The studies were approved by the University of Nottingham Medical School Ethics Committee. Initially, 20 female medical undergraduate volunteers, 19–21 years of age, were recruited from Nottingham University Medical School. The purpose of the study was explained both verbally and in writing and all recruits were given the opportunity to ask further questions before self-consenting to participate. Each volunteer completed a questionnaire which sought information about the onset date of last menstruation (to estimate their menstrual cycle stage) and current oral contraceptive usage (to account for variations in hormone concentrations between students). All subjects were classed as dextral (right-handed) on the basis of volunteered informed about hand preference. The duration of the study was eight weeks to cover two separate menstrual cycles for every individual. One volunteer was withdrawn from the study due to lack of a suitably comprehensive dataset. Eventually, students were divided into two groups: those not taking oral contraceptives (the –OC group, n = 13) and those taking various types of oral contraceptive (+OC group, n = 6).
In order to obtain digit length measurements for all participants, photocopies of left and right hands were made. Photocopying has been shown to be effective in previous human 2D:4D studies (Manning et al. 2001; Peters et al. 2002; Rahman & Wilson, 2003). The method produces measurements which correlate well with those obtained directly from the hands with callipers (Robinson & Manning, 2000) and tends to give lower 2D:4D ratios but with enhanced inter-observer reliability (Peters et al. 2002; Manning et al. 2005). At least for the present longitudinal studies on the same digits, any relative biases should be equal and comparisons should retain their comparative worth. However, it has been suggested recently that photocopies may distort left–right asymmetries (Manning et al. 2006). Clearly, further studies are required, including direct measurements on digits, to test the possibility of such biases influencing outcome measures. For consistency, photocopies were made according to a strict protocol designed on the basis of a pilot study, which compared the effect of different hand positions on digit length measurements.
The pilot study was designed to test the null hypothesis that wrist position (abducted, adducted or straight) does not alter 2D or 4D length measurements for an individual. Twelve independent participants (n = 6 males and n = 6 females) were selected randomly from a set of medical undergraduates in the library of the University of Nottingham Medical School. The purpose of the study was explained and each student consented to partake. Any student who had fractured, suffered trauma to, or had arthritis in, their 2D or 4D was excluded and students were asked to remove any rings from these digits.
The photocopier contrast was altered initially so that the proximal crease and distal finger tip were clearly visible. Three photocopies were taken from each student, the first with both hands in full abduction (radial deviation) at the wrist, the second with both hands in full adduction (ulnar deviation) and the third with a straight wrist, such that a line drawn along the length of the third digit extended between the radius and the ulna. In each position, the lengths of 2D and 4D were measured from the midpoint of the proximal crease, to the distal finger tip point, using digital callipers accurate to 0.01 mm. Differences between the 2D and 4D measurements were calculated for each of the three photocopies. For any individual, the minimum difference between photocopies was 0.05 mm and the maximum difference was 3.00 mm.
Despite the small sample sizes, significant differences between length measurements of the same digit were found for the three different hand positions. These indicated that abduction at the wrist favoured the 2D and adduction favoured the 4D. In a larger study (see Peters et al. 2002), similar significant differences were established. On this basis, the third straight wrist position was selected for use in this study. This allowed us to avoid favouring either the 2D or 4D and maintained consistency between participants.
Having established a satisfactory protocol, we embarked upon the definitive study. The photocopier contrast was adjusted to a setting which produced a clear hand image, where both the proximal crease and distal finger tips of 2D and 4D were visible and distinct from the background. The female participants were asked to remove any rings or jewellery that might hinder measurements of 2D and 4D lengths from the proximal crease to the finger tip. Any participants who had previously fractured, suffered trauma, or had arthritis in the relevant digits were excluded from the study, along with participants who had been diagnosed with a condition affecting hormone concentrations, bone growth, or oedema. This was intended to remove potential confounding variables from the study.
Participants were asked to place the palmar surface of both hands on the photocopier, making full contact with the photocopy plate but without applying firm pressure. Adopting the method used by Peters et al. (2002), participants were instructed to hold their three middle digits straight and adducted (without overlapping) but with first and fifth digits slightly splayed. Their upper limbs were extended at the elbow with the third digit in line with the midpoint of the radius and ulna. The photocopier lid was lowered to avoid excessive light exposure and to improve image quality. After photocopying, the image quality was checked and, if necessary, the photocopying was repeated. All photocopies were on Mondays and Thursdays between 1–2 pm to minimise possible diurnal and circadian variations in digit lengths. Participants remained anonymous and the only other information taken was their age and hand preference. This was recorded on the reverse of the photocopy to conceal the information from the measurer.
Identical digit length measurements were taken from every photocopy, using digital callipers accurate to 0.01 mm. To eliminate bias due to inter-observer errors, all digits were measured by one observer. The lengths of 2D and 4D were measured from the midpoint of each proximal crease to the distal finger tip. These measurements were used to calculate the 2D:4D ratio (obtained by dividing 2D by 4D length) and the 2D-4D difference (obtained by subtracting 4D from 2D). Another measure, referred to as the ‘distal tip extent’ (DTE) was taken. In effect, this affords a different method of estimating the 2D-4D difference and was adapted from the ‘tip measure’ described by Manning et al. (2000). A line was drawn between the midpoints of the 2D and 4D proximal creases to account for hand shape. Two further lines were then drawn parallel to this, one at the point of the 2D distal finger tip and one at the 4D distal finger tip. The distance between the 2D and 4D distal tip lines was then measured at 90° to the lines, giving the ‘distal finger tip extent’. The purpose of using both the DTE and 2D-4D difference as outcome measures was to assess the comparative worth of alternative digit measures.
Where appropriate, measurements were averaged over both cycles so as to account for possible within-subject cycle-to-cycle differences.
The mean, standard deviation (SD) and standard error of mean (SEM) were calculated for each group of subjects. To better describe subject-to-subject variation within groups, the observed coefficient of variation was calculated (oCV = SD/mean). Comparisons between digits on the same hand, and between the same digits on left and right hands, were drawn using paired Student's t tests (Sokal & Rohlf, 1981) and based on measurements obtained from both cycles. To test for differences on the same hand during different stages of the cycle, we applied Page's L trends tests for related samples (Miller, 1975). To this end, the menstrual cycle was divided into 7 stages each of which comprised an equal interval of time (4 days). Thus, the first stage of the cycle (marked from the onset of menstruation) covered days 1–4 and the last covered days 25–28. All statistical tests were undertaken using Unistat v5.5 software (Unistat Ltd, London, UK). For all comparisons, the corresponding null hypothesis was rejected at the P < 0.05 level of significance.