That’s wrong! Right?
The effects of language errors in online and offline advertisements on brand
recall and recognition, text evaluation, author perceptions and persuasiveness
Master thesis
Student: C.A. (Claudia) Hop
Student name: Claudia Hop
Supervisor: B.C. Hendriks
Second reader: W.F.J. van Meurs
Date: 22 June 2019
Abstract
Building on studies concerning language errors, the present study aimed at uncovering the effects of different types of errors across media representations on brand recall, brand recognition, text evaluation, author perceptions and persuasiveness. This was done to examine whether marketeers could use language errors to break through the advertising clutter or whether language errors negatively affect the ad’s persuasiveness.
To date, studies concerning language errors were often limited to a single language error or texts containing different types of errors. These studies were missing comparisons between different types of errors or media. The current study added to research concerning the subject by comparing the effects of spelling and d/t-errors. Furthermore, online and offline advertising were compared to see whether effects differ across media representations. Participants (N = 207) evaluated two advertisements within one error condition, one on Facebook (online) and one in a magazine (offline).
It was found that overall salience of language errors was lacking as errors, especially d/t-errors, were not always noticed. Furthermore, results showed that, for both media, actual errors had no effect on the dependent variables, while perceived errors did. Participants who thought the ads contained errors negatively evaluated the ad’s attractiveness, the writer as well as the company. They also were less likely to purchase products than participants that did not notice any errors. Comprehensibility of the text was unaffected by error perception. Concerning brand recall and brand recognition, this study found no effects of error perception. Medium, however, did have an effect, but only on brand recognition. Participants were more likely to recognize the brand from an online ad than from an offline one. Thus, the findings would suggest that intentionally using language errors to attract attention may not be advisable since error perception negatively affects the overall ad including its author and persuasiveness.
Key words: advertising, L1 errors, perceived errors, text evaluation, author evaluation, persuasiveness, brand recall, brand recognition.
The persuasiveness of advertisements is essential for organisations to attract consumers to their brand. Advertisers put a great deal of time and effort into correct language use with the belief that errors would be detrimental to the goal of the text (i.e., persuasiveness; behavioural intentions) (Kloet, Renkema & van Wijk, 2003). At the same time, advertisements regularly contain language errors. Wyckham, Banting and Wensley (1984) found, for example, that 75% of the commercials that they examined contained at least one syntactic or stylistic irregularity. Furthermore, 64% of the offending commercials included more than one irregularity. More recent studies concerning the frequency of such errors are lacking but there are countless examples of advertisements containing language errors (e.g., Quest Braintainment, 2018; Santa Maria, 2018; Van der Zwet, 2018; Zijlstra Architecten, 2018) (see Appendix 1). For these examples, it is unknown whether the errors are intentional or unintentional. The purpose of the current study is twofold: this study aims to uncover whether (unintentional) language errors in ads negatively affect their persuasiveness or whether these errors could be strategically (i.e., intentionally) used by businesses. An example of an organisation intentionally using language errors is Starbucks, a company that is notorious for using spelling errors as part of their marketing strategy (Riepema, 2014; Wessels, 2014). In this strategy, employees purposely misspell consumers’ names on coffee cups. As a result, consumers share photos of the cups with their misspelled names online, creating free publicity for the business. Furthermore, this strategy contributes to social proof (Cialdini et al., 1999; Wessels, 2014) and elicits a mere exposure effect (Wessels, 2014; Zajonc, 1968).
Previous research has shown that advertisements regularly contain language errors (Wyckham et al., 1984). However, there has been limited research into the effects of language errors, especially in the context of advertising (Mozafari, El-Alayli, Kunemund & Fry, 2017). There has been extensive research on deviating use of language (e.g., a foreign language) showing positive effects in terms of attracting attention and increasing brand awareness (e.g., Domzal, Hunt & Kernan, 1995; Hornikx & van Meurs, 2015). Concerning the effects of language errors in advertisements, Mozafari et al. (2017) explored the field by examining whether language errors would reduce the effectiveness of a business advertisement. Having focused exclusively on traditional print advertisements, they indicated that it is important for future research to examine the effects of language errors across different forms of media presentation (e.g., to compare print ads with online ads). They also urged future research to make comparisons between different types of language errors. The current study will therefore add to research regarding the subject by examining the effects of different types of language errors in advertisements in both online and offline advertising.
Theoretical framework
Attract attention
Marketers go to great lengths in order to break through the advertising clutter and attract attention to their advertisement, hoping that consumers will be persuaded to buy their products (Urwin & Venter, 2014). One possible positive effect of language errors in advertisements could be the attention they might attract. This is supported by the Markedness Model, which states that something unexpected will attract attention (Krishna & Ahluwalia, 2008; Myers-Scotton, 1998). The Markedness Model explains that a language that is widely expected to be used is unmarked and that the message content in this language will be processed literally. An example of unmarked language usage is the use of the Dutch language in Dutch television advertisements, where the information given in the ad is processed literally. In contrast, a marked language deviates from expectations, draws attention and elicits language related associations. To illustrate, marked language usage would be that of French accents or language in Dutch television advertisements. Given that this is not a commonly used language in the Netherlands, this will draw attention and elicit associations that viewers have with France, such as sophistication in the case of perfume ads (Hornikx, van Meurs & Hof, 2013; Kelly-Holmes, 2005). Studies concerning the use of foreign languages in advertisements widely agree that this deviating use of language attracts the consumers’ attention (e.g., Domzal et al., 1995; Hornikx & van Meurs, 2015; Petrof, 1990).
Similarly, marketers have used shockvertising to use norm breaking stimuli in advertisements in an attempt to break through the advertising clutter (Dahl, Fankenberger & Manchanda, 2003). Shockvertising is a form of advertising that intentionally uses appeals to ‘startle and offend their audience’ by violating norms (e.g., obscenity or profanity). Dahl and colleagues found that advertisements containing shocking content increased attention, and positively affected memory as well as behavioural intentions. More recent research suggests that shockvertising is ‘not so shocking anymore’ (Urwin & Venter, 2014). Urwin and Venter (2014) examined the effectiveness of shockvertising on Generation Y in today’s society. The results of their study suggest that shockvertising has become obsolete in today’s society among Gen Y and that marketers need to find new ways to break through the clutter. Strategic use of language errors could be explored as a new mean to this end.
Brand awareness
Another possible positive effect of language errors in advertisements specifically for fast moving consumer goods (FMCG) (i.e., products that sell quickly at relatively low cost like groceries) might be brand recognition (Percy & Rossiter, 1992). Percy and Rossiter (1992) distinguish brand recognition and brand recall which are two types of brand awareness. Brand awareness could be described as the ‘buyer’s ability to identify a brand within a category in sufficient detail to make a purchase’ (Percy & Rossiter, 1992, p. 264). Sufficient detail does not necessarily mean identification of a brand name but could also be the packaging of a product that is recognized. The two types of brand awareness (i.e., brand recognition and brand recall) differ in which occurs first in the buyer’s mind: brand awareness or category need (i.e., the buyer’s perception of requiring something to fulfil a need, e.g., hunger) (Percy & Rossiter, 1992). With brand recognition, brand awareness precedes a category need in the customer’s mind. This means that recognition of a certain brand or product triggers a certain need, like the hunger someone might feel after seeing their favourite food in the supermarket. Typically, brand recognition is more easily achieved for FMCG because it is more natural in the way that consumers often make grocery lists describing products rather than brands (Percy & Rossiter, 1992). In the store, brand recognition might then function as a reminder of a need. Conversely, for brand recall to occur, a category need has to be experienced first, after which the customer thinks of a certain brand. An example of a business that tries to accomplish brand recall is Snickers with the slogan “Hungry? Grab a Snickers!”, attempting to link the category need of being hungry to their brand (Percy & Rossiter, 1992).
