Cultural directional preferences across countries
As expected, the CDPQ showed that preferences were generally LR in German; however, surprisingly, they were not generally RL in Iran (see Faghihi & Vaid,
2023 for a review), but rather, indicating no preference overall. We could conceive two possible explanations for this.
First, it has been argued that LR directionality is initially rather innate (e.g., Felisatti et al.,
2020) and only later overwritten by culture (e.g., McCrink & de Hevia,
2018). It is possible that this overwriting occurs only partially so that responses in the CDPQ are influenced by both innate LR directionality and cultural RL directionality, leading to an actual null effect. Note in this respect that the LR/RL directionality for Iranians differs greatly depending on the focus of attention. While object naming and line drawing were from LR, even in most Iranian participants, the action moving items, horizontal timeline, and bicycle (drawing) were predominantly from RL. Thus, it is possible that certain activities are more influenced by culture, and others, are less so.
Second, the world culture portrayed in movies and video games, experienced in international airports, computers, and mobile phone systems comes predominantly from LR cultures (e.g., the US). Individuals exposed to it are more likely to be confronted with LR culture in images and experience, which was the subject of our CDPQ. This global LR directionality might influence cultural habits in addition to their own culture. Some cultural aspects might be more RL because they are influenced more by their own culture (e.g., rarely encountered via international media, travel, and technology), while those more frequently encountered in international culture may be more LR. Since global culture and language influence many aspects of national languages (e.g., via loan words) and culture (e.g., via international music), it is clearly conceivable that this holds for directionality as well. It is therefore impossible to distinguish the two hypotheses in urban samples exposed to global culture, and future research should disentangle these by studying remote populations with RL culture and restricted or no access to global culture.
Overall, the preferences of German and Iranian participants for most items were in line with the previous findings of LR and RL cultures, respectively. The horizontal timeline and action moving items generally followed the respective culture direction as demonstrated previously (see Tversky et al.,
1991; Mass & Russo, 2003). In the bicycle drawing, German participants reported a greater preference to draw the front-facing leftward compared to other samples. When the front is leftward, the bicycle appears moving RL, which might seem counterintuitive for an LR culture. On the other hand, this finding is in line with previous research with adults and children that demonstrated the tendency to draw the objects so that they appear moving RL in LR cultures, and so that they appear moving LR in RL cultures (Kebbe & Vinter,
2013; Vaid,
1995). The significant difference between samples in bicycle drawing disappeared in the expression of bicycle liking, suggesting that this directional bias difference across samples was specific to drawing habits, but not necessarily generalized to liking for objects. One possible explanation of this discrepancy between drawing and liking on directional preferences is that since people tend to start from the left side of the paper in LR cultures, they might place the most important and detailed part (i.e., the front) on the left side and continuing the drawing (i.e., the back) toward the right side. Correspondingly, we would expect the exact opposite pattern in RL cultures. Note that, in the present study, participants did not actually draw but only reported how they would draw the bicycle, which might underestimate their directional bias compared to lab studies. Interestingly, there was an overall LR preference in the object naming, including for Iranian participants. This finding is interesting since Shaki et al. (
2012) showed RL preference in a similar task (i.e., object counting) in RL cultures (see Shaki & Fischer, 2023 for the influence of object counting direction on the attentional SNARC effect). On the other hand, a study by Shaki and Gevers (
2011) suggested that cultural characteristics can manifest themselves in reverse directions for magnitude and ordinal information processing. In their study, magnitude information was associated with LR processing but ordinal information was with RL processing in Hebrew participants. In addition to that, although the tasks in the mentioned studies seem superficially similar, object naming in the present study and object counting in Shaki et al. (
2012) might include different directionalities: object counting requires active object numbering by pointing at objects, whereas in the present study, participants identified pictures of the objects on a computer screen and wrote the words for each object, an activity which might be influenced by LR computer screen scanning habits on a more global scale. Although significantly less prominent in Iranian participants, there was an overall LR preference in line drawing, in contrast to Lieblich et al.’s (
1975) study, and again, this difference could be due to the online setup of the questionnaire.
Our initial expectation was that preferences in Turkish culture would be closer to Iranian rather than German culture, but this did not seem to be the case in the present study. The total scores were negative (i.e., indicating LR direction) for both German and Turkish participants and close to zero (i.e., indicating no preference) for Iranian participants. More specifically, the preferences in object naming, horizontal timeline, and action-moving items were similar to those in Germany. On the other hand, Turkish and Iranian preferences were closer to each other compared to German preferences in bicycle drawing and line drawing. This pattern suggests that certain cultural directional preferences in Turkish culture more tightly linked to the reading direction (i.e., habits that require scanning of horizontally aligned objects/events) tend more to LR. Other preferences that are less influenced by reading direction and more ingrained in the culture may tend towards more RL patterns, because the LR reading direction has only been relatively recently established in Turkey, i.e., for about four generations.
