Individuals’ attitudes are often influenced by the cognitive feelings triggered by the dynamics of their mental operations during information processing (Schwarz, 2004, 2015; Avnet et al., 2012). One particularly potent cognitive feeling is processing fluency, that is, the subjective feeling of mental ease that people experience while processing information (Reber et al., 2004; Alter and Oppenheimer, 2009). Mental ease can be elicited by different aspects of information, such as repetition, perceptual clarity, or writing font (for a review, see Alter and Oppenheimer, 2009). A high compared to low processing fluency triggers positive feelings (Garcia-Marques and Mackie, 2000), higher liking judgments (Reber et al., 1998), increases the truth-value ascribed to trivia statements (Reber and Schwarz, 1999; Silva et al., 2016, 2017) and elicits facial muscular activity specific to positive experiences (Harmon-Jones and Allen, 2001; Winkielman and Cacioppo, 2001; Lee and Labroo, 2004; Labroo et al., 2008; Topolinski et al., 2009). Thus, fluency seems to be inherently positive and that aura of positivity leads to more favorable evaluations of fluent as compared to disfluent information (the hedonic marking of fluency, Winkielman et al., 2003).

Regarding the name of a seller, one key determinant of its fluency is its pronounceability. In psychological research, word pronounceability has been shown to affect consumer attitudes. For instance, Alter and Oppenheimer (2006) showed that the pronounceability of ticker codes of shares predicts short-term stock prices fluctuations both in the laboratory and with real-world stock market data, with stocks with easy-to-pronounce names outperforming stocks with difficult-to-pronounce names. Moreover, Song and Schwarz (2009) found that food additives are judged as more dangerous when they have difficult-to-pronounce rather than easy-to-pronounce names. Regarding person names more directly, Laham et al. (2012) showed that persons with easy-to-pronounce names are judged more positively and, according to a real world data analysis, they seem to occupy superior positions in companies’ hierarchies than persons with difficult-to-pronounce names. Also relevant to the present work, Newman et al. (2014) found that statements about different topics are given higher levels of truth when attributed to individuals with easy-to-pronounce names, and very recently Zürn and Topolinski (2017) showed that in economic trust games individuals trust more in partners bearing easy-to-pronounce names (for related effects of language articulation on attitudes, see Topolinski et al., 2014, 2015a,b; Topolinski and Boecker, 2016a,b).

Most recently Topolinski et al. (2016) have explored the impact of another, independent, contribution to processing fluency of nonsense words, namely the mere length of a word. They found that short anagrams were judged as more likely to be solved and as requiring less solvability effort than long anagrams. This demonstration, however, did not address attitudes but judgments of cognitive solvability of abstract intellectual problems.

The present research explored whether mere complexity of seller names would influence perceived trustworthiness of those sellers. We chose the context of online marketplaces, such as the auction website eBay (Roth and Ockenfels, 2002; Ockenfels and Roth, 2006), because name complexity most likely plays an important role in this domain, due to the following. Usually, one searches in eBay for a certain product and the search generates a list of all available offers. When clicking through the offers, the obvious parameters of price and seller reputation vary across offers, but also the names of the sellers. Imagine you come across the seller SIBU and the seller VLEGTIQCLAPL. Both usernames are fairly meaningless and could simply be the result of a contraction of the sellers’ given and family names. However, given all other trustworthiness parameters being constant, would you expect one of them to be more trustworthy than the other?

Building on the evidence cited above, we predicted that: (1) Online seller profiles with easy- as compared to difficult-to-pronounce usernames will be evaluated as more trustworthy, and (2) Online seller profiles with short as compared to long usernames will be evaluated as more trustworthy. We tested these hypotheses in a series of experiments in which participants were presented with screenshots of ostensibly real profiles of eBay private sellers, rating them for trustworthiness.

