A Survey of Visual Impression of Urban Space: Findings from Empirical Studies, Challenges and Opportunities

1. Introduction

Predicting how an urban space will be experienced by humans can enable us to align our designs with the needs of users, and promote well-being in the built environment. As vision plays a dominant role in our interaction with the environment [1], understanding the relationship between objective visual features of the environment and the subjective experience of users is of great importance to urban design research. Verbal language is a powerful medium for conveying one’s subjective experience to others. Therefore, studying how users respond verbally to what they see is important for the progress of urban design. The primary aim of this article is to organize recent empirical findings concerning users’ visual experience of urban spaces, as captured in verbal form.

According to Norman, who famously stated that “attractive things work better”, the visual experience of artifacts goes beyond aesthetic pleasure, and contributes to the success (or failure) of designs [2]. In the context of urban spaces, the importance of considering the user’s experience as a critical part of the design was highlighted in early works, e.g. by Lozano who specifically discussed visual factors [3], by Williams who emphasized their aesthetics [4], among others.

Since then, researchers have been constructing a growing body of knowledge concerning how users perceive, interpret and experience urban space. Common areas of exploration include wayfinding [5], visual interest [6,7], aesthetic experience [8], neurocognitive response [9] ,and association of visual features with linguistic descriptions [10] (the latter being the focus of this study).

Past studies have produced valuable findings regarding various links between visual features of spaces and the way they are experienced by users. For example, Ohno has investigated the “frame effect” in buildings and found that the shape of an exit opening can affect the descriptions assigned to the view framed by it [11]. Further, Nishio and Ito have found that the degree to which an outdoor environment feels “calm” can vary by the sky factor (the degree of daylighting) [12]. Synthesizing these findings could be useful for various purposes, such as creating design guidelines. However, the wide range of studies conducted obscures the underlying structure of the field, and thus poses challenges for cross-study syntheses. To address this challenge, we are systematically reviewing works of this subfield.

In this work, we review literature that reports on links between visual features of urban spaces with users’ impressions. To structure our review, we have formulated three research questions: (RQ1) What aspects of visual impression does research aim to understand? (RQ2) How has visual impression in urban space been empirically studied? (RQ3) What relationships between visual features of the urban environment and users’ experience have been established?

2. Materials and Methods

To answer our research questions, we executed a systematic literature review using PRISMA — a widely used protocol for conducting literature reviews [13] which increases robustness and reproducibility. PRISMA has been applied in a wide range of domains, such as psychiatry [14], biology [15], computer science [16,17], pedagogy [18,19], and design [20,21,22]. PRISMA reviews are conducted in two major steps of (1) record selection, and (2) record classification and synthesis. We elaborate on each of these steps below.

2.1. Selection

To retrieve a set of candidate records for screening and determination of eligibility, we searched the complete Web of Science Core Collection dataset which includes over 90 million records to date (March 2026). Our search string was devised to retrieve any record which reports on an empirical investigation of visual impression in the context of urban space, while excluding articles that focus on other modalities, on the aspect of deriving impression from memory without the aid of visual stimulus, or on design pedagogy. Additionally, we refined the search by focusing only on articles written in English which are publicly available (open-access) and are in relevant research areas. The search yielded a total of 199 records for further review and screening. The details of the search procedure are summarized below:

• Record Type: Article
• Language: English
• Availability: Open access
• Research Areas: Areas with low relevance (surgery, meteorology, etc.) were excluded
• Search String: ("impression" OR "perception" OR "experience") AND ("urban" OR "city" OR "cityscape" OR "streetscape" OR "built environment" OR "built-environment") AND ("empirical" OR "experiment" OR "survey" OR "questionnaire" OR "interview" OR "user feedback" OR "user response" OR "field study" OR "observational study") AND (NOT ("auditory" OR "sound" OR "memory" OR "tactile" OR "education" OR "educational" OR "pedagogical"))

Next, we screened the collection of retrieved papers manually by reading their titles and abstracts, using the following inclusion criteria:

