<%BANNER%>

Attentional Contrast During Sequential Judgments: An Explanation for the Number-of-Levels Effect


PAGE 1

ATTENTIONAL CONT RAST DURING SEQUENTIAL JUDGMENTS: AN EXPLANATION FOR THE NU MBER-OF-LEVELS EFFECT By ELS DE WILDE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Els De Wilde

PAGE 3

This dissertation is dedicated to my parents.

PAGE 4

ACKNOWLEDGMENTS This dissertation leaves me indebted to many people. First, I would like to thank my advisor, Chris Janiszewski, for taking up the mentorship role and guiding me through the dissertation process. Interactions with other members of my committee were also invaluable in shaping this dissertation. In particular, I would like to thank Alan Cooke for his conceptual guidance and Richard Lutz and Asoo Vakharia for their feedback. My thanks also go to my friends in Belgium and in the doctoral program and my colleagues at HEC Montral for their support. I would like to thank JoAndrea Hoegg and Andrew Kuo for their assistance with the data collection while I was at long distance. I am extremely thankful to my family members for their encouragement throughout the entire duration of my doctoral studies. Finally, I would like to thank Roberto for giving me the best reason to finish this dissertation. iv

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iv LIST OF TABLES ............................................................................................................vii LIST OF FIGURES .........................................................................................................viii ABSTRACT ......ix CHAPTER 1 INTRODUCTION........................................................................................................1 2 THEORETICAL BACKGROUND..............................................................................4 The Number-Of-Levels Effect......................................................................................4 Existing Accounts of the Number-Of-Levels Effect....................................................4 Attentional Contrast......................................................................................................9 3 EXPERIMENT 1: DEMONSTRATION OF THE NUMBER-OF-LEVELS EFFECT WITH EXPANDED DESIGN....................................................................16 Method........................................................................................................................16 Results.........................................................................................................................17 Discussion...................................................................................................................20 Experiment 1A............................................................................................................21 4 EXPERIMENT 2: TEST OF ATTENTIONAL CONTRAST ACCOUNT WITH NO DIFFERENCE IN NUMBER OF LEVELS........................................................28 Method........................................................................................................................28 Results.........................................................................................................................30 Discussion...................................................................................................................31 5 EXPERIMENT 3: TEST OF ATTENTIONAL CONTRAST ACCOUNT WITH REVERSED NUMBER OF LEVELS........................................................................35 Method........................................................................................................................35 Results.........................................................................................................................36 Discussion...................................................................................................................38 v

PAGE 6

6 GENERAL DISCUSSION.........................................................................................41 7 LIMITATIONS AND FUTURE RESEARCH..........................................................45 LIST OF REFERENCES...................................................................................................47 BIOGRAPHICAL SKETCH.............................................................................................49 vi

PAGE 7

LIST OF TABLES Table page 1. Number-of-levels manipulations used in published studies..........................................15 2. Attribute levels from experiment 1................................................................................26 3. Mean difference scores and simple effects for experiment 1........................................26 4. Mean difference scores and simple effects for experiment 1A......................................27 5. Mean difference scores and simple effects for experiment 2........................................34 6. Attribute levels from experiment 3................................................................................40 7. Mean difference scores and simple effects for experiment 3........................................40 vii

PAGE 8

LIST OF FIGURES Figure page 1. Design for experiments 1 and 1A..................................................................................25 2. Design for experiment 2.................................................................................................33 3. Design for experiment 3.................................................................................................39 viii

PAGE 9

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ATTENTIONAL CONTRAST DURING SEQUENTIAL JUDGMENTS: AN EXPLANATION FOR THE NUMBER-OF-LEVELS EFFECT By Els De Wilde August 2005 Chair: Chris Janiszewski Cochair: Alan D. J. Cooke Major Department: Marketing Normatively, the importance of an attribute does not depend on the number of levels on which it is defined, exterior levels being fixed. This dissertation examines a bias identified in the conjoint literature known as the number-of-levels effect, which occurs when consumers derived importance weights increase with the number of attribute levels. Borrowing from research on attention during category learning, this dissertation examines psychological mechanisms that underlie the number-of-levels effect. In contrast to earlier attentional accounts that have been offered for the bias, I propose that the weighting effect found in the number-of-levels studies is due to the attentional processes that are inherent to the sequential evaluation procedure. Analogously to attentional processes during category learning, consumers divide their attention between attribute dimensions and attribute levels when making sequential evaluations. Within a given trial, a mechanism of attentional contrast directs attention ix

PAGE 10

away from attribute levels that have been associated with an outcome. I suggest that the number-of-levels effect is caused by this attentional contrast strategy. When attributes differ on the number of levels on which they are defined, they also differ on the number of times a given attribute level has been associated with a judgment (i.e., attribute level novelty). A process of attentional contrast away from the more familiar attribute level within each offer results in differential importance weights for the attributes. In a series of experiments, I examine this attentional contrast account for the number-of-levels effect. Consistent with this account, the weighting effect is shown to appear when attentional contrast is created in a situation in which there is no difference in number of levels. Moreover, the number-of-levels effect reverses when attentional contrast is created on the attribute with the lower number of levels. x

PAGE 11

CHAPTER 1 INTRODUCTION Normatively, the importance of an attribute should not depend on the number of levels on which it is defined, exterior levels being fixed. Yet, several studies from the conjoint literature show that derived importance weights increase with the number of attribute levels, a bias which has generally been labelled the number-of-levels effect (e.g., Currim, Weinberg and Wittink 1981; Verlegh, Schifferstein and Wittink 2002; Wittink et al. 1992b). In searching for a method to eliminate this bias, several accounts have been offered for it. Methodologically, there is evidence that the number-of-levels effect is related to the data collection method (e.g., Wittink et al. 1992b), the measurement scale for the dependent variable (e.g., Steenkamp and Wittink 1994; Wittink et al. 1992b), and the parameter estimation procedure (Wittink, Krishnamurthi and Reibstein 1989). Although tests of these methodological accounts reduced the number-of-levels effect, they have never completely eliminated the bias. The possibility that psychological processes are at least partly responsible for the number-of-levels effect has been acknowledged since the effect was first documented. Currim et al. (1981) state that the respondent may react to the number of levels used for a given attribute by consciously or unconsciously weighting an attribute more heavily as the number of levels increases (p. 72). Although a test of this account has been included in several studies, it has not received much support during the past two decades. An assumption underlying tests of this attentional account is that there should be a number-of-levels effect not only in sequential judgments or choices (e.g., full-profile rank orders, 1

PAGE 12

2 full-profile paired comparisons, full-profile ratings), but also in self-explicated priors (Wittink et al. 1992a) and stated importance measures (Steenkamp and Wittink 1994; Verlegh et al. 2002). However, I propose that the weighting effect found in the number-of-levels studies may be due to the attentional processes that are inherent to the sequential evaluation procedure. Studies in which people learn new categories by making sequential classifications distinguish between attention allocated to attributes and to attribute levels (Kersten, Goldstone and Schaffert 1998). They show that within a given trial, a mechanism of attentional contrast directs attention away from attribute levels that have been associated with an outcome. I propose that this attentional contrast strategy can at least partly account for the number-of-levels effect. When attributes differ on the number of levels on which they are defined, they also differ on the number of times a given attribute level appears across offers during a sequential judgment task (henceforth referred to as relative attribute level novelty). More specifically, the levels of attributes that are defined on more levels are overall relatively more novel than the levels of attributes that are defined on fewer levels. A process of attentional contrast away from the less novel attribute level within each offer may result in differential importance weights for the attributes. This dissertation proceeds as follows: Chapter 2 discusses the number-of-levels effect and the methodological and psychological accounts that have been offered for it in the literature. It also introduces the attentional contrast account from the category learning literature. In chapters 3 to 5, I examine if attentional contrast can account for the number-of-levels effect. In chapter 3, I extend the designs that have typically been used

PAGE 13

3 in studies that examine the number-of-levels effect and find evidence for an account in terms of attentional contrast. In chapter 4, I manipulate attentional contrast directly to examine its possible impact when there is no difference in number of levels. In chapter 5, I create attentional contrast on the attribute with the lower number of levels. Finally, chapter 6 discusses the findings and chapter 7 examines the limitations and avenues for future research.

PAGE 14

CHAPTER 2 THEORETICAL BACKGROUND The Number-Of-Levels Effect When an attribute has a higher number of levels (exterior values being fixed), derived importance weights for that attribute increase. For example, Currim et al. (1981) examined subscription series to performing arts events, using three attributes that were defined on two levels and three attributes that were defined on three levels. They asked participants to make trade-offs between pairs that were defined on two attributes: a two-level attribute and a three-level attribute. The two-level attributes had derived importance weights between 0.36 and 0.45, whereas the three-level attributes had derived importance weights between 0.55 and 0.66. Originally identified between attributes using part-worth estimation within conjoint analysis, this number-of-levels effect has been replicated using a variety of data collection methods, measurement scales for the dependent variable, and estimation methods. Accounts for the number-of-levels effect include methodological and psychological factors. I will first review the existing accounts for the number-of-levels effect and then propose a new account based on attentional contrast from the category learning literature. Existing Accounts of the Number-Of-Levels Effect The greater share of research on the number-of-levels effect has dealt with methodological accounts. This research attributes the number-of-levels effect to the 4

PAGE 15

5 properties of the measurement scale, the usage of the measurement scale, or the method of parameter estimation. Properties of the Measurement Scale Several accounts of the number-of-levels effect relate to the properties of the measurement scale. Wittink, Krishnamurthi, and Reibstein (1989) hypothesized that the number-of-levels effect was partly due to the use of a rank-order measurement procedure. They showed analytically why the effect occurs for rank-order preferences using least-squares analysis on a multitude of full factorial and fractional factorial designs. Since it is impossible to analytically show the same effect exists for ratings, Wittink et al. (1989) examined whether the number-of-levels effect exists for rating scales empirically, comparing the results from full-profile ratings with the results from rank-order preferences obtained by paired comparisons and trade-off matrix rank orders. The full-profile preferences were obtained using a ten-point scale. Wittink et al. (1989) compared derived importances of a three-level and five-level price attribute in five categories and found that the number-of-levels effect was equally strong using rank-order or rating scale measurement procedures, the implication being that rank-order measurement was not the sole source of the number-of-levels effect. Steenkamp and Wittink (1994) proposed that the number-of-levels effect occurs with a rating scale measurement procedure because the preference ratings lack metricity. They hypothesized that rating scales properties are closer to ordinal than to interval measurement. Since the method of magnitude estimation has been claimed to result in interval-scale data, Steenkamp and Wittink (1994) used this method to test their hypothesis. The method of magnitude estimation consists of asking participants to match their evaluation of a profile to the perceived magnitude expressed on a physical medium.

PAGE 16

6 Steenkamp and Wittink (1994) used numerical estimation (i.e., participants assign numbers to profiles) and line production (i.e., participants draw lines) as response modalities. They found that a nine-point rating scale and a magnitude estimation procedure provide number-of-levels effects of comparable size. However, the magnitude estimation procedure allows for a test of the metricity of the results of each respondent separately. Categorizing the respondents as metric versus non-metric, Steenkamp and Wittink (1994) found a reduced number-of-levels effect in participants whose responses satisfied the criteria for metricity. In other words, the lack of metricity of the rating method per se had no impact on the number-of-levels effect, but individual differences in metricity were identified as a source of the effect. Usage of the Measurement Scale Other measurement procedure explanations relate to respondents use of the measurement scale. Wittink et al. (1992b) included a balancing manipulation in their study (i.e., they constructed a condition in which dominated pairs were eliminated from the procedure), provided respondents made pair-wise comparisons among alternatives. Using adaptive conjoint analysis, Wittink et al. (1992b) gathered full factorial data on one product category (notebook computers) defined on six attributes, of which four attributes alternated between two and four levels, depending on the condition. Wittink et al. (1992b) found that the number-of-levels effect is considerably reduced when dominated pairs are eliminated from the procedure. Wittink et al. (1992b) explain this result in terms of participants tendencies to give preference ratings toward the midpoint of the scale. Since preference ratings for dominated pairs are expected to be more extreme than those for non-dominated pairs, the exclusion of dominated pairs provides less opportunity for distortion.

PAGE 17

7 Verlegh et al. (2002) propose the number-of-levels effect may be a consequence of a persons tendency to perceive a constant, absolute distance between levels of an attribute (i.e., exhibit a uniform distribution tendency). They tested this hypothesis by comparing results on five different measures of attribute importance, using one product category (color TVs). The product category was defined on six attributes, of which two alternated between two and four levels, depending on the condition. Verlegh et al. (2002) found significant number-of-level effects on two measures of attribute importance: difference in part worths in a full-profile conjoint and attractiveness ratings for individual attribute levels. On the remaining three measures (stated importance of a set of attributes, linear regression coefficients, and paired comparisons varying one attributes exterior levels at a time), no significant number-of-levels effect was observed. Verlegh et al. (2002) interpreted their results as evidence for respondents tendency to distribute stimulus representations uniformly over a restricted internal continuum (i.e., the distance between each level of an attribute was constant) in full-profile and attribute level ratings. Ceteris paribus, the addition of intermediate levels will lead respondents to distribute the exterior attribute levels more toward the extremes of the internal continuum. Method of Parameter Estimation A second potential source of the number-of-levels effect is the method of parameter estimation. In their seminal paper, Currim, Weinberg, and Wittink (1981) mentioned the method of parameter estimation as a possible source of the number-of-levels effect. For the 2x3 trade-off matrix they used to obtain their empirical results, they estimated the parameters based on permissible configurations of preferences using ordinary least squares. They derived the importance weights for two extreme cases: a dominant two-level attribute and a dominant three-level attribute. They found that the two-level

PAGE 18

8 attributes importance weights varied between 0.20 (i.e., when the three-attribute level attribute was dominant) and 0.60 (i.e., when the two-level attribute was dominant). Analogously, the three-level attributes importance weights ranged from 0.40 (i.e., when the two-level attribute was dominant) to 0.80 (i.e., when the three-level attribute was dominant). In other words, while the range of possible derived importance weights was not affected by the number of levels, its size was affected by it. Currim et al. (1981) used this analytical evidence as support for their hypothesis that the method of parameter estimation is a source of the number-of-levels effect. Subsequent studies have provided additional support for the method of parameter estimation account of the number-of-levels effect. Wittink et al. (1989) showed analytically why the number-of-levels effect occurs on rank-order preferences on a multitude of full factorial and fractional factorial designs. Assuming an additive, main-effect model, they calculated raw attribute importance weights for any number of attributes and any attribute levels using least-squares analysis. Based on the raw attribute importance weights, they estimated maximum relative weights for each attribute, and the differences and ratios between the attributes within a design. They found that maximum relative weights increase with the number of levels and concluded that the method of parameter estimation is a source of the number-of-levels effect. Wittink, Krishnamurthi, and Nutter (1982) gathered participants preference data for hypothetical summer jobs, defined on four attributes, of which two attributes alternated between two and four attribute levels, depending on the condition. Two data collection methods were used: a fractional factorial profile evaluation and a trade-off matrix. Wittink et al. (1982) estimated the parameters using a non-metric method (LINMAP, Srinivasan and Shocker

PAGE 19

9 1973) and found that the derived importance weights vary with the number of levels on which an attribute is defined. They concluded that the number-of-levels effect is at least partially accounted for by the method of parameter estimation. Perceptual Processes Since the initial identification of the number-of-levels effect, researchers have acknowledged that the effect could be a consequence of perceptual processes (Currim et al. 1981). More specifically, it has been proposed that respondents may consciously or unconsciously assign more weight to attributes with more levels. Several studies have included one or more tests for this weighting explanation, but have failed to find support for the hypothesis (Steenkamp and Wittink 1994; Verlegh et al. 2002; Wittink et al. 1992b). For example, Verlegh et al. (2002) hypothesized that increased attention should affect the stated importance of a set of attributes and paired comparisons varying one attributes exterior levels at a time. However, no number-of-levels effect was observed on these two measures. Attentional Contrast Several of the proposed accounts of the number-of-levels effect can partially explain the effect. However, no single explanation can wholely account for the number-of-levels effect. I propose that attentional contrast may be an additional source of the number-of-levels effect. The proposed attentional contrast explanation is based on two observations. First, the occurrence of the number-of-levels effect depends on the preference elicitation method. Decompositional methods ask respondents to provide multiattribute judgments for the full-profile descriptions of alternatives and then derive individual attribute level values from the judgments. The number-of-levels effect is typically found using

PAGE 20

10 decompositional methods. Compositional methods ask respondents to assign values to each level of an attribute. The values are combined to construct an overall evaluation of an alternative. The number-of-levels effect is typically not found using compositional methods. There are three studies that have used a combination of decompositional and compositional methods. Wittink et al. (1992a) showed that the magnitude of the number-of-levels effect was greater for a full-profile procedure than for an adaptive conjoint analysis. Whereas full-profile procedure is an entirely decompositional method, adaptive conjoint analysis is a combination of a decompositional and compositional procedure. No number-of-levels effect was observed in the compositional data. Verlegh et al. (2002) used a decompositional and compositional procedure and found evidence for a number-of-levels effect in the compositional data. Yet, the compositional data were collected after the decompositional data, suggesting the decompositional procedure may have biased the values provided in the compositional procedure. Steenkamp and Wittink (1994) also found minimal evidence for a number-of-levels effect on stated attribute importances that were gathered after the decompositional data. Johnson (1992) found evidence for a number-of-levels effect using only a compositional method. He asked participants to assess attribute values directly. However, the values of two attributes were assessed at the same time. More specifically, participants were told that a base TV with 17 inch screen and monophonic sound was worth $200. Then, they were asked to evaluate what other options were worth: 19 inch screen, 21 inch screen and excellent stereo sound in the 3+2 condition and 21 inch screen, good stereo sound and excellent stereo sound in the 2+3 condition. Hence, participants

PAGE 21

11 could interpret this task as full-profile evaluation. For example, when evaluating the 19 inch screen, it is likely that they imagined a monophonic sound TV. Second, the designs used in studies on the number-of-levels effect invariably compare conditions in which the number of levels of more than one attribute is varied at once (i.e., varying the number of levels of attribute 1 is confounded with varying the number of attribute levels of attribute 2). For example, consider an experiment in which participants respond to a 2x4 design (i.e., attribute 1 has two levels and attribute 2 has four levels) or a 4x2 design. If one finds that the derived importance weight on attribute 1 is larger in the 4x2 condition than the 2x4 condition, then the weighting difference may be due to the increased number of levels in attribute 1 or to the accompanying decrease in the number of levels in attribute 2. In other words, the weighting effect that has been interpreted as a number-of-levels effect may not (solely) be caused by the increased number of levels in that attribute. It may also be caused by the reduced number of levels of other attributes. It is important to note that all published demonstrations of the number-of-levels effect have this characteristic built into their manipulations. An overview of the designs can be found in Table 1. Based on these two observations, I will now propose a new account for the number-of-levels effect: the attentional contrast account. The attentional contrast account depends on the idea that people direct attention away from attribute levels they have seen more often and towards attribute levels they have seen less often. For example, in a 4x2 design, the first attribute has four levels and each of these levels will be experienced twice in a full-factorial design. The second attribute has two levels and each of these levels will be experienced four times in a full-factorial design. Thus, the levels of the first attribute will

PAGE 22

12 be relatively more novel than the levels of the second attribute. The attentional contrast account proposed that attention will contrast away from the more common levels (i.e., attribute 2) and toward the less common levels (i.e., attribute 1). Thus, the attribute with more levels will be perceived as more important. The attentional contrast hypothesis has its foundation in the categorical learning literature. The most successful theories in categorization and learning incorporate some notion of selective attention (Nosofsky 1986, 1991; Sutherland and Mackintosh 1971). More specifically, consumers increasingly pay attention to attributes and attribute levels that are important, and decreasingly pay attention to irrelevant attributes, irrelevant attribute levels, or both (Goldstone and Steyvers 2001). For example, Goldstone and Steyvers (2001) showed that selective attention can be directed toward arbitrary dimensions created by randomly pairing faces. They selected four photographs of bald heads and generated two dimensions by morphing between the photographs. Each of the sixteen created faces varied in the proportion of each of the four faces. Based on category relevance, both sensitization of relevant dimensions and desensitization of irrelevant dimensions were observed. Moreover, Goldstone and Steyvers (2001) showed that once a participant had learned to direct attention toward or away from a dimension, this learning extended to different stimuli. Kersten, Goldstone, and Schaffert (1998) provide additional evidence that differential selective attention mechanisms operate on attributes and attribute levels. They show that within a given trial, a mechanism of attentional contrast directs attention away from attribute levels that have been associated with an outcome. In a series of experiments, participants learned verbs that were defined on two attributes: path and

PAGE 23

13 manner. Each verb was always accompanied by a particular value of one or both of these attributes. One group of participants learned conjunctive verbs (i.e., verbs that could be distinguished on path and manner). Another group first learned path verbs (i.e., verbs that could only be distinguished on the basis of path). When this second group learned new conjunctive verbs (i.e., verbs that could be distinguished based on manner as well as the same values of path), they attended more to level of manner than the participants who first learned conjunctive verbs. In other words, since the path values had been associated with an outcome (i.e., a verb), attentional contrast was directed toward the new manner values within subsequent trials. The attentional contrast account is consistent with the observation that the occurrence of the number-of-levels effect depends on whether the method is compositional or decompositional. In a compositional method, each level of an attribute is repeated only once and it is repeated out of context, so no attentional contrast can operate. In a decompositional method, attribute levels are repeated and this repetition occurs in the context of other attribute levels that have been more or less repeated, so attentional contrast can operate. The attentional contrast account also makes specific predictions in the event the levels of attributes are unconfounded in an experimental design. For example, imagine three designs are investigated: a 4x2, a 2x4, and a 4x4. The derived importance weight of attribute one (two) should be greater (smaller) in the 4x2 than the 2x4 studies, as has been observed in prior studies. However, the attentional contrast account predicts that the derived importance weight of attribute one should be greater in the 4x2 design than the 4x4 design because the number of levels of attribute two are more frequent in the 4x2

PAGE 24

14 condition. The attentional contrast account also predicts that the derived importance weight of attribute two should be equal in the 4x2 and the 4x4 conditions because the number of levels of attribute one are equivalent. Similar predictions can be made when comparing the 2x4 and the 4x4 conditions. The predictions of the attentional contrast hypothesis are not consistent with the predictions that might be expected if one were simply to count the levels of an attribute. For example, if the predictions were based solely on the number of levels of an attribute, the derived importance weight of attribute one should be the same in the 4x2 condition and the 4x4 condition because the number of levels of attribute one are equivalent. The derived importance weight of attribute two should be less in the 4x2 condition than the 4x4 condition because the number of levels of attribute two is less in the 4x2 condition. Thus, an explanation based solely on attribute levels makes a set of qualitatively different predictions than the attentional contrast hypothesis. In a series of four experiments, I will test this attentional contrast account for the number-of-levels effect. In experiments 1 and 1A, I will extend the designs that have typically been used to examine the number-of-levels effect to test an attentional contrast account. In experiment 2, I will manipulate attentional contrast directly to examine its potential impact when there are no differences in number of levels. In experiment 3, I create attentional contrast on an attribute with a lower number of levels.

