Data were collected by Biodiversity Initiative at two locations in Equatorial Guinea: Cuidad de la Paz (hereafter ‘La Paz’) during the sunny dry season (December-January) from 2016 to 2023 and one cloudy dry season in 2022 (1128 captures) and about 65 km south at Altos de Nsork National Park (hereafter ‘Nsork’) in 2014 (65 captures) for a total of eight field seasons. In La Paz, mist nets in the primary forests were over 500 m from the nearest road and at least >1.5 km from the closest settlement; the canopy is closed. The secondary forest site is adjacent to the city; the north and west side are bounded by ~70-m wide paved roads, a ~50-m wide dirt road on the southern portion, and high-tension power lines on the east (30 m wide with vegetation 2-3 m high which is regularly trimmed) effectively isolated the eastern portion (Barrie et al., in review). The canopy of the secondary forest plot remains closed, but it has been commercially logged in the past decades and continues to be regularly selectively logged; it was most recently logged in about 2018. Both forests had similar canopy height (primary: 14.1 m ± 0.5 SE m; secondary 14.1 ± 0.4 SE m). Although primary forests had a higher canopy cover and visibility through the understory when compared to the secondary forests (canopy cover: primary: 88 ± 1.3 SE m; secondary 83 ± 2.2 SE m; visibility: primary: 9.5 ± 0.5 SE m; secondary 7.5 ± 0.4 SE m). At Nsork, primary forest sites were located approximately 800 m into the national park (i.e., north and west) from a road bordering the southern edge of the park. Secondary forest was located <500 m from a main road and had been selectively logged over at least the last three decades. Based on LANDSAT satellite imagery, primary forests at both locations in Nsork were selectively logged as late as the 1980’s but have remained intact since.

We captured birds using mist nets (12 m x 2.5 m); net lanes (Nsork n=6, La Paz n=20) were set up at two sites in each forest type and moved every two days. We opened the nets at about 6:30 AM and closed after six hours. We identified every individual captured to species level when possible, and we recorded standard morphometric measurements (mass [grams], natural wing chord [mm], and tail length [mm]), breeding condition (presence of a brood patch, degree of cloacal protuberance), sex (when possible), and molt status (symmetrical flight feather molt only). We collected a single rectrix (tail feather) from adult birds to be used in fCORT analysis.

We defined the bird’s age following a molt-based aging system (Wolfe-Ryder-Pyle, hereafter ‘WRP’ [Wolfe et al., 2010, Wolfe et al., 2021]). All species captured exhibit a complex basic molt strategy where juveniles undergo a preformative molt into a unique formative (post-juvenile) plumage prior to adopting their adult basic plumage (Howell et al., 2003). Species exhibiting this molt strategy are further divided into two groups: those with a complete preformative molt (where individuals replace all of their juvenile feathers) or a partial-to-incomplete preformative molt (where individuals replace only some of their juvenile feathers). In species that undergo complete preformative molts, the formative and adult basic plumage are virtually indistinguishable. Therefore, we used different age groups for species that undergo complete or partial preformative molts. For species with complete preformative molts, juveniles were defined as individuals that had not undergone their preformative molt (i.e., birds in their juvenile plumage) and the rest were grouped as “adult” birds. For birds with partial preformative molts, we defined first year birds as juvenile birds (juvenile plumage) or adolescent birds (formative plumage) and the rest were defined as adults.

We used data from ten understory insectivorous species, including a subset that are specialized ant-following species, as both groups are particularly vulnerable to anthropogenic disturbance at a Pantropical scale (Peters and Okalo, 2009; Powell et al., 2015). The three specialized ant-followers were Alethe castanea, Chamaetylas poliocephala, and Neocossyphus poensis. Of the African ant-following species, these three have been identified as the specialized ant-followers, depending most heavily on driver ant swarms to forage (Willis, 1986; Peters et al., 2008; Craig, 2022). Furthermore, at our site, these species attended swarms at disproportionally higher rates compare to other regular to occasional ant-following insectivores and displayed behavioral adaptations for tracking and locating ants similar to obligate ant-followers in the Neotropics (Rodrigues, 2024). Understory insectivores included Illadopsis rufipennis, I. fulvescens, I. cleaveri, Hylia prasina, Sheppardia cyornithopsis, Bleda notatus, and B. syndactylus. These species forage for insects and other invertebrates near the forest floor and are similarly vulnerable to anthropogenic disturbance. While these species sometimes also attend ant swarms, they do not specialize on driver ants. We pooled I. rufipennis and I. fulvescens captures as Illadopsis sp. due to their similar ecological niches and challenges in differentiating them in the field.

