PaPI 2005 (Barcelona, 20-21 June) The perception of stress patterns by Spanish and Catalan infants Ferran Pons (University of British Columbia) Laura Bosch.

PaPI 2005 (Barcelona, 20-21 June)

The perception of stress patterns by Spanish and Catalan infants

Ferran Pons (University of British Columbia)Laura Bosch(Universitat de Barcelona)


Rhythmic properties of the input language

Infants’ early sensitivity to prosodic information (rhythm and intonation) in the prelexical stage: language discrimination studies

Stress as a cue to bootstrapping syntax

Stress as a cue to word segmentation: orientation towards the predominant stress pattern of the target language


Previous data (I)

English-learning infants

Between 6 and 9 months of age growing knowledge about the distribution of stress within words in the language

Jusczyk, Cutler & Redanz (1993)9 month-olds (but not 6 month-olds) prefer to listen to disyllabic trochees (real words, also low-pass filtered).


Previous data (II)

Echols et al. (1997) SW grouping when pauses precede a

strong syllable (WSW stimuli)

Höhle (2002) SW preference in German-learning

infants (6 month-olds)


Word segmentation in English

Around 90% content words (disyllabic) are trochaic; monosyllabic words are stressed.

Metrical segmentation strategy

Stress cues can reliably be used for segmentation purposes, as they help infants find word boundaries

• Jusczyk et al. (1999): earlier segmentation for trochees than for iambs (10 ½ mo.)

• Houston et al. (2000): cross-language SW segmentation by 9 month-olds (English/Dutch)


Word segmentation in English

Multiple cues to segmentation, although prosodic and statistical ones would be used before phonotactic and allophonic cues

Mattys et al., 1999 Statistical regularities earlier than

prosodic cues? Thiessen & Saffran (2003)


Word segmentation in languages other than English

French: late word segmentation? Gout, 2001 9 month-olds do not segment

iambic words (predominant pattern).

Dutch: SW words are not segmented by 7 ½ months of age, but by 9 months. Kuijpers et al., 1998 differences in the

relative prominence of the SW foot between English and Dutch would explain the discrepancy

These studies shed light on the generability of the metrical segmentation strategy


Stress in Spanish

Most frequent word shape is disyllabic, but only around 65% of these words begin with a strong syllable.

Lexical stress is not fixed, but stress on the penultimate syllable is the predominant pattern.

Syllable weight is a relevant factor


Stress in Catalan

Catalan, like Spanish, is also considered a trochaic language.

Lexical stress is variable. Around 66% of disyllabic words are

trochaic Moraic throchee: minimal word

pattern (Cabré, 1993)


Infant data assessing infants’ capacity to

discriminate SW vs. WS patterns

exploring preferences in Spanish-learning and Catalan-learning infants


Experiment 1: Discrimination

Material: CVCV stimuli Vowels: a, i, u Consonants: p, t, k, b, d, m, n, l Examples: buki, nila, luta, taki, datu… Lists: 8 different tokens, presented

twice (trials’ length: 24”)



Procedure: Familiarization-preference procedure

(adaptation from Jusczyk & Aslin 1995) A 2-minute’s familiarization phase to

trochaic or iambic lists followed by 4 test trials (2 same & 2 switch).

Participants: Sixteen 8 ½-month-old infants from a

monolingual environment (Spanish or Catalan).



Results

0

2

4

6

8

10

12

14

16

Me

an

loo

kin

g t

ime

s (

s)

Same Switch

*

* p = .022


Experiment 1: Discrimination Discussion:

8 ½ -month-old infants from Spanish or Catalan monolingual families are able to discriminate between trochaic and iambic disyllabic word shapes.

This discrimination capacity, which is possibly present even at an earlier age, does not seem to have been modified by language exposure.

Sensitivity to stress patterns is required if stress cues are to be used for word segmentation purposes.


Experiment 2: Preference Stress pattern preferences arise during the

second semester of life from exposure to the ambient language, possibly after some words have already been segmented by means of statistical learning mechanisms applied to the continuous speech signal

Question:

Will a trochaic bias be also present in the two Romance languages under study?


Experiment 2: Preference

Hypothesis: As in the case of English-learning infants,

9 month-old infants, but not 6 month-olds, will show a preference for the dominant word stress pattern in their language (trochaic).


Experiment 2a: Preference

Stimuli: 42 different disyllabic CVCV non-words,

either trochaic or iambic (lists of 18 items (6x3) per trial).

Participants: 16 6-month-old Spanish monolingual

infants. 16 6-month-old Catalan monolingual

infants.



