New insights into the impact of digital signage as a retail atmospheric tool

This paper investigates the previously little-researched role of digital signage (DS) in retail atmospherics, using an environmental psychology framework, drawing support from the Limited Capacity Model of Mediated Message Processing (LCM). DS consists of screen displays in public spaces showing video. The method consisted of a structured questionnaires quasiexperiment (n=357), comparing before and after DS installation against an unchanged control mall. Results demonstrate the effectiveness of DS, which has a positive effect on shoppers’ approach behaviours such as spending, mediated by perceptions of the retail environment and positive affect. Results are limited as the DS screens content was information-based, whereas according to LCM, people pay more attention to emotion-eliciting communications. The results have practical implications as digital signage appeals to employed shoppers. This study contributes to

track record of research projects for EPSRC, Knowledge Transfer Partnership and commercial sponsors. Currently, Newman serves as a council member for the British Academy of Management (BAM), and is also chair of the Marketing and Retailing BAM Special Interest Group.
Eleftherios Alamanos is a Lecturer in Marketing at Lincoln Business School, the University of Lincoln, UK.

INTRODUCTION
The retail landscape is changing rapidly. In the UK, for example, traditional high street spending is stagnant or even falling (BRC, 2011); with famous brand names struggling to survive (Sunday Times Business, 18 December, 2011) whilst e-shopping continues double digit growth (IMRG, 2011) from currently 14 percent of non-food sales predicted to reach 34 percent by 2020 (Javelin, 2011). At the same time, offline and online channels are becoming increasingly interrelated, with shoppers, for example, researching in store and buying online (Javelin, 2011), which can be aided by a smartphone price comparison app allowing shoppers to scan a barcode (Carmody, 2010). Major retailers are increasingly providing shopping assistant systems for consumers, either via shopping trolleys or apps on consumers' own smartphones (Pantano and Naccarato, 2010). These apps are catching on with consumers, with around one-sixth of the US and European consumers who use location-aware smartphones using them for location-based shopping or coupons (Microsoft, 2011). Thus, there is pressure on traditional retailers to improve efficiency, often by the use of technology (Pantano, 2010;Pantano and Timmermans, 2011) and to improve the shopping experience (Pantano and Naccarato, 2010).
This study concerns a technology development with the potential to improve the retail experience: digital signage or digital communications network (DCN). The study evaluates the impact of digital signage on shoppers' perception of the retail environment, positive affect, and approach behaviour in a shopping mall context. Digital signage consists of 'screen displays located in public spaces showing video material (or private TV channels)' (Clarke, 2003), Content typically includes (e.g.) advertisements, community information, entertainment and news. TV screens have been used in retail environments for some time but since the advent of digital control and flat screens, the use of networks of screens has made digital signage available as an effective, easily controlled communication medium. Referring to digital billboards, the Outdoor Advertising Association of America describes them as: ' … updated electronically through a variety of methods. Some are networked together, most are operated remotely, and all of them can be updated quickly, sometimes with just the click of a mouse. This ability gives digital [signage] flexibility and nimbleness. This nimbleness gives local businesses a unique and powerful way to reach a large number of geographically targeted consumers very quickly' (Outdoor Advertising Association of America, 2009).
Digital signage content may include, for example, advertisements, community information, entertainment and news. Such screen networks go by many names but we use the terminology 'digital signage' here as being most commonly used internationally. Similarly, we use 'shopping mall' (or simply 'mall') as the term becoming more accepted internationally for what has formerly been referred to in some retail literature as a 'shopping centre', i.e. a 'planned retail development … managed and marketed as a unit' with a 'pedestrian precinct covered from the weather' (Dennis, 2005, quoting Guy, 1994and citing Reynolds, 1993.
Digital signage aims to talk to shoppers while they are captive and in the mood to buy.
Retailers in countries including the US (Albertson's, Target, Kroger), the UK (Tesco, Asda, Sainsbury) and China (Carrefour) have launched digital signage networks. In addition to pushing merchandise, digital signage also generates hefty advertising revenues. Brand manufacturers pay anywhere from to $60,000 to $293,000 for a four-week campaign on Wal-Mart's TV network connecting more than 2,500 stores (The Economist, 2006). Although research figures are sparse, industry insiders estimate that digital signage is currently worth around $2billion in the US (Computerworld.com 2007).
Digital signage might be considered as contributing to retail atmospherics. Leo J. Shapiro & Associates, the firm that conducts store atmospherics surveys for Chain Store Age (Wilson, 2005) categorises in-store TV among interactive atmospheric elements helping retailers building a competitive advantage. Research indicates that shoppers tend to consider that they would benefit from technological innovations such as electronic shelf-edge displays (a special case of digital signage) and product information kiosks (which parallel digital signage) (Burke, 2002). This paper examines the effect of a digital signage network in contributing to retail atmospherics by influencing shoppers' perceptions of the overall retail environment and approach/avoidance responses. The research takes place in a shopping mall environment which differs from the retail store in not being aimed primarily at promoting a single retailer. Rather, digital signage in the mall environment is similar to the outdoor digital billboard, where it is often used to display breaking news, community information and promote a range of local retailers (Outdoor Advertising Association of America, 2009). As outlined by Underhill (2004), the mall is a store of stores, and better provision of information and perceptions of the atmosphere and environment of a mall should enhance shoppers' experiences such that they are likely to stay longer and spend more money (Wright et al., 2006). Digital signage is thus an important tool for retail atmospherics, with particularly important potential for shopping malls. Yet, there is a paucity of scholarly research into digital signage (for exceptions, see Dennis et al., 2010;Newman et al., 2006) and previous studies (Grewal et al., 2011;Kalyanam et al., 2010) highlight the need for further research in this area.
This study therefore sets out to address this research gap. The paper aims to make a theory contribution by exploring processes by which digital signage influences perception of a mall environment, affect and approach/avoidance behaviours. There are important implications for mall owners and for retailers, as, if digital signage can be demonstrated to enhance perceptions of a mall environment, research demonstrates that shoppers transfer perceptions of the mall environment to the store images of individual retailers (Chebat et al., 2006), which may significantly impact revenue.

