Publication
Experiments on temporal learning : from intra and interdimensional gradients to a generalization-based account of complex behavior
| Summary: | In a temporal double bisection task pigeons learn two temporal discriminations. They learn to choose Red over Green after 1-s samples and Green over Red after 4-s samples (Type 1 discrimination) and to choose Blue over Yellow after 4-s samples and Yellow over Blue after 16-s samples (Type 2 discrimination). Subsequently, pigeons choose between Green and Blue after samples ranging from 1 to 16 s and, even though the two choices were previously associated with the same duration of 4 s, the probability of choosing Green increases with sample duration – the context effect. This complex result seems to suggest relational learning because Green was associated with the longer sample in the Type 1 discrimination and Blue was associated with the shorter sample in the Type 2 discrimination. The first goal of this thesis was to investigate whether the context effect could be explained from the combination of the temporal generalization gradients for Green and Blue induced by the discriminative training in a double bisection task, following a strategy used in other stimulus dimensions. The timing model Learning-to-Time (LeT) instantiates this generalization-based account of the context effect. The second goal of this thesis was to investigate whether both the temporal generalization gradients and the context effect were consistent with LeT. In contrast to more familiar stimulus dimensions, such as light wavelength or tone frequency, little research has been directed to the study of generalization gradients in the temporal domain. Little is known about their features and the variables that affect them. The third goal of this thesis was to explore the effects of different training protocols, namely the intradimensional and interdimensional protocols, on the temporal generalization gradient. The thesis comprises three studies. Study 1 and Study 2 addressed the first two goals and also investigated the shape of intradimensional temporal generalization gradients. Study 3 explored the effects of interdimensional training on the temporal generalization gradient. In Study 1, pigeons learned a simplified version of the double bisection task, whereby only the two responses critical to the context effect, Green and Blue, were trained. Namely, pigeons learned to peck Green after 4 s but not to peck it after 1 s and to peck Blue after 4 s but not to peck it after 16 s. Next, temporal generalization gradients for Green and Blue were obtained by varying the sample duration from 1 to 16 s. Finally, pigeons were given a choice between Green and Blue after samples ranging from 1 to 16 s. Study 2 extended Study 1 to a standard double bisection task and, additionally, it involved testing with durations outside the training range. First, pigeons learned to choose Red over Green after 2 s and Green over Red after 6 s and to choose Blue over Yellow after 6 s and Yellow over Blue after 18 s. Second, sample duration was varied from 0.7 to 51.4 s and temporal generalization gradients for Green (over Red) and for Blue (over Yellow) were obtained. And third, pigeons were given a choice between Green and Blue after samples ranging from 0.7 to 51.4 s. The results of the two studies showed that a) the generalization gradients for Green and Blue were either ramp-like or bitonic; b) the preference for Green over Blue increased with sample duration – the context effect; c) the two generalization gradients predicted the observed context effect well; and d) the LeT model accounted for the major trends in the data. In Study 3 pigeons were trained to discriminate the presence and the absence of a t-s sample: they learned to choose Green after the illumination of a houselight for t s and to choose Red after a period in the darkness. Subsequently, the houselight duration was varied and the preference for Green was assessed to obtain temporal generalization gradients. The results showed negative-exponential-like temporal generalization gradients, with proportion Green increasing with houselight duration from the shortest test duration to t s and remaining high for longer durations. Additionally, these gradients were shown from the beginning of testing, suggesting that temporal control can be established without explicit discrimination training along the temporal continuum. Together the three studies showed that phenomena observed in more familiar stimulus dimensions also occur in the temporal domain. |
|---|---|
| Main Authors: | Castro, Ana Catarina Gonçalves Vieira de |
| Year: | 2013 |
| Country: | Portugal |
| Document type: | doctoral thesis |
| Access type: | open access |
| Associated institution: | Universidade do Minho |
| Language: | English |
| Origin: | RepositóriUM - Universidade do Minho |
| Summary: | In a temporal double bisection task pigeons learn two temporal discriminations. They learn to choose Red over Green after 1-s samples and Green over Red after 4-s samples (Type 1 discrimination) and to choose Blue over Yellow after 4-s samples and Yellow over Blue after 16-s samples (Type 2 discrimination). Subsequently, pigeons choose between Green and Blue after samples ranging from 1 to 16 s and, even though the two choices were previously associated with the same duration of 4 s, the probability of choosing Green increases with sample duration – the context effect. This complex result seems to suggest relational learning because Green was associated with the longer sample in the Type 1 discrimination and Blue was associated with the shorter sample in the Type 2 discrimination. The first goal of this thesis was to investigate whether the context effect could be explained from the combination of the temporal generalization gradients for Green and Blue induced by the discriminative training in a double bisection task, following a strategy used in other stimulus dimensions. The timing model Learning-to-Time (LeT) instantiates this generalization-based account of the context effect. The second goal of this thesis was to investigate whether both the temporal generalization gradients and the context effect were consistent with LeT. In contrast to more familiar stimulus dimensions, such as light wavelength or tone frequency, little research has been directed to the study of generalization gradients in the temporal domain. Little is known about their features and the variables that affect them. The third goal of this thesis was to explore the effects of different training protocols, namely the intradimensional and interdimensional protocols, on the temporal generalization gradient. The thesis comprises three studies. Study 1 and Study 2 addressed the first two goals and also investigated the shape of intradimensional temporal generalization gradients. Study 3 explored the effects of interdimensional training on the temporal generalization gradient. In Study 1, pigeons learned a simplified version of the double bisection task, whereby only the two responses critical to the context effect, Green and Blue, were trained. Namely, pigeons learned to peck Green after 4 s but not to peck it after 1 s and to peck Blue after 4 s but not to peck it after 16 s. Next, temporal generalization gradients for Green and Blue were obtained by varying the sample duration from 1 to 16 s. Finally, pigeons were given a choice between Green and Blue after samples ranging from 1 to 16 s. Study 2 extended Study 1 to a standard double bisection task and, additionally, it involved testing with durations outside the training range. First, pigeons learned to choose Red over Green after 2 s and Green over Red after 6 s and to choose Blue over Yellow after 6 s and Yellow over Blue after 18 s. Second, sample duration was varied from 0.7 to 51.4 s and temporal generalization gradients for Green (over Red) and for Blue (over Yellow) were obtained. And third, pigeons were given a choice between Green and Blue after samples ranging from 0.7 to 51.4 s. The results of the two studies showed that a) the generalization gradients for Green and Blue were either ramp-like or bitonic; b) the preference for Green over Blue increased with sample duration – the context effect; c) the two generalization gradients predicted the observed context effect well; and d) the LeT model accounted for the major trends in the data. In Study 3 pigeons were trained to discriminate the presence and the absence of a t-s sample: they learned to choose Green after the illumination of a houselight for t s and to choose Red after a period in the darkness. Subsequently, the houselight duration was varied and the preference for Green was assessed to obtain temporal generalization gradients. The results showed negative-exponential-like temporal generalization gradients, with proportion Green increasing with houselight duration from the shortest test duration to t s and remaining high for longer durations. Additionally, these gradients were shown from the beginning of testing, suggesting that temporal control can be established without explicit discrimination training along the temporal continuum. Together the three studies showed that phenomena observed in more familiar stimulus dimensions also occur in the temporal domain. |
|---|