Stakeholders in many municipalities worldwide are committed to promoting cycling and improving cycling provision. Scarcity of space is a major issue in most of these cities, especially for main streets with multiple demands, including the movement of pedestrians, cyclists and motorised vehicles as well as activities that take place on the pavement (also called sidewalk) such as communicating, eating, shopping, sitting in the sun or waiting at public transport stops. Requiring cyclists to share the carriageway with motorised vehicles might be the only possible solution for cycling provision in these contexts, which is applied in many German cities. Mixed traffic solutions for cycling in this study mean either no cycling markings at all or bicycle pictograms consisting of a bicycle and one or more arrows (also known as sharrows; Ruf et al. (2023)). In some references, marked advisory lanes are also considered as mixed traffic solution (Hantschel, 2022). Advisory lanes are cycle lanes demarcated by a broken white line that allows motorists to drive on the cycle lane if necessary, e.g., when two motorised vehicles encounter. However, advisory cycle lanes are not included in this study.

From the cyclists' perspective, mixed traffic is the second best solution for the majority of cyclists who prefer to be separated from motorised traffic. This preference has been consistently found in previous studies investigating perceived safety and comfort, mental stress, physiological markers such as galvanic skin response, and route choice, as stated in survey responses and revealed in field observations and naturalistic driving studies, among others (Chataway et al., 2014; Hu et al., 2023; Huber, 2022; Rossetti et al., 2019; Zhang et al., 2024). For the city of Berlin, FixMyCity (2020) finds in their online survey that only 11% of the surveyed cyclists perceive cycling in mixed traffic with a speed limit of 50 kilometres per hour and on-street parking to be safe or rather safe, this proportion increases to 28% with a speed limit of 30 kilometres per hour and no parking. Avoidance strategies are identified in the literature including the use of alternative routes with cycle facilities (Chataway et al., 2014; O'Connor & Brown, 2010) and the decision not to cycle at all (Chataway et al., 2014; Pearson et al., 2022).

Despite these concerns and the reluctance of cyclists to cycle in mixed traffic, they still use these facilities, mainly due to a lack of alternatives. Although it is illegal for cyclists to ride on the pavement in such streets, they might move onto the pavement to be better separated from motorised traffic and thus to increase their perceived level of safety. Very few studies could be identified that have observed cyclists' in such situations. Compliance levels, measured as the proportion of cyclists using the carriageway or cycle facility, were observed in the studies conducting on-site observations (Figure 1), including different types of cycle provision and street sections with two and four lanes. All four studies from the USA (Birk et al., 2004; Hunter et al., 2010; Mills et al., 2010; Pein et al., 1999) examine the effects of bicycle pictograms, which is also the focus of Koppers et al. (2021) from Germany and Vasilev et al. (2017) from Norway. Studies on the other facility types (except pictograms) are based in Germany (Alrutz et al., 2009; Alrutz et al., 2015; Kaulen et al., 2014; Ohm et al., 2015; Richter et al., 2019; Schüller et al., 2023) and Switzerland (Dietiker et al., 2012). The high variability in the proportion of cyclists using the carriageway with mixed traffic, advisory cycle lanes and bicycle pictograms shows that these facilities can achieve high compliance levels. At the same time, some studies show very low proportions, suggesting problems with perceived safety and/or comfort. The compliance level is significantly higher and variances are lower for the two dedicated cycle facilities, cycle lanes and cycle tracks. The mean values are lowest for mixed traffic and bicycle pictograms. Even marking an advisory cycle lane already leads to a significant increase of cyclists on the carriageway, which is even higher for cycle lanes and cycle tracks.

The wide range of compliance levels for mixed traffic provisions in Figure 1 indicates the existence of systematic determinants, which are mainly analysed using descriptive statistics. Factors related to lower compliance levels include high volumes of motorised vehicles (Ohm et al., 2015; Schüller et al., 2023; Zweibrücken et al., 1999), high shares of heavy goods vehicles, a speed limit of 50 kilometres per hour compared to lower values, the active use of the buildings adjacent to the street which might be a proxy for high pedestrian volumes (Schüller et al., 2023), the existence of on-street parking (Zweibrücken et al., 1999), and frequent school traffic (Zweibrücken et al., 1999). Higher compliance levels are found for study sites with higher cyclist volumes (Schüller et al., 2023).

