Are compounds words or phrases? A phonetic investigation of

Are compounds words or phrases?
An acoustic investigation of Polish
Adjective + Noun Compounds
LINGUIST 1
Olek Główka
October 15, 2014
What makes a word?
grapefruit
travel agency
good-for-nothing
son-in-law
money-hungry
stand by
Orthographic criterion:
a word is what occurs
between spaces in writing
What makes a word?
grapefruit
travel agency
good-for-nothing
son-in-law
money-hungry
stand by
Semantic criterion:
a word has semantic coherence
(i.e. expresses a unified concept)
What makes a word?
grapefruit
travel agency
good-for-nothing
son-in-law
money-hungry
stand by
Phonological criterion:
each word has an
independent primary stress
What makes a word?
grapefruit
travel agency
good-for-nothing
son-in-law
money-hungry
stand by
Morphological criterion:
external (but not internal)
modification by a prefix/suffix
signals lexical category
membership
travel
agencies
son-in-law’s
grapefruits
good-for-nothings
stands
bybirthday
*travels
agencies
two
*grapesfruits
sons-in-law
*worse-for-nothings
*stand
bys
What makes a word?
grapefruit
travel agency
good-for-nothing
son-in-law
money-hungry
stand by
Syntactic criterion:
a word must behave as a unit
*grapevery
andhungry
fruit
*money
What makes a word?
criterion
orthographic
semantic
phonological
morphological
syntactic
grapefruit
travel agency
good-for-nothing
son-in-law
money-hungry
stand by
Polish Adjective+Noun Compounds
• Why are compounds important to our understanding of language?
• Linguistic theory has traditionally distinguished between two major
components of language:
lexicon
syntax
the grammar of words
the grammar of sentences
• Each of these two components of language is argued to have a
distinct phonological interpretation.
• Compounds challenge the clear-cut distinction between the lexicon
and the syntax.
Polish Adjective+Noun Compounds
• morphological compounds:
(1)
(2)
Adj + -o- + N
linking morpheme
(derivational)
all Polish words
are stressed on
the penultimate
syllable
all Polish phrases
are stressed on the
rightmost element
Polish Adjective+Noun Compounds
• phonological compounds
(3)
L-AdjN compounds:
Prosodic variation
(4)
• Orthographically
non-cohesive
• Semantically
cohesive
• Morphologically
non-cohesive
• Syntactically
cohesive
• How about the
phonology?
The Puzzle
• Phonological compounds ≠ phonological words
lefthand variant is not stressed on the penultimate syllable
• Phonological compounds ≠ phonological phrases
lefthand variant is not stressed on the rightmost element
...so what is the phonological identity of the lefthand variant?
The Prosodic Hierarchy
Intonational Phrase (ι)
Phonological Phrase (ϕ)
Phonological Word (ω)
Foot (f)
Syllable (σ)
Polish phonological
compounds?
Recursion or Independent Unit?
Hypothesis: Polish compounds form an independent phonological unit.
Challenge: show that they are not recursive phonological words or phonological phrases.
Phonetic Correlates of Recursive Embedding
• Several studies have reported gradient differences in prosodic boundary strength
for constituents hierarchically embedded within constituents of the same type for
large prosodic domains.
• Ladd (1988) and Wagner (2005) found increased pitch peaks and increased
duration in clauses B and C after a but-boundary than after an and-boundary:
(5)
a. [[A and B] but C]
b. [A but [B and C]]
• No phonetic study has yet investigated the phonetic correlates of phonological
word recursion, word recursion is widely assumed in phonological theory (Ito &
Mester 2007, Selkirk 2011).
Acoustic experiment
• An acoustic experiment looked at correlates of prosodic prominence in lexically stressed
vowels of lefthand AdjNs, righthand AdjNs, typical phrases.
