My typologically isolating conlang Ronc Tyu 
( does not strictly make a big 
deal of evidentiality (i.e. information about the source of knowledge 
regarding a statement) - it's generally optional so that the vast 
majority of sentences in the language are not marked for information 
source, and it's not morphologically marked anywhere either. However, 
it's an interesting topic. I've been thinking about strategies for 
expressing evidential information in Ronc Tyu lately, and I think the 
constructions I've come up with are worth mentioning here.

The first such strategy is not very remarkable at all and consists 
simply of the main statement as a normal complement clause introduced 
with the regular complementizer *tenc*, embedded within a matrix clause 
which describes the source of information:

Ne  wuc  tenc nà     twíc       myè.
1SG hear SUB  mother argue_with father
I heard that mother had an argument with father.

The second construction, however, is more interesting. It's more common 
in speech than the other type, it's used specifically for statements of 
evidential value, and it's a few steps further forward on the way to 
becoming grammaticalized. It employs a special kind of complement 
clause, which is subordinated to the impersonal verb *trà* ‘exist’ and 
introduced by one of several specialized complementizers that indicate 
different levels of evidentiality. (These are diachronically derived 
from former possessed nouns, along the lines of "there is [the 
information source of]..."):

Trà   nrù  ki   pei ha      kú      twí   yu mun.
exist SENS this boy step_on destroy spear of 2SG
This boy stepped on your spear and broke it [he was seen doing it].
(sensory evidence)

Trà   nèi  mae  zúc  fou  dzic       twinc.
exist PHYS deer walk pass go_through here
A deer must have passed by here [there are traces on the ground].
(physical evidence)

Trà   rei  ugei      ma.
exist QUOT chieftain die
The chieftain has died [people say so].
(communicative evidence, hearsay)

Trà   zrin nrèn      brinc     kóan  tóc    taoc.
exist HYP  foreigner be_secret seize remove horse
The stranger might have stolen the horse [it is suspected].
(no concrete evidence, speculation)

It is worth noting that none of the evidentials distinguish between 
direct and indirect evidence, i.e. they only assert that the relevant 
knowledge was indeed acquired in the specified way, but not necessarily 
by the speaker himself. By whom exactly can only be deduced from context.

The more modal semantics of two additional special complementizers do 
not quite fit in with the canonical definition of evidentiality, having 
to do more with the attitude of the speaker towards the proposition, but 
they function in the same way syntactically so I'll mention them here too:

Trà   yu  dyao    tèin     dric   màc     ló  hei.
exist AFF enemies then.FUT attack village DAT 1PL
The enemies will attack our village [I'm sure of this].

Trà   mè  Gèn  móc  bloun nù!
exist MIR NAME kill lion  really
Gèn really killed a lion! [I didn't expect him to].

Evidential complementizers can also be used in a certain type of polar 
questions, which are formed with the impersonal auxiliary verb *hrec* 
‘is it true that...?’. In such a situation, it is not the truth value of 
the proposition as such that is being asked for, but the information 
status expressed by the evidential complementizer:

Hréc  nrù  pei fwei nonc?
Q.AUX SENS boy hit  girl
Did you witness the boy hitting the girl [or did you get that 
information from somewhere else]?

Hréc  yu  ugei      ma  nù?
Q.AUX AFF chieftain die really
Are you sure that the chieftain has died [or might he still be alive 
after all]?

Hope this is of interest to some of you. How do your conlangs form 
evidential statements? How many and which levels of evidentiality do 
they have? Are they expressed morphologically or periphrastically?

-- Jan