How Can We Know What Works? – Review

Pages: 16-27

In the Notes on S.I. 87/1 Wolf Wagner presented a study under the title “How can we know what works?”. He later specified the question, reducing it in scope, to: “Do photographs show Rolfing changes in such a way that Rolfers can identify them and distinguish them from changes that have other causes?”

The study was set up in its first part, which will be discussed here, in the following way. Of 36 models, none of which had been Rolfed before, 9 had a Rolfing session (A), 9 a Reiki treatment (B), 9 a Massage (C), and 9 “Nothing” (D), which consisted in “a walk around the table”. Before and After photographs from all four sides were taken of each model, then photocopies were made and sent out to 8 other Rolfers, the “judges”. They were asked to assign the models to the respective treatment groups. The main interest lay with whether they would be able to find out which models had been Rolfed. The results are given in Table I (from Notes on S.I. 87/1). Since each of the 8 judges determined 9 models to have been Rolfed, 72 “nominations” were made. 26 times a Rolfed model was identified correctly; and incorrectly 23 times a model of the Reiki group was named Rolfed, 15 times a Massage was mistaken as Rolfing, and 9 times a member of the Nothing group was “nominated”(1).

The result is summarized in the following statements:

– Rolfers are not able to distinguish a Rolfing from a Reiki session using photographs.

– They have a slightly better than even chance to distinguish a Rolfing session from a Massage.

– They have a much better than even chance to distinguish a Rolfing session from “Nothing”. This statement can be misleading, however. With regard to the preceding two it should better be formulated as “Rolfers can fairly well determine if some kind of treatment was given or not”.

These findings are explosive; but the judgment immediately calls for specification of what is exploded, and how, and what not. Ida Rolf?s theory for one is not touched by the results of the study. Its central concept is that of “normal structure” which permits the human body to function most economically. It is flanked by the notion that the body is plastic and can therefore be changed toward normal if appropriate means are applied. These concepts do not need to be supported empirically. They derive by means of consistent reasoning from facts in the fields of physics and biology. These facts, which can be called premises for the theory of Structural Integration, have been corroborated empirically in abundance in their respective fields. No valid argument is raised against the theory by the study.

However, the conclusions and tenets of the theory are only “theoretical”, so-to-speak. They only affirm that a normal structure can function more economically than an “average” or “random” structure, and that the real structure of a real person can be changed toward normal. Specifically it doesn’t follow that a structure does change toward normal when a Rolfer puts his hands into it. The crucial formulation used above is “appropriate means”. This implies an effective technique and – even more important – a framework of evaluation which allows to determine by criteria which are observable visually whether the “means” were “appropriate”: whether a real structure was actually integrated or not. This practical aspect, whether Rolfers actually Rolf, whether Rolfers actually integrate structure in reality while the theory only states that it is possible, is not supported by the study. Very much in contrast to the theory, the practice, the concrete realization of what the theory states as a potential only, is subject to empirical confirmation. More than that, it is extremely in need of such empirical support if the practice – not the theory! – is to achieve and maintain credibility.

The practice of Rolfing claims that it fulfills the promise given by the theory at least to some noticeable degree. The study does not support that claim in any way. In fact, it disclaims it. It states, put pointedly, that Rolfers don’t Rolf, or if they Rolf that Rolfers don’t recognize it. Either interpretation is unacceptable to a Rolfer, at least the more critical one who needs some sort of empirical confirmation that he indeed does what he proclaims he does. His first impulse will be to simply ignore the study. This strategy is crude, time-honoured, and amazingly successful. But not wanting to know what he knows he can know violates his intellectual honesty. So he might turn to look for empirical studies which contradict this study and support that he indeed integrates structure.

Unfortunately, he encounters a nearly empty fields. To my knowledge, the only comparable study dates as far back as 19743. It concludes that “highly trained practitioners (of Rolfing, H.F.) are no more accurate in discriminating Rolfed from un-Rolfed bodies than totally unsophisticated undergraduates.” Naturally the author comes to believe that “the present findings raise questions about the implicit and explicit normative systems which are being utilized by practitioners of Structural Integration for … assessment of results.” He finishes with the obvious truth that “… value can never be properly assessed without development of a conceptual framework having clear observable referents.” This is paralleled by Wolf Wagner’s own conclusion – which he finds “surprising” – that “we first have to define and describe very clearly and in detail how that which we want to change should look afterwards …”

The purpose of this paper is to examine critically the questions raised by the set-up of the study and interpret its results in order to point out the problems more clearly which hamper such empirical studies designed to support the claim that “Rolfers Rolf’. It must first be admitted that any attempt for such a study is a little daring and full of traps. This is because very little theoretical groundwork has been laid, and one can’t draw on solidly defined terms, meaningful concepts understood and shared by colleagues, and established relationships and applications tested for validity and relevance. Any such study is “planted in the middle of nowhere”, so-to-speak. So it is perhaps useful to situate the question the study asks in a context. This is given in the form of a descending scale of implied premises and hypotheses, progressing from the general to the specific.

1. Structure can change.

This is an undisputed fact from physics and biology.

