Predicting maximal performance from a submaximal effort in resistance training is a useful method to avoid exposure to extreme physical requirements, thereby minimizing the risks of injury (Mayhew et al. 2008). This approach uses muscle endurance performance through a number of repetitions performed to predict maximal strength performance, replacing a 1-repetition maximum protocol (Mayhew et al. 2008; Knutzen et al. 1999). These strength predictions also aid in exercise prescription and manipulation of training load (Mayhew et al. 2008). Unlike muscle strength, prediction equations currently do not exist for flexibility capacity represented by joint maximal range of motion (ROM_MAX). The ROM_MAX is defined as the maximum range of joint motion tolerated by the participant during a stretching maneuver (Halbertsma et al. 1996; Magnusson et al. 2000; Blazevich et al. 2012).

While ROM_MAX is a variable associated with the maximal tolerance of the subject to stretching, another variable related to the individual's tolerance to stretching is the first sign of pain (i.e., first detection of pain), that was first analyzed by Halbertsma and Göeken (1994). In spite of the use of the term “pain”, the authors instructed the volunteers instead to note the first sensation of tension in the musculature, generated during passive stretching. A similar procedure was performed in other studies that also associated the first sensation of tension in the musculature during stretching to the individual's tolerance of stretching (Cabido et al. 2014; Halbertsma et al. 1999; 2001; Ylinen et al. 2009). In these studies, the ROM corresponding to the first muscular sensation of stretching (FSS) was the measure used (ROM_FSS). Afferent communication from the muscle-tendon unit (MTU) and joint kinesthetic receptors is responsible for the signaling of ROM_MAX and ROM_FSS by the individual during a stretching maneuver. The modulation of information by individual anatomical and physiological differences and, consequently, the impacts on ROM variables are currently not addressed in the literature. Some authors have reported that of the various putative neuromuscular factors, higher ROMS are associated with the individual's ability to tolerate higher torque values (Ben and Harvey 2010; Magnusson 1998; Weppler and Magnusson, 2010). According to Blazevich et al. (2012), this capacity and volitional stretch termination may be related to subconscious responses to afferent stretch (I and IIa), pressure (III) and pain (IV), as well as differences in the supraspinal registration of afferent signals. These authors also suggest that future studies are warranted to explain each responses relative influence on ROM.

Considering the multi-factorial influences described above, ROM_MAX indicates the maximal tolerable stretching sensation and ROM_FSS represents the first sensation of stretching associated with MTU stretch-associated tension. Thus, ROM_MAX and ROM_FSS could represent the limits of a continuum of the individual's tolerance to stretching. Postulating the existence of this continuum is based on the results of studies that have analyzed the effects of stretching on these two variables. Furthermore, finding that ROM_MAX and ROM_FSS both change in the same direction indicates similar factors may be related to their changes after acute (Cabido et al. 2014; Halbertsma et al. 1999; 2001; Ylinen et al. 2009) and chronic muscle stretching (Halbertsma and Göeken 1994). Although the aforementioned studies do not substantiate the level of relationship between ROM_FSS and ROM_MAX, an analysis based on data presented by Halbertsma and Göeken (1994) demonstrates a strong and significant correlation between ROM_FSS and ROM_MAX (r = 0,94; p = 0,001; R² = 0,88), supporting the possible existence of this continuum. Recently, Chagas et al. (2016) through an exploratory factor analysis reported ROM_FSS and ROM_MAX were inserted into the same factor. These findings reinforce the expectation that these variables are interrelated and that they may represent a common aspect associated with tolerance to stretching.

In this context, the use of ROM_FSS to predict ROM_MAX may help in the development of evaluation methods with a lower risk of injury and discomfort. In addition, a greater tolerance to the ROM_FSS could improve the monitoring of flexibility training. However, differences between males and females for ROM_MAX and the individual’s tolerance to stretching could influence the prediction capability via ROM_FSS (Hoge et al. 2010; Gajdosik et al. 1990; Gajdosik, 2006; Marshall and Siegler 2014; Wiesenfeld-Hallin 2005). Thus, given the possible differences between males and females, two different sex-specific prediction equations for ROM_MAX may need to be developed, as has been proposed in the case of other capacities, such as muscular strength (Mayhew et al. 2008).

