Planning units in Tagalog sentence production: Evidence...
1
Introduction To transform thoughts into speech, speakers must encode preverbal messages linguistically. What is the timecourse of message and sen- tence formulation for sentences like “The child kicks the ball”? • In English, this can happen one word at a time. Speakers can begin uttering sentences after having encoded only the sentence initial NP ([1]). • How does formulation proceed in languages where the sentence initial element overtly marks a dependency with other elements of the sentence, such as Tagalog? Tagalog Tagalog is spoken by approx. 23 million speakers in the Philippines. It belongs to the Western Malayo- Polynesian branch of the Austronesian language family. Task • 53 native Tagalog speakers described drawings of transitive events ([2]). • Their gaze and speech were recorded (Tobii T120 eye tracker, 120 Hz sampling rate). More information: [email protected] Planning units in Tagalog sentence production: Evidence from eye tracking Sebastian Sauppe, Elisabeth Norcliffe, Agnieszka E. Konopka, Robert D. Van Valin, Jr and Stephen C. Levinson References [1] Gleitman et al. (2007). JML, 57, 544–596. [2] Griffin and Bock (2000). Psych. Sci. , 11, 274–279. [3] Kroeger (1993). Phrase structure and grammatical relations in Tagalog. verb AM patient OBJ actor SUBJ s <um> isipa ng= bola ang= bata <AM> kick OBJ= ball SUBJ= child “ The child kicks the ball .” 0.0 0.2 0.4 0.6 0 500 1000 1500 2000 2500 3000 time (ms) proportion of fixations N = 384 sentences mean speech onset (1584 ms) “early” planning Sentence structure in Tagalog • Tagalog has a verb-initial word order. • Transitive events can be described using different syntactic structures (see a–c below). • The verb can carry actor marking (i.e., the verb agrees in semantic role with the actor) or patient marking (i.e., the verb agrees in semantic role with the patient). • There are no syntactic constraints on post-verbal NP ordering ([3]). Questions • Verb-initial word order implies that speakers must encode information about the relation between event participants very early in the sentence formulation process to select a suitable verb. • Verb marking also implies that an event participant has to be chosen to be the subject very early in order to select appropriate verbal affixes. To what extent do speakers encode the subject at the outset of sentence formulation? • Is this influenced by the type of verb marking, i.e. actor marking vs. patient marking (Analyses 1 and 2)? • Is this influenced by post-verbal NP order (Analyses 2 and 3)? 0.0 0.2 0.4 0.6 0 200 400 600 time (ms) proportion of fixations (a) actor in verb AM patient OBJ actor SUBJ Analysis 1 Sentences differing in verb marking and word order: (a) actor marking with verb AM patient OBJ actor SUBJ word order vs. (c) patient marking with verb PM actor OBJ patient SUBJ word order (c) actor in verb PM actor OBJ patient SUBJ “Early” window (0–600 ms after picture onset): more fixations to actor if it is marked on the verb 0.0 0.2 0.4 0.6 0 500 1000 1500 2000 2500 3000 time (ms) proportion of fixations verb PM actor OBJ patient SUBJ s <in> isipa ng= bata ang= bola <PM> kick OBJ= child SUBJ= ball “ The child kicks the ball .” mean speech onset (1716 ms) N = 787 sentences “early” planning 0.0 0.2 0.4 0.6 0 200 400 600 time (ms) proportion of fixations Analysis 3 Sentences differing in word order: (a) actor marking with verb AM patient OBJ actor SUBJ word order vs. (b) actor marking with verb AM actor SUBJ patient OBJ word order “Early” window (0–600 ms after picture onset): no difference in fixations to actor if verb marking is identical (a) actor in verb AM patient OBJ actor SUBJ (b) actor in verb AM actor SUBJ patient OBJ Analysis 2 Sentences differing in verb marking: (b) actor marking with verb AM actor SUBJ patient OBJ word order vs. (c) patient marking with verb PM actor OBJ patient SUBJ word order 0.0 0.2 0.4 0.6 0 200 400 600 time (ms) proportion of fixations “Early” window (0–600 ms after picture onset): more fixations to actor if it is marked on the verb (b) actor in verb AM actor SUBJ patient OBJ (c) actor in verb PM actor OBJ patient SUBJ Conclusions Two observable phases in Tagalog sentence production: • early phase involving planning of the dependency relation between the verb and the subject (i.e., the verb marking) • later phase involving incremental encoding of characters in the order of mention, as in English Fixation patterns provide evidence for: • temporal dissociation of mapping of message-level concepts to syntactic functions and lexical encoding of these concepts • linguistic guidance in the earliest stages of sentence production in Tagalog verb AM actor SUBJ patient OBJ s <um> isipa ang= bata ng= bola <AM> kick SUBJ= child OBJ= ball “ The child kicks the ball .” 0.0 0.2 0.4 0.6 0 500 1000 1500 2000 2500 3000 time (ms) proportion of fixations N = 67 sentences mean speech onset (1668 ms) “early” planning (a) ACTOR MARKING (b) ACTOR MARKING (c) PATIENT MARKING * * n.s.
Transcript of Planning units in Tagalog sentence production: Evidence...
