Language is conceived in modern behaviour analysis as a large network of contextually controlled interconnecting stimulus relations. One process in particular, the derived transfer of response functions, is a central feature of these verbal networks. According to this process, the functions of conditioned stimuli (e.g., words) can emerge spontaneously for other stimuli in the language network (e.g., other words). Given this, it is not difficult to see how fear and avoidance can quickly become a clinical issue for verbally able humans once fear and avoidance have been established through direct conditioning experiences in the real world. Researchers within the associative conditioning field have recently become excited by the possibility that conditioned fear can generalise through non-formal stimulus relations. However, their interest in this is recent, their paradigm differs significantly from the behaviour-analytic one, and no studies from that field have directly tested the idea that natural language networks can produce and maintain spontaneous emergence of fear for unconditioned stimuli (i.e., along a semantic or symbolic stimulus continuum). This thesis represented an attempt to produce and control the transfer of fear and avoidance using existing words as conditioned and novel probe stimuli. In doing so, it attempted to build bridges between the methodologies and nomenclature of associative learning theory and behaviour analysis. Experiment1 used an operant conditioning procedure to establish an avoidance response for a real word, and then probed for a derived transfer of avoidance to a categorically related word. Avoidance was not observed to transfer through these verbal relations. Experiment 2 employed a similar paradigm, but with an enhanced US and using concurrent physiological measures of fear. It also employed synonyms as conditioned and probe stimuli. Significant levels of transfer of fear, avoidance and US expectancies were observed. Correlations between physiological and behavioural measures produced ambiguous but conceptually interesting outcomes. These are discussed in terms of the nature of the relationship (i.e., causal of otherwise) between fear, overt avoidance and stimulus function appraisals recorded as US expectancy ratings. The implications of these findings for our understanding of the interface between language and anxiety are considered.