We investigated students’ ability to learn from multiple-representations display (MRs) in comparison to learning from a text-alone display (TR). The study novelty evolves from the examination of students' learning in the complexity of an authentic context (homework assignments) rather than in a controlled context. It also employed a rich complex MRs display, to resemble a textbook chapter. Students' learning was examined through their ability (a) to process information presented in the MRs display, and (b) to retain that information. The display complexity was expressed in the following factors: (i) numerous representations (18), (ii) representations of diverse types, (13), (iii) familiar and unfamiliar representation-types, and (iv) optimal partitioning of text and images in the display. Representations were presented on distinct cards to increase contiguity effect and enable a non linear processing, each MR card having a "twin" TR card. Three tasks were designed to involve students in learning through information manipulation, including the need to utilize all representations in the display, to employ various numbers of cards for each task, and to experience performance at different processing levels. A 32-item multiple-choice retention and processing test was administered 2 weeks after task completion without any notice. Appropriate analysis and coding systems were developed. Overall, for all the investigated aspects of the tasks and test, the performance of the experimental group students significantly surpassed that of their control group counterparts, supporting other reported findings despite differences in complexity and context. However, a closer look at specific questions revealed particular phenomena as visual tagging and use of anchor words, which students employed to overcome the task as related to the display difficulties. In addition, integration of unfamiliar representation-types was found to frequently offset the advantages of visual representations to learning. Such evidences have to be accounted for in the design of curriculum and instruction.

Cumulative findings suggest that learning with MRs is enhanced due to the MRs roles (Ainsworth, 1999, 2006) and is characterized, among other things, by: (a) enhanced retention and retrieval of encoded information (Levin, Anglin, & Carney, 1987) according to the theory of dual coding (Clark & Paivio, 1991); (b) increased comprehension and transfer (e.g., Mayer & Moreno, 1998; Schnotz & Kulhavy, 1994) due to interactions among the constructed visual and textual mental representations and with prior knowledge (Mayer, 2003; Schnotz & Bannert, 2003); and (c) difficulties in integrating MRs, due to the need to map between them (e.g., Ainsworth, 1999; Yerushalmy, 1991).

Our aim was to extend existing knowledge regarding learning with rich complex MRs display in authentic-like situations.

Hypotheses

Learning from MRs will:

1)      Enhance certain aspects of lower level processing of information and knowledge and impede higher level processing, in comparison to learning from texts alone.

2)      Enhance retention of the acquired knowledge in comparison to learning from texts alone.

Method

Participants: Undergraduate students of education (N = 150) were randomly assigned to homework tasks either using MRs (the experimental MR group – MRG, n1 = 82) or using a textual-representation (TR) display (the control TR group – TRG, n2 = 68).

Procedure: Authentic-like conditions. Each student received a hard copy of the display (either MR or TR) and performed three learning tasks individually as course-administered homework assignments. Students received each new task after completing and submitting the previous one. Time devoted for task performance was reported by students. Without advance notice, two weeks after tasks submission, a test was administered.

Research tools' design:

Information displays. The display topic was "cell phones" chosen to control for prior knowledge. A large number of representations of high diversity was included, to resemble an average textbook chapter and to examine the effect of representations familiarity on performance. Minimal captions and legends for images were introduced to achieve a complete partition of text and images. To decrease the effect of contiguity and split-attention, and to enable non-linear learning we displayed each representation on separate card. Cards' numbering permitted the analysis to identify specific information sources utilized by each learner. The MR display included 13 different visual and textual presentation modes placed on 18 cards. The TR display resembled the MR display exactly with respect to topic, contents, and division of information among the 18 cards, but constituted a “twin” textual presentation describing (rather than interpreting) the corresponding MR card. Such a design sharpened the differences emerging from the encoding of information in two distinct symbolic languages.

Tasks were designed to involve students in learning through information manipulation, including the need to utilize all representations in the display, to employ various numbers of cards for each task, and to experience performance at different processing levels. The three tasks contained 4, 9, and 18 questions, respectively.

Testing. A 32-item multiple-choice test was designed to elicit the products of students' learning processes, including 12 questions assessing students’ retention and 20 questions assessing knowledge processing. 

Data analysis: Data constituted students’ responses to the tasks and the test questions. We performed an initial data reduction by coding the task responses for two predetermined criteria and scoring them accordingly: (a) accuracy and comprehensiveness (A/C), with subcriteria of (i) incorrect/correct and, for correct responses only, (ii) partial/full comprehensiveness; and (b) number of information cards used (NIC). Average inter-coder agreement of 93% was calculated for a random task sample. Test scores were dichotomous (correct/incorrect). Differences between groups were calculated using t-tests and chi-square tests. Pearson correlations were also calculated

Results and Discussion

General performance on the tasks: Overall, the MRG performed significantly better than the TRG both for the A/C (t₁₄₈ = 12.10, p < .001) and NIC criteria (t₁₄₈ = 3.05, p < .01). This results support other reported findings concerning learning from MRs, despite the richness and complexity of the displays, and the authenticity of the performance context. Significantly better MRG performance for both criteria also emerged for each task alone. No significant differences emerged between groups regarding time devoted to task performance.

Lower level processing: In questions that required identification/location of specific information or gathering bits of information concerning a specific topic, the MRG performed significantly better than the TRG. Two modes of identifying/locating information in the cards were revealed and used for gathering too: (a) visually tagging the cards' information – as expressed in the significantly higher number of categories elicited by the MRG than by the TRG, and (b) using "anchor task words" (Eilam, 2001) for identifying information, a mode that was applied mostly by the TRG, possibly because they learned using a 'fluent text'.

Higher level processing: We found an inconsistent pattern of differences between the MRG and TRG concerning their ability to integrate the various representation modes. Integration among highly familiar representations resulted in significantly better performance by the MRG, suggesting that the advantages of learning with MRs nonetheless exceeded the difficulties inherent to integration. However, no differences emerged between the two groups' performance regarding integration among unfamiliar representations, suggesting that integration difficulties offset the MR's advantages for novel representations. Thus, high level processing appeared to be affected by the additional factor of students' prior knowledge and experience regarding the involved representations.

Students’ test performance: The MRG exhibited significantly better performance regarding both retention of the acquired knowledge (t₁₄₈ = 2.01, p < .05), and its processing (t₁₄₈ = 3.86, p < .001), than did the TRG (t₁₄₈ = 13.01, p < .001), again supporting current research findings. Effect size for t-test suggested that the effect of learning with MRs was larger regarding processing of acquired knowledge (d=.61)  than for its retention (d=.33).

Conclusions

Generally, the present study, employing a rich complex display and authentic context, corroborated other findings regarding learning with a limited display in lab-like conditions. A closer look revealed that these findings were sometimes constrained by various circumstances that may affect students' learning (e.g., integration between unfamiliar representations). The study holds implications for education and curriculum.