Phytoacoustics: A Systematic Review and Structured Database of Plant Bioacoustics

1. Introduction

1.1. A Field Without a Map

The physiological effects of acoustic and vibrational stimulation on plants have been a subject of scientific investigation since the late 1920s [1,2]. Since then, the field of phytoacoustics has grown substantially. While some researchers maintain that plants cannot perceive sound as an environmental signal [3], others have reported evidence to the contrary. The latter view is supported by hundreds of experiments and publications, encompassing a wide range of plant taxa exposed to varying acoustic frequencies, exposure durations, intensities, modalities, and propagation media.

As acoustic parameters have been varied, scientists have documented effects spanning multiple biological scales. At the molecular level, sound and vibration have been shown to alter gene expression [4,5,6,7,8,9,10,11,12], modulate hormone biosynthesis and signaling [13,14,15,16], and trigger mechanotransduction [17,18,19]. At the cellular level, treatments can shift enzyme activity [20,21,22,23] and cause cellular damage at sufficient intensities [24,25,26,27,28]. At the plant level, reported effects include accelerated germination [29,30,31,32,33,34], altered growth and biomass accumulation [35,36,37,38], changes in root architecture and phonotropic behavior [39,40,41,42], modulation of photosynthetic and stomatal activity [43,44,45,46], and induction of broader stress-tolerance traits [47,48,49,50,51]. Several rigorous studies have also reported null or marginal effects under conditions where positive responses might be expected [29,36,52,53,54], underscoring the inconsistency of outcomes across the parameter space.

Ironically, the breadth of acoustic and biological parameters available to researchers may actually impede the field’s progress. Despite nearly a century of investigation, no review has systematically mapped the combinations of species, frequencies, intensities, exposure durations, and delivery modalities that have been tested and those that have not. In the absence of such a map, the literature has accumulated unevenly causing certain parameter combinations to be revisited repeatedly across decades and laboratories, while adjacent spectro-temporal regions remain entirely unexplored. Replicated experiments provide phenomenological validation and are valuable in their own right, but they also stall exploration and broader coverage, trading coverage for confirmation. As a result, new research struggles to build meaningfully on what has come before and is less likely to expand into unexplored phytoacoustic territory.

1.2. Why a Structured Database?

Since Wood’s 1928 publication, phytoacoustic review papers have appeared at an average rate of one publication every 18 months [3,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116], with publication rates accelerating in recent decades. While these reviews summarize recent work, none provide a systematic cross-study comparison of the experimental parameters used across the field. Without a machine-readable database of published parameters, meta-analysis remains prohibitively laborious, and researchers have no reliable means of determining which parameter combinations have already been tested.

This work establishes such a database. It is intended to complement, rather than replace, existing reviews. Whereas prior reviews synthesize findings, this database quantifies the structure of the experimental record itself (e.g., what has been tested, how thoroughly, where gaps remain) and is designed to be updated continuously as new studies are published. The initial literature search culminated in March 2026 after identifying 477 candidate publications. Of these, 404 met inclusion criteria and were entered into the database, parsed into 2,991 unique experimental conditions. The remaining 73 publications were deemed out-of-scope (e.g., architectural acoustics, non-scientific commentary, plant-animal sensory studies without direct acoustic treatment) and excluded.

1.3. Scope

The dataset covers acoustic and vibratory effects on plants, drawn from any academically rigorous study in which sound or vibration interacts with a living plant in a meaningful way. This includes externally applied treatments such as airborne sound from speakers [8,9,29,39,117,118,119,120,121], liquid sonication [27,35,122,123,124,125,126,127,128], organism-produced vibrations such as insect buzz pollination and herbivore chewing [69,129,130,131,132,133,134,135,136], and passive characterization of plant-emitted signals such as xylem cavitation acoustics [59,137,138,139,140,141,142,143,144,145]. The first three categories inform the parameter space of sound impinging on plants, while the fourth captures detected acoustic emissions from plants themselves, providing context for understanding stress responses and other physiological states.

2. Methods

2.1. Literature Identification and Inclusion Criteria

Relevant research publications were identified through systematic searches of academic databases including Google Scholar, PubMed, Web of Science, and Semantic Scholar. Search queries combined plant-related terms (plant, botany, agriculture, biology, phytoacoustics) with acoustic-related terms (acoustic, sound, sonic, ultrasonic, infrasonic, vibration, music). Additional publications were identified through backward citation chaining from retrieved articles.

Once identified, a paper was included in the database if its full text could be retrieved and it described at least one empirical interaction between sound or vibration and a living plant. Qualifying interactions fall into two groups (sound or vibration applied to a plant and sound or vibration emitted by a plant) comprising four categories in total. Sound or vibration applied to a plant includes externally applied acoustic treatments (e.g., airborne sound from speakers, liquid sonication), mechanically induced vibrations (e.g., direct contact, substrate-borne vibration), and organism-produced vibrations on plant tissue (e.g., insect buzz pollination, herbivore chewing). Alternatively, sound or vibration emitted by a plant includes passive detection of plant-emitted acoustic signals (e.g., xylem cavitation, ultrasonic stress emissions). Papers were excluded from the database if they lacked empirical data or if the acoustic component had no interaction with a living organism. For example, studies using plant material purely as an inert acoustic reflector or absorber [146,147,148,149] with no biological response measured would not be included.

Each paper matching the inclusion criteria is decomposed into one or more rows. Each row represents a single unique set of experimental conditions, preserving within-paper parameter variation that would otherwise be lost in a paper-level summary. For every row, publication metadata (year, author, title, journal) is recorded along with acoustic parameters (frequency, intensity, duration), experimental context (delivery medium, organism, tissue type, temperature), and a methodological quality score that captures the completeness and rigor of the original reporting.

