If you look online for advice on piano sight-reading skills, you will find plenty of resources. In the vast majority of methods, the primary focus is based around keeping the rhythm going at any cost, even if it involves a great deal of educated guesswork regarding some notes (or even omitting things altogether). This is for very good reason. If someone asks you to accompany them with an unfamiliar score, you're not going to be of any use if you regularly stop in response to any kind of doubt. You have to keep going ahead to stay with them. For that reason "never stop" has become a golden rule for practising sight reading.
It's truly unquestionable that any good sight-reader will have to learn to "fake" their way around some tight corners, rather than freeze in a self-defeating hope for perfection. So why am I writing a post to challenge the almost universal notion that this is the definitive mantra for all practise of sight-reading skills? Quite simply, it's because there's also a minimum level of accuracy required.
How much accuracy counts as enough?
How much accuracy counts as enough?
Let's suppose that two pianists both manage to stay in time, while sight-reading a modestly difficult accompaniment for a singer. One fakes around 5 percent of the written notes- slightly rearranging the distribution of chords and skipping a few inner notes, while keeping a clear harmonic skeleton. Not a problem. It sounds fine. The other one manages around 70 percent accuracy, with failures including a missed change of key signature and a whole phrase where one hand is played in the wrong clef. Is it still fine as long as they remember not to stop?
Okay, I've plucked those theoretical percentages out of thin air. However, as a very rough working estimate I'd say that anything short of 90 percent accuracy is usually going to be enough to cause at least some awkwardness. That's open to debate, but let's say we were to fall back to as little as 70 percent accuracy. Could that sound fine? I don't want to try to define an exact boundary, so much as encourage you to ask your own serious questions about what would be feasible. However, if we get to a 50/50 success rate and you still think it might sound okay then you must be kidding. Not even Les Dawson fell to that.
https://youtu.be/DtK7Hg7PnnQ?t=407
In another field, 70 percent wouldn't actually be a notably poor success rate. In a maths exam you'd get a strong pass for that, if not necessarily the absolute top grade. However, let's say I offer 70 percent accuracy to a musician for whom I have been paid to accompany a diploma exam. After the exam, not only do I remind them that I did I not stop, but I also "boast" about how proud I was to have played roughly twice as many right notes as wrong ones. Will they book me again? I doubt it.
(Actually, as a brief aside, I'll just mention modern atonal music as the exception to this. I have actually fallen well below 50 percent accuracy while accompanying the ludicrously difficult final section of Hindemith's tuba sonata. In that case, I heard afterwards that the examiners felt it was the best they'd heard anyone manage with it before, so it's far from an ordinary case. I also recall agreeing to accompany a modern piece for a friend at very short notice, in which I probably managed to play around half the notes correctly in one hand, while playing literally anything in the other hand. While I doubt that the composer of either work would have been terribly appreciative of what I managed to muster up, you *can* sometimes get away with murder in atonal music. Apply a similarly poor level of accuracy to classical or baroque repertoire, however, and it's a very different story).
My own sight-reading captured- warts and all
Let's link those hypothetical figures more to reality, via a couple of examples of myself sight-reading. I can assure you that the first two are genuine first attempts, which I hope will be evidenced by the handful of errors in the Bach, plus the rather more numerous ones in the Rachmaninoff. I also deliberately avoided any mental reading through (save for a very brief glimpse, solely to choose each excerpt), in order to go in as unprepared as possible. The third film illustrates the same Rachmaninoff excerpt as the second, but after around four or five minutes of quick study practise (in which all rhythmic pressure was removed, in order to feel around the notes, before a return to performance style playing).
I don't think it would be unreasonable to label myself as a pretty good sight-reader, but these neither represent elite level sight-reading nor even some of my own most focused reading. I made a couple of silly lapses in the Bach, although I think it only strengthens my arguments- if I connect them to an airing of "dirty laundry", rather than select some of my best work. Now, there are two notable lapses of accuracy- particularly at around 48 seconds, where I should have planned ahead. Instead I had to skip a couple of right hand notes, while the left carried the rhythm. In the last line I assumed an F sharp would resolve to up a G, in spite of the fact it doesn't- providing an excellent reminder of how hugely overrated prediction is. There's a really good way to avoid being fooled by the fact that composers don't always do the most obvious thing- which is to actually read the notes they wrote down, rather than assume what might be likely! I also rushed the last couple of notes. While careless, this was a smaller issue, given that they are just a slight addition after the main cadence.