We measured fCORT concentrations and performed extraction using the protocol described by Bortolotti et al (2008) and ELISA kits (Corticosterone ELISA kit; Neogen Corporation, Ayr, UK), as validated by Carbajal et al. (2014). For A. castanea, B. notatus, and C. poliocephala, we used 20 rectrices (ten from each forest type). For B. syndactylus we used 16 rectrices (six from primary and ten from logged forest forest), and for Illadopsis spp. We used 14 rectrices (eight from primary six in logged forest).

We used generalized linear mixed models (GLMMs) to determine how life history and population traits differed between forest types (R version 4.2.2 and the lme4 package (Bates et al., 2015; R Core Team, 2022). We fit seven models with breeding status (binary), molting stage (binary), breeding-molt overlap (binary), age ratio for species that undergo complete preformative molt (binary), age ratio for species that undergo partial preformative molt (binary), fCORT (numerical) and body condition (numerical) as response variables. The explanatory variables were the individual effects of forest type and species, and their interaction. We included year (i.e., field season) as a random effect. To avoid pseudoreplication, we only included the first capture of each individual per season (individuals are often captured repeatedly in a day or season) leaving us with 792 birds of our focal species in primary forest and 401 in secondary forest. Best-fit models were selected using likelihood ratio tests, starting with the most complex model, followed by subsequent reverse step-wise deletion of non-significant terms. We thought it important to model the age ratio separately depending on the molting strategy as the age groups were different depending on the molt strategy of the species. Models were used to predict (predictMerMod function in the lme4 package, v1.1-26; (Bates et al., 2015) life history and population traits for each species.

We found no effect of forest type on the probability that an individual was breeding (94/792 in primary and 38/401 in secondary) or undergoing molt-breeding overlap (10/325 in primary and 12/190 in secondary) for any of the focal species (LRT vs best-fit: X² = 8.14, P = 0.52, df = 1192; X² = 0.44, P = 0.52, df = 514) ( Table 1 ). However, the best fit model showed that forest type had a significant effect on the probability that a bird was molting (337/792 in primary and 178/401 in secondary) (LRT vs best-fit: X² = 25.69, P < 0.01, df = 1192) ( Table 1 ). Two ant-following species were significantly more likely to be molting in primary forests than in secondary forests: A. castanea and C. poliocephala ( Figure 1 ).

For birds that undergo a complete preformative molt, we found no effect of forest type on the probability that a bird was a juvenile bird (22/290 in primary and 17/170 in secondary) (LRT vs best-fit: X² = 6.26, P = 0.39, df = 459) ( Table 1 ). In contrast, the best fit model for species that undergo a partial preformative molt showed that forest type had a significant effect on the probability of a bird being on its first year (91/369 in primary and 60/139 in secondary) (LRT vs best-fit: X² = 20.82, P < 0.0001, df = 514) ( Table 1 ). A. castanea, C. poliocephala, N. poensis and S. cyornithopsis had higher probability of a bird being on its first year in secondary forests compared to primary forests ( Figure 2 ).

The best fit model for body condition showed that forest type had a significant effect (LRT vs best-fit: X² = 19.84, P < 0.05, df = 1192) ( Table 1 ). The only species that showed a significant difference in body condition was I. cleaveri, which weighed on average 7 g/mm more (6.67%) in secondary forest ( Figure 3 ). We detected no effect of forest type on the amount of fCORT for any species (LRT vs best-fit: X² = 0.88, P = 0.97, df = 89) ( Table 1 ).

Breeding and molting are both costly life history traits (Romero, 2002). We expected to find a higher proportion of birds breeding, molting in primary forests as they have access to more food resources, more cover, and lower densities of predators (Sekercioglu, 2002; Cosset et al., 2021). We also expected a higher proportion of birds undergoing molt-breeding overlap in secondary forest. However, we did not identify any differences in the proportion of breeding birds or molt-breeding overlap between the two forest types. This might suggest that forest disturbance does not affect the breeding status of our focal species; in other words, the ability to physiologically mobilize the body into breeding condition per se does not appear to be limiting population growth/maintenance at our study system. This was also the case in tropical rainforests in Malaysia where understory birds in primary and selectively logged forests showed no difference in their breeding status (Yap et al., 2007).

With that said, relatively little is known about the annual cycle of Central African birds. Regional variation may be possible, particularly in species that occupy the entire Congo Basin as well as forests of East Africa and/or West Africa (in the case of the latter species); regional influences on annual life cycles could be present for other more geographically restricted species as climatic influences on seasonality across the Congo Basin and into the Lower Guinea Forest. Though differences may vary subtly, they may appreciably impact the timing and duration of breeding and molting periods. It is possible that our study did not capture the optimum breeding period for these species due to the temporally limited nature of our study. Incorporating data from the cloudy dry season and shoulder seasons where feasible may provide further insights into how the annual life cycle may be influenced by habitat degradation.

