autoria: valter afonso vieira, claudio vaz torres, rogerio ... · validity of nft scale, since...

16
1 Haptic Information Processing: Assessing the Need for Touch Scale Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio Gava Abstract: Some people could not buy products without touching them at first, believing that doing so might generate more assurance, information and reduce uncertainty. The literature suggests an instrument to measure consumers’ necessitate of touching – Need for Touch (NFT). This paper had as main goal to analyze if its structure is consistent. The method was a survey with 171 people. Based on literature review, six research propositions were suggested in order to asses the nomological, convergent and discriminant validity of the construct. The outcomes suggested that four assumption were supported in the predict direction. The results also found the dual characterization of NFT as multi-dimensional construct with two underlying factors, Instrumental and Autotelic Touch. In addition, the outcomes also found that the NFT Autotelic dimension showed its role as moderator variable, given that when the consumer has a higher NFT autotelic tendency; the experiential motivation for shopping plays a more important role on impulsive motivation. Introduction Consumer behavior decision making process suggests that individuals decide buy products across different methods, such as brand name, relevant attributes, quality, convenience, etc. One of these ways is touching in the products. Touching is a way of getting more information thought the process of touching. In this context, the sense of touch can help consumer decision making process by providing sensory forms of pertinent information. For instance, individuals’ confidence in product may differ depending on whether a shopper has the opportunity to touch and examine a product. In addition, attitude toward a product may be more positive, depending on whether a shopper has the opportunity to touch and to experience pleasurable sensory feedback (e.g. rub a soft leather coat) before purchase (Peck and Childers 2003). It is because the touch might create on the consumer a better certainty and familiarity with the product. In fact, some people could not buy products without touching them at first, believing that doing so might generate more assurance and reduce uncertainty. Peck and Childers (2003) comment that some “consumers touch products to simply place them in shopping carts, other consumers spend more time exploring products with their hands before ultimately making a purchase decision. [Thus, it] seems likely that some individuals would prefer information available through the sense of touch” (p.430) rather than information obtained through reading or listening. Literature suggested that information available through the sense of touch is called “haptic” attributes. For Klatzky and Lederman (1992, 1993) and Lederman and Klatzky (1987) touching with the hands, or activating the haptic system, has been reported to be particularly adept at encoding the material object properties corresponding to texture, hardness, temperature and weight information. Therefore, reading the information printable on the package is not the only way of getting information from the product, but also touching can be the other choice. Based on the necessity of a better decision making process, touching in the products could help to explain why individuals created more confidence, attitude, assurance, and more ability to make their judgment. For that reason, marketing strategies could be generating in order to facilitate consumer’s touch and product acquisition inside stores. In order to do that, marketing professionals and consumer analysts need at first to measure the phenomenon in

Upload: others

Post on 18-May-2020

5 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

1

Haptic Information Processing: Assessing the Need for Touch Scale

Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio Gava

Abstract: Some people could not buy products without touching them at first, believing that doing so might generate more assurance, information and reduce uncertainty. The literature suggests an instrument to measure consumers’ necessitate of touching – Need for Touch (NFT). This paper had as main goal to analyze if its structure is consistent. The method was a survey with 171 people. Based on literature review, six research propositions were suggested in order to asses the nomological, convergent and discriminant validity of the construct. The outcomes suggested that four assumption were supported in the predict direction. The results also found the dual characterization of NFT as multi-dimensional construct with two underlying factors, Instrumental and Autotelic Touch. In addition, the outcomes also found that the NFT Autotelic dimension showed its role as moderator variable, given that when the consumer has a higher NFT autotelic tendency; the experiential motivation for shopping plays a more important role on impulsive motivation. Introduction Consumer behavior decision making process suggests that individuals decide buy products across different methods, such as brand name, relevant attributes, quality, convenience, etc. One of these ways is touching in the products. Touching is a way of getting more information thought the process of touching. In this context, the sense of touch can help consumer decision making process by providing sensory forms of pertinent information. For instance, individuals’ confidence in product may differ depending on whether a shopper has the opportunity to touch and examine a product. In addition, attitude toward a product may be more positive, depending on whether a shopper has the opportunity to touch and to experience pleasurable sensory feedback (e.g. rub a soft leather coat) before purchase (Peck and Childers 2003). It is because the touch might create on the consumer a better certainty and familiarity with the product. In fact, some people could not buy products without touching them at first, believing that doing so might generate more assurance and reduce uncertainty. Peck and Childers (2003) comment that some “consumers touch products to simply place them in shopping carts, other consumers spend more time exploring products with their hands before ultimately making a purchase decision. [Thus, it] seems likely that some individuals would prefer information available through the sense of touch” (p.430) rather than information obtained through reading or listening. Literature suggested that information available through the sense of touch is called “haptic” attributes. For Klatzky and Lederman (1992, 1993) and Lederman and Klatzky (1987) touching with the hands, or activating the haptic system, has been reported to be particularly adept at encoding the material object properties corresponding to texture, hardness, temperature and weight information. Therefore, reading the information printable on the package is not the only way of getting information from the product, but also touching can be the other choice. Based on the necessity of a better decision making process, touching in the products could help to explain why individuals created more confidence, attitude, assurance, and more ability to make their judgment. For that reason, marketing strategies could be generating in order to facilitate consumer’s touch and product acquisition inside stores. In order to do that, marketing professionals and consumer analysts need at first to measure the phenomenon in

