|Title:||The Effect of Attribute Order on Judgment in Chinese and English|
Tavassoli, Nader T., Marketing Subject Area, London Business School,
London, United Kingdom
Lee, Yih Hwai, Department of Marketing, National University of Singapore, Singapore
|Source:||Journal of Experimental Psychology: Applied, Vol. 10(4), December 2004. pp. 258-266.|
|Publisher:||American Psychological Association|
|Digital Object ID:||10.1037/1076-898X.10.4.258|
|Article Type:||Journal Article|
|Abstract:||The authors found that the order of attribute presentation had a stronger effect on judgment in English than in Chinese. In Experiment 1, with a sample of 102 female and 63 male bilingual Singaporeans, the authors found that participants' memory-based judgments showed a stronger primacy effect in English than in Chinese that was mediated by recall from long-term memory. In contrast, participants' online (immediate) judgments showed a primacy effect in both languages that was unmediated by recall from short-term memory. In Experiment 2, with a sample of 67 female and 53 male bilingual Singaporeans, the authors found that participants' online judgments were more influenced by the attribute order of a previously seen competitive advertisement in English than in Chinese. A cross-cultural field study in Mainland China and the United Kingdom provided external validity for the experimental results.|
|Persistent link to this record:||http://search.epnet.com/login.aspx?direct=true&db=pdh&an=xap104258|
|* * *|
Marketing Subject Area, London Business School
Yih Hwai Lee
Department of Marketing, National University of Singapore
The authors thank Gordon Pincott and Millward Brown for making available the cross-cultural field data. Nader T. Tavassoli gratefully acknowledges funding from the Centre of Marketing at the London Business School.
Correspondence concerning this article should be addressed to: Nader T. Tavassoli, London Business School, Regent's Park, London, NW1 4SA, United Kingdom. Electronic mail may be sent to:firstname.lastname@example.org
Imagine evaluating an advertised product or service. When consumers judge an ad online (i.e., immediately as it unfolds), it is evaluated item by item as attribute information is acquired (Ash, 1946; Anderson & Hubert, 1963; Hastie & Park, 1986; Wyer & Srull, 1986). When judgments are formed some time after the information is acquired, they are memory-based; that is, they are based on information retrieved from memory (Bargh & Thein, 1985; Chattopadhyay & Alba, 1988; Hastie & Park, 1986; Lichtenstein & Srull, 1987; Unnava, Burnkrant, & Erevelles, 1994). Judgments can also be of a hybrid form, which is when some information is at hand and some is retrieved from memory (Alba, Marmorstein, & Chattopadhyay, 1992; Biehal & Chakravarti, 1983; Kardes, 1986; Keller, 1987; Lynch, Marmorstein, & Weigold, 1988; Lynch & Srull, 1982). For example, consumers may evaluate product label information while retrieving information from memory about a competitive brand they encountered previously.
In this study, we examined whether there are differences in how the same information is evaluated across these judgment situations when only the information presentation order is changed, which is one of the most basic considerations in the design of a persuasive message (Belch & Belch, 2001, p.184). In particular, we examined this issue across the alphabetic English and logographic Chinese scripts. One fourth of the world population, including speakers of Chinese, Japanese, and Korean, read logographs that are based on meaning rather than units of sound. English and most other languages rely instead on alphabetic scripts that were developed to represent the sounds of words and not their meanings. These scripts include the Latin alphabet (e.g., English and Spanish), Arabic, Hebrew, and Cyrillic scripts (e.g., Russian), as well as alphabetic scripts that are used in Japanese and Korean that were developed to complement the use of logographs.
We argue that the order of presentation should have a stronger effect on judgments in English when these judgments are shaped by information retrieved from memory, such as in memory-based and hybrid judgments. This is because order memory for logographs is not routinely encoded to the same degree as it is for alphabetic English words (Tavassoli, 1999; Tavassoli & Han, 2001). In Experiment 1, we examined the effect of attribute order on online and memory-based judgments in Chinese and English. In Experiment 2, we examined hybrid judgments for which an online evaluation is made subsequent to an exposure to a competitive ad. We then present the cross-cultural field data with which we examined the external validity of the experimental findings. We conclude by discussing the theoretical contribution and practical implications.
