From Decoration to Syntax: Emoji Sequences as Emerging Grammatical Structures in Digital Communication

Abstract

Emojis entered everyday writing as decorative additions to digital messages, yet many users now arrange multiple emojis into patterned sequences that look less like ornamentation and more like structure. This study asks whether multi-emoji strings exhibit recurring positional regularities, ordering preferences, and combinatorial constraints that parallel (in limited ways) natural language syntax. Using a corpus-linguistics pipeline applied to a large public dataset of computer-mediated communication, we extract emoji-only sequences and analyze their length distributions, positional asymmetries, and co-occurrence patterns using information-theoretic and distributional measures. Results show that emoji sequences are not merely random bundles: they cluster into a small set of high-frequency “constructions,” including repetition-based intensification, peripheral “framing” with sparkles/hearts, and face-to-gesture orderings that resemble discourse-pragmatic packaging. We propose an Emoji Construction Grammar perspective in which stable, community-level templates emerge through use, supporting classic claims that grammar can be an emergent property of repeated social action (Hopper, 1987; Bybee, 2010). While emojis remain tightly integrated with surrounding text and context, their internal sequencing demonstrates systematicity significant for humanities-centered accounts of meaning-making in digital communication and for linguistic theories of how grammatical structure arises in new semiotic systems.

Keywords: emoji, digital communication, emergent grammar, corpus linguistics, CMC, languages & linguistics

Introduction

Why emoji sequences matter (beyond “playful decoration”)

Emojis are often discussed as decoration—tiny images that add emotion, humor, or “tone” to otherwise plain text. That view is partly right: emojis are frequently used to manage interpersonal nuance in contexts where gesture, facial expression, and prosody are absent or reduced (Dresner & Herring, 2010; Crystal, 2006). Yet everyday digital writing has also produced something more structurally interesting: people combine multiple emojis in ways that are not arbitrary. Consider how quickly many readers recognize differences between the following:

• 😂😂😂 (repeated laughter, often stronger than a single 😂)

• ✨new phone✨ (sparkle “frame” around a phrase, implying emphasis or stylization)

• 😭🙏 (often read as pleading or grateful intensity; order feels “right” to many users)

• 🙏😭 (similar elements, but many readers report a different feel, sometimes less conventional)

Even when meanings remain context-dependent, such sequences show regularities that resemble familiar linguistic notions: preferred ordering, peripheral markers, intensification via repetition, and limited “slots” that are more acceptable for certain emoji types than others. In other words, we are not only seeing new symbols; we may be seeing the beginnings of emergent grammar —regular patterns that arise from use rather than from top-down design (Hopper, 1987).

This question is not merely technical. In the humanities, grammar is historically tied to authority (standards, correctness, schooling) and to social identity (who is “legitimate” as a speaker/writer). Digital communication repeatedly disrupts those expectations by normalizing forms that are simultaneously informal and highly patterned (Thurlow & Mroczek, 2011; Tagg, 2015). Emoji sequencing provides a vivid case in which ordinary users—not institutions—shape regularities that become socially recognizable.

Background: emojis, CMC, and the emergence of structure

Computer-mediated communication (CMC) has long been a site where paralinguistic cues get re-invented. Emoticons (e.g., 🙂 ) and typographic choices (ALL CAPS, repeated punctuation) historically carried pragmatic force—marking irony, friendliness, teasing, or intensity (Dresner & Herring, 2010; Crystal, 2006). Emojis extend this work with a standardized, widely shared iconography maintained by the Unicode Consortium (Unicode Consortium, 2024). Still, standardization of characters does not determine how people use them. Users innovate, and patterns propagate socially.

Prior research has established that emojis contribute to sentiment, stance, and interpersonal meaning, and that interpretation varies by context and by rendering differences across platforms (Kralj Novak et al., 2015; Miller et al., 2016). Research has also documented historical shifts, such as increasing emoji use alongside decreasing emoticon use in some environments (Pavalanathan & Eisenstein, 2016). These studies clarify why emojis matter for meaning. The present study focuses on a narrower but theoretically consequential phenomenon: the internal organization of emoji sequences .

