<%BANNER%>

Attribute Level Distributions and Consumer Goals affect Subsequent Attribute Use

Permanent Link: http://ufdc.ufl.edu/UFE0024862/00001

Material Information

Title: Attribute Level Distributions and Consumer Goals affect Subsequent Attribute Use
Physical Description: 1 online resource (73 p.)
Language: english
Creator: Itzkowitz, Jesse
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: attribute, categorization, decision, goals, marketing, perception, psychology
Marketing -- Dissertations, Academic -- UF
Genre: Business Administration thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This dissertation examines an unexplored influence on consumers? product perceptions: the distribution of attribute values within a product market. In line with a goal-directed theory of perception, I show that consumers? tasks prime the perceptual accessibility particular product attributes, depending on the attributes? distributions. This increased sensitivity persists post task completion, resulting in greater perceptions of attribute importance (pilot study) and greater sensitivity to attribute level tradeoffs (Experiments 1-3). Experiment 3 also shows how consumers? dispositional biases (need for cognition) can prime particular attributes, and, in some cases magnifies the effects of task related priming.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Jesse Itzkowitz.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Janiszewski, Chris A.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024862:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024862/00001

Material Information

Title: Attribute Level Distributions and Consumer Goals affect Subsequent Attribute Use
Physical Description: 1 online resource (73 p.)
Language: english
Creator: Itzkowitz, Jesse
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: attribute, categorization, decision, goals, marketing, perception, psychology
Marketing -- Dissertations, Academic -- UF
Genre: Business Administration thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This dissertation examines an unexplored influence on consumers? product perceptions: the distribution of attribute values within a product market. In line with a goal-directed theory of perception, I show that consumers? tasks prime the perceptual accessibility particular product attributes, depending on the attributes? distributions. This increased sensitivity persists post task completion, resulting in greater perceptions of attribute importance (pilot study) and greater sensitivity to attribute level tradeoffs (Experiments 1-3). Experiment 3 also shows how consumers? dispositional biases (need for cognition) can prime particular attributes, and, in some cases magnifies the effects of task related priming.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Jesse Itzkowitz.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Janiszewski, Chris A.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024862:00001


This item has the following downloads:


Full Text

PAGE 1

1 ATTRIBUTE LEVEL DISTRIBUTIONS AND CONSUMER GOALS AFFECT SUBSEQUENT ATTRIBUTE USE By JESSE ITZKOWITZ 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 2009

PAGE 2

2 2009 Jesse Itzkowitz

PAGE 3

3 For Jen

PAGE 4

4 ACKNOWLEDGMENTS I have been truly fortunate to have had incred ible support on my path towards this PhD. It has been a journey filled with highs and lows, and I know that I certainly would not have been able to finish without the am azing support system th at I have been given. It is my pleasure to acknowledge some of my mentors, friends, and family that have c ontributed to this dissertation. First, thanks to my supervisory committee me mbers who made this dissertation possible. I am very grateful for the time and energy they took to help me through the doctoral program. I am especially thankful to Joe Alba. Joe introduced me to the field of consumer behavior, and was the first one to plant the idea of a second PhD in my mind. If you were not the amazing teacher and scholar that you are, I doubt that I would have ever consid ered doing it again. I was also lucky to have Alan Cooke on my committee. Al an, your statistical and methodological guidance was invaluable, and the glasses, a nd glasses, and glasses of wine certainly helped take the sting out of bad data days. It wa s also an honor to have Joel Houston on my committee. Your presence really forced me to make sure that the work that I was doing had some relevance and merit within the larger business discipline. Also, thanks for keeping Jen busy with her dissertation. This was critical for me at times. Finally, to Chris, throughout this process, you have known when to push, when to help, and when to watch from a distance. Your insights, while occasionally frustrating, help ed make me a better researcher in countless ways. You were always willing to give your time and support when ever I asked, and your faith in me gave me the strength to keep going even when there was not hing I wanted to do more than give up. I am also grateful to my friends for thei r endless patience. Tha nks to Joey Hoegg and Elise Chandon. You helped recruit me to the pr ogram, and made me feel like I was home from the very start. I also want to thank Xiaoqing Ji ng. You have been an incredible officemate and friend. You were always willing to listen to me and help me out with advice. Our lunches

PAGE 5

5 together will be something that I truly miss once I leave here. Tha nks to Melissa Minor as well. Melissa, you are a great friend, a nd I have really enjoyed worki ng on research with you. Your enthusiasm for projects is c ontagious, and our adventures at conferences (and beyond!) have provided much needed diversions from school life. Julia Belyavsky also deserves special thanks. Julia, you are like a sister to me, and I could not have made it out of this program alive without you. It has been amazing getting to know you, and I really cherish your friendship. I am truly envious of your energy and your selflessness when it comes to your friends. I look forward to many great times together in the future. Fina lly, big ups to Baler Bilgin, Dan Rice, Steven Sweldens, and Juliano Laran. You guys are my boys I could not have imagined making four better friends in such a quick time. There are no words to describe how cr ucial all of you are to me. I will be your wingman anytime. Finally, I want to thank my family who has provided much moral support during my education. I am very appreciative of my Mo m and Dad for always encouraging, understanding, and being there for me. I would like to thank my brother for being there to listen to me and having the right thing to say. Thanks also to my in-laws, Marv and Trudee, for your love and support. I am really lucky to have you. Last but not least, I would like to thank my wife Jennife r. I cannot believe you let me do this again. I love you very much.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4LIST OF TABLES................................................................................................................. ..........8LIST OF FIGURES.........................................................................................................................9ABSTRACT...................................................................................................................................10CHAPTER 1 INTRODUCTION..................................................................................................................11Attribute Value Distributions and Attribute Use....................................................................13Attributes with Category Diagnostic Distributions......................................................... 14Attributes with Unique Distributions of Values.............................................................. 16Consumer Goals Affect Attribute Use.................................................................................... 182 THE PILOT STUDY..............................................................................................................22Method....................................................................................................................................22Results.....................................................................................................................................23Discussion...............................................................................................................................243 EXPERIMENT 1....................................................................................................................27Experiment 1A.................................................................................................................. ......27Method.............................................................................................................................28Results........................................................................................................................ .....31Discussion..................................................................................................................... ...32Experiment 1B.................................................................................................................. ......33Method.............................................................................................................................33Results........................................................................................................................ .....35Discussion..................................................................................................................... ...354 EXPERIMENT 2....................................................................................................................45Method....................................................................................................................................45Results.....................................................................................................................................47Discussion...............................................................................................................................475 EXPERIMENT 3....................................................................................................................51Experiment 3A.................................................................................................................. ......51Method.............................................................................................................................52

PAGE 7

7 Results........................................................................................................................ .....53New product groupings............................................................................................ 53Initial product viewing times.................................................................................... 54Experiment 3B.................................................................................................................. ......55Method.............................................................................................................................56Results........................................................................................................................ .....56Discussion...............................................................................................................................586 GENERAL DISCUSSION..................................................................................................... 63Applications and Extensions...................................................................................................65LIST OF REFERENCES...............................................................................................................69BIOGRAPHICAL SKETCH.........................................................................................................73

PAGE 8

8 LIST OF TABLES Table page 3-1 Schematic for attribute value as signm ents for the exposure stimuli.................................. 39 3-2 Product categories and attribute types used in experim ent 1.............................................40 3-3 Schematic representation of the te st phase stim uli in experiment 1.................................. 41 4-1 Schematic of dependent measures used in the test portion of experim ents 2 and 3.......... 49

PAGE 9

9 LIST OF FIGURES Figure page 2-1 Tee-shirts used as stimuli in the pilot experim ent............................................................. 25 2-2 Reported importance ratings for the pilot experiment. ...................................................... 26 3-1 Example of counterbalancing in the tee-shirt stimuli used experim ent 1.......................... 42 3-2 Results of experiment 1A................................................................................................... 43 3-3 Results of experiment 1B...................................................................................................44 4-1 Results of experiment 2.................................................................................................... .50 5-1 Results of experiment 3A................................................................................................... 61 5-2 Results of experiment 3B...................................................................................................62

PAGE 10

10 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ATTRIBUTE LEVEL DISTRIBUTIONS AND CONSUMER GOALS AFFECT SUBSEQUENT ATTRIBUTE USE By Jesse Itzkowitz August 2009 Chair: Chris Janiszewski Major: Business Administration This dissertation examines an unexplored in fluence on consumers product perceptions: the distribution of attribute values within a produc t market. In line with a goal-directed theory of perception, I show that consumers tasks prime the perceptual accessibility particular product attributes, depending on the attri butes distributions. This increased sensitivity persists post task completion, resulting in greater perceptions of attribute importa nce (pilot study) and greater sensitivity to attribute level tradeoffs (Expe riments 1-3). Experiment 3 also shows how consumers dispositional biases (need for cognition) can prime particular at tributes, and, in some cases magnifies the effects of task related priming.

PAGE 11

11 CHAPTER 1 INTRODUCTION Changing consum ers weighting of product attribut es is one of the most critical elements of marketing strategy. At a most basic level, positioning strategies based upon differentiation require that some product attri butes are more accessible than ot hers when consumers consider particular products (Dickson and Ginter 1987). Changing the accessibility of product attributes also affects product choice. Products are typicall y dense with attributes, forcing consumers to utilize only a subset of produc t attributes in the decision making process (Bettman, Luce, and Payne 1998). Obviously, managers would like consumers to use attributes where their product possesses a competitive advantage. Thus, the abili ty to understand and control which attributes enter consumers decision calcu li is of fundamental theoreti cal and practical interest. Previous research has indicated that an at tribute enters the decision calculus when the attribute is accessible or diagnostic (Feldman and Lynch 1998; Lynch, Marmorstein, and Weigold 1998). Of these factors, accessibility is a primary requirement. One way that particular attributes can gain accessibility is through priming. Traditionally, priming has been thought of as a memory based phenomena, whereby prime-asso ciated concepts become more accessible in memory after exposure to a prime. Building on th at general theme, recent work has also shown that priming increases the accessibility of external prime-associated cues as well. These priming effects have been shown to influence consumers external search for information (Mandel and Johnson 2002) and stimulus-based ch oices (Chernev 2004; Yi 1990). However, extant research has centered on co mpatibilities between the qualitative nature of the prime and the attributes of products in stimulus-based c hoice. For example, in Mandel and Johnson (2002) individuals were exposed to either a safety or frugality prime. They were then given information about two cars, one which was safer, but more expensive, and one which was

