More than Just Figures: Structural and Visual Complexity in Soil Science Articles

Abstract

The structure of a scientific article is crucial for clearly conveying research findings. Modern scientific publications combine text with various elements—such as tables, graphs, images, diagrams, and maps—that support the narrative and aid data interpretation. Understanding how these components influence a publication’s reception and scientific impact is essential. This study analyzes differences among 15 soil science journals (indexed in the Web of Science) in terms of visual elements, tables, number of authors, and article length. The journals had a 5-year Impact Factor (2023) ranging from 0.9 (Soil and Environment) to 10.4 (Soil Biology and Biochemistry). The Kruskal–Wallis test and Bonferroni-adjusted Dunn’s post hoc tests revealed statistically significant differences across all variables (p < 0.05). The relationships were further assessed using Pearson’s correlation, based on the median number of authors and article length, as well as the percentage of articles that include at least one element of a given type (e.g., table, graph, image, diagram, or map). Key findings show that journals with a higher impact factor tend to publish articles with more authors (r = 0.62, p = 0.014), use diagrams more frequently (r = 0.69, p = 0.004), and include fewer tables (r = –0.85, p < 0.001). These results suggest that journals with a higher 5-year IF tend to include articles with a greater number of authors and a higher frequency of diagram use, while relying less on tables.

1. Introduction

An academic article is a complex structure whose primary task is to present a specific research issue. To achieve this, researchers employ various means of communication, including text, tables, and visual elements. The format and clarity of presentation can significantly influence how effectively a message is conveyed and understood. Well-structured presentation is particularly valuable in disciplines like soil science, in which topics can be highly complex.

This structure is sometimes guided by editorial policies (e.g., Instructions to Authors, Manuscript preparation), which may set limits on article length or the number of graphical elements. For example, Nutrient Cycling in Agroecosystems prefers manuscripts of 7000 words (excluding references), approximately 14 pages, and 4–8 display items (figures or tables). In contrast, Revista Brasileira de Ciência do Solo states that the results “should be presented using tables or, preferably, figures containing graphics, images, or schematic models. The use of more than four tables and four figures should be avoided, as well as restating numeric values that have already been presented in tables and figures”. Soil advises against using colored cells in tables. These editorial guidelines are generally flexible compared to strict formatting requirements (e.g., font size, graphic file format, or reference style). Ultimately, authors have discretion over how to structure content, which raises important questions: What determines the method of presenting information in scientific articles? Do larger research teams use more graphical elements? Does the presentation style reflect the journal’s scope or prestige (e.g., impact factor—IF)?

Previous research shows that visual content —especially graphs—varies significantly across disciplines. Cleveland [1] compared 57 journals from various fields and found that the fractional graph area (i.e., the percentage of page space devoted to graphs) differed strongly between disciplines and among journals within a discipline. Smith et al. ([2,3]) extended this by linking visual usage to the perceived “hardness” of scientific disciplines, with “hard” sciences like chemistry and physics devoted more space to graphs than “soft” fields like sociology or economics (r = 0.97). Within psychology, quantitatively oriented journals—described as “hard”—also featured significantly more graphical content (r = 0.93). Arsenault et al. [4] found a similar pattern for non-graph visuals (e.g., photographs, diagrams, equipment readouts): natural and exact sciences employ more visual material than the social sciences or humanities.

Temporal trends and visual strategies have also been examined. Schild and Voracek [5] found that meta-analyses in medicine, psychology, and management differed in their use of visualizations over 30 years, with what the authors describe as an overall underutilization of available graphic techniques, as reflected in the paper’s title “Less is Less”. Hegarty and Walton [6] studied psychology journals and found that the JIF (Journal Impact Factor) correlated positively with cited references (r = 0.35) and article length (r ≈ 0.3), but negatively with the number of tables (r = –0.14), while showing no significant correlations with the graphs count (r = 0.04). Similarly, 5-year scientific impact (i.e., the number of references that a given scientific article received within 5 years of publication) correlated positively with cited references (r ≈ 0.4), article length (r ≈ 0.3), or table count (r = 0.13), but negatively with graph count (r = –0.14). Meanwhile, Lindsey [7] reported a moderate positive correlation (r ≈ 0.33) between the number of graphs and citation rates in psychology. Other studies also analyzed the influence of features such as IF, page count, or citation frequency [8,9].