Moreover, considering the hypothesis of the continuum of stretching tolerance, if an individual can reliably determine a specific point within this continuum then this data may reinforce a possible linearity between ROM_FSS and ROM_MAX. In this way, the ROM registration corresponding to a specific point, (i.e., a tension sensation corresponding to halfway (ROM_50%) between the lower (ROM_FSS) and upper (ROM_MAX) limits of the continuum), may allow for verification of whether the individual is able to reproduce an adequately determined sensation of stretching. Stretching sensation is used as an indicator of the intensity of the elongation stimulus in experimental studies involving flexibility training (Kay et al. 2015; Freitas et al. 2015). Despite this, the reliability of the stretching sensation associated with a submaximal stimulus of the stretching stimulus has not yet clearly been investigated. Another aspect to consider is the comparison of the registered value of ROM_50% and the expected value of this variable obtained by the predicted values of ROM_MAX using a measured ROM_FSS.

Thus, the objectives of the present study were to: (a) verify whether ROM_FSS and ROM_MAX are correlated and whether ROM_FSS is able to accurately predict ROM_MAX in males and females; (b) verify the reliability of the ROM_50% measure; and (c) compare the observed and expected ROM_50% value. We hypothesized that: (a) ROM_FSS and ROM_MAX will be significantly and strongly correlated, and the first sensation of stretching (ROM_FSS) will be a predictor of maximal performance (ROM_MAX); (b) ROM_50% will exhibit high reliability values in a test-retest procedure; and (c) there will be no significant difference between the observed and expected mean values of ROM_50%.

Based on correlation r values between ROM_FSS and ROM_MAX from a pilot study, sample size was calculated using GPower 3.1 software (Dusseldorf, Germany). The calculation resulted in 58 individuals for a statistical power above 80% and a priori significance level was set at α = 0.05 (Beck, 2013). The final sample consisted of 69 individuals: 37 males (mean ± SD: age = 23.3 ± 3.8 years, body mass = 75.2 ± 14.2 kg and height = 1.76 ± 0.6 m) and 32 females (mean ± SD: age = 23.0 ± 4.0 years, body mass = 58.4 ± 9.5 kg, height = 1.61 ± 0.5 m). The study was approved by the ethics committee of the University of Minas Gerais, Belo Horizonte, Brazil (Project # CAAE 25468114.7.0000.5149) in accordance with the Declaration of Helsinki, and all volunteers signed an informed consent form prior to any testing. None of the volunteers had undergone flexibility or strength training prior to testing, or had musculoskeletal injury of the lower limbs, spine, or pelvis, in the previous six months.

All subjects met the following inclusion criteria: a) an absence of history of neurological or orthopedic pathologies in the lower limbs within the last six months and no recent or chronic low back pain (Halbertsma et al. 1999); b) an absence of musculoskeletal injury or pain in the lower limbs, spine or pelvis in the last six months (Blackburn et al. 2004); and c) subjects had not participated in activities involving flexibility training during the three months prior to the study and therefore were considered untrained (Halbertsma et al. 1996). The exclusion criteria for this study included (a) the practicing of stretching exercises during the period of data collection; b) missed testing attendance on the scheduled day and time; and c) the ability to achieve full knee extension in the pre-test phase on the provided equipment during ROM_MAX measurements.

