Motor unit recruitment patterns 2: the influence of myoelectric intensity and muscle fascicle strain rate

Muscle fibres within a single motor unit have similar biochemical and contractile properties (Burke et al.,1971), meaning that individual motor units have distinct physiological and mechanical properties. Mammalian skeletal muscles are composed of a mixture of motor unit types. For a muscle to effectively contribute to smooth, co-ordinated movement it must therefore selectively activate an appropriate number and combination of motor units to generate the stresses and strains required. Differences in muscle fibre physiology(Burke et al., 1973),metabolism (Peter et al.,1972) and myosin heavy chain (MHC) isoforms(Schiaffino et al., 1989)suggests that muscle fibres are specialised to perform distinct functional tasks. For example it is suggested that the maintenance of posture is best achieved by activation of muscle fibres that shorten at low strain rates,develop modest power and use minimal energy for their contractile activity(He et al., 2000). By contrast, powering rapid movements is best achieved by recruitment of muscle fibres that shorten at higher strain rates, developing greater power, but at a greater metabolic cost (He et al.,2000).

Muscles can alter the force they produce by changing the firing frequency of the active motor units and changing the number of motor units that are active at any one time (Adrian and Bronk,1929). In a classic series of experiments the stretch reflex response of decerebrate cats revealed that motor units were recruited in an orderly fashion, termed the `size principle'(Henneman et al., 1974; Henneman et al., 1965a; Henneman et al., 1965b). Since the size principle was first described its effects have been observed in a large number of in vivo studies. These have come from sources as varied as the respiratory muscles of chickens(Fedde et al., 1969),voluntary contractions in humans (Freund et al., 1975; Hogrel,2003; Milner-Brown et al.,1973; Tanji and Kato,1973) and walking cats (Hoffer et al., 1987). There is, however, a growing body of evidence suggesting that the size principle does not always hold true during rapid contractions, and that the recruitment of faster motor units without prior activation of slower motor units can occur. These examples come from glycogen depletion studies of supra-maximal cycling in humans(Gollnick et al., 1974) and jumping in the bushbaby (Gillespie et al.,1974) and from studies of reflex inhibition in the cat(Sokoloff and Cope, 1996). More recently, electromyographic studies of humans running(Wakeling, 2004) and cycling(Wakeling et al., 2006) and running rats (Hodson-Tole and Wakeling,2007; Hodson-Tole and Wakeling, 2008) have also reported preferential recruitment of faster motor units. Henneman and co-workers(Henneman et al., 1965b)themselves reported that recruitment of slower motor units prior to faster ones occurs in approximately 85% of pairs of motoneurons tested, a finding supported by other reports (Sokoloff et al., 1999). Patterns of motor unit recruitment other than those predicted by the size principle must therefore exist. It is consequently of interest to determine what other factors, in addition to the level of muscle activity, govern patterns of motor unit recruitment and the conditions under which these factors become significant, and this topic forms the basis of the study presented here.

Variations in sarcomere structure and organisation lead to differences in the mechanical properties of different motor units(Hill, 1950; Johnston, 1991). The intrinsic speed or maximum unloaded strain rate, determines a number of contractile properties within a muscle. In humans, maximum mechanical power has been shown to occur at higher strain rates in faster motor units [0.083 s^–1 type I; 0.23 s^–1 type IIA; 0.28 s^–1 type IIA/IIB at 12°C(He et al., 2000)]. In addition, the maximum mechanical efficiency also occurs at higher strain rates in faster motor units [0.05 s^–1 type I; 0.15 s^–1 type IIA at 12°C(He et al., 2000)]. Producing maximum mechanical power at high efficiency during fast movements would,therefore, be best achieved by preferentially recruiting faster motor units(Rome et al., 1988). In addition to the differences in motor unit force–velocity relationship,differences also exist in activation and relaxation rates between the motor unit types. For a muscle to effectively contribute force during a cyclic motion, it must be able to generate and relax its force at a suitable rate for the movement. Faster motor units have faster activation and relaxation rates(Burke et al., 1973), and are therefore better suited to situations where rapid force development is required.