Previous studies have shown that both brand recall and brand recognition can be stimulated through a deviating use of language, such as a foreign language (Ahn & La Ferle, 2008; Domzal et al., 1995, Hornikx & van Meurs, 2015; Petrof, 1990). For the use of a foreign language, Petrof (1990) found that advertisements in which a foreign language was used (versus no foreign language display) drew more attention to them and increased overall advertisement recall (i.e., brand, colour of the ad and ad message). More specifically, Ahn and La Ferle (2008) found that in the Korean advertising context a brand name in English increased recall and recognition as opposed to a Korean brand name. However, they also found that for the body copy of the advertisement, Korean language elicited an increased recall and recognition as opposed to English. Ahn and La Ferle state that this difference is accounted for by the Revised Hierarchical Model (RHM) (Ahn & La Ferle, 2008; Kroll & Stewart, 1994) which explains how different languages are processed. Simply put, because the body copy contains important information that has to be directly understood, L1 is required as the link between L1 and
conceptual meaning is stronger than for L2. Conversely, information such as brand names of which the direct meaning is unimportant to readers, can be more easily remembered in L2 because it attracts their attention. In contrast, Domzal et al. (1995) argued that because information given to readers in a foreign language has to be processed on a deeper level in order to understand that information, this increases the likelihood that the ad is recalled. Moreover, Hornikx and van Meurs (2015) presented mixed effects for recall concerning ads containing foreign languages. In conclusion, the unexpected use of language (e.g., a foreign language) might lead to an increased recall, even though possibly limited to brand name and other ad elements that do not require to be deeply understood. The current study will try and explore whether unexpected use of language in the form of language errors might increase brand awareness in terms of brand recall and brand recognition.
Effects of language errors
Even though research concerning the effects of language errors in advertisements is scarce, research has consistently shown certain negative effects of language errors in general. Studies have often shown effects not only on (1) evaluation of the text (Appleman & Bolls, 2011; Figuerdo & Varnhagen, 2005; Kloet et al., 2003; Jansen, 2010; Jansen & de Roo, 2012; Mozafari et al., 2017; Planken, van Meurs & Maria, 2019; Raedts & Roozen, 2015; Schloneger, 2016), but also on (2) perceptions of the author (Brandenburg, 2015; Figuerdo & Varnhagen, 2005; Jansen, 2010; Jansen & de Roo, 2012; Jessmer & Anderson, 2001; Planken et al, 2019; Raedts & Roozen, 2015; Schloneger, 2016; Stiff, 2012), and (3) persuasiveness of the text (i.e., willingness to comply with the offer; behavioural intentions) (Jansen, 2010; Mozafari et al., 2017; Stiff, 2012). In the marketing field, Mozafari et al. (2017) were one of the first to examine language errors in print advertisements. They found that advertisements containing language errors significantly decreased perceived business/advertisement quality as well as employee quality. Furthermore, they found that participants had less interest in using the business when the advertisement was error-laden, but only when the ad was for a white-collar service (i.e., specialized work that typically requires higher education, e.g., computer memory upgrade). Language errors did not affect interest in using blue-collar services (i.e., manual work that can be performed by less educated individuals, e.g., automobile oil change). Concerning product recall advertisements, Raedts and Roozen (2015) showed that language errors negatively affected perceptions about the text and the associated business. Language errors did however not affect behavioural intentions. For direct mail promotions, Jansen (2010) found negative effects of language errors on text evaluation, perceptions of the writer as well as compliance
with the offer. Similarly, Kloet et al. (2003) examined language errors in direct mail soliciting funding and goodwill and found a negative effect on text evaluation, but only on comprehensibility; language errors did not affect the attractiveness of the text. They found no significant effects on the image of the sender or the persuasiveness of the mail.
The aforementioned effects of language errors in advertisement are in line with similar research concerning other text formats. Appleman and Bolls (2011) showed that grammatical errors in news stories negatively affected the comprehensibility, credibility and recall of the articles. For an English petition to make free downloading legal, Planken et al. (2019) found that language errors had negative effects on the evaluation of the text as well as the author for both native and non-native (German) speakers of English. Judges who thought the text contained language errors, evaluated the text as less attractive and the writer as less trustworthy, friendly and competent. They found no effect on persuasiveness of the text. In the field of government communication, Jansen and de Roo (2012) examined the effects of grammatical errors in municipal folders. They showed that these errors negatively influenced the perceived quality and attractiveness of the text as well as the writer and the municipality. Concerning internal communications, Brandenburg (2015) and Jessmer and Anderson (2001) found negative effects of language errors on perceptions about the sender. Brandenburg (2015) examined the effects of language errors in memos sent by an operations manager to employees and found that the presence of errors negatively influenced the writer’s ‘ethos’ (i.e., credibility; reputation). Jessmer and Anderson (2011) investigated internal e-mail messages concerning the personal use of printers and copiers and showed that ungrammatical messages had a negative effect on sender perceptions. Figuerdo and Varnhagen (2005) and Schloneger (2016) examined essays written by university students to understand the effects of spelling errors. Figuerdo and Varnhagen (2005) showed that such errors negatively affected perceptions about the writer’s abilities as well as the quality of the written product. Similarly, Schloneger (2016) manipulated the number of spelling errors in three conditions (containing 0, 5 or 10 errors) and found that as more spelling errors were found by participants, both the intelligence of the author and the quality of the text were perceived less favourably. A final study, Stiff (2012), examined feedback comments on e-commerce websites containing grammatical or spelling errors. In this study, errors negatively affected the reputation of the target business, intentional behaviours as well as favourability and trustworthiness of the writer. Interestingly, these effects only occurred when the examined feedback comments had a positive valence. When feedback comments were negative, the presence of errors had no influence.
Explanations for the aforementioned negative effects are provided by the Elaboration Likelihood Model (ELM) (Petty & Cacioppo, 1986) and the Language Expectancy Theory (Burgoon & Miller, 1985). According to Mozafari et al. (2017), it is likely that advertisements are processed at the peripheral level given the lack of processing time, attention or interest that people often experience while being exposed to print advertising. At the peripheral level, heuristic cues direct behaviour. A widely adopted example of this is the use of a white coat in dental commercials as a cue indicating expertise, positively influencing consumers’ attitudes and behaviours towards the advertised products or brands. Following this logic, language errors could serve as a peripheral cue, indicating a low level of competence and expertise, lowering perceived credibility. This may cause readers to reject the message while developing negative evaluations of the advertisement and the company behind it, interfering with the persuasiveness of the advertisement. The ELM may thus offer an explanation for the negative effects that language errors could have on evaluations of the text, as well as perceptions of the writer and the persuasiveness of the text. Regarding the possible negative evaluation of the writer, the Language Expectancy Theory may offer further support (Burgoon & Miller, 1985). This theory argues that readers develop expectations regarding the language a specific sender should use based on their estimation of the sender. Whether these expectations are confirmed or disconfirmed determines their evaluation of the writer. This means that when readers perceive a company as professional or an expert, they would expect no language errors. As such, language errors are in that case violations of the readers’ expectations, negatively affecting the company’s image.
