Attention to multilingual job ads: an eye-tracking study on the use of English in German job ads

1.1 Background

The discussion of the use of English loanwords or anglicisms appears at regular intervals in the German media (cf. Hunt 2019). This discussion is fostered by organisations such as the Verein Deutsche Sprache (German Language Association) that promotes language purity (see Verein Deutsche Sprache 2021). The use of English loanwords or anglicisms in German has also inspired a multitude of scientific studies.^[1] There are studies that deal with the fundamental problem of defining what anglicisms are (e.g. Carstensen 1992; Galinsky 1977), studies that look at the way English loanwords are used as for example Eisenberg (2018). He provides a detailed analysis of the (degree) of integration of English loanwords (but also that of other languages) into the German language in terms of their phonology, inflection, word formation and orthography. Another example is the study by Onysko (2007), who investigated the occurrence of English loanwords and discussed their lexical productivity while also taking a closer look at German-English code-switching in his corpus. Another area of interest is the (stylistic) function of English loanwords (Galinsky 1977). Some studies take a more global approach by looking at similarities in the use of (English) loanwords across different languages in order to establish linguistic patterns of use and identify so-called interlexemes, words that are similar in form and meaning and appear in several languages (see Braun et al. 1990, 2003).

Most of these studies are based on corpora of written language, mainly corpora of media texts (see Onysko 2007), but there are also a number of studies that focus on specific text genres, e.g. Fink (1997) and Kupper (2007) looked at the use of English loanwords in advertising. Few studies, however, looked at the readers’ perception of the use of English loanwords in German and the effect this use might have on either comprehension or effectiveness of the text. These studies are restricted to surveys in which participants have to judge the use of English loans in general (Schmidlin 2008), or to isolated words and phrases without further context (Hofmann 2002; Stickel and Volz 1999). To our knowledge, there are no studies on the use of English loanwords that tests the effect of these words experimentally. For the present paper, a text genre was chosen that allowed for testing such an effect, namely job advertisements.

1.2 English loanwords in job advertisements

A number of studies on languages other than German (Dogancay-Aktuna 1998 (for Turkish); Hurme 2013 (for Finnish); Van Hooft 2015; Van Meurs 2010; Van Meurs et al. 2006 (both for Dutch); Zenner et al. 2013 (for Flemish) have shown that English is frequently used in job advertisements in non-English-speaking countries. The use of English can be found on a macro-level (the entire ads are in English) as well as on a micro-level (part of the advertisements are in English) (Van Meurs 2010; Zenner et al. 2013). Van Meurs (2010) showed that the use on micro-level appeared to be more frequent than that on macro-level in The Netherlands. Whereas the macro-level use is often motivated by factors related to the job itself, this is not always the case for the use on a micro-level.

Van Meurs (2010) found that in The Netherlands almost 90 % of the jobs advertisements with English loanwords appear in advertisements for higher – level or academic jobs, mainly in sectors such as public administration, health, commercial services and education. A closer look at these job ads shows that in some cases the use of English is directly related to the job advertised, in the sense that the job involves a certain knowledge of English, the applicant has to function in an international setting, or the type of job is in an environment that involves a lot of English jargon (e.g. IT sector). However, for most of these advertisements the reason for the use of the English does not seem to be related to the job advertised, given the fact that the job is not based in a English-speaking country, does not involve international contacts, or any knowledge of the English language. A lack of an appropriate word in the recipient language has been proposed by many studies on the use of English loanwords (Hock 2021; Wetzler 2006) as an explanation for their use. The German ‘Federal Institute for Vocational Education and Training’ (BiBB) that provides information on all job training programmes available in Germany, for example, lists Controller or Fachwirt Public Relations (business management specialist public relations) as being without a German equivalent. However, a lack of an appropriate equivalent does not always seem to be the reason for the use of an English loanword either. In most cases, good equivalents for the English words were available in the native language of the country in which the advertisement appeared (Van Meurs 2010; Van Meurs et al. 2006). So, the question is why companies use English in their job advertisements.