The predictions of the size principle provide a good explanation of motor unit recruitment for sustained low force contractions e.g. postural control. In these instances recruiting slower, fatigue-resistant motor units first has several functional advantages. In situations where rapid force production is required, however, there is strong evidence to suggest that there should be a mechanical basis for force production and hence motor unit recruitment. Such a relationship has already been shown to exist in the medial gastrocnemius muscle of humans when cycling (Wakeling et al., 2006). In the companion paper(Hodson-Tole and Wakeling,2008), we showed that patterns of motor unit recruitment changed significantly in response to changes in locomotor velocity and incline and that preferential recruitment of faster motor units does occur in the rat. Here we investigate the intrinsic factors that may govern the patterns of motor unit recruitment recorded. The aims of the current study were therefore to: (1) determine the influence of the size principle on motor unit recruitment patterns by determining the relationship between motor unit recruitment and myoelectric intensity; (2) identify if preferential recruitment of faster motor units may provide a mechanical advantage by determining the relationship between motor unit recruitment and muscle fascicle strain rate; and (3) identify differences in the associations described above between muscles with distinctly different muscle fibre populations.

Wavelet analysis has been shown to provide a highly sensitive method of assessing myoelectric data (Hodson-Tole and Wakeling, 2007; Wakeling and Rozitis, 2004). The size principle predicts that faster motor units, with their higher frequency signals, will be progressively recruited as the strength of the muscle contraction, and hence myoelectric activity,increases. A positive association between the intensity and frequency of the myoelectric signal would therefore be expected(Wakeling and Rozitis, 2004). Preferential recruitment of faster motor units would also result in a relative shift of the myoelectric signal to higher frequencies(Hodson-Tole and Wakeling,2008). Faster motor units have greater mechanical power outputs and efficiency at faster shortening velocities(He et al., 2000). If preferential recruitment of faster motor units is associated with faster shortening velocities, it may therefore be suggested that there is a mechanical advantage to such a recruitment strategy. The previous study demonstrated that patterns of motor unit recruitment vary between muscles and in response to changes in locomotor velocity and incline(Hodson-Tole and Wakeling,2008). In this study we investigated whether, in addition to the relationship between myoelectric intensity and the recruitment of faster motor units, there could be a mechanical advantage for the variations in motor unit recruitment patterns reported. We therefore hypothesised that there would be a significantly positive association between myoelectric signal frequency content and muscle fascicle shortening strain rate. We also hypothesised that there would be a significantly positive association between myoelectric signal frequency content and myoelectric intensity.

A detailed description of the methodology is provided in the accompanying paper (Hodson-Tole and Wakeling,2008). A brief outline only is therefore provided here.

Myoelectric and sonomicrometric data were collected from the soleus,plantaris and medial gastrocnemius muscles of the right hind limb of 19 female Sprague Dawley rats [approximate age 5–6 months; mass 250.07±18.57 g (mean ± s.d.); Table 1]. All rats had undergone a 5 week training programme during which time they were habituated to run on a custom-built motorised treadmill at nine speed (20–50 cm s^–1) and incline (0–25°) combinations. The rats were housed in pairs in cages, maintained on standard rat feed and kept in a temperature-controlled room (20°C) with a 12 h:12 h light:dark cycle. All procedures were conducted in accordance with current UK Home Office regulations.

Offset twist-hook silver wire electrodes (0.1 mm diameter; California Fine Wire Inc., Grover Beach, CA, USA) were surgically implanted into two of the muscles of interest in each subject under sterile surgical conditions. The electrodes were placed in the mid-belly region of the soleus and plantaris muscles, while in the medial gastrocnemius muscle the medio-caudal region was used. This area was specifically targeted as it has been shown to contain a predominant proportion of MHC type IIB fibres(Armstrong and Phelps, 1984). In addition, two sonomicrometry crystals (1.0 mm +38 gauge stainless steel lead wires; Sonometrics Corp., London, ON, Canada) were placed in the third muscle of interest to measure muscle fascicle length changes. Excess wire from the myoelectric and sonomicrometric transducers externalised in the region of the shoulder blades was placed into a small cotton pouch, which was secured under a small jacket fashioned for each subject from elasticated bandage(Vetrap™; 3M United Kingdom PLC, Bracknell, UK). The jacket protected the wound and the wires, and enabled the animals to be kept in their pairs during the recovery period. All subjects received post-operative analgesia(buprenorphine, 0.01 mg kg^–1, subcutaneously) during the 48 h recovery period.

