Oldal 42 [42]
CHAPTER 4 THEORETICAL FRAMEWORK
other, striving to become the actual output realization, the optimal candidate
complying with the rule of well-formedness. Ihe number of potential
competing candidates (inputs) is infinite. However, through the derivational
process, their number is reduced to only one, which ends up as the optimal
candidate, the actual surface realization. The actual surface realization is the
candidate that complies the most optimally with the universal rule of well¬
formedness.
The evaluative part of universal grammar (EVAL) evaluates the competing
candidates, the potential output representations. The candidates are inputs
with a corresponding output representation. The output representation is
selected out of the set of all possible output representations or candidates.
The competing candidates have to undergo a set of violable and hierarchically
ranked constraints, and the EVAL part of generative grammar will select, out
of an indefinite number of inputs, the optimal one. The optimal candidate
is the one that violates the lowest ranked constraint(s) but not the highest
one. There is a strict hierarchy in each language, meaning that the order of
constraints cannot be changed in a given language, and the higher ranked
constraint has absolute priority over the lowest ranked one(s). However, since
not necessarily all constraints are activated in a given linguistic situation,
only the relevant ones are arranged into hierarchy. The derivational speech
production process, through which the particular underlying representation is
turned into the corresponding surface realization, activates only the relevant
constraints.
The constraints are violable because even the optimal candidate, the actual
output representation, may violate some of them. The only inviolable rule
in terms of the constraints is that a candidate violating the highest ranked
constraint activated in the relevant speech production process cannot be
the optimal one.
The constraints of well-formedness are universally applicable, but their
actual ranking is always language specific. Hence, it is the actual language
specific ranking of constraints that determines the optimal candidate. Fora
language specific ranked set of constraints, a candidate A is more harmonic
than candidate B if A is more harmonic with respect to the highest ranked
constraint on which the two candidates differ. The optimal candidate (the
selected output) is the candidate that is more harmonic than all the others
with respect to the ranked constraints. There is no cumulative effect of
constraints, which means that no matter how many lower-ranked constraints
one candidate violates if it does not violate the highest ranked constraint, it
will end up as the optimal candidate.
The ranking of constraints is based on an algorithmic computational
process applied on empirical data. Although there are linguistic characteristics
rendering particular languages more salient toward a specific ranking, the
+ 40 +