Recent research has demonstrated that order memory is superior for words written in alphabetic scripts than for words written in logographic scripts (Tavassoli, 1999; Tavassoli & Han, 2001). The explanation for this effect is that people primarily engage a subsystem of working memory known as the phonological loop to encode and rehearse alphabetic words; however, people do not predominantly involve the phonological loop when processing logographs, but, instead,they rely more on visual working memory (e.g., Hung & Tzeng, 1981; Tan et al., 2001; Tavassoli & Lee, 2003; Zhou & Marslen-Wilson, 1999). Information is rehearsed in the phonological loop in a serial fashion, whereas it is rehearsed in visual working memory spatially and in parallel (Baddeley, 1986; Paivio, 1986). As a result, the encoding of order memory for alphabetic words is accentuated compared with logographs, as can be demonstrated by a superior performance on direct tests of order memory (Tavassoli, 1999; Tavassoli & Han, 2001).
It is not obvious, however, if and how the highly theoretical results of Tavassoli (1999) and Tavassoli and Han (2001) would transfer to consumer judgments. Those authors examined the differential ability of readers of alphabetic and logographic scripts to purposefully reconstruct the presentation order of a list of words that had been presented in sequence. For the memory test, participants received a shuffled deck of cards in which each card contained one of the presented words. Participants were instructed to reorder these cards into the same order the words had been presented in. This type of card sort isolates order information from memory for the items themselves, as the items are printed on the cards during the task. Moreover, in this direct test of explicit order memory, participants are intentionally attempting to reconstruct the order of information and this may not reflect routine retrieval strategies typical of consumer settings. This is especially true as the stimuli were a list of nonsense words (Tavassoli & Han, 2001) and a list of unrelated nouns (Tavassoli, 1999). In consumer judgments, on the other hand, participants are not intentionally trying to reconstruct the order. Instead, judgments capture the degree to which participants incidentally rely on order in retrieving item information from memory. In these situations, it is the interplay between item information and contextual order information that is critical, determining the accessibility of items from different input positions.
How does order memory affect the accessibility of item information? Order memory is a distributed property of the entire set of information (Murdock, 1960), and the affect on memory retrieval depends on the specific retrieval and encoding context. Retrieval in recall is influenced by an item's relative distinctiveness, which is a function of item-specific cues (e.g., temporal, visual, and semantic), relational cues (e.g., associations to already retrieved words), and encoding effects (e.g., attention and interference; Burgess & Hitch, 1992). Although temporal order memory is one of the most important retrieval cues, it can result in a primacy effect, a whole-list effect, or a recency effect depending on the precise nature of the memory task. For example, when recall memory is assessed immediately after presentation, order memory results in a recency effect. As the interval between learning and retrieval increases, there is a shift from recency to primacy. Although there are various memory models that attempt to account for this recency-to-primacy shift, the expression below provides a simple illustration of the basic effect (e.g., Murdock, 1960; Neath & Crowder, 1990):
The explicit memory tests of Tavassoli and colleagues (Tavassoli, 1999; Tavassoli & Han, 2001) isolated order memory in a card sort, and because there was no item retrieval involved, the results demonstrated a whole-list advantage for alphabetic over logographic words. In contrast, the memory model just described illustrates that temporal distinctiveness should be highest for end-of-list items when a memory task occurs immediately. The results of Tavassoli and colleagues, therefore, predict a stronger recency effect for alphabetic than for logographic words in immediate recall. After as little as a 10-s delay, recall starts exhibiting both primacy and recency, and after typically no more than 2 min, only a primacy and no recency effect remains (Wright et al., 1985). The results of Tavassoli and colleagues, therefore, predict a stronger primacy effect for alphabetic over logographic words from long-term memory. These hypotheses are tested in Experiment 1, where we code recency (primacy) as the event of the last (first) word presented being the first word recalled (Kahana, 1996).
In Experiment 1, we also examined the effect of memory retrieval on online versus memory-based judgments. We refer to memory-based judgments as (a) those made subsequent to a person simply having learned product information at the initial exposure; (b) those for which information is retrieved because an initial judgment is not diagnostic (Lynch & Srull, 1982); or (c) those where consumers lack the processing capacity for making an online judgment (Bargh & Thein, 1985). The findings of research in English that have relied on a delayed attitude assessment have shown that information presented first within a message has an undue influence on memory-based judgments (Lichtenstein & Srull, 1987; Unnava et al., 1994). This is because judgments are based on information retrieved from memory. Retrieval from long-term memory that is based on the order of presentation should result in a primacy effect (Knoedler et al., 1999; Wright et al., 1985), such that the first words learned in a list are more likely to be recalled and recalled early in the recall protocol (Kahana, 1996). These attributes would be top of mind and have a disproportionately large impact on judgments (Feldman & Lynch, 1988). Because order memory is more pronounced, it should be a more potent retrieval cue in English than in Chinese. As a result, memory-based judgments should exhibit a stronger primacy effect for alphabetic English ads than for logographic Chinese ads. We assessed primacy in judgment by examining to what degree the valence of product attributes presented at the beginning of an ad affected product evaluations (e.g., Unnava et al., 1994).