All data were collected 48 h post surgery in an electrically shielded room. Two cameras (A602f; Basler, Ahrensberg, Germany) connected to the data collection computer via IEEE 1394 ports and running off digital triggers were used to collect kinematics data (100 frames per second) from the right hind limb during each trial. Myoelectric signals (3200 Hz) were collected through a 16-bit data acquisition card (PCI-6221, National Instruments Corp., Austin, TX, USA) having been amplified (CP511 AC amplifier,Astro-Med Inc., West Warwick, RI, USA), with a bandpass filter of 30–1000 Hz. Each rat ran a three block, randomised exercise protocol incorporating nine speed and incline conditions (0° at 20, 30, 40 and 50 cm s^–1; 10° at 20, 30 and 40 cm s^–1;20° at 20 cm s^–1 and 25° at 20 cm s^–1), a design that minimised bias in the results due to muscle fatigue or temperature. All data were collected for periods of 30 s,following synchronous triggering via the data acquisition card. Myoelectric and video data streams were synchronised using custom-written software (LabView 7.1, National Instruments Corp.). On completion of the trial, animals were euthanased with an intraperitoneal overdose of pentobarbitone, and dissection carried out to confirm the location of the fine-wire electrodes and sonomicrometry crystals in each of the muscles.

Sonomicrometry signals were partitioned into complete strides defined,using video data, by consecutive foot on times of the right hind limb (total number of strides analysed soleus: 1375; plantaris: 2052; medial gastrocnemius: 2035). Within each subject sonomicrometry traces for each condition were grouped and, based on least squares minimisation, a three harmonic Fourier series was fitted to the data, to give a quantifiable line of best fit (Fig. 1). Strain was calculated, from the Fourier-fitted data, as the fractional length change relative to resting length. Resting length was defined as the mean of the maximum and minimum lengths recorded(Gabaldon et al., 2004). Strain rate was determined as the first differential of strain. Data from each condition were partitioned into 20 equal time windows and mean strain and strain rate calculated for each time window.

A filter bank of 20 non-linearly scaled wavelets, indexed by 0⩽k⩽19, were used to decompose the myoelectric signals into their intensities, as a function of time and frequency. Each wavelet domain was described by its frequency bandwidth, centre frequency and time resolution using the methods described by (von Tscharner, 2000). As previous analysis of Fourier-transform-derived power spectra had revealed a quantity of low frequency noise (<100 Hz) in the signal, the first four wavelet domains were excluded from further analysis(Hodson-Tole and Wakeling,2007). The frequency band 69.92–1325.00 Hz (wavelets 4⩽k⩽19) are therefore presented in the analysis here. This ensures that signals from slow motor units [183.3±7.9 Hz(Wakeling and Syme, 2002)] are included in the analysis, and fits well with the cut-off frequency used in other fine-wire myoelectric studies (Daley and Biewener, 2003; Gabaldon et al., 2004; Gillis and Biewener, 2001; Gillis and Biewener, 2002).

Myoelectric intensity was calculated at each wavelet domain, at each time point, from the magnitude and the first-time derivative of the square of the convoluted signal (von Tscharner,2000). The total intensity of the signal at a given time is given by summing the intensities over all the wavelets. As with the sonomicrometry data, myoelectric signals were partitioned into complete strides, based on the kinematics video data (total number of strides analysed in soleus: 4373;plantaris: 3533; medial gastrocnemius: 4388). These data were then partitioned into 20 equal time-windows and the mean intensity for each time window calculated (Fig. 2).

To determine time periods when muscles were active and when they were not,the mean myoelectric intensity was calculated for a 10% stride duration window over which muscle activity was at its lowest point (55–65% stride duration) in each condition. The lowest mean value was found in the medial gastrocnemius muscle (0.0001) and twice this was used as the threshold value. This represented 13.33% of the maximum mean myoelectric intensity value recorded in all the muscles (0.0015). When myoelectric intensity was above this value the muscle was categorised as active, when intensity was below this value the muscle was deemed to be inactive. Within each muscle, data points in which myoelectric intensity was below the threshold value were not included in any further analysis.

The data set were aligned into a p×N data matrix A, where p=16 dimensions (wavelet domains) and N=245 740 spectra included in the analysis (12 287 total number of strides×20 partitioned time windows). The principal components, defined in terms of eigenvector-eigenvalue pairs, were then calculated from the covariance matrix of the data matrix A. Calculations were made on the total intensity data without prior subtraction of the mean, to ensure that the whole signal would be described and not just its variance(Wakeling and Rozitis, 2004). The first few principal components were able to explain a large proportion of the original spectra (PCI 89.64%; PCII 5.07%; PCIII 1.47%; PCIV 0.89%), making it possible to express the spectra in fewer terms than the original suite of wavelets used (Wakeling and Rozitis,2004).