PAGE 25

15 Table 1. Number-of-levels manipulations used in published studies Authors Number-Of-Levels Manipulation Currim, Weinberg and Wittink (1981) number-of-levels manipulation between attributes (2 or 3 levels) Johnson (1992) 3+2 vs. 2+3 (used lists) Orme (1998) number-of-levels manipulation between attributes (2 to 5 levels) Steenkamp and Wittink (1994) 2x4 vs. 4x2 Verlegh, Schifferstein and Wittink (2002) 4x2x2x2x3x2 vs. 2x4x2x2x3x2 Wittink, Huber, Fiedler and Miller (1992a) 3x4x2x2x4x2x2x2x2 vs. 3x2x4x4x2x2x2x2x2 Wittink, Huber, Zandan and Johnson (1992b) 2x2x4x4 vs. 4x4x2x2 Wittink, Krishnamurthi and Nutter (1982) 4x4x2x2 vs. 4x2x4x2 Wittink, Krishnamurthi and Reibstein (1989) 3 vs. 5 levels

PAGE 26

CHAPTER 3 EXPERIMENT 1: DEMONSTRATION OF THE NUMBER-OF-LEVELS EFFECT WITH EXPANDED DESIGN The objective of the first experiment was to test the attentional contrast account of the number-of-levels effect by expanding the experimental design that has typically been used to examine the bias. More specifically, the number of levels of only one attribute was varied across conditions. The attentional contrast hypothesis (Kersten et al. 1998), predicted that the number-of-levels effect would be sensitive to the relative novelty of each attribute level. Method Design The design was a 3 cell between-subjects design (4x2, 2x4, and 4x4 attribute levels) with three between-subject product replicates (refrigerators, LCD televisions, and digital cameras). Each offer was defined on two attributes. Participants were asked to provide their reservation price for each product offer. This dependent variable should not be sensitive to the uniform distribution problem associated with a fixed scale (Verlegh et al. 2002) or suffer from a lack of metricity (Steenkamp and Wittink 1994). The design of experiment 1 is presented in Figure 1. Stimuli The product categories were refrigerators, LCD televisions, and digital cameras. Each product offer was defined on two attributes for which participants were known to have monotonic preferences. The attributes were the following: annual energy cost and 16

PAGE 27

17 capacity (refrigerators), screen size and horizontal viewing angle (LCD televisions), imaging sensor resolution and image storage capacity (digital cameras). Each attribute had two or four possible attribute levels (see Table 2 for an overview of the attribute levels). These attribute levels were either exterior (i.e., the two values in a 2-level attribute or the two extreme values in a 4-level attribute) or interior (the two inserted levels in a 4-level attribute). Exterior offers consisted of the exterior levels of the two attributes (e.g., [1,1], [1,4], [2,1], [2,4] in a 2x4 design). Intermediate offers consisted of the interior levels of the two attributes (e.g., [2,2], [2,3], [3,2], [3,3] in a 4x4 design), or of an exterior level of one attribute and an interior level of a second attribute (e.g., [1,2], [1,3], [2,2], [2,3] in a 2x4 design). Procedure Participants were seated at personal computers. Participants read instructions stating that they would be stating reservation prices for products described on two dimensions. Participants were asked to assume that the products were identical on all non-defined attributes and were informed of the average price of the product. Then, the product offers were presented. First, the participants stated their reservation price for the intermediate offers. These four stimuli were presented in random order. Next, participants stated their reservation price for the exterior offers. Again, the stimuli were presented in random order. Thus, each participant provided reservation price judgments for eight stimuli with the last four being the identical exterior stimuli. Results The analysis was conducted using the mean difference scores of the exterior offers. For example, attribute ones differences score in a 2x4 design is the difference between the average of [1,1] and [1,4] and the average of [2,1] and [2,4]. Similarly, attribute twos

PAGE 28

18 differences score in a 2x4 design is the difference between the average of [1,1] and [2,1] and the average of [1,4] and [2,4]. The four exterior offers were identical across conditions, so these difference scores could be compared. The results of participants whose ratings on the exterior offers were non-monotonic were removed from the dataset. Table 3 presents the mean difference scores per attribute for each condition. Refrigerators Data of 91 participants were retained for analysis. The analysis for each attribute was performed separately. The first analysis was on attribute one (annual energy cost). The mean difference scores of the 2x4 condition (X = $78.97) and the 4x2 condition (X = 144.24) were significantly different (F(1, 88) = 10.38, p < .01). This replicates the number-of-levels effect using a design that is common in the academic literature. The mean difference scores of the 2x4 condition (X = $78.97) and the 4x4 condition (X = $63.84) were not significantly different (F(1,88) = 0.63, p > .05). This result is inconsistent with the hypothesis that the number-of-levels effect is sensitive only to the number of levels of an attribute. The mean difference scores of the 4x2 condition (X = $144.22) and the 4x4 condition (X = $63.84) were significantly different (F(1, 88) = 17.16, p < .01). This result is consistent with the attentional contrast hypothesis. The second analysis was on attribute two (capacity). The mean difference scores of the 2x4 condition (X = $122.55) and the 4x2 condition (X = $120.17) were not significantly different (F(1, 88) = 0.01, p > .05). The mean difference scores of the 2x4 condition (X = $122.55) and the 4x4 condition (X = $100.93) were not significantly different (F(1, 88) = 0.96, p > .05). The mean difference scores of the 4x2 condition (X = $120.17) and the 4x4 condition (X = $100.93) were not significantly different (F(1, 88) = 0.73, p > .05)

PAGE 29

19 LCD Televisions Data of 90 participants were retained for analysis. The first analysis was on attribute one (screen size). The mean difference scores of the 2x4 condition (X = $176.20) and the 4x2 condition (X = $257.37) were significantly different (F(1, 87) = 3.98, p < .05). The mean difference scores of the 2x4 condition (X = $176.20) and the 4x4 condition (X = $196.82) were not significantly different (F(1, 87) = 0.26, p > .05). The mean difference scores of the 4x2 condition (X = $257.37) and the 4x4 condition (X = $196.82) were not significantly different (F(1, 87) = 2.21, p > .05). These results are not supportive of the number of levels hypothesis or the attentional contrast hypothesis. The second analysis was on attribute two (horizontal viewing angle). The mean difference scores of the 2x4 condition (X = $140.64) and the 4x2 condition (X = $130.37) were not significantly different (F(1, 87) = 0.26, p > .05). The mean difference scores of the 2x4 condition (X = $140.64) and the 4x4 condition (X = 122.82) were not significantly different (F(1, 87) = 0.78, p > .05). The mean difference scores of the 4x2 condition (X = $130.37) and the 4x4 condition (X = $122.82) were not significantly different (F(1, 87) = 0.14, p > .05). Digital Cameras Data of 117 participants were retained for analysis. The first analysis was on attribute one (imaging sensor resolution). The mean difference scores of the 2x4 condition (X = $134.90) and the 4x2 condition (X = $169.31) were significantly different (F(1, 114) = 3.18, p = .08). The mean difference scores of the 2x4 condition (X = $134.90) and the 4x4 condition (X = $131.49) were not significantly different (F(1, 114) = 0.03, p > .05). The mean difference scores of the 4x2 condition (X = $169.31) and the

PAGE 30

20 4x4 condition (X = $131.49) were significantly different (F(1, 114) = 4.07, p < .05). These results are supportive of the attentional contrast hypothesis. The second analysis was on attribute two (image storage capacity). The mean difference scores of the 2x4 condition (X = $110.42) and the 4x2 condition (X = $106.81) were not significantly different (F(1, 114) = 0.07, p > .05). The mean difference scores of the 2x4 condition (X = $110.42) and the 4x4 condition (X = $95.86) were not significantly different (F(1, 114) = 1.15, p > .05). The mean difference scores of the 4x2 condition (X = $106.81) and the 4x4 condition (X = $95.86) were not significantly different (F(1, 114) = 0.63, p > .05). Discussion The data from experiment 1 replicate the number-of-levels effect that has been reported in the literature and provide some new insights as to the potential source of this effect. First, all three replicates showed a number-of-levels effect on attribute one when the analysis was restricted to the 2x4 and 4x2 conditions. This type of design is common in the number-of-levels literature. Second, when the 4x4 condition was also included in the analysis, the mean difference score for attribute one was similar to the difference score in the 2x4 condition, but less than the difference score in the 4x2 conditions. Although this pattern of results held for only the refrigerator and digital camera replicates, and only for attribute one, the result is consistent with the attentional contrast hypothesis. The most perplexing result in experiment 1 is the apparent attentional contrast effect on attribute one and the lack of the attentional contrast effect on attribute 2. One possible reason for this finding is that attentional contrast only operates on attributes that are perceived to be important (i.e., worth attending to). In other words, an attribute must

PAGE 31

21 be perceived to be important if its attribute levels are to exhibit an attentional contrast based number-of-levels effect. The implication is that attribute two could also exhibit a number-of-levels effect if it were to be perceived as more important. Experiment 1A will examine this hypothesis. Experiment 1A The goal of experiment 1A was to assess whether the differential effects observed on attributes one and two in experiment 1 could be a consequence of a (perceived) difference in importance between the two attributes. In Experiment 1A, the (relative) importance of the second attribute was increased in two ways. First, the range of the attribute twos levels was increased and the range of attribute ones levels was decreased. Second, participants were asked to elaborate on product usage. This second strategy was adopted from Huffman (1997), who showed that elaboration on product usage increases the weight of usage-related attributes whereas elaboration on product purchase increases the weight of purchase-related attributes. Of the three replicates, only the refrigerator category had a purchase-related attribute (annual energy cost) and a usage-related attribute (capacity). Given the failure to find a number-of-levels effect on the usage-related attribute in this category, participants were asked to elaborate on usage. Method The design was identical to experiment 1 with the limitation that only one of the replicates (refrigerators) was investigated. The stimulus were adjusted so that the range of attribute two (capacity) was doubled and the range of attribute one (annual energy cost) was reduced. The attribute levels for capacity were 13, 17, 21, and 25 cubic feet. The attribute levels for annual energy cost were $90, $85, $80, and $75.

PAGE 32

22 The procedure was identical to experiment 1 except for one change. Prior to being exposed to the eight product offers, participants were asked to elaborate on product usage. Participants were told to Think back to the times you have used a refrigerator. Picture, in your minds eye, the situations in which you have used it. Below, please describe your thoughts and feelings about using the refrigerator. What was important to you when you were using the refrigerator? Results Table 4 presents the mean difference scores per attribute for each condition. Data of 120 participants were retained for analysis. The first analysis was on attribute one (annual energy cost). T the mean difference scores of the 2x4 condition (X = $74.25) and the 4x2 condition (X = $98.37) were significantly different (F(1, 117) = 4.09, p < .05). The mean difference scores of the 2x4 condition (X = $74.25) and the 4x4 condition (X = $71.30) were not significantly different (F(1, 117) = 0.05, p > .05). The mean difference scores of the 4x2 condition (X = $98.37) and the 4x4 condition (X = $71.30) were significantly different (F(1, 117) = 5.70, p < .05). These results are supportive of the attentional contrast hypothesis. The second analysis was on attribute two (capacity). The mean difference scores of the 2x4 condition (X = $265.13) and the 4x2 condition (X = $180.93) were significantly different (F(1, 117) = 7.10, p < .05). The mean difference scores of the 4x2 condition (X = $180.93) and the 4x4 condition (X = $220.72) were not significantly different (F(1, 117) = 1.75, p > .05). The mean difference scores of the 2x4 condition (X = $265.13) and the 4x4 condition (X = $220.72) were not significantly different (F(1, 117) = 1.76, p > .05). These results are supportive of the attentional contrast hypothesis.

PAGE 33

23 Discussion Experiment 1A shows that a number-of-levels effect can be produced on both attributes in a two-factor conjoint design, provided the participant perceives each attribute as important. When the importance of a second attribute was increased by increasing its range and by getting consumers to elaborate on situations in which the attribute might be important, a number-of-levels effect was obtained. In addition, the mean difference score for attribute one in the 4x4 condition was similar to the difference score in the 2x4 condition, but less than the difference score in the 4x2 conditions. Similarly, the mean difference score for attribute two in the 4x4 condition was similar to the difference score in the 2x4 condition, but less than the difference score in the 2x4 conditions. These results are consistent with the attentional contrast hypothesis. Experiments 1 and 1A show that the number-of-levels effect does not occur because of increasing levels on the focal attribute, but because of the decreasing levels on the accompanying attribute. Moreover, the effect depends on the (perceived) importance of an attribute. These results are consistent with an attentional contrast account, which predicts that within subsequent offers, attentional contrast is produced on the attribute that has more levels, as its levels are relatively more novel than the attribute levels of the attribute with the lower number of levels. However, experiments 1 and 1A were designed to manipulate the number of levels and are, therefore, not the most direct test of the attentional contrast account. While the conditions that were compared differed on attentional contrast on the focal attributes, they also differed on the number of levels of the accompanying attribute. Experiment 2 provides a more direct test of the attentional contrast account by showing that it can

PAGE 34

24 create a weighting difference while the number of levels on both focal and accompanying attributes is kept constant.

PAGE 35

25 A Attribute 2 1 2 3 4 1 2 3 Attribute 1 4 B Attribute 2 1 2 3 4 1 2 3 Attribute 1 4 C Attribute 2 1 2 3 4 1 2 3 Attribute 1 4 Note.Exterior offers are colored in black, intermediate offers are shaded in grey, and interior offers are hatched. Figure 1. Design for experiments 1 and 1A. A) 4x2 Condition. B) 4x2 Condition. C) 4x4 Condition

PAGE 36

26 Table 2. Attribute levels from experiment 1 Attribute level Product category Attribute 1 2 3 4 Refrigerators annual energy cost ($n) 100 90 80 70 capacity (n cubic feet) 16 18 20 22 LCD televisions screen size (n inches) 20 24 28 32 horizontal viewing angle (n off-center) 45 55 65 75 Digital cameras imaging sensor resolution (n MegaPixels) 3.2 4.4 5.6 6.8 image storage capacity (up to n MB) 128 256 384 512 Table 3. Mean difference scores and simple effects for experiment 1 Mean difference score F values Attribute 1 4x2 2x4 4x4 4x2 vs 2x4 4x2 vs 4x4 2x4 vs 4x4 Annual energy cost (Refrigerators) 144.24 78.97 63.84 10.38 a 17.16 a 0.63 Screen size (LCD televisions) 257.37 176.20 196.82 3.98 b 2.21 0.26 Imaging sensor resolution (Digital cameras) 169.31 134.90 131.49 3.18 c 4.07 b 0.03 Attribute 2 4x2 2x4 4x4 4x2 vs 2x4 4x2 vs 4x4 2x4 vs 4x4 Capacity (Refrigerators) 120.17 122.55 100.93 0.01 0.73 0.96 Horizontal viewing angle (LCD televisions) 130.37 140.64 122.82 0.26 0.14 0.78 Image storage capacity (Digital cameras) 106.81 110.42 95.86 0.07 0.63 1.15 a implies significance at 99% level of confidence b implies significance at 95% level of confidence c implies significance at 90% level of confidence

PAGE 37

27 Table 4. Mean difference scores and simple effects for experiment 1A Mean difference score F values Attribute 1 4x2 2x4 4x4 4x2 vs 2x4 4x2 vs 4x4 2x4 vs 4x4 Annual energy cost (Refrigerators) 98.37 74.25 71.30 4.09 b 5.70 b 0.05 Attribute 2 4x2 2x4 4x4 4x2 vs 2x4 4x2 vs 4x4 2x4 vs 4x4 Capacity (Refrigerators) 180.93 265.13 220.72 7.10 a 1.75 1.76 a implies significance at 99% level of confidence b implies significance at 95% level of confidence

PAGE 38

CHAPTER 4 EXPERIMENT 2: TEST OF ATTENTIONAL CONTRAST ACCOUNT WITH NO DIFFERENCE IN NUMBER OF LEVELS The goal of experiment 2 was to examine if attentional contrast can create a weighting effect when there is no difference in number of levels. In experiments 1 and 1A, we found evidence for an attentional contrast account by varying the number of levels of only one attribute at a time. However, attentional contrast was confounded with the number of levels manipulation. By manipulating attentional contrast in a context in which there is no difference in number of levels, we want to obtain more direct evidence that the attentional contrast by itself can account for the weighting difference. In addition, the three replicates and procedure from experiment 1 were used in the experiment. It was felt that demonstrating a robust attentional contrast effect on one attribute was superior to demonstrating an attentional contrast effect on two attributes of a single replicate. Method Design and Stimuli The design was a 3 cell between-subjects design (contrast on attribute 1, contrast on attribute 2, balanced contrast) with three between-subject replicates (refrigerator, LCD televisions, digital cameras). In all three experimental conditions, each attribute was defined on 4 levels (see Figure 2). The cells differed on the attribute on which attentional contrast was created. In the contrast on attribute 1 condition, attentional contrast was created on attribute 1. This contrast was created via the selection of offers that were initially rated by the participant. The intermediate offers were [2,1], [2,2], [2,3], [2,4], 28

PAGE 39

29 [3,1], [3,2], [3,3], and [3,4]. Thus, when the participant saw the exterior offers of [1,1], [1,4], [4,1], and [4,4], the values of the first attribute were novel and the values of the second attribute were familiar. In the contrast on attribute 2 condition, the intermediate offers were [1,2], [1,3], [2,2], [2,3], [3,2], [3,3], [4,2],and[4,3]. Thus, when the participant saw the exterior offers of [1,1], [1,4], [4,1], and [4,4], the values of the second attribute were novel and the values of the first attribute were familiar. In the balanced contrast condition, the intermediate offers were [2,2], [2,3], [3,2], and [3,3]. Thus, when the participant saw the exterior offers of [1,1], [1,4], [4,1], and [4,4], the values of both attributes were on average equally novel, hence there was no contrast. Note that this last condition is particularly important because it will show that the number-of-levels effect is not an issue of novelty (i.e., seeing levels that have not been seen before), but an issue of attentional contrast (i.e., seeing levels that are novel relative to the levels seen on the other attribute). In all other respects, the procedure was identical to the one used in experiment 1. Predictions The attentional contrast hypothesis makes predictions for attributes one and two. With respect to attribute one, the attentional contrast hypothesis predicts that the mean difference score should be greater in the contrast on attribute 1 condition than in the contrast on attribute 2 condition or the balanced contrast condition. With respect to attribute two, the attentional contrast hypothesis also predicts that the mean difference score should be greater in the contrast on attribute 2 condition than in the contrast on attribute 1 condition or the balanced contrast condition. Given the results of experiments 1 and 1A, it was anticipated that the attentional contrast effect would be obtained on attribute 1, but not attribute 2.

PAGE 40

30 Results Table 5 presents the mean difference scores per attribute in each condition. Refrigerators Data of 104 participants were retained for analysis. The analysis for each attribute was performed separately. The first analysis was on attribute one (annual energy cost). The mean difference scores of the contrast on attribute 1 condition (X = $99.37) and the contrast on attribute 2 condition (X = $68.60) were significantly different (F(1, 101) = 4.96, p < .05). The mean difference scores of the contrast on attribute 1 condition (X = $99.37) and the balanced contrast condition (X = $70.04) were significantly different (F(1, 101) = 4.87, p < .05). The second analysis was on attribute two (capacity). The mean difference scores of the contrast on attribute 2 condition (X = $94.02) and the contrast on attribute 1 condition (X = $126.58) were not significantly different (F(1, 101) = 2.65, p > .05). The mean difference scores of the contrast on attribute 2 condition (X = $94.02) and the balanced contrast condition (X = $99.79) were not significantly different (F(1, 101) = 0.08, p > .05). LCD televisions Data of 111 participants were retained for analysis. The first analysis was on attribute one (screen size). The mean difference scores of the contrast on attribute 1 condition (X = $286.04) and the contrast on attribute 2 condition (X = $228.72) were not significantly different (F(1, 108) = 2.13, p > .05). The mean difference scores of the contrast on attribute 1 condition (X = $286.04) and the balanced contrast condition (X = $192.99) were significantly different (F(1, 108) = 5.27, p < .05). The second analysis was on attribute two (horizontal viewing angle). The mean difference scores of the contrast on

PAGE 41

31 attribute 2 condition (X = $133.74) and the contrast on attribute 1 condition (X = $146.52) were not significantly different (F(1, 108) = 0.26, p > .05). The mean difference scores of the contrast on attribute 2 condition (X = $133.74) and the balanced contrast condition (X = $101.04) were not significantly different (F(1, 108) = 1.75, p > .05). Digital cameras Data of 133 participants were retained for analysis. The first analysis was on attribute one (imaging sensor resolution). The mean difference scores of the contrast on attribute 1 condition (X = $190.12) and the contrast on attribute 2 condition (X = $143.15) were significantly different (F(1, 130) = 5.78, p < .05). The mean difference scores of the contrast on attribute 1 condition (X = $190.12) and the balanced contrast condition (X = $144.37) were significantly different (F(1, 130) = 6.14, p < .05). The second analysis was on attribute two (image storage capacity). The mean difference scores of the contrast on attribute 2 condition (X = $128.43) and the contrast on attribute 1 condition (X =$114.12) were not significantly different (F(1, 130) = 0.89, p > .05). The mean difference scores of the contrast on attribute 2 condition (X = $128.43) and the balanced contrast condition (X = $119.46) were not significantly different (F(1, 130) = 0.34, p > .05). Discussion Experiment 2 manipulated attentional contrast directly by varying the relative novelty of the attribute levels while keeping the number of levels of the attributes constant. The results showed that attentional contrast can create a weighting difference when there is no difference in number of levels. While this effect was consistently produced on attribute 1, the results on the second attribute were non-significant. Since the stimuli used in this experiment were identical to the ones used in experiment 1, we

PAGE 42

32 attribute this lack of significance to the fact that attribute 2 was perceived to be less important (see experiment 1A).

PAGE 43

33 A Attribute 2 1 2 3 4 1 2 3 Attribute 1 4 B Attribute 2 1 2 3 4 1 2 3 Attribute 1 4 C Attribute 2 1 2 3 4 1 2 3 Attribute 1 4 Note.Exterior offers are colored in black, intermediate offers are shaded in grey. The hatched offers are presented twice. Figure 2. Design for experiment 2. A) Contrast on Attribute 1 Condition. B) Contrast on Attribute 2 Condition. C) Balanced Contrast Condition.

PAGE 44

34 Table 5. Mean difference scores and simple effects for experiment 2 Mean difference score F values Attribute 1 contrast on attribute 1 contrast on attribute 2 balanced contrast contrast on attribute 1 vs contrast on attribute 2 contrast on attribute 1 vs balanced contrast Annual energy cost (Refrigerators) 99.37 68.60 70.04 4.96 b 4.87 b Screen size (LCD televisions) 286.04 228.72 192.99 2.13 5.27 b Imaging sensor resolution (Digital cameras) 190.12 143.15 144.37 5.78 b 6.14 a Attribute 2 contrast on attribute 1 contrast on attribute 2 balanced contrast contrast on attribute 1 vs contrast on attribute 2 contrast on attribute 2 vs balanced contrast Capacity (Refrigerators) 126.58 94.02 99.79 2.65 0.08 Horizontal viewing angle (LCD televisions) 146.52 133.74 101.04 0.26 1.75 Image storage capacity (Digital cameras) 114.12 128.43 119.46 0.89 0.34 a implies significance at 99% level of confidence b implies significance at 95% level of confidence

PAGE 45

CHAPTER 5 EXPERIMENT 3: TEST OF ATTENTIONAL CONTRAST ACCOUNT WITH REVERSED NUMBER OF LEVELS The goal of experiment 3 was to show that the number-of-levels effect can be reversed when attentional contrast is created on the attribute with the lower number of levels. In experiments 1 and 1A, attentional contrast was shown to create weighting differences but the number of levels of the accompanying attribute was varied, but not the number of levels of the focal attribute. In experiment 2, attentional contrast created weighting differences when there was no difference in number of attribute levels. In experiment 3, the test of the attentional contrast account is made more rigorous. More explicitly, I test whether the number-of-levels effect can be obtained on the attribute with the lower number of levels. Method Design and Stimuli The design was a 3 cell between-subjects design (4x6, 6x4, and 6x6 attribute levels) with three between-subject product replicates (refrigerators, LCD televisions, digital cameras). Each attribute had four or six possible attribute levels. For an overview of the attribute levels, see Table 6. The intermediate offers were [3,1], [3,2], [3,3], [3,4], [3,5], [3,6], [4,1], [4,2], [4,3], [4,4], [4,5], and [4,6] in the 4x6 condition, [3,1], [4,1], [3,2], [4,2], [3,3], [4,3], [3,4], [4,4], [3,5], [4,5], [3,6], and [4,6] in the 6x4 condition, and [1,3], [1,4], [2,2], [2,5], [3,1], [3,6], [4,1], [4,6], [5,2], [5,5], [6,3], and [6,4] in the 6x6 condition. The four exterior offers were [1,1], [1,6], [6,1], and [6,6]. The design of 35

PAGE 46

36 experiment 3 is presented in Figure 3. In the 4x6 condition, the levels of attribute 1 were more novel than the attribute levels of attribute 2 at the time the participants evaluated the exterior offers. In other words, attentional contrast was created on attribute 1. In the 6x4 condition, the levels of attribute 2 were more novel than the attribute levels of attribute 1 at the time the participants evaluated the exterior offers. In other words, attentional contrast was created on attribute 2. In the 6x6 condition, the attribute levels of attribute 1 and attribute 2 were on average equally novel when participants evaluated the exterior offers. In sum, attentional contrast was created on the four-level attributes. If the number-of-levels effect depends only on the number of levels, the four-level attributes will have lower derived importance weights than the six-level attributes. If the number-of-levels effect depends on attentional contrast, then the four-level attributes will have higher derived importance weights. Predictions The attentional contrast hypothesis makes predictions for attributes one and two. With respect to attribute one, the attentional contrast hypothesis predicts that the mean difference score should be greater in the 4x6 condition than in the 6x4 condition or the 6x6 condition. With respect to attribute 2, the attentional contrast hypothesis predicts that the mean difference score should be greater in the 6x4 condition than in the 4x6 condition or the 6x6 condition. Given the results of experiments 1, 1A, and 2, it was anticipated that the attentional contrast effect would be obtained on attribute 1, but not on attribute 2. Results Table 7 presents the mean difference scores per attribute for each condition.

PAGE 47

37 Refrigerators Data of 87 participants were retained for analysis. The analysis for each attribute was performed separately. The first analysis was on attribute one (annual energy cost). The mean difference scores of the 4x6 condition (X = $119.89) and the 6x4 condition (X = $77.59) were significantly different (F(1, 84) = 7.52, p < .05). The mean difference scores of the 4x6 condition (X = $119.89) and the 6x6 condition (X = $76.75) were significantly different (F(1, 88) = 6.72, p < .05). The second analysis was on attribute two (capacity). The mean difference scores of the 6x4 condition (X = $110.09) and the 4x6 condition (X = $132.15) were not significantly different (F(1, 84) = 1.34, p > .05). The mean difference scores of the 6x4 condition (X = $110.09) and the 6x6 condition (X = $94.92) were not significantly different (F(1, 84) = 0.55, p > .05). LCD televisions Data of 96 participants were retained for analysis. The analysis for each attribute was performed separately. The first analysis was on attribute one (screen size). The mean difference scores of the 4x6 condition (X = $246.95) and the 6x4 condition (X = $201.55) were not significantly different (F(1, 93) = 1.07, p > .05). The mean difference scores of the 4x6 condition (X = $246.95) and the 6x6 condition (X = $228.70) were not significantly different (F(1, 93) = 0.15, p > .05). The second analysis was on attribute two (horizontal viewing angle). The mean difference scores of the 6x4 condition (X = $138.15) and the 4x6 condition (X = $129.30) were not significantly different (F(1, 93) = 0.16, p > .05). The mean difference scores of the 6x4 condition (X = $138.15) and the 6x6 condition (X = $133.70) were not significantly different (F(1, 93) = 0.05, p > .05).

PAGE 48

38 Digital cameras Data of 123 participants were retained for analysis. The analysis for each attribute was performed separately. The first analysis was on attribute one (imaging sensor resolution). The mean difference scores of the 4x6 condition (X = $175.06) and the 6x4 condition (X = $169.20) were not significantly different (F(1, 120) = 0.09, p > .05). The mean difference scores of the 4x6 condition (X = $175.06) and the 6x6 condition (X = $153.29) were significantly different (F(1, 120) = 1.17, p > .05). The second analysis was on attribute two (image storage capacity). The mean difference scores of the 6x4 condition (X = $157.11) and the 4x6 condition (X = $126.92) were not significantly different (F(1, 120) = 2.50, p > .05). The mean difference scores of the 6x4 condition (X = $157.11) and the 6x6 condition (X = $111.00) were significantly different (F(1, 120) = 5.54, p < .05). Discussion The results of experiment 3 point in the direction of an attentional contrast explanation. On the first attribute, the means are in the direction predicted by the attentional contrast account, though the results are significant in only one of three product categories (refrigerators). On attribute 2, the results of the refrigerators category differs from the predictions of both number-of-levels effect and attentional contrast. The means of the LCD televisions category are in the direction predicted by attentional contrast, though non-significant. The results for the digital cameras category are in the direction predicted by attentional contrast. Since this experiment used the same exterior levels as experiments 1 and 2, the partial lack of significance on attribute 2 may be attributed to the perceived lack of importance of attribute 2 (see experiment 1A).

PAGE 49

39 A Attribute 2 1 2 3 4 5 6 1 2 3 4 5 Attribute 1 6 B Attribute 2 1 2 3 4 5 6 1 2 3 4 5 Attribute 1 6 C Attribute 2 1 2 3 4 5 6 1 2 3 4 5 Attribute 1 6 Note.Exterior offers are colored in black, intermediate offers are shaded in grey. Figure 3. Design for experiment 3. A) 4x6 Condition. B) 6x4 Condition. C) 6x6 Condition.

