How the Multimedia Communication of Strategy Can Enable More Effective Recall and Learning

Many claims have been made about the learning benefits of communicating strategies in multimedia picture-plus-text formats, rather than monomedia text-only formats. However, there is little theoriz...


INTRODUCTION
A multi-media combination of pictures and words has been promoted as a more effective way to enable recipients to learn about strategies for three decades (Huff, 1990;Fiol & Huff, 1992;Kaplan & Norton, 1992;1996a;2000;2005;Mintzberg & van der Heyden, 1999;Kim & Marbourgne, 2005;Osterwalder & Pigneur, 2010;Meyer, et al., 2013;Wright et al., 2014;Jarzabkowski & Kaplan, 2015). In the past decade, the strategic management classroom has become a site of particular pedagogical interest, as strategic management courses have been criticized for being mere 'kit bags' of theories and frameworks (Mintzberg, 2004;Bell et al., 2018) and calls have gone out for more practically applicable strategy education (Albert & Grzeda, 2015;Clegg, Jarvis, & Pitsis, 2013). To this end, some researchers have advocated less emphasis on theory-based approaches and more on practical skills and implementation (Grant & Baden-Fuller, 2018;Lindsay, Jack, & Ambrosini, 2018). Others have advocated using better theory (Buckley, 2018).
Bridges between theory and practice in the strategic management classroom have traditionally been built by utilizing picture plus text frameworks such as SWOT, VRIO and the Value Chain to 'unpack' and discuss cases toward making practical recommendations.
Michael Porter (1991: 98), developer of many popular strategy frameworks, related the value of these multi-media conceptions to their ability to help "identify the relevant variables and the questions which the user must [then] answer in order to develop [their own] conclusions." While strategic management courses have tended to use frameworks as diagnostic aids, there has been little reflection on why they are used, what the benefits of them are, how their results help communicate strategy to students, and how we might use them more effectively to address the demand for practically applicable strategic management courses.
This article reports on a study that increases our understanding of how presentation and engagement with multi-media, defined here as drawings and text, can improve learning in the strategic management classroom. It helps reveal the benefits and drawbacks of presenting strategies using a multi-media format, and provides recommendations to strategy educators about enhancing student capability in dealing with strategy communications.