Each myoelectric spectrum can be visualised by its principal component loading score (Ramsay and Silverman,1997; Wakeling,2004) with the magnitude of the PCI loading score indicating the level of myoelectric intensity and the angle formed between the PCI and PCII loading scores (θ) providing a quantitative measure of the frequency content of the myoelectric signal(Wakeling and Rozitis, 2004). A small θ, caused by a positive PCII loading score relative to the PCI loading score, indicates a proportionally higher amount of high frequency content in the signal. A large θ, related to a negative PCII loading score relative to the PCI loading score, indicates a proportionally higher amount of low frequency signal content. Mean PCI and PCII loading scores were calculated for each of the sectioned portions of the stride, enabling changes in their relative contributions to be defined for different time points within each stride.

From the principal component analysis, θ was used as a measure of relative frequency content within the myoelectric signal, as defined above. To identify conditions in which measurements were consistent with the predictions of the size principle the association between θ and myoelectric intensity was assessed. To identify conditions in which there was a mechanical basis for the change in myoelectric frequency content the association betweenθ and muscle fascicle strain rate was assessed. To test this association it was important to ensure that the results were not confounded by any electromechanical delay. Partitioning the strides into 20 equal time windows meant that each point represented 5% of stride duration. In the fastest strides (0°, 50 cm s^–1), where stride duration was 299.8±5.1 ms, each window represented 14.99 ms. The time between the occurrence of a muscle action potential and the beginning of force contraction has been reported as 1.11±0.14 ms in fast and 2.82±0.16 ms in slow rat motor units (Rannou et al.,2007), therefore, any relationship identified between strain rate and myoelectric frequency content would not be affected by this factor. Assessments were initially made on data points from all time windows within all conditions. In addition, data points were categorised as occurring in early stance, late stance, early swing or late swing phases, with stance and swing durations defined on the basis of kinematics data, and the split into early or late phases being exactly half the duration of each. This strategy enabled us to determine if the association(s) between myoelectric intensity and θ, and between muscle fascicle strain rate and θ, vary as the functional demands placed on the musculoskeletal system vary over the time course of a stride (Kaya et al.,2005).

As sonomicrometric and myoelectric data were not simultaneously collected from the same muscles within an individual, means of all the data for each muscle during each condition were calculated and used for statistical analyses. Differences in myoelectric intensity and muscle fascicle strain and strain rate between muscles and between stride phases were identified using general linear model ANOVA, with muscle, time window and stride phase identified as fixed factors in each case. When significant differences were identified, Bonferroni post hoc tests were applied to the data set to determine the location(s) of the differences. The presence of a significant association between θ and intensity, and θ and muscle fascicle strain rate were determined using general linear model ANCOVA. The test was initially applied to the whole data set with muscle and time window defined as fixed factors and covariates defined as muscle fascicle strain and either myoelectric intensity or muscle fascicle strain rate. Muscle fascicle strain was included in the analysis to ensure that results were not confounded by changes in strain, which have been shown to affect myoelectric frequency content (Doud and Walsh, 1995). As greater strains are associated with a decrease in frequency content, if a positive association was found between muscle fascicle strain and θ(i.e. greater strains associated with lower signal frequency content) the results were discarded and not analysed any further. To identify differences in the relationships between θ and intensity, and θ and muscle fascicle strain rate, that occurred across the time course of a stride within each muscle, general linear model ANCOVA were applied to data grouped according to muscle and stride phase. For these tests, time window was defined as a fixed factor with covariates defined as muscle fascicle strain and either myoelectric intensity or muscle fascicle strain rate. In cases where a significant association between θ and intensity or θ and muscle fascicle strain rate occurred Pearson product moment correlations were used to identify the strength and direction of the association. Model II linear regression was then used to determine the line of best fit for those data(Sokal and Rohlf, 2000). Significant differences between the slopes of the calculated regression lines for each condition and each muscle were identified using an analysis of covariance, a test of equality described by Sokal and Rohlf(Sokal and Rohlf, 2000). In all cases differences were judged to be significant when P⩽0.05.