A large-scale study by Lee et al. [10] used machine learning to analyze over 8 million figures from the PubMed Central database (Viziometrics project). Visuals were categorized (e.g., diagrams, plots, photographs, tables), confirming strong disciplinary differences: for example, molecular biology and pathology had many photographs, while mathematical biology and medicinal chemistry relied heavily on diagrams and plots. The findings also revealed a positive correlation between visual richness and scholarly impact: highly cited papers and those in leading journals—particularly those containing diagrams—used more visual elements overall. While the number of figures per article remained relatively stable, their complexity (e.g., multi-panel images) increased.

In other disciplines, visual and structural features have also been linked to journal prestige. Kotiaho et al. [11] examined 65 journals in biology, ecology, and evolution, finding a strong positive correlation (r = 0.7) between journal impact factor and the average number of authors per article. In the agricultural sciences, Tartanus et al. [12] analyzed 21 journals and found a moderate positive relationship (r ≈ 0.40) between impact factor and graph usage. According to their findings, journals with higher IF published more articles that included graphs, and those articles contained more graphs per paper. In some cases, 100% of the articles in a given journal included at least one graph; in others, this rate was closer to 50%.

Abramo and D’Angelo [13] studied Italian publications indexed in the WoS core collection, analyzing the relationship between author count, citation rates, and journal impact factor. According to their findings, an increase in co-authors was associated with higher citation rates across most fields, especially in the social sciences and arts and humanities. However, the expected relationship between author count and journal prestige (measured by IF) was weaker than expected—the authors did not find strong evidence that papers with more co-authors were published in significantly higher-IF journals. They suggest that although teamwork papers may receive more citations, this does not necessarily translate into publication in journals with a higher impact factor.

In addition to discipline and journal prestige, other studies have examined how article or author characteristics relate to the use of graphs and tables. Seglen [14] identified three main factors influencing article length: number of pages (usually not exceeding 7–8), number of tables and figures (typically one per page), and the data density within each table or figure (which had no upper limit). Cabanac et al. [15], analyzing over 5000 articles from the physical and social sciences, found that collaborative papers contained more graphs and tables than those written by a single author. The authors suggest that this may reflect the tendency of larger research teams to generate more data, thereby require additional visual tools for presentation.

Moreover, author gender may impact the data presentation format. Hartley and Cabanac [16], explored gender-based differences in data presentation using ~2000 articles from over 200 peer-reviewed journals in the sciences and social sciences. Their findings revealed that male authors produced 26% more figures than females authors, although no gender-based differences were observed in the use of tables. No significant gender difference were found in the use of graphs, figures, or tables within social science articles.

Although previous studies have examined structural and visual components of scientific articles—either within individual disciplines or across broader domains—soil science journals have rarely been analyzed in such comparative contexts. At the same time, bibliometric studies suggest that soil science has undergone dynamic growth in recent decades, both in terms of publication volume and citation impact. Analyses focusing on specific subfields—such as those focused on soil classification [17], nutrients [18], erosion [19], and soil health [20]—demonstrate an upward trend in research output. Moreover, thematic shifts in soil science have been observed from traditional topics like physicochemical soil characterization to integrative and globally relevant challenges such as climate change, pollution, and the use of advanced techniques like artificial intelligence, predictive modeling and spectroscopy [17,18,19].

In light of these developments, this study investigates whether the author count, article length, and use of individual elements (images, diagrams, maps, tables, and graphs) vary across soil science journals. We also examine whether these variables are correlated with journal prestige, represented by the 5-year impact factor.

2. Materials and Methods

This study is based on data from scientific articles published in 15 soil science journals (see Figure 1), all indexed in Clarivate Web of Science (formerly Thomson Reuters’ Web of Knowledge). We studied all papers published in these journals in 2023 (except editors’ letters and similar items; n = 1448 articles). The 5-year IF ranged from 0.9 for Soil and Environment to 10.4 for Soil Biology and Biochemistry. The number of published articles in 2023 varied greatly across journals (Figure 1), from 20 papers in Acta Agriculturae Scandinavica, Section B to 461 papers in Geoderma.

The dataset included the following variables:

• 5-Year Impact Factor (2023).
• Number of authors per article.
• Article length (number of pages).
• Number of graphical and non-graphical elements, categorized as follows:

  ○

  Tables (any tabular format with text or numerical data).

