Bimodal Contour Grammar — v0.2 User Guide

What this does

Bimodal Contour Grammar generates a musical contour by recursively expanding a simple formal grammar. The grammar is applied in a loop, building a sequence of phrases (each composed of an onset, nucleus, and coda). For every generated “point” (time, pitch), two actions occur simultaneously:

• Audio mode: the point is added to a PitchTier, which is later used to resynthesize the original sound via PSOLA (pitch‑synchronous overlap‑add).
• Visual mode: the point is drawn in the Praat picture window as a line segment or dot, with colour and thickness/size modulated by intensity and pitch class.

The method – generative contour grammar

The underlying model

The method treats melodic shape as a kind of syntax. Instead of saying “compose note by note,” it says:

• a whole contour can be recursively expanded into phrases
• a phrase can be expanded into functional components
• each component can be realized by one of several gesture types

So the contour is not arbitrary. It is rule‑governed, but still variable. In this sense, it is close to:

• generative linguistics, where sentences are built from phrase rules
• gesture‑based music theory, where shape matters more than exact pitch content
• procedural composition, where structure is generated from reusable primitives

Quick start

1. In Praat, select exactly one Sound object (any duration). This sound provides the base timbre – the grammar will impose a new pitch contour on it.
2. Run script… → Bimodal_Contour_Grammar.praat.
3. Adjust parameters:
   • randomSeed – 0 = auto‑seed from current time; any other number = fixed seed for reproducible results.
   • Image_width / height – dimensions of the visual output in the Praat picture window.
   • colorScheme – Pitch+Loudness Rainbow, PitchClass+Loudness Wheel, Intensity Heatmap, Octave Spiral.
   • lineStyle – Thin continuous line, Thickness varies with loudness, Dots with size varies with loudness.
   • baseLoudness / loudnessVariation – controls intensity modulation of the generated contour.
4. Click OK. The script runs the grammar, draws the contour, resynthesizes the sound via PSOLA, and creates a new Sound object named originalname_grammar.

Grammar rules

Terminal rules

The grammar uses a seeded random number generator (LCG) to ensure reproducibility. All random choices (duration, pitch offsets, rule selection) are derived from this seed.

What counts as a phrase

A phrase has three functional zones, forming a classic arch of activation → sustain → release:

1. Onset – the entry gesture

Its role is to establish motion and direction. In this method, the onset tends to rise – either by a jump upward or by a gliding ascent. So the phrase begins with activation or lift.

2. Nucleus – the main body

Its role is to stabilize or elaborate the pitch region reached in the onset. The nucleus is either a plateau (sustained pitch) or a wobble (local oscillation around a centre). This gives the phrase its central identity: steadiness or internal activity.

3. Coda – the release gesture

Its role is to close the phrase. In this method, the coda is always downward – either a moderate fall or a deep drop. Thus each phrase follows a broad shape of upward activation → sustain/fluctuation → downward closure. This is the real compositional core of the method.

Bimodal primitives – the addPoint procedure

This single procedure ensures that every grammar event has a dual representation: a sound‑modifying pitch point and a visual mark. The visual styles are applied live as the grammar runs.

Theoretical claim – what the method embodies

The strongest claim behind this method is:

That is, a contour is not just something heard in melody or seen in graphics; it is an abstract organizational pattern that can be generated once and rendered in more than one modality.

What is being composed: notes or shapes?

Primarily shapes. The method is not about choosing exact pitches first, but about choosing contour archetypes and then instantiating them numerically. The important musical object is a profile like “rise quickly, hover, descend strongly” – privileging direction, curvature, registral behavior, and phrase morphology over discrete note succession.

Why randomness is used

Randomness here is not the method itself; it is a way to produce variation within constraints. The grammar specifies allowable kinds of motion; random choice determines which variant occurs, how large the rise or fall is, how long each segment lasts, and where the local fluctuations go. The result is outputs that are recognizably from the same family, but never exactly the same – balancing coherence with novelty.

What “bimodal” adds conceptually

The most distinctive part is that the contour is treated as a shared abstract object across two sensory domains. The same form – height over time, intensity over time, segmentation into phrase units – is rendered as pitch movement in audio and as line/color/point behavior in graphics. This makes the method a kind of cross‑modal mapping, where pitch height maps to vertical position or color, intensity maps to brightness/thickness/dot size, and phrase segmentation appears both as heard articulation and visible shape.

What it is not

It is not primarily harmonic composition, tonal voice‑leading, motivic development in the classical sense, or analysis of an existing melody. It does not infer deep structure from the source sound; instead, it imposes a newly generated contour structure on the sound. Methodologically, it is closer to contour synthesis, gesture grammar, and procedural phrase generation than to transcription or conventional composition.

The method in one sentence

This method treats melodic contour as a grammar of phrase‑level gestures and realizes the resulting structure simultaneously as sound and image, making contour itself the central compositional object.

Parameters & defaults

Random seed

Visual display

Color & style

Audio synthesis

Display options

Visual styles – colour schemes

The line styles affect how the contour is drawn: continuous lines (with optional thickness variation) or discrete dots. The dot size maps intensity linearly between minDotSize and maxDotSize.

FAQ / troubleshooting