Performing PowerPoint lectures: examining the extent of slide-text integration into lecturers’ spoken expositions

ABSTRACT The PowerPoint-assisted lecture (slide-lecture) is a common lecturing approach in higher education, in spite of much criticism of its use. Its popularity is facilitated by its affordances for multimodal instructional design, e.g. text with images and speech. Little is known about the integration of different semiotic modalities within the instructional communication practices of slide-lectures nor the learning conditions that they create. Given that text bulletpoints are ubiquitous in slide-lectures, and may impose linearity into instructional communications, this study explores the extent to which lecturing speech is systematically coordinated with slide-text. Eleven slide-lectures given in psychology departments across the UK were recorded and transcribed. Patterns of semantic matches between speech and slide-text were analysed to produce similarity scores for each lecturer. Lectures were scored using an integration scoring system of 0–1, with 1 indicating a perfect match of speech and slide-text. There was significant departure from a systematic voicing of the slide text (i.e. reading off the slides). Two characteristic speech–slide relationship styles were identified. The ‘referent’ style is one in which the slide is an object of reference for the lecturer to comment on and the ‘scaffolding’ style is one in which the slide-text is blended into the spoken narrative. Consequences of the lecturer’s coordination with presentational slides are discussed in terms of the learning environment it might produce. It is suggested that whichever relationship a lecturer has with their slide-text, students might benefit from the integration being consistent. Abbreviations: DA: Discourse Analysis; SPP: Secondary Pointing Procedures; ANOVA: Analysis of Variance

slide is achieved during lecturing. We shall focus on the interplay between speech and slide-text during slide-lectures in order to illustrate the performance of lecture-slide mediation through examining how far lecturers systematically mirror their slide-text in their speech. The research questions guiding this study are as follows: to what extent do lecturers integrate slide-text in their speech in a systematic manner and are there characteristic ways in which this integration is achieved?

Materials and methods
Data collection took the form of a naturalistic video observation of 12 undergraduate psychology lectures given across the UK during the academic year 2009/10. In a discipline that deals with a mix of methods and representations, a single, iconic and introductory topic was sought for observation. Attachment theory in developmental psychology was selected as it is typically covered using a wellestablished expository pattern. Ethical approval was confirmed prior to carrying out data collection.
Rather than focus on a single institutionwhich may impose idiosyncratic lecturing practices across its programmesan opportunity sample was gathered from a mix of UK universities. Lecturers in 18 psychology departments were approached and, of those, 12 fit the criteria of teaching first-year attachment theory and were willing to participate.
Participants were given the option of either making a recording of their attachment theory lecture using a small, portable video-recorder sent to them in the post or allowing a researcher to visit and record the lecture using the same device. The video-recorder was sent to 11 participants in advance of their lecture, together with instructions. The instructions requested that the recorder be set up in a position that captured the main display screen or focal point in the lecture theatre, along with the lecturer's speech, but not necessarily including the lecturer themselves (unless unavoidable). It was also required that students' faces were not visible on the recording, and that students were made aware of the recording prior to the start of the lecture. In one instance, the researcher made the recording by positioning themselves at the front of the lecture theatre with the recorder pointed at the display screen. In total, 12 lecture videos were collected from 12 different lecturers, each comprising a slide presentation accompanied by the lecturer's speech.
All lectures were transcribed verbatim, with the exception of one that was excluded from the analysis owing to the exceptionally poor quality of the recording. During the transcribing process, slide transitions were used to split the speech into sections, such that each slide was displayed side-by-side with the accompanying speech. Where a lecturer changed slides mid-sentence, that sentence was divided between the slides at the point of transition. The slide-text was also transcribed for ease of analysis.
It was noted that slides included many other types of representation: graphs, diagrams, photographs, images, numbers, formulae, videos, web links and dynamic diagrams. Of the 2095 distinct slide-elements found on 413 slides, text-elements (in the form of bulletpoints, structural headings and quotations) comprised 91.57% of the slide-elements (1923 of 2095 total slide-elements). The next most prevalent representation was photographs (appearing 68 times; that is, 3.24% of the total slide-elements), which are dealt with separately in Hallewell and Lackovic (2017). Table 1 describes the characteristics of the lectures.
To consider the integration of slide-text with speech, only the slides that contained these 1923 text-elements were selected for analysis. Thus, 336 out of 413 slides were analysed.

