Why is prosodic structure needed Evidence from pphrase construction in Romance

1
Why is prosodic structure needed? Evidence from
p-phrase construction in Romance

• Pilar Prieto
• ICREA Research Professor
• and Universitat Autònoma de Barcelona

2
Factors affecting phrasing

• (Direct or indirect effect of) syntax (Selkirk,
  1978, 1980 Nespor and Vogel, 1986)
• Information structure, constituent weight, speech
  rate (Nespor and Vogel, 1986 Steedman, 1991
  Oliva, 1992 Truckenbrodt, 1999, Selkirk, 2000)
• Language-specific tendencies (Prieto, 1997
  Nibert, 2000 Frota, 2000 DImperio, 2002).

3
The Prosodic Phonology view

• Work on the prediction of phonological phrases
  (?- or p-phrases) has highlighted the role of
  syntax
• Alignment to syntactic constituents or heads
  (Selkirk 1986, 1995, 2000, Nespor Vogel 1986)
• Cohesional demands on maximal projections. XPs
  should be contained in a phonological phrase
  (Truckenbrodt 1995, 1999)
• Non-branching of constituents in p-restructuring
  (Nespor Vogel 1986)

4
End-based theory of syntax-prosody relation

• Alignment of left or right edges of XPs (max.
  projections) or heads of XPs with edges of ? or
  p-phrases (Selkirk 1986, 1995, 2000)
• Constraints couched in OTs Generalized Alignment
  (McCarthy and Prince 1993)
• ALIGN XP,R Align right edge of XP to right edge
  of P
• ALIGN XP,L Align left edge of XP to left edge
  of P

5
Cases of suppression of edge alignment

• Tonal migration in Chizigula lexical floating H
  tones migrate to the penult position of the
  phrase (Kenstowicz and Kisseberth 1990)
• H
• a. nambikilaVmvyeleNPnyamaNP VP
• b. (nambikila mvyele nyáma)?
• I am cooking the woman meat

6
Cohesional constraints (Truckenbrodt 1995, 1999)

• Extension to edge-alignment. ALIGN-XP is a ranked
  constraint whose effects may be suppressed by
  another syntax mapping constraint (Truckenbrodt
  1995)
• WRAP-XP Each syntactic XP should be contained in
  a p-phrase (Truckenbrodt 1995, 1999).
• the same size or a larger size

7

• Align XP, R gtgt Wrap XP (Chimuiini)
• Wrap-XP gtgt Align XP, R (Chizigula)

8
Nespor Vogels mapping rules for Italian

• ?-construction Xhead (N, V, A)
• ?-restructuring Phonological phrases can undergo
  restructuring if they happen to be the first
  nonbranching complement on the right side
  (recursive side in Italian)
• (Hanno)? (dei caribú piccoli)?
• (Hanno)? (dei caribú)? (molto piccoli)?

9
Ghinis (1993) observations

• Importance of notions such as average weight and
  balancing in Italian
• a. (Prenderà granchi)? (di ogni specie)?
• He will catch every kind of crab
• b. (Comprerá mappe)? (di cittá molto
  antiche)?
• He will buy maps of very old towns
• c. (Soffrirá pane)? (incredibilmente dure)?
• He will feel a lot of pain
• d. (Osserverá stelle)? (di ogni dimensione)?
• He will observe stars of any dimension

10
Ghinis proposal

• ?-Formation (Ghini 1993a96)
• ?-domain formation. The domain of ?-formation is
  delimited by right-edge Xmax-boundaries
• ?-formation. Phonological words included in a
  string delimited by ?-domain formation are
  distributed according to the principles of
• i. Uniformity and average weight (binary length
  preference)
• ii. Symmetry (weight symmetry at both edges)
• iii. Increasing units (p-phrases on the
  recursive side are longer)
• Example (?)? (?)? (?)? (?)? ? (??)? (??)?
• (Prenderá granchi)? (di ogni
  specie)?