Anyway, my point is that, even with these issues, the accuracy level was far above 90%. The rule of never stopping for uncertainty/mistakes did not directly produce any of the things that were genuinely good about my read-through. They merely saved those things which were not so good, from turning into outright disaster.
For the second film I deliberately chose something where I expected to have a notably lower hit rate, for illustration purposes. At an estimate, I'd say I'm probably above 80% accuracy there, but perhaps not too much more. If you compare to the version I did after some brief practise, you'll realise that while it's not a terrible approximation, the number of splashed/guessed chords is more than enough to show up to someone who knows how they should really sound. The fact it sort of works is not a bad argument for that "golden rule" of just keeping the rhythm going at any cost. Had I not kept going onward, it would only have been far worse still. However, let's think just how much accurate processing was necessary for even this dodgy version. Had I only read the lower chords at all and played random notes above (which admittedly happened fairly literally on a couple of occasions) I'd still be having to process a large amount of information correctly. Dwelling narrowly on the need to play past errors does nothing to shed any light on the majority of notes which I was able to both read and execute correctly (not to mention how a truly elite sight-reader still could certainly do a far more accurate job again).
Why top sight-readers often give the least helpful advice to amateurs
https://youtu.be/DtK7Hg7PnnQ?t=407
In another field, 70 percent wouldn't actually be a notably poor success rate. In a maths exam you'd get a strong pass for that, if not necessarily the absolute top grade. However, let's say I offer 70 percent accuracy to a musician for whom I have been paid to accompany a diploma exam. After the exam, not only do I remind them that I did I not stop, but I also "boast" about how proud I was to have played roughly twice as many right notes as wrong ones. Will they book me again? I doubt it.
(Actually, as a brief aside, I'll just mention modern atonal music as the exception to this. I have actually fallen well below 50 percent accuracy while accompanying the ludicrously difficult final section of Hindemith's tuba sonata. In that case, I heard afterwards that the examiners felt it was the best they'd heard anyone manage with it before, so it's far from an ordinary case. I also recall agreeing to accompany a modern piece for a friend at very short notice, in which I probably managed to play around half the notes correctly in one hand, while playing literally anything in the other hand. While I doubt that the composer of either work would have been terribly appreciative of what I managed to muster up, you *can* sometimes get away with murder in atonal music. Apply a similarly poor level of accuracy to classical or baroque repertoire, however, and it's a very different story).
My own sight-reading captured- warts and all
Let's link those hypothetical figures more to reality, via a couple of examples of myself sight-reading. I can assure you that the first two are genuine first attempts, which I hope will be evidenced by the handful of errors in the Bach, plus the rather more numerous ones in the Rachmaninoff. I also deliberately avoided any mental reading through (save for a very brief glimpse, solely to choose each excerpt), in order to go in as unprepared as possible. The third film illustrates the same Rachmaninoff excerpt as the second, but after around four or five minutes of quick study practise (in which all rhythmic pressure was removed, in order to feel around the notes, before a return to performance style playing).
I don't think it would be unreasonable to label myself as a pretty good sight-reader, but these neither represent elite level sight-reading nor even some of my own most focused reading. I made a couple of silly lapses in the Bach, although I think it only strengthens my arguments- if I connect them to an airing of "dirty laundry", rather than select some of my best work. Now, there are two notable lapses of accuracy- particularly at around 48 seconds, where I should have planned ahead. Instead I had to skip a couple of right hand notes, while the left carried the rhythm. In the last line I assumed an F sharp would resolve to up a G, in spite of the fact it doesn't- providing an excellent reminder of how hugely overrated prediction is. There's a really good way to avoid being fooled by the fact that composers don't always do the most obvious thing- which is to actually read the notes they wrote down, rather than assume what might be likely! I also rushed the last couple of notes. While careless, this was a smaller issue, given that they are just a slight addition after the main cadence.
Anyway, my point is that, even with these issues, the accuracy level was far above 90%. The rule of never stopping for uncertainty/mistakes did not directly produce any of the things that were genuinely good about my read-through. They merely saved those things which were not so good, from turning into outright disaster.
