Remembering and Musing: or Conan, the Barbarian

On last Thursday's Talk of the Nation Neal Conan had a conversation with Farhad Manjoo about password security.

There was some good advice on the show. For instance, don't choose <123456> as your password. No really. Don't.

The best passwords are of course also the hardest to remember. Numbers and symbols and primarily unintelligible strings of letters are the way to go. So how to remember such a random string? According to Manjoo you should use a mnemonic device. He uses a pronunciation somewhere between [nəmɑnɪk] and [nʊmɑnɪk]: between a schwa and the vowel in book. Not quite the pronunciation with first syllable rhyming with boo zoo and you (that [u] possibly from the influence of a word like pneumatic).

As I pronounce mnemonic the first vowel falls somewhere between the [i] in beet and the [ɪ] in bit. Manjoo and I pronounce the word almost identically with a slight difference in the front/back place of the first vowel. The sequence of vowels and consonants is the same: CVCVCVC.

Conan also mentioned mnemonic devices. His pronunciation is quite distinct from mine. In fact it's a pronunciation I'd never heard before [mɛmnɑɾɪk] as if the word was spelled <memnotic>.

I thought this might be a simple performance error, not a fixed pronunciation, but he uses the same pronunciation later in the segment. It appears to be his somewhat intentional pronunciation of the word.

There's an obvious effect of the initial <mn> in the spelling of the word. An onset cluster not possible in English. Of course when the [m] is the coda of one syllable and the [n] is an onset of another that's a perfectly acceptable sequence because it's no longer a true cluster. That's one possible fix for a spelling confound on phonology.

But I suspect it's more than just a fix. I wonder if it's not also the result of a couple other influences: memory and hypnotic. Memory is obviously a relevant word probably assumed to be a related form. Hypnotic isn't as direct a connection, but the relevance of psychological terms and 'mindwork' types of tricks and strategies pushes it forward as an influence. Once you've got the initial mem- and you need to put the 'mn' sequence in there and you know the word ends with [ɑCɪk], the [ɑɾɪk] of nearby hypnotic can easily push aside the [ɑnɪk] of already corrupted mnemonic. Perhaps.

I didn't find loads out there. A Google™ search for "memnotic" brings up almost 5000 items as raw hit results, not all of them relevant or even accurate. "Memnotic device" brings up only about a dozen.

It's an interesting blend.

You two behave or I'm gonna separate you

The pronunciation of many place names is difficult to deduce because they are long which confuses the scansion; they are full of digraphs which combined with confused scansion confuses the vowels; and they are unfamiliar. It's the same thing that makes my last name so daunting for so many people. I say it [ˌkʰow.vəˈɻuː.bjəs]

But consider a place name like Hamtramck a city geographically embraced by Detroit. There's no real ambiguity there. It looks pretty clear: [hæmtræmk]. But the final consonant cluster is a problem in English so it has to be fixed. That's where the ambiguity comes in. We don't like a nasal and following adjacent consonant to have contrasting places of articulation when syllabification forces us to make them a cluster. So the labial [m] and velar [k] doesn't work for us word finally. How to fix it?

It could be any of several repairs including:

• simple progressive place assimilation--[hæmtɹæŋk]: this has worked in tank and rank and tin cup (even tho it's not word final) and is one of the most common repairs. But that's a coronal [n] assimilating. Labial [m] is not as willing to move.


• excrescent progressive place assimilation--[hæmtɹæmŋk]: but this is very unlikely because there's now a more complex cluster and the adjacent mŋ still needs to be fixed. One likely repair would be to delete the [m] leading to the first repair as a final form. It's not a likely extra step.


• excrescent progressive manner assimilation--[hæmtɹæmpk]: the m takes the voiceless stop feature of the [k] which sometimes occurs with words like tense [tɛnts] and answer [æntsɻ] but isn't a fix because it leaves adjacent [pk]. If the [k] is then deleted we would come around to another possible repair.