  ○

  Images (e.g., microscopy photographs, radiology images, fluorescence imaging, soil profiles, equipment photographs).

  ○

  Diagrams (e.g., schematics of the experimental design, conceptual diagrams, flow charts, phylogenetic tree, Venn diagram, Structural Equation Modeling—SEM).

  ○

  Maps (e.g., satellite images, geographic maps of study areas or countries, maps with numerical overlays such as forest cover percentage or heat maps).

  ○

  Graphs (e.g., bar plots, boxplot, scatter plots—including 3D version, bubble plots, corrplot, ternary plots).

The identification and classification of visual and structural elements were based largely on the scientific literature (e.g., [21]) and conducted systematically by a single researcher. The coding process followed established typologies and was supported by the researcher’s knowledge.

Statistical Analysis and Data Visualization

Normality of distributions was assessed using the Shapiro–Wilk test. Differences between journals were tested with the Kruskal–Wallis test, followed by Dunn’s post hoc tests with Bonferroni correction for pairwise comparisons. These tests were performed on raw article-level data (n = 1448) to compare journals across all studied variables. Relationships between variables (medians for author count and article length, and percentage of articles that contained at least one graphic element of a given type) were calculated using Pearson’s correlation coefficients (one value per journal, n = 15). The Scatterplot Matrix (SPLOM) was completed using a graphical method called the CI thermometer [22]. All statistical analyses and visualizations were conducted using R.

3. Results

The Shapiro–Wilk test was conducted for each variable within each group (i.e., journal). The results indicated that most distributions deviated from normality (p < 0.05). Violin plots were used to illustrate the distribution (range, median, and interquartile range) for the number of authors and pages (Figure 2), as well as for elements such as the number of images, diagrams, maps, tables, and graphs (Figure 3). For variables such as the number of images or maps, the majority of the articles in many journals had a value of zero, resulting in highly skewed distributions. For these variables, the Shapiro–Wilk test rejected the assumption of normality in all 15 journals. Variables such as the number of authors or pages were less skewed, but in most cases, they also failed to meet the assumption of normality (e.g., the page count distribution was non-normal in 9 out of 15 journals).

Extreme values can be clearly observed:

• One article in Geoderma listed 54 authors and an article in Arid Land Research and Management was 58 pages long (Figure 2);
• Individual articles featured up to 186 graphs (Soil), 160 images (Geoderma), and 116 maps (Revista Brasileira de Ciência do Solo), as shown in Figure 3.

Because none of the variables met the assumptions of normality (or was strongly skewed), nonparametric Kruskal–Wallis tests were used.

3.1. Differences Between Journals

The number of authors per article varied significantly between journals (H = 120.6, p < 0.001). Post hoc Dunn comparisons (Bonferroni-adjusted) showed that Biology and Fertility of Soils (median = 7; Figure 4), or Soil Biology and Biochemistry, Geoderma, Nutrient Cycling in Agroecosystems, or Revista Brasileira de Ciência do Solo (each median = 6) had significantly more authors per article than Eurasian Soil Science, Arid Land Research and Management, or Acta Agriculturae Scandinavica, Section B (median = 4). Similarly, article length varied strongly (H = 483.1, p < 0.001). Arid Land Research and Management (median = 20) and Revista Brasileira de Ciência do Solo (median = 18 pages) published significantly longer papers than, for example, Pedobiologia (median = 9 pages) or Soil Science and Plant Nutrition (median = 10 pages); Figure 4.

