 In the following, I would like to demonstrate the main principles of morphological analysis using an example from Latin. Not only because Latin is a synthetic language that is morphologically very rich, but also because the example will once more show that the basis for morphological analysis must be phonology and not the orthography of a language. The example I have chosen concerns nominal variants of the so-called consonantal declension, and I will confine myself to the singular forms. Here are my examples. The first one is the Latin equivalent of king in the nominative, genitive, dative, accusative and ablative case. The next one is the Latin equivalent of English knight, nox, noctis, etc. Then we have the equivalent of city, orbs, orbes, etc. And finally, here is the Latin word or paradigm for snow, nicks, newies and so on. Already, we can see here that the orthographical representation does not reflect the phonetic facts. Now, look at this orthographical symbol here. The X does not stand for its phonetic equivalent, the voiceless vela fricative ch. So, it is not rech, noch or nich, but rex, nox and nicks. Another example is this V symbol. Now, the V is not pronounced as a labiodental fricative, so we do not say nivis, but nuis. So, the real pronunciation is that of a labiodental approximate. And finally, let's assume your mother tongue is English. Now, if this is so, then you would possibly pronounce this as herbs, and not as you should do orbs, where, in fact, the B does not stand for a voiced bilabel plosive, but for a voiceless one, orbs. Furthermore, the vowel, such as this O symbol, is not pronounced noctis, but noctis. Possibly, if you are the variant of English, you speak American English, then you would have to say naktis. Well, what we do need is a phonetic representation in terms of phonemes. Well, here it is. Then we have nox, noctis, nocti, noctem and nocta. The third paradigm is urps, urbis, urbi, urbem, urbe. And finally, we have nix, nuis, nui, nuem, nuwe. Now, we want to analyze this data morphologically. Our first task is to isolate the morphs involved. Well, quite clearly we can see that the suffixes are identical across the paradigms. The nominative is always represented by the element sir. The genitive is realized by the suffix is. The dative is realized by attaching the suffix i to the stem. The accusative uses m and the ablatibus instrumentalis uses e. And this is the same across the four paradigms. Let's now look at the stems. Well, and here we can clearly see that we have a different form in the nominative case and that the remaining cases are realized by the same form. So the forms for the genitive, the dative, the accusative and the ablatib constitute the sort of regular pattern, whereas the forms that are used for the nominative case are to some extent exceptional. So we are confronted with an example of stem alternation. The affixes for the various case forms are identical. We do not have to say anything about them, but what about the stem forms? Well, as we've already said, in the nominative case we have these forms, and elsewhere we have Now let's look at their distribution. Well, in a distributional analysis it becomes obvious that the nominative forms are followed by s, which in phonetic terms is a voiceless alveolar fricative. And the other forms occur elsewhere, but what does that mean? They occur as soon as s, e, m or e follow, that is, as soon as the first segment of the suffix is a vowel and the vowel is a mid-high or mid-low front vowel. Can we generalize this situation? Well, we can. The final voiceless fricative in the nominative cases has most remarkably the feature minus voice. Well, and in all other case forms we have a vowel that follows, and vowels by definition are normally voiced. So we have a contrast between a voiceless initial element of the suffix that follows the stem in the nominative case, and we have a voiceless initial element in the suffix that follows the elements elsewhere. With this in mind we can now apply a principle that has long been known and was first introduced by Charles Hockett, the American linguist, in the mid-1940s and he called this normalization or, to be precise, preliminary normalization. Now what is it? Well, we first define the most general morph of variants of morphs as a default. So here we are. The most general ones are the ones that occur in the genitive, dative, accusative and ablative cases. Reg, nokt, urd and new. And then we define alternation rules. So in our case we have to do the following. We have to define a rule that says as soon as one of these default stems, these normalized stems, is followed by a suffix whose first element is voiceless, we have a stem change. Well, and elsewhere we do not change at all, so nothing will happen. So all we have to do now is define one of these stem changes in terms of alternation rules. Well, let's do it. Here are the alternation rules, some linguists call these rules of alomophie. Here's the first one that applies to reg or urb. It simply says, if we have a stem whose final consonant, well, it's marked it, whose final consonant is voiced, has the feature plus voice. Here we are, plus voice. This is a symbol here for the word boundary, for the stem boundary. So if the final element before the morph boundary is voiced, then this final element becomes voiceless. If, and that's a condition, if the first element of the suffix is also voiceless. This here is the position of the final element, the consonant. This is the boundary, and the element after the boundary has the feature minus voice. So this is the precise formulation of the morphological realization rule, which says if a consonant in the stem is voiced, it becomes voiceless if the first element of the suffix is also voiceless. So reg plus s becomes regs. Urb plus s becomes urbs. Well, what about the next rule? Well, the noctis pattern is quite similar. Now here we have two consonants. We have noct. So the first consonant, this would be our k in noct, and this would be our t, the c0. And c0, the last consonant, this one becomes zero. That is, it is elided. It is dropped if the first element of the suffix, that is added, has the feature minus voice. So here we have two phonological conditioning rules. The first one in reg and urb is an assimilation rule. The second one in noct is an elision rule. Well, and what about the last one? Well, the fact that new becomes nick in nix-newis. Well, that's a morphological conditioning rule because here we have to remember that in the nominative case, and only in the nominative case we have this relatively unconditioned change where we cannot really postulate a phonological rule unless we would postulate a very strange one. Let's now summarize. Morphological analysis has to be carried out first by converting whatever you have into a phonemic representation. The next step of morphological analysis is the isolation of the morphs involved. And wherever you get some sort of irregularity, you have to try to normalize either the base forms or the affixes. In the case of our Latin examples, the affixes were totally straightforward, so we had to do something about the base forms. And since we were talking about inflection, we were talking about stems. Well, and since we had stem variation, we had to formulate some sort of alternation rules. That can be phonological alternation, morphological alternation, or even lexical alternation in order to come to grips with these variant forms. That is what you have to do in morphological analysis.