ADVOCACY WITHOUT EVIDENCE
Beyond the development of many of strategic management's best-known theoretical frameworks in the 1970s and 80s, academic discussion on the value of presenting strategy by using pictures and text can be traced to the early 1990s. It was noted then that "managers have long recognized the importance of map-like products" and that "maps used as [management] tools [would become] increasingly important in an uncertain world that requires managerial judgement" (Fiol & Huff, 1992: 273). The benefits of using pictorial forms to enhance strategy communication was attributed to their ability to help managers and students make sense of complexity, focus attention and trigger memories, signal priorities and supply missing information, simplify and aid the communication of complex ideas, and divorce ideas from specific speakersmaking them more accessible to debate and modification (Huff, 1990;Fiol & Huff, 1992).
The framework of this era most associated with communicating strategy pictorially was Kaplan and Norton's 'Balanced Scorecard'. Kaplan and Norton (2005: 72) argued that the typical lengthy and text-laden (or mono-media) forms of communicating strategy were not effective: "Our research reveals that, on average, 95% of a company's employees are unaware of, or do not understand, its strategy". More recently, Sull et al.'s (2015) study showed that only about half of the managers they studied could name any of their organization's strategic priorities. Kaplan and Norton claimed that their picture and text "comprehensive snapshots" would help people "view strategies in a comprehensive, integrated and systematic way" (Kaplan & Norton, 2000: 60), and enable strategy to be 'bottled' "so that everyone could three categories of assumed effects (cognitive, social, and emotional) of multi-media strategy communication.
CLT is attractive for understanding strategy communication as it focuses on complex learning environments and real-life settings (Sweller, 1988). Through extensive studies, Sweller and his colleagues showed how communications that result in heavy cognitive loads are associated with errors in recipients' recall and interference with tasks, particularly when instructions and tasks are subjective and complex (Sweller & Cooper, 1985;Paas, 1992;Sweller et al., 1998;. We reason that strategy communication can be subjective and complex and can result in confusion for recipients, causing heavier cognitive load, especially when multiple strategy dimensions are combined or integrated. Cognitive load could be reduced in recipients through the incorporation of pictures in communications as these provide a vehicle for 'mental integration', allowing the integration of concepts which cannot be achieved as effectively through the use of text-only formats, such as bullet points (Sweller et al., 2011).
Richard E. Mayer and his associates at the University of California built on Sweller's work on CLT to concentrate on the role of multi-media or multi-channel approaches. The primary focus of Mayer's Cognitive Theory of Multi-media Learning's (CTML) is that human working memory has sub-components that work in parallel and that learning can be improved if multiple channels are used for information processing at the same time. Based on many decades of findings, as evidenced by measures such as recall or the ability to subsequently solve problems using what students have learned, Mayer and his followers now have significant evidence for their 'multimedia principle' that people learn more from words and pictures than from words alone and CTML studies continue to seek to better understand how this works, and to design more effective instructional materials (Mayer & Anderson, 1991;Mayer, 1997;Mayer, 2002;Moreno & Mayer, 1999). CTML theorizing relates multi-media communication to three avenues of inquiry in cognitive research. The first is the Dual-channel Assumption, which proposes that we have separate channels in our working memory to process pictures and text, and that engaging both can enhance learning. This idea was first suggested by Baddely and Hitch in the 1970s (they referred to a phonological loop system channel and a visuospatial sketchpad channel - Baddely & Hitch, 1974) and developed into Paivio's (1991) 'Dual-coding Theory'. This assumption is analogous to Cognitive Load Theory's acknowledgement of "separate channels for dealing with auditory and visual material (Sweller, 1999: 128)." While much of the early research conducted in this avenue looked at multi-media communication using visual plus audio content (Drewnowski & Murdock, 1980), it was suggested that, within the visual channel, different approaches (e.g., text and pictorial communication) could activate different mental responses and processes, so that a multiplicity of visual media using pictures and text might be superior to singular text-based communication in enhancing learning.
The second avenue of CTML theorizing is the Active-processing Assumption. This claims that humans are dynamic learners who attend to incoming information by organizing it into mental representations developed in relation to prior knowledge. This view conflicts with the conventional assumption that humans are passive processors who pile incoming information on to what is already held, as if inputs were discrete files to be retrieved or downloaded at a later stage. The CTML theorization suggests that multi-media picture plus text instruction may aid effectiveness if it assists learners in building representative schemas that combine the information communicated with knowledge held in the long-term memory, and is related to what is often referred to as 'Constructivist Learning' approaches (Huang, et al., 2010).
The third avenue of CTML in this regard is the Limited Capacity Assumption.
Research dating back to Miller's Information Processing Theory (which identified that humans could effectively hold only sevenplus or minus twounits of information in their short-term memory) showed that our channels have inherent working memory capacity load limits (Miller, 1956;Chase & Simon, 1973;Linden, 2007). Cognitive load theorists have defined three types of cognitive load to help us understand how Limited Capacity works in practice: intrinsic, extraneous, and germane. Intrinsic cognitive load is the inherent level of difficulty associated with a specific instructional topic (Chandler & Sweller, 1991;. As it is inherent to the nature of what is being conveyed an instructor cannot alter it. Germane cognitive load is associated with the processing and construction of representative schemas (Sweller, et. al., 1998). It relates to elements that an instructor can introduce in delivering the intrinsic load that helps the learner remember and understand what is being conveyedso, it is 'germane' to the learning process. Extraneous cognitive load refers to the mode or way in which information is presented which is not germane, and may become a distraction by unnecessarily adding to the cognitive load. Since individuals' channels have limited capacity and a lot of information may be difficult to retain over time (Peterson & Peterson, 1959), too much information across the channels causes cognitive overload, which can lead to increased stress, decreasing confidence, and a declining ability to think effectively about subsequent tasks. So, to convey information effectively, the extraneous load should be limited and the germane load should be enhanced, especially when the intrinsic load is high (Ginns, 2006). CTML suggests that the addition of pictures in the communication of germane load may enhance learning if it reduces apparent intrinsic load; but it may be detrimental if it adds extraneous load.
Using these theories, we can surmise that the way strategy communication is received will be affected by its presentation, its extraneous cognitive load, particularly if the content, or intrinsic cognitive load and germane load, remains relatively constant. Cognitive load theorists suggest that communication modes that show interdependencies, such as pictures, are likely to be more effective in reducing cognitive load than discrete separated components, such as bullet points. While CLT has focussed more on researching the benefits of incorporating audio communication in relation to communications that solely use the visualization of text, both CLT and CTML suggest that multi-media communications are more effective than single mode communications in promoting learning and recall. In terms of communicating strategy, we might contend that picture plus text can convey added information, allowing recipients to understand its complexity more effectively than through the use of text alone, and facilitates better recall (Kosslyn, 2007). This leads us to our first hypothesis that multi-media communication of strategy, using picture and text, will provide a cognitive benefit in aiding recipients' recall more effectively than text-only presentation.
Hypothesis 1: Multi-media picture plus text presentations of a strategy are more effective in enabling recipient recall than mono-media text-only presentations.
CLT and CTML also suggest that, when a communication presents a high intrinsic load that is difficult to process and recall, it may lead recipients to be discouraged and their selfconfidence to be eroded (Elen & Clark, 2006). However, where communications are able to buttress the recipients' thinking, in terms of constructing linkages to known representative schemas, this cognitive load may be mitigated: in this way, self-confidence may be preserved or even boosted. Therefore, our second hypothesis posits that multi-media picture and text communications of a strategy help reduce recipients' cognitive load and promote their feelings of confidence with respect to discussing or acting on that strategy.
Hypothesis 2: Multi-media picture plus text communications are more effective than mono-media text-only presentations of strategy in promoting recipients' confidence to discuss the strategy.
As introduced above, another stream of the research carried out in CLT and CTML exploring the workings of germane cognitive load, relates to the processing and construction of representative schemas in recipients. This research leads us to posit that strategy communication may convey intrinsic load more effectively when it mitigates cognitive challenges through tracking, and showing interdependencies and connections between strategy elements, in such a way that make these interrelationships easy to see and recall (Sweller et. al., 1998). Therefore, our third hypothesis focuses on whether strategy communication mode affects recipients' ability to perceive connectivity between different elements in, or parts of, a strategy, or what we term here 'integration'.
Hypothesis 3: Multi-media picture plus text presentations of a strategy enable recipients to more effectively perceive interrelationships between that strategy's elements than mono-media text-only presentations.
These three hypotheses broadly map on to the three categories of benefits attributed to the multi-media communication of strategies espoused by Eppler and Platts (2009) and described earlier: cognitive benefits (e.g., recall), emotional benefits (e.g., creating involvement, engagement and providing inspiration or confidence), and social benefits (e.g., integrating and seeing interdependencies).
While CLT argues that the greater the working load needed to process information, the greater the errors, and CTML responds to this problem by theorizing that multi-media communications can be used in teaching to allow more meaningful learning (Mayer, 2005) (Mayer, 2005), and this could lead to higher-quality learning. In a reverse case, if a participant receives a multi-media communication and only reproduces it in one mode, information might be lost by choosing to deploy just one mental pathway. This line of reasoning extends CTML to argue that multi-channel processes are not only important for reducing cognitive load in acquiring knowledge but equally important for recipient recall, confidence to discuss, and ability to show interconnections, the latter being a necessary part of indicating schemas in a germane load. Consequently, in testing our three hypotheses, we will comment on the modes of reproduction chosen voluntarily by participants to see if they have any moderating effects on learning outcomes.

