PaperlayerView exact revision on arXiv

Source-linked AI summary

How Prolific Sellers Self-Present: Dissecting the Communication Patterns of 1.6 Million Reverb Listings

David M. Markowitz

arXiv:2608.29952v1cs.CL

TL;DR

The paper asks how language in online marketplace listings reflects seller status and examines this question through content, style, and structural analysis of 1.6 million Reverb listings. Prolific sellers emphasized objective and functional product information more than subjective characteristics, while also writing more narrative-like and longer descriptions; the study’s observational design limits causal claims and does not establish effects on sales.

  • Problem

    The paper examines how language patterns in online marketplace product descriptions differ between prolific and everyday sellers and relate to seller self-presentation.

  • Method

    The study analyzes 1.6 million Reverb listings for content, style, and structural differences between prolific and everyday sellers.

  • Results

    Prolific sellers emphasized objective and functional product information more, subjective characteristics less, narrative-like style more, and longer descriptions than everyday sellers.

  • Takeaways & Limitations

    Prolific sellers’ descriptions were technically focused in content but more story-like in communication style.

  • Takeaways & Limitations

    The observational study cannot support causal claims or determine whether language differences translate into sales.

Abstract

from arXiv · show

The current paper draws on self-presentation theory and warranting theory to evaluate how the language patterns in an online marketplace reflect seller status (i.e., a prolific seller versus an everyday seller). Using 1.6 million musical instrument listings from Reverb.com in search of content, style, and structural differences in seller product descriptions, the evidence suggested prolific sellers tend to focus more on objective and functional aspects of a product (e.g., its features and specifications) and less on subjective characteristics like tone, relative to everyday sellers. Prolific sellers also communicated in a more narrative-like style, which was driven by an elevated use of personal pronouns, and they used longer descriptions than everyday sellers. Therefore, what prolific sellers focus on tends to be quite technical, but how they communicate this information is typical of a story that is told to potential buyers. Implications for self-presentation theory and warranting theory are discussed.

Affiliations

The paper examines how language in Reverb.com product descriptions can reveal differences between prolific and everyday sellers, connecting large-scale marketplace communication to self-presentation, impression management, and warranting theory.

  • The study compares prolific sellers, defined as those selling above-average numbers of products, with everyday sellers through their product descriptions.
  • Using Reverb listings, the paper evaluates what sellers communicate, how they communicate it, and how their words are structured.
  • Product descriptions function both as accounts of the item being sold and as reflections of the seller’s identity and intended impression.
  • The prolific-versus-everyday comparison offers a naturally occurring, large-scale contrast in seller expertise that is difficult to create in laboratory settings.
  • The work uses marketplace language to extend theoretical understanding of self-presentation, impression management, and warranting processes.

A Primer on Psychology of Language Research: The Words-As-Attention Framework

Psychology of language research distinguishes what people communicate from how they communicate and how language is organized. The words-as-attention framework uses these dimensions to approximate psychological characteristics from language.

  • Content: Content words, including nouns and verbs, describe what people communicate and can cluster into coherent themes through topic modeling.A cited example found more home-cooking language and less alcohol-related language in healthy versus unhealthy cities.
  • Style: Style words, including articles, prepositions, and pronouns, represent how people communicate rather than what they communicate about.Although style words are drawn from a relatively small vocabulary, they make up nearly half of English words and relate to social and psychological dimensions.
  • Style: First-person singular pronouns have been associated with psychological attention, distress from romantic breakups, academic performance, and interest in deep thinking.The cited relationships vary by outcome, including positive and negative associations.
  • Style: The analytic thinking index is a bipolar continuum that approximates communicative and thinking style from linguistic patterns.Higher-performing college students used more articles and prepositions but fewer storytelling words, including auxiliary verbs, personal pronouns, and conjunctions.
  • Structure: Verbal structure concerns how words and sentences are organized and is commonly measured with readability and lexical-diversity metrics.Examples include Flesch readability and type-token ratio; structurally complex, less readable writing has also appeared in deception research.
  • Framework: Psychological information can be inferred from language content, communication style, and the general arrangement of words and sentences.The words-as-attention framework has received substantial support across many studies.

Self-Presentation, Warranting, and Language Patterns

Self-presentation theory frames language as a way to manage impressions online, while warranting theory explains why verifiable claims receive more evidentiary value. Reverb marketplace descriptions connect these ideas because sellers’ claims can be checked through products and feedback.

  • Self-Presentation: Self-presentation involves strategically managing the impressions people give others, and language patterns help reveal this process online.Prior studies linked concrete information, self-references, and verbal confidence with trustworthiness, warmth, or competence judgments.
  • Warranting: Warranting theory proposes that receivers value information more when it is difficult for communicators to manipulate.Claims about product specifications have higher warranting value than claims generated entirely by the seller.
  • Warranting: Instrument specifications are objective and verifiable after delivery, whereas claims about sound are subjective and difficult to falsify before purchase.This asymmetry predicts greater impact for objective content and a stronger dynamic among prolific sellers who face verification more often.
  • Marketplace Setting: Reverb’s bidirectional feedback and longitudinal seller histories make accurate, complete descriptions consequential for perceived trustworthiness and credibility.Incomplete or inappropriate descriptions can put sellers at risk for decreased future sales.
  • Marketplace Setting: The marketplace enables comparison of prolific and everyday sellers’ verbal descriptions independently of inventory size.Reverb tracks selling volume and exposes self-presentation markers at scale.