As a result, the CDPQ successfully captured differences in cultural directional preferences across samples in an online setup. This suggests the relative effectiveness of this approach, namely, the pseudo-experimental manipulation of testing of three different cultures, with one being predominantly LR in cultural directional preferences, one being predominantly RL, and the third being mixed between reading direction and cultural directional preferences. Based on these results, the next question was whether such cultural differences would affect SNAs.
Directional SNAs are influenced by cultural directional preferences
The differences between German and Turkish culture in the SNARC effect, despite the common LR reading direction, strongly suggest that factors other than reading direction are potentially involved in the emergence of SNAs. Turkish CDPQ scores are less LR than German ones, suggesting that these factors could be cultural directional preferences. The influence of cultural directional preferences was previously proposed in several studies (see Nuerk et al.,
2015; Patro et al.,
2016a,
2016b; Göbel et al.,
2018) to explain the emergence of SNAs in preschool children. In the present study, we further suggest that the influence of cultural directional preferences is not merely a transient phase during early preschool development, where reading direction is not (well) developed. Instead, it seems that, even in adults, cultural directional preferences influence SNAs beyond reading direction.
The hypothesized reason for the regular but weaker (compared to German culture) SNARC effect in Turkish culture is that there are RL cultural directional preferences (at least more RL as compared to Germany), even though the reading direction is LR. The underlying reason for the absence of strong and consistent LR directionalities in Turkish culture could be related to the script change around a hundred years ago. However, it could be argued that participants in the present sample were not influenced by this change directly. Note that in 1928 (i.e., just before the time of the Latin alphabet revolution), only 8% of the Turkish population was literate (Yılmaz,
2011), indicating that, even at that time, only a small portion of the population was directly influenced by the change. Importantly though, the reading direction of a culture infiltrates many organizations in the cultural environment and typically covaries with the cultural directionalities. Therefore, (semi-) illiterate individuals can be still influenced by their socio-cultural environment, which naturally involves directionalities (see Nuerk et al.,
2015, for a review). Before the Latin alphabet revolution, Turkish culture was probably a more consistent RL culture due to not only the RL Arabic-Farsi alphabet but also the Middle-Eastern and Islamic RL influence on the cultural environment. The literacy rate increased dramatically after the Latin alphabet revolution, eventually influencing the directional preferences in the cultural environment. Nevertheless, some RL habits might have remained, especially those not directly linked to reading direction. Considering that the reading direction is LR, these remaining RL habits might create an inconsistency in associations of space and sequence or space and magnitude (see Patro et al.,
2016a, for how such influences may develop), which may be the cause of weaker LR SNAs in Turkish culture.
The differences between Turkish and Iranian culture in the SNARC effect in both tasks suggest that reading direction might also be involved in the emergence of the SNAs. As demonstrated previously in the literature, RL reading habit has a weakening influence on the regular SNARC effect compared to LR reading habit (e.g., Dehaene et al.,
1993; Shaki & Fischer,
2008; Lopiccolo & Chang,
2021) and has even been claimed by some to reverse the effect (e.g., Shaki et al.,
2009; Zebian,
2005). Nevertheless, when we evaluate the literature overall, the influence of reading habits on the SNARC effect remains unclear due to inconsistent findings among RL readers as well as the absence of evidence of a causal relationship between reading activity and number processing (see Pitt & Casasanto,
2020).
The group-level regular SNARC effect was significant for German and Turkish participants in both PJ and MC tasks. This finding was unsurprising for the German sample since the effect has been demonstrated in LR cultures several times with both tasks (see Wood et al.,
2008). In contrast, Bulut et al. (
2023) reported no SNARC effect in a Turkish sample with a PJ task¸ however, it should be noted that there are three major differences between Bulut et al. (
2023) and the current study. First, in the present study, the number of participants was considerably higher, second, the number of trials per response-to-key assignment was two-fold as in Bulut et al. (
2023). These two parameters are highly influential in increasing the sensitivity to detect an existing SNARC effect (Cipora & Wood,
2017). Finally, Bulut et al. (
2023) conducted a lab study, while the present study was run online. However, it is unlikely that the difference in the results would originate from online or lab administration because the SNARC was successfully replicated in an online setup previously (e.g., Cipora et al.,
2019a). Future investigations need to explore the reason for the difference between the results, but at present, lower power and lower measurement precision seem likely candidates for previous null effects. Also, one should consider that the Turkish SNARC was considerably smaller in size compared to the German one, therefore, the reason for previous null effect findings in the Turkish sample (Bulut et al.,
2023) is likely a conjunction of lower power and the effect being smaller (consequently, even more difficult to detect with low power).