Myers-Scotton (1991, 1993, 1998 in her Markedness Model suggested that the use of English loanwords or code-switching is a way to communicate group membership or interpersonal relations, which could be either real or desired. The choice to use English rather than Dutch in The Netherlands for example is marked. Onysko and Winter-Froemel (2011) distinguish between catachrestic and non-catachrestic borrowings. The former introduce new concepts that are absent in the recipient language (the above example controller does not have an equivalent in German). The latter usually have a near equivalent in the recipient language, but the use of a loanword in a particular context is marked, indicating that something about the meaning conveyed is not normal or stereotypical. In an advertising context, this usually means that the product or service advertised is communicated as belonging to a group of products or services with better or more desirable characteristics. For the use of English loanwords, Domzal et al. (1995), Fink (1977), Kelly-Holmes (2005), Piller (2001), and Schütte (1996) have argued that they, rather than their equivalents in the recipient language, lend the advertised product or service an air of modernity and globalization. This in turn makes a product or service more desirable and renders the advertisement more effective than the same ad in the recipient language. This is in line with a study by Van Meurs (2010) on the motivation of recruitment professionals for using English in job advertisements. He indicated that, among other reasons, the job advertised using English would appear more attractive to prospective applicants than the Dutch equivalent. So, the linguistic choice made is considered to render a job advertisement more effective in the sense that the job as well as the company are perceived to be more attractive to prospective applicants, which could in turn result in a job application.

A lot of research has been done in The Netherlands. Here, English is frequently used on a daily basis, especially by people with higher or more advanced education. In addition, foreign film on TV and in the cinema are typically subtitled rather than dubbed and, according to the ‘EF English Proficiency Index’ (EF EPI), Dutch speakers of English are the most proficient among the 35 countries studied. The situation in Germany is somewhat different. English is less readily spoken on a daily basis, Germans are tenth on the EF EPI (EF.edu 2021), and have less contact with the English language via media such as film and TV. This could mean that the choice to use English loanwords is even more marked than it is in The Netherlands.

The study of English loanwords calls for a clear identification of what constitutes an English word. A task that is more complicated than it would appear at first sight (for a recent discussion see Roberts et al. [2021]). A common approach is to use dictionaries of English and the recipient language to determine whether a word is English or not. However, this is problematic for words that appear in both dictionaries. These words are either not taken into consideration (e.g. Van Meurs 2010) or end up in a separate category (Gerritsen et al. 2007b) and are further categorized depending on their degree of integration into the recipient language. Since most of the research on the use of English in job advertisements has been done among (job-seeking) graduate students, words from the latter category of loanwords (such as manager, team, or marketing) were included because they often feature in advertisements that are attractive to a large group of participants.

1.3 The effect of the English language use in job ads

The perception of English loanwords in job advertisements has been studied in terms of their influence on the communicative goal of the ads, rather than the perception of the use of the loanword itself. Renkema et al. (2001) investigated the effectiveness of job ads with English loanwords compared to those in Dutch in The Netherlands. They studied readers’ reactions in terms of their attitude towards the ad, the evaluation of the ad, the company image, and the exclusiveness of the job. They found no differences between the Dutch ad and the ad using English loanwords or phrases. The only differences they found was that the use of English loanwords was considered to be less natural than the equivalents in the recipient language. Van Meurs et al. (2004) and Van Meurs et al. (2006) came to similar conclusions in their studies on English in Dutch job ads. Van Meurs et al. (2007) compared the effectiveness of five job titles in English and Dutch among Dutch students. In their study, the attitude towards the job title was more positive for English job titles for only two of the five ads used. Jobs advertised using English job titles were considered to be more international, but as to the associations readers had with the job title (positive, negative, neutral), no differences between English and Dutch were found. Dutch was more successful with respect to the evaluation of the job for three of the five job titles. So overall, the results by Van Meurs et al. (2007) were not very conclusive about the effectiveness of English job titles on the Dutch job market. Van Meurs et al. (2015) studied whether the use of English loanwords in job ads in The Netherlands was more persuasive than the use of their Dutch equivalents. They did not find any evidence for such an effect. Although the job in their study was considered to be more attractive when advertised using English loanwords, the estimated salary and the attractiveness of the jobs were higher for jobs advertised without English loanwords in their peripheral condition.^[2] Thus, the results of this study did not provide any clear evidence for a positive effect of the use of English loanwords in job ads. This is also true for an experiment carried out by Van Meurs et al. (2017) on the effect of English loanwords in job titles in The Netherlands and Flanders. They concluded that “that English loanwords (in this case, job titles) were not perceived as more prestigious than their Dutch equivalents” (Van Meurs et al. 2017: 104).