The Shopping Mall
Consumers patronise shopping malls for purposes other than mere convenience (Woodruffe-Burton et al., 2002). Shopping frequency in malls is correlated with deal proneness, recreation and demographic characteristics (Roy, 1994). Personal life values and ethnic identification also influence mall patronage (Shim and Eastlick, 1998). Shoppers patronise shopping malls for walking and exercise (Hangland and Cimbalo, 1997) and as a social and recreation meeting place (Graham, 1988). The shopping mall is considered as a public place for community development among non-shoppers (Lewis, 1990), for the construction of social links (Aubert-Gamet and Cova, 1999), a city within a city (Backes, 1997) and as an ecological habitat for consumers (Bloch et al., 1994).
Many traditional shopping malls have difficulty competing against newer shopping malls that target fashion-oriented, value-oriented or time constrained shoppers (Reynolds et al., 2002).
Mall image, ease of spatial navigation (way finding), and entertainment stimulate shoppers' visits. The importance of the physical environment in a retail store setting has long been recognised (Baker et al., 2002;Bitner, 1990;Bitner, 1992;Bloemer and De Ruyter, 1998;Theodoridis and Chatzipanagiotou, 2009) and has more recently been extended to that of the shopping mall (e.g. Chebat and Morrin, 2007).

Retail Atmospherics
Retail atmospheric stimuli should be recognisable by consumers' senses, lead to positive shopping behaviour and be capable of being effectively manipulated by retailers (Turley and Chebat, 2002).
Reviews of many prior studies (Bakamitsos and Siomkos, 2004;Turley and Milliman, 2000) underscore a wide spectrum of shopping behaviours that can be influenced by specific atmospheric stimuli in a variety of retail formats. A selection of these and more recent studies is included in Table 1. Atmospheric stimuli that have been demonstrated to positively affect patronage behaviours include music (Garlin and Owen, 2006), lighting (Summers and Hebert, 2001), colour (Babin et al., 2003), design (Sherman et al., 1997) and digital signage (Dennis et al., 2010). Comments on selected studies follow in the sections below.
[Take in Table 1 here]

Mall Atmospherics
Compared with store atmospherics, there are fewer studies about how consumers perceive or respond to a mall's environment, particularly with respect to specific stimuli ( Table   1). The limited research available indicates that, similarly to stores, mall atmospherics influence mall image, shopper affect (e.g. Wakefield andBaker, 1998) andpatronage (e.g. Finn andLouvière, 1996). Interestingly, mall image influences store image (Chebat et al., 2006).
We predict that mall atmospherics will not only contribute to building mall traffic, but also promote sales and additional spending. Based on the environmental psychology approach (Foxall and Soriano, 2005;McGoldrick and Pieros, 1998;Mehrabian and Russell, 1974), a shopping-congruent atmosphere is expected to put shoppers in a favorable mood, have them stay longer in the mall, and encourage them to spend more.

