Sufficient seed numbers were obtained from three to five mother plants (Table 1) of five very common species (Anemopaegma oligoneuron, Arrabidaea trialii, Mimosa guilandinae, Passiflora nitida, Senna sp. 2) for an analysis with statistically meaningful replicates. These species are mainly found in disturbed areas or secondary regrowth [23]. For all other species seeds were only obtained from one mother plant and were therefore pseudoreplicates or had a very small number of replicates (Table 1). The five common species all had desiccation tolerant seeds; as was revealed in the germination test and the probability calculation for desiccation sensitivity P(D-S) following Daws et al. [22], which is based on the seed coat ratio and seed mass; all P(D-S) values were <0.5. Light was not necessary for germination: the relative light germination (RLG, [24]), which is the ratio of germination in light and the sum of germination in light and darkness, was around 0.5 (Table 2). Alternating temperatures inhibited germination in P. nitida, but did not affect the other four species (Table 1). Except P. nitida, all species had germinated seeds within two weeks. The range of seed mass varied only five-fold, and seed shape was mainly flat or oval-flat, three species had winged dispersal units or seeds (Table 2). The results of these five species will be discussed in context with the other studied species.

A wide variety of morphological traits and physiological responses to environmental factors were found among the 20 study species. The range in dry seed mass covered three orders of magnitude from 0.0085 g (Matelea badilloi) to 11 g (Mucuna sp.). The length of the longest seed axes (wingless species) differed eight-fold in the extremes, from 5.1 mm (Cissus sicyoides) to 43.2 mm (Gnetum nodiflorum, Table 2). Seed moisture content at dispersal for all species ranged between 13–68% and was not significantly correlated to dry seed mass (Pearson’s R = 0.22 P = 0.45, data of seed mass log transformed).

Amongst all species, the time until the first seed showed radical protrusion at the constant temperature of 25 °C was between two and 155 days; in a similar time range 50% of all germinated seeds had developed a radical (2–157 days, Table 2). No meaningful pattern in germination velocity related to biometric data could be found. The type of dormancy that delayed germination was not determined in this study. For P. nitida, which was one of the slowest germinating species in this study, a dormancy period of several months and low final germination are known [26].

For eight species final germination of the seeds at constant 25 °C with light was ≥90%. The lowest final germination was found in P. nitida (20%) and G. nodiflorum (7%) (Table 1). Seed viability of all seeds at the end of the germination trials (excluding the desiccation treatment) was >91% for six species, 78–88% for four species (Coccoloba sp., M. badilloi, M. guilandinae, Strychnos glabra), and two species could not be tested (P. nitida, G. nodiflorum).

We assessed desiccation tolerance using two techniques (germination trials and probability equation based on biometric data [22]), which should add reliability to our results. In fact, results of both techniques were consistent for 18 out of 20 species (Table 2, Figure 1). The ten species that have smaller seeds (up to 0.07 g dry seed mass) germinated after desiccation to a moisture content <9% (3.7–8.8%) and had a calculated probability of desiccation sensitivity P(D-S) < 0.5 (Table 2). These species all belong to families or genera with seeds that are known to be tolerant to desiccation, and species records exist for Cissus sicyoides [20]. An intermediate desiccation tolerance (seeds tolerate desiccation not below 5% moisture content, neither temperatures below 0 °C, [27]), as found in some Passiflora species, e.g., P. edulis [28], could not be excluded for Passiflora acuminata or P. nitida in our experimental design.

Eight species with larger seeds (dry seed mass > 0.27 g) and P(D-S) > 0.5 (Anomospermum solimoesanum, G. nodiflorum, Mucuna sp., Paullinia rugosa, Paullinia capreolata, Smilax sp., Strychnos amazonica, S. glabra) did not germinate after desiccation and seeds were dead. This corresponds with the finding of Hong and Ellis [29] that the percentage of species with desiccation sensitive seeds increases with increasing seed size.

