
78f3de0f4a129379d3cbf3f41c81a0c8.ppt
- Количество слайдов: 29
Prosodic facilitation and interference in the resolution of temporary syntactic closure ambiguity Kjelgaard & Speer 1999 Kent Lee Ψ 526 b 16 March 2006
Prosodic structure Potential cues for parsing: 1) Fundamental frequency (F 0) – glottal 2) Amplitude & spectral info
Prosodic structure Potential cues for parsing: 3) Duration 4) Lexical, sentential, focal stress (from duration & F 0) 5) Segmentation; e. g. : “let me kiss the sky” “let me kiss this guy”
Prosodic structure n Phonological phrase (PP) often maps onto XPs (NPs, VPs. . . ) ¨ demarcated by phrase accent [L-, H-] ¨ n Intonational phrase (IP) often maps onto clause (S) ¨ demarcated by boundary tones [L%, H%] ¨ n n Pitch (stress) accents [H*, LH*, L*H, H-!H] Prosodic structure is non-recursive, thus less complex than syntax (which allows nested structures / recursion: S´ ⊂ VP ⊂ NP = N+RC = S ⊂ VP ⊂ NP)
Prosody-syntax mapping: Consistency & GP
Prosody-syntax mapping: Consistency & GP 1) Early Closure a. Consistent (“cooperating”) prosody When Tim is presenting, the lectures are interesting. b. Baseline When Tim is presenting the lectures are interesting. c. Conflicting When Tim is presenting the lectures, are interesting. 2) Late Closure a. Consistent (“cooperating”) When Tim is presenting the lectures, they’re interesting. b. Baseline When Tim is presenting the lectures they’re interesting. c. Conflicting When Tim is presenting, the lectures they’re interesting.
Prosody-syntax mapping: Consistency & GP n n Consistent prosodic-syntactic mapping should facilitate parsing Inconsistent mapping should create interference effects, and GPing with ambiguous sentences Previous studies found some such effects, but didn’t examine early vs. late closure Interference & facilitation compared with baseline condition (single IP for both clauses)
Materials n n 18 sets of 6 sentence variants 2 X 3 design: early/late closure X consistent (“cooperating”), baseline, conflicting prosody norming study used to generate sentences with transitivity index; phonetic analysis & manipulation pretests for acceptability / intelligibility
Experiment 1 n n n 66 subjects speeded grammaticality judgment DVs: judgments & error rates weak transitivity effect (e. g. , r=-. 5, p<. 05) – should have been used as control variable? Insertion of extra function words (late closure) did not lead to significant length effects
Expt. 1 1) Sig. main effects of prosody & syntax in RTs & errors a) Errors | Prosody: ¨ Conflicting > Baseline ¨ GP effects only in Baseline ¨ (Early > Late Closure) b) RTs: facilitation but no interference effects -¨ Coop < Baseline, Conflicting
Expt. 1 n n Prosody affected metalinguistic judgments Consistent faster than Baseline (w/o IP boundaries), no GP, no early/late closure differences Early closure in Baseline & Conflicting slower, induced GP But: no clear interference effect on RTs in Conflicting cond. (? ), hence, Expt. 2
Expt. 2 Same design & materials n Speeded “end of sentence” comprehension task (comprehended = yes/no) n Avoids metalinguistic judgments n
Expt. 2 n n n RTs showed prosodic effects Facilitation: Cooperating < Baseline Interference: Conflicting > Baseline GP processing difficulty for early closure in Baseline & Conflicting No GP in Cooperating
Expt. 2 n “no comprehension” reported in 4. 6% trials
Expt. 2 n n Consistent IP boundary facilitated parsing, cf. ambiguous Baseline Significant interference effect in Conflicting cf. Baseline Consistent prosody early facilitation for early closure; no difference in early cf. late closure in Cooperating cond. , cf. Baseline & Conflicting Jives with Expt. 1 & overall predictions
Expt. 2 n n n Rules out later or minor role of prosody Rules out prosody interpreted based on underlying syntax Can assist with ambiguous or dispreferred parses Can lead to interference with parsing if incongruous But: drawbacks to speeded comprehension task
Expt. 3 Cross-modal priming n to measure point of disambiguation n avoids “contamination” or interference of metalinguistic processing n same materials & basic design n
Expt. 3 n n n Ss hear syntactically ambiguous fragment over headphones Previous sound files were truncated to ambiguous region (e. g. , When Roger left the house. . . ) Ss see & name disambiguating word that is visually presented (is/its) Ss complete the sentence (unnatural task? ) In Conflicting cond. , is/its opposite closure type In Baseline cond. , IP boundary tone was truncated
Expt. 3 n n Most errors in Baseline conditions & Conflicting early closure Cooperating EC < Baseline EC Cooperating LC = Baseline LC Interference: Conflicting < Baseline conditions
Expt. 3 n n n GP difficulty for EC cf. LC in Baseline & Conflicting conditions, not for Cooperating cond. No correlations b/ naming times & # syllables / words in control study Show similar interference & facilitation effects as in Expts. 1 -2, at point of disambiguation
Expts. 1 -3 Expts. show prosody can mislead the parser n Results are consistent with early processing n Location of IP boundaries guide assignment of syntactic boundary locations, until reanlysis required n
Expt. 4 n n Aims to examine PP boundaries If PP boundaries are used by syntactic parser, then PP facilitation / interference effects should be possible Similar design, materials, norming, analysis, & pretesting of stimuli as before Cross-modal naming task used for temporal resolution in detection of PP effects in parsing & disambiguation
Expt. 4 n n Stimuli with subtle tone changes used to rule out possibility that processing advantage is attributable to acoustically exaggerated boundaries Stimuli without final pauses used to rule out possibility that processing effects derive from extra duration; silences were deleted. . . ]PP ]IP
Expt. 4 n n n 18 sentence sets, digitally truncated to ambiguous region 48 subjects 2 X 3 design: early/late closure X 3 prosodies Cooperating: PP boundary consistent with clause boundary Baseline: phonetically ambiguous phrase accent (L-) & brief break Conflicting: misleading visual targets to PP boundary tones
Expt. 4 Most errors in Baseline & Conflicting EC n Main effects for prosody, syntax, & interaction effect n
Expt. 4 n n Facilitation in Cooperating cf. Baseline cond. , but only in EC as before Interference in Conflicting cf. Baseline GP processing difficulty in Baseline (EC, LC) & Conflicting (EC, LC) No GP in Cooperating cond.
Expt. 4 n n Results consistent with those for IPs in Expts. 13 Consistent PP boundaries facilitated parsing of dispreferred syntactic analysis Inconsistent PP boundaries led to interference for EC & LC (preferred / dispreferred syntax) Even subtle boundaries can influce parsing
Discussion No GP with consistent prosodic-syntactic boundaries n Disadvantage for less preferred EC in ambiguous & inconsistent conditions n Facilitation & interference early in syntactic processing n Supports interactive processing models n
Discussion Potential problems: n Expts. 1 & 2 lack temporal resolution n Expts. 3 & 4 have better resolution, but lack naturalness n Still perhaps not so much resolution as previous studies, e. g. , studies of verb bias effects n Baseline condition stimuli sound like utterances with special prosodic focus; hence, may be problematic as control or wholly ambiguous stimuli