2. Rolfing changes structure.

Although the nature of the change in the connective tissue is not well understood, it is an uncontested fact that mechanical forces can change its shape. Gravity as well as a car accident cause the fascial network to change in a geometrical sense. It does not follow from the statement that the degree of change is proportional to the mechanical force.

3. Rolfing change is integrative.

With this statement we enter the field of Structural Integration. The notion of an integrated structure contains as its central concept that of normal structure. While it is deduced by consistent reasoning from elements in the fields of mechanics, anatomy, and physiology, the result is specific for the field of Structural Integration. One can arrive at other concepts of normalcy with different premises. But for “normal” and “integrated structure” in the Rolf sense the definitions as well as the “observable referents” must be developed within the field.

Table I

The code numbers of the 9 Rolfed models are given, assorted with their respective Rolfer. The method assigned to them by the 8 judges is indicated.

AR or A is “Advanced Rolfer”, ER “Experienced Rolfer”, BR or B “Beginning Rolfer”, T is “Teaching Rolfer”.

<img src=’’>

Perhaps one reason why this appears to be so difficult lies with a peculiar feature of the Rolfing process which is expressed by the special relationship between quantitative and qualitative change. On the material level, when the immediate change of the tissue brought about by a move is considered, this change is quantitative. A tendon is a little longer, a fascia a little broader. But also on a more complex level the change is at first perceived quantitatively: there is more space behind the trochanter, or the forearm is a little less supinated. But sooner or later – when the legs are rotated less, when the pelvis is more horizontal, when the spine is elongated, which are all observations of a quantitative nature – the structure changes qualitatively. Everything shifts around a little and arranges itself not only more straight or erect but in a different manner. The structure is not the same kind of structure anymore but something else, where all the parts relate to each other in a completely different manner and the whole behaves in a completely different way with respect to the gravity field. It is this qualitative change a Rolfer works for although he usually can’t pinpoint the moment when it happens and doesn’t always realize that in fact it has happened. Nor is it understood which combination and sequence of quantitative changes has actually brought about the qualitative step. Using a formulation of Ida Rolfs we can say that the normal pattern which was dormant, unnoticeable, not realized, and actually only believed to be there because of experience, emerges and materializes somehow sometime. The properties of this completely different normal structure have been described extensively and in detail by Ida Rolf. But it is extremely hard and uncertain to describe such properties in actual bodies. It is definitively easier to describe quantitative changes – like a leg being sraighter after a session than before – and to measure such quantitative changes is at best a technical problem. A qualitatively different system however cannot be described by comparing measurements. A whole new set of data relating differently is needed. But already clinically it is rather daring to state with confidence that e.g. a leg that was compressed before now provides lift up through the body. It is nearly impossible to date to support such a contention by “observable referents”. The situation becomes altogether unintelligible when we suppose that not only the large step from “random” to “normal” is at issue but that the process breaks down into smaller steps of such qualitative change.

This study like the earlier one cited above both come to the conclusion that before everything else development on this basic level is necessary. Otherwise all contributions made rest on very flimsy ground and are more founded in a belief system – which must be defended against all doubts vigorously of course – than on rational argument. It should have become clearer that this is probably a double task. A framework of reference, criteria which permit to analyze a structure and its degree of integrity or lack of it, must be developed along quantitative lines. With such reliable data different systems of organizing them should become possible which would allow to identify different levels of integrity.

4. Rolfers Rolf.

The question turns up sometimes and becomes acute when a client says that he doesn’t feel any differences. The Rolfer can then try to convince him – and himself by using photographs e.g. This is often successful. At other times it is not. And if he has forgotten to mark the photographs or crumple the client’s hair when doing neckwork in order to be sure not to mix up before and after pictures, the dilemma cannot be covered up.

Some comfort is provided by a frequent confusion that results from the fact that “Rolfing” is used to designate at least two very different things. In the strict sense it means “integrating structure”. In a much larger sense it stands for what Rolfers do. The confusion can be illustrated by a simple but false syllogism:

– Rolfing means Structural Integration.

– I am Rolfing because I am a Rolfer.

– Therefore I am integrating structure.

Under the circumstances a Rolfer has really only his own word to believe in that he is actually integrating structure. It is hard to remain critical toward oneself and tempting to do away with such doubts once and for all. Such a “solution” leads to a stand-still however, and the development of one’s skills and understanding which lies almost exclusively with oneself is quelled. For my personal practice I have acquiesced in considering 60% of my clients definitively better integrated. 30% don’t show sufficient change to really be called Rolfed. Perhaps 10% are actually worse off. They are those clients who come back to be fixed or who develop symptoms which I must ascribe to imbalances created. Fortunately they often break off treatment or they consider the experience worthwhile for some other gain. This “confession” is of course not worth much. But it can and sometimes does prove profitable when a little piece of knowledge finally turns up after digging into a failure and trying to understand what has happened. Such experiences could be made even more productive if a medium for examining them existed and if the little piece of knowledge gained could be communicated.