Additionally, our study did not address other demographic markers such as hatching or fledging success. With higher densities of predators and fewer resources, it is likely that primary forest specialists relegated to selectively logged forests will have more failed breeding attempts (Thiollay, 1999; Cosset et al., 2021).

In contrast, we found that the proportion of molting birds in two ant-following bird species was higher in primary forests. In Malaysia, species sensitive to forest degradation molted in higher proportions in primary forest compared to secondary forests (Yap et al., 2007). We expected this to be the case, since prior research indicated that species could alter their annual molt in habitats according to food availability (Marini and Durães, 2001). Foster (1975) suggested that suspension of molt could be an environmental adaptation to ensure successful breeding attempts. If secondary forests are indeed more challenging environments for these birds (e.g., fewer Dorylus ants to follow), this may increase the likelihood that a bird would alter their molt cycle to ensure a successful breeding attempt and avoid the consequences of molt-breeding overlap. All of the above could be caused due to reduce raid activity of Dorylus ant activity during the dry season is in secondary forests reducing the foraging opportunities for ant-following birds (Kumar and O’Donnell, 2009). This is likely because secondary forests have a more open canopy and a more unstable microclimate making the birds more vulnerable during the dry season (Cao and Sánchez-Azofeifa, 2017). As a result, these birds would delay their molt cycle to a later season with a higher resource availability, although we are unable to test for that as our data was collected during the dry season. The two ant-following species that had the higher molting probability in primary forests (A. castanea and C. poliocephala) are also captured about twice as often in primary forest at our study site, suggesting that they are sensitive to forest disturbance (Barrie et al., in review).

Among the study species we could age more precisely (due to partial preformative molts creating distinct molt limits), we found a higher proportion of first-year birds in second growth forest. Variation in age ratios across different forest types can be attributed to several mechanisms. First, obligate ant-following birds in the Neotropics hold large, non-exclusive home ranges and although they exhibit low territoriality, they are, however, aggressive and compete for the most profitable regions at swarm fronts (Willson, 2004). Although little is known about their behavior, low territoriality and high competition at swarm fronts have similarly been observed in African ant-followers (Brosset and Erard, 1986). This competition for space at the swarm front intensifies in denser populations (Willson, 2004). Therefore, it is likely that some species may adhere to an ideal despotic distribution where young and potentially dispersing individuals select habitats with fewer antagonistic individuals (Fretwell and Calver, 1969) resulting in young and dispersing birds occurring in second growth forest, habitats characterized by lower densities of ant-following birds (Barrie et al., in review). Regardless of the mechanism, a population with a higher proportion of first year birds is less productive, as first year birds have a lower survival rate and are first-time breeders, which results in lower population growth rate (Pyle et al., 2020).

We did not detect any differences in body condition for all but one species; contrary to our predictions, Illadopsis cleaveri had a higher body condition in secondary forests ( Figure 3 ). We speculate that changes in vegetation structure in the secondary forest may have benefitted I. cleaveri—the smallest-bodied and smallest-billed Illadopisis at our site—perhaps releasing it from competition from the other, larger Illadopsis species that also spend much time on the forest floor looking for arthropods. Indeed, I. fulvescens/rufipennis is nearly absent from the secondary forest; I cleaveri is also quite rare there (Barrie et al., in review) but perhaps is better adapted to the secondary forests’ more cluttered understory. Jones et al. (2022) describe similar results in a study looking at changes in the body condition of 20 understory birds across a gradient of deforestation in tropical forests in Western Andes of Colombia. They found that some species benefited from the changes in the vegetation structure allowing them to increase their body condition, which then had a positive effect on their breeding condition.

We investigated the fCORT levels of six species as an integrated measure of physiological stress aggregated over the period of feather growth (Messina et al., 2020). However, we did not identify any differences in fCORT levels between the forest types for our focal species. Previous work (Bortolotti et al, 2008) has indicated fCORT levels are most closely associated with the frequency or magnitude of a CORT stress response (i.e., acute stress). Therefore, the lack of a difference suggests there is no difference between habitats in physiological stressors, specifically during the period of feather molt. It is possible that differences do exist between habitats at other points in the annual cycle, or in baseline CORT levels, where elevated levels would indicate chronic stress, such as that caused by prolonged food shortage (e.g., Marra and Holberton, 1998). Future work could consider blood CORT levels at multiple timepoints and challenge experiments to both quantify baseline levels and the magnitude of the stress response between habitats, although this approach presents methodological challenges in the field (e.g., Wingfield et al., 1992).