11
Binary size effects

• Ghini (1993a, b) for Italian
• Selkirk (2000) for English
• (i) Binary Maximum (MaP)
• A major phrase may consist of at most two
  minor/accentual phrases
•   (ii) Binary Minimum (MaP)
• A major phrase may consist of at least two
  minor/accentual phrases
• Sandalo Truckenbrodt (2002) for Brazilian
  Portuguese
• Max-Bin (Sandalo Truckenbrodt 2002295)
• P-phrases consist of maximally two prosodic
  words

12
Why study Romance p-phrasing?

• Importance of prosodic factors in phrasing.
• In VP and NP projections, size effects matter
• (Menjaré pastissos)? (de xocolata amarga)?
• I will eat cakes of bittersweet chocolate
• Catalan SVO utterances phrased as (SV)(O) when
  objects are long (Prieto 1997, Elordieta et al
  2003)
• (La Maria bevia)? (aigua destillada)?
• Mary used to drink distilled water
• Catalan phrasing reported to be affected by clash
  (Oliva 1992)
• ?(La Maria beu)? (aigua destillada)? ? (beu
  aigua)?
• Mary drinks distilled water

13
Two case studies

• Case study I. Investigate the role of prosodic
  and syntactic constraints on phrasing in Catalan.
• Prieto, Pilar. Syntactic and eurhythmic
  constraints on phrasing decisions in Catalan,
  Workshop on Intonational Boundaries, GLOW 2003,
  special volume of Studia Linguistica.
• Case study II. Investigate the influence of
  syntactic/prosodic branching and constituent
  length (in ?s) on phrasing in 4 Romance
  languages
• DImperio, M, G. Elordieta, S. Frota, P.
  Prieto and M. Vigário Intonational Phrasing and
  Constituent Length in Romance, ICPHS 2003, PaPI
  2003.

14
Case study I. Catalan

• Investigate the role of eurhythmic conditions in
  Catalan and their interactions with syntactic
  constraints.
• To examine the influence of both syntactic and
  prosodic factors on prosodic boundary placement
  in Catalan.
• Test syntactic alignment conditions on SVO
  groupings, VP and NP projections. Test behavior
  of PP and AdjP with different level attachments.
• Test size and clash effects on SVO groupings, VP
  and NP projections, etc

15
Sources of data

• Main source
• Judgments/productions on phrasing divisions of a
  wide variety of sentence structures.
• Four linguists native speakers of Central Catalan
  acted as judges
• Indications possible phrasings for normal speech
  rate, out-of-the-blue sentences.
• A production database of SVO sentences for
  Catalan, Italian, Portuguese, and Spanish
  (reported in Elordieta, Frota, Prieto, Vigário
  2003).

16
Arguments for p-constituency in Catalan

• Tonal and stress phenomena
• Final prominence More prominent stress at end of
  p-phrase
• Continuation rise (or other boundary tones) at
  end of p-constituents.
• Other phonetic boundary effects at end of
  p-phrase lengthening, lower intensity, optional
  pause, resetting of pitch.
• No blocking or triggering of postlexical rules
  (It. Radop. Sintattico or Port Stress Retraction).

17
Waveforms and Fo contours of the utterances (La
nena mirava)? (la melmelada meravellosa)? and
(La boliviana de Badalona)? (rememorava la noia)?.
18
VP Projections, I

• Wrapping effects (similar to Brazilian
  Portuguese Sandalo Truckenbrodt 2002)
• V NPNPVP ? (V NP)? V N PPNPVP ?
  (V)? (N PP)?
• a. (Comprava mapes)? a. (Comprava)? (mapes de
  Barcelona)?
• He used to buy maps He used to buy maps
  of Barcelona
•  
• b (Menjaré pastissos)? b. (Menjaré)? (pastissos
  de xocolata)?
• I will eat cakes I will eat chocolate
  cakes
• c. (El nomenaren professor)? c (El
  nomenaren)? (professor de filologia)?
• They appointed him professor They
  appointed him philology professor
• d. (Sortiran els volums)? d. (Sortiran)? (els
  volums importants)?
• The volumes came out The important
  volumes came out

19
VP Projections, II

• Binarity effects (similar to Italian Ghini
  1993a,b)
• V N PP APAPPPNPVP ? (V N)? (PP AP)?
• a. (Comprava mapes)? (de la Barcelona antiga)?
• He used to buy maps of old Barcelona
•  
• b. (Menjaré pastissos)? (de xocolata amarga)?
• I will eat cakes of bittersweet chocolate
•  
• c. (El nomenaren professor)? (de filologia
  romànica)?
• He was appointed professor of Romance
  philology
•  
• d. (Sortiran els volums)? (importants de
  veritat)?
• The really important volumes will come out
•  

20
VP Projections, III

• Alignment effects non-local PP attachment
• V NPNP PPPPVP ? (VNP)? (PP)?
•   a. (Comprava mapes)? (per a lAnna)?
• He/she used to buy maps for Anna
•   b. (Va donar el llibre)? (a la Maria)?
• He/she gave the book to Mary
• c. (Va enviar petonets)? (a lAnna)?
• He/she sent kisses to Anna
• Cf. (Comprava)? (mapes de Barcelona)? vs.
  (Comprava mapes)? (per a lAnna)? ((Comprava)?
  (mapes per a lAnna)?)