The role of error perception
An important boundary condition that is revealed in research concerning language errors is error perception. Planken et al. (2019) for example show that actual errors do not matter but when judges perceived errors, it negatively affected their evaluations of text and author. This means that these negative effects only occur if readers correctly perceive errors in error-laden texts. Other studies support this finding (e.g., Mozafari et al., 2017; Raedts & Roozen, 2015). Furthermore, Raedts and Roozen (2015) showed that readers who incorrectly perceived errors in error-free advertisements also evaluated the text and business behind it less favourably. This suggests that error perception might function as a moderator for the negative evaluations that readers could develop when encountering language errors in a text. Brandenburg (2015) found that of the six manipulated errors, readers noticed only two on average, suggesting that readers do not always notice errors in texts. Similarly, Mozafari et al. (2017) also found a surprising
lack of salience of the language errors; many participants did not notice any of the seven errors in the error-laden condition. They suggest that consumers who do not consciously notice language errors are not influenced by them unconsciously.
Types of language errors
Studies concerning the effects of language errors have examined different types of language errors. Table 1 provides an overview of different types of language errors matched with the studies that have examined them.
Table 1. Overview types of language errors examined in current research on their effects.
Type of language error Studies
Grammatical errors Appleman & Bolls (2011); Jessmer & Anderson (2001); Mozafari et al., (2017); Planken et al., (2019); Raedts & Roozen (2015); Stiff (2012).
Spelling errors Figuerdo & Varnhagen (2005); Kloet et al. (2003); Jansen (2010); Mozafari et al., (2017); Planken et al., (2019); Raedts & Roozen (2015); Schloneger (2016); Stiff (2012).
Punctuation Jansen (2010); Mozafari et al., (2017); Planken et al., (2019). Vocabulary Planken et al., (2019).
Wrong conjunctions Kloet et al. (2003).
D/t error (Dutch) Kloet et al. (2003); Jansen & de Roo (2012); Raedts & Roozen (2015).
Table 1 shows that the most commonly examined errors are grammatical and spelling errors. Firstly, grammatical errors range from an incorrect pronoun, incorrect subject-verb agreement (Appleman & Bolls, 2011), to the use of a simple present instead of a progressive (Planken, van Meurs & Maria, 2019) and run-on sentences (Appleman & Bolls, 2011; Raedts & Roozen, 2015). Each study seems to operationalize grammatical errors differently, resulting in a range of different errors. In contrast, spelling errors could more easily be classified in a category that shows lack of knowledge (e.g., Figuerdo & Varnhagen, 2005) and a category that shows sloppiness (e.g., Schloneger, 2016; Stiff, 2012). As an example of the latter, Stiff (2012) manipulated spelling errors by replacing letters in words with another near it on the keyboard, resulting in errors such as ‘compyter’ instead of ‘computer’. Examples of spelling errors that show lack of knowledge rather than a typo being made are found in Figuerdo and Varnhagen
(2005). In this study, spelling errors are operationalized as homophone errors (i.e., using a similar sounding word with a different meaning, e.g., ‘their’ instead of ‘there’) and misspellings that are phonologically acceptable, meaning that the pronunciation stays correct, without the newly created word actually existing (e.g., ‘vyle’ instead of ‘vile’). Both of these categories show a certain lack of knowledge rather than a wrong letter being used on the keyboard. Table 1 also shows that some studies examine both grammatical and spelling errors. However, these studies either use the errors simultaneously in one condition (e.g., Mozafari et al., 2017; Planken, van Meurs & Maria, 2019) or separate them into distinct conditions but do not find differences among these conditions (e.g.; Raedts & Roozen, 2015; Stiff, 2012).
Noteworthy is the category ‘d/t errors’ concerning a typically Dutch language error. The reason that it is a separate category is because it is questionable whether it is a grammatical error given that it relates to verb conjugations or a spelling error, as classified by Kloet et al. (2003). The reason that this type of error is singled out and used in research is found in Jansen and de Roo (2012). They state to have opted for this error in their research because: (1) such errors are realistic; (2) they occur often, and (3) despite the high frequency of such errors, they are still perceived as severe. This would make this type of error of particular importance in research using Dutch texts.
Media representation
All of the aforementioned studies have not included different forms of media representation. Table 2 provides an overview of the stimulus material used in current research. This table shows that (1) no study is represented in both the offline and online category; (2) the offline category is overrepresented by current research compared to the online category; and finally that (3) the online category has a great focus on e-mail specifically, whereas stimuli in the offline category are more diverse in nature. Concerning advertising, Table 2 shows that it is only discussed in terms of traditional print advertising and product recall advertisements, which are both offline media. However, digital ad spending has already become more dominant in some countries, including the UK, China, Norway, Canada, the U.S. and the Netherlands, in which the current study was carried out (Enberg, 2019). As of 2019, digital ad spending accounts for 52.6% of total ad spending in the Netherlands. The largest online ad sellers in the world are Google, accounting for a revenue of $103.73 billion, and Facebook, with a revenue of $67.37 in terms of advertising. Facebook reported a revenue of $16.6 billion for the last quarter of 2018, which is a 30% increase year-over-year, indicating that social advertising is growing (Facebook, 2019). In that quarter, Facebook had 1.52 billion daily active users and 2.3 billion monthly
active users, indicating the enormous reach of the medium. Facebook thus is an important medium in advertising, as shown in these numbers. This study will therefore add to current research by making a comparison between offline and online advertising in the form of print advertising and Facebook advertising, respectively.
Table 2. Overview stimulus material in current research categorized in online and offline.
Offline Print advertising Mozafari et al. (2017)
Product recall advertisements Raedts & Roozen (2015)
Petition (L2) Planken et al. (2019)
Municipal brochure Jansen & de Roo (2012)
University essays Figuerdo & Varnhagen (2005);
Schloneger (2016)
Workplace memos Brandenburg (2015)
News stories Appleman & Bolls (2011)
Online Direct e-mail letters Jansen (2010)
Direct mail soliciting funding and goodwill Kloet et al. (2003)
E-mail message within organization Jessmer & Anderson (2001) Feedback comments on e-commerce websites Stiff (2012)
Research question
As stated previously, language errors frequently occur in advertisements. Nevertheless, research concerning the effects of such language errors is scarce, specifically in the marketing field. There have been studies that show possible effects of deviating language use in terms of attracting attention as well as enhancing brand awareness (i.e., brand recall and brand recognition). Conversely, there has been extensive research on language errors, and some studies provide support for negative effects that language errors have on the text, the author and the persuasiveness of the text. These studies often have a limited scope regarding the types of language errors used. It is not uncommon that studies only examine the effects of a single type of language error. Studies that do examine multiple error types often use the errors combined within one text or did not incorporate multiple forms of media representation. The present study will try to add to current research by examining multiple types of language errors across
different forms of media representation. This study will address the following research question:
To what extent do d/t and spelling errors in Dutch advertisements for fast moving consumer goods affect (1) brand recall, (2) brand recognition, (3) text evaluation, (4) author perceptions, and (5) the persuasiveness of the text across online and offline advertising?
