An investigation into the effect of formatting manipulations on reading and information retrieval from text documents

The BBC Audience Lines, a call centre based in Glasgow, voiced an interest into whether their system, of responding to queries from the general public via telephone could be improved. In response to this, a detailed Task Analysis was conducted, which identified information retrieval as the specific area in which time was misused during call answering. A series of experiments was then conducted to investigate whether manipulating the layout of a text document could influence participants' reading behaviour or their strategies and latencies in a subsequent search task. Studies 2 and 3 illustrate that more fluent reading experiences and faster search latencies are evident when participants first read plain, unformatted text. Initially reading blanket bulleted text disrupts the reading process by affecting reading time and all eye movement measures, and also creates an improverished memory of the text as evidenced by poorer search latencies. This pattern of results held during both paper and computer screen search of a document. Study 4 revealed that many of the problems associated with blanket bulleted text were seen to be resolved by the use of stem sentences in combination with the bullet points, resulting in fluent reading and a faster search latency than plain formatting. The reason for this advantage for stem bulleted formatting was probed during Study 5 by investigating whether there is a stronger visual representation of the text created in memory during the reading of stem bullet formatting. Weak results were obtained, suggesting the task used in Study 5 may not actually probe the same representation created in memory as Studies 2 to 4 did. This is discussed in depth. It is concluded that during reading both semantic and visual information is encoded and play a role in information retrieval. In addition, the way the document is formatted can alter the way a person reads a piece of text and hence encodes its information content, particularly in leading participants to have a greater of lesser semantic representation of the passage. However, all formatting influences occur at the time of encoding, and no difference is seen in the ability of participants to locate information within a page of text they have not first read, regardless of format. The implication of these findings are discussed in relation to the work of the BBC Audience Lines