Crucially, we explored the impact of username complexity in the face of more diagnostic information, with seller reputation being the most important one. Previous studies showed that private eBay sellers with an established good reputation receive a premium of up to 8.1 percent of the average selling price of their products, have a higher probability that individual bidders enter their auctions and also receive more bids (e.g., McDonald and Slawson, 2002; Bajari and Hortacsu, 2003; Resnick et al., 2006). More relevant to our study, buyers’ perceptions of eBay sellers’ trustworthiness are also positively affected by their reputation (Ba and Pavlou, 2002). Therefore, we manipulated the sellers’ reputation independently from name complexity and predicted that: (3) Online seller profiles with better as compared to worse reputation will be evaluated as more trustworthy.

Given the strong positive effect that seller reputation seems to have on seller success, it is possible that the effect of username complexity on trustworthiness evaluations is rather weak. Thus, finding evidence of such an effect will be quite remarkable. It suggests that the subjective experience of fluency associated with username complexity can influence buyers’ judgments of seller trustworthiness despite the presence of strong objective information attesting or disconfirming it. However, we hypothesize the effect of username complexity on consumers’ attitudes to be associated with the bias to evaluate fluently processed stimuli more positively (e.g., Winkielman et al., 2003; Lee and Labroo, 2004; Labroo et al., 2008), exerting its effects independent of the more systematic, conscious and deliberative use of objective arguments such as a seller’s prior reputation (Maheswaran et al., 1992; Bohner et al., 1994; Smith and DeCoster, 2000; Suri and Monroe, 2003; Strack and Deutsch, 2004; Strack et al., 2006; Novak and Hoffman, 2009). Moreover, these two different types of influences seem to relate to two different and independent types of trust, namely emotional trust (elicited by positive feelings about the target person, product or brand) and cognitive trust (elicited by rational arguments about the person, product or brand characteristics) (Lewis and Weigert, 1985). Thus, we also predicted that: (4) Name complexity effects will not be substantially qualified by seller reputation, in that seller reputation will not alter or reverse the predicted pattern of username pronounceability and length effects on trustworthiness (the two factors should have additive rather than interactive effects).

We report all measures and conditions that were run in each Experiment. We report and justify exclusion of data (if any). We calculated required sample sizes a priori using G^∗Power (Faul et al., 2007). To estimate the required sample size in a conservative fashion, we used the smallest effect of pronounceability found in Topolinski et al. (2016, Experiment 5) for the basic effect of pronounceability on anagram solvability difficulty, which was ηp2 = 0.19. To replicate this effect with a power of 0.80, required sample size is N = 37. As the effect of pronounceability on perceptions of trustworthiness in consumer choices is unknown, we arbitrarily over-powered most of the present Experiments, with the smallest sample size being N = 38. Across all Experiments, analyses were run only after the full final sample size was collected.

Participation in all Experiments was voluntary. Participants gave their consent either when approached directly on the University campus or by clicking on the web link leading to the Experiments when invitations were disseminated online. Participants did not provide any identifying information and they could terminate their participation at any time. Experiments 8–10 were reviewed and approved by the Institutional Review Board at the University of Southern California (approval numbers: UP-15-00402 and UP-14-00630). Experiments 1–7 took place at the University of Cologne, Germany. These Experiments were reviewed and approved as part of the funding process by expert committees at the Department of Innovation, Science, and Research of the state of North Rhine-Westphalia (the German system does not require subsequent ethics review of individual studies and does not maintain institutional review boards for psychological experiments at the local university level).

All materials and databases are available online at https://osf.io/7mgaq.

The mean trustworthiness ratings for the six Pronounceability × Reputation conditions are shown in Figure 1 (across repetition levels). A 3 (Pronounceability: easy- vs. medium- vs. difficult-to-pronounce usernames) × 2 (Reputation: good vs. bad) × 2 (Repetition: first evaluation vs. second evaluation) repeated measures ANOVA (all factors within-participants) found the predicted main effects for Pronounceability, F(2,114) = 52.27, p < 0.001, ηp2 = 0.48, and for Reputation, F(1,57) = 94.13, p < 0.001, ηp2 = 0.62. There was also an interaction between the two factors, F(2,114) = 17.38, p < 0.001, ηp2 = 0.23. The main effect of repetition was not statistically significant, F(1,57) = 1.50, p = 0.227, as well as none of the interactions involving this factor (all Fs < 2, p > 0.300).