In each locomotor condition, each muscle showed one burst of myoelectric activity, which generally began during the swing phase and ended late in the stance phase of the following stride [fig. 1 in (Hodson-Tole and Wakeling, 2008)]. There were significant differences in myoelectric intensity between muscles and stride phases. The soleus and plantaris muscles had significantly lower mean intensity values than the medial gastrocnemius muscle (GLM ANOVA: P<0.001; Bonferroni post hoc: P<0.001 both cases), but did not differ from each other. Significant differences in intensity occurred between each of the stride phases (GLM ANOVA: P<0.001; Bonferroni post hoc: P<0.001 all cases), with the highest values occurring during initial stance and late swing phases(7×10^–4±2×10^–5mV², N=147;6×10^–4±3×10^–5mV², N=123, respectively). Values for late stance(4×10^–4±2×10^–5mV², N=135) and early swing(2×10^–4±1×10^–5mV², N=108) were much smaller. In addition there was a significant positive correlation between myoelectric intensity and the PCI loading score (P<0.001, r=0.95). Mean muscle fascicle strain rate did not differ significantly between the muscles. There was,however, a significant difference between strain rates recorded during the two stance phase periods and those recorded during the two swing phase periods(GLM ANOVA: P<0.001; Bonferroni post hoc: P<0.002 all cases). Mean values showed muscle fascicle shortening strain rates were predominant during the stance phases (initial stance:–0.24±0.03 s^–1, N=147; late stance:–0.06±0.03 s^–1, N=135) with lengthening strain rates occurring during the swing phase (initial swing: 0.23±0.06 s^–1, N=108; late swing: 0.168±0.07 s^–1, N=123). Muscle fascicle strain did not differ significantly between late stance (–0.0129±0.002, N=135)and early swing (–0.0138±0.004, N=108) phases, but significant differences did occur between all other stride phases (initial stance: 0.0048±0.002, N=147; late swing: 0.0145±0.003, N=123; GLM ANOVA: P<0.001; Bonferroni post hoc: P<0.001 all cases). There were no significant differences in muscle fascicle strain between muscles.

General linear model ANCOVA showed that for the whole data set there was a significant association between myoelectric intensity and θ(P<0.001), with Pearson product moment correlation showing that r=–0.59. There was no association between θ and strain(P=0.582), so changes in frequency content were not confounded by changes in muscle length. No significant association between θ and myoelectric intensity occurred during the initial stance (P=0.880) or initial swing (P=0.269) phases(Fig. 3). Significant associations did, however, occur during the late stance phase(P=0.002) and late swing phases (P<0.001; Fig. 3). Pearson product moment correlation showed that both of these relationships were significantly negative (late stance P<0.001, r=–0.30; late swing P<0.001, r=–0.63). Comparison of the model II linear regression lines showed that there was no significant difference in the slope of the lines.

When muscles were assessed individually, results for the plantaris muscle were confounded by changes in strain in all the stride phases except the late swing phase. In this phase, general linear model ANCOVA showed a significant association between θ and myoelectric intensity (P<0.001),with Pearson product moment correlation showing the association to be significantly negative (P<0.001, r=–0.82). Results for the soleus muscle were not confounded by changes in strain, with general linear model ANCOVA finding significant associations during the initial stance(P=0.004), late stance (P=0.049) and late swing phases(P<0.001). Pearson product moment correlation found a significant negative association during late stance (P=0.002, r=–0.41) and late swing (P<0.001, r=–0.85), but did not identify an association during the initial stance phase (P=0.244, r=0.17). A similar pattern was also seen in the medial gastrocnemius muscle. There were significant associations between θ and myoelectric intensity during the early stance(P=0.003), late stance (P=0.026) and late swing(P<0.001) phases. Pearson product moment correlation results showed a negative association during late stance (P<0.001, r=–0.46) and late swing (P<0.001, r=–0.83) phases.

For the whole data set a significant association existed between muscle fascicle strain rate and θ (P=0.008), but not between strain and θ (P=0.860). This again meant that changes in θ could be directly attributed to changes in myoelectric frequency content, and were not confounded by changes in muscle fascicle length. Pearson product moment correlation showed that the association between strain rate and θ was positive, with r=0.25. General linear model ANCOVA showed that there was a significant association between θ and muscle fascicle strain rate during the late stance (P<0.001) and late swing(P<0.001) phases (Fig. 4). The association identified during the initial swing phase was confounded by changes in strain and was therefore discounted. Pearson product moment correlation showed a significant positive association during the late stance phase (P=0.001, r=0.26), but did not quantify the relationship during the late swing phase(Fig. 4). There were no significant differences between the slopes of the lines.

