When we describe the emotional atmosphere of a poem, we usually reach for words such as joyful, somber, anxious, tender, angry, or calm. Those descriptions are useful, but they can also flatten experiences that are more complicated than a single emotion. Two poems may both feel sad, for instance, while producing very different effects. One may be quiet and resigned; the other may be frantic, furious, or overwhelmed. Two poems may both use positive language, but one may feel peaceful while the other feels ecstatic. A poem may sound threatening without sounding frightened, or vulnerable without sounding passive.
Valence, arousal, and dominance offer another way of talking about those differences. Together, they form a widely used framework for describing the emotional associations of language. Rather than assigning every word to a single emotional category, the framework places words along three dimensions: how positive or negative they are, how calm or activated they are, and how powerful or powerless they seem.
These dimensions are often abbreviated as VAD, which is where VerseVAD gets its name. VerseVAD uses established affective lexicons—large datasets in which words have been rated along these dimensions—to examine emotional patterns in poems and other texts. The resulting scores do not tell us what a poem ultimately means or how every reader will experience it. They describe tendencies in the poem’s vocabulary. Used carefully, those tendencies can give us a more precise way to compare poems, investigate surprising results, and return to the text with better questions.
Valence: Positive and Negative Association
Valence is probably the most immediately familiar of the three dimensions. It describes whether a word tends to carry a positive or negative emotional association. Words such as joy, love, beautiful, and peace generally receive relatively high valence ratings. Words such as grief, murder, disease, and despair generally receive lower ones. Many other words fall somewhere between those poles.
It is tempting to interpret a poem’s average valence as a direct measurement of whether the poem is “happy” or “sad,” but that would be too simple. Valence measures the associations of the words found in the text. It does not fully capture how those words interact within syntax, metaphor, irony, negation, or narrative context.
Consider the difference between the statements I am happy and I am not happy. A word-level lexicon recognizes happy as positive, but it may not fully account for the negation created by not. A sentence such as What a beautiful disaster brings positive and negative associations into direct tension. A poem may describe a beloved person through luminous, affectionate language while also making clear that the person has died. The vocabulary may retain considerable positive valence even though the poem’s larger emotional effect is elegiac.
This is not necessarily a failure of the analysis. In fact, it may identify one of the poem’s central features: grief expressed through the language of beauty, love, memory, or gratitude. A lower-valence poem may use openly painful language, while another poem about the same subject may avoid that vocabulary almost entirely. The two poems may both concern loss, but they construct loss differently.
Valence is therefore most useful when treated as a description of lexical tendency rather than a final diagnosis of tone. A relatively high score tells us that the poem contains more positively associated language, not that it is uncomplicatedly cheerful. A relatively low score indicates a greater presence of negatively associated language, not that the poem is artistically bleak or emotionally uniform.
The next step is always to look at the poem itself. Which words contribute most strongly to the score? Where do they appear? Do positive and negative terms cluster in separate sections, or are they intertwined? Does the poem move from one region of the valence spectrum toward another? Does repetition give a particular word unusual influence? These questions turn a numerical result into an interpretive starting point.
Arousal: Calmness and Activation
Arousal describes emotional or physiological activation. At the lower end of the scale are words associated with quiet, stillness, rest, softness, and calm. At the higher end are words associated with excitement, danger, speed, panic, violence, intensity, or exhilaration.
Arousal should not be confused with positivity. A highly aroused word can be positive or negative. Ecstasy and terror are both emotionally intense, although their valence differs dramatically. Likewise, low-arousal language can be pleasant, unpleasant, or emotionally ambiguous. Peaceful suggests positive calm, while numb may suggest a diminished or troubling state.
This is one reason the VAD framework is more informative than a simple positive-versus-negative scale. Valence alone cannot distinguish between quiet sorrow and explosive rage. Both may be negative, but one is low in activation while the other is high. Similarly, peaceful contentment and exuberant celebration may both be positive, yet they occupy very different emotional spaces.
In poetry, arousal can help us describe the energy of the vocabulary. A battle poem may contain language of motion, impact, noise, danger, and command, producing a relatively high-arousal profile. A pastoral poem may rely more heavily on stillness, distance, quiet observation, and slow movement. But subject matter does not determine the result automatically. A poem about violence may describe it in detached or restrained language, while a poem about an ordinary domestic moment may feel emotionally charged through its diction.
