Adult specimens used in this study were reared from galls in small plastic vials, which were maintained under controlled conditions (14 h light/10 h dark; 25 ± 2 °C) in the Laboratório de Morfologia e Comportamento de Insetos (LMCI), Departamento de Zoologia, Universidade Federal do Rio Grande do Sul (UFRGS), Porto Alegre city, RS. Dehiscent galls (approx. 20 in total) were collected from the ground, in the surroundings of S. weinmannifolius plants of an old grass field, located in a farm belonging to Antonio Malta, Coxilha das Lombas, 30°01'46"S, 50°36'40"W, 86m, 29.V.2012, Santo Antônio da Patrulha Municipality, Rio Grande do Sul State (RS), Brazil. Pupae were obtained later (September) by dissecting some galls under a stereomicroscope in the laboratory. Larvae were obtained by dissecting S. weinmannifolius branches, either from galls located under swollen bark (early instars) or erupted from the stem (later instars). Adults were pin-mounted and dried. Immature stages were fixed in Dietrich´s fluid and preserved in 75% ethanol. Larvae used for DNA extraction came from several additional populations (listed below), and were preserved in 100% ethanol at -20 °C.

For descriptions of adult morphology the specimens were cleared in a 10% potassium hydroxide (KOH) solution, stained with Chlorazol black E and slide-mounted in either glycerine jelly or Canada balsam. Last instar larvae were prepared similarly for description of chaetotaxy. Observations were performed with the aid of a Leica® M125 stereomicroscope. Structures selected to be drawn were previously photographed with a Sony® Cyber-shot DSC-H10 digital camera attached to the stereomicroscope. Vectorised line drawings were then made with the software Corel Photo-Paint® X7, using the corresponding digitalized images as a guide. Additional specimens were used for scanning electron microscope analyses. They were dehydrated in a Bal-tec® CPD030 critical-point dryer, mounted with double-sided tape on metal stubs, coated with gold in a Bal-tec® SCD050 sputter coater and examined and photographed in a JEOL® JSM6060 scanning electron microscope at the Centro de Microscopia Eletrônica (CME) of UFRGS.

Mitochondrial and nuclear DNA sequences were used for two different levels of analysis of the undescribed genus and species: 1) to infer the phylogenetic status and relationships within Cecidosidae, and 2) to describe the genetic diversity and population structure of this rare taxon. For the first approach we used representative species of all members of Cecidosidae except Xanadoses, the corresponding samples coming from the tissue collection of LMCI: i.e., C. eremita, Dicranoses congregatella Kieffer & Jörgensen, Eucecidoses minutanus Brèthes, O. argentinana, an undescribed lineage from Chile (previously known to be closely related based on morphology) and Scyrotis (Scyrotis sp. and S. granosa Meyrick), a genus from Africa included for the first time in a molecular phylogeny. For the second approach we sampled 10 populations across the distribution range of Cecidonius pampeanus sp. n. (P1 to P10), including six individuals per site (Suppl. materials 3, 5). Previous analyses indicated there was no substantial addition of genetic variation by increasing corresponding sample size. Total genomic DNA was purified from fresh collected larval tissue of all Cecidosidae surveyed except Scyrotis (dried museum adult specimens were used), using the PureLink genomic DNA kit (Life, Invitrogen, USA) following the manufacturer’s instructions.

For cecidosid phylogeny we used nucleotide sequences obtained from different molecular markers, selected because they evolve at different rates and provide phylogenetic resolution at different, overlapping taxonomic levels: two mitochondrial (1421 bp of the cytochrome oxidase subunit I [COI] and 474 bp of the 16S ribosomal RNA [16S] genes), and one nuclear (395 bp of the Wingless [Wg] gene) loci. For the genetic structure and variability approach, we amplified COI in 60 individuals, six from each population sampled. The selected molecular markers were amplified by polymerase chain reaction (PCR); primers and conditions used are described in the supplementary material (Suppl. material 1). PCR products were purified using the enzymatic method (exonuclease and alkaline phosphatase), sequenced with BigDye chemistry, and analysed in an ABI3730XL (Applied Biosystems Inc.). Chromatograms obtained from the automatic sequencer were read and sequences were assembled using the software CodonCode Aligner (CodonCode Corporation). Sequences generated in this study were deposited in the GenBank database (Suppl. material 3).