From the results of the studies discussed above, it is difficult to draw a clear picture of the actual effect of the use of English loanwords in The Netherlands, and in other countries where English loanwords are frequently used. Possibly, with the growing importance of the internet and with it, the increased availability of English language material and increasing contact, the situation may have changed in the past 10–15 years. This may have had an influence on the effect of English loanwords in job ads in terms of reader attitudes. Another effect that has been attributed to the use of English loanwords is their attention-getting function.

1.4 Attracting the reader’s attention

The ability to attract the readers’ attention has been suggested as a motivation for the use of English loanwords in job ads in non-English-speaking countries (cf. Domzal et al. 1995; Seitz 2008; Winter-Froemel et al. 2014). Domzal et al. (1995: 100) argued regarding product ads that “[f]oreign words stand out – they represent an incongruity relative to the language used in the rest of the advertisement – and this distinctiveness makes them noticed”. Thus, the use of English is supposed to have an effect which is similar to that of other aspects of an ad that are considered to be original and therefore attract the readers’ attention (cf. Pieters et al. [2002], but see also Hornikx and Van Meurs [2020]). This form of attention has been considered the key factor for the success of job ads (e.g. Barber 1998; Blackman 2006; Kaplan et al. 1991; Redman and Mathews 1992; Wiley 1992), or as Barber (1998: 38) puts it, “[e]ffective recruitment materials must first attract the attention of potential applicant, and then persuade them to act (i.e. to apply)”. Thus, initial attention is crucial to the selection of which ad to look at and which not. In addition to attracting readers’ initial attention, it has been suggested that the English part of a job ad might also get more attention in terms of cognitive processing, because processing a foreign language generally takes more cognitive effort than processing the same text in the reader’s native language (Piller 2001). Given the extra effort readers make to understand the English part of the ad, they might even care to read the rest of the advertisement. And this is precisely what organisations want to achieve with their job ads.

However, there is little empirical evidence for the attention-getting function of the use of English in job ads. Hilberink-Schulpen et al. (2016) tried to find evidence that does not require participants’ judgements or self-reflection. They carried out an eye-tracking study to investigate the attention-getting function of English language use in Dutch job ads. They used ‘attention’ in the sense of Moray (1969), who defined it as “the ability to select part of the incoming stimulation for further processing” (as paraphrased in Eysenck and Keane 2002: 363). They used the eye-tracking paradigm to measure attention in terms of first fixation, average viewing time, number of returned views and number of fixations. Participants saw job ads in different conditions, one at the time. The ads were manipulated with respect to the language used in the job title and the company information. Hilberink-Schulpen et al. (2016) manipulated two parts of the job ad, because the job title is generally considered the most important part of a job ad, which might make it difficult to disentangle the effect of the foreign language and that of an important part of the ad. The manipulation of the ads resulted in job ads that were entirely in Dutch, ads with an English job title, and ads with the company information in English. The results did not confirm the attention-drawing function of the use of English in job ads. The company information received even fewer fixations when it was in English compared to the native Dutch version. In this eye-tracking study, participants looked at individual job ads, in which the attention-getting capacity of the ads was determined by comparing the viewing patterns for the individual job ads in the different conditions. Yet, in an average newspaper reading situation, readers would see a number of different job ads on the same page competing for the readers’ attention. The question is whether in this more natural context, the use of English is able to draw the reader’s attention to one rather than another job ad. This is why in our study we used the eye-tracking methodology to look at competing job ads and their ability to attract viewers’ attention. Based on the assumption made by Domzal et al. (1995), we would expect that English job titles that are more deviant from what the reader is used to in the equivalent job titles in the recipient language, would draw initial attention more frequently. In order to test whether unexpectedness is really the trigger for attention, we also included an English ad with a German job title, a combination that hardly ever occurs and is consequently even less expected. To control whether an English text would actually require a heavier cognitive load to be processed, we also included a completely English job ad.