Arousal also becomes interesting when it changes within a poem. A poem may begin in relative calm and gradually accumulate more activated language. Another may open in crisis and slowly settle. Those shifts may correspond with changes in imagery, syntax, line length, rhythm, or narrative situation. In a corpus, arousal can help us ask whether certain collections, periods, speakers, or subjects are associated with greater emotional activation.
As always, the score is not the whole story. A poem can create intense suspense using relatively ordinary words. Rhythm, pacing, punctuation, repetition, and sound may generate energy that a lexical rating does not fully represent. Conversely, a poem may contain individually high-arousal words while placing them inside a reflective, distant, or ironic structure. VerseVAD measures one important source of emotional intensity—the associations carried by the vocabulary—but poetic intensity can emerge from many interacting features.
Dominance: Power, Control, and Agency
Dominance is often the least immediately intuitive dimension, but it can be one of the most revealing. It concerns the degree of power, control, autonomy, or agency associated with a word. Higher-dominance words tend to suggest strength, authority, command, capability, or control. Lower-dominance words tend to suggest vulnerability, weakness, submission, fear, dependence, or powerlessness.
Dominance is not the same as moral goodness. A tyrant may be highly dominant. A victim may be low in dominance without being negatively portrayed. Nor is dominance identical to confidence, although the two can overlap. The dimension is concerned broadly with whether the emotional position associated with a word feels powerful or constrained.
This can help distinguish poems that might otherwise appear similar. Two poems may both have low valence and high arousal, but their dominance profiles may differ. One might use the language of threat, command, violence, and control; the other might use the language of fear, helplessness, injury, and entrapment. Both are intense and negative, but they position power differently.
Dominance can be especially useful when studying poems concerned with authority, resistance, captivity, illness, gender, warfare, nature, spiritual submission, political power, or the relationship between human beings and forces larger than themselves. A speaker may begin in a position of confidence and gradually lose control. Another may appear powerless but reclaim agency through speech. A poem may describe an overwhelming natural landscape through words that diminish the human figure, or it may depict the speaker attempting to master that landscape through naming and observation.
The dimension can also complicate our first impressions. A poem that sounds quiet may nevertheless have high dominance if its vocabulary conveys certainty, possession, command, or self-control. A louder, more emotionally activated poem may have lower dominance if its speaker is panicked, trapped, pleading, or overwhelmed. This helps us separate emotional intensity from emotional power.
Because dominance is less common in everyday literary vocabulary than terms such as positivity or excitement, it can open particularly fresh lines of inquiry. It asks us to notice not merely what a poem seems to feel, but how much agency its language appears to possess. Who acts? Who is acted upon? Does the vocabulary suggest command or dependence? Is power stable, contested, surrendered, imagined, or lost?
Why the Three Dimensions Work Best Together
The real value of the VAD framework appears when the three dimensions are considered together. Each one answers a different question:
Valence: Is the language more positively or negatively associated?
Arousal: Is the language calmer or more activated?
Dominance: Does the language suggest greater power and control, or greater vulnerability and constraint?
A poem with high valence, low arousal, and moderate dominance might feel peaceful, secure, and gently affirmative. A poem with high valence and high arousal might feel ecstatic, celebratory, or passionately joyful. Low valence combined with low arousal may suggest sorrow, exhaustion, loneliness, or resignation. Low valence with high arousal may suggest terror, fury, crisis, or violent conflict.
Dominance adds another layer. High-arousal negative language with high dominance may sound threatening, wrathful, or forceful. High-arousal negative language with low dominance may sound frightened, desperate, or overwhelmed. Two poems can therefore have nearly identical valence scores while differing substantially in their emotional energy and sense of power.
These combinations should not be treated as rigid categories. Poetry is too context-dependent, historically situated, and linguistically complex for any three numbers to define it completely. Still, the dimensions create a useful map. They help us describe emotional language with more precision than a single label can provide.
They are particularly valuable for comparison. A score that seems unremarkable in isolation may become meaningful when placed beside another poem, a collection average, or the poet’s broader corpus. We may find that a seemingly gentle poem is unusually high in arousal compared with the poet’s other work, or that a famous poem is much lower in dominance than the collection surrounding it. The numbers become most informative when they reveal relationships.