Sequence data were used for the reconstruction of a concatenated phylogenetic tree (COI+16S+Wg) with the Bayesian method in BEAST 2.02 (Bouckaert et al. 2014). The tree prior was set as a Yule calibrated process, using GTR + I for the COI partition and TN92+G for both 16S and Wg, selected with the Bayesian information criterion (BIC; Schwarz 1978) for each data set in jModelTest 2.1.2 (Darriba et al. 2012). The branch lengths were allowed to vary under a relaxed clock model with an uncorrelated log-normal distribution (Drummond et al. 2006). To adjust the molecular clock we used the fossil calibration point of Adeloidea (sensu van Nieukerken et al. 2011), about 120±10mya, with a log-normal distribution (Walhberg et al. 2013). The analysis was run for 10,000,000 Markov Chain Monte Carlo (MCMC) cycles and parameters were sampled every 1,000 cycles; this was repeated four times to test for MCMC convergence, and priors exceeded 200 to ensure effective sample sizes (ESS). Burn-in was determined in Tracer 1.5 (Drummond and Rambaut 2007) based on ESS and parameter trajectories, and the first 20% of trees were then removed with TreeAnnotator. Trees were observed and edited in FigTree v1.3.1 (Rambaut 2009). Clades with Bayesian Posterior Probability (BPP) ≥ 95% were considered strongly supported. Pairwise genetic distances (p-distances) among lineages were calculated in MEGA 7 (Tamura et al. 2013).

Nucleotide and haplotype (gene) diversity indices were estimated for individuals grouped into ten populations (P1 to P10) with DnaSP 5.1 (Librado and Rozas 2009). To investigate the evolutionary relationships among COI-haplotypes a median-joining haplotype network (Bandelt et al.1999) was constructed in NETWORK 5 (http://www.fluxus-engineering.com/sharenet.htm). Levels of genetic structure among populations were characterized using φST with Arlequin3.5 (Excoffier and Lischer 2010). Additionally, to investigate spatial patterns of genetic structure we assessed the correlation between genetic and geographic distances for all pairs of sampled individuals using a Mantel test (Mantel 1967). We also performed an Analysis of Molecular Variance (AMOVA; Excoffier et al. 1992) with Arlequin to assess more detailed quantitative differentiation among populations, performing two rounds of AMOVA employing different geographic clustering strategies: i) taking into account the vicariate effect of the Jacui River, and ii) using the genetic distance and haplotype relationship results. Finally, we investigated the demographic history of the new genus and species using neutrality tests (Tajima’s D, Fu & Li’ D* and F*, Fu’s Fs) and mismatch distribution analysis (Rogers and Harpending 1992, Schneider and Excoffier 1999) with DnaSP.

Abbreviations of the institutions from which specimens were examined are:

DZUP Coll. Padre Jesus S. Moure, Departamento de Zoologia, Universidade Federal do Paraná, Curitiba, Paraná, Brazil.

LMCI Laboratório de Morfologia e Comportamento de Insetos, Universidade Federal do Rio Grande do Sul, Porto Alegre, Rio Grande do Sul, Brazil.

The phylogenetic reconstruction corroborated our hypothesis of monophyly (well supported by posterior probability) for the new proposed genus (Fig. 1). Its sister taxon is the undescribed lineage from Chile (Cecidosidae sp.); it was close to O. argentinana among the described species of cecidosids. The dated phylogeny revealed the new genus as the youngest lineage among cecidosids, which emerged around 23.8 Mya (95% HDP 10.2–34.6 mya). Genetic distances of the new genus to other cecidosids ranged from 9 to 25%; less divergence was observed in relation to the sister species Cecidonius sp. and highest to D. congregatella (Table 1).