Digital signage
As a relative newcomer to the retail environment, digital signage networks are now found in the marketing toolbox. Digital signage networks are used in many retail contexts, including main street shopping areas, malls and individual stores. Most commonly they consist of flat LCD or plasma screens with content linked digitally. They are used for many purposes, including advertising; provision of news and community information; and to enhance image. In the outdoor arena, screens can be very large, matching the largest conventional billboards. In most retail applications size is more modest, often less than two meters, although some can be much larger digital billboards. Shopping malls use digital signage particularly to generate advertising revenue (which is important but beyond the scope of this study); and to improve customer satisfaction and image.
We are able to cite only two published papers on the marketing aspects of digital signage in scholarly journals. First, a qualitative study by Newman et al. (2006) reports on the acceptability of digital signage to shoppers. That paper reports the results of eight focus groups (51 participants in total) with shoppers of varying age and gender recruited in a mall in which digital signage had just been installed. The consensus was that the digital signage created an ambience that influenced participants' perceptions of the mall environment, giving it a more modern image. The participants reported that the screens added enjoyment to their shopping experiences and provided useful information, informing their shopping choices. There were few objections to the digital signage but a minority of the participants considered it to be boring and not attention-grabbing. Second, Dennis et al. (2010) report a survey of mall consumers (n = 315).
That study suggests that digital signage has a significant, positive, total effect on approach behaviours such as consumer spending, mediated by positive affect and (arguably) perception of the mall environment, although as a cross-section study rather than an experiment, the reported effects of digital signage must be treated with caution.
The impact of digital signage on shoppers' perception of the environment and shoppers' responses falls in general within the environmental psychology approach (McGoldrick and Pieros, 1998;Mehrabian and Russell, 1974) and, more specifically to the digital signage stimulus, the Limited Capacity Model of Mediated Message Processing (LCM), which models how people process television communications, predicting the effectiveness of vivid moving visual images (Lang, 2000). The LCM (Lang, 2000) holds that people have a limited capacity to process information and therefore allocate processing resources to those most demanding stimuli that have a high information rate and distinctive features such as movement, colour and vividness (Li and Bukovac, 1999). Digital signage should therefore act as a more effective atmospheric stimulus, with higher recall of messages than those that are static or less vivid (Taylor and Thompson, 1982). Moving images attract viewers' attention (Reeves and Nass, 1996). The findings of Newman et al. (2006), mentioned above, that few people object to digital signage and most perceive it positively support the LCM in this context. We therefore consider that digital signage will constitute an effective marketer-manipulable atmospheric stimulus.

RESEARCH HYPOTHESES
There is ample evidence summarised above and in Table 1 to confirm that various environmental stimuli induce emotions that in turn influence approach/avoidance behaviour.
There is limited research on the effect of digital signage specifically but as the LCM predicts that people pay more attention to emotion-eliciting communications, we expect that any emotioneliciting content on the digital signage screens is likely to have positive effects, helping to justify our use of an environmental psychology framework.
According to the principle of cognitive mediation, the effects of cues (such as atmospherics) on people's emotions and behavioural responses are initially mediated by cognition (Lazarus, 1991). We propose that the cognitive construct of perception of the mall environment mediates the effects of the stimulus-emotion links (Chebat and Michon, 2003).
According to the LCM (Lang, 2000), the moving images of digital signage should constitute an effective atmospheric stimulus that may influence shoppers' images of the shopping environment, for example providing information. Therefore, following the LCM: Digital signage providing information to shoppers will positively influence perceptions of a mall's environment.
The environmental psychology approach has been previously tested in a retail setting.
The perception of a pleasant shopping environment (Dubé and Morin, 2001) should elicit positive emotions such as pleasure and arousal (Ang et al., 1997) and result in higher spending (Spies et al., 1997). Our model also draws on cognitive mediation (Lazarus, 1991) in that the specific stimulus (digital signage) influences the cognitive perception of other attributes of the mall environment which then influence affect and in turn influence behaviour. This mediation has been found to hold in retail atmospheric studies. For example, Sherman et al. (1997) find that the retail environment influences shopping behaviour but this influence is mediated by pleasure and arousal. In a more specific parallel, Chebat and Michon (2003) find that the influence of ambient aroma on shopper spending is mediated first by the cognitive perception of a store environment, then also by pleasure and arousal. In both the theory of cognitive mediation and the empirical results of Chebat and Michon (2003), mediation is full rather than partial. Following the principle of cognitive mediation and in analogy with the aroma stimulus, we therefore predict that: The effect of digital signage on shoppers' affect will be fully mediated by the perception of the mall environment.
Research propositions are summarised in Figure 1.
[Take in Figure 1 here] Digital signage is a relatively new, technological innovation. Therefore, in line with Rogers's (1995) Diffusion of Innovations, we expect age and income-earning status to have a moderating effect, such that digital signage has a greater positive impact for shoppers who are younger and earning, vs. those who may be retired or on state benefits.
The hypotheses above are based on the principle of cognitive mediation (Lazarus, 1991), in that the specific stimulus (digital signage) influences the perception of other attributes of the store environment which then influence emotions and in turn influence behaviour. This is the cognition → emotion model. Notwithstanding this, an alternative argument holds that atmospheric stimuli influence emotion (Donovan and Rossiter, 1982) and that emotion influences cognitive perceptions (Zajonc and Markus, 1984). In line with this, Puccinelli (2006) finds that people who are in a good mood before shopping have a better perception of products that they see and are willing to spend more, i.e. the emotion → cognition model. In the case of the digital signage stimulus, the two competing approaches are not necessarily mutually exclusive but are both consistent with the Elaboration Likelihood Model (ELM) of Petty and Cacioppo (1986). The message appeal can be either rational or emotional. The rational appeal may be more effective when the elaboration likelihood of the communication situation is high, i.e. when shoppers' processing resources are unrestricted and they can stop to watch the digital signage and perceive specific information (notwithstanding that customers may still process information via the peripheral route, particularly if they have not perceived relevance). Under high elaboration likelihood conditions, a person's cognitive responses will determine the behavioural outcome -the 'central route' -and these cognitive responses may result in controlled, higher-order Type III affective appraisals of the stimulus (Cohen and Areni, 1991;Pham, 2004;Shiv and Fedorikhin, 1999). Alternatively, when the elaboration likelihood is low, i.e. when shoppers' processing resources are restricted and the digital signage is perceived as background 'wallpaper', shoppers will not process messages cognitively but may still be influenced emotionally -the 'peripheral route' -and this affect, either Type I that is based on triggering hardwired programs and conceptually similar to the affect that Zajonc (1980) detected in his studies on "mere-exposure" effect or Type II that is based on the activation of emotional schemas and acquired through conditioning (Cohen and Areni, 1991), may still positively influence approach behaviour. Therefore, digital signage content might usefully be designed specifically to increase positive emotions. In this study, the researchers had no influence on the content, which was mundane and information based (as detailed in the 'Method' section below), and no control over shoppers' processing resources. However, there was no reason to believe that shoppers' processing capacity was consistently reduced on average. Therefore, we would expect that, in this instance, the model in which cognition gives rise to a higher-order affect would be superior to the model in which a low-order affect precedes cognition. Notwithstanding our hypothesised direction, when analyzing the findings, we acknowledge that our method is unlikely to resolve the direction conclusively and consequently examine both models.