PAGE 50

40 Table 6. Attribute levels from experiment 3 Attribute level Product category Attribute 1 2 3 4 5 6 Refrigerators annual energy cost ($n) 100 95 90 80 75 70 capacity (n cubic feet) 16 17 18 20 21 22 LCD televisions screen size (n inches) 20 22 24 28 30 32 horizontal viewing angle (n off-center) 45 50 55 65 70 75 Digital cameras imaging sensor resolution (n MegaPixels) 3.2 3.8 4.4 5.6 6.2 6.8 image storage capacity (up to n MB) 128 192 256 384 448 512 Note.The underlined attribute levels are the levels that are added when an attribute has 6 levels. Table 7. Mean difference scores and simple effects for experiment 3 Mean difference score F values Attribute 1 4x6 6x4 6x6 4x6 vs 6x4 4x6 vs 6x6 Annual energy cost (Refrigerators) 119.89 77.59 76.75 7.52 a 6.72 a Screen size (LCD televisions) 246.95 201.55 228.70 1.07 0.15 Imaging sensor resolution (Digital cameras) 175.06 169.20 153.29 0.09 1.17 Attribute 2 4x6 6x4 6x6 4x6 vs 6x4 6x4 vs 6x6 Capacity (Refrigerators) 132.15 110.09 94.92 1.34 0.55 Horizontal viewing angle (LCD televisions) 129.30 138.15 133.70 0.16 0.05 Image storage capacity (Digital cameras) 126.92 157.11 111.00 2.50 5.54 b a implies significance at 99% level of confidence b implies significance at 95% level of confidence

PAGE 51

CHAPTER 6 GENERAL DISCUSSION Experiment 1 provides evidence that the weighting effect as observed in published studies may not be a number-of-levels effect, but that it may be due to attentional contrast away from the less novel attribute level within each trial. Experiment 2 shows that attentional contrast can create a weighting effect when there is no difference in the number of levels. Experiment 3 shows that attentional contrast can overrule the number-of-levels effect, i.e., that it can result in greater derived importance weights on the attribute with the lower number of levels. In the three experiments, results were consistent on attribute 1, but did not reach significance on attribute 2. In experiment 1A, I showed that this lack of significance was due to the perceived relative unimportance of the second attribute. In experiment 1A, the distinction between attribute level importance and attribute importance is highlighted. Simonson (1993) shows that consumers often do not possess well-formed beliefs about attribute importance. Instead, they construct their preferences during judgment and choice (Payne, Bettman, and Johnson 1992). First, the experiments in this dissertation show that attribute importance is influenced by the attribute levels that are presented. Second, experiment 1A shows that this influence only occurs when the attribute is perceived to be important. While in experiments 1, 2 and 3 attentional contrast only created a weighting effect on attribute 1, increasing the accessibility of consumers product experiences in experiment 1A resulted in a weighting effect on attribute 2. In the number-of-levels effect literature, price was often used as one of the attributes. Since this 41

PAGE 52

42 dissertation uses price as a dependent variable, one of the product attributes that are usually perceived as being inherently important was not eligible, which may have reduced the overall perceived importance of the attributes used in this dissertation compared to attributes used in earlier studies. This effect may be enhanced by the fact that the choice of attributes was restricted to attributes that have numerical values and for which evaluations are known to be monotonically distributed. The attentional contrast explanation is consistent with all the results that have been published in the number-of-levels effect literature. First, all empirical studies (except Johnson 1992) used a sequential evaluation procedure. Attentional contrast can only operate when attribute levels are repeated and when this repetition occurs in the context of other attribute levels that have been more or less repeated. Second, attentional contrast is consistent with the observation that the number-of-levels effect was not produced in the compositional measures that were added before the sequential evaluation procedure (Wittink et al. 1992a) and that only minimal effects were produced when the compositional measures were added after the sequential evaluation procedure (Steenkamp and Wittink 1994; Verlegh et al. 2002). Finally, the reason why attentional contrast was not detected in earlier studies is that they invariably compared conditions in which the number of levels of more than one attribute was compared at once (see Table 1). While the main contribution of this dissertation is the proposition of a new account for a repeatedly published effect, two additional contributions can be mentioned. First, until now, the process of attentional contrast had only been documented in a category learning task. Based on the shared characteristic of sequential trials between the

PAGE 53

43 classification learning and conjoint or multiattribute tasks, I showed that attentional contrast also occurs in an evaluation (versus categorization) task. Second, in previous studies (Kersten et al. 1998), attentional contrast was documented using pictorial values. I generalized its occurrence to the use of metric values. While I conclude that attentional contrast is at least partly responsible for the weighting effect that was observed in studies examining the number-of-levels effect, the data do not provide information as to how attentional contrast creates a weighting difference. One possibility is that participants attend longer to the relatively novel attribute levels, and that attribute levels that are attended to for longer durations receive more weight. However, it is possible that it is not mere attention which results in the weighting difference. It is likely that participants elaborate more on the relatively novel attribute levels. For example, when asked to evaluate an LCD television with a horizontal viewing angle which is 60 degrees off-center, participants may imagine themselves in their apartments and try to assess mentally from which area they would have a decent view of the screen. Aside from the question how attentional contrast works, one may ask when it works. First, the impact of the attentional contrast effect is likely to be dependent on the assumptions with regards to the total pool of attention. In the experiments presented in this dissertation, students evaluated the trials at their own pace. In other words, the pool of attention was variable. This means that increased attention to one attribute is not necessarily accompanied by decreases in attention to a second attribute. In contrast, in a model in which the total pool of attention is fixed, increased attention to a first attribute results in decreases in attention to the second attribute. A model with a fixed pool of

PAGE 54

44 attention is more closely operationalized by a computer-paced experiment. Using a computer-paced experiment may enhance the chances that attentional contrast creates a weighting effect on both attributes. Second, as shown in experiment 1A, the occurrence of the weighting effect depends on the importance of the attribute. The prior level of attention to an attribute may influence the impact of changes in attention towards the attribute levels. Attributes that normally receive a low level of attention may show small reductions in attention as a result of attentional contrast away from the attribute levels. Analogously, attributes that tend to receive high levels of attention may show small increases in attention as a result of attentional contrast towards the attribute levels. Therefore, attentional contrast is likely to have the greatest impact when attributes are moderately important. Finally, a main focus of many number-of-level effect studies was to find a way to eliminate the bias. Based on the account offered in this dissertation, one should try to keep the relative novelty of attribute levels in subsequent trials constant. Since this is impossible when attributes are defined on unequal numbers of levels, one may prefer to use a data collection method that does not involve subsequent trials. For example, the use of a trade-off matrix may be preferred in case of an unequal number of levels.

PAGE 55

CHAPTER 7 LIMITATIONS AND FUTURE RESEARCH While this dissertation takes an important first step towards proposing a new account for the number-of-levels effect, there are several issues that remain unexamined and that could be extended to potentially useful future research. There are several characteristics of the experiments that limit the generalizability of the findings. First, only one dependent measure was used in all studies. Participants stated their reservation price for product offers. While the results of these studies supported an attentional contrast explanation, the number-of-levels effect has been identified using a variety of dependent measures. To generalize the results of this dissertation, it would be useful to examine whether attentional contrast can also explain the results of studies that use other dependent measures such as paired comparisons and rating scales. Second, no compositional measures were collected in this dissertation. While there is considerable evidence from the literature that the number-of-levels effect is not produced on compositional measures, it could be useful to gather additional evidence that shows that the number-of-levels effect is inherent to the sequential evaluation procedure. Moreover, it could be useful to examine if the weighting effect caused by attentional contrast carries over to a posteriori collected compositional measures. A final extension of this dissertation has the potential of being useful. This research shows that attentional contrast does occur not only during category learning tasks, but also during multiattribute judgment tasks. The role of attentional contrast during other 45

PAGE 56

46 phenomena that have been identified in the multiattribute judgment literature (e.g., frequency effect) would be worth examining.

PAGE 57

LIST OF REFERENCES Currim, Imran S., Charles B. Weinberg, and Dick R. Wittink (1981), The Design of Subscription Programs for a Performing Arts Series, Journal of Consumer Research, 8 (June), 67-75. Goldstone, Robert L. and Mark Steyvers (2001), The Sensitization and Differentiation of Dimensions During Category Learning, Journal of Experimental Psychology: General, 130 (March), 116-139. Green, Paul E. and V. Srinivasan (1978), Conjoint Analysis in Consumer Behavior: Issues and Outlook, Journal of Consumer Research, 5 (September), 103-123. Green, Paul E. and V. Srinivasan (1990), Conjoint Analysis in Marketing Research: New Developments and Directions, Journal of Marketing, 54 (October), 3-19. Huffman, Cynthia (1997), Elaboration on Experience: Effects on Attribute Importance, Psychology & Marketing, 14 (August), 451-474. Johnson, Richard M. (1992), Comment on Attribute Level Effects, Second Annual Advanced Research Techniques Forum Proceedings, Rene Mora (ed.), Chicago: American Marketing Association, 62-64. Kersten, Alan W., Robert L. Goldstone, and Alexandra Schaffert (1998), Two Competing Attentional Mechanisms in Category Learning, Journal of Experimental Psychology: Learning, Memory, and Cognition, 24 (6), 1437-1458. Nosofsky, Robert M. (1986), Attention, Similarity, and the Identification-Categorization Relationship, Journal of Experimental Psychology: General, 115 (March), 39-57. Nosofsky, Robert M. (1991), Tests of an Exemplar Model for Relating Perceptual Classification and Recognition Memory, Journal of Experimental Psychology: Human Perception and Performance, 17 (February), 3-27. Orme, Bryan K. (1998), Reducing the Number-of-Attribute-Levels Effect in ACA with Optimal Weighting, 1998 Sawtooth Software Conference Proceedings, Ketchum, ID: Sawtooth Software. Payne, John W., James R. Bettman, and Eric J. Johnson (1992), Behavioral Decision Research: A Constructive Processing Perspective, Annual Review of Psychology, 43, 87-131. 47

PAGE 58

48 Simonson, Itamar (1993), Get Closer to Your Customers by Understanding How They Make Choices, California Management Review, 35 (4), 68-84. Srinivasan, V. and Allan D. Shocker (1973), Estimating the Weights for Multiple Attributes in a Composite Criterion Using Pairwise Judgments, Psychometrika, 38 (December), 473-493. Steenkamp, Jan-Benedict E.M. and Dick R. Wittink (1994), The Metric Quality of Full-Profile Judgments and the Number-of-Attribute-Levels Effect in Conjoint Analysis, International Journal of Research in Marketing, 11 (June), 275-286. Sutherland, N.S. and N.J. Mackintosh (1971). Mechanisms of Animal Discrimination Learning. New York: Academic Press. Verlegh, Peeter W.J., Hendrik N.J. Schifferstein, and Dick R. Wittink (2002), Range and Number-of-Levels Effects in Derived and Stated Measures of Attribute Importance, Marketing Letters, 13 (February), 41-52. Wittink, Dick R., Joel Huber, John A. Fiedler, and Richard L. Miller (1992a), Attribute Level Effects in Conjoint Revisited: ACA versus Full Profile, Second Annual Advanced Research Techniques Forum Proceedings, Rene Mora (ed.), Chicago: American Marketing Association, 51-61. Wittink, Dick, Joel Huber, Peter Zandan, and Richard M. Johnson (1992b), The Number of Levels Effect in Conjoint: Where Does It Come From and Can It Be Eliminated? 1992 Sawtooth Software Conference Proceedings, Ketchum, ID: Sawtooth Software. Wittink, Dick R., Lakshman Krishnamurthi, and Julia B. Nutter (1982), Comparing Derived Importance Weights Across Attributes, Journal of Consumer Research, 8 (March), 471-474. Wittink, Dick R., Lakshman Krishnamurthi, and David J. Reibstein (1989), The Effects of Differences in the Number of Attribute Levels on Conjoint Results, Marketing Letters, 1, 113-123.

PAGE 59

BIOGRAPHICAL SKETCH Els De Wilde was born in Gent, Belgium, on July 18, 1975. After attending high school at the Onze-Lieve-Vrouw-Presentatie in Sint-Niklaas, Belgium, she attended the Katholieke Universiteit Leuven in Louvain, Belgium. At this university, she obtained the degrees of Kandidaat in de Psychologie (1996), graduating with distinction, and Licentiaat in de Psychologie (1999), for which she graduated with great distinction. As part of her studies for the Licentiaat degree, she spent five months at the Universidad Complutense de Madrid, Spain. She wrote a thesis on categorization under the guidance of Professor Gert Storms. She spent seven months on a research internship, split between the Laboratory of Experimental Psychology at the Catholic University of Louvain, where she worked with Karl Verfaillie, and the marketing research company TNS Dimarso in Brussels, Belgium, where she was a trainee at the Fast Moving Consumer Goods unit. These research experiences were instrumental in shaping her interest in pursuing doctoral studies in the area of marketing, which she started at the University of Florida in August 1999. During her four years in the Ph.D. program at the Department of Marketing, she collaborated with Professors Chris Janiszewski and Alan Cooke. In 2001, Els taught an undergraduate course in consumer behavior at the University of Florida. Since Summer 2003, Els has been working as an Assistant Professor of Marketing at HEC Montral, Canada. 49


