A quick scan of the available literature was carried out in order to identify search terms. The list of search terms increased during the search for literature and includes the following 32 terms (including several translations of bicycle streets in other languages), see Table 1.

Several resources were used to find studies: Science Direct, Google Scholar, SWOV library, CROW-Fietsberaad publications, Google, and reference lists of found studies. Note that the literature review was part of a more extensive literature review presented in a report published by SWOV (Dutch Institute for Road Safety Research) (Nabavi Niaki et al., 2023), which also includes a discussion of local design manuals and guidelines. In this study, however, we focus solely on the safety-related studies. In total, the search terms led to almost one million hits. For each search term, and only for the relevant hits (generally the first 30 to 50 hits), the titles and abstracts were scanned. This scan led to a list of 90 most relevant studies, chapters, reports, and websites of which the abstracts were read more thoroughly. Based on the review of the abstracts, 67 hits were indicated as most relevant related to the topic of design and safety of bicycle streets. Since the number of safety studies related to bicycle streets is limited, grey literature is used in addition to scientific papers.

In general, subjective safety of bicycle streets can be investigated relative to the subjective safety on other cycling infrastructure types. For example, the study byvon Stülpnagel et al. (2022) in Munich, Germany, compared different types of cycling infrastructure based on both objective and subjective safety. To examine the subjective safety, cyclists were asked to add crashes, near-crashes, and conflicts on a map. These data were compared with official reported crash data, bicycle volumes, and road characteristics (type of cycling infrastructure and speed limit). Compared to other types of cycling infrastructure, bicycle streets have a lower subjective and objective risk (corrected for bicycle volume), especially compared to roads without cycling infrastructure.

Focussed on the safety of children in particular, the results in the study byVanZerr (2010) are contradictory. A survey was carried out among residents living close to a bicycle street in Portland, Oregon (USA). The questionnaire contained one safety-related question, about the observed safety for children. Results were contradictory: 37% believed that bicycle streets have a positive impact on the safety of children, 22% believed there is a negative impact, and 41% indicated no impact on the safety of children. However, one must be careful in drawing conclusion based on this study, since it is difficult to rely on the results from one survey question only.

Table 2 shows a summary of the most important factors that affectboth subjective and/or objective safety of cyclists on bicycle streets. The table illustrates that the most important factor is the presence of motorised vehicles. On the one hand, the design of bicycle streets or bicycle volume on those streets minimise the negative impact of motorised vehicles, since some studies found a lower volume and/or speed. On the other hand, other studies found that the motorised vehicle volume and speed are still (too) high on some streets. Additionally, the overtaking behaviour by drivers of motorised vehicles is often found as a negative factor since they overtake cyclists too frequently and closely. Conversely, other drivers avoid to take over and drive to closely behind the cyclists, which is experienced as tailgating. On top of that, motorised vehicles standing on the carriageway (e.g. to load or unload) negatively affect both the objective and subjective safety. On top of that, it is noteworthy that the impact of bicycle streets on cycling safety may be context-specific, since part of the differences in the results addressed in this literature review may be explained by the fact that various study areas are considered. For example, the cycling and driving culture and the infrastructural design in North American cities may strongly differ from European cities, and, for example, specifically from the Dutch context.

 It is also important to stress that the literature that examines the impact of specific design elements on safety is limited. The most important design element identified in the literature is the width of the carriageway, since width has a strong impact on the speed and overtaking behaviour of drivers of motorised vehicles. A wider carriageway may invite for speeding, while a too narrow carriageway may lead to dangerous overtaking manoeuvres. The effective width of the carriageway is decreased by implementing too wide rumble strips. Other design elements that affect the safety of cyclists, mainly discussed in the subjective safety literature, are a red coloured carriageway, ‘Cars are guests’ signage, and corresponding road markings. Studies that discuss the design of bicycle streets in relation to objective safety are even more limited and if they do, often only the traffic calmed character of bicycle streets is mentioned.