Experiment Design
A randomized experiment was viewed as the most appropriate method for addressing our three hypotheses because it creates comparable treatment groups that are less likely to differ on any measured or unmeasured variables (Cook, Shadish & Wong, 2008). It was implemented as follows by the authors of this study, during strategy courses that they were teaching in universities in various locations around the world. The experiment took place before any strategy tools and frameworks had been taught in their courses. Each professor's role was to hand out an envelope to all students in the course and read out a set of standardized instructions at the beginning and end of the exercise. Specifically, students were asked to each open an envelope, take out the page inside and read the instructions. Students were assured that this experiment was not part of any assessment and that they would not be required to provide biographical information, meaning the results would be anonymous. 1 Inside the envelope was a representation of a company's strategy simplified to highlight five distinct but interrelated core elements, each with two sub-parts, making 10 data points in total. While each page depicted the same strategy (i.e. the same 10 data points), different envelopes contained these in different presentation formats (see Appendix 1): two text only formats in 1) bullet points and 2) a paragraph (the paragraph contained more detail and contextual load); and one multi-media presentation comprising a pictorial expression of the strategy with texta hand-drawn adaptation of a Value Chain (Porter, 1980). The envelopes were distributed to students randomly, and students were not aware that they did not all receive the same strategy presentation format.
Students were asked to look at the strategy presented in their envelope and memorize it without consulting with other students and without making notes. After one minute had elapsed, they were asked to put the page back in the envelope, seal it, and place it to one side.
After a pre-determined interval (of either 5, 10, 15, 20, 30, 40 or 50 minutes during which students were focussed on other tasks within the lecture), students were instructed to reproduce the elements of the strategy in whatever form they wished on the back of the original envelope (to avoid them tampering with the document), again without consulting others.
With this design, the experiment was randomized in that students may not have received the strategy in the presentation format most inclined or suited to their preferences, but they were also not channelled into a particular format of response when recalling the strategy. We recorded both the mode in which the students received the strategy as well as the mode with which students recalled the strategy.

Sample
The experiment was conducted by the researchers across numerous courses over three years involving a breadth of management students (N=1140, see Table 1), spanning eight countries with most completing the course as part of a business degree. The sample was spread as follows due to the varying course sizes: Austria: 4%; China: 2%; France: 5%; Morocco: 12%; New Zealand: 32%; Tunisia: 9%; UK: 29%; USA: 7%. The primary nationality within each course matched the course location, although the UK and NZ undergraduate and UK/European-based MBA cohorts included a greater mix of nationalities. 70% of the courses surveyed (18/26) were MBA-type classes, the other 8 groups were undergraduate classes, with the eventual number of responses from each being 578 MBAs and 562 undergraduates.
The average age of the MBAs was 34 years compared with 20 years for the undergraduates, although some MBA courses included a few students who were only slightly older than the 19-22 age range typical of undergraduate courses. The MBA students had typically worked in managerial roles and had greater prior familiarity with strategy frameworks such as the Value Chain. The MBA courses tended to have a higher proportion of males (average = 70%), whereas this average was lower for the undergraduate courses (56%). The average interval was slightly higher for the undergraduate courses (21.9 vs. 16.5 minutes), because some of these courses provided the best opportunity for including longer intervals in the experiment.
The diverse nature of the sample, containing a breadth of nationalities and associated first languages, ages and genders, suggests that any systematic differences associated with the strategy presentation and recall in the sample are likely to be generalizable to students in most countries in the world. *** insert Table 1 about here *** 42% of the students received the strategy in a multi-media picture and text format, and 58% received the strategy in a mono-media textual format (a bullet pointed list or paragraph), with 23% of the mono-media group receiving the strategy via a text paragraph containing more words than the bullet point version. The paragraph version was introduced later in the data-collection phase to investigate if the picture plus text version might have contained more information about the strategic issues and their integration than bullet points alone. The paragraph explicitly included discussion about strategic issue integration. The proportion of students receiving the multi-media version ended up slightly lower than 50% in order to gather more data about the effect of paragraphs.
Fewer than 2% of all students chose not to participate, did not complete, or "spoiled" the responses on the envelopes and less than 9% did not answer all of the survey questions.
Failure to answer all questions was highest in the MBA cohorts (for the question related to confidence to discuss). There was no readily identifiable reason for this, with the response rate of over 90% suggesting that the tasks were understandable and achievable within the timeframes utilized. A total of 1060 students completed all tasks/questions in the experiment and this slightly reduced sub-sample was used in the regression analyses.

Measures
There are two main measures of learning effectiveness used in CTML: student recall and the ability to solve problems with what has been learned. While we recognize that recall relates to lower level aspects of Bloom's Taxonomy (Anderson & Krathwohl, 2001), we used it as our first measure of effectiveness as it could be more reliably operationalized and assessed independently in a large-scale experiment. Confidence to discuss and ability to perceive interrelationships (associated with our second and third hypotheses), which relate to higher levels of learning, could not be assessed on an independent basis given the scale of the experiment. Therefore, we relied on students' self-assessments in this regard.
Recall scores. Student recall was assessed on a ten-point scale based on the extent to which the ten elements of the strategy were evident in their responses (+1 for a correct recall and 0 otherwise). This scale provides a sufficiently broad gradation of the extent to which different elements of the mono-/multi-media formats were being accurately recalled and enabled the reliability of coding. Two different coders, who assessed approximately 50% of the overall sample each, scored each of the students' envelopes and recorded the data. A randomized sample of approximately 30% of all envelopes across the two coders (N=326) was assessed by a third coder to evaluate inter-rater reliability for recall scores, with the tests indicating a high level of agreement (86% equal or +/-1 of the original recall score and 98% within 2; Cohen's weighted kappa value of .723 indicating good agreement as well as Pearson's and Kendall's tau correlation coefficients of 0.95 and 0.88).