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E20101129_AAAACC INGEST_TIME 2010-11-30T04:47:05Z PACKAGE UFE0011080_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 12427 DFID F20101129_AABGTT ORIGIN DEPOSITOR PATH dewilde_e_Page_34.jpg GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
39669f04f6679eb5a155e9ad7e2bd91c
SHA-1
04185018aefee4cfd50173779d976926177d1131
6131 F20101129_AABHFN dewilde_e_Page_29thm.jpg
d78b3f9296bbcc75a88b0a037a9da57e
f6bd0038fe4577f3c91948c652f3957c2443b3da
8423998 F20101129_AABGYQ dewilde_e_Page_49.tif
9d317825a0705523615e53ca05bc31d7
7a983ac0c9f492f6d949b05b5bc2d87fa275b678
26790 F20101129_AABHAQ dewilde_e_Page_44.pro
d2e02c3bd0fa557a9d998157d4665022
13d32160c8b694038680c7b75829ad48f4af1ad0
47246 F20101129_AABGTU dewilde_e_Page_36.jpg
bf23868243812c4fc0dc58e2cc27865a
f1f21c5e7bae6ca1ed3804ee67986a67d82ddc78
22478 F20101129_AABHFO dewilde_e_Page_30.QC.jpg
8983be4f9e214dec1c573cc2d9e933df
0f1a03ea58cf4339481b57ea3d5fb72d877d9b9a
1053954 F20101129_AABGYR dewilde_e_Page_50.tif
c02f3135166deb0d9a7bacd32cd5fa81
82a0bc871714b3a90eb9237b2a7116b5b96c1d71
42155 F20101129_AABHAR dewilde_e_Page_45.pro
34857089fb5ebab4b61aef4adb37e60a
27af965c0b3bfb651deedae41cbbf3cb1c673596
25139 F20101129_AABGTV dewilde_e_Page_37.jpg
817721768a9d8b2c290b8e589f276680
c445fe79a7bd319d2bb121dc3cd75cce6b402cdb
6405 F20101129_AABHFP dewilde_e_Page_30thm.jpg
facc09d6f629119ca3bafe56aea13711
8eb6900c2913dc4e9e287367e861ffba41b1c5f5
F20101129_AABGYS dewilde_e_Page_51.tif
1555b11ba86b83e169e9707a442e6a9a
2753f5eeeb8ba521e30de23f3d7d0603e310a467
46647 F20101129_AABHAS dewilde_e_Page_46.pro
9271cd9e6fa17728d9cafb72732b2e7d
30d3728bc41954ba0df7775f2151d955db08c004
63363 F20101129_AABGTW dewilde_e_Page_38.jpg
01e9649f7b5e19ca42701520518556bc
cc1da7abc8b59ad3170bedb9117e3d6f79cce08c
21592 F20101129_AABHFQ dewilde_e_Page_31.QC.jpg
d1a1ea4ae14b3ed04a55ead942a44913
f114dbe8e0a0b1807eb19034811cd063b04cd0aa
F20101129_AABGYT dewilde_e_Page_52.tif
51b32d89b0ec8bfe272aaa4e24566021
3b690b6c54d320f694557fe2c057186cd3b4631b
46524 F20101129_AABHAT dewilde_e_Page_47.pro
fecc55ddfc757946f826673267353286
2983d54728649e1659b0d3ad7f88e9f1b2d134f6
70576 F20101129_AABGTX dewilde_e_Page_39.jpg
a225ae0e4ed80c6f06f4387e57457d3a
76eed6655841e767501060234f30a411c53a938a
6189 F20101129_AABHFR dewilde_e_Page_31thm.jpg
898ccd56218a1db2e8cceca1ef3c514c
858d8d0cc01980cb79a9c1643952406848bfe1bf
F20101129_AABGYU dewilde_e_Page_53.tif
385c0a164b4760e7a69ac809a0c2a157
52b285f6a1491a23f57ac13a24cd83d7914927f6
47230 F20101129_AABHAU dewilde_e_Page_48.pro
79f19343f6c8fd83fe7de5826ba2f35a
c134cf7238a97f629206d92a9dcbf45d6cedbb59
66002 F20101129_AABGTY dewilde_e_Page_40.jpg
d49f3d0daa34578ae7f304c1527410f2
95993d6fd2faa4d506dea1e889a5e76753c5f247
22528 F20101129_AABHFS dewilde_e_Page_32.QC.jpg
162045d5d57dd0202b0b6130da413549
de968f817d628ea07f90c440a798850433529187
F20101129_AABGYV dewilde_e_Page_54.tif
06efa5d49799957879313666f064a4f0
de1dd45e6ce297a83f8be84a85a006e5f5268eb9
8571 F20101129_AABHAV dewilde_e_Page_49.pro
967c5ccc1eeedf7499aa2c65f7d882b8
dc633d2718446c77f825f76427b01acfdce5b3ff
6211 F20101129_AABHFT dewilde_e_Page_32thm.jpg
113083c09104b5c6176d82aa572ce989
e37a6aac10e93a00614b14842ec4ec458d207810
F20101129_AABGYW dewilde_e_Page_55.tif
81c4ad7ec1f3613234e278596052a651
3bbfe3ea635c40055bf1490f4592bb2d8255e24a
37433 F20101129_AABHAW dewilde_e_Page_50.pro
47528a59ed6999f3b24bf81d329f6685
833411739b5b523ead1dbf17d1160f2387140d39
22775 F20101129_AABHFU dewilde_e_Page_33.QC.jpg
be79da932cd2137ce5d700355807eaf2
8142db2a64f9ceee5d6e508c50c5dd2a67821a63
F20101129_AABGYX dewilde_e_Page_56.tif
dfb10e44ece7cd044939209ffebf062e
4d6b52e824ab9c357c6c85b833754c3b0bc147ee
46384 F20101129_AABHAX dewilde_e_Page_51.pro
7623fb47d5c14a7ef22922e7371e6709
1424ff089a6455af07071c385e26fa2309cff9f5
69123 F20101129_AABGTZ dewilde_e_Page_41.jpg
2ea6a2cbb3589e8b395cbe781da6897e
a0b0cdae6ab1849baaf7ebfb0c4cc002b00b074a
6214 F20101129_AABHFV dewilde_e_Page_33thm.jpg
d5658cbcff41d8ab1b8ef08816a5e673
aff646fc385ae2a7e522989b2e432e00f74c11ae
F20101129_AABGYY dewilde_e_Page_57.tif
9e4841c5a422ac477cc176d361d0a248
0b084efbe2c733c42aa1d19ef191beffe63ebe1a
50110 F20101129_AABHAY dewilde_e_Page_52.pro
4c22e5877cf826ef0abbec64c0982586
daeb00a58407113492d1fbd232658cafba40402f
4192 F20101129_AABHFW dewilde_e_Page_34.QC.jpg
5feb45608a677e80400dc79fda9d0246
0e17c000c10e28b8a2bbdbe248a6d52940271eda
93009 F20101129_AABGWA dewilde_e_Page_38.jp2
2e27e8d42c22a4f493e6b385b464b377
5b76952beafaac8b3ffb478f657529b7d8f8b7ad
F20101129_AABGYZ dewilde_e_Page_58.tif
45e91708b98de2b5b2bb0187d6d3c665
331539120bf89bec457a39064e49c1fd3b0f3fcb
51501 F20101129_AABHAZ dewilde_e_Page_53.pro
2039820c3690806ccfbe5b6e1ed31ede
b193c715809e5d5dc7daf49559ebe8e67d90284d
1601 F20101129_AABHFX dewilde_e_Page_34thm.jpg
bb647d203ff3f41af8720e25439dd11f
e958f1bcb3549a32e01b5fe3391c42eb03837757
105829 F20101129_AABGWB dewilde_e_Page_39.jp2
98369ceb494220f64a84607000e89fd2
86eedb60868690ee3b59a3e4be5c7bb679a34d69
10011 F20101129_AABHFY dewilde_e_Page_35.QC.jpg
e7fb20b26369fbda9e724d3f02430096
85f37ad282e5df97af60136ff06bd1a9df7f920a
105580 F20101129_AABGWC dewilde_e_Page_41.jp2
6fc8be58cfa2784bf6927eec7d495619
e55064f25e9d9bba7baf1c769dcfe98c7ff4eca1
1895 F20101129_AABHDA dewilde_e_Page_51.txt
14163abcf1531132a409288ccfed8a52
32cdfc00dc495e5dd7d06d3a478e4125931637a3
3611 F20101129_AABHFZ dewilde_e_Page_35thm.jpg
5a794aa86d2957c7ac1910f398d0931d
50eae066bafde0f029f64a2ccbfa043bb6116cea
10626 F20101129_AABGWD dewilde_e_Page_42.jp2
a95199ac773624e382e9fa3850b40416
f3c7da2e0a58094f8d3cb2d7ef3e135f440555d3
1973 F20101129_AABHDB dewilde_e_Page_52.txt
50aeeb3d148062248ccfe17b767a1e76
234f542659aeb9f07e7059110380583242c56d24
289482 F20101129_AABGWE dewilde_e_Page_43.jp2
eb013533117318cc83d3bb41c7942148
6e85292d92255a905b8bb03ad4e0850a8a4d7d10
2025 F20101129_AABHDC dewilde_e_Page_53.txt
630f1a368a660aea6322152c117d8f06
0eb93d784f172a7ba00f8db5e6091bbdc8e65c12
49605 F20101129_AABGWF dewilde_e_Page_44.jp2
3d8afe223a6b08e1f342d5bf88841974
25b6276b455837a30b523e140a1a0e5e0b36fe4c
1487 F20101129_AABHDD dewilde_e_Page_54.txt
63b8a79c8b002a3132a76d3c0c0095ca
773272e242e338b10be36fa329e2e965d766ac5d
87035 F20101129_AABGWG dewilde_e_Page_45.jp2
abb649aa4de90e019d7891ec12cb2ed3
2fe6ee1fd45b06732f5eca7c3dde531da3856780
1827 F20101129_AABHDE dewilde_e_Page_55.txt
c7dc9546916ad17fedac5b9a7ba4191f
30bc17f125954c473861fb502ef0a8ccd1d3b96f
97293 F20101129_AABGWH dewilde_e_Page_46.jp2
800cbb0d3d35ba547d9e96a84f223081
5c123c4c2cd37b214e97350e02264866e410c9d7
191 F20101129_AABHDF dewilde_e_Page_56.txt
8ae2aaa261b21f900725105cfd4b76d6
fc01a2988f19d7cce79958a4ba142a2c54a9b65a
102472 F20101129_AABGWI dewilde_e_Page_47.jp2
a83c9da6fb2ddeac0a8472ede583a209
da1e721ee489d76486e5a1fc0beaebb6f418f9f2
2169 F20101129_AABHDG dewilde_e_Page_57.txt
27df1e8d5e4bcfb96fa255e9b792a683
052471311230ccfc5a47e101206dca0c6ddb0303
102279 F20101129_AABGWJ dewilde_e_Page_48.jp2
b6f71a3f6c4acb2f83731c214ab42fe4
b6f396e5f5113fbf5a778b2700dada2535267a08
1930 F20101129_AABHDH dewilde_e_Page_58.txt
d06c8fb6ca4e320316d77382f8bba9bf
99f021f30fafde9fe55979b7aee23920cf4fb83d
273123 F20101129_AABGWK dewilde_e_Page_49.jp2
b9e10fcaf204cd7237424215814fbb80
a8583b5403003d5a6cbd96967dc02843ffbf4521
69454 F20101129_AABGWL dewilde_e_Page_50.jp2
491c1203e6b4d44863051d855ad70333
14e2d14182afa4569b2370f5edee39852be45a6d
1604 F20101129_AABHDI dewilde_e_Page_59.txt
487bb8a57048d63613b90d37a30b0e5b
e0ef2f032b2af8afa2268c366baf6bba468f0ae0
97817 F20101129_AABGWM dewilde_e_Page_51.jp2
05f43b585277acbd57a659d5ec7d6ae7
6e3ce7d6f36f3d7eb6191d76e408819dffdde879
354556 F20101129_AABHDJ dewilde_e.pdf
7fd1c7cd2f4a7634d22cfcff99444fce
4a891692fb66e19247b548501453d06536b7d703
106480 F20101129_AABGWN dewilde_e_Page_52.jp2
2514052638eed174945d1ccf2d60e043
d047ff1c999903d726de34819d7b446991e28f6d
7467 F20101129_AABHDK dewilde_e_Page_01.QC.jpg
4f51eee5d201f6c7ec57c431da78ccd1
ba1ae4766e163a4901ff1e0f5ddd08d07b17f668
107861 F20101129_AABGWO dewilde_e_Page_53.jp2
6b07134eddbf1c078e24c386e46157d1
c4d52cff61248c3339164e0ef3250e26dd8712c8
2387 F20101129_AABHDL dewilde_e_Page_01thm.jpg
5084ad4a52c2cc414965fcf265187c50
94bbb2cf9d2402826262fd4370870eec6fe30033
80814 F20101129_AABGWP dewilde_e_Page_54.jp2
2bf860a407670aca83092f4fcac14222
973b78a8143c5ea782280b4c5d131c7ed14a6039
3262 F20101129_AABHDM dewilde_e_Page_02.QC.jpg
2536e6759025e6d96440764dffc7a37f
82181a7258342a993425aec808386b3c13a76735
94952 F20101129_AABGWQ dewilde_e_Page_55.jp2
09cbb38aeedd25463b641ceda0ec77c6
cbfe150fdcd39c7383a686797752a273592ecc38
1367 F20101129_AABHDN dewilde_e_Page_02thm.jpg
b019b7bea16205371a17bddfba2efd96
9525cce3c0d58e5c084a54468a6557b24b39b077
11176 F20101129_AABGWR dewilde_e_Page_56.jp2
b35fa422d1d80899e3b80caaf2925fb7
14a88c44a60e24d174cfc92fdf63c1a0a37d921b
52027 F20101129_AABGRU dewilde_e_Page_18.pro
8424b739b4010c45986800e476018ef1
3323631878ec655090f56c14497f758054485a22
3188 F20101129_AABHDO dewilde_e_Page_03.QC.jpg
42b00bf96cf420c74e17a0d32f73990b
917e6c78d0e84d5d657c957ebc64511201cd8829
119936 F20101129_AABGWS dewilde_e_Page_57.jp2
d9d60e5c832dc00d58654fcc0e7b022c
72590ec06bd21608e877b52c1d71e03f374b309a
534 F20101129_AABGRV dewilde_e_Page_08.txt
e754a826c68a812fa8439f2fc3277425
05506374593b5d1bbf42853ee8836ccbd6e4c7f0
1360 F20101129_AABHDP dewilde_e_Page_03thm.jpg
f7d57f1b8ef1272378c569504a69d23e
f9ff0f5c7804e69e8d152a02be995bdb4467bc72
103489 F20101129_AABGWT dewilde_e_Page_58.jp2
01ef287f6495e824035ffa34281945ad
d0891319f4a670172645f6133fea612fe75a27c3
696538 F20101129_AABGRW dewilde_e_Page_07.jp2
72aa420a855e85f866b9c0ef1e23872d
5b672f91efdea2bc3aa7f424c47a4502a081513b
13359 F20101129_AABHDQ dewilde_e_Page_04.QC.jpg
bdc4cb5ac6cff1bf900088bdc1762b59
c8f242161d8661a8dd47dfe7af5c3530e13a250a
88305 F20101129_AABGWU dewilde_e_Page_59.jp2
e1730703b0308c4788d47fc6e7f1e5f7
a82973b85922380e186fc0ad439292724bd9e18e
3976 F20101129_AABHDR dewilde_e_Page_04thm.jpg
dda89b78bfd4ebaac0c9fc9c3824fdf1
0c98a2138751a17b6f88b30e4f34cff48217c7c5
924 F20101129_AABGRX dewilde_e_Page_07.txt
54ec64324e4eb37e63bfb934e264e8e4
c193bef651d0e2309e5c25f21e77e5f3f3db4252
16699 F20101129_AABHDS dewilde_e_Page_05.QC.jpg
ea1a2bb76c73ede01640a36ac2e5a34d
7a0c27980772da65105660504485f786bebe694a
F20101129_AABGWV dewilde_e_Page_01.tif
5f77054cfe2ce8e3a4656da14398a7b8
72c0b8b6e72e4b8aa3a8240463951c092f06f44b
34854 F20101129_AABGRY dewilde_e_Page_36.pro
9d27cdd380cfa4a226cae17471cec3a3
4ebebab91e4c808bf5c5e28ce815522faee3834a
4785 F20101129_AABHDT dewilde_e_Page_05thm.jpg
e4cbd3941ef602c8fd643a2b06de8ca1
004dc2ba8f22a805fb62f37dbf7537460b9e82e4
F20101129_AABGWW dewilde_e_Page_02.tif
d73345bdffc066004591cd73e5992a89
48f68bfdcbf0084a912141844d06449527f39b9d
22581 F20101129_AABGRZ dewilde_e_Page_19.QC.jpg
a29023ed28ac4547569f589e22d9f11b
23970fa0c14360931c9487ec4e58c23b340dc0c1
5568 F20101129_AABHDU dewilde_e_Page_06.QC.jpg
7def295cb1e70c3212c0f8aba741bab5
97298b75590781ecee3dcaf5f3d2130c2793dfa9
F20101129_AABGWX dewilde_e_Page_03.tif
1ea39f4a7b0395dd41f344b17871522a
5a0faffa9d41c875eabeee5148e3a7fcee2442a2
1834 F20101129_AABHDV dewilde_e_Page_06thm.jpg
97598da94a7605bf401e7799873e6fb5
66749d0d864977ea3222155b4d00cf4feb7914a7
F20101129_AABGWY dewilde_e_Page_04.tif
ce729a8b0df430f267831b498d539295
e3bca1885891f17ce9d9de9444104e2edcc0316f
9429 F20101129_AABHDW dewilde_e_Page_07.QC.jpg
0524d6158c16e0ed17ab39b959094f62
49db8f15aa2b136bca669dcc532a518462ef59cd
12631 F20101129_AABGUA dewilde_e_Page_42.jpg
93ef4c9878431fcd6c721c0702faac0f
b6b8d081cd1b68df2e2e9513aa499bec1403964e
25271604 F20101129_AABGWZ dewilde_e_Page_05.tif
002f4dcb1c82ce76b9a2af33be18cd62
e52b14934b73197def0c5ad67c08f46771255d60
2952 F20101129_AABHDX dewilde_e_Page_07thm.jpg
defec98895425c8cdd771f126a45a61b
8f3d0f3e3143550ea84dcc44ede1655c800dd0a9
31208 F20101129_AABGUB dewilde_e_Page_43.jpg
23420b9481465ffab9ea4730b6ed1e47
13bd8c37a41ac893f26e638a361e182a5f8f7eba
37343 F20101129_AABHBA dewilde_e_Page_54.pro
b5e213529616e47db1657413568f9784
0cd5bdd7497d6230775cd5f739f021ee81aeb7f8
5325 F20101129_AABHDY dewilde_e_Page_08.QC.jpg
19128cc9fb1f878b6833ee0d97273d13
131c0fa2e827c076f706e9141102f9fa791527b8
35940 F20101129_AABGUC dewilde_e_Page_44.jpg
2228fb4002c25af55c2d5b6ef2ff92e3
ae3c7e511397220ca6517e83770f8dba634940f7
F20101129_AABGZA dewilde_e_Page_59.tif
0674981528bf3593360b8ced08c2d90b
b9f43c70e7bef55a3b902aed33891aa61524f971
44517 F20101129_AABHBB dewilde_e_Page_55.pro
4dddb5fe04e0b0b8bbd9c884aad4729b
5aa20194aa6d00cb3b629e07651d3ebf6f6f84c5
1976 F20101129_AABHDZ dewilde_e_Page_08thm.jpg
59df6fa0ad03823ac3b62a641747497e
03e4226f02d27bb8b90b652607acda987602e6b0
62499 F20101129_AABGUD dewilde_e_Page_45.jpg
4e0917354847d1195b4f856aee020b72
9ce8990116a7981844faf29671586bd12067b3ec
8631 F20101129_AABGZB dewilde_e_Page_01.pro
9cf51e83975d065003c3b91c858e7ee7
647a20f11fa6a9e8caea21fcfcd6b9376e03249e
3715 F20101129_AABHBC dewilde_e_Page_56.pro
c5a7e9eb3ab166ea938883e8da02fb3c
5aed40a9aff389ab39a33d78c446c0997d3d531c
64790 F20101129_AABGUE dewilde_e_Page_46.jpg
76a0265338409097647197a8a23f795b
775cb3d158808c85cd02e41c7f93cc0e275a275d
1096 F20101129_AABGZC dewilde_e_Page_02.pro
4bade33fb23835a578f3d2a091bd10a8
c4626c02714c302a255e31a5d52651875722e222
53350 F20101129_AABHBD dewilde_e_Page_57.pro
ece2f73f178c0346ad1ee8287fb00b34
1a939588e22e4179a250802c49e88b7ada638fce
67178 F20101129_AABGUF dewilde_e_Page_47.jpg
58e3570f70c0cdc3d9a97ecdd4a2e82d
e34e39b9cb0595f12560e99bf8f9578c0c3a8e24
14947 F20101129_AABHGA dewilde_e_Page_36.QC.jpg
30cad9d317ac65d27d33bd025e3280de
6a46e837863e69206d9cac3c76a577ac95482ac5
1456 F20101129_AABGZD dewilde_e_Page_03.pro
046e45b596688c37cc00a4e87e61f69c
8edd910a548afa39abb1f82c4e1112ec72f628f6
46983 F20101129_AABHBE dewilde_e_Page_58.pro
42b86ab661b661ba3681f722af4515f1
fe78eaf8875be762b1d0b1b61432585009df2372
67790 F20101129_AABGUG dewilde_e_Page_48.jpg
29e6626c59e1a01bead14c7c9b208b3b
b942b11fb24b2fd10b18309a89a836a21486dd54
4675 F20101129_AABHGB dewilde_e_Page_36thm.jpg
85822cb4aaf74c3159582df51e5f76e2
08666e0bb038221800c65511cb0b149b2da4e214
24242 F20101129_AABGZE dewilde_e_Page_04.pro
5f610e999f539d78a18f812c1af042aa
33cba2820640719e8494d5f2b439a09c3a39e3e1
39542 F20101129_AABHBF dewilde_e_Page_59.pro
603a8ea0850bb2ad4cef6bcf7f9a84b0
af596beb5014e04411154ea5e30d9f5738032680
48699 F20101129_AABGUH dewilde_e_Page_50.jpg
ed38af924273f63be3db0f24142c0b61
83eea8cb6d00f340b5ea5a08e23632657b0e519a
7531 F20101129_AABHGC dewilde_e_Page_37.QC.jpg
762da1c0c40cddd8ec89fb04f31bdaf0
305cc6b76dee4732d2cd6e17e4cdc2ad43e54c60
68906 F20101129_AABGZF dewilde_e_Page_05.pro
1c0428172b5371bb2af13e38ffadecf5
f238b55f47176069daadf251daf96c20a22216c1
65147 F20101129_AABGUI dewilde_e_Page_51.jpg
ae80287b4771a3d929023deb3f5b4827
398c4cc6cfac52c4a0df970f9ed8061e9b7d7ba0
2353 F20101129_AABHGD dewilde_e_Page_37thm.jpg
2b00a97ccf758827b64e1883ded5f2d6
fedefa89329e5d024c1ab02e28a1f152e75c6a39
12577 F20101129_AABGZG dewilde_e_Page_06.pro
5cceed944af9c1e76d3e9b77ba6fb699
268784167a02082305c988337c3dfd7eff408b0d
459 F20101129_AABHBG dewilde_e_Page_01.txt
d7f2e05a06e1fefb0ab8d595c2b6ab36
6c80e5534348e5f7cbc9d8e12a3c5c5de02d5ba6
70088 F20101129_AABGUJ dewilde_e_Page_52.jpg
8a82bc4463760e3b4a059c8f046e4bb0
684c2d8a4e8fa472a3a232518f4c243c360b9bc9
20244 F20101129_AABHGE dewilde_e_Page_38.QC.jpg
075b24c521b7b7117a2d826b6777ca65
594e81f8e7d75f9fa253d613c135a9dfd3ebd402
21136 F20101129_AABGZH dewilde_e_Page_07.pro
2f9180178277b00a03e98c641770178b
07e2c2a461fcf4add4d5cabe072190617f1114a9
107 F20101129_AABHBH dewilde_e_Page_02.txt
e9ffca868aabd415e26b5d5017bddcc9
54ece406b0df8c929b5ae672758cfa12553a5b36
70458 F20101129_AABGUK dewilde_e_Page_53.jpg
9d14951baf6ea526f914e98a03d25aa2
b8a653850eb19450cc47fa8e389de6d0423a99d3
5866 F20101129_AABHGF dewilde_e_Page_38thm.jpg
80825f11ebd0393975ada7bf9cdbcf9d
aa2e4ece29845071608e1bb8393d4a90dcb4b618
11742 F20101129_AABGZI dewilde_e_Page_08.pro
0c9de1569b591569759b38b94963a858
8514e235b5b5a8d4e27732dfbe2056277c6652f0
108 F20101129_AABHBI dewilde_e_Page_03.txt
8c67b1ec0a8b3eb330b64a859aa0aa2f
a3e2d106a875f7454133a4ecb53e745581197fed
53840 F20101129_AABGUL dewilde_e_Page_54.jpg
185462e0782241372ec490d94b9f426d
806d12f74db068838af507c975e305a279b15f3c
23155 F20101129_AABHGG dewilde_e_Page_39.QC.jpg
724234b220610df831b644c09449bfc7
4dd6ec686fb132a180af188477d025aa01f9393c
37330 F20101129_AABGZJ dewilde_e_Page_09.pro
628389ce45a2df891e2ac4e4558f586b
b3f73cd74aa3dc267ee8d50decd3276a67359471
1012 F20101129_AABHBJ dewilde_e_Page_04.txt
47adb10aa35f54c9b1000e23f3837296
7686df0bf1cddd0ce3c05dd7eb9e9ccc691c2d0c
63155 F20101129_AABGUM dewilde_e_Page_55.jpg
69cae607bc3078c98e6d8f14944820fd
2dd16648825154a8deafac40f37ba9132d3f0c12
6415 F20101129_AABHGH dewilde_e_Page_39thm.jpg
fd8d60be8a7796961f58272fa5357565
c58160e849852639b1d84d811ca8dfc00fdded7e
24955 F20101129_AABGZK dewilde_e_Page_10.pro
1fab7266aef8af2dc0c7cb209621e6b6
3755fabc8171c95d7b725f15e1846256e39d1cae
2727 F20101129_AABHBK dewilde_e_Page_05.txt
8a758d42b1256bfd1376c75bb71fc646
781f34ecff95189d117d93a87ed613d48a4f3de4
13096 F20101129_AABGUN dewilde_e_Page_56.jpg
cdf7bde275a984611cc3d5960d191941
b8bff65f348f3bef0d259d5bf251df9d0d55b707
21210 F20101129_AABHGI dewilde_e_Page_40.QC.jpg
252c41a67e308ef2999d79b83b306552
c3aaddd8e043b2fa1ab2357cef2d2d937afdfa25
46549 F20101129_AABGZL dewilde_e_Page_11.pro
f56a4e9c2335c327105bfd7ff47fade4
6ec82d1ab782cfb799a3ec6dbede043eaa80056e
499 F20101129_AABHBL dewilde_e_Page_06.txt
e606d2ea5c2e3d1118c383120ba37367
d9f65be6f67090da5ecf6d8e7ffda3c80d04b2e7
74957 F20101129_AABGUO dewilde_e_Page_57.jpg
52f8289f42037a1b8e4cbf55853ef8f8
3eaa2de48e8c928740bee47c01b6dcdcfcd01187
6195 F20101129_AABHGJ dewilde_e_Page_40thm.jpg
0ac954db4ef64f56a2a2b1adead73b38
e654563b64993a4fc0f2b2965cedfe0f8fb13fd7
50166 F20101129_AABGZM dewilde_e_Page_12.pro
739f6b8b39ea7845390d64f33e2bc167
cc2222100ad4f18e32689ae142a576363d87194f
1652 F20101129_AABHBM dewilde_e_Page_09.txt
a9fc1fe64e8e0a3a9e866e93ed29d2ac
ce56de1cd3ae2d7c36384b8dd490e9432176e529
72505 F20101129_AABGUP dewilde_e_Page_58.jpg
ec8cd0822d66acb43b3c4a235596d6c8
901b956bd2749318c1b71966165886042726e58c
22519 F20101129_AABHGK dewilde_e_Page_41.QC.jpg
bd51ec0e5ca1b0f5eeee36a77f5dccbd
62c2dea5a03eee043a9b54acbe6b5b4e726542b8
12461 F20101129_AABGZN dewilde_e_Page_13.pro
75b36ab528f540065ba7bfc261af38ac
1cabb0fbb9865aba488f2ccc2179e2f13f7a0a02
996 F20101129_AABHBN dewilde_e_Page_10.txt
6b2ed1c36df3327f254e96cbe84b169a
708a0e7353e748fc2ed0ff9e8a31e2c5b4407cb0
60176 F20101129_AABGUQ dewilde_e_Page_59.jpg
160d7958f8c425b712c30692d8a4b1c5
d2578bc285de943388d08759a61806d6c7d886ee
38554 F20101129_AABGZO dewilde_e_Page_14.pro
ec31aa32cef5410140b20e2a54d93ac2
bca0567b19b05bf3af7c2363d93314d23fd886a0
1903 F20101129_AABHBO dewilde_e_Page_11.txt
1f336d8a6919663a8052cad4c1eace95
1552f11d1738f251951b6ca019971be165d63910
24496 F20101129_AABGUR dewilde_e_Page_01.jp2
1039f76eaee163e45a733e052d10cbe3
440c056824d851ed5badd8203a3d671889161db0
6477 F20101129_AABHGL dewilde_e_Page_41thm.jpg
ca2dcc27425b384e8616b3de9588b3c0
5771658ecf0fc503013d85fae2dfc21025791f90
51482 F20101129_AABGZP dewilde_e_Page_15.pro
4533aeffbf62186c930b34b7551578aa
c1185278bb27183d842ca5ad74383fa85ae8df05
538 F20101129_AABHBP dewilde_e_Page_13.txt
0e08aacc069aca8861a3f2c7fa4450bf
1be14076098fbfb5aed9bd9e0d75f673a6ebd9ad
5431 F20101129_AABGUS dewilde_e_Page_02.jp2
e0b5064e935052de55f946dcd42f609e
39265e242de56a0fd4ab61cd0b14156dca953112
4291 F20101129_AABHGM dewilde_e_Page_42.QC.jpg
71fd884b6a2d023438c67cf38f46d055
2429f47f0bf0f21e3f509fa9a833bf64d74366a5
51046 F20101129_AABGZQ dewilde_e_Page_16.pro
e5a234414039ef6b71972e1a2744ec86
274b449ea68d431aa60ffa4c379878cba4b5bb1f
1651 F20101129_AABHBQ dewilde_e_Page_14.txt
c9c3406ddf2aea91a76517de11ebf343
4012537ff5b105b020ddb3002ab227cdaccd2ba1
6071 F20101129_AABGUT dewilde_e_Page_03.jp2
e2c8ab6599bdd15fc1aa3afe6e53aeda
0efe0695b512550309e38b4ebc6cae33610747c6
1611 F20101129_AABHGN dewilde_e_Page_42thm.jpg
6635567f7ea661557d1c722fb9c1a2e9
18aedbb38e79de2c40522269e1484164160f91d1
2022 F20101129_AABHBR dewilde_e_Page_15.txt
ff82a159331ed182e9e4fa12ed915a51
488fe478630f5fb2a25c3ab14d5c05c8da200e4c
54927 F20101129_AABGUU dewilde_e_Page_04.jp2
a790e5781d4c1515ed8ac6cb2e4f1914
55162181d8d7584cf1d8d7581739bb6366b7c756
10477 F20101129_AABHGO dewilde_e_Page_43.QC.jpg
51ec3381c5e9742c21e5ae95e6928653
e2b4bcd9639e5b5e183bdce947d263eca0944ce4
54421 F20101129_AABGZR dewilde_e_Page_17.pro
a93dcdc17dd55988fd8852073b9b0ed9
a291a39e22b95f06466eeba3be88fcadef81f8f6
2004 F20101129_AABHBS dewilde_e_Page_16.txt
1dd17bc0330bac060333fdb3c7230c46
0a2e52c046619ee44982be5d1b30030f09096f02
1051976 F20101129_AABGUV dewilde_e_Page_05.jp2
efc6fe3cae16bd13938277febf099b68
b77565e376747acd0df30d769ed96b65a1798d09
3689 F20101129_AABHGP dewilde_e_Page_43thm.jpg
d71f141ed95d0fe2c7d2bb7045c91be9
87038b5b32eaeab6cd7fc3732e66c5b1029f4600
47864 F20101129_AABGZS dewilde_e_Page_19.pro
d4ee6dce3e669af812002cc764acd1eb
7f191e80cdf348a046a6fe5ef063209c169b33e9
2135 F20101129_AABHBT dewilde_e_Page_17.txt
b42d75e5d08e1b39fd29279b238a4d74
7abc85df4395cc0789c395ad580a61d5c691b4b3
313159 F20101129_AABGUW dewilde_e_Page_06.jp2
4a43cdb70e1ca950a4d246a6af555842
701deea0de2bece2e68980a3a265c95e0e78c6bc
10795 F20101129_AABHGQ dewilde_e_Page_44.QC.jpg
002031d55861af716afa08128097018e
f46e176d1cedd3e20e4359d8ea16f30ab21d83af
49437 F20101129_AABGZT dewilde_e_Page_20.pro
dda062fd3e4768f5fd16ca735ba4f355
c223a4010aec8032d6f6561ddee4faa69631b2a1
2041 F20101129_AABHBU dewilde_e_Page_18.txt
bf4144248f949dadd1bd02b9b824d38d
56745cd4a3d564f9968c074d3bb810351e12887b
253769 F20101129_AABGUX dewilde_e_Page_08.jp2
e0f55717027f9266b82833c5d126595d
5c55d167ecbdbcf0df417fdb3dee8eed44cc4252
3518 F20101129_AABHGR dewilde_e_Page_44thm.jpg
a86884a2ae1518aacf59139e414c8119
495a8e20d27e1dd82f4e96aa856dcb6c3dac89a5
52872 F20101129_AABGZU dewilde_e_Page_21.pro
212acdef3f64d989d2a7ae3aae298c49
96543494120df44939d5bf67fba33606a8adca95
1923 F20101129_AABHBV dewilde_e_Page_19.txt
a2ca50cb04e66e17fa11995911f8698b
9f44d4dfafa2ef31816b44f53a5e8a3170b65ba8
81993 F20101129_AABGUY dewilde_e_Page_09.jp2
7064fd6695e2fca58fcc3ec75af96947
d958b854e20d0824a0e6b0d3f99aeee95e537299
19739 F20101129_AABHGS dewilde_e_Page_45.QC.jpg
c0d43e63a64cdb5e5c055880f4b83915
525a8c2368c43d7f36eb8b160b33cad82ceeeaee
50644 F20101129_AABGZV dewilde_e_Page_22.pro
85779742fa0f901a9e4865ec30f4bdd1
7dfdf008f4ddb2f38b731355d57df26670663e11
1947 F20101129_AABHBW dewilde_e_Page_20.txt
49ab6d7fd650e3ba369d18c0f5497e9c
074ba4816a0d996bdeffc39d1634ef8c71fd8c70
70258 F20101129_AABGSA dewilde_e_Page_27.jpg
611f306766a1c5228cfca6d18b72689b
333821287e8b2e09dfa4331db4cd92bc820da029
54412 F20101129_AABGUZ dewilde_e_Page_10.jp2
9b532b8f4cc394a6e0d0d0998605a81f
d29142ea53715e96135228a61b640c9d702462d9
5527 F20101129_AABHGT dewilde_e_Page_45thm.jpg
c172361f74f87935b03afe246a83a9f3
ed9e10dfd1aa2f9b0881daf6524f870c35de3519
50541 F20101129_AABGZW dewilde_e_Page_23.pro
acec4c7a43a7e84f04ea91d44cd62f15
0dcfd6b0577b1600d5dff4f5833b8559c05ef1e7
2077 F20101129_AABHBX dewilde_e_Page_21.txt
1b1aeeafb3d12fb3911758c27de0c9bb
0bcf6e854282eb91a2f657270f128bad519e1951
6384 F20101129_AABGSB dewilde_e_Page_22thm.jpg
98861b723d3ce8617459ffd7baa59dee
76723cd98b89c40f29579c749a13ebc8e4be00fe
21769 F20101129_AABHGU dewilde_e_Page_46.QC.jpg
985db6291a4dd25c930352f446f83423
4484826136bca333daec87eaff8ca0dea6a190b5
41637 F20101129_AABGZX dewilde_e_Page_24.pro
331340b2969ab686d8a99de3e584d9f9
984afdd188a8bccfedd4614a1ef5272f58f853e2
111492 F20101129_AABGSC dewilde_e_Page_18.jp2
94359abc24045d6144a2f0f37d7587df
61074f01eff6985fdfb67ed3b668f4bba5da82c5
6008 F20101129_AABHGV dewilde_e_Page_46thm.jpg
40310d183c690dbea7bd0408d2031db6
fb3fbf44bd5f59acf042c3d212a0d978c79adb0a
20044 F20101129_AABGZY dewilde_e_Page_25.pro
6055febbde4fc60cabd20cbd1a7f52ef
c26957fc56c77409d1eb02efad15fbea50f131a5
1993 F20101129_AABHBY dewilde_e_Page_22.txt
012758f849f6e0ed5bfff7a581e01eda
04e0731212e406b1f7d2858ea6c3683987df8961
4390 F20101129_AABGSD dewilde_e_Page_56.QC.jpg
bcc9eaf4c3175461576c60e15fe6774f
e29e5684eb73fde33a8ff61ee257e654d5197a86
21826 F20101129_AABHGW dewilde_e_Page_47.QC.jpg
ddbab585a5d0eefccd52695bd23b9963
d105a014fa745eb10f33a15518bb0832d357cafb
F20101129_AABGXA dewilde_e_Page_06.tif
71f5213a0669763acfd2f0bba29906f8
17838f71da39f5ac802a7158368a3152e64c7ab4
37317 F20101129_AABGZZ dewilde_e_Page_26.pro
d6fb29bddf25a0902538a73b0ca9d863
9ee5c69eeb82bebd8b8c26ede100812fc8b9a494
1989 F20101129_AABHBZ dewilde_e_Page_23.txt
49b58db7539bdf4bc4540983e0797523
18d027bb1efbbc260edf2abc207c68e30fc50b35
5985 F20101129_AABHGX dewilde_e_Page_47thm.jpg
6aafb9c6408021e1b94a66a5319af8b9
efa49b4f7db0f4bb3f11d07c55920de615ecca9b
F20101129_AABGXB dewilde_e_Page_07.tif
f6f7ec423322c78543defde2252ce8e6
471f023b9af80b8eb79bb377d7bfd93f018a7070
98837 F20101129_AABGSE dewilde_e_Page_40.jp2