In contrast, we did not expect cross-linguistic differences in online judgments. This is because there is typically no relationship between recall and judgments when these occur online and when there is a short delay between the presentation of information and judgment (Hastie & Park, 1986; Wyer & Srull, 1986). Memory does not mediate online judgments, which are anchored by the first few attributes encountered (Ash, 1946; Anderson & Hubert, 1963; Wyer & Srull, 1986). This appears to be largely a result of first-encountered information receiving more attention (Anderson & Hubert, 1963; Rundus, 1971) and affecting the way in which subsequent information is evaluated (Ash, 1946). We therefore expected that online judgments would exhibit a primacy effect both in Chinese and in English.
Participants were 63 male and 102 female bilingual Singaporeans (M age = 18.6 years) who participated in the 2 (Script: alphabetic English vs. logographic Chinese) × 2 (Attribute Order: positive-negative vs. negative-positive) × 2 (Judgment: memory-based vs. online) between-participants experiment. We expected that judgments would be more positive when the positive attributes are presented first than when the negative attributes are presented first. Moreover, this difference should be strongest for memory-based judgments in English.
On the one hand, the use of Singaporean bilinguals restricts the generalizability of any results. On the other hand, this population provides for commonality across a host of factors (e.g., educational styles, product preferences) that differ in cross-country comparisons and that would confound the script conditions. The participants learned both languages in childhood and were carefully screened as to their abilities to read and write Chinese and English, both in their recruitment as well as through the use of self-assessment scales. Early bilinguals have been shown to directly access a single conceptual system through either language (for a review of over 100 cognitive studies of language integration by bilinguals, seeFrancis, 1999). In our experiments, it is therefore not the concepts themselves that are manipulated but their representation, whether they are written in an alphabetic or logographic script. This provides a strong test of the hypothesis that writing systems affect differences in memory and judgment.
We selected positive (charming, unique, cozy, famous) and negative (noisy, greasy, eclectic, busy) attributes from a larger set that was pretested with 32 bilinguals different from the main experiment (16 participants rated the attributes in each language). Participants were told to imagine a restaurant dining experience and to rate each attribute's positivity (1 = very negative; 7 = very positive) and importance in influencing their overall judgment about a dining experience (1 = not at all important; 7 = very important). Across both scripts, the set of positive attributes chosen was rated as significantly more positive (M = 5.48, SD = 0.58) than the set of negative attributes chosen (M = 3.27, SD = 0.78), paired t(31) = 11.12, d = 3.25,p <.01. The set of positive attributes was rated as equally positive in Chinese (M = 5.48, SD = 0.60) and English (M = 5.47, SD = 0.58), t(30) =.05, d =.02, ns, and as equally important in Chinese (M = 4.77, SD =1.32) and English (M = 4.74, SD = 1.01), t(30) =.08, d =.03, ns. The set of negative attributes was rated as equally negative in Chinese (M = 3.23, SD = 0.91) and English (M = 3.32, SD = 0.65), t(30) = -.31, d =.11, ns, and as equally important in Chinese (M = 4.55, SD = 1.22) and English (M = 4.53, SD = 1.05), t(30) =.05, d =.02, ns.
At learning, attributes were individually projected onto a 6-ft (1.85-m) screen for 2 s each and interspersed by a blank screen for 1 s. In the memory-based judgment condition, participants were instructed to try and remember the words. After exposure, all participants were engaged in the same 2-min nonverbal, audiovisual task on which they had been pretrained. This task involved responding to a series of nonverbal sounds and shapes such that participants had to recognize when a stimulus was the same as the one that occurred two stimuli previously. After this task, participants had 2 min to rate other information they had been exposed to and then 2 min to rate the restaurant dining experience on five 7-point scales. Because participants received only memory instructions and not impression formation instructions at learning, the judgment task came as a surprise thus judgments can be considered memory-based (Lichtenstein & Srull, 1987). The online judgment task differed in that participants were told beforehand that they would be asked to judge the restaurant dinner but not that they would be asked to recall information. There was also no filler task, and attitude ratings were made immediately after exposure to the attribute information. In both judgment conditions, the attitude ratings were followed by 1 min of free recall. Participants received a piece of paper containing eight blank lines and were instructed to start with the first line and to only write one word per line. They were further instructed to write down the words in the order in which they came to mind.