21

• Alignment effects
• V NP PPPPNP PPPPVP ? (V)? (NP PP)?
  (PP)?
• a. (Comprava)? (mapes de Barcelona)? (per a
  lAnna)?
• Also (Comprava mapes de Barcelona)? (per a
  lAnna)? He/she used to buy Barcelona maps for
  Anna
• b. (Va donar )? (el llibre de contes)? (a la
  Maria)?
• Also (Va donar el llibre de contes)? (a la
  Maria)?
• He/she gave the story book to Mary
• c. (Va enviar )? (petonets dolços)? (a
  lAnna)?
• Also (Va enviar petonets dolços)? (a lAnna)?
• He/she sent sweet kisses to Anna

22
OT Analysis of VP Projections
 

• Align-XP,R ltgt Max-Bin gtgt Wrap-XP (free
  ranking)
• Subordinate effects of Wrap-XP Max-Bin gtgt
  Wrap-XP

 
 
 
23

• Align-XP,R gtgt Wrap-XP
• Align-XP,R, Max-Bin gtgt Wrap-XP
• Restricted version of Max-Bin (to the end of the
  sentence) would favor candidate c.

24
NP Projections, I

• Wrapping effects same effects as in VPs
• N PPPPNP ? (N PP)? N PPPPPPPPNP ?
  (N PP)? (PP)?
• a. (Una pila de mapes)? a. (Una pila)? (de mapes
  de Barcelona)?
• A stack of maps A stack of maps of old
  Barcelona
•  
• b. (Una ampolla de vi)? b. (Una ampolla)? (de vi
  rosat)?
• A bottle of wine A bottle of rose wine
•  
• c. (Un anell de plata)? c. (Un anell)? (de
  plata suïssa)?
• A silver ring A ring made of Swiss silver
•  
• d. (Iogurts de la Fageda)? d. (Iogurts)? (de la
  Fageda den Jordà)?
• La Fageda iogurts La Fageda den Jordà
  iogurts

25
NP Projections, II

• Binarity requirements
• N PP PP APAPPPPPNP ? (N PP)? (PP AP)?
•   a. (Una pila de mapes)? (de la Barcelona
  antiga)?
• A stack of maps of old Barcelona
• b. (Una ampolla de vi)? (força rosat)?
• A bottle of pretty rose wine
• c. (Un anell de plata)? (força bonica)?
• A ring made of very nice silver
• d. (Iogurts de llimona)? (de la Fageda den
  Jordà)?
• La Fageda den Jordà lemon iogurts  

26
NP Projections, III

• Alignment effects local vs. non-local AP
  attachments
• Non-local attachment Local attachment
• N PPPP APAPNP N PP APAP PPVP
  
•   (N PP)? (AP)? (N)? (PP AP)?
• a. (Una ampolla de vi)? (rosada)? a. (Una
  ampolla)? (de vi rosat)?
• A pink bottle of wine A bottle of rose wine
• b. (Un barret de nen)? (argentí)? b. (Un
  barret)? (de nen argentí)?
• A boys Argentinian hat A hat from an
  Argentinian boy
•   
• c. (Un anell de plata)? (suís)? c. (Un anell)?
  (de plata suïssa)?
• A Swiss silver ring A ring made of Swiss
  silver

27

• More alignment/weight effects
• N PP APAPPP PPPPNP ? (N)? (PP AP)?
  (PP)?
• a. (Una pila)? (de mapes antics)? (per a
  lAnna)?
• also (Una pila de mapes antics)? (per a
  lAnna)?
• A stack of old maps for Anna
•  
• b. (Una ampolla)? (de vi rosat)? (de Londres)?
• also (Una ampolla de vi rosat)? (de Londres)?
• A bottle of rose wine from London
•  
• c. (Un anell)? (de plata suïssa)? (per a la meva
  mare)?
• also (Un anell de plata suïssa)? (per a la
  meva mare)?
• A ring made of Swiss silver for my mother

28
OT Analysis of NP Projections

• Align-XP,R ltgt Max-Bin gtgt Wrap-XP (free
  ranking)
• Subordinate effects of Wrap-XP Max-Bin gtgt Wrap-XP

29

• Align-XP,R gtgt Wrap-XP
• Align-XP,R, Max-Bin gtgt Wrap-XP
• Restricted version of Max-Bin (to the end) would
  favor candidate c.