For the second film I deliberately chose something where I expected to have a notably lower hit rate, for illustration purposes. At an estimate, I'd say I'm probably above 80% accuracy there, but perhaps not too much more. If you compare to the version I did after some brief practise, you'll realise that while it's not a terrible approximation, the number of splashed/guessed chords is more than enough to show up to someone who knows how they should really sound. The fact it sort of works is not a bad argument for that "golden rule" of just keeping the rhythm going at any cost. Had I not kept going onward, it would only have been far worse still. However, let's think just how much accurate processing was necessary for even this dodgy version. Had I only read the lower chords at all and played random notes above (which admittedly happened fairly literally on a couple of occasions) I'd still be having to process a large amount of information correctly. Dwelling narrowly on the need to play past errors does nothing to shed any light on the majority of notes which I was able to both read and execute correctly (not to mention how a truly elite sight-reader still could certainly do a far more accurate job again).
Why top sight-readers often give the least helpful advice to amateurs
The problem with the advice given by most good sight-readers is that an evolution-like process of natural selection has taken place. If they hadn't got the right things down you wouldn't get to hear their advice. Instead, they'd be out looking for a new job. As part of that, if you don't learn not to stop at the first sign of uncertainty, you will be weeded out fast. Worthy accompanists know that once you're in performance conditions, you must continue at any cost. Their livelihood literally depends on sticking to this rule every day. The problem is that they first had to spend years developing the background skills for a minimum threshold of accuracy. This is more of a long and complex slow grind, that couldn't even begin to be summarised via a catchy soundbite. Perhaps this is why we hear phrases like "never stop" so much, yet scarcely anything about how to cement deeper foundations? Anyway, if the "cost" of pressing on literally proved to be as severe as ignoring the key signature for a whole section, they would not get paid work. Easy as it is to talk about the importance of playing past mistakes, it's a lot harder to explain how to be in a position to keep errors down to an acceptable level.
Moving away from a professional standard, now- when an amateur does not have the same core skills, there's a point where no amount of willingness to press on via guesswork can compensate for that missing backbone. Good sight-readers might be very good at creatively coping with potential emergencies, but they are also very good at having as few emergencies as possible. We need to think far more about the origins of how they became able to process notes with sufficient accuracy.
It's based on this that I have come to feel very strongly that a lot of weaker sight-readers are being given completely the wrong advice for building the foundations they are most urgently in need of. If we say you can only train sight-reading under performance conditions (where the rhythm must go onward at all times and you are never allowed to go back over anything), it may do more harm than good- simply perpetuating the feeling of being a terrible reader who cannot cope under pressure.
The problem with using pressured testing as a basis for learning
I'd suggest that it's irrational to assume that test conditions are either the only way to learn, or even the most important part of overall learning, in just about any field. Would someone expect to learn to drive by simply taking driving tests over and over? Or to learn trigonometry by sitting maths exams over and over without any training?
It's based on this that I have come to feel very strongly that a lot of weaker sight-readers are being given completely the wrong advice for building the foundations they are most urgently in need of. If we say you can only train sight-reading under performance conditions (where the rhythm must go onward at all times and you are never allowed to go back over anything), it may do more harm than good- simply perpetuating the feeling of being a terrible reader who cannot cope under pressure.
The problem with using pressured testing as a basis for learning
I'd suggest that it's irrational to assume that test conditions are either the only way to learn, or even the most important part of overall learning, in just about any field. Would someone expect to learn to drive by simply taking driving tests over and over? Or to learn trigonometry by sitting maths exams over and over without any training?
Testing skills under pressure is certainly important, some of the time, but it's not where first foundations are built. I remember hearing the conductor/pianist Andre Previn speak of learning his sight-reading skills by spending a lot of time doing pressured sight-reading of duets, with a teacher (who would carry on if he was floundering and expect him to catch up). I believe this can easily mislead though. Such stories are appealing, as they present a romanticised notion of a "hero" being resilient enough to endure challenge and come out stronger for it. However, I strongly believe that he already had excellent reading skills. It's a make or break scenario, in which you don't automatically become better unless the main foundation has largely been laid. The situation would chiefly test how his existing reading skills fared under greater pressure. But it was doubtless contrasted against plenty of less pressured reading, in which accuracy was the main goal. This is where the core foundations are consolidated, before being pushed to the limit under performance conditions.