Two conceptual anchors guide the analysis. First, usage-based linguistics argues that grammatical structure can arise from repeated patterns in everyday interaction; frequency and conventionalization matter (Bybee, 2010; Tomasello, 2003). Second, construction-based theories treat grammar as a network of learned pairings of form and function—ranging from highly schematic templates to fixed idioms (Goldberg, 1995). These ideas are well-suited to emoji sequences because many appear as recurring templates with partially predictable “slots” (e.g., a repeated emoji, a framing pair, or a face followed by a gesture).

Research questions

This corpus-based study examines whether multi-emoji strings exhibit structural regularities that plausibly parallel (in a limited, domain-specific way) aspects of natural language syntax. The goal is not to claim that emoji sequences constitute a full independent language; rather, the goal is to test whether they show grammar-like constraints that are stable enough to be modeled systematically.

We address four research questions (RQs):

1. RQ1 (Position): Do certain emoji types disproportionately occur at the beginning or end of emoji sequences (i.e., positional asymmetry)?

2. RQ2 (Ordering): Are there recurring orderings between emoji classes (e.g., “face → gesture” vs. “gesture → face”), suggesting a preference akin to constituent ordering?

3. RQ3 (Combinatorial constraints): Are some emojis or classes “sticky” (often repeated or framed) while others resist certain combinations, indicating constraints comparable to selectional restrictions?

4. RQ4 (Modeling): Can a small set of reusable templates (“emoji constructions”) account for a substantial portion of frequent sequences, consistent with an emergent grammar approach?

Methodology

Corpus source and ethical considerations

Because private messaging data (SMS, WhatsApp, iMessage) raise serious privacy concerns and are rarely available for research, we focus on public CMC where large-scale textual evidence can be examined transparently. The corpus was derived from Reddit comments made available through the Pushshift dataset, a widely used resource for computational social science and digital discourse research (Baumgartner et al., 2020). Reddit is not “representative” of all digital communication; it is, however, large enough to support robust corpus-linguistic generalizations about recurring forms.

Ethical practice in internet research requires attention to context collapse, user expectations, and the potential harms of re-identification even when data are technically public (Association of Internet Researchers [AoIR], 2019). Accordingly:

• We analyze only de-identified text (no usernames) and report only aggregated patterns.

• We avoid quoting rare or potentially identifying sequences and instead use either high-frequency patterns or clearly labeled illustrative examples.

• We treat emoji sequences as cultural artifacts produced in social contexts rather than as purely computational tokens.

Operational definitions

To study emoji sequences as potential grammatical structures, we must define what counts as an “emoji,” what counts as a “sequence,” and what counts as “structure.”

• Emoji token: A Unicode emoji grapheme cluster as defined by Unicode emoji properties and sequences (Unicode Consortium, 2024), including single-codepoint emojis (e.g., 😂), skin tone variants (e.g., 👍🏽), and ZWJ sequences (e.g., 👩‍💻).

• Emoji-only sequence: A contiguous run of two or more emoji tokens with no intervening letters or numbers. Punctuation and whitespace were treated as boundaries except where punctuation is itself an emoji-like symbol (e.g., ❤️ vs. ♥) and for directional arrows that are encoded as emoji characters in Unicode emoji sets.

• Structure: Regularities in (a) position (e.g., an emoji disproportionately appears as a final element), (b) ordering (e.g., A→B is common while B→A is rare), and (c) combinatorial templates (e.g., repetition, framing, sequencing with arrows).

Data processing pipeline (corpus linguistics workflow)

Corpus linguistics emphasizes systematic sampling, explicit preprocessing, and quantitative summaries grounded in attested usage (Biber et al., 1998; McEnery & Hardie, 2012). Our workflow follows that tradition, adapted for Unicode emoji segmentation.

Emoji tokenization and extraction (algorithm)

Emoji processing is non-trivial because many emojis are multi-codepoint sequences (e.g., family emojis, profession emojis, skin-tone modifiers). We treat each emoji grapheme cluster as a token. In practice, researchers commonly implement this via Unicode-aware segmentation and the Unicode emoji property list (Unicode Consortium, 2024).

# Pseudocode (illustrative) for extracting emoji-only sequences

for comment in corpus:
    text = normalize_unicode(comment)
    tokens = split_into_grapheme_clusters(text)  # handles ZWJ and modifiers
    emoji_mask = [is_emoji(tok) for tok in tokens]

    # Extract contiguous runs of emoji tokens
    runs = []
    current = []
    for tok, is_e in zip(tokens, emoji_mask):
        if is_e:
            current.append(tok)
        else:
            if len(current) >= 2:
                runs.append(current)
            current = []
    if len(current) >= 2:
        runs.append(current)

    store(runs)

Two practical notes matter for interpretation:

• Unicode “morphology” is real morphology for users. Skin-tone modifiers and ZWJ sequences are not random: they obey composition rules. Even before we consider ordering in a string, emoji writing already contains a system of lawful combinations at the character level.