• regressive assimilation of place--[hæmtɹæmp]: this is more simple than the previous mentioned repair; it would require one step: k → p. But that's not a likely repair. The [k] likes to stay put. It's not like those peregrine coronal nasals that assimilate to place of articulation with relative ease.

But the local fix is none of these. Call epenthesis to the rescue. The city is pronounced as if it was spelled 'Hamtramick' with primary stress on the 2nd syllable--[hæmˈtɻæmɪk]. Think of the rhythm of ham sandwich.

This is not so surprising a repair. About a year ago Eric Baković put up a post on phonoloblog describing his wife's playful truncation of berry names. When she encountered a similar problem she avoided assimilation and turned to epenthesis as a repair. Baković writes

assimilation (*[bɔɪzəmbz]) is independently blocked by whatever is responsible for the lack of word-final noncoronal nasal-voiced stop clusters ([mb] and [ŋg]) in English, deletion (as in bomb [ba:m]) is also blocked because the point of the truncation is for the [b] of berries to be retained, and so epenthesis is employed as the last resort. At least, that's an interesting way to look at what may just be Karen’s funny way of talking.

Epenthesis is not a favoured repair for illegal clusters. It may be too 'obvious' a fix. But it apparently has its place. Stretching at least from San Diego to Detroit.

------------
Lately there's been a tiny burst of discussion regarding place names and their pronunciation. Recently the LA Times ran an article about MissPronouncer.com: Jackie Johnson's website meant to help with those hard-to-pronounce Wisconsin names. Ben Zimmer and Nancy Friedman alerted Mr. Verb to the article and he covered it in a post. Then the topic showed up on the ADS-L board and subscribers are still throwing out all sorts of observations and questions about the words and attitudes involved.

Aspiration. It's all about persbective.

A friend of mine recently told the story (on her own web log) about teaching her ESL class. For some reason she thought it appropriate to teach them the word "supercalifragilisticexbealidocious." At the very least this word can generate an interesting discussion of what makes a word a word. Would we consider this a lexeme? What is semantic content to this word? I'm pretty sure it's an adjective. And it probably means something like "super." I can't remember the lyrics of that part of Mary Poppins but I remember the tune clearly. It gives me headache. Is there a lyric something like 'lumdiddle-iddle' in there?

So what makes this of interest to me is my friend's admission that she doesn't know how the word is spelled but when her students asked her to write it out for them she chose to spell it "supercalifragilisticexbealidocious." And sure. It works. I knew what word she meant when I read it.

But it's not the spelling I would have used. There's a letter sequence in there that I noticed when I read it: the "xbe". My guess would have chosen "kspi" for that sequence. Why do she and I see that sequence differently? Because I'm a big believer in onset maximization. To say it as simply as I know how...

In English, when stops are the initial sound in a word, they are aspirated. There's a slight puff of air audible right after the release. When an 's' comes before the stop there is no puff of air. (Well -- it's less audible.) Compare the following:

The /p/ in pin and spin
The /k/ in kit and skit
The /t/ in tab and stab

Since English does not differentiate between an aspirated and unaspirated voiceless stop the +/- aspiration alternation is more likely seen as a voiceless/voiced alternation. That is, since voiced stops are not aspirated, an initial unaspirated voiceless stop sounds like a voiced stop.

Now consider the two spellings I suggested for our long word above. Her use of 'x' makes me think that she syllabified the 'ks' as a coda cluster. So her unaspirated bilabial stop sounded like a voiced bilabial stop because she analyzed the onset as a single consonant [...eks.bi.æ.lə...]. Analyzing the [s] as part of the coda she would have expected a [p] to be aspirated [...eks.p^hi.æ.lə...].

Because I love onset maximization I have just assumed that the syllabification is [...ek.spi.æ.lə...]. After the [s] an unaspirated [p] is what I'd expect.

After searching around I find that "xp" is probably the correct spelling. Even so...I syllabify "expert" with a maximized onset and an unaspirated [p].