Bicycle pictograms (sharrows) on the carriageway are used in several countries to increase the acceptance of cycling in mixed traffic and improve motorists' awareness of cyclists on the carriageway. Examples of countries using sharrows include the USA and Canada (AASHTO, 2012; MTO, 2021), although they have only been evaluated in the USA to date. Sharrows are also used in some cities in Germany, but are not regulated by law. Using German street sections as an example, Koppers et al. (2021) show that the proportion of cyclists on the carriageway without bicycle pictograms is on average around 50%, and with pictograms around 60%. Birk et al. (2004), Pein et al. (1999), and Hunter et al. (2010) also find an increase in the proportion of cyclists on the carriageway with the marking of pictograms, whereas Mills et al. (2010) and Vasilev et al. (2017) find no effect.

These studies provide important insights into the willingness of cyclists to comply with mixed traffic cycle provision, but there are still open research questions. The analyses in previous studies are mainly based on descriptive statistics and do not disentangle the complex interrelationships between the various relevant factors. Schüller et al. (2023) is the only exception that applies a linear regression model providing first indications of these interrelationships. In addition, the number of cases in the existing studies is often small, which might be one main reason for the exclusive use of descriptive statistics. Exposure data are limited, particularly for cycling.

These research gaps are of high relevance because we need to understand cyclists' on-street behaviour in order to design mixed-traffic streets that are perceived as safe and attractive, and therefore actually used by cyclists as intended. This means that cyclists ride on the carriageway and do not move onto the pavement to avoid being mixed with motorised traffic. Pedestrians benefit from such compliance because cyclist volumes on the pavement and related conflicts are likely to be reduced. Cyclists may also profit by being able to cycle at higher speeds on the carriageway.

This study combines the video data collected at 273 study sites in 13 German cities by Koppers et al. (2021), Ohm et al. (2015), and Schüller et al. (2023) into a unique dataset that allows the identified research gaps to be addressed. The following three research questions are formulated for this study: (1) What are the proportions of cyclists using the carriageway versus the pavement at specific types of street sections with mixed traffic in German cities? (2) What are the determinants of the probability for cycling on the carriageway in terms of exposure and design characteristics? (3) Which recommendations can be derived from this analysis for designing mixed traffic provisions for cyclists to achieve high levels of compliance?

To answer these research questions, detailed data on design characteristics, surrounding land use and volumes of motorised vehicles, cyclists and pedestrians are collected for each street section and combined with data on cyclists' general on-street behaviour and, in particular, their observed choices of whether to cycle on the carriageway or the pavement. In the following, we first describe the German legal framework and introduce criteria for mixed traffic provision from international guidelines in section 2. Section 3 introduces data and methods for this study, followed by the presentation of the results in section 4, which includes descriptive and model-based analyses. Results are discussed in section 5 as the basis for drawing conclusions and for developing recommendations for the future design of mixed traffic provisions for cyclists in section 6.

Mixed traffic in Germany means that no separate cycle facilities are provided on the street. Pavements are for pedestrians only and are not open to cyclists. Cyclists must cycle on the carriageway and share the space with motorised vehicles; cycling on the pavement is prohibited and is against the German Road Traffic Regulations. The only exception to this rule is made for children. According to section 2 (5) of the German Road Traffic Regulations (Leue & Bouska, 2018), children up to the age of eight must cycle on the pavement; they may use the pavement up to the age of ten. Children up to the age of eight may be accompanied on the pavement by a person of any age.