• The record has not been retracted
• The record is a research article (not a review article)
• The record contains a descriptive scientific contribution (as opposed to normative papers which propose new indices, methods, etc.)
• Empirical data regarding how users experience urban space was analyzed
• The focused aspect of the environment is its visual quality (as opposed to auditory quality, etc.)
• Visual impression is captured by asking users to report it, i.e. it is not derived based on physiological indicators such as EEG, etc. Note that, papers which investigated physiological factors alongside verbal descriptions, such as Chinazzo et al. [23], were included
• The user group investigated is representative of the general population (not declared to be domain experts) and are physiologically healthy (not visually impaired)

Only articles which fulfilled all criteria were included, and the rest (N=47) were screened out, resulting in (N=72) records for further consideration and determination of eligibility .Eligibility of the remaining records was determined by reading the full text and confirming that the study fulfills all the following criteria:

• Empirical results are reported concerning the visual impression of human users
• The empirical results establish a relationship between visual elements and users’ subjective feedback (reported in any form)
• The stimulus presented to users represents an urban environment, in a broad sense. For example, environments defined as “urban forests” were included, while those that were defined as “nature” without reference to an urban environment were excluded.
• If the investigation was done in a lab setting, the stimulus presented to users was realistic (e.g. photograph, immersive model) and not an abstraction (e.g., line drawing representing the environment or a visual pattern in it)

Using these criteria, we excluded approximately one third of the remaining records (N=26) and identified a collection of highly relevant records (N=46) for a detailed review, classification and synthesis, as reported below. The process of paper selection is summarized below in Figure 1.

2.2. Classification &Synthesis

Classification of the selected records was done by four independent researchers, following the method proposed by Ucci et al. [24], to increase robustness and ensure validity [25]. The four classifiers were split into two pairs according to their time zones, to ease communication. Each pair classified exactly half of the records (N=23), and each record was classified twice – once by each team member. This enabled us to evaluate inter-classifier reliability, and to minimize bias in the results of the classification process.

Each record was classified by considering according to three dimensions: (1) the purpose of the work, (2) the method used, and (3) the primary findings reported. These dimensions correspond with our three research questions. The information extracted from each record is summarized below in Table 1.

Following the classification of all records, synthesis was conducted in two steps. First, synthesis was done within teams, by resolving intra-team differences in classification. In this, discrepancies were discussed in both asynchronous and synchronous settings, and the final arbitration authority rested with the first author. Second, intra-team differences were resolved by adding or merging categories, as needed, aiming to produce a classification that was (1) as simple as possible, (2) inclusive of the findings in all of the records surveyed.

3. Results

This section is organized into three subsections (3.1-3.3), each addressing one of our research questions (RQ1-RQ3). Regarding inter-rater reliability, we report on agreement rates for two of the classification tasks which employed discrete choice, and can therefore be quantitatively evaluated. Inter-rater reliability was first calculated within each team and then across teams. (Table 2).

3.1. Foci of Past Works (RQ1)

The selected studies can be classified into three major types based on their research focus (i.e. the kind of information that the work attempts to extract):

• Studies investigating users’ preference concerning the visual qualities of a space
• Studies investigating users’ impression as reflected in descriptions attributed to a given artifact or environment.
• Hybrid studies which combine 1 and 2

We organize the surveyed records into these three categories below. For each category, we provide a concise characterization of typical studies, as well as an overview of the classification outcome in the form of a table. An overview of the distribution across these categories is given in Figure 2.

Studies of type 1 attempt to answer the following question: given an artifact (or a set of artifacts), would users consider it preferable? While some studies explicitly indicate that their focus is on studying preference, e.g. [26], terms that strongly imply preference such as “desirability” of a design are included within the category as well. Table 3 summarizes the studies focused users’ preferences, which shows that researchers are studying preference across a wide range of scales, from specific objects such as green roofs [27] or wind turbines [28], and up to large outdoor environments such as urban forests [26] or streetscapes [29]. A notable portion of the studies focused on green spaces or elements, as illustrated in Figure 3.

Studies of type 2 attempt to answer the following question: given an artifact (or its visual representation), how would a human user describe it verbally? When investigating the ways that we describe designs, researchers may focus on associating an urban space with a single specific description, e.g. “safe” [38] or with multiple description which may or may not be directly linked, e.g. [“safe”, “lively”, “beautiful”, “wealthy”, “depressing”, “boring”] [39]. The studies investigating users’ experience in terms of verbal descriptions are organized in Table 4. The table includes the details for each study, the object which the user was required to view or consider, and the specific aspect emphasized in the investigation. Studies investigating multiple aspects of the human visual experience were classified as “diverse”.