We used an alpha level of.05 to determine statistical significance. We usedCohen's (1988)f to indicate effect sizes, where f values of.10,.25, and.40 correspond to small, medium, and large effect sizes, respectively.
The five 7-point attitude scales were anchored by very positive-very negative, very enjoyable-not enjoyable at all, very good-very bad, very pleasant-very unpleasant, and would like to try-would not like to try. The scales did not contain any numbers but they contained seven boxes, one of which participants marked. These were coded from 1 to 7, such that more positive ratings received a higher score. The ratings on these five scales were averaged to provide an index for the overall attitude toward the dining experience (α =.94).
An analysis of variance (ANOVA) was performed on the attitude index for the between-participants factors script (alphabetic English vs. logographic Chinese), order (positive-negative vs. negative-positive), and judgment (memory-based vs. online). The means and ANOVA results are summarized inTables 1 and 2, respectively. There was a significant main effect showing that attitudes were more positive in the positive-negative presentation order and a significant Script × Order interaction suggesting that this effect is stronger in English. A significant Script × Order × Judgment three-way interaction qualified these effects. As predicted, the linguistic attitude differences across the order conditions were larger for memory-based judgments than for online judgments.
Condition Means and Standard Deviations for Attitude Index for Experiment 1
ANOVA Table for Analysis of Attitude Index Dependent Variable for Experiment 1
The pattern of recall was the same across the positive-negative and negative-positive order conditions, and we collapsed the recall data across the order conditions. A greater reliance on temporal order memory predicts a stronger recency effect for recall from short-term memory, but a stronger primacy effect for recall from long-term memory (Murdock, 1960; Neath & Crowder, 1990). We examined recency effects by observing whether the last item presented was the first item recalled, and we examined primacy effects by observing whether the first item presented was the first item recalled (Kahana, 1996).
In recall from short-term memory, there was not a significant difference in recency effects in English (M =.23, SD =.42) and in Chinese (M =.08, SD =.27), t(78) = 1.90, f =.21, ns. In recall from long-term memory, there was a significantly stronger primacy effect in English (M=.67, SD =.47) than in Chinese (M =.36, SD =.49), t(83) = 3.05, f =.33, p <.01. Therefore, our hypotheses, which predicted stronger effects in English than in Chinese for recency in short-term memory and for primacy in long-term memory, received only partial statistical support for the long-term memory condition.
We also examined the relationship between memory and judgment by considering the valence of the first attribute recalled. The positivity of this attribute was coded as +1 if it was a positive attribute and as −1 if it was a negative attribute (Unnava et al., 1994). As expected, the correlation between the positivity of recall from short-term memory and online judgments was not significant in Chinese, r (38) = −.23, or in English, r (38) = −.08, and we do not consider this relationship any further. However, as expected, the correlation between recall from long-term memory and memory-based judgments was significant in Chinese, r (40) =.63, p <.01, and in English, r (41) =.65, p <.01. Therefore, we used Baron and Kenny's (1986) test of mediation to evaluate whether memory mediated memory-based judgments.
Baron and Kenny's (1986) test of mediation specifies three regression models to be run: the first model examines the effect of the mediators (the positivity of the first attribute recalled) on the independent variables (the script and order conditions); the second examines the effect of the dependent variable (attitude index) on the independent variables; and the third examines the effect of the dependent variable on both the independent variables and on the mediators. As shown in Table 3, the mediator is significantly predicted by the independent variables (Equation 1), which significantly predict the dependent variable (Equation 2). Most important, the effects of independent variables become insignificant when the mediator is included in the model (Equation 3). Therefore, the positivity of the first word recalled had a significant mediating effect on attitudes (Söbel  test: z = 2.46, p <.05), lending support to the notion that judgments were memory-based.
Summary of Regression Analysis for Test of Mediation in the Memory-Based Judgment Condition of Experiment 1
The results of Experiment 1 demonstrate that both in Chinese and in English there was a primacy effect in online judgments that were uncorrelated with recall from short-term memory. In contrast, memory-based judgments were mediated by attribute recall both in Chinese and in English. However, this recall was based significantly more on the order of presentation in English than in Chinese, resulting in a stronger primacy effect in memory-based judgments in English than in Chinese. These results suggest that the importance of making a positive first impression is important for online judgments in both scripts, but greater for memory-based judgments in English than in Chinese. They also suggest that the placement of information in an ad should have a stronger effect on the accessibility of information in English and a stronger effect on key measures of advertising effectiveness such as top-of-mind awareness. It is surprising to note that leading advertising testing agencies that we surveyed do not currently assess the impact of attribute order per se on advertising effectiveness.
