A limited number of tests, older as well as more recent ones, are described in detail below. Parts A through H pertain to intelligibility tests, ordered from tests based on meaningless items of a few phonemes to tests based on meaningful sentences, parts I and J describe two prosody tests, and finally parts K and L are related to two overall quality tests. Tests were included that have been frequently used, or were developed specifically, for evaluating synthesis output. Four tests (SAM Standard Segmental Test, Cluster Identification Test , Semantically Unpredictable Sentences , and the SAM Overall Quality Test ) were developed within the ESPRIT-SAM (Speech Assessment Methods) project (1989--1992). Software for these tests is available in SOAP (Speech Output Assessment Package) (cf. =1 (
; Howard-Jones 1992b) ). SOAP also includes programs to add noise to speech files, to automatically generate test sequences, to randomise test sequences, and to equalise speech levels. Test descriptions are given below.
A. SAM Standard Segmental Test
B. CLuster IDentification Test (CLID)
D. Diagnostic Rhyme Test (DRT)
F. Haskins Syntactic Sentences
G. SAM Semantically Unpredictable Sentences (SUS)
H. Harvard Psychoacoustic Sentences