• Platform rendering differences remain a limitation. The same Unicode string can appear differently across platforms, which can affect interpretation (Miller et al., 2016). Corpus findings thus describe forms and their distribution; meanings remain context-sensitive.

Emoji classification for structural analysis

To test positional and ordering tendencies, we group emojis into broad semantic-pragmatic classes informed by Unicode categories and prior emoji interpretation research (Kralj Novak et al., 2015; Unicode Consortium, 2024). The classification is intentionally coarse, designed for general public interpretability and robust aggregation:

• Faces & emotion (e.g., 😂 😭 😡)

• Gestures & body (e.g., 👍 🙏 🙌)

• Hearts & affect symbols (e.g., ❤️ 💕 ✨)

• Objects & activities (e.g., 🍕 🎮 🎉)

• Directional / linking symbols (e.g., ➡️ 🔁)

• Flags and identity symbols (e.g., 🇺🇸 🏳️‍🌈)

These classes are not “universal semantics.” They are analytic tools for measuring distributional patterns—much like parts of speech in traditional grammar, which are also abstractions over varied meanings.

Quantitative measures

To evaluate whether sequences show grammar-like constraints, we use three families of measures: positional bias, association strength, and predictability.

Positional bias

For each emoji token e , we estimate whether it favors the sequence-final position (“closure”), the initial position (“launch”), or neither. We define a log-ratio positional bias score:

Positive values indicate a tendency toward final position; negative values indicate a tendency toward initial position. The log-ratio form is common in corpus work because it expresses asymmetry in an interpretable way and avoids misleading conclusions from raw counts alone (McEnery & Hardie, 2012).

Association strength via pointwise mutual information (PMI)

To identify “sticky” adjacent pairs (bigrams) that occur more than expected by chance, we compute PMI:

PMI is widely used in corpus linguistics to detect collocations—units that behave like conventionalized pairings rather than accidental neighbors (Biber et al., 1998). In emoji strings, high PMI can indicate an emergent constructional pairing.

Predictability via conditional entropy

Grammar is partly about predictability: if certain tokens strongly constrain what can follow, the system has structure. We estimate conditional entropy of the next emoji given the current emoji:

Lower conditional entropy means stronger constraint: after seeing x , fewer y are plausible. In natural language, function words and grammatical morphemes often reduce entropy by creating expectations; we ask whether certain emoji types play analogous roles.

Methodological limitations (declared up front)

Three limitations shape interpretation:

• Platform ecology: Reddit differs from private messaging and from platforms like WhatsApp or TikTok. Our results should be treated as evidence that emoji sequencing can become structured in public CMC, not as a universal claim about all digital environments (Tagg, 2015).

• Meaning is context-bound: Emojis can shift meaning dramatically by context, community, and irony. Corpus regularities indicate patterned form; they do not fix a single meaning (Miller et al., 2016).

• Unicode vs. perception: The analysis operates on Unicode strings. Rendering differences can change perceived similarity and may influence which sequences are favored (Miller et al., 2016).

Results

Overview: emoji sequences are short, frequent, and patterned

Across the extracted emoji-only runs, most sequences are short (typically 2–4 tokens), with a long tail of much longer strings. This resembles a familiar pattern from language: a small set of short constructions accounts for a large share of everyday usage, while longer utterances exist but are comparatively rare (Bybee, 2010).

This length pattern matters because it constrains what “syntax” can look like: in very short strings, structure is likely to appear as local ordering preferences and peripheral markers rather than as deep hierarchical embedding. Emoji sequencing, in this sense, is a plausible site for micro-syntax —shallow but socially stable templates.

RQ1: Positional asymmetries suggest “openers” and “closers”

Positional bias scores (Eq. 1) show that some emoji categories disproportionately appear at the right edge of sequences, functioning as closers. Faces expressing affect (e.g., laughter, tears) and hearts frequently behave this way, aligning with long-standing observations that CMC often places affective cues at the end of an utterance—similar to sentence-final particles or punctuation-like markers (Dresner & Herring, 2010).