Desiccation sensitive seeds have been recorded for Paullinia cupana (Sapindaceae) [30], a well studied species of economic importance in the Amazon region, which supports our finding in the same genus for P. capreolata. There are no records of desiccation sensitive seeds for the genera Smilax, Strychnos, Gnetum and Mucuna, but there are records of some desiccation tolerant congeners [20,31]. No seed information was found for the genus Anomospermum, the few species investigated from the same family (Menispermaceae) all had desiccation tolerant seeds [20]. The mentioned records of desiccation tolerant congeners often included smaller seeded species and some are from temperate regions. Desiccation sensitive seeds are most common in the tropics, and, therefore, the probability of finding a species with desiccation sensitive seeds of a certain genus or family is higher in tropical than in temperate regions [32].

Seed shape was correlated to the probability of desiccation sensitivity: flat seeds all had low P(D-S) values, rounder seeds had a wider range of probability values; but overall the correlation was positive (Spearman’s rho = 0.62, p = 0.003). Hong and Ellis [33] found seed shape is associated with desiccation tolerance within the Meliaceae family: desiccation sensitive species being round, desiccation tolerant species being generally flat.

Another trait related to desiccation sensitivity is seed moisture content at dispersal [33]. However, no relationship between moisture content and desiccation sensitivity was found in the tested species. As seeds were weighed immediately after collection or extraction from the fruits, fresh weight and moisture content of the winged or hair tuft (coma) bearing species (A. oligoneuron, Anemopaegma floridum, A. trailii, Aristolochia silvatica, M. badilloi) might be slightly overestimated, as these seeds would probably dry some hours or days later in the opened fruit before wind dispersal. In the case of Senna sp. 2 (35% moisture content), it can be assumed that seeds were still water permeable during collection and might have dried more after dispersal. Hong and Ellis [29] also mentioned that predicting desiccation tolerance is not possible for seed moisture content in the range of 25% to 55% at dispersal time. Desiccation sensitive seeds are expected to shed during the rainy season to avoid any risk of severe water loss. In our study, seeds of five desiccation sensitive species were shed at the end or beginning of the rainy season (June, July and November) and three shed during the rainy season (April, May). Desiccation tolerant seeds were shed during the rainy season (3 species), the dry season (4 species) at the transition from wet to dry season (2 spesies), and also over several months (3 species).

Two species (Coccoloba sp. and A. floridum) revealed contradicting results in the germination test and calculated probability of desiccation tolerance: they did not germinate after desiccation (0% germination) but should be desiccation tolerant according to the calculated probability P (D-S) = 0.22 and 0.06, respectively. The phylogenetic relationship of both species suggests they could have desiccation tolerant seeds; only species with desiccation tolerant seeds are known from the Bignoniaceae family (A. floridum) (80 species records, [20]), and for Polygonaceae (Coccoloba sp.) so far only one tree with desiccation sensitive seeds (Triplaris cumingiana) has been recorded, but 159 species with desiccation tolerant seeds are known [20,34]. The lack of germination after drying might be the result of using immature seeds in the treatments, which may not yet have acquired the tolerance to desiccation at the end of maturation drying [35]. In doubtful cases, the calculation of P (D-S) after Daws et al. [22] seemed to be more reliable for determining desiccation tolerance than germination tests, and can also support uncommon findings, e.g., that the Mucuna species is desiccation sensitive, where smaller seeded congeners are usually known for desiccation tolerant seeds and hard seed coats [36]. Biometric data may be more robust considering the difficulty in recognizing the state of seed maturation at harvest of species with unknown phenology.

Two species (M. badilloi 0.009 g, A. silvatica 0.03 g) needed light for germination (photoblastic seeds), RLG was ≥ 0.99, as there was no germination in darkness and final germination in light was 61% and 90% respectively. C. sicyoides (0.02 g), with 25% germination in darkness and 93% in light, had a RLG of 0.79. The high RLG suggest that seeds of these three species could be considered photoblastic. Replication size was sufficient (at least 3 × 20 seeds), yet results cannot be considered as totally reliable since seeds were harvested from only one liana plant per species. To our knowledge this is the first record indicating photoblastism of A. silvatica and M. badilloi. Photoblastism of C. sicyoides had already been documented in an earlier field study [10]. All three species had seeds with 0.009–0.03 g dry mass. The chance that light is required soon after germination to support growth for photosynthetic self sufficiency is higher in small seeded species with restricted resources than in large seeded species. Thus with increasing seed size, the share of species with light depending seed germination is reduced as shown for different growth forms (herbaceous plants, trees) in tropical and temperate zones [17,24,37].