Figure 1. 

Molecular phylogeny of Cecidosidae. Bayesian time-calibrated consensus tree based on cytochrome oxidase subunit I (COI), r16S ribosomal (16S) and Wingless (Wg) genes. Prodoxidae (Greya enchrysa and Tegeticula antithetica) was used to root the tree. Colored branches indicate posterior probability support for the equivalent node following the legend. Dark gray bar indicates confidence interval for each node age estimate, presented in millions of years ago (Mya).

Since it was proposed as a family by Brèthes (1916), the position of Cecidosidae remained for a long time uncertain until its affinity to the superfamily Adeloidea was clarified by Becker (1977), who regarded the group as endemic to South America. The affinity of Scyrotis with South American cecidosids was proposed later by Davis (1987). Molecular data provided here give further support to this taxonomic affinity, and show that the African Scyrotis are much older (ca. 90 Myr) than South American genera. Results also suggest there could exist more than one cecidosid lineage in Africa, since the two species we sequenced were 27% apart from each other in terms of genetic divergence in our analyses. The first studies on African cecidosids conducted by Meyrick (1909, 1913, 1928) clearly suggested the existence of at least three lineages of cecidosids, which can be separated by differences in the buccal apparatus of the adults. In fact, these lineages were initially grouped by him into different genera, which were later considered by Gozmány and Vári (1973) as synonyms of Scyrotis and treated as such ever since (for a detailed discussion, see Mey 2007). Two of these lineages are represented in our analysis by Scyrotis sp. and S. granosa; the former presents a rudimentary proboscis and maxillary palpi and the latter lacks such buccal structures. This question should be taken into account in revising the taxonomy and phylogeny of the family in the future, which is much needed. Although not linked to any Scyrotis species in particular, a field survey of galls conducted by van Noort et al. (2007) found them in association with several species of Searsia and suggested the existence not only of a wide variety of gall morphology but also considerable variation in life history styles among the African Scyrotis. It is unlikely that such variation will be conciliated within a single genus, which should be further explored. Unfortunately, this revision is pending upon description of the immature stages and gall morph types they induce, but as already said these aspects are still unknown for any African cecidosid species. The present study showed how valuable the inclusion of immature stages and galls is in taxonomic studies of cecidosids, whose adults in particular have relatively uniform morphology, especially regarding the genitalia (Mey 2007). In addition, our results support an accelerate evolutionary rate in all Cecidosidae lineages, as mentioned by Pellmyr and Leebens-Mack (1999) when a cecidosid (C. eremita) was used for the first time in a molecular phylogeny of Adeloidea. Similarly, Regier et al. (2015) in a family-level phylogenetic study based on 19 genes, found a high substitution rate in Cecidosidae when including C. eremita and D. capsulifex, which they found in Incurvariidae as well. This faster evolution of cecidosids makes it difficult to resolve some internal evolutionary relationships within the group, generating phylogenetic uncertainties that are hard to overcome even by increasing markers, and should be further explored.

Cecidonius gen. n. resulted as a unique lineage in the present study from both morphological and molecular analyses. Also, interestingly, it appeared as one of the most recent lineages (ca. 24 Myr) to be evolved within the extant cecidosids. It diverged ca. 16% from the closest related lineage, an additional undescribed cecidosid taxon existing in Chile and Argentina, which was included in the present study for comparison. This undescribed taxon differs from Cecidonius by having adults that lack a rudimentary proboscis and having a three-segmented maxillary palpus, pupae bearing a gall-cutter with a different shape, larvae without long hair on thorax and galls with completely developed wall without basal orifice, and will be described elsewhere. Molecular findings also showed that although described as monotypic, there is at least one more species belonging to Cecidonius, associated with Schinus terebinthifolius Raddi, which is still awaiting description. This undescribed species diverged from C. pampeanus by more than 9% in DNA sequences. Its galls are conspicuous, morphologically different, and larger than those of C. pampeanus. They are relatively common in populations of S. terebinthifolius existing in southern Brazil. Unfortunately, we have no pupae or adults of this species yet, which apparently shares a similar life-history style and associated difficulties regarding rearing of C. pampeanus.