Although the median number of images was 0 in most journals (Figure 3), the distributions varied significantly (H = 49.9, p < 0.001). For example, Arid Land Research and Management (mean ≈ 2.9 images) or Geoderma (mean ≈ 2.8) contained significantly more images than journals like Nutrient Cycling in Agroecosystems (mean ≈ 0.05). In nine out of fifteen journals, at least 20% of the articles included at least one image (Figure 5). The highest proportion was observed in Arid Land Research and Management (41%), while the lowest was in Nutrient Cycling in Agroecosystems (only 3%). Use of diagrams also varied by journal (Figure 3). In some journals (Soil, Vadose Zone Journal, Soil Biology and Biochemistry, or Geoderma), diagrams were relatively common, while in others they were rare, resulting in non-normal distributions (Shapiro–Wilk p < 0.05 in most groups). These journals contained significantly (H = 141.5, p < 0.001) more diagrams (averaging ≈ 2.5 diagrams per article), compared to <0.5 diagrams in Soil and Environment, Arid Land Research and Management, or Soil Science and Plant Nutrition (which had <0.5 diagrams on average). It is worth noting, however, that a minimum of 24% of articles contained at least one diagram (Nutrient Cycling in Agroecosystems; see Figure 5), while in Soil Biology and Biochemistry, the rate was as high as 64%. The Kruskal–Wallis test also showed significant differences between journals in terms of the number of maps (H = 126, p < 0.001). The average number of maps per article exceeded five in Vadose Zone Journal or Arid Land Research and Management, while in Biology and Fertility of Soil or Acta Agriculturae Scandinavica, Section B, the average was below 0.5 (Figure 3; the median number of maps was 0 for most journals). Only 8% of articles in Biology and Fertility of Soils contained at least one map (Figure 5), in contrast to 62% in Soil and Water Research.

Tables and graphs were used much more frequently than other methods of presenting information (Figure 5). In many journals, more than 90% of articles included at least one table; the exceptions were Soil Biology and Biochemistry and Biology and Fertility of Soils (69–72%) and Soil, Geoderma, and Pedobiolgia (86–88%). It is worth noting that 100% of the articles in Vadose Zone Journal included at least one graph, while the lowest rate was noted in Eurasian Soil Science (86%). Tables per article also showed significant variation between journals (H = 188.3, p < 0.001): Arid Land Research and Management (median = 4) had more tables than journals such as Soil Biology and Biochemistry (median = 1), with most journals in between (median 2–3 per article). Although most of the journals contained graphs, the rates for the presence of these elements per article differed significantly (H = 203.9, p < 0.001): Biology and Fertility of Soils had the highest number of median graphs (20), and Eurasian Soil Science had the lowest (6). In all journals, more than 24,500 graphs were used. Most were one-dimensional (≈ 73% e.g., bar plots, histograms, line plots, boxplots), fewer were two-dimensional (≈ 25%; mainly scatterplots for correlations and regressions), and only 1.8% were multi-dimensional (e.g., ternary graph, bubble plot, and 3D scatterplot). It is worth noting that in many of the analyzed articles, graphs, regardless of type, were presented as multi-panel. Such visualizations, also known as Trellis plots or facets, are among the most effective ways of presenting structures and patterns across multiple variables or categories [23,24,25,26]. Figure 1, Figure 2, Figure 3, Figure 4 and Figure 5 in this paper also use this technique. While the information presented in Figure 1 and Figure 4 could also be conveyed in tabular form (see Appendix A,

Table A1. Soil science journals included in this study, sorted by 5-year Impact Factor; ^a copyright by Thomson Reuters; ^b number of research articles published in 2023.

Journal	5-IF 2023 a	Papers (n) b
Soil Biology and Biochemistry	10.4	251
Soil	7.1	38
Geoderma	6.7	461
Biology and Fertility of Soils	6.2	71
Soil Use and Management	3.9	105
Nutrient Cycling in Agroecosystems	3.5	74
Vadose Zone Journal	3.3	43
Soil Science And Plant Nutrition	2.7	34
Acta Agriculturae Scandinavica, Section B	2.1	20
Arid Land Research and Management	2.1	29
Pedobiologia	2.1	34
Soil and Water Research	2.1	24
Revista Brasileira de Ciência do Solo	1.8	51
Eurasian Soil Science	1.4	166
Soil and Environment	0.9	21

and

Table A2. Median number of authors and pages for 15 soil science journals, based on articles published in 2023. Journals are ordered by 5-year Impact Factor.

Journal	Median Number:
	Authors	Pages
Soil Biology and Biochemistry	6	11
Soil	5.5	16
Geoderma	6	12
Biology and Fertility of Soils	7	14
Soil Use and Management	5	14
Nutrient Cycling in Agroecosystems	6	15.5
Vadose Zone Journal	5	15
Soil Science And Plant Nutrition	5	10
Acta Agriculturae Scandinavica, Section B	4	12
Arid Land Research and Management	4	20
Pedobiologia	4.5	9
Soil and Water Research	5	11.5
Revista Brasileira de Ciência do Solo	6	18
Eurasian Soil Science	4	12
Soil and Environment	5	12

), as is the case for Figure 5 (Appendix A,

Table A3. Percentage of articles in soil science journals (published in 2023) included at least one of the following elements: images, diagrams, maps, tables, or graphs. Red dashed lines indicate the mean within each panel. Journals are sorted by 5-year Impact Factor.