The experimental parameters were compiled into a machine-readable database and analyzed. Because of the database’s complexity, detailed analysis cannot be presented in a single summary graphic; instead, the analysis proceeds through a series of targeted visualizations, each examining a different dimension of the field’s development and illustrating relationships between various experimental parameters.

2.2. Database Schema

Within the database, each unique parameter set is treated as a row-splitting criterion, characterizing each paper by its unique organism × treatment combinations. For instance, a report detailing the impact of four unique acoustic frequencies on one taxon of wheat [150] will produce four rows. In some cases, an additional row may be added to account for a control specimen. Units for treatment frequency, intensity, power, exposure duration, and temperature are recorded in hertz (Hz), decibels (dB), watts (W), minutes (min), and degrees Celsius, respectively.

Each row is further characterized by 24 columns: Year, Author, Title, Journal, Publication Type, Organism Common Name, Organism Latin Name, Organism Category, Tissue Type, Treatment Method, Low Frequency, High Frequency, Specific Frequencies, Frequency Description, Exposure Duration, Power, Intensity, Temperature, Sample Size, Parameter Quality, Study Type, Notes, Parameters, and Delivery Medium. Although some publications provide no data (e.g., review papers), they are still assigned one database row describing the review itself, but its metrics are not included in the analysis provided below. Each reference obtained from literature reviews is found and analyzed separately, ensuring completeness, eliminating duplication, and increasing overall database fidelity and transparency.

2.3. Controlled Vocabulary Definitions

Of the 24 columns used to characterize each entry, most are continuous or free-text fields. Metadata like publication year, title, journal, and the last name of the first author identify each unique source. Organism fields (common name, Latin name, and category) specify the specimen and its biological classification (e.g., plant, bacterium, fungus). Although only plant data were considered in this analysis, several survey-style publications applied the same acoustic apparatus to multiple kingdoms within a single study [2,151,152,153], with plant tissues tested alongside protozoa, blood cells, and microorganisms in the same experimental design. “Tissue type” characterizes which part of the organism was targeted (e.g., seed, whole plant, stem); acoustic parameters are captured in three frequency columns based on the low frequency bound, high frequency bound, and specific frequencies cited in the text. Sample size, experiment notes, and a summary of reported parameters round out the free-text fields. The remaining columns (Publication Type, Parameter Quality, Study Type, and Delivery Medium) are categorical with fixed definitions provided below.

Publication Type

This category describes the medium in which the information was published. Its labels include:

• Journal Article – Primary research published in a peer-reviewed scientific journal
• Review Article – A published synthesis paper with little to no unique experimental data
• Conference Proceedings – Information presented at a conference that may or may not mandate peer review
• Thesis/Dissertation – A graduate research document examined and approved by an academic committee
• Preprint – A manuscript posted to a preprint server without formal peer-review at time of retrieval
• Book Chapter – Information published in an edited volume or handbook
• Report – A technical, government, or institutional report published outside the journal system
• Other – Information published outside of the above definitions (e.g., patents)

Parameter Quality

This category describes the completeness of each publication’s parameter reporting. Its labels include:

• High – All parameters needed to replicate the acoustic or vibratory treatment are reported (frequency/range, intensity, duration, delivery, source-to-specimen distance, waveform). Instrument calibration is explicitly described.
• Medium – The paper reports some but not all key parameters, with common gaps including sonic intensity without calibration, substrate-vibration frequency reported without an acceleration amplitude, or vague delivery descriptions The treatment is partially reproducible.
• Low – Minimal experimental parameters are reported; the paper relies on vague descriptions, unverified or estimated values, or data from sources with known credibility concerns (e.g., low-tier conferences).

Study Type

This category describes the outcome observed during individual experiments. Its labels include:

• Stress/Damage – An acoustic or vibratory treatment is applied to a living organism and the primary measured outcome is injurious, destructive, or lethal (e.g., cell death or tissue damage).
• Stress/Stimulation – An acoustic or vibratory treatment is applied to a living organism and the primary measured outcome is positive or neutral (growth promotion, germination acceleration, yield increase, enhanced metabolic activity, etc.).
• Mechanistic – The paper investigates a cellular, molecular, or physiological mechanism (gene expression, enzyme activity, pollen-discharge mechanics).
• Emission Detection – An acoustic or vibratory signal originating from the living organism is detected (e.g., insect wing vibrations, xylem cavitation).
• Review – The paper presents no unique experimental data of its own; instead, it synthesizes, summarizes, or meta-analyzes existing literature.
• N/A – The paper presents no experimental data and is not a review (e.g., theoretical or computational pieces).

Delivery Medium

This category describes the medium through which the acoustic or vibratory treatment is delivered to the specimen. Its labels include:

• Airborne Acoustic – Sound is delivered to the organism through air, typically via loudspeakers positioned at a distance from the organism. The organism is not in physical contact with a vibrating surface or submerged in liquid.
• Substrate Vibration – Vibration is delivered through direct physical contact between a vibrating surface or object and the organism or its substrate (e.g., the vibration pathway is solid-contact rather than air pressure).
• Liquid Sonication – The organism or tissue is submerged or suspended in liquid and vibration is applied through the liquid medium, typically via ultrasonic bath or probe sonicator (commonly seen in seed germination experiments).
• Passive/Detection – The vibration or sound being measured is organism-produced and the apparatus is used to detect, record, or characterize it rather than to deliver a sonic treatment.
• Non-acoustic – The primary treatment is not acoustic or vibratory in nature (e.g., electromagnetic stimulation); the row appears in the database because it belongs to a multi-treatment study that also includes acoustic arms.