Confidence to discuss.
Once students had reproduced the strategy on their envelopes, they were asked 2 additional questions. The first addressed their confidence to discuss the strategy, specifically: " You are about to go in to a meeting to discuss this strategy with a group of managers from the company. On a scale of 1 to 5, how confident are you to discuss the strategy in a convincing way?", with 1 being not at all confident and 5 being extremely confident. Students were informed at this point that this is not asking about their confidence in the particular strategy, but their confidence in debating its merits with others. Stankov et al. (2015) report that such measures of confidence in one's knowledge related to a particular cognitive act have been used extensively. While empirical evidence shows individual differences can affect confidence ratings, internal consistency and reliability coefficients typically range between .75 and .90 across these studies. While a single item as we use here has some limitations, Stankov and Lee (2008) concluded that confidence is a separate concept that can vary across individual characteristics such as national culture and its measurement reliability remains high even when different numbers of test items have been employed (Stankov et al. 2015). The observed correlation between recall score and confidence (.33, see Table 5) is consistent with, although slightly lower than, previous validity assessments reported by Stankov et al. (2015) for similar relationships and supports convergent validity for our measure of confidence.

Extent of integration perceived.
Given that all students received representations of the same strategy (varying only in terms of which of three formats was in their envelope), differences in the extent of integration perceived should provide a reliable measure for whether the format has a systematic impact. Thus, students were asked, again on a scale of 1 to 5: "How integrated do you think the elements of strategy were?", with 1 being not at all integrated and 5 being extremely integrated. They recorded answers to both questions on their envelopes. Integration and inter-relationships across various strategy dimensions (e.g., customers, suppliers, internally) have long been argued to affect organizational performance positively (see Swink et al., 2005 for a review). This characteristic had been explained as part of the strategic management courses where the experiment was conducted in terms of alignment across functions and cooperation/collaboration with stakeholders, which matches how integration has been measured previously (O'Leary-Kelly & Flores, 2002;Swink et al., 2005).

Other measures.
Binary variables (where 0 = mono-media and 1 = multi-media) recorded both the presentation (From P + T) and the recall mode (To P+ T). Given that the sample covers multiple countries and strategy courses, a variety of factors associated with course demographics could be associated with differences in recall, confidence to discuss and integration perceived, such as whether English was the first language of instruction for most students in the course, their prior familiarity with strategic management concepts such as the Value Chain, their work experience involving the implementation of strategies, as well as the interval before the strategy was recalled. Thus, we also coded the sample for country location of the course (with the UK used as the contrast country), the programme level (0 = undergraduate, 1 = MBA), and the interval before students were asked to recall the strategy (measured in minutes).

RESULTS
Our analysis occurred in several stages. This initially involved understanding whether recall, confidence and integration may have differed in terms of locational demographic factors noted in Table 1 as well as the mode in which the strategy was received and recalled. This was followed by the more specific tests associated with Hypotheses 1-3.

Recall
Average recall across different geographical locations ranged from 3.0-5.9 (out of 10) when receiving a text-only version of the strategy and 6.3-8.8 for the picture and text format. The highest average scores when receiving text were observed in an undergraduate class in the United States and the Executive MBA course in the UK. The geographical variation in average scores (between receiving picture plus text versus from text only) matches the overall contrasts reported earlier and was at or above +3 points across each sub-sample (except for Tunisia, where it was +1.7 points). This consistent difference suggests that the subsequent regression analyses should include explanatory variables to account for potential countrylevel effects as well as measures of covariate factors noted above (programme level and interval before recall). Table 2 provides the average recall scores across the full sample as well as by different formats of strategy communication and recall. There was no evidence that the students did not undertake the task of memorizing and recalling the strategy conscientiously, given only 5.5% of the sample failed to recall at least 1 of the 10 strategy components (matching the 6.4% who recalled all 10). Table 2 about here *** Average recall for those students receiving the strategy in a multi-media format (7.08 out of 10) was 1.1-3.1 points higher than those receiving the text only version (averaging 3.92 out of 10). Numerous t-tests for differences in these averages (see Table 2) are significant at a p <.01 level, with effect sizes at times quite large, ranging from 0.5 to 1.4 (using the pooled standard deviations). For those recalling the strategy using pictures and text, the average score was 7.28 versus 4.04 for those recalling the strategy via text only, with the t-value (and effect size Cohen's d = 1.4) again suggesting a substantial effect across recall modes. While there appear to be substantial gains for those recalling the strategy in a multi-media format, the average recall in our sample for those receiving it in this format (7.37) is also significantly higher (p=0.002) than when receiving it in a mono-media text format (6.25) (effect size 0.5).

*** insert
Table 2 also illustrates that most students recalled the strategy in the same format as it was received (i.e., a strategy received in bullet point form would generally be reproduced using bullet points). While we provided two mono-media textual formats when communicating the strategy, overwhelmingly bullet point lists were the mode used for recall in this format. Very few students recalled the strategy in a richer extended textual form.
However, some recipients reproduced the strategy in a different format (e.g. Figure 1). While 57.6% of the total sample had received the strategy in a text only mode to consider and memorize, a significantly higher 62.5% of the recalled strategy representations were in the form of mono-media text, showing a potential predisposition towards text for strategy reproduction (Pearson χ 2 =692, df=1, p<0.001). *** insert Figure 1 about here*** Most marked in terms of recipients reproducing the strategy in a different format were those who received the strategy as a picture. Of the 483 students that received the strategy in a pictorial form, 89 (18.4%) reproduced the strategy using only text. This switching was more limited for recipients receiving bullet points or a paragraph where only 4% and 8.5% respectively shifted to recalling it by including a picture and text. Students' individual preferences or inclinations for text and pictures are unlikely to account for this shift in recall mode, since strategy presentation formats were distributed to the students randomly. This suggests that many students, when faced with the task of recalling a strategy, view a textual description (and particularly lists of bulleted points) as the logical, most legitimate or acceptable format, even when the strategy has been communicated to them in another mode.
Average recall scores were the lowest for those receiving the strategy in the mono-media paragraph form of text (irrespective of the recall format chosen), and significantly higher for both receipt and recall in the picture plus text format with effect sizes greater than 1 in both cases (see Table 2). The reverse also appears true. Incorporating pictures into a reproduction generates better average recall results than text alone. Students that received the strategy in picture plus text form and then reproduced it with only text exhibited lower recall accuracy relative to those who matched recall with the multi-media format that they received it in (5.78 and 7.37 respectively, d = 0.8). Overall, with respect to Hypothesis 1, this initial analysis suggests preliminary support, noting though that other potential influences are not controlled for in the t-tests.

