Table 3 shows the Intha names recorded for the Inle Lake fish species, along with the number of respondents who were able to correctly identify each species. Of the 43 stimulus species used for elicitation with Intha respondents, four (Chaudhuria caudata, Glyptothorax rugimentum, G. siamensis and Oryzias uwai) could not be identified by anyone, while a handful of other species (such as the endemic Poropuntius schanicus) were only recognised by a small minority (less than ten) of respondents. These were mostly smaller species, with a size range of around 10–13 cm. There were, however, a number of smaller species that many people recognised: these included the tank goby Glossogobius giuris, the loaches Lepidocephalichthys berdmorei and Petruichthys brevis, the cyprinids Inlecypris auropurpureus, Pethia stoliczkana and Puntius sophore and the glassfishes Parambassis spp. A number of these were reported as forming an important part of the locally consumed ‘dried whitebait’ by Annandale [19] (it is known locally as ngət̪əphwɛ3 chauʔ), and this tradition continues to this day. An important difference between Annandale’s observations and the present study lies in the abundance of the small endemic species Sawbwa resplendens, with a maximum length of 3.5 cm. Annandale reports it as being ‘extremely abundant all over Inle Lake’ and ‘of economic importance’ (p. 49), but only 18 older individuals were able to identify it in the present study.

Another endemic species, Physoschistura shanensis, sampled by Kano et al. [20] from streams to the west of Inle Lake proved not to be very well known, with only 22 respondents giving it a name. The small Microrasbora rubescens fared slightly better, as it was identified by 34 respondents. Even this higher number is at odds with Annandale’s observation that ‘this little fish is very abundant all over Inle Lake’ (p. 51), and that it also forms an important constituent of dried whitebait.

Many larger fish species are widely known and represent ethnotaxa that are presumably still common and/or culturally important. These include the nga3 phein3 (Cyprinus intha), the most highly prized of all fish to Intha people, and almost universally known, along with other large carp (some of which are native, such as Systomus rubripinnis, while others are introduced aquaculture species, such as the Java/silver barb Barbonymus gonionotus), catfishes (Clarias spp.), eels (Monopterus spp.), snakeheads (Channa spp.) and the bronze featherback (Notopterus notopterus). Not surprisingly, the highly invasive and fecund Nile tilapia Oreochromis niloticus is also universally recognised. This fish probably dominates the Inle Lake ecosystem now, as nearly all fishing boats encountered by the authors during data collection (except for those focusing on shrimp) contained nothing but tilapia.

The Rakhine fish names recorded in Kyaukphyu and Kyaukpyauk are presented in Table 4, along with the number of respondents who were able to correctly identify the species in the Core list. While many ‘local’ names are presented in Psomadakis et al. [23], there is no information on where the names were collected, and it is likely that very few are in the Rakhine language. Moreover, many of the names are direct translations of the official English Food and Agricultural Organisation (FAO) names into Burmese, as in ‘Taung Pan Mae Set Nga Pyan’ for ‘spotfin flyingfish’ (p. 341). Thus, while Psomadakis et al. is resource of outstanding value, the local names presented within this publication cannot be considered as a substitute for primary language documentation in fishing communities.

It is more difficult to make generalisations about the Rakhine fish name dataset, compared to the Intha names, due to the large number of species involved. However, there seems to be a tendency for pelagic species (such as Selar crumenophthalmus) and species associated with coral reefs (Sargocentron praslin) and deep water (Antigonia emanuela, Argyrops spinifer, Hapalogenys merguiensis) to be less well known than nearshore, intertidal or brackish water species (examples of well-known fish that meet the latter criteria include Eleutheronema tetradactylum, Lates calcarifer, Scatophagus argus). As expected, Rakhine people largely could not identify predatory reef-associated fish that are commonly consumed in other parts of the world, such as snappers (Lutjanus spp.) and emperors (Lethrinus spp.).

A noteworthy feature of the Rakhine dataset is the large number of synonyms for many fish ethnotaxa/species. It is not uncommon for a species to be labelled with four or five names, and some have as many as eight distinct names. Most of the names presented in Table 6 are very likely correct, as the correspondences between pairs or sets of names were stated explicitly during interviews by (often multiple) expert fishermen. For instance, a 49 y.o. fisherman from Kyaukpyauk said that the Indo-Pacific king mackerel Scomberomorus guttatus was called ngənyo1 loun3 in his village, but was referred to as ngəshwan3 in town (i.e. Kyaukphyu). Both claims were verified through interviews with other fishermen from the two locations. Other forms of naming variation include simple phonological variation, semantic variation, where different aspects of the fish are referred to in the name, and internal lexical variation, where two variant forms of a fish name have the same meaning, but use different lexemes or morphemes. The sociolinguistic aspects of fish nomenclature are highly interesting but complex and will be explored in detail in a future publication.

Following are the results of nonparametric statistical analyses on the responses from the speakers of the two languages (the data were not normally distributed and frequently had unequal variances.) The dependent variable Total (the total number of stimuli named by each consultant) is only mentioned in case of a significant result.

A Kruskal–Wallis test on the three Age categories returned a significant result for the dependent variable Unique (H = 8.43, d.f. = 2, p = 0.01), while post hoc pairwise comparisons revealed that it was the ‘younger’ age group that was able to produce significantly fewer Unique fish names than the other two groups (Fig. 2a). The variable Total was also significant, (H = 11.05, d.f. = 2, p < 0.01), with the performance of the ‘younger’ group being significantly worse than that of the ‘older’ group.

a Mean Unique and Total responses according to Age category (Intha). Y-axis shows average number of responses (names) ± S.E; b, c Mean Unique responses according to the Occupation category (Intha). b all respondents; c ‘younger’ respondents excluded

Overall, the variable Occupation did not have a significant effect on Unique responses (Kruskal–Wallis H = 6.72, d.f. = 3, p = 0.08). There is some indication in Fig. 2b that sellers might outperform men who have never fished, but this is possibly due to the fact that there were no ‘younger’ respondents among the sellers (the mean age for sellers was 44.36 y.o., compared to 37.7 for the men who have never fished). Excluding younger respondents from the analysis reduces the differences in Unique names among the Occupation categories (Fig. 2c).