30
Infl/Comp Projections

• Minimality effects
• Catalan Braz Port. (ST 200286)
• a. (La nena demana)? a. (Café)? (queima)?
• The little girl wants some Coffee burns
•  
• b.    (El president pateix)? b. (O José)?
  (dança)?
• The president is suffering José dances
• Minimality requirements on utterances
• Min-Phrase
• An utterance has to be minimally parsed into one
  p-phrase containing two prosodic words (that is,
  ((??)?)U better than ((?)? (?)?)U

31

• Minimality requirements on utterances

32

• ALIGNMENT/WEIGHT effects
• (S)? (VO)?
• a. (La nena)? (demana els regals)?
• The little girl asks for her gifts
• b. (El periodista)? (comunicà la notícia)?
• The journalist reported the news
•  
• (S)? (V)? (O)?
•   a. slow speech (La nena)? (demana)? (els
  regals de Reis)?
• normal/rapid speech (La nena demana)? (els
  regals de Reis)?
• (La nena)? (demana els regals de
  Reis)?
• b. slow speech (El periodista)? (comunicà)? (la
  notícia del dia)? normal/rapid speech (El
  periodista comunicà)? (la notícia del dia)?
• (El periodista )? (comunicà la notícia del
  dia)?
•  

33

• Basic hierarchy
• Slow speech phrasing derived by basic hierarchy

34

• Normal/rapid renderings are derived by adding a
  constraint which reduces the number of phrases in
  an utterance (Min-N-phrases)

35

• WRAPPING effect on Subjects
• (S)? (VO)?
• a. (La casa de Pineda)? (ha quedat buida)?
• The Pineda house is empty
•   b. (El president de la Generalitat)? (pateix)?
• The president of the Generalitat is suffering
•   c. (La nena de la Maria)? (no vol venir)?
• Marys daughter does not want to come
•    
• a.  (La casa de Pineda de Mar)? (ha quedat
  buida)?
• The Pineda de Mar house is empty
•   b.    (El president de la Generalitat de
  Catalunya)? (pateix)?
• The president of the Generalitat is suffering
•   c.   (La nena boliviana de la Maria)? (no vol
  venir)?
• Marys daughter does not want to come
•  

36

• Long subjects are grouped into very long phrases
  (reminiscent of Align-Focus, L).
• WRAP-MIN-SUBJECT
• Wrap the subject if it has more than one
  prosodic word

37
Effects of clash

• Clash avoidance plays a role on phrasing
  decisions (optional effects)
• NP and VP Projections
• No clash Clash
• a. (Un sac)? (de patates dOlot)? a.(Un sac
  dherbes)? (aromàtiques)?
• A bag of Olot potatoes A bag of aromatic
  herbs
• b. (Un anell)? (de plata suïssa)? b. (Un anell
  dor)? (suís)?
• A ring made of Swiss silver A Swiss golden
  ring
• No clash Clash
• a. (Menjaré)?(pastissos de xocolata)? a.
  (Menjaré flams)? (de xocolata)?
• I will eat chocolate cakes I will eat
  chocolate custard
• b. (Sortiran)? (els volums importants)? b.
  (Sortiran cartes)? (importants)?
• The important volumes will come out The
  important letters will come out
•  

38

• IP/CP Projections
• No-clash Clash
• (La Maria)? (bevia)? (aigua destillada)? (La
  Maria)? (beu aigua)? (destillada)?
• Mary used to drink distilled water Mary used
  drinks distilled water
• (En Joan)? (llegeix )? (novelles de por) (En
  Joan)? (llegeix llibres)? (de por)?
• Joan reads horror novels Joan reads horror
  books
•  
• (La Maria)? (venia)? (peres de Lleida)? (La
  Maria)? (ven peres)? (de Lleida)?
• Mary used to sell pears from Lleida Mary
  sells pears from Lleida
•  

39
OT analysis of clash effects

• Default case with no clash
• Clash resolution optionality is obtained through
  free ranking of No-clash ltgtWrap)
•  

•  

40

• Default case
• Optionality in clash resolution through free
  ranking of No-clash ltgtWrap)
•  

•  

41
Conclusions

• Alignment and wrapping conditions are not
  sufficient to predict ?-groupings in Catalan.
• Syntactic branching conditions do not play a
  crucial role in ?-grouping
• In Catalan size and eurhythmic constraints play a
  crucial role in phrasing. Active constraints
  Max-Bin, Min-Phrase, No-Clash.