Conversely, emojis that function as scene-setting or topical cues (objects, celebration items, identity symbols) show a weaker tendency toward initial positions in emoji-only strings, consistent with a “topic-first” packaging strategy: introduce the domain (🎮) and then add affect/commentary (😂).

In natural language terms, this resembles a pragmatic ordering: “content” first, “stance” last. It is not identical to syntax (there is no evidence here of nouns and verbs), but it is structurally analogous to how discourse packaging works in many languages—where the end of an utterance is a privileged site for interpersonal and evaluative marking (Goffman, 1981).

RQ2: Ordering preferences (face → gesture; object → affect)

Bigram analysis indicates that certain cross-class orderings recur more than their reverse. Two especially robust tendencies appear:

• Object/activity → affect: Sequences like 🎉😂, 🍕😍, or 🎮🔥 (topic then evaluation) are more conventional than reversing the order. This mirrors a common pragmatic structure: “here is what we’re talking about” followed by “here is how I feel about it.”

• Face → gesture in stance clusters: Pairs like 😭🙏 or 😂🙌 appear more conventional than the reverse in many contexts, suggesting that users treat the face as the core stance marker with the gesture as reinforcement, or vice versa depending on community norms. Importantly, this is a statistical preference, not a categorical rule.

These findings align with the broader idea that CMC devices evolve conventional placements for paralinguistic cues, much as emoticons were often used to disambiguate illocutionary force (Dresner & Herring, 2010; Searle, 1969).

RQ3: Combinatorial constraints and “construction families”

Three high-frequency construction families account for a substantial share of repeated patterns. They are not “grammar” in the traditional schoolbook sense, but they are form–function pairings that are stable enough to be recognized and reused—precisely what usage-based construction approaches predict (Goldberg, 1995; Bybee, 2010).

1) Repetition constructions (intensification and stance)

The simplest and most pervasive template is repetition:

• Template: X X (or X X X …)

• Function: intensification, prolongation, or heightened stance

In text, repetition has long marked emphasis (e.g., “soooo good”), and CMC normalizes it further through expressive spelling and punctuation (Crystal, 2006). Emoji repetition behaves similarly: 😂😂😂 tends to read as more intense than 😂, and ❤️❤️ often reads as warmer than ❤️. This is not arbitrary; it is an iconic strategy in which “more form” maps to “more meaning,” a well-known semiotic principle (Danesi, 2017).

Repetition also acts as a stabilizer: if users are unsure which emoji best fits, repeating a single conventional emoji can produce a clear stance without inventing a novel combination.

2) Framing constructions (peripheral markers)

A second family uses the same emoji at both ends of a string, framing either a short phrase (in mixed text-emoji usage) or other emojis:

• Template: F … F (where F is often ✨, 💖, 🔥)

• Function: highlight, stylize, or “quote” a segment; mark affective or aesthetic stance

Framing is structurally significant because it creates a two-slot dependency: the first F “projects” an expectation that another F will close the unit. That expectation is grammar-like in the sense captured by predictability (Eq. 3): seeing an initial ✨ increases the probability of a later ✨. Comparable projection effects occur in language with paired punctuation and brackets.

Framing also resonates with multimodal literacy: visual design often uses borders and symmetry to mark a unit as special (Kress & van Leeuwen, 2006). Emoji framing imports that visual logic into linear text.

3) Linker constructions (arrows, repetition marks, transitions)

A third family uses directional or linking symbols to form mini-narratives or transformations:

• Template: A ➡️ B (or A ➡️ B ➡️ C)

• Function: sequence, causality, before/after, journey, escalation

Even when playful, these constructions have a compositional logic that resembles clause chaining: symbols like ➡️ behave like connectors that constrain what comes next—again lowering conditional entropy for the following token relative to unconstrained sequences (Eq. 3). This parallels how discourse connectives structure relations in language, though here the relations are often schematic or iconic rather than propositional.

Collocational “hot spots”: high-association pairs as emergent units

PMI (Eq. 2) highlights pairs that behave like collocations—elements that are more strongly associated than their individual frequencies would suggest. In emoji practice, many such pairs are affect clusters (face + gesture; heart + sparkle) or conventional intensifiers.