Research Setting
The quasi-experimental research concerns two similar-sized and comparable shopping malls in West London (UK), one being used as control. It was carried out in two phases, before and after the introduction of a digital signage network in the mall under study. The first phase was carried out immediately before the installation of the digital signage followed by the second phase, which was held off until six months later in order for the effects of the installation of the digital signage to have stabilised. The role of the control mall was to neutralise the effects of other exogenous variables such as seasonality and validate measurement changes attributable to the introduction of digital signage. Changes in the test mall were compared against the control mall. This method improves validity by overcoming the potential flaw of standard "before and after" longitudinal studies -possible changes in uncontrollable confounding conditions. With this quasi-experimental design, incremental changes in perceptions are measured more precisely by comparison with the control mall (Cook and Campbell, 1979). No observable environmental changes, marketing communications or public relations activity took place at the control mall (or at the test mall, other than the installation of the digital signage) during the period.
The digital signage consisted of nineteen 1.07-metre plasma screens distributed around the public areas of the mall including the café (but not particularly placed at the entrances to the mall and not in the retail stores). The content consisted of one-third community information such as what was on at the theatre; one-quarter news, weather and sport; one-quarter advertising for the mall as a whole; and one-sixth local advertising. There was no entertainment other than in those categories, no national advertising and no sound.
Our sample consisted of actual customers that patronize both malls. In order to ensure that the sample was as representative as practicable of local people who may shop frequently or otherwise at the malls, it was sourced in two ways: post and email (eschewing mall intercept as less suitable for a long questionnaire (Frost-Norton, 2005)). The postal sample was intended to be representative of the residents in the area, obtained as a random selection from a commercially-available permission-based list sourced from the register of electors, selecting residents' living within 6 miles of the test mall (1000 before and 1000 after). The email sample was sourced from a commercially-available consumer panel who have given permission to be emailed with surveys and offers, again selecting residents living within 6 miles of the test mall.
The email sample reflected the slightly higher-than-average socio-economic profile of the typical mall shopper (1000 before and 1000 after).
Noting that McGoldrick and Pieros (1998) demonstrate shortcomings of previous retail atmospherics studies that use student samples, in order to maintain the integrity of our nonstudent shopper sample, any responses self-classified as 'student' were not considered in the analysis of results. Participants were offered a chance to win a shopping voucher worth £100 as an incentive to respond within five days. The response rates varied between 6.7 percent and 11.7 percent and responses totaled 357 usable questionnaires (see Table 2a). Seventy-six percent of the responses were received back within three days of the first response arriving. We therefore consider responses of four days and over to be late responders. The means of all the main variables for the late responders do not differ significantly from the means of those variables for the non-late responders; hence we conclude that there is likely little non-response bias in our model.
The sample achieved a profile approximately matching that of a typical upper socioeconomic suburban mall with 68 percent females (the mall owner's figure from proprietary data is 73 percent), median age of 44 years (same as mall owner's figure), 77 percent actively income-earning (not recorded in mall owner's data), 76 percent of households in the higher socio-economic employment categories (mall owner's figure 74 percent) and a median household income of £35,000 (not recorded in mall owner's data). There were no significant differences between the "before" and "after" samples in key characteristics such as gender, age, socio-economic status, income and whether earning vs. retired / not earning (see Table 2b). We therefore consider that the sampling procedure was as effective as practicable in matching the profile of shoppers at the mall and the before and after samples.