Results
Each individual text-element was labelled alphabetically according to the position in which it appeared on the slide. It was assumed that slides are intended to be read from top-to-bottom and left-to-right. For instance, Figure 1 shows an example of such reading along with the coding of its four text-elements.
Identification of the integration of slide-text with speech was carried out using a discourse analysis (DA) framework (Coulthard 2014) that compared the semantic content of the speech with the semantic content of the text. The identifiers of speech-text integration are described in Table 2. This table is an expansion of Knoblauch's (2008) 'secondary pointing procedures' (SPPs) in which speech points to text without the use of physical pointing methods. These identifiers were selected from the limited frameworks available for analysing the communicational practices of slide-lectures. Some identifiers were added to reflect the practices of lectures specifically, as Knoblauch's identifiers were developed from business presentations. It should be noted that, although the observation of animation schemes would have been possible for some lecturers, the inconsistency of their use rendered them unreliable as a pointing procedure. Using these SPPs, the speech transcripts were scrutinised alongside the slide-text to establish where integration occurred. The speech that integrated the text-element(s) was alphabetically coded accordingly, such that the coding produced an 'expected' pattern (slide) and an 'observed' integration pattern (speech). Figure 2 is an example of such coding.
Where a lecturer integrated two text-elements, for instance by saying 'these two points', the speech was coded with both associated letters alphabetically. Where the speech integrated more than two text-elements, for instance by saying 'this slide', it was considered integration of the whole slide and therefore was not coded.
Note that this analysis concerns only speech in which the slide-text was being addressed. Speech that develops on the slide-text, explains and expands on it (in which case the speech is related to the slide-text but is not immediately identifying the text to be attended to), housekeeping interactions ('can everybody hear me?') and tangential speech was disregarded. Although this kind of speech is relevant to the student's general topic understanding, what is important to this study is the extent to which the speech assists in the initial identification of relevant slide-text.

Reliability of the SPPs
The SPPs were given to a colleague who coded a randomly selected 10% of the slides and their associated speech for each lecturer. It was noted for each text-element whether both the researcher and this colleague believed the speech was integrated with the slide-text (scoring 1) Spoken words that are also present in the slide-text (Knoblauch 2008, 87); for example, reading entire sentences from the slide or simply stating the significant words present in the text.

Itemisations
Providing there is more than one element present on the slide, the speech addresses the structure of the slide and the pattern of the elements within. For instance, when displaying a list, by saying 'first' the speaker points to the first text-element and by saying 'then' they point to the next. Direction and demonstratives The speech directly addresses the element such as 'this notion' or 'these things'. Reformulating the text/'mangling' Although the concepts are the same in speech and text, the speech can be so different in structure and terminology that they are two separate entities that provide the same semantic message.  or not integrated (scoring 0) with the speech for each coder. The specific SPP used was not noted. Substantial agreement was found using the Kappa statistic to determine consistency amongst coders: Kappa = 0.844 (p < 0.001). Thus, confidence was high that the process for judging integration was reliable.

Expected versus observed pattern of integration
The 'expected' and 'observed' patterns represent the integration of the text-element on each slide for comparison. For example, the expected pattern of the slide in Figure 2 would be A, B, C, D, but the observed pattern was B, C, D. Figure 3 represents the 'expected' and 'observed' patterns for the most consistently matching and least consistently matching lecturer. The strings of letters representing the expected and observed patterns were converted into a statistic, here referred to as an 'integration score', which expressed whether and how closely a lecturer matched or did not match the pattern of their slides with their speech. Note that in referring to 'matching' or 'not matching' the slide's pattern, there is no implication that these results directly support judgements on the pedagogical value of matching.

