國
立立 政 治 大
㈻㊫學
•‧
N a tio na
l C h engchi U ni ve rs it y
b. [a.tóm] *! *
c. a[tóm] *!
d. [á.tom] *! * *
A possible concern of the dominance of NON-FIN may arise in tackling monosyllabic nouns. To be more specific, will an unfooted monosyllabic foot surpasses a footed counterpart? However, such a problem can be easily overcome by placing STRESS over NON-FIN, which then forces monosyllabic nouns to bear main stress.
XV-ending nouns
In the previous section, the elicitation data of stress assignment in English simplex nouns show that L2 learners tended to place the primary stress on the XVO ultima, otherwise the penult. The preference for a rightmost moraic trochee is universally unmarked. Harris (1983) states that a wide range of evidence has supported that the most unmarked nonverb stress are in two kinds in (17):
(17) Unmarked nonverb stress
a. penultimate stress in vowel-final word b. ultimate stress in consonant-final words
Based on the proposed algorithm, universal unmarkedness in the interlanguage can be obtained via the bidirectional model. L2 learners’ preference can be attributed to the interlanguage markedness constraints, FT-BIN(µ), TROCHEE(µ), and ALIGN-HEAD in the undominated position at the first stage. In the bidirectional model, grammar construction is divided into two stages. In the first stage, interlanguage markedness constraints are promoted to the undominated position. At the same time, L2 constraints join interlanguage grammar construction from a lower rank than L1 constraints. Constraints activated in the interlanguage are generally effectuated for a more simplified and unified process in dealing with L2 tokens before a solid interlanguage grammar is formed. These constraints are originally ranked low or even inactivated in the L1 or L2. Yet, only through promotion can they become undominated and determinant enough to regularize the interlanguage forms.
Expectably, at the first stage, the promotion of interlanguage constraints yields learner’s forms that are distinct from L1 or L2 forms. At the second stage, the
•‧
國
立立 政 治 大
㈻㊫學
•‧
N a tio na
l C h engchi U ni ve rs it y
intermediate grammar is modified by error-driven constraint demotion. Since grammar modification at the second stage is error-driven, revision suspends when correct L2 forms are successfully selected but continues instantly when learners’
forms are wrongly predicted.
In the following analysis, I will account for the elicitation data based on the bidirectional model in the extended EDCD. The extended EDCD holds that the initial state of interlanguage grammar is L1 grammar, which can be illustrated in (18). As can be noticed, there is no L2 input at this point, and therefore the application of L1 grammar to L2 stimuli is impossible.
(18) The initial state of simplex noun stress grammar FT-BIN(σ) ALLFTL PARSE TROCHEE(σ)
Next, with the exposure to L2 stimuli, the first stage is initiated. In the bidirectional model, interlanguage markedness constraints are activated and promoted to the undominated position. In the case of simplex nouns, the interlanguage constraints to be promoted to the undominated position is ALLFTR. ALLFTR is triggered under the preference for a head-foot on the right of a prosodic word, which tallies with universal typology. According to Gordon (2002), 59.6% of single stress languages have final or penultimate stress. The emergence of the unmarked is supported by the studies by Broselow, Chen, & Wang (1998). At the same time, L2 constraints participate interlanguage grammar construction from the lowest rank.
Tableau (19) illustrates the evaluation of stage one interlanguage grammar. As can be seen, ALLFTR is promoted to the highest rank, followed by L1 constraints and L2 constraints sequentially. Since the promoted constraints are also L2 constraints, they do not occupy the rank like the other L2 constraint, NON-FIN, repeatedly.
The evaluation in tableau (18) demonstrates how the interlanguage grammar in stage one operates. Here, candidate (a) survives as the learner’s form; the violations of having the ultima footed and the ultima unstressed are minor. Candidate (b) is ruled out as its iambic foot incurs a serious violation of TROCHEE(µ). The target form, candidate (c), although shares the same penultimate stress as the winner, does not survive in the evaluation because the head foot is monomoraic, which is punished by FT-BIN(µ) and FT-BIN(σ). Besides, the head foot of candidate (c) does not coincide
•‧
國
立立 政 治 大
㈻㊫學
•‧
N a tio na
l C h engchi U ni ve rs it y
with word on the right edge, which results in one violation of ALLFTR. The unparsed ultima incurs another two violations of PARSE and ALIGN-HEAD. TROCHEE(µ) and TROCHEE(σ) are incurred, for TROCHEE(µ) prefers LL and H only, while TROCHEE(σ) prefers HL, LL, or LH only. Therefore, candidate (c) is defeated by candidate (a) in the evaluation. Candidate (d) violates the all of the constraints except ALLFTR and ALIGN-HEAD.
Even though the correct footing has not been achieved at stage one, the result tallies with the elicitation data that almost all L2 learners at lower level could stress nouns composed of two light syllables penultimately at the very beginning. Almost no students assign the main stress ultimately in LL nouns.