Measurement Scales
Three scales were used in this study: perception of the mall environment, affect, and reported approach/avoidance behaviour. These scales were based on the literature and used multiple-item measurements. Respondents rated both the test mall and the control mall based on their most recent visit and the items used are based on the differences between them. Scales were first subjected to exploratory factor analysis before being re-screened through confirmatory analysis and introduced in the structural model. Table 3 outlines the measurement scales with selected items, alpha coefficients, factor loadings and sources.
Perception of the mall environment. Retail image has been studied for some decades, formerly having been considered as a formative index, incorporating many attributes (e.g. Gentry and Burns, 1977). More recently, researchers have recognised that the perception of a retail environment can be considered as a latent variable reflected by a modest number of items (Bloemer and De Ruyter, 1998;Chebat et al., 2006). Nevertheless, there is little consistency in previous research as to which are the most salient attributes. The questionnaire therefore included the 15 attributes that previous studies found to be most salient (Finn and Louviere, 1996;Hackett and Foxall, 1994;McGoldrick and Thompson, 1992;Severin et al., 2001;Sit et al., 2003). In accordance with Bollen (1989) and Kline (2005), the top four loading perception of mall environment items were retained in the SEM model. The unused items are listed below Table 3. The indicators of perception of the store environment thus consist of 'How does [this mall] rate on the following on a scale of 1 -5 (where 1 is very poor and 5 is very good)? The satisfactory loadings empirically justify following the most recent authors, treating this variable as reflective.
Affect. The dimensions pleasure (Ang et al., 1997), arousal (Wakefield and Baker, 1998) and satisfaction (Rigopoulou et al., 2008) capture the range of affective states relevant to shopping approach/avoidance behaviour (Russell, 1979). The following four items, originally from Mehrabian and Russell (1974), which have been validated many times in retail applications, were used as indicators of affect: (i) unhappy / happy; (ii) melancholic / contented; (iii) unstimulated / stimulated; and (iv) dissatisfied / satisfied. Despite pleasure and arousal being considered as separate variables in many prior studies, our model seeks an overall latent variable to capture shoppers' affective state. This conceptual approach finds empirical support from, e.g., the pragmatic correlation of the error terms of pleasure and arousal by Sherman et al. (1997), notwithstanding those authors including pleasure and arousal as separate variables in their model. Our conceptual approach could have been satisfied by entering affect into the model as a second-order variable of pleasure and arousal but in the event, as we expected and hoped, the consistent loadings on the single latent variable rendered such an awkward model unnecessary.
The indicators of affect thus consisted of 'to what extent does (this mall) make you feel … ' (five point scale anchored by, e.g., unhappy -happy).
Digital Signage. We also wished to measure shoppers' perception of digital signage, even though this cannot be used in our model as it has no value in the 'before' condition. We are unaware of any prior scale for this variable, so we based this measurement on those previously reported for other stimuli such as aroma (e.g. Ellen and Bone, 1998). In reporting these results, we chose to report the before and after results together and therefore digital signage entered the model as a dichotomous variable: digital signage present / not present.
[Take in Table 3  'Spending on non-food shopping'. Each of these items is the measure of the test mall utilization relative to the total of the test mall and the control mall.
All respondents answered all the perception of the mall environment; affect; and approach/avoidance questions about both the test and control malls. Numerically, the items other than approach/avoidance are based on the value for the control mall subtracted from the value for the test mall (with 5 added so as to be always positive).