42
Case study II. Goal of the study

• Investigate the influence of syntactic/prosodic
  branching and constituent length (in ?s) on
  phrasing in 4 Romance languages Catalan,
  Italian, European Portuguese (2 varieties), and
  Spanish.

43
Language-specific tendencies
Catalan (SV)O
Italian (SVO)
44
Corpus (Elordieta, Frota, Prieto Vigário 2003)

• Comparable database for different languages, i.e.
  SVO sentences with 2 varying factors
• Constituent length
• short 3 ?
• long 5 ?

• Syntactic branchingness
• non-branching
• branching (object, subject or both)
• double branching (object, subject or both)

45
Examples from Italian corpus
46
Recording and labeling

• 2 speakers, 3 repetitions, broad focus (Whats
  happening?).
• 456 utterances per language/variety (76
  utterances x 2 speakers x 3 repetitions)
• Digitized and prosodically annotated using
  language-specific version of ToBI.

47
Italian

• In most cases, the end of a phrase was marked
  with a clear F0 continuation rise or sustained
  pitch, a prominent phrase-final H accent and an
  optional pause.

48

• Though (SVO) is still the preferred grouping,
  when a non-branching subject is followed by a
  long object, the number of (SVO) renderings
  decreases only for one speaker (while other
  phrasing strategies increase).

Table 2 non-branching subject / branching object
49
Catalan

• In most cases, the end of a phrase was marked
  with a clear F0 continuation rise and a prominent
  phrase-final accent

Waveforms and Fo contours of the utterances (La
nena mirava) (la melmelada meravellosa) and (La
boliviana de Badalona) (rememorava la noia).
50

• Table 2 non-branching subjects followed by
  branching objects
• (SV)(O) renderings increase before long branching
  objects (around 25 for both speakers) length
  vs. syntactic complexity.
• Tendency to have more (SV)(O) phrasings when S is
  short rather than long (especially for NM)

51
European Portuguese

• (a NOra maraviLHAva veLHInhas LINdas)
• the daughter-in-law marveled beautiful old
  women
• (a LOUra)(memoriZAva uma melodia) the blond
  girl learned a song by heart

SEP
NEP


52
EP results

• Lenght or syntactic complexity of O is not
  important for SEP but it is for NEP.

Table 2 non-branching subject / branching object
53
Spanish

• In most cases, the end of a phrase was marked
  with a continuation rise, final lengthening,
  pitch reset and an optional pause.

(La niña) (miraba la noria) The girl watched the
turning-wheel
(La niña de Lugo) (miraba la mermelada) The girl
from Lugo watched the marmalade
54

• Branchingness or syntactic complexity plays a
  role in intonational phrasing, as the difference
  between long nonbranching objects and short
  branching objects indicates (taking the 2
  speakers together, Long Nonbranching Os 75.5,
  Short Branching Os 96).

Table 2 non-branching subject / branching object
55
Major Findings

• CAT SEP NEP SP
  IT
• Non Br (S)(VO) (SVO)
  (S)(VO)/ (S)(VO)/ (SVO)
• 
  (SVO)
  (SVO)
• Sh. Br. S (S)(VO) (SVO)
  (S)(VO)/ (S)(VO) (S)(VO)
• 
  (SVO)
• L. Br. S (S)(VO) (SVO)/
  (S)(VO)/ (S)(VO) (S)(VO)
• (S)(VO)
  (SVO)
• Sh. Br. O (S)(VO) (SVO)
  (S)(VO)/ (S)(VO) (SVO)
• (SVO)
• L. Br. O (SV)(O) (SVO)
  (S)(VO)/ (S)(VO) (SVO)/
  (SVO) (SV)(O)

56
Conclusions

• Clear crosslinguistic differences in phrasing in
  our five languages.
• Weight differences in number of syllables make a
  difference in phrasing decisions (up until now,
  only number of prosodic words have been taken
  into account with constraints such as Max-Bin,
  Min-Bin, etc.
• Prosodic branchingness is what matters, at least
  for the speakers analyzed up until now. This
  suggests that the traditional 'syntactic
  branching' condition (vg. Nespor Vogel) should
  probably be replaced by a prosodic branching
  condition.

57

• Thank you!