It took us a few years to obtain the small number of C. pampeanus pupae and adults used for description in this study. Although relatively abundant as young larvae when still under the bark of swollen stems, later instars of C. pampeanus occur at low density in the field. Collection of mature, dehiscent galls during later summer, either using cloth bags attached to the plants or picking by hand those that had naturally dropped to soil, always led to failure regarding development under laboratory conditions. Dissection of these galls demonstrated that larvae do not pupate, remaining alive in the last larval instar for months, eventually dying without any apparent cause. Interestingly, similar difficulties regarding rearing of C. pampeanus are also mentioned by Meyrick (1917) in relation to the African Scyrotis. Cecidonius pampeanus apparently diapauses for months in the last instar larva, which stays motionless within its dehiscent gall in the soil until pupation and adult emergence occur in the next growing season. Probably this species presents a seasonal adaptation (sensu Tauber, Tauber and Masaki 1986) to overcome the unfavourable low temperatures that prevent growth during winter, and also adjust its life cycle to the host plant phenology. As already mentioned, new growth shoots that are required for gall induction (Raman 1994, Yukawa 2000) start appearing on S. weinmannifolius plants during the spring. This time of the year coincides with adult emergence in the field, and supposedly also with oviposition in C. pampeanus. A group of approximately 20 galls were collected in the field by the first author during winter and kept under room temperature in the laboratory within plastic vials containing moist soil from the type locality. A few were dissected at fifteen-day intervals, rendering only last instar larvae. The first pupa in this case appeared in spring (October), and the adults, which were used in the present description, ca. one month later. The token stimuli that initially trigger and later break the diapause in C. pampeanus remain to be determined. We may speculate from above that the corresponding stimuli may be received during autumn by the dehiscent galls that are already in the soil.

Allorhogas species are among a few braconid wasps having a phytophagous feeding habit. They are apparently relatively common and widespread in the Neotropics, all associated with galls, occurring in several plant families including Burseraceae, Fabaceae, Melastomataceae, Polygonaceae, Rubiaceae, and Solanaceae (e.g., Macedo and Monteiro 1989, Marsh et al. 2000, Marsh 2002, Penteado-Dias and Carvalho 2008, Chavarría et al. 2009, Centrella and Shaw 2010, 2013, Martínez et al. 2011, Martínez and Zaldívar-Riverón 2013). However, their biology is largely unknown, and it is still uncertain whether they are primary gall inducers or inquilines. A clear pattern always emerged during dissections of hundreds of galls from several S. weinmannifolius populations in the present study, demonstrating that they are inquilines. First, they were never found inside young galls that were located under swollen stem bark, where only young larvae of C. pampeanus were always present. Second, erupted galls bearing either free-living C. pampeanus larvae or those attacked by Lyrcus ectoparasitoids do not change their shape, but only those bearing Allorhogas that turn from cylindrical into globular galls. Third, Allorhogas immatures were always found within older, erupted and much larger, shaped-modified galls, where larvae of C. pampeanus were found dead. Fourth, as already described, a progressive transition in shape between galls free from such inquilines (cylindrical) to those attacked by them (globular) is found in the field, always in association with early development of Allorhogas larvae. Most of the Allorhogas studies listed above have a taxonomic bias and are based on descriptions of adults reared during extensive surveys, without including descriptions of immature stages. They lack information on gall ontogeny, and most importantly, about identification of trophic levels of insects present within these galls. Thus, the biological status of Allorhogas in those gall systems should be re-examined, since some of them may not induce galls but act as inquilines, the true gall inducers being either underexplored or missed in such cases.

