|
Abstract:
lthough the visual recognition of morphologically complex
words has been well-studied, there has been little attention for
their spelling process. Yet the spelling of Dutch regularly
inflected verb forms presents a challenge for sentence processing
research, as experienced writers often make errors on these forms,
even though (i) their spelling invariably involves a rule of the
simple "add suffix" type, and (ii) they involve a single rule only
(hij biedt
[he offers]:
bied + t ).
The persistence of these errors is surprising, considering the
time invested in verb spelling in education and the social
unacceptability of the errors. Clearly, they must inform us on
underlying spelling processes.
We investigated the spelling of homophonous verb forms, i.e.,
whose pronunciation corresponds to two spelling patterns ([vInt]:
vind - vindt),
and manipulated the frequency relationship between the target form
and its homophone (higher, lower, equal frequency). Relative to the
frequency-matched condition, more intrusion errors were expected
when the homophone frequency was highest and fewer when it was
lowest. The distance between the target and the morpheme
determining its spelling was also manipulated (zero or four words).
We hypothesised more errors at the longer distance than at the
shorter one. Adult language users (18-year old university students)
and two groups of adolescents (12 and 14 years old)
participated.
In Experiment 1 we used the homophonic relationship between the
first and third person singular in the present tense of some verbs
(vind - vindt).
Both the frequency and distance factors determined error rates and
were additive. In Experiment 2 the homophonic relationship obtained
between the third person singular present tense and the past
participle of certain prefixed verbs
(versiert - versierd).
We replicated the results for the present tense and found a third
source of intrusions for the participles: the co-occurrence
frequency of the prefix and the suffix. Although the children's
pattern was not the same as the adult one, there was a clear
evolution towards it over the two adolescent groups.
The data are compatible with a race model of the spelling
process, in which a time-consuming rule-driven process operates
alongside frequency-driven retrieval processes within an
orthographic lexicon. Factors delaying the rule-process
(within-sentence distance) increase the likelihood of intrusion
errors originating in retrieval processes. In retrieval,
high-frequency whole-word orthographic targets are less error-prone
than low-frequency ones. Suffix spellings are also retrieved on the
basis of co-occurrence frequencies with other word parts (e.g.,
prefix). The relative speed of all these processes determines the
speller's eventual output.
These spelling results will also be compared to self-paced
reading data for the same stimuli, to see whether the implicit
spelling check during reading is subject to the same factors.
|