2.1 Materials

For this experiment, different job advertisements were created in German. They focussed on positions for people entering the job market (trainee or entry-level positions) and for sectors that were shown to most frequently include English in online job ads, such as public administration, commercial services etc. in The Netherlands (van Meurs 2010), which are attractive to a wider audience. In order to be of interest to as many respondents as possible, the information on the required candidate profile and the job branch were kept rather general. The texts for these advertisements were taken from job advertisements that appeared on online job sites such as Stepstone.de and Monster.de or the Frankfurter Allgemeine Zeitung, a national German newspaper. A number of job advertisements that were directed at the target group of this experiment (highly educated graduates starting on the job market) were collected to get an idea about their make-up. These job advertisements appeared to make frequent use of the words trainee and management/manager, English loanwords that lack an equivalent in German. Given the characteristics of the real-life job advertisements, a decision had to be made between a clean design (no words like manager and trainee in the German job title, advertisements not reflecting the real life situation) or a realistic, more ecologically valid design. A decision was made for the latter combined with an extra analysis of these words in the experiment.

Based on the small collection of job ads, a total of 24 advertisements were created in German for positions such as an International Management Trainee Sales, Project Assistant Trainee, Marketing Trainee or Junior Buyer – Graduate Entry Level. The English advertisements were translated into German (back translated to control the translation). This resulted in a total of 48 job advertisements (see Appendices A and B for examples of the advertisements and all job titles). The lay-out of the advertisements was changed to a more uniform lay-out in terms of composition, length, font type and size to prevent differences in lay-out from influencing the general attractiveness of the advertisement and the viewing behaviour. All advertisements had a logo, a job title, information on the company and the candidate requirements, as well as the job offer. These were the sections of job advertisements that according to Feldman et al. (2006) contribute to the attractiveness of these advertisements. The advertisements did not contain any information on the location of the company, on compensation and benefits to prevent these aspects from attracting readers’ attention and influencing their attitude. In addition, the original logos and company names were replaced by fictitious logos and company names to prevent well-known companies from attracting more attention than those of less well-known companies. The logos themselves reflected an image, and the fictitious company name, image size and font size were kept similar, but colour and shape differed to create an ecologically valid stimulus. The fact that the logo colours might attract readers’ attention was controlled by the fact that the four conditions of one advertisement had the same logo. On average, the advertisements contained 90 words, the number of bullet points in the job ad was counterbalanced over stimuli (the use of bullet points is common in online job advertisements).

The resulting 48 advertisements were pre-tested by six native speakers of German. The participants of the pre-test evaluated all 48 advertisements in a random order on their naturalness, intelligibility and whether the job advertisements are a good example of a job advertisement on a scale from 1 (strongly agree) to 5 (strongly disagree). The 16 job advertisements which received the highest scores in both the English and German version for all three measures were used for the actual experiment.

On the basis of these two language versions of these 16 job ads (all English and all-German advertisements), two additional versions were created: a German version with an English job title and an English version with a German job title. The all-English ads and the German ads with an English job title functioned as control stimuli. This resulted in a total of four versions for each of the advertisements (for an example of the four conditions of one advertisement see Appendix B). For each of the advertisements, two interest areas (AOIs) for the eye-tracking part of the experiment were created. These AOIs were the job title and the entire advertisement. The advertisements were presented in four sets of four advertisements to imitate a page of an online job platform with a number of job advertisements on one page. Each set consisted of four different advertisements in four different conditions (all English, English with German job title, all German, and German with English job title), as can be seen in Figure 1. The positions of the advertisements in the different conditions varied in the four sets. An advertisement in one language condition did not appear in the same position in one of the other sets an individual participant saw. This resulted in four different sets of stimulus material.

Figure 1:

Example of a stimulus set of four advertisements with the job title highlighted as AOI (this boxed highlight was not included in the actual stimulus material).

2.2 Design

For the eye-tracking part of this experiment, a 4 × 4 mixed design was used. There were 16 different advertisements, each in four language conditions (all English, English with German job title, all German, and German with English job title), and the advertisements in the different language conditions appeared in four different positions (top left, bottom left, top right and bottom right). Given earlier evidence for the influence of reading direction on viewing behaviour (Göbel et al. 2011; Ossandón et al. 2015), the position of the advertisement on attention position was added as a factor. The advertisements were presented in sets of four advertisements. Participants saw all conditions of the advertisements in all four positions, but they only saw one condition of the individual advertisements.