In Iranian culture, the regular group-level SNARC effect was significant in PJ and, even though, descriptively, being regular (i.e., negative slope), it was not significant in the MC task. Note that Bayesian statistics provided inconclusive evidence for both tasks in the Iranian group-level SNARC effect. Previous studies also demonstrated conflicting evidence for the SNARC effect at the group level in RL cultures; some reported a reverse SNARC effect (Shaki et al.,
2009, in Palestinian participants; Zebian,
2005, in Lebanese participants), others reported null effects (Shaki et al.,
2009, in Israeli participants; Dehaene et al.,
1993, Exp. 7, in Iranian participants; Lopiccolo & Chang,
2021, in Jordanian participants), and yet others reported a regular SNARC effect (Hochman et al.,
2024; Zohar-Shai et al.,
2017 in Israeli participants; Feldman et al.,
2019 in Israeli children; Supplementary Material of Cipora et al.,
2019a, in Iranian participants; Shaki & Gevers,
2011, in Israeli participants). Considering these and the present study’s findings, it seems that more recent studies report regular SNARC effects more frequently, even in RL cultures, than earlier ones. How could a regular SNARC effect be explained in an RL culture? As we discussed above in cultural directionalities, a possible explanation is that LR directionality is innate and overwritten by culture (Adachi,
2014; Drucker & Brannon,
2014; McCrink & de Hevia,
2018; Rugani et al.,
2015). Alternatively, instead of overwriting (by deleting the innate biases), cultural directionalities may diversify SNAs. This suggests that innate biases on SNAs are still present in adult life (see
The Brain’s Asymmetric Frequency Tuning hypothesis in Felisatti et al.,
2020) and coexist with culture-dependent representations.
Even though culture clearly influences individuals’ responses, this influence might not be strong enough to reverse the SNARC effect, but only weaken it (i.e., null or weak regular effects). Furthermore, global culture is dominated by LR cultures, as discussed above. With increasing globalization, populations from both LR and RL cultures are exposed to globalized cultural directional preferences. Support for this notion is that more recent studies (i.e., from a more globalized world) found a weakening of the effect, rather than a complete reversal. Moreover, some studies demonstrated the influence of experience with LR culture on the SNARC effect. For instance, Deheane et al. (1993) demonstrated a significant relationship between the time spent in France and the SNARC effect of Iranian participants; the more time in France, the stronger the regular SNARC effect. Furthermore, the level of SNARC effect was shown to be related to participants’ status as either monoliterate (in RL languages) or biliterate (in both LR and RL languages) (see Zebian,
2005; Lopiccolo & Chang,
2021). In the present study, more than half of the Iranian participants reported speaking English as a foreign language and most stated they were somewhat familiar with LR cultures. Although these findings were in line with the globalization explanation of the absence of reverse SNARC effects in RL cultures, the familiarity of the Iranian participants with LR cultures and LR languages overshadows the influence of the cultural directionalities. Therefore, future studies should focus on SNAs among monolinguals/participants who have no/limited experience with LR cultures in general.
Finally, we should also note that Eastern Arabic numbers are written/read in LR fashion, and therefore, during a number-related task, LR activation of the number system might interfere with the responses. The above-mentioned factors may result in inconsistent directionalities in RL cultures, depending on the sample’s characteristics, which in turn resulted in weakened, but not reversed, SNARC effect. A weak SNARC effect could be more susceptible to the power of the study, precision of measurement, and the task at hand, and consequently might produce inconsistent findings, which could explain the divergence of the findings of the SNARC effect in RL cultures. In sum, both the innateness of LR directionality and also, at a more individual level, experience with a predominantly LR global culture may explain why the SNARC effect is merely weakened, not reversed in an RL culture.
Markedness influencing response codes
Previously, effects of markedness on spatial association with response codes have frequently been reported. One such was the association of parity with response code (even-right and odd-left), which was first reported in German adults (Huber et al.,
2015; Nuerk et al.,
2004). Morphologically, German includes linguistic markedness in terms of even (“gerade”) being the standard version of the word, and odd (“ungerade”) being the opposite as marked by the prefix (“un”). This corresponds to formal markedness (see Nuerk et al.,
2004) and is a strong determinant of the markedness grade. Therefore, we would expect a strong MARC effect in German culture. In the present study, there was some marginally significant (
p = 0.06) standardized MARC effect, which may hint towards small markedness influence in the German sample. Importantly, Nuerk et al., (
2004; see also Iversen et al.,
2004,
2006) showed that the MARC effect is more consistent when the stimuli are presented in verbal notation, rather than in the Arabic notation used in this study.