3. Results

Table 1 provides a numerical characterization of the analyzed dataset.

Of the 477 candidate publications, 404 met inclusion criteria; subsequent row parsing resulted in 2,991 unique experiment conditions. Three hundred forty-four of the qualifying publications focused on plants, 46 of which were reviews or otherwise reported no experimental parameters of their own and could not be rated.

3.1. Frequency and Duration

Figure 1 plots unique combinations of sonic frequency and exposure duration for five plant categories. Each dot is color-coded by taxonomic category and dot size scales linearly with the number of independent literature reports at that frequency-duration coordinate. Both axes are logarithmic, and the background delineates three acoustic regimes: infrasound (< 20 Hz), audible sound (20 Hz–20 kHz), and ultrasound (> 20 kHz).

In total, 825 reports were plotted across 277 unique frequency-duration coordinates. Of these, 577 reports are vegetable/herb taxa and only two are algal or aquatic, a marked imbalance that bounds any claim of cross-clade generality. The relatively empty 10–20 kHz band represents the seam between loudspeaker and piezoelectric hardware. In this database, no single study was found to bridge the audible-to-ultrasonic transition and mechanistic continuity across the gap remains assumed rather than tested.

The large number of data points represents all database rows containing a plant organism, acoustic frequencies greater than zero Hz, and non-zero exposure duration. However, there remains a large number of database rows that cannot be plotted due to incomplete reporting. For instance, 533 rows contain non-plant organisms, 693 rows characterize plant organisms but lack a parsable frequency (e.g., generically described musical playback), and 346 contain frequency data but have no recorded exposure duration. The smallest dots in the figure represent single experiments that were run with a particular frequency-duration combination and the largest dots indicate highly repeated parameter combinations.

Ultrasonic studies cluster at short exposure duration ($\le 10$ min) and at frequencies between 20–50 kHz, consisting almost entirely of liquid sonication experiments alongside a small number of plant emission studies and electromagnetic control conditions [121]. Similarly, the infrasonic regime is strikingly sparse despite wind-loading [89], seismic activity [154], thunder [155,156], and wind turbine vibrations [157] constituting the dominant low-frequency acoustic environment of living plants. Only five data points (1, 2, 3, 4, and 5 Hz, each at 60-minute duration) were identified in this regime.

3.2. Frequency Coverage

Figure 2 shows the distribution of stimulus frequencies across the dataset, with stacked colors in each bin to indicate study type. The low and high frequencies are found to be 1 Hz [158] and 20 MHz [159], respectively. The infrasonic, audible, and ultrasonic regimes are provided for reference.

Stress/stimulation studies dominate the dataset, followed closely by mechanistic studies, together creating the overwhelming majority of reported treatments. Emission detection is sparser, with both audible-range components (organism-produced sounds like chewing and buzzing) and ultrasonic components (xylem cavitation and plant-emitted ultrasonic signals). Stress/damage studies and reviews follow, while uncategorized entries are negligible. The vertical axis indicates counts of distinct publications, not database rows; this means a paper testing several frequencies contributes to multiple bins, and a paper spanning more than one study type (e.g., a dose-response study reporting both stimulation and damage) is counted under each type, so the category totals are not mutually exclusive and are not expected to sum to the length of the dataset. Three review publications report a frequency range drawn from the work they synthesize; these are plotted for completeness but contribute no experimental parameters of their own. The heaviest concentration of stimulation and mechanistic studies is centered near 500 Hz and 1 kHz. Damage-inducing stress, by contrast, is reported across a wide frequency range from 250 Hz to several MHz.

3.3. Taxonomic Coverage

Figure 3 shows the 20 most frequently studied plant taxa in the corpus ranked by paper count. Three taxa dominate the corpus: tomato (Solanum lycopersicum) [9,14,120,145,160,161,162,163,164,165,166,167,168,169,170,171], thale cress (Arabidopsis thaliana) [6,8,9,17,41,49,50,118,132,134,136,172,173,174,175,176,177,178], and chrysanthemum (Chrysanthemum morifolium) [4,10,11,13,18,19,20,21,22,179,180,181,182,183]. The next tier is dominated by globally important staple crops, including rice (Oryza sativa) [5,44,167,184,185,186,187,188,189,190,191,192,193,194], cucumber (Cucumis sativus) [53,162,163,185,195,196,197,198,199], soybean (Glycine max) [136,189,200,201,202,203,204,205,206], potato (Solanum tuberosum) [128,207,208,209,210,211], and wheat (Triticum aestivum) [29,53,145,150,189,196,212,213,214,215].

3.4. Treatment Delivery

Figure 4 illustrates the distribution of treatment modes within the dataset. The five categories together account for 284 of the 344 plant publications; the remaining 60 are reviews, commentaries, and studies for which no single delivery medium could be assigned. Airborne acoustic stimulation is the most common approach, followed by liquid sonication – a split that reflects two methodologically distinct research traditions based in agricultural and growth-promotion studies on intact plants, and ultrasonically mediated treatments of cells, tissues, and germinating seeds in suspension. Grouped by coupling physics rather than by label, mechanically driven delivery (airborne plus substrate) outweighs cavitation-driven liquid sonication by roughly two to one, yet conclusions about “sound” routinely pool the two. Here, “Non-acoustic” is a control/multi-treatment bucket rather than a treatment mode, so it is not commensurable with the other four delivery media. The same category also encompasses control experiments where no sonic treatment is applied to any experimental specimen. Papers applying multiple treatment methods to plant specimens also fall under this heading since the overall effect on the plant is likely coupled.