The guidelines for cycle facilities in Germany (FGSV, 2010) recommend mixed traffic provision with a speed limit of 50 kilometres per hour up to a volume of 400 motorised vehicles per hour and with a speed limit of 30 kilometres per hour up to a volume of 800 motorised vehicles per hour. Table 1 provides an overview of the criteria for mixed traffic provision for cyclists in urban areas according to international guidelines. The table shows that other countries, such as the Netherlands, are stricter than Germany and only recommend the mixing of cyclists and motorised vehicles on residential streets with low volumes (< 500 motorised vehicles per hour) and speeds (30 or 40 kilometres per hour) of motorised traffic. Only a few countries are identified that recommend mixed traffic provision at a speed limit of 50 kilometres per hour. These countries use lower thresholds for the volume of motorised traffic. Austria is the only exception where mixed traffic solutions are recommended up to 10 000 motorised vehicles per hour. The high thresholds for mixing cyclists and motorised vehicles in Germany raise questions about possible negative impacts on cyclists' safety and comfort but also provide an opportunity to examine cyclists' compliance with mixed traffic provision, as is done in this study.

The results of this study confirm and expand previous research. The large number of study sites made it possible to estimate a binary logistic regression model, taking into account the exposure of cyclists, pedestrians and motorised vehicles, as well as detailed infrastructure characteristics. In terms of exposure, the volume of motorised vehicles becomes significant, which is consistent with the descriptive findings from the literature on cyclists' compliance with mixed traffic provision (Koppers et al., 2021; Schüller et al., 2023; Zweibrücken et al., 1999) and also from the literature on cyclists' perception and evaluation of cycle facilities (Hu et al., 2023), and on cyclists' route choice (Rossetti et al., 2019).

Cyclist volume has a positive effect on the probability of cyclists using the carriageway, which we explain by individual cyclists feeling safer on the carriageway when they are part of a larger group. This indicates a positive self-reinforcing effect: Mixed traffic provision for cyclists in appropriate conditions increases the probability of cycling on the carriageway. When this increased probability leads to higher absolute numbers of cyclists on the carriageway, then cyclist volumes will in turn lead to higher compliance levels. These mechanisms must not be translated into guidelines recommending mixed traffic provision for higher volumes of motorised vehicles when cyclist volume is high, as this could lead to a negative downward trend: Some cyclists might avoid the carriageway because of the high volume of motorised vehicles, reducing compliance levels and thus the number of cyclists on the carriageway, etc. Instead, the positive effect of cyclist volumes on compliance levels should only be understood as a positive side effect of appropriate conditions for mixed traffic provision, which should be chosen to ensure high proportions of cyclists using the carriageway, which will then have a positive self-reinforcing effect, increasing cyclist volumes and thus again the proportion of cyclists on the carriageway.

Pedestrian volume does not become significant in the model. Still, due to the (moderate) correlation between pedestrian and cyclist volumes in our data, we cannot exclude that pedestrian volumes also influence the probability of cyclists using the carriageway. Previous survey-based studies (Hu et al., 2023), as well as descriptive findings from on-site studies on cyclists' compliance with mixed traffic provision (Schüller et al., 2023; Zweibrücken et al., 1999), suggest such effects of pedestrian volume, finding higher proportions of cyclists using the carriageway on streets with actively used buildings (e.g. shops), which may be a proxy for higher pedestrian numbers on the pavement. It appears that motorised vehicles push cyclists onto the pavement and pedestrians push cyclists onto the carriageway, which is plausible but further research based on data with a lower correlation between pedestrian and cyclist volumes is needed to confirm this hypothesis and determine the extent of these opposing effects.

In terms of infrastructure characteristics, the probability of cycling on the carriageway is highest on narrow profiles with a width of less than three metres compared to the two groups of three to 3.5 metres and with more than 3.5 metres. A possible explanation could be that overtaking manoeuvres become more frequent as lane width increases (Ohm et al., 2015), which might be perceived as more uncomfortable. The low explanatory power of lane width in the model should be noted, meaning that the above explanations describe trends that would need to be validated in further research. Pavement width has no significant effect on the proportion of cyclists on the carriageway in our data, which might be due to interactions with pedestrian volume.

No significant effects are found for other infrastructure characteristics, such as the median (marked or raised), the presence of tram tracks and parking. However, findings from the descriptive analyses, as shown in Table 3, indicate trends for the median and the presence of parking. Study sites without a marked or raised median have a higher probability of cyclists using the carriageway, which could be explained by an interaction with lower numbers of motorised vehicles and lower lane widths. The only correlations found in our data are between the presence of any type of median and volumes of motorised vehicles, which is higher at study sites with marked or raised medians, and a higher volume of motorised vehicles negatively affects the probability of cycling on the carriageway, as shown.