•
-
Adjective-Noun Compound
Noun-Adjective Phrase
mit grecki
‘Greek myth’
mroczny mit
‘dark myth’
kot perski
‘Persian cat’
krępy kot
‘stocky cat’
The study examined three correlates of prosodic prominence:
fundamental frequency (F0) (a.k.a “pitch”): the rate of vibration of vocal folds.
vowel duration
Intensity (loudness)
Experimental Setup
• 4 speakers of Standard Polish with no reported history of speech or hearing problems.
• Stimuli: 12 Adjective-Noun compounds embedded in carrier phrases
• Subjects were presented with a visual stimulus followed by an auditory question. They responded
to the question after a given signal.
• 4 dependent variables: mean F0, F0 change, duration, and intensity
• The presence of variation was verified in a pre-experiment, but was not controlled for in the main
experiment. Overall, the subjects produced 135 L-AdjNs and 142 R-AdjNs.
Example Spectrogram
Results: Mean F0
Figure 1: Mean 𝐹0 values for V1 & V2 in L-AdjNs, R-AdjNs, and Phrases
• A relatively large pitch excursion signals a pitch
accent, the phonetic correlate of primary stress.
• Single pitch accent in L-AdjNs, two pitch accents
in R-AdjNs and Phrases. This indicates the
phonological uniqueness of L-AdjNs and
similarity between R-AdjNs and phonological
phrases.
• Non-gradient differences between the lefthand
and righthand constituent of L-AdjN speak
against recursive embedding.
Results: F0 Change
Figure 2: Example pitch contour showing the minima Figure 3: 𝐹0 change values for V1 & V2 in L-AdjNs, RAdjNs, and Phrases
and the maxima in V1 & V2 in a L-AdjN
• L-AdjNs exhibit a large 𝐹0
change in V1, R-AdjNs exhibit
large 𝐹0 changes in V1 & V2.
• Phrases show only one large
𝐹0 change in V2.
• All three conditions are
prosodically distinctive.
• Speakers normalize for pitch
declination, so V2 is
perceptually lower.
Results: Mean Intensity
Figure 4: Mean intensity values for V1 & V2 in LAdjNs, R-AdjNs, and Phrases
• Mean intensity co-varies with F0 in L-AdjNs and
Phrases, but not in R-AdjNs.
• All three-conditions are prosodically distinctive.
Results: Vowel Duration
Figure 5: Duration values for V1 & V2 in L-AdjNs,
R-AdjNs, and Phrases
• L-AdjNs & R-AdjNs did not exhibit distinctive
duration patterns, but differed significantly
from Phrases.
• Duration provides no support for the distinctive
constituency frames for L-Adjs & R-AdjNs.
• Possible confound: phrase-final lengthening
might have lengthened V2 and distorted the
underlying relationship between V1 and V2.
Results: Summary
Conclusion
• Lefthand prominent Adjective-Noun compounds are prosodically distinct from recursive phonological
words and phonological phrases and can be more adequately interpreted as an independent
phonological constituent in the prosodic hierarchy.
• In terms of their phonological properties, lefthand prominent Adjective-Noun compounds cross-cut
the distinction between words and sentences. Although morphologically they represent two
independent words, phonologically they form a more cohesive unit.
• This indicates that phonology is not a strictly interpretive component of the lexical or the syntactic
component of language but has an independent constituent structure that may or may not parallel
other components of the grammar.
Selected References
Itô, Junko & Armin Mester (2007). Prosodic Adjunction in Japanese Compounds. In
Formal Approaches to Japanese Linguistics 4. MIT Working Papers in
Linguistics 55: Cambridge, MA. 97–111.
Ladd, D. Robert (1988). Declination ‘reset’ and the hierarchical organization of
utterances. JASA 84. 530-544.
Selkirk, Elizabeth O. (2011). The Syntax-Phonology Interface. In John A. Goldsmith,
Jason Riggle, Alan Yu, Handbook of Phonological Theory. Oxford: WileyBlackwell. 435–484.
Wagner, Michael (2005). Prosody and Recursion. PhD Dissertation, MIT.