The results of the online-judgment condition are also important in addressing potential confounds. Bicultural individuals in Hong Kong who, similar to our participants, have internalized two cultures can switch between cultural frames in response to culturally laden symbols (Hong, Morris, Chiu, & Benet-Martinez, 2000). Language could be a cue that primes different cultural frames that may lead to different product-attribute evaluations (Aaker & Sengupta, 2000). Language could have also primed different response styles. For example, Chinese students (in Taiwan) have been found to be more likely to use scale midpoints, whereas North American students were more likely to use extreme values (Chen, Lee, & Stevenson, 1995). These potential confounds can be ruled out, however, because online attitudes were similarly extreme in both languages.
The results of Experiment 1 suggest that online judgments are similar in Chinese and in English because they are not based on product attribute information retrieved from memory. However, in real-life settings many evaluations are of a hybrid form where some information is available for online evaluation and other information is retrieved from memory (e.g., Alba et al., 1992; Biehal & Chakravarti, 1983; Kardes, 1986; Keller, 1987; Lynch et al., 1988; Lynch & Srull, 1982). For example, memory accessibility can affect hybrid judgments when a new brand is being evaluated against a previously encountered brand. Comparability in these hybrid judgments is defined jointly in terms of attribute overlap and attribute memory: More accessible attributes become more diagnostic and thereby define the criteria by which brands are compared (Lynch et al., 1988). Our theorizing predicted that the order in which a competitor presents attribute information in an ad should have a more pronounced effect on the evaluation of a new product in English than it would in Chinese.
In Experiment 2, we tested this hypothesis by first presenting a competitive ad and later asking participants to evaluate a new product. The products were described such that the competitive brand dominated on certain attributes, whereas the new product dominated on others. The attributes in the competitive ad were either ordered such that they began with attributes on which the performance of the competitive brand was relatively stronger and ended with those on which it was relatively weaker, or vice versa. We expected that the evaluation and preference of the new product would be superior when the competitive brand was first described by the relatively weaker attributes than it would be when the competitive brand was first described by the relatively stronger attributes. We expected this effect to be stronger in English than in Chinese. This is because attributes learned first should be more accessible in English and thereby should define the criteria by which the brands are compared. We also avoided the use of negative product attributes that may be more typical of a restaurant review (e.g., Experiment 1) than of most advertisements. This provides a more general test of our theory because a mix of positive and negative attributes can result in a narrow set of processing strategies, such as with a polarizing approach (e.g., Aaker & Sengupta, 2000).
We also varied the order in which we presented the new product information. The new product was described beginning with either the relatively stronger attributes first, or in the reverse order. These attributes were listed in bullet form on the questionnaire to mimic the presentation of product information at the point of sale (e.g., on packaging). We expected that the evaluation, preference and choice of the new product would be superior when it was first described by the relatively stronger attributes than when it was first described by the relatively weaker attributes. We also expected that there would be an interaction between the order of attribute information in the competitive ad with the presentation order of new product information. Specifically, because of an attribute overlap, we expected that the order of new product information would act as a memory prime and increase the accessibility of the attributes first presented in the competitive ad. Attribute-order matching should thereby polarize the ratings between the order conditions of the competitive brand.
Participants were 53 male and 67 female bilingual Singaporeans (M age = 20.5 years) who participated in the 2 (Script: alphabetic English vs. logographic Chinese) × 2 (Order of Competitive Brand Attributes: relatively stronger first vs. relatively weaker first) × 2 (Order of New Product Attributes: relatively stronger first vs. relatively weaker first) between-participants experiment.
