feedback and the correct cue-target pairs were never shown to the participants again.
1.4.2 Crude assessment of the change in performance across the retention interval
Since the majority of sleep and memory studies used the paired-associate task to examine the role of sleep in offline declarative memory processing, the most common, if not the only, behavioural improvement index was an increase in the
number of correctly recalled word pairs from before to after a sleep retention interval^, while a decrease in the number of correct responses indicated forgetting. This traditional, global approach, though simple and efficient, focuses on the recall of all the word pairs, instead of examining the consistency of successfully retrieving
individual pair across the retention interval. This word pair-level analysis can
provide more information about the offline processing of declarative memory. For instance, at retest, correct recall responses may consist of word pairs that have never been retrieved successfully during leaming (‘memory gain’), and some of the word pairs recalled at leaming have been forgotten during the retention interval (‘memory loss’). When the gain is equal to the loss, the global analysis will show maintenance of recall performance, implying a stabilization of declarative memory. But, obviously, memory of word pairs ‘gained’ and those ‘lost’ has undergone different offline processing. Specifically, ‘memory gain’ can be attributed to the offline enhancement of word pairs never recalled at leaming, resulting in its successful retrieval at retest. On the other hand, ‘memory loss’ reflects a lack of stabilization of the previously acquired materials during the retention interval, leading to retrieval failure at retest. In other words, memory stabilization will allow word pairs recalled at leaming to be retrieved again at retest.
Since enhancement can only be shown with word pairs that have been poorly leamed and not recalled in the leaming session, while stabilization can only occur to those word pairs acquired successfully, the first step to distinguish between these two
^ Reaction time was measured in several studies, but it was indicated by the time elapsed from stimulus onset (presentation of cue) to button pressing before the verbal response was actually produced by the participants (Gais et al., 2007; Gorfine et al., 2007; Schabus et al., 2004; Schabus, Hôdlmoser, Pecherstorfer, & Klôsch, 2005). The limitation of this reaction time measure will be discussed in more detail in Chapter 5.
offline processes of declarative memory will be to categorize word pairs based on how well they have been acquired, i.e. level of leaming. The impact of level of leaming at a word pair levef on the sleep-declarative memory relationship has been reported in only two studies three to four decades ago. Lovatt and Warr (1968) who found that the more frequently an unrelated word pair was recalled in the leaming session, i.e. the better leamed the word pair was, the more likely it would be recalled in a delayed recall test after a 9-hour retention interval. More importantly, for those word pairs that were moderately leamed and only recalled once during leaming, sleep helped increase their recall probability in the delayed recall test as compared to an equivalent period of wakefulness (.92 vs. .59). However, as the level of leaming increased, the benefit of sleep over wakefulness in delayed recall probability narrowed, and in fact, no longer existed for extremely well acquired materials. Since these data were based on word pairs that had been recalled at least once in the leaming session, it can be concluded that sleep facilitates the stabilization of moderately leamed word pairs but not very well leamed declarative materials. It is unfortunate that Lovatt and Warr (1968) did not report the delayed recall probability of those poorly leamed word pairs which were not recalled at all in the leaming session, and hence, the effect of sleep on declarative memory enhancement could not be determined. Contrary to Lovatt and Warr’s (1968) results, Benson and Feinberg
^ The moderating role of level of performance in the sleep-declarative memory consolidation relationship has been investigated in two recent studies, but level of performance was based on the number of correctly recalled word pairs in cued recall tests instead of the consistency in recall as in the word pair level approach. Tucker and Fishbein (2008) categorized their participants into better and poorer learners with a median split of the number of correct recalls at the end of the leaming session. They showed that sleep improved the recall of semantically unrelated word pairs in better learners, but not in poorer learners. Drosopoulos and colleagues (2007) manipulated the level of performance either by interference or systematically varying the duration of stimulus presentation and leaming criterion. Their results indicated that sleep selectively facilitated the consolidation o f weaker memory associations, i.e. those impaired by interference and those acquired with a shorter stimulus presentation duration and to a lower leaming criterion.
(1977) found that the memory stabilizing effect of sleep on semantically unrelated word pairs decreased with level of leaming.
In light of these earlier findings (Benson & Feinberg, 1977; Lovatt & Warr, 1968), in all the five studies reported in this thesis, the paired-associate leaming session involved three presentations of word pairs, each followed by an immediate cued recall test. Thé level of leaming of each word pair was based on the number of immediate recall tests in which it was correctly recollected (immediate recall frequency or IRF). The retest session of Studies 1-4 consisted of three delayed cued recall tests, and how well each word pair was remembered after the retention interval was determined by the number of delayed recall tests where it was correctly recalled (delayed recall frequency or DRF).
For those word pairs never recalled during leaming, i.e. IRF = 0, the more its memory has been enhanced during the retention interval, the more frequently they will be remembered in the delayed cued recall tests (high DRF). If the enhancement of declarative memory depends on sleep, DRF of these word pairs will be higher after sleep than after wakefulness retention.
On the other hand, for those word pairs that have been correctly recalled at least once in the leaming session (IRF >= 1), their successful retrieval after the retention interval indicates a maintenance of performance and memory stabilization. On the contrary, an absence of stabilization will result in forgetting, and at retest, the word pair will not be consistently recollected, or in the most severe case, no longer be retrievable. Low DRFs will, therefore, be resulted. If sleep facilitates the stabilization
of declarative memory, DRF after sleep will be higher than that found after wakefulness.
However, even when no feedback is given in a recall test, performance in subsequent tests may improve (Bjork, 1975). Therefore, a more stringent way to quantify memory enhancement and stabilization is to consider the amount of word pairs at each level of leaming that are correctly recalled in the first test after the retention interval. Specifically, the extent of enhancement is indicated by the percentage of word pairs never recalled at leaming that could be remembered in the first delayed recall test. To quantify the extent of forgetting or a lack of stabilization, among those word pairs that had been consistently recalled throughout the leaming session (i.e. in all the three tests at leaming), the percentage that was not remembered in the first delayed test was calculated. A similar percentage was calculated for the forgetting of word pairs correctly recalled twice (i.e. in the second and the third tests at leaming), and those correctly recalled once (i.e. only in the final recall test at leaming). By comparing these three percentages which represent the forgetting of word pairs acquired to various extents, the effect of level of leaming on memory stabilization can be determined. The impact of sleep can also be investigated by contrasting these percentages after sleep and after wakefulness retention. It should be highlighted that to ensure that forgetting of those word pairs recalled in only some of the tests in the leaming session did occur during the retention interval, only those pairs successfully retrieved at the end of leaming were included. This is essential because for those word pairs correctly recalled initially during acquisition but was no longer remembered later in the leaming session, it is not clear whether forgetting took place before or during the retention interval.
Due to the additional details provided by a word pair level analysis over the traditional method, in this thesis, these two more sophisticated, word pair level approaches to define declarative memory stabilization and enhancement, both of which involving the examination of the consistency of word pair retrieval at leaming and at retest, were explored. The role of sleep in stabilizing and enhancing declarative memory as assessed with these two new approaches as well as the traditional method were contrasted.