When individual muscles were assessed, the plantaris muscle had a significant association between θ and muscle fascicle strain rate during the late stance (P<0.001) and late swing (P=0.002)phases. The associations were shown to be positive in both phases (late stance P<0.001, r=0.53; late swing P=0.041, r=0.32). In the soleus muscle, general linear model ANCOVA found a significant association between θ and muscle fascicle strain rate during the initial and late stance phases (P<0.001; P<0.001,respectively). Pearson product moment correlation quantified a significant negative association during the initial stance phase (P<0.001, r=–0.54), and a significant positive association during the late stance phase (P<0.001, r=0.63). A significant association was identified between muscle fascicle strain rate and θduring the late swing phase, but was discarded due to the significant association identified between θ and muscle fascicle strain. In the medial gastrocnemius muscle general linear model ANCOVA revealed a significant association between θ and muscle fascicle strain rate during the initial and late swing phases (P<0.001; P<0.001,respectively). In both instances Pearson product moment correlation revealed the association to be positive (initial swing P=0.018, r=0.39; late swing P<0.001, r=0.63).

The above results showed striking differences in the association between myoelectric frequency and myoelectric intensity and muscle fascicle strain rate, occurring over the time course of a stride. Relative signal frequency content was quantified by θ. Larger values of θ represented relatively more low frequency content, and smaller values represented relatively more high frequency content. The significant negative association between θ and myoelectric intensity during the late stance and swing phases therefore represented a positive association between myoelectric signal frequency content and myoelectric intensity(Table 2). An association between θ and muscle fascicle strain rate was identified in each of the stride phases (Table 3). In all except one instance, the association was positive. As shortening strain rates are represented by negative values and smaller values of θ represent a shift to higher frequency signal components, this relationship indicated that,as predicted, there was a positive association between higher frequency signals and faster shortening strain rates.

Several factors must be taken into consideration when interpreting myoelectric signals. Signal frequency content can be affected by several factors, which must be controlled for if valid interpretations are to be made. In the present study, bias, which would have been introduced by changes in muscle temperature and fatigue status, was minimised by using a three-block randomised exercise protocol. Changes in signal frequency content, which occur as a result of changes in muscle fascicle length(Doud and Walsh, 1995), were controlled for by including strain as a covariate in the statistical analyses conducted. Any results where strain was a significant positive factor (i.e. longer lengths associated with greater θ values and hence relatively lower frequency content) were discarded. The fact that strain was a positive factor in some instances does not mean that there was not a significant association between θ and either myoelectric intensity or muscle fascicle strain rate. A conservative approach has therefore been taken to overcome this problem, and despite this some interesting results have been identified. It has been suggested that higher myoelectric frequencies are produced by faster motor units, as a result of their larger diameter. We have been unable to find any in vivo experimental evidence that supports this claim. By contrast, the electrical properties of the sarcolemma have been shown to vary between fast and slow fibre types within mammals(Luff and Atwood, 1972). As these properties will determine the conduction velocity of an action potential(Buchtal et al., 1955), it can be predicted that faster motor units will have faster conduction velocities and hence generate higher myoelectric frequencies(Gerdle et al., 2000; Wakeling et al., 2002). We therefore interpret that a higher value of θ, representing relatively more low frequency signal content, can be associated with the recruitment of slower motor units. A smaller θ value, associated with relatively more high frequency content, can be associated with the recruitment of faster motor units.

Motor unit recruitment strategies have been extensively investigated since Henneman and co-workers first proposed the size principle theory of recruitment. The quantity and quality of the work put forward to test the predictions of the size principle are testament to its robustness and stature. The majority of studies to date have focussed on identifying an association between recruitment order and either the motoneuron portion or the muscle fibre portion of the motor unit. In work focussing on the muscle fibres, motor unit size has been typically estimated based on either the twitch amplitude or contraction speed or the amplitude of the motor unit action potential. In the present study motor unit recruitment has been quantified by the analysis of the interference patterns of the action potentials from active motor units. In agreement with the large body of work that exists to support the size principle, our results show that there was a positive association between recruitment of faster motor units and myoelectric intensity. The association was, however, only significant during the late stance and swing phases(Table 2). In each case the strength of the association was similar between muscles, with much higher r values found during the late swing phase. When comparing θand muscle fascicle strain rate the late stance and swing phases were again identified to have significant associations. There was, however, much more variation between r values reported for the individual muscles, and both the initial stance and swing phases also had a significant association within one muscle (Table 3). The influence of muscle fascicle strain rate on motor unit recruitment therefore appears to be a much more variable factor than myoelectric intensity within the muscles studied here. It should be noted that, as sonomicrometric and myoelectric data were not simultaneously recorded from individual muscles,assessment here was based on mean values. This is a limitation in our study design and is likely to reduce the amount of correlation recorded. Despite this, we have been able to show that significant correlations do exist between muscle fascicle strain rates and the recruitment of faster motor units, and we suggest that these associations are likely to be greater if measured simultaneously within an individual.