e6293d0f36f77608b30c3b2b5c76db39
6f30a20177bfe362e9b23592ac72d427a48b62a3
17557 F20101129_AABHEA dewilde_e_Page_09.QC.jpg
3c862cfd42b65755727cb9d0033dd353
de08b6fe21b809f70c899c27858ea4d282d6c51e
21902 F20101129_AABHGY dewilde_e_Page_48.QC.jpg
9b3653d776b074daee92c3c21d334acf
2c26bbcfa2b9d15653a7d60a3efda37076110a1c
F20101129_AABGXC dewilde_e_Page_08.tif
e8b2cc72d90ba7f9fdda19b1d59ea153
673a42859da45c567f7932ad4e4d954193b12fee
F20101129_AABGSF dewilde_e_Page_27.tif
a2d4b806ffb52f374e2b030912414bb7
b90587c393fc6297795b21be03218158d61af401
5154 F20101129_AABHEB dewilde_e_Page_09thm.jpg
4b7f4ec1eba44e9b74756a4d246014f4
2dcae969edb029e91b3bb55707e0fbee60f8922c
5925 F20101129_AABHGZ dewilde_e_Page_48thm.jpg
3266a0cf4e4ea2207a5ee81332188778
70b44c1ebe119c8c4c49b27cd830a13e0d97fcad
F20101129_AABGXD dewilde_e_Page_09.tif
e6d8de9ed8836fe114028a0b406568ac
851c6f141faf21b5ea893c26a866123f19036899
1764 F20101129_AABGSG dewilde_e_Page_45.txt
fb8b5831d9422578c665b48fc4bf5e48
b6fdf471f6e59ef0164ee676f05224d7840f18f0
12886 F20101129_AABHEC dewilde_e_Page_10.QC.jpg
a7aab392cca5c5e4b55db7df2a5d99c0
07a8a8b58097b5e13d4a2294938391061ed04052
F20101129_AABGXE dewilde_e_Page_10.tif
0a5869fae0629f759499df3894b3d46e
ab0fac302961ff6171db78129be67d4cae30b350
29223 F20101129_AABGSH dewilde_e_Page_35.jpg
9981c28278924e1440b1007471593825
e772b70cf2949a2c37cabf7a79a69b243292fa78
3739 F20101129_AABHED dewilde_e_Page_10thm.jpg
d686452c6b32075d93f3f8c371138a8a
c58fd8cb0b7533b7ddfcfced335a378437328088
F20101129_AABGXF dewilde_e_Page_11.tif
3d2dd2e885182052e78752f44b38d345
aa0ed0ddad924b14a873734411d56c6db53644f7
1974 F20101129_AABGSI dewilde_e_Page_12.txt
7d364cb9824b04c1c5d48a7109faff59
3b19f7a0827062723a4c6c063d3bceabaabb058a
21640 F20101129_AABHEE dewilde_e_Page_11.QC.jpg
0a888c46f9a7a1acdd49a7b40396b81e
31717159f711ceaa982d2b2d33702d1f96fa033d
F20101129_AABGXG dewilde_e_Page_12.tif
5b47a2ecda7ee5941453b19f132e0c7a
a129e513454f074990e172f82b423652e311e4ac
28734 F20101129_AABGSJ dewilde_e_Page_49.jpg
5fde28304c83fa1b12d53ca03ab7cbc4
39db62294cd53639e5bec07ae6843dddb7bb8fd4
6127 F20101129_AABHEF dewilde_e_Page_11thm.jpg
7291fee93bad0abc5d1e9ef949b4c923
b27d441a3851344a4b8e90239d384cf1c4a2df9d
F20101129_AABGXH dewilde_e_Page_13.tif
f5b6bd314f50db2d3cf3328bc06463f5
f9ce64b90a0cf5c1198bce0bbf63708eb4277e72
91153 F20101129_AABGSK UFE0011080_00001.xml FULL
8b0e249330ac2fc2a3bbcf3fc63b5d71
ad981400275f201a0c3202b0b9d5d45b22fbbe17
22831 F20101129_AABHEG dewilde_e_Page_12.QC.jpg
8fb606abc93027dadd2e3b24eb2d0cb3
c69f1e1bc305e5a0c6ee67b6e84b4a060a74cd81
F20101129_AABGXI dewilde_e_Page_14.tif
ecbfb2dc2fcc75c05f5b1f62dff1b3d3
5a73b11d7631b21d880e167c0c3ccb87e9d654b2
6344 F20101129_AABHEH dewilde_e_Page_12thm.jpg
7e8593fd1829a176f03989993dfde8c4
4ee305c25810e835487e8833998d393dc95df7b0
F20101129_AABGXJ dewilde_e_Page_15.tif
7c5bcccace0cd47572487272c3e26d88
54da182f79103667ce177e98002a279100a70cec
7624 F20101129_AABHEI dewilde_e_Page_13.QC.jpg
ecf0841d098676c6c35ed8220a69cca0
fc45aeb681b21e82d324ef8071e2fb523d4e1e67
F20101129_AABGXK dewilde_e_Page_16.tif
f1cfdb54e4e9532834d2ebeeab6b6031
d9a26fdcd95c733616e5f20732d9b9778e38c987
24384 F20101129_AABGSN dewilde_e_Page_01.jpg
a3b90ca7deb0d05e537806cd035c8b08
9dacb1c4ebb92efe3827ae1db24bf88dc9074269
F20101129_AABGXL dewilde_e_Page_17.tif
10b85f386dfef50d0a8d416714208f93
c989e4fac33106582f159490400814332374770c
10193 F20101129_AABGSO dewilde_e_Page_02.jpg
489db329cf704fa5737304d2a14802c6
e6cbd6c7c8fdb522f99430a233693300cfbc326c
2403 F20101129_AABHEJ dewilde_e_Page_13thm.jpg
75e89c53a37a42a6bbc07f828cbc536f
444e5e600599f2a797bf62a46cbf1b028b7935be
F20101129_AABGXM dewilde_e_Page_18.tif
2e3095c6590fc3a4adb2d9b0120d4843
267d732aeb57b30278ae754fb8760f41e6e27a9c
10543 F20101129_AABGSP dewilde_e_Page_03.jpg
54fdb9eb537be6e486df7065c392ae44
3c789cabf144a1b01c3ce8477fe43b4d49f82dd9
18474 F20101129_AABHEK dewilde_e_Page_14.QC.jpg
d2283e3be6753c6d3a65181f4ca26865
02a62f3c27990a026675987b68cb7cdb75b03942
F20101129_AABGXN dewilde_e_Page_19.tif
fe58501179c034b1073751232aa923b2
1b6a1e16146471f576ac49fb279a2a740ce51670
40047 F20101129_AABGSQ dewilde_e_Page_04.jpg
e7ff925d1d5ddd9f01ded522fb234fc8
a92de0b34ae31a56a30b16b9b4b3343871c62fff
5419 F20101129_AABHEL dewilde_e_Page_14thm.jpg
ad512d0bd78d4147d80a4b8434b7ee0c
160af67c0e78cd41b1c18ab207ada1587fa141bc
F20101129_AABGXO dewilde_e_Page_20.tif
0d116f6e7df70928e2600bd9844a32e5
3fe627add66d760452bcc5528c9ff0443f351f35
66786 F20101129_AABGSR dewilde_e_Page_05.jpg
6b076d22984a1e796f754683fee5e999
2b050b7c7488c2eeef2e71703fc98007ab81f5fb
24077 F20101129_AABHEM dewilde_e_Page_15.QC.jpg
16fe5759180a4027fedcc62a072870b8
507081d6811df76eb1a20d9acb84f66ad6dce409
F20101129_AABGXP dewilde_e_Page_21.tif
1fa523d914d067db0b4c329d036e0d05
229a26c56e1a81bf6d4e62bef1fdb5d17e856b7d
18022 F20101129_AABGSS dewilde_e_Page_06.jpg
ee38186c03480c6b35e23b96705feb89
5ad714fcde96cd75334ce16ae104ac66fe160f65
6609 F20101129_AABHEN dewilde_e_Page_15thm.jpg
8b4b8d11eee1d767ad3d1952e427e668
a266cba972872853a23c5085aa464225a7e116a9
F20101129_AABGXQ dewilde_e_Page_22.tif
c8f555f6fb2be08d0c5551b220453351
70870054f5875f0746e6bc9e02961095a6ada6b4
30967 F20101129_AABGST dewilde_e_Page_07.jpg
2c08e2507be5a30919c547917a408085
2ad90a5a6c6cd062560deca101526bb5f0a6b708
23467 F20101129_AABHEO dewilde_e_Page_16.QC.jpg
945d28ad3eabeb740897db1873e3ded9
2f272266fae22a1af436fbecde587f56ba7a64dc
F20101129_AABGXR dewilde_e_Page_23.tif
e4f6b1b9b65069c2702e5675e0337454
04e0798c06f1d9047b640e957c1c4728d3840da2
17067 F20101129_AABGSU dewilde_e_Page_08.jpg
4a84b43e5abbcb1eed008bf49ba406ac
4a3cf718b5091e37ed0fbffbdaa02967b9e2e6b7
6527 F20101129_AABHEP dewilde_e_Page_16thm.jpg
2f37b818ec88619efe0d153561def487
7b4aeddbb2d1a5aadcd253bc525334e1cc86b31d
F20101129_AABGXS dewilde_e_Page_24.tif
4cee472d4eeeb15087e4872dab0d3461
0aed417721442e6c9c98914f3b10aa81806edc97
56753 F20101129_AABGSV dewilde_e_Page_09.jpg
29af3912807c25ec8eca6b9efc06767e
b34fded573b3c65505dd67c33568f052d30eb827
24753 F20101129_AABHEQ dewilde_e_Page_17.QC.jpg
33ff7ef319e54beb61bd76ac4f4f2f0f
96d0227d478dcbb1df8fc6639e31d1e99b007461
F20101129_AABGXT dewilde_e_Page_25.tif
0ca4874c3de59c05dde4f428f75287fb
91514d3ab92cc12dc25d0b11a9863cea5d438a10
39235 F20101129_AABGSW dewilde_e_Page_10.jpg
7cca3c5791f8e6170e2cba81cb42be7f
8684c28140d4e77370c09406b5c7c064fad14554
6725 F20101129_AABHER dewilde_e_Page_17thm.jpg
8b5e3d4a649b386ee0c44b01a66ef5e4
ee163337e59d8df4e995513e4e208307a649e692
F20101129_AABGXU dewilde_e_Page_26.tif
3f2e6adf6f98055925702dafd4503c70
62eab4d0566cf7a3077d8f55928b819588c19ebf
66017 F20101129_AABGSX dewilde_e_Page_11.jpg
361b8364673383b51ca56fc451c1131e
78bb8339a3c5fce827d23c7368ff3aeb1c2f1423
23921 F20101129_AABHES dewilde_e_Page_18.QC.jpg
d316f433ac0cd329a4ab31e758585f48
9805c52bd2d452f5e96dea0c9f5c3c0a0b67f69d
F20101129_AABGXV dewilde_e_Page_28.tif
eef4805d7010bc37c832e65d4da87cd2
a98f9e177dff2d1acad68e8b9458d8cb31de7a47
6480 F20101129_AABHET dewilde_e_Page_18thm.jpg
6e6472bdf342cee3ec01be5eed919ea4
02cd6f8553efd054fa85536fd96cd8f35a174880
F20101129_AABGXW dewilde_e_Page_29.tif
250c68a709c0e21e81fa334474395d90
5cdd30bb1ff7f18864de9d999aad874e54ddab05
70142 F20101129_AABGSY dewilde_e_Page_12.jpg
c9b5729bfed7b13860309bfd5786f928
9dff134684bf18f884245bcb1ad451349fb95d36
F20101129_AABGXX dewilde_e_Page_30.tif
24dcc01d4405ce8b5e6b354c75455f7a
a9c42453c9235d7d1f662678502dacf4ff46aa7e
22949 F20101129_AABGSZ dewilde_e_Page_13.jpg
5a84be3925822605c69267a491ea801d
b4faccfbdf6305f10e5193c5d40bed36122c3c09
6352 F20101129_AABHEU dewilde_e_Page_19thm.jpg
2b8b691d3d07d935f1a63f50ccf9878c
8bb304cc16130e22abfeb81d5809d5d145578119
F20101129_AABGXY dewilde_e_Page_31.tif
d4295ea20e7bd42211d953a7618f2489
f5a5f644b43c2772cab5e91af1dd7936e0b30472
23290 F20101129_AABHEV dewilde_e_Page_20.QC.jpg
324ee71e46324154a618d6e0c26440ac
e94bdebdd7e4ebb1edb1d617e1b669a9676cf234
F20101129_AABGXZ dewilde_e_Page_32.tif
cea3bfa199cb30161af7eb4b593ab7d3
6165480af34de4fbadf7f4e3d3bd3dede986097c
6476 F20101129_AABHEW dewilde_e_Page_20thm.jpg
c228fdadab814bdbc071c95c0521fbe5
2f77c18a3cd68d51036dad195c76349ec238e49a
98958 F20101129_AABGVA dewilde_e_Page_11.jp2
e2d39642df90e4c79bd281097757c6ee
6b0b664070ee2137e6c0a7db15d0cb27e3ac82c4
23866 F20101129_AABHEX dewilde_e_Page_21.QC.jpg
6761718ed26e8ecf59837faeb6275b99
6800c561741988e96a1dad6563c279ceb177a6ac
106706 F20101129_AABGVB dewilde_e_Page_12.jp2
ca873c87941b404e00113cf8daf7ccbe
5fc216cde3e7c0bd97879fefdc7ebd6ab83f2d3c
1663 F20101129_AABHCA dewilde_e_Page_24.txt
a904fcb9fd7b4e75fc2cfe7c0f93e767
9c253282e3d15bab4204364cae35a2c1bcb416a9
6582 F20101129_AABHEY dewilde_e_Page_21thm.jpg
10172cf4a08e17727e36f230f7d3c90c
3c2c11a7b2153b749b593946239b05f44fd2c404
28632 F20101129_AABGVC dewilde_e_Page_13.jp2
02be7db446b5ee88d094415b63689576
ce698dec4ed198919ca5108993ffedf6de952473
975 F20101129_AABHCB dewilde_e_Page_25.txt
1a34e9db288053cdd3661d90c117eb41
bf9e7cce34eca33c254de2f0b34cd1c6a388d2e8
23111 F20101129_AABHEZ dewilde_e_Page_22.QC.jpg
f2164814382fcd43046c514831f319e1
a8b9d429b3d216a21a371fe39f9d964eaacc0736
84789 F20101129_AABGVD dewilde_e_Page_14.jp2
f5881bac69d1fea56ed6e206f867b327
adf947f18d58a52dfc7d5fc70e9d24f7596a570a
1594 F20101129_AABHCC dewilde_e_Page_26.txt
2087ea15ec4d5de0641c21c5c59d02b2
2ab0d41df8f7011a878fd9999f41331f1af364e5
111003 F20101129_AABGVE dewilde_e_Page_15.jp2
69dc1180f4c073ba7c2119a4bb4556d2
6bf6f6609ad7329cbf43f0d076001b355efb3904
10776 F20101129_AABHHA dewilde_e_Page_49.QC.jpg
ca0ba00f05ff46ae2ce25f61ece70d55
3c49c6d5066b74745c9dcd859cb197f62fcf21fe
2044 F20101129_AABHCD dewilde_e_Page_27.txt
af0c3dbf75921f867001cd6e09d8141c
00cbc047c41432a0fba74f0731cd28c8e1baa8cb
110256 F20101129_AABGVF dewilde_e_Page_16.jp2
ac9485d124abeea6dc8702837230458c
8f37b22f2e9c85b38e195866dc56eda6b3cd3352
3523 F20101129_AABHHB dewilde_e_Page_49thm.jpg
66d57b0bec4ee7eb93210478ac26f6d5
67cb88e82d8eeb0e6798ce2f9a8f26ca4a371454
2011 F20101129_AABHCE dewilde_e_Page_28.txt
ed4433b6a0d23a0a2ac0c61f7e06498e
e0ae62511f5f21603da575e9926e3aee772d01fb
114797 F20101129_AABGVG dewilde_e_Page_17.jp2
773dadaaf4239423413a115eafe5977f
e59d5ad88449253b46150755d032a4b0f2d7abe5
15199 F20101129_AABHHC dewilde_e_Page_50.QC.jpg
ace35137031b7c7e2c2bb7a056f2ab90
e3fe72c6d10d2a685c3ad20fdd5dd8065c20f94b
1873 F20101129_AABHCF dewilde_e_Page_29.txt
ddc31dc8a81259b52da037364882d948
a5977ec7206e4117ab3bc6136b82fe0d3e1c2a3a
103622 F20101129_AABGVH dewilde_e_Page_19.jp2
03c6633695d37e8f3feddda7fd9b860b
3a686e04f0e1d61a8dbcd70511cc142936140ea2
4828 F20101129_AABHHD dewilde_e_Page_50thm.jpg
46dff502a1b2b836c17017fd993141ef
8b9ac50f7abc814a9821981ad2857eb865c7f077
2013 F20101129_AABHCG dewilde_e_Page_30.txt
fe7462779057f3dbc332ecc8715a7257
d5439d72c69284802e22a7c1060e8252bb476ff1
107904 F20101129_AABGVI dewilde_e_Page_20.jp2
80e758a2444b77b13e033a3159617cd0
0d937550db8f12c92c0083148765d69bd18c78ad
21380 F20101129_AABHHE dewilde_e_Page_51.QC.jpg
2d345ef154d5ba06e3ba4abcc24612f6
27be3fcf270bc279a01b952151476ce58487b565
109263 F20101129_AABGVJ dewilde_e_Page_21.jp2
7af6632e97235e6a50910d16180b0eba
89ba9c2034cdf2b86320f9ec6d717a9c3e5e132c
6111 F20101129_AABHHF dewilde_e_Page_51thm.jpg
ce98d056f06049ea2023f94314e18488
56bb5ed405d9ac8248eca0f56bbadde350346967
1899 F20101129_AABHCH dewilde_e_Page_31.txt
0411b5767b7a209604a5ffdd1ac77d3a
168ff573cc0fbe5e95fb0442eb4d646d88ea9b1f
109234 F20101129_AABGVK dewilde_e_Page_22.jp2
a5b5d830a314c6ee2f86db9c1281fb54
393e98e482901b2d9991e191dd07c22be7621060
22351 F20101129_AABHHG dewilde_e_Page_52.QC.jpg
7500b8fc900022ba520957c4df0ea286
10049fc21b7d6b2da757fd55d5cf10767f24333c
1913 F20101129_AABHCI dewilde_e_Page_32.txt
964ecfae0924161f6644d615b32baa06
e8e2ea99c98b1cd5417b03ae99830da6e27a31ae
107567 F20101129_AABGVL dewilde_e_Page_23.jp2
3507d0e7a655433befa133ba2b75a73d
63ce6ef07f4d180bfedd1280a2a8e998be8f2927
6244 F20101129_AABHHH dewilde_e_Page_52thm.jpg
387320c3df7702435de68d920b02951d
a5fc8a2e9a4c6f3335b29a50f98720075f6d7ecd
F20101129_AABHCJ dewilde_e_Page_33.txt
abe4813d36a4c369270903878fa30624
ad51b81260ae9ff1ce5bef5950ec2e6e31c08d4d
88712 F20101129_AABGVM dewilde_e_Page_24.jp2
5f689c672a4d390dcf20736301a896b7
f9f9b0ab058bf809636f675ef48be10070d42356
22876 F20101129_AABHHI dewilde_e_Page_53.QC.jpg
e10383a69ea751e834e31b087362c1f1
4f49966f191da7ad857491b2d86a019a49b87e68
178 F20101129_AABHCK dewilde_e_Page_34.txt
d5b9de643eee8bda7b625cbb1bf920e1
c9342d40ba2e6a96ad6dc81aa10f5f5844b918cc
41772 F20101129_AABGVN dewilde_e_Page_25.jp2
312e09fa434e9d685e1fff928a00953e
ad21f2789bc9bafe72dcf77003d460ba69d29721
6472 F20101129_AABHHJ dewilde_e_Page_53thm.jpg
fa45619092b25dfff7fde6aa57ab0a7c
62a4e5292f8e6d72c09bf9c05159547a664fde70
510 F20101129_AABHCL dewilde_e_Page_35.txt
f0825fad2cbf9336129c172b59ddd47d
d97051b5642fc6b079bb48efff6ccf8df0e30afc
83755 F20101129_AABGVO dewilde_e_Page_26.jp2
678454d820f3179fe03e5e28ba31943e
7ae5fd348526113b4b3b99ecc1b5488fb06e027c
17457 F20101129_AABHHK dewilde_e_Page_54.QC.jpg
bac27723e80d7c78a81e40916cdc4276
66946ceec4209cccf8f358b80e402ecab0808b43
1590 F20101129_AABHCM dewilde_e_Page_36.txt
7a0e2e03c7108c810c44c99c1f07aaa0
3bd3a5b049e46ffc781f5b7a5898a89d355c752f
106087 F20101129_AABGVP dewilde_e_Page_27.jp2
7f2e2f55ae7b61827df5eac0b20bc41e
07450a3835c70e52990aa8e6ff2eabd609ba124c
5087 F20101129_AABHHL dewilde_e_Page_54thm.jpg
b0d1b212b2f3085e2c141fe658de892a
ed5209fc0f154ea8a2182d73b0f2a4013f850c5f
609 F20101129_AABHCN dewilde_e_Page_37.txt
d9449ffb9aea40e1a759dd51a492a20b
0dc0bf5b6a79c29581f8746a73afa897d4db9b44
109908 F20101129_AABGVQ dewilde_e_Page_28.jp2
48bfa40c51c63b3a4ba9a8c104cfad73
f02445100122c50fa50691892b2bcee465dc970d
1823 F20101129_AABHCO dewilde_e_Page_38.txt
43b9509b8f95f4ad8bfed244dc6d3b56
2d6a4424ba71c37c3bab9b8b978f921cd18d6507
103195 F20101129_AABGVR dewilde_e_Page_29.jp2
dd592a07a04d1f452170ff96105f56e7
637ccf0afe6ec80e0faf250a47ec6d896579492a
21036 F20101129_AABHHM dewilde_e_Page_55.QC.jpg
8f4c6919b9d35cde5570f4bafea3df68
142d1da821e56ac4fc790d523108cbff83c45e5d
2045 F20101129_AABHCP dewilde_e_Page_39.txt
3c164a39667edb332e442cf701fadf28
d8dc6ee42e62f821452470ff83c0fa92887cea94
105120 F20101129_AABGVS dewilde_e_Page_30.jp2
c4b1524b1870e6ba52946a8826642a2c
bae91bc71fa26804a828922741f73e875677871d
5842 F20101129_AABHHN dewilde_e_Page_55thm.jpg
89c5b9ecf7173c8fa82feaf61f7134fc
8fea7fc15bfdfcb1d2f3fd0094cf1011462fc6e7
1870 F20101129_AABHCQ dewilde_e_Page_40.txt
f68bd34b86a5554a84c2de25afd12bca
bf37e8eb2093e082bc10b5cee6b618a2297b45f1
103014 F20101129_AABGVT dewilde_e_Page_31.jp2
ea9a155038bf33797d15104df7126569
ee90c12cbd9ed09ae18b1b423ef1bf071fae8705
1653 F20101129_AABHHO dewilde_e_Page_56thm.jpg
ca6d2885b2a941cc1f5ce7065f227f72
0afc5fd134bd531d4a487764bfda7ed3145f209e
1991 F20101129_AABHCR dewilde_e_Page_41.txt
20b899bf498a426b5a5f32cd47f32955
d1141098d57e7746d37dbe65f2686bf271ac33d2
104133 F20101129_AABGVU dewilde_e_Page_32.jp2
4cb43a212d42117fd7a4750a79c7169b
bbf977d6c661d25633e0611b9b6b934496b82e5d
22578 F20101129_AABHHP dewilde_e_Page_57.QC.jpg
d915f15abf3d16a4d4dae918183e5f85
7d05ae35998809cf47ebd96f7cccee2242c1ef54
185 F20101129_AABHCS dewilde_e_Page_42.txt
f2f1a315631ea65fa942ef7c60cbc0de
c79d62a5acfff93b153ecc3007d052595fda2b1c
105607 F20101129_AABGVV dewilde_e_Page_33.jp2
d55a130c8996e6ff31a73d7679fda3cd
e25710d3b00db0e73b2f59be79af54984341d6b4
5994 F20101129_AABHHQ dewilde_e_Page_57thm.jpg
b27dbd2a62e1d9b54c8cfef1356a1230
40ce9be7c014137d723285b539cdbf31003a20fd
498 F20101129_AABHCT dewilde_e_Page_43.txt
7436b3c09e9240c12412f79df2a6676e
963355a68ae66a2ba58e08bf19b0b6062567d3db
10712 F20101129_AABGVW dewilde_e_Page_34.jp2
d27d4b73dab5d3ebf4116cedfec900aa
0244cbb897bb2e634b1d1d1c9393cc5a6f41852f
20727 F20101129_AABHHR dewilde_e_Page_58.QC.jpg
28c19af9823dbf22c31054c896bf0660
107c50e2b182e970796cadaa229f4e42c5e8a4d0
1156 F20101129_AABHCU dewilde_e_Page_44.txt
4e7d2afd8b7a76e12b2c12dc3b8ade40
bb87d15028953c74d3b0f8ca1761907631a7bff0
261829 F20101129_AABGVX dewilde_e_Page_35.jp2
9e051e0e118cf7e5a23e5d640d078dbb
f1e501b9e846b552c1e7249cac05c58939fa5b5e
5458 F20101129_AABHHS dewilde_e_Page_58thm.jpg
6b97b263cd3806a8deb7dafdbab7ca12
d0062c53d3947da9c49dd80e380ff5359496bd26
1891 F20101129_AABHCV dewilde_e_Page_46.txt
320c88497b23039a9db1a011b4adfced
27111363da1d5c38fba4d4fc803539bdb9260dad
66238 F20101129_AABGVY dewilde_e_Page_36.jp2
eba46d22773066d8a40ca0ba34611ddf
1b09b0379665bb3d3c346506faf1301aedbb0d8e
19723 F20101129_AABHHT dewilde_e_Page_59.QC.jpg
d60dda533e0b7326840ce17b64481ec4
d6584c8f1ad552bd4c85faff6676482e90649065
1841 F20101129_AABHCW dewilde_e_Page_47.txt
f92344e1ae73e77765a7e56734fa1cf3
f789dd94ca5d8e4385348bf28fe24b5cfc12b5fe
57480 F20101129_AABGTA dewilde_e_Page_14.jpg
25de2840f9c3f5ea779b2628a0f3e785
a67ed7881154abedae866f1cac96b738323803ef
28042 F20101129_AABGVZ dewilde_e_Page_37.jp2
e151d09443faaa95ba4f6e9900d3c1e9
fdbb17251a0e63aa20565127f5ae3895914d17cb
5679 F20101129_AABHHU dewilde_e_Page_59thm.jpg
d504458bbf085b0d9f7a100007304db6
cb85eafbdc7175bc2a4af15d3a0a30cdd85db021
F20101129_AABHCX dewilde_e_Page_48.txt
b021b04be67e2d9ef5f5e2e081af3c8a
66500f0ece51da3dc1aa345406bbf46f3ba5e335
72553 F20101129_AABGTB dewilde_e_Page_15.jpg
a4597f7aecbc288d44239be9b7c7a3a1
eb84fc1000fb6ac5f2a7fb981ac3bf8af690dea7
70581 F20101129_AABHHV UFE0011080_00001.mets
986307b518d85375658499f1127396e8
f46e7191cb13ec5c9ddee956d7e3c9f4018203c4
50838 F20101129_AABHAA dewilde_e_Page_27.pro
f8fa0f679628d4c62172fa083a76bc79
a600c21584a04c666a643b1c584c8009242b4088
939 F20101129_AABHCY dewilde_e_Page_49.txt
1524e912952ed054a60921e73625a67b
1a27fdee164c0ea65c9554fd47f31d178d662bdc
73128 F20101129_AABGTC dewilde_e_Page_16.jpg
0c15951b099353b6f5384a6245098b7b
071da4ce24e8ec5261f7455f8291c6909c8b5310
51061 F20101129_AABHAB dewilde_e_Page_28.pro
96ab396668f3789fb7a75585aebb894f
4f6afa231ae6ac10b259db977aa46c93cdebcbd4
1888 F20101129_AABHCZ dewilde_e_Page_50.txt
d067cd172b667d7b046566bd08c8f5e6
55b3c8e7c3d1f414b53dbe92803adaea5f240cc6
74816 F20101129_AABGTD dewilde_e_Page_17.jpg
649808bef8dac82067d99d462e3d077f
9cb2a7058c30f5d23e62efc4e137b1fae62c44a8
F20101129_AABGYA dewilde_e_Page_33.tif
68a8e165be6c75979db3e46ab19036c4
9d34285b23af1732bf13bd9ac6fa4fa83132a3e1
47302 F20101129_AABHAC dewilde_e_Page_29.pro
23c89273e5a3d49836a75ca6b836231e
5161c792da71adea73585f916c69fecc791bad75
72992 F20101129_AABGTE dewilde_e_Page_18.jpg
80b28b334aa1580149b7ceefc9b1e68a
f72334957ac1808281083c960eb393287b44146e
F20101129_AABGYB dewilde_e_Page_34.tif
e31b2f1e1315f51257e7dfdaf00b8a1e
84b7e14f3205d10799b7c8cfbe4a76973fe4fd0f
50012 F20101129_AABHAD dewilde_e_Page_30.pro
db4c1f464ee482b026e971fefbcc7319
a42b0bb67209e3e65d5bf990913ea2ff3d84590c
68241 F20101129_AABGTF dewilde_e_Page_19.jpg
858e8fe05f07779da32ba7c82b07c5e8
9d8c21591635fbbb6e2b302c0e29c1a9d8b633ec
23234 F20101129_AABHFA dewilde_e_Page_23.QC.jpg
f9db3190f841312004f6188969304e98
0289d0e34dbaef66f57f88437ea766c8d2b79f85
F20101129_AABGYC dewilde_e_Page_35.tif
a61e6126fbb4240acf1e10d76a7702a0
01febf9f52ac7e99031f71d363a6481e20b42a45
70587 F20101129_AABGTG dewilde_e_Page_20.jpg
02283e04f5f04c6d6a7fdec477fc80fb
a22b8b023393d047dac600826a36fdbff4286b0b
6504 F20101129_AABHFB dewilde_e_Page_23thm.jpg
216831c78d7c2b3c179d79e5da288564
6ee6278cfd8ef2b5cd7d9e6288e29d7ad52775ba
F20101129_AABGYD dewilde_e_Page_36.tif
a0040526b644d60a853c6de2b7118a65
ce2b8362b3786d8b5f3e3f7bbefe0d6948d5c659
47198 F20101129_AABHAE dewilde_e_Page_31.pro
cc9379773f52a52dd56e538e23b905c2
381d8692bac9c0a89ad30a3d53bda1777c8cd53c
73540 F20101129_AABGTH dewilde_e_Page_21.jpg
ef7300dfcbe14da656ac352d356d113e
7f50e9bd85d4235c808afbb00f491073649a4ba3
19772 F20101129_AABHFC dewilde_e_Page_24.QC.jpg
9884b264053d1ce9d8f46b853136fa27
bb4486b0a0558ad2a8b83113da3fabe7272c3bdd
F20101129_AABGYE dewilde_e_Page_37.tif
80700bc55bdaddc808ea3166bc50806e
8bcd8738312c59fa8aa1b6881822dccf861da4a0
70882 F20101129_AABGTI dewilde_e_Page_22.jpg
597a1583100afea1c88a00dd07d04022
e6447c62e69766307961a870b2c2b9c2353c9626
5546 F20101129_AABHFD dewilde_e_Page_24thm.jpg
1621b0f3a223b62ad41678b26cac4bed
0c01b48af5177bcd74dd46d263db3d6b9e6d9655
F20101129_AABGYF dewilde_e_Page_38.tif
2bd626925d94e4dd91ea79ecfcf5c362
fc9dab495f84297e5e7a4d0c8130b070b61e45a4
48202 F20101129_AABHAF dewilde_e_Page_32.pro
e2d0f75167c1957ad3ec7a6ec0e3c64f
9d0eed2d2fa800bffdbc85c13b5a54a7ec02f87f
70523 F20101129_AABGTJ dewilde_e_Page_23.jpg
1e91893d584cd21f7e2b90b1956b8854
4e024707567336419f610074dd606662e91d4653
10023 F20101129_AABHFE dewilde_e_Page_25.QC.jpg
001b67474a7c40e84e8af9e18fde477e
a0f674cf73585056b1ccc44e58b22368a9c166aa
F20101129_AABGYG dewilde_e_Page_39.tif
1629643edef591185aa42800218d4be1
6d7da71db2a780b92d9cacd49b7636908d37dca0
49960 F20101129_AABHAG dewilde_e_Page_33.pro
dab0c34a049e2ae51a74f245dfa72f2c
5cc2d5c400223f7d270656b07e022ef3aef44525
59461 F20101129_AABGTK dewilde_e_Page_24.jpg
3ffc07151981985d363a3fa01eaed316
1f0ffbab9f7f372a89cd63cd06677ee25406288f
2934 F20101129_AABHFF dewilde_e_Page_25thm.jpg
16b88aa2ac23d9461c14df369f558e08
89325445fcb18e97042d381e5fe7d981f01da884
F20101129_AABGYH dewilde_e_Page_40.tif
a8810a2836a06d0c32b08735cff4ae52
880dde3c6070c8526c1bc7c8cb6a828bdcccc7d9
3403 F20101129_AABHAH dewilde_e_Page_34.pro
1531b59a9c1763e21ae43cd6fad65beb
dca6a2c94070ac6102d5cb0a1b3d8b0b2d3b9a33
33049 F20101129_AABGTL dewilde_e_Page_25.jpg
71c851ca21a64fafd81837d6e84098cb
443d342523a907a7e511b1037d7f9bb83bbb34ce
17817 F20101129_AABHFG dewilde_e_Page_26.QC.jpg
b2008e74fb160dfe7e545056897e801a
ceab236dd7d7b873d77717a173c6decd6e3838d5
F20101129_AABGYI dewilde_e_Page_41.tif
933a44191809d7ceefe6be77f148990e
b5b300e41a6355aed2301f4db9dcea2d0fe42935
8931 F20101129_AABHAI dewilde_e_Page_35.pro
4f990f3270023b21f9d6a827c64f4a0a
587408824ab7ecdd739c7a27d2142581e970f43e
56425 F20101129_AABGTM dewilde_e_Page_26.jpg
3fa1f396acfc555ebc59b4bcc3c38f64
23806a99c954521e3a7cbcc2781e631a04107ff2
5320 F20101129_AABHFH dewilde_e_Page_26thm.jpg
1e8df84f707e71b17382cb6b31eff17f
a71e4c1b60060b9bf09ef6e5a3eba79e9ecf344e
F20101129_AABGYJ dewilde_e_Page_42.tif
3c5a5d9fdd7f7986216045fe784aa46d
61dba8e89e3bbbb24a220460620153fbd96775ac
12823 F20101129_AABHAJ dewilde_e_Page_37.pro
71de584b67382cad7d0f2e1dbc6ebb58
5371e6f97bb2bf4b9a97bdc948c37539ee3fba5b
70818 F20101129_AABGTN dewilde_e_Page_28.jpg
15bd16557380d4872635107ed0d0a976
f59d8643a2d680dd0c4bdd32b84c1f4cb65123a4
23077 F20101129_AABHFI dewilde_e_Page_27.QC.jpg
2304b9918fa163105cd1d63e3420f639
b939a73383b6a1e837930c03023b150f10e3d55f
F20101129_AABGYK dewilde_e_Page_43.tif
64c4a59b80f957841cb491f1d4be5106
556a02606041136047c569333d49695086a0586d
43683 F20101129_AABHAK dewilde_e_Page_38.pro
fdcb5e0bd914f1cdacf9f447e11ac81f
35423bc4a50f56979aadaec198d4e4656328eab2
68602 F20101129_AABGTO dewilde_e_Page_29.jpg
372c479a44a80ebe667db82024c9a719
76d89156dd4b105214fe6c1e5badeef78599b531
6516 F20101129_AABHFJ dewilde_e_Page_27thm.jpg
ca306c7723cf69c67511d14434d85f8c
415a2719001bd0b07b4a71cee7d8af020808ea5c
F20101129_AABGYL dewilde_e_Page_44.tif
bb2651d72ece9da6cefba308d7482018
0e7e451849b698db52807930d1df2f358ecac8b4
52154 F20101129_AABHAL dewilde_e_Page_39.pro
d7029bd0b5bfbe95e0a2408efc3068e6
b1cfa261277d2c6363a0385fe08798faba26dbe3
69614 F20101129_AABGTP dewilde_e_Page_30.jpg
720cc77c431046294ccc80a66d42fcc2
120c205e8e5c06f0e68edbeec79c2786af617186
F20101129_AABGYM dewilde_e_Page_45.tif
93dc20d1a60859f7c7d84c302d9c38b8
ccf4784abb61fe656de775307ff04fcc8a96a093
46302 F20101129_AABHAM dewilde_e_Page_40.pro
3b4f5b4c28ad8f00e1f85fcd58c0f1bc
b4dee1153a416742642f536221845e69ce4a75f9
67221 F20101129_AABGTQ dewilde_e_Page_31.jpg
8dbf14d75ed45c549e9528f28e50cf46
029a627156e545c8d35471801d7be700125d512b
23225 F20101129_AABHFK dewilde_e_Page_28.QC.jpg
a7681525c2325f2a8b1fbd5b7479f668
30cfe7d3a65344083094113d02fd97b33f4eee78
F20101129_AABGYN dewilde_e_Page_46.tif
70f1b0ef7cb262fb7302c877773925d4
c73f8d223d02cde67b9ade3e2141a258c09d87d6
49649 F20101129_AABHAN dewilde_e_Page_41.pro
70958349b3c1b69e0d4b7f40128f633e
4518ee43345713c76c99eba53c19a28fee3fc9c2
68481 F20101129_AABGTR dewilde_e_Page_32.jpg
b19d150c8951993fe39c541c25db083c
66a454092ac7a52e8679a4262c087b172ae04a05
6358 F20101129_AABHFL dewilde_e_Page_28thm.jpg
46c731ead3340b8480997aa4873eb1b0
0bb6619312d4b7ff80a1db4f427b5e0025f4b603
F20101129_AABGYO dewilde_e_Page_47.tif
cc813fd218fede261c5e3ec43be4a451
29a29f7d19f6437018b051a95e3755a3450207a2
3507 F20101129_AABHAO dewilde_e_Page_42.pro
51d6f56dc4b3ed40ff30db2c2632a065
602fbb7fc2ce42dbf8803ae731743b9d354f98c7
70534 F20101129_AABGTS dewilde_e_Page_33.jpg
e931d56ed1c925ce0b57b1649e60ec0e
47eda527d73cfffff6162cc87f554721389913bc
22059 F20101129_AABHFM dewilde_e_Page_29.QC.jpg
c1c274136e8fdddd1fdf19816a5337ee
4157f1210169089a4a6c5528f87d7d39312ca6b8
F20101129_AABGYP dewilde_e_Page_48.tif
f6c03e4f69ec239c08383677655402b7
43a0d8522374144f9517bd678a96c05cad124650
10093 F20101129_AABHAP dewilde_e_Page_43.pro
66b374a28debe0a0267269aa01a17293
2e2105ba6711145107620d8dd4d05dca6762b368


