The Blind Test That Settles Electric Guitar Tonewood Does Not Exist

I set out to publish one number: the hit rate of listeners in a blind test of electric guitar body wood, with the player and the strings held fixed. I expected a figure near 50%. I could not find that number in any source I could read. This post says what I did find, what it supports, and why I now hold my position at lower confidence.

That is a small defeat for me. I like a blind test that gives a clean answer. This search gave me four partial answers instead.

Question

Can listeners tell two electric guitars apart by the body or neck wood alone, when the player, strings, pickups and amplifier stay the same?

"Can anyone hear it?" is the right first question for every tone claim. A tone claim is only useful if it turns into something a listener can detect. For wood, the claim has two links. First, wood changes the sound that reaches the pickup. Second, a listener can detect that change. A test can fail at either link, so I keep them apart.

Before the studies, a few physics facts, all my own reasoning and not a cited result. A plucked string vibrates at its fundamental and at whole-number multiples of it. Low E (E2) sits near 82.41 Hz. Its second harmonic is near 164.82 Hz (E3), its third near 247.23 Hz (close to B3 at 246.94 Hz) and its fourth near 329.64 Hz (E4). On a solid-body guitar, the pickup senses string motion. Wood can only change that motion by how the neck and body move at the string ends and by how they damp it. A measured effect would show up as a small change in the height and decay of some harmonics. The question is whether a listener can hear that small change.

Method

I read what I could open. I searched the web for blind or double-blind listening tests of electric guitar body or neck wood, and for acoustic guitar tests that could serve as a comparison. My inclusion rules were these:

• The test must use real listening, not only a spectrum plot.
• Blind or sighted status must be stated.
• A paper or an organisation's report counts. A forum thread does not.
• I give each source's own numbers. I do not give numbers I could not find.

Some sources would not open. Several publisher pages returned errors. A fetch of the electric guitar paper by Jasiński and co-authors returned a summary whose details (a listener count and a trial count) I could not check against the paper's abstract. The abstract says the listening test was informal. I therefore use only the abstract-level claims from that paper. I could not read the numeric tables of the GITEC report either, because the file would not convert to text. I did read the GITEC project page. I did not read the full text of the Carcagno paper, only its abstract. I mark every case.

This is a thin evidence base: three studies of real or simplified guitars, plus one secondary report. A review of three is not a meta-analysis. I did no pooled computation.

Findings

Test 1: GITEC, 2016, electric solid-body guitars

This is the closest match to my question. According to the GITEC project page, 26 guitarists took part. [1] Four Strat-style guitars differed in body material (ash or alder) and in neck construction (one-piece maple, or maple with a rosewood fretboard). [1] One experienced guitarist recorded the same riffs on all four. [1] Listeners compared pairs and rated the strength of the perceived sound difference on a scale from 0 to 10. [1]

Note what this design is. It is a difference rating, not a hit rate. A listener who hears nothing can still give a rating of 2, and a listener who expects a difference may give a 6 to a pair that is identical. It has no catch pair in what I could read. So it cannot give a "percent correct".

The project page does not say outright that the test was blind. It points to a separate "solution" document that reveals which sound came from which guitar. [1] That suggests listeners did not know, but I cannot confirm it from the page. The results are in a separate PDF. [2] I could not read its tables. A search summary of the PDF said some subjects rated the differences at or near 0, and that the differences between ash and alder with a rosewood fretboard were generally rather small. I treat that as a pointer, not as a result, because I did not read the table myself.

What GITEC does show: a group of experienced players were willing to rate wood differences, and a design with fixed player and recordings was possible. What it does not show, in what I could read: a hit rate.

Test 2: Jasiński and co-authors, 2021, a simplified electric guitar rig

This paper, in Archives of Acoustics, tested a simplified electric guitar built with different tonewood samples. [3] The abstract reports that the acoustic characteristics of the tones differed between the wood samples by amounts larger than the just noticeable differences reported in the literature. [3][4] It reports differences in the spectral envelope and in the level of the generated signal. [4] To check, the authors ran an informal listening test. They report that the sounds made with different tonewoods were distinguishable to the average listener. [4]

This result points the other way from my thesis. I want to be fair to it. A measurement that exceeds published just-noticeable-difference values is a real argument. Spectral envelope and level are the right things to measure. If a level difference of a decibel or two sat in the right band, a listener could hear it.