As Figure 1 shows, eBay seller profiles with easy-to-pronounce usernames (M = 5.18, SE = 0.23) were considered more trustworthy than profiles with medium-difficult-to-pronounce usernames (M = 4.55, SE = 0.24), t(57) = 5.58, p < 0.001, d_z = 0.73, 95% CI [0.40, 0.85], which in turn were considered more trustworthy than profiles with difficult-to-pronounce usernames (M = 3.97, SE = 0.26), t(57) = 7.71, p < 0.001, d_z = 1.01, 95% CI [0.43, 0.74]. The main effect associated with reputation shows that across all pronounceability conditions profiles with good reputation (M = 5.44, SE = 0.30) were considered more trustworthy than profiles with bad reputation (M = 3.69, SE = 0.19). The pronounceability × reputation interaction shows that the impact of pronounceability was more extreme in the case of good reputation sellers. But the differences between easy- and medium-to-pronounce, as well as between medium- and difficult-to-pronounce usernames were highly significant in each of the reputation levels (all ts > 4.00, ps < 0.001).

Previous research shows that gender influences trusting behavior in economic contexts such as the Investment Game or the Trust Game, namely that men seem to trust more than women and women seem to be more trustworthy than men (Croson and Buchan, 1999; Buchan et al., 2008). Thus, we tested whether gender differences would emerge in our experimental paradigm. For each participant, we calculated a difference score between the trustworthiness ratings given to the easy-to-pronounce usernames and the difficult-to-pronounce usernames, collapsing across all other factors. We ran an independent samples t-test on this difference score, entering Participant gender (male vs. female) as the between-participants factor. Results showed that men and women did not differ in their trustworthiness evaluations, t < 1.²

These results clearly suggest that username pronounceability has a strong impact on how trustworthy an eBay seller may seem to potential buyers. The effect sizes we found were much larger than those in previous studies on pronounceability (e.g., Topolinski et al., 2016). Moreover, this effect emerged in a context with clear and unambiguous objective information about the sellers’ reputation. The absence of a repetition effect, however, was quite surprising in light of extensive work on the mere exposure effect (see Bornstein, 1989). But recent work has clarified that mere exposure effects are strong when repetition is varied within-participants, but weak or absent when repetition is varied between-participants (Dechêne et al., 2009, 2010; see also Hansen et al., 2008). This presumably reflects that the fluency experience varies from stimulus to stimulus in within-participants designs, where some stimuli are repeated and others are not. In contrast, repeating all or none of the stimuli reduces the variation in participants’ fluency experiences, thus making the experiential signal less informative. From this perspective, a repetition effect was not obtained in Experiment 1 because repetition did not vary from stimulus to stimulus. In contrast, pronounceability did and so it was able to elicit fluency effects. Experiment 2 was designed to address this issue and at the same time replicate the pronounceability effect.

We analyzed only the trustworthiness ratings of the second block of stimuli (with repeated and new profiles) because the effect of repetition could be assessed only in this block. The conditional means are shown in Figure 2. A 3 (Pronounceability: easy vs. medium vs. difficult to pronounce usernames) × 2 (Reputation: good vs. bad) × 2 (Repetition: repeated vs. new profiles) ANOVA (all factors within-participants) revealed again the main effect for Pronounceability, F(2,114) = 75.41, p < 0.001, ηp2 = 0.57, the main effect for Reputation, F(1,57) = 88.25, p < 0.001, ηp2 = 0.61, and the same interaction between the two factors, F(2,114) = 27.77, p < 0.001, ηp2 = 0.33. Replicating Experiment 1, eBay sellers with easy-to-pronounce usernames (M = 5.50, SE = 0.23) were considered more trustworthy than sellers with medium-difficult-to-pronounce usernames (M = 4.51, SE = 0.24), t(57) = 9.01, p < 0.001, d_z = 1.18, 95% CI [0.77, 1.21], which in turn were considered more trustworthy than sellers with difficult-to-pronounce usernames (M = 3.95, SE = 0.26), t(57) = 6.56, p < 0.001, d_z = 0.86, 95% CI [0.39, 0.72]. Also as in Experiment 1, profiles with good reputation (M = 5.58, SE = 0.31) received higher ratings of trustworthiness than profiles with bad reputation (M = 3.73, SE = 0.19), across all pronounceability levels. The interaction between Pronounceability and Reputation factors also replicated the pattern found in Experiment 1, showing that the differences between the different levels of pronounceability were larger for good reputation sellers. Paired sample t-tests again demonstrated that the differences between easy- and medium-, as well as between medium- and difficult-to-pronounce usernames were highly significant in each of the two reputation levels and both for repeated and new profiles (all ts > 5.00, ps < 0.001). The main effect of repetition was now statistically significant, F(1,57) = 6.13, p = 0.016, ηp2 = 0.10. Repeated profiles (M = 4.70, SE = 0.24) were considered more trustworthy than new profiles (M = 4.61, SE = 0.23). None of the interactions involving Repetition was significant (all Fs < 1).

