The fossil material comprises the cast section of tracks G1, G2, and G3. Three prints of G1 (37, 36, 35) and one print of G3 (26) were used for geometric morphometry.

The living material comprises the tracks of nine adult humans (four males, five females) of European origin, and the track of one adult female chimpanzee (Table 1). The chimpanzee (Tiby, 14 years old, 45 kg) is a hybrid of Pan paniscus (father) and P. troglodytes (mother). The animal was born in captivity, lives with humans and other chimpanzees (Kino's Circus) and adopted a permanent bipedalism during experiment. In total, 62 human prints, and five chimpanzee prints have been measured. Because of the small number of fossil prints (4), we reduced the human sample to 9 prints (one per individual) to have a mean shape (consensus) that is intermediary between humans and fossils. We calculated two possible shapes for each fossil print and we found that differences in shape between two data captures were negligible in multivariate space.

The walkway (length: 5 m; breadth: 45 cm; height: 4 cm) was made of thin wet clay. We maintained constant humidity to obtain a print depth comparable to those made in Laetoli ground. Measurement for ground density is not available because print depth also depends on individual characteristics, such as velocity, body weight, body proportions, sex, and peculiar type of foot stance. In human experiment, a video camera (24 frames/s) was placed perpendicular to the long axis of the walkway at a distance of 3.5 m (lateral view of two strides). In chimpanzee experiment, three wide angle video cameras (24 frames/s) were situated at 3.5, 6, and 4 m, to obtain lateral, posterior and superior views of the complete walkway respectively. We experimented with five or six consecutive trails to try to obtain different types of walk (but to no avail). As previously noticed in Pan paniscus [34], footfall pattern is relatively stable for each individual, with some differences between right and left feet. Finally, we retained a trail with five measurable footprints, i.e. with hallux and lateral-toes contact. Other prints with no toes contact were used for observation and other calculations.

For each individual, we calculated hindlimb length (greater trochanter to heel contact), stride velocity, stride length (stride: between two initial heel contacts of the same foot), stride length relative to hindlimb length, foot angle (between foot symmetrical axis and track direction), feet gap (between two successive heel posterior landmarks measured perpendicularly to the track direction), and Froude number (Table 1). The Froude number (Fr) allows us to compare gaits in terms of dynamic similarity [2], [3] and [25]. Calculation is the following one: Fr = V ² g ^–1 L ^–1, where V is velocity during one stride, g gravitational acceleration, and L hindlimb length.

Hindlimb length was measured at heel contact on films. Hindlimb was extended in humans, and flexed in chimpanzee [see also 11]. For Laetoli, we calculated an approximate hindlimb length from the femur length of Australopithecus afarensis Lucy (AL 288), considering that australopithecines had an extended hindlimb at heel contact, and a knee-to-foot/ femur ratio similar to humans [4] and [5].

Eighteen homologous landmarks have been defined in humans and Laetoli (Table 1, Fig. 1). Twelve of them draw the outline of prints (landmarks 1 to 12), and six (landmarks 13–18) body weight transfer. Because the shape of prints is very different in our chimpanzee, the number of landmarks was reduced to 14 for the chimpanzee comparison (Table 1). Landmarks were digitalized with the 3D Revpro DX Microscribe (precision: 0.1 mm).

Two analyses were carried out from Laetoli prints, comparing them either with humans (18 landmarks), or with the chimpanzee (14 landmarks). For each analysis, and after Procrustes superimposition, a principal-component analysis was computed from Procrustes residuals. We calculated the discriminant vector that best separates the two groups (Laetoli vs. humans, Laetoli vs. chimpanzee) in the multivariate space of shape (see [6], [23] and [24]). Shape changes along discriminant vectors may be viewed in three planes: horizontal, sagittal and frontal. Here they are given in horizontal plane (superior view of prints) and sagittal plane (lateral view of prints). To compare distances from Laetoli to humans and from Laetoli to chimpanzees, we computed an additional Laetoli-humans analysis with 14 landmarks, and we calculated the Mahalanobis D ² in the two 14-landmark analyses. Superimposition, graphics, parametric tests were calculated with APS Software [22], and D ² with Matlab 5.3.

Bipedal gait on a soft ground leads to various styles of walk in humans (Table 2). Froude numbers vary from 0.01 to 0.25, indicating that there is no real dynamic similarity (velocity is not proportional to hindlimb length) within humans walking in such conditions. Stride length relative to hindlimb length is also very variable (SL/HL: 120.0 to 166.7). In comparison to humans, the chimpanzee walked with a very large stride length, high relative velocity, and consequently a large Froude number (see other locomotor parameters in [1], [10], [17], [18] and [21]). In proportion to its short (and flexed) hindlimb, stride is particularly long (SL/HL: 218.2). Similar values have been measured in Pan paniscus walking in normal conditions of bipedalism [1]. On the contrary, stride was very short in Laetoli printmakers. When associated to Lucy hindlimb reconstruction, Laetoli stride length turns out to be very human-like in proportions (SL/HL: 137.5).