Page 2: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

2

question. Measuring the construct can help to find answers such as, does touching in products help consumer decision making process? Does touching in products influences behavioral intention? And is the touching process a way of reduces uncertainty? Thus, looking for fulfilling a literature gap of lacking a validity instrument, Peck and Childers (2003) propose a 12-item instrument to measure consumers’ Need for Touch (NFT). Scholars frequently need new scales for measuring new or old phenomena in marketing. Thus, the necessity of the construction and the test of new instruments for measuring these phenomena is very critical. In that sense, Jacoby (1978, p.91) comments that “most of our measures are only measures because someone says that they are, not because they have been shown to satisfy standard measurement criteria (validity, reliability and sensitivity)”. Therefore, this paper has a main goal to analyze if its structure is consistent, validity, reliability and multidimensional in its configuration. Hence, the paper is structured as follow. In the next part, literature review defines Need for Touch construct and its dimensions. Subsequently, from the literature review, five propositions are suggested in order to assess discriminant, nomological and convergent validity. After that, the method used in the field research is presented. Data analysis shows the results of the scale and general conclusions end the paper. Literature Review In this part of the literature, the construct and its supposed two dimensions are defined. Follow it, six research propositions are suggested in order to asses the nomological validity of the phenomenon. Initially, Peck and Childers (2003) defined NFT “as a preference for the extraction and utilization of information obtained through the haptic system” (p.431). It means that using the haptic system consumers can get product information and use it for make judgments. The NFT construct is based on motivational vs. ability differences among individuals (Johansson 1978). It suggests two factors that underline the construct. The need to examine products haptically can be driven by intrinsic motivations that are either consumer problem solvers or consumers seeking fun, fantasy, arousal, sensory stimulation and enjoyment. The dichotomy “motivational vs. capability” has been represented in the retail context by the themes of shopping as work vs. the festive perspective of shopping, as fun (Peck and Childers, 2003). In fact, the dual characterization of NFT from both the retail as well as the psychological literature on motivations is consistent with the perspective of NFT as multi-dimensional construct with two primary factors. They are defined as instrumental and autotelic touch, or the ability/work goals vs. motivational/fun. The NFT Instrumental Factor. It represents those aspects of prepurchase touch that reflect outcome-directed touch with a salient purchase goal (Peck and Childers, 2003). In this factor, people touch in the products looking for fulfilling a gap, which can be either getting more information from the products or analyzing its physical structure. Contained within the domain of this form of touch are goal-driven evaluative outcomes related to the consumer (e.g. eliminate doubt) as well as the target product (e.g. quality, durability, etc.). For instance, picking up a notebook computer and holding it in one’s hand to assess its weight, product’s texture, hardness, properties, and deriving an inference with respect to its portability would be an example of an instrumentally driven haptic product evaluation (Peck and Childers, 2003).

Page 3: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

3

The NFT Autotelic Factor. Autotelic touch dimension corresponds to the sensory aspects of product touch, with no purchase goal necessarily salient, but with spontaneous investigation of multisensory psychophysical product relationships (Holbrook and Hirschman, 1982). Peck and Childers (2003) suggested that central to defining the domain of autotelic touch are its hedonics and the compulsive or irresistible need to engage in exploratory variety seeking via. For example, hedonics may mean for consumers enjoyment, affect, fun, arousal and sensory stimulation, and compulsive may mean lack of control and indiscriminate processing. Some evidence for appreciating the experiential aspect of consumer behavior is found in museums that offer multisensory environments, including music and hands-on displays of sculpture, which individuals can touch and interact with things (Fiore, Moreno and Kimble, 1996). Summarizing, the autotelic dimension of NFT relates to touch as an end in and of itself (Peck and Childers, 2003). Research Propositions In this part, six propositions are created to show the convergent, discriminant and nomological validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid measurement is the sine qua non of science. In a general sense, validity refers to the degree to which instruments truly measure the constructs which they are intended to measure” (p. 6, original italic). In that logic, for achieving such proposes, three procedures were followed. They are (1) convergent validity that is the extent to which the scale correlates highly with other methods designed to measure the same construct (Churchill, 1979); (2) discriminant validity, which for Churchill (1996) is when an instrument “behaves as expected with respect to some other construct to which it is theoretically related” (p.538) and (3) nomological validity – it is to link the conceptual/theoretical real with the observable one, because this is the central concern of construct validity (Trochim, 2002)”. Convergent Validity. Citrin, Stem, Spangenberg and Clark (2003) developed a six-item scale to measure Need for Tactile Input (NTI) and defined this construct as “the desire or need for tactile input to make brand/product evaluations” (p.918). Peck and Childers (2003) believe that NTI domain overlaps with the instrumental dimension of the NFT scale, because most of the items in the NTI are very similar with NFT’s. For example, “I need to touch a product in order to create a general evaluation of it” vs. “I feel more confident making a purchase after touching a product”. Basing on that similarity, since the instrumental dimension of the NFT scale refers to those aspects of prepurchase touch that reflect outcome-directed touch with a salient purchase goal, and since NTI represents a tactile way to make brand/product judgments, it is assumed that these two constructs are positively associated. Thus, the first proposition is P1: NTI scale will have a higher positive relationship with the instrumental dimension of NFT vs. the autotelic dimension of NFT. Discriminant Validity. Need for Cognition is the necessity of extracting the best of reasoning in the decision making process (Cacioppo and Petty, 1982). In psychology, it is a personality variable reflecting the extent to which people engage in and enjoy effortful cognitive activities. Peck and Childers (2003) commented that “although NFC, like NFT, overlaps in the domain of information-acquisition behavior, NFC taps this domain at a more macro level that is also not specific to the consumption context” (p.434). It is assumed because the NFC scale does not have always the need of consumption contained in its womb. For instance, the need for cognition might be just by the eventual reasoning of reading a newspaper or interpreting a movie. This does not requite a specific purchase moment, dispensing the

