# Algorithm for figuring out the pronunciation of a word

Wondering how we pronounce words. I feel like I learned this when I was a kid in school with all the language rules, but now I can't remember.

I am trying to think about how we pronounce words. How we actually mentally parse the letters and "sound it out". It seems a bit difficult, because it doesn't just move from left to right as you might imagine. From first look it seems we have to "look ahead" to figure out the sound of the current letters, even to the point of having to look at the entire word to figure out the sounding out.

For example, the word `phase`. We know what `ph` is always the `f` sound. But `ase` is `aze`, like `haze`. So we only know this I think because it is the word "phase", and that's how the whole word is pronounced. Or maybe it's because the word `face` already takes up that sound.

The word `chase` has the `face` sound. So `chase` and `phase` have a `s` vs. a `z` sound.

Then there are other words like `tie` and `field`. There are lots of `ie` words that are pronounced "fear" or "tear", "pierce", "fierce", "field", "siege", etc. But `tie` is like "I". You only know this because the word is short.

So I don't see the rules for pronouncing words. I imagine there are a lot of rules, but still, I am wondering what they are.

So my question is, if there is a resource to look at that outlines the rules for sounding out words, or if it is purely a learned thing that there are too many rules to write down.

Or maybe instead of all the specific rules (like `ph` is `f`, etc.), just the algorithm for moving through the word and applying the rules.

Then there are words like `tear` which have two pronunciations! `teer` and `tair`.

So my question is, if there is a resource to look at that outlines the rules for sounding out words, or if it is purely a learned thing that there are too many rules to write down.

I would say that the rules can't be written down, not exactly because there are too many of them, but because it's too difficult to ensure that the rules account for all the possible rare words and variant pronunciations. Any set of rules that you can find or make will probably have some accidental omissions. And of course, as you pointed out, there are words like tear that simply have ambiguous spellings: there just isn't enough information in the spelling alone to completely determine the pronunciation of the word. Situations like this are fairly common in English (e.g. the digraph "ow"/"ou" also has two pronunciations, so "cow" does not rhyme with "crow", and there's no way to figure this out without just memorizing it or memorizing the etymology of these words).

The glib answer to your question about a resource for this would be "a pronouncing dictionary". A lot of people are fond of stressing the irregularities of English spelling (no offense to people who like it, but I've grown tired of seeing people recommending the poem "The Chaos" as a supposed description of what English spelling is like) and it's true that there are a lot of specific words with unpredictable pronunciations. And this isn't only a feature of high-frequency words. For example, I only learned a few months ago about the word geoduck, which is pronounced "GOO-ey duck": there isn't any practical way to know this other than memorizing the spelling and pronunciation of this particular word. So any time you are faced with an unfamiliar word, there is a chance (although not a very large one) that it will have a pronunciation that just doesn't follow the rules.

That said, I think user6726 is right to say that many spelling–pronunciation pairs that seem unpredictable are not actually completely arbitrary.

There are a number of good books that give an overview of the most common spelling–pronunciation patterns in English (not a comprehensive description), and these patterns cover the pronunciation of most words.

I would recommend Greg Brooks' Dictionary of the British English Spelling System, which you can read for free online.

Another similar resource is A Survey of English Spelling, by Edward Carney (1994).

• Well, because we have geoducks aplenty here, I knew the word and its pronunciation well before I could write. Originally I thought it was a kind of duck. Aug 28, 2018 at 16:13

Unfortunately, any algorithm for determining the pronunciation from its spelling would be rife with special cases and exceptions. That's simply the nature of our spelling system. In the distant past (Old English) it was fully regular, but many, many things have changed since then.

There are some general rules, certainly, which can be found in resources for teaching English pronunciation (either materials for children or for ESL learners). But an unfortunate number of exceptions simply need to be memorized. For just a few of these, see the famous poem The Chaos.

• Wondering if you know of any thorough resources listing out pronunciation rules in this case. Will take a look at the poem, thank you. Aug 26, 2018 at 0:12
• Also wondering then how the metaphone algorithm works in this case, but that's probably another question. Aug 26, 2018 at 0:13
• @LancePollard Metaphone isn't designed to come up with the actual pronunciation: modern versions don't make any distinctions whatsoever between vowels, for example, which is how they deal with your tie/field problem. Similarly it doesn't distinguish voicing, which is how it deals with chase/phase. The purpose of Metaphone is rather to group words into vague categories based on how they sound.
– Draconis
Aug 26, 2018 at 0:28
• Some years ago a computational linguist friend told me it required ~2,000 rules to create a system that could consistently and successfully convert the front page text of a daily newspaper into speech. Aug 26, 2018 at 22:55

It depends on which word, and what person. At the most basic level, you have learned the various rules of English (subjects before verbs, adjectives before nouns, and so on), and also the morphemes of the language, like "cat". When you learn "cat" you learn the pronunciation ([kæt]) and the real-world referent of the word. If you grew up speaking English, you did that as a child, by age 3. If you learned English as a second language, then I'd have to ask a bunch of questions. You did not learn the word "phase" the same way you learned "cat". You probably learned the verb "tear" and the noun "tear" the same way you learned "cat", though later.

A lot of vocabulary is only marginally picked up from ordinary speech ("phase" might be), and certain words ("hammamelidanthemum") are almost always learned from writing. Some of these words are so infrequent in speech that there is no one common pronunciation. There are at least two ways to pronounce that word (həˈmæmələˌdænθəməm ~ ˈhæməmæləˌdænθəməm). There are strategies that English speakers acquire for converting spelling into pronunciation, but they are highly variable (e.g. is "Pegnataro's" pronounced [pɛgnəˈtɛroz] or [pɛnjəˈtɑroz]? – depends on whether you know any Italian). It would be an interesting research project, to try to sort out a system of rules that computes the probability of a given pronunciation given a spelling. Although those rules would be complex, I think they could reduce the seeming chaos, if the conditioning factors were make explicit. The difference [həˈmæmdələˌdænθəməm] versus [ˈhæməmæləˌdænθəməm] stems from treating the word as bi-morphemic in the former case (hammamelid-anthemum) and as a very long root in the latter (with alternating stress). Ultimately, there will be completely arbitrary cases such as the fact that "Sequim" is pronounced [skwɪm], but I actually think the completely arbitrary part of spelling-to-pronunciation is not huge. So for example, "horse" vs. "worse" may seem like random chaos, unless you look at other words with "wor..." like work, wort, worry, worm – there is a sub-generalization that makes "worse" regular.

You write: "I feel like I learned this when I was a kid in school with all the language rules, but now I can't remember." But your feeling about this must obviously be wrong, since people knew how to pronounce words before there were schools. And even today, many people who never went to school can pronounce words perfectly well.

It's quite a puzzle.

This article attempts to explain the rules of mapping a spelling onto pronunciation in English. http://www.zompist.com/spell.html

• Welcome to Linguistics! Whilst this may theoretically answer the question, it would be preferable to include the essential parts of the answer here, and provide the link for reference. Sep 2, 2018 at 18:10