Scoring speech-slide integration
A statistical model was employed to represent the extent to which the observed pattern of integration matched or deviated from the expected pattern. The strings of letters were compared using a string-matching or edit-distance algorithm, such as those designed for spell-checking or text-matching, and which can easily be adapted for different purposes such as for measuring errors on text-entry tasks (Soukoreff and MacKenzie 2001).
The Levenshtein string distance statistic measures the 'minimal number of insertions, deletions and substitutions to make two strings equal' (Navarro 2001, 37) where all 'operations', or differences between strings (for example, the word 'Levenshtein' versus the word 'Lvenshfeins'), gain a score of 1 (in the example, the deleted 'e', the substitution of the 't' with 'f' and the insertion of an 's' imply a score of 3). The higher the score, the more changes would need to be made to one string to make it match the other. Although there would be no insertions, as the analysis focused only on what existing text-elements were integrated rather than examining addition of material, it was useful to be able to measure together 'deletions' (where a lecturer missed out a text-element in their speech) and 'substitutions' (addressing text-elements out of order).
The slide provided the 'expected' string of letters against which the 'observed' string of the lecturer's speech integration was compared. These two strings were processed in a spreadsheet containing a macro for the Levenshtein edit distance algorithm: each slide was thereby given a Levenshtein distance score for comparison. For example, comparing the strings in Examples 1 and 2 in Table 3, there is a greater difference between the expected and observed strings in 1 than in 2. In 1, the string was not repeated at all, meaning the Levenshtein distance equals the length of the string (10). On the other hand, in 2, there is only one letter missing from the observed string, meaning the distance between the two strings is 1.
This score alone does not take into account the respective length of the strings; there will be a greater difference if one item is deleted from a short sequence than from a long sequence (Ainsworth, Clarke, and Gaizauskas 2002). To account for length, Levenshtein distance scores were scaled (divided) by the length of the expected string sequence (in Examples 1 and 2: 10 and 4, respectively) to allow comparisons for the patterns on a 0-1 scale. In order to provide a more meaningful statistic, this 'similarity' measure, which accounts for string length, was subjected to a reordering of the scores to provide a 0-1 scale, where 1 represents an exact match and 0 represents infinite difference. This was achieved by adding 1 to the scaled Levenshtein score and then dividing 1 by this sum. The formula for the similarity measure is as follows: It should be noted that absolute zero is impossible here, as to receive zero, the scaled Levenshtein distance needs to be above 1. This score is only achievable if there are different letters added to the observed string: for example, adding KLMN to Example 2. This would not represent integration of the existing slide-text; rather, it would represent the addition of slide-text in the speech. This is impossible in the present case because, although a lecturer can add speech that is not related to the slide-text, such as explaining, introducing or tangential speech, there is no associated slide-text to reference. The lowest similarity score received for any of the slides was 0.33 and the highest was 1. Mean scores were calculated for all of the slides for each lecture to give an integration score in ascending order, as shown in Table 4. It was considered that the closer to 1 the mean, the more systematic the lecturer was in their integration.
A one-way analysis of variance test (ANOVA; e.g. Lazaraton 2002) was applied to the similarity scores for each slide of each lecture to test for differences between the lecturers in the extent to which their observed patterns matched their expected patterns, using the individual slides as the population and the lecturer as the factor. Lecturers differed significantly from each other in the similarity of the pattern of speech to the slide-text: F(10, 364) = 3.801, p = < 0.001.

Considering the lecturer's relationship with the slide
The next analytical step was to describe qualitative differences in slide-lecture practice. Schnettler (2006) identified two activities by which a presentation can be orchestrated: 'translating' (deciphering the slide-text) and 'conducting attention' (directing attention to particular elements). It was noted during the analysis of integrations that there were other practices not covered by these two categories: the lecturer contradicting the slide-text or highlighting why it was important. In order to consider whether the speech might reveal anything about the lecturers' relationship with their slides, a qualitative DA approach was taken. This involved analysing examples of integration and the speech surrounding integrations, focusing on what the lecturer appeared to be doing with their integration of text.