Models
Two models were used to assess the effect of digital signage on shoppers' perception of the environment and the subsequent behavioural response. ANOVA captured the change in shoppers' perception following the introduction of the digital signage in the test mall. Observed changes in the test mall (M) were adjusted for any change in the control mall (C) (e.g. Cook and Campbell, 1979): For measurement purposes, the equation was algebraically modified to: The next step modeled the influence of digital signage on shoppers' positive affect and approach behaviour and investigated mediation through a latent path structural equation model (SEM) using SPSS AMOS (Arbuckle, 2006).
The constructs were used in two different ways. First, as the Cronbach alpha coefficients are satisfactory (> .8), for simplicity of presentation, we subjected the means of the indicators (rescaled 0-1) to ANOVA to compare the values before and after the installation of digital signage at the test mall (Table 4). Second, to investigate mediation and illustrate the relationship between latent variables representing shoppers' perception of the mall environment, positive affect, and approach behaviour, we carried out a path analysis (Figure 2).

RESEARCH FINDINGS
First, it can be reported that respondents' opinions on the digital signage were mainly results. In the interests of parsimony, the details of these tests are not reported here.
[Take in Table 4 here] The latent variable path analysis outlines the relationships between the environmental cue (digital signage) and shoppers' response. The SEM exhibits an excellent fit (CFI = .99, RMSEA = .026, χ 2 = 73.1, df = 59, p = .10, χ 2 /df 1.2). When all responses are considered, the influence of digital signage on shoppers' perception of the mall environment is significant in the (hypothesised) one-tailed test [Standardised Coefficient = .11, Critical Ratio (C.R.) = 1.9, p (two tailed) = .06]. These results support H 1 , digital signage providing information to shoppers positively influences perceptions of a mall's environment. In the interests of parsimony, this model is not reported diagrammatically but rather, we proceed to the model for the incomeearning shoppers.
When the results for income-earning shoppers only are considered, digital signage has a more significant direct influence on shoppers' perception of the mall environment (Standardised Coefficient = .14, C.R. = 2.2, p = 0.03). Shoppers' perception of the mall environment influences shoppers' emotions (Coefficient = .79, C.R. = 11.2, p < .001). In turn, shoppers' affect impacts approach behaviour (Coefficient = .66, C.R. = 89.1, p < .001) (Figure 2). These results for income-earning shoppers also exhibit an excellent fit (CFI = .98, RMSEA = .037, χ 2 = 81.2, df = 59). As is common for reasonable size samples, p is significant at .002 but other measures confirm fit with χ 2 /df 1.6. In line with H 1 , digital signage providing information to shoppers positively influences perceptions of a mall's environment. In line with expectations, this finding is more significant with the income-earning shoppers than with the total sample. Sample size considerations preclude a between-groups analysis of the moderation effect.
[Take in Figure 2 here] The significant path from the dichotomous variable, digital signage to the perception of the mall environment confirms the differential effect of the digital signage between the two conditions. According to McArdle (2001), latent mean scores are more reliable for comparing means over a time series than are direct change score analyses. Accordingly, using a multi-group model, we investigate whether the latent means of the constructs vary between the two conditions. First, we confirm that the measurement weights do not vary significantly between these conditions (∆χ 2 = 6.61, ∆df =9, p = 0.67). The latent means in the 'with digital signage' condition are significantly higher for: the perception of the mall (C.R. = 2.26, p = 0.024); and affect (C.R. = 1.94, p = 0.053, significant in the one-tail test). This means that respondents rate the perception of the mall significantly higher and have significantly higher positive affect in the 'with digital signage' condition compared with the 'without digital signage' condition. The latent mean of approach is also in the expected direction but not sufficiently large for significance (C.R. = 1.46, p = 0.14). In the interests of parsimony, we do not illustrate the latent mean models diagrammatically.
Testing for cognitive mediation (Baron and Kenny, 1986), we first demonstrate that digital signage has a positive effect (one-tailed test) on affect in the absence of the mediating variable, perception of the mall, using the latent mean test in the paragraph above, i.e. affect is significantly higher (one-tailed p = 0.025) in the 'with digital signage condition. Second, when the mediator, perception of the mall, is included, the direct relationship between digital signage and affect becomes non-significant (C.R. = -1.16, p = 0.25). Third, the R 2 of affect is only 0.01 in the unmediated condition compared with 0.63 in the final mediated model. These three tests fulfill Baron's and Kenny's (1986) conditions for mediation, supporting H 2 , the effect of digital signage on shoppers' affect is fully mediated by the perception of the mall environment, demonstrating that Lazarus's (1991) principle of cognitive mediation is valid for digital signage.
Similarly, it is also noted that including a direct path in the SEM between digital signage and approach is non-significant (C.R. = -0.40, p = 0.69). This is in line with most retail atmospherics models that test specific stimuli (e.g. Chebat and Michon, 2003, who test aroma in a mall), notwithstanding that authors seldom draw attention this aspect.
In the 'Conceptual Framework' section above, we introduced a possible alternative conceptualization based on the argument that atmospheric stimuli may influence emotion (Donovan and Rossiter, 1982) and that emotion influences cognitive perceptions (Zajonc and Markus, 1984). This route is not hypothesised, on the grounds that the digital signage content used in this study was designed to influence cognitions rather than emotions. Nevertheless, we also evaluate that model for comparison. When the results for income-earning shoppers only are considered, digital signage has a marginally significant direct influence on shoppers' perception of the mall environment (Standardised Coefficient = .12, C.R. = 1.8, p = 0.07). Shoppers' affect influences shoppers' perception of the mall environment (Coefficient = .82, C.R. = 11.7, p < .001). In turn, shoppers' perception of the mall environment impacts approach behaviour (Coefficient = .56, C.R. = 7.4, p < .001). These results for income-earning shoppers exhibit a good fit (CFI = .96, RMSEA = .063, χ 2 = 124.4, df = 60. p = .000, χ 2 /df 2.1. The influence of the digital signage on affect is only marginally significant and the fit of the hypothesised model is significantly (p < 0.001) better than this alternative conceptualization. Nevertheless, this remains a promising area for future research. In the interests of parsimony, we do not include this alternative model diagrammatically.