Studies of type 3 attempt to answer questions that pertain both to the general experience of a design and to specific preferences. Therefore, participants in these studies are requested to associate a verbal description with a visual object, and to assess the desirability of the object. Studies in this category are organized in Table 5.

3.2. Approaches for Studying Visual Impression of Urban Space (RQ2)

Visual impression in urban space is being studied using a diverse set of approaches. We characterize these by focusing on the materials used as stimuli, the procedure for data collection and the analysis of the data. An overview of the types of stimuli used and the different methods employed for data collection is given in Figure 4.

3.2.1. Visual Stimuli

The most common material used as stimuli is two-dimensional photographs of urban spaces shown to participants, which have been used in over half of the studies (N=28). Within this group, one quarter of the studies (N=7) have also employed photomontage techniques using software such as Adobe Photoshop, to produce design alternatives, e.g. [34].

Second are studies that used the actual space as the stimulus, commonly referred to as studies “in-the-wild” (N=12), e.g. [60]. A few studies conducted in-the-wild employed a hybrid approach, which also included other stimuli, such as digitally produced drawings of alternatives [26], or a photomontage of the location the participant is situated at [64].

Third are studies that involved the display of virtual yet realistic stimuli. In one case, panoramic virtual images of real spaces were shown to participants in virtual reality [39]. In another case, a virtual model of an existing environment was constructed, and multiple alternatives were rendered [66], and in another a virtual model was constructed without a specific real-world referent (yet inspired by North-Italian landscape) [52].

In three cases, the investigation did not employ a visual stimulus. Despite this, they relied on past visual experiences and the visual impression of participants, and were thus retained in the review.

3.2.2. Method of Data Collection

Methods for data collection are divided into two major categories, following the distinction between the studies which investigate users’ preferences concerning a visual scene and those that investigate the descriptions that users may attribute to it (introduced in 3.1).

To extract users’ preferences, human subjects were requested to conduct four kinds of tasks. First, choice survey tasks in which users are presented with a set of design alternatives to choose from [29]. Second, rating tasks, e.g. using a Likert scale, in the form of a questionnaire administered in a laboratory setting [28] or in-situ [26]. Third, passive observation tasks, in which eye movements are recorded while viewing a stimulus, and linked to rating data [34]. Fourth, usage tasks, in which walking behavior is recorded and analyzed to infer the users’ subjective preferences from observable data, such as walking behavior [30].

When extracting descriptions assigned to visual scenes, four different tasks were used. First, rating tasks in which a subject evaluates the correspondence between a visual representation of a space and a verbal description, which can be done via a Likert scale, .e.g. [54] or via a semantic differential scale, e.g. [51]. Among these, one study also included questions regarding one’s attention to specific visual elements [47]. Second, matching tasks were administered. In these, a subject is asked to focus on a given verbal description, and selects the best match out of a set of visual representations, typically via pairwise comparison, e.g. [42]. The latter was also done without collecting new data from human participants, by utilizing existing open datasets [38] such as Place Pulse [72]. Third, in a single study, a sorting task was administered, which required subjects to group sets of photos based on shared attributes, then order them according to the extent they match a given description [44]. Finally, a single study has employed a semi-structured interview for extracting descriptions from users [55].

Studies that attempted to extract both preferences and descriptions have commonly used a combination of two or more of the above approaches. Additionally, one study involved its participants in the data collection activity by asking them to take photos that captured their impressions [56].

3.2.3. Verbal Description Studies

The complete set of verbal descriptions investigated (in applicable studies) is summarized in Figure 5 below. Closely related terms are grouped under a single term, e.g. “integrated” and “unified” are both included within “integrated”. Note that one study [48] has used the positive-negative affect schedule (PANAS) [73] and its 20 descriptors to extract visual impression, yet these referred to the individual rather than the spaces experienced (e.g. “scared”, “distressed”, etc.), and were thus excluded from this figure.

3.3. Current Empirical Findings Regarding the Visual Impression of Urban Space (RQ3)

3.3.1. Findings Concerning Preference

With respect to users’ preferences, we organize our findings into three types of studies, based on the overarching research question pursued. Findings for each type are organized in the form of a table.