(19) Stage 1: constraint promotion and L2 constraint incorporation
IL L1 L2
visa
ALLFTRIL=L2 FT-BIN(σ)L1 ALLFTLL1 PARSEL1 TROCHEE(σ)L1 NON-FIN(σ)L2 NON-FIN(FT)L2 FT-BIN((µ)L2 TROCHEE(µ)L2 ALIGN-HEADL2
→ a. [vísa] * *
b. [visá] *! * * *
← c. [ví]sa *! * * * * * *
d. vi[sá] *! * * * * * * *
In tableau (19), the target form has not been selected, so interlanguage grammar will undergo readjustment in stage two. In the second stage, constraints that disfavor the target form will be demoted rank by rank. With the violations that candidate (c) commits in tableau (19), the constraints to be demoted are all the constraints except ALLFTL, NON-FIN(σ), and NON-FIN(FT). The sequence of constraint demotion is determined by the frequency of the corresponding L2 stimuli, which can be detected by their ranking in L2. A constraint ranked higher in L2 is assumed to result in more corresponding outputs than a lower-ranked counterpart. The more often a certain type of stimuli appears, the more the constraint’s status will be reinforced, which makes the constraint less mobile. In other words, the higher-ranked constraints in L2 grammar will be demoted later in the interlanguage grammar. Considering that
•‧
國
立立 政 治 大
㈻㊫學
•‧
N a tio na
l C h engchi U ni ve rs it y
English is quantity-sensitive and exclude ultima in footing, target-form disfavoring L1 constraints, including TROCHEE(σ), PARSE and FT-BIN(σ), are more violable in English than the L2 constraints, TROCHEE(µ), FT-BIN((µ), and ALIGN-HEAD.
Therefore, target-form disfavoring L1 constraints should be demoted prior to target-form disfavoring L2 constraints.
Tableaux (20) to (22) illustrate the step by step demotion of L1 constraints that disfavor the target form. In tableau (19), TROCHEE(σ), PARSE and FT-BIN(σ)are demoted by one rank respectively. Since the target form does not surface, another constraint demotion proceeds in a similar way.
(20) Stage 2A: error-driven constraint demotion of TROCHEE(σ), PARSE and FT-BIN(σ)
visa
ALLFTR IL=L2 ALLFTLL1 FT-BIN(σ)L1 PARSEL1 TROCHEE(σ) L1 NON-FIN(σ)L2 NON-FIN(FT)L2 FT-BIN((µ)L2 TROCHEE(µ)L2 ALIGN-HEADL2
→ a. [vísa] * *
b. [visá] *! * * *
← c. [ví]sa *! * * * * * *
d. vi[sá] *! * * * * * * *
In tableau (21), error-driven constraint demotion takes one more step by downgrading PARSE, TROCHEE(σ) and FT-BIN(σ) by one more rank, uncrucially ranked with NON-FIN(σ), and NON-FIN(FT). As expected, the new constraint ranking still fails to predict the target form as these constraints are too low ranked to influence the result. However, even though the constraint demotion does not alter the result instantly, a quick solution by demoting higher-ranked constraints first is theoretically groundless and should not be applied.
•‧
國
立立 政 治 大
㈻㊫學
•‧
N a tio na
l C h engchi U ni ve rs it y
(21) Stage 2B: error-driven constraint demotion of TROCHEE(σ), PARSE and FT-BIN(σ)
visa
ALLFTR IL=L2 ALLFTL L1 FT-BIN(σ)L1 PARSEL1 NON-FIN(σ)L2 NON-FIN(FT)L2 TROCHEE(σ)L1 FT-BIN((µ)L2 TROCHEE(µ)L2 ALIGN-HEADL2
→ a. [vísa] * *
b. [visá] * *! * *
← c. [ví]sa *! * * * * * *
d. vi[sá] *! * * * * * * *
In tableau (22), all the violated constraints by the target form have all been demoted to the lowest rank, except for ALLFTR. As the target form fails in the competition, another cycle of demotion initiates with the other target-disfavoring constraints.
(22) Stage 2C: error-driven constraint demotion of PARSE and FT-BIN(σ)
visa
ALLFTRIL=L2 ALLFTLL1 NON-FIN(σ)L2 NON-FIN(FT) L2 FT-BIN(σ)L1 PARSEL1 TROCHEE(σ)L1 FT-BIN((µ)L2 TROCHEE(µ)L2 ALIGN-HEADL2
→ a. [vísa] * *
b. [visá] * *! * *
← c. [ví]sa *! * * * * * *
d. vi[sá] *! * * * * * * *
In tableau (23), ALLFTR is lowered by one rank, uncrucially ranked with ALLFTL. However, it is obvious that the alignment of head foot with the prosodic word on both edges prevents candidate (a) from being beaten by candidate (c).