Caricatures of the slide-lecture
The analysis focused on the actions that the speech performed in relation to the slide-text, using Schnettler's (2006) activities as a starting point, thus the extent to which the speech 'translated' or 'conducted attention' was identified. This analysis was also intended to uncover any further activities that were carried out by the speech in relation to the slide-text. The highest-and lowest-scoring lecturers were treated as the two extremes in approaches to slide integration. Here, we considered the two lecturers who were quantitatively different in their approach to integration to consider what the qualitative differences between them might be. Jackson had the highest integration score: he integrated his slide-text most consistently. This consistency might imply that his approach was closest to the 'reading off the slide' practice, or to Schnettler's 'performer' approach. Indeed, on closer inspection, this was often the case; for instance, the slide in Figure 4 shows very close matches between the speech and slide-text. This is particularly evident in his integration of elements E, F and G.
It appears that the elements are being spoken about, rather than that the text is somehow being performed. For instance, where the speech first integrates slide-element E, the lecturer says, 'and here's some fairly obvious basic facts about it', which provides a context for the following speech: he will be reading out these obvious and basic facts. Such an act might be considered to be conducting attention to the slide-element, using a directive to identify that the text will be integrated. Yet he wanted to do so in order to classify the information as being 'obvious' and 'basic'. However, once the slide-text is read out, or verbalised, he follows it by questioning the text using information previously learned. He says, 'now we've got something else, well hold on a minute', before verbalising the slide-text again, which suggests to the audience that what is written on the slide is questionable in some way. He follows this with suggesting that if we attempt to pick it apart it is rather difficult. The lecturer is almost disagreeing with the slide-text, or else pointing out that, although such a point has been made, it does not necessarily mean that it is a simple fact to be digested. Rather, students should be considering it in light of what they learned in the last lecture.
Jackson includes some degree of translation into the speech, i.e. in explaining the text or otherwise deciphering it for the audience. For instance, where he integrates element C, he follows with a translation, 'so, you know. . .'. Mostly, the lecturer seems to signal that the text is self-explanatory, and as such he is not using the text-element as an object that needs to be explained to students. Rather, his relationship with the slide appears to be based on his indicating or referring to specific elements in order to assess them. There were many further examples of this lecturer talking about the slide-text in such a way, for example (instances of integration are in italic): Here's some features of it I think are worth just reminding you about. Attachments are not just to anyone. Now we notice that, he's saying that but remember at the very beginning you get this social responsiveness for the first couple of months, but that's not attachment. They are selective, focused on specific individuals who elicit attachment behaviour in a manner, form and extent that is not found in the child's interactions with other people. I know what it means because my mother can still really get to me. Ha-ha, I shouldn't admit that, should I? Attachments provide comfort and security, the outcomes of being close to the attachment object. That's fair enough.
Before reading out the list of features that appear on the slide the lecturer notes that he thinks these features are important to remember. In this way, the lecturer might be signalling the importance of the slide-text to the general thesis of the lecture. The lecturer follows the reading of the first item on the list by linking back to what was previously learned to help explain or translate the statement; that responsiveness in attachment is more than just general social responsiveness. It is noted that later the lecturer agrees with the text: 'that's fair enough'.
Although this lecturer most consistently addresses his slide-text, he does so predominantly to question or provide an assessment of it. The lecturer does not use his slide-text as a script to tell him what to talk about; rather, he uses it as an artefact of reference. This 'referring' style is particularly salient when compared with the practices of the lowest-scoring lecturer in terms of integration, Leaman, who did not regularly assess the slide-text. Instead, the slide-text is more subtly woven into her speech. An illustration of this is provided in Figure 5. Here, if the speech were read alone, it might be impossible to tell that there was text on the slide at the time.
The boundaries between slide-text elements in the speech are less marked, as evidenced by the first speech-sentence. Element A is merged together with B in the same speech-sentence. This merging is also evident in her integration of elements B and C. Here, she skips the majority of the text to merge the two phrases written in italics on the slide -'Safe haven' and 'Safe base'before going on to define or translate these phrases separately afterwards: 'so safe haven behaviours are. . .'. Here, the lecturer is stating the concept before explaining it, such that the students need not see the concepts on the slide. What is more evident here is the extent of translation of the slide-text being carried out, as much of the slide-text is explained in other terms without explicitly referring to it. Also, the lecturer appears to make more of an effort to combine the speech and text into a single story, compared to Jackson's approach which served to separate them. This is clear in the extract below in which the lecturer more subtly integrates the words appearing on the screen (slide-text integrations are shown in italic): Now when we talk about attachment, often people know what we mean when we talk about parent-child attachments or child-parent attachments. And most of this work is based on how the infant expresses emotion and how the caregiver responds to that emotion. So what we need to think about is when we're looking at parental and child interactions and we're looking at this dyad interacting together, how do we conceptualise what the attachment is? This bond between parents and their children, how do we conceptualise it?
The slide-text is woven into the lecturer's sentence, serving to convert the text from solitary phrases to a more articulate narrative. The lecturer adds some information to the narrative as an extended translation. Overall, this lecturer seems to be using the slide-text as some form of flexible scaffold for the lecture, in which the text is not assessed but, rather, becomes a part of the speech. Further, the role of the speech as a translator of the slide-text is more obvious.

