Research philosophies
  1. Experimental control. It is often claimed that so-called laboratory speech is unnatural and hence cannot be used to learn about natural speech. However, without rigorous experimental control, one can never be certain whether an observed pattern is a functional structure of speech or a byproduct of several different mechanisms. As long as the assumptions and claims of laboratory studies are properly made, the risk of finding speech patterns that occur only in the laboratory but never in natural conversation is much smaller than the risk of failing to find the basic mechanisms by directly observing natural speech without proper experimental control. Also, just as importantly, experimental control should not be equated to laboratory/instrumental observation. The latter is only about the means of observation, thus should not be taken as synonymous to experimental control.
  2. Always look for actual mechanisms. One of the ultimate goals of science is to understand the world rather than to just describe it. The same is true for speech science: it is not enough to simply describe the phenomena we observe and give each a name; rather, it is more important to understand the mechanisms that generate the observed phenomena. Furthermore, whether we can discover the actual mechanisms depends much on whether we are consciously and persistently looking for them. Naming ≠ understanding.
  3. Spare no details. Scientific understanding can be only as accurate as the level of detail we choose to use in making our observations. If one takes measurements from only a limited number of points, such as peaks, valleys, the middle point, etc., in the speech samples they study, although certain gross patterns can be observed, the causal relations between the contributing factors and observed patterns is difficult to establish. For this reason, as long as feasibility allows, which itself should also be a goal for improvement, observations and measurements should be as fine-grained as possible.
  4. One step at a time. To effectively tease apart the tightly intertwined mechanisms of speech, it is critical to first isolate the most tangible one, and then extend the understanding to the less tangible ones. It is critical, however, that once a mechanism is established, variations clearly due to it should not be re-attributed to other mechanisms.
For further elaborations, please see Xu (2006a).
Goal of studying speech production & perception
  1. To recognize all individual communicatively functional units
  2. To understand the mechanisms of their transmission.
    Basic Criteria for Recognizing Communicatively Functional Units
  1. They need to be concrete rather than symbolic or featural; otherwise their execution would not be possible.
  2. They need to be distinct from each other in form, otherwise they could not be recognized in perception.
  3. They need to be stable and remain relatively constant across various conditions, otherwise their production and perception would both be difficult.

    Basic Criteria for Recognizing Transmission Mechanisms
  1. Speech is produced by a physical device — the articulatory system; so it necessarily reflects the mechanical properties of the articulators.
  2. Speech production is controlled by the Central Nervous System (CNS); so it necessarily bears the properties of CNS and the characteristics of its interaction with the articulatory system.
  3. Speech is perceived by the auditory system of human listeners; so it is necessarily constrained by the properties of the auditory system.
For further elaborations, please see Xu (2006b).