fill in the missing word

[Virginia Marchman]

Dear info-childes,

We have used a version of this task in our work with several
groups of children 3 1/2 years and up, and I would like to
make a few comments/clarifications in the context of the ongoing discussion
about task demands and the frequency of zero-marking vs. overregularization
errors.

In the form of the task that we use, the procedure involves
picture-supported
prompts similar to the original Berko study (also Bybee & Slobin),
for example, "This boy is running. He runs everyday. Yesterday,
he.....". While
it officially could be described as a "fill in the missing word" task,
this is not
really the best way to describe the task demands.  As Jean Berko points out,
the fill in portion comes at the end of the prompt, and so the task is
really more analogous
(in terms of memory load) to the "finish what i've started" version
outlined by Thomas.  Children are not required to repeat any part of the
prompt and generally one word responses are all that are needed.

My intuition is that the main impact of the "finish what I've started"
manipulation
(versus "fill in at the end") would be to reduce the number of times that
the child provides a response that fits with the semantics of the prompt
but does not
use the target word (e.g., "this boy is running.  Yesterday, he...."
"Fell down").
Because the child does not know what sound they should start their
response with, they are potentially free to provide something new. In
all of our analyses, of course, responses in which the child provides
their own new verb are excluded from final counts of
particular error types (e.g., zero-marking vs. overregularization).

To comment on a minor point made by Brian, this task should not be
considered to
be "primarily dependent on verbal presentation" in that the children are
always
shown pictures depicting the action in conjunction with the prompt.  The
verbal
presentation is important in that we try to get the child "use the same
word that
I do", but the child has visual input to support their response as well.

Most substantively, it is important to point out that at least some
aspects of
the theoretical issues at hand are crucially dependent on *which* items are
subject to which error types. The goal of our studies (see references)
was to determine the degree to which item features (e.g., phonological
"shape")
predicts zero-markings versus overregularizations. We found that regardless
of the overall frequency with which zero-marking or overgeneralizations
were produced (e.g. more zero-markings in children with SLI than NL),
certain *items* were more (or less) likely  to be vulnerable to those
errors in both groups of children.

For example, if a zero-marking is going to occur, it is more
likely to occur on a verb that ends in an alveolar stop consonant
(for both regular and irregulars);  irregular verbs are more likely
to be overregularized when they have similar phonological  features
("neighbors) as regular verbs, etc.  These findings would suggest  that
even though task demands may make zero-marking or overregularization
more or less likely to occur in general, it is item characteristics that
predict
the likelihood of particular error types.  Interestingly, it may be
possible that changes
in zero-marking vs. overregularization frequency in the two tasks  (reported
by Thomas) is the consequence of how the task demands influence
*which items* are produced erroneously, rather than zero-marking or
overregularization
per se.

We should note that some recent studies (e.g., van der Lely & Ullman,
2001) do not
find a substantive role for item characteristics in determining error
type.  However,
readers should evaluate the degree to which groups of items are balanced for
these types of item characteristics (e.g., if zero-marking is strongly
predicted
by presence of stem-final alveolar stop consonant, then this feature
should be
represented in both regular and irregular items, high vs. low frequency
items, etc).

-Virginia Marchman

References follow:

Marchman, V., Wulfeck, B., & Ellis Weismer, S. (1999). Morphological
productivity in children
with normal language and SLI: A study of the English past tense. Journal
of Speech, Language
and Hearing Research, 42, 206-219.

Marchman, V. (1997). Children's productivity in the English past tense:
The role of
frequency, phonology, and neighborhood structure. Cognitive Science,
21(3), 283-304.


Michael Thomas wrote:

