I found this book to be entirely satisfying, which is not something I typically say about a linguistics book. This little gem is tiny, under 100 pages, but covers a lot of ground. Herein you can find the locus classicus of many fundamental ideas about how language works and how best to analyze its inner workings.
The book is divided into two sections. The first (co-authored with Morris Halle) charts the relationship between phonetics and phonology. This phonology part is pre-generative, although Chomsky's Syntactic Structures would appear in the same Janua Linguorum series a year later. Significantly, the book emphasizes the hearer's perspective, both in the definition of features in terms of auditory as well as articulatory profiles and in terms of how features are organized according to criteria such as loudness/volume, pitch/frequency, or duration/length. Jakobson and Halle make interesting claims about how phonetic systems are picked up by children and lost, in the same order but reversed, by victims of aphasia. The primary downside to this fairly terse treatment of linguistic sound systems is that there are no extended examples of analysis showing that binary features are the best way to go. Instead, the only examples are isolated words yanked out of Jakobson's polyglot brain. It would take later work (such as Chomsky and Halle's Sound Pattern of English) to make the strong case that binary features are not only adequate but optimal for the analysis of phonemic systems. If you read this book, I suggest you keep an eye out for two exceptionally modern ideas: 1) physiognomic indices, which point out the speaker to the hearer through cues in pronunciation, timbre, relative pitch, etc. thereby showing that the information contained in a message's sound is not exhausted by the message's decomposition into binary features. 2) they see the dead end which emerges when an attempt is made to treat synchrony without an understanding of change, or to see diachrony of features apart from their systemic function in the sound pattern of the language in question. To this day, phonologists are likely to represent synchronic systems which are in fact internal reconstructions.
The second part of this book is purportedly about aphasia but is in fact an exegesis on semantic systems we all (hopefully) share. Jakobson's ideas about aphasia derive from two poles, contiguity and similarity, between which lies the continuum of aphasia cases. Those with the similarity disorder can only speak about items in context; they cannot express what is topical. They speak in rheme or predicate, unable to express a topic. Their sentences tend to be well formed, but they have difficulty substituting names, instead tending to make an association by way of spatiotemporal contiguity. On the other hand, those lacking contiguity are able to name things but have no ability to link the names together into any kind of proposition. Ideas can only be elaborated through metaphorical linkages in the lexicon of the speaker. Contiguity aphasics apparently lack morphology and syntax (i.e. suffer from agrammatism). They might be able to say a set phrase (idiom chunk). If they utter a morphologically complex word, that word in unanalyzable to the speaker and cannot serve as an analogy for forming new words. Jakobson ties this all together with a comparison of metaphor and metonymy. He sees metaphor as being more basic to poetic forms in all languages and all cultures, while metonymy is foundational to unmarked language use such as simple prose narrative. He even talks about film vs. theater in a semiotic discussion of contiguity as represented in time and place.
To sum up, this book would make a great introduction to the field of linguistics. If you already have a background in linguistics, then this little book should, in a day or two, lead you to rethink some fundamentals of language.