In this study we used a repeated-measures design to assess hamstring flexibility via an isokinetic knee extension test as previously described (Cabido et al. 2014; Peixoto et al. 2015). Participants were seated on the Flexmachine with the distal third of the thigh supported at 45º hip flexion relative to the ground. Positioning prevented complete knee extension and ensured elongation of the muscle-tendon unit was measured and performance was not limited by joint mechanics. Subjects performed the same procedures for each experimental session consisting of five valid repetitions of passive knee extensions, at a speed of 5°.s-1 with an interval of approximately 15 s between each repetition. In each repetition participants were asked to reach maximum knee extension and to press an analog button when experiencing ROM_FSS and estimating the mid-point between the first sensation and ROM_MAX. Therefore, ROM_FSS, ROM_50%, and ROM_MAX were measured during each maximal attempt.

Participants visited the laboratory on four different days. The first day was a familiarization trial, and the subsequent three days consisted of identical experimental sessions. The time interval between sessions one and two was 48 hours and between sessions one and three was 28 days to verify both the reliability of the measurements and the accuracy of the prediction equation. In sessions one to three ROM_FSS, ROM_50%, and ROM_MAX, corresponding to the lower limit, midpoint, and upper limit of the continuum of stretching tolerance, were measured. For each volunteer, all sessions were conducted at the same time of the day (< 2 hours). Both lower extremities of the volunteers were assessed.

ROM_FSS was recorded by asking subjects to indicate when they first felt stretching in the hamstrings by pushing a button. This FSS was recorded as a function of the knee extension angle, generating the ROM_FSS variable (Cabido et al. 2014). Together, the ROM_MAX and ROM_FSS variables determined the continuum of stretching tolerance for each individual, allowing a perceptual scale to be created equally for all participants. Thus, the determination of these variables allowed for the establishment of the minimum and maximum values within the continuum of stretching tolerance. This procedure was based on the study of Gearhart et al. (2001). ROM_50% was defined and recorded at the sensation of stretching corresponding to the midpoint between the lowest (ROM_FSS) and highest (ROM_MAX) limits of the continuum.

Each participant was familiarized with and properly positioned on the isokinetic dynamometer termed the Flexmachine that was built in the Biomechanics Laboratory of the Federal University of Minas Gerais (UFMG, Belo Horizonte, Brazil) (for details see Peixoto et al. 2015). All equipment settings were recorded for future use (Cabido et al. 2014). Familiarization consisted of two stages. Initially, subjects were instructed on how the equipment worked, how the procedures were to be conducted and how the data was to be collected. Then subjects were asked to perform, on average, five trials, during which the ROM_FSS, ROM_50%, and ROM_MAX variables were collected and recorded.

Subjects performed the same procedures for each experimental session consisting of five valid repetitions of passive knee extensions of each leg, at a speed of 5°.s-1 with an interval of approximately 15s between each repetition. Schematic illustration of the testing protocol is shown in Figure 1.

Subject characteristics are presented as mean and standard deviation of the variables. The strength of relationship between ROM_FSS and ROM_MAX was determined from a Pearson Product Moment linear correlation coefficient (r), and the ROM_MAX prediction by ROM_FSS was verified using simple linear regression. The fit quality of the regression model was assessed by the coefficient of determination (R2). The coefficient of determination is interpreted as the proportion of total variance of the dependent variable that is explained by the variance of the independent variable. Moreover, since the value of R2 is defined by a ratio between the variance explained by the model and the total variability that represent the sum of the explained variance and the unexplained variance (error), the larger the value of R2, the smaller the model error (SEE). Together, these two statistics (i.e., R2 and SEE) measure the adequacy of the regression model (Field et al. 2013).

The simple linear regression equation is indicated as follows:

where y is the dependent variable, the predicted value of ROMMAX; x is the independent variable (ROMFSS); and a and b are the estimated coefficients of the regression line using the least squares calculation method. To verify the accuracy of the prediction model, the standard error of estimate (SEE) was computed as the square root of the unpredictable portion of the variance in a set of observations as follows:

(2)

where, Š· is the predicted score from the regression line; y is the actual score; and n is the number of subjects. In addition, a repeated measures ANOVA was used to compute the difference between the observed and expected ROMMAX values.