The Current Paper

The paper analyzes Reverb listings to compare prolific and everyday sellers across content, style, and verbal structure. It predicts greater attention to verifiable product attributes and examines style and structure exploratorily.

  • Marketplace Setting: Because buyers cannot physically inspect or play products before purchase, listing descriptions carry substantial evaluative weight.Descriptions function both as product descriptions and as acts of self-presentation.
  • Data and Approach: The analysis evaluates seller language across content, communicative style, and verbal structure dimensions.Content uses bottom-up thematic extraction; style uses an analytic thinking index; structure uses readability and type-token ratio metrics.
  • Hypotheses and Questions: H1 predicts that prolific sellers will attend more to verifiable, objective product attributes and less to unverifiable, subjective attributes than everyday sellers.The prediction follows warranting logic because prolific sellers encounter product verification more frequently.
  • Hypotheses and Questions: RQ1a-b asks how style and structural patterns in product descriptions differ between prolific and everyday sellers.The paper does not receive a direct warranting-theory prediction for these dimensions, so they are examined exploratorily.
  • Data and Approach: The study analyzes 1.6 million Reverb listings while accounting for visual metadata and seller characteristics to isolate language differences.Reverb’s open-ended descriptions allow sellers to choose what to emphasize, how to communicate, and how much of themselves to include.

Data Collection and Preprocessing

The study collected Reverb marketplace listings and retained a large English-language sample of live listings for analysis.

  • Collection: 2.1 million initial Reverb listings were collected through Reverb’s dedicated API.The extracted fields included descriptions, listing duration, price, category, visual metrics, listing state, and engagement measures.
  • Preprocessing: Only live listings were retained because ended and sold listings were often unavailable from the API.
  • Preprocessing: Non-English descriptions were automatically removed before analysis.
  • Final sample: 1,628,876 listings remained, covering over 300 million words.Descriptions averaged 201.97 words, with a median of 137 and a range from 1 to 11,460 words.

Automated Text Analysis

The paper used LIWC and bottom-up topic modeling to quantify linguistic categories and extract interpretable content themes from product descriptions.

  • LIWC analysis: LIWC counts words as percentages of each text’s total word count across validated social, psychological, and grammatical categories.The example illustrates counts for self-references, articles, and money-related terms.
  • Topic modeling: The Meaning Extraction Method removes function and low-frequency words before representing content-word presence with binary scores.Unigrams, bigrams, and trigrams were used as textual units.
  • Topic modeling: Principal Component Analysis with varimax rotation clustered content words into coherent themes.Themes were retained when they explained meaningful variance, had eigenvalues above one, and were face-validly interpretable.
  • Topic modeling: Eight themes were retained and saved as standardized regression weights for later statistical models.The weights had M = 0 and SD = 1.

Style Patterns: Analytic Thinking

The analysis operationalized communication style and structural properties with analytic thinking, readability, lexical diversity, and seller-type comparisons using marketplace covariates.

  • Analytic thinking: The analytic thinking index combines eight function-word categories into a standardized communication-style measure.Lower scores indicate more dynamic, storytelling, informal communication, whereas higher scores indicate hierarchical, formal communication.
  • Structural measures: Flesch Reading Ease measures readability using words per sentence and syllables per word.Higher scores indicate more readable text.
  • Structural measures: Type-token ratio measures lexical diversity by comparing unique, varied terms with reused terms.Both readability and TTR were measured using quanteda.textstats in R.
  • Analytic plan: Models compared prolific sellers with everyday sellers while controlling for price, inventory, visual content, listing age, category, condition, seller status, and engagement.Standard errors were clustered at the seller level because seller-level characteristics could correlate with feedback volume.

Results

Results showed that prolific sellers emphasized technical product information and used less subjective tone language, while the extracted themes covered specifications, post-purchase information, and acoustics.

  • Thematic results: Eight themes were grouped into product specifications, post-purchase information, and acoustics.Product specifications included guitar body, drums, electronics, response latency, and power supply; post-purchase information included shipping and customer service; acoustics comprised tone.
  • Statistical results: Two findings were statistically significant at the 5% level in the seller-type analyses.
  • Thematic results: Prolific sellers emphasized guitar-body details such as frets, scale length, and the bridge rather than appearance or feel.
  • Thematic results: Prolific sellers focused on functional, structural, and playability-related features instead of aesthetics.
  • Thematic results: Prolific sellers used fewer tone-related terms such as rich, warm, smooth, and clarity than everyday sellers.The tone association was statistically significant at p = .039, and H1 was partially supported.

Style Patterns

Prolific sellers differed from everyday sellers in analytic-thinking-related language, personal-pronoun use, and description structure. Their listings were more narrative-like and longer, while the apparent lexical-diversity difference was explained by description length.