Table 1: High-frequency “emoji construction” families observed in the corpus workflow (examples are illustrative and not intended as fixed meanings).

For general readers, the crucial point is not which specific pair is “most common.” The crucial point is that many users converge on shared templates. That convergence is what makes grammar a plausible term here—not because emojis suddenly become a full language, but because social repetition creates recognizable form–function pairings.

RQ4: Toward an Emoji Construction Grammar (ECxG)

The recurring templates motivate a construction-based model. In classic construction grammar, speakers learn a network of constructions ranging from fixed idioms (“by and large”) to schematic patterns (“the X-er, the Y-er”) (Goldberg, 1995). Emoji sequences appear to develop a comparable range:

• Fixed micro-constructions: highly conventional clusters (often repetition or paired symbols)

• Semi-schematic templates: frames and linkers with variable internal slots (✨ X ✨; A➡️B)

• Open-ended strings: ad hoc sequences with weaker predictability

We therefore propose Emoji Construction Grammar (ECxG) as a descriptive framework: emoji sequences are treated as a set of community-learned templates whose “grammar” is probabilistic, shallow, and strongly tied to discourse function.

Discussion

Are emoji sequences really “syntax”?

If “syntax” is defined narrowly as hierarchical phrase structure with categorical parts of speech and recursive embedding, then emoji sequences do not straightforwardly qualify. They rarely show stable categories comparable to noun/verb distinctions, and many sequences remain interpretively underspecified without text. However, if we treat syntax more broadly as a system of constraints on combination and ordering—constraints that are socially learned and used for meaning—then emoji sequences exhibit syntax-like behavior.

In this respect, emoji sequencing resembles other domains where limited but real combinatorial structure emerges: gesture sequences, conventionalized interjections, and formulaic discourse markers. Gesture scholars have long noted that visible action can form structured contributions to communication, especially when coordinated with speech (Kendon, 2004; McNeill, 1992). Emojis function as written proxies for some of those cues, and their sequencing can reflect structured “packages” of stance and evaluation.

Emergent grammar: why regularities arise without formal instruction

Hopper (1987) argued that grammar is not only a static system; it can be an emergent phenomenon arising from recurrent patterns in usage. The present findings fit that view: repeated emoji practices stabilize into templates because they solve recurring communicative problems in CMC.

Several pressures encourage such stabilization:

• Speed and efficiency: Short, reusable emoji templates are quick to type and easy to recognize.

• Social alignment: Shared patterns index in-group competence; using conventional emoji sequences can signal belonging (Zappavigna, 2012).

• Ambiguity management: Emojis are often ambiguous; stereotyped sequences reduce interpretive uncertainty by providing redundancy (e.g., repetition and clusters).

Usage-based linguistics predicts exactly this: frequent patterns become entrenched, and entrenched patterns become available as units (Bybee, 2010; Tomasello, 2003). In emoji practice, the “units” are often not single emojis but multi-emoji bundles.

Emoji sequences as discourse grammar (stance, framing, and closure)

The strongest positional effects in the corpus concern utterance-level stance: faces and hearts often function as closers. This is consistent with CMC research showing that nonverbal-like markers can disambiguate illocutionary force—what an utterance is doing socially (Dresner & Herring, 2010; Searle, 1969). In a sense, many emoji sequences operate less like noun phrases and more like pragmatic particles or evaluative punctuation.

This also helps reconcile two truths that can seem contradictory:

• Truth 1: Emojis are playful and flexible; their meaning shifts across contexts and communities.

• Truth 2: Emoji usage is not chaotic; it shows community-level regularities.

Discourse markers in spoken language show the same duality: “like,” “you know,” and “okay” are flexible and context-sensitive, yet they have distributional constraints and conventional placements. Emoji sequences appear to be developing similar distributional norms within digital communication.

What counts as a “grammatical constraint” in a new semiotic system?

One risk in discussing emoji grammar is over-claiming. Not every statistical preference is a rule, and not every conventional pattern is a grammar. For that reason, the study emphasizes constraints rather than categorical well-formedness.

We suggest three grades of constraint, each with different theoretical implications:

1. Hard constraints (Unicode composition): Some combinations are formally governed by Unicode sequences (e.g., skin tone modifiers attach to certain base emojis). These constraints are machine-enforced but also socially meaningful: users experience them as part of what an emoji “is” (Unicode Consortium, 2024).

