Palms are important components of tropical forest plant communities, due both to their abundance (Henderson et al. Reference HENDERSON, FISCHER, SCARIOT, WHITAKER PACHECO and PARDINI2000) and to the network of interactions with their pollinators and dispersers (Henderson Reference HENDERSON2002, Zona & Henderson Reference ZONA and HENDERSON1989). Forest fragmentation alters the biotic and abiotic conditions of habitats (Ewers & Didham Reference EWERS and DIDHAM2006, Fahrig Reference FAHRIG2003) and it has been observed that Attalea palms increase their densities in disturbed sites (Aguiar & Tabarelli Reference AGUIAR and TABARELLI2009, Andreazzi et al. Reference ANDREAZZI, PIMENTA, PIRES, FERNANDEZ, OLIVEIRA-SANTOS and MENEZES2012, Lorenzi et al. Reference LORENZI, SOUZA, MEDEIROS-COSTA, CERQUEIRA and FERREIRA2004). Increased light availability (Salm Reference SALM2005, Souza & Martins Reference SOUZA and MARTINS2004), changes in seed dispersal and predation patterns (Andreazzi et al. Reference ANDREAZZI, PIMENTA, PIRES, FERNANDEZ, OLIVEIRA-SANTOS and MENEZES2012, Pimentel & Tabarelli Reference PIMENTEL and TABARELLI2004, Wright et al. Reference WRIGHT, ZEBALLOS, DOMÍNGUEZ, GALLARDO, MORENO and IBAÑEZ2000), and ability to recover after disturbance (Souza & Martins Reference SOUZA and MARTINS2004) are among the main mechanisms that have been proposed to explain enhanced palm densities. However, how altered conditions following disturbances influence the dynamics of flower and fruit production is still little understood.

Attalea palms have a complex reproductive system in which male, female and mixed inflorescences are produced according to the size and age of the adult individuals (Henderson Reference HENDERSON2002). Temporal differentiation in the expression of unisexual flowers affects the reproductive output and can result in phenological differences in space and time (Henderson et al. Reference HENDERSON, FISCHER, SCARIOT, WHITAKER PACHECO and PARDINI2000). Forest fragmentation modifies most of the proximate environmental cues (e.g. water stress, radiation) that regulate plant phenology and reproductive patterns (Laurance et al. Reference LAURANCE, RANKIN-DE MERONA, ANDRADE, LAURANCE, D'ANGELO, LOVEJOY and VASCONCELOS2003). Increased light availability, higher temperatures and reduced humidity are usually found in small fragments, mostly due to edge effects and reduced canopy height and cover (Brothers & Spingarn Reference BROTHERS and SPINGARN1992, Murcia Reference MURCIA1995). For this reason we may expect that small forest fragments should present distinct phenological patterns as compared with larger undisturbed sites.

Attalea humilis Mart. ex. Spreng. (Arecaceae) is a monoecious and solitary palm, endemic to the central-south-eastern region of the Brazilian Atlantic Forest (Lorenzi et al. Reference LORENZI, SOUZA, MEDEIROS-COSTA, CERQUEIRA and FERREIRA2004). Populations of this large-seeded palm seem to be favoured in small forest fragments (< 60 ha) in south-east Brazil due to the higher fruit production per individual, associated with reduced seed predation (Andreazzi et al. Reference ANDREAZZI, PIMENTA, PIRES, FERNANDEZ, OLIVEIRA-SANTOS and MENEZES2012). In the present study we analysed the reproductive phenology of A. humilis in the same sites sampled by Andreazzi et al. (Reference ANDREAZZI, PIMENTA, PIRES, FERNANDEZ, OLIVEIRA-SANTOS and MENEZES2012) in order to investigate which mechanisms could explain the differences in productivity between small and large fragments. We expected that the mean dates of flowering and fruiting should not differ between the two classes of fragments because the timing of these events should be influenced by macroclimatic factors such as rainfall and photoperiod (Haman Reference HAMAN2004, van Schaik et al. Reference VAN SCHAIK, TERBORGH and WRIGHT1993) which vary at larger spatial scales. However, the duration of flowering and the production of female inflorescences should be higher for palms in small fragments, resulting in a greater fruit production.