Page 4: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

4

necessity of NFT. In contrast, the NFT scale is more molecular in terms of its sense-specific focus and narrower in tapping consumption behavior. Thus, it is predicted the scales will not be related, because they are unequal constructs. Therefore, the second proposition is P2: NFC scale will not have relationship with NFT. Evaluation is defined as the assessment of the positive and/or negative qualities of an object, which is assumed to be among the most pervasive and dominant human responses, and based on that Need to Evaluate construct is defined as the assessment of the positive and/or negative qualities of an object (Jarvis and Petty, 1996). Peck and Childers (2003) argue that Need to Evaluate (NTE) represents information acquisition, but in the context of chronic evaluation. This chronic form of evaluation across such objects as social issues, groups, and future behaviors is in contrast to the product-specific nature of NFT, which is more peculiar to goods. Evaluation represents only one aspect of the NFT, particularly when instrumental is contrasted with autotelic touch. NTE represents a more global necessity of assess things. Then, P3: NTE scale will not have relationship with NFT. Nomological Validity. As evidence of nomological validity, it was considered three direct-marketing media, such as shopping via catalog/mail, over the telephone, and on the Internet. Klatzky, Lederman and Matula (1993) evidenced a “visual preview model” that vision provides a quick glance that results in broad but coarse information about the haptic properties of an object, information that is useful in directing further processing. Consequently, viewing a catalog or a web page may reveal that more detailed information about a haptic property is available because it is write, yet not readily attainable. Supposedly, a consumer who values haptic information would be expected to be less likely to purchase products via marketing channels, where direct product touch is prohibited (Peck and Childers, 2003). For these reasons, (P4a) is expected a negative relationship between purchasing via these nontouch media and instrumental dimension of NFT (goal oriented). Additionally, when consumers purchase via these direct media, it is more likely that they are engaged in shopping behavior with a salient purchase goal rather than for fun (Peck and Childers, 2003). For example, a prominent goal could be safe time. Nevertheless, autotelic touch is concerned with touch without a leading purchase aim, which is contrary to the assumption that consumers purchase via these direct media with a salient goal. For this reason, it’s expected (P4b) there to be no relationship between purchasing via these media and autotelic NFT. Dawson, Bloch and Ridgway (1990) proposed the experiential motivation for shopping as a latent construct to represent social or recreational motivations rather than to purchase products as obligation. It suggested that the experience of buying would be driven more by the desire for fun/diversion rather than be the necessity of acquiring information to purchase a product. In this context, it is expected (P5a) that experiential shopping would relate positively with autotelic NFT, since that dimension has no purchase goal necessarily salient, and (P5b) not be significantly related to instrumental NFT, because that factor reflects outcome-directed touch with a salient purchase end. As a last evidence for nomological validity, Rook and Fisher (1995) salient that “highly impulsive buyers are more likely to experience spontaneous buying; their shopping lists are more ‘open’ and receptive to sudden, unexpected buying ideas” (p.306). In that sense, the impulsive purchase trait is characterized by the lack of salient purchase goal, at least at the start of the shopping experience, and because of that (P6a) it is supposed to be positively

Page 5: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

5

related to autotelic NFT, since it is more spontaneous and characterized by a nonsalient purchase objective (Peck and Childers, 2003). Contrarily, given that instrumental NFT is more reflective and concerns a salient purchase goal, (P6b) it’s expected that buying impulsiveness will be unrelated to instrumental NFT. Research Design First, back translation was used to create the Brazilian version of the NFT instrument. Two Portuguese fluent academics did this job (Malhotra, 2001). As a consequence, the questionnaire was pre-tested with 3 students. In depth interviews were conducted with each student to identify the problems, looking for ambiguities and misleading of the instrument. Modifications were implemented and a final version of the scale was shaped. A total of 171 observations was possible, using a faculty as application place. All people were undergraduate business. The sample was defined as non-probabilistic by convenience (Malhotra, 2001). Measurement. The questionnaire contained 12-item NFT instrument from Peck and Childers (2003) that can be viewed on Appendix I, 16-item NTE scale from Jarvis and Petty (1996), 18-item NFC instrument from Cacioppo and Petty (1982), 5-item NTI instrument from Citrin et al., (2003), 3-item to non-touch buying style from Peck and Childers (2003), such as “My tendency to buy by catalog, (…) by telephone and by internet is”, 6-item of Impulsive behavior from Rook and Fisher (1995) and 5-item Experiential instrument from Dawson, Bloch and Ridgway (1990). All scales were managed as likert 7-point style; varying “strongly disagree to strongly agree”. Constructs Conceptual Definitions. Basing on theoretical review, Need for Touch construct is defined as a preference for the extraction and utilization of information obtained through the haptic system (Peck and Childers, 2003). Need to Evaluate construct is defined as the assessment of the positive and/or negative qualities of an object (Jarvis and Petty, 1996). Need for Tactile Input construct is defined by examining the role of tactility in making product choices on the Internet (Citrin et al 2003). Need for Cognition is the necessity of extracting the best of reasoning in the decision making process (Cacioppo and Petty, 1982). Impulsive Buying Behavior is defined as consumer’s tendency to buy spontaneously, unreflectively, immediately and kinetically (Rook and Fisher, 1995). Experiential Shopping relates to social or recreational motivations of buying rather than to purchase products by obligation/duty (Dawson, Bloch and Ridgway, 1990). Data Analysis Initially, all variables were analyzed in terms of missing values (observation and variable), outliers univariate and multivariate, multicoliniarity, skewness and curtosis and normality. Albeit ten observations were excluded because of missing values, many of them were found within the boundaries suggested by literature (Kline, 1998; Hair et al., 1998). After that, exploratory factor analysis looked for analyzing the NFT structure. The criteria used for retaining the variables were factor loading and communalities both above 0.45; Tabacknick and Fidell (2001) suggested λ ≥ .32) and eigenvalues over one. The extraction method choosen was non-orthogonal. It is justified because (a) the factors are supposed to be correlated in social science, (b) oblique rotations will always meet the simple structure criterion better than orthogonal rotations and (c) some research supports a slight superiority of oblique rotations in terms of factor replicability (Reise, Waller and Comrey, 2000).

Page 6: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

6

In the beginning, the first exploratory factor analysis showed three dimensions of the NFC scale (Kaiser-Meyer-Olkin = 0,879; p < 0.000). This structure presented the first dimension as autotelic, the second dimension as instrumental and the third dimension also as instrumental (variance explained 46%, 12% and 8% respectively). The results can be viewed on Table 1. The three-dimension structure is so different from the original two-dimension model. Initially, the configuration suggested that the second factor showed the confident, trust and comfortable aspects of purchasing a product by physically examining it. It means that an “Instrumental-Trust” dimension was supposed to exist. On the other hand, the third factor showed that the indicators represent that only through touching them the consumer can make sure that a product is worth to be purchased. It means “Instrumental-Touch as Duty” dimension. In other words, it suggests that if someone cannot touch in a product, he/she would be reluctant to buy that. Hence, the initial finds induced to two dimensions of the Instrumental factor, instead of one. However, some censure need to be done in this initial result.

Table 1: First Exploratory Factor Analysis of NFT

It is important to comment that there are some limitations of this first exploratory factor analysis. Initially, some authors suggested that eigenvalues over one, as the criterion for determining the number of components to retain, is problematic since this method has eighty percent of be wrong, when compared to scree plot, parallel analysis, χ²-Bartlett and MAP (Zwick and Velicer, 1986). Basing on that limitation, Zwick and Velicer (1986), in their Monte Carlo simulation, discovered that the best procedure to found the number of components is Parallel Analysis, since it is correct in eighty percent of the cases vs. twenty from eigenvalue over 1. According to Zoski and Jurs (1996) “parallel analysis compares the eigenvalues of the correlation matrix to those of a matrix of randomly generated variables, where eigenvalues from the research data should be greater than those from the random data” (p.444).

