would you say master the language? Or is it hard to say? Can anyone join the forums where the language is spoken?

Alan: This is a tough question, and I doubt anyone has an accurate count. I personally know at least 30-40 people who would answer to their FoodTongue names; of those, up to a dozen are in my IM contacts and regularly converse with me in FoodTongue. But these are mostly Mathcamp alumni from 2007-2008; I know next to nothing about those from 2004-2006, and I'm also not the best-connected in my corner of Mathcamp. In theory, there's nothing preventing anyone from learning FoodTongue. However, the three main barriers are knowledge, interest, and teaching.

First, FoodTongue has probably spread little beyond Mathcamp, so many people who may be interested have never heard of it. Second, although many speakers try to teach friends the language, they are unfortunately often uninterested. Finally, as I noted, FoodTongue is best taught in person, and very difficult otherwise unless pictures are used (assuming the principle is adhered to). Most teaching thus goes on at Mathcamp, where each summer a handful of campers learn FoodTongue from alumni. This seems sufficient to keep the language alive, and indeed it shows no signs of dying just yet, but it is still limited in its reach.

Brit: It is obviously fun to learn a new language and to be able to speak a different language. But do you think that FoodTongue might also serve any practical purposes?

Alan: I have not seen any uses of FoodTongue for which it is particularly suited. That said, I think generally FoodTongue has as much practical use as any other natural language. It has sufficient vocabulary to describe everyday life concisely, especially at Mathcamp. But FoodTongue is probably spoken more for its principles and for itself than for any external purpose. Speakers enjoy being part of the community and communicating within the restrictions -- which is good, because short of aliens visiting Earth who share none of our physical features but somehow eat the same foods, I can't think of any situation in which it would uniquely useful.

Brit: Do you know of any other languages which are like FoodTongue? How does it compare to international languages like Esperanto? How does it compare to secret codes used throughout history by e.g. intelligence agencies?

Alan: So first of all, FoodTongue is certainly not an international language. It's spoken by people who already know English and much of it, from the vocabulary to word order, comes from English. The words are easy to learn because they're English food names, and the meanings are easy to remember largely because they have associations in English. (Of course, it's possible to create foreign equivalents by translating all of the food names; for example, "Pomme langue baguette-langue" is a valid sentence in French FoodTongue.)

I think the property that mapping a subset of a language (food words) into a complete language in itself sets FoodTongue apart from many languages. It may be similar to some codes, but it's not suitable for secrecy as it is -- many words are easy to figure out, and even if disguised as a grocery list it's usually obvious there's a hidden meaning. FoodTongue is probably most similar to games played by children in which they redefine words (for example, Tolkien's Animalic).

Brit: I once compared FoodTongue to so-called Lagadonian languages in which objects, properties, reletions etc. are names of themselves. Obviously, FoodTongue is not a Lagadonian language. However, I wonder whether there is any interesting similarity between FoodTongue and Lagadonian languages? Any thoughts?

Alan: As I noted before, it's possible to translate all the food words into another language, while expressing the same ideas. There may have been some dispute over the issue, but my view is that FoodTongue is theoretically independent of the words used to express it. That is, the concept of me is expressed by the concept of apple, not the English word "apple" (or the French word "pomme"), even if the concept was apple was chosen to represent the first-person pronoun because of associations of the English word "apple". It is thus possible to create a salad that's a valid FoodTongue sentence, though it may be a rather weird salad. Although the foods are abstract -- the concept of me is represented by the concept of apple, not a specific apple -- this interpretation is similar to the original Lagado, in that representations of concepts are not tied to any specific spoken, written, or physical form.

Brit: How do you predict that FoodTongue will evolve in the future?

Alan: As I understand it, after FoodTongue was invented at Mathcamp 2004, almost nobody learned it in 2005, but it became much more popular in 2006. In the first dictionary there are many words that have fallen out of use (some of which have been redefined), and there are many words that I use for which it includes no equivalent. Unlike with natural languages, once campers leave the population (graduate from Mathcamp), they maintain contact with friends from camp. A possible scenario is that every few years the FoodTongue spoken at Mathcamp will have shifted, and the alumni from those years as they move on will still speak it as they did as campers. So while the language continues to evolve at camp, multiple age-separated groups of speakers will form, still mutually intelligible but with visible differences in vocabulary and perhaps philosophy. I don't know how likely this is, though.