The data collection took place at the 11^th International Cycling Safety Conference 2023 (ICSC 2023) in the Hague, the Netherlands, during a workshop from 09:00 to 12:30 about the safety of bicycle streets. This workshop started with an introduction, followed by presentations by CROW (Dutch knowledge platform that develops practical guidelines for infrastructure and transport) and University of Freiburg about safety of and design guidelines for bicycle streets in the Netherlands and Germany. Afterwards, the 49 participating experts were randomly divided over ten groups of about five persons. Table 3 shows there is a slight overrepresentation of male participants (57%) and that most participants are from institutes in Europe (82%), in particular from Germany (24%) and the Netherlands (18%). Furthermore, most of the participant are from a university (67%), followed by research institute (14%).

The assessment to fulfil by the groups was to categorise nine international examples of bicycle streets in three piles: 'Safest', 'In between', and 'Least safe'. Each pile must consist of three bicycle streets, which are entered in LimeSurvey by the groups. Per group, experts were also asked to provide a maximum of two arguments per bicycle street about the reason of putting this bicycle street in a specific pile. These arguments were written down on a form that was handed out together with the printed images of the bicycle streets. The printed images included a text box with information about the city and country of the bicycle street on the image.

The nine examples of bicycle streets come from Odense (Denmark), Palo Alto (USA), Luxembourg (Luxembourg), Dresden (Germany), Zurich (Switzerland), Jyväskylä (Finland), Utrecht (the Netherlands), Berlin (Germany), and Kalmthout (Belgium). The images of these examples are presented in Section 4.1.2 and originated from online news articles and official documents about bicycle streets (e.g. design manuals from local governments) that were looked up on Google Street View. Note that in the instruction for the experts, the bicycle and motorised vehicle volumes are fixed across the examples, all with a fictive peak hour flow of 120 cyclists per hour and 60 cars per hour. The reasoning behind this fixed volume is that volumes may vary substantially across the international examples and to prevent that participants are too distracted by the volume on the images. Moreover, it is assumed that, since bicycle streets are specifically designed to prioritise cyclists, all of the nine examples have a considerable amount of cyclists relative to the standards of the city or country of origin and that, at least during the peak hours, the cycling volumes are higher than the car volumes. For the exact volumes, it was argued that they should not be unreasonably high (like Dutch or Danish standards) or unreasonably low (like in countries with very low cycling rates). At the beginning of the group work during the workshop it was once more stressed that participants should not be distracted by the volume on the image and to keep in mind the fixed volume. All other characteristics to assess are left open in order to provide that participants are free to decide for themselves how they assess the safety of bicycle streets.

To come to a final safety score per bicycle street, every time a bicycle street is put in the ‘Safest’ pile it got 3 points, for the ‘In between’ pile it got 2 points, and for the ‘Least safe’ pile it got 1 point. These scores were summed to get a final score per bicycle street. Additionally, all written arguments were collected in one document and it was assessed whether an argument is positive, neutral, or negative. This was aggregated per bicycle street, which shows if a bicycle street receives primarily positive, neutral, or negative arguments or whether the arguments are more mixed. Moreover, the arguments are compared per bicycle street to find out whether these mention similar design elements to put a bicycle street in a specific pile. Lastly, to examine which design elements are important to cycling safety professionals for assessing the safety of bicycle streets, all arguments were categorised. In order to illustrate which design elements are most important, the categories were aggregated to get a frequency per category. These safety related design elements were compared to existing literature in order to provided future research directions.

Besides evaluating which arguments are given to assess the safety per bicycle street, the arguments were categorised. Such categorisation helps to examine which design elements are important to cycling safety professionals to assess the safety of bicycle streets. Note that for this categorisation, the direction of the arguments (positive, neutral, negative) is ignored such that the categories just show what kind of design elements cycling safety professionals look out for. This is due to the fact that the design elements itself have no ‘direction’, as, for example, in terms of ‘width’, a bicycle street can be too narrow or too wide, but the element of interest is ‘width’. Table 5 shows the categorisation and frequency of the arguments.

The two most frequently mentioned design elements are the ‘width’ of the street, and that it should be recognised from the design that cyclists have priority over motorised vehicles (cycling priority design). Almost as often mentioned are ‘road markings’ and ‘dooring’. Road marking relates to proper indication of a bicycle street. The possibility of dooring relates to the risk of having a crash as a result of drivers opening their door when being parked parallel to the street.