PAGE 12

12 cheaper, but less safe. Subjects then chose the ca r they favored. As predicted, subjects receiving a safety prime chose the car with superior values on the safety attribute and subjects receiving the frugality prime chose the cheap er car. Similarly, subjects in Chernev (2004), who were either primed with a prevention or promotion goal, chose different products, depending on how these products attributes afforded the comp letion of the primed regulatory goal. This paper investigates an alternative way that stimulus-b ased attributes are primed. Similar to the way that goals prime individuals to search for and use qua litatively related product attributes, I examine how consumers tasks act as primes, increasing the accessibility (and subsequent use) of attributes th at enable task completion, regardle ss of the qualitative nature of the attribute. Specifically, drawing elements from the theories of goal-based perception (Janiszewski 2007), event coding (Hommell, Mssler, Aschersleben, and Prinz 2001), and diagnostic recognition (Schyns 1998), I propose a mechanism that describes how consumers tasks are best afforded by attribut es with particular distributions lead them to rely on attributes that afforded initial task completion during subsequent judgments. One pilot study and five experiments demonstr ate that consumers tasks prime perceptual accessibility of certain product attributes, dependi ng on their distributions. This sensitization is shown to affect perceptions of attribute importance (pilot study) and sensitivity towards attribute level tradeoffs (Experiments 1-3). Experiment 1A shows that task related attribute priming leads to differential sensitivity towards attribute leve l tradeoffs, resulting in changes in consumers judgments about product novelty. Experiment 1B extends this finding to differences in consumers willingness to pay for new products. Experiment 2 tests attribute level sensitivity differently, and provides critical insight into how attribute use differs depending on consumers tasks. Experiments 3A and 3B provide evidence that post-exposure sensitivity towards specific

PAGE 13

13 attribute distributions is the result of priming rather than consumers motivation. These studies also discuss an important moderato r to the task-attribute relations hip. They show that consumers need for cognition (NFC), an individual disposi tional bias having effects on information search, affects the influence of task related priming on subsequent judgments of product similarity. The paper concludes with a discussion of both the th eoretical and applied im plications of these results, providing several directions for future inquiry. Attribute Value Distributions and Attribute Use Within any product category, individual brands often possess similar attributes. This is even more prevalent for individual models of a product within a part icular manufacturers product line. Although brands (models ) within a product category (product line) may all have the same attributes, each brand (model) may possess a different attribute value along these attributes. For example, all cars have some doors (shared a ttribute), but the specifi c shape of each car door (value along this attribute) may vary between different brands. Likewise, there may also be variation in values along an attribute between different models of the same brand. Along shared attributes, indivi dual items may have similar or different values. On some attributes, all of the items may have the same va lue. The structural alignment literature defines such attributes as commonalities (Markman and Gentner 1993). On other attributes, a group of products may share one value while another grou p shares a different value. Because these attributes describe sets of pr oducts, they can be thought of as category diagnostic Finally, there may be some product attributes where each item possesses a unique value. Both of these types of attribute distributions represent alignable differences between the items (Markman and Gentner 1993), which will be the focus of this pape r. Before discussing ho w these attributes are primed, it is useful to understand th e baseline salience of the attributes. The next sections of this paper explore the general perceptual accessibil ity of category diagnosti c and unique attributes.

PAGE 14

14 Attributes with Category Diagnostic Distributions There may be reason to believe that the human perceptual system may be naturally attuned to perceive attributes with category diagnostic distri butions of attribute values. Categorization and concept formation are inhere nt perceptual processes critical to human cognition (Fodor 1981; Murphy 2002). One function of the conceptual system is to simplify the information in the environment (Atick 1992; Chater 1999; Rosch, Mervis, Grey, Johnson, and Boyles-Braem 1976). Efficiency is accomplished through the representation of attributes that allow the most accurate groupings of the largest number of similar items, known as basic-level categories (Rosch et al. 1976). Within the memo ry literature, high accessibility for basic-level categories is well documented (Mervis and Rosch 1981; Rosch et al. 1976), but less evidence exists regarding the accessibility (eit her perceptual or in memory) of the attributes that define them. While Schyns (1998) notes that the perceptu al accessibility of physical features defining basic-level membership (i.e., category diagnosti c attributes) provides a compelling explanation for the speedier recognition of ba sic-level categories, there has been limited work detailing this hypothesis explicitly. The existing literature regard ing accessibility of category diagnostic attributes mainly relates to memory based phenomena. Schyns a nd Murphy (1994) contended that individuals preferentially represent category diagnostic attribut es in memory because of their contribution to efficient concepts and la beled this phenomenon the functionality principle Schyns and Rodet (1997) provided a test of the functionality principle In the preliminary experiment of their study, subjects were exposed to te n pictures. Half of the pictur es were from one category and the other half were members of an alternate ca tegory. Category membership was defined by the exclusive presence of an object feature such that features common to all the members of the first category were absent from members of the second category, and vice-versa. Additionally, all of

PAGE 15

15 the stimuli had some unique features. Subjects were not given any information about the category structure of the items in the exposure se t. They did not know that the set could be divided into categories, nor were they made aware of the number of possible categories that existed. Following exposure to the entire set of items, subjects were shown a new picture and were asked to delineate features which they h ad seen during exposure. Consistent with the functionality principle subjects delineated features that prev iously defined membership in either category to a greater extent than the other stim ulus features. This was interpreted as higher accessibility for the category diagnostic attributes. Other evidence also suggests that category diagnostic attributes become perceptually salient after exposure, leading in dividuals to rely on them in fu ture decisions. Tversky (1977) found that attributes that allowe d for accurate classifications of the presented stimuli were more heavily weighted in subjects similarity judgm ents. Similarly, Goldstone (1994) demonstrated that following exposure, perceptual sensitivity towards category dia gnostic attributes is increased. In his experiment, a lthough subjects could attend to either (or both) of two stimulus attributes, subjects only demonstrated increased sensitivity (as indicated by their performance on same/different judgments) on the attribute whic h accurately predicted correct categorization, even when both attributes were present an equal number of times and were tightly controlled to be equally salient. There are several implications to these studies. First, they suggest that the distribution of values along an attribute aff ects that attributes perceptual accessibility. Second, category diagnostic attributes may be more perceptually accessible and be the most likely to affect individuals subsequent decisions than attributes with other distributions of values.

PAGE 16

16 Attributes with Unique Distributions of Values Aside from operating efficiently, it is critical th at our conceptual system be able to identify differences between individual category members. Individual item identification requires representation of attributes th at have differentiating informa tion. While representing category diagnostic attributes facilitates storage and retrieval of many items using a single cue, it does not afford specific identification or recognition processes within the category itself. Identification is dependent on the recognition of at tribute values that define indi vidual items. Attributes with unique values for each item best afford this proc ess. Due to this functions importance, unique attributes may be more perceptually sa lient than other attribute distributions. Some evidence for greater perceptual accessib ility comes from the perceptual learning literature. One learning mechanism, predifferentia tion, sensitizes individuals towards features of stimuli that enable distinctions to be made between them (Gibson 1991). Indeed, predifferentiation has been shown to be an in tegrated, automatic process that accompanies perception itself. In their cl assic study, Gibson and Walk (1965) demonstrated that stimulus differentiation occurs through exposure alone and that attributes that distinguish individual objects guide future behavior even when the ini tial exposure to the set of items is not governed by any high-level strategy. Attributes with unique distributions of va lues for may also be more perceptually accessible due to the increased attention paid to them during the exposure process. As shown by Lurie (2004), unique attributes contain the largest amount of information (of any distribution type) due to the number of possible attribute valu es and the equiprobabili ty of these attribute values occurrences. The large am ount of information contained within this attribute may draw attention towards it du ring exposure. Pirolli and Card (1999) theorize that atten tion to particular patches of a stimulus is proportional to the amoun t of information that the patch contains. If

PAGE 17

17 attentional resources are zero-sum, then the most variant attributes would receive more attention than other product attributes. This initial attention may impact the attributes later use. Recent work regarding the number of levels effect in co njoint analysis supports this premise. In their study, De Wilde, Janiszewski, and Cooke (2008) f ound that unique attribute values tended to draw increased attention, aff ecting subjects importance ratings. Other evidence also shows that unique attrib utes may be more likely to be used in decisions than other attributes. Tversky (1977) reported the exis tence of the extension effect whereby attributes with values th at varied between items were h eavily weighted in judgments of differences between items. Likewise, Me din, Goldstone, and Gentner (1993) found that variation along an attribute was a core determin ant of the attribute be ing noticed. Building on this work, Goldstone, Medin, and Halberstadt (1997) demonstrated that the importance given to a particular product attribute was re lated to the variability of va lues along it. In their study, attributes that were otherwis e ignored had a major influence on judgment when variation in values was introduced along them. Additional support for the use of unique attributes comes from the consumer choice literature. Johnson (1989) determined that the im pact of an attribute on a decision was increased when the attribute wa s shared among the alternatives considered and when each alternative possessed a unique value along it. Like before, these studies further contribute to the notion that there is a relationship between the distribution of attribut e values along an attribute and its use. However, the literature reviewed here builds a case that unique attribut es are more perceptually available and have greater impact in subsequent decisions than othe r product attributes. This accessibility may be derived from two sources: a natural sensitivity to perceive these attr ibutes due to their affordance

PAGE 18

18 of identification and recognition pro cesses or the increased attention paid to them as a result of their higher information value. Consumer Goals Affect Attribute Use The previous section described how different attribute distribu tions are critical to conceptual formation and indi vidual item differentiation. Th e reviewed evidence provides conflicting support for the greater ac cessibility of category diagnostic and unique attributes. This section proposes a framework that disentangles the disparate findings. Specifically, drawing elements from the theories of goal-based per ception (Janiszewski 2007), event coding (Hommell, Mssler, Aschersleben, and Prinz 2001), and diagnostic recognition (Schyns 1998), I hypothesize that consumers task s during product exposure prime a ttention towards attributes with different distributions of values, depending on how these distributions afford initial task goals. The resulting increase in attribute accessibi lity persists after exposur e, resulting in greater reliance on this attribute duri ng subsequent decisions. It is clear from the reviewed literature that there is disagreement about the link between the distribution of values along an attribute, its accessibility, and its weight in future choices. The lack of a parsimonious body of evidence sugge sts the presence of a m oderating factor. The most obvious difference between the studies reported are the tasks employed in the experiments. When subjects learned sets of items, category diagnostic attr ibutes became more accessible (Schyns and Rodet 1997) and affected future decisions (Goldstone 1994; Tversky 1977). However, when subjects made judgments about individual items or made choices between items, unique attributes were more accessible and sh aped judgments and choices (Goldstone, Medin, and Halberstadt 1997; Johnson 1989; Medin, Goldstone, and Gentner 1993). Thus, it seems that temporary task goals affect use of attribut es, depending on these attributes distributions.