Confidence to discuss and integration
Confidence to discuss varied across countries and programme levels, with Moroccan MBAs indicating particularly high levels, and NZ undergraduates low levels. Table 3 shows the average scores again reported by the format received and used for recall. In contrast to the numerous differences observed with respect to recall, we found only two statistically significant differences in average recipient confidence to discuss the strategy. These were for differences in aggregate, with slightly higher confidence when there was multi-media receipt or reproduction of the strategy (with 3.03 vs 2.76 and 3.04 vs. 2.74, t > 3.5, p<0.001). These  (Table 4) again show few significant differences in averages, with the effect sizes for these correspondingly smaller than for the recall scores, ranging from 0.1 to 0.6. Recipients receiving paragraphs (N=152) tended to perceive slightly more integration than those receiving bullet points (average scores of 3.16 and 2.91: t=2.5, p=0.012). Thus, the use of paragraphs may assist with perceiving integration across the strategy elements.
Overall, receiving the strategy as picture plus text exhibited higher average perceptions of integration amongst the strategy elements and these scores were significantly different to receiving text only (3.36 vs. 2.97, t = 6.0, with an effect size of 0.4)thus, some initial support for H3. This low effect size, however, indicates that the probability is only about 0.6 that the integration perceived by a random student receiving the strategy in a multi-media form is higher than a student presented with a text only version (see McGraw & Wong, 1992).

Regression Analyses
While the univariate analyses provide insight and suggest some support for the hypotheses, a more robust assessment of the relationships can be achieved through multiple regression, since the effects of some other factors can be controlled for. Table 5 presents the correlation matrix, including means and standard deviations for the variables used in the different models. *** insert Table 5 about here *** The multiple regression models were developed in stages (see Table 6) initially adding the control and then the focal independent variables associated with the hypotheses. Most country-level effects were statistically significant and may simultaneously capture a range of facets associated with the locations of the courses as noted above. Most models explain a significant proportion of the variation in the dependent variable (F-value > 20), although it is much lower for Model 6. In line with this lower adjusted r-square, differences in Integration perceived were not as evident across countries. All models were checked to determine if receiving the paragraph text-only form of the strategy had a separate systematic effect. Only in Model 6 was this estimated coefficient different (although only at a 5% level), suggesting that it is valid to aggregate the bulleted lists and paragraph into a pooled category of 'From Text' in all models. *** insert Table 6 about here *** Tests for multicollinearity demonstrated little evidence of this having an impact on estimates, with variance inflation factors below 2.0 and condition indices below 10 (Hair et al., 2014) in the models reported. Endogeneity issues were also assessed (Semadeni, Withers to reduce this. The potential for simultaneous relationships between the focal dependent variables is likely the only concern, that is, Integration affecting Recall and vice versa.
However, the use of a two-staged least square regression procedure did not indicate different signs or significance levels for key coefficients. Checks for heteroscedasticity and nonnormality of residuals indicate that these potential problems were not present either.
The regression coefficient estimates for programme level were positive and significant for models 1-4. This would indicate that experience in a business environment and greater familiarity with strategy concepts (as would be expected of MBA/executive students) enhances an individual's ability to comprehend, remember and recall presentations of strategies, especially when the strategy has close alignment with a framework like the Value Chainwhich most MBA students would have been more familiar with than the undergraduates in the sample when the experiments were conducted. This, in turn, could boost confidence. Therefore, this finding suggests that the Active Processing Assumption may hold, whereby multi-media presentation may aid effectiveness and confidence if it assists learners in connecting new knowledge to prior knowledge. The effect of programme level is, though, small relative to the variation in the dependent variable, about 40% of one standard deviation for recall.
The interval between experiment distribution and participant recall had a small negative estimated effect [significant in Model 1 only], likely reflecting memory lapses particularly when there was over 20 minutes before being asked to recall the strategy. The relationship appears linear with any effect size quite small. With regard to Integration perceived (Model 6), a longer interval was associated with slightly higher integration scores, suggesting that a greater delay may lead students to forget aspects of intrinsic load relative to germane load; therefore simplifying the message retained to one that appears more integrated on recall.
Interval length had a similarly small effect size in accounting for differences in integration perceived.