Results perfectly replicate Experiment 1 and show also an effect of repetition. Together, Experiments 1 and 2 suggest that fluency effects are more robust when fluency is manipulated within-participants and varies across stimuli (see Dechêne et al., 2009, 2010). Again, eBay profiles with easy-to-pronounce usernames were considered more trustworthy than profiles with difficult-to pronounce usernames, independently of their reputation and their familiarity. In Experiment 3 we explored the effect further, disentangling two independent parameters of name complexity, namely pronounceability and length.

The mean trustworthiness ratings for the Pronounceability × Length × Reputation conditions are shown in Figure 3. A 2 (Pronounceability: easy- vs. difficult-to-pronounce usernames) × 2 (Length: short vs. long) × 2 (Reputation: good vs. bad) repeated measures ANOVA (all factors within-participants) found a main effect for Pronounceability, F(1,47) = 18.16, p < 0.001, ηp2 = 0.28, a main effect for Length, F(1,47) = 17.04, p < 0.001, ηp2 = 0.27, and a main effect for Reputation, F(1,47) = 73.61, p < 0.001, ηp2 = 0.61. Sellers with easy-to-pronounce usernames (M = 4.92, SE = 0.20) were again considered more trustworthy than sellers with difficult-to-pronounce usernames (M = 4.51, SE = 0.24), as were sellers with good reputation (M = 5.54, SE = 0.27) as compared to sellers with bad reputation (M = 3.89, SE = 0.16). The effect of Length was also as predicted, with short-username profiles (M = 4.88, SE = 0.18) being considered more trustworthy than long-username profiles (M = 4.55, SE = 0.21). None of the interactions between the three independent variables reached statistical significance (all Fs < 2).

This Experiment shows that username pronounceability and length have strong, independent effects on the level of perceived trustworthiness attributed to online sellers. By disentangling the two dimensions and manipulating them orthogonally, the effect sizes were somewhat smaller than in the previous Experiments (except for the effect of Reputation), suggesting that indeed length might have contributed to the stronger effects previously observed for pronounceability. But note again that the two effects persisted despite the presence of the much more informative and seemingly powerful information of seller reputation.

After demonstrating the basic existence of the effects of username complexity (pronounceability and length) on trustworthiness, the remaining Experiments tested possible driving mechanisms.

The mean trustworthiness ratings for the Pronounceability × Length × Reputation conditions are shown in Figures 4–6, (pronounceability-correction, length-correction and no-correction, respectively). A 2 (Pronounceability: easy- vs. difficult-to-pronounce usernames) × 2 (Length: short vs. long usernames) × 2 (Reputation: good vs. bad) × 3 (Correction instructions: pronounceability-correction vs. length-correction vs. no-correction) mixed ANOVA (last factor between-participants) revealed once more strong main effects associated with Pronounceability, F(1,82) = 34.17, p < 0.001, ηp2 = 0.29, Length, F(1,82) = 31.41, p < 0.001, ηp2 = 0.28, and Reputation, F(1,82) = 192.49, p < 0.001, ηp2 = 0.70. These effects replicate what was found in the previous Experiments: participants attributed higher trustworthiness to sellers with easy- (M = 5.50, SE = 0.13) as compared to difficult-to-pronounce usernames (M = 5.05, SE = 0.15); to sellers with short- (M = 5.35, SE = 0.13) as compared to long-usernames (M = 4.15, SE = 0.14); and to sellers with a good (M = 6.14, SE = 0.18) as compared to bad (M = 4.36, SE = 0.11) reputation.