,786 ,259 -,549 M=3.03

,761 ,312 -,485 M=3.49

,750 ,285 -,607 M=3.13

,746 ,177 -,509 M=3.01

,527 ,432 -,374 M=3.98

,499 ,439 -,273 M=3.81

,441 ,874 -,357 M=5.19

,286 ,809 -,368 M=5.23

,505 ,669 -,648 M=4.62

,544 ,298 -,855 M=3.77

,547 ,388 -,751 M=3.65,477 ,263 -,659 M=3.40

NFT_9_A NFT_7_A NFT_12_A NFT_5_A NFT_2_A NFT_1_A NFT_3_I NFT_4_I NFT_8_I NFT_10_I NFT_11_I NFT_6_I

1a 2b 3cFactor

Extraction Method: Principal Axis Factoring.Rotation Method: Oblimin with Kaiser Normalization.

KMO = 0,88; p <0.000; M = mean; scale of 7-pointsa.

Eigenvalues = 5.5; 1.5 and 1.05b.

Acumulated Variance Explained = 67%c.

Page 7: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

7

Second, according with Table 1, inherently three variables became below 0.45 in communalities (indicated as minimum by Clarck and Watson, 1995). According to Floyd and Widaman (1995), “communality of a variable is the variance that variable shares with the latent variables underlying the set of observable measures” (p.290). The result under 0.45 means that the item either not be related to the other items, or suggest an additional factor that should be explored (Costello and Osborne, 2005). The communalities values were “NFT_1_A”, h² = .33; “NFT_2_A”, h² = .34; and “NFT_6_I”; h² = .44. Thus, these variables that did not achieve the minimum should be excluded. Because of that, the three variables under 0.45 in communalities were excluded and a second factor analysis was done. RanEingen Program Syntax to determine the number of random eigenvalues was used in this analysis (Enzmann, 1997). Eigenvalues from the research data (5.5; 1.5 and 1.05) when compared to random data (1.47; 1.34 and 1.24) showed that the ideal dimensions should be just two dimensions of NFT scale, which is in agreement with the one suggested by literature. These results represent perfectly the dichotomy view of the utilitarian view (consumers are concerned with purchasing products in an efficient and timely manner to achieve their goals with a minimum of irritation) and the adventure view, which reflects shopping’s potential entertainment value and the enjoyment of any prespecified end goal (Peck and Childers, 2003). In consequence, the two dimensions are showed much separated. The results can be viewed in Table 2. Between brackets are the standardized factor loadings from the confirmatory factor analysis. The variance explained was 51% in the first dimension and 16% in the second dimension (67% total). Streiner (1994) suggested that factors should explain at least 50% of the total variance. Using again Parallel Analysis, this second factor analysis showed that eigenvalues from the research data (4.57 and 1.48) when compared to random data (1.37 and 1.25) perfectly present two dimension of NFT scale.

Table 2: Second Exploratory Factor Analysis of NFT

The 4-item autotelic scale had a coefficient α of 0.85, while the 5-item instrumental scale had a coefficient α of 0.83. These two alpha coefficients are according with the ≥ .80 suggested by Nunnally (1978) and each construct has more than three variables (Costello and Osborne, 2005). The other reliability indicators also showed that the scales sound good. For example, in Average Variance Extracted (AVE) the 4-item autotelic scale had a value of 0.59 and

,79 [0,78] ,364

,78 [0,80] ,318

,73 [0,77] ,350

,71 [0,72] ,226

,679 ,46 [0,67],652 ,51 [0,67],284 ,81 [0,66],402 ,80 [0,71],588 ,78 [0,85]

NFT_12_A

NFT_9_A NFT_7_A NFT_5_A NFT_10_I

NFT_11_I

NFT_4_I NFT_3_I NFT_8_I

1 a 2

Factor

Extraction Method: Principal Axis Factoring.Rotation Method: Oblimin with Kaiser Normalization.

KMO=0.848; p<0.000; all communalities areabove 0.45; Eigenvalues are 4.57 and 1.48.

a.

Page 8: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

8

instrumental of 0.53; and in the Composite Reliability (CR) autotelic = 0.85 and instrumental = 0.84 (Fornell and Larcker, 1981). According to Table 2, it is important to comment that two instrumental variables showed problems (“NFT_10_I” and “NFT_11_I”), since they had high loads on both dimensions. Because of that, done as alternative option, a third rotation, now fixing the number of factors in two, instead of eigenvalue over one, showed exactly that these two items had problems and loaded near to autotelic dimension (see Figure 1). Figure 1 presents the factor plot of the two dimensions, where the regression prediction line (diagonal) and mean of Y reference line (horizontal) are showed in red. This third rotation did not improve or damage the second model. Thus, to decide which model is best to represent the NFT phenomenon, we decided by the second factor analysis, given that the first model had problems with communalities and overextraction of components and the third model did not improve the second one.

Figure 1: Factor Plot in Rotated Factor Space

NFT - Autotelic

1,0,50,0-,5

NFT

- In

stru

men

tal

1,0

,5

0,0

-,5

nft_12_a

nft_11_inft_10_i

nft_9_a

nft_8_i

nft_7_a

nft_6_i

nft_5_a

nft_4_inft_3_i

nft_2_anft_1_a

The next step was to assess scale structure, confirmatory factor analysis was used. Concurrent models using the variance covariance matrix was estimated using AMOS (Byrne, 2001). Competing measurement models were examined and the results can be viewed on Table 3. The model “B” had a r of 0,67 (p <0.001; r² = 0,45). This result is the same one found by Peck and Childers (2003) in their study (r = 0,64; p < 0.001), where the association value (0,67 vs. 0,64) is very close in both studies. However, the problem is that the structural indexes were poor. For instance, AGFI, GFI, NFI were all under 0.90 and RMSEA was above 0.08 (McDonald and Ho, 2002). It denotes that the theoretically model does not fit well on the real data. In this sense, Hox and Bechger (1998) explain that “perfect fit may be too much to ask for, instead, the problem is to assess how well a given model approximates the true model” (p.9).