Four more species had RLG ≥ 0.67 (Table 2), but replicate number (A. floridum, A. solimoesanum, P. acuminataTable 1) and/or germination of the control was very low (G. nodiflorum, one seed) and do not allow an affirmation of light dependence of germination.

The understory of old growth tropical forests present a habitat with small daily temperature alternations compared to open areas such as gaps [17], borders or secondary forests. In this study, the different environments were represented by a constant temperature and two regimes of alternating temperatures. Germination at alternating temperatures was tested with 14 species. In general, high alternating temperatures (15/35 °C) had no influence or reduced germination; low alternating temperatures (20/30 °C) induced various responses (Table 1). Replicates and seed numbers were sufficient for statistical analysis for only nine species (≥3 replicates, ≥60 seeds per treatment, Table 1, Figure 2). Of these nine species, final germination differed significantly among treatments in four species. P. nitida germinated only at constant 25 °C. A. silvatica and Coccoloba sp. reached the same final germination result under both a constant temperature and small daily temperature cycle (30/20 °C); however, with the larger daily temperature cycles (35/15 °C) germination dropped to 0% and 5%, respectively. Germination increased in only one case (by 1.6-fold in M. badilloi, 9 mg dry seed mass) under alternating temperatures of 20/30 °C (Table 1), but no germination occurred under large temperature cycling (35/15 °C). Increased germination at alternating temperatures has also been reported for a related vine, Matelea maritima (setting: 25/35 °C, [38]). No related records were found for the other three species.

In former studies, the influence of alternating temperature on germination was related to seed size: Small seeded species (0.04–0.68 mg) were found to be intolerant to temperature fluctuations whereas for relatively larger seeds (>2 mg) germination rate increased with temperature fluctuations [17]. Ecologically, this intolerance may prevent small seeded species germinating in unfavourable conditions for seedling establishment, e.g., high irradiation and drought in large clearings, though these species still depend on smaller vegetation gaps for seedling establishment [17]. However, no specific pattern of germination response to temperature fluctuations was found with increasing seed resources (embryo and endosperm dry mass) in our study. This was probably because the range of seed sizes in our study did not include species with extremely small seeds (<1 mg).

As summarized by Probert [39] temperature cycles stimulated germination for many species and were often linked to photoblastism. Alternating temperatures may enhance stimulation through light or even enable dark germination of initially photoblastic seeds. However, in the present study one species with photoblastic seeds (C. sicyoides) tolerated alternating temperatures and the two others did not (M. badilloi, A. silvatica); the combination of dark germination and alternating temperatures was not tested here. This indicates that species of the same growth form with photoblastic seeds still differ in their niche of regeneration.

Mature fruits and seeds of 20 liana species belonging to 12 families were collected close to their natural time of dispersal in the surrounding area and up to 80 km north of Manaus (3°–2.25° S, 60°–59.5° W). The climate is tropical-humid, with 2,285 mm annual rainfall [40], and a dry season from June to October. Species were identified by comparing specimens at the herbarium of the Instituto Nacional de Pesquisas da Amazônia (INPA), Brazil. Names follow the database TROPICOS [26] (Table 1). Some of the species are subwoody-woody persistent climbers and one is a climbing shrub (Senna sp. 2), but they were nevertheless included here, following the liana census of Gerwing [41]. Seed collection was restricted to one plant for most species, but seeds were obtained (used here in a broader sense for diaspore) from three to five individual plants at different time points for six species (A. oligoneuron, A. solimoesanum, A. trailii, M. guilandinae, P. nitida, Senna sp. 2, Table 1). The time between collection and assessment of biometry or implementation of the germination tests did not exceed 10 days, during this time fruits and seeds were maintained at 15 °C. Immediately after extraction from the fruits, where succulent pulp was eliminated if necessary, biometric measurements were made with at least 30 seeds. For two species fewer seeds were available (Table 2). Maximal length, width and height were measured with a digital caliper (0.01 mm). Fresh seeds were weighed individually (0.001g) and subsequently dried until constant dry mass was reached, at 105 °C for 24–62 h (depending on seed size) and weighed again. Seed moisture content was calculated as the percentage of fresh mass. Seed coat ratio (SCR; seed coat mass versus whole seed mass) was determined by dissecting 7–20 dried seeds.