An in-depth analysis of possible interactions between Age and Occupation could not be carried out for Intha, due to the small number of sellers, and the very unequal age distributions for both sellers and fishermen. However, it was possible to combine the Unique responses of current and previous fishermen, to see if the categorical Age variable had an effect on people with at least some fishing experience. While there was a very slight trend for Unique responses to increase with increasing age (Table 5), the differences were not statistically significant (Kruskal–Wallis H = 1.04, d.f. = 2, p = 0.59).

The variable Zone did not have an overall effect on consultants’ responses (Kruskal–Wallis H = 4.80, d.f. = 2, p = 0.09). However, a significant difference did exist in a subset of the data, but this needs to be verified with a larger sample: when regarding just those respondents who have some fishing experience (i.e. current and previous fishermen), the Unique responses from the ‘north’ zone were significantly higher than those from the ‘central’ and ‘south’ zones combined (Kruskal–Wallis H = 4.08, d.f. = 1, p = 0.04). The difference was marginal, however (Table 6).

Finally, the responses of male and female individuals from the ‘never fished’ category were compared, to determine if there were gender effects in fish knowledge. The dataset was unfortunately small (11 females, 9 males), and although females on average knew more Unique names (f: 23, m: 19.7), there was no statistically significant difference (Mann–Whitney U = 32, p > 0.05).

As with the Intha data, the categorical Age variable was found to have a significant effect on the Total number of fish species named by Rakhine speakers (Kruskal–Wallis H = 6.06, d.f. = 2, p = 0.048) (Fig. 3). However, the p value indicates that this effect is marginal, possibly due to the high level of inter-individual variation. Age did not have a significant effect on the dependent variables Unique (Kruskal–Wallis H = 5.19, d.f. = 2, p = 0.07) and Core (Kruskal–Wallis H = 4.39, d.f. = 2, p = 0.11).

Mean Total, Unique and Core responses according to Age category for all respondents (Rakhine). N.S., not significant

Next, the effect of Age on the responses of the two largest occupation categories was analysed separately. Looking at just the data from the fishermen, the categorical Age variable was found to have a significant effect on all three dependent variables: Total (Kruskal–Wallis H = 14.44, d.f. = 2, p < 0.001), Unique (Kruskal–Wallis H = 16.92, d.f. = 2, p < 0.001) and Core (Kruskal–Wallis H = 10.48, d.f. = 2, p < 0.01). In all cases, it was the younger fishermen who performed worse than the intermediate and older fishermen (Fig. 4a). The responses of the sellers showed a completely different pattern, as Age category did not have a significant effect on any of the dependent variables (Total, Kruskal–Wallis H = 0.412, d.f. = 2, p = 0.81; Unique, Kruskal–Wallis H = 4.723, d.f. = 2, p = 0.09; Core, Kruskal–Wallis H = 0.50, d.f. = 2, p = 0.77). However, the Unique scores for the sellers did show an increasing trend with Age (Fig. 4b), and so a correlation analysis between age as a continuous variable and the Unique responses for sellers was carried out. The result turned out to be significant (Kendall’s Tau = 0.40, p < 0.01). It is also noteworthy that sellers of all age categories had nearly identical responses for the Core fish names (Fig. 4b).

a Fishermen only: Mean Total, Unique and Core responses according to Age category (Rakhine). b Sellers only: Mean Total, Unique and Core responses according to Age category (Rakhine). N.S., not significant

The Occupation variable was found to have a significant effect on Rakhine people’s responses, for all three dependent variables measured: Total, Kruskal–Wallis H = 39.84, d.f. = 2, p < 0.001; Unique, Kruskal–Wallis H = 48.73, d.f. = 2, p < 0.001; Core, Kruskal–Wallis H = 37.44, d.f. = 2, p < 0.001. A graphical representation of the combined responses for the Unique names can be seen in Fig. 5. In the case of Total and Core, respondents in the consumer category performed significantly worse than the fishermen and sellers, while the latter two groups were statistically similar (Fig. 5c). The Unique responses showed a different pattern, as the sellers’ performance fell between that of the fishermen and the consumers (Fig. 5). In other words, fishermen were able to produce significantly more Unique names than sellers. The brokers’ data were excluded from this and the previous analysis, but their performance was generally similar to that of the fishermen (the brokers’ data can be seen in Fig. 5).

a Distribution of Unique responses based on Occupation (Rakhine). b Distribution of Core responses based on Occupation (Rakhine). c Mean Total, Unique and Core responses based on Occupation (Rakhine)]

Finally, there were gender differences among the group of consumers, i.e. people with no direct connection to fishing or the fish trade. Both males and females volunteered similar numbers of Total identifications (Mann–Whitney U = 40, p > 0.05) and knew roughly similar numbers of Unique names (Mann–Whitney U = 47.5, p > 0.05), but female respondents performed significantly better at identifying species from the Core group (Mann–Whitney U = 11, p < 0.05) (Fig. 6). On average, female consumers were able to correctly identify 12 more culturally and/or commercially important fish ethnospecies than male consumers.

Mean Total, Unique and Core responses based on gender among the consumer group (Rakhine)