Page 9: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

9

Table 3: Structural Equation Model Fits Model χ²/d.f. p-value AGFI GFI NFI RMSEA

A. Two factor without correlation 6,98 0,000 0,67 0,80 0,75 0,19 B. Two factor with correlation 4,96 0,000 0,71 0,83 0,83 0,16 C. One general factor 7,79 0,000 0,58 0,58 0,72 0,20 D. Second order two factors 4,96 0,000 0,71 0,83 0,83 0,16 E. Three factors with correlation 2,33 0,000 0,82 0,88 0,87 0,09 F. Three factors without correlation 4,79 0,000 0,70 0,79 0,73 0,15 G. Second order three factors 2,33 0,000 0,82 0,88 0,87 0,09

Convergent Validity. The first assumption stated that NTI scale will have a higher positive relationship with the instrumental dimension of NFT vs. the autotelic dimension of NFT. A structural model was built to test this hypothesis. Initially, the model fits were χ²/d.f. = 3,19; p-value = 0,000; AGFI = 0,748; GFI = 0,82; NFI = 0,828 and RMSEA = 0,11. The structural adjustment is not so good; however, it is important to comment the main goal here is to verify the variable association, and not just the structural model. The 5-item NTI scale (α = 0,89) was related positively to instrumental (β = 0.77; p < 0.001; r = 0.75; p < 0.001) and autotelic dimensions (β = 0.63; p < 0.001; r = 0.65; p < 0.001), where the instrumental factor is higher in beta value than autotelic. These consequences support P1. It suggested that the desire or Need for Tactile Input to make brand/product evaluations has a strong impact on those aspects of prepurchase touch that reflect outcome-directed touch with a salient purchase goal – instrumental NFT. Discriminant Validity The second proposition declared that the 18-item NFC scale would not have a significant association with neither the instrumental dimension of NFT nor the autotelic dimension of NFT. First, an exploratory factor analysis showed problems with the NFC dimensionality. The results suggested five dimensions to a scale that in the Cacioppo and Petty (1982) theoretical construction has just one´ (Kaiser-Meyer-Olkin = 0,71; p < 0.000). Because of that problem, it was decided to fix the number of factors in one, instead of using the eigenvalue over one (Kaiser-Meyer-Olkin = 0,70; p < 0.000). Thus, four variables were retained (α = 0,71) in the first dimension and theses ones were used to represent the NFC latent construct. The structural model fits for the second hypothesis were χ²/d.f. = 2,533; p-value = 0,000; AGFI = 0,78; GFI = 0,83; NFI = 0,72 and RMSEA = 0,10. The reduced NFC scale, as a predicted, was neither related to instrumental (β = 0.14; n.s.; r = 0.17; n.s.) nor autotelic (β = 0.19; n.s.; r = 0.10; n.s.), supporting the hypothesis. Here, the result indicates that although both NFT and NFC overlap in the domain of information-acquisition behavior, they are discriminant in their features and goals. P3 affirmed that the 16-item NTE instrument will not have a significant association with neither the instrumental dimension of NFT nor the autotelic. It is a similar statement of the second hypothesis. Originally, an exploratory factor analysis showed troubles with the NTE dimensionality. It recommended five dimensions to a instrument that is theoretically created in just one factor (Kaiser-Meyer-Olkin = 0,70; p < 0.000). Because of that problem, it was decided to fix the number of factors in one (Kaiser-Meyer-Olkin = 0,70; p < 0.000) rather than use the eigenvalue over one. Thus, four items were retained (α = 0,71) and theses indicators were used to test the premise. Structural model fits (χ²/d.f. = 3,57; p-value = 0,000; AGFI = 0,75; GFI = 0,83; NFI = 0,77 and RMSEA = 0,13) showed that the 4-item NTE scale was not only related positively to instrumental (β = 0.44; p < 0.000; r = 0.39; p < 0.01), but also to autotelic (β = 0.22; p < 0.03; r = 0.10; n.s.). The result does not support P3.

Explanation of the P3 result. Peck and Childers (2003) hypothesized that Need to Evaluate represents information acquisition, but in the context of chronic evaluation and that Need for

Page 10: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

10

Touch is not in the context of chronic evaluation. The evidences indicated that both NTE and NFT represent information acquisition, since they are positively associated. Specifically, Table 5 shows that NTE is positively associated with Instrumental dimension of NFT, which reflects the information acquisition to decision making. We hypothesized that NFT instrumental dimension represents the “subconscious” aspect of touching with a goal of evaluating the products and do better judgments. Because of that these two constructs have similar aspects within them, i.e. chronic evaluation. Subconscious because NFT and NFC did not associated. Peck and Childers (2003, p.434) comment that saying that “Evaluation represents only one aspect of the NFT, particularly when instrumental is contrasted with autotelic touch”. Nomological Validity. The next propositions believe that a consumer who values haptic information would be expected to be less likely to purchase products via marketing channels where direct product touch is prohibited. Thus, catalog, telephone and internet were the ways that consumer could buy its products without touching in them. These three observable variables were regressed against autotelic and instrumental dimensions, where it was expected a non-significant relation with autotelic and a negative relation with instrumental. The structural model fits were poor, being χ²/d.f. = 6,29; p-value = 0,000; AGFI = 0,62; GFI = 0,75; NFI = 0,63 and RMSEA = 0,18. The results can be viewed on Table 4.

Table 4: Results of P4a and P4b

Endogenous Exogenous Autotelic - P4b Instrumental - P4a Catalog β = 0.05; p>.05 β = 0.11; p>.05 Telephone β = 0.13; p>.05 β = 0.05; p>.05 Internet β = -0.02; p>.05 β = -0.11; p>.05

