Gems

[Margaret Fleck]

All methods of collecting child language data have fairly serious limitations.
   To
even out the rock throwing, why don't I point out that audio recording provides
only
a very small sample of a child's speech, tends to lose important contextual
information
or require absurd amounts of observer time, and has a bias towards situations
that
are nice for audio recording (low noise, civilized hour of the day, few
strangers who
would engender permissions problems, limited water play).

Rather than fighting over which method is best, or trying to pretend that one
of these
methods is bias-free, it would seem more productive to find out how we can use
different types of data to complement one another.

We might also think about whether these methods share any common biases.
For
example,  one might wonder whether any of these methods successfully captures
the
real-world frequency of (say) swear words  or amateur-quality  singing in
parental
speech.    I also suspect we know little about what happens in situations such
as
grocery shopping or freeway journeys or beach expeditions or that hectic period
in
the morning when everyone has to be gotten ready for school/work,   where audio
and written recording are both difficult and parents are highly distracted.

Margaret Fleck
     U. Illinois, Urbana-Champaign (in 2 weeks)