5. One Rolfing session produces visible integrative change.

The view is sometimes held that the changes in one session are too subtle to be seen. The logical objection is of course given by the question after how many sessions a change is supposed to become visible. The position seems to grossly overstate the case of Rolfing as an art. But it must be admitted that some clients prove very hard to get the process of change going. It would again be much more productive, however, to attempt to identify such difficult structures and understand why they are so difficult. Another, more “historical” and obsolete position holds that structure must become worse at first, that it must “be taken apart” first to be “put together” later again. I don’t believe that nowadays anybody would want clients to walk out of a session disintegrated. The logic of such a view is more than doubtful. It implies too much a view of the body as a machine, an apparatus of parts wired and bolted together. Of course it happens to every Rolfer now and then that a client is “taken apart” after a session and not better integrated. But this cannot really be ascribed to the nature of the process but must be accepted as a characteristic of a lacking collective and individual understanding of structure and the process of its integration.

Now and then this state of the understanding also leads to hold a structure to be worse off when it really isn’t. Classically this is the case after the “third hour” when rotations of girdles and spine are often more obvious and marked. Especially beginners tend to become fixed on symmetry which is the one clear and observable criterion available. However, not infrequently the structure is actually better so that it could be said that symmetry is worse, but balance is better.

6. Rolfing changes are visible on photographs.

An additional premise would be that “Rolfers are able to see these changes on photographs”. It is covered by 6. though, and indirectly by 4. If these hypotheses were true, it would be possible for Rolfers to identify the Rolfed models out of the whole group. Finding and defining the criteria which permitted them to be successful would permit to compare the Rolfing sessions as to how effective they were. Relating this evaluation to the procedures used by the Rolfers doing the Rolfing would finally make it possible to assess which approaches proved to be more effective. This would then allow to give an answer to the question asked in the title of the study: to know how we can find out what works and, more poignantly, to “know what works”.


It is near to impossible for a Rolfer to hold the view that photographs don’t show relevant properties of structure. Ida Rolf states that the photographs in her book “show clearly a lack of symmetry and balance” (Rolf, p. 16). They should also show symmetry and balance clearly. Rosemary Feitis in her Foreword says that “the photographs give at least a static impression of the changes resulting from Structural Integration …” (Rolf, p.12). Rolfers are encouraged to study structure using photographs, and many explain to their clients the structural changes habitually by using photographs. Declaring that photographs don’t represent structural changes would threaten dangerously the Institute’s and Rolfers’ credibility.

This is not a valid argument for photographs, of course. It rests on a belief in good faith. The case for photographs can be made more rationally, however. Since structure is a statement about the spatial relationships of the parts of the body – however defined – which are material and visible, and because photography reflects exactly such spatial relationships of material bodies, it should represent structure very accurately. Nonetheless, there often prevails an uneasiness and uncertainty when judging structure from photographs. It is sometimes argued that the third dimension is lost as well as that of time, meaning movement. The author of the study lists these restrictions, but he also states that observation of live bodies usually doesn’t result in greater insights or clear analysis. I concur with this in view of the often helpless attempts at saying something relevant about models in classes as well as of my own similar observations with clients.

Another opinion is interesting. It holds that photographs don’t show structure but only illustrate it. Since illustrations are used to throw an additional light on a description, analysis, or demonstration established in another medium, the question is legitimate what photographs are supposed to illustrate. The answer is structure, the structure of a person. The implication of the opinion is then that this structure is already described in another way, in a manner more accurate than photographs. This would have to be in words, and so the position points again exactly to the lack of such a verbal system discussed under 3.

But the main problem with using photographs stems probably from a certain confusion about structure and function. It is understandable that Rolfers would like to depict structure visually, to present a clear image of structure in the literal sense. Differences in that image would then represent directly structural differences. This is not possible, however, because structure cannot be shown in a naive sense. In a living body, structure and a functional element are always both present muddling the picture by necessity. The reality is always a structure/function whole which can be separated into its two parts only mentally, artificially. The case is made evident by the models of the Nothing group whose structure was by definition identical in the Before and After pictures. 9 times such models were believed to be Rolfed, and the basis for this judgment must lie with the fact that the shape of the bodies as shown by the photographs is really different in the two sets of pictures. These differences are exclusively functional differences.

Since we cannot eliminate function, the intention becomes to keep this function constant. Differences of form in the photographs could then be justly ascribed to structural change. In the study the clients were asked “to stand easy without any conscious pose”. The instruction contains a paradox because it calls “pose” to attention of which the client should not be aware. It is doubtful whether such an instruction can be followed at all. But if so, the result is not an absence of posing. What it does at best is removing that factors which influence and determine function – posing- from consciousness to the unconscious. And once there they can of course not be controlled anymore because they are not known. The most promising approach is the “structural point on the postural curve” (Notes on S.I. 87/1, p.31). It is not so easy to apply in practice, however, and the concept needs to be developed further. Anyway, empirically it seems necessary to examine first how constant or how variable stance is under various conditions. Only a study with several models photographed several times on consecutive days e.g. could shed light on this elementary question. Without some idea founded in reality about what the answer is, photographs will always be a hazardous undertaking if rationality is intended.