A while ago on the ADS listserve there was discussion of a local pronunciation of "Wisconsin" heard by some as "Wisgonsin." Several people called attention to the aspiration alternation and suggested that it's primarily an alternation of syllabification. Where most will syllabify the word [wis.k^han.sən] some locals (and surely some non-locals as well) will syllabify it [wi.skan.sən]. Note the difference in aspiration.

Rain in Espain Estays Mainly in the Plain

To answer my own question - my niece (not the youngest, Olivia; the second youngest, Calen) used to say tutumber instead of cucumber. and instead of saying Buffy she says Bussy. In fact I've noticed that she always switches the /f/ to /s/.

This switch from /k/ to /t/ and /f/ to /s/ seems almost random. but let's note the obvious methodology first. All these sounds are voiceless. The voiceless /k/ remains voiceless (the voiced /k/ would be /g/ the voiced /t/ would be /d/). The f remains likewise (voiced it would be /v/ and the /s/ would be /z/). The stop /k/ remains a stop. The fricative /f/ remains a fricative.

Here's the interesting similarity. /t/ and /s/ belong to a very powerful class of sounds: coronals. A coronal is articulated using the tip of the tongue. in our english language the coronals are /t/ /d/ /s/ /z/ /sh/ /zh/ /ch/ /j/ /n/ (and depending on your analysis /l/ /r/). These are powerful sounds because of the rules they seem to ignore. Basically coronals often occur where other sounds just like them - with the excepetion of coronality - would not. Such as before another consonant at the beginning of a word.

E.g. - in English we never begin a word /fk/. fkit is difficult to say and many would argue it's impossible to say it (it's not of course - it's just outside the bounds of English phonological grammar). But skit is perfectly acceptable. What is the difference between /f/ and /s/? All we need in order to explain this pattern the coronal distinction. /s/ is coronal and /f/ is labial.

Further evidence of the same characteristics: flip the letters around at the end of a word. Kicks is easy to say - kickf is not so natural (though not so difficult as fkit is it? - more on that in another post). Mixing the data around - rats: easy. ratf: not so easy. And let's look beyond /s/ and /f/.

act - pact/packed - stacked - tact/tacked: all these allow the /kt/ ending. Can anyone think of a word that ends /kp/? how about /kf/? I can't.

Simply put: in most languages coronals show up in a greater variety of places than other classes. Note: they don't have a complete pass. For instance Spanish does not allow the sp- st- sk- onsets like English does. That's why an accented speaker will often throw an e- before that cluster. So stop becomes estop.

But compare languages that have more flexibility than does English - German allows sht- very easily at the beginning of a word while there are only a few words in English that use the cluster - all recently borrowed; Russian allows fs- at the beginning of words - English does not. And the fascinating Imdlawn Tashlhiyt dialect of Berber (ITB) allows such odd words as txznt tmsxt and tftkt. All 2 syllable words. Why are examples like this so rare?

There is plenty of evidence that this has to do with a naturally articulate tongue tip in all humans. So while mama is a very common first word - dada is also common. And all this probably led to these neonatal phonetic hobbyhorses becoming our words for mothers and fathers - or grandmothers and grandfathers. These words - or very similar ones - are universally associated with these same concepts. And although the /m/ and /p/ sounds are so common in early speech they are not found as ubiquitously as these peregrine coronals.

Why? Probably because ease of articulation does not translate directly to flexibility of articulation. Babies can say mama and papa easily enough - but the lips don't move far. It's hard to imagine the lips moving up or down or forwards or backwards to fit easily alongside other sounds. But the tip of the tongue is as agile and nimble as a fingertip. People can curl their tongue - they can twist it - they can make a cloverleaf (my latest trick) - they can touch the front of their incisors and the back of their wisdom teeth.

I like to brag that i can even touch my uvula. I've been able to ever since I was in junior high.

Now try doing all that with the side or the back of your tongue. With your lips.

Try saying pkat.