Procedure: A slightly modified version of the Head-turn

Preference Procedure (HPP) was used. 4 familiarization trials 12 test trials (6 trochaic & 6 iambic)

Measures: Mean attention time to trochaic vs. iambic test trials



Setting



Results (6-month-olds)

0

2

4

6

8

10

12

Mea

n lo

oki

ng

tim

es (

s)

MC ME

Group

Trochaic

Iambic


Experiment 2b: Preference

Participants: 16 9-month-old Spanish monolingual infants. 16 9-month-old Catalan monolingual infants.

Stimuli and procedure: The same as in experiment 2a. Setting: infants sat on their parents’ lap.


Experiment 2b: Preference

0

2

4

6

8

10

12

Mea

n lo

oki

ng

tim

es (

s)

MC ME

Group

Trochaic

Iambic



Studies of preference (2 a & b)

Discussion Neither 6 month-old nor 9 month-old

Catalan and Spanish infants showed a preference for the trochaic pattern.

Acoustic analysis of the test stimuli revealed clear markers of stress that signaled trochaic versus iambic items (intensity, duration and pitch)


Acoustic correlates of stress in test stimuli

Pitch

0

50100

150

200

250300

350

w eak strong

Hz

Duration

0

0.1

0.2

0.3

w eak strong

seco

nd

s

Intensity

72

74

76

78

80

82

w eak strong

dB


Studies of preference (2 a & b)

These results suggest that infants exposed to trochaic languages, such as Catalan or Spanish, which have nevertheless a less predominant stress pattern than English, may not develop an early sensitivity for this metrical property.

However, stress information may be linked to other properties (syllable structure, phonotactics…) preference might then be observed and this information could be used for word segmentation purposes.


Lexical stress and syllable weight

Quantity sensitive languages use syllable weight when assigning stress: Syllable weight hypothesis preference

for SW patterns would not emerge for words with a lax vowel in the stressed syllable

Stress pattern hypothesis preference for SW patterns before sensitivity to the principle of syllable weight


Syllable weight studies

Turk, Jusczyk & Gerken (1995) SW preference in non-words, where S

syllables were heavy or had a tense vowel.

SW preference, even when S syllables were not heavy and had a lax vowel (stress pattern hypothesis)

No preference for CV cvc / cv CVC (competing preferences cancelled each other)


Syllable weight studies: conclusion

Turk, Jusczyk & Gerken (1995) In English, syllable weight is not a

necessary component of the SW preference observed in infant studies

Sensitivity to surface patterns and underlying linguistic principles seems to follow a separate time course in development


Syllable weight in Spanish(based on LEXESP measures)

Spanish: CV·CVC words are mainly iambic (72%-84%,

depending on the segments in the endings) -or 98% -on 84% -al 97% -el 97% -in, un < 60%

CVC·CV words are mainly trochaic (95%)


Experiment 3: Syllable weight

Procedure: A slightly modified version of HPP- Head-turn

preference procedure was used.

Stimuli: 86 bisyllabic non-words CV·CVC, with either

trochaic or iambic stress (lists of 12 items per trial).

4 familiarization trials 12 test trials (6 trochaic & 6 iambic)



Participants: 16 9-month-old Spanish monolingual infants.

Hypothesis: If syllable weight is a relevant factor in lexical

stress assignement, then 9-month-old infants will show a preference for the cv·CVC lists (iambic pattern) vs. CV·cvc lists (trochaic pattern, but highly infrequent in Spanish).



0

2

4

6

8

10

12

Mea

n lo

okin

g tim

es (s

)

ME

Trochaic

Iambic


*

* p = .048



Discussion By 9 months of age, infants have acquired

knowledge about the most frequent stress pattern in cv·cvc words.

They prefer cv·CVC iambic words, even though the language has a predominant trochaic pattern.

At least in Spanish, syllable weight seems to be an important component linked to lexical knowledge and possibly relevant for stress segmentation.


Final comments The present results differ from previous

work in English and support the syllable weight hypothesis.

Additional research exploring the preference for CVC·cv trochaic words in Spanish strong support for the present hypothesis.

Segmentation studies with 9 month-old infants would reveal whether this information is used in word segmentation tasks.

PaPI 2005 (Barcelona, 20-21 June) The perception of stress patterns by Spanish and Catalan infants Ferran Pons (University of British Columbia) Laura Bosch.

Documents

barcelona slide

sw words

late word segmentation

earlier segmentation

lexical stress

crosslanguage sw segmentation

predominant stress pattern

segmentation purposes