  • Analytic thinking: Seller type negatively predicted analytic thinking, with a statistically significant relationship (B = -1.52, SE = 0.65, t = -2.35, p = .019, Cohen’s f2 = 0.001).The analytic-thinking measure was an eight-category composite, and eight fixed-effect models were used to identify the dimensions driving the overall effect.
  • Communication style: Prolific sellers used a more informal, dynamic, and narrative-like communication style than everyday sellers.The paper links this style pattern to elevated use of personal pronouns.
  • Personalization: Personal pronouns were the significant analytic-thinking dimension after multiple-comparisons correction, with prolific sellers using more terms such as I, you, and she.The reported relationship was B = 0.38, SE = 0.13, t = 3.01, p = .003, corrected to p = .021.
  • Structural patterns: Seller type predicted type-token ratio before controlling for token count (B = -0.02, SE = 0.01, t = -2.13, p = .033), but not after control (B = 0.003, SE = 0.003, t = 1.00, p = .317).TTR was strongly negatively correlated with log token count (r = -0.90, p < .001), supporting description length as the relevant structural factor.
  • Structural patterns: 16% more text appeared in prolific sellers’ listings, as prolific seller status positively predicted token count (B = 0.15, SE = 0.05, t = 2.75, p = .006).Description length may account for the observed structural difference between prolific and everyday sellers.

Discussion

Across content, style, and structural dimensions, prolific sellers differed from everyday sellers in both what they communicated and how they organized descriptions. Their descriptions were more technical yet more narrative-like, longer, and oriented toward claims that could withstand repeated verification.

  • Content: Prolific sellers emphasized functional, structural, and playability-related features rather than subjective aspects such as an instrument’s look, feel, or sound.Examples included frets, scale length, and bridge hardware.
  • Style: Their descriptions were less analytic and more narrative-like, with especially frequent use of personal pronouns.Thus, technical product information was communicated in a personal, story-like style.
  • Theoretical implications: The findings suggest that prolific sellers can present themselves as technical and personal simultaneously, potentially signaling competence associated with prolific selling status.The authors note that future experimental work would help validate this assertion.
  • Warranting implications: Warranting logic offers one explanation for prolific sellers’ emphasis on specifications: objective claims are more verifiable against the received product than subjective acoustic claims.Repeated exposure to buyer verification may favor claims that can withstand scrutiny.
  • Theoretical implications: The study extends self-presentation research from one-shot encounters to marketplaces where sellers present the same self hundreds or thousands of times to an accumulating audience.The authors identify whether these patterns generalize across other marketplaces and settings as an open question.
  • Structure: Prolific sellers wrote longer descriptions, roughly 16% more text after accounting for covariates.Once description length was accounted for, differences in lexical diversity measured by TTR were not obtained.

Practical Implications

The study offers practical guidance for sellers and platforms based on how prolific sellers’ descriptions differ from everyday sellers. Prolific sellers emphasize verifiable product information while combining technical content with personal, narrative-like communication.

  • Practical Implications: The results provide empirical evidence about what and how prolific sellers communicate compared to everyday sellers.The observed differences concern sellers’ language patterns rather than a causal effect on marketplace outcomes.
  • For Sellers and Platforms: The findings provide practical evidence about how more experienced sellers write compared to less experienced sellers.The study describes differences in communication patterns, but does not establish that these differences produce more sales.
  • For Sellers and Platforms: Prolific sellers lean toward verifiable content and specifications over speculation.This suggests that objective details may help buyers assess claims before purchase.
  • For Platforms: Platform-supplied advice could encourage sellers to present claims that buyers can verify and assess objectively before purchase.The proposed guidance follows from the association between prolific seller status and emphasis on specifications and objective details.
  • For Sellers and Platforms: Subjective claims such as tone may be more characteristic of novice sellers than prolific sellers.The paper presents this as an implication of the tone result rather than a causal conclusion.

Limitations and Future Directions

The study’s conclusions are constrained by small effects, an indirect seller-status measure, an observational design, and a corpus limited to English Reverb listings. Future work should separate experience from sales output and test whether the patterns generalize across marketplaces with less expert buyers.

  • Limitations: Effect sizes were small in magnitude, limiting the strength of the observed seller-type differences.This is identified as the first limitation of the work.
  • Limitations: Prolific seller status was operationalized through accumulated feedback rather than a direct measure of sales volume.Feedback reflects both transaction volume and platform tenure, which the present data cannot separate.
  • Future Directions: Future transaction-level data could disentangle seller experience from actual sales output.The paper notes that these factors may relate to language in different ways.
  • Limitations: The observational design cannot establish causal claims, and sold items were excluded because Reverb’s API does not systematically index them for analysis.The study therefore describes differences in self-presentation between seller types rather than effects on selling outcomes.
  • Scope and Generalizability: The corpus was restricted to English descriptions from a single specialized marketplace.Reverb buyers may possess domain knowledge that makes specifications more valuable than they would be in general-purpose marketplaces.
  • Future Directions: Whether these content, style, and structural patterns generalize to marketplaces with less expert buyers remains an open question.The paper calls for additional research across other online marketplaces and self-presentation settings.
Loading 2608.29952v1…