Not all of the information held in SIS can be utilized in later stages of memory. Because the attending response can be given to only parts of the information in SIS, a sequence of attending responses is necessary for all information in SIS to be extracted. However, the "image" in SIS decays rapidly, and frequently this occurs before the sequential attending responses can be made. Hence some information may be lost due to lack of attention. Another problem, too, contributes to the loss of information from SIS; the STM has a relatively small information storage capacity capable of retaining only a few units or "chunks" of information.

Thus, for instance, the following list of words is difficult to recall if one reads them one after another at a rate of about one per second: the, happy, cage, open, now, bright, to, door, and, possible, yellow, see, had, that, to, canary, that, been, it, fly, was, the, left, was, away. After seven or so words, our STM capacity is approached; we need to perform a control process called rehearsal (i.e., repeating the list several times) in order to finally store the total list in LTM for subsequent recall.

However, if the list of words is rearranged as follows: "the bright, yellow canary was happy to see that the cage door had been left open, and that it was now possible to flyaway," we are able to change the size of the "chunks" in STM from words to meaningful phrases and substantially increase the amount of information we can serially extract from the visual image in SIS and retain in STM for immediate reproduction. Recoding of SIS information into language information is a control process performed in STM to improve the amount of information that can be processed in a unit period of time.

Concerning STM then, we consider that STM processes are under direct control of the individual; they govern the actual flow of information in the memory system. Also, short-term memory has a relatively small capacity capable of retaining only a few units or "chunks" of information, and a storage survival duration of approximately 15-30 seconds. The retrieval of information from this temporary working memory is rapid, accurate, and reliable.

In contrast to the STM, the LTM is considered to be a virtually permanent memory store with no practical limit in terms of capacity. Through such STM control processes as rehearsal or recoding, information may be transferred from STM and retained in LTM. Information retrieval from LTM is a complicated, and sometimes unsuccessful, process which requires the activation of searching and scanning strategies in STM. It is thought that much information apparently lost is actually stored in the LTM system, but has become inaccessible (Greeno and Bjork, 1973). In the present model, LTM is referred to as the cognitive content, and contains the conceptual base and the language system as subcontents. Since these aspects of LTM have been discussed in Chapter III, we will not pursue them further at this point.