Slightly less frequently mentioned are the orientation of ‘parking (i.e. parallel vs. perpendicular)’, and ‘speeding’. The orientation of parking is also related to dooring, since dooring is only possible with parallel parking. When parked perpendicular, the professionals are concerned about sight obstruction of other parked cars, street furniture, or vegetation and about backwards entering the bicycle street. Speeding is mentioned in relation with width, since wider streets may provoke speeding. The following design elements are ‘signage’, ‘intersection design’, and ‘parking (i.e. number of cars)’. Signage relates to the presence of a sign that indicates that the street is a bicycle street as well as signs to indicate the speed limit, intersection design relates to how complex an intersection is for cyclists and sometimes also pedestrians, and parking to how many cars are parked along the bicycle street.

The category ‘other’ contains various arguments which were difficult to attach to the other categories. It includes temporary loading zone, volume of cyclists, sight obstruction due to road works, facilities for pedestrians (e.g. crossing facilities), perceived safety, and that the bicycle street is complicated. Some arguments are about the ‘road surface’, whether it is red coloured and whether it is smooth. Only three arguments are about the attractiveness of the environment surrounding the bicycle streets, such as vegetation.

In the eyes of the groups of cycling safety professionals, the most frequently mentioned characteristic to assess the safety of bicycle streets is the width of the carriageway. Width on the one hand can increase the safety of a bicycle street, because narrow streets help to decrease the speed of motorised vehicles, while on wide streets the opposite may occur. This is in line with existing studies, showing that width is an important factor for both the objective and the subjective safety of bicycle streets. In terms of subjective safety, cyclists experience the speed of cars more negative on wider streets and narrower streets may slow down cars (CROW-Fietsberaad, 2021; Jörgensen, 2020). Narrow streets, on the other hand, impact the experience of a bicycle street negatively for side-by-side cyclists, caused by motorised vehicles being unable to overtake (CROW-Fietsberaad, 2021). Studies of objective safety of bicycle streets also found width as an important factor, but that the impact of width is highly correlated with motorised vehicle volume. Width and volume together determine how many critical encounter occur on bicycle streets (CROW-Fietsberaad, 2021). In addition, it was found that, compared to narrower bicycle streets, wider bicycle streets more provoke for speeding (Delbressine, 2013).

Cycling priority design is related to what extent it is clear that cyclists have priority over motorised traffic. Two studies indicate that a ‘cycling priority design’ increases the subjective safety of cyclists (Baert et al., 2021; von Stülpnagel et al., 2022). In terms of objective safety, Goldenbeld and van Schagen (1997) show that car drivers adapt their behaviour in favour of cyclists on newly implemented bicycle streets. Similarly, Schaefer et al. (2022) show that, after a decrease in speed limit, the driven speed is reduced more on bicycle streets than the same decrease in speed limit on regular arterial roads. Presumably, traffic calming measures on the bicycle street play a role here; however, it is unknown what these traffic calming measures are. Note that both road markings and signage also apply on ‘cycling priority design’, but since these are assessed separately by some of the groups of experts and in existing studies, these are discussed in separate sections (5.1.3 and 5.1.6).

Another often mentioned characteristic is road markings. The groups of cycling safety professionals paid attention to whether the bicycle street is properly marked with marking indicating it is a bicycle street. In existing studies of subjective safety, Swedish and Belgian studies found that road markings with bicycle symbols improve the perception of safety of cyclists (Baert et al., 2021; Jörgensen, 2020; Rivera Olsson & Elldér, 2023). In terms of objective safety, Teschke et al. (2012) found a lower injury risk on local roads that are designed as bicycle routes by means of road markings that indicate the road as a bicycle route. Moreover, Schaefer et al. (2022) found a higher reduction in driven speed as a result of a decrease in speed limit on bicycle streets compared to regular arterials. This may be achieved as a consequence of both traffic calming measures and road markings to indicate that the road is shared with cyclists.

Parking is divided over three factors: dooring, the orientation of parked cars (i.e. parallel vs. perpendicular), and the number of parked cars. The chance of being hit by a opened door is found to be mentioned most by the cycling safety professionals to assess the safety of bicycle streets. Therefore, it is even more striking that no existing studies of the safety of bicycle streets found in the literature review (Section 2.3) specifically evaluates the consequences of providing too limited space at parking lots along bicycle streets for drivers to open their door. However, it is known from other studies that opening of doors by drivers increases both crash risk and injury risk of cyclists (Hagemeister & Kropp, 2019; Johnson et al., 2013; Jänsch et al., 2015; Pai, 2011; Schimek, 2018).