Type (1) studies examine whether a specific visual element is desirable or undesirable in binary terms – e.g., does the existence of a visual element X (or its absence) correlate with positive or negative preference? Such studies answer the following question: if a certain element is added or removed, what would be the impact on users’ preference? For example, Hamza et al. found the presence of wind turbines in specific environments (landscape, seascape) desirable [28]. Current findings concerning the relations between presence/absence and preference are summarized in Table 6.

Type (2) studies report on hierarchies of elements in terms of desirability – e.g. is a visual element X preferred over another visual element Y? These studies answer the following question: given a set of elements to choose from, which one would be preferable from the perspective of users? For example, a study of urban green spaces has found that adolescents preferred recreational infrastructure over the biophysical environment [33]. Current findings concerning visual element identified as preferable among sets of elements are given in Table 7.

Type (3) studies examine the desirable amount or degree of certain elements – e.g., will a large amount of element X be preferable over a smaller amount of it? For example, focusing on urban forests, Buckwitz et al. have found a medium amount of ground vegetation to be desirable [26]. Current findings concerning the desirable amount or degree of elements are given in Table 8. For reference, results prior to synthesis, including sources, are given in Appendix A (Table A1).

The above findings can aid planners and decision-makers in designing urban spaces by pointing to desirable alternatives and by highlighting non-desirable ones which they may want to avoid. For instance, based on [27], planners of extensive green roofs can consider reducing gaps in vegetation, and removing weeds. At the same time, the above findings should be interpreted with caution, as the relationships reported may be associated with the specific group under investigation, and since findings occasionally conflict (discussed in 4.3).

3.3.2. Findings Concerning Description

With respect to users’ description of urban space, primary findings that link visual elements with descriptions were extracted and synthesized. These findings (Table 9) are organized on two levels – first by the description which was assigned to the visual environment (“safe”, “clean”) then by the type of environment studied. Findings reported here are those which do not display clear dependency on the specific group sampled.

Alongside these findings, works have reported on various factors which pose challenges in trying to establish robust knowledge concerning visual impression. First, context dependency of certain elements (e.g. vegetation, sky) was observed [38,43]. Second, asymmetries among different groups were noted. For instance, gender differences in using the description “safe” with respect to a streetscape were found (women evaluating the same streetscape as less safe than men) [41]. Third, visual elements may interact nonlinearly, e.g. greenery can mitigate the negative effects of dense buildings or roads, up to certain threshold [42]. Fourth, visual impression is affected by cognitive processes such as visual attention. Worse yet, attentional processes can be in-turn affected by past knowledge or experience – e.g. higher site familiarity was found to increase or redirect attention to children playing [47], and this experience was deemed as preferable/positive.

Finally, in addition to teaching us about positive/negative relations between visual factors and verbal description, studies have deepened our understanding of visual impression in at least three ways: (1) by identifying visual elements which are potentially neutral – e.g., in one study, hue did not play an important role in creating an environment which is visually comfortable [46]; (2) by helping us evaluate the extent that users’ impression have (or have not) matched the design intention [56]; (3) by demonstrating that certain types of features correlate more with a given description – e.g., Florio et al. have found bottom-up visual features (contrast, edges, low-frequency structure, attention-grabbing regions) to correlate more strongly with perceived complexity than top-down semantic features [44].

4. Discussion

This section is organized into four subsections. The first three correspond to our research questions (4.1, 4.2, 4.3). The last provides recommendations for future study of visual impression of urban space, and highlights the challenges involved in understanding the response of human users to their environment (4.4).

4.1. Characterization of Past Works Based on the Focus of the Investigation

Overall, three primary types of studies were identified, based on their goal:

Goal 1: Understanding users’ preference with respect to an urban environment

Goal 2: Understanding the descriptions assigned by users to an urban environment

Goal 3: Understanding both preferences and descriptions.

The primary difference between these studies is the question that they attempt to answer. Given a visual element X, type 1 studies are interested in finding out whether users will deem X desirable, type 2 studies are interested in finding out how users will describe it, and type 3 studies are interested in both.