Journal	Papers (%) with at Least One of the Following:
	Images	Diagrams	Maps	Tables	Graphs
Soil Biology and Biochemistry	15	64	15	69	97
Soil	32	53	42	87	89
Geoderma	27	64	43	86	97
Biology and Fertility of Soils	13	58	8	72	93
Soil Use and Management	21	42	38	95	97
Nutrient Cycling in Agroecosystems	3	24	11	91	96
Vadose Zone Journal	26	63	37	91	100
Soil Science And Plant Nutrition	18	32	18	91	97
Acta Agriculturae Scandinavica, Section B	25	35	20	95	90
Arid Land Research and Management	41	28	41	93	90
Pedobiologia	26	44	24	88	97
Soil and Water Research	8	42	62	96	88
Revista Brasileira de Ciência do Solo	31	51	47	94	94
Eurasian Soil Science	24	31	39	95	86
Soil and Environment	29	33	10	95	95

), using this approach for Figure 2 and Figure 3 would have been less effective due to the complexity and size of the data. Although multi-panel visualizations are widely praised for their clarity and ability to present complex data structures, some studies have also highlighted their potential limitations. For example, Kozak [27] demonstrated that when data are split into separate panels by groups, certain global patterns or messages may become obscured. Combining data into a single plot with appropriate visual encoding (e.g., color) can sometimes reveal structures that remain hidden in faceted displays. The author suggests that the effectiveness of multi-panel figures depends on how the data are structured and that such designs should be evaluated with consideration of the analytical goals.

3.2. Relationships Between Variables

To evaluate relationships between variables across the fifteen journals, Pearson’s correlation coefficients were calculated for the following variables: medians (for authors and article length) and the percentage of articles that contained at least one element of a given type (images, diagrams, maps, tables, or graphs). The Scatterplot Matrix (SPLOM) in Figure 6 shows scatterplots (one value per journal, n = 15) in the top panel and, in the bottom panel, CI Thermometer, a visualization that helps present correlation matrices accompanied by additional information on confidence intervals. In the CI Thermometer, the horizontal edges of the box (top and bottom) represent scales for the correlation value between –1 and +1. A white dashed line represents an estimate of the correlation coefficient (r), and the color area shows the coefficient interval (in this case 95%): blue indicates negative values of the correlation coefficient, and red its positive values. The coefficient is significant when the rectangle contains only a single color (the confidence interval does not include zero). When the rectangle is bicolored, the coefficient is nonsignificant. This visualization method was selected due to its enhanced interpretability compared to standard formats. In their study, Wnuk et al. [22] proposed the CI Thermometer as a novel tool for visualizing correlation matrices, and as part of their research, they conducted an expert evaluation comparing three formats: a traditional correlation table with confidence intervals, a corrplot-style matrix with confidence intervals, and a hybrid format combining a classical correlation table with the CI Thermometer. The hybrid format received the highest expert ratings for clarity, usefulness, and overall effectiveness. Its main strength lies in its ability to intuitively convey the direction, strength, and statistical significance of correlations in a single compact graphic—provided the number of variables remains moderate. To support the visual approach, a correlation table was added in Appendix A (

Table A4. Correlation matrix for selected variables (n = 15): medians for number of authors and article length, and the percentage of articles containing at least one element of a given type (images, diagrams, maps, tables, or graphs). The upper-diagonal section shows 95% confidence intervals for the corresponding correlation coefficients and the p-values are presented in the lower-diagonal section.