PAGE 19

19 A number of recent (and related) theories speak to this possibility. Janiszewski (2007) proposed a goal-directed account of perception. He contends that the goal directed nature of the perceptual system causes individuals to only perceive the events in the environment that are instrumental to an ongoing task and to not perceive irrelevant elements. In other words, active goals increase the perceptual salience of cues that enable goal fulfillment and decrease the salience of cues not critical to goal comple tion. This link between perception and action is explained using the Theory of Ev ent Coding (Hommel et al. 2000). At the core of the Theory of Event Coding is the notion that repres entations of goals (or planned actions) share a common, ye t abstracted, language with perc eptual representations. This is different from traditional information processing accounts where perception and action planning operate in different, distinct doma ins. As the result of sharing a common representational language, intera ctions between goals and percep tual activity are possible and routine. For example, when individuals inte nd to perform an action (i.e., possess a goal), perceptual features in the indivi duals environment related to this action are primed in parallel. This has two effects. First, the priming results in greater salience for features relevant to the active goal. Second, features unrel ated to the goal become less pe rceptually accessible. Most, Scholl, Clifford, and Simons (2005) provide substantial evidence for these effects. The framework of diagnostic recognition (S chyns 1998) predicts similar outcomes. Schyns noted that the ability to recognize and identify elements of the environment depends on the interaction of two factors: available perceptual information and task constraints. Through a review of experimental results, Schyns (1998) builds the case that ambiguous stimuli are classified using cues that afford accomplishment of previous or current tasks. Further, cues that were not diagnostic with regard to the attempted task did not ente r conscious awareness. Finally,

PAGE 20

20 but importantly, his experiments de monstrated the persistent sensit ization of particular stimulus features as a result of their previous dia gnosticity. This differs somewhat from the aforementioned work (Hommel et al. 2000; Jani szewski 2007; Most et al. 2005), because the studies presented by Schyns (1998) indicate that the increased sensitization of goal-related perceptual cues persists even after a particular task has been completed. Interestingly, this explains the continued of sensitization for cate gory diagnostic attributes in Goldstone (1994). Therefore, subsequent to product exposure, pr oduct judgments should demonstrate a greater reliance on attributes with distributions of values that afford completion of consumers goals at the time of initial product exposure. It is easy to hypothesize how consumers co ncrete tasks link to more ephemeral goals. This paper investigates two consumer tasks: consideration set formation and product choice. When forming consideration sets, consumers atte mpt to simplify their choice environments and engage in a categorization-like process (Chakr avarti and Janiszewski 2003; Chakravarti, Janiszewski, and Ulkumen 2006). When consumers choose, they must select a single item and therefore engage in individual item recognition and identification in order to engage in a compensatory process. The central hypothesis of this paper is that the tasks consumers perform during the initial exposure to a set of products prime the perceptu al accessibility of particular product attributes when the distributions afford the task goal. This sensitization persists ev en after task completion, resulting in greater sensitivity to attribute level tradeoffs along th ese attributes during subsequent decisions. Category diagnostic attributes will dominate judgments when exposure is accompanied by tasks that require simplification of the choice environment (consideration set

PAGE 21

21 formation). Conversely, unique attr ibutes will drive future produc t perceptions when individual item identification is critical during initial exposure (choice).

PAGE 22

22 CHAPTER 2 THE PILOT STUDY The pilot study was designed to determine th e basic effects of the relationship between how consumers tasks and distri butions of attribute values. He re, their impact on perceived attribute importance is investig ated. Consumers perceptions of attribute importance are a fundamental issue for managers. Changes in per ceived importance affect both the accessibility and perceived diagnosticity of particular attributes, which influence the likelihood that a particular attribute will be uti lized in future product judgments and choices (Feldman and Lynch 1988; Lynch, Marmorstein, and Weigold 1988). In line with the genera l hypotheses of this paper, category diagnostic attr ibutes should be perceived as more important when initial exposure is accompanied by consideration set fo rmation and unique attributes should gain importance when exposure is coupled with a choice task. Method Participants and Design Seventy-five undergraduate stude nts from a large southeastern university participated in this experiment in exchange for extra credit. A two (shopping task: form a consideration set vs. choose the best product) by tw o (attribute dist ribution type: category diagnostic vs. unique) mixed design was used. Shopping task was the between-subjects factor and attribute distribution ty pe was the within-subjects factor. Stimuli This experiment used four actual te e-shirts during the exposure phase (Figure 21). Two of the tee-shirts had a rounded crew co llar and two of the tee-shirts had a wedged vneck collar. All of the tee-shirts were a di fferent color (grey, white, blue, and red). By definition, neck type was the category diagnostic attribute and color was the unique attribute. The shirts were identical in every other way. Th e shirts were hung at the front of the laboratory

PAGE 23

23 and were easily viewed from any of the computer stations in the lab. The shirts remained visible throughout the entire experiment. The test phase of the experiment utilized a pen and paper questionnaire. Here, subjects were asked to circle the shirts that they included in their consideration set or the shirt that they chose during the initial product exposure. Next, the subjects were presen ted with a list of product attributes typical of tee-sh irts including: size, material t ype, brand, price, color, shape, weight, and thickness. Subjects were asked to rate how important each attribute was to them when thinking about tee-shirts. Of these attr ibutes, shape (category diagnostic) and color (unique) were the cri tical attributes. Procedure. Upon entering the laboratory, subjects we re seated and were told to wait for further instructions. Once all subjects were seated, the group of subj ects was given a shopping task to perform while examini ng the tee-shirts. The group receiv ing the consideration set task was told to form a considerati on set of the two shirts they liked the best, while the group receiving the choice instruction was told to choo se their favorite shirt. The subjects were allowed to examine the shirts for as long as th ey desired. Following this step, the subjects finished other, unrelated experiments, using the computers located in front of them. Upon completion of these experiments, the subjects we re given a paper questionnaire which they were asked to complete. Depending on their initial task, the questionna ire asked them to circle the brand(s) composing their consideration set or ch oice. Next, subjects were asked to rate the importance of various attributes of tee-shirts, using a 1-10 scale. The experiment concluded once subjects had finished the one-page survey. Results A 2 (shopping task: form a consideration se t vs. choose the best product) by 2 (attribute distribution type: ca tegory diagnostic (shape) vs. uni que (color)) repeated-m easures

PAGE 24

24 MANOVA was performed on the importance ratings, with shopping task as the between-subjects variable, and attribute di stribution type as the w ithin-subjects factor. In line with the hypotheses, there was a two-way interaction between shopping task and attribute distributions on subjects ratings of attribute importance ( F (1, 73) = 3.74, p = .057). Simple effect tests revealed a significant effect of instruc tion on importance rati ngs of the unique attribute (color) ( MConsider = 7.30, MChoose = 8.25; F (1, 73) = 3.92, p = .051). However, there was no effect of instruction on importance ratings for the categor y diagnostic attribute (shape) ( MConsider = 7.67, MChoose = 7.77; F < 1). There were no effects of instruction alone ( F (1, 73) = 1.56, p = .20) or attribute type alone ( F < 1) on subjects ratings of the critical attributes. Fi gure 2-2 depicts these results. Discussion The pilot experim ent provides evidence that consumers tasks and distributions of attribute values affect perceptions of attribute importance. The experiment also shows that sensitization towards particular attributes persis ts, even after task completion. Indeed, the fact that the attribute importance ratings did not take place until 20-25 minutes after the initial consideration set formation (choice), demonstrates the strength of the effe ct. Interestingly, there was a larger effect of initial task on importan ce ratings for the unique attribute than for the category diagnostic attribute. This may have occurred due to the methodological imperfections of this experiment. Here, the shape of the shir t (neck-type) was always category diagnostic and the color of the shirt was always unique. One possibility is that the unique colors of the shirts somehow made individuals more sensitive to differences in task. If this is the case, then counterbalancing assignment of attr ibute values distributions to a particular attribute should lead to effects (or null effects) for both the category diagnostic and unique attributes. Experiments 1a and 1b test for this possibility.

PAGE 25

25 Figure 2-1. Tee-shirts used as st imuli in the pilot experiment.

PAGE 26

26 6 6.5 7 7.5 8 8.5 9 Consider ChooseImportance Rating Shape Color Figure 2-2. Reported importance rati ngs for the pilot experiment.

PAGE 27

27 CHAPTER 3 EXPERIMENT 1 One limitation of the pilot experiment was th at it did not counte rbalance assignment of attribute type to attribute value distributions. This experiment corrects this limitation in two ways. First, assignment of attribute value distri bution to attribute type is counterbalanced, which should counteract any spurious interaction between attribute type and distribution. Second, experiment 1 utilizes different types of products, which is not only more ecologically valid, but also allows more attribute type/attribute distribution situations to be examined. Further, while the pilot experiment demonstrated that attribute va lue distributions and consumers tasks affect percepti ons of attribute importance, the experiment does not show that consumers will actually rely on these attributes when making a decision. Experiments 1a and 1b show that attribute distributions have an effect on consumers actual judgments and decisions. Specifically, experiment 1A examines consum ers perceptions of new product novelty, and experiment 1B examines consumers willingness to pay for new products. Experiment 1A When managers introduce a new product to the market, they face a critical decision regarding product positioning. An elem ental part of this decision is to determine how new they want the product to seem to consumers. Ind eed, the degree of perceived product novelty has a host of implications for consumers judgment s about the product (Alexander, Lynch, and Wang 2008). This experiment examines how a cons umers task during pr oduct exposure and the distribution of attribute values along a product attribute affect s the perceived novelty of new product introductions. In line with the prev ious hypotheses, consumers engaged in consideration set formation at the time of exposure should sensitize towards th e category diagnostic attribute. This will cause

PAGE 28

28 them to be more apt to notice changes in va lues along that attribute, leading to greater perceptions of product novelty for items that differ along that attribute than for new products that differ on the unique attribute. Conversely, when c onsumers initially engage in choice, they will become sensitized towards unique attributes, and will report greater novelty for new products that differ along the unique attribute than for ne w products that differ on the category diagnostic attribute. Method Participants and Design. Forty-eight undergraduate stude nts from a large southeastern university participated in this experiment in ex change for extra credit. The design was a three (product category: tee-shirts, sunglasses, fris bees) by two (shopping task: form a consideration set vs. choose the best produc t) by two (new pr oduct attribute difference type: category diagnostic vs. unique) mixed desi gn. Product category and new produc t attribute difference were within-subject factors, and task inst ruction was a between-subjects factor. Stimuli. Pictures of products, each consisting of three product attr ibutes, were used as the main experimental stimuli. This served two purposes. First, it prevented consumers from using information about particular products that could influence th eir future decisions. Second, creating the pictures allowed for tight control over the attribute va lues that would be shown to the subjects. The product categories used and attr ibutes manipulated are shown in Table 3-1. Because the main interest of this study was how different distributions of attribute values interact with consumer task goals, each product attribute was assigned a different distribution of attribute values. Three types of di stributions were used. The first type of distribution, labeled the common distribution, gave each brand the same attribute value. The second type of distribution, labeled the categor y diagnostic distribution, gave the first two brands the same