How VerseVAD Produces the Scores
VerseVAD compares the words in a text with established VAD lexicons. These lexicons contain words that have been rated by human participants for valence, arousal, and dominance. The individual ratings are combined to create a standardized score for each word.
When VerseVAD analyzes a poem, it identifies words that appear in the selected lexicon and combines their scores to produce poem-level averages. It also reports match rates so users can see how much of the text was represented in the dataset. This is important because no lexicon contains every possible word, especially when analyzing archaic language, specialized terminology, invented words, unusual spellings, or proper names.
The software can also distinguish between token-weighted and type-weighted analysis. In token-weighted analysis, every occurrence of a word contributes to the result. If a word appears twenty times, it is counted twenty times. This allows repetition to influence the poem’s profile, which often makes sense in poetry because repeated language can shape the reader’s experience.
In type-weighted analysis, each unique word is counted once, regardless of how often it appears. This provides a different view: the general emotional range of the poem’s vocabulary without allowing repeated words to dominate the average. Neither method is universally better. They answer different questions.
Stopword controls provide another useful comparison. Stopwords are highly frequent grammatical words such as the, and, to, of, or be. These words are essential to language, but they may carry affective ratings that influence a result simply because they occur often. An analysis that includes all matched words captures the poem’s full lexical stream. A content-focused analysis that removes common function words may better highlight nouns, verbs, adjectives, and adverbs. Comparing both versions can reveal how much the result depends on grammatical language versus semantically prominent vocabulary.
These choices matter because computational analysis is not a neutral button that produces one inevitable answer. Every result depends on the selected lexicon, weighting method, preprocessing decisions, and coverage of the text. VerseVAD is designed to make those choices visible rather than hiding them behind a single authoritative-looking number.
What VAD Scores Cannot Tell Us
VAD analysis has real limitations, especially when applied to poetry. Lexicons typically assign ratings to words in isolation. Poems rarely use words in isolation.
Context may reverse or complicate a word’s emotional effect. Negation can change a statement’s meaning. Irony can create distance between the literal vocabulary and the poem’s intended force. Metaphor can transfer emotional associations from one domain to another. A historically positive word may function differently in an older text. A word may also have several meanings, each with different emotional implications.
VAD scores do not independently capture speaker, narrative development, sound, meter, syntax, imagery, allusion, or form. They do not know whether a frightening image is being embraced, rejected, remembered, imagined, or mocked. They cannot determine whether the reader trusts the speaker. They cannot tell us whether a poem succeeds artistically.
This does not make the method useless. It tells us what kind of evidence it provides. VAD analysis measures patterns in the affective associations of vocabulary. It is strongest when used comparatively, transparently, and alongside close reading.
An unexpected score should not automatically be dismissed as an error, nor should it be accepted as the poem’s hidden truth. It should be investigated. Which words produced it? Does repetition explain it? Is the score capturing a genuine tension between subject and diction? Is irony creating a gap between lexical association and overall meaning? Does the poem use pleasant language to describe something disturbing, or forceful language to conceal vulnerability?
Those questions are often more interesting than a perfectly predictable result.
A Starting Point for Reading, Not an Ending
Valence, arousal, and dominance give us a structured vocabulary for discussing emotional language. Valence helps us examine positive and negative association. Arousal helps us distinguish stillness from activation. Dominance helps us notice power, control, vulnerability, and agency. Together, they allow for a richer account of emotional texture than any one dimension could provide alone.
But the purpose of VerseVAD is not to translate a poem into three numbers and leave it there. The purpose is to make patterns visible and then return those patterns to the act of reading.
A high-arousal score may lead us to examine the poem’s verbs, rhythms, or images of movement. A low-dominance score may draw our attention to grammatical passivity, confinement, fear, or dependence. An unexpectedly positive valence score may reveal that a poem of grief is built from the vocabulary of love and beauty. The data points toward the language; the language remains the real object of study.
Poetry has always asked readers to attend closely to words. VAD analysis is one more way of practicing that attention. It does not replace intuition, interpretation, or human response. It gives those responses another set of coordinates—and sometimes shows us that the emotional landscape of a poem is stranger, more divided, and more interesting than we first assumed.