For the questionnaire part of the experiment, a between-subject design was used. Participants saw one advertisement in one of the four language conditions (all English, English with German job title, all German, and German with English job title) and answered questions concerning their attitude towards the advertisement, recruitment outcomes and their job pursuit intentions.

2.3 Participants

A total of 60 native German students from the Radboud University Nijmegen^[3] took part in this study, divided over 4 groups. They represented the intended target group, since van Meurs (2010) has shown that in online job advertisements English is most frequently used in higher-level positions. Their mean age was 22.50 (range = 18–31, SD = 3.07). All participants had normal to corrected vision and German as their mother tongue; 22 of the participants were male (36.7 %) and 38 were female (63.3 %). There was no difference in age between the four groups (F (3, 59) = 1.68, p = 0.182). In addition, we tested whether age had an influence on the dependent variables, given that more experience on the job market might have an influence. We also tested whether gender had an influence on the results. The gender of the participants was not equally distributed between the four groups (χ ² (3) = 10.64, p = 0.014): group 1 had fewer men than group 2.^[4] There was no difference with respect to gender between the remaining three groups. For this experiment, participants judged their overall proficiency in English on a scale from 1 (not proficient) to 5 (very proficient) (for the items used see below). There was no difference between the four groups with respect to their proficiency in English (F (3, 56) < 1). Their mean proficiency in English was relatively high M = 3.87 (SD = 0.68). The intended target group was highly educated and therefore had a high level of proficiency in English. This is similar to the studies discussed above, creating a high comparability.

2.4 Instrumentation

An eye-tracking study was used to determine whether the language choice in job advertisements (English, English with German, German, and German with English) that targeted the German job market led to differences in the way the respective advertisements attract the attention of potential applicants. Participants’ eye movements were recorded using the EyeLink II from SR Research. Attention was determined by three of the different measures the eye tracker is able to record (Poole and Ball 2006; Rayner 1998; Salvucci and Goldberg 2000). First, the attention-grabbing function was measured by the number of first fixations for the job ads in the different positions and conditions. Second, attention in general for the job title and the remaining parts of the job ad were measured by the proportional average viewing time in the different positions and conditions, and the number of fixations in the two AOIs defined for the different advertisements. The advertisements were visible for 45 s. This time limit was determined in a pre-test; it gave participants the opportunity to skim the advertisements but did not allow them to read them in detail since we were mainly interested in the attention the job advertisements in the different conditions was able to attract. In a normal real-life situation, job ads are hardly ever read one by one, they are usually skimmed through and only the relevant ads receive more attention (cf. Jones et al. 2006). This is the situation we tried to imitate by setting a time limit.

After the eye-tracking part of the experiment was completed, participants were asked to fill in a questionnaire to measure their attitude towards the advertisements, the recruitment outcomes, participants’ job pursuit intentions, and their proficiency in English as a control variable.

The attitude towards the advertisement was measured in terms of the intelligibility of the advertisement, its attractiveness and its naturalness (cf. Van Meurs 2010). Five-point Likert scales (1 = I totally disagree, 5 = I totally agree) introduced by “The text is …” were used for all items. The intelligibility of the advertisement was measured with five items (easy, clear, well-organised, logically structured and concise). The reliability measured with Cronbach’s alpha for these items was acceptable (α = 0.78), meaning that all five items measured the same concept of intelligibility. The attractiveness of the advertisement was measured using four items (interesting, appealing, inviting and varied). The reliability for these four items was good (α = 0.80). The naturalness of the advertisement was measured with the single item The advertisement is natural.

The recruitment outcomes were measured in terms of the attitude towards the company, the attitude towards working for the company, and the attitude towards the job. Five-point Likert scales (1 = I totally disagree, 5 = I totally agree) were used for the three constructs. The attitude towards the organisation was measured using a scale that was introduced by “The company is …” followed by seven items: reliable, professional, honest, prestigious, attractive, innovative, international. The reliability of these seven items was inadequate (α = 0.26); leaving out the item international, however, resulted in an adequate reliability (α = 0.70). Consequently, this item was not included in the calculation of the concept attitude towards the organisation. The attitude towards working for the company was measured with three items (interesting, exciting, gratifying) introduced by “Working for this company is …”. The reliability of these three items was adequate (α = 0.79). The attitude towards the job was also measured with three items: involves a great responsibility, is varied and is interesting. These items were introduced by “The job is …”. The reliability of these three items was unacceptable (α = 0.45). When the item ‘involves a great responsibility’ was left out, the reliability of the remaining two items was questionable (α = 0.61). However, the inter-item correlation mean of the 2 items was 0.452. This score is just within the optimal range short scales suggested by Briggs and Cheek (1986: 115). Therefore, the attitude towards the job was calculated using these two items.