PAGE 29

29 attribute value, and the second two brands a different attribute value. The third distribution, labeled the unique distribution, ga ve each brand a unique value. Table 3-2 depicts a schematic representation of the attribute value assignment in the exposure phase of the experiment. Each brand s attribute values were dependent on both the product category used and the attribute type. In Ta ble 3-2, a value of one represents the smallest attribute value possible for that product category and attribute while a value of four represents use of the largest attribute value for that product category and attribute. Assignment of attribute value distributions to particul ar attributes was counterbalanced to average the effects of interactions between the distribution type and a specific attribute. An example of this for the tee shirt category is shown in Figure 3-1. In Schedul e 1, shirt length is the common attribute, neck type is the category diagnostic at tribute, and sleeve strap widt h is the unique attribute. In Schedule 2, neck type is the co mmon attribute, sleeve strap wi dth is the category diagnostic attribute, and shirt length is th e unique attribute. In Schedule 3, sleeve strap width is the common attribute, shirt length is the categ ory diagnostic attribute, and neck type is the unique attribute. The test phase of the experiment utilized a different set of stimuli. During this phase, the subjects were shown a new model of the pr oduct category alongside one of the products that had been present during the exposure phase of the experiment. The new product differed from the old product in one of two ways, depending on which attribute the new product possessed a different value upon (in both cas es, the new product differed fr om the old product attribute values by only one step, along one attribute). Ne w stimuli where the attribute value difference occurred on the attribute which had been previously category diagnostic were named category diagnostic products (NewCatDiag), and stimuli that differed on the attribute that previously had a unique distribution were named unique products (NewUnique). Table 3-3 provides a schematic

PAGE 30

30 representation of the new products and also indica tes the products from the exposure phase of the experiment to which the new products were compared. Using the same technique, new products were created for all three product categories. Procedure. Upon entering the laboratory, experi mental participants were randomly placed into one of the two instruction conditions. The participants assigned to the consider task received the following instructions: In the next portion of the experiment, you will be shown pictures of four different models of a speci fic type of product. When you are shown these pictures, imagine that you are considering a futu re purchase, and that you want to narrow your selections down to 2 items. Based upon the pictures alone, FORM A CONSIDERATION SET of 2 models that you are most interested in. You may be asked to explain the way you grouped the pictures in order to FORM A CONSIDERATION SET Participants assigned to the choose condition received these instructions: In the next portion of the experiment, you will be shown pictures of four different models of a speci fic type of product. When you are shown these pictures, imagine that you will be buying one of these specific models. Using the information provided, CHOOSE THE BEST PRODUCT. You may be asked to explain the reasons you used to CHOOSE THE BEST PRODUCT of the group. After the subjects had received the task instructions, they we re shown four models of one of the three product categories (s elected randomly). The models were presented simultaneously in a side by side manner. The leftmost product was labeled Brand A, and the rightmost product was labeled Brand D. After ten seconds, a c ontinue button appeared at the bottom of the screen. Although this was done to prompt them to move to the test phase of the experiment, subjects could continue viewing the products until they were ready to move forward.

PAGE 31

31 Following the exposure phase of each product cat egory, subjects entered the test phase of the experiment (for that categor y). Here, they received the following instruction: Now that you have (formed a consideration set/chosen the best product) we are interest ed in what you think about 8 new models of (product category name). On each screen, the model on the left will be a new version of this brand of (product category name) and the one on the right will be one that you have previously seen. We are interested in how new the new model seems compared to the old model you are being shown Please indicate this using the scale presented below the pictures. The pictures will remain pres ent while you make your decision. Following the instructions, one of the eight possible comparisons was randomly selected to be presented. During presentation, the new product was shown on the left portion of the screen and the old product was pr esented on the right portion of the screen. On the bottom of the screen was a scale that ranged from 1-10, with th e phrase Not New at All anchoring the left hand side and the phrase Very New anchoring the right hand side of the scale. After this decision had been made, another of the remaini ng comparisons was randomly selected to appear, and presentation and judgments c ontinued until all of the remaining comparisons had been made. At this point, one of the remaining product catego ries was selected, after which the exposure and test processes occurred again, unt il the subjects had completed all three of the product categories. Results A 3 (product category: tee-shirts, sung lasses, frisbees) by 2 (shopping task: form a consideration set vs. choose the best product) by 2 (new pr oduct attribute difference type: category diagnostic vs. unique) repeated-m easures MANOVA was performed on subjects novelty ratings, with shopping task as the betw een-subjects variable, and product category and new product attribute difference type as within -subjects factors. There was no interaction between product category, shopping task, and ne w product difference type on judgments of

PAGE 32

32 novelty ( F (1, 45) = 1.90, p = .16), so the rest of the analys es were collapsed across the product category factor. After the subjects judgments had been a ggregated across product category types, a 2 (shopping task: form a consider ation set vs. choose the be st product) by 2 (new product attribute difference type: cat egory diagnostic vs. unique) repeated-measures MANOVA was performed on the data, with shopping task as a between-subjects factor and new product attribute difference type as a within-subjects factor. The analysis did not re port a significant effect of new product attribute difference type (F (1, 46) = 2.03, p = .16) or shopping instructions alone ( F (1, 46) = 1.67, p = .20) on judgments of new product novelty. However, the analysis did reveal the exp ected two-way interacti on between shopping task and new product attribute difference type on the perceived novelty of new products ( F (1, 46) = 12.64, p = .001). Figure 3 depicts these results. Simple effects tests revealed a significant effect of instruction on new products that di ffered along the unique dimension (NewUnique) ( MConsider = 2.14, MChoose = 2.68; F (1, 46) = 4.59, p = .04), indicating changes in its use. However, these tests also reported no effect of instruction when the new products differed on the category diagnostic dimension (NewCatDiag) (MConside r = 2.46, MChoose = 2.54; F < 1). Discussion This experiment demonstrated that increased sensitivity caused by the interaction between the distribution of values along an attribute and consum ers tasks affect consumers subsequent attribute use. As before, the result s confirmed the experimental hypotheses. The task goals active during exposure affected the perceptual availability of particular product attributes, depending on the distribution of values along th at attribute during exposure. This caused subjects to be more sensitive to attribute level differences along this attribute when judging new products, affecting their percepti ons of product novelty. As in the pilot experiment, the results

PAGE 33

33 indicated that the strongest eff ects were for items that differed along the attribute with the unique distribution. Subjects who formed consideration sets did not perc eive these products to be as novel as subjects who made a choice during exposure. Experiment 1B Experiment 1A found a link between consumer tasks during initial product exposure and the distribution of values along an attribute on consumers later judgments of product novelty. This experiment extends the ecol ogical validity of this experime nt by demonstrating that these same factors influence consumers willingness to pay for new items. Assuming that individuals are willing to pay more for items that are perceived as novel, a consumer should be willing to pay more for items th at differ on the attribute that they sensitize to as a result of her task during exposure. Thus, consumers should pay greater amounts for products that differ on the category diagnostic attribute when they initially form consideration sets than when they choose. Conversely, consumers should pay more for products that differ on the unique attribute when they are asked to make a c hoice than when they form consideration sets. Method Participants and Design Sixty-six undergraduate students from a large southeastern university participated in this experiment in ex change for extra credit. The design was a three (product category: tee-shirts, sunglasses, fris bees) by two (shopping task: form a consideration set vs. choose the best produc t) by two (new pr oduct attribute difference type: category diagnostic vs. unique) mixed desi gn. Product category and new produc t attribute difference were within-subject factors, and task inst ruction was the between-subjects factor. Stimuli. The stimuli used in this exposure and te st phases of this experiment were exactly the same as those used in experiment 1A.

PAGE 34

34 Procedure. Upon entering the laboratory, experi mental participants were randomly placed into one of the two instruction condition s. The instructions a nd exposure phase of the experiment were identical to that of experi ment 1a. Following the exposure phase of each product category, subjects entered th e test phase of the experiment (for that category). Here, they received the following in struction: Now that you have (for med a consideration set/chosen the best product) we are interested in what you think about 8 new m odels of (product category). On each screen, the model on the left will be a new version of this brand of (product category) and the one on the right will be one that you have prev iously seen. We are interested in how much more you would be willing to pay for the new model compared to the old model you are being shown Please indicate this using th e numbers presented below the pictures. The pictures will remain present while you make your decision. Following the instructions, one of the eight possible comparisons was randomly selected to be presented. During presentation, the new product was shown on the left portion of the screen and the old product was pr esented on the right portion of the screen. On the bottom of the screen was a scale that ranged from -$2.50 to +$ 2.50 coupled with the que stion: Compared to the model on the right, how much more would y ou be willing to pay (in dollars) for the NEW item? After this decision had been made, anothe r of the remaining comparisons was randomly selected to appear, and presentation and j udgments continued until all of the remaining comparisons had been made. At this point, one of the remaining product categories was selected, after which the exposure and test processes occurred again, until all three of the product categories had been completed.