Passive detection represents a fundamentally different research orientation by characterizing organism-produced or stress-induced acoustic emissions (e.g. xylem cavitation in drought-stressed plants, or volatile-stress emissions of the kind used to remotely monitor crop water status [164]) rather than applying acoustic treatment to elicit a biological response. The presence of Passive/Detection alongside the larger stimulation-focused literature reflects an emerging research interest in observing and listening to plants as well as stimulating them.

The last group, substrate vibration, includes mechanical vibration from touch, wind, a shaker, direct contact, or piezoelectric stimulation. The generic nature of this category means it is almost certainly under-resolved in the figure. Pooling such a wide variety of stimulation sources under a single label merges stimuli with very different coupling and acceleration profiles, and a finer contact-delivery taxonomy would be required to separate them.

3.5. Intensity Distribution

Figure 5 shows the distribution of reported decibel sound pressure level (dB SPL) across plant treatment studies. The distribution spans 40 dB [162] to 120 dB [216] across 94 publications. The resulting distribution is sparse and broad, with a single conspicuous mode at 100 dB accounting for roughly one third of these publications, nearly twice as many as the next-largest bin at 80 dB. The remainder of the distribution is sparse and irregular, reflecting both the modest size of the experiment corpus and chronic under-reporting of acoustic parameters across the field. A relatively small number of airborne and music-playback studies report a sound-pressure level comparable enough to be placed on this axis, and intensity of any kind is reported in roughly one quarter of the 344 plant publications (Figure 9).

3.6. Duration Distribution

Figure 6 plots the exposure duration of treatments across the dataset. In total, durations span more than five orders of magnitude ranging from a 20-second airborne ultrasound exposure on tobacco cell cultures [28] to a 47-day chronic vibratory noise regimen on Arabidopsis [217]. The 141 papers shown here are those reporting a single, parsable numeric duration; papers providing exposure durations as ranges, “continuous,” or multi-session schedules that could not be reconstructed are counted among the duration-reporting corpus of Figure 9 but cannot be placed on the axis of this plot. The duration distribution is not symmetric and peaks between 50-200 minutes, reflecting the dominance of single-session or short-multi-session airborne speaker protocols. Roughly one fifth of the papers sit above 1,000 minutes, corresponding to multi-day and multi-week chronic-exposure regimens drawn primarily from substrate-vibration and music-playback studies.

3.7. Publication Timeline

Figure 7 plots the annual growth of phytoacoustics literature over the past century. Publication count remained low throughout the mid-20th century with fewer than 20 papers per decade through the 1980s. Early studies were isolated and dominated by two methodological threads (seed germination experiments and ultrasonic exposure studies) supplemented by a smaller body of work on music and continuous-tone playback. Activity grew modestly through the 1990s before expanding substantially in the 2000s and 2010s, with the 2010–2019 decade alone contributing more publications than the entire pre-2000 corpus combined. The apparent decline after 2023 reflects the lag between publication and indexing, together with the practical limits of the literature search underlying this database. Journal access limitations must also be kept in mind: not only were there fewer journals to publish in during earlier years, but accessing indexed articles from 90 years ago remains difficult in the modern day.

3.8. Report Quality

Matching the shape of Figure 7, Figure 8 presents the parameter-reporting quality of each paper as a stacked bar chart, with each segment colored by tier corresponding to the definitions in section 2.3.The two plots are constructed from nested subsets of the corpus: Figure 7 counts all 344 plant papers, while Figure 8 restricts these to the 298 that report experimental parameters of their own and could therefore be rated. The difference consists almost entirely of review articles and a small number of commentaries.

Two trends are visible across the time axis. First, high-quality parameter reporting has not scaled with the field’s growth; high-quality papers remain a minority of the rated corpus in every era examined, never approaching a majority, and absolute annual counts of high-quality papers have remained in the single digits since 2017. Second, the share of low-quality reporting rose across the 2000s and 2010s and has since held near that elevated level, so that growth in publication volume has been absorbed disproportionately by the lowest reporting tier.

3.9. Parameter Reporting Completeness

Figure 9 characterizes the reporting completeness of the corpus by parameter. Organism-level information is reliably reported since species, tissue type, and treatment method appear in the majority of publications. However, reporting drops noticeably for the acoustic and environmental parameters that define the stimulation the plants receive and drops sharply for the parameters most directly tied to delivered dose. Source intensity is described in roughly one quarter of the publications, and source power in roughly one paper in eight. The parameters reported most often are the least dose-determining, while those that define the stimulus dose are reported least. Undefined stimulus parameters also cause issues with comparative analyses because the experiments cannot easily be represented in each figure.

Figure 9. Reported experiment parameters by paper count.

Figure 9. Reported experiment parameters by paper count.

3.10. Delivery Medium

Similar in shape to Figure 2, Figure 10 shows the distribution of stimulus frequencies across the dataset, with stacked colors in each bin to indicate delivery medium. Unsurprisingly, airborne acoustic delivery dominates the audible regime while liquid sonication dominates the ultrasonic regime, though neither is exclusively confined to that region. Substrate vibration is found in each of the three sonic regimes. Passive/Detection and Non-acoustic are omitted here, since neither delivers a stimulus at a controlled frequency; the bin heights are therefore not directly comparable to those of Figure 2.