The reliability of ROM_50%, in the experimental sessions was determined by the intra-class correlation coefficient (CCI) and the standard error of measurement (SEM). The ICC values were interpreted as weak (<0.4), moderate (0.4-0.59), good (0.6-0.74) and excellent (0.75-1.0) (Cicchetti, 1994). Finally, the comparison between the ROM50% value recorded by the volunteer and the expected value for the variable as defined by (ROM_MAX - ROM_FSS) / 2 + ROM_FSS was performed using a repeated measures ANOVA. Normality and sphericity were verified using Shapiro-Wilks and Mauchly's tests, respectively. When necessary, a post-hoc Tukey HSD test was used to identify the differences reported in the ANOVA. Statistical analyses were conducted using the Statistical Package for the Social Sciences (SPSS version 22.0; Chicago, IL, USA) with a significance level of 5%.

The results from this study are reported in three parts. First, the correlation and variance values between ROM_FSS and ROM_MAX and the prediction equations with the accuracy values (SEE) are presented. Secondly, the reliability data of ROM_50% (ICC and SEM) from each experimental session are conveyed. The final part presents the repeated measures ANOVA for the comparison of the observed and expected values of ROM_50%. In each trial, we recorded the ROM_FSS, ROM_50%, and ROM_MAX variables (Figure 2).

The correlation values between ROM_MAX and ROM_FSS, for female and male subjects, and the ROM_MAX prediction equations through ROM_FSS in all experimental sessions, with the coefficient of determination and standard error of the estimate, are depicted in Table 1.

In Table 1, prediction equations, for females and males, in the three experimental sessions (S1, S2 and S3) accompanied by Pearson Product Moment correlation values (r), coefficient of determination (R²), standard error of the estimate (SEE) and p values of the means comparison between the observed and expected values of joint maximal range of motion (ROM_MAX).

The repeated measures ANOVA results between the observed and expected ROM_MAX values are depicted in the Table 1 above (p-value). Interestingly, there were no significant differences between the values observed and predicted in the three experimental sessions in the case of either male or female. Figure 3 illustrates the prediction of ROM_MAX by means of ROM_FSS in experimental session 1.

An intraclass correlation coefficient (ICC) for ROM_MAX and ROM_FSS was calculated for both male and female from measurements during sessions one and two; these inter-session values were 0.93 and 0.88 for the males, and 0.98 and 0.97 for the females, respectively. In addition, percent standard error of measurement (SEM%) values were reported to reflect the reliability (i.e., 4.44% for ROM_MAX and 2.21% for ROM_FSS in male; and 2.70% for ROM_MAX and 2.70% for ROM_FSS in female).

Table 2 depicts intraclass correlation coefficient (ICC) and standard error of measurement (SEM) values for the range of motion corresponding to halfway (ROM_50%) between the lower (ROM_FSS) and upper (ROM_MAX) limits of the continuum of the individual's tolerance to stretching, in each of the three experimental sessions (S1, S2 and S3). ROM_50% reliability was strong for the three experimental sessions’ ICC values in both males and females. In addition, SEM values found for ROM_50% varied between 3.15º and 3.62º for female subjects and between 3.34º and 3.69º for male subjects.

The results of the repeated measures ANOVA did not indicate significant differences between the observed ROM_50% value and predicted values for the three experimental sessions, for either males or females (p = 0.65; η² = 0.15).