Functions of the speech-slide-text relationship
Through considering the practices of both the most and the least integrative lecturers, it is proposed that there are two functions of the speech-slide-text relationship in slide-lectures: (1) The referent function, characterised by the lecturer providing an assessment of the slide-text.
(2) The scaffolding function, characterised by the lecturer's speech blending and translating the slide-text within the spoken narrative.
The described analysis was carried out on the remainder of the slides for each of the two lecturers to consider the extent to which the lecturers display characteristics of one kind of relationship over another. The speech acts were separated into the two relationships that they appear to indicate, and instances in which they occurred were recorded throughout the whole lecture transcript. Here, 'conducting attention', 'questioning', 'agreeing/disagreeing' and 'signalling importance' were considered as acts used when a lecturer refers to their slide-elements, as they serve to separate speech from slide as two distinct aspects of the presentation. Verbalising also fits here, as it was considered that, in verbalising the text, the lecturer draws attention to the text on the slide, which again highlights the distinctness of speech and slide. 'Merging', 'translating' and 'combining' are considered to be aligned with the scaffolding relationship, as these serve to combine the speech and slide information into a single message: the speech and slide-text are not identified as distinct messages. This analysis is detailed in Table 5. A χ 2 analysis was carried out to compare the total number of speech acts within each relationship type that the lecturers produced. The difference in relationship indicators between lecturers was significant: χ 2 (df: 1, n = 291) = 63.08, p < 0.001. It was concluded that the lecturers differed significantly in terms of the relationship indicators they employed in their lectures.

Reliability of the indicators
The indicators were checked for reliability by employing a colleague who was given two slide examples from each of the two lecturers (approximately 10% of the total number of slides), along with descriptions of the speech acts. The coder was asked to identify whether the speech acts were present in the examples and how frequently they occurred. This information was then compared with the same coding performed by the authors. It should be noted that the second coder was given the entire slide-speech transcript for the slides, yet the examples above are clipped.
Inter-rater reliability, using the Kappa statistic to determine consistency amongst the two coders, was found to be in substantial agreement: Kappa = 0.846 (p < 0.001). There was a high level of agreement that the same speech acts were being carried out for the four slides above. Observed pattern (speech) Figure 3. Visualisations of the extent of matching between speech and slide-text of the most and least consistently-matching lecturers.