Method
Material
The stimulus material consisted of advertisements for non-existing FMCG brands in different media forms representing both online and offline advertising. The choice to use imaginary brands was made to eliminate existing attitudes concerning brands that respondents were familiar with. Furthermore, this study used FMCG brands because of its relevance to a large public (i.e., all consumers who do groceries) and the plausible bigger effect of brand recognition for these products (Percy & Rossiter, 1992). The fictional FMCG brands used, were one for plant-based butter and one for biological coffee. These were chosen because they are both staple supermarket products relating to sustainability. This is an increasingly important theme to consumers (World Business Council for Sustainable Development, 2016) and therefore to brands that supply for this demand by creating more sustainable options (Unilever, n.d.). The media forms represented an online advertisement (Facebook) and a traditional print ad (magazine). Both of these advertisements were manipulated in three different conditions for both brands: error free, containing a d/t error, or containing a spelling error. This resulted in a total of 12 advertisements (see Appendix 4). Concerning the selected errors, d/t errors were chosen because they show a lack of knowledge and are generally perceived as severe despite their high occurrence (Jansen & de Roo, 2012). Spelling errors on the other hand were used to show sloppiness rather than a lack of knowledge. They were manipulated in line with the example from Santa Maria (2018); words misspelled were those that are misspelled in terms of double letters resulting in a similar word but with a different meaning (e.g., but instead of butt or bot instead of boot). The reason for this is that these types of misspellings would not be recognized by spell check and are therefore more easily overlooked, making them more realistic.
Pre-test. Pretesting was conducted to ensure that the errors used in the advertisements were noticed and that they were perceived equally in terms of severity, how often they are made and if the respondents would make the errors themselves. In the pre-test, a sample of Dutch respondents (N = 26) viewed seven sentences containing a d/t-error and seven sentences containing a spelling error, which were displayed in random order. For each sentence, they were first asked whether they noticed any errors (yes/no). If respondents answered yes, they were asked to type or copy-paste the error and what the correction would be. Next, they were asked to rate the errors on seven-point scales in terms of how severe they found them, how frequently they thought they were made and how likely it was for themselves to make them. If respondents answered that they did not notice any errors, they were immediately redirected towards the next sentence. Repeated measures analyses with Bonferroni corrections were used to select errors within each category that did not differ significantly in terms of severity, perceived frequency and likelihood in which the participant would make the error. Pairwise comparisons showed that d/t-errors ‘wordt’ instead of ‘word’ as imperative and ‘onderscheid’ instead of ‘onderscheidt’ as direct verb did not differ on the three measures (p = 1.000). For spelling errors, pairwise comparisons showed that the errors ‘voorruit’ instead of ‘vooruit’ and ‘verassen’ instead of ‘verrassen’ did not differ on the three measures (p = 1.000). These selected errors are highlighted in Table 3 and were used in the main experiment. Across categories (d/t and spelling errors), the four selected errors did not differ significantly on severity (p = 1.000, except for the difference between ‘wordt’ and ‘verassen’, in that case p = .081). In terms of frequency, only ‘voorruit’ differed significantly from both d/t errors (p = .003/p = .026). When indicating whether respondents would make the errors themselves, only the spelling error ‘voorruit’ differed significantly from ‘wordt’ as imperative (p = .023). Despite these differences across categories, ‘voorruit’ was still used in the main experiment because of its similarities with ‘verassen’. Table 4 shows the sentences used in the main experiment, based on the results of the pre-test.
Participants
The participants (N = 207) were L1 Dutch speakers. Initially, a total of 289 respondents started the experiment. The data of 82 participants were excluded from analysis because they did not complete the questionnaire (n = 78) or were not native speakers of Dutch (n = 4). The average age of the final sample (n = 206) was 29.9 (SD = 13.16) and ranged from 17 to 81 years old; one participant did not submit her age. In total 83 men (40.1%), 123 women (59.4%) and one
Table 3. Pretest results: means and standard deviations (between brackets) of indicated severity, perceived frequency and likelihood of making such errors measures on a 7-point scale.
Language error
Sentences as presented to respondents with correction between brackets.
n Severity Perceived
frequency
Could make the mistakes themselves
D/t-errors (N = 26)
1. Dit bedrijf bied (biedt) de beste prijs. 24 5.71 (1.30) 5.25 (1.29) 3.08 (2.06) 2. De woonkamer is versiert (versierd). 23 5.57 (1.38) 5.43 (1.24) 3.04 (1.97) 3. Ik leidt (leid) een team van acht. 24 5.83 (1.37) 5.04 (1.55) 2.33 (1.63) 4. Word (wordt) u ook altijd blij van lage
prijzen?
22 5.14 (1.25) 5.59 (1.10) 3.32 (1.67)
5. Wordt (word) verrast! 17 4.94 (1.39) 5.76 (.66) 2.94 (1.68)
6. Het aankoopbedrag word (wordt) teruggestort op uw rekening.
24 5.50 (1.35) 5.75 (.94) 2.71 (1.83)
7. De organisatie onderscheid (onderscheidt) zich door de sfeer.
23 5.09 (1.35) 5.52 (.90) 2.83 (1.80)
Spelling errors (N = 26)
1. Alleen de mand (maand) december, extra voordelig
26 5.65 (1.62) 2.85 (1.80) 2.12 (1.31)
2. Bijzonder, en toch hele (heel) gewoon. 26 5.58 (1.24) 3.38 (1.68) 2.12 (1.14) 3. Coca-cola, ten alle tijden (te allen tijde)
de beste keuze.
17 3.24 (1.20) 6.41 (.71) 5.29 (1.11)
4. Planaardige (plantaardige) boter, goed voor jou en het milieu.
25 5.12 (1.20) 3.20 (1.41) 2.24 (1.36)
5. Vanaf nu kijken we alleen nog maar voorruit (vooruit).
24 4.88 (1.68) 4.33 (1.55) 2.33 (1.17)
6. Laat u verassen (verrassen)! 22 5.59 (1.22) 5.27 (1.35) 2.59 (1.65) 7. Verkrijgbaar in alle kleren (kleuren) van
de regenboog.
other (0.5%) participated in the study. The great majority were highly educated: 50 participants (24.2%) had finished Higher Vocational Education, 68 (32.9%) held a Bachelor’s degree and 29 (14.0%) a Master’s degree. Of the remainder, 38 participants (18.4%) had finished secondary school and 22 (10.6%) had finished Intermediate Vocational Education. Of the participants included, 98.6% does their own groceries. The majority goes to the grocery store multiple times each week (58.9%). Of the remainder, 16.4% does groceries daily, 15.5% weekly, 8.2% less than weekly and 1.0% never.
Table 4. Sentences containing language errors as used in the stimulus.
Planta True Brew
Error-free Duurzaam vooruit, dat is hoe Planta zich onderscheidt.
Word verrast door de biologische koffie van True Brew.
D/t-error Duurzaam vooruit, dat is hoe Planta zich onderscheid.
Wordt verrast door de biologische koffie van True Brew.
