We can detect laterality in the manufacture of stepped tools (Fig. 2a) by New Caledonian crows (Corvus moneduloides) because this differs according to whether the left or the right leaf-edge is used1. Birds carry out a well-defined sequence of precise cutting and ripping actions on both edges, with bill action being away from the tree trunk: they must therefore work from either right to left (left edges) or left to right (right edges). The right side of the head is generally towards the leaf edge in left-edge usage, and the left side in right-edge usage. A tool's finished shape is recorded in its unmistakable 'counterpart', the outline remaining on the leaf edge (Fig. 2a).

Figure 2: Lateralization in the manufacture of stepped tools from Pandanus sp. leaves by New Caledonian crows.
figure 2

a, A stepped tool (arrow), together with its counterpart shape on the left edge of a section of pandanus leaf (about 5 cm wide). Pandanus leaf rips easily along the strong, longitudinally parallel fibres, but must be cut laterally across the fibres. The tool is removed from right to left, beginning at the pointed end. Stepped-tool counterparts averaged 18.5 ± 6.30 cm long (mean ± s.d.m., n = 3,418) and 0.84 ± 0.24 cm wide (n = 3,128). b, The number of tool counterparts on the left (blue bars) and right (green bars) edges of leaves for each leaf-spiral direction. A 2 × 2 contingency test on the data (edge × spiral direction) found a significant access effect due to leaf-spiral direction (G-test of in- dependence: G1 = 330.3, P < 0.0001, n = 3,727), but there still exists a highly significant preference (χ21 = 430.3, P < 0.0001, n = 3,727) for manufacturing tools on left edges (n = 2,463) rather than right edges (n = 1,264). As individual birds are likely to have contributed more than one counterpart, this sample size is inflated. A conservative alternative is to treat each site as the unit of analysis. A Wilcoxon signed-ranks, matched-pairs test showed that the left-edged bias was still significant across the 19 sites (T = 44, two-tailed P < 0.05, n = 19). Sample sizes for tool counterparts varied between sites (196.2 ± 120.8, n = 3,727), but the number of trees sampled with clockwise (n = 369) and anticlockwise (n = 364) leaves were almost identical (38.6 ± 24.4 trees per site, n = 733).

We collected 3,727 stepped-tool counterparts from 733 pandanus trees at 19 sites scattered throughout mainland Grande Terre, New Caledonia, the only island where crows are known to make these tools. The sites were mostly at least 10 km apart and distributed along 300 km of the narrow, 400-km-long island at altitudes of 160–1,600 m above sea level. This sampling covered the geographical range of pandanus-toolmaking (G. R. H., personal observation).

The crow's use of left or right leaf-edges depends in part on the direction in which the leaves spiral. Clockwise-spiralling leaves provide easier access to left edges, and anticlockwise-spiralling leaves provide easier access to right edges. This access effect was overridden, however, by an island-wide preference for manufacturing tools from left edges (Fig. 2b).

Population-level footedness in manipulating food while feeding is known in parrots4. In great apes, population-level handedness has been reported in several manipulative activities in wild gorillas5 (Gorilla g. beringei) and in captive chimpanzees (Pan troglodytes)6,7,8, but has not been shown so far in chimpanzees in the wild9,10. Although there have been no systematic comparisons, it has been suggested that the level of complexity of a motor task and the associated neural-processing demands may correlate with the degree of lateralization8,11.

The rather simple nature of chimpanzees' tool-making contrasts with the greater complexity of crows' stepped-tool manufacture1. The shape of stepped tools is highly regular and mostly determined by the crows themselves, rather than by the raw material, implying that a neural programme is involved in their manufacture1. Laterality in crows may therefore reflect the complexity of stepped-toolmaking and a specialization of the right-eye/left-hemisphere system for non-spatial, sequential actions, as previously demonstrated in birds and mammals8.

It has been proposed that right-handedness in humans may be a consequence of the evolution of language, which is also predominantly left-hemispheric3. Our results favour the more general possibility that species-level lateralization is an adaptation for the efficient neural programming of complex sequential processing8,12, of which language and right-handedness in humans and stepped-tool manufacture in crows are examples.