Hypothesis 1.
As can be seen in Models 2 and 3, recall scores for receiving the strategy via a picture plus text (From P+T = 1) were significantly and systematically higher (β=1.5) relative to the default category (From P + T = 0, when the strategy was given to the students as a text onlyeither bulleted points or a paragraph). Similarly, the multi-media reproduction of the strategy (To P+T) coefficient was significant and positive. Separate tests for mixed receipt/recall modes indicate that these outperformed receiving and recalling the strategy with text only. This suggests that the process of working with information about a strategy in textonly form, either as part of memorizing it or recalling it, has a significant effect (a 1-3 point decrease in recall on the 10-point scale).
Even where some students may have been inclined to recall the strategy textually, a multi-media pictorial presentation of the strategy benefitted them significantly in aiding their retention of the strategy elements in comparison to learning about the strategy via a paragraph or bulleted points and recalling it in this same type of format. Reproducing the strategy via a picture and text showed clear benefits for recall (evidenced in the positive significant To P+T coefficient), both when given the strategy in text or in multi-media picture plus text format matching the findings from Table 2.
The effects of textual presentation and recall of strategy were also tested separately in the portion of the sample where there was a shorter time interval (< 15 minutes) between receiving and memorizing the strategy and recall, with the significant positive coefficient estimate for From P+T being reproduced with a similar effect size. Thus, the constraining effects of textual presentation and recall dominate even when memory retention issues should be less present. We therefore find strong support for H1 that multi-media representations of a strategy are more effective than mono-media in enabling recipient recall. Our analysis adds to this finding, though, by suggesting that multi-media recall also has benefits for strategy recall effectiveness.
Model 3 assessed whether a student's perception of the strategy's elements being interrelated was associated with greater recall. This variable was positively related to recall and highly significant with an effect similar in size to programme level for the average integration perceived. It represents a clear avenue for future research into ways to develop enhanced recall.
Hypothesis 2. Models 4 and 5 assess the students' confidence to discuss the strategy that had been presented to them. The strong association between the picture plus text communication format and recall score (Models 2 and 3) did not generate indications of multicollinearity problems when including both sets of variables here, with coefficient estimates remaining stable and variance inflation factors low. Checks with models using different combinations of variables found no substantial changes in coefficient estimates.
Confidence scores for MBAs were at consistently higher levels than undergraduates, when other factors such as country factors are controlled for. This suggests that confidence may vary more based on each individual's characteristics as has been noted by Stankov et al. (2005). As with Models 1-3, the coefficient for interval suggests that a longer delay since first viewing the strategy had little effect on whether students were confident.
As would be expected, those students who were able to recall the strategy more accurately exhibited significantly higher confidence, with this variable accounting for about 20% of the variation explained. However, when this effect is compared with the larger magnitude for the coefficient for perceptions of integration in the strategy (p<0.001), it suggests that an individual's confidence to discuss a strategy is only in part associated with their ability to recall the strategy elements accurately and to a greater extent dependent on perceptions about the strategy as a whole and connections between elements. In model 4, there was no significant effect associated with receiving the strategy in a graphical format, which does not support H2 that multi-media presentation is more effective than mono-media in promoting recipient confidence.
Additional tests were undertaken to understand this result, including considering whether some relationships with confidence might be moderated (Model 5). Two primary effects were considered: first, whether programme level may interact with variables such as presentation mode or interval, and second, if the effect of a picture + text presentation mode may be moderated by other factors such as recall score or whether integration is perceived.
Standardized values were used for all variables associated with the interaction terms. These checks indicate that the negative effect of an increased interval before recall occurred mostly for undergraduate students in our sample (see Figure 2), with the MBAs not experiencing this decline in confidence. Stable confidence for MBAs again suggests that the Active Processing Assumption (mentioned earlier) may be occurring, that is, these students seemed able to build on connections between prior (such as the Value Chain) and current knowledge (greater experience). It is also possible that students in MBA/Executive programmes could feel expected to demonstrate confidence when involved in strategic discussions irrespective of other facets, a social desirability effect. In either case, this result suggests that aspects of cognitive load may operate differently depending on familiarity with strategy, frameworks and communication modes, which is consistent with cognitive load theory. A smaller effect of the combination of how the strategy was received and the student's level of recall on confidence was also found, indicating that multi-media presentation boosted confidence in combination with the student's recognition as to whether they had more effective recall. It is also possible that recall accuracy and integration perceived mediate the effect of strategy presentation on confidence but such analysis went beyond our hypothesis. *** insert Figure 2 about here *** Hypothesis 3. The regression analysis for the 'extent of integration perceived' as the dependent variable [Model 6] resulted in a relatively low adjusted r-square (0.08). It should, however, be noted that this variable exhibited less variance than the dependent variables in the other models. Significant estimated coefficients were primarily a few locational effects and interval before recall, but this effect was again very small. The multi-media presentation mode for the strategy coefficient (From P+T) was positive and significant, suggesting that receiving a strategy pictorially aids in perceivng interrelationships and integration between the strategy elements. The coefficient estimated for recalling the strategy with pictures and text (To P+T) was positive but only significant at a 5% level. Thus, there is some support for H3 that multimedia presentations are more effective in enabling recipients to perceive strategy interrelationships. The small effect sizes and low amount of variation explained, though, suggest that future research seeks to better understand how central characteristics of a strategy as a whole, such as integration, can be communicated most effectively. This could include adding questions that assess the different types of strategy integration that are perceived to be present.

DISCUSSION
Overall, our experiment provides strong evidence in support of using multi-media picture plus text presentation, rather than text only, for communicating strategy, but not necessarily in the ways that we hypothesized. It has enabled us to test assumptions, and advance our theoretical understanding, of how students' learning can benefit from the multi-media presentation of strategy, but also identified some of the challenges for doing so.
Firstly, our results confirm Hypothesis 1 that picture plus text communication of strategy is superior to text-only communication in terms of student recall. This was as much as 3.3 points (out of 10) better overall than with bullet points or paragraph forms of text-only communication. There are some variations across the sample in terms of different educational levels, with those with more experience (MBAs) showing greater recall, and some countries exhibing higher scores; but these effects do not alter the veracity of this general finding.
In addition, we established that variations in recall were explained by the mode in which participants chose to reproduce the strategy. Where students received the strategy as a text, and reproduced it by combining drawing and text, their recall performance was significantly higher than those who reproduced the strategy in a text-only format. Conversely, where students received the strategy as picture plus text but chose to reproduce it as text only, their recall performance was significantly lower than those reproducing a picture plus text.
Secondly, the extent to which the integration of strategy elements was perceived had an effect on recall scores. Picture plus text communication was associated with higher perceived integration scores, confirming Hypothesis 3, and this echoes findings from clinical trials in the psychology literature (Dansereau & Simpson, 2009). Overall, superior recall of strategy communication is explained by the mode of receiving the strategy, the mode of recall, and perceptions of integration.
In contrast, for our second hypothesis, we posited that the receipt of a strategy communication in the form of a picture plus text would lead to greater student confidence, but we found no support for this. This is surprising as students receiving strategy communications as picture plus text are likely to have superior recall to those receiving the communications as text only, and are likely to have a higher ability to perceive interdependencies. This suggests that the advantages of using multi-media picture plus text are more social (in terms of perceiving integration benefits) and cognitive (in terms of easing recall), than emotional (in terms of building confidence). However on closer inspection, our results also show that students demonstrated lower confidence with the multi-media format, than with text only, when recall levels are low.
One explanation for this is that the text-only recipients' lower level of competence robs them of the metacognitive ability to recognize their low recall or its significance. Our results suggest that text-only communication provides fewer clues than picture and text to the complexity of a strategy, to poor performing recipients, and so presents more opportunity for false optimism and errors. This is now referred to as the Dunning-Kruger effect (Kruger and Dunning, 1999) whereby a subject's lack of understanding actually increases their confidence.
Or, as Charles Darwin (1871: 3) put it, "Ignorance more frequently begets confidence than does knowledge."