In addition, a significant interaction between Pronounceability and Length was found, F(1,82) = 5.70, p = 0.019, ηp2 = 0.07. The pattern of means associated with this interaction shows only that although the difference between easy- and difficult-to-pronounce usernames is always significant, it is larger for short (M_easy = 5.59, SE = 0.13, M_difficult = 5.11, SE = 0.15, t(82) = 5.62, p < 0.000, d_z = 0.61, 95% CI [1.86, 3.90]) than for long usernames (M_easy = 5.31, SE = 0.14, M_difficult = 4.98, SE = 0.15, t(82) = 4.96, p < 0.000, d_z = 0.54, 95% CI [1.17, 2.74]). This interaction was further qualified, albeit only marginally, by Reputation, F(1,82) = 3.77, p = 0.056, ηp2 = 0.04, in that only for sellers with good reputation was the Pronounceability × Length interaction significant [F(1,82) = 6.12, p = 0.015; interaction for bad reputation sellers, F < 1].

The independent effect of the Correction instruction was not significant (F < 1). Most importantly for this experiment, it also did not qualify the effects promoted by pronounceability, length and reputation (all Fs < 2). As Figures 4–6 show, the exact same pattern of effects was found in the three Correction instruction conditions. Instructing participants to not let their judgments be influenced by the pronounceability or length of the usernames did not eliminate the seemingly heuristic, irrational bias that these dimensions bear on trustworthiness evaluations.

This Experiment showed that username pronounceability and length influence evaluations of seller trustworthiness even when participants were explicitly asked to correct for them. These results suggest that name complexity effects on attitudes toward online sellers are not caused by a conscious and deliberative use of pronounceability or length as cues to infer trustworthiness. The route through which username complexity exerts its effects seems to be rather experiential (Novak and Hoffman, 2009), stemming from the positivity that is associated with fluent processing (e.g., Winkielman and Cacioppo, 2001; Winkielman et al., 2003; see also Topolinski and Strack, 2009a,b,c).

An alternative explanation is that participants maybe attributed the differences in username pronounceability and length to other characteristics of the sellers. If so, simply being warned that the usernames differed in how easy they were to pronounce or in their length could not eliminate the effects of other possible attributions participants might have made during their judgments. In the next three Experiments we addressed possible rational attribution processes, to further clarify the nature of name complexity influences on perceived trustworthiness.

The mean trustworthiness ratings for the Pronounceability × Length × Reputation × Seller origin conditions are shown in Figure 7. A 2 (Pronounceability: easy- vs. difficult-to-pronounce usernames) × 2 (Length: short vs. long usernames) × 2 (Reputation: good vs. bad) × 2 (Seller origin: German vs. Poles) repeated measures ANOVA (all factors within-participants) was performed on the averaged trustworthiness ratings. The analysis revealed the same three main effects associated with Pronounceability, F(1,74) = 29.14, p < 0.001, ηp2 = 0.28 (M_easy = 4.82, SE = 0.18; M_difficult = 4.34, SE = 0.20), Length, F(1,74) = 25.31, p < 0.001, ηp2 = 0.26 (M_short = 4.71, SE = 0.19; M_long = 4.45, SE = 0.19), and Reputation, F(1,74) = 69.95, p < 0.001, ηp2 = 0.49 (M_good = 5.25, SE = 0.25; M_bad = 3.91, SE = 0.14). The only significant effect associated with the Seller origin factor was a weak Length × Seller origin interaction, F(1,74) = 4.85, p = 0.031, ηp2 = 0.06. However, this interaction only shows that the statistically significant difference between short and long usernames was larger when sellers were from Poland (M_short = 4.76, SE = 0.19, M_long = 4.44, SE = .19, t(74) = 5.20, p < 0.000, d_z = 0.60, 95% CI [0.78, 1.75]) rather than from Germany (M_short = 4.66, SE = 0.20, M_long = 4.46, SE = 0.20, t(74) = 3.64, p < 0.000, d_z = 0.42, 95% CI [0.36, 1.23]). No other effects reached significance (all Fs < 2.5, all ps > .125).³