Spelling error Duurzaam voorruit, dat is hoe Planta zich onderscheidt.
Wordt verast door de biologische koffie van True Brew.
Distribution of participants over the different conditions. Respondents were randomly assigned to one of six conditions (see Table 5). A Chi-square test showed no significant relationship between condition and gender (χ2 (10) = 9.43, p = .492), nor between condition and level of education (χ2 (20) = 16.25, p = .701). Furthermore, a one-way analysis of variance showed no significant difference in age across conditions (F (36, 170) = 1.29, p = .144). Participants thus were evenly distributed over the different conditions in terms of gender, level of education and age. Finally, it was checked if participants were equally distributed across conditions in terms of product usage for both brands (i.e., butter and coffee). A paired samples t-test showed a significant difference between the product usage of butter/margarine and that of coffee (t (206) = 2.53, p = .012) where more people purchase butter (M = 1.73, SD = .45) than coffee (M = 1.62, SD = .49). However, a Chi-square test showed no significant relationships between product usage and error condition (butter: χ2 (2) = .15, p = .927; coffee: χ2 (2) = .45, p = .798), nor between product usage and order in which advertising format was displayed (butter: χ2 (1) = .09, p = .767; coffee: χ2 (1) = .06, p = .811). This means that participants were evenly distributed over the different conditions in terms of their product usage of butter and coffee.
For this reason, product type will further be ignored in the results section and all variables will be reported including both brands and products.
Table 5. Distribution of participants over the different conditions.
Error-free D/t-error Spelling error Total
Facebook ad first n = 36 n = 31 n = 40 n = 107
Magazine ad first n = 37 n = 35 n = 28 n = 100
Total n = 73 n = 66 n = 68 N = 207
Design
The experiment had a 3 (text version: error free, d/t error, spelling error) x 2 (media form: Facebook advertisement, print ad) design: text version as between-subject and media form as within-subject factor (where order of advertisements was between-subject; i.e., online followed by offline or reverse). To maximize the number of responses to each message (Jessmer & Anderson, 2001) whilst ensuring the questionnaire to not be too long, it was chosen to combine between and within-subjects. The reason to opt for text version as between-subject factor was to ensure the question regarding error perception could be asked at the end, not influencing the participant during the experiment. Practically, this meant that each participant answered questions based on two advertisements (Facebook and print ad), both containing for example a spelling error. The text version was assigned randomly.
Instrumentation
The dependent variables that were included in this study are text evaluation, author perceptions, persuasiveness of the text and brand awareness. These variables were measured in an online questionnaire with seven-point semantic differentials, unless otherwise indicated. For each variable, these differentials were randomized so that the order in which they were displayed differed between respondents. Because all variables were measured after both advertisements and products, alpha is reported twice.
Text evaluation. Two factors of text evaluations were measured: comprehensibility (Figuerdo & Varnhagen, 2005; Jansen & de Roo, 2012; Kloet et al., 2003; Planken et al., 2019; Schloneger, 2016) and attractiveness (Figuerdo & Varnhagen, 2005; Jansen, 2010; Jansen & de Roo, 2012; Kloet et al., 2003; Planken et al., 2019; Raedts & Roozen, 2015). Comprehensibility was measured by two semantic differentials: easy/difficult and simple/complicated ( = .87/
= .83). Attractiveness was measured with four items: interesting/uninteresting, attractive, unattractive, nice to read/not nice to read, and creative/uncreative ( = .83/ = .86). Furthermore, participants were asked to mark the advertisement on a scale from one to ten (Jansen, 2010; Raedts & Roozen, 2015).
Author perceptions. Author perceptions were measured for the writer as well as for the company behind the ad. Writer perceptions were measured in terms of trustworthiness (Planken et al., 2019), friendliness (Jansen & de Roo, 2012; Jessmer & Anderson, 2001; Planken et al., 2019) and competence (Figuerdo & Varnhagen, 2005; Jansen & de Roo, 2012; Jessmer & Anderson, 2001; Planken et al., 2019). Trustworthiness was measured through three differentials: honest/dishonest, reliable/unreliable and sincere/unsincere ( = .87/ = .90). Three items measured friendliness: friendly/unfriendly, pleasant/irritating and likeable/unlikeable ( = .89/ = .91). Lastly, competence was measured with four items: competent/incompetent, intelligent/stupid, educated/uneducated and professional/ unprofessional ( = .91/ = .94). Company image was measured in terms of attractiveness (Kloet et al., 2003; Jansen, 2010), competence (Kloet et al., 2003; Jansen & de Roo, 2012; Raedts & Roozen, 2015) and trustworthiness (Kloet et al., 2003; Jansen, 2010; Jansen & de Roo, 2012; Raedts & Roozen, 2015). Three items measured attractiveness: attractive/unattractive, good/bad and sympathetic/unsympathetic ( = .82 / = .88). Competence was measured through two items: competent/incompetent and customer friendly/not customer friendly ( = .66 / = .72). Trustworthiness was measured with three items: reliable/unreliable, honest/dishonest and credible/incredible ( = .88 / = .89).
Persuasiveness of the text. Persuasiveness of the text was operationalized in terms of willingness to buy the advertised product. This was measured through three statements: ‘I would buy this product’ (likely/unlikely) (Jansen & de Roo, 2012; Kloet et al., 2003; Mozafari et al., 2017; Planken et al., 2019); ‘I find buying this product (wise/unwise)’ (Jansen & de Roo, 2012; Planken et al., 2019); and ‘This product (appeals to me/does not appeal to me)’ (Planken et al., 2019) ( = .87 / = .88).
Brand awareness. Brand awareness was measured in terms of brand recall and brand recognition using the same scales as in Ahn and La Ferle (2008) and Lerman and Garbarino (2002). Brand recall was measured with an open question: ‘write down the name of the brands that were advertised’. Brand recognition was measured through a multiple-choice question: ‘select the brand name that was advertised in the ad’. The MC question consisted of six brands: the correct one as well as five (similar) filler brands. Brand recall and recognition were
measured separately so participants could not change their answer after seeing the brand recognition options.
Error perception. Error perception was measured only at the end of the questionnaire. Firstly, participants were asked whether they came across any errors while reading the texts (yes/no) (Brandenburg, 2015; Planken et al., 2019; Raedts & Roozen, 2015). Participants indicating yes were asked what kind of error they noticed in a multiple-choice question (Raedts & Roozen, 2015). This MC question contained four types of language errors (i.e., spelling errors, incorrect capitalization, d/t-errors, interpunction errors) as well as the option ‘other, namely…’. Next, participants were asked to do a detection task where they viewed the ads again and were asked to click on the error(s) they saw (Kloet et al., 2003; Planken et al., 2019; Schloneger, 2016).
Language sensitivity. Language sensitivity was measured with five statements (completely disagree/completely agree): ‘Correct language usage is important to me’ (Jansen & de Roo, 2012); ‘I do not mind it when there are errors in a text’ (recoded) (Jansen & de Roo, 2012); ‘I think less of people who make language errors’ (Schloneger, 2016); ‘I never make language errors’ (Jansen & de Roo, 2012); and ‘I often rely on spellcheck to catch spelling errors’ (recoded) (Schloneger, 2016) ( = .69).