4. Discussion

The four principal findings of this review are as follows: (1) the field’s apparent parameter consensus is largely an artifact of inheritance, amplified by the disproportionate weight of a small number of prolific single-laboratory programs; (2) physically incomparable delivery media are routinely pooled; (3) dose-determining parameters are chronically under-reported; and (4) large regions of the stimulus space (the infrasonic regime, the sub-70 dB range, and the audible-to-ultrasonic seam) remain untested. None of the four is independent, since each arises from the same dynamic. A field that grows by replicating convenient parameter sets, rather than by systematically covering the stimulus space, concentrates its evidence where earlier work was performed, favors the groups that report most consistently, and treats any study labeled “sound” as comparable with any other. The first three also reinforce one another, since inherited parameters supply the baseline for replication, the prolific groups supply most of those replications, and the shared vocabulary of “sound” obscures the delivery-medium boundaries that would otherwise expose the mismatch; the fourth is their cumulative footprint on the parameter map. Each can be addressed directly through coverage-driven experimental design, calibrated dosimetry at the tissue, and meticulous reporting.

4.1. Frequency Selectivity

Publications in the phytoacoustics literature typically expose a given taxon to one or more frequencies, measure a set of plant responses, and conclude that a particular frequency, chord, or musical genre is more favorable than others for a desired outcome. The manner in which these frequencies are chosen introduces three structural problems.

First, airborne-acoustic studies overwhelmingly select round integer frequencies that are convenient for record-keeping rather than values motivated by the plant’s acoustic environment. Counted by publication, the four most common airborne stimulus frequencies are 1000 Hz, 500 Hz, 2000 Hz, and 5000 Hz. The ultrasonic regime concentrates near 40 kHz, contributed almost entirely by liquid sonication. Each of these frequency values falls on or near the grid lines of Figure 1 and bears no evident relationship to the frequencies a plant actually experiences in its surroundings, such as buzz-pollination vibrations near 374 Hz [218] or the broadband chewing vibrations of insect herbivores, whose energy extends to several kilohertz [134]. Direct mechanical evidence supports this point: the leaf trichomes of Arabidopsis thaliana preferentially resonate in the frequency band of their primary insect herbivore [177], and modeling of tomato trichomes identifies them as plausible structures for acoustic transduction at comparable frequencies [168]. The micro-structures most plausibly implicated in mechanoreception thus appear tuned to ecologically specific frequencies that the tonal and polyphonic literature has largely passed over.

Second, the stimuli are almost always steady single tones. Harmonics, frequency sweeps, and other polyphonic signals that dominate natural soundscapes are rarely tested, so researchers often characterize plant responses to a stimulus class that is itself ecologically uncommon. Visual data presentation further compounds the problem, since a broadband or polyphonic stimulus cannot be assigned a single frequency coordinate; such signals are therefore absent from Figure 1 and under-counted in Figure 2. Caterpillar chewing vibrations [134], bumblebee sonication [219], Vedic chant [220], Buddhist meditative and Pirith chanting [221,222], Quran recitation [183], Javanese gamelan [45], and “classical music” [223] all fall into this unplottable category. Because of this limitation, the frequency-analysis figures should be read as describing the tonal sub-literature rather than the field as a whole. Unlike the tonal, substrate, and cavitation work whose transduction pathways are hypothesized, a subset of the structured acoustic literature (Vedic and Buddhist chant [220,221,222], Quran recitation [183], mantra recitation [224], and Agnihotra fire-ritual studies [188]) proposes no physical mechanism by which the stimulus would couple to plant tissue. For completeness, each such publication is nonetheless retained in the database, but carries a low parameter-quality rating.

Third, the handful of discrete frequencies tested in any single study is far too sparse to reconstruct a frequency-response function. True response curves, such as twenty logarithmically spaced frequencies across a decade, are essentially absent from the literature. As a result, the recurring claim that a particular frequency is “optimal” for a given taxon rests on a comparison among a few convenient tones rather than on a resolved tuning curve, and cannot distinguish a genuine biological resonance from an arbitrary sampling artifact.

4.2. A Classical Problem

The imprecise characterization of acoustic signals presents a parallel limitation in the research comparing musical genres, which routinely concludes that plants benefit from one music genre more than another. The colloquial label “classical music”, for instance, is applied without distinguishing between the Baroque (c. 1600-1750), Classical (c. 1750-1820), and Romantic (c. 1820-1900) periods, which are acoustically non-equivalent due to the considerable variation in reference tuning standards across these eras. No universal pitch standard existed before the French diapason normal fixed A at 435 Hz in 1859, and the modern A = 440 Hz was not agreed internationally until 1939 [225]. Baroque performance pitch instead varied by city, instrument, and repertoire across roughly a semitone; the A = 415 Hz used in historically-informed performance today is a later convention adopted for convenience, not a period standard. Compositional pitch structure differs systematically across the "classical" periods as well; an analysis of 1,876 MIDI-encoded compositions from Bach through Chopin reported mean pitch values ranging from 314 Hz (Chopin) to 435 Hz (Mozart), with pitch-fluctuation magnitude increasing from the Baroque to the Romantic era [226].

These differences are not trivial, since they represent variation in dominant frequency content, dynamic range, and melodic-interval structure, each of which would independently influence a frequency-dependent biological response. Without the composer, period, piece, tempo, instrumentation, and playback conditions, a study citing "classical music" as a treatment can be neither meaningfully replicated nor compared with another. Because the treatment variable cannot be defined, this becomes a measurement-validity problem rather than a labeling inconvenience.