The product category we selected was that of a student backpack. We selected the attributes from a larger set that we had pretested with 59 bilinguals that were not involved in the main experiment. Twenty-nine participants rated the attributes in English, and 30 rated them in Chinese. Participants were told to imagine a backpack described by each attribute and to rate each attribute on four 7-point scales anchored by a good product-a bad product, high quality-low quality, highly desirable-not at all desirable, and I would buy this product-I would not buy this product. The ratings on these scales were averaged to provide an overall positivity rating (α =.89). We chose three attributes on which the competitive brand (comes with a three-year warranty, tear resistant fabric, sturdy leather base, reflective patches for safety) was judged as more positive than the new product (comes with a two-month warranty, lightweight plastic, padded shoulder straps for comfort) in both Chinese, t(28) = 4.21, d = 1.54, p <.01; and in English, t(27) = 4.23, d = 1.58, p <.01. We also chose three attributes on which the competitive brand (available in blue or black, handle for easy carrying, two storage areas) was judged as less positive than the new product (available in red, blue, or black; equipped with a side pocket for a water bottle and a front pocket for pens and pencils, five convenient storage areas) in both Chinese, t(28) = 4.16, d = 1.44, p <.01; and in English, t(27) = 3.74, d = 1.39, p <.01.
Participants were told that we were interested in how they typically viewed television ads and that we would ask them some questions about the ads at a later time without specifying the target ad. They were thus not given any specific processing instructions prior to viewing a series of ads presented by a computer and projected onto a 6-ft (1.85-m) screen. One of these was the competitive ad and was repeated twice. This ad contained visuals of a fictitiously branded backpack with the attributes described above the image and embedded in sentences. The exposures to the competitive ads were each preceded and followed by filler ads from different product categories. After viewing the ads, participants participated in a 2-min math quiz to clear short-term memory. Next, they evaluated the new backpack on five 7-point attitude and purchase intention scales anchored by very good-very bad, high quality-low quality, very desirable-not at all desirable, would like to own-would not like to own, and would buy-would not buy. We averaged the ratings on these five scales to provide an index for the overall evaluation of the new backpack (α =.93). In addition, participants were asked to indicate whether they preferred the new backpack to the competitive brand and whether they would choose to receive a brochure for either the new backpack or the competitive brand.
Results and Discussion
The order of the attributes describing the new product (relatively stronger first vs. relatively weaker first) did not have any significant effects on any of the dependent variables, and the data was collapsed across these conditions. ANOVAs were performed on the attitude index, brand preferences, and the request for more information. As the means inTable 4 and the ANOVA results in Table 5 show, the significant interaction effects suggest that the order of attributes in the competitive ad had a stronger effect on the evaluation of the new backpack in English than it did in Chinese. Evaluations of the new backpack in English were less positive when the attributes in the competitive ad had been ordered from relatively stronger to relatively weaker than they were when ordered from relatively weaker to relatively stronger. This suggests an important interplay between memory and online evaluations, such that a competitive ad can provide a mental lens through which new product information is judged.
Condition Means and Standard Deviations for Experiment 2
ANOVA Tables for Experiment 2
There are at least two routes by which a competitive ad can cause this effect. First, in an attribute-based comparison, participants may have attempted to recall matching attributes about the competitive brand from memory while they were evaluating the new product information online (Lynch et al., 1988). Because the temporal order of presentation is a stronger determinant of memory accessibility in English than it is in Chinese, this may have lent more weight to the evaluation of the attributes presented first in the competitive ad. Alternatively, in a brand-based comparison, participants may have formed memory-based judgments for the competitive brand and used this as a standard of comparison to the new backpack. The evaluation of the competitive brand should have been more strongly affected by the presentation order of the attributes in English than in Chinese. However, even though consumers may form global judgments of competitive brands, it is likely that a judgment task that implicitly compared products from the same category facilitated retrieval of discrete attribute information due to attribute overlap (Kardes, 1986). Consumers are more likely to use global judgments when choices are not directly comparable across attributes, such as in the choice of seeing a movie versus going ice skating (Kardes, 1986).
Against our expectations, the order of the attributes in the new product description had no effect on any of the dependent measures. This is surprising, because in Experiment 1 there was a primacy effect for online judgments in both languages. However, whereas attributes in Experiment 1 (and in previous research dating back to Ash  and Anderson and Hubert ) were presented one-by-one in serial order, they were listed as a group with bullet points and could be examined and re-examined in any order before making a judgment. We speculate that this may have weakened any order effects.
To examine the external validity of our results, we sought actual consumer responses to television ads. On the basis of the results of our research, one of the world's leading advertising-testing agencies, Millward Brown, agreed to make available field data from 60 advertising field tests that had been conducted in China and in the United Kingdom. Each ad was tested on approximately 120 consumers. The ads were for frequently purchased consumer goods, and they varied between 15 s and 60 s in length. The ads were not matched across countries, and it was agreed that all client and ad-copy data would remain confidential.
