Encoding Self-Embedded Sentences

Are encoding and decoding mirror image processes? Previous studies show that semantically supported self-embedded sentences are easier to decode than semantically neutral sentences. A procedure for testing encoding is developed. The hypothesis that semantically supported self-embedded sentences are not easier to encode than semantically neutral sentences is supported. An analysis of the errors made indicates that although errors increase with increasing sentence complexity, subjects can produce the correct orderings of subjects and predicates in sentences with up to five degrees of self-embedding. It is concluded that the scarcity of complex self-embedded sentences is not based primarily on memory limitations in the encoding of sentences, but more likely on the memory requirements in decoding.