Similar to what was described for the Scyrotis galls attacked by Rhoophilus Ioewi (Hymenoptera: Cynipidae) inquilines (van Noort et al. 2007), space for a C. pampeanus larva within a given gall is progressively diminished with the development of tissues induced by Allorhogas larvae. In fact, in several cases in the present study during the dissections of medium-sized developing galls bearing Allorhogas inquilines, a dead C. pampeanus larva was found within a compressed space inside. From a gross morphology perspective, tissues induced to develop by Allorhogas are clearly different from those induced by the original inducer C. pampeanus, regarding thickness, consistency, and colour. In general, tissues present in insect galls are complex, and may structurally vary even within a given gall lineage (Stone and Schönrogge 2003). Specially the nutritious ones, which are absent on ungalled host plants, may also vary in complexity at a very fine scale not only among but also within galls. For example, when tissues produced by lepidopterans and hymenopterans are compared, differences between them emerge at the cell level in relation to the type, quantity, local and disposition of chemicals they store, among other characteristics (e.g., Ferreira and Isaias 2013, Vecchi et al. 2013). These tissues are used for feeding by the corresponding inducers independently, that is within their own distinctly located galls. This is not the case in the present system, where such tissues are induced by distantly related insect lineages and occur within the same gall. Thus, we suggest that tissues induced by Allorhogas species may inhibit feeding by C. pampeanus, whose larvae, by being confined in space, completely surrounded by tissues unsuitable for feeding, would be lead to death by inanition.

Additional field observations suggest that the existence of an inquiline association between Allorhogas species and galls of other cecidosids is common in southern Brazil. This is the case of the gall induced in S. terebinthifolius by the undescribed, additional species of Cecidonius already mentioned, as well as of those induced in S. polygamus by C. eremita and E. minutanus. Thus it seems that these braconid wasps parallel in the Neotropics the cynipd wasps that are inquilines of cecidosid galls in Africa (van Noort et al. 2007). Cynipids are found in South America, not acting as inquilines but as primary gall inducers, as for example in Fabaceae (e.g., Nieves-Aldrey and San Blas 2015). Unfortunately, little is known about the biology of the Neotropical species of Lyrcus. They are diverse and difficult to identify in the Nearctic region, where many are important parasitoids of agricultural pests, primarily belonging to Coleoptera and Diptera (Gibson GAP, Agriculture and Agri-Food Canada, pers. comm.). According to preliminary observations there are additional insect species yet to be explored in association with C. pampeanus galls, including cecidophages, predators, hyperparasitoids and successors that use them as shelter. The latter may include other arthropods and are common in cecidosid galls, since some of these galls may last for years after adult emergence and thus be used by other insects for shelter and even for nesting (e.g., Wille 1926, Laroca 1972). We hope this study will stimulate additional studies on this topic, thus revealing fully the hidden diversity existing in association with these galls.