This tripartite distinction is important since verbal feedback from users can contain descriptive information (which may be neutral) and information about their preferences. For example, three out of the four most commonly studied terms are ones that imply preference (“beautiful”, “safe”, “pleasant”). This means that such studies contribute to our understanding users’ preferences, even when preference is not stated as the focus of the study. By contrast, neutral terms (e.g. “layered”), which dominate the list of terms studied, do not necessarily contain information concerning users’ preference.

Therefore, we recommend considering these three goals when defining the scope of future studies on visual impression, to clearly communicate the research aims and contributions to readers.

4.2. How Have We Empirically Studied Visual Impression in Urban Space? (RQ2)

4.2.1. Stimuli

The primary stimulus was two-dimensional photos of the space under investigation. Studies conducted in the wild have emerged, but still consist of a little over a quarter of the total, despite the rise in such studies in other fields, e.g. in human-computer interaction [74]. To produce ecologically valid results, researchers can attempt to reproduce past results in real-world settings, or conduct new experiments in which the visual stimulus presented is the real object, rather than a representation of it. An alternative to the “in-the-wild” approach is to form realistic 3D virtual representations – a technique which is becoming increasingly feasible with reduced costs of immersive headsets. However, the assumption that our response to virtual and physical representations of the environment has recently been called into question [75]. Thus, it is important that studies which rely on virtual stimuli are validated by comparing users’ responses to virtual versions of the urban environment and their physical counterparts. Nonetheless, our ability to do so is limited, since one of the usages of virtual stimuli is to present users with counterfactual situations, i.e. ones which do not yet exist in physical space [66]. A low-cost intermediate solution is to produce counterfactual images then present them to participants in-situ [64].

4.2.2. Data Collection

Among the seven approaches identified, the least employed methods were interviews (used in a single study) and requesting participants to take photos of visual elements of interest, thereby enabling greater freedom in terms of attention and selection (also used in a single study). These approaches can complement highly structured tasks which are characteristic of quantitative studies. While the value of highly constrained tasks is clear (e.g. choice tasks facilitate comparability) they are limited at least in two ways: (1) given any verbal description, they assume that its interpretation is comparable among different individuals, yet linguistics tells us that this is not necessarily true [76,77]; (2) given the environment under investigation, visual stimuli that may have been of interest are excluded a priori during material preparation. Interviews can enable researchers to inquire into differences in their understanding of verbal terms and how they can be matched with urban environments [78], thereby helping to address (1). Encouraging users to take photos and participate in the collection of visual stimuli can help reduce bias when selecting visual elements for study. Photo collection may be further structured by methods such as “photo diary” [79], in which the viewer both collects the visual stimulus and comments on it at different levels (both objectively and subjectively). Since both interviews and photo collection are costly, however, they cannot be expected to replace other data collection approaches, but rather to support and enrich them.

4.2.3. Verbal Descriptions Investigated

The most common description was “safe”, reflecting the importance researchers assign to designing safe urban environments. Overall, descriptions commonly appear in positive form, where applicable (e.g. “beautiful” rather than “ugly”), showing a tendency to focus on the existence/absence of positive experience. Yet the absence of a positive experience does not necessarily imply a negative experience (absence of beauty does not necessarily imply presence of ugliness). Therefore, future work can further study the links between negative descriptions and visual impression of urban space. Also, while a significant portion of the studies focused on green environments, the description “natural” occurred only once. The emerging approach of biophilic design has established that nature-like forms correlate with positive experience among humans [80,81]. Therefore, there is room to expand our investigation into what is visually experienced as natural and what is not, such that designed elements can draw on the benefits of biophilia, despite their artificial nature.

Additionally, our findings have shown that more than half of the descriptions studied were employed only in a single study, reflecting potential fragmentation in the literature. Considering this, future studies can refer to past literature for examining the verbal descriptions employed, to select ones which are suitable for their purposes. In this manner, we can build a body of knowledge regarding how specific descriptions are used in multiple contexts and by different groups, which is essential for building a robust understanding of the relation between language and space (as both words and visual environments are information-rich).

Finally, studies exploring neutral verbal descriptions can benefit from the inclusion of data concerning the users’ emotional states (e.g. valence [82]), to help resolve ambiguity regarding the meaning one assigns to the description. This can be done in tandem with qualitative methods such as interviews, suggested earlier (4.2.2), to understand how the description assigned reflects one’s subjective experience.