Limitations
While a strength of our study is that it is based upon a large number of responses, a limitation is that the measures associated with hypotheses two and three are based on self-reporting of ability. Podsakoff and Organ (1986) note that such self-report responses are asking students to engage in a higher-order cognitive process and can be affected by consistency motifs and social desirability problems. A mitigating factor in our experiment is that these perceptions are not removed in time from the task on which they are based and, thus, in terms of selfreported confidence this may be less of a limitation as students were immediately reflecting on their ease/unease with engaging in a proposed discussion. Social desirability (being seen to be confident) is more likely to be a concern, particularly for MBA students, and could have led to some systematic escalation in scores for both measureshowever, this could be controlled for to a degree in our regression analyses. Problems related to a consistency motif are less likely given only two questions are asked and confidence and integration should not be associated with an immediate connection for students within the short timeframe for response. Nonetheless, these caveats suggest that our results for hypotheses two and three should recognize that studies with students indicate a cautious "yes" (Pike, 1996) for self report measures being closely aligned with independent measures. Finally, the typical length of teaching sessions (50 minutes) limited our ability to assess recall over longer periods of time.

Theory development
Cognitive Load Theory enabled us to theorize how the oft-promoted use of multi-media picture plus text forms of communicating strategy may be more effective than conventional text-only modes. Our results provide general support for CLT and CTML of communication in learning strategy. The strong finding that multi-media communications of strategy are superior to single media communications supports the Dual-channel Assumption, that we have separate channels in our working memory to process pictures and text, and engaging both can enhance learning by reducing overall cognitive load. In addition, we find support for the theory's Active Processing Assumption, as MBAs (who in our study were more familiar with the generic Value Chain frameworkthe basis of our picture), outperformed others in terms of recall. Using multi-media communications helps recall by assisting learners to connect the intrinsic load of the task to knowledge already held in their long-term memory, and this helps to build representative schemas. This can help us understand the value of, and reliance upon, frameworks in strategy development.
However, our findings did not confirm all our hypotheses, nor were our findings completely explained by CLT and CTML. CLT and CTML seek to explain learning when exposed to different media of communication, but do not take into account how students may voluntarily recall communications potentially using different modes. Our results show that students do not always reproduce communications in the form they are received in as, in many cases, students chose to 'switch modes'. This resulted in significant variation in recall performance. In particular, students who chose to reproduce a text communication as a picture and text, demonstrated superior performance both in terms of recall and perceptions of integration. Although CTML has not taken into account voluntary recall modes, it may nevertheless help explain this result in terms of whether students chose to use one or two mental channels. Where students draw on two mental channels to better understand a strategy they may also find it beneficial to use the two channels to reproduce it, as two cognitive channels may reduce overall cognitive load. For instance, performance improvement likely occurs because drawing a picture in addition to writing encourages people to construct the data in ways that enable them to better perceive and articulate interrelationships. Using multiple channels can highlight omissions in their memory, prompting a guided search to piece things together to fill the gap, and sparking memory recall.
This effect might also work in a detrimental way if recipients receiving a picture and text strategy choose to reproduce it in a text-only format. Over 18% of those recipients in our experiment chose to do this and it resulted in a significantly lower recall performance than those who reproduced a picture and text. This confirms our thinking, related to CTML, that reducing the number of mental channels to just one in the act of reproducing a strategy leads to a loss of information. Therefore 'mode switching' in terms of recall can affect learner performance both positively and negatively and should be taken into account in CTML.
CLT and CTML also do not take into account student levels of confidence once the communication of something as complex as a strategy has been received. Our results highlight an inverse relationship between overall performance (recall and integration awareness), and student confidence, which is significant for weaker students. Whilst the cognitive load of understanding a strategy may be reduced through the use of dual-mode communications, we theorize that it may also convey a greater sense of the complexity of the strategy. The Dual-channel Assumption may create greater dissonance in students by prompting them to think that they cannot remember everything they should.
In other words, having seen a multi-media picture plus text representation of a strategy may make it easier for students to recall that they have forgotten key elements of that strategy. In this sense, there may be an interaction effect between different channels so that the Dual-channel Assumption may not have entirely positive outcomes if student confidence is an important consideration in reproducing learning. This insight directs attention to a social element to the learning process; that students may need to further communicate their learning to others and this may cause them to 'mode switch'. In our experiment, most 'mode switching' occurred from the receipt of a multi-media strategy communication to a text reproduction (18%), with many fewer switching from text receipt to picture reproduction (5%). The dominance of 'mode switching' from picture and text to text-only reproduction may have been a result of students seeking a format with which they feel more confident. By reproducing information at the lowest level on Bloom's Taxonomy, students may have felt more confident of accuracy about the 'main points' and felt they were avoiding the complexity of higher-order concepts of interconnections. This might have reduced their social exposure to error and there may even be the thought that, by presenting knowledge as bullet points, the audience will 'join the dots' for themselves, thus avoiding potential conflicts of interpretation. Furthermore, bullet points may feel more comfortable as a legitimate communication technique for strategies, unlike drawing, as their experience would predominantly have been textual descriptions. Of course, it might also be that some of the students in question failed to see the importance of integration, a crucial part of understanding any strategy, perhaps due to a 'spot-light' effect of learning to see data points only, rather then relationships, and this would be critical for teachers to recognize and address.
The tension for students exhibiting superior performance in terms of recall and integration, and yet not showing higher levels of confidence, highlights a limitation of CLT and CTML, as these theories do not take into account the ways in which students may prefer to communicate their learning. Those theories focus on individual student performance in terms of recall competence and ignore student confidence to communicate to, and engage with, others, which are higher levels of outcome in Bloom's Taxonomy. We might suggest, therefore, that there may be other student characteristics as well, such as emotional states, that also play a role in students' capacity to communicate their learning. CLT and CTML could be extended to take into account student capacity for conveying learning.