Permanent Link: http://ufdc.ufl.edu/UFE0011080/00001

Material Information

Title: Attentional Contrast During Sequential Judgments: An Explanation for the Number-of-Levels Effect
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0011080:00001

Permanent Link: http://ufdc.ufl.edu/UFE0011080/00001

Material Information

Title: Attentional Contrast During Sequential Judgments: An Explanation for the Number-of-Levels Effect
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0011080:00001


This item has the following downloads:


Full Text












ATTENTIONAL CONTRAST DURING SEQUENTIAL JUDGMENTS:
AN EXPLANATION FOR THE NUMBER-OF-LEVELS EFFECT

















By

ELS DE WILDE


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2005

































Copyright 2005

by

Els De Wilde
































This dissertation is dedicated to my parents.















ACKNOWLEDGMENTS

This dissertation leaves me indebted to many people. First, I would like to thank

my advisor, Chris Janiszewski, for taking up the mentorship role and guiding me through

the dissertation process. Interactions with other members of my committee were also

invaluable in shaping this dissertation. In particular, I would like to thank Alan Cooke for

his conceptual guidance and Richard Lutz and Asoo Vakharia for their feedback. My

thanks also go to my friends in Belgium and in the doctoral program and my colleagues

at HEC Montreal for their support. I would like to thank JoAndrea Hoegg and Andrew

Kuo for their assistance with the data collection while I was at long distance. I am

extremely thankful to my family members for their encouragement throughout the entire

duration of my doctoral studies. Finally, I would like to thank Roberto for giving me the

best reason to finish this dissertation.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ......... ................................................................................... iv

LIST OF TABLES .............. ............. .......... ............ vii

LIST OF FIGURES ........................ .. .................... viii

ABSTRACT ........................ ............... ................... .........ix

CHAPTER

1 IN TR O D U C TIO N ......................................................................... .... .. ........

2 THEORETICAL BACKGROUND....................................................................4

The Number-Of-Levels Effect.................................. .............................. 4
Existing Accounts of the Number-Of-Levels Effect ................................................4
A ttentional C ontrast........... ................................................................ .......... .. ..

3 EXPERIMENT 1: DEMONSTRATION OF THE NUMBER-OF-LEVELS
EFFECT WITH EXPANDED DESIGN .............................................................16

M eth o d ............................................................................................. 16
R e su lts ...................................... .................................................... 1 7
D isc u ssio n .............................................................................................................. 2 0
E x p e rim e n t 1A ............................................................................................................ 2 1

4 EXPERIMENT 2: TEST OF ATTENTIONAL CONTRAST ACCOUNT WITH
NO DIFFERENCE IN NUMBER OF LEVELS .............. ................................ 28

M eth o d ..............................................................................................2 8
R e su lts ...................................... .................................................... 3 0
D iscu ssio n ...................................... ................................................. 3 1

5 EXPERIMENT 3: TEST OF ATTENTIONAL CONTRAST ACCOUNT WITH
REVERSED NUMBER OF LEVELS.................................................................35

M eth o d ............................................................................................. 3 5
R e su lts ...................................... .................................................... 3 6
D discussion .................................... ................................... .......... 38









6 G E N E R A L D ISC U SSIO N ................................................................ ....................4 1

7 LIMITATIONS AND FUTURE RESEARCH ................................. ...............45

L IST O F R E F E R E N C E S .......................................................................... ....................47

B IO G R A PH IC A L SK E TCH ...................................................................... ..................49
















LIST OF TABLES

Table p

1. Number-of-levels manipulations used in published studies .......................................15

2. A tribute levels from experim ent 1 ........................................... ......................... 26

3. Mean difference scores and simple effects for experiment 1 .............. ...................26

4. Mean difference scores and simple effects for experiment 1A...................................27

5. Mean difference scores and simple effects for experiment 2 .....................................34

6. A tribute levels from experim ent 3 ........................................... ......................... 40

7. Mean difference scores and simple effects for experiment 3 .....................................40
















LIST OF FIGURES

Figure page

1. D esign for experim ents 1 and 1A ....................................................... .....................25

2 D design for experim ent 2 ......................................................................... .................. 33

3. Design for experiment 3............. ................................................................39















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


ATTENTIONAL CONTRAST DURING SEQUENTIAL JUDGMENTS:
AN EXPLANATION FOR THE NUMBER-OF-LEVELS EFFECT

By

Els De Wilde

August 2005

Chair: Chris Janiszewski
Cochair: Alan D. J. Cooke
Major Department: Marketing

Normatively, the importance of an attribute does not depend on the number of

levels on which it is defined, exterior levels being fixed. This dissertation examines a bias

identified in the conjoint literature known as the number-of-levels effect, which occurs

when consumers' derived importance weights increase with the number of attribute

levels. Borrowing from research on attention during category learning, this dissertation

examines psychological mechanisms that underlie the number-of-levels effect.

In contrast to earlier attentional accounts that have been offered for the bias, I

propose that the weighting effect found in the number-of-levels studies is due to the

attentional processes that are inherent to the sequential evaluation procedure.

Analogously to attentional processes during category learning, consumers divide their

attention between attribute dimensions and attribute levels when making sequential

evaluations. Within a given trial, a mechanism of attentional contrast directs attention









away from attribute levels that have been associated with an outcome. I suggest that the

number-of-levels effect is caused by this attentional contrast strategy. When attributes

differ on the number of levels on which they are defined, they also differ on the number

of times a given attribute level has been associated with a judgment (i.e., attribute level

novelty). A process of attentional contrast away from the more familiar attribute level

within each offer results in differential importance weights for the attributes.

In a series of experiments, I examine this attentional contrast account for the

"number-of-levels" effect. Consistent with this account, the weighting effect is shown to

appear when attentional contrast is created in a situation in which there is no difference in

number of levels. Moreover, the number-of-levels effect reverses when attentional

contrast is created on the attribute with the lower number of levels.














CHAPTER 1
INTRODUCTION

Normatively, the importance of an attribute should not depend on the number of

levels on which it is defined, exterior levels being fixed. Yet, several studies from the

conjoint literature show that derived importance weights increase with the number of

attribute levels, a bias which has generally been labelled the number-of-levels effect (e.g.,

Currim, Weinberg and Wittink 1981; Verlegh, Schifferstein and Wittink 2002; Wittink et

al. 1992b). In searching for a method to eliminate this bias, several accounts have been

offered for it. Methodologically, there is evidence that the number-of-levels effect is

related to the data collection method (e.g., Wittink et al. 1992b), the measurement scale

for the dependent variable (e.g., Steenkamp and Wittink 1994; Wittink et al. 1992b), and

the parameter estimation procedure (Wittink, Krishnamurthi and Reibstein 1989).

Although tests of these methodological accounts reduced the number-of-levels effect,

they have never completely eliminated the bias.

The possibility that psychological processes are at least partly responsible for the

number-of-levels effect has been acknowledged since the effect was first documented.

Currim et al. (1981) state that "the respondent may react to the number of levels used for

a given attribute by consciously or unconsciously weighting an attribute more heavily as

the number of levels increases" (p. 72). Although a test of this account has been included

in several studies, it has not received much support during the past two decades. An

assumption underlying tests of this attentional account is that there should be a number-

of-levels effect not only in sequential judgments or choices (e.g., full-profile rank orders,









full-profile paired comparisons, full-profile ratings), but also in self-explicated priors

(Wittink et al. 1992a) and stated importance measures (Steenkamp and Wittink 1994;

Verlegh et al. 2002).

However, I propose that the weighting effect found in the number-of-levels studies

may be due to the attentional processes that are inherent to the sequential evaluation

procedure. Studies in which people learn new categories by making sequential

classifications distinguish between attention allocated to attributes and to attribute levels

(Kersten, Goldstone and Schaffert 1998). They show that within a given trial, a

mechanism of attentional contrast directs attention away from attribute levels that have

been associated with an outcome. I propose that this attentional contrast strategy can at

least partly account for the number-of-levels effect. When attributes differ on the number

of levels on which they are defined, they also differ on the number of times a given

attribute level appears across offers during a sequential judgment task (henceforth

referred to as relative attribute level novelty). More specifically, the levels of attributes

that are defined on more levels are overall relatively more novel than the levels of

attributes that are defined on fewer levels. A process of attentional contrast away from

the less novel attribute level within each offer may result in differential importance

weights for the attributes.

This dissertation proceeds as follows: Chapter 2 discusses the number-of-levels

effect and the methodological and psychological accounts that have been offered for it in

the literature. It also introduces the attentional contrast account from the category

learning literature. In chapters 3 to 5, I examine if attentional contrast can account for the

number-of-levels effect. In chapter 3, I extend the designs that have typically been used






3


in studies that examine the number-of-levels effect and find evidence for an account in

terms of attentional contrast. In chapter 4, I manipulate attentional contrast directly to

examine its possible impact when there is no difference in number of levels. In chapter 5,

I create attentional contrast on the attribute with the lower number of levels. Finally,

chapter 6 discusses the findings and chapter 7 examines the limitations and avenues for

future research.














CHAPTER 2
THEORETICAL BACKGROUND

The Number-Of-Levels Effect

When an attribute has a higher number of levels (exterior values being fixed),

derived importance weights for that attribute increase. For example, Currim et al. (1981)

examined subscription series to performing arts events, using three attributes that were

defined on two levels and three attributes that were defined on three levels. They asked

participants to make trade-offs between pairs that were defined on two attributes: a two-

level attribute and a three-level attribute. The two-level attributes had derived importance

weights between 0.36 and 0.45, whereas the three-level attributes had derived importance

weights between 0.55 and 0.66.

Originally identified between attributes using part-worth estimation within conjoint

analysis, this number-of-levels effect has been replicated using a variety of data

collection methods, measurement scales for the dependent variable, and estimation

methods. Accounts for the number-of-levels effect include methodological and

psychological factors. I will first review the existing accounts for the number-of-levels

effect and then propose a new account based on attentional contrast from the category

learning literature.

Existing Accounts of the Number-Of-Levels Effect

The greater share of research on the number-of-levels effect has dealt with

methodological accounts. This research attributes the number-of-levels effect to the









properties of the measurement scale, the usage of the measurement scale, or the method

of parameter estimation.

Properties of the Measurement Scale

Several accounts of the number-of-levels effect relate to the properties of the

measurement scale. Wittink, Krishnamurthi, and Reibstein (1989) hypothesized that the

number-of-levels effect was partly due to the use of a rank-order measurement procedure.

They showed analytically why the effect occurs for rank-order preferences using least-

squares analysis on a multitude of full factorial and fractional factorial designs. Since it is

impossible to analytically show the same effect exists for ratings, Wittink et al. (1989)

examined whether the number-of-levels effect exists for rating scales empirically,

comparing the results from full-profile ratings with the results from rank-order

preferences obtained by paired comparisons and trade-off matrix rank orders. The full-

profile preferences were obtained using a ten-point scale. Wittink et al. (1989) compared

derived importance of a three-level and five-level price attribute in five categories and

found that the number-of-levels effect was equally strong using rank-order or rating scale

measurement procedures, the implication being that rank-order measurement was not the

sole source of the number-of-levels effect.

Steenkamp and Wittink (1994) proposed that the number-of-levels effect occurs

with a rating scale measurement procedure because the preference ratings lack metricity.

They hypothesized that rating scales' properties are closer to ordinal than to interval

measurement. Since the method of magnitude estimation has been claimed to result in

interval-scale data, Steenkamp and Wittink (1994) used this method to test their

hypothesis. The method of magnitude estimation consists of asking participants to match

their evaluation of a profile to the perceived magnitude expressed on a physical medium.









Steenkamp and Wittink (1994) used numerical estimation (i.e., participants assign

numbers to profiles) and line production (i.e., participants draw lines) as response

modalities. They found that a nine-point rating scale and a magnitude estimation

procedure provide number-of-levels effects of comparable size. However, the magnitude

estimation procedure allows for a test of the metricity of the results of each respondent

separately. Categorizing the respondents as metric versus non-metric, Steenkamp and

Wittink (1994) found a reduced number-of-levels effect in participants whose responses

satisfied the criteria for metricity. In other words, the lack of metricity of the rating

method per se had no impact on the number-of-levels effect, but individual differences in

metricity were identified as a source of the effect.

Usage of the Measurement Scale

Other measurement procedure explanations relate to respondents' use of the

measurement scale. Wittink et al. (1992b) included a balancing manipulation in their

study (i.e., they constructed a condition in which dominated pairs were eliminated from

the procedure), provided respondents made pair-wise comparisons among alternatives.

Using adaptive conjoint analysis, Wittink et al. (1992b) gathered full factorial data on one

product category (notebook computers) defined on six attributes, of which four attributes

alternated between two and four levels, depending on the condition. Wittink et al.

(1992b) found that the number-of-levels effect is considerably reduced when dominated

pairs are eliminated from the procedure. Wittink et al. (1992b) explain this result in terms

of participants' tendencies to give preference ratings toward the midpoint of the scale.

Since preference ratings for dominated pairs are expected to be more extreme than those

for non-dominated pairs, the exclusion of dominated pairs provides less opportunity for

distortion.









Verlegh et al. (2002) propose the number-of-levels effect may be a consequence of

a person's tendency to perceive a constant, absolute distance between levels of an

attribute (i.e., exhibit a uniform distribution tendency). They tested this hypothesis by

comparing results on five different measures of attribute importance, using one product

category (color TVs). The product category was defined on six attributes, of which two

alternated between two and four levels, depending on the condition. Verlegh et al. (2002)

found significant number-of-level effects on two measures of attribute importance:

difference in part worths in a full-profile conjoint and attractiveness ratings for individual

attribute levels. On the remaining three measures (stated importance of a set of attributes,

linear regression coefficients, and paired comparisons varying one attribute's exterior

levels at a time), no significant number-of-levels effect was observed. Verlegh et al.

(2002) interpreted their results as evidence for respondents' tendency to distribute

stimulus representations uniformly over a restricted internal continuum (i.e., the distance

between each level of an attribute was constant) in full-profile and attribute level ratings.

Ceteris paribus, the addition of intermediate levels will lead respondents to distribute the

exterior attribute levels more toward the extremes of the internal continuum.

Method of Parameter Estimation

A second potential source of the number-of-levels effect is the method of parameter

estimation. In their seminal paper, Currim, Weinberg, and Wittink (1981) mentioned the

method of parameter estimation as a possible source of the number-of-levels effect. For

the 2x3 trade-off matrix they used to obtain their empirical results, they estimated the

parameters based on permissible configurations of preferences using ordinary least

squares. They derived the importance weights for two extreme cases: a dominant two-

level attribute and a dominant three-level attribute. They found that the two-level









attribute's importance weights varied between 0.20 (i.e., when the three-attribute level

attribute was dominant) and 0.60 (i.e., when the two-level attribute was dominant).