Two research assistants who were blind to the hypotheses complemented the field-test data by recording whether the advertised brand was mentioned by name in each 5-s interval of an ad. The mentioning of a brand name included both text and voice, as this was confounded to a large degree. This coding scheme allowed us to examine the existence of a primacy effect—coded as whether the brand name was mentioned within the first 5 s of an ad—on brand appeal and brand name memorability, two measures that had been collected by the research agency a day after an ad was aired. This field test therefore represents the long-term memory and memory-based judgment condition of Experiment 1.
Brand appeal was measured as the proportion of consumers that agreed with the statement “This advert makes (full brand name) much more appealing to me.” Brand appeal data was missing for five ads, resulting in a reduction in sample size for that measure. Brand name memorability was measured as the proportion of consumers that agreed with the statement “You couldn't fail to remember the advert for (full brand name).” The field data did not include the recall of specific attributes. ANOVAs were performed on the brand name memorability and the brand appeal proportions for the between-ads factors language (Chinese vs. English) and early brand name mention (within the first 5 s vs. after the first 5 s). The means and ANOVA results are summarized in Tables 6 and 7, respectively. For the brand appeal measure there was a significant interaction effect demonstrating that the mentioning of a brand name in the first 5 s of an ad had a more positive effect in English than in Chinese. Whereas there was a negative effect in Chinese, there was a 55% increase in brand appeal in English. The interaction was not significant for the memorability measure. However, under the assumption that each of the 60 ads was tested on 120 consumers, significance can be assessed with a logit model on 7,200 binary observations converted from the proportions data: In this model the interaction effect for memorability is significant, χ2(1, N = 7,200) = 16.20, p <.01. The coarse field data thus provided external validity for the experimental results.
Condition Means and Standard Deviations for Field Test
ANOVA for Field Test
The literature on consumer behavior and on social cognition has only displayed a sporadic and nonsystematic interest in the role that language plays in memory and judgments (cf.Semin & Zwier, 1997) and only recently in the effect of writing systems (Pan & Schmitt, 1996; Schmitt, Pan, & Tavassoli, 1994; Tavassoli, 1999, 2001, 2002; Tavassoli & Han, 2001; Tavassoli & Lee, 2003; Zhang & Schmitt, 2001). However, the idea that language influences cognition has a long tradition in Western thought. The original idea goes back to at least von Humboldt (1836/1988) and gained notoriety in the United States through Whorf's (1940/1956) linguistic relativity hypothesis. Most debates on the relationship between language and thought have centered on the idea that the semantic structures of different languages are incommensurable, with the consequence that speakers of specific languages might think and act differently. Several early empirical tests of this hypothesis have been called into question, however, and it has been argued that natural languages may be too ambiguous and schematic to be functional as a mental code, and that we share a universal “mentalese” (Pinker, 1994). This position implies that language does not affect thought and that humans learn and evaluate verbal information through the use of similar mental processes across languages.
More recently, the Whorfian hypothesis has been reconceptualized in terms of how linguistic forms are represented, how they operate in the mind, and how they affect the concepts and categories that denote objects and relations in the world (Hunt & Agnoli, 1991). One of the basic premises of this idea is that “different languages pose different challenges for cognition and provide differential support for cognition” (Hunt & Agnoli, 1991, p. 387). Our ideas are compatible with those of Hunt and Agnoli (1991) who argue that language may not alter perception but that it can affect classification and linguistically based records of events. We similarly focus on the role of memory and the idea that “computational” ease can favor some thought processes over others; that relative differences in the temporal coding of information makes information differentially accessible from long-term memory and, therefore, differentially available for judgments.
Our arguments, however, do not concern linguistic coding or language per se but the writing systems used to represent language. Thus, whereas Hunt and Agnoli's (1991) analysis concerns the representational level of cognition, our analysis describes the mechanistic effects (Pylyshyn, 1984) that influence the manipulation of concepts without regard to their meaning. In other words, it is not the concepts' meanings but their accessibility from memory that we believe created linguistic differences in our setting.
In Experiment 1, we found that there was a primacy effect in recall from long-term memory and memory-based judgments that was stronger in English than in Chinese. Experiment 1 also found that retrieval from long-term memory mediated these cross-linguistic judgment differences. The online judgment data further support this interpretation by ruling out a potential confound, namely, that the effect was a by-product of language priming different cultural schemas that the bilingual consumers may have internalized. Experiment 1 also suggests that the script difference we found is limited to judgments that involve a substantial memory-based component and should not be generalized to online judgments. Experiment 2 cautions, however, that mnemonic processes can influence even online judgments. Experiment 2 demonstrated that the online evaluation of, preference of, and choice to receive, information about a new product were sensitive to the order in which competitive advertising information had been presented. This effect was stronger in English than it was in Chinese.