Practical implications for strategy educators
Strategy is complex and difficult to communicate and there have been suggestions that visual representations aid learning for managers and students (Kaplan & Norton, 2000;2005). Our results provide strong empirical support for the use of multi-media communication of strategy by strategy educators. Interacting with picture plus text modes of strategy communication, like frameworks, will greatly improve recipient recall over the use of text alone. In addition, as strategy is generally multi-dimensional in nature and connections are critical for understanding the whole, our results also show that multi-media communication enables students to perceive higher levels of integration than using text alone. For these reasons, strategy educators should be encouraged to make greater use of picture and text presentations of strategy and resist recoursing to mono-media presentations, such as the lists of bullet points, common in PowerPoint presentations.
We also found that students perform better if they subsequently communicate their learning in multi-media forms although, contrary to what advocates of multi-media strategy communication have promoted, there can be a loss of confidence amongst weaker students who are receiving and communicating strategy in multi-media formats that include pictures.
The art of drawing is generally marginalized in formal business education (Bridgman et al., 2016), so students may be far less comfortable using this form of expression. Our study revealed a general predilection for students to recourse to text-only reproductions of strategy.
It may be that they are less familiar with the media, feel that drawing is more difficult, and could interfere with their estimates of the correctness of their solutions (Efklides, 2013). We show that this automaticity of representation in text is damaging to learning about strategy, as text-only communication can reduce recall, reduce perceptions of integration, and may instill false optimism, or the Dunning-Kruger effect.
To mitigate this, strategy educators should legitimate students, not just at receiving strategy in multi-media formats, but also in communicating it: drawing their understandings to help them develop their skills. Just as proponents of 'design thinking' advocate learning through prototyping, questioning and pivoting (Dunne & Martin, 2006), strategy students can be encouraged to prototype solutions to cases or problems in pictorial formats, perhaps using strategy frameworks as a basis, but customizing these to incorporate their own insights and discussions. While this can be challenging and lead to declining confidence in the solutions presented at first, we believe it is far better to build up confidence in this way than unwittingly promote a false optimism based on superficial mono-channel understandings.
Indeed, if the aim of strategy educators is to enable students to become more competent and self-assured in understanding, developing, and communicating strategy, then much greater engagement with drawing as a method of communication and reproduction could be an extremely helpful pedagogical approach. There has been much written in recent years about strategic management courses having become mere 'kit bags' of generic theories and frameworks, suffering from poorly thought-out theoretical grounding, and how strategy education should be made more practically applicable. Some researchers have advocated less emphasis on theory-based approaches and more on practical skills and implementation toward this aim, others have advocated using better theory. The research reported on here suggests that the greater use of multi-media communication of strategy by teachers and students can help us achieve both of these solutions.
We can achieve a more theoretically grounded and practically applicable education that enables students to see similarities and differences between organizational strategies and to understand general themes, but also to probe and question them in the light of particular differences and practical realities. Strategy educators can do this by actually using and applying multi-media communication, and encouraging students to actively draw and annotate their workings. Indeed, we argue that presenting and interrogating a strategy using pictures and frameworks, using technology like whiteboards, flip charts, tablets, and document cameras, would be an excellent way of promoting "critical thinking that can lead to sound judgements" in a strategy classroomas has been called for recently by others writing in this journal (Priem, 2018: 1).
There is a strong argument here for students drawing, rather than reducing content to bullets and text, and drawing 'freehand' on paper or a tablet, and not being constrained by generic graphics packages. The freedom to think and express through drawing also has the benefit of allowing students to interrogate strategy communications received as bullets or text only through the construction of pictures to identify linkages, and also to reconceive pictures and text, where received pictures may be reproductions of generic strategy frameworks.
These multi-channel processes encourage higher levels of engagement from students so improving learning and critical thinking, and these are a higher order on Bloom's Taxonomy than recall alone. This approach can mitigate the criticisms made against the current state of strategy teaching as merely generic boxes of tools that participants memorize but do not know how to use (Mintzberg, 2004;Bell et al., 2018), the need for more practically applicable strategy education (Albert & Grzeda, 2015;Clegg, Jarvis, & Pitsis, 2013), and an emphasis that is less theory-based and more oriented to practical skills (Grant & Baden-Fuller, 2018;Lindsay, Jack, & Ambrosini, 2018). Encouraging students (and managers) to engage in multi-media strategy communication with their peers, incorporating pictures and text, will allow them to benefit from what cognitive psychologists are starting to describe as a 'drawing effect' (Wammes et al., 2016). This drawing effect promotes better recall, increased critical insights, greater awareness of connectivity, enhanced creativity (to think in addition to and beyond generic frameworks), may build confidence and should make people less subject to false optimism: all things that will help them become more capable strategists.

APPENDIX 1 Different representations of the same strategy used in experiment Option 1. Strategy in bullet point text form
• Continue to outsource distribution to customers to reduce costs • Develop preferred supplier arrangements with three key suppliers • Investigate cutting out retailers and selling direct to customer groups • Work with famous designer to create high-end complementary brand • Engage in viral marketing to drive demand from customers to our retailers

Option 2. Strategy in paragraph form
Our strategy for the future is to continue to outsource our distribution through partner organizations to help us to reduce our costs and pass on lower prices to customers. We will continue to develop preferred supplier arrangements with three of our key suppliers. We will further investigate cutting out retailers in areas no longer regarded as a priority in order to sell direct to customers in these areas. Another key aspect of our strategy will be to employ a famous designer to create a high-end brand that will be complementary to our existing brands. And, finally, we will engage in a viral marketing campaign to drive demand from some customer groups who will put pressure on existing retailers to stock our products.     Reported t-values refer to differences in 2 adjacent cell averages either above or to the left.    Values reported are unstandardized β with standard errors in parentheses. + p<0.05, ** p<0.01, *** p<0.001 (two-tailed test).