The same terminological hazard extends to the word “sonication”, which recurs across the corpus with at least three distinct meanings. In the pollination literature, the term refers to the buzz a bee produces to release pollen from poricidal anthers [69,135]; in transducer studies, “sonication” refers to the mechanical vibration a researcher applies to a flower to simulate that buzz, and in the seed- and cell-treatment literature, the same term indicates generic ultrasonic exposure in a liquid bath, a physically unrelated cavitation process. Considering the context-heavy variation in definitions, a reader who encounters “sonication” in a title may find it challenging to understand which of these was performed without consulting the methods. To make matters worse, a keyword search conflates all three.

4.3. Cavitation, Propagation Medium, and the Damage Threshold

Comparing delivery methods across the literature is difficult due to incompatibility between their reported intensities. Under ideal free-field conditions, airborne sound pressure falls by 6 dB per doubling of distance from a point source [227] – a stimulus reported as “100 dB at the speaker” may reach the tissue at 80-90 dB depending on the experiment geometry. The effective dose is further perturbed by enclosure reverberation, frequency-dependent atmospheric absorption, frequency-dependent speaker output, and the standing-wave structure typical of small chambers. Atmospheric absorption spans roughly three orders of magnitude between 500 Hz and 120 kHz [227], meaning airborne ultrasonic stimuli are attenuated over centimeter path lengths, along which audible-range tones are essentially unaffected. A reference-pressure and impedance mismatch further compounds the problem, since liquid-sonication studies use a different reference pressure (1 $\mu$Pa versus 20 $\mu$Pa in air, a 26 dB offset) and operate in a regime where the air-water impedance mismatch imposes roughly 30 dB of transmission loss at the interface [228]. Together, these factors make reported dB values from liquid- and airborne-acoustic studies difficult to compare directly, even before the distinguishing physical mechanism is considered.

One such mechanism is acoustic cavitation. In liquid media at sufficient intensity, the formation and collapse of micro-bubbles generate intense local shear, transient temperature and pressure spikes, and reactive oxygen species at the cavitation site, each of which may damage plant cells [58]. The dose-response framework for this regime, including the role of tissue gas bodies as cavitation nucleation sites, was developed in the ultrasound-biophysics literature of the 1970s and 1980s and applies directly to the plant sonication record [27]. These processes, however, have no direct analog in airborne- or substrate-coupled delivery at biologically relevant intensities where the dominant interaction is mechanical wave propagation through tissue rather than cavitation in a surrounding fluid. Damage is nonetheless not exclusive to the liquid regime, since short-duration airborne ultrasound has been shown to induce cell death and cell-wall remodeling in tobacco cell cultures [28].

The database compiled for this study flags 410 treatment conditions across all organisms (347 of them on plants) as “Cellular Damage Risk”, primarily based on liquid-media intensities reported at or above 1 ${\mathrm{W}/\mathrm{cm}}^2$ or airborne-ultrasonic exposures of the kind characterized by [28]. Several publications in this category report a textbook dose-response relationship, in which growth stimulation occurs at sub-cavitation doses and inhibition or cell death occurs at supra-cavitation doses. Treating liquid-sonication and airborne studies as interchangeable lines of evidence for a single phenomenon (Figure 10), however, is a category error that may have shaped the field’s published claims and its replication failures. The separation is visible in the figures themselves: liquid sonication accounts for roughly a third of the acoustically delivered treatments (Figure 4) yet occupies almost exclusively the ultrasonic band above 20 kHz, which airborne and substrate work do not meaningfully reach. What a frequency-only projection renders as a frequency/study-type structure is largely a frequency/delivery-medium structure since airborne and liquid delivery barely share a frequency range.

4.4. Commercial Confounds and the Intensity Mode

Delivery media are not the only axis along which incomparable studies are pooled; commercial stimulation protocols present the same problem. Phytoacoustic Frequency Technology (PAFT) [229] and related commercial systems such as Sonic Bloom [204] and Agri-Wave [161] combine acoustic stimulation with the simultaneous application of nutrients or growth stimulants. In systems that vary two parameters at once, a growth response cannot be attributed to acoustic stimulation alone. The database contains 132 treatment arms in this category, tagged “Sonic Bloom/PAFT/Agri-Wave”. In nearly all of them the frequency is unreported and the intensity, where provided, is a nominal source level rather than a calibrated value. Nominal reporting contaminates the intensity axis of Figure 5, though not where one would expect: only a single paper in the 100 dB mode is a PAFT/Sonic Bloom study. The commercial systems that report an intensity level instead cluster near 80 dB, their median reported level, and account for several of the papers in the secondary 80 dB bin.

Further inspection reveals that the popular 100 dB and 1000 Hz parameter set does not come from a commercial signature, but rather an academic lineage responsible for half of the mode’s papers. Within this group, most papers report on chrysanthemum phytoacoustics, many at 1000 Hz. The dominant mode in Figure 5 is therefore substantially amplified by parameter inheritance (Section 4.1); 100 dB remains the modal value even when this lineage is excluded, so inheritance sharpens a pre-existing preference rather than creating it.

The intensity distribution is equally informative in what it omits. With a floor near 40 dB and almost all of its mass at or above 80 dB, the corpus has scarcely sampled the sub-70 dB range that characterizes most of a plant’s everyday acoustic environment. Through this lens, the tested range is not merely clustered but systematically loud by ecological standards, and the field’s dose-response picture is anchored well above the levels a plant ordinarily encounters.