DISCUSSION
The environmental psychology paradigm is not new to retail atmospherics. Ample research has shown that environmental cues will impact shoppers' cognition and emotion, and trigger some approach behaviour (Turley and Milliman, 2000). What is new here is the advent of digital signage or digital communications networks in the retail atmospheric toolbox as a stimulus with a significant effect, as predicted by the LCM. Digital signage has a dual usage: it conveys information when and where shoppers are in the mood to shop, and it can have a strong entertainment component. These results indicate that digital signage is an effective stimulus, adding to positive perceptions of the mall environment, emotions and approach behaviour such as spending, as predicted by the LCM. This study cannot claim to be a test of the LCM, which would entail measuring the availability and allocation of cognitive resources; and the information rate of the stimulus; and comparing with a lower-information rate stimulus. Nevertheless, the effectiveness of digital signage as an atmospheric stimulus is consistent with the LCM, which predicts the effectiveness of vivid moving visual images. The LCM holds that people have a limited capacity to process information and therefore allocate processing resources to those most demanding stimuli that have a high information rate and distinctive features such as movement, colour and vividness. The moving images of digital signage should attract viewers' attention and act as a more effective atmospheric stimulus, with higher recall of messages than those that are static or less vivid. The confirmation of digital signage as an effective stimulus therefore extends the LCM from television to digital signage.
The research shows that digital signage is effective with income-earning shoppers, who are generally in a hurry to complete their shopping chores. Digital signage will inform them about product offerings and promotions. This may simplify their shopping experience, which Kalitcheva and Weitz (2006) have found to be preferred by high task-orientated shoppers. These shoppers might also be more familiar with plasma screen technology. This more-evaluative pattern for income-earning shoppers parallels findings of Raajpoot et al. (2008), who report that the effect of the overall evaluation of a shopping mall on repatronage is greater for working women than non-working.
On the other hand, digital signage may not enhance the perception of the mall environment among non-working or retired patrons (although their representation in the total sample is too small to justify reporting in detail). The latter are already spending time in the mall for other purposes than shopping, namely entertainment or passing time (e.g. Bloch et al., 1994).
It may well be that this particular digital signage may not work so well with retired, unemployed or inactive people because it is purely informational (e.g. prices, promotions, special offers and community news) and has no or little entertaining effect on those just spending time in the mall.
If the digital signage content were entertaining, the story might be totally different.
Mediation testing (Baron and Kenny, 1986) illustrates that the impact of digital signage on shoppers' emotion and approach behaviour is fully mediated by shoppers' perception of the mall environment and positive affect (Figure 2). The standardised total effects are reported in Table 5. The total effect of digital signage on approach behaviour (.07) and emotions (.11) is significant.
In order to explore the alternative emotion → cognition conceptualization, we considered the ELM of Petty and Cacioppo (1986). In this instance, with the digital signage content being designed to influence cognition rather than emotions, the hypothesised cognition → emotion model fitted the data significantly better. Nevertheless, the results provide food for thought, suggesting future research into the alternative conceptualization and the potential for digital signage to influence approach behaviour by priming through the peripheral 'wallpaper' route rather than (or in addition to) providing cognitive information.
[Take in Table 5 here] The impact of the digital signage network in the mall was cross-validated by traffic counting. Twelve months before the installation of the digital signage network, traffic tracking was undertaken. Footfall traffic (vs. a national benchmark) was 6.1 percent higher after the installation of the digital signage compared with the same period in the previous year.