Introduction
To transform thoughts into speech, speakers must encode preverbal messages linguistically. What is the timecourse of message and sen-tence formulation for sentences like “The child kicks the ball”?
• In English, this can happen one word at a time. Speakers can begin uttering sentences after having encoded only the sentence initial NP ([1]).• How does formulation proceed in languages where the sentence initial element overtly marks a dependency with other elements of the sentence, such as Tagalog?
Tagalog
Tagalog is spoken by approx. 23 million speakers in the Philippines. It belongs to the Western Malayo- Polynesian branch of the Austronesian language family.
Task
• 53 native Tagalog speakers described drawings of transitive events ([2]).• Their gaze and speech were recorded (Tobii T120 eye tracker, 120 Hz sampling rate).
More information: [email protected]
Planning units in Tagalog sentence production:Evidence from eye tracking
Sebastian Sauppe, Elisabeth Norcliffe, Agnieszka E. Konopka, Robert D. Van Valin, Jr and Stephen C. Levinson
References[1] Gleitman et al. (2007). JML, 57, 544–596.[2] Griffin and Bock (2000). Psych. Sci., 11, 274–279.[3] Kroeger (1993). Phrase structure and grammatical relations in Tagalog.
verbAM patientOBJ actorSUBJ s<um>isipa ng=bola ang=bata <AM>kick OBJ=ball SUBJ=child “The child kicks the ball.”
0.0
0.2
0.4
0.6
0 500 1000 1500 2000 2500 3000time (ms)
prop
ortio
n of
fixa
tions
N = 384 sentences
mean speech onset (1584 ms)
“early” planning
Sentence structure in Tagalog• Tagalog has a verb-initial word order. • Transitive events can be described using different syntactic structures (see a–c below).• The verb can carry actor marking (i.e., the verb agrees in semantic role with the actor) or patient marking (i.e., the verb agrees in semantic role with the patient).• There are no syntactic constraints on post-verbal NP ordering ([3]).
Questions• Verb-initial word order implies that speakers must encode information about the relation between event participants very early in the sentence formulation process to select a suitable verb.• Verb marking also implies that an event participant has to be chosen to be the subject very early in order to select appropriate verbal affixes.
To what extent do speakers encode the subject at the outset of sentence formulation?• Is this influenced by the type of verb marking, i.e. actor marking vs. patient marking (Analyses 1 and 2)?• Is this influenced by post-verbal NP order (Analyses 2 and 3)?
0.0
0.2
0.4
0.6
0 200 400 600time (ms)
prop
ortio
n of
fixa
tions
(a) actor inverbAM patientOBJ actorSUBJ
Analysis 1Sentences differing in verb marking and word order:(a) actor marking withverbAM patientOBJ actorSUBJ word ordervs.(c) patient marking with verbPM actorOBJ patientSUBJ word order
(c) actor inverbPM actorOBJ patientSUBJ
“Early” window (0–600 ms after picture onset): more fixations to actor if it is marked on the verb
0.0
0.2
0.4
0.6
0 500 1000 1500 2000 2500 3000time (ms)
prop
ortio
n of
fixa
tions
verbPM actorOBJ patientSUBJ s<in>isipa ng=bata ang=bola <PM>kick OBJ=child SUBJ=ball “The child kicks the ball.”
mean speech onset (1716 ms)
N = 787 sentences
“early” planning
0.0
0.2
0.4
0.6
0 200 400 600time (ms)
prop
ortio
n of
fixa
tions
Analysis 3Sentences differing in word order:(a) actor marking withverbAM patientOBJ actorSUBJ word ordervs.(b) actor marking with verbAM actorSUBJ patientOBJ word order
“Early” window (0–600 ms after picture onset): no difference in fixations to actor if verb marking is identical
(a) actor inverbAM patientOBJ actorSUBJ
(b) actor inverbAM actorSUBJ patientOBJ
Analysis 2Sentences differing in verb marking:(b) actor marking withverbAM actorSUBJ patientOBJ word ordervs.(c) patient marking with verbPM actorOBJ patientSUBJ word order
0.0
0.2
0.4
0.6
0 200 400 600time (ms)
prop
ortio
n of
fixa
tions
“Early” window (0–600 ms after picture onset): more fixations to actor if it is marked on the verb
(b) actor inverbAM actorSUBJ patientOBJ
(c) actor inverbPM actorOBJ patientSUBJ
ConclusionsTwo observable phases in Tagalog sentence production:• early phase involving planning of the dependency relation between the verb and the subject (i.e., the verb marking)• later phase involving incremental encoding of characters in the order of mention, as in English
Fixation patterns provide evidence for:• temporal dissociation of mapping of message-level concepts to syntactic functions and lexical encoding of these concepts• linguistic guidance in the earliest stages of sentence production in Tagalog
verbAM actorSUBJ patientOBJ s<um>isipa ang=bata ng=bola <AM>kick SUBJ=child OBJ=ball “The child kicks the ball.”
0.0
0.2
0.4
0.6
0 500 1000 1500 2000 2500 3000time (ms)
prop
ortio
n of
fixa
tions
N = 67 sentences
mean speech onset (1668 ms)
“early” planning
(a)ACTORMARKING
(b)ACTORMARKING
(c)PATIENTMARKING
* *n.s.