PAGE 35

35 Results A 3 (product category: tee-shirts, sung lasses, frisbees) by 2 (shopping task: form a consideration set vs. choose the best product) by 2 (new pr oduct attribute difference type: category diagnostic vs. unique) repeated-m easures MANOVA was performed, with product category and new product attribute di fference as within-subject fact ors, and task instruction as the between-subjects factor. There was no interaction betwee n product category, shopping task, and new product attribute difference type on subjects willingness to pay ( F (2, 128) = 1.96, p = .145), so the results were collapsed along the product category factor for the rest of the analyses. Further analyses revealed the expected two-way interaction betw een shopping task and new product attribute difference type on subj ects willingness to pay for new items (F (1, 64) = 5.54, p = .022). Figure 4 shows these results. Simple effects tests revealed that instruction had an effect on subjects willingness to pay for ne w items that differed on the unique attribute (NewUnique) ( MConsider= .203, MChoose = .487; F (1, 64) = 8.24, p = .006), but not on subjects willingness to pay for items that differed along the category diagnostic attribute (NewCatDiag) ( MConsider= .215, MChoose = .325; F (1, 64) = 1.11, p = .295). There was also a significant effect of new product attribute difference type such that subjects were, on average willing to pay more for unique products than for cat egory diagnostic products ( MUnique = .358 MCatDiag = .275; F (1, 64) = 4.07, p = .048). Subjects were also willing to pay more for products (regardless of type) when they initially chose than when th ey formed consideration sets (MConsider = .209 MChoose = .406; F (1, 64) = 4.34, p = .041). Discussion Experim ent 1B demonstrates that the ta sk active during product exposure and the distribution of values along a product attribute prime subjects to be more sensitive towards attribute level differences along the task compatible attribute. Here, this was reflected in

PAGE 36

36 subjects willingness to pay for new items, depending on which attribute the new product differed along. Depending on their initial task, subjects were willing to pay more for new products that differed along the unique attribute than for new products that differed along the category diagnostic attribute. When the new pr oduct differed along the uni que attribute, they were willing to pay, on average, approximately $0.28 more for the same item when they had made a choice than when they formed a cons ideration set during ini tial product exposure. Interestingly, subjects shopping tasks did not change their per ceptions of attribute importance (pilot study), perceived product novelty (experiment 1A), or willingness to pay for new products (experiment 1B) for category diagno stic attributes or products. While it was initially hypothesized that this difference was due to a spurious interaction between a color type attribute having the unique dist ribution, the counterbalanced de sign of experiments 1a and 1b would have negated this effect. However, ther e are several explanati ons for this phenomenon. The first possibility is that formation of consideration sets does not lead to use of the category diagnostic attribute. A second possibility is that choos ers sensitize to both category diagnostic and unique attributes in their decisions. Previous research has shown that choice is a staged process where consumers examine the entire array of options, narrow the options into an acceptable group, and then make a final decision from this subset (Bettman and Park 1980; Chakravarti, Janiszewski, and Ulkumen 2006; Nedunga di 1990). The first phase of this process is the formation of a consideration set, which is typically composed of products that are similar to one another either because they share the same attribute values or because they have comparable values along the same attribut e (Chakravarti and Janiszewski 2003). Thus, individuals engaged in choice, in addition to using unique attr ibutes, would also incorporate information about the category diag nostic attribute in their product judgments. Specifically, these

PAGE 37

37 individuals first utilize the categ ory diagnostic attribute to form their consideration set and then use the unique attribute to make a choice. Becau se the sensitization of a particular attribute persists even after goal completion (Schyns 1998), the initial formation of a consideration set, even when followed by a choice task, could increas e use of the category diagnostic attribute. If choosers utilize both cat egory diagnostic and unique attr ibutes, then the design of experiments 1A and 1B may have contributed to the lack of an effect for products that differ on the category diagnostic attribute. In these stud ies, subjects new product evaluations could have been based on one or any combination of the ne w products attributes. T hus, placing consumers in a choice situation where attribute use is mutual ly exclusive may provide better insight into the use of category diagnostic attri butes by choosers because it would force them to use only the most important attribute to them. Expe riment 2 examines this possibility. The use of multiple attributes by those maki ng choices raises an additional question as well: Is the use of multiple produc t attributes due to goal-base d priming, or is it due to motivational issues? While the premise of this pa per contends that goals make some attributes more perceptually available, the description of the staged choice process given above makes attribute use seem motivational. This explan ation suggests that consumers making choices examine category diagnostic attributes before examination of unique at tributes. The initial utilization of the category dia gnostic attribute carries over to future product judgments, which leads choosers to appear sim ilar to considerers when eval uating products that differ only along the category diagnostic attribute. Alternatively, choice causes category diagnostic attributes and unique attributes to be primed in parallel. As said earlier, making a choice, by definition, requires the selection of a product that is different from all of the other products. With rega rd to distributions of attribute

PAGE 38

38 values, both category diagnostic and unique distributions provide differentiating information. This is obviously more granular for unique attr ibutes, but category diagno stic attributes also differentiate products albeit in a broad manner. Thus, in line with a goal-directed account of perception, choice may prime receptivity towards all differentiating dimensions simultaneously. In other words, whereas a motivational account states that choosers move from category diagnostic to unique attributes in their deci sion process, the priming account posits that sensitization to both attribute di stributions occurs at once. Expe riment 3 examines this question explicitly.

PAGE 39

39 Table 3-1. Schematic for attribute value assignments for the exposure stimuli. Model 1 Model 2 Model 3 Model 4 Common1111 Category Diagnostic1144 Unique1432

PAGE 40

40 Table 3-2. Product categories and attr ibute types used in experiment 1. Tee Shirts Sunglasses Frisbees V-Neck LengthShade DarknessDisk Size Torso LengthArm LengthNumber of "Speed Holes" Shoulder Strap WidthBridge HeightSize of "Speed Holes"

PAGE 41

41 Table 3-3. Schematic representation of the test phase stimuli in experiment 1. Differing Attribute New Old Category DiagnosticCommon1 1 Cat Diag2vs.1Brand A Unique1 1 Common1 1 Cat Diag2vs.1Brand B Unique4 4 Common1 1 Cat Diag3vs.4Brand C Unique3 3 Common1 1 Cat Diag3vs.4Brand D Unique2 2 UniqueCommon1 1 Cat Diag1vs.1Brand A Unique2 1 Common1 1 Cat Diag1vs.1Brand B Unique3 4 Common1 1 Cat Diag4vs.4Brand C Unique4 3 Common1 1 Cat Diag4vs.4Brand D Unique1 2

PAGE 42

42 Brand ABrand BBrand CBrand D Schedule 1 Schedule 2 Schedule 3 Figure 3-1. Example of counter balancing in the tee-shirt stimuli used experiment 1.

PAGE 43

43 1 1.5 2 2.5 3 ConsiderChooseNovelty Rating Category Diagnostic Unique Figure 3-2. Results of experiment 1A.

PAGE 44

44 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5Consider ChooseWTP ($) Category Diagnostic Unique Figure 3-3. Results of experiment 1B.

PAGE 45

45 CHAPTER 4 EXPERIMENT 2 Experiments 1A and 1B demonstrate that consumers shopping tasks and attribute distributions affect new produc t judgments and willingness to pay. Moreover, both experiments showed that consumer tasks had more of an influence when the new products differed along the unique attribute. One possible r eason why the earlier experiments did not find differences in use of the category diagnostic attribut e may be that choosers used th e category diagnostic attribute in addition to the unique attribute in their judgmen ts. This made it difficult to determine whether subjects tasked to cons ideration set formation used the cate gory diagnostic attribute less than predicted or if subjects that we re tasked to make a choice used the category diagnostic attribute more than expected. This experiment attempts to make this distinction more obvious by creating a choice environment that requires subjects to use either the category diagnostic or the unique attribute exclusively. Here, we examine the situation where, after a task-directed exposure to a set of products, three new products are seen. Two of these products possess more similar attribute values along the category diagnostic attribute (e.g., Brand A and Brand B), and two have more similar values along the unique attribute (e.g., Brand B and Brand C). Subjects are then asked to choose which two of the three new products are most similar to each other. We predict that when consumers initially engage in consideration set formati on, that they will group the new products based on their similarity along the category diagnostic attri bute more than when they make choices during initial product exposure. Method Participants and Design. Fifty-one undergraduate stude n ts from a large southeastern university participated in this experiment in ex change for extra credit. The design was a three

PAGE 46

46 (product category: tee-shirts, sunglasses, fris bees) by two (shopping task: form a consideration set vs. choose the best produc t) by two (dependent measure setup) mixed design. Product category and dependent measure types were within -subject factors, and ta sk instruction was the between-subjects factor. Stimuli. This experiment used the same stimu li as experiment 1 in the exposure phase. Different stimuli were used in the test portion of the experiment. Because the goal of the experiment was to determine how tasks present during exposure affected the exclusive use of product attributes, the stimuli at test phase created a conflict be tween the uses of two of the attributes in forming a decision. He re, subjects were presented with pictures of three brands from a given product category. Brands 1 and 2 possess ed more similar values on either the category diagnostic (or unique) attribute, and Brands 2 and 3 had more similar values on the unique (or category diagnostic) attribute. Along the common attribute, all three brands possessed the same value. Table 4-1 depicts the assignment of attrib ute values for both sets of dependent measures used. Note that the attribute/di stribution labels refer to the ty pe of distribution a particular attribute had in the exposure phase of the experiment. Procedure. Upon entering the laboratory were randomly assigned to one of the two shopping task conditions. The instructions and exposure phase of the experiment were identical to that of experiments 1a and 1b. After the su bjects were done viewing the exposure stimuli, they moved into the test phase of the experiment Here, subjects were presented with three new brands of the product. Two of the brands were more similar on the attribute that had been previously category diagnostic, and two were more similar on the attribute where each brand had possessed a unique value. The subjects were inst ructed that their task was to select the two brands that were most similar to each other (Brand 1 and Brand 2, or Brand 2 and Brand 3).

PAGE 47

47 Following their selection, the subjec ts were presented with the s econd set of three brands, from which they had to again make a similarity deci sion. The order of the two sets of three products was randomized. Following this, another produc t category was randomly selected for exposure and test until all three product categories had been seen. Results A 3 (product category: tee-shirts, sung lasses, frisbees) by 2 (shopping task: form a consideration set vs. choose the best produc t) by 2 (dependent measure type) repeatedmeasures MANOVA was performed, with shopping task as a between-subjects variable and product category and dependent meas ure type as within-subjects f actors. The percentage of groupings made using the categ ory diagnostic attribute wa s the critical measure. Multivariate tests revealed that neither product category ( F (2, 98) = 1.41, p < .25) nor dependent measure type ( F < 1) significantly interacted with shopping task on the type of groupings made so the results were collapsed across both within-subject measures for the remainder of the analyses. Figure 4-1 depicts the critical result. As hypothesized, shopping ta sk had a significant effect on product groupings. Subject s who formed consideration se ts were more likely to group the new products using the categor y diagnostic attribute than subjec ts who were asked to make a choice during initial product exposure ( MConsider = .53, MChoose = .42; F (1, 49) = 5.64, p = .022). Discussion This experim ent again shows that the ta sk goals during product exposure affect individuals sensitivity towa rds attribute level differences depending on that attributes distribution. When subjects formed consideration se ts during exposure, they were more likely to group the new products based on the products sim ilarity along category diagnostic attribute But, when subjects made choices during initial exposure, they were more likely to group the new

PAGE 48

48 products based upon their similarity upon the unique at tribute. This supports the claim that the lack of a difference in product judgments seen in experiments 1A and 1B was most likely due to choosers incorporating category diagnostic information in their judgments.

PAGE 49

49 Table 4-1. Schematic of dependent measures used in the test portion of experiments 2 and 3. DV 1 Model A Model B Model C Model A Model B Model C Common111111 Category Diagnostic124223 Unique134233 DV 2

PAGE 50

50 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 ConsiderChoose% New Groupings Category Diagnostic Unique Figure 4-1. Results of experiment 2.