As a result, proposition P4a was not supported; since all associations were non significant. It was expected negatively significant. It was expected a negative relationship between purchasing via these nontouch media and instrumental dimension of NFT (goal oriented). It can be hypothesized that the purchasing via these nontouch media can be much more by impulse than by a goal oriented focus, and because of that the association was not significant. On the other hand, supposition P4b was supported, because all associations of the autotelic were non-significant, as predictable. In sequence, the experiential buying behavior was the next construct to be analyzed in the model. It was chosen just one dimension from Dawson, Bloch and Ridgway (1990) scale used in the questionnaire. The exploratory factor analysis showed that all variables loaded in just one dimension, harmonizing with literature (Kaiser-Meyer-Olkin = 0,79; p < 0.000). The structural model fits were χ²/d.f. = 3,55; p-value = 0,000; AGFI = 0,75; GFI = 0,82; NFI = 0,76 and RMSEA = 0,13. The 5-item scale had coefficient α of 0.84, was not related to instrumental (β = 0.15; n.s.; r = 0.14; n.s.), and was positively associated to autotelic (β = 0.19; p < 0.05; r = 0.18; p < 0.05). These two results supported both P5a and P5b. The last assumption stated that the impulsive scale will (P6b) not have a significant relation with the instrumental dimension of NFT, (P6a) but a positive and significant with autotelic. A exploratory factor analysis showed that the instrument was load in just a factor (Kaiser-Meyer-Olkin = 0,84; p < 0.000). The structural model fits were χ²/d.f. = 3,40; p-value = 0,000; AGFI = 0,75; GFI = 0,81; NFI = 0,76 and RMSEA = 0,12. The 6-item impulsive scale had coefficient α of 0.85, was not related to instrumental (β = .12; n.s.; r = 0.11; n.s.), and was

Page 11: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

11

positively associated to autotelic (β = .20; p < 0.05; r = 0.19; p < 0.05). These outcomes supported hypotheses P6a and P6a. As complement of the propositions test, the correlation matrix is showed in Table 5. Some results deserve eminence at a first glance. Initially, the NTE did not associate with NFT autotelic, although the regression model showed significance (β = 0.22; p < 0.03). Second, the NFC scale did not have association with the NFT, as predicted in the literature and supported by the tests. In the end, impulsive behavior showed a positive association with (a) the NFC (r = 0.24; p < 0.05) and with (b) the experiential conduct (r = 0.60; p < 0.001). First, this result suggested that even the individual is buying by the momentum or motivated by the pulse she/he needs to use cognition to decide. Second, since experiential conduct relates to social or recreational motivations rather than to purchase products or shopping for the sake of the “experience” (Peck and Childers, 2003), it appear to be related to impulsive buying, which is not an organized buying. Buying by phone and by Internet presented a positive association with experiential behavior. It could indicate that phone and internet are ways of a buying system provoked by leisure and entertaining trait.

Table 5: Correlation Matrix of the Constructs 1 2 3 4 5 6 7 8 9 1. Instrumental 1[.53] 2. Autotelic 0.67*** 1[.59] 3. NFC 0.17 0.10 1[.51] 4. NTE 0.39** 0.13 0.02 1[.39] 5. Experiential 0.14 0.18* 0.19* 0.03 1[.51] 6. Impulsive 0.11 0.19* 0.24* -0.07 0.60*** 1[.50] 7. Catalog 0.09 0.12 0.10 -0.06 0.14 0.19* 1[n.a.] 8. Phone 0.06 0.16 0.25* 0.02 0.43*** 0.34** 0.53*** 1[n.a.] 9. NTI 0.75*** .65*** 0.11 *0.27 0.16 .13 0.11 0.07 1[.62] 10. Internet -0.03 0.07 0.16 -0.10 0.41*** 0.41*** 0.43*** 0.50*** 0.02

*p < 0.05; **p < 0.01; ***p < 0.001; Between brackets are AVE; n.a. = not available Moderating Effects The moderating effects were tested by multigroup analysis in structural equation modeling. The full sample was divided in two groups using a median split of the NFT autotelic scale, which is a common procedure suggested by literature (Bell and Luddington, 2006; Evanschitzky and Wunderlich, 2006; Voorhees and Brady, 2005). In general terms, a moderator is a qualitative (e.g. sex, race) or quantitative (e.g. level of reward) variable that affects the direction and/or strength of the relation between and independent and a dependent (Sauer and Dick, 1993). Baron and Kenny (1986) comments that “it is desirable that the moderator variable be uncorrelated with both the predictor and the criterion to provide a clearly interpretable interaction term” (p.1174). This criterion is partially fulfilled the case in the first moderating test. For instance, Experiential vs. Autotelic (r = 0.18; n.s.); Impulsive vs. Autotelic (r = 0.19; p < 0.05); NFC vs. NTI (r = 0.11; n.s.) and totally accomplished in the second moderating test NFC vs. Autotelic (r = 0.17; n.s.); NFC vs. Instrumental (r = 0.10; n.s.); NFC vs. NTI (r = 0.11; n.s.). Dawson, Bloch and Ridgway (1990) proposed the experiential motivation for shopping is related to social or leisure motivations. On the other hand, Rook and Fisher (1995) salient

Page 12: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

12

those highly impulsive buyers are more likely to experience spontaneous buying. Thus, it is supposed that since consumer is walking because of recreational, desire for fun and relaxation motivations (i.e. experiential), some impulsive/spontaneous buying behavior could happen. Is is because that the individuals are more ‘open’ and friendly to unexpected buying in that moment. Consequently to that association, given that NFT autotelic is guided by a no purchase goal preexisted, it could moderate the association between experiential motivation for shopping and impulsive motivation. Specifically, the proposition suggest is (P7) if the consumer has a higher (lower) NFT autotelic tendency, the experiential motivation for shopping will play a more (less) important role on impulsive motivation. Initially, the result of the association between experiential motivation and impulsive buying behavior was supported (β = .59; p < 0.000; R² = .35). The effect of experience on impulsive buying behavior was different across high autotelic (i.e. high enjoyment) vs. low autotelic (i.e. small fun and arousal). The Δχ2 value were = 47.35, Δd.f. = 1, sig = p < 0.001. A positive influence was found for both groups, but this relationship was stronger for the high autotelic (β = .61; p < 0.01; n = 75) when compared to the low autotelic (β = .50; p < 0.01). The next assumption assumes that NFC might moderate the relationship between NFT and NTI. Given that the association between these two constructs was presented in the literature review part, it won’t be showed again (see P1). Hence, explicitly the hypothesis is that (P8a) if the consumer has a higher NFC (low NFC), the instrumental NFT dimension will play a more (less) important role in predicting NTI, since instrumental is reflect of outcome-directed touch with a salient purchase goal and it presupposes more cognition to achieve a specific outcome previously established. On the contrary, (P8b) if the consumer has a smaller (higher) NFC, autotelic NFT will play a more (less) important role in predicting NTI, since autotelic factor corresponds with no purchase aim necessarily salient. Thus, autotelic factor could not demand a more severe thinking or information processing from the individual (vs. instrumental). The effect of NFT autotelic (β = .26; p < 0.000) and NFT instrumental (β = .58; p < 0.001) on NFI was significant. These two dichotomous construct were different not only across high autotelic vs. low autotelic but also across high instrumental vs. low instrumental. The two models showed significance between them Δχ2 = 33.05; d.f. = 1, sig = p < 0.001. First, (P8a) a positive influence was found for both groups, but this relationship was stronger for the high NFC (β = .54; p < 0.01; n = 77) when compared to the smaller NFC (β = .48; p < 0.01). Second, (P8b) a positive influence was found for just one group. Thus, this relationship was positive and significant for the low NFC (β = .38; p < 0.05; n = 78) when compared to the high NFC (n.s). The results from the moderating effects can be viewed on Table 6.