Quick-study as the gauge of true reading skill
Quick-study as the gauge of true reading skill
Stepping back from that pressured type of sight-reading, I believe the most important foundations are best developed through what I call "quick-study"- the missing link between performance style sight-reading and slower long term projects. This involves working under more moderate pressure rather than under literal performance conditions. It comes with expectation that errors and uncertainties should be fully corrected at source, rather than done badly once (in strict time) and then left behind forever. In short, you look to understand exactly what the score means and then expect to translate it into practical reality- *not* by imprinting it slowly into habitual muscle memory, but by playing to a clear and precise intention. This is the same fundamental skill as that for your baseline of accuracy in sight-reading- to read well enough to get a clear mental image of what you want (and to have the physical skill to realise that intention immediately).
When learning repertoire over months, it should actually be very little different. You shouldn't start by playing long sections badly over and over and assume they'll eventually become right. However, neither should learning be based on reading details one at a time and spending days before understanding how they fit together. Efficient use of practise is based on having a clear picture of a group of notes (be it one bar or a whole phrase) and then trying to turn that thought into reality as directly as possible. If you cannot command a clear result when given a little planning time, you cannot realistically expect to sight-read well under pressure either.
What can you accomplish in half an hour?
What can you accomplish in half an hour?
As a thought experiment, let's say that a pianist is given a fairly easy score and told to learn to play it well within a month. If they can, it tells us nothing much either way about how likely they are to be a good sight-reader. Alternatively, if they couldn't do it well by then, it's just plain obvious they'd have done even worse at first sight. But what if we gave them just half an hour? This is far more interesting, because it neither imposes the sheer pressure of performance sight-reading, nor allows the luxury of building things slowly into habit. This makes for a very good indicator of general reading skill. It's just a simple way to check you can both read something correctly and act upon it.
For those who could meet the challenge of playing well after half an hour, it doesn't necessarily mean that they will also be good at first sight. Even if they can figure things out this efficiently, it's still possible to be insufficiently used to either to processing literally immediately, or playing on through uncertainty. The classic advice would likely be appropriate for them. It might well be time to start to focus more on letting go of the need for one hundred percent accuracy and loosen up into some educated guesses.
For those who could meet the challenge of playing well after half an hour, it doesn't necessarily mean that they will also be good at first sight. Even if they can figure things out this efficiently, it's still possible to be insufficiently used to either to processing literally immediately, or playing on through uncertainty. The classic advice would likely be appropriate for them. It might well be time to start to focus more on letting go of the need for one hundred percent accuracy and loosen up into some educated guesses.
However, what about those players who could not build a convincing performance of an easy piece, with half an hour to throw at it? Clearly the results would again be substantially worse still at first sight- no matter how willing they are to fudge details in favour of trying to keep time! If you're not a good sight-reader (and want to know how to improve), the very first thing you should look at is whether you have this crucial foundation. If half an hour isn't enough to master some easier repertoire, it shows there is simply not enough fluency with translating written instructions in general, never mind under pressure. Although quick-study skills don't fully guarantee good sight-reading skills, it would be fair to say that a lack of quick-study skills will always be complemented by even worse sight-reading skill.
The background ability to translate a score into correct execution is not something that should be developed solely under intense pressure, but neither is it something where you should need days or even weeks to figure something out. We seem to have become lost in a false dichotomy- where either you learn pieces very slowly over an extended period of time, or you have a single make or break shot at sight-reading and then have to move on. It is highly artificial to have such extremes without middle ground, and it does not serve development of the best reading skills. To say that you should never go back and make corrections (because "it's no longer sight-reading" second time around) is simply missing how learning works.