There was no significant MARC effect in Turkish (see also Bulut et al.,
2023) or Iranian samples. The morphological structure of markedness is lacking in Turkish (“tek”—odd, “çift”—even) and Farsi (“fard”—odd, “zoj”—even), which could be an explanation for the absence of MARC in these cultures. At the same time, the morphological structure of the language for parity has not always been found to affect the MARC effect (Cipora et al.,
2019a, Supplementary Material).
Overall, the MARC effect with Arabic digits was not always replicated, as the effect is small in size and shows interindividual variability (see Huber et al.,
2015). The current study adds to these inconsistent results.
Inter- and intraindividual prevalence of the SNARC effect across cultures
Intraindividual prevalence of the cognitive effects is typically overlooked. To investigate the reliability of an individual SNARC slope, Cipora et al., (
2019b) suggested the H0 bootstrapping approach (see also Roth et al.,
2024b). Using this method, we found that the reliable SNARC effect was observed most frequently in German (57% in PJ and MC), less in Turkish (38% in PJ and 46% in MC), and the least in Iranian (29% in PJ and 39% in MC). The proportion of reliable SNARC effect was found to be higher in German culture compared to the previously reported values from LR cultures (e.g., ≤ 45% in Cipora et al.,
2019b; 42% in Hohol et al.,
2022 for PJ, 42% in Hohol et al.,
2020 Supplementary Material for MC). Although the reason for this higher proportion is not clear, it could be that participants performed both PJ and MC (see Supplementary Material Tables 15 and 16), whereas, in previous studies, participants only performed the PJ. Overall, these findings indicated that cultural directionalities, directional SNAs (i.e., the SNARC effect), and the prevalence of participants revealing a reliable SNARC effect systematically differed between cultures.
One important point we should address is that although the SNARC and CDPQ findings were comparable regarding the cultural differences, the CDPQ total scores did not correlate with individual SNARC slopes. This finding is not surprising considering that Roth et al. (
2024b) demonstrated that a SNARC effect was present at the group level in their study, but was not stable over time for individual participants. Therefore, the absence of correlation between CDPQ and SNARC slopes is not necessarily a result of CDPQ’s failure to predict the SNARC effect, but rather the poor stability of the individual SNARC slopes. A correlation between two measures in which one fluctuates unsystematically will be weak at best. In line with Roth et al. (
2024b), we successfully captured the cross-cultural differences at the group level in both CDPQ and the SNARC, while correlations based on individual data did not show the same pattern. Note that further research is needed to fully establish reliability and validity of the CDPQ therefore conclusions should be made with caution. Nevertheless, the findings are promising and suggest that CDPQ could be a new tool to assess the cross-cultural differences in horizontal directional preferences.
Finger counting habits cannot explain the sample differences in SNAs
Finger-counting habits are considered to influence the SNARC effect (Fischer,
2008; Fischer & Brugger,
2011). Although there is no conclusive evidence for this hypothesis (e.g., Prete & Tommasi,
2020; Cipora et al.,
2019a, Supplementary Material; Hohol et al.,
2022), we identified individuals’ finger-counting direction to examine its influence on the SNARC effect, since such habits show variance across cultures (see Lindemann et al.,
2011).
We were unable to replicate Fischer’s (
2008) findings in any of the cultures. In line with other studies (Fabbri,
2013; Hohol et al.,
2022; Prete & Tommasi,
2020), the SNARC slope of left- and right-starters did not differ significantly. Bayesian analysis supported this lack of differences between the two groups. Although most participants reported stable finger-counting habits (88% in German, 77% in Turkish, and 95% in Iranian), no link was found between the stability of finger-counting habits and the level of the SNARC effect, similar to Hohol et al., (
2020; 75% in Polish participants). It is important to note that in the present study, finger-counting habits were collected via self-report in an online setup, an approach that was shown to be inadequate in some studies (Lucidi & Thevenot,
2014; Morrissey & Hallett,
2018; Wasner et al.,
2014). Therefore, more controlled lab studies in which participants physically count their fingers repeatedly are needed to accurately capture these habits and their stability for studies focusing on the relationship between finger-counting and the SNAs.