The study was carried out at two large (> 2000 ha) and two small (< 60 ha), lowland Atlantic Forest remnants located in south-eastern Brazil. The climate is tropical humid, with mean annual temperature 25.5 °C. Mean annual precipitation is 1995 mm and there is a moderate seasonality in rainfall with a less rainy period from April to September. All areas sampled are within a 30-km radius and located on small hills, mostly with clay soils, at low elevations (< 100 m). Additional information about them can be found in Andreazzi et al. (Reference ANDREAZZI, PIMENTA, PIRES, FERNANDEZ, OLIVEIRA-SANTOS and MENEZES2012).

The reproductive phenology of A. humilis was evaluated in each fragment along a pre-existing 1-km linear transect, in which 20 adult palms (individuals that were reproducing or had signs of previous reproductive events) were marked and checked from March 2006 to February 2008. Every month the numbers of spathes, male and female inflorescences, immature and mature infructescences and fruits were directly counted.

The proportion of flowering and fruiting palms was compared between small and large fragments using chi-square tests. Deviations from the expected ratio of 1:1 in the number of male and female inflorescences observed in small and large fragments were tested using the Wilcoxon signed-ranks test for paired samples. The seasonality of the reproductive events was analysed using circular statistics. For each circular histogram, a mean vector length (r) was calculated which represents the concentration of the phenological events around the mean angle; the vector length varies from 0 to 1, zero indicating absence of seasonality (Zar Reference ZAR1999). The mean dates of male and female flowering and fruit branches maturation and the concentration of these events around the mean were calculated for small and large fragments.

The Rayleigh test was used to test the hypothesis that flowering and fruiting in each fragment class were uniformly distributed over the year. When seasonality of a phenological event was detected for both fragment classes, the Watson–Williams test for two samples (F_ww) was used to test if it differed between large and small fragments (Zar Reference ZAR1999). Analyses were performed using the software package ORIANA (Kovach Computing Services, Pentraeth, UK).

Small fragments had higher proportions of palms producing inflorescences of both sexes (χ² = 4.94, df = 1, P = 0.03) and infructescences (χ² = 17.3, df = 1, P < 0.0001). Inflorescence development time – from the emergence of inflorescence buds to flower anthesis – ranged from 2 to 6 mo (3.69 ± 1.32 mo; n = 39) and did not differ between small and large fragments (Welch's t = −0.739, df = 37, P = 0.467). Male inflorescences were more frequent in both large (Wilcoxon signed-ranks test rank sums for positive and negative differences: T+ = 213, T− = −18.0, P = 0.0002) and small fragments (T+ = 383, T− = −112, P = 0.0062), but this bias was more marked in large ones (Figure 1). Inflorescences of this sex were more frequent in the wet season (Table 1, Figure 1) and the mean date of this phenophase did not differ between fragment classes (F_ww = 0.043, P = 0.836). Female inflorescences corresponded to 30% of the total in small fragments, but only to 18% in large ones (Table 1). In small fragments female inflorescences were produced throughout the year without difference between seasons (P = 0.084), while in large fragments flowering was observed only during the wet season (Table 1, Figure 1).

Table 1. Mean dates, lengths of the mean vectors (r), concentration and Rayleigh test results for Attalea humilis male and female inflorescences and ripe fruit branches in small and large Atlantic Forest fragments in northern Rio de Janeiro. N is the number of reproductive events recorded (infructescences or inflorescences).

Figure 1. Circular histograms of Attalea humilis male (a, b) and female inflorescences (c, d) and ripe fruit branches (e, f) from March 2006 to February 2008 in small and large Atlantic Forest fragments in northern Rio de Janeiro. The arrow represents the length of mean vector. The wet season is from October to March.