The presence of parking is not significant in the model but is associated with a higher probability of cyclists using the carriageway in the descriptive analysis in this study, contrary to the results of Zweibrücken et al. (1999), who find lower compliance levels in street sections with parking spaces on the carriageway, and also to studies investigating cyclists' perceived safety (Chataway et al., 2014; FixMyCity, 2020). Possible explanations are that parking is associated with more intensive use of the buildings, which is used as a proxy for pedestrian volumes in some studies and is correlated with the presence of shops and pedestrian volumes in this study. It has also been shown in the literature that the presence of parking reduces the average speeds of motorised vehicles (Schüller, 2010). It is also conceivable that parking visually narrows the width of the carriageway. This could lead to fewer overtaking manoeuvres and, in turn, lower average speeds of motorised vehicles (Schüller, 2010). In addition, vehicles parked on the carriageway might be a barrier for cyclists, preventing them from moving onto pavement. It should also be noted that around 30% of accidents on street sections with mixed traffic are related to parking (Hantschel, 2022).

The highest explanatory power in terms of operational characteristics is found for the speed limit. A speed limit of < 50 kilometres per hour (40 or 30 kilometres per hour) significantly increases the probability of cycling on the carriageway in mixed traffic conditions, which is consistent with the literature. For example, Rossetti et al. (2018), Rossetti et al. (2019) and FixMyCity (2020) find negative effects of high speeds of motorised vehicles on cyclists' route choice and on cyclists' perceived safety.

Markings with bicycle pictograms significantly increase the probability of cyclists using the carriageway compared to mixed traffic without any markings, which is consistent with the literature (Birk et al., 2004; Hunter et al., 2010; Koppers et al., 2021; Pein et al., 1999) but should be interpreted with caution, as the number of street sections with pictograms is very low at 13. Conclusions from these consistently positive effects of pictograms on cyclists' compliance with mixed traffic provision in the literature and this study should be carefully considered. In principle, all streets should be designed so that all users, including cyclists, use them intuitively and in accordance with the traffic rules.

City type is an interesting variable that has not been considered before. Streets in upper-level centres show a significantly higher probability of cyclists using the carriageway than those in mid-level centres. A possible explanation for this might be that the modal share of cycling is generally higher in upper-level than mid-level centres (Gerike et al., 2020). This may lead to a higher presence of cyclists in the city (regardless of the cyclist volume at the considered study sites). A possible explanation for this might be that the modal share of cycling is generally higher in upper-level than mid-level centres (Gerike et al., 2020). This may lead to a higher presence of cyclists in the city (regardless of the cyclist volume at the considered study sites), which in turn could mean that cyclists are more familiar with motorised vehicles and vice versa. The higher modal share of cycling in upper-level centres also means that more motorists do also cycle, which might have a positive impact on cyclists' view of motorists and vice versa, which has been shown to be relevant in previous studies on how aggressive both user groups behave in the street (Kaplan et al., 2019; Piatkowski et al., 2017).

This study is based on on-site observations; one limitation is that we do not consider personal characteristics. Age, gender, cycling practice, and earlier incidents experienced by cyclists are shown in previous studies to impact how safe cyclists feel. They can also be assumed to affect how likely cyclists are to use the carriageway. The consideration of coping strategies is limited to whether or not cyclists use the carriageway. Further possible coping strategies, such as choosing other modes of transport, alternative routes to avoid street sections with mixed traffic provision or other departure times for a trip to avoid peak-hour traffic, are not considered in this study focusing on on-site behaviour. The high correlation between pedestrian and cyclist volumes is another limitation of this study. Previous studies have found that higher pedestrian volumes negatively influence cyclists' perception and compliance with mixed traffic provision. This study finds that higher cyclist volumes positively influence the proportion of cyclists using the carriageway. Further research based on data with lower correlations between pedestrian and cyclist volumes would allow these effects to be disentangled. Studies in countries with different legal frameworks would be interesting to better understand the relevance of the legal requirement to use the carriageway in Germany compared to cyclists' perceptions of the street environment.