The first hypothesis of the present study was that ROM_FSS and ROM_MAX would be strongly correlated, and that the first sensation of stretching (ROM_FSS) would predict maximal performance (ROM_MAX) in male and female subjects. The results revealed a strong and significant Pearson Product Moment correlation between ROM_FSS and ROM_MAX, with a statistic that was higher than 0.80 in all experimental sessions (Table 1). Our results support the data from a previous study (r = 0.94; p = 0.001) (Halbertsma and Göeken 1994). In addition, the strong correlation between ROM_FSS and ROM_MAX determined in the present study corroborates the results of Chagas et al. (2016). These authors performed an exploratory factor analysis (EFA) involving variables often used to evaluate the response of the MTU to stretching exercises and found that the ROM_FSS and ROM_MAX represent a common construct (factor). Thus, considering EFA theory, factors are composed of variables measuring common aspects. The strong correlation found between ROM_FSS and ROM_MAX in this study allowed linear prediction modelling in this analysis and verification for the ability of ROM_FSS to predict ROM_MAX with accuracy in male and female subjects. The results of the coefficient of determination (R²) of the prediction model were between 67.2% and 92.0% (average of 76.4%), for both males and females (Table 2). It is reasonable to conclude that on average, 76.4% of the variability of ROM_MAX are explained by the variability of ROM_FSS. For a model to be considered a good predictor, it must have R² values above 70% (Hair et al. 1999). In the present study, the mean accuracy of the model given by the SEE for males and females was 9.01º (range 6.54º - 10.72º) (Table 2). The SEE measures the mean deviation (error) between the observed and predicted values of ROM_MAX and is interpreted as the standard deviation of residuals, since these residuals are normally distributed and what is sought is the achievement of the lowest possible value of SEE (Field et al. 2013). In our study, the mean value of SEE was 9.01º that together with the non-verification of a significant difference between the observed and predicted ROM_MAX values (p > 0.05) for the three experimental sessions, allows us to reasonably infer that the prediction model rendered high accuracy.

These results together reinforce the idea that ROM_FSS and ROM_MAX may represent points at the ends of a continuum of individual tolerance for stretching exercises ranging from the initial through final individual tolerance, respectively (Chagas et al. 2016). However, an explanation that involves the structures and mechanisms associated with the modulation of individual tolerance to stretching (i.e., the beginning and end of the continuum) is not yet fully accessible. The tension perceived by an individual during stretching is due to mechanical stimulation of the existing receptors in the musculotendinous unit, especially the free nerve endings (afferents of group III and IV) (Stacey 1969). Different studies have reported that these free nerve endings are stimulated during stretching, in particular, triggering depolarization of type III afferent fibers (Hayes et al. 2005; Mense and Meyer 1985; Paintal 1960). Findings from the study of Cleland et al. (1990) have revealed that the response frequency of type III afferent fiber (i.e., the change in the number of action potentials) increases with the increase in passive force that results from stretching. Furthermore, the stimulus originated by the afferent pathways penetrates the posterior horn of the medulla reaching the gelatinous substance. According to the gate theory of Melzack and Wall (1965), this would modulate the afferent signal that is sent to the higher centers. This modulated signal is perceived by the superior centers as a process, and not as a single event that is interpreted on the basis of the attention, emotion and memories of previous experiences (Melzack and Wall 1965). As such, this mechanism indicates the possibility of modulating stretching sensation through these afferent pathways. In addition, this mechanism reinforces the perspective of a continuum of the individual's tolerance to stretching, since lower levels of activation of the afferent pathways during stretching would be associated with lesser sensation of stretching (i.e., ROM_FSS, while ROM_MAX would be related to a greater level of activation of these afferent pathways). Thus, increased activation of these mechanoreceptors results in a highly uncomfortable sensation that could limit the achievement of a greater range of motion (e.g., ROM_MAX). In addition, results from previous studies that investigated ROM_FSS and ROM_MAX, also corroborate this perspective of a continuum of tolerance. Cabido et al. (2014) and Ylinen et al. (2009) both found that an increase in ROM_MAX was accompanied by a corresponding increase in ROM_FSS.

The second hypothesis, that ROM_50% would show high reliability values in a test and re-test procedure, was confirmed by ICC and SEM values (Table 2). The ICC determines the capacity to differentiate between subjects, and is calculated by dividing the variance between subjects by the total variability (i.e., variance between subjects plus error) (Weir 2005). The SEM quantifies the error in the same unit of measure as the original variable, and is not influenced by the variability between individuals (Weir 2005). Therefore, small ICC values may indicate that the sample is homogeneous, via a consideration of the study variable and not that the reproducibility is small (Weir 2005). The SEM value reflects the degree of fluctuation of the scores of an individual in a test or condition, indicating the expected natural variability (i.e., random error) for the response of a given variable (Weir 2005). Higher SEM values indicate higher ranges of expected scores for a certain variable making it difficult to perceive significant changes in the scores of an individual following a systematic intervention (e.g., training). ICC results (~ 0.94) and SEM (~3º) for the three experimental sessions that corroborate the second hypothesis (i.e., that ROM_50% indicated a high reliability).