Variable	5-IF	Authors Median	Pages Median	Images %	Diagrams %	Maps %	Tables %	Graphs %
5-IF	–	[0.16, 0.86]	[−0.56, 0.46]	[−0.65, 0.34]	[0.28, 0.89]	[−0.63, 0.38]	[−0.95, −0.59]	[−0.31, 0.67]
Authors Median	0.62 p = 0.014	–	[−0.44, 0.58]	[−0.77, 0.10]	[0.08, 0.84]	[−0.68, 0.30]	[−0.88, −0.22]	[−0.18, 0.74]
Pages Median	−0.07 p = 0.817	0.10 p = 0.732	–	[−0.14, 0.76]	[−0.56, 0.46]	[−0.26, 0.70]	[−0.38, 0.62]	[−0.64, 0.36]
Images %	−0.20 p = 0.465	−0.43 p = 0.106	0.40 p = 0.135	–	[−0.48, 0.54]	[−0.26, 0.70]	[−0.28, 0.69]	[−0.60, 0.41]
Diagrams %	0.69 p = 0.004	0.57 p = 0.027	−0.07 p = 0.808	0.04 p = 0.875	–	[−0.39, 0.62]	[−0.86, −0.17]	[−0.17, 0.74]
Maps %	−0.17 p = 0.551	−0.25 p = 0.370	0.29 p = 0.296	0.29 p = 0.292	0.15 p = 0.593	–	[−0.08, 0.78]	[−0.75, 0.17]
Tables %	−0.85 p < 0.001	−0.66 p = 0.007	0.16 p = 0.565	0.27 p = 0.336	−0.63 p = 0.012	0.45 p = 0.092	–	[−0.69, 0.28]
Graphs %	0.24 p = 0.383	0.37 p = 0.181	−0.18 p = 0.514	−0.13 p = 0.645	0.37 p = 0.173	−0.38 p = 0.166	−0.27 p = 0.336	–

).

There is a significant positive correlation (red color only in CI Thermometer) between 5-IF and the author median (r = 0.62, p = 0.014) and the percentage of articles that contain at least one diagram (r = 0.69, p = 0.004), and a negative correlation (blue color only in CI Thermometer) with the proportion of articles that contain tables (r = –0.85, p < 0.001). Thus, journals with a higher impact factor were more likely to publish multi-authored articles, in which diagrams were used more frequently and tables less frequently. Furthermore, there is a negative correlation between the contents of tables and diagrams (r = –0.63, p = 0.012), and between tables and the author median (r = –0.66, p = 0.007), but a positive correlation between the author median and diagrams (r = 0.57, p = 0.027). It is worth noting, however, that diagrams appeared in a maximum of 64% of the articles (Soil Biology and Biochemistry), but more than half of the articles in each journal contained tables (minimum 69%, although most of them were above 90%), and a minimum of 86% of papers contained at least one graph. These findings could reflect functional differences: diagrams present different information from those in tables or graphs—they are often used to visualize experimental designs, conceptual frameworks, or model structures. Tables typically present precise numerical data, while graphs illustrate patterns or relationships—and they can be used interchangeably.

4. Discussion

This study examined 1448 scientific articles published in 15 soil science journals indexed in the Web of Science, highlighting statistically significant differences across all of the analyzed variables, including article structure and graphical composition. Journals such as Biology and Fertility of Soils (IF = 6.2) or Soil Biology and Biochemistry (IF = 10.4) were more likely to publish multi-authored papers, while others, such as Arid Land Research and Management (IF = 2.1), favored longer texts (median = 20). Pearson’s correlation analysis showed that journals with a higher impact factor tend to publish articles with more authors and use diagrams more frequently while relying less on tables. Interestingly, article length did not show a clear relationship with impact factor. Arid Land Research and Management (with a relatively low IF) published the longest articles.

The significant variation observed in article length, author count, and use of various elements across journals suggests that structural and visual formatting decisions are influenced not only by journal prestige, but also by factors such as journal scope, editorial culture, and individual author strategies. Moreover, the frequency of graphic elements (such as images, diagrams, maps, and graphs) and non-graphic elements (tables) varied across journals. Graphs and tables were present in most of the articles analyzed in this study (Figure 5). Over 69% of the articles in the journals studied contained at least one table, and over 86% contained at least one graph, respectively—diagrams, maps, and images were far less common.

Although tables and graphs may convey overlapping content, they serve different cognitive purposes, and their effectiveness depends on the specific task. The presentation or interpretation of data in tables is accurate, but it is most effective when datasets are relatively small [28,29]. Smaller-scale tables and plots may be used interchangeably depending on the authors’ preferences or readers’ needs, or on whether it is necessary to prioritize precise numerical values or visual patterns. In the present study, violin plots were used to summarize multi-journal variability, as presenting the same information in table form would have been unwieldy due to the number of individual values required per journal.