Background variables. The background variables included in the questionnaire were age, gender, education, nationality and native language. Furthermore, participants were asked some questions regarding their grocery shopping behaviour ‘Do you (sometimes) do your own groceries?’ (yes/no), ‘How often do you do groceries?’ (MC: every day, multiple times each week, once a week, less than once a week, never). Finally, relating to the advertisements in the experiments, participants were asked ‘Do you (sometimes) buy coffee?’ (yes/no) and ‘Do you (sometimes) buy butter/margarine?’ (yes/no).
Procedure
Data for this study was collected through an online questionnaire that was distributed between 30 May and 11 June 2019. The complete questionnaire is to be found in Appendix 2 and completing it took the participants an average of 9-10 minutes. Participants first saw an introduction to the experiment, thanking them for their participation and informing them about their anonymity and voluntariness with the right to stop the experiment at any given time. In the introduction, they were also briefly instructed that they would be shown two advertisements about which they were asked questions. If they agreed to take part in the experiment, they were instructed to view the ads as they would normally do. Participants were randomly assigned to
one of six conditions (see Table 5). They first saw the two advertisements and after each ad they were asked questions regarding Text evaluation, Author perceptions and Persuasiveness. Next, they were asked to answer the questions measuring Brand recall and Brand recognition. This was followed by Error perception, Language sensitivity and Background information. Afterwards, participants were again thanked and if they were looking for respondents themselves, they could leave a link on this page. In total, fourteen participants left a link of their study.
Statistical treatment
A MANOVA was conducted to examine the effects of error condition, error perception and order of advertisement on text evaluation, author perceptions and persuasiveness. Chi-square tests were carried out to test differences in error perception across text versions (Planken et al., 2019), to test differences in gender, education and product usage across conditions. Furthermore, Chi-square tests were also used to test differences in brand recall and brand recognition across text versions (Ahn & La Ferle, 2008). McNemar’s tests were performed to test the effect of medium (i.e., online vs. offline) on brand awareness and brand recall. The effect of medium on the dependent variables was measured through repeated measures analyses. Because these repeated measures analyses had to be performed separately for each variable, the choice was made to exclude medium from the MANOVA. Finally, a linear regression was conducted to test language sensitivity as a predictor of persuasiveness, the goal of advertising.
Results
In the results section, the different effects of language errors and medium type will be explored. These results will be given in the order in which they were discussed in the theoretical framework. First, results will be given concerning the attention language errors attract, i.e., the extent to which participants found errors in the different text versions. Next, effects on brand awareness will be given, in terms of brand recall and brand recognition. Third, the effects of actual and perceived errors will be shown on text evaluation, author perceptions and persuasiveness. Finally, results concerning the influence of language sensitivity will be discussed.
Error perception
After having seen and evaluated both advertisements, participants were asked whether they noticed any errors or not. Table 6 shows the number of participants that indicated noticing errors or not for each text version.
Table 6. Error perception as a function of text version.
Text version Error-free D/t-errors Spelling errors
Total
Yes Count 11a 19a, b 31b 61
% 15.1% 28.8% 45.6% 29.5%
No Count 62a 47a, b 37b 146
% 84.9% 71.2% 54.4% 70.5%
Total Count 73 66 67 207
% 100% 100% 100% 100%
*Each subscript letter denotes a subset of categories that do not differ significantly from each other at the .05 level.
A significant relationship was found between text version and error perception (χ2 (2) = 15.80, p < .001). Significantly more participants (45.6%) detected errors in the text containing spelling errors than in the text version without errors (15.1%). Surprisingly, there was no significant difference in participants’ detection of errors in the d/t-error condition compared to the text version without errors. These results indicate that the manipulation was successful, but only for spelling errors, contrasting with the pre-test results.
With respect to the text version containing d/t-errors, the vast majority (73.7%) of the participants who noted errors correctly identified the errors as d/t-errors. The remainder either categorized the error as incorrect capitalization (15.8%) or as spelling error (10.5%). Consistently, in the text version with spelling errors, most participants (80.6%) who noticed errors correctly identified them as spelling errors. However, there were also participants that categorized them as d/t-errors (6.5%), as incorrect capitalization (3.2%), as an interpunction error (3.2%) or as other (6.5%). These results suggest that the manipulation of both error types was successful. For the text version without errors, 11 participants (15.1%) indicated that they noticed an error in the ad. These errors were categorized as d/t-errors (45.5%), spelling errors (18.2%), incorrect capitalization (9.1%), and other (27.3%). Error perception in the following
analyses will include all participants who perceived errors in the ads. More specifically, it will include: (1) participants who correctly perceived and identified the errors in the error-laden ads; (2) those who correctly perceived but wrongly identified the errors in the error-laden ads; as well as (3) participants who incorrectly perceived errors in the error-free ads.
Brand awareness
In the theoretical background, it was discussed that language errors like foreign language use might make a brand stand out more. This was measured in terms of brand recall and brand recognition. Participants were first asked to note down the brands they saw (brand recall) and then to select the brand name they saw from in a multiple-choice question (brand recognition).
Brand recall. A Chi-square test showed no significant relationship between error condition and brand recall (offline: χ2 (2) = 1.13, p = .568; online: χ2 (2) = 3.79, p = .150). Similarly, there was no significant relationship between error perception and brand recall (offline: χ2 (1) = .80, p = .370; online: χ2 (1) = .66, p = .415). A McNemar test determined that there was no significant difference in brand recall between the online and offline ad (p = .556). Brand recognition. A Chi-square test showed no significant relationship between error condition and brand recognition (offline: χ2 (2) = 2.72, p = .256; online: χ2 (2) = 2.04, p = .361). Similarly, there was no significant relationship between error perception and brand recognition (offline: χ2 (1) = 2.14, p = .144; online: χ2 (1) = .14, p = .706). A McNemar test showed that brand recognition was influenced by medium (i.e., online/offline) of the ad (p = .020). Brand recognition was significantly higher for the online advertisements (M = .89, SD = .32) than for the offline advertisements (M = .81, SD = .40).
Text evaluation, author perceptions and persuasiveness
This section will explore the results concerning the effects of the independent variables (i.e., error condition and order of advertisements) and error perception on text evaluation, author perceptions and persuasiveness. Regarding the independent variables, a multivariate analysis for text evaluation, author perceptions and persuasiveness, with error condition, error perception and order of advertisements as factors, found a significant main effect of error perception (F (20, 176) = 4.55, p < .001, 2 = .34) and of order of advertisements (F (20, 176) = 2.15, p = .005, 2 = .20). There was no significant main effect of error condition on the dependent variables (F (40, 352) = 1.12, p = .291). Furthermore, there were no interaction effects found between the factors included in analysis. There was no interaction effect between error perception and order of advertisements (F (20, 176) < 1), between error perception and
error condition (F (40, 352) = 1.24, p = .155), nor between error condition and order of advertisements (F (40, 352) = 1.17, p = .225). The three-way interaction between error perception, error condition and order of advertisements was also found to be insignificant (F (40, 352) < 1).