PAGE 51

51 CHAPTER 5 EXPERIMENT 3 Experiment 2 provided evidence that categor y diagnostic and unique attributes are used when initial exposure is accompanied by choice. Ho wever, it does not clarify whether the use of both attribute distribution types is a motivatio nal or priming issue. Indeed, one can easily imagine how choosers could be more involved or more meticulous in their information search than those who were simply forming consider ation sets. Using two different techniques, experiments 3A and 3B examine whether at tribute use is motivational or primed. Experiment 3A This exper iment attempts to determine whethe r the use of category diagnostic attributes and unique attributes by those forming choices is due to subjects sequential pursuit of them (motivational account) or if the choice task pr imes receptivity towards both attribute distribution types simultaneously. This experiment tests these hypotheses by examin ing subjects initial product viewing times. If motiva tional influences are present, individuals in the choice task should take longer to view the in itial products than subjects form ing consideration sets, because they have to form considerati on sets prior to choice. However, if there are no differences in viewing times, then it seems more likely that both category diagnostic an d unique attributes are primed when individuals prepare to choose. In addition to task-related motivations to ut ilize more product inform ation, individuals may also have dispositional biases that affect informa tion search behavior. This bias is often reflected in individuals need for cognition (NFC). Gene rally, an individuals score on the need for cognition scale determines her tendency to engage in and enjoy effortful cognitive endeavors (Cacioppo and Petty 1982; Cacioppo, Petty, and Kao 1984) such that higher scores reflect a greater willingness to engage in these activities. I ndeed, individuals with high NFC have been

PAGE 52

52 shown to incorporate a greater number of product attributes in their decisions (Chuna and Laran 2009; Levin, Huneke, and Jasper 2000; Wood an d Swait 2002). If greater NFC leads to a conscious motivation to search for more informati on then we would expect that individuals with high NFC should have longer initial viewing times than low NFC individuals. Another possibility is that the dispositiona l bias provided by NFC may simply act as a baseline prime which pre-sensiti zes individuals towards either e fficient or accurate attributes. Along these lines, previous research has shown that high NFC consumers place greater emphasis on accuracy than consumers with low NFC (Levi n, Huneke, and Jasper 2000). If NFC operates through priming, then high NFC subjects should be more likely to group the new products based on their similarity along unique at tribute while subject s with low NFC subjects should group the new products using the category di agnostic attribute. Moreover, in opposition to the motivational account, the priming explanation predicts no differences in viewing time due to subjects NFC. Method Participants and Design. Thirty-eight undergraduate stude nts from a large southeastern university participated in this experiment in ex change for extra credit. The design was a three (product category: tee-shirts, sunglasses, fris bees) by two (shopping task: form a consideration set vs. choose the best produc t) by two (need for cognition: high vs. low) by two (dependent measure setup) mixed design. Product categor y and dependent measure types were withinsubject factors, and task inst ruction and need for cognition were between-subjects factors. Stimuli. This experiment utilized the same stimuli as in Experiment 2. Procedure. Before beginning the experiment, s ubjects entering the laboratory were randomly assigned to one of the two shopping task conditions. The instructions and exposure phase of the experiment were identical to the previous expe riments. However, in this experiment, subjects viewing of the initial produ cts in the exposure phase was timed. The test

PAGE 53

53 phase of the experiment was the same as experi ment 2. After all of th e product categories had been exhausted, the subjects completed the 18 it em need for cognition scale (Cacioppo, Petty, and Kao 1984). Results Before any analyses were conducted, subjec ts were placed into a need for cognition category (high vs. low) based upon their res ponses to the NFC scale. The mean was ( MNFC = 87.13). Subjects who scored lower than the mean were marked as low NFC and subjects scoring higher than the mean were marked as high NFC. New product groupings A 3 (product category: tee-shirts, sungla sses, frisbees) by 2 (shopping task: form a consideration set vs. choose the best product) by 2 (need fo r cognition: high vs. low) by 2 (dependent measure type) repeated-measures MANOVA was performed, with shopping task and need for cognition as between-subjects variable s, and product category and dependent measure type as within-subjects factors. The percenta ge of groupings made usi ng the category diagnostic attribute was the critical measure. Multivariate tests revealed that neither product category ( F (2, 68) = 1.90, p = .16) nor dependent measure type ( F < 1) significantly interacted with shopping task on the type of groupings made. Likewise, no significant interac tion existed between pro duct category type and need for cognition (F < 1) on new product groupings. Additionally, neither the interaction between shopping tasks, need fo r cognition, and product category ( F < 1) nor the interaction between shopping tasks, need for co gnition, and dependent measure type ( F < 1) nor the interaction between shopping tasks, need for cognition, product category, and dependent measure type were significant ( F < 1), so the results were collapsed across both within-subject measures for the remainder of the analyses.

PAGE 54

54 Figure 5-1 depicts the new product groupings. Results indicated a marginally significant two-way interaction between subjects initial task instruction and need for cognition on the types of new product groupings (F (1, 34) = 3.74, p =.06). Within the low need for cognition group, task instructions signif icantly affected subjects new produc t groupings; subjects grouped the new products based on their similarity along the ca tegory diagnostic attribute more when they initially formed consideration sets than wh en they made choices during initial exposure ( MNFCLow/Consider = .77, MNFCLow/Choose = .39; F (1, 35) = 18.98, p < .001). Subjects tasks also affected the product groupings within the high need for cognition group. Again, new product groupings demonstrated greater reliance on the ca tegory diagnostic attribut e when consideration sets were initially formed than when choices were made during product exposure ( MNFCHigh/Consider = .51, MNFCHigh/Choose = .33; F (1, 35) = 4.32, p < .05). Further, low NFC subjects were more likely to group the new produc ts using the category di agnostic attribute than high NFC subjects when asked to form consideration sets ( F (1, 35) = 4.42, p = .043). There was no effect of NFC for subjects that were asked to make choices ( F < 1). Again, there was an effect of task instruction on subjects new product groupings. More new product groupings were based on the new products similarity along the category diagnostic attribute when subjects formed consideration sets ( MConsider = .62) than when they were initially asked to make a choice ( MChoose = .36; F (1, 34) = 29.21, p < .001). In line with the priming hypothesis, subjects NFC also affected new pr oduct groupings. Low NFC subjects were more likely to group the new products using the category diagnostic attribute ( MNFCLow = .57) than high NFC subjects ( MNFCHigh = .44; F (1, 34) = 9.02, p = .005). Initial product viewing times Before analysis, all viewing tim es were subj ected to a natural log transformation. These times were then used in a 3 (product categor y: tee-shirts, sunglasses, frisbees) by 2 (shopping

PAGE 55

55 task: form a consideration set vs. choose the best product) by 2 (need for cognition: high vs. low) repeated-measures MANOVA, with shoppi ng task and need for cognition as betweensubjects variables, and product category as a w ithin-subjects factor. There was no interaction between product category a nd task instructions ( F < 1), no interaction between product category and NFC ( F < 1), or between task instru ction, NFC, and product category ( F < 1) on subjects viewing times, so the product category factor was collapsed for the rest of the analyses. Providing additional evidence for the priming hypothesis, these analyses revealed that task instruction, NFC, and the interaction between the two had no effect on subjects viewing times during initial product exposure ( F < 1). Experiment 3B This experiment also attempts to determin e if task goals prime sensitivity to product attributes or if task goals provide motivation fo r additional information search. As opposed to the previous studies, where subjects viewing times of the initial stimuli was self-determined, this experiment artificially limits in itial product exposure (10 seconds). If dual attribute use by those forming choices is motivati onal, then those forming choices should require more time to process this information. Because the viewing time in this experiment is significantly less than the av erage viewing times reported in experiment 2b (14.88s), subjects making choices during initial product exposure may not have time to move from category diagnostic to unique attributes. Under these conditions, the motivational account would predict no differences in new product groupings between those asked to form consideration sets and those asked to make choices. However, if post-exposure changes in attribute sensitivity are due to priming, then we would expect the task instructions to affect subjects new product groupings.

PAGE 56

56 This design also allows re-evaluation of th e effects of consumers NFC. If NFC simply provides a motivational bias to process more in formation, then, due to the time constraint, no differences in new product groupings would be expected between low and high NFC subjects. However, if NFC primes efficiency (low NFC) or accuracy goals (high NFC), then low NFC subjects should make more new product groupin gs using the category di agnostic attribute then high NFC subjects. Method Participants and Design. Fourty-four undergraduate students from a large southeastern university participated in this experiment in ex change for extra credit. The design was a three (product category: tee-shirts, sunglasses, fris bees) by two (shopping task: form a consideration set vs. choose the best produc t) by two (need for cognition: high vs. low) by two (dependent measure setup) mixed design. Product categor y and dependent measure types were withinsubject factors, and task inst ruction and need for cognition were between-subjects factors. Stimuli. This experiment utilized the same stimuli as in Experiment 2b. Procedure. Upon entering the laboratory subjects were randomly assigned to one of the two shopping task conditions. The instructions and exposure phase of the experiment were identical to the previous experi ments with one exception. In this experiment, subjects viewing of the initial products in the exposure phase was limited to ten seconds. Once this time had expired, the products in the ini tial exposure set disapp eared and the subjects entered the next phase of the experiment, which was identical to experiment 2b. After a ll the product categories had been exhausted, the subjects completed the 18-item need for cognition scale. Results Before any analyses were conducted, subjec ts were placed into a need for cognition category (high vs. low) based upon their res ponses to the NFC scale. The m ean was ( MNFC =

PAGE 57

57 86.95). Subjects who scored lower than the mean were marked as low NFC and subjects scoring higher than the mean were marked as high NFC. A 3 (product category: tee-shirts, sungla sses, frisbees) by 2 (shopping task: form a consideration set vs. choose the best product) by 2 (need fo r cognition: high vs. low) by 2 (dependent measure type) repeated-measures MANOVA was performed, with shopping task and need for cognition as between-subjects variable s, and product category and dependent measure type as within-subjects factors. The percenta ge of groupings made usi ng the category diagnostic attribute was the critical measure. Multivariate tests revealed that neither product category ( F < 1) nor dependent measure type ( F < 1) significantly interact ed with shopping task on the type of groupings made. Likewise, no significant inte raction existed between produc t category type and NFC (F (2, 80) = 1.48, p = .233) or between dependent measure type and NFC ( F (1, 40) = 1.45, p = .236) on new product groupings. Additionally, neither the in teraction between shopping tasks, need for cognition, and product category ( F (2, 80) = 1.41, p = .251) nor the interaction between shopping tasks, need for cognition, and dependent measure type ( F (1, 40) = 3.56, p = .066) nor the interaction between shopping tasks, need for cognition, product category, and dependent measure type were significant ( F (2, 80) = 1.39, p = .255), so the results were collapsed across both within-subject measures for th e remainder of the analyses. The critical results are shown in Figure 52. There was no interaction between subjects initial task instruction and need for cognition on the t ypes of new product groupings (F < 1). For low NFC subjects, task instruc tions had a marginal effect on subjects new product groupings; new products were grouped based on their similarity along the category di agnostic attribute more by the subjects that formed consideration sets than by subjects who initially chose