There is considerable difference between humans and chimpanzees in the way body weight is displaced during bipedal walk (Table 2). The orthograde bipedal walk of humans is characterized by a reduced displacement of body weight from one supporting foot to the other, with reduced rotational movements of pelvis and shoulders around the vertebral column, and alternate arm swinging. In soft ground as here, we observed that stride lengths and velocity may vary greatly, although body weight transfer is relatively invariant. In humans, foot contact begins with the heel, continues with lateral edge of foot, instantly followed by metatarsal heads of digits (ball of the foot) and extremities of digits (hallux and lateral toes). Digits act together at ground contact and toe-off. At ground contact on a soft subtract, human digits are flexed. At toe-off, human digits are passively hyperextended under the effect of body weigh although they are submitted to flexor muscles (the muscles flexors digitorum and hallucis proprius are strongly stretched). On a soft ground, humans energetically push on hallux and lateral toes to propel themselves.

In our chimpanzee experiment, the walk in soft ground was not very different from descriptions of normal walks made on hard ground [1], [10], [11], [15], [17] and [34]. The chimpanzee walked with feet apart, high velocity, and wide stride lengths (Table 2). We observed on films that stride length is increased by the forward displacement of the supporting hip. Body weight is strongly laterally displaced alternately on each supporting foot, with wide swinging movements of arms in the opposite direction, and without movements of shoulders and head. Initial contact of the foot with the ground begins with the heel. Roll-off involves displacement of the centre of pressure along the lateral edge of the foot, down to the 5th metatarsal head. Then, later, the centre of pressure shifts towards the extremity of the hallux. In some cases observed here, foot contact stopped after lateral edge of mid-foot, digits leaving no print on the ground. At foot-off, the chimpanzee preserves flexed or curled-underneath toes, the propulsive force being generated mainly by heel and midfoot. Here toe-off is very variable (hallux first or hallux and lateral toes simultaneously or lateral toes first).

The Procrustes analysis uses 18 homologous landmarks to compare Laetoli with human footprints. Calculation of a discriminant vector between humans and fossils indicates that the two patterns are clearly different (R ² = 0.95, F = 41.6, p < 2 × 10^–5, with 4 PC). Fig. 2 gives the coordinates of specimens in terms of shape differences and multivariate size. Prints of G1 (35-36-37) and G3 (26) have a similar shape but different sizes. As compared with the consensus (mean shape), humans show a narrow footprint characterized by a marked narrowing at midfoot, closer hallux and toe II, and lateral toes shorter than the hallux (Fig. 3A). Centres of pressure are aligned from heel to lateral toes. The distance between the hallux centre of pressure and the lateral-toes centre of pressure is short, not only because toe II and hallux are close together, but also because the maximal pressure corresponding to the lateral-toes falls frequently on toe II (sometimes between toe II and III; in one case on toe IV). In lateral view, the plantar vault is well-marked (landmark 14 higher than landmarks 13, and 15–16). Body weight pressure is more marked at the level of metatarsal heads (ball of the foot), and digit extremities (hallux and lateral-toes centres of pressure) than at the level of the heel. Laetoli footprints display the opposite shape (Fig. 3B). In superior view, the print is very broad with no narrowing at midfoot, separated hallux and toe II, and lateral toes longer than hallux. The successive centres of pressure indicate that body weight pressure is internal at heel contact, then laterally displaced to the external edge of the foot and toes. We also observe that pressure on metatarsal heads is proportionally lower situated, and pressure on hallux and lateral toes more spaced than in humans. The lateral view of the print indicates that fossils had a small plantar vault, with a stronger pressure on heel than on metatarsal heads. Pressure was smaller on hallux and lateral toes than in humans. In frontal view (not represented here), lateral toes left a deeper print in the ground than hallux, whereas it is generally the reverse in humans.

The Procrustes analysis uses 14 homologous landmarks to compare Laetoli with the chimpanzee footprints. Discrimination is very high (R ² = 0.99, F = 311.6, p = 10^–6 with 2 PC). In comparison to the consensus, the chimpanzee footprint is much larger at the level of digits than at the level of heel (Fig. 3C). The chimpanzee walked with flexed or curled-underneath toes, which are superimposed on the lateral edge. Thus, maximal pressure corresponding to lateral toes is on phalanges 1 or 2 of toe V, placed above toe extremity. Centres of pressure from heel to lateral toes are situated on the lateral edge of the foot. The hallux centre of pressure is further from lateral toes' centre of pressure. In lateral view, the print of the chimpanzee has no plantar vault. Only heel print is deep, sometimes hallux extremity. In comparison, the Laetoli footprint is narrower at the level of digits (Fig. 3D). Body weight transfer is much more internal from midfoot to toes' extremities. Hallux and lateral-toes centres of pressure are close together. In lateral view, the print is not completely flat as in the chimpanzee but indicates the presence of a vault with a marked pressure at the level of metatarsal heads (ball of the foot) and extremities of digits (hallux and lateral toes).

We calculated the Malahanobis D ² with 4 PC in the two 14-landmark analyses (Laetoli-humans, Laetoli-chimpanzee). The distance between Laetoli and humans corresponds to D ² = 95.4, whereas the distance between Laetoli and chimpanzee corresponds to D ² = 615.7, that is to say six times more distant from the chimpanzee than from humans. In Table 2, gait parameters seem to increase the likeness between Laetoli hominids and modern humans. Contrary to the chimpanzee, Laetoli hominids walked with small stride lengths (absolute and relative measurements), and close feet.