Concerning the effects of error perception on text evaluation, it was found that participants who noticed errors in the ad, regardless of the error condition, evaluated the ads as significantly less attractive (offline: M = 3.66, SD = 1.29; online: M = 3.45, SD = 1.08, see Table 7) than participants who did not notice any errors (offline: M = 4.16, SD = 1.14; online: M = 4.02, SD = 1.04) (offline: F (1, 195) = 4.59, p = .033, 2 = .02; online: F (1, 195) = 9.21, p = .003, 2 = .045). Error perception also negatively affected the quality of the ad (offline: F (1, 195) = 4.39, p = .038, 2 = .02; online: F (1, 195) = 14.16, p < .001, 2 = .07). Participants who noticed errors, gave the ad a lower grade (offline: M = 5.48, SD = 1.72; online: M = 5.21, SD = 1.54) than those who did not notice any errors (offline: M = 6.08, SD = 1.46; online: M = 6.05, SD = 1.29). Comprehensibility of the ad, however, was not influenced by error perception (offline and online: F (1, 195) < 1).
Next, author perceptions were measured for the writer as well as the company behind the advertisement. Regarding writer evaluations, the perception of errors negatively influenced the writer’s trustworthiness (offline: F (1, 195) = 5.50, p = .020, 2 = .03; online: F (1, 195) = 4.10, p = .044, 2 = .02). Participants who indicated that they perceived errors in the text evaluated the writer as less trustworthy (offline: M = 4.19, SD = 1.19; online: M = 4.11, SD = .94) than those who did not (offline: M = 4.54, SD = .86; online: M = 4.40, SD = .88). Similarly, participants who noticed errors thought of the writer as less friendly (offline: M = 4.27, SD = 1.13; online: M = 4.24, SD = .87) compared to participants who did not notice any errors (offline: M = 4.58, SD = .88; online: M = 4.58, SD = .89) (offline: F (1, 195) = 5.17, p = .024, 2 = .03; online: F (1, 195) = 7.22, p = .008, 2 = .04). Lastly, when participants noticed errors, they evaluated the writer to be less competent (offline: M = 3.55, SD = 1.51; online: M = 3.37, SD = 1.33) than when they did not notice any errors (offline: M = 4.52, SD = 1.05; online: M = 4.52, SD = .90) (offline: F (1, 195) = 17.65, p < .001, 2 = .08; online: F (1, 195) = 46.10, p < .001, 2 = .19).
Regarding company image, participants who noticed errors in the ads perceived the company behind them as significantly less trustworthy (offline: M = 4.07, SD = 1.39; online: M = 3.96, SD = 1.02) than those who did not notice any errors (offline: M = 4.72, SD = 1.02; online: M = 4.58, SD = .93) (offline: F (1, 195) = 12.44, p = .001, 2 = .06; online: F (1, 195)
= 23.53, p < .001, 2 = .11). Similarly, when errors were noticed, it led participants to view the company as less competent (offline: M = 4.10, SD = 1.26; online: M = 3.92, SD = 1.01) than when they were not (offline: M = 4.58, SD = .95; online: M = 4.54, SD = .79) (offline: F (1, 195) = 5.37, p = .022, 2 = .03; online: F (1, 195) = 25.81, p < .001, 2 = .12). Surprisingly, error perception only negatively affected the attractiveness of the company for the online ad (F (1, 195) = 7.25, p = .008, 2 = .04); participants who noticed errors found the company behind the online ad less attractive (M = 4.07, SD = 1.09) than participants who did not find any errors (M = 4.45, SD = .96). For the offline ad, the company’s attractiveness was unaffected by error perception (F (1, 195) = 3.57, p = .060).
Moreover, participants who indicated that they noticed errors in the advertisements, were significantly less likely to buy the advertised product (offline: M = 3.38, SD = 1.42; online: M = 3.10, SD = 1.29) than those who did not (offline: M = 4.07, SD = 1.40; online: M = 3.77, SD = 1.38) (offline: F (1, 195) = 9.37, p .003, 2 = .05; online: F (1, 195) = 10.98, p = .001, 2 = .05).
Besides the main effect found of error perception, the multivariate analysis for text evaluation, author perceptions and persuasiveness, with error condition, error perception and order of advertisements as factors also found a significant main effect of order of advertisement (F (20, 176) = 2.15, p = .005, 2 = .20). The univariate analyses showed an effect of order of advertisement on the attractiveness of the online ad (F (1, 195) = 4.06, p = .045, 2 = .02) as well as the persuasiveness of the online ad (F (1, 195) = 3.92, p = .049, 2 = .02). Participants evaluated the online ad as more attractive (M = 4.05, SD = 1.05) and more persuasive (M = 3.78, SD = 1.44) when it was shown to them after the offline ad than when it was shown to them first (attractiveness text: M = 3.67, SD = 1.08; persuasiveness: M = 3.38, SD = 1.30).
Table 7. Text evaluation, author perceptions and persuasiveness as a function of error perception and medium
Error
Perception Offline Online
M SD M SD
Text Comprehensibility Yes 5.17 1.23 5.20 1.22
No 5.36 1.32 5.32 1.21
*p < .050, **p < .010
Error
Perception Offline Online
M SD M SD
Text Attractiveness Yes 3.66** 1.29 3.45** 1.08
No 4.16** 1.14 4.02** 1.04
Total 4.01 1.20 3.85 1.08
Quality Yes 5.48* 1.72 5.21** 1.54
No 6.08* 1.46 6.05** 1.29
Total 5.90 1.56 5.81 1.41
Writer Trustworthiness Yes 4.19* 1.19 4.11* .94
No 4.54* .86 4.40* .88 Total 4.44 .97 4.32 .91 Friendliness Yes 4.27* 1.13 4.24* .87 No 4.58* .88 4.58* .89 Total 4.49 .97 4.48 .90 Competence Yes 3.55** 1.51 3.37** 1.33 No 4.52** 1.05 4.52** .90 Total 4.23 1.28 4.18 1.17
Company Attractiveness Yes 4.22 1.25 4.07* 1.09
No 4.63 1.00 4.45* .96 Total 4.51 1.09 4.34 1.01 Trustworthiness Yes 4.07** 1.39 3.96** 1.02 No 4.72** 1.02 4.58** .93 Total 4.53 1.17 4.40 1.00 Competence Yes 4.10** 1.26 3.92** 1.01 No 4.58** .95 4.54** .79 Total 4.44 1.07 4.36 .91 Persuasiveness Yes 3.38** 1.42 3.10** 1.29 No 4.07** 1.40 3.77** 1.38 Total 3.87 1.44 3.57 1.38
Online versus offline advertisements
To test the effect of medium, repeated measured analyses were used with medium as within subject factor and error condition as between subject variable. As Table 8 shows, there are two dependent variables which are influenced by medium. First, medium affected company attractiveness, where participants evaluated the company more favorably after the offline ad (M = 4.51, SD = 1.09) than after the online ad (M = 4.34, SD = 1.01) (F (1, 204) = 3.91, p = .049). Secondly, the online and offline advertisements differed in terms of persuasiveness (F (1, 204) = 5.87, p = .016). Participants indicated being more likely to buy the advertised product after the offline ads (M = 3.87, SD = 1.44) than after the online advertisements (M = 3.57, SD = 1.38).
Table 8. Reported t-tests testing differences as a function of medium (offline versus online).