Getting things correct RIGHT NOW- the truest testament to good reading
In quick-study you should not keep ploughing on past uncertainty until you reach the end. You're looking to build a foundation of accuracy (both for the piece being studied and, by extension, for more general reading skill). You should imagine exactly what you intend to do for a small segment, with ample thinking time, and then check for the accuracy after execution. This could be as little as a bar or so (although it is usually best to add the first note of the bar after, in order to avoid dividing bars into separate entities) and then you try to execute the plan. If it goes wrong, you plan once again and then you try again. If it still doesn't work, you must now stop and have a big rethink. It's probably necessary to reduce it to something smaller until you have found a manageable sized unit that you can produce confidently on command. You're looking for the things you can get right with minimal fuss. You should never just "have a go" and hope for a fluke. Guessing is not part of this approach. Done well, most things should be right first time and nothing should ever take more than three attempts. If it's not working, you should find a smaller musical unit or even go back to easier repertoire, if need be. As you get better, the mental preparation becomes ever more efficient. You won't always need to start from smaller units. Actually, I often begin as normal sight-reading, but then stop shortly after anything I wasn't happy with. The idea of working with small units can then be used as something to fall back on for correcting things- in order to get to the heart of the issue. You can also practise trying to take just brief glimpses of a unit, before immediately looking away and trying to execute it perfectly from that mere glimpse. As you become sharper at this, the same processing skill feeds very directly into your pressured sight-reading.
Going back to get things right on a second execution will do far more for your core sight-reading skills than just doing it incorrectly once alone and moving on. Finishing with a correct version means we are meaningfully connecting every detail of the score to its execution, rather than to wild guesses which were fuelled by panic. There are lots of ways to go wrong (including guessing, incorrect reading, moving the hands without adequate control etc) but very few ways to get everything right. What kind of delusional optimism would lead us to imagine that the ability to work accurately is best developed with an exclusive routine of guessing things under pressure and then never looking back? This trains willingness not to stop, but nothing else at all. To further the ability to read correctly and then instantly do things correctly, we should at least expect to get things right on a second (less pressured) attempt.
Getting things correct RIGHT NOW- the truest testament to good reading
Although I'm generally a good sight-reader, I used to struggle with composers like Bach and Mozart- where everything is highly exposed. I went through a period where I deliberately took a much more careful approach to reading their music. These can be some of the hardest to sightread well, because everything is so exposed that details really do matter. Rather than plough on ahead through errors, I'd now start to slow down a little for the most dense passages, to find a pace where I could process everything in advance rather than have to guess. I wouldn't just stop dead in the middle of ideas, but I'd happily stretch the tempo out- thus making more time, without ever destroying the rhythmic flow. If something went badly I might stop and think about that bar before correcting it and only then go on after resolving uncertainty. I wasn't exactly "practising" these pieces in the full way I would if learning them, but neither was I settling for any old rubbish- just because it's "only" sight-reading and "everyone knows" that you're not meant to stop.
In spite of breaking almost all the rules, this approach made me far more accomplished back in pressured situations. I got used to allowing fewer holes in the basic processing. I wasn't allowed to just skim over problems in my reading, but had to face them and solve them. In the end, I found myself needing fewer guesses in general. When I have to accompany another musician, I can still snap into performance style sight-reading at any time I need to. However, by having tilted the balance towards accuracy rather than always time pressure, I built a far stronger foundation in general.
In quick-study you should not keep ploughing on past uncertainty until you reach the end. You're looking to build a foundation of accuracy (both for the piece being studied and, by extension, for more general reading skill). You should imagine exactly what you intend to do for a small segment, with ample thinking time, and then check for the accuracy after execution. This could be as little as a bar or so (although it is usually best to add the first note of the bar after, in order to avoid dividing bars into separate entities) and then you try to execute the plan. If it goes wrong, you plan once again and then you try again. If it still doesn't work, you must now stop and have a big rethink. It's probably necessary to reduce it to something smaller until you have found a manageable sized unit that you can produce confidently on command. You're looking for the things you can get right with minimal fuss. You should never just "have a go" and hope for a fluke. Guessing is not part of this approach. Done well, most things should be right first time and nothing should ever take more than three attempts. If it's not working, you should find a smaller musical unit or even go back to easier repertoire, if need be. As you get better, the mental preparation becomes ever more efficient. You won't always need to start from smaller units. Actually, I often begin as normal sight-reading, but then stop shortly after anything I wasn't happy with. The idea of working with small units can then be used as something to fall back on for correcting things- in order to get to the heart of the issue. You can also practise trying to take just brief glimpses of a unit, before immediately looking away and trying to execute it perfectly from that mere glimpse. As you become sharper at this, the same processing skill feeds very directly into your pressured sight-reading.