When the dataset is large or involves complex interrelations, tables are often used to support effective interpretation. Humans are generally more adept at decoding well-designed graphics than at interpreting large numerical tables [30,31,32,33,34], and the main task of each graph is the effective presentation of data and underlying phenomena [12,23,31,35]. As Few [36] remarked: “As the saying goes, ’a picture is worth a thousand words’—often more—but only when the story is best told graphically rather than verbally and the picture is well designed. You could stare at a table of numbers all day and never see what would be immediately obvious when looking at a good picture of those same numbers. This also applies to other types of visual elements.”

Although this paper could have used a traditional correlation table instead of Figure 6 in the main text, presenting the same amount of information would have required a format with correlation coefficients below the diagonal and 95% confidence intervals above. While such a table would have provided numerical precision, its interpretation would have been more demanding (see Appendix A, Table A4), particularly in assessing statistical significance (i.e., whether the confidence interval includes zero). In contrast, Figure 6 allows readers to easily distinguish significant relationships: when the rectangle is bicolored (does not include zero), the coefficient is nonsignificant, which helps highlight meaningful findings. While asterisks are often used to denote statistical significance, they also have their limitations [22,37,38,39].

This also applies to other types of visual elements. Tufte [28] noted the following: “Only a picture [in the context of maps] can carry such a volume of data in such a small space. Furthermore, all that data, thanks to the graphics, can be thought about in many different ways at many different levels of analysis”. Maps present a huge amount of information, but their use is not as universal as tables or graphs (the same applies to images or diagrams). Naturally, journals whose aim and scope include interdisciplinary approaches or emphasizing spatial data—such as studies using geographic, satellite, or field-mapping techniques—are more likely to employ maps than those focused primarily on biochemical or conceptual research. For instance, maps were more frequently observed in journals such as Vadose Zone Journal and Soil and Water Research than in Soil Biology and Biochemistry. However, care must be taken to avoid redundant presentation. It is important to ensure that tables do not duplicate the content of graphs—an issue occasionally observed in the analyzed material.

Beyond technical considerations related to data presentation, several broader limitations of this study should also be acknowledged. Although the analysis included all of the research articles published in the selected soil science journals in 2023 (excluding editorials and similar items; n = 1448), some heterogeneity related to article type may still have influenced the results. Furthermore, while journal prestige (measured by 5-year Impact Factor) was the variable used to compare differences across journals, content-related differences between journals—such as their emphasis on biochemical, ecological, or technical aspects of soil science—may also have shaped the structure and content of the articles. As previously mentioned, the identification and classification of visual and structural elements were based on established typologies (e.g., [21]) and were carried out systematically by a single researcher. Nevertheless, a certain degree of subjectivity is inevitable. In ambiguous cases, classification was guided by the authors’ own descriptions or in-text references to the elements. These limitations should be taken into account when interpreting the findings and may inform future efforts to refine methodological approaches in this field.

5. Conclusions

This study confirms that scientific journals in the soil science field significantly differ not only in terms of the number of authors or article length but also in how research findings are communicated through graphical and tabular elements. While nearly all of the journals regularly use tables and graphs —often appearing in over 50% of articles— other visual tools such as images, diagrams, or maps are used less consistently and vary across journals. For example, journals with a biochemical and conceptual focus included more diagrams, whereas those oriented toward agronomic or applied research more often used tabular formats to convey results. Ultimately, the choice of data presentation remains at the discretion of the author(s). However, journals may suggest the use of one method over another (e.g., Revista Brasileira de Ciência do Solo recommends presenting results using figures containing graphs, images, or schematic models). Our analysis also showed that journal prestige (expressed as 5-year IF) is positively correlated with the number of authors and the use of diagrams, and negatively correlated with the use of tables.

Finally, it is important to consider that some visual elements may fall short in terms quality or clarity, even when formally correct. During our review, we observed several issues related to visual quality and readability—for example, low resolution, leftover editing artifacts (e.g., fragments of internal author communication), or graphs that were difficult to read or interpret (e.g., almost the entire area of the scatterplot was covered by the results of the statistical analysis, or the number of categories on the graphs marked in color did not allow them to be effectively distinguished). Some tables spanned several pages without clear segmentation, while others contained only two rows. Some figures grouped dozens of graphs under a single caption without using multi-panel layouts. Of course, we do not intend to question the validity of the authors’ choices, but to draw attention to the existence of certain extremes. In the future, it would be worth examining the quality (or validity) of using such elements in soil science journals.