Job pursuit intentions were measured with two items (“Would you like to know more about the possibilities to work for this company?” and “If you completed your degree would you consider applying for a job at this company?”) using 5-point Likert scales (1 = I totally disagree, 5 = I totally agree). The reliability for the two items was good (α = 0.81).

Proficiency in English was measured in terms of participants’ own judgement of their proficiency based on four items: rating their speaking, writing, reading and listening proficiency. Although self-rated proficiency might be less accurate, studies have shown that self-rated and objective proficiency are correlated (e.g. Scholl et al. 2021). For the four items, 5-point Likert scales (1 = not at all, 5 = very good) were used. An overall mean of these four measures of proficiency was calculated. The overall proficiency was M = 3.87 (SD = 0.68)

At the end of the questionnaire, participants were asked to provide some personal details (such as age, gender, handedness, mother tongue, education and current job).

2.5 Procedure

The experiment was conducted by a German native speaker, who instructed participants in German. There were two parts to this experiment. The first part was the eye-tracking part. Participants were told that they were going to see different job advertisements in six different sets (two practice sets and four experimental sets). They were instructed to look at these advertisements as if they were looking for a new job in a newspaper. They were also told that they would not have enough time to read every detail of the advertisements but would only have time to skim them. To familiarise participants with this time restriction and the eye-tracking device, they saw two practice sets first.

Before participants saw the practice set, calibration of the eye–tracking setup took place. After participants saw the practice set, the actual experiment started. Each job advertisement was shown for 45 s followed by a fixation point, before participants saw the next set of job advertisements. The order in which the different job advertisements were presented and their position within a set were the same for all participants. However, the conditions (all English, English with German job title, all German, and German with English job title) varied in all four versions of the material.

After the eye-tracking part of the experiment, participants continued with the second part. Participants saw one of the job advertisements they had previously seen in the eye-tracking part. The condition of this ad (all English, English with German job title, all German, and German with English job title) varied across the different participants. After having seen the advertisement, participants answered questions concerning their attitude towards the advertisement, recruitment outcomes as well as their job pursuit intentions. In total, the experiment lasted between 20 and 30 min.

3.1 Attitude towards the advertisement

The attitude towards the advertisement was measured in terms of the attractiveness, the intelligibility and the naturalness of the advertisement. Three one-way ANOVA analyses were carried out to see whether the language choice in the advertisement had an influence on the three measures of the attitude. The means for the three variables are displayed in Table 1. The attractiveness of the ad was not influenced by the language condition F (1, 29) = 1.13, p = 0.296. The same was true for intelligibility F (1, 29) = 1.80, p = 0.191 and naturalness F (1, 29) < 1. The language choices made in the ads did not have an influence on the attitudes towards the ad.

3.2 Job pursuit intentions and recruitment outcomes

Four measures were used to see whether the language used in the job ad had an influence on recruitment outcomes. The measures were the attitude towards the organisation, towards the job, the responsibility that comes with the job, and job pursuit intentions. Mean scores can be found in Table 2. One-way ANOVA analyses were carried out to determine the influence of the language conditions. The language choice made in the ad did not influence any of the four measures (attitude organisation F (1, 29) < 1; attitude job F (1, 29) < 1; responsibility F (1, 29) < 1, and job pursuit intentions F (1, 29) < 1). This means that job pursuit intentions were not influenced by the language choice in the advertisement.

3.3 First fixation

Before the analysis of the eye tracking data, we checked whether the viewing pattern for the advertisements with the German job titles were different for the advertisements with words such as trainee and manager in the job title compared to the German job titles without these words; we did not find any differences (p > 0.050 for all comparisons between job titles with English words in the German job title and purely German job titles), therefore all stimuli were used in the analyses below.