4.5. Reproducibility

Underlying each of these issues is a documentation problem. Ideally, a complete report of a phytoacoustic experiment would include the acoustic stimulus (frequency content, waveform, intensity, and exposure schedule), the experiment geometry (source-to-tissue distance, propagation medium, and enclosure parameters), and the biological context (taxon, tissue type, ambient temperature, and sample size). As the analysis shows, omission of any one of these introduces an ambiguity that propagates into every downstream comparison. A study reporting only “periodic playing of classical music”, for instance, does not contain enough information to reproduce the experiment or to place its parameters on the same axis as any tonal study in Figures 1 through 5.

The published record is also a filtered one. As in any field, it represents the studies that passed peer review rather than all those conducted. Positive results tend to reach print more readily than null or marginal ones across the natural sciences, and null or marginal results appear in only a handful of studies in the present corpus. This publication pattern is worth bearing in mind when interpreting apparent consistencies of positive responses at a given frequency, and it is a limitation that no amount of parameter reporting within the published studies can address.

4.6. The Parameter Reporting Crisis

Constructing the database for this study provided a clear view of how often parameters are reported incompletely. Organism-level fields (taxon, tissue type, treatment method) are provided in the great majority of papers, but the parameters that determine what the plant actually experiences are not. Parameters like frequency, duration, sample size, and temperature are each reported in roughly half the corpus, source intensity in roughly one quarter, and source power in roughly one paper in eight (Figure 9). Part of the frequency gap is benign since polyphonic stimuli such as music and recorded soundscapes have no single meaningful frequency to report. The under-reporting of intensity and power is more troublesome, since both are straightforward to characterize and are the parameters most directly tied to whether a treatment falls in a stimulatory, neutral, or damaging regime.

A further difficulty is that documentation has not improved as the field has grown. The post-2010 expansion reflects both the rise of plant mechanosensory biology [63] and renewed agricultural interest in non-chemical growth stimulation, with annual output peaking in 2020 and the 2000s inflection tracking the emergence of a few model-system programs (Arabidopsis and Chrysanthemum callus). However, the field’s growth (Figure 7) has not been matched by a comparable rise in reporting quality (Figure 8). The proportional share of low-tier reporting rose from the pre-2010 era through the 2010s and has since remained near that elevated level. In terms of publication volume, the field appears to be growing by accretion of low- and mid-tier work rather than by a rise in rigor and meticulously detailed reporting, which places a binding constraint on future synthesis. In this sense, more publications do not translate into more usable evidence.

A more subtle distortion arises when a small number of prolific groups report their parameters consistently while the surrounding literature reports incompletely. When this occurs, well-documented work accrues disproportionate apparent weight in any parameter-level synthesis – set against sparse reporting elsewhere, detailed reporting by one academic program can establish experimental parameters that the field begins to accept as default. Incomplete reporting does not merely shrink the analyzable dataset, it biases the analysis toward the parameter choices of groups that happen to report their findings well.

Incomplete reporting also affects statistical power. Sample size is reported in just over half the corpus (Figure 9) and effect sizes are rarely given. In instances where a frequency-dependent difference is claimed, the reader may find it challenging to judge whether the result is statistically supported. A study that omits its sample size cannot readily sustain a quantitative claim about a physiological phytoacoustic effect. Likewise, without intensity data, it is impossible to determine whether two studies testing the same frequency are comparable, or whether one sits in a sub-threshold regime and the other in a cavitation regime. Exposure duration complicates the matter, since it is an incomplete proxy for delivered energy. Total dose depends on the product of intensity, exposure duration, and number of exposures, yet studies differ in whether they report a per-session duration or a cumulative total, and frequently do not say which.

Based on these findings, it is recommended that future plant acoustic studies adopt a minimum reporting standard, including: frequency, the delivered stimulus level measured at the tissue expressed in the unit appropriate to the medium (sound pressure level for airborne delivery, acceleration in $m·s^{-2}$ for substrate delivery, and spatial-peak temporal-average intensity in $W·c{m}^{-2}$ for liquid delivery), waveform, exposure duration with an explicitly stated dosing convention (per-session versus cumulative), a labeled schematic of the experiment geometry, propagation medium, and equipment models. The responsibility of maintaining highly detailed publication records also rests with journal editors, since acceptance of under-specified methods allows non-reproducible work to enter the permanent academic record. Separately, the database reveals the absence of any shared positive control. A reference stimulus known to elicit a defined response would enable cross-calibration of apparatus and dosimetry between laboratories. Without such an anchor, even fully reported studies cannot be placed on a common effect scale, since there is no way to confirm that two experiments deliver the same dose.

4.7. Research Opportunities

The taxonomic landscape of Figure 3 is shaped by economics, research sociology, and biology. Setting aside the model organism Arabidopsis thaliana, tomato (Solanum lycopersicum) and chrysanthemum (Chrysanthemum spp.) are the two most-studied crops, but their popularity has different origins. The popularity of tomato plants in the literature likely reflects its economic and nutritional importance [120] and is split across mechanistic and airborne-acoustic work. Chrysanthemum’s popularity, on the other hand, rests largely on the output of a single sustained research program working in the airborne-acoustic regime with a focus on callus tissue. The two therefore carry different meanings despite adjacent bar heights – tomato’s papers span multiple groups and study types, while chrysanthemum’s issue almost entirely from one academic lineage. Read uncritically, Figure 3 implies a breadth of community interest in chrysanthemum that does not exist, again demonstrating the need to evaluate each plot beyond its face value. Rice is placed as the fourth leading taxa, despite its role as a staple for more than three billion people [44].