Most groups that mention the orientation of parked cars argue it is safer when cars are parked in parallel direction rather than perpendicular. However, none of studies referred to in the literature review (Section 2.3) evaluates the impact of the direction of parked cars. Nevertheless, one can argue that with perpendicular parking, sight of the car driver may be obstructed by other parked cars, especially when the car is leaving the parking bay in backwards direction (Cicchino, 2019). On the other hand, the abovementioned studies show that dooring is an issue with parallel parking.

Lastly, the number of parked cars seems an issue for some of the groups of cycling safety professionals. It is argued that more parked cars may lead to more conflicts. This may be related to the turnover rate: the number of cars parking or leaving a parking spot. Although not specifically related to bicycle streets, existing studies found increased crash risk with a higher number of parked cars along the road (Greibe, 2003) and the parking movement itself is also found to be risk increasing to cyclists (Vandenbulcke et al., 2014).

In terms of speeding, it is argued by the groups of cycling safety professionals that speeding is an issue on some of the bicycle street examples. This is often related to the fact that some examples provoke speeding due to their width or the absence of a cycling friendly design. Several existing studies also pay attention to speed in relation to the safety of bicycle streets. High speeds of cars at bicycle streets are perceived as negative in terms of subjective safety, especially on wider roads (Baert et al., 2021; CROW-Fietsberaad, 2021; Jörgensen, 2020). Moreover, in objective safety studies it is also observed that speeding can be an issue on bicycle streets and that this indeed is related to the width of the carriageway (Delbressine, 2013). In Norway, the implementation of a bicycle street had no effect on the driven speed of cars (Fyhri et al., 2020). However, when comparing the speed on bicycle streets after a speed limit reduction to the same reduction on regular arterials, it is found that the driven speed decreased more on the bicycle streets (Schaefer et al., 2022). Besides speeding of cars, speeding of (light) mopeds on bicycle streets in the Netherlands is addressed in Section 2 Literature review (Delbressine, 2013). Given the increased popularity of new, electrically powered micromobility modes, speeding is a topic that should be further explored related to bicycle streets since several studies show that mixing of these micromobility modes with cyclists or pedestrians leads to increased numbers of conflicts and crashes (Porojkow & Lißner, 2024; Šucha et al., 2023).

Signage is an issue that some groups of cycling safety professionals looked at. In most cases, the presence of signage is argued as something positive, but sometimes the signs are unclear. Most signs indicate that ‘Cars are guests’ on a bicycle street. Accordingly, one of the studies from the literature review (Section 2.2.1) found a positive effect on subjective safety of the sign with ‘Cars are guests’ (Jörgensen, 2020). The respondents from that study indicated that mixing cyclists with motorised vehicles is best possible when, amongst others, there is a sign with ‘Cars are guests’.

Intersection design is also mentioned by some groups of cycling safety professionals to assess the safety of bicycle streets. Several elements of intersections are discussed, such as complexity, road markings and corner islands. In the studies from the literature review (Section 2.3), however, intersections are discussed in only one study. That study found, compared to main roads without cycling infrastructure, a lower injury risk on local roads that are designed as bicycle routes and therefore have separate traffic lights for cyclists at intersections with main roads and road markings to indicate the road as bicycle route (Teschke et al., 2012). For this characteristic it is important to note that not all images of the bicycle streets contain intersections or are focussed on an intersection, which might affect the importance of this characteristic.

The last characteristic to discuss is road surface, but the importance of surface to assess the safety of bicycle streets is limited in the eyes of the groups of cycling safety professionals. Road surface is mostly about the colour of the pavement or the type or smoothness of the pavement. A red, smooth road surface made from asphalt is most in favour of cyclists, since the street can be identified as a cycling friendly environment.Baert et al. (2021) argue that a red surface helps to recognise the street as bicycle street. In Jörgensen (2020), respondents argue that a red pavement can help to decrease the speed of motorised vehicles.