Seeds were dried in two sub-samples in an air conditioned room (25 ± 4 °C, 60 ± 10% relative humidity) above a fan between five to eight days until a constant mass was reached. Afterwards seeds were dried for another seven days over the same amount of silica gel as seed mass in three layers of hermetically closed plastic bags. Seed moisture after desiccation was determined for a part of the seeds (n= 4–50), as described above, and the other seeds were re-hydrated slowly for seven days in an atmosphere with saturated relative humidity, avoiding imbibition damage by direct contact to water [18], before the assessment of germination. A species was classified as desiccation sensitive when all dried seeds in the germination test were dead.

Seeds were sown on water saturated vermiculite (fine or medium grain, Eucatex Agro^®) in transparent plastic containers (11 cm × 11 cm × 3 cm), or in Petri dishes (9 cm diameter) for small seeded species (Senna sp. 2, Acacia multipinnata) with vermiculite as substrate. The containers were wrapped in loosely closed transparent plastic bags to avoid desiccation. Three temperature regimes were provided, two incubators for each temperature regime were used (FANEM Mod. 347 CDG São Paulo, Brazil and LMS cooled incubator, York, UK): (i) constant temperature at 25 °C with either a 12 h photoperiod (white fluorescent light with approximately 70 µmol m⁻² s⁻¹ PAR) or in complete darkness, (ii) 12/12 h alternating temperatures of 20/30 °C and (iii) 12/12 h alternating temperatures of 15/35 °C, the period of higher temperature coincided with the 12 h light period. The average thermal input for all three temperature conditions was 25 °C. For the dark treatment, the containers were wrapped in two layers of aluminum foil and germination was assessed only after the germination curve under light conditions reached an asymptote, for slow germinating species the substrate was remoistened in dim green light. For one slow germinating species (P. nitida), seedlings in the dark treatment died, but were still accurately countable. Dried seeds were only tested at constant 25° C with 12 h light. The number of repetitions and seeds per repetition depended on seed availability and ranged between 2 × 3 seeds to 25 × 10 seeds per treatment (Table 1). Germination in light was assessed after radical protrusion (<1 mm). Observations were made daily, three times per week, or weekly, depending on the germination speed of the species. Seeds that did not germinate were tested for viability at the end of the germination trials via a staining (2,3,5 triphenyl tetrazolium chloride, 1% solution) or cutting test.

The probability of desiccation sensitivity or tolerance P (D-S) was calculated following Daws et al. [22]. This equation (1) includes data of the seed coat ratio (SCR, in formula: a) and the log10 value of the whole seed mass (b). If P (D-S) is > 0.5 seeds are probably desiccation sensitive.

The requirement for light during germination was calculated by using the relative light germination index (RLG) based on Milberg [30]. The index was calculated as RLG = Gl/(Gd + Gl) where Gl = the germination percentage in light, and Gd = the germination percentage in darkness. RLG = 1 when germination is obligate light depending. The ratio of shortest to longest seed axis was used to create an index for seed shape (1 = round seed, towards 0 = flat seed).

Differences in germination between control and alternating temperatures were tested via Kruskal Wallis Rank Sum Test and subsequent pair wise Wilcoxon Test. In one case (A. silvatica, alternating temperature), ANOVA and subsequent Tukey Post Hoc Test with arcsine transformed values was used instead of Kruskal-Wallis Rank Sum Test, due to small replicate numbers (n = 3). For correlation, Pearson’s correlation was used, or Spearman rank correlation in case of non-normal distribution of data. Statistical analyses were carried out with R-2.13.2 (R Core Development Team 2011).