Analogously, the three-level attribute's importance weights ranged from 0.40 (i.e., when

the two-level attribute was dominant) to 0.80 (i.e., when the three-level attribute was

dominant). In other words, while the range of possible derived importance weights was

not affected by the number of levels, its size was affected by it. Currim et al. (1981) used

this analytical evidence as support for their hypothesis that the method of parameter

estimation is a source of the number-of-levels effect.

Subsequent studies have provided additional support for the method of parameter

estimation account of the number-of-levels effect. Wittink et al. (1989) showed

analytically why the number-of-levels effect occurs on rank-order preferences on a

multitude of full factorial and fractional factorial designs. Assuming an additive, main-

effect model, they calculated raw attribute importance weights for any number of

attributes and any attribute levels using least-squares analysis. Based on the raw attribute

importance weights, they estimated maximum relative weights for each attribute, and the

differences and ratios between the attributes within a design. They found that maximum

relative weights increase with the number of levels and concluded that the method of

parameter estimation is a source of the number-of-levels effect. Wittink, Krishnamurthi,

and Nutter (1982) gathered participants' preference data for hypothetical summer jobs,

defined on four attributes, of which two attributes alternated between two and four

attribute levels, depending on the condition. Two data collection methods were used: a

fractional factorial profile evaluation and a trade-off matrix. Wittink et al. (1982)

estimated the parameters using a non-metric method (LINMAP, Srinivasan and Shocker









1973) and found that the derived importance weights vary with the number of levels on

which an attribute is defined. They concluded that the number-of-levels effect is at least

partially accounted for by the method of parameter estimation.

Perceptual Processes

Since the initial identification of the number-of-levels effect, researchers have

acknowledged that the effect could be a consequence of perceptual processes (Currim et

al. 1981). More specifically, it has been proposed that respondents may consciously or

unconsciously assign more weight to attributes with more levels. Several studies have

included one or more tests for this weighting explanation, but have failed to find support

for the hypothesis (Steenkamp and Wittink 1994; Verlegh et al. 2002; Wittink et al.

1992b). For example, Verlegh et al. (2002) hypothesized that increased attention should

affect the stated importance of a set of attributes and paired comparisons varying one

attribute's exterior levels at a time. However, no number-of-levels effect was observed on

these two measures.

Attentional Contrast

Several of the proposed accounts of the number-of-levels effect can partially

explain the effect. However, no single explanation can wholely account for the number-

of-levels effect. I propose that attentional contrast may be an additional source of the

number-of-levels effect.

The proposed attentional contrast explanation is based on two observations. First,

the occurrence of the number-of-levels effect depends on the preference elicitation

method. Decompositional methods ask respondents to provide multiattribute judgments

for the full-profile descriptions of alternatives and then derive individual attribute level

values from the judgments. The number-of-levels effect is typically found using









decompositional methods. Compositional methods ask respondents to assign values to

each level of an attribute. The values are combined to construct an overall evaluation of

an alternative. The number-of-levels effect is typically not found using compositional

methods.

There are three studies that have used a combination of decompositional and

compositional methods. Wittink et al. (1992a) showed that the magnitude of the number-

of-levels effect was greater for a full-profile procedure than for an adaptive conjoint

analysis. Whereas full-profile procedure is an entirely decompositional method, adaptive

conjoint analysis is a combination of a decompositional and compositional procedure. No

number-of-levels effect was observed in the compositional data. Verlegh et al. (2002)

used a decompositional and compositional procedure and found evidence for a number-

of-levels effect in the compositional data. Yet, the compositional data were collected

after the decompositional data, suggesting the decompositional procedure may have

biased the values provided in the compositional procedure. Steenkamp and Wittink

(1994) also found minimal evidence for a number-of-levels effect on stated attribute

importance that were gathered after the decompositional data.

Johnson (1992) found evidence for a number-of-levels effect using only a

compositional method. He asked participants to assess attribute values directly. However,

the values of two attributes were assessed at the same time. More specifically,

participants were told that a base TV with 17 inch screen and monophonic sound was

worth $200. Then, they were asked to evaluate what other options were worth: 19 inch

screen, 21 inch screen and excellent stereo sound in the 3+2 condition and 21 inch screen,

good stereo sound and excellent stereo sound in the 2+3 condition. Hence, participants









could interpret this task as full-profile evaluation. For example, when evaluating the 19

inch screen, it is likely that they imagined a monophonic sound TV.

Second, the designs used in studies on the number-of-levels effect invariably

compare conditions in which the number of levels of more than one attribute is varied at

once (i.e., varying the number of levels of attribute 1 is confounded with varying the

number of attribute levels of attribute 2). For example, consider an experiment in which

participants respond to a 2x4 design (i.e., attribute 1 has two levels and attribute 2 has

four levels) or a 4x2 design. If one finds that the derived importance weight on attribute 1

is larger in the 4x2 condition than the 2x4 condition, then the weighting difference may

be due to the increased number of levels in attribute 1 or to the accompanying decrease in

the number of levels in attribute 2. In other words, the weighting effect that has been

interpreted as a number-of-levels effect may not (solely) be caused by the increased

number of levels in that attribute. It may also be caused by the reduced number of levels

of other attributes. It is important to note that all published demonstrations of the number-

of-levels effect have this characteristic built into their manipulations. An overview of the

designs can be found in Table 1.

Based on these two observations, I will now propose a new account for the number-

of-levels effect: the attentional contrast account. The attentional contrast account depends

on the idea that people direct attention away from attribute levels they have seen more

often and towards attribute levels they have seen less often. For example, in a 4x2 design,

the first attribute has four levels and each of these levels will be experienced twice in a

full-factorial design. The second attribute has two levels and each of these levels will be

experienced four times in a full-factorial design. Thus, the levels of the first attribute will









be relatively more novel than the levels of the second attribute. The attentional contrast

account proposed that attention will contrast away from the more common levels (i.e.,

attribute 2) and toward the less common levels (i.e., attribute 1). Thus, the attribute with

more levels will be perceived as more important.

The attentional contrast hypothesis has its foundation in the categorical learning

literature. The most successful theories in categorization and learning incorporate some

notion of selective attention (Nosofsky 1986, 1991; Sutherland and Mackintosh 1971).

More specifically, consumers increasingly pay attention to attributes and attribute levels

that are important, and decreasingly pay attention to irrelevant attributes, irrelevant

attribute levels, or both (Goldstone and Steyvers 2001). For example, Goldstone and

Steyvers (2001) showed that selective attention can be directed toward arbitrary

dimensions created by randomly pairing faces. They selected four photographs of bald

heads and generated two dimensions by morphing between the photographs. Each of the

sixteen created faces varied in the proportion of each of the four faces. Based on category

relevance, both sensitization of relevant dimensions and desensitization of irrelevant

dimensions were observed. Moreover, Goldstone and Steyvers (2001) showed that once a

participant had learned to direct attention toward or away from a dimension, this learning

extended to different stimuli.

Kersten, Goldstone, and Schaffert (1998) provide additional evidence that

differential selective attention mechanisms operate on attributes and attribute levels. They

show that within a given trial, a mechanism of attentional contrast directs attention away

from attribute levels that have been associated with an outcome. In a series of

experiments, participants learned verbs that were defined on two attributes: path and









manner. Each verb was always accompanied by a particular value of one or both of these

attributes. One group of participants learned conjunctive verbs (i.e., verbs that could be

distinguished on path and manner). Another group first learned path verbs (i.e., verbs that

could only be distinguished on the basis of path). When this second group learned new

conjunctive verbs (i.e., verbs that could be distinguished based on manner as well as the

same values of path), they attended more to level of manner than the participants who

first learned conjunctive verbs. In other words, since the path values had been associated

with an outcome (i.e., a verb), attentional contrast was directed toward the new manner

values within subsequent trials.

The attentional contrast account is consistent with the observation that the

occurrence of the number-of-levels effect depends on whether the method is

compositional or decompositional. In a compositional method, each level of an attribute

is repeated only once and it is repeated out of context, so no attentional contrast can

operate. In a decompositional method, attribute levels are repeated and this repetition

occurs in the context of other attribute levels that have been more or less repeated, so

attentional contrast can operate.

The attentional contrast account also makes specific predictions in the event the

levels of attributes are unconfounded in an experimental design. For example, imagine

three designs are investigated: a 4x2, a 2x4, and a 4x4. The derived importance weight of

attribute one (two) should be greater (smaller) in the 4x2 than the 2x4 studies, as has been

observed in prior studies. However, the attentional contrast account predicts that the

derived importance weight of attribute one should be greater in the 4x2 design than the

4x4 design because the number of levels of attribute two are more frequent in the 4x2









condition. The attentional contrast account also predicts that the derived importance

weight of attribute two should be equal in the 4x2 and the 4x4 conditions because the

number of levels of attribute one are equivalent. Similar predictions can be made when

comparing the 2x4 and the 4x4 conditions.

The predictions of the attentional contrast hypothesis are not consistent with the

predictions that might be expected if one were simply to count the levels of an attribute.

For example, if the predictions were based solely on the number of levels of an attribute,

the derived importance weight of attribute one should be the same in the 4x2 condition

and the 4x4 condition because the number of levels of attribute one are equivalent. The

derived importance weight of attribute two should be less in the 4x2 condition than the

4x4 condition because the number of levels of attribute two is less in the 4x2 condition.

Thus, an explanation based solely on attribute levels makes a set of qualitatively different

predictions than the attentional contrast hypothesis.

In a series of four experiments, I will test this attentional contrast account for the

number-of-levels effect. In experiments 1 and 1A, I will extend the designs that have

typically been used to examine the number-of-levels effect to test an attentional contrast

account. In experiment 2, I will manipulate attentional contrast directly to examine its

potential impact when there are no differences in number of levels. In experiment 3, I

create attentional contrast on an attribute with a lower number of levels.











Table 1. Number-of-levels manipulations used in published studies
Authors Number-Of-Levels Manipulation
number-of-levels manipulation
Currim, Weinberg and Wittink (1981) between attributes (2 or 3 levels)
Johnson (1992) 3+2 vs. 2+3 (used lists)
number-of-levels manipulation
Orme (1998) between attributes (2 to 5 levels)
Steenkamp and Wittink (1994) 2x4 vs. 4x2
Verlegh, Schifferstein and Wittink (2002) 4x2x2x2x3x2 vs. 2x4x2x2x3x2
3x4x2x2x4x2x2x2x2 vs.
Wittink, Huber, Fiedler and Miller (1992a) 3x2x4x4x2x2x2x2x2
Wittink, Huber, Zandan and Johnson (1992b) 2x2x4x4 vs. 4x4x2x2
Wittink, Krishnamurthi and Nutter (1982) 4x4x2x2 vs. 4x2x4x2
Wittink, Krishnamurthi and Reibstein (1989) 3 vs. 5 levels














CHAPTER 3
EXPERIMENT 1: DEMONSTRATION OF THE NUMBER-OF-LEVELS EFFECT
WITH EXPANDED DESIGN

The objective of the first experiment was to test the attentional contrast account of

the number-of-levels effect by expanding the experimental design that has typically been

used to examine the bias. More specifically, the number of levels of only one attribute

was varied across conditions. The attentional contrast hypothesis (Kersten et al. 1998),

predicted that the number-of-levels effect would be sensitive to the relative novelty of

each attribute level.

Method

Design

The design was a 3 cell between-subjects design (4x2, 2x4, and 4x4 attribute

levels) with three between-subject product replicates (refrigerators, LCD televisions, and

digital cameras). Each offer was defined on two attributes. Participants were asked to

provide their reservation price for each product offer. This dependent variable should not

be sensitive to the uniform distribution problem associated with a fixed scale (Verlegh et

al. 2002) or suffer from a lack of metricity (Steenkamp and Wittink 1994). The design of

experiment 1 is presented in Figure 1.

Stimuli

The product categories were refrigerators, LCD televisions, and digital cameras.

Each product offer was defined on two attributes for which participants were known to

have monotonic preferences. The attributes were the following: annual energy cost and









capacity (refrigerators), screen size and horizontal viewing angle (LCD televisions),

imaging sensor resolution and image storage capacity (digital cameras). Each attribute

had two or four possible attribute levels (see Table 2 for an overview of the attribute

levels). These attribute levels were either exterior (i.e., the two values in a 2-level

attribute or the two extreme values in a 4-level attribute) or interior (the two inserted

levels in a 4-level attribute). Exterior offers consisted of the exterior levels of the two

attributes (e.g., [1,1], [1,4], [2,1], [2,4] in a 2x4 design). Intermediate offers consisted of

the interior levels of the two attributes (e.g., [2,2], [2,3], [3,2], [3,3] in a 4x4 design), or

of an exterior level of one attribute and an interior level of a second attribute (e.g., [1,2],

[1,3], [2,2], [2,3] in a 2x4 design).

Procedure

Participants were seated at personal computers. Participants read instructions

stating that they would be stating reservation prices for products described on two

dimensions. Participants were asked to assume that the products were identical on all

non-defined attributes and were informed of the average price of the product. Then, the

product offers were presented. First, the participants stated their reservation price for the

intermediate offers. These four stimuli were presented in random order. Next, participants

stated their reservation price for the exterior offers. Again, the stimuli were presented in

random order. Thus, each participant provided reservation price judgments for eight

stimuli with the last four being the identical exterior stimuli.

Results

The analysis was conducted using the mean difference scores of the exterior offers.

For example, attribute one's differences score in a 2x4 design is the difference between

the average of [1,1] and [1,4] and the average of [2,1] and [2,4]. Similarly, attribute two's









differences score in a 2x4 design is the difference between the average of [1,1] and [2,1]

and the average of [1,4] and [2,4]. The four exterior offers were identical across

conditions, so these difference scores could be compared. The results of participants

whose ratings on the exterior offers were non-monotonic were removed from the dataset.

Table 3 presents the mean difference scores per attribute for each condition.

Refrigerators

Data of 91 participants were retained for analysis. The analysis for each attribute

was performed separately. The first analysis was on attribute one (annual energy cost).

The mean difference scores of the 2x4 condition (X = $78.97) and the 4x2 condition (X =

144.24) were significantly different (F(1, 88) = 10.38, p < .01). This replicates the

number-of-levels effect using a design that is common in the academic literature. The

mean difference scores of the 2x4 condition (X = $78.97) and the 4x4 condition (X =

$63.84) were not significantly different (F(1,88) = 0.63, p > .05). This result is

inconsistent with the hypothesis that the number-of-levels effect is sensitive only to the

number of levels of an attribute. The mean difference scores of the 4x2 condition (X =

$144.22) and the 4x4 condition (X = $63.84) were significantly different (F(1, 88) =

17.16, p < .01). This result is consistent with the attentional contrast hypothesis.

The second analysis was on attribute two (capacity). The mean difference scores of

the 2x4 condition (X = $122.55) and the 4x2 condition (X = $120.17) were not

significantly different (F(1, 88) = 0.01, p > .05). The mean difference scores of the 2x4

condition (X = $122.55) and the 4x4 condition (X = $100.93) were not significantly

different (F(1, 88) = 0.96, p > .05). The mean difference scores of the 4x2 condition (X =

$120.17) and the 4x4 condition (X = $100.93) were not significantly different (F(1, 88) =

0.73, p> .05)









LCD Televisions

Data of 90 participants were retained for analysis. The first analysis was on

attribute one (screen size). The mean difference scores of the 2x4 condition (X =

$176.20) and the 4x2 condition (X = $257.37) were significantly different (F(1, 87) =

3.98, p < .05). The mean difference scores of the 2x4 condition (X = $176.20) and the

4x4 condition (X = $196.82) were not significantly different (F(1, 87) = 0.26, p > .05).

The mean difference scores of the 4x2 condition (X = $257.37) and the 4x4 condition (X

= $196.82) were not significantly different (F(1, 87) = 2.21, p > .05). These results are

not supportive of the number of levels hypothesis or the attentional contrast hypothesis.

The second analysis was on attribute two (horizontal viewing angle). The mean

difference scores of the 2x4 condition (X = $140.64) and the 4x2 condition (X = $130.37)

were not significantly different (F(1, 87) = 0.26, p > .05). The mean difference scores of

the 2x4 condition (X = $140.64) and the 4x4 condition (X = 122.82) were not

significantly different (F(1, 87) = 0.78, p > .05). The mean difference scores of the 4x2

condition (X = $130.37) and the 4x4 condition (X = $122.82) were not significantly

different (F(1, 87) = 0.14, p > .05).

Digital Cameras

Data of 117 participants were retained for analysis. The first analysis was on

attribute one (imaging sensor resolution). The mean difference scores of the 2x4

condition (X = $134.90) and the 4x2 condition (X = $169.31) were significantly different

(F(1, 114) = 3.18, p = .08). The mean difference scores of the 2x4 condition (X =

$134.90) and the 4x4 condition (X = $131.49) were not significantly different (F(1, 114)

= 0.03, p > .05). The mean difference scores of the 4x2 condition (X = $169.31) and the









4x4 condition (X = $131.49) were significantly different (F(1, 114) = 4.07, p < .05).

These results are supportive of the attentional contrast hypothesis.

The second analysis was on attribute two (image storage capacity). The mean

difference scores of the 2x4 condition (X = $110.42) and the 4x2 condition (X = $106.81)

were not significantly different (F(1, 114)= 0.07, p > .05). The mean difference scores of

the 2x4 condition (X = $110.42) and the 4x4 condition (X = $95.86) were not

significantly different (F(1, 114) = 1.15, p > .05). The mean difference scores of the 4x2

condition (X = $106.81) and the 4x4 condition (X = $95.86) were not significantly

different (F(1, 114) = 0.63, p > .05).

Discussion

The data from experiment 1 replicate the number-of-levels effect that has been

reported in the literature and provide some new insights as to the potential source of this

effect. First, all three replicates showed a number-of-levels effect on attribute one when

the analysis was restricted to the 2x4 and 4x2 conditions. This type of design is common

in the number-of-levels literature. Second, when the 4x4 condition was also included in

the analysis, the mean difference score for attribute one was similar to the difference

score in the 2x4 condition, but less than the difference score in the 4x2 conditions.

Although this pattern of results held for only the refrigerator and digital camera

replicates, and only for attribute one, the result is consistent with the attentional contrast

hypothesis.

The most perplexing result in experiment 1 is the apparent attentional contrast

effect on attribute one and the lack of the attentional contrast effect on attribute 2. One

possible reason for this finding is that attentional contrast only operates on attributes that

are perceived to be important (i.e., worth attending to). In other words, an attribute must









be perceived to be important if its attribute levels are to exhibit an attentional contrast

based number-of-levels effect. The implication is that attribute two could also exhibit a

number-of-levels effect if it were to be perceived as more important. Experiment 1A will

examine this hypothesis.

Experiment 1A

The goal of experiment 1A was to assess whether the differential effects observed

on attributes one and two in experiment 1 could be a consequence of a (perceived)

difference in importance between the two attributes. In Experiment 1A, the (relative)

importance of the second attribute was increased in two ways. First, the range of the

attribute two's levels was increased and the range of attribute one's levels was decreased.

Second, participants were asked to elaborate on product usage. This second strategy was

adopted from Huffman (1997), who showed that elaboration on product usage increases

the weight of usage-related attributes whereas elaboration on product purchase increases

the weight of purchase-related attributes. Of the three replicates, only the refrigerator

category had a purchase-related attribute (annual energy cost) and a usage-related

attribute (capacity). Given the failure to find a number-of-levels effect on the usage-

related attribute in this category, participants were asked to elaborate on usage.

Method

The design was identical to experiment 1 with the limitation that only one of the

replicates (refrigerators) was investigated. The stimulus were adjusted so that the range of

attribute two (capacity) was doubled and the range of attribute one (annual energy cost)

was reduced. The attribute levels for capacity were 13, 17, 21, and 25 cubic feet. The

attribute levels for annual energy cost were $90, $85, $80, and $75.









The procedure was identical to experiment 1 except for one change. Prior to being

exposed to the eight product offers, participants were asked to elaborate on product

usage. Participants were told to "Think back to the times you have used a refrigerator.

Picture, in your mind's eye, the situations in which you have used it. Below, please

describe your thoughts and feelings about using the refrigerator. What was important to

you when you were using the refrigerator?"

Results

Table 4 presents the mean difference scores per attribute for each condition.

Data of 120 participants were retained for analysis. The first analysis was on

attribute one (annual energy cost). T the mean difference scores of the 2x4 condition (X

= $74.25) and the 4x2 condition (X = $98.37) were significantly different (F(1, 117) =

4.09, p < .05). The mean difference scores of the 2x4 condition (X = $74.25) and the 4x4

condition (X = $71.30) were not significantly different (F(1, 117) = 0.05, p > .05). The

mean difference scores of the 4x2 condition (X = $98.37) and the 4x4 condition (X =

$71.30) were significantly different (F(1, 117) = 5.70, p < .05). These results are

supportive of the attentional contrast hypothesis.

The second analysis was on attribute two (capacity). The mean difference scores of

the 2x4 condition (X = $265.13) and the 4x2 condition (X = $180.93) were significantly

different (F(1, 117) = 7.10, p < .05). The mean difference scores of the 4x2 condition (X

= $180.93) and the 4x4 condition (X = $220.72) were not significantly different (F(1,

117) = 1.75, p > .05). The mean difference scores of the 2x4 condition (X = $265.13) and

the 4x4 condition (X = $220.72) were not significantly different (F(1, 117) = 1.76, p >

.05). These results are supportive of the attentional contrast hypothesis.









Discussion

Experiment 1A shows that a number-of-levels effect can be produced on both

attributes in a two-factor conjoint design, provided the participant perceives each attribute

as important. When the importance of a second attribute was increased by increasing its

range and by getting consumers to elaborate on situations in which the attribute might be

important, a number-of-levels effect was obtained. In addition, the mean difference score

for attribute one in the 4x4 condition was similar to the difference score in the 2x4

condition, but less than the difference score in the 4x2 conditions. Similarly, the mean

difference score for attribute two in the 4x4 condition was similar to the difference score

in the 2x4 condition, but less than the difference score in the 2x4 conditions. These

results are consistent with the attentional contrast hypothesis.

Experiments 1 and 1A show that the "number-of-levels" effect does not occur

because of increasing levels on the focal attribute, but because of the decreasing levels on

the accompanying attribute. Moreover, the effect depends on the (perceived) importance

of an attribute. These results are consistent with an attentional contrast account, which

predicts that within subsequent offers, attentional contrast is produced on the attribute

that has more levels, as its levels are relatively more novel than the attribute levels of the

attribute with the lower number of levels.

However, experiments 1 and 1A were designed to manipulate the number of levels

and are, therefore, not the most direct test of the attentional contrast account. While the

conditions that were compared differed on attentional contrast on the focal attributes,

they also differed on the number of levels of the accompanying attribute. Experiment 2

provides a more direct test of the attentional contrast account by showing that it can






24


create a weighting difference while the number of levels on both focal and accompanying

attributes is kept constant.












Attribute 2
2 3


Attribute 1 2
3
4


Attribute 2


1
2
3
4


Attribute 1


Attribute 2


Attribute 1 [ 43



Note.-Exterior offers are colored in black, intermediate offers are shaded in grey, and
interior offers are hatched.

Figure 1. Design for experiments 1 and 1A. A) 4x2 Condition. B) 4x2 Condition. C) 4x4
Condition












Table 2. Attribute levels from experiment 1


Product category
Refrigerators


LCD televisions


Digital cameras


Attribute
annual energy cost ($n)
capacity (n cubic feet)


screen size (n inches)


Attribute level
1 2 3 4
100 90 80 70
16 18 20 22


20 24 28 32


horizontal viewing angle (n off-center) 45


imaging sensor resolution (n MegaPixels) 3.2


55 65 75


4.4 5.6 6.8


image storage capacity (up to n MB) 128 256 384 512



Table 3. Mean difference scores and simple effects for experiment 1


Attribute 1
Annual energy cost
(Refrigerators)
Screen size
(LCD televisions)
Imaging sensor resolution
(Digital cameras)


Attribute 2
Capacity
(Refrigerators)
Horizontal viewing angle
(LCD televisions)
Image storage capacity
(Digital cameras)


Mean difference score
4x2 2x4

144.24 78.97


F values
4x2 vs 2x44x2 vs 4x42x4 vs 4x4


63.84 10.38a 17.16a 0.63


257.37 176.20 196.82 3.98b


169.31


2.21 0.26


134.90 131.49 3.18c 4.07b 0.03


120.17 122.55 100.93


130.37

106.81


a implies significance at 99% level of confidence
b implies significance at 95% level of confidence
c implies significance at 90% level of confidence


4x2 vs 2x44x2 vs 4x42x4 vs 4x4

0.01 0.73 0.96


140.64 122.82 0.26 0.14 0.78


110.42 95.86 0.07 0.63







27




Table 4. Mean difference scores and simple effects for experiment 1A


Mean difference score


F values


Attribute 1
Annual energy cost
(Refrigerators)


4x2

98.37


2x4

74.25


4x4 4x2 vs 2x4 4x2 vs 4x4 2x4 vs 4x4


71.30


Attribute 2 4x2 2x4 4x4
Capacity
(Refrigerators) 180.93 265.13 220.72
a implies significance at 99% level of confidence
b implies significance at 95% level of confidence


0.05


4x2 vs 2x4 4x2 vs 4x4 2x4 vs 4x4


7.100














CHAPTER 4
EXPERIMENT 2: TEST OF ATTENTIONAL CONTRAST ACCOUNT WITH NO
DIFFERENCE IN NUMBER OF LEVELS

The goal of experiment 2 was to examine if attentional contrast can create a

weighting effect when there is no difference in number of levels. In experiments 1 and

1A, we found evidence for an attentional contrast account by varying the number of

levels of only one attribute at a time. However, attentional contrast was confounded with

the number of levels manipulation. By manipulating attentional contrast in a context in

which there is no difference in number of levels, we want to obtain more direct evidence

that the attentional contrast by itself can account for the weighting difference. In addition,

the three replicates and procedure from experiment 1 were used in the experiment. It was

felt that demonstrating a robust attentional contrast effect on one attribute was superior to

demonstrating an attentional contrast effect on two attributes of a single replicate.