When I went back to practise the Rachmaninoff, I just played the chords slowly and in free time, confirming every single note. Rhythm is essential to a lot of practise but, particularly in chords, there can be great value in simply making room to stay with each chord. No, rhythm shouldn't always come first (as is often claimed). Merely sometimes. At times you're better off stopping to clarify notes, before immediately bringing it back in. I already understood the rhythms but I didn't want to leave a single chord on a misunderstanding or on a guess. Admittedly, the 2nd execution is far from electrifying musically and could have been sped up plenty. My intent was not to show an ideal final performance however, but to show I could take command over the basic reading issues at once, rather than continue to approximate. Had I pressured myself into spirited guesses a second time over, I could have scrambled my perceptions. People sometimes say "Practise like you would wish to perform". Sometimes you should indeed, but I was much better off "practising like I would wish to practise". I just calmly built a link between reading each chord and then feeling/confirming the match between reading and result. Sometimes, I will even linger on the surface of a chord before playing, to confirm every last note before playing it. Then I linger once again after playing, to confirm a second time. This allows me to keep checking both that the reading is flawless and that the activity always matches what I see. After all, if I couldn't first take command of accuracy in freer time, how could I expect to be ready to process similar chords in future pressure situations?
Foundations before instincts
Think again about how we develop ability to work under time pressure in other fields. When a child first begins arithmetic, we don't encourage them to take wild guesses, under the assumption that this will later make them perform best at speed. No, it will fill their head with confusion. We get them to count on their fingers until they reliably know the sums on a more instinctive level.
This is not remotely unusual in human endeavour. If you want to learn to do something well under time pressure, first you must actually be able to do it well at all. We routinely apply this to learning a specific piece (eg starting slow for accuracy before speeding up) but we should also appreciate how this relates to the broader skill of reading something new and knowing exactly what it means.
The main skill behind sight-reading is not truly "different" to that involved in quick study, except in terms of having less thinking time. If I look at a small enough segment of music and imagine it in full detail, I should then be ready to execute it straight away. The core of sight-reading comes from the very same place. You understand it and then you do it. Yes, we also need an emergency plan B, to scrape through those situations where there is a more than we can fully be sure of. Willingness to keep going when uncertain is largely just something you can either do or you can't, however. Yes, you must practise it often enough to be sure you can indeed do it. The problem is that there's no further summit to be reached here. Once you have that tool, further progress is almost chiefly about building on the deeper core.
Effective sight-reading skill is built by by training two opposite personalities side by side- one that demands absolute precision/excellence and another that is happy to occasionally scrape by on any old crap that stays in rhythm. It's only a balanced amalgam of the two that defines a truly great sight-reader. Think of the approximation aspect of sight-reading as being a bit like keeping a cupboard stocked up with ancient cans of cheap tinned spaghetti. It's always good to know the option is there, for a time when you have nowhere else to turn. But I'd far sooner take something fresh from the fridge.
This is not remotely unusual in human endeavour. If you want to learn to do something well under time pressure, first you must actually be able to do it well at all. We routinely apply this to learning a specific piece (eg starting slow for accuracy before speeding up) but we should also appreciate how this relates to the broader skill of reading something new and knowing exactly what it means.
The main skill behind sight-reading is not truly "different" to that involved in quick study, except in terms of having less thinking time. If I look at a small enough segment of music and imagine it in full detail, I should then be ready to execute it straight away. The core of sight-reading comes from the very same place. You understand it and then you do it. Yes, we also need an emergency plan B, to scrape through those situations where there is a more than we can fully be sure of. Willingness to keep going when uncertain is largely just something you can either do or you can't, however. Yes, you must practise it often enough to be sure you can indeed do it. The problem is that there's no further summit to be reached here. Once you have that tool, further progress is almost chiefly about building on the deeper core.
Effective sight-reading skill is built by by training two opposite personalities side by side- one that demands absolute precision/excellence and another that is happy to occasionally scrape by on any old crap that stays in rhythm. It's only a balanced amalgam of the two that defines a truly great sight-reader. Think of the approximation aspect of sight-reading as being a bit like keeping a cupboard stocked up with ancient cans of cheap tinned spaghetti. It's always good to know the option is there, for a time when you have nowhere else to turn. But I'd far sooner take something fresh from the fridge.
