This study aims at examining linguistic devices used in the examination in chief, cross-examination, and re-examination in trial. The data elicited includes 25 criminal cases. It is found that attorneys adopt various linguistic devices to achieve their ends. In the examination in chief, the seven devices adopted are making a request, repeating the witness's answer, stating a short-question, restating a question, using a follow-up question, repeating a question, and summarizing the gist. As for the cross examination, the eleven devices adopted are using certain words to discredit the testimony of the witness, asking the information that witness is not supposed to know, attacking the testimony of the witness, using a hypothetical question, making a statement opposite to the fact for information-checking, pressing for a certain answer, cutting an irrelevant response, restating a question, using a follow-up question, repeating a question, and summarizing the gist, In the re-examination, the three linguistic devices adopted are guiding to a certain answer, repeating the witness's answer, making a request. Most of these devices are used for asking. The acts of asking found in the data are both in the interrogative form and the affirmative form. Yet, the interrogative form appears to be preferred. A contrastive study reveals that the direct examination and the cross-examination are different in three aspects: That is, the cooperation of the participants, the linguistic devices adopted, and the types of question adopted. In direct examination, attorneys and witnesses are cooperative throughout the conversation. Therefore, attorneys adopt only those devices that help the witnesses answer the questions or to guide them to a certain answer. The question-word question is the most preferred in this type of examinations. In the cross-examination on the other hand, witnesses and attorneys are less cooperative. Thus, the devices adopted are those that discredit the testimony of the witnesses. The Yes-No question is the most preforred in the cross-examination. This study shows that the conversation in Thai courtrooms is schematized, constrained by procedural rules, and allocated to specified parties at specified times. That is, attorneys speak in questions, witnesses in answers; witnesses do not ask, attorneys do not answer. The linguistic form and strategy choice is determined by the three factors--the participants, the end, and the legal provision.