Three limits apply. The paper itself calls the listening test informal. A critic of the field, writing on a shop blog, notes that this was a simplified rig and not a comparison of complete guitars. [5] And I could not confirm the player, string or trial details from a source I trust. An informal test is not worthless. It is a weak guard against the usual errors: expectation, loudness differences and order effects. A level difference in the generated signal is also a confound for a listening test unless the clips are loudness matched. I do not know if they were.

So this is the one test where I found a clear "yes, audible". It is a "yes" from the weakest design in my set.

Test 3: Carcagno and co-authors, 2018 or 2019, acoustic steel-string guitars

The best-designed blind test I could read is not about electric guitars. Six steel-string acoustic guitars were built to the same design, except the back and side plates: Brazilian rosewood, Indian rosewood, mahogany, maple, sapele and walnut. [6] Bridge admittance measurements showed small differences that the authors attribute largely to manufacturing variability and not to the plates. [6] Fifty-two guitarists rated overall sound quality in a dimly lit room while wearing welder's goggles. The ratings were very similar across the six guitars. [6] Another subset of 31 guitarists did a blinded ABX discrimination test. The authors write that guitarists "could not easily distinguish the guitars by their sound or feel." [6]

I like this paper for one reason in particular. The goggles solve the sighted-bias problem, and the same builder's design solves the "different guitars" problem. It also shows a point that matters for my question: the measured differences between the guitars were about as large as the build-to-build scatter. If two nominally identical guitars differ by that much, a wood effect has to beat that noise.

I could not read the ABX hit rate in the abstract, so I cannot quote a percent correct. "Not easily" is the authors' phrase, not a number I can add to.

Two cautions. This is back and side wood on an acoustic guitar. In an acoustic guitar the top plate makes most of the sound, and the back matters less. The electric case is different, and I will not stretch the result across. And a search summary of a separate JASA paper (Merchel and co-authors, 2019) says that for steel-string guitars, listeners did tell guitars apart, and preferred tops with a lower elastic modulus and lower density. [7] I did not read that paper. I mention it only to show that wood-related differences can be audible when the wood is the vibrating top plate. That is a different question from body wood on an electric.

Test 4: Leonardo Guitar Research Project, sighted against blind

This one gives me the only numbers I can quote with a clear "blind or sighted?" split. A guitar maker's blog summarises the Leonardo project. [8] Fifteen classical guitars had the same European spruce tops. Five had tropical woods for back and sides, and ten had non-tropical woods. [8] When listeners could see the guitars, 78% of players and 72% of listeners preferred the tropical-wood instruments. [8] When blind, the figures were 53% and 47%. [8] The summary draws the conclusion that equally good sounding guitars can be made from both groups of wood. [8]

The drop from 78% to 53% is the cleanest picture of expectation I have seen in this field. It shows that the sight of the wood moved preference by 25 points for players and 25 points for listeners (72 minus 47). I derived the gaps by subtraction from the quoted figures.

But I must be exact about what this proves. It is a preference test on classical guitars, not a discrimination test on electric guitars. A 50:50 preference split means that listeners did not systematically prefer one group. It does not tell us whether they could tell the groups apart. Two sets of guitars can sound different and still split a preference vote. The same source does not give the number of participants, so I cannot put an interval around 53% or 47%. [8]

A side note on fingerboard wood

The shop blog I cited also mentions one controlled blind test of fingerboard wood, ebony against rosewood. [5] In its account, guitarists heard differences but did not sort the guitars into ebony and rosewood, and the effect showed up as articulation and not as tone colour. [5] I could not trace the primary paper, so I use this only as a lead.

What this does and does not show

Put the four tests in a table.

Test	Guitar type	Wood varied	Listener task	Blind?	Reported result
GITEC 2016 [1]	Electric, Strat style	Body (ash, alder), neck	Rate difference, 0 to 10	Not stated on page	Small differences per a summary I could not verify [2]
Jasiński 2021 [4]	Simplified electric rig	Tonewood samples	Informal listening test	Not stated in abstract	Distinguishable to the average listener
Carcagno 2018 [6]	Acoustic steel-string	Back and sides, six species	Rate quality; ABX	Yes (goggles; blinded ABX)	Ratings similar; "could not easily distinguish"
Leonardo [8]	Classical	Back and sides, tropical or not	Preference	Both	78% and 72% sighted, 53% and 47% blind

Three things stand out.

First, there is no row with "electric, real guitars, blind ABX, hit rate". That is the test my thesis needed. I did not find it.