Finally, the third hypothesis of this study considered was that there would be no significant difference between the observed and expected ROM_50% mean values. This was confirmed by repeated measures ANOVA (F_(2,32) = 0.44, p = 0.65; η² = 0.147). These findings indicate that the individuals who participated in the study were accurate in determining the sensation of tolerance to stretching, since the value registered by these individuals was not different from the expected value.

The concept of a continuum of the individual's tolerance to stretching is strengthened by the reliability of ROM_50%, the absence of statistical difference between the observed and expected mean values for ROM_50%, and data from ROM_FSS and ROM_MAX. These results allow us to assert some assumptions. The first assumption is that the continuum of the individual's tolerance to stretching can be considered a perceptual range. Furthermore, this range of perception can be established equally for all individuals by determining a lower (ROM_FSS) and higher (ROM_MAX) limit of the continuum of tolerance to stretching. Thus, even if there is inter-individual variation in the perception range, all individuals should be able to undergo the same process of anchoring the range limits. Since the definition for this range limits is reliable and consistent over time, an individual should also be able to reliably and consistently indicate a point within the range of perception that would represent a particular “intensity” of individual perception. This assumption was tested in the present study through having the subject indicate an “intensity” corresponding to the half of the lowest (ROM_FSS) and highest (ROM_MAX) limits of the perception range (ROM_50%) during the stretching maneuver. The individual’s capacity to accurately and reliably determine this point provides support for the assumption of the linearity of the perception response to stretching stimulus. Interestingly, results from the present study confirmed this assumption. In this way, the present study demonstrates need for the development of a standardized scaling procedure of individually perceived “intensity” during stretching. Future studies are warrantied, however, to verify the impact of using this standardized scaling procedure on the response of the muscle-tendon unit submitted to a training process in both more controlled conditions (e.g., in the laboratory), and in those with a greater ecological validity.

Any extrapolation of the results found in the present study may need to consider the sample of the population studied (i.e., young males and females). Furthermore, extrapolation can only be made for individuals without musculoskeletal injury in the joint of interest. In the scientific literature, the elderly and individuals with musculoskeletal injuries have a lower maximum passive torque, indicating a lower tolerance to stretching in these populations (Magnusson 1998; Bressel and McNair 2001). Consequently, it is possible that the difference in tolerance changes the relationship between ROM_FSS and ROM_MAX. Thus, group differences for tolerance still needs to be investigated in future studies to determine valid prediction equations for specific populations. Finally, additional research is necessary to verify if the results found in this study are reproducible in other muscle groups.

The present study reports results that support the perspective of a continuum of tolerance to stretching and a proposal that ROM_MAX may be predicted by the first sensation of stretching (ROM_FSS). Such findings are of great importance in the context of sports and rehabilitation, since they will allow sports scientists, coaches, and physiotherapists to predict an individual's ROM_MAX, without performing a maximal physical test, thereby minimizing the risk of injury. However, at this time, this recommendation is limited to young male and female subjects, as well as to the muscle group investigated in this study. Thus, an important aspect to be considered is that the relationship was found using a method that allowed for the control of several factors that could interfere in the ROM_FSS and ROM_MAX including the standardization of the individual's positioning, the speed of the stretch, the muscle group, and the way to operationalize the variables. For greater external validity, future studies should investigate measures similar to those performed in a clinical situation. In addition, future studies should investigate whether the linearity found in the extremes of the continuum of stretching tolerance, given by ROM_FSS and ROM_MAX, would also be true in the case of the submaximal values that lie within that continuum.