Re: Measuring readability

[Damien Braniff <Damien_Braniff -at- PAC -dot- CO -dot- UK>]

At the ISTC conference last year there was an interesting presentation on
readability by Sandra Harrison (Coventry University) and Paul Bakker
(Eindhoven University of Technology).  It was based on a new package for
determining readability.  It works on Word documents and included
information explaining the statistics generated by Word, how to interpret
them and how to use (and not use!) them.  This was all fairly standard but
it also reported on research carried out on two other readability
descriptors - lexical density and distribution of sentence and "packet"
length.  Conclusions of early research indicated that:

   Lexical density of text may give a better indication of readability
   than many of the common formulae.

   The effective breaking up of sentences into "packets" may be as
   important to readability as the actual sentence length.

   In looking at mechanical variables of texts, we should not only be
   concerned with averages, but with distributions and most frequently
   occurring values.

The package was demonstrated at the conference and the idea was to make it
available as shareware once the online help had been completed (not sure
how far along this is yet).  The demonstration was quite impressive on a
variety of texts, both technical and non-technical.  The speakers said it
could not be used to indicate good writing but could give a good indication
of bad writing!  Basically a "good "score" didn't mean the text was well
written but a "bad score" probably indicated that the text would not be
easy reading.

Damien Braniff
Technical Author
PAC International