Lab study

Vocal space: does naming reverb character move the vocal back?

An eight-generation test design for a single question — whether adding explicit space language to a style line changes where the vocal sits in the mix.

Hypothesis

H1. Adding an explicit space descriptor — a named reverb character plus a placement word — to an otherwise identical style line produces vocals that are consistently heard as further back in the mix.

H0. The descriptor makes no reliable difference; apparent changes are within the normal variation between takes of the same prompt.

This matters because space language is cheap in characters and often the first thing people cut when trimming a prompt. If it does nothing, cut it without regret. If it works, it is one of the highest-value phrases you can spend characters on.

Prompt pair

Both arms use the same genre, instrumentation, tempo and vocal type. The only difference is the trailing space clause. Lyrics are identical across all eight takes — a two-verse, one-chorus sheet with no tag beyond the section markers.

Paste both lines into Prompt Diff before generating. Any difference beyond the trailing clause invalidates the study.

ARM A (control)
indie pop, jangly electric guitar, warm bass, brushed drums, female vocal, mid-tempo 104 BPM, bittersweet mood

ARM B (variant)
indie pop, jangly electric guitar, warm bass, brushed drums, female vocal, mid-tempo 104 BPM, bittersweet mood, wide plate reverb on the vocal, vocal set back in the mix

What to listen for

Score each take on the three criteria below, 0–2, deciding the meaning of each score before you press play. The point of writing this down in advance is that "sounds better" is not a criterion and will quietly absorb your expectations.

The third criterion is deliberately a counterweight: a variant that wins on depth but loses intelligibility has not improved the song, and the sheet should be able to say so.

Criterion	0	1	2
Perceived depth — how far behind the instruments the vocal sits	Vocal pinned to the front, dry	Slight separation	Clearly behind the band, sitting in a space
Tail character — audible reverb quality after phrase ends	No audible tail	Short, generic tail	Distinct plate-like tail with a sustained decay
Intelligibility cost — did the words survive the space	Words obscured	Occasional loss on dense lines	Fully intelligible throughout

Procedure

The downloader exports all eight takes with predictable filenames, and the parser can give you RMS and peak figures per take if you want a rough objective companion to the ear scores. The steps:

Generate in the order A, B, A, B, A, B, A, B — four takes per arm.
Export each take and name the files by take number only, so the arm is not visible while scoring.
Shuffle the playback order.
Listen on one system for the whole study. Switching between headphones and speakers mid-session ruins depth judgments.
Score all eight, then reveal which take belonged to which arm.
Compute arm means per criterion and total.

Verdict framework

Fill the table below from your own run. The rows describe how to read a result, not what we observed — the study is published as a design, and any numbers you see here are placeholders.

Four takes per arm cannot separate small effects from chance. Report what you heard, and resist turning a weak-effect result into a rule.

Pattern in your score sheet	Verdict	What to do next
B leads on depth by ≥1.0 mean, direction consistent across takes	Clear effect	Keep the space clause as standard in this genre
B leads by 0.3–0.9, or direction flips on one or two takes	Weak effect	Keep it, but do not spend more characters elaborating
Means within 0.3, no consistent direction	No effect	Cut the clause; spend the characters on instrumentation
B wins on depth but loses ≥1.0 on intelligibility	Unstable	Try a shorter descriptor, or move space language to production only
Both arms score 2 on depth	Ceiling	Redesign with a drier control; the genre is supplying the reverb

Variations worth running next

Swap plate for hall, spring or room and see whether the character words are distinguishable at all.
Drop the placement clause and keep only the reverb name — which half of the phrase is doing the work?
Rerun in a dense genre such as metal, where the mix has far less room for depth cues.
Rerun with a male vocal, and with a choir, to test whether the effect depends on vocal type.

FAQ

Why not just measure reverb with an analyser?

Because the question is perceptual. A measurable tail that nobody notices in the song has not achieved anything. Objective figures from the parser are a useful companion, not a replacement.

Do I have to score blind?

Not strictly, but expectation bias on depth judgments is severe. Renaming files to numbers costs a minute and materially improves the result.

Can I use my own style line instead?

Yes, and you should if indie pop is not your area. Keep the structure — one clause changed, everything else frozen — and the design still works.

How long does the study take?

About an hour, most of it listening. Generating the eight takes is the quick part.