The things that you can "see" in a spectrogram can arise from all sorts of things, including things produced by a speaker and things produced by other noise sources. By "speech phenomena" I assume you intend to exclude those other noise sources, and also non-speech noises from the speaker like coughing, burping, sighing and so on. Sort of by definition, speech sounds involve "articulators", although the vocal folds are not usually talked about as "articulators". Still, one can say that the articulators are all of the organs involved in speech. However, not all aspects of the speech signal involve movement: things that you can "see" in the speech signal can be there because of the state of some articulator (which affects vocal tract geometry, thus acoustic output). The acoustics of a sustained "s" can be seen, and is not the result of a movement, it's the result of a state.
A spectrogram simply reports what is there and does not care about perception. You can thus see facts of speech which do not aid perception and in fact hinder it. The only way to tell if a feature aids perception or is neutral / antagonistic is to design a word-recognition experiment that manipulates acoustic properties.
You can't directly "see" vowels, fricatives, stops, but you can infer that they are present from the distinctive acoustic properties of those (and other) sounds. The inference is quite fallible, so you may not actually "see" a vowel or stop that is there. You can't really "see" specific sounds unless you are really experienced, but you can see that something is probably a fricative (but which one?). That is, you can detect features, to some extent. You can likewise detect voicing, aspiration, place of articulation (up to a point; practice necessary), nasality, breathy voicing, and so on.