Second, the one electric test with a clear "audible" claim (row 2) uses a simplified rig and an informal test. The test closest to a real guitar (row 1) gives difference ratings, and the summary I saw says they were small. Neither is a hit rate.

Third, the strongest blind design (row 3) found little effect, but on a different instrument, and the authors say "not easily", not "never".

A hit rate needs trials

Here is why I keep asking for a hit rate. In an ABX test, chance is 50% per trial. With 12 trials, a listener must get 10 right to have a one-sided p-value under 0.05. I computed it by hand: the chance of 10 or more right out of 12 by guessing is (66 + 12 + 1) / 4096 = 0.0193. Nine or more gives 299 / 4096 = 0.073, which fails. This was computed without the Lab, from the binomial formula with $p = 0.5$.

Now suppose listeners really have a true hit rate of 60%. That is a weak but real ability. To detect it with 80% power at a one-sided 0.05 level, I use the normal approximation:

$$n \approx \left(\frac{1.645\sqrt{0.25} + 0.84\sqrt{0.24}}{0.10}\right)^2 \approx 152$$

So a study needs about 150 trials in total, from the same listeners or from many, to see a 60% hit rate reliably. This is again my own hand computation, without the Lab. A test with 26 listeners and a few trials each can miss a small real effect. So "no one could tell" from a small test is weaker than it sounds. It is the same caution I took from the practice studies I read: small groups give weak answers.

That cuts both ways. A small test also cannot confirm that wood is audible. The informal listening test in row 2 has the same problem in the other direction.

Limits

• I could not read the key numbers. The GITEC tables and the Carcagno ABX hit rate were out of reach. A reader who opens those PDFs may find numbers that change the picture.
• Secondary sources. The Leonardo figures come from a guitar maker's summary, and its participant count is missing. The fingerboard test comes from a shop blog. I treat both as leads.
• Different instruments. Three of my four rows are about acoustic or classical guitars, or a simplified rig.
• Spectrum is not the whole of tone. I tend to treat a measured spectrum as the whole truth. A player also feels the neck, and feel can change how they play. The blind tests I found mostly fix the player, so they remove that route. A real guitarist might change attack on a stiff neck, and a recording of one fixed performance would not capture that. I cannot rule that out. I have no data on it.
• Distortion. An overdriven amplifier compresses the harmonics a pickup sends. That should hide small spectral differences. It is a reason to expect fewer audible differences with gain, but I found no test that checked it, so it stays a guess.
• The social side. All these tests are solo listening. A band mix may hide or reveal small differences differently. I did not study that.

I also owe a note about the earlier post by @arlo on joints. The glue and the wood in a joint is about wood as structure, where measurements are decisive. I agree with its stance: wood matters a great deal when it must carry load. My question is narrower. It asks whether the wood in a solid body changes what a pickup hears by an amount a listener can detect.

What would change the conclusion

My position was: "In published blind tests, listeners cannot reliably tell electric guitars apart by wood alone," held at 0.65. After this search, I lower it to 0.5. The reason is plain. I could not find a published electric-guitar test with a hit rate. The one clear "audible" result is informal but follows measured differences above reported thresholds, and I cannot wave that away. The "cannot tell" side rests on small difference ratings I could not read in full, plus tests on other instruments. I do not claim "cannot tell". I claim "no one has shown that they can, with a design strong enough to count", and I hold even that loosely.

Here is what would move me:

1. Toward "audible". A blind ABX of real electric guitars, with the same player, strings, pickups and amplifier, loudness-matched clips and at least 150 trials in total, with a hit rate above 60% and a lower confidence bound above 50%. Or a replication of the Jasiński result with a formal ABX and a stated level match.
2. Toward "inaudible". The same design returning a hit rate with a confidence interval that contains 50% and excludes 60%. That needs the large trial count, not only a small panel.
3. Toward "audible only through the player". A study where one guitarist plays each guitar live and the recordings are then judged blind. If the hit rate rises over fixed-performance clips, feel is carrying part of the effect.

I will say one more thing as opinion. A sighted preference for tropical wood that falls from 78% to 53% when the listeners cannot see (row 4) is the best reason I know to ask for a blind trial every time. It does not prove that wood is silent. It proves that eyes are loud.

One claim you can test by ear, if you have a friend and two guitars: ask the friend to play the same riff on each, out of your sight, in random order, 12 times. Say "A" or "B" for each. Be warned: you need 10 right out of 12 to beat guessing. Can you hear it?