Discussion
This study quantified and described lecturers' differing relationships with their slides. Simply looking at the 'observed' patterns side-by-side in Figure 3 gives an idea of the difference in approach to integration between two lecturers. Moreover, this inconsistency of approach varies across a considerable rangefrom the most consistent and systematic lecturer, Jackson, who received an integration score of 0.89, to Leaman, who received an integration score of 0.69indicating significantly differing levels of deviation from the slide pattern with their speech. However, the way in which lecturers integrated text was also characteristically different.
It was identified that there are two relationship styles that were employed by the highest-and lowest-scoring lecturers in terms of integration: referent and scaffolding. Here slide-text is treated as either an object to comment on or is blended into the speech. Jackson employed indicators of a referent relationship in 70.45% instances of integration throughout the lecture, whereas Leaman employed them in 21.67% of instances of integration. On the other hand, Jackson employed indicators of a scaffolding relationship in 29.55% of instances of integration, whereas Leaman employed them in 78.33% of instances of integration. It may be suggested, then, that a lecturer who employs a referent relationship might be more concerned with systematically addressing each at that time because it was saying: hold on a minute, there could be a real problem here with mothers going to work. And the problem is to do with this business of attachment. So he caused a lot of trouble and he wasn't very popular. But it was a very controversial book at that time.

His major influences on this,
on his writing, which really, he was really the first person in psychology, apart from Freud of course much earlier, to really begin to pull out the significance of this relationship and he did for two reasons, one a lot of his ideas came from psychoanalysis, but also from another branch of the natural sciences called ethology.
And OK, here's some fairly obvious basic ideas about it, a young child is biologically biased to develop attachment to its caregivers given its genetic endowment.

Now the second point I wanted to make was this whole business of the function is simply the protection of the young.
Right, it looks like a fairly, is it therefore he was asking, a kind of automatic phenomena And the function of it psychologically is to provide security. element on the slide in the expected order, whereas the lecturer employing a scaffolding relationship might be less concerned with such following of the slide-text. It must be noted that, although the two lecturers showed significant preferences for different approaches, they did not consistently display characteristics of only one relationship. Thus the function of slide-text might vary both between and within slide-lectures, and might depend heavily on how the lecturer intends to use each text-element. As a consequence, the lecturers' relationship with their slides is not immediately evident, thus further empirical examination of these relationships is needed to ascertain their relative impacts on the learning environment.
The proposed referent and scaffolding relationships appear to be reflected in literature commenting on PowerPoint practice, yet it seems that most often it is the scaffolding relationship that is described. For instance, Adams (2006) identifies the role of slides as being where the lecture resides: the information contained is elaborated by the lecturer through their verbal exposition. Further, Maxwell's (2007) critical account of the prevailing role of PowerPoint is that it provides a summary for the lecture, which is repeated by the lecturer's speech. Yet, as our analysis has identified, this predictable level of integration is not always the case within the scaffolding relationship, and often the integration is much more intricate.
In terms of the referent relationship, there is comparatively less commentary on its use. Rather, the literature that discusses this kind of relationship often calls for more lecturers to adopt it in preference to the scaffolding relationship, suggesting that it is a less common strategy. For example, Maxwell (2007) argues that the role of the slideshow is an artefact to be commented and elaborated upon. Others focus on the adoption of multimedia design principles; that is, a succinct headline, along with visual evidence for that headline (Alley and Neeley 2005;Johnson and Christensen 2011;Nagmoti 2017). Pate and Posey (2016) revealed that slides designed following multimedia principles (i.e. images with labels) contributed to the superior performance and satisfaction of students to a greater degree than traditional text-  based slides. Here, the lecturer shows visual evidence for their arguments, rather than text summaries of their lecture. It must be noted that the suggested lecturing model in these and similar studies (e.g. Ari et al. 2014; de Koning, van Hooijdonk, and Lagerwerf 2017;Schüler, Scheiter, and Gerjets 2013) calls for greatly reduced levels of slide-text compared to the slides examined in the current study.
Although the data were collected during the 2009/10 academic year, it is highly likely that the results are applicable to lectures today owing to the significant 'constraining qualities' of PowerPoint/slideware (Kernbach, Bresciani, and Eppler 2015) that potentially force lecturers into the kinds of relationship discussed here. The constraining qualities that Kernbach et al. describe -'bulleting', 'sequencing', 'monotonous ritualizing', 'reading slide text aloud'speak of similar concepts to the relationship indicators employed in this article. Although lecture practices are gradually shifting towards more active teaching methods, for example the flipped classroom (e.g. Bergmann and Sams 2012) or 'active lectures' (e.g. Pickering and Roberts 2018), which may make less use of the slide-lecture format, slide-lecture components are still a predominant feature of instructional messages that may accompany these approaches. Students may be asked to watch recorded slide-lectures online within the flipped classroom approach. As Kernbach et al.'s (2015) constraining qualities are embedded in the software, it is likely that a lecturer's integration approach will be driven to one of the relationships whenever there is text on screen, even if the text-lecture is significantly reduced.
There are many integration options open to lecturers when conducting slide-text lectures and their integration choices might depend on personal preferences, familiarity with the subject and attitudes and beliefs regarding lecture pedagogy. Furthermore, their academic discipline might carry its own idiosyncrasies regarding the kinds of information displayed on their PowerPoint screens (Garrett 2016), which might influence the ways in which slide-text is spoken about. Indeed, the cultural studies lectures discussed by Zhao and van Leeuwen (2014) appear to adopt the referent approach as standard. Cultural studies, as described by Zhao and van Leeuwen, is a discipline in which students are invited to engage with the ideas presented, using images and other modalities as triggers for a personally meaningful experience, compared to the 'lecturerexplaining-concepts' model typical of psychology lectures. This is reflected in the proportion of Zhao and van Leeuwen's seven lectures' worth of slides that were image-rather than text-based (of 268 slides, only 24.25% were exclusively text-based; in this article, 91.57% of slide-elements were textual). It is important to note that slides are conceptualised by Zhao and van Leeuwen as semiotic resources that enable knowledge to be recontextualised in the classroom. In other words, in cultural studies classrooms slides do not just convey information, they transform meanings. This practice is more suggestive of our referent relationship.