The Goals of the First Session and the Goal of Rolfing

In my role as a judge in the study every now and then a problem crept up and blocked the process of evaluation. It was in the form of the question whether I should look for better structural integrity or the “goals of the first session”. There should be no conflict of course. But if there is conflict, the overall goal of structural integration takes priority. In other words: should the “goals of the first session” lead to disintegration, they must be considered false.

The recipe organizes the basic series of ten sessions along a certain sequence. The overall goal of Structural Integration is broken down into steps leading to that end. The recipe has two major features. First and more important theoretically, it can be understood as a concretization of a general theory of the integrating process for human structure. Of that theory itself very little is known. The recipe derives exclusively from Ida Rolf?s intuitive understanding of the process. Secondly, it is the basic tool with which students are introduced to Rolfing. It provides something like a roadmap in a highly unfamiliar and exotic territory.

A host of problems besets the “recipe”. Some have to do with the fact that it exists exclusively in the oral form of tradition. Others derive from the fact mentioned above that its background, the rationale of it, is not really understood. As it is, there exist as many variations of the recipe as there are Rolfers, and these versions not infrequently contradict each other when they are formulated in an concrete and binding manner. The original however is lost, if it ever existed at all in a set way. Moreover, it cannot be reproduced from its many versions in an unequivocable way.

The recipe is situated in a field with specific dynamics. On one side, the recipe derived from an abstract theory of integrative change calls for a certain sequence of moves to fulfill the prescribed goals. The sequence may vary some within certain limits as long as the goals are reached. On the other hand, everyone’s structure has its own specific configuration and asks for a special procedure suited to the demands of that individual structure. From this point of view and narrowing the focus down to a single move the following can be stated. There probably exists one special move in every situation with every structure which is most beneficial for integrating it. The thesis presupposes a very sharply discriminating system of evaluation which does not exist. Then there is a multitude of moves which also serve to integrate structure although to a lesser degree than the “best move”. Below, another group of moves can be stated hypothetically which could be called irrelevant. They neither help integration much nor do they any harm. Below zero, so-to-speak, still another group of moves would actually disorder the body. The recipe can by nature not indicate the best move. That is neither its intention nor obligation. If it points out and leads to the moves of the second group it fulfills its duty and is immeasurably helpful. Moves of the third group can be tolerated, but if the recipe led to disorder by moves of the last group something would be wrong about the recipe, not the structure!

When one goes from the structure at hand however, looking for the best move is the natural strategy. This is conditional to a very advanced and evolved understanding of structure.

Another source of irritation with the recipe comes from a confusion about levels of intention. The following tries to structure the goals as they could be formulated for the first hour. The proposal is strictly formal and does not attempt to formulate correct goals. So the content of the suggestions should be disregarded.

0-level: the goal is structural integration, horizontalizing the pelvis, front-to-back balance, etc.

1-level: the goal is to relate the thoracic and pelvic segments better to each other and the Line.

x-level: “up the pectoralis major to … to three times down the fascia lata.”

x+1-level: “free the pectoralis attachment to the 5th rib, spread the pectoralis fascia evenly when narrow or take it medial if too wide, …, clear the pectoralis tendon and lengthen especially with internally rotated humerus, …”

The 0-level does not constitute a recipe yet. It contains all the goals that are true and an issue in every session. Horizontalizing the pelvis is such a goal for every session and so does not help to break down the overall goal into smaller steps. On the 1-level the goals of each session differ but are formulated in the most general manner. Between that and the x-level as many levels as are sensible and can be discerned can be inserted. The x-level represents about how a beginner learns the recipe. An important change can be noted between 1and x-level. On the first the goal is formulated in structural terms, essentially directing intention towards normal in a geometrical sense. The second merely indicates a course through the geography of the body. It summons the student to work at certain places. And indeed “work” is the term used most on this level. It can mean all and anything and is therefore meaningless. The student knows where to put his hands but has no idea what to do there with them. The x+1-level would add structural considerations to the geography. But it can be said without difficulty that the goals of the first session described this way would fill a book since it would have to allow for all possible configurations – if those were known in detail.

So the characteristic trait of the scale of levels descending to the concrete and specific is a switch from geometrical goals to geographical ones. Another is that the formulations become longer. The first observation can lead to state that there exists a reciprocal relationship between “truth” and “usefulness”. The goal proposed for the 1-level is probably almost always correct. It could be reasoned that for model 16 it might be a superior goal to relate the legs to the pelvis in a first session, however. There, structural limitations appear most obvious in the legs if rigidity is taken as an indication while the upper body seems relatively soft. But the 1-level is not useful for beginners. They wouldn’t know what to do to achieve the goal formulated in such a general manner. Even experienced Rolfers might get into a predicament when questionned why and how exactly what they are doing should be the most logical path toward that goal.

On the other hand, the x-level is eminently useful, but it is also certainly more often wrong. It is not very convincing that working (!) on the pectoralis major should help integration best with whatever the structure of an actual client is like. The x+1-level could remedy this unsatisfactory state of matters partly. Structural considerations on a very concrete and local level would be added to geography which should make the prescription which has a certain value more effective and rationally understandable.

