Within the psychology of language and cognition, a problem is demonstr
ated that is generally important and that relates to object references
: On the one hand, there is good cause to speak in a detailed way in o
rder to supply the communication partner with lots of information. On
the other hand, there are assumptions, e.g. the principle of informati
vity (Grice, 1979), according to which it appears desirable to enrich
one's utterances with no more information than needed. However, if we
look at object references, we often find more attributes specified tha
n would be necessary with respect to the partner's unambiguous identif
ication of the intended object. In many cases, this additional specifi
cation of certain attributes (e.g., color) facilitates the partner's i
dentification of the object in question. However, facilitation is effe
ctive only in cases in which the additionally specified attribute is p
erceptually salient and the partner's field of search is narrowed as w
ell. This condition holds if the object to be named involves a value o
n the specified attribute that distinguishes the object from at least
some of the other objects in its context. Empirical studies revealed t
he color of an object to be most important among the redundant, but sa
lient attributes that are often specified additionally. However, the c
olor of an object tends to be not specified in cases where it is chara
cteristic of the object and, therefore, can be easily inferred by the
partner. We report on two experiments from which we learned three thin
gs: (1) The color of an object is specified more frequently if it allo
ws for the discrimination of the object from (at least some) other obj
ects in the context. (2) The color of an object is specified less freq
uently if it is characteristic of the object. (3) The aforementioned r
esult remains unchanged if an experimental manipulation is introduced
that makes the speakers doubt whether the partner can easily infer the
color of the object. Thus, the specification of attributes when namin
g objects appears to be systematic, but more complicated than was prev
iously assumed.