Conclusions
Whichever relationship the lecturer has with their slide-text, it should be acknowledged that live slide-lectures represent an unpredictable sub-type of multimedia instruction. This unpredictability might result in a potentially fraught situation when the speech and slide are not integrated as expected, i.e. what the lecturer is saying is not related to the expected textelement on screen. Rowley-Jolivet (2002) describes the audience's task during a slide-presentation as a difficult one when characterised using Mayer et al.'s notions of dual processing and cognitive load (e.g. Mayer 2005;Moreno 1998, 2003;Mayer et al. 1999). Rowley-Jolivet suggests that the speaker must follow, or 'synchronise', their verbal commentary with the slides, whilst the audience must negotiate both the slide content and the synchronisation of it by detecting and processing the 'linear progression and semiotic mix imposed by the speaker' (2002,21). Applied to a slide-lecture, the student expects the lecturer to systematically address the slide-text. Those who expect that the lecturer will do so in order to expand on the slide-text could easily be confused by a lecturer who uses the referent approach. Similarly, those who are seeking an appraisal of the information on screen might be frustrated by a scaffolding approach. Lecturers who are consistent in their integration of slide-text might go some way towards making slide-lectures easier to follow. Yet it remains unclear which kind of relationship (referent versus scaffolding) is better for learning. Either way, it seems that copious amounts of slide-text might be problematic.
We suggest that further research should examine how learning might be influenced by different types of speech-slide relationship by, for example, examining student reflections on the different relationships. It would be useful to examine how student attention can be optimally managed in a slide-lecture, and whether particular integration practices can facilitate this management. An investigation into the extent to which lecturers employ the slide-lecture approach for entire sessions or as part of a range of approaches during a session would provide useful insights into how much lecture time a student employs in attempts to synchronise speech and slide-text. Arguably, there is a need for some consensus, discipline or guidance in relation to the practices of coordinating speech and slide-text.