Our study provides strong evidence that C. pampeanus is under threat of extinction, and protection measures should be taken to conserve its remaining populations. The reasons are based primarily on the destruction of the host plant habitat. Open savannas of southern Brazil (= Brazilian ‘Campos’) where populations of S. weinmannifolius are found have been suffering from anthropic influence for decades, mostly caused by agriculture in general and/or cattle ranching, and recently from widespread expansion of Eucalyptus L’Heritier, Acacia Martius and Pinus Linnaeus plantations (Overbeck et al. 2007, Cordeiro and Hasenack 2009). A search by the first author for populations of S. weinmannifolius of which older dried material is preserved in the main herbaria in the region (e.g., UFPR/ Curitiba, Barbosa Rodrigues/Itajai; and UFRGS/Porto Alegre), suggested that most of these have disappeared since. In Parana state, for example, extant populations seem to be restricted to a few places, including the preserved area of Parque de Vila Velha, Ponta Grossa municipality. In Rio Grande do Sul scattered populations were located on high elevation steppes, as for example in Canela and São Francisco de Paula municipalities, and particularly at low elevations in the western portions of the Pampa biome. However, as already mentioned, extant populations of S. weinmannifolius bearing galls of C. pampeanus were restricted to small patches in the latter area. More importantly, these populations are distant and isolated from each other. Most of them are located at higher elevation, such as on hilltops and hill slopes interspersed with small bushes as already mentioned, where they are relatively more protected from anthropic influence. At least two of these areas (Morro do Osso and Morro Santana) are officially protected areas already, but the remaining populations are located on private property. Schinus weinmannifolius is considered a pasture weed, supposedly unpalatable to livestock, the reason for which we presume it has disappeared from most low elevation areas, where agriculture and cattle ranching prevail as the main economic activities. Schinus weinmannifolius is apparently heliophilous, and in consequence does not grow satisfactorily within plantations such as those composed of Eucalyptus, Acacia or Pinus, also common in the region.

There is no indication that adults of cecidosids feed actively, last long and disperse much; oviposition supposedly occurs on the plants surrounding those where they developed as immatures (e.g., San Blas and Davis 2013). The limited dispersal together with low connected patches of S. weinmannifolius resulted in ‘island’ populations of C. pampeanus with reduced variability. Moreover, high genetic structure and partition of variation based on geography corroborated a pattern of isolation by distance. The restricted distribution and small population sizes are important causes of reduced genetic diversity (Hamrick and Godt 1989), since the effects of natural selection and/or demographic changes may be more pronounced in such populations (Ellstrand and Elam 1993, Gibson et al. 2008). The low variability found in C. pampeanus possibly makes the species vulnerable for novel selection pressure. Whether populations would be affected by stochastic processes, particularly genetic drift, depends on gene flow within and among populations, among other ecological factors. We found significantly low levels of gene flow among populations of C. pampeanus. Haplotypes were mainly unique to each locality, except between Morro do Osso/Morro Tapera and Lombas de Viamão/Lombas de Santo Antônio; even so, the latter presented significant high F_ST values.

The low number of nucleotide differences between the haplotype pairs (except for H12, H13 and H14) and a multimodal curve in the mismatch distribution analysis of C. pampeanus indicate that population expansion is unlikely to have occurred. In contrast, the population of Lombas de Viamão presented an expansion pattern. According to Rogers and Harpending (1992) and Tajima (1989), bottlenecks may generate waves in the distribution of pairwise nucleotide differences. However, contrary to expansion, a population contraction leads to maintenance of genetic diversity over time. In a bottleneck model individuals differ in the average number of nucleotide changes when taken randomly from a given population. Such an effect leads to multipeak nucleotide distributions, as well as large pairwise differences between them (Harpending et al. 1998, Rogers and Harpending 1992). Additionally, when estimated by median-joining, the haplotype topology also did not support a population expansion scenario, as it did not fit into a typical star-like model (Harpending et al. 1998, Slatkin and Hudson 1991). The results suggest that demographic changes in populations of C. pampeanus are a consequence of ancient historical processes and recent decline, likely due to landscape disturbance.