MANAGERIAL IMPLICATIONS AND FURTHER RESEARCH
This quasi-experimental study of the little-researched atmospheric stimulus of digital signage indicates that it significantly enhances the perception of the mall environment and approach behaviour among income-earning busy shoppers, who are generally in a hurry to complete their shopping. For these shoppers, digital signage may facilitate their tasks and inform them about product offerings and promotions. On the other hand, digital signage is not highly evaluated by retired and non-working patrons who are already spending time in the malls for entertainment or passing time (Bloch et al., 1994). The particular digital signage in this study was used to convey information (e.g. advertising) and had little or no entertaining effect on those just spending time in the malls. Mall managers therefore need to consider market segmentation in the design of digital signage content. We have demonstrated in this study that informationbased content is suitable for targeting the higher-spending income-earning shoppers.
Entertainment-based content may well be preferred by retired and non-working shoppers, although further research is recommended to confirm this. In this study, the formal testing of moderation effects was precluded by considerations of sample size but remains an objective for future studies.
Results confirm expectations that the cognitive construct of mall perception mediates the effects of stimulus-emotion links. The results demonstrate that customer information strategies using digital signage can be useful and effective for mall managers.
These results can be seen in the wider context of improvements to shopper information, viz the convergence of online, mobile and in store shopping (Pantano, 2010;Pantano and Timmermans, 2011), as shoppers are increasingly using smartphone apps for shopping information and to take advantage of special offers (Microsoft, 2011). Taking  The results are also limited in as much as the content shown on the digital signage screens was almost entirely information-based, having limited entertainment value, whereas the LCM, supported by evolutionary psychology, predicts that people pay more attention to emotioneliciting communications. If processing resources are limited (e.g. if busy shoppers perceive digital signage only as 'wallpaper'), emotion rather than cognition may be expected to have the stronger effect on consumer choices (Shiv and Fedorikhin, 1999). Given that hedonic benefits are more likely to increase positive affect and loyalty than utilitarian ones (e.g. Chitturi et al., 2008) (Anderson and Gerbing, 1988). The fit of this combined model to the data is significantly (p = < 0.001) worse than the hypothesised model (CFI = .92, RMSEA = .083, χ 2 = 175.0, df = 61, p = .000, χ 2 /df 2.9), justifying discriminant validity of the two constructs and preference for the hypothesised model. Moreover, the criterion that we used to show the discriminant validity is stricter than those used in recent construct development papers (e.g., Thomson et al., 2005).  Herrington and Capella (1996) Music Sales in supermarkets In-store survey, n=140 (89 aware of background music) Yalch and Spangenberg (1990) Music Sales in department stores In-store experimental design, n = 86 (foreground music = 33, background music = 32, and control =21) Hui et al. (1997) Music Responses to waiting in banks Retail banking video setup, n = 116 undergraduate students, experimental design (4 types of music plus control) North et al. (1999) Music Sales in wine shops In-store 2x2 experimental display; German and French music, and German and French wines, n = 82 Dubé and Morin (2001) Music Positive affect and approach behaviors such as spending In-store survey, n = 110 shoppers aware of background music post-categorised in the low pleasure intensity (48) and high pleasure intensity (62) conditions Spangenberg et al. (2006) Genderappropriate aroma Perceptions of apparel store environment, merchandise and approach behaviors such as spending Pretesting of feminine and masculine scents (n = 300 students, faculty and staff); in-store field experiment (82 males, 99 females) in congruent and incongruent scent conditions. Chebat and Michon (2003)  Cognitive influences Beverland et al. (2006) Music "fit" Perceptions of an apparel brand 20 in-depth consumer interviews Smith and Burns, 1996 Store layout Price perceptions Warehouse grocery store intercept before and after manipulation (n = 182), with control (n = 198) Baker et al.
General environment Store image of a card and gift store N = 297 undergraduates in a laboratory experiment (2x2x2) opposing prestige to various discount conditions Machleit et al. (1994) Crowding Shopper (dis)satisfaction 1) University bookstore video simulating high and low crowding situations (n = 76 undergraduates) 2) Actual bookstore under various crowding conditions (n = 140) 3) Two grocery stores under various crowding conditions (n = 232 shoppers) Yoo et al. (1998) Facilities and product assortment Shoppers' positive affect Shoppers intercept (n = 294) in two large Korean department stores

Perception of mall environment
Mall image and atmosphere Mall image influences store image Video mall simulation (n = 200 shoppers) in an experimental factorial design; store types (2), mall image (2), shoppers SES (2). Dependent variable: self-congruity and store image