An implicit conclusion following directly from Kay’s (1990) groundbreaking analysis of the scalar additive focus particle even is that propositions marked by this particle are not necessarily located at the endpoint of a pragmatic scale, contrary to the findings of prior research (e.g. Fauconnier 1975). Nonetheless, in spite of this result—or perhaps due to its implicitness—even scholars who agree wholeheartedly with Kay’s findings continue to assume that even marks the "extreme end of a scale" (Traugott 1995:43) or the "least likely member of [a] set" to have realized some action described by the predicate (Horn 1996:306).

The objective of this paper is to demonstrate that Kay’s implicit finding is supported in explicit fashion by data from Spanish and Hindi, and that these data moreover provide evidence for two distinct types of scalar particles. Both languages have pairs of particles translatable into English as even: hasta/incluso (Spanish), -tak/-bhii (Hindi). I argue that the relevant distinction between the two particles in each language depends crucially on their sensitivity, or lack thereof, to scalar propositions already accessible in the discourse context. I also argue that, as a result of their differing sensitivity to context, the particles hasta/-tak invariably mark scalar endpoints, while the particles incluso/-bhii do not. The latter particles are, nevertheless, compatible with endpoint interpretations when their discourse-contextual requirements are met. Consider (1a) from Spanish and (1b) from Hindi (# = pragmatic infelicity):

(1a) A: ¿Qué comiste en la cena?

‘What did you eat at the dinner?’

B: Comí hasta/#incluso caviar.

‘I ate HASTA caviar.’

(1b) A: tum-ne paarTii-meN kyaa khaayaa?

‘What did you eat at the party?’

B: mai-ne-to caviar-tak/#-bhii khaayii

‘I ate caviar-TAK.’

Because of the presence of hasta/-tak, the responses in (1) must be interpreted as marking the endpoint of a pragmatic scale ranking foods eaten at the party. The asymmetrical behavior of the focus particles in (1) shows that incluso/-bhii, but not hasta/-tak, require a discourse-accessible context proposition. Compare (1) with the yes-no question context in (2), where the assertion underlying the question is accessible as a scalar context proposition, thereby allowing incluso/-bhii to appear felicitously.

(2a) A: ¿Comiste gambas en la cena?

‘Did you eat shrimp at the dinner?’

B: Comí hasta/incluso caviar.

‘I ate HASTA/INCLUSO caviar.’

(2b) A: tum-ne paarTii-meN biryaanii khaayii?

‘Did you eat biryani (meat and fried rice) at the party?’

B: mai-ne-to caviar-tak/-bhii khaayii

‘I ate caviar-TAK/-BHII.’

Similar evidence for distinguishing the particles is provided by their usage possibilities in other syntactic constructions, e.g. concessive ("even if") conditionals and comparative sentences ("X is even more/less than Y"). In each language, it is the non-endpoint-marking particle which is favored in such constructions.

The differences between these pairs of scalar focus particles show that there is robust cross-linguistic evidence for distinguishing two types of scalar particles. Absolute particles like hasta/-tak do not require an accessible context proposition and as a result invariably signal scalar endpoints. These contrast with relative particles like incluso/-bhii, which require an accessible context proposition on the same pragmatic scale, and do not invariably indicate endpoints. A prediction that follows from this analysis is that if a given language has only one scalar particle, it will be relative, not absolute, in nature, since the former, but not the latter, is compatible with both endpoint and non-endpoint interpretations.