Brit: Do you have any advice to people interested in learning the language?

Alan: The best advice to people interested in FoodTongue is to contact someone who knows it. The wiki is designed so that it's possible to learn from scratch entirely using that site, but it's much easier to have a speaker walk you through (preferably in person). One of the things I learned from making the wiki is that there are significant numbers of people even at Mathcamp who are interested in the language but have not had a chance to learn -- so talk to a speaker, or someone who can refer you to one.

Brit: Thank you, Alan! To the interested reader: You can contact me via email if you want me to pass on your contact info to a speaker. No anonymous requests will be considered.

Inconsistency Theories of Semantic Paradox

Here is a short review I just wrote of "Inconsistency Theories of Semantic Paradox".

Relative Truth

I recently put together a special issue of Synthese on relative truth. It has just come out in print. Contributors include: David Capps, Andy Egan, Michael Glanzberg, Steven Hales, Max Kolbel, Peter Lasersohn, Michael Lynch, John MacFarlane, Daniel Massey, Sebastiano Moruzzi, Stephen Neale, Duncan Pritchard, Brian Weatherson and Crispin Wright.

MacFarlane, Kratzer-Conditionals and Knowing What You Ought to Do

At the AOC conference John MacFarlane gave a paper (co-authored with Niko Kolodny), defending an assessment-sensitive semantics for 'ought'. In MacFarlane-style assessment-sensitive semantics the truth-value of utterances will depend on features determined by the context of assessment and not just on features determined by the context of utterance and circumstance-shifting operators. Thus, my utterance of 'John ought to pay his bill' may have different truth-values, depending on who is considering it for truth.

The paper is very rich. I can't cover all the details. But the paper includes, as part of its defense, a Kratzer-inspired theory of indicative conditionals. On the theory in question, conditional 'ought'-statements are always narrow-scope rather than wide-scope. So, the conditional 'if I get drinks in the bar, then I ought to pay' has the correct form with the 'ought' taking narrow scope. 'It ought to be the case that if I get drinks in the bar, I pay', on the other hand, must be paraphrased as 'if I get drinks in the bar, I ought to pay'.

I have one concern about this account of conditional 'oughts', which piggybacks on John Broome's concern about narrow-scope conditional requirements. Consider the conditional 'ought' statement:

(1) If I (sincerely) assert the sentence 'there is a department meeting this afternoon', then I ought to believe that there is a department meeting this afternoon.

(1) seems true. After all, there is supposed to be a close connection between assertion and belief. But now consider the following scenario.

Scenario:
I have strong evidence that there is no department meeting this afternoon but I assert 'there is a department meeting this afternoon'.

Since I asserted that there is a department meeting this afternoon, it would seem that I ought to believe that there is a department meeting this afternoon. And since I have strong evidence that there is no department meeting this afternoon, it ought to be the case that I don't believe that there is a department meeting this afternoon. So, it ought to be the case that I believe and don't believe that there is a department meeting this afternoon. But this can't be right. So, something must have gone wrong.

MacFarlane rejects modus ponens. So, the following inference form is not unrestrictedly valid:

If I assert 'there is a department meeting this afternoon', then I ought to believe that there is one.
I assert 'there is a department meeting this afternoon
So I ought to believe that there is one

So he can avoid the unfortunate consequence. But there is an alternative move available. Broome's move. Broome allows for wide-scope 'ought' statements (or 'requirement' statements but I am here focusing on 'ought' statements). So, it is not quite right that if I assert 'there is a department meeting', then I ought to believe that there is one. Rather, it ought to be the case that if I assert 'there is a department meeting', then I believe that there is one. Moreover, detachment fails. So the following inference is invalid:

It ought to be the case that if I assert 'there is a department meeting', then I believe that there is one
I assert 'there is a department meeting'
Hence, I ought to believe that there is one

Of course, Broome could allow the following inferences:

It ought to be the case that if I assert 'there is a department meeting', then I believe that there is one
I ought to assert 'there is a department meeting'
Hence, I ought to believe that there is one

It ought to be the that if I assert 'there is a department meeting', then I believe that there is one
I assert 'there is a department meeting', and it is not the case that I ought not to assert 'there is a department meeting'
Hence, I ought to believe that there is one

My main concern with the MacFarlane move (i.e., his theory of conditionals) is that on the assumption that the first sentence is a narrow-scope 'ought' statement, the following instance of modus ponens seems exceedingly plausible:

If I assert 'there is a department meeting this afternoon', then I ought to believe that there is one.
I assert 'there is a department meeting this afternoon'
So I ought to believe that there is one

However, as we have seen, if we accept just this one instance of modus ponens, then we can derive a contradiction in the envisaged circumstances.