Table 6: Moderating Effects of NFT and NFC constructs Smaller Higher Relation Moderator β t β t

P7 Experiential → Impulsive Buying NFT: Autotelic .50 2.88** .61 3.16** P8a NFT Instrumental → NTI NFC: Cognition .48 2.96** .53 3.30*** P8b NFT Autotelic → NTI NFC: Cognition .38 2.37* .24 1.73

NOTE: Maximum Likelihood (ML) estimation was used; moderation tests analysis were used with unstandardized coefficient for comparation as suggested by Kline (1998). For reading propose standardized coefficient are presented. * p < 0.05 ** p < 0.01 *** p < 0.001

General Conclusion This paper had as a main goal to analyze the psychometric proprieties of the 12-item NFT scale. As a main conclusion, the instrument really possessed the hypothesized two-factor

Page 13: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

13

structure, demonstrated high reliability in each of the two dimensions. The dual characterization of NFT from both the retail as well as the psychological literature on motivations (and empirically found here) is consistent with the perspective of NFT as multi-dimensional construct with two underlying factors, Instrumental and Autotelic Touch. The 4-item autotelic scale had a coefficient α of 0.85, while the 5-item instrumental scale had a coefficient α of 0.83. In addition, even though P4a was not supported, P4b was confirmed. This result, added to convergent validity using Garver and Mentzer (1999) t-values above 1.96, indicates that the instrument has convergent validity. For example, all t-values in the confirmatory factor analysis were above 8.56; p < 0.000. Other results also deserve prominence. Need for Tactile Input was related positively to instrumental and autotelic dimensions of NFT, where the instrumental factor is higher in beta value than autotelic’. The reduced NFC scale, as a predicted, was neither related to instrumental nor autotelic, supporting the discriminant validity and demonstrating that they are not the same phenomenon. Impulsive instrument was positively associated to autotelic, since this last one denotes the touch as an end in and of itself. NFT Autotelic dimension showed its role as moderator variable, since when the consumer has a higher NFT autotelic tendency (e.g. touching by fun), the experiential motivation for shopping plays a more important role on impulsive motivation. On the other hand, if the consumer has a higher NFC, instrumental NFT dimension plays a more important function in predicting NTI, given that instrumental is a reflect of outcome-directed touch with a prominent purchase goal. On the opposing, if the consumer has a smaller NFC, autotelic NFT will play a more important role in predicting NTI, since autotelic factor corresponds with no purchase aim necessarily salient. Summarizing, it is important to note that the moderator role of NFT Autotelic dimension needs to be more explored in future studies, as well as its mediator role. Future research. Future investigations should test other associations with NFT, looking for strengthening the phenomenon in a nomological network. These studies can include (a) attitude toward product and confidence in purchasing moment, as consequents of the touch. Besides, other investigations (b) might analyze the NFT construct and its relationship with attention and memory. For instance, higher NFT individuals might be able to be more readily retrieve information from memory than smaller NFT. Finally, (c) according to Table 2, it is important to comment that two instrumental variables showed problems (“NFT 10 I” and “NFT 11 I”), since they had high loads on both dimensions. In addition, an “Instrumental-Trust” dimension and “Instrumental-Touch as Duty” factors were indicated by this study. In that sense, future research can created other new indicators to overlap those two variables with problems and to measure these two constructs hypothesized. Then, basing on this new instrument, it will be possible to test if this new scale created will have an “Instrumental-Trust” and “Instrumental-Touch as Duty” factors and/or problems of high loads on more than one dimension.

Appendix 1: NFT Correlation Matrix

Page 14: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

14

References Baron, R.M. & Kenny, D. A. (1986). The Moderator-Mediator Variable Distinction in Social

Psychological Research: Conceptual, Strategic and statistical considerations, Journal of Personality and Social Psychology, 51 (6), p.1173-1182, December.

Bell, S. J. & Luddington, J.A. (2006). Coping with Customer Complaints. Journal of Service Research, 8 (3), p.221-33.

Byrne, B. (2001) Structural Equation Modeling with AMOS: basic concepts, applications and programming. Mahwah, NJ: Lawrence Erlbaum Associates.

Cacioppo, J.T. & Petty, R.E. (1982). The Need for Cognition. Journal of Personality and Social Psychology, 42 (1), p.116-131.

Churchill, G. A. Jr. (1979). A Paradigm for Developing better Measures of Marketing Constructs. Journal of Marketing Research, 16 (1), February, p.64-73.

Churchill, G. A. Jr. (1996). Marketing Research: Methodological Foundations. 3rd. The Dryden Press, Chicago.

Citrin, A., Stem, D., Spangenberg, E. & Clark, M. (2003). Consumer Need for Tactile input: an internet retailing challenge. Journal of Business Research. 56, p.915-922.

Costello, A.B. & Osborne, J.W. (2005). Best Practices in Exploratory Factor Analysis: four recommendations for getting the most fro your analysis. Practical Assessment, Research & Evaluation, 10 (7), July, p.1-9.

Dawson, S., Bloch, P.H. & Ridgway, N.M. (1990). Shopping motives, emotional states and retail outcomes, Journal of Retailing, 66 (winter), p. 408-427.

Enzmann, D. (1997). RanEigen: a program to determine the parallel analysis criterion for the number of principal components. Applied Psychological Measurement, 21 p.232.

Evanschitzky, H. & Wunderlich, M. (2006). An Examination of Moderator Effects in the Four-Stage Loyalty Model. Journal of Service Research, 8 (4), p.330-345.