But one could also start at the other end. The higher levels could be complemented by more concrete and geographically specific content. The question would here be which moves or which local changes would benefit a given structure best for coming closer to the general and abstract goal. But whichever approach is chosen, a better understanding of structure on more general and more specific levels is conditional. This points again to the essential need of developing premise 3.

The author of the study names four goals for the first session: “more length in front”, “separation of pelvis and thorax”, “more horizontality in the pelvis”, and “a more harmonious pattern of breathing”. The first impression is that of a relatively arbitrary collection of unrelated goals. A logic of the recipe can hardly be suspected from it.

“Separation of pelvis and thorax” is perhaps on the 1-level. It illustrates a peculiar feature of “loose language” in Structural Integration. Separation is not an end goal because it lacks geometrical specifity. It is rather a means to be able to relate the two segments differently in a more ordered way. Separation by itself can also lead to disintegration, which would look as more freedom or flexibility between the two but less relationship. But if the segments are too much bound up, which they often are, separation or differentiation is certainly an essential prerequisite for integration.

The usefulness of the goal is limited because it presupposes an understanding of what needs to be done to achieve it. The thorax could be mostly bound at the costal arch be it because it is too drawn in or too blown out. The shoulder girdle in back or in front might bear down on it. The lumbodorsal fascia might present the main restriction. The ribs could be bound together too much, or maybe they need to be freed from the spine. The x-level would simply cover all these stations – and all others which are known to restrict the thorax – and “work” them. It should have become evident that freeing the thorax as part of “separation” needs to be regarded with respect to “normalizing”. And this in turn depends on a concrete understanding of the structural mechanics of local areas.

The “more harmonious pattern of breathing” belongs perhaps to the 0-level but presents even more of a problem. It first implies that the pattern which is there only needs to be “more harmonious”. There is a certain conflict with the concept that Rolfing is about changing one pattern, “random”, to another one, “normal”. It is again often necessary, as a precondition, to make the existing pattern more “harmonious” if this means more ease and fluidity of movement and more freedom. But if the intention stops there, this is not Rolfing; it is a step back to before Ida Rolf. This is of course not a value statement but a technical point. The LDH can be used as an example: when it goes forward with inspiration as it so often does in “random bodies”, it is senseless to harmonize that. The LDH needs to go back with inspiration. Matters should perhaps be stated clearly where possible which means that instead of a more “harmonious pattern” the term “normal pattern” should be preferred.(5)

The two goals discussed so far can be misleading. This can be made clearer by the statement: “I have separated thorax and pelvis and harmonized the breathing pattern; therefore this client’s structure is better integrated”. It could be true – and probably it often is -, but not necessarily so. The judgment on structural integrity must be made using other terms.

“More horizontality in the pelvis” is undisputed and certainly true. It depends of course on the tilt having been diagnosed correctly. It is probably better subsumed under 0- level goals than on the 1-level.

“More length in front” is the most questionable of the four goals. Treated formally, we begin with front-to back balance which is certainly a 0-level goal and always true. From this follows that when and where the front is shorter than the back, “more length in front” is needed and correct. But when and where the back is shorter than the front the length is needed in the back. One of the author’s models (55), whose photographs are not available for print, illustrates the point somewhat ironically. The model is external, and the session seems on the brink of having gone wrong. It was saved to actually produce a good result by vigorous hamstring work for which the author is known. Length in the hamstrings is certainly length in back and not in front.

The recipe merits to be examined more and in depth. Not only is it in dire need of clarification, but certainly it also deserves such attention as the primary and extremely valuable tool it is. But it is also of practical importance because the intention of the Rolfers doing the Rolfing in such a study should influence greatly the result.

<img src=’’>
Tight internal with little depth front-to-back. Locked knees, posterior lower legs. The main constriction is around the hip joints which makes the pelvis look wider than it is. The lumbar segment is collapsed forward down, the lower back is compressed. The left shoulder and upper ribcage are markedly high and posterior. The head is displaced to the right, the lower thorax to the left, all of which indicates strong rotation. The left leg has an external aspect.

In the After picture the body is held up more from the head. The lumbar segment is less forward with tension visible in the abdomen. The pelvic tilt is less with concomitant strain in the legs. Rotation of the upper body is less marked in the front picture, with a little strain showing in the neck. Better appearance is due to the holding of the head and the weight which was left more collapsed on the shorter right leg and is now more on top of the longer left leg.

<img src=’’>
The upper body is internal, the lower external. The posteriorly tilted pelvis is shifted forward extremely, the thorax tilted posteriorly. Marked pelvic torsion with the right hip and leg drawn in. The thorax is rotated markedly with the left shoulder high.

In the After picture the posterior tilt of the pelvis is reduced. There is a little support visible about up to the iliac crest. The pelvis is less anterior, and the diagonal from the pubes, which are less up and out, to the dorsal hinge is a little longer and less slanted. Upper thorax and shoulder girdle are better balanced.