This Experiment ruled out the hypothesis that the effects of name complexity on trustworthiness ratings are due to participants attributing complex usernames to foreigners or outgroup members (Laham et al., 2012). Seller origin did not qualify the effects of pronounceability or length in any conceptually relevant way. Thus, the ingroup bias hypothesis seems to be discarded as an alternative explanation. In Experiments 6 and 7, we explored two other possible seller attributes that participants might infer from our manipulations of username pronounceability and length.

As in Experiment 5, the alternative explanation that username complexity affected trustworthiness ratings due to some strategic attribution was shown to be unlikely. We observed reliable and strong significant effects for both name pronounceability and name length independent of whether participants thought the usernames had been created by the sellers themselves or by an automatic algorithm. It is true that the pronounceability effects was even stronger when participants thought the sellers had created the usernames themselves, but it was still significant when participants thought the usernames had been created by an algorithm. Also, the other determinant of name complexity, length, did not interact with belief of username generation method. So, once more, a mechanistic assumption is bolstered in that the fluency experience caused by pronounceability and length seems to directly influence trustworthiness ratings in an experiential manner (Novak and Hoffman, 2009).

We found no moderation of the effects of name complexity by the Age of accounts factor. This suggests that the biasing effects of username pronounceability and length did not affect trustworthiness via attributions of less experience to sellers with more complex usernames.

The ratings given to the two target product descriptions were analyzed with a mixed design repeated-measures ANOVA with Seller (B vs. C) as a within-participants factor and Processing ease condition (Seller B easy-to-read/Seller C difficult-to-read vs. Seller B difficult-to-read/Seller C easy-to-read) as a between-participants factor. Mean ratings for the fluent and disfluent seller, the interaction terms of the ANOVA and the follow-up simple t-tests are presented in Table 1. Confirming the effectiveness of the fluency manipulation, results showed that participants rated the offer in the easy-to-read style as easier to process than the offer in the difficult-to-read style (see bottom row of Table 1).

Table 1 further shows that the seller with the easy-to-read product description had an advantage over the seller with the difficult-to-read description on each of the dependent variables. Most importantly, participants trusted seller B more than seller C when offer B was easier to read, but trusted seller C more than B when offer C was easier to read. Going beyond the measures used in the earlier Experiments, participants also reported a higher likelihood to purchase from the seller with the easier to read offer, considered the product information more reliable, and their credit card information in safer hands. They also liked the camera, as well as the webpage, more when the product offer was easy to read, consistent with earlier findings on the beneficial effects of processing fluency on esthetic appreciation (Reber et al., 2004).

Whereas the preceding Experiments manipulated ease of processing by varying the ease with which a seller’s username could be pronounced, Experiment 10 varied the ease of reading product descriptions. It found beneficial effects of processing fluency across a range of dependent variables. Consistent with the preceding Experiments, when the product description was easy to read, participants reported more trust in the seller, a higher likelihood of buying from the seller and fewer concerns about the safety of their credit card information than when the same product description was difficult to read. To our knowledge, this is the first study to document such pervasive effects of the ease of processing product information on the perceived trustworthiness of the seller. As observed in other domains (for a review, see Schwarz, 2015), people are likely to bring their metacognitive experiences to bear on any judgment to which they are applicable, giving the print font of a product description the potential to affect the trustworthiness of the seller (as observed in this Experiment) and the ease of pronouncing the seller’s username the potential to affect judgments of the product.