Fiore, A.M., Moreno, J.M., & Kimble, P.A. (1996). Aesthetics: a comparison of the state of the art outside the inside the filed of textile and clothing. III. Appreciation process, appreciator and summary comparisons. Clothing and Textiles Research Journal, 14 (3), p.169-184.

Itens 1 2 3 4 5 6 7 8 9 10 11 12

1NFT_A Quando vou a uma loja, não posso deixar de tocar nos produtos 1

2NFT_A Tocar nos produtos pode ser divertido .36 1

3NFT_IEu tenho mais confiança nos produtos quando eu posso tocarneles antes da compra .42 .44 1

4NFT_IEu me sinto mais confortável comprando produtos depois deexaminá-los fisicamente .38 .34 .68 1

5NFT_A Quando estou nas lojas é importante para mim tocar todos ostipos de produtos .33 .42 .23 .11 1

6NFT_ICaso eu não possa tocar num produto na loja, eu resisto em comprá-lo .24 .32 .24 .27 .44 1

7NFT_A Gosto de tocar nos produtos mesmo sem intenção de comprá-los .35 .41 .38 .19 .44 .33 1 8NFT_I

Eu me sinto mais confiante em fazer uma compra depois de tocar no produto .29 .36 .63 .61 .37 .45 .45 1

9NFT_A Quando eu estou na loja, eu gosto de tocar em muitos produtos .39 .35 .30 .23 .57 .42 .64 .40 1

10NFT_I A única maneira de estar certo sobre um produto é tocando ele .25 .33 .30 .30 .43 .55 .45 .57 .46 1

11NFT_IExistem muitos produtos que eu poderia comprar ou ter comprado se eu pudesse tocar ou ter tocado neles .28 .37 .42 .35 .39 .53 .42 .50 .49 .66 1

12NFT_A Eu me identifico tocando nos produtos nas lojas. .37 .37 .30 .27 .59 .40 .57 .44 .62 .55 .51 1All correlations are p<0,01

Page 15: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

15

Floyd, F.J., & Widaman, K.F. (1995). Factor Analysis in the development and refinement of clinical assessment instruments. Psychological Assessment, 7 (3), p.286-299.

Fornell, C., & Larcker, D.F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18 (1), p.39-50, February.

Garver, M.S., & Mentzer, J.T. (1999). Logistics research methods: employing structural equation modeling for construct validity. Journal of Business Logistics, 20 (1), p.33-57.

Hair, J.Jr., Anderson, R.E., Tatham, R.L., & Black, W. (1998). Multivariate Data Analysis. New Jersey: Prentice Hall.

Holbrook, M.B., & Hischman, E.C. (1982). The experiential aspects of consumption: consumer fantasies, feelings and fun. Journal of Consumer Research, 9 (2), p.123-140.

Hox, J.J., & Bechger, T.M. (1998). An introduction to Structural Equation Modeling. Family Science Review, 11, p.354-373.

Jacoby, Jacob (1978). Consumer Research: A State-of-the-Art Review. Journal of Marketing, 42 (2), April, p.87-96.

Jarvis, W.B.G., & Petty, R. (1996). The need to evaluate. Journal of Personality and Social Psychology, 70 (1), p.172-194.

Johansson, R.S. (1978). Tactile sensibility in the human hand: receptive field characteristics of mechanoreceptive units in the Glabrous Skin Are. Journal of Physiology, 281 (2), p.101-123.

Klatzky, R.L. & Lederman, S.J. (1992). Stages of manual exploration in haptic object identification. Perception & Psychophysics, 52 (6), p.661-670.

Klatzky, R.L. & Lederman, S.J. (1993). Toward a computational model of constraint-driven exploration and haptic object identification, Perception, 22 (5), p.597-621.

Klatzky, R.L., Lederman, S.J. & Matula, D.E. (1993). Haptic exploration in the presence of vision. Journal of Experimental Psychology: Human Perception and Performance, 19 (4), p. 726-743.

Kline, R. B. (1998). Principles and practice of structural equation modeling. 1st. New York: Guilford Press.

Lederman, S.J., & Klatzky, R.L. (1987). Hand movements: a window into haptic object recognition. Cognitive Psychology, 19 (3), p.342-368.

Malhotra, N. (2001), Pesquisa de Marketing. 3rd. Porto Alegre, Bookman. McDonald, R.P., & Ho, M.R. (2002). Principles and practice in reporting structural equation

analyses. Psychological Methods, 7 (1), p.64-82. Nunnally, J.C. (1978). Psychometric Theory. 2nd. New York: McGraw-Hill. Peck, J., & Childers, T.L. (2003). Individual differences in haptic information processing: the

“Need for Touch” scale. Journal of Consumer Research, 30, (3), p.430-442. Peter, P.J. (1979). Reliability: a Review of Psychometric Basic and Recent Marketing

Practices. Journal of Marketing Research, 16 (1), February, p.64-73. Reise, S.P., Comrey, A.L., & Waller, N.G. (2000). Factor analysis and scale revision.

Psychological Assessment. 12 (3), p.287-297. Rook, D.W., & Fisher, R.J. (1995). Normative influences on impulsive buying behavior.

Journal of Consumer Research, 22 (3), p. 305-313. Sauer, P.L., & Dick, A. (1993). Using moderator variables in structural equation models. In:

Advances in Consumer Research, 20, p.636-640. Streiner, D.L. (1994). Figuring out factors: the use and misuse of factor analysis. Canadian

Journal of Psychiatry, 39, p.135-140. Tabacknick, B.G., & Fidell, L.S. (2001). Using Multivariate Statistics. Boston: Allyn and

Bacon.

Page 16: Autoria: Valter Afonso Vieira, Claudio Vaz Torres, Rogerio ... · validity of NFT scale, since these steps seek to assess the construct validity. According to Peter (1979) “valid

16

Trochim, W.M. (2002). Research Methods knowledge Base. Available at http://www.socialresearchmethods.net/kb [Accessed in April 2007].

Voorhees, C. M. & Brady, M.K (2005). A Service Perspective on the Drivers of Complaint Intentions. Journal of Service Research, 8 (2), p. 192-204.

Zoski, K.W., & Jurs, S. (1996). An objective counterpart to the visual scree test for factor analysis: the standard error scree. Educational and Psychological Measurement, 56, p.443-451

Zwick, R., & Velicer, W.F. (1986). Comparison of five rules for determining the number of components to retain. Psychological Bulletin, 99, p.432-442.