Here is a second potential worry about MacFarlane's (and Kolodny's) account. For MacFarlane, possible world-states are states compatible with what is known (by the assessor). The ideal world-states are a subset of the possible world-states (the set of the most ideal states). Now, consider 4-year old Mary. Her mom has told her that if her pants are on fire, she ought to pour water on them. She truly believes but doesn't *know* that if her pants are on fire, she ought to pour water on them, and she doesn't know that liquid hydrogen is distinct from water. So, in some of the possible world-states (where Mary is the assessor), water = liquid hydrogen and in others water is not identical to liquid hydrogen (since she doesn't know about the identity). Moreover, since she doesn't know that if her pants are on fire she ought to pour water on them, water = liquid hydrogen in some of the ideal possible world-states, and in some of those states, Mary's pants are on fire and she pours liquid hydrogen on her pants (causing an explosion). Assuming 'she ought to' and 'she is permitted to' are duals, it follows that if Mary's pants are on fire, she is permitted to pour liquid hydrogen on her pants, which seems unintuitive (even when she is the assessor).

For further discussion of MacFarlane's paper, click here and here.

S & P Open for Submissions

The new journal, Semantics and Pragmatics (S&P), edited by David Beaver and Kai von Fintel, is now open for submissions!

Strawsonian Definite Descriptions

Though I am definitely still a Russellian with respect to definite descriptions, I thought that Peter Lasersohn's "The Temperature Paradox as Evidence for a Presuppositional Analysis of Definite Descriptions", which I just read, made a quite convincing case for a Strawsonian treatment. I won't go into his reasons for preferring a Strawsonian treatment to the Russellian alternative but I do want to make a couple of remarks about Lasersohn's take on the Strawsonian approach.

The Strawsonian view is considered implausible by Millians because Millians naturally assume that if definite descriptions refer, then they contribute an individual to truth-conditions. So, 'the president' would contribute Bush to truth-conditions, which is quite implausible. Lasersohn's approach, however, allows for a more natural referential treatment of definite descriptions. The rough idea is that 'the' takes the intension of its argument and returns an extension at the world in question. So for the case of unembedded occurrences of 'the president', 'the' takes the intension of 'president' and returns Bush at the actual world. This version of the Strawsonian approach is considerably more plausible than the Millian version of the view. Add to this Lasersohn's very convincing arguments for adopting the Strawsonian approach and we have a very strong competitor to the Russellian stance.

I have often wondered the same thing

The Google Search function is a great resource for philosophers of language. Just realized that, contrary to what many thinkers have argued, 'wonder' does indeed select determiner phrases with a question meaning in natural language. Here are some search results:

(1) I’ve often wondered the same thing.
(2) Ever wondered the answer to any of these questions?
(3) Ever wondered the "right" way to describe swordfights?
(4) Stratford wondered the time line.
(5) If you've ever wondered the value of partner programs from software/hardware companies, let me give you my perspective from the front lines.
(6) Geisenberger wondered the price range.
(7) Hopkinson wondered the height of the tallest building in this neighborhood.

This datum, of course, has intrinsic interest. But what's the broader impact? Well, some thinkers have argued that determiner phrase complement clauses cannot be interpreted as concealed questions, because if they were, we should expect 'wonder' to select determiner phrases as complements. As 'wonder' does select determiner phrases as complements, determiner phrase complement clauses can (just maybe) be interpreted as concealed questions. Needless to say I was thrilled to discover this.

New Semantics Journal

Kai von Fintel and David Beaver recently announced that they are going to start a new semantics journal called Semantics and Pragmatics. S & P will be open access, which means that there will be free online access to published articles. von Fintel and Beaver have also started a new blog devoted to the project. Speaking of which, Beaver just reported that S & F has been officially accepted as an "affiliated journal of the Linguistic Society of America as part of their eLanguage initiative". So, if you're working in semantics, there are now two respectable open access journals where you can send your work, the other being Philosophers' Imprint.