The late swing and stance phases, respectively, represent the main periods of motor unit recruitment and de-recruitment, represented by the rise and fall of myoelectric activity (Hodson-Tole and Wakeling, 2008). The finding that motor unit recruitment during these two phases is influenced by myoelectric intensity and muscle fascicle strain rate may therefore not be surprising. It is possible that once a muscle is initially activated, using the size principle, modulation of force production occurs on the basis of the mechanical properties of the motor units. Previous work has shown that different combinations of motor units are recruited to produce the same or similar myoelectric intensities during different time points of a stride(Hodson-Tole and Wakeling,2007; Hodson-Tole and Wakeling, 2008; Wakeling,2004; Wakeling et al.,2001; Wakeling et al.,2006). Muscles are therefore able to develop and maintain a given myoelectric intensity using a number of combinations of motor units. One of the proposed functional advantages of the size principle is that it provides a strategy by which a smooth increment in force magnitude can be achieved. Larger motor units have been reported to contain larger numbers of muscle fibres and are hence capable of producing more force than smaller motor units(Milner-Brown et al., 1973). Orderly recruitment would therefore mean that the force increment, as a proportion of the force being generated, would always be similar(Zajac and Faden, 1985). Maintenance of a particular force magnitude could, however, be achieved by selectively de-recruiting active, slower motor units whilst maintaining or increasing activity in faster motor units. Such a mechanism is possible within vertebrates because of the presence of Renshaw cells in the ventral horn of the spinal chord. These are inhibitory neurons that regulate the firing rate of α-motoneurons, causing the recruitment of faster motor units to have a disfacilitatory influence on already active slower motor units(Broman et al., 1985).

Analysis of data from individual muscles enabled comparisons of the influence of myoelectric intensity and muscle fascicle strain rate on motor unit recruitment to be made between populations of distinctly different muscle fibre types. The soleus muscle is predominantly composed of slow, MHC type I fibres, the plantaris has a mixed population of MHC type I, IIA and IIB fibres, whereas the medial gastrocnemius, in the area data were collected from, is predominantly composed of MHC type IIB fibres(Armstrong and Phelps, 1984). We were, therefore, able to determine whether recruitment strategies remained the same between these fibre type populations. The relationship betweenθ and both myoelectric intensity and muscle fascicle strain rate varied across the time course of the stride within each muscle. Distinct differences were also apparent between the muscles, indicating some differential control occurring between them.

During the late stance phase motor unit recruitment in both the soleus and plantaris muscles was influenced more by muscle fascicle strain rate than myoelectric intensity. The medial gastrocnemius had a significant association between muscle fascicle strain rate and motor unit recruitment during the initial swing phase, which was the only significant association identified. In these instances it is likely that preferential recruitment of faster motor units provides a mechanical advantage. Rome et al.(Rome et al., 1988) suggested that generating mechanical power at a high efficiency is best achieved by using faster motor units for faster tasks. The strain rates reported here,however, fall short of values estimated to produce maximum mechanical output in rodent muscle (Swoap et al.,1997). This may reflect differences between intrinsic speed measures taken in vivo compared to those taken in situ, or may reflect the range of velocities incorporated in this study (20–50 cm s^–1), which included walk and trot gaits but not the faster gallop. In addition, it must also be considered that the values reported will have been influenced by the definition of the resting length. Here it was defined as the mean of the maximum and minimum lengths recorded in each stride, following previously reported methods(Gabaldon et al., 2004). Ideally, the length should correspond with the plateau region of the stress–strain relationship, which was not determined here. Although this will not affect the patterns of change in strain rate over time, the absolute values will have been influenced.

Analysis of muscle fascicle strain rate data during the initial stance phase resulted in the identification of a negative association between θand strain rate in the soleus muscle. This result was unexpected, as it indicates that slower motor units were recruited in response to faster shortening strain rates. The highest myoelectric intensities recorded in this muscle occurred during the initial stance phase in the fastest locomotor conditions (Hodson-Tole and Wakeling,2008). The limited range of muscle fibre types within the soleus muscle and its limited size may therefore force it to recruit as many motor units as possible in an attempt to meet the force requirements of the stride phase. Alternatively, this trend may reflect the differences in deactivation times that exist between slow and fast fibre types. Myoelectric activity was high during the initial stance phase and dropped off to its lowest points during the late stance phase (Hodson-Tole and Wakeling, 2008). Slower muscle fibres, with their longer deactivation times, may therefore need to be deactivated during the initial stance phase to ensure they had sufficient time for force production to end. By contrast, during the late stance phase the association between θ and muscle fascicle strain rate in the soleus was positive and one of the highest found (r=0.63). Such a change can only highlight the different force demands and loads that must be placed on an individual muscle during the course of a stride and the complex pattern of control that must occur.

