Declarative Memory & Models

Declarative Memory & Models

1. Declarative memory: Memory that is accessible to awareness (consciousness) and can be verbalised e.g. Facts (semantic) and events (episodic)

   • How do we test it? Free recall and recognition memory
     • Was the picture presented before (Old) or not (New)?

2. Explaining Declarative Memory

   • Signal Detection Theory
     • Green & Swets, 1966)
       • Each item in test is associated with particular memory strength / familiarity
         • Strength is normally distributed. The mean of the strength of old items is greater than the mean of new items. There is OVERLAP between the two distributions
           • If the strength of an item exceeds the criterion ,C, then the item is judged "old", if not it is judged as "new"
             • Threshold needs to be exceeded to confirm familiarity
             • C is different across individuals due to being conservative / liberal etc
               • E.g conservative - less willing to respond "old"
                 • C moves to the right - proportion of false alarms and hits decreases
                 • Liberal = more willing to respond "old"
                   • d' = How much the distributions overlap (lower the value = more overlap, recognition is poor)
                     • Confidence ratings can also be used
                       • Multiple criterion
                         • Greater strength leads to greater confidence that an item is old
                         • C3 & C4 less likely to be accurate
                       • Predictions: -Accuracy -Decision RTs - ROCs
                         • Accuracy
                           • Mickes et al (2007)
                             • Presentation of new or old words and had to identify ones seen
                               • As people became more confident in making judgements (old / new) they became more accurate
                         • Decision Response Time
                           • Murdoch (2007)
                             • Higher confidence rating = faster response as strength value is translated into reaction time
                         • Receiver Operating Characteristics (ROCs)
                           • Hit rate (correctly recognised) Vs False alarm (incorrectly recognised)
                             • ROC curve is to be plotted
                   • Item needs a lower value of strength for me to say old
   • Squire, Wixted & Clark (2007)
     • Declarative memory varies in strength - strength signals in MTL
       • Performance in memory tasks such as recognition can be modelled

3. Equal-variance signal detection theory

   • All items have the same strength boost
     • Unequal-variance signal detection model
       • Variance of old item distribution is greater than new item distribution
         • Items have different strength boosts - more plausible?
     • In reality, the ROC slope is not equal to one- so this goes against SDT

Images and files in the map