The above-mentioned higher trophic level-associated fauna may be also under threat, considering that its existence depends on the success of C. pampeanus, the primary gall inducer. In other words, a whole community associated with C. pampeanus galls may go extinct in South Brazil, even before species that integrate it have been described, in the case of extinction of the primary gall inducer. A survey should be carried out to identify the unknown fauna associated with these galls. We also suggest that additional studies should examine the degree of specificity and inter-dependence of this fauna with C. pampeanus and its host plant. These actions should be prioritized when planning the corresponding conservation measures, since they are prerequisite to their implantation. Protection measures have been scarcely taken in relation to the lepidopteran species that are under threat of extinction in the Neotropical region. In Brazil, actions in this regard have involved primarily the butterflies, in total 55 species that are officially considered under threat of extinction (Freitas and Marini-Filho 2011). However, microlepidoptera and associated plants are largely unknown in this country due to a corresponding taxonomic impediment (Aguiar et al. 2009), and thus they have been completely neglected from a conservation biology perspective. Within the gall-inducer and leaf-miner micromoths there are many species that are specialists on rare and/or endemic plants, particularly in the Brazilian Atlantic Forest, most yet to be discovered and/or described (Luz et al. 2014, Moreira et al. 2017). By being dependent on endemic hosts at a regional scale, these species in particular are under comparatively greater threat, because most of such plants are restricted in distribution (Lewinsohn et al. 2005). This study is apparently the first to suggest that a micromoth and its associated fauna should be taken into account in this regard in Brazil. It is also important to emphasize that the restricted number of extant C. pampeanus populations are located within the southern Brazilian “Campos” (= Pampean savanna) that is considered a diverse but neglected biome from a conservation biology perspective (Overbeck et al. 2007), and where no moth has ever been targeted from a conservation biology perspective.

This study also showed how important intensive, integrative taxonomic studies are to identify accurately the role of a cecidosid species in a given gall community. Cecidonius pampeanus attracted our attention ca. 10 years ago as a cecosid lineage by comparison of DNA sequences extracted from the larval stage, dissected from under the bark of swollen stems of S. weinmannifolius. For several years, its identification remained provisional, tied only to DNA similarity to other cecidosids, since for this new species morphology of the last larval instar, found later in the field, was also atypical compared to any known cecidosid. Full confirmation of the existence of this new lineage came when we finally obtained their pupae and reared them to adults. We inferred that the absence of such an approach led Tavares (1909: 8) to identify the true inducers of such S. weinmannifolius galls as “… probabiliter Cynips incognita” [… probably an undescribed Cynips Linnaeus species]. This action has prevented unraveling not only the true gall inducer, but also the diversity of fauna associated with such galls for more than a century, since his rationale was followed without being questioned by other authors (e.g., Wille 1926, Houard 1933, Sáiz and Núñez 1997). In other words, from Tavares’ original description until the present study, such galls have been treated as two trophic level systems, and their induction was erroneously associated with an unidentified species of Cynipidae (Hymenoptera). The Portuguese Jesuit priest Joaquim da Silva Tavares, also a naturalist who first described these galls, was a pioneer in the study and description of Brazilian cecidology during the first quarter of the last century. His descriptions were accurate and finely illustrated, but most of them were based on the gall morph type only, not always being associated with precise identification of the corresponding inducers. We suppose the large and colourful Allorhogas- bearing galls, which appear as neat black and white photographs in his publication (Plate VIIi, figs 22, 23), called his attention to S. weinmannifolius plants at first sight in the field. As already mentioned, free-living external galls bearing C. pampeanus larvae are rarely found on S. weinmannifolius plants in the field, most being killed by Lyrcus parasitoids, and thus they may never have been encountered by him. That is, the gall phenotype that is modified by the Allorhogas inquilines would have misguided him and led him to suggest the primary induction of such galls to be by cynipid wasps, based on the immature stages obtained when dissecting such galls, since those dissections were also illustrated by him (Plate VIII; figs 24, 25). He apparently did not rear to the adult stage of the assumed cynipid species at that time, since later on when working with the Brazilian melastomatacean galls he made comments on his disappointment about not ever having had a Brazilian cynipid specimen in his collection (Tavares 1917, p.19). In the same publication, he indirectly admitted having erroneously thought at first that these melastomatacean galls also looked like those induced by cynipids in Europe, but that he had changed his mind after having surprisingly obtained the first adult Lepidoptera reared from them.