PAGE 58

58 ( MNFCLow/Consider = .63, MNFCLow/Choose = .5; F (1, 41) = 3.14, p = .084). Subjects tasks also affected the product groupings within the high need for cognition group. Again, new product groupings demonstrated greater reliance on the ca tegory diagnostic attribut e when consideration sets were initially formed than when choices were made during product exposure ( MNFCHigh/Consider = .54, MNFCHigh/Choose = .37; F (1, 41) = 4.83, p < .05). But, even while directionally appropriate with our hypothesis, low NFC subjects were not significantly more likely to group the new products using the category diagnostic attribute than high NFC subjects when asked to form consideration sets (F (1, 41) = 2.01, p = .164). Similarly, there was no effect of NFC for subjects that were asked to make choices ( F (1, 41) = 2.68, p = .109). This experiment again demonstrated that subjects tasks during initial exposure shape their later product decisions. Like before, re sults indicated a signifi cant effect of task instructions on subjects new product groupings. More new product groupings were based on the new products similarity along the category di agnostic attribute when subjects formed consideration sets ( MConsider = .59) than when they were initially asked to make a choice ( MChoose = .43; F (1, 40) = 7.46, p = .009). In line with the primi ng hypothesis, subjects NFC also affected new product groupings. Low NFC subjects were more likely to group the new products using the category diagnostic attribute ( MNFCLow = .57) than high NFC subjects ( MNFCHigh = .44; F (1, 40) = 4.24, p = .046). Discussion Experiment 3 generates a number of insight s. The significant effect of task on new product groupings again demonstrates that consumers tasks sensit ize consumers towards particular distributions of attribute values, de pending on how those distributions afford their initial task. Subjects who initially formed consider ation sets were more likely to show sensitivity

PAGE 59

59 to attribute level tradeoffs along the category diag nostic attribute whereas subjects who initially chose were more sensitive to attribute level differences along the unique attribute. The results also show that the link between ta sk and attribute sensitivity and use is more likely to be the result of priming than subj ects motivation to simply search for more information. Specifically, because both category diagnostic and unique attributes contribute to identification processes, the instruction to choos e primes sensitivity for both unique and category diagnostic attributes simultaneously. Because the unique attribute does not contribute to efficient representations, the instruction to form a consideration set only primes the category diagnostic attribute. Instead, the motivationa l account states that consumer s tasks prompt them to first screen the available products into a consider ation set using the categ ory diagnostic attribute before moving on to making a selection from that se t using the unique attribut e. If this were the case, then subjects should have vi ewed products longer when they were choosing than when they were forming consideration sets. However, no differences in viewing times were found as a function of the subjects initial task instructions Further, if attention to attributes occurs sequentially then significantly truncating the viewing times (3B) should have limited choosers ability to view the unique attrib ute after the category diagnostic attribute, leading to no task differences in their new product groupings. But, resu lts of this experiment revealed an effect of task on new product groupings, which is what the priming account would have predicted (simultaneous sensitization). Investigation of the effects of subjects need for cognition (NFC) also provide support for a priming account. The priming account of NF C posits that NFC dictates individuals type of information search. Low NFC consumers search for efficient attributes, while High NFC consumers search for accurate attributes. The mo tivational explanation argues that NFC dictates

PAGE 60

60 individuals amount of information search. Under th is explanation, high NFC consumers examine both category diagnostic and unique attr ibutes allowing them to have the most information possible. Three concrete predicti ons flow from the motivational hypothesis. First, task differences would not be expected to affect hi gh NFC subjects, because they would always incorporate all available product information. But, both experiments show that task had a significant effect on high NFC subjects. Second, if high NFC leads to a greater search for information, then they would be expected to take more time viewing the products during initial exposure than low NFC subjects. The resu lts of experiment 3A show no differences in viewing time as a result of subjects NFC. Third, the motivational account would predict that when initial product viewing times were truncat ed, that there should be no difference between the product groupings of high and low NFC subjects. In contrast, the results of experiment 3B show clear differences in product groupings as a function of subjects NFC. While these results clearly make the case against a motivational account, other evidence provides support for a priming explanation. If NF C acts as a baseline prime that pre-sensitizes individuals towards either efficient or accurate attributes, then low NFC individuals should be more likely to use the category diagnostic attr ibute and high NFC should be more likely to use the unique attribute to form their new product groupings. Results from both experiments confirm this prediction. Additiona lly, if NFC primed sensitivity to wards efficient or accurate information, then we would expect NFC level to magnify the effect of task instructions on subjects groupings. While results were only signif icant for the low NFC group in experiment 3a, the means in both experiments are directionally consistent with this premise.

PAGE 61

61 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 ConsiderChooseConsiderChoose LowNFC HighNFC % New Groupings Category Diagnostic Unique Figure 5-1. Results of experiment 3A.

PAGE 62

62 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 ConsiderChooseConsiderChoose LowNFC HighNFC % New Groupings Category Diagnostic Unique Figure 5-2. Results of experiment 3B.

PAGE 63

63 CHAPTER 6 GENERAL DISCUSSION This work c ontributes to nascent literature exploring how attribute distributions affect consumers information processing and attribute use. While Lurie (2004) demonstrated a link between attribute distributions and amount of information pro cessing, he did not explain how different attribute value distributi ons affected consumers actual de cisions. Existing research has investigated the link between attribute value di stributions, attribute acce ssibility, and attribute use (Goldstone 1994; Goldstone, Medin, and Halberstadt 1997; Johnson 1989; Medin, Goldstone, and Gentner 1993; Schyns and Murphy 1994; Schyns and Rodet 1997) but the literature is divided; one can find evidence that both category diagnostic and unique attributes will be used in future decisions. This pape r coheres this evidence by proposing a mechanism that accounts for both sets of results. Using the framework of goal-directed perception, this paper shows that an individuals future use of at tributes is dependent on the consumers task and how the attributes distributions of values afford task completion. When consumers form considerations sets they use cate gory diagnostic attributes in late r decisions, but when consumers initially choose, they utilize unique attr ibutes in their subsequent judgments. The results of one pilot study and five expe riments support the claim that both attribute distributions and the tasks perf ormed at the time of product ex posure interact to determine perceptions of attribute importance and attribut e use (as shown by their sensitivity to attribute level tradeoffs). The pilot study demonstrated that consumers tasks during initial product exposure interact with the distribu tion of attribute values along an attribute such that attributes possessing distributions that afford task completi on are later perceived to be more important. In experiment 1A, results showed that the task goa ls active during exposure sensitized individuals towards particular product attri butes, depending on the distribution of values along that attribute.

PAGE 64

64 This led them to be more reac tant to attribute level tradeoffs along the primed attribute when judging the novelty of a previously unseen product. Experime nt 1B extended this finding by showing that consumer tasks and attribute value distributions also affect consumers willingness to pay for new items. Experiment 2 showed that the similar sensitivity to tradeo ffs along the category diagnostic attribute between those forming consider ation sets and those choosing in experiment 1 was due to choosers incorporat ing both category diagnostic a nd unique attributes in their judgments. Experiment 3 ruled out a motivation-based explanation for our results, and also found that individuals dispositional biases (need for cognition) may also serve as a prime to sensitize consumers towards attributes with particular attribute distributions. Whereas low NFC consumers focus on category diagno stic attributes, high NFC consum ers attend to attributes with unique distributions. Together, these studies contri bute to the notion th at perception is goal-directed. When a goal is active, either due to a particular task, or dispositional bias, th e perceptual system is primed to notice elements in the environmen t that are goal related. Furthermore, this sensitization persists into future judgments a nd choices. Indeed, the theory of goal-directed perception aptly explains the prior goal-att ribute links detailed by Yi (1990), Mandel and Johnson (2002), and Chernev (2004). This dissertati on provides a unique contribution to this area by demonstrating that goal based priming extends beyond the qualitative nature of the prime and the product attribute. Specifically we show that attributes with par ticular distributions of values may differentially afford consumers tasks such that when task -attribute distribution compatibility is present, future use of these attributes occurs. Further, consumers NFC may

PAGE 65

65 naturally sensitize them towards particular attrib ute types as well, and in some cases, may even magnify the effect of task related primes. Applications and Extensions These findings can be extended in several different ways. Interestingly, the theory presented here may contribute to work examining consumers perceptions of the variety of an assortment (Hoch, Bradlow, and Wansink 1999), which consumers rank as one of the most important factors they consider when choosing a store. One of the most critical elements contributing to these perceptions is the informati on structure of the enviro nment, which relates to the actual differences in attribute values between different products in th e assortment. As shown in the present study, the tasks that individuals engage in during exposure interact with actual information structure, and could therefore lead to different perceptions of variety for identical sets of products, depending on an individuals task. When cons umers form consideration sets during exposure, they focus on cat egory diagnostic attributes, but when they are choosing, they place greater emphasis on unique attributes. Because, by definition, category diagnostic attributes have less variety al ong them than unique attributes it would be expected that individuals engaged in consideration set form ation would have lower perceptions of product variety than individuals that initially choose from the product assortment. Determining the impact of consumers goals and attribute di stributions on these perceptions would be an interesting avenue of future research. There may also be links between the current work and the temporal construal literature (Trope and Lieberman 2003). In general, construal level theory holds that the greater the temporal distance between an individual and an event, the more general and abstract representations of that event become. In cont rast, when an event is temporally close, its representations become more conc rete and specific. It may be the case that these changes of

PAGE 66

66 representation also affect which attributes individuals sensitize towards depending on when that consumer planned to purchase the item. When distance to purchase was lengthy, we would expect these consumers to be more likely to atte nd to category diagnostic attributes, because of their more general nature. However, when deci sion was imminent, consumers would be expected to sensitize towards unique attributes, because they are more concrete and specific in nature. Similarly, it may be possible to lead consumer s to have either simplification or accuracy goals using environmental primes. While this could be accomplished through signage with goal related words (Choose the best. vs. Make li fe easy.), there are also more subtle (and arguably more interesting) ways that this co uld be accomplished as well. Recently, work by Meyers-Levy and Zhu (2007) found that changes in ceiling height influenced consumers processing styles. When ceilings were high, cons umers engaged in more relational processing, which led them to be more focused on general as pects of the products. However, when ceilings were low, consumers engaged in a more analy tic processing style, focusing on more specific product attributes. These changes in processing style have the potential to interact with attribute value distributions on attribute use. When ceilin gs are high, and consumers engage in relational processing, they should be more likely to utilize category diagnostic attrib utes. However, when ceilings are low, and consumers engage in analyt ic processing, they would be expected to be more likely to rely on unique attributes in their future decisions. Indeed, there may be a number of ways in which accuracy or efficiency can be primed by elements present in the retail environment. Future work in this area is warranted. These findings also have important managerial implications. First, this work provides a specific methodology that managers can use to ma nipulate consumers use of product attributes. For businesses on the web, it is likely that a ma nager can deduce a consumers propensity to