Method

Design and Stimuli

The design was a 3 cell between-subjects design (contrast on attribute 1, contrast

on attribute 2, balanced contrast) with three between-subject replicates (refrigerator, LCD

televisions, digital cameras). In all three experimental conditions, each attribute was

defined on 4 levels (see Figure 2). The cells differed on the attribute on which attentional

contrast was created. In the contrast on attribute 1 condition, attentional contrast was

created on attribute 1. This contrast was created via the selection of offers that were

initially rated by the participant. The intermediate offers were [2,1], [2,2], [2,3], [2,4],









[3,1], [3,2], [3,3], and [3,4]. Thus, when the participant saw the exterior offers of [1,1],

[1,4], [4,1], and [4,4], the values of the first attribute were novel and the values of the

second attribute were familiar. In the contrast on attribute 2 condition, the intermediate

offers were [1,2], [1,3], [2,2], [2,3], [3,2], [3,3], [4,2],and[4,3]. Thus, when the participant

saw the exterior offers of [1,1], [1,4], [4,1], and [4,4], the values of the second attribute

were novel and the values of the first attribute were familiar. In the balanced contrast

condition, the intermediate offers were [2,2], [2,3], [3,2], and [3,3]. Thus, when the

participant saw the exterior offers of [1,1], [1,4], [4,1], and [4,4], the values of both

attributes were on average equally novel, hence there was no contrast. Note that this last

condition is particularly important because it will show that the number-of-levels effect is

not an issue of novelty (i.e., seeing levels that have not been seen before), but an issue of

attentional contrast (i.e., seeing levels that are novel relative to the levels seen on the

other attribute). In all other respects, the procedure was identical to the one used in

experiment 1.

Predictions

The attentional contrast hypothesis makes predictions for attributes one and two.

With respect to attribute one, the attentional contrast hypothesis predicts that the mean

difference score should be greater in the contrast on attribute 1 condition than in the

contrast on attribute 2 condition or the balanced contrast condition. With respect to

attribute two, the attentional contrast hypothesis also predicts that the mean difference

score should be greater in the contrast on attribute 2 condition than in the contrast on

attribute 1 condition or the balanced contrast condition. Given the results of experiments

1 and 1A, it was anticipated that the attentional contrast effect would be obtained on

attribute 1, but not attribute 2.











Results

Table 5 presents the mean difference scores per attribute in each condition.

Refrigerators

Data of 104 participants were retained for analysis. The analysis for each attribute

was performed separately. The first analysis was on attribute one (annual energy cost).

The mean difference scores of the contrast on attribute 1 condition (X = $99.37) and the

contrast on attribute 2 condition (X = $68.60) were significantly different (F(1, 101) =

4.96, p < .05). The mean difference scores of the contrast on attribute 1 condition (X =

$99.37) and the balanced contrast condition (X = $70.04) were significantly different

(F(1, 101) = 4.87, p < .05). The second analysis was on attribute two (capacity). The

mean difference scores of the contrast on attribute 2 condition (X = $94.02) and the

contrast on attribute 1 condition (X = $126.58) were not significantly different (F(1, 101)

= 2.65, p > .05). The mean difference scores of the contrast on attribute 2 condition (X =

$94.02) and the balanced contrast condition (X = $99.79) were not significantly different

(F(1, 101) = 0.08, p > .05).

LCD televisions

Data of 111 participants were retained for analysis. The first analysis was on

attribute one (screen size). The mean difference scores of the contrast on attribute 1

condition (X = $286.04) and the contrast on attribute 2 condition (X = $228.72) were not

significantly different (F(1, 108) = 2.13, p > .05). The mean difference scores of the

contrast on attribute 1 condition (X = $286.04) and the balanced contrast condition (X =

$192.99) were significantly different (F(1, 108) = 5.27, p < .05). The second analysis was

on attribute two (horizontal viewing angle). The mean difference scores of the contrast on









attribute 2 condition (X = $133.74) and the contrast on attribute 1 condition (X =

$146.52) were not significantly different (F(1, 108) = 0.26, p > .05). The mean difference

scores of the contrast on attribute 2 condition (X = $133.74) and the balanced contrast

condition (X = $101.04) were not significantly different (F(1, 108) = 1.75, p > .05).

Digital cameras

Data of 133 participants were retained for analysis. The first analysis was on

attribute one (imaging sensor resolution). The mean difference scores of the contrast on

attribute 1 condition (X = $190.12) and the contrast on attribute 2 condition (X =

$143.15) were significantly different (F(1, 130) = 5.78, p < .05). The mean difference

scores of the contrast on attribute 1 condition (X = $190.12) and the balanced contrast

condition (X = $144.37) were significantly different (F(1, 130) = 6.14, p < .05). The

second analysis was on attribute two (image storage capacity). The mean difference

scores of the contrast on attribute 2 condition (X = $128.43) and the contrast on attribute

1 condition (X =$114.12) were not significantly different (F(1, 130) = 0.89, p > .05). The

mean difference scores of the contrast on attribute 2 condition (X = $128.43) and the

balanced contrast condition (X = $119.46) were not significantly different (F(1, 130) =

0.34, p> .05).

Discussion

Experiment 2 manipulated attentional contrast directly by varying the relative

novelty of the attribute levels while keeping the number of levels of the attributes

constant. The results showed that attentional contrast can create a weighting difference

when there is no difference in number of levels. While this effect was consistently

produced on attribute 1, the results on the second attribute were non-significant. Since the

stimuli used in this experiment were identical to the ones used in experiment 1, we






32


attribute this lack of significance to the fact that attribute 2 was perceived to be less

important (see experiment 1A).










Attribute 2



Attribute 1

j4


Attribute 2
1 2 3 4


Attribute 1


Attribute 2


Attribute 1__I431


Note.-Exterior offers are colored in black, intermediate offers are shaded in grey. The
hatched offers are presented twice.

Figure 2. Design for experiment 2. A) Contrast on Attribute 1 Condition. B) Contrast on
Attribute 2 Condition. C) Balanced Contrast Condition.













Table 5. Mean difference scores and simple effects for experiment 2


Mean difference score


Attribute 1
Annual energy cost
(Refrigerators)
Screen size
(LCD televisions)
Imaging sensor resolution
(Digital cameras)


contrast
on
attribute 1


contrast
on balanced
attribute 2 contrast


F values
contrast on
attribute 1 vs
contrast on
attribute 2


99.37 68.60 70.04 4.96b

286.04 228.72 192.99 2.13

190.12 143.15 144.37 5.78b


contrast
on


contrast
on balanced


Attribute 2 attribute 1 attribute 2 contrast
Capacity
(Refrigerators) 126.58 94.02 99.79
Horizontal viewing angle
(LCD televisions) 146.52 133.74 101.04
Image storage capacity
(Digital cameras) 114.12 128.43 119.46
a implies significance at 99% level of confidence
b implies significance at 95% level of confidence


contrast on
attribute 1 vs
contrast on
attribute 2


0.26


contrast on
attribute 1 vs
balanced
contrast


6.14a


contrast on
attribute 2 vs
balanced
contrast


4.87b














CHAPTER 5
EXPERIMENT 3: TEST OF ATTENTIONAL CONTRAST ACCOUNT WITH
REVERSED NUMBER OF LEVELS

The goal of experiment 3 was to show that the "number-of-levels" effect can be

reversed when attentional contrast is created on the attribute with the lower number of

levels. In experiments 1 and 1A, attentional contrast was shown to create weighting

differences but the number of levels of the accompanying attribute was varied, but not the

number of levels of the focal attribute. In experiment 2, attentional contrast created

weighting differences when there was no difference in number of attribute levels. In

experiment 3, the test of the attentional contrast account is made more rigorous. More

explicitly, I test whether the number-of-levels effect can be obtained on the attribute with

the lower number of levels.

Method

Design and Stimuli

The design was a 3 cell between-subjects design (4x6, 6x4, and 6x6 attribute

levels) with three between-subject product replicates (refrigerators, LCD televisions,

digital cameras). Each attribute had four or six possible attribute levels. For an overview

of the attribute levels, see Table 6. The intermediate offers were [3,1], [3,2], [3,3], [3,4],

[3,5], [3,6], [4,1], [4,2], [4,3], [4,4], [4,5], and [4,6] in the 4x6 condition, [3,1], [4,1],

[3,2], [4,2], [3,3], [4,3], [3,4], [4,4], [3,5], [4,5], [3,6], and [4,6] in the 6x4 condition, and

[1,3], [1,4], [2,2], [2,5], [3,1], [3,6], [4,1], [4,6], [5,2], [5,5], [6,3], and [6,4] in the 6x6

condition. The four exterior offers were [1,1], [1,6], [6,1], and [6,6]. The design of









experiment 3 is presented in Figure 3. In the 4x6 condition, the levels of attribute 1 were

more novel than the attribute levels of attribute 2 at the time the participants evaluated the

exterior offers. In other words, attentional contrast was created on attribute 1. In the 6x4

condition, the levels of attribute 2 were more novel than the attribute levels of attribute 1

at the time the participants evaluated the exterior offers. In other words, attentional

contrast was created on attribute 2. In the 6x6 condition, the attribute levels of attribute 1

and attribute 2 were on average equally novel when participants evaluated the exterior

offers. In sum, attentional contrast was created on the four-level attributes. If the number-

of-levels effect depends only on the number of levels, the four-level attributes will have

lower derived importance weights than the six-level attributes. If the number-of-levels

effect depends on attentional contrast, then the four-level attributes will have higher

derived importance weights.

Predictions

The attentional contrast hypothesis makes predictions for attributes one and two.

With respect to attribute one, the attentional contrast hypothesis predicts that the mean

difference score should be greater in the 4x6 condition than in the 6x4 condition or the

6x6 condition. With respect to attribute 2, the attentional contrast hypothesis predicts that

the mean difference score should be greater in the 6x4 condition than in the 4x6 condition

or the 6x6 condition. Given the results of experiments 1, 1A, and 2, it was anticipated that

the attentional contrast effect would be obtained on attribute 1, but not on attribute 2.

Results

Table 7 presents the mean difference scores per attribute for each condition.









Refrigerators

Data of 87 participants were retained for analysis. The analysis for each attribute

was performed separately. The first analysis was on attribute one (annual energy cost).

The mean difference scores of the 4x6 condition (X = $119.89) and the 6x4 condition (X

= $77.59) were significantly different (F(1, 84) = 7.52, p < .05). The mean difference

scores of the 4x6 condition (X = $119.89) and the 6x6 condition (X = $76.75) were

significantly different (F(1, 88) = 6.72, p < .05). The second analysis was on attribute two

(capacity). The mean difference scores of the 6x4 condition (X = $110.09) and the 4x6

condition (X = $132.15) were not significantly different (F(1, 84) = 1.34, p > .05). The

mean difference scores of the 6x4 condition (X = $110.09) and the 6x6 condition (X =

$94.92) were not significantly different (F(1, 84) = 0.55, p > .05).

LCD televisions

Data of 96 participants were retained for analysis. The analysis for each attribute

was performed separately. The first analysis was on attribute one (screen size). The mean

difference scores of the 4x6 condition (X = $246.95) and the 6x4 condition (X = $201.55)

were not significantly different (F(1, 93) = 1.07, p > .05). The mean difference scores of

the 4x6 condition (X = $246.95) and the 6x6 condition (X = $228.70) were not

significantly different (F(1, 93) = 0.15, p > .05). The second analysis was on attribute two

(horizontal viewing angle). The mean difference scores of the 6x4 condition (X =

$138.15) and the 4x6 condition (X = $129.30) were not significantly different (F(1, 93) =

0.16, p > .05). The mean difference scores of the 6x4 condition (X = $138.15) and the

6x6 condition (X = $133.70) were not significantly different (F(1, 93) = 0.05, p > .05).









Digital cameras

Data of 123 participants were retained for analysis. The analysis for each attribute

was performed separately. The first analysis was on attribute one (imaging sensor

resolution). The mean difference scores of the 4x6 condition (X = $175.06) and the 6x4

condition (X = $169.20) were not significantly different (F(1, 120) = 0.09, p > .05). The

mean difference scores of the 4x6 condition (X = $175.06) and the 6x6 condition (X =

$153.29) were significantly different (F(1, 120)= 1.17, p > .05). The second analysis was

on attribute two (image storage capacity). The mean difference scores of the 6x4

condition (X = $157.11) and the 4x6 condition (X = $126.92) were not significantly

different (F(1, 120) = 2.50, p > .05). The mean difference scores of the 6x4 condition (X

= $157.11) and the 6x6 condition (X = $111.00) were significantly different (F(1, 120) =

5.54, p <.05).

Discussion

The results of experiment 3 point in the direction of an attentional contrast

explanation. On the first attribute, the means are in the direction predicted by the

attentional contrast account, though the results are significant in only one of three product

categories (refrigerators). On attribute 2, the results of the refrigerators category differs

from the predictions of both number-of-levels effect and attentional contrast. The means

of the LCD televisions category are in the direction predicted by attentional contrast,

though non-significant. The results for the digital cameras category are in the direction

predicted by attentional contrast. Since this experiment used the same exterior levels as

experiments 1 and 2, the partial lack of significance on attribute 2 may be attributed to

the perceived lack of importance of attribute 2 (see experiment 1A).





















Attribute 1


A

Attribute 2


1

2
3

4

5


B
Attribute 2

22 3 4 5


2

Attribute 1
4
5
6


C
Attribute 2




2

Attribute 1



6

Note.-Exterior offers are colored in black, intermediate offers are shaded in grey.

Figure 3. Design for experiment 3. A) 4x6 Condition. B) 6x4 Condition. C) 6x6
Condition.












Table 6. Attribute levels from experiment 3
Attribute level
Product
category Attribute 1 2 3 4 5 6
Refrigerators annual energy cost ($n) 100 95 90 80 75 70
capacity (n cubic feet) 16 17 18 20 21 22

LCD
televisions screen size (n inches) 20 22 24 28 30 32
horizontal viewing angle (n off-center) 45 50 55 65 70 75


Digital imaging sensor resolution (n
cameras MegaPixels) 3.2 3.8 4.4 5.6 6.2 6.8
image storage capacity (up to n MB) 128 192 256 384 448 512
Note.-The underlined attribute levels are the levels that are added when an attribute has
6 levels.


Table 7. Mean difference scores and simple effects for experiment 3


Mean difference score


F values


Attribute 1
Annual energy cost
(Refrigerators)
Screen size
(LCD televisions)
Imaging sensor resolution
(Digital cameras)


4x6 6x4 6x6 4x6 vs 6x4 4x6 vs 6x6

119.89 77.59 76.75 7.52a 6.72a

246.95 201.55 228.70 1.07 0.15

175.06 169.20 153.29 0.09 1.17


Attribute 2 4x6 6x4
Capacity 132.15 110.09
(Refrigerators)
Horizontal viewing angle 129.30 138.15
(LCD televisions)
Image storage capacity 126.92 157.11
(Digital cameras)
a implies significance at 99% level of confidence
b implies significance at 95% level of confidence


6x6

94.92


4x6 vs 6x4 6x4 vs 6x6

1.34 0.55


133.70 0.16

111.00 2.50


0.05

5.54b














CHAPTER 6
GENERAL DISCUSSION

Experiment 1 provides evidence that the weighting effect as observed in published

studies may not be a "number-of-levels" effect, but that it may be due to attentional

contrast away from the less novel attribute level within each trial. Experiment 2 shows

that attentional contrast can create a weighting effect when there is no difference in the

number of levels. Experiment 3 shows that attentional contrast can overrule the number-

of-levels effect, i.e., that it can result in greater derived importance weights on the

attribute with the lower number of levels. In the three experiments, results were

consistent on attribute 1, but did not reach significance on attribute 2. In experiment 1A, I

showed that this lack of significance was due to the perceived relative unimportance of

the second attribute.

In experiment 1A, the distinction between attribute level importance and attribute

importance is highlighted. Simonson (1993) shows that consumers often do not possess

well-formed beliefs about attribute importance. Instead, they construct their preferences

during judgment and choice (Payne, Bettman, and Johnson 1992). First, the experiments

in this dissertation show that attribute importance is influenced by the attribute levels that

are presented. Second, experiment 1A shows that this influence only occurs when the

attribute is perceived to be important. While in experiments 1, 2 and 3 attentional contrast

only created a weighting effect on attribute 1, increasing the accessibility of consumers'

product experiences in experiment 1A resulted in a weighting effect on attribute 2. In the

number-of-levels effect literature, price was often used as one of the attributes. Since this









dissertation uses price as a dependent variable, one of the product attributes that are

usually perceived as being inherently important was not eligible, which may have

reduced the overall perceived importance of the attributes used in this dissertation

compared to attributes used in earlier studies. This effect may be enhanced by the fact

that the choice of attributes was restricted to attributes that have numerical values and for

which evaluations are known to be monotonically distributed.

The attentional contrast explanation is consistent with all the results that have been

published in the number-of-levels effect literature. First, all empirical studies (except

Johnson 1992) used a sequential evaluation procedure. Attentional contrast can only

operate when attribute levels are repeated and when this repetition occurs in the context

of other attribute levels that have been more or less repeated. Second, attentional contrast

is consistent with the observation that the number-of-levels effect was not produced in

the compositional measures that were added before the sequential evaluation procedure

(Wittink et al. 1992a) and that only minimal effects were produced when the

compositional measures were added after the sequential evaluation procedure

(Steenkamp and Wittink 1994; Verlegh et al. 2002). Finally, the reason why attentional

contrast was not detected in earlier studies is that they invariably compared conditions in

which the number of levels of more than one attribute was compared at once (see Table

1).

While the main contribution of this dissertation is the proposition of a new account

for a repeatedly published effect, two additional contributions can be mentioned. First,

until now, the process of attentional contrast had only been documented in a category

learning task. Based on the shared characteristic of sequential trials between the









classification learning and conjoint or multiattribute tasks, I showed that attentional

contrast also occurs in an evaluation (versus categorization) task. Second, in previous

studies (Kersten et al. 1998), attentional contrast was documented using pictorial values. I

generalized its occurrence to the use of metric values.

While I conclude that attentional contrast is at least partly responsible for the

weighting effect that was observed in studies examining the number-of-levels effect, the

data do not provide information as to how attentional contrast creates a weighting

difference. One possibility is that participants attend longer to the relatively novel

attribute levels, and that attribute levels that are attended to for longer durations receive

more weight. However, it is possible that it is not mere attention which results in the

weighting difference. It is likely that participants elaborate more on the relatively novel

attribute levels. For example, when asked to evaluate an LCD television with a horizontal

viewing angle which is 60 degrees off-center, participants may imagine themselves in

their apartments and try to assess mentally from which area they would have a decent

view of the screen.

Aside from the question how attentional contrast works, one may ask when it

works. First, the impact of the attentional contrast effect is likely to be dependent on the

assumptions with regards to the total pool of attention. In the experiments presented in

this dissertation, students evaluated the trials at their own pace. In other words, the pool

of attention was variable. This means that increased attention to one attribute is not

necessarily accompanied by decreases in attention to a second attribute. In contrast, in a

model in which the total pool of attention is fixed, increased attention to a first attribute

results in decreases in attention to the second attribute. A model with a fixed pool of









attention is more closely operationalized by a computer-paced experiment. Using a

computer-paced experiment may enhance the chances that attentional contrast creates a

weighting effect on both attributes. Second, as shown in experiment 1A, the occurrence

of the weighting effect depends on the importance of the attribute. The prior level of

attention to an attribute may influence the impact of changes in attention towards the

attribute levels. Attributes that normally receive a low level of attention may show small

reductions in attention as a result of attentional contrast away from the attribute levels.

Analogously, attributes that tend to receive high levels of attention may show small

increases in attention as a result of attentional contrast towards the attribute levels.

Therefore, attentional contrast is likely to have the greatest impact when attributes are

moderately important.

Finally, a main focus of many number-of-level effect studies was to find a way to

eliminate the bias. Based on the account offered in this dissertation, one should try to

keep the relative novelty of attribute levels in subsequent trials constant. Since this is

impossible when attributes are defined on unequal numbers of levels, one may prefer to

use a data collection method that does not involve subsequent trials. For example, the use

of a trade-off matrix may be preferred in case of an unequal number of levels.














CHAPTER 7
LIMITATIONS AND FUTURE RESEARCH

While this dissertation takes an important first step towards proposing a new

account for the number-of-levels effect, there are several issues that remain unexamined

and that could be extended to potentially useful future research.

There are several characteristics of the experiments that limit the generalizability of

the findings. First, only one dependent measure was used in all studies. Participants stated

their reservation price for product offers. While the results of these studies supported an

attentional contrast explanation, the number-of-levels effect has been identified using a

variety of dependent measures. To generalize the results of this dissertation, it would be

useful to examine whether attentional contrast can also explain the results of studies that

use other dependent measures such as paired comparisons and rating scales. Second, no

compositional measures were collected in this dissertation. While there is considerable

evidence from the literature that the number-of-levels effect is not produced on

compositional measures, it could be useful to gather additional evidence that shows that

the number-of-levels effect is inherent to the sequential evaluation procedure. Moreover,

it could be useful to examine if the weighting effect caused by attentional contrast carries

over to a posteriori collected compositional measures.

A final extension of this dissertation has the potential of being useful. This research

shows that attentional contrast does occur not only during category learning tasks, but

also during multiattribute judgment tasks. The role of attentional contrast during other






46


phenomena that have been identified in the multiattribute judgment literature (e.g.,

frequency effect) would be worth examining.















LIST OF REFERENCES


Currim, Imran S., Charles B. Weinberg, and Dick R. Wittink (1981), "The Design of
Subscription Programs for a Performing Arts Series," Journal of Consumer
Research, 8 (June), 67-75.

Goldstone, Robert L. and Mark Steyvers (2001), "The Sensitization and Differentiation
of Dimensions During Category Learning," Journal ofExperimental Psychology:
General, 130 (March), 116-139.

Green, Paul E. and V. Srinivasan (1978), "Conjoint Analysis in Consumer Behavior:
Issues and Outlook," Journal of Consumer Research, 5 (September), 103-123.

Green, Paul E. and V. Srinivasan (1990), "Conjoint Analysis in Marketing Research:
New Developments and Directions," Journal ofMarketing, 54 (October), 3-19.

Huffman, Cynthia (1997), "Elaboration on Experience: Effects on Attribute Importance,"
Psychology & Marketing, 14 (August), 451-474.

Johnson, Richard M. (1992), "Comment on Attribute Level Effects," SecondAnnual
Advanced Research Techniques Forum Proceedings, Rene Mora (ed.), Chicago:
American Marketing Association, 62-64.

Kersten, Alan W., Robert L. Goldstone, and Alexandra Schaffert (1998), "Two
Competing Attentional Mechanisms in Category Learning," Journal of
Experimental Psychology: Leiaruiu Memory, and Cognition, 24 (6), 1437-1458.

Nosofsky, Robert M. (1986), "Attention, Similarity, and the Identification-Categorization
Relationship," Journal ofExperimental Psychology: General, 115 (March), 39-57.

Nosofsky, Robert M. (1991), "Tests of an Exemplar Model for Relating Perceptual
Classification and Recognition Memory," Journal ofExperimental Psychology:
Human Perception and Performance, 17 (February), 3-27.

Orme, Bryan K. (1998), "Reducing the Number-of-Attribute-Levels Effect in ACA with
Optimal Weighting," 1998 .Sen tiI ,th Software Conference Proceedings, Ketchum,
ID: Sawtooth Software.

Payne, John W., James R. Bettman, and Eric J. Johnson (1992), "Behavioral Decision
Research: A Constructive Processing Perspective," Annual Review ofPsychology,
43, 87-131.









Simonson, Itamar (1993), "Get Closer to Your Customers by Understanding How They
Make Choices," California Management Review, 35 (4), 68-84.

Srinivasan, V. and Allan D. Shocker (1973), "Estimating the Weights for Multiple
Attributes in a Composite Criterion Using Pairwise Judgments," Psychometrika, 38
(December), 473-493.

Steenkamp, Jan-Benedict E.M. and Dick R. Wittink (1994), "The Metric Quality of Full-
Profile Judgments and the Number-of-Attribute-Levels Effect in Conjoint
Analysis," International Journal ofResearch in Marketing, 11 (June), 275-286.

Sutherland, N.S. and N.J. Mackintosh (1971). Mechanisms of Animal Discrimination
Learning. New York: Academic Press.

Verlegh, Peeter W.J., Hendrik N.J. Schifferstein, and Dick R. Wittink (2002), "Range
and Number-of-Levels Effects in Derived and Stated Measures of Attribute
Importance," Marketing Letters, 13 (February), 41-52.

Wittink, Dick R., Joel Huber, John A. Fiedler, and Richard L. Miller (1992a), "Attribute
Level Effects in Conjoint Revisited: ACA versus Full Profile," Second Annual
Advanced Research Techniques Forum Proceedings, Rene Mora (ed.), Chicago:
American Marketing Association, 51-61.

Wittink, Dick, Joel Huber, Peter Zandan, and Richard M. Johnson (1992b), "The Number
of Levels Effect in Conjoint: Where Does It Come From and Can It Be
Eliminated?" 1992 .i eni ,,th Software Conference Proceedings, Ketchum, ID:
Sawtooth Software.

Wittink, Dick R., Lakshman Krishnamurthi, and Julia B. Nutter (1982), "Comparing
Derived Importance Weights Across Attributes," Journal of Consumer Research, 8
(March), 471-474.

Wittink, Dick R., Lakshman Krishnamurthi, and David J. Reibstein (1989), "The Effects
of Differences in the Number of Attribute Levels on Conjoint Results," Marketing
Letters, 1, 113-123.















BIOGRAPHICAL SKETCH

Els De Wilde was born in Gent, Belgium, on July 18, 1975. After attending high

school at the Onze-Lieve-Vrouw-Presentatie in Sint-Niklaas, Belgium, she attended the

Katholieke Universiteit Leuven in Louvain, Belgium. At this university, she obtained the

degrees of Kandidaat in de Psychologie (1996), graduating "with distinction," and

Licentiaat in de Psychologie (1999), for which she graduated "with great distinction." As

part of her studies for the Licentiaat degree, she spent five months at the Universidad

Complutense de Madrid, Spain. She wrote a thesis on categorization under the guidance

of Professor Gert Storms. She spent seven months on a research internship, split between

the Laboratory of Experimental Psychology at the Catholic University of Louvain, where

she worked with Karl Verfaillie, and the marketing research company TNS Dimarso in

Brussels, Belgium, where she was a trainee at the Fast Moving Consumer Goods unit.

These research experiences were instrumental in shaping her interest in pursuing doctoral

studies in the area of marketing, which she started at the University of Florida in August

1999. During her four years in the Ph.D. program at the Department of Marketing, she

collaborated with Professors Chris Janiszewski and Alan Cooke. In 2001, Els taught an

undergraduate course in consumer behavior at the University of Florida. Since Summer

2003, Els has been working as an Assistant Professor of Marketing at HEC Montreal,

Canada.