<img src=’’>
Massive, top-heavy body with hips constricted laterally. The spine is strongly internal. The body is generally tight as shown by ankles, flexed knees, elbows, and neck. The chest has an external aspect.

In the After picture the pelvis suggests a posterior tilt, helped by some holding in the lower abdomen. But the legs appear a little easier, the area of the groins seems a little freer. Back and neck appear a little longer, the chest and the shoulder girdle have relaxed. It is not sure whether all this is due to muscular relaxation, however. It can be suspected that the generally tight fascial system has released a little, too.

<img src=’’>
Tight congruent internal except for some conflict in the chest. Good sense of posture despite anterior lumbar segment, compression L5/S1. The pubes are pulled down markedly. The main limitation comes from the restriction around the hip joints. Very compressed left wrist, elbow, and TMJ.

In the After picture the release of the tensor compartment gives good length all the way up to the occiput and down the fascia lata with decompression of the knees. The lumbar segment has come back considerably, the pelvis is much more horizontal. Less drag on upper ribs results in better horizontals through the ribcage. Pelvis torsion is less obvious. The costal arch could be freer.

Judges’ Judgment

Of the 8 judges, 5 recognized 3 Rolfing sessions each, one 2, another 4, and one 5. This homogenous distribution suggests at first sight that the judges could have agreed on at least some aspects of integrated structure in the sense that they recognized those models with the clearest structural change. However, the count is so close to that which would have been expected by guessing (2,25) that the interpretation is more convincing that they were in fact pretty much guessing, which guessing may have been tinted slightly by a semi-conscious notion of structural change. This is further borne out when one searches for a pattern. This starts out promising with the “leader”, model 23, which was recognized 6 times, by all the judges except the two “Beginning Rolfers” (Table I). But already with model 35, which was recognized 5 times, 2 of the 3 “dissenting voices” came from the two probably most experienced judges who both assigned the model to the Nothing group. Except for an obvious tendency towards “high-scoring” models I cannot detect a pattern. The idea that some of the judges could have named the same set of models indicating similar and shared criteria of evaluation cannot be confirmed: of the 5 judges recognizing 3 models no two of them named the same 3.

A completely different and interesting observation can be made when the sessions of all four methods are examined as to how many “nominations” they received. Table II lists under each method the number of sessions receiving the same number of “nominations” (given by the left column). One would naturally expect a distribution somewhat along a Gauss curve. That means that most sessions of each method would receive the number of “nominations” around their average, and the farther away from average the fewer sessions would be receiving that number of “nominations”. If Table II is read from right to left, it is evident that the distribution of the Nothing group conforms to this expectation. But already with the Massage group the picture changes. There appears a “dip” around the average number of “nominations” (1,7). This tendency becomes clearer when going to the left so that for the Rolfing group we actually find no session receiving the -average number of “nominations” (3,0). The curve becomes very suggestively “two-peaked”.

The phenomenon can be explained tentatively by the following hypothesis. We suppose that from right to left the “impact of change” increases as indicated by the average for each method. This means only that differences of stance become more obvious from right to left but that we neglect the direction of change for this view. The change is certainly partly functional as proved by the Nothing group. It must also be assumed that this functional component of the “impact of change” increases towards the left, but it should be granted that a structural component adds to it although there is no proof. Then we could suspect that with rising “impact of change” the unselected collective of models splits into two groups. They are for now called “responders” and “non-responders” for obvious reasons.

This calls to mind a paper by Julian Silverman of some years ago(6). It was based on the Hunt study where extensive biochemical and physiological measurements were taken of a group of models before and after 10 Rolfing sessions. A “cluster analysis” was applied to these data which resulted in three groups of models representing three different patterns of data. The first group showed what one would consider favorable physiological changes while the other two did not. The Before and After photographs of all models were ordered according to the three groups and submitted to a Rolfer for structural analysis(7). He judged the first group to show good structural integration while the other two groups showed little structural change. The first group he called “balanced bodies”, the other two “soft body types” and “soft core/hard sleeve types” respectively. This emerging structural typology is used by some Rolfers in their own idiosyncratic way but suffers from the fact that it was never developed descriptively nor defined and so is not applicable in a rational sense. But the conclusion is very seductive that the “balanced bodies” correspond to the “responders” while the other two “types” represent the “non-responders”(8). Furthermore, the observation is in accordance with the experience of the practitioner that clients often seem to fall into two categories: the “easy” ones and the “difficult” ones, with a relatively clear separation between the two. These observations have shifted the focus already from “recognizing structural change” to “producing structural change”, which also merits attention.

<img src=’’>
Tight internal with conflicted shoulder girdle which holds the body up and displaces head and. neck forward. Anteriorly collapsed lumbar segment on pelvis with marked anterior tilt. The chest is collapsed, the upper back correspondingly kyphotic. The body stands very much in front of the ankles.

In the After picture, holding by the shoulder girdle is more manifest. The pelvis is less tilted, and the upper body seems to rise out of it better, being a little longer through the center and better supported. The neck is also a little longer and the head is more back. Feet and legs seem more tensed.