Similar gaps are found in other vegetation types. First, although legumes as a group are well represented in Figure 1, that total is carried almost entirely by soybean (Glycine max) and mung bean (Vigna radiata). Globally important food-security legumes such as cowpea (Vigna unguiculata), lentil (Lens culinaris), and chickpea (Cicer arietinum) are nearly absent, appearing in at most one or two studies each despite their physiological differences from both the well-studied legumes and the temperate cereals. Second, the infrasonic regime is nearly empty; only five infrasonic conditions were identified in this review, even though wind-loading [89], seismic activity [154], thunder [155,156], and wind-turbine vibration [157] constitute the dominant low-frequency acoustic environment of living plants. Data scarcity in this regime is understandable, since an infrasonic wavelength (roughly 17 meters at 20 Hz) vastly exceeds the dimensions of any bench-scale chamber, so a free-field pressure wave cannot be easily established. Substrate-coupled delivery sidesteps this constraint entirely. Third, aquatic-plant, fungal, and algal responses have been examined only in isolated studies and lack systematic frequency-response characterization. The gap here is not merely one of missing treatments but a substantial missed opportunity, since marine algae are known to produce sound during photosynthesis [230]. In this case, an acoustically active group of organisms sits almost wholly outside the corpus as both emitter and potential receiver.

4.8. Future Work

Beyond these biological gaps, the database itself defines a methodological program, creating a standardized parameter map that is easy to share, validate, and expand as the literature grows.

Frequency-response curves The absence of finely discretized frequency sweeps is a clear deficiency in the existing record, and a tractable one. A program of logarithmically spaced exposures across the audible decade, on a small number of well-characterized taxa, would convert the field’s scattered single-tone claims into resolved response curves. In principle, frequency sweeps could be drawn in Figure 1 as horizontal segments since the database records their lower and upper bounds. But a segment of uniform width would nonetheless misrepresent the unequal energy distribution across a swept band, and would occlude the dot-area encoding of replication. Polyphonic and broadband stimuli admit no such representation at all, and a separate spectral-coverage figure, keyed to the Frequency Description field, would be required to place them alongside the tonal literature.

Calibrated substrate-vibration delivery Substrate vibration is the smallest treatment category in Figure 4, yet with calibrated acceleration reporting it offers a cleaner dose definition than airborne delivery, which itself is complicated by distance, enclosure acoustics, and frequency-dependent speaker response. Controlled tactile-transducer platforms with closed-loop intensity verification at the tissue would enable the field to report dose in a medium-appropriate, reproducible unit, and directly test whether substrate and airborne delivery produce equivalent biological outcomes at matched effective dose.

Cross-medium dosimetry conventions The reference-pressure and transmission-loss differences quantified in Section 4.3 should be formalized into a conversion and reporting convention. This would enable airborne, substrate, and liquid studies to be placed on comparable axes, with cavitation-regime studies explicitly partitioned rather than silently pooled.

Standardized confound tagging and pre-registration The flags developed in this study (Sonic Bloom/PAFT/Agri-Wave, Cellular Damage Risk, Dual Outcome, Low-tier Venue), together with the three parameter-quality tiers, provide a starting vocabulary for machine-readable meta-analysis. Coupled with pre-registration of the minimum reporting standard above, standardized tagging would enable field growth by addition rather than by repetition.

5. Conclusions

This study presents a structured, machine-readable database of 2,991 experimental conditions drawn from 404 publications in plant acoustics, spanning nearly 100 years from 1928 to 2025. By decomposing each study into its constituent organism × treatment conditions, the database makes the parameter space of the field directly visualizable and reveals four principal findings.

First, the field’s apparent consensus on stimulus parameters is largely an artifact. The conspicuous 100 dB intensity mode and the clustering of treatments on round-number frequencies appear to reflect parameter inheritance and experimental convenience, amplified by the disproportionate output and consistent reporting of a small number of prolific single-laboratory programs, rather than independent convergence on a biological optimum. The substantial body of music and polyphonic-playback work cannot be placed on these axes and is treated separately; the frequency- and intensity-level findings therefore characterize the tonal, single-frequency sub-literature.

Second, the literature tends to pool physically incomparable delivery media. Liquid sonication, whose dominant mechanism is acoustic cavitation and whose reported intensities use a different reference pressure, is routinely cited alongside airborne and substrate delivery as evidence for a single phenomenon. This is a category error that may have shaped both the field’s published claims and its replication failures.

Third, methodological rigor has not scaled with the field’s growth. The parameters that determine delivered dose, namely calibrated intensity at the tissue, source-to-target distance, exposure schedule, and propagation medium, remain chronically under-reported, rendering most cross-study comparisons unsupportable and gating quantitative meta-analysis. Addressing methodological rigor through a minimum reporting standard, rather than through additional under-specified experiments, may be one of the most pressing near-term needs.

Fourth, the parameter space itself remains largely unexplored, and the database makes the unexplored regions enumerable rather than merely suspected. The infrasonic regime contains five experimental conditions in the entire century-long record, despite wind loading, seismic activity, thunder, and wind-turbine vibration constituting the dominant low-frequency acoustic environment of living plants; no study bridges the 10–20 kHz seam between loudspeaker and piezoelectric hardware; the sub-70 dB range that characterizes a plant’s ordinary acoustic surroundings is scarcely sampled; algal and aquatic taxa account for two of the 825 plotted reports; and finely discretized frequency-response curves are essentially absent, so no claim of an optimal frequency rests on a resolved tuning curve. These gaps are tractable rather than fundamental, since each can be addressed with existing instrumentation, and together they define a coverage-driven experimental program that would grow the field by addition rather than by repetition.