It should be noted that this study only considered people who already cycle. It would be interesting to assess how cycle-friendly mixed traffic provisions affect the attitudes and/or concerns of ‘interested but concerned’ cyclists.

The results of this study provide new insights into the complex interdependencies between the different factors influencing cyclists' behaviour in mixed traffic conditions. Based on a case study in Germany, where guidelines allow mixed traffic for cyclists on streets with a speed limit > 30 kilometres per hour and where cyclists are requested by law to use the carriageway, this study is to our best knowledge the first to quantify the effects of exposure and street characteristics on cyclists' compliance with mixed traffic provision.

Returning to the three research questions, this study finds an average proportion of cyclists using the carriageway of 80% across all 273 study sites, which is relatively high compared to the literature presented in Figure 1, where mean proportions of appr. 65% are found for mixed traffic provision and appr. 85% for advisory cycle lanes. One reason for this difference might be that the studies in Figure 1 include 2-lane and 4-lane street sections, whereas this study only analyses 2-lane street sections. Other possible reasons include different countries with specific legal frameworks, small case numbers in some studies, and city size and type. On the one hand, the value of 80% is low considering that cycling on the carriageway is a legal requirement in Germany for mixed traffic. It means that one in five cyclists does not obey the traffic rules even though cyclists in Germany generally have a good knowledge of the legal framework (Schüller et al., 2023); they know where they are expected to cycle. On the other hand, it can be assumed that the cyclists in this study use the carriageway because they feel sufficiently safe and comfortable but also because they are legally obliged to do so. There may be cyclists in our sample who would prefer to cycle on the pavement but do not do so to avoid violating the traffic rules. The proportion of cyclists using the carriageway in mixed traffic provision might be even lower in countries where cyclists are free to choose where they cycle.

In terms of the determinants of the probability of cyclists using the carriageway, the high relevance of volume and speed of motorised traffic found in this study is consistent with previous studies that are mainly survey-based and analyse cyclists' perceptions of different street environments as well as route choices. Previous studies consistently show that cyclists feel unsafe on the carriageway with high volumes of motorised vehicles sharing the same space. This study finds that cyclists are less likely to use the carriageway when it is heavily loaded with motorised vehicles. It thus describes the consequence of the negative perceptions found in previous studies. Hardly any studies have considered the effect of cyclist volumes. The positive relationship between cyclist volumes and the probability of cyclists using the carriageway in this study positively reinforces compliance levels in streets with mixed traffic provision under appropriate conditions. Speed limit, lane widths and bicycle pictograms are the significant infrastructure characteristics in the model, while the presence of a median (marked or raised), tram tracks, or parking shows no significant effect on compliance levels. City type as a proxy for city size and city characteristics in terms of transport provision and travel behaviour of residents is also significant.

The developed model is directly applied to derive recommendations for mixed traffic provision, which can only be recommended up to a maximum volume of 400 motorised vehicles per hour and a speed limit of < 50 kilometres per hour, with narrow profiles being preferred. These thresholds are necessary to achieve a 90% probability of cycling on the carriageway; they can and need to be adjusted according to political preferences. These recommendations are in line with those of the Netherlands, Great Britain, France and Ireland (CROW, 2016; Cerema, 2020; DfT, 2020; NTA, 2023) and call for a clear hierarchy in the cycle network: Mixed traffic designs should only be provided for cyclists where the volume and speed of motorised vehicles are low. Separate cycle facilities should be provided where volumes of motorised vehicles are high and/or motorised vehicles travel at higher speeds.

Overall, this study contributes to the comprehensive understanding of the mechanisms of cycling provision, cyclists' perceptions and behaviour. It shows that mixed traffic provision for cyclists on main streets can achieve high compliance levels, it shows the key determinants of cyclists' willingness to comply with mixed traffic provision, and it also shows the consequences when exposure and infrastructure characteristics do not match cyclists preferences and perceived safety.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for profit sectors.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.