<img src=’’>
Rather soft-looking internal. Tight lower legs which are posterior, tight iliotibial tract and overworked hamstrings. The pelvis seems pulled down on the remora, the lower lumbars are compressed forward. Pronated forearms which seem in conflict with external upper arms.

In the After picture the shortness in the legs seems to pull down the upper body more clearly. The pelvis is less tilted anteriorly, but anterior compression of the lower lumbars is more obvious. The body stands more forward on the feet. There is a little more length through the midbody. The chest has dropped, the shoulders have relaxed forward some. The front of the body from the pubes up seems less supported. The head is tilted back less.

Rolfers’ Rolfing

The study as it is set up asks the question: are Rolfers able to recognize integrative changes on photographs? The formulation contains a hidden premise in that it presupposes that these structural changes are actually produced. It can be said that the variable of the Rolfers’ ability for recognizing structural change is tested against the constant of that change. This relationship can be reversed, and the question is then: are Rolfers able to produce structural changes so that they can be recognized by other Rolfers? The Rolfing side of the study is here the variable and the judges are the constant. It is granted that the performance of the judges doesn’t exactly recommend them as a “constant”. But it shall be shown that the implicit premise that “Rolfers Rolf’ rests also on a shaky foundation.

3 Rolfers Rolfed each 3 of the 9 models of the Rolfing group. They are called “Beginning Rolfer” (BR), “Experienced Rolfer” (ER), and “Advanced Rolfer” (AR) respectively. An analysis of Table I shows that their sessions were recognized to an extremely different degree. BR received 14 of possible 24 “nominations”, ER 9, and AR 3. An extrapolation shows the situation in a clearer way. If we assume that BR had done all 9 sessions with the identical rate of recognitions, he would have received 42 “nominations”. We further assume that the remaining 30 “nominations” would have been distributed among the three other methods the same way as in the study. Then the result would have looked the following way for BR, expressed in percentages of all “nominations” (figures rounded):

Rolfing – 58%

Reiki – 20%

Massage – 13%

Nothing – 8%

If AR had done all 9 sessions with the same assumptions as taken for BR, the result would have been:

Reiki – 42%

Massage – 28%

Nothing – 16%

Rolfing – 12%

In the first case the result would be judged satisfying although certainly not exhilarating. The second case would have to be judged a desaster.

This finding makes the study almost useless in a traditional sense because a “constant” may vary a little but must certainly not show such extreme variances. It definitively renders statistical evaluation which does not distinguish between the three Rolfers illusionary.

<img src=’’>
Table II

number of nominations of a session

number of sessions receiving the same number of nominations, ordered according to methods. Average number is indicated for each method (figures rounded).

Rolfing Reiki Massage Nothing 3,0 2,6 1,7 1,0

Ill.1 Some photographs of models in the study. The number on top is the code number used in the study. It is followed by the method applied. The number after “R” indicates how many of the 8 judges called the session Rolfing.

This very large difference in the outcome of the evaluation demands discussion. It cannot be said that it was due to the difference in how long these Rolfers have been working with respectively different degrees of experience; nor can it be ascribed to the different amount of training. For, the outcome is exactly the opposite way: the Beginning Rolfer scored highest, the Advanced Rolfer worst, and the Experienced Rolfer in the middle. This suggests to examine such rather exotic hypotheses as “the longer a Rolfer works and the more training he has, the less is his work recognized by colleagues”. It would be compatible with the discouraging statement that “no matter how long one works or how much training one has, one is born as a Rolfer or not”.

Such reasoning goes of course from the fictitious assumption that all the premises mentioned had been accepted, including the judges being able to evaluate objectively. This is more than in doubt, of course. Another point could be made: it is possible that BR got all the “easy” models, the “responders”, and AR all the “difficult” ones, “non-responders”. When one examines the models with this objection in mind, it does seem that AR’s models were the most difficult group. However, I would judge ER’s group similarly difficult. The argument is a little thin anyway because it calls for a relative improbability in the distribution of models to restore probability of an average outcome only. Of course we would expect that AR would score best! It does pose the important question however of what makes some clients difficult while others are easy.

The situation, brought to the point, permits two opinions to be held. It can be held that the judges were unable to recognize structural changes or to differentiate between these and non-structural differences. Or it can be held that the Rolfers were unable to produce such structural changes. The truth might lie somewhere in between, but just where that would be is anyone’s guess.


Returning to more fundamental objections, a third opinion which can be held shall be discussed again. It contends that this study is impossible or senseless. The most radical version would be the statement that the true goals of Rolfing are on much higher levels, personal growth or spiritual enlightment e.g. Such results are of course not visible on photographs. The argument can easily be dismissed, however. It doesn’t have anything to do with Rolfing as such. Of course, Ida Rolf also had such goals in mind, but she insisted repeatedly on any such change for the better having to be rooted in structure, in the normal structural pattern having been evoked. Otherwise, such gains were superficial phenomena in her view which was rather radical in this respect.