2. Soft constraints (community convention): Ordering tendencies and framing expectations are not enforced but are socially recognizable; violating them can feel odd or can create new effects.

3. Interpretive constraints (contextual anchoring): Many sequences require textual or situational context to stabilize meaning. Here the “grammar” is interactional: the sequence’s function depends on what it is responding to (Goffman, 1981).

This graded approach is consistent with a humanities perspective that treats communication as social practice. “Grammar” is not only an abstract system; it is also a set of normative expectations negotiated in communities.

Comparison with earlier paralinguistic systems: emoticons and punctuation

Emoticons provided early evidence that CMC users invent conventional visual cues with pragmatic force (Dresner & Herring, 2010). Pavalanathan and Eisenstein (2016) showed that emojis increasingly compete with emoticons for paralinguistic functions. Our findings extend that story: emojis are not only replacing emoticons as single markers; they are forming multi-unit patterns that have no direct analogue in traditional punctuation.

Repetition, framing, and linkers resemble punctuation in that they structure messages, but they also resemble gesture in their iconicity and affect. This hybrid character may explain why emoji sequences are fertile ground for emergent grammar: they recruit both visual design logics (Kress & van Leeuwen, 2006) and conversational logics (Goffman, 1981).

Interpretation variability and the limits of corpus inference

Miller et al. (2016) demonstrated that emoji interpretation varies across platforms and users. That variability places an important limit on any grammar claim: if the same emoji can be read differently, then sequences built from those emojis can also diverge in meaning. Corpus statistics cannot solve that problem by themselves.

However, variability does not eliminate structure. Natural language also contains ambiguity, polysemy, and context-dependent meaning; grammar constrains forms without fully determining interpretation. In emoji sequences, we can similarly separate:

• Form-level regularities (ordering, framing, repetition)

• Meaning-level inferences (what the sequence implies in a specific conversation)

A key contribution of a corpus-linguistics approach is to identify stable form-level tendencies even when meaning remains flexible.

Novel contribution: a humanities-centered “construction inventory” for emoji

Many computational studies focus on emoji prediction or sentiment classification (Felbo et al., 2017; Kralj Novak et al., 2015; Barbieri et al., 2018). Those approaches are valuable, but they can treat emojis as features rather than as culturally meaningful forms that develop conventions. The present study contributes a complementary, humanities-centered inventory approach:

• Identify recurring templates as cultural artifacts.

• Connect those templates to discourse functions (stance, emphasis, framing).

• Model the emergence of regularities as a product of social repetition and interactional needs.

This approach supports interdisciplinary conversation: linguistics supplies tools for pattern detection, while humanities perspectives supply the interpretive lens for understanding what those patterns do in social life.

Future research directions

Several research avenues would deepen and test the claims made here:

• Cross-platform comparison: Compare Reddit with private messaging corpora (where ethically and legally possible), and with platforms where emoji aesthetics are central (e.g., Instagram captions). This would test whether the same constructions generalize.

• Longitudinal change: Track whether specific constructions (e.g., ✨…✨ framing) rise and fall over time, and whether new templates replace older ones—an empirical lens on emergent grammar as historical process (Hopper, 1987).

• Comprehension experiments: Combine corpus findings with user studies testing whether readers judge some sequences as more “natural” than others, and whether position changes interpretation. This triangulation would connect distribution to cognition.

• Community-sensitive analysis: Different subreddits may develop distinct emoji micro-grammars. Mapping that variation would illuminate how grammar and identity co-evolve in digital communication (Zappavigna, 2012).

Conclusion

This study set out to evaluate whether emoji sequences in digital communication show emerging grammatical structure. Using a corpus-linguistics workflow applied to large-scale public CMC, we find that multi-emoji strings exhibit robust non-randomness: they are short but patterned, show positional asymmetries (especially affective closers), display ordering preferences between semantic-pragmatic classes, and cluster into recurring construction families such as repetition, framing, and linkers.

These regularities support a usage-based and emergent-grammar interpretation: users repeatedly solve communicative problems—tone, emphasis, stance, and narrative compression—and their solutions stabilize into conventional templates (Bybee, 2010; Hopper, 1987). The resulting system is not a full language, but it is more than decoration. It is an evolving set of social conventions with grammar-like constraints, offering a concrete window into how new semiotic resources can become structured through everyday practice.

References