Background: Deficits in the production of complex structures have been widely documented in non-fluent forms of aphasia. Nevertheless, the data available on fluent deficits are scarcer. In both cases, reduced complexity is attributed to syntactic factors. In the related field of syntactic theory, there exist a number of studies on the production of non-brain damaged (NBD) subjects which try to account for the existence of two alternative constructions in embedded complement clauses in English (as in I think that the situation will improve/I think the situation will improve). The absence of that in the embedded clauses of verbs like say, know or think in colloquial English is very frequent and this suggests that verbs of this type may select a clause lacking a complementiser phrase (CP) layer, namely tense phrase (TP). The presence of that is taken to be the result of insertion, which is triggered by cues associated to contextual factors like register.Aims: To compare the presence and absence of the complementiser that in the speech of English subjects diagnosed with aphasia with the same phenomenon in NBD subjects with the objective of clarifying the nature of the phenomenon of that-omission.Methods & Procedures: We carried out an analysis of spontaneous speech that included the performance of 200 individuals brought together by the AphasiaBank project. Two groups were included in the study, an experimental group comprising 100 individuals diagnosed as aphasic according to the standards of the Western Aphasia Battery (WAB), and a control group including 100 non-brain damaged control subjects (NBDs).Outcomes & Results: This study shows that the asymmetries across populations (NBDs vs. subjects with aphasia) are restricted to the number of occurrences of subordinate clauses. NBD subjects produce more embeddings than subjects diagnosed with fluent aphasia (although they do produce embedded clauses and crucially prefer the "omission" option) and subjects with non-fluent aphasia.Conclusions: Our results confirm the findings on fluent aphasias as for the presence of deficits with complex constructions. These results may be regarded as evidence for the claim that TP is the default selection for the verbs analysed.