A Pearson’s chi-square analysis was conducted to analyse whether there was a relation between the first fixation for the four different language conditions (English job ad, English job ad with German job title, German job ad, and German job ad with English job title) and the four different positions of the advertisements (top left, bottom left, top right, and bottom right). The number of first fixations per ad per condition can be seen in Table 3.

The results of the analysis showed that there was no significant relation between the language choice in the job ad (English job ad, English job ad with German job title, German job ad, and German job ad with English job title) and the first fixations in the different positions of the ad (top left, bottom left, top right and bottom right) (χ ² (9) = 9.94, p = 0.355). This means that the first fixation was not affected by the language choice, nor by the position of the ad.

Another Pearson’s chi-square analysis was conducted to see whether, irrespective of the position of the ad, there was a relation between first fixation for the AOI job title and the four different language conditions (English job ad, English job ad with German job title, German job ad, and German job ad with English job title). The number of first fixations can be seen in Table 4.

The results revealed no significant relation between the AOI job title and the four conditions of the ads (χ ² (3) = 4.03, p = 0.258) (nonparametric one sample test). In other words, first fixations for the job title were not affected by the language used in the ads. In order to see whether there are more first fixations in one of the four possible positions of the ad irrespective of the language used, another Pearson’s chi square test was carried out. The number of first fixations in the different ad positions can be seen in Table 5. The test revealed a significant relation between the position of the ad and whether it received a first fixation (χ ² (3) = 34.20, p < 0.001). Column proportions (Z-scores) revealed that ads in the top left position received significantly more first fixations (39.1 %) than ads in the other positions (top right 18.7 %, bottom left 22.1 %, bottom right 22 %). This means that the position of the ad has an influence on the number of first fixations an ad received, with most first fixations being in the top left position.

In order to see whether the position of the ad overruled the influence of the language manipulation, another Pearson’s chi-square analysis was conducted without ads in the top left position to see whether there was a relation between the first fixation for the three remaining positions of the advertisements (bottom left, top right, and bottom right) and the four different language conditions (English job ad, English job ad with German job title, German job ad, and German job ad with English job title). However, there was no significant relation between the different language conditions and the three positions of the ad (χ ² (6) = 2.17, p = 0.306). There was no effect of the language used on the location of the first fixation. The first fixation was predominantly in the left upper corner, which is illustrated in the heatmap in Figure 2. The heatmap is a visualisation of the accumulated first fixations of all participants across the different conditions to show the areas that received most first fixations.

Figure 2:

Heatmap of first fixations, irrespective of language of advertisement and job title.

3.4 Number of fixations

To establish whether an English job title can hold more attention for that job title, a repeated measures analysis was run with language (all German and German with English job title) as within-subject factor for the number of fixations. The analysis showed no significant main effect of language (F (1, 59) < 1).

In addition, to see whether English took more cognitive effort than German, a repeated measures analysis was run for the ad as a whole with language (all German, all English, and German with English job title) as within-subject factor. The analysis showed no main effect of language for the number of fixations (F (1.84, 108.62) = 2.39, p = 0.101). This analysis is reported with the alternative Huynh-Feldt F-statistic because the assumption of homogeneity of variance was violated. The mean number of fixations per language condition can be found in Table 6.

3.5 Duration of fixation (proportional duration)

To determine whether the use of a job title in English leads to longer processing times of that job title or the ad as a whole, two analyses were carried out. First, a repeated measures analysis for the proportion of duration fixations of the job title only with language (all German and German with English job title) as within-subject factor, showed no significant main effect of language (F (1, 59) < 1). Second, a repeated measures analysis for the proportion duration of fixations of the whole ad with language (all German, all English and German with English job title) as within-subject factor, also showed no significant main effect of language (F (2, 118) = 2.23, p = 0.112). The proportional duration of fixations per language of the ad can be found in Table 7.

In order to be able to see whether job ads in one position were viewed longer than ads in other positions, a repeated measures analysis for the four different ad positions was carried out. The proportional duration of the views in the different positions can be seen in Table 8.

The analysis showed that there was a significant effect of the position of the job ad (Huynh-Feldt correction; F (2.42, 148.57) = 4.92, p = 0.005). A pairwise comparison with Bonferroni correction showed that job ads in top left position (M = 22.05 %, SD = 0.10) were viewed longer than ads in bottom left position (M = 16.74 %, SD = 0.06). There were no differences between the ads in the other positions.