· Click HERE for the abstract.
A PDF version of the whole thesis can be retrieved by clicking HERE, or a compressed postscript version HERE. Note that the postscript file is 1.7 MB when uncompressed, and 315 pages when printed.
Alternatively, PDF versions of selected chapters can be viewed by clicking on the highlighted areas in the overview below:
In Chapter 1, the problem of serial order is introduced, together with three possible solutions: chaining theory, positional theory and ordinal theory. Some experimental paradigms that have been used to test these theories are discussed, such as serial learning, probed recall and serial recall, and it is argued that the analysis of errors in serial recall is the most informative approach. A comprehensive classification of the typical errors in short-term, serial recall follows, together with a summary of the empirical, statistical and computational methods employed in the thesis.
In Chapter 2, the predictions of chaining theory are tested in immediate serial recall of lists of phonologically confusable and nonconfusable items. The data from Experiment 1, together with those in Henson et al. (1996), provide no support for chaining theory.
In Chapter 3, the effects of list length, grouping (Experiment 2) and proactive interference (Experiment 3) are examined. The pattern of errors found between groups and between trials is explicable by a positional theory of serial order, but not an ordinal theory.
In Chapter 4, meta-analyses of error data are performed on a number of experiments performed recently at the Applied Psychology Unit. These analyses produce a set of empirical constraints that any model of short-term, serial recall must meet. No previous model can.
In Chapter 5, a new, computational model of serial recall is developed, the Start-End Model (SEM), which meets the empirical constraints of Chapter 4, and provides quantitative fits to the data from Experiments 1, 2 and 3. This model is an example of a positional theory. Extension of SEM to other phenomena in STM is discussed, as is its relationship to previous models. Most importantly, SEM predicts an new property of positional errors in serial recall.
In Chapter 6, Experiments 4 and 5 confirm the predictions of SEM, and pose a serious challenge to other positional models, particularly those that assume positional codes are generated by internal oscillators in memory.
In Chapter 7, Experiments 6, 7 and 8 examine the effects of repeated items in serial recall. The results, together with those in Henson (1996b), are consistent with the basic assumptions of the SEM, but suggest that several additional processes are involved in memory for repeated items.
In Chapter 8, the more general assumptions of a positional solution to the problem of serial order are discussed. It is concluded that, while not denying other representations of serial order, particularly in procedural memory, positional theory appears a promising approach to the problem of serial order, particularly in episodic memory.
Bibliography contains the references.
Appendix 1 summarises some statistical techniques used in the thesis.
Appendix 2 lists details of the experiments in the meta-analyses of Chapter 4.
Appendix 3 gives a formal definition of the Start-End Model, together with four additional qualitative fits to the effects of retention interval, list length, articulation rate and modality.