Our research suggests that copy presented early in an ad should be more likely to be “top of mind” in English than it should be in Chinese. This affects what information consumers recall in advertising tracking studies and the memory-based judgments they report and rely on to make purchasing decisions. From the advertiser's perspective, this implies that order is an important message strategy decision that is under copywriters' control to a greater degree in English than it is in Chinese. It also implies that global advertising campaigns may not be able to generalize copy-test results across alphabetic and logographic language groups.
The findings have important implications for advertising research where testing occurs both in online and in memory-based settings. In copy tests, attitudes and recall are typically assessed immediately after participants are exposed to an ad. In contrast, these measures are assessed after a substantial delay in field tests of advertising effectiveness. Interestingly, the order in which information is presented in an ad is currently not explicitly considered by several leading ad testing agencies that we surveyed. This is surprising because advertising tracking studies are highly concerned with developing instruments that are sensitive to detecting memory for an ad—for example, by providing ad cues such as photos, music, or headlines—and temporal order is one of the most potent memory retrieval cues (e.g.,Burgess & Hitch, 1992; Kahana, 1996; Knoedler et al., 1999; Murdock, 1960; Wright et al., 1985).
Our findings also have implications for thought processes in nonadvertising settings. Consider, for example, reading arguments concerning a political candidate in an Internet chat room. The order in which chat-room members discuss the pros and cons of this candidate should have a stronger effect on memory-based judgments concerning that candidate as well as subsequent online (hybrid) judgments of a competitor, in English than in Chinese. Similarly, consider a friend recommending a list of restaurants in an e-mail. Assuming that this list is read from top to bottom, restaurants that were on the top of the list should be more likely to be retrieved from memory and considered for choice at a later point in time in English than in Chinese, where the likelihood of being “top of mind” is not as sensitive to the order of presentation. Our findings may therefore have broad implications for everyday thought.
Finally, our findings also have implications for academic research and caution that knowledge based on alphabetic stimuli may not apply directly to logographic scripts. This is an implication that is not to be trivialized: Logographic scripts are used in Chinese, Japanese, and Korean, all of which are language groups that represent more consumers than English, German, Spanish, Russian, and Hindi combined. The large body of literature on the effects of temporal order of presentation in English, therefore, needs to be reevaluated for logographic writing systems.
To summarize, bilingual consumers do not appear to think the same way about the same conceptual information when this is represented in an alphabetic or logographic script. In interpreting these findings, we assume that psycholinguistic universals do exist and that processing Chinese and English draws on the same mental processes. However, reading alphabetic and logographic scripts taxes these mental processes to different degrees, making differential use of the same basic mechanisms for perceptual processing, encoding, rehearsal, and retrieval. Similar toWhorf's (1940/1956) original idea that language influences unconscious habitual thought rather than limiting thought potential, we propose that writing systems unconsciously affect everyday thought based on memory processes routinely invoked in the processing and retrieval of written language.
6. Baron, R. M. & Kenny, D. A. (1986). The moderator-mediator distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182.
7. Bargh, J. A. & Thein, R. D. (1985). Individual construct accessibility, person memory, and the recall-judgment link: The case of information overload. Journal of Personality and Social Psychology, 49, 1129-1146.
25. Lynch, J. G., Marmorstein, H. & Weigold, M. F. (1988). Choices from sets including remembered brands: Use of recalled attributes and prior overall evaluations. Journal of Consumer Research, 15, 169-184.
35. Semin, G. R. & Zwier, S. M. (1997). Social cognition. In J. W. Berry, M. H. Segall, & C. Kagitçibasi (Eds.), Handbook of cross-cultural psychology, Volume 3: Social behavior and applications (2nd ed., pp. 51–78). Boston: Allyn and Bacon.
36. Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. In S. Leinhart (Ed.), Sociological methodology 1982 (pp. 290–312). San Francisco: Jossey-Bass.
44. von Humboldt, W. (1988). On language: The diversity of human language-structure and its influence on the mental development of mankind (P. Heath, Trans.). England: Cambridge University Press. (Original work published 1836)
Received: May 19, 2004 Revised: September 2, 2004 Accepted: September 2, 2004
|Top of Page|