Green fruits took 3–9 mo (6.87 ± 1.60 mo; n = 15) to mature, independent of fragment size (Welch's t = 0.363, df = 13, P = 0.731). Infructescences harbouring ripe fruits showed no seasonal pattern in any fragment (Table 1), but they were more frequent and abundant in small fragments (Figure 1). Besides, branches had four times more fruits in small fragments, resulting in the 16-fold increase in fruit production reported by Andreazzi et al. (Reference ANDREAZZI, PIMENTA, PIRES, FERNANDEZ, OLIVEIRA-SANTOS and MENEZES2012).

Reproduction of A. humilis was recorded throughout the year, inflorescences were produced more often in the warmer and wettest months in both fragment classes, as observed in other studies on Attalea (Almeida & Galetti Reference ALMEIDA and GALETTI2007, De Steven et al. Reference DE STEVEN, WINDSOR, PUTZ and DE LEON1987, Henderson et al. Reference HENDERSON, FISCHER, SCARIOT, WHITAKER PACHECO and PARDINI2000, Voeks Reference VOEKS2002). Although the number of records in large fragments was not large enough to warrant firm conclusions, the main difference between small and large fragments in our study was that in dry months female flowers were produced only in small ones. As the canopy is more open in these fragments (Carvalho et al. Reference CARVALHO, NASCIMENTO and BRAGA2006, Reference CARVALHO, NASCIMENTO and OLIVEIRA-FILHO2008) and palms are closer to forest edges, light incidence would be enough to allow female flower production even with the shorter photoperiod of the dry season, whereas, in large ones, female flower production could be afforded only in the months with most light. In fact, other studies have shown that palm individuals exposed to higher light incidence produce more female flowers (Barot et al. Reference BAROT, MITJA, MIRANDA, MEIJA and GRIMALDI2005, Berry & Gorchov Reference BERRY and GORCHOV2006, Cunningham Reference CUNNINGHAM1997, Scariot et al. Reference SCARIOT, LLERAS and HAY1995, Voeks Reference VOEKS1988, Reference VOEKS2002). The production of male inflorescences, however, is possible in all fragments even in the months with shorter photoperiod because pollen is cheaper to produce than fruits.

Our results indicate that forest fragmentation can have a great influence on palm reproductive dynamics. The altered abiotic conditions of small fragments resulted in increased and longer female flowering, which boost the reproductive output of A. humilis in these places. This explains the increased A. humilis densities in the small fragments studied here (Andreazzi et al. Reference ANDREAZZI, PIMENTA, PIRES, FERNANDEZ, OLIVEIRA-SANTOS and MENEZES2012) and could favour this palm in other disturbed sites. The same process can occur with other palm species which become dominant in altered landscapes (Lorenzi et al. Reference LORENZI, SOUZA, MEDEIROS-COSTA, CERQUEIRA and FERREIRA2004) or in natural palm patches (Aguiar & Tabarelli Reference AGUIAR and TABARELLI2009, Barot et al. Reference BAROT, MITJA, MIRANDA, MEIJA and GRIMALDI2005, Fragoso Reference FRAGOSO, Burslem, Pinard and Hartley2005). Studies suggested that some forest tree species can also increase their reproductive output following disturbance (Cunningham Reference CUNNINGHAM2000), thus similar results could be found for other plant families. These findings are important for understanding changes in plant species composition and abundance following fragmentation. Most studies on this subject have focused on the disruption of animal–plant interactions, especially seed dispersal and pollination, in small fragments. However, our results show that the importance of abiotic factors cannot be neglected when investigating the net effects of fragmentation on plant regeneration. Therefore, changes in plant abundance in small fragments can be better understood as the result of a complex interplay of the whole set of factors, both biotic and abiotic, in which a fragment differs from the original forest.