Our findings have important practical implications in many consumer-related contexts. The most obvious is of course connected with name composition and choice. In many everyday situations, the first information that is available about others is their name, be it in form of an email, when the employee of a customer care service identifies himself, or in the nametags worn during professional meetings. Globally, our results suggest that individuals should strive for simple, easy-to-pronounce names as a way to increase perceived trustworthiness, which in turn may contribute to positive outcomes. Similar to Laham et al. (2012) findings that individuals with easy-to-pronounce names are liked more and attain higher positions in their companies’ hierarchy, it is possible that consumers are more persuaded to try a new product in their local store when approached by a saleswoman with an easy-to-pronounce name depicted in her nametag.

The importance of choosing a fairly simple, short and easy-to-pronounce name might be even greater in e-commerce and online markets. Because in these contexts there are even fewer pieces of information both about transaction partners and the quality of the products (Degeratu et al., 2000), an easy-to-pronounce username might just boost perceptions of trustworthiness enough to enable more opportunities for successful transactions. An easy-to-pronounce name might thus work as a success-enabling tool. One way in which these effects could be assessed is by gathering real data from auction sites like eBay or Amazon.com to analyze the correlation between sale success, price premiums, and username pronounceability (as often done in research on seller reputation; Cabral and Hortaçsu, 2010). Future research should also take the form of field experiments and actively manipulate the complexity of seller usernames to address the causal relation between name complexity and sale success. Field experiments may also allow overcoming the fact that our participants did not face any real (financial) risk by rating one seller higher on trustworthiness than another one, and such risk may lead to different choices/outcomes in real commercial contexts. Future experiments could also explore the effect of name complexity in the presence of other variables that influence trustworthiness in consumer domains, such as the consistency between the valences of product ratings and of the text in the written reviews about the product (Tsang and Prendergast, 2009), or the popularity of the product being sold and level of consumer Internet experience (Zhu and Zhang, 2010). Related to consumer experience in online contexts, it may be interesting to explore individuals’ experience in online marketplaces and their knowledge of how the contact with a partner develops.

The relevance of name complexity in online environments should not be limited to the case of seller usernames. Name complexity can be relevant also in the domain of the products that are transacted. In fact, pronounceability effects in other contexts have also been found with product names. For example, stock market shares with easy-to-pronounce ticker codes outperform those with difficult-to-pronounce codes (Alter and Oppenheimer, 2006), food additives with easy-to-pronounce names are considered less dangerous than their counterparts (Song and Schwarz, 2009), and pharmaceutical drugs with fluent names are perceived as less hazardous, assumed to have less side effects and increase willingness to buy (Dohle and Siegrist, 2014; Dohle and Montoya, 2017).

Another aspect of online environments in which name complexity may play a relevant role are the online platforms themselves – the websites, mobile apps, search engines, among others (Corritore et al., 2003). Because accessing the online platforms that support the commercial or social transactions is the first step to actually engage in those interactions, the name that companies or entrepreneurs choose to give to their online sites needs to be extremely well crafted, otherwise individuals may not even register with the website or download the app (Keller et al., 1998; Yorkston and Menon, 2004; Pranata et al., 2012; Kuehnhausen and Frost, 2013).

There might be, however, important exceptions to the general conclusion of our Experiments. While for most instances of online environments the general take home message of the paper is make names short and easy, in some contexts disfluency may be desired. Previous research shows that creating contexts where processing fluency is low increases preference for special occasion products and services (Pocheptsova et al., 2010) and perceptions of product innovativeness (Cho and Schwarz, 2006). Thus, it is likely that the effects of name complexity that we observed are different in contexts related to online privacy and security, for example (Bleha et al., 1990; Miyazaki and Fernandez, 2001; Ni et al., 2003). Although our results suggest that individuals are judged as more trustworthy when the usernames they create are short and easy-to-pronounce, the recommendations for the creation of safe passwords go exactly in the opposite direction (Gehringer, 2002; Huth et al., 2012): make it long and difficult! Thus, online contexts that intend to create a feeling of security and safety may gain advantage when the agents that contact with users bear complex names.