RUMINATION ON AFFECT: CAUSE FOR NEGATIVE
ATTENTION BIASES IN DEPRESSION?

A Thesis
Presented to the
Faculty of
San Diego State University

In Partial Fulfillment
of the Requirements for the Degree
Master of Science
in
Psychology

by
Greg Jeremy Siegle
Fall 1996

THE UNDERSIGNED FACULTY COMMITTEE APPROVES
THE THESIS OF GREG JEREMY SIEGLE

	Rick E. Ingram, Chair	Date
	Georg E. Matt
	Peter F. Saloman

SAN DIEGO STATE UNIVERSITY
Fall 1996

Copyright 1996
by
Greg Jeremy Siegle

DEDICATION

This work is dedicated to Monica Barback and to any depressed organic or silicon beings who may be helped by its contents.

I think therefore I am is the statement of an intellectual who underrates toothaches. I feel therefore I am is a truth much more universally valid and it applies to everything that's alive." - Milan Kundera, Immortality, p. 200.

ACKNOWLEDGEMENTS

     Many many thanks to the folks who've helped me through their advice support, time, understanding, and encouragement. First let me thank, and publicly herald the wisdom and guidance of my advisors and committee members, Rick Ingram, Jörg Matt, Eric Granholm, and Peter Salomon. Rick, for sharing a profound and very human insight into depression. Jörg for helping me to see so many different ways of making research stronger. You have all been constant sources of inspiration and information for me. Monica, thank you for your well placed and much needed support, acts of genius when I thought all was lost, and healthy cynicism. Thanks also to Javier Movellan - over the course of a few weeks you reformed my perception and use of neural network and human information processing models. Without 6 extremely dedicated research assistants, none of this would have been possible. Mark Shibley, Daniel Grant, Ivan Nepomunko, Maureen Flaherty, and Sean Gyll, each of you has been more help than you can know. Your ideas directly contributed to many of the directions this work has taken, and will take in the future. Your many hours of work on this project is well noticed and much appreciated. Thanks also to the 168 conscientious research participants who gave their hours and a glimpse into their souls such that this project could be completed. Mom and Dad, thanks so much for your encouragement at each stage of my training leading up to this project, as well as your support throughout its gestation. The support and unwavering faith of my inlaws has also been more valuble than they can know. There are a number of other people whose suggestions, presence around the labs, comments, reassurances, and offhand remarks are most appreciated. Mark Dombeck, Chris Bernet, Christie Scher, Bob McGivern, Yukari Takari, and the other folks in Rick's, Jörg's, and Eric's labs, thanks. Thanks also to my advisors and teachers in the past who have helped to shape me into someone who could create this document.

PREFACE

This thesis is a model for what I hope will be a more and more common style of research. Often experimental psychopathologists design information processing experiments based on hypotheses generated from a literature review, and analyze their data via statistical procedure designed to test how well a general linear model can account for differences between group means. Then, the ``Discussion'' section of the thesis serves as a forum in which to summarize the ``Results'' section and integrate results with speculations from the ``Introduction'' or literature review section. This organization implicitly suggests that ideas in the ``Discussion'' section are data-driven, and are developed after/out of the acquired results. In reality, theories and models presented in the Discussion section are often the very ideas which guided the development of the thesis, and in fact, were developed concurrently with the writing of the introduction, and collection of the results.

This thesis has been created using such an iterative methodology. Initial ideas regarding information processing in depression were culled from the literature and operationalized in computer simulations meant to loosely simulate processes assumed to operate in depressed people. The simulations suggested experiments which might validate not only the computational model but the ideas which drove its creation. In this way, the computational simulations discussed in this thesis, while presented after the human data, were developed concurrently with the experimental methodology. The computational simulations are used not, as an adjunct to the human data, but as a springboard for its collection, a method of creating nonlinear models to fit aspects of the data, and as a platform from which to generate new theories of the interaction of emotional and semantic information processing in depression.

Still, a caution involving the wholehearted interpretation of computational results as indicative of mechanisms operating in depression is in order. The computational models presented in this paper are extremely simple, and represent very few of the intricate details assumed to be operative in human information processing. Though anthropomorphic terms such as ``depressed'' are used to describe the network models, no claim is being made that the models, or computers on which they are running, are actually ``depressed.'' Instead, such terms, when applied to the models merely represent labels indicating the human states or behaviors which guided the creation of a particular model. These labels are used primarily to save space. Rather than writing out ``Network overtrained on multiple vectors having a particular random configuration of activations, herein labeled 'negative' after having been trained on another set of vectors having a different set of random activations, herein labeled 'positive', and which is subsequently expected to generate characteristically biased outputs'', such a network is sometimes referred to as ``depressed.'' The reader is encouraged to willfully suspend some disbelief when reading suggestions for mechanisms operating in humans, culled from these models, but is, at the same time, admonished not to interpret the labels given to computational simulations too literally.

ABSTRACT

A research program designed to clarify the relationship of depression to biased attention towards negative stimuli is described. A series of published experiments have examined performance by depressed and nondepressed people on an affective lexical decision task in which research participants are asked to assess whether lexical stimuli spell a word, with varying results. This literature is integrated through a series of meta-analytic comparisons which suggest that depressed people are delayed in responding to negative but not positive or neutral words on the task. Results of the meta-analysis are interpreted to form a physiologically plausible theory which accounts for the observed aggregate data as well as inter-experimental differences. The theory suggests that depressed people systematically attend to the negative content of stimuli, regardless of their semantic content, and as such, identification of semantic content of negative stimuli may be delayed in depressed people. Depressed people are hypothesized to better recognize negative stimuli which are specifically depressotypic for them than other negative stimuli. An experiment employing both an affective lexical decision task and an affective valence identification task, in which participants with and without features common to depression are asked to judge whether stimuli are positive, negative, or neutral, is interpreted to support this theory. Implications of the theory for future research and treatment are generated using a cognitively and physiologically motivated computational neural network model designed to simulate performance by depressed and nondepressed people on these tasks. A simulated analog of "rumination" on the affective valence of a stimulus is shown to serve as a vulnerability and maintaining factor for the expression of information processing biases in the model. It is also shown to prevent the network from overcoming these biases upon exposure to simulated analogs of positive stimuli.