Stress patterns as formal languages March 5, An alphabet for stress Formal language: set of strings over some alphabet Human language: a system mapping between sound and meaning; it allows certain outputs and not others Insight: every human language can be seen as a formal language, by simply looking at its set of allowable outputs We have seen that, if you look at strings of words in a sentence, in at least some languages, the set does not seem to be a regular language; for this reason (among others) we think that the mechanism that handles syntax must be more powerful than an FSA Linguistics is divided up into different sub-fields for a reason: we find interesting, complex systems of mostly unconscious knowledge when we focus on different aspects of sentences (even things, like meaning, that don t have to do so much with what is uttered, but what is understood or intended) and we find they look interestingly different Let s probe how with a test case: word stress What aspects of the utterance does stress deal with? We said it s relative prominence, and we meant relative prominence of syllables (we mostly subbed in vowels for syllables) So let s look at a word as a sequence of syllables, each either stressed or not, and each stressed one either primary or secondary stressed Alphabet {,,} for unstressed, primary, and secondary stressed Some examples Maranungku tíralk saliva C V CVCC 3 mérepèt beard C V CVC V C 4 yángarmàta the Pleiades C V CVC V CV 5 lángkaràtetì prawn C V CCVC V CVC V 6 wélepènemànta type of duck C V CVC V CVC V CCV
Set of all strings () (ε + ). Let s try and write an FSA for it: q q q q q q Actually, we know something else about Maranungku, which is that there aren t any monosyllabic words. So we could even fit that in, too. Araucanian q q q q q q 3 q 3 q wulé tomorrow CVC V 3 tipánto year CVC V CCV 4 elúmuyù give us CVC V CVC V 5 elúaènew he will give me CVC V V V CVC 6 kimúbalùwulày he pretended not to know CVC V CVC V CVC V C Weri Nintíp bee CVCC V C 3 kùlipú arm hair C V CVC V 4 ulùamít mist VC V VC V C 5 àkunètepál times V CVC V CVC V C
Warao tíra woman C V CV 3 koránu drink it! CVC V CV 4 rùhunáe he sat down C V CVC V V 5 yiwàranáe he finished it VC V CVC V V 6 7 (no examples) 8 yàpurùkitàneháse verily to climb V CVC V CVC V CVC V CV 9 enàhoròahàkutái the one who caused him to eat VC V CVC V VC V CVC V V Quantity-sensitivity If we just had this alphabet, we could not handle quantity-sensitivity. Let s then use the following alphabet: { L, L, L, H, H, H}. Now the lower element indicates whether the syllable is light () or heavy (). Remember that, in some languages, heavy means long vowel ; in other languages, it means consonant in the syllable coda (which, when we were trying to capture these things as rules that operated on consonants and vowels, we captured as a sequence CC or C# although we said that was probably only an approximation, because in some cases we also need information about what the sequence of consonants is). Now, before, we were trying to write rules that would give us the stress patterns just by looking at the sequence of C/V/V:, without assuming that we had syllabified input marked as heavy/light. So, if the language was 3
quantity sensitive, we had to put long or followed by a consonant right in the rules. But now we are assuming that we don t have to do that. Does that mean that whatever we find out from looking at the FSA version of these rules is invalid? Think about this. We will think about this later. Aklan pitú seven CVC V sugúd room CVC V C bítbìt carry C V CC V C hámbàg speak C V CC V C gástà spend C V CC V bisá kiss CVC V 3 pàligús bathe C V CVC V C 3 Pasírtàr lucky CVC V CC V C 4 sugùgupún servant CVC V CVC V C 4 PatùbaNán genitals CVC V CVC V C 4 màpanísdàp will go fishing C V CVC V CC V C 4 màgmagánhùd more than two siblings C V CCVC V CC V C 5 màtinàmarún being lazy C V CVC V CVC V L L L H H H H H H L L L L L H L H H L L L H L L L H L L H H HL H H LL L L H We ve assumed that this generalizes to the pattern described in Bruce Hayes dissertation: [Secondary stress] is assigned recursively to the syllable immediately preceding a stress if that syllable is long, and two syllables to the left of a stress if the syllable immediately preceding the stress is short. Notice that apart from the last two syllables, there is no evidence above for heavy syllables attracting stress, though. That is, there s no evidence above for heavy syllables attracting stress in the middle of the word. Let s assume this is true about Aklan. Then we predict forms such as: C V CC V CCVC V C V CC V CVC V C V CC V CC V C C V C V CVC V C V C V CC V CC V C H H L L H L L L H H H LL L L L H H H The intuition is as follows: H is special. It can appear in the same positions L can appear in (L _L, H _L ), but it can also appear in special positions (H _H, L _H ). However, the sequence L H should always be ungrammatical. Here is an FST for Aklan: 4
It s easier to see if we reverse it. After we have taken care of the initial stress, we want to say that we are in a state where we read alternating stresses. If we read an H in that state, we stay in that state. 5
ḋarábt CVC V CC LS ijaqmá:l CVCC V: C HS 3 wálad C V CVC L H 4 ráija: C V CV : L H 5 híya C V CV L L 6 3 ka:tába CV:C V CV HL L 7 3 mustášfa: CVCC V CCV: HH H 8 3 muqállim CVC V CCVC LH H 9 3 muqá:til CVC V :CVC LH H 3 ša:bá:tun CV:C V :CVC HH H 3 qattálat CVCC V CVC HL H 3 maktábah CVCC V CVC HL H 3 4 katabáta: CVCVC V CV: LLL H 4 4 šajarátun CVCVC V CVC LLL H 5 5 ijadwiyatúhu CVCCVCVC V CV HLLL L 6 5 murtabiṫátun CVCCVCVC V CVC HLLL H 7 5 bulahníyatun CVCVCC V CVCVC LHL LH 8 3 kátaba C V CVCV L LL 9 3 šájarah C V CVCVC L LH 4 ijinkásara CVCC V CVCV HL LL Some examples of impossible stress systems Quaternary stress 6
Binary/ternary stress Limit four stresses in a word 7