Myoelectric signals were specifically collected from the area of the medial gastrocnemius that has been reported as being predominantly composed of fast MHC type II fibres (Armstrong and Phelps,1984). It has previously been reported that the correlation between motor unit axonal conduction velocity and fused tetanic tension is weak when data are collected from large mixed fibre muscles(Burke et al., 1973; Burke and Rymer, 1976; Fleshman et al., 1981; Proske and Waite, 1976; Stephens and Stuart, 1975). From these reports it appears that the lack of correlation is most apparent when numerous fast contracting motor units are present. In the present study a significant association between myoelectric intensity and motor unit recruitment in the medial gastrocnemius was present during the late stance and swing phases, and was strongest during the late swing phase. In parallel with this there was also a significant association between motor unit recruitment and muscle fascicle strain rate within this muscle during the initial and late swing phases. Again this was particularly strong during the late swing phase. It would therefore appear that in this population of fast contracting fibre types myoelectric intensity and muscle fascicle strain rate both play a significant role in determining motor unit recruitment. It is probable that the population of fibre types within this muscle derive a greater mechanical advantage (in terms of mechanical power output) from preferential recruitment of faster motor units. Indeed, in humans it has been found that during cycling there is a significant association between the recruitment of faster motor units and maximum muscle fascicle strain rate in the medial gastrocnemius muscle that is not found in the soleus or lateral gastrocnemius muscles(Wakeling et al., 2006).

Although it has been noted that generating mechanical power during faster tasks is more efficiently achieved by faster motor units(Rome et al., 1988), it should also be considered that the mechanical behaviour of a muscle is not fixed and can alter in response to changes in locomotor condition. For example, some muscles, for example in the turkey(Gabaldon et al., 2004; Roberts et al., 1997) undergo much greater strains during incline locomotion than during locomotion on the flat. Other muscles have been shown to adapt to changes in locomotor conditions by changing the timing of myoelectric activity in relation to force production and hence adapting mechanical work output(Daley and Biewener, 2003; Gabaldon et al., 2004). Motor unit recruitment patterns have also been shown to change in response to changes locomotor velocity and incline(Hodson-Tole and Wakeling,2008) and it is therefore probable that the association(s) between motor unit recruitment patterns and muscle fascicle strain characteristics will vary during different locomotor tasks. This serves to indicate that current knowledge of musculoskeletal biomechanics is limited and is an area that warrants further investigation.

The results show that θ, a measure of motor unit recruitment, is significantly and differently related to myoelectric intensity and muscle fascicle strain rate. Motor unit recruitment must therefore be a multifactorial phenomenon. Previous work has shown that high and low values ofθ can be associated with myoelectric activity in populations of slow and fast motor units, respectively(Hodson-Tole and Wakeling,2008). Our results therefore indicate that there were times when motor unit recruitment was either related to the predictions of the size principle (θ vs myoelectric intensity) or had a mechanical basis (θ vs muscle fascicle strain rate). The predictions of the size principle did not hold true for all periods of the stride cycle. In addition, some periods of activity were accounted for by a combination of size principle and mechanical factors. The change in recruitment strategies that were found across the time course of the stride may reflect different functional demands that are placed on the muscles as the limb cycles between stance and swing phase positions. This supports the suggestion that motor units may form task groups, which are selectively recruited for different kinematic conditions within a stride(Loeb, 1985; Von Tscharner and Goepfert,2006; Wakeling,2004; Wakeling et al.,2001). This is only the second report we are aware of that has identified a second influential factor related to motor unit recruitment, and the first example in rat muscle. Further work is needed to ascertain how widespread the phenomenon of mechanically driven motor unit recruitment strategies is between species and locomotor tasks.

The authors would like to thank Michael Boyd and John Thurlborne for their assistance with surgical procedures and the care of the animals. Karin Jespers and Pattama Ritruechai were crucial to the data collection process; the use of sonomicrometry equipment was by kind permission of Prof. Alan Wilson.