PAGE 67

67 make a purchase by determining when the consumer last visited her site, the other websites most recently visited by the consumer, and the type of behaviors (click patterns) exhibited. By determining a consumers likeli hood of purchase, the system can determine if a consumer is more likely to be forming a consideration set or to be making a choice. Using this information, product displays could be presented in a way that makes the retailers chosen attribute either category diagnostic or unique, t hus affecting the consumers pe rceptions of th at attributes importance, and increasing its chance of being used in the consumers future decisions. A second implication of this work relate s to product positioning. By surveying the current product market, a manager can determine which attributes of the product category have a category diagnostic distribution a nd which have a unique distributi on. It is also assumed that managers have some information about their product segments. Using this information, a manager can then determine which attributes to either mimic or change in order to achieve a desired level of similarity to or differentiation from the current products on the market. For example, assume that a manager wants to engage in a copycat strategy and position their product similarly to a product currently on the market. If the segment being targeted is likely to be high in need for cognition, then a manager should make sure that their produ ct has an identical attribute value to the existing produ ct along the unique attribute. Ho wever, if they are targeting a segment that is thought to have low need for co gnition, then it would be beneficial to mimic attribute values along the categ ory diagnostic attribute. Similarly, if a manager is introducing his product during the introduc tory stage of the product life cycle, then the majority of cons umers (i.e., non early-adopters) may be forming consideration sets of products th at they would examine when they were ready to purchase. In this case, managers would want to make sure that their product possessed a si milar (different) value

PAGE 68

68 on the category diagnostic attribut e if they wanted to pursue a c opycat (differentiation) strategy. However, if the product was bei ng introduced into a more matu re product market, where the majority of consumers were ready to make purch ases, then the manager should ensure that their brand is similar (different) on the unique at tribute, depending on th eir positioning strategy.

PAGE 69

69 LIST OF REFERENCES Alexander, David L., John G. Lynch Jr., and Qi ng W ang (2008), As time goes by: Do cold feet follow warm intentions for really-n ew vs. incrementally-new products? Journal of Marketing Research 45 (June), 307-319. Atick, Joseph J. (1992), Could information theo ry provide an ecological theory of sensory processing? Network: Computation in Neural Systems 3, 213-251. Bettman, James R., Mary Francis Luce, and J ohn W. Payne (1998), C onstructive consumer choice processes, Journal of Consumer Research 25 (December), 187-217. Bettman, James R. and C. Whan Park (1980), Effects of prior knowledge and expertise and phase of choice process on consumers deci sion processes: A pr otocol analysis, Journal of Consumer Research 7 (December), 234-238. Cacioppo, John T. and Richard E. Petty (1982), The need for cognition, Journal of Personality and Social Psychology 42, 116-131. Cacioppo, John T., Richard E. Petty, and Chuan Fe ng Kao (1984), The efficient assessment of need for cognition, Journal of Personality Assessment, 48, 306-307. Chakravarti, Amitav and Chris Ja niszewski (2003), The influence of macro-level motives on consideration set composition in novel purchase situations, Journal of Consumer Research 30 (September), 244-258. Chakravarti, Amitav, Chris Janiszewski, and Gulden Ulkumen (2006), The neglect of prescreening information, Journal of Marketing Research 43 (November), 642-653. Chater, Nick (1999), The search for simp licity: A fundamental cognitive principle? Quarterly Journal of Experimental Psychology 52 (2), 273-302. Cunha, Jr., Marcus and Juliano Laran (2009), "Asymme tries in the sequential learning of brand associations: Implications for the early entrant advantage," Journal of Consumer Research 35 (February), 788-789. De Wilde, Els, Alan D. J. Cooke, and Chris Janiszewski (2008), Atte ntional contrast during sequential judgments: An examinati on of the number-of-levels effect, Journal of Marketing Research 45 (August), 437-449. Dickson, Peter R. and James L. Ginter (1987) Market segmentation, product differentiation, and marketing strategy, Journal of Marketing 51 (April), 1-10. Feldman, Jack M. and John G. Lynch Jr. (1988), Self-generated validity and other effects of measurement on belief, attitude intention, and behavior, Journal of Applied Psychology, 73, 421-435.

PAGE 70

70 Fink, Michael and Shimon Ullman. Characteriz ing the features encoded during multiple category acquisition, Working Paper, Ce nter for Neural Co mputation, Hebrew University of Jerusalem, Jerusalem 91904, Israel. Fodor, Jerry A. (1981). Representations Cambridge, MA: The MIT Press. Gibson, Eleanor J. (1991), An odyssey in learning and perception Cambridge, MA: The MIT Press. Gibson, Eleanor J. and Richard D. Walk (1965), The effect of prolonged exposure to visually presented patterns on learning to discriminate them, Journal of Comparative and Physiological Psychology 49, 239-242. Goldstone, Robert L. (1994), I nfluences of categor ization on perceptual discrimination, Journal of Experiment al Psychology: General 123 (2), 178-200. Goldstone, Robert L., Douglas L. Medin, and Jami n Halberstadt (1997), Similarity in context, Memory and Cognition, 25 (2), 237-255. Hoch, Stephen J., Eric T. Bradlow, and Brian Wa nsink (1999). The variety of an assortment, Marketing Science, 18 (4), 527-546. Hommel, Bernhard, Jochen Musseler, Gisa As chersleben, and Wolfgang Prinz (2001), The Theory of Event Coding: A Framewor k for Perception and Action Planning, Brain and Behavioral Sciences 24, 849-878. Janiszewski, Chris (2007). G oal-directed perception, In The Handbook of Consumer Psychology, Curtis Haugvedt, Paul Herr, and Frank Kardes, (Eds.) (forthcoming) Johnson, Michael D. (1989), The differe ntial processing of product category and noncomparable choice alternatives, Journal of Consumer Research 16 (December), 300-309. Kotler, Philip and Kevin Lane Keller (2006). Marketing Management Upper Saddle River, NJ: Prentice Hall. Lehmann, Donald R. and Russel S. Winer (2005), Analysis for Marketing Planning 6th Edition New York, NY: McGraw Hill Corporation. Levin, Irwin P., Mary E. Huneke, and J.D. Jasper (2000), Information processing at successive stages of decision making: Need for c ognition and inclusion-exclusion effects, Organizational Behavior and Human Decision Processes 82 (July), 171-193. Lurie, Nicholas H. (2004), "Decision making in information-rich environments: The role of information structure," Journal of Consumer Research 30 (March), 473-486.

PAGE 71

71 Lynch, John G. Jr., Howard Marmorstein, and Mi chael F. Weigold (1988), Choices from sets including remembered brands: Use of recalled attributes and prior overall evaluations, Journal of Consumer Research 15 (September), 169-184. Mandel, Naomi and Eric J. Johnson (2002), When web pages influence choice: The effect of visual primes on experts and novices, Journal of Consumer Research 29 (September), 235-245. Markman, Arthur B. and Dedre Gentner (1993) Structural alignmen t during similarity comparisons, Cognitive Psychology 25, 431. Medin, Douglas L., Robert L. Goldstone, and De dre Gentner (1993), Respects for similarity, Psychological Review 100 (2), 254-278. Mervis, Carolyn and Eleanor H. Rosch (1981) Categorization of natural objects, Annual Review of Psychology 32, 89-115. Mervis, Carolyn, J. Catlin, and Eleanor Rosch (1976), Relationships among goodness-of example, category norms, and word frequency, Bulletin of the Psychonomic Society 7, 283-284. Meyers-Levy, Joan and Rui (Juliet) Zhu (2007), The influence of ceiling height: The effect of priming on the type of processing people use, Journal of Consumer Research 34 (August), 174-186. Most, Steven B., Brian J. Scholl, Erin Cliffor d, and Daniel J. Simons (2005), What you see is what you get: Sustained inattentional blindness and the capture of awareness, Psychological Review 112 (1), 217-242. Murphy, Gregory (2002), The big book of concepts Cambridge, MA: The MIT Press. Nedungadi, Prakash (1990), Recall and consumer c onsideration sets: Influencing choice without altering brand evaluations, Journal of Consumer Research 17 (December), 263-276. Pechmann, Cornelia and S. (Ratti ) Ratneshwar (1991), The use of comparative advertising for brand positioning: Association versus differentiation, Journal of Consumer Research 18 (September), 145-160. Pirolli, Peter L. and Stuart K. Card (1999), Information foraging, Psychological Review 106 (4), 643-675. Rosch, Eleanor H., Carolyn B. Mervis, Wayne D. Gray, David M. Johnson, and Penny BoyesBraem (1976), Basic objects in natural categories, Cognitive Psychology 8, 382-439. Schyns, Philippe (1998), Diagnostic Recognition: Task constraints, object information, and their interactions, Cognition, 67 (2), 147-79.

PAGE 72

72 Schyns, Philippe and Gregory Murphy (1994), The ontogeny of part representation in object concepts, In Medin, D. L. (Ed.), The psychology of learning and motivation (Vol 31, pp. 301-349), San Diego, CA: Academic Press. Schyns, Philippe and Luc Rodet (1997), Cat egorization creates functional features, Journal of Experimental Psychology: Learning, Memory, & Cognition 23 (3), 681-696. Trope, Yaacov and Nira Lieberma n (2003), Temporal construal, Psychological Review 110 (3), 403-421. Tversky, Amos (1977), Features of Similarity, Psychological Review 84 (4), 327-352. Tybout, Alice M. and Brian Sternthal (2001), Brand positioning, In Iacobucci, D. (Ed.) Kellogg on Marketing, p. 31-58, New York, NY: John Wiley and Sons, Inc. Wood, Stacy L., and Joffre Swait (2002), Psyc hological indicators of innovation adoption: Cross-classification based on need fo r cognition and need for change, Journal of Consumer Psychology, 12 (1), 1-13. Yi, Youjae (1990), The effects of contextu al priming in print advertisements, Journal of Consumer Research, 17 (September), 215-222.

PAGE 73

73 BIOGRAPHICAL SKETCH Jesse Itzkow itz obtained his undergraduate degr ee in psychology from the University of Florida in 2000. In 2005, he received his doctora l degree in cognitive psychology, also from UF. He will receive his second doctoral degree in marketing from the Warrington College of Business Administration in August 2009. In fall 2009, Jesse will begin as an assistant professor at the Sy Syms School of Business at Yeshiva University in New York City.