4.3. What Links Have Been Established Between Features of the Urban Environment and Descriptions Attributed to Them by Human Users? (RQ3)

This subsection integrates findings concerning preferences and descriptions. The surveyed works offer a broad perspective on the relations between urban spaces and visual impression, as studies were conducted in a wide range of environments, and with different foci in terms of the human experience under investigation. The largest body of knowledge identified is concerned with the design of urban green spaces. However, even in this case, the amount of evidence we currently have regarding visual impression in within each category is relatively small and sporadic, since the category of urban green spaces includes a diverse set of sub-categories (e.g., urban forest, urban parks), each with its specific characteristics (e.g., functions and visual elements). Nonetheless, several observations can be made on the basis of current evidence.

First, a notable number of elements were identified as potentially desirable, and such information can be instructive in designing urban spaces. For example, in the context of urban parks, a simple perimeter shape was preferred over a complex one [59]. As the study was executed in a single country, however, it remains to be clarified whether the findings are culturally dependent. Given the relatively small number of studies conducted in similar/identical spaces, the need for further validation extends beyond the specific study mentioned here, and can be taken as a general recommendation for the subfield as a whole.

Second, when examining the list of features associated with verbal descriptions (Table 9), a difficulty emerges concerning interpretation – while certain features are based on clear-cut definitions, e.g. background vegetation reaches up to 50% of the height of the building [66], other features are not readily quantifiable or definable. For instance, features such as “fragmented (visual) regions” or “sharp edges” [44] may be employed differently in different studies, which poses challenges during synthesis across the field. Therefore, researchers can consider creating a catalogue of qualitative visual features in verbal form paired with operational definitions to be used by the community, to help align our efforts to understand how specific visual aspects of the environment are experienced.

Third, while studies of visual impression typically assume that visual features directly determine our experience, the survey revealed that conceptualization and interpretation can also play an important role. For example, in the context of rural greenways, it was found that elements which provide opportunities for social/physical activity were evaluated as desirable [52]. Such findings highlight the importance of understanding the underlying reasons for users’ responses, and their contribution to forming impressions.

Finally, several challenges in studying visual impression exist. Among these, individual differences pose a significant hurdle. As the experience of visual stimuli may vary with background knowledge [47], culture [83] and other factors, it may be fruitful to aim to study the range of descriptions that an urban environment affords, rather than attempting to determine whether a single description is its best match. Indeed, research has shown that a single verbal description (“oku” in Japanese, meaning depth or mystery) can be associated with different visual features by different people (layers, barriers and other visual features) [78], thus pointing to the importance of employing an inclusive lens in terms of the descriptions and visual features studied.

Furthermore, evidence shows that findings concerning preference may conflict, even for environments of the same type, thus posing challenges in its application. For example, in the context of urban forests, one study reported that high amounts of deadwood were evaluated as desirable [26], while in another study users were pleased with having no deadwood at all [71]. Such conflicting evidence may be attributed to a large number of factors such as the user group, the specific type of forest being studied, or the activity in which participants engaged, which poses challenges for resolution. Finally, differences across groups have been established (e.g., the public preferred flower beds more than professionals [32]), thus highlighting the difficulty of identifying universal preferences.

5. Conclusions

Based on a review of recent studies, we propose guidelines and highlight opportunities and challenges for studying the visual impression of urban space.

First, to improve clarity, researchers are encouraged to clearly state the specific type of research question their study addresses:

• How do users experience and describe a given urban space?
• What visual features of urban spaces do users prefer?
• Both of the above

Second, ecological validity can be improved by including in-the-wild experiments (currently emerging) or by using immersive virtual environments for presenting visual stimuli display (as opposed to the common method of using photos).

Third, researchers have produced a broad and diverse body of knowledge concerning visual impression in different settings. Yet, if we aim for generalizable insights, existing works must be further validated and extended. Two useful strategies are:

• replicating past findings (none were identified in this review)
• conducting additional experiments in previously studied environments concerning other features or verbal descriptions

Finally, to overcome differences in the definition of visual features across studies, a catalogue of visual features paired with operational definitions may be developed in future work.