A milder objection is raised by contending that Rolfing is a slow process, that it takes ten sessions and a long time afterwards to unfold structural integrity. This is very probably true, of course, but the argument is unfortunately not supported empirically in any way. It only merits being taken serious when there is some effort linked to it for creating an appropriate system of evaluation which would allow to base the claim on some sort of substantial evidence. The problem is the same as with the study in question. But it seems more practical to concentrate on recognizing short-term changes reliably – which certainly exist -, which could then form the base for developing the more difficult analysis of long-term changes. But it must be acknowledged that there are difficult clients whose photographs one wouldn’t want to display in public after just one session, and sometimes not even after ten.

The opinion that photographs don’t represent structural changes doesn’t need to be discussed further. But the insecurity that comes with them is legitimate, and it may be emphasized again that they probably result largely from a lack of knowledge about function influencing the appearance of the body. Since affective states determine this functional side to a great extent, the following speculation can be made. If the members of the Nothing group had not only walked around the table but taken a good hour’s nap, or been asked to take a brisk walk along the river, they might have shown even greater differences in the After pictures. Discrimination of the Nothing group would even have been worse. It would not be unfair to suggest such a procedure because the members of the other groups can certainly be assumed to have been in markedly different affective states after their treatment. On the other hand, if the After photographs had been taken 24 hours later, functional differences between Before and After could be thought to have mostly passed, “randomizing” the groups in this respect. The structural differences could perhaps have been distilled out more clearly and the result would have been better. The argument calls to attention again stance which is not understood well enough. Even differences in appearance between inspiration and exspiration are not known e.g.

So despite the disappointing result this study has great merits. It calls to mind specific white spots in the structural territory, and it directs attention to hidden premises and hypotheses not examined and even hardly conscious until now. This paper has attempted to point out some of them more clearly and help prepare the ground for studies which focus more accurately on such questions. That they are not only desirable but also absolutely necessary is proved by the outcome of the study.

Photographs of some of the models of the study are supplied. They are accompanied by a short analysis which was done in cooperation with Peter Schwind. It may seem paradoxical to do that when the point of view of this paper can be understood to believe this impossible – and misunderstood as holding such an undertaking futile. Certainly our annotations must not be mistaken as being objectively true. They are the result of our combined idiosyncratic ways of looking at structure. If we present them, we do so because only by stating such structural observations of subjective relevance an analytical system can be developed eventually. Four aspects need to be mentioned.

For one, the observations are on different levels of evidence. Some are clear in a geometrical sense and can be checked directly. Others are more interpretative and require some experience and subjective knowledge, especially concerning the external/internal system. They rest on premises and assumptions. Still others are more intuitive and little defined. There is no question that the first kind is preferable and that the other kinds should be based on them. Only when they acquire sufficiently sharp contours and concrete content can they possibly become meaningful means of communication.

A second consideration is that of validity. It is not certain that our observations are really correct, that what we think we perceive is actually so. Because there exists no apparatus which produces “objective data” for us, such validity of observations can only result from a prolonged process of making them, checking them with reality, and criticizing them. The outcome depends very much on the success of rendering them operational.

Still another issue is that of relevance. Again it is not certain that our observations actually describe important features and properties of the given structures. They might be rather marginal and not touch on the essence of these structures. They may even be outright wrong.

And finally, our analyses are certainly incomplete. Other approaches and observations different in nature are possible and necessary to gain a deeper and more extensive understanding of the geometry of structure and its mechanics.

Such somewhat binding analysis needs to be developed much more broadly to confront it with the questions these four aspects pose. Only this will eventually make the essential step possible when Rolfing turns into a relatively rational and universally applicable method which can be evaluated. It is still very much a “hit or miss” procedure where it is even only darkly known whether it “hit or missed”. The reader is invited to check our observations critically and find weak spots and essentials we missed, and correct mistakes. All the information we had is at his disposal. We furthermore encourage everyone to contribute.


1. The very critical reader will count 73 nominations.

2. There is a number of studies which associate positive medical, physiological, biochemical, or psychological effects with Rolfing. For reasons of elementary logic however they do not support that the subjects’ structures were actually integrated. There is an immense multitude of methods and interventions “between talk and touch” which produce such effects.

3. Richard Carrera, in “Psychotherapy: Theory, Research, and Practice”, 1974; as communicated by Anthony M. Zimkowski in “Rolf Lines”, Vol. XIII, No. 3, March 1985 (“Forum” section), Rolf Institute, Boulder.

4. The reverse is not conclusive although welcome. When a client feels better or moves easier, this is not proof that the structure has been integrated. Such effects are produced by most methods and many non-therapeutical activities which are not considered to integrate structure.

5. The case for the normal movement of the LDH as backwards is argued in the paper on “Breathing” in this issue.

6. Silverman, Julian: “Dr. Rolfs Agnews Project”, unpublished paper available at the Rolf Institute, Boulder.

7. Emmett Hutchins.

8. Since the Hunt study was done relatively long ago, it can also be suspected that the “non-responders” overlap partly with externals whose different mechanics were not known then.

To have full access to the content of this article you need to be registered on the site. Sign up or Register. 

Log In