Only a minority of searchers in both databases reformulated their query. Satisfaction with initial search results as well as general motivation are prime factors in determining whether or not a user will try to improve a query in some way. Since end-users generally seem to be satisfied with so little, it may be that those who bother to or need to reformulate will always be in a minority. Nevertheless it is of paramount importance that those who do wish to reformulate should be given appropriate support from the system so that they can do so confidently and with the least possible effort.
The analysis of the logs of this and previous Okapi projects, and of other systems, indicate that it is not uncommon for users to repeat queries and start over for no apparent reason [13]. In addition some manipulate the query and then fail to search [5]. This observed behaviour may be due to a lack of confidence or knowledge of how systems really operate but it appears to be indicative of an underlying need which is not being met. Given better support or understanding of the system's functionality, the benefits of reformulating queries ought to become more apparent to users --- the challenge is still how to achieve this objective through interface design.
In spite of seemingly better initial search results, a quarter of INSPEC users, as opposed to only 14% of LBS users, reformulated queries (see section 5.2.4). Individual searchers in both groups tended to adopt one single strategy only --- the dominant one being to add their own terms. Elsewhere this has proved to be an effective strategy [14]; in the current study the tendency was more prevalent for LBS users, whereas INSPEC users were also likely to add suggested thesaural terms. The variation may reflect differences between the nature or type of queries in the two subject areas, and / or the quality of the respective thesauri.
Removing terms automatically extracted from relevant items can also be classed as interactive query reformulation. INSPEC users were just as likely to remove candidate terms generated by this process as to add suggested terms from the thesaurus. Overall there was a greater degree of interactivity in searching the INSPEC database than the LBS catalogue --- reformulation was more dependent on the mixed initiative of user and system.