F p Text Comprehensibility < 1 Attractiveness 2.56 .111 Quality < 1 Writer Trustworthiness 3.25 .074 Friendliness < 1 Competence < 1 Company Attractiveness 3.91* .049 Trustworthiness 2.40 .123 Competence < 1 Persuasiveness 5.87* .016 *p < .050 Language sensitivity
Language sensitivity was measured with statements concerning the importance of correct language and the participants’ self-reported ability to use correct language. A regression analysis showed that language sensitivity did not explain the variance in persuasiveness (offline: F (1, 205) = 2.28, p = .132; online: F (1, 205) = 3.39, p = .067).
Conclusion
The aim of this study was twofold: this study wanted to uncover whether language errors could be used strategically (i.e., intentionally) or whether (unintentional) language errors in ads negatively affect their persuasiveness.
It was explained that intentionally using language errors in advertising might attract attention and increase brand awareness, similar to studies concerning foreign language use in advertisements. Results show that participants noticed more errors in the text version containing spelling errors than in the error-free text. However, participants did not notice more errors in the d/t-error condition compared to the error-free one. This means that d/t-errors are less noticeable than spelling errors. In terms of attracting attention, these results indicate that it might be inadvisable for companies to use language errors intentionally to break through the advertising clutter. Furthermore, results have shown no effects of language errors on brand awareness in terms of both brand recall and brand recognition. It seems that intentionally using language errors thus has no positive effects for businesses.
With regards to the effects of (unintentional) errors on the ad’s persuasiveness, it has been shown that actual errors did not affect the dependent variables, while perceived errors did. Participants that reported that they noticed errors in the advertisements evaluated the ad less favorably in terms of attractiveness and quality than participants who did not notice any errors, while error perception had no influence on the comprehensibility of the text. Regarding the author perceptions, error perception negatively affected evaluations concerning the writers’ friendliness, trustworthiness and competence as well as the company’s attractiveness, trustworthiness and competence. Furthermore, participants that noticed errors were less likely to buy the advertised products than those who did not. In conclusion, the actual presence of errors does not seem to matter, while the readers’ perception of errors does.
Besides the effects of language errors, the research question included effects of different advertising formats, in this study a Facebook ad (online) and a print ad (offline). Results show that medium affected persuasiveness as well as brand recognition. Participants were more likely to buy the advertised products based on offline ads compared to online ads. Conversely, participants recognized the advertised brand better from online ads than offline ads. Medium did not influence the other dependent variables. Furthermore, the order in which the different advertising formats were displayed to the participant affected evaluations of the online advertisement in terms of attractiveness and persuasiveness. The online ad was found to be more attractive and persuasive if the online ad was preceded by an offline ad.
Lastly, results show that these effects occur for all participants, regardless of the importance they attach to the use of correct language. This means that the perception of errors leads to less favorable evaluations, regardless of someone’s language sensitivity.
Discussion
The current study showed that errors matter only when they are perceived by the public in terms of their effects on text evaluation, author perceptions and persuasiveness. Even though the lack of effects of actual errors contrasts with certain studies (e.g., Brandenburg, 2015; Figuerdo & Varnhagen, 2005; Kloet et al., 2003), the important role found of error perception is in line with more recent research concerning the effects of language errors (Mozafari et al., 2017; Planken et al., 2019; Raedts & Roozen, 2015). It thus seems that recent studies have found that error perception serves as a boundary condition for errors to affect evaluations, which the current study supported. Even though actual errors do not influence the public’s evaluations, the effects of perceived errors imply that companies should put an effort into making sure their communications are at least perceived as error-free.
An interesting finding concerning error perception was that readers do not always notice errors in texts, which is in line with other studies such as Brandenburg (2015) and Mozafari et al. (2017). Results of the current study have shown that especially d/t-errors are not noticed often. This is in contrast with the pre-test, in which d/t-errors were often correctly noticed. There is a good methodological explanation for this difference. In the pre-test, participants were given a sentence with beneath it the question whether or not they noticed any errors. This might have led participants to look at the sentence more critically, looking for errors. In the main experiment, however, the participants were first shown two advertisements to evaluate and were asked only at the end whether or not they noticed any errors. This might have led participants to not look at the text in the advertisements as critically as they would when immediately asked if they noticed any errors. Furthermore, the advertisements provide a more natural context for the text than a sentence on itself, which might make the d/t-error error stand out less. While spelling errors immediately stand out, d/t-errors are frequently made errors as they require more thought (Jansen & de Roo, 2012). Another possible explanation for the lack of salience of language errors, especially d/t-errors, include not reading the texts carefully (cf. Jansen & de Roo, 2012; Raedts & Roozen, 2015). Moreover, research has shown that language errors stand out less in an electronic version than on paper (Wharton-Michael, 2008, as cited in Raedts & Roozen, 2015). This could mean that normally, people might perceive errors in offline
advertisements more easily than those in online ads, implying that not proofreading Facebook posts could be less harmful to the company, compared to magazine ads.
The overall negative effects of perceived errors mostly coincide with effects found in current literature as discussed in the theoretical framework. These effects include text evaluation (i.e., attractiveness and quality), writer perceptions (friendliness, trustworthiness and competence), company perceptions (attractiveness, trustworthiness and competence) as well as persuasiveness (e.g., Brandenburg, 2015; Figuerdo & Varnhagen, 2005; Jansen, 2010; Jansen & de Roo, 2012; Mozafari et al., 2017; Planken et al., 2019; Raedts & Roozen, 2015). Concerning the negative effect of errors on persuasiveness, studies have reported different results. Similarly to Jansen (2010), the current study showed a negative effect of errors on persuasiveness. Mozafari et al. (2017) only found a negative effect on persuasiveness for a white-collar service (i.e., computer update) and not for a blue-collar service (i.e., car oil change). Similarly, in examining errors in feedback comments on e-commerce websites, Stiff (2012) found that errors only negatively affect intentional behaviors for positive feedback comments, whereas errors had no influence on persuasiveness for positive comments. Moreover, other studies have not found an effect on persuasiveness, even though these studies did find negative effects on evaluations regarding text, writer and/or company (Kloet et al., 2003; Planken et al., 2019; Raedts & Roozen, 2015). The only effect not found in the current study was that of perceived errors on the comprehensibility of the text. This finding is in line with the results of Jansen and de Roo (2012) and Planken et al. (2019). It is, however, noteworthy that research is not consistent in this, as other studies find comprehensibility to be affected by language errors (e.g., Appleman & Bolls, 2011; Kloet et al., 2003). In the current study, each ad contained a language error created by a single supplement or missing letter. In contrast, both studies (Appleman & Bolls, 2011; Kloet et al., 2003) use multiple errors (10 and 5/10, respectively) in a single text. Moreover, these errors were often bigger than a single letter including wrong vocabulary or run-on-sentences. The difference in frequency and size of the error between the aforementioned studies and the current one might account for the different findings of effects on comprehensibility.
Besides the language errors, this study also tried to compare different types of media with each other. Current research focused on either offline (e.g., Mozafari et al., 2017; Planken et al., 2019; Schloneger, 2016) or online media (e.g., Jansen, 2010; Stiff, 2012), without comparing the two. The current study adds to current research by including a comparison between online (i.e., Facebook) and offline (i.e., magazine) advertising. This study found that brands are better recognized after exposure in a Facebook advertisement compared to one in a