>> In the Thomas et al. paper, we were purposely replicating other work
>> to make a direct comparison. But we did use two different tasks to
>> compare the results.  There is also the general issue about what lack
>> of generalisation to nonce terms in atypical populations really means.
>> Annette
>
>
> Just to reinforce Annette's comments, in our work with children and adults
> with Williams syndrome, we were concerned that aspects of the task demands
> in the past tense elicitation task might differentially affect our
> atypical and control populations.
>
> Our initial aim was to replicate work by Harold Clahsen and Mayella
> Almazan (Cognition, 2000) who used a past tense elicitation task on a
> small sample of children with WS. Our aim was to replicate these findings
> with a larger sample size, so we stuck to their stimuli and procedures. To
> address our concerns about task demands, we also we included a past tense
> elicitation procedure developed by Lorraine Tyler and William
> Marslen-Wilson for use with adults with aphasia. In contrast to the "fill
> in the missing word" procedure, this task required no repetition of
> previous words, and participants were prompted in their response by the
> provision of the initial sound of the word. We called this procedure
> "finish the word I started":
>
> The participant was told "I'm going to say a sentence, and then I'll start
> another one and stop in the middle.  Your job is to finish off the word
> that I've started."
>
> For example: "The bull sometimes kicks. Yesterday, it k ___"
>
> This is similar to the procedure used by Jean Berko Gleason, (although
> compared to her procedure, we omit an intervening question, i.e. The bull
> sometimes kicks. What did the bull do yesterday? Yesterday it ___ ) As
> Jean comments, putting the required word at the end of the sentence may
> provide a context which helps to constrain the response to the inflected
> form of the verb.
>
> By incorporating a second task, we were then able to examine how the
> respective task demands affected the responses, and whether any effects
> differed between atypical and control groups. The results showed a slight
> advantage in performance on existing regular and irregular verbs in the
> "finish the word I started" task over the "fill in the missing word" task,
> but no interaction with participant group.
>
> The interesting result here was in the pattern of errors, which showed up
> most clearly on irregular verbs. For both the participants with Williams
> syndrome and the typically developing children (our sample here was 6 year
> olds through to 10 year olds), the "fill in the missing word" procedure
> produced predominantly zero marking errors, whereas the "finish the word I
> started" procedure produced predominantly over-regularisation errors. Thus
> the task demands appeared to alter the pattern of errors from one of
> omission to commission, while the level of correct performance was broadly
> similar between the two tasks.
>
> This is consistent with Brian's suspicion that the form of the task itself
> may encourage zero marking errors. As Brian points out, the level of zero
> marking errors is of theoretical importance in studying disorders such as
> Williams syndrome and Specific Language Impairment, so evidence of the
> role of task demands is significant here. However, currently we are not
> clear whether the higher level of zero marking errors in the "fill in the
> missing word" task was due to that task's greater memory load (i.e.,
> repetition of a sentence fragment prior to offering the verb form), or due
> to a context that constrained a past tense response less strongly than in
> the "finish the word I started" procedure.
>
> In addition, we found that this task demand worked *differentially* on our
> atypical and control groups. The shift from zero marking to
> over-regularisation errors was characteristic of the younger control
> children. However, participants with Williams syndrome persisted in
> showing this pattern at verbal mental ages at which one would expect it to
> disappear. That is, they persisted with an immature pattern of response to
> the task demands.
>
> Unfortunately, we weren't able to compare task effects on the production
> of nonce terms, since our second elicitation task comprised a larger set
> of regular and irregular verbs to allow examination of frequency and
> imageability effects. We do think it is important to collect data on the
> effect of task demands on nonce terms, because we believe it is still an
> open question about how one interprets failure to generalise to nonce
> terms in atypical populations, and in particular, atypical populations
> with lower IQs.
>
> We also agree with Brian's suggestion that a comparison of question
> answering and sentence completion paradigms would offer an important
> clarification, especially in the light of ours and Brian's initial
> evidence for the effect of task demands in these elicitation tasks.
>
> If anyone would like further details of the error data produced by our
> Williams syndrome and control populations, I'd be happy to provide them.
> (Incidentally, if you're trying to replicate these results, it's worth
> noting the statistical difficulties that arise in comparing
> non-independent proportions of error types!)
>
> cheers,
>
> Michael
>
>
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Dr Michael Thomas
> Senior Research Fellow
> Neurocognitive Development Unit
> Institute of Child Health
> 30 Guilford Street
> London WC1N 1EH,  U.K.
> tel:      +44 (0)20 7905 2747
> fax:     +44 (0)20 7242 7717
> http://www.ich.ucl.ac.uk/units/ncdu/NDU_homepage.htm
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
>
>
>
>

--
*******************************************
Virginia A. Marchman, Ph.D.
P.O. Box 830688
School of Human Development GR 41
The University of Texas at Dallas
Richardson, TX  75083-0688
Office:	(972) 883